The biosynthesis, degradation, and function of cell wall β‐xylosylated xyloglucan mirrors that of arabinoxyloglucan

Summary Xyloglucan is an abundant polysaccharide in many primary cell walls and in the human diet. Decoration of its α‐xylosyl sidechains with further sugars is critical for plant growth, even though the sugars themselves vary considerably between species. Plants in the Ericales order – prevalent in human diets – exhibit β1,2‐linked xylosyl decorations. The biosynthetic enzymes responsible for adding these xylosyl decorations, as well as the hydrolases that remove them in the human gut, are unidentified. GT47 xyloglucan glycosyltransferase candidates were expressed in Arabidopsis and endo‐xyloglucanase products from transgenic wall material were analysed by electrophoresis, mass spectrometry, and nuclear magnetic resonance (NMR) spectroscopy. The activities of gut bacterial hydrolases BoGH43A and BoGH43B on synthetic glycosides and xyloglucan oligosaccharides were measured by colorimetry and electrophoresis. CcXBT1 is a xyloglucan β‐xylosyltransferase from coffee that can modify Arabidopsis xyloglucan and restore the growth of galactosyltransferase mutants. Related VmXST1 is a weakly active xyloglucan α‐arabinofuranosyltransferase from cranberry. BoGH43A hydrolyses both α‐arabinofuranosylated and β‐xylosylated oligosaccharides. CcXBT1's presence in coffee and BoGH43A's promiscuity suggest that β‐xylosylated xyloglucan is not only more widespread than thought, but might also nourish beneficial gut bacteria. The evolutionary instability of transferase specificity and lack of hydrolase specificity hint that, to enzymes, xylosides and arabinofuranosides are closely resemblant.

Xyloglucan is synthesised by glycosyltransferases (GTs) and acetyltransferases in the Golgi apparatus before being further modified by glycosidases and transglycosylases in the cell wall (Pauly & Keegstra, 2016).The biosynthesis of Arabidopsis (fucogalacto)xyloglucan is particularly well characterised: The glucan backbone is made by CSLC4,5,6,8 or 12 (CAZy family GT2), backbone glucosyl residues are xylosylated by XXT1-5 (GT34), xylosyl residues are galactosylated by XLT2 and MUR3 (GT47 subclade A), and galactosyl residues are fucosylated by FUT1 (GT37) and acetylated by XGOAT1/2 (Fig. 1b; Campbell et al., 1997;Coutinho et al., 2003;Pauly & Keegstra, 2016;Zhong et al., 2018;Kim et al., 2020;Drula et al., 2022;Julian & Zabotina, 2022).XLT2 and MUR3 are specific for the second and third xylose in the XXXG repeat, respectively, though XLT2's activity has not yet been confirmed biochemically.The polysaccharide is then trimmed and remodelled in the cell wall by enzymes such as XTH endohydrolases/transglycosylases, AXY8 a-fucosidase, BGAL10 b-galactosidase, and XYL1 a--xylosidase (Pauly & Keegstra, 2016).However, this model cannot be easily extrapolated to other species, as, despite a close evolutionary relationship, homologues of the XLT2 and MUR3 galactosyltransferases in family GT47-A exhibit a remarkable variability in substrate specificity, both between paralogues and between orthologues from across the plant kingdom.For instance, two XLT2 orthologues in tomato, XST1 and XST2, decorate xylosyl residues with arabinofuranose, rather than galactose, to create the S structure (Schultink et al., 2013), whereas a paralogue in Physcomitrium patens, XDT, instead transfers arabinopyranose, creating the D structure (Zhu et al., 2018).Additionally, the hypothetical b-xylosyltransferase needed to synthesise the U structure has been suggested to belong to this family (Schultink et al., 2014).Even in Arabidopsis, an XLT2/ MUR3 paralogue expressed in root hair cells decorates the xylosyl residues with galacturonic acid, resulting in the Y sidechain (Peña et al., 2012).Furthermore, some enzymes in the GT47-A family specifically recognise different polysaccharide acceptors: MBGT1 decorates galactoglucomannan with b-galactose for instance (Yu et al., 2022).Currently, there is no way to predict the activities of these enzymes from amino acid sequence, and current phylogenetic classifications fail to delineate the different donor substrate specificities.Without more functional characterisation, the inability to assign a substrate specificity from sequence prevents us from predicting the carbohydrate structures present in any given plant.Furthermore, we do not yet know whether the rapid evolution of GT47-A activity is purely a product of inconsequential drift (perhaps resulting from a weakly defined substrate specificity that is easily perturbed by active site mutations, or the promiscuity of a recent common ancestor) or whether it is driven Fig. 1 Xyloglucan sidechain nomenclature and biosynthesis.(a) To describe xyloglucan structures in shorthand, a letter is assigned to every backbone glucosyl residue according to its sidechain (Fry et al., 1993).For instance, unsubstituted glucose is represented by 'G', whereas glucose substituted with a single xylosyl residue (Xyl-a1,6-Glc) is termed 'X'.Further letter codes: It is plausible that such structural diversity could provide resistance against cell wall-degrading pathogens (Malinovsky et al., 2014) and constitute an evolutionary selection criterion.The xyloglucan degradation system of Xanthomonas citri pv citri 306 (Vieira et al., 2021) is currently the best-characterised pathogen model for xyloglucan utilisation.Interestingly, this gene cluster appears only to encode enzymes for degradation of fucogalactoand acetoxyloglucan (though two co-regulated GH43-family hydrolases exhibit activity on synthetic a-arabinofuranosides).Hence, it remains to be determined whether the specialised plant pathogens in this genus are able to metabolise other types of xyloglucan, such as the b-xylosylated xyloglucan of plants in the Ericales order.
With xyloglucan making up perhaps as much as 20% of the dry weight of some fruit and vegetables (Toushik et al., 2017), the same structural complexity ought to have consequences for the fibre-degrading microbiota in our guts (and therefore indirectly for human health; Koropatkin et al., 2012).Interestingly, the ability to degrade xyloglucan appears to be a rare trait amongst gut micro-organisms (Hartemink et al., 1996;Larsbrink et al., 2014).The best-studied gut xyloglucan degradation system is that encoded by the xyloglucan utilisation locus (XyGUL) of Bacteroides ovatus ATCC 8483 (Larsbrink et al., 2014).The B. ovatus XyGUL encodes both secretory endo-xyloglucanases, which cleave extracellular xyloglucan into xyloglucan oligosaccharides (XyGOs), and exo-glycosidases, which disassemble XyGOs after their import into the periplasm.Amongst the exoglycosidases are two GH43 a-arabinofuranosidases: BoGH43A and BoGH43B.Importantly, these are the only enzymes demonstrated to exhibit a-arabinofuranosidase activity on xyloglucan or XyGOs.Although the B. ovatus XyGUL does not encode an afucosidase, other Bacteroidetes XyGULs encode functional GH95 a-fucosidases, leading to the suggestion that individual XyGULs are tailored to specific types of xyloglucan (Larsbrink et al., 2014;D ejean et al., 2019).However, it is currently unknown whether gut micro-organisms are capable of metabolising xyloglucans other than fucogalacto-and arabinoxyloglucan.
The diversity of xyloglucan structures across the plant kingdom has yet to be fully explored.At the same time, the range of activities exhibited by GT47-A family members has not been completely characterised.Furthermore, despite the widespread interest in the role of dietary polysaccharides such as xyloglucan in gut health, we lack comprehensive knowledge of their metabolic fate.These factors limit not only our ability to understand xyloglucan function and versatility in planta but also to predict, select, and design xyloglucans with beneficial material and health properties.Here, making use of molecular phylogeny, recently released AlphaFold structures (Jumper et al., 2021;Varadi et al., 2022), and functional characterisation in planta, we explore the evolution and diversity of donor sugar specificities in the GT47-A family.Accordingly, we identify a xyloglucan b-xylosyltransferase (CcXBT1) from Coffea canephora (robusta coffee, from outside of the Ericales order) that is closely related to previously characterised xyloglucan aarabinofuranosyltransferases.We use CcXBT1 to make a novel xyloglucan in Arabidopsis composed almost entirely out of XUXG units that can functionally replace native fucogalactoxyloglucan.
To demonstrate this, we present a glycosidase-based system for analysing xyloglucan structure by electrophoresis.Furthermore, we use the novel xyloglucan structure to investigate the activity of BoGH43A.We find that BoGH43A, and perhaps BoGH43B, are promiscuous and therefore likely act as dual-purpose aarabinosidases/b-xylosidases in B. ovatus xyloglucan metabolism.Hence, both the synthesis and degradation of b-xylosylated xyloglucan are closely related to those of arabinoxyloglucan, suggesting that, in general, the evolutionary leap between activity on L- arabinosides and activity on D-xylosides requires relatively few mutations, therefore constituting a shortcut in the landscape of enzyme evolution.
Exo-glycosidases (Table S4) were added directly to 100 ll resuspended oligosaccharides (polysaccharide analysis by carbohydrate electrophoresis (PACE) implied that XyGO concentration was in the low micromolar range).Reactions were carried out at 37°C for 18 h, except in the case of CgGH3 digestions, which required 48 h for completion.Since we found CgGH3 to exhibit a secondary b-galactosidase activity, this enzyme could not be used to analyse XyGOs that had not already been treated with Fam35 b-galactosidase.In sequential digestions, a-xylosidase was deactivated by ethanol precipitation (as described above), whereas other enzymes were removed by passing the sample through a 3 kDa NanoSep centrifugal filter device (VWR, Radnor, PA, USA).

Polysaccharide analysis by carbohydrate electrophoresis
Polysaccharide analysis by carbohydrate electrophoresis was carried out as described by Goubet et al. (2002).One hundred microlitres of oligosaccharide mixture was used for each sample, along with 5 ll labelling reagent and 10 ll 6 M urea for final resuspension.The 1000-V running step was extended to 150 min.Cello-oligosaccharide standards (50 pmol each) were obtained from Megazyme (Bray, Ireland).

Size exclusion chromatography of oligosaccharides
Size-exclusion chromatography was carried out by a similar method to Yu et al. (2022) in 50 mM ammonium acetate, pH 6.0 (Methods S2).Eluates required clean-up with a PD MiniTrap G-10 column (Cytiva, Marlborough, MA, USA).

Monosaccharide analysis
Following NMR, c. 25 lg of the purified XUXG oligosaccharide was dried, resuspended in 400 ll 2 M trifluoroacetic acid (TFA), and incubated at 120°C for 1 h.After drying, the sample was resuspended in 200 ll MilliQ water and injected into a Dionex ICS300 HPLC system alongside 10-400 lM monosaccharide standards.High-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) then proceeded as described previously (Bromley et al., 2013).

Kinetic enzyme assays
Initial substrate screens on pNP-a-Araf (Merck, Darmstadt, Germany), pNP-b-Xyl (Merck), and pNP-a-Arap (Glycosynth, Warrington, UK) were performed in 10 mM HEPES, pH 7, 250 mM NaCl using 1 lM enzyme and 1 mM pNP-glycoside (4nitrophenyl-glycoside), measuring the product absorbance at A 405 every 5 s for 5 h and then once again after 19 h in a plate reader.To determine pH optima, the activities of BoGH43A and BoGH43B were measured over a range of pH conditions.The following buffers were prepared at 100 mM concentration with New Phytologist (2023) 240: 2353-2371 www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation pH 0.5 increments between the given ranges: citrate-phosphate (pH 3.0-7.0),HEPES (pH 7.0-8.0),Tris-HCl (pH 8.0-9.0), and glycine (pH 9.0-10.5).Initial reaction velocities were recorded by measuring A 405 with 1 mM substrate (pNP-b-Xyl for BoGH43A and pNP-a-Araf for BoGH43B, respectively) and 2 lM enzyme in 50 mM buffer.Activities were normalised to the highest reaction velocity measured for each enzyme and plotted against the pH.Detailed kinetic experiments to obtain Michaelis-Menten parameters for pNP-a-Araf and pNP-b-Xyl were performed in 50 mM HEPES, pH 7.5, with 1 lM enzyme at 20°C.Reaction progress was followed by measuring A 405 every 1 min in a plate reader to measure initial rates.Controls for nonenzymatic hydrolysis were carried out in the absence of enzyme and the background rates subtracted from the initial rates measured for the enzymatic reactions.

Results
The GT47-A III clade is expanded in asterid genomes and exhibits subgroups with altered putative sugar-binding residues Glycosyltransferases from GT47-A exhibit an unusually wide range of substrate specificities, with the sugar nucleotides UDP-Gal, UDP-GalA, UDP-Araf, and UDP-Arap all known to be employed as donor substrates by different members.This variability likely underlies the variety of xyloglucan sidechain structures observed in the cell walls of asterids (Schultink et al., 2014;Pauly & Keegstra, 2016).Within asterid GT47-A enzymes specifically, activities have already been demonstrated for tomato galactosyltransferase SlMUR3 (from group GT47-A VI , as classified by Yu et al., 2022) and arabinofuranosyltransferases SlXST1 and SlXST2 (GT47-A III ; Schultink et al., 2013); however, the enzymatic basis for asterid xyloglucan structural diversity is yet to be fully explored.To investigate whether asterids might possess an enlarged complement of GT47-A enzymes, we calculated a phylogeny of asterid GT47-A protein sequences, with a focus on Ericales and lamiids, which are the source of most of the known xyloglucan structural diversity (Fig. S1).Inclusion of sequences from the Cornales and Ericales, which are basal with respect to the two main asterid groups (lamiids and campanulids), ensured coverage over the full breadth of the asterid clade, while sequences from Arabidopsis (rosids), Aquilegia coerulia (Ranunculalesbasal eudicots) and rice (monocots) provided reference points for other angiosperm groups (previously mapped in Yu et al., 2022).Interestingly, in line with previous phylogenies (Schultink et al., 2013;Yu et al., 2022), the asterid proteomes were observed to exhibit a substantial expansion in group GT47-A III (the clade containing AtXLT2, OsXLT2, and SlXST1/2) relative to the nonasterid references.This expansion hints at potential neofunctionalisation within the clade.
To investigate further, we examined the GT47-A III subtree in more detail.Three distinct subclades were plainly visible within the asterid sequences; we named these subclades a-c (GT47-A III a-c; Figs 2a, S1, S2).AtXLT2 was apparently grouped in subclade a, whereas subclades b and c were specific to asterids.Subclade b contained the tomato sequence Solyc02g092840, which was previously expressed in Arabidopsis but not found to exhibit an obvious activity (Schultink et al., 2013).Within the lamiid sequences, subclade c was clearly split into two lower-level subclades (c1 and c2).Although GT47-A III c1 contains both SlXST1 and SlXST2, no tomato sequences were grouped into GT47-A III c2.
To screen for potential new donor sugar specificities within these groups, we sought to identify likely substrate-binding residues, using the AlphaFold-predicted structure of AtXLT2 (Fig. S3) as a model.From DALI searches (http://ekhidna2.biocenter.helsinki.fi/dali/),the most similar experimentally determined sugar-nucleotide-bound structures were found to be from families GT90, GT10, and GT4, and superimposing them on the AtXLT2 model consistently aligned their ligands in a conserved pocket (Fig. 2b-e).In each case, the aligned nucleotide was placed below the sidechain of Asp407a highly conserved residue that has since been shown to bind sugar nucleotides in other GT47 family members (Leisico et al., 2022;Wilson et al., 2022;Li et al., 2023); hence, this is the likely binding pocket for UDP-Gal in AtXLT2.
Following this argument, we focused on residues in AtXLT2 within 5 A of any of the three aligned donor sugars, which comprised five different regions of interest (Arg343 was omitted for steric reasons).Together, they formed a QF-RI-IE-Y-GDSFTRR pattern, which was fully conserved in OsXLT2 and almost fully conserved in GT47-A VI members AtMUR3, SlMUR3, and OsMUR3 (which exhibited a QF-RI-VE-Y-GDSYTRR pattern).Hence, conservation of these residues may be important for the shared galactosyltransferase activity (Schultink et al., 2013;Liu et al., 2015) of all five enzymes.
To investigate the variability of these residues within the different subclades of GT47-A III , we created a sequence logo for each subclade.Interestingly, while the general pattern seen in AtXLT2 and other characterised galactosyltransferases was strongly conserved in subclades a and b, subclade c exhibited noticeable differences, not only with respect to subclades a and b but also between its lower-level subclades (Fig. 2a).Notably, and consistent with our hypothesis, the arabinofuranosyltransferases SlXST1 and SlXST2 exhibited patterns of MY-RL-VE-Y-GD-GFTRR and MY-RL-VE-Y-GDGLTRR, respectively.Hence, we predicted that family members with unusual substitutions in these regions could potentially harbour novel activities.

GT47-A III c enzymes CcXBT1 and VmXST1 are xyloglucan pentosyltransferases capable of rescuing xyloglucan galactosyltransferase mutant phenotypes
To explore the range of enzyme activities in GT47-A III , we selected three sequences for further characterisation based on their putative donor-sugar-binding residues: (1) Cc07_g06550 (GT47-A III c2; VmXST1.To screen for potential xyloglucan-modifying activity, we tested the capacity of each enzyme to rescue the phenotype of the Arabidopsis mur3-3 mutant, which lacks MUR3 activity and exhibits a dwarfed, cabbage-like growth phenotype (Madson et al., 2003;Kong et al., 2015).CcXBT1, Cc07_g06570, and VmXST1 were each expressed under the promoter of XXT2 (one of the main xyloglucan a-xylosyltransferases in Arabidopsis (Julian & Zabotina, 2022)).We deleted a stretch of repetitive and New Phytologist (2023) 240: 2353-2371 www.newphytologist.comÓ 2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation putatively disordered residues (residues 167-193) from the VmXST1 stem domain in order to maximise chances of expression.After 6 wk of growth, we compared the growth phenotypes of T 1 mur3-3 transgenic plants to wild-type and untransformed mur3-3 plants.Although expression of Cc07_g06570 did not appear to alter the phenotype, expression of CcXBT1 rescued growth almost to wild-type levels (Fig. S4).Expression of VmXST1, however, had an intermediate effect on growth, which was variable between different lines.These results are consistent with the idea that CcXBT1 and potentially VmXST1 are able to alter xyloglucan structure.
To better characterise the products of CcXBT1 and VmXST1, we wanted to express these enzymes in a background lacking galactosylation from both MUR3 and XLT2.Because of the prohibitively strong phenotype of mur3-3, we crossed xlt2 plants with a weaker allele, mur3-1, whose growth is less severely stunted (mur3-1 plants express a MUR3 point mutant with very low activity; Madson et al., 2003;Tamura et al., 2005).The resultant xlt2 mur3-1 double homozygous mutant has been characterised previously (Jensen et al., 2012;Kong et al., 2015) and used in many similar analyses (Schultink et al., 2013;Liu et al., 2015;Zhu et al., 2018).To maximise product yield, we expressed both CcXBT1 and VmXST1 under the strong, primary-cell-wall-synthesis-specific promoter of cellulose synthase 3 (CESA3).As a positive control, we made similar transgenic lines expressing SlXST1.Once more, we compared the phenotypes of 6-wk-old T 1 transgenic plants.Expression of either SlXST1 or CcXBT1 was sufficient to restore growth to wild-type levels (Figs 3a, S5a).However, as before, expression of VmXST1 resulted in only partial complementation.Nevertheless, these results are consistent with the idea that both enzymes modify xyloglucan.
To analyse the xyloglucan structures in the cell walls of these plants, we prepared alcohol-insoluble residue (AIR) from their rosette leaves and extracted hemicellulose using alkali.XyGOs were released by treatment with xyloglucan-specific A. aculeatus endo-b1,4-glucanase (AaXEG; hereafter endo-XGase), which requires an unsubstituted backbone glucosyl residue at the À1 subsite (Pauly et al., 1999).The products were analysed by PACE (Figs 3b, S5, S6) and matrix-assisted laser desorption/ionisationtime of flight (MALDI-TOF) mass spectrometry (Fig. 3c).With reference to previous XyGO PACE band assignments (Yu et al., 2022), we determined that the predominant oligosaccharides released by endo-XGase from wild-type hemicellulose were XXXG, XXFG, and XLFG, as well as a smaller amount of XXLG and XLLG.By contrast, xlt2 mur3-1 material released almost exclusively XXXG, as well as a very small amount of XXFGas reported previously (Kong et al., 2015).However, compared with untransformed xlt2 mur3-1, SlXST1-and CcXBT1-expressing plants exhibited strikingly different XyGO profiles.The PACE profile from SlXST1-expressing plants was dominated by a single band that corresponded to a MALDI-TOF ion with m/z 1217likely representing a (sodiated) octasaccharide composed of four pentoses and four hexoses.From the original characterisation of SlXST1 in Arabidopsis, this XyGO is already known to be XXSG (Schultink et al., 2013), and interestingly, VmXST1-expressing plants yielded a XyGO co-migrating with (and of identical mass to) the main SlXST1 product (albeit of much lower abundance).
However, although the products released from CcXBT1expressing plants also exhibited a dominant band corresponding to the same mass, this main XyGO exhibited substantially different mobility in PACE analysis, indicating that it possesses a different structure.Furthermore, SlXST1-and CcXBT1-expressing plants differed in the presence/absence of several other bands and ions.Interestingly, two XyGOs from the CcXBT1 material had higher mobility than XXXG in PACE analysis.Their smaller implied size indicates that they most likely originate from XXGG-type xyloglucan, which (when deacetylated by alkali) can be cleaved on the reducing-end side of either unsubstituted glucose residue by endo-XGase, producing XXG, XXGG, or decorated versions thereof.
CcXBT1 is a b1,2-xylosyltransferase that transfers xylose to the second a-xylosyl residue in the XXXG repeat To further characterise these potentially novel structures, we tested their sensitivity to two different exo-glycosidases.The first of these was CjAbf51, an a-arabinofuranosidase whose primary activity is on a1,2and a1,3-linked terminal arabinofuranosyl residues on xylan and arabinan (Beylot et al., 2001).Interestingly, both XXSG and the co-migrating oligosaccharide from Fig. 2 Despite their close phylogenetic grouping, members of GT47-A III exhibit an unexpected variability in activity that might stem from subtle changes in three regions of interest.(a) Maximum-likelihood phylogeny of GT47-A subclade III (GT47-A III ; XLT2-related) sequences from a range of asterid species, as well as the following nonasterid references: Arabidopsis, rice, and Aquilegia coerulia.Known donor substrate specificities are annotated with sugar symbols (as designated in the key).Ericales GT47-A III c (c ERI. ) sequences are loosely and tentatively labelled with xylose because the glycosyltransferases responsible for b-xylosylating xyloglucan in argan and bilberry have not yet been identified but are likely to belong to this clade.Sequences were truncated to the GT47 domain before phylogenetic inference using IQ-TREE.Horizontal branch lengths indicate average number of subsititutions per site (refer to scale bar).Support values at important splits represent percentage replication within 1000 ultra-fast bootstrap pseudo-replicates.The tree shown is a subtree of the full GT47-A phylogeny (Supporting Information Fig. S1).See Fig. S2 for taxon labels and further support values.For sites corresponding to putative donorsugar-binding residues in AtXLT2, the conservation was examined by plotting a sequence logo for each phylogenetic subgroup using WebLogo 3. VmXST1-expressing plants could be converted to XXXG by this enzyme (Fig. 4a).Hence, we concluded that VmXST1 is likely an arabinofuranosyltransferase with weak levels of activity, but the same specificity as SlXST1 (at least when expressed in Arabidopsis).By contrast, the main XyGO from CcXBT1-expressing plants was insensitive to CjAbf51 treatment.
New To confirm this hypothesis, we purified the main CcXBT1 XyGO by size-exclusion chromatography and subjected it to two-dimensional solution NMR experiments, monosaccharide analysis, and collision-induced dissociation-mass spectrometry/ mass spectrometry (CID-MS/MS).The results of 1 H-13 C HSQC NMR experiments enabled us to assign 1 H and 13 C chemical shifts to seven distinct monosaccharides within the octasaccharide (Table S5); signals corresponding to the reducing-end glucosyl residue were not detected, presumably due to its anomeric heterogeneity.Cross-peaks produced in 1 H-1 H NMR experiments then confirmed the linkages between them (Fig. 4b).In particular, a cross-peak at (4.53, 3.62) ppm in NOESY experiments provided evidence of a b-xylosyl residue linked to the C2 hydroxyl of the second a-xylosyl residue of an XXXG structure.Furthermore, HPLC analysis of the monosaccharides released after total hydrolysis with TFA revealed that the octasaccharide is composed of a 1 : 1 ratio of xylose to glucose (Fig. 4c; exact ratio = (0.94 AE 0.04) : 1 Xyl : Glc as an average of three technical replicates).Finally, after derivatising the oligosaccharide with 2aminobenzamide, CID-MS/MS confirmed that it possesses a tetrahexosyl backbone with a dipentosyl sidechain attached to the second hexosyl unit and single pentosyl substitutions on the first and third (Fig. 4d).Therefore, the results of all experiments are consistent with the proposed XUXG (Xyl 4 Glc 4 ) structure.Integrating data from the preceding experiments, the xyloglucan in CcXBT1-(over)expressing plants is therefore likely made up predominantly of XUXG repeats in addition to a much smaller number of XUGG, XXXG, XUUG, and XUFG units.

BoGH43A is a dual-function, promiscuous aarabinofuranosidase/b-xylosidase
The presence of an apparent xyloglucan b-xylosyltransferase in C. canephora, from outside the Ericales order, suggests that bxylosylated xyloglucan could be more widespread in the plant kingdom (and therefore the diet) than previously thought.Furthermore, the potential presence of other pentosyl xyloglucan decorations (such as the arabinopyranosylated D structure) has yet to be fully eliminated in higher plants.We were curious to see whether such structures can be metabolised by gut micro-organisms, as no xyloglucan b-xylosidases or aarabinopyranosidases have yet been identified.Interestingly, the two B. ovatus xyloglucan a-arabinofuranosidases, BoGH43A and BoGH43B (encoded by neighbouring genes in the B. ovatus XyGUL; Fig. 5a,b), have markedly different levels of activity on a-arabinofuranosides, and it has been suggested that BoGH43B may have evolved a different function to BoGH43A (Larsbrink et al., 2014).Hence, we set out to analyse comprehensively the activities of the two enzymes.
Kinetic data have already been collected for BoGH43A on the synthetic substrates para-nitrophenyl a-arabinofuranoside (pNP-a-Araf ) and pNP-b-Xyl, whereas data on BoGH43B have only been published for pNP-a-Araf (Larsbrink et al., 2014).To obtain a more complete picture of their activities, we purified both enzymes (Fig. S7) and tested their activity on pNP-a-Araf, pNP-b-Xyl, and pNP-a-Arap.Both BoGH43A and BoGH43B were active on pNP-b-Xyl and pNP-a-Araf, but not pNP-a-Arap, with BoGH43A appearing as a better catalyst for both substrates (Figs 5c, S8), consistent with previous data (Larsbrink et al., 2014).
After determining the pH optimum for both enzymes at pH 7.5 in 50 mM HEPES buffer, we collected Michaelis-Menten kinetic data for both enzymes on pNP-a-Araf and pNP-b-Xyl under these buffer conditions (Fig. S9).Similarly to previous results (Larsbrink et al., 2014), we found the catalytic efficiency (k cat /K M ) of BoGH43A for pNP-a-Araf (20 M À1 s À1 ) to be only twofold higher than that for pNP-b-Xyl (10 M À1 s À1 ).This was in spite of much higher affinity for pNP-a-Araf (K M = 0.49 mM, vs 8.9 mM for pNP-b-Xyl), which was balanced out by a lower turnover number (k cat = 0.01 s À1 vs 0.1 s À1 for pNP-b-Xyl).Although BoGH43B exhibited catalytic parameters in the same order of magnitude for both substrates, they were almost two orders of magnitude lower than those of BoGH43A (Table S6).
To ascertain whether BoGH43A and BoGH43B are active on bona fide b-xylosylated xyloglucan substrates, we incubated endo-XGase products from SlXST1-and CcXBT1-expressing plants with each enzyme in turn.Not only was BoGH43A able to efficiently convert the a-arabinofuranosylated oligosaccharide XXSG to XXXG, but it was also able to convert bxylosylated XUXG to XXXGto at least the same, if not greater, completeness (Fig. 5d).By contrast, the same amount of BoGH43B exhibited only a small amount of activity on the XXSG oligosaccharide and no detectable activity on XUXG (Fig. 5e).Hence, we were able to confirm that BoGH43A, at least, can act both as a xyloglucan a-arabinofuranosidase and as a xyloglucan b-xylosidase.
However, interestingly, one of the blueberry XyGOswhich co-migrated with XUXG and XXLG in PACE analysisappeared to be completely resistant to all three enzymes (Fig. 6e).We enriched this oligosaccharide by size exclusion but were unable to purify it sufficiently for NMR characterisation.Nevertheless, CID MS-MS experiments indicated that its structure is likely XXXG with a pentosyl decoration on the third xylose (Fig. S13).A b1,2-linked xylosyl decoration on the third xylose has been previously reported in bilberry (Hilz et al., 2007); hence, we hypothesised that the recalcitrant blueberry XyGO might be XXUG, and that BoGH43A might not tolerate the structure of this oligosaccharide.However, the recalcitrant blueberry XyGO was also resistant to CgGH3 b1,2-xylosidase (as well as to CjAbf51 a1,2/3-arabinofuranosidase; Fig. S12c).Furthermore, removal of the nonreducing terminal isoprimeverosyl unit by sequential digestion with a-xylosidase and b-glucosidase did not alter its insensitivity to BoGH43A (Fig. S14).Although more work will be required to identify this oligosaccharide, our results suggest that blueberry xyloglucan could be decorated with a hitherto unknown xyloglucan side chainperhaps even the speculative a-xylosylated xylose disaccharide that was previously reported in tobacco and aubergine (Sims et al., 1996;Kato et al., 2010).These results highlight the importance of using comprehensive glycosidase-and/or NMR-based experiments for characterising xyloglucan structure, as opposed to assignments based on mass spectrometry alone.
b-Xylosylated xyloglucan may be restricted to specific, undetermined tissues in coffee and other asterid species Since b-xylosylated xyloglucan has so far been detected in only two species from the Ericales order, we were also prompted to investigate the true prevalence of b-xylosylated xyloglucan in asterids.Since expression of CcXBT1 mRNA has been detected most successfully in roots and leaves (Denoeud et al., 2014), we focused our efforts on these tissues.Surprisingly, our methods were unable to detect b-xylosylation of xyloglucan extracted from leaves of the lamiids C. arabica 'Catimor' (Arabica coffee; Figs S15, S16), C. roseus (Gentianales; Fig. S17), or B. officinalis (borage, Boraginales; data not shown), nor the roots of C. arabica 'Catimor' or C. arabica 'Catuai Amarelo' (Figs S18-S20).Interestingly, the xyloglucan extracted from root samples of the two C. arabica varieties differed dramatically in structure, with 'Catimor' yielding almost exclusively XXFG and XXXG and 'Catuai Amarelo' yielding a complex mixture of arabinosylated and/or galactosylated XyGOs (not previously seen in the Gentianales).This difference highlights the remarkable potential variability in  S5.(c) Monosaccharide analysis of the purified XUXG oligosaccharide following NMR.Monosaccharides were liberated using trifluoracetic acid (TFA) hydrolysis before separation and quantification with high-performance anion-exchange chromatography.Upper trace: 400 lM monosaccharide standards (representative of three technical replicates).Centre trace: monosaccharides released by TFA hydrolysis of the XUXG oligosaccharide; '?', unknown contaminant (representative of three technical replicates).Lower chart: relative quantification of glucose and xylose peak areas from XUXG hydrolysis (using standards as references).(d) Spectrum from tandem mass spectrometry (MS-MS) with collision-induced dissociation (CID) of purified XUXG oligosaccharide following reducing-end derivatisation with 2-aminobenzamide.Assigned carbohydrate fragments are labelled according to the nomenclature of Domon & Costello (1988).We also investigated the structure of xyloglucan from the exocarp of kiwi (Actinidia chinense), a member of the Ericales.Endo-XGase products from kiwi exocarp exhibited a similar size distribution to those from tomato fruit, which is known to exhibit an XXGG-type arabinoxyloglucan (Fig. S21a).Unlike tomato XyGOs, however, kiwi XyGOs were sensitive to CgGH3 b-1,2-xylosidase and BoGH43A, but not CjAbf51 a1,2/3arabinofuranosidase (Figs 7a,b,S21b).With aid from MALDI-TOF MS analysis of untreated and BoGH43A-treated endo-XGase products (Fig. 7c), we assigned the main two XyGOs as XUGG and XUG (the latter likely arising from alternative cleavage of the deacetylated backbone by endo-XGase).Hence, kiwi fruit skin xyloglucan is likely composed mainly of repeating units

Discussion
It was initially speculated that the b-xylosyl residues of argan leaf xyloglucan might be the result of a galactosyltransferase side activity (Ray et al., 2004).However, in this work, we have identified a dedicated xyloglucan b-xylosyltransferase (CcXBT1), which, when expressed in Arabidopsis, is specific to the second xylosyl residue in the XXXG repeat.Interestingly, the presence of XUGG and XUG endo-XGase products from Arabidopsis plants overexpressing CcXBT1 suggests that the high level of CcXBT1 activity might somehow alter the a-xylosylation pattern laid down by XXTs in the Golgi.We speculate that early modification of the second a-xylosyl residue by the overexpressed CcXBT1 enzyme might inhibit the addition of the third.Expression of CcXBT1 in Arabidopsis galactosyltransferase mutants mur3-3 and xlt2 mur3-1 rescued growth, revealing that, like arabinoxyloglucan (Schultink et al., 2013), b-xylosylated xyloglucan can functionally replace fucogalactoxyloglucan in Arabidopsis cell biology.Furthermore, the fact that (over-)expression of either AtXLT2 (Kong et al., 2015) or CcXBT1 (which are both strongly specific for the second a-xylosyl residue) can rescue growth of mur3-3 and xlt2 mur3-1 demonstrates that decoration of the third a-xylosyl residue is not necessary for normal growth in Arabidopsis.Therefore, it seems that it is undecorated xyloglucan in general that is somehow dysfunctional or toxic.This is consistent with the previous proposal that undecorated xyloglucan aggregates in the Golgi, preventing its secretion (Tamura et al., 2005;Kong et al., 2015).
In addition to CcXBT1, we also identified a putative xyloglucan arabinofuranosyltransferase from cranberry (V.macrocarpon, Ericales): VmXST1.The activity of this enzyme is consistent with previous reports that arabinofuranosylated XyGOs can be directly extracted from cranberry fruits (Hotchkiss et al., 2015;Sun et al., 2015;Auker et al., 2019).When expressed in Arabidopsis, VmXST1 exhibited only a very small amount of activity; the reason for this is not clear, but it is possible that, in planta, it is redundant to other, more active paralogues.Nevertheless, the expression of VmXST1 in Arabidopsis galactosyltransferase mutants afforded substantial (albeit incomplete) complementation, suggesting that the level of xyloglucan decoration needs only to reach a low threshold to permit normal growththough CcXBT1 and VmXST1 are members of GT47-A, as was originally proposed for the Ericales b-xylosyltransferase by Schultink et al. (2014).Furthermore, after posting of this manuscript in preprint form, a xyloglucan b-xylosyltransferase from blueberry, VcXBT, was identified in the same family (Immelmann et al., 2023).More specifically, all three enzymes are members of GT47-A III (alongside AtXLT2, OsXLT2, SlXST1, and SlXST2).
Here, we have comprehensively mapped the expansion of this clade in asterid genomes, establishing the presence of as many as four paralogues (a, b, c1, and c2) in many lamiid genomes.Our analysis grouped CcXBT1 in GT47-A III c2 together with orthologues from many other lamiid species, including olive and C. roseus.However, we were unsuccessful in detecting bxylosylated xyloglucan in any lamiid; hence, it is unclear whether the b-xylosyltransferase activity is particular to CcXBT1 or whether b-xylosylated xyloglucan is restricted to a particular organ or tissue in these species.Precedence for the latter lies in the fact that XXXG-type fucogalactoxyloglucan has only been found in the roots of grasses and the pollen tubes of Solanaceaefamily plants (Lampugnani et al., 2013;Dardelle et al., 2015;Liu et al., 2015); similarly, Arabidopsis galacturonoxyloglucan appears limited to root hairs (Peña et al., 2012).A third explanation could be that b-xylosylated xyloglucan is for some reason difficult to extract from the cell walls of these species.
Our data suggest that changes to certain residues surrounding the donor sugar-binding site may explain the apparent neofunctionalisation seen in GT47-A III c.Despite this, we were unable to identify a compelling pattern relating the observed activities to the corresponding protein sequences.Although studying the evolution of these sites could still accelerate the discovery of new and unexpected activities, it is possible that GT47-A nucleotide sugar specificity is too subtle to be predicted from amino acid identities alone; characterisation of further GT47-A III activities and point mutation experiments will be required to confirm this.In particular, we suggest that the distinction between UDP-b-L-Araf and UDP-a-D-Xyl may be especially subtle.This could explain the apparent rapid fluctuations between arabinofuranosyltransferase and xylosyltransferase activity as illustrated in Fig. 2(a).Supporting this idea, the same phenomenon has been observed in the unrelated GT61 family (Cenci et al., 2018;Zhong et al., 2022).Generally, many glycosyltransferases exhibit some level of donor substrate promiscuity (Laursen et al., 2018;Ohashi et al., 2018;Biswas & Thattai, 2020;Ehrlich et al., 2021), so the effective activity could depend to some extent on the availability of the relevant sugar nucleotides.Interestingly, however, and in contrast to glycosyl hydrolases, there is not yet any evidence for Araf /Xyl promiscuity in any member of the GT47 or GT61 families.
Many species in the Ericales order produce commonly consumed fruits, nuts, and leaves; hence, b-xylosylated xyloglucan is likely a substantial component in human diets.Here, we have revealed the probable pathway for its disassembly in B. ovatus and similar species, which are thought to convert xyloglucan into beneficial short-chain fatty acids (Cantu-Jungles et al., 2019;Liu et al., 2020).Although we initially hypothesised that BoGH43B might have a novel activity, we found that BoGH43A is probably the major enzyme responsible for processing both arabinoxyloglucan and bxylosylated xyloglucan.Similar promiscuity is also found in related GH43-family xylan a-arabinosidases/b-xylosidases (Rogowski et al., 2015), suggesting that a-L-arabinosides and b-D-xylosides could also have similar chemical properties.However, although BoGH43B exhibited promiscuity for the two pNP-labelled sugars (Figs 5c, S8, S9), it did not exhibit activity on the b-xylosylated XUXG oligosaccharide (Fig. 5d,e).This discrepancy might stem from differences in substrate concentration or fundamental kinetic parameters between the model and native substrates.Nevertheless, it cannot yet be eliminated that some xyloglucan GH43 enzymes are specific for one of the two sugars; for instance, the putative xyloglucan a-arabinofuranosidase XacAbf43A from X. citri pv citri exhibits tenfold lower activity on pNP-Xyl than on pNP-Araf (Vieira et al., 2021).This could provide an incentive for plants to develop more diverse xyloglucan structures as a means of pathogen resistance.In turn, such adaptations likely incentivise promiscuity in microbial cell wall-degrading enzymes.
To conclude, our results not only expand the existing knowledge of xyloglucan diversity, but also, by identifying an enzyme capable of adding b-xylosyl decorations, make a significant step towards understanding and predicting it from sequence level.In terms of methodology, the glycosidase assays and PACE experiments that we present here also represent the most comprehensive system for analysing xyoglucan structure to date.Combined with knowledge of microbial metabolism in the gut, these results will ultimately allow us to select the best dietary fibres for a balanced gut microbiome and thereby empower us to improve our digestive health.Table S1 Genomic data sources.
Table S2 De novo-synthesised coding sequences for Golden Gate assembly.
Table S3 PCR primers used to amplify DNA parts for Golden Gate assembly.
Table S4 Exo-glycosidases used in this work.
Table S5 1 H and 13 C NMR assignments for XUXG oligosaccharide.
Table S6 Kinetic parameters for BoGH43A/B activity on paranitrophenyl glycosides.
Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors.Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Ó
2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation New Phytologist (2023) 240: 2353-2371 www.newphytologist.com Fig.2Despite their close phylogenetic grouping, members of GT47-A III exhibit an unexpected variability in activity that might stem from subtle changes in three regions of interest.(a) Maximum-likelihood phylogeny of GT47-A subclade III (GT47-A III ; XLT2-related) sequences from a range of asterid species, as well as the following nonasterid references: Arabidopsis, rice, and Aquilegia coerulia.Known donor substrate specificities are annotated with sugar symbols (as designated in the key).Ericales GT47-A III c (c ERI. ) sequences are loosely and tentatively labelled with xylose because the glycosyltransferases responsible for b-xylosylating xyloglucan in argan and bilberry have not yet been identified but are likely to belong to this clade.Sequences were truncated to the GT47 domain before phylogenetic inference using IQ-TREE.Horizontal branch lengths indicate average number of subsititutions per site (refer to scale bar).Support values at important splits represent percentage replication within 1000 ultra-fast bootstrap pseudo-replicates.The tree shown is a subtree of the full GT47-A phylogeny (Supporting Information Fig.S1).See Fig.S2for taxon labels and further support values.For sites corresponding to putative donorsugar-binding residues in AtXLT2, the conservation was examined by plotting a sequence logo for each phylogenetic subgroup using WebLogo 3. (b-d) AlphaFold model of AtXLT2 (blue) aligned by its C-terminal subdomain to the three most similar nucleotide-sugar-bound GT-B structures according to the DALI webserver (http://ekhidna2.biocenter.helsinki.fi/dali/):(b) UDP and glucose bound to Drosophila POGLUT1 (PDB: 5F84; GT90), (c) GDP-fucose bound to Helicobacter pylori FucT (PDB: 2NZY; GT10), (d) UDP-N-acetylglucosamine bound to Campylobacter jejuni PglH (PDB: 6EJI; GT4).The conserved ribose-binding aspartate/glutamate residue is indicated where present.For clarity, the transmembrane and stem regions of the model are hidden, and only the C-terminal subdomain and bound nucleotide sugar is shown for each of the aligned structures.(e) Close-up of putative AtXLT2 donor sugarbinding pocket.Grey circle represents approximate position of donor sugar in aligned structures.Grey dotted lines trace the main chain backbone and are numbered according to the five regions of interest.

Fig. 3
Fig. 3 Expression of CcXBT1 in xlt2 mur3-1 Arabidopsis introduces a novel cell wall xyloglucan structure and rescues the mutant growth phenotype.(a)Representative 6-wk-old heterozygous transformants (T 1 generation) expressing SlXST1 (from tomato), CcXBT1 ('Robusta' coffee), or VmXST1 (cranberry) under the strong primary wall-specific promoter of CESA3.Expression of SlXST1 or CcXBT1 fully rescues the mildly stunted phenotype of xlt2 mur3-1, whereas expression of VmXST1 results in only partial complementation.See Supporting Information Fig.S5for further transgenic lines.(b) Polysaccharide analysis by carbohydrate electrophoresis (PACE) analysis of endo-xyloglucanase (endo-XGase) products from transgenic cell wall material.Hemicellulose was extracted by alkali treatment of leaf alcohol-insoluble residue (AIR) and digested with AaXEG endo-XGase (which cleaves the xyloglucan backbone regularly at unsubstituted glucosyl residues).Products were subsequently derivatised with a fluorophore and separated by electrophoresis.Band assignments made later on in this work are marked with an asterisk.See Fig.S5for further transgenic lines and Fig.S6for no-enzyme controls.Xyloglucan oligosaccharides are labelled according to the previous nomenclature(Fry et al., 1993).(c) Matrix-assisted laser desorption/ionisation-time of flight (MALDI-TOF) mass spectrometry of the same endo-XGase products.Ions assigned in this work are marked with an asterisk.Note that, while PACE is fully quantitative, MALDI-TOF is only semi-quantitative.WT, wildtype.

Fig. 4
Fig. 4 Characterisation of the XUXG endo-xyloglucanase product from CcXBT1-expressing Arabidopsis plants.(a) Sensitivity of endo-xyloglucanase products to a-arabinofuranosidase vs b-xylosidase polysaccharide analysis by carbohydrate electrophoresis (PACE).Alkali-extracted hemicellulose from transgenic plants was digested with AaXEG endo-xyloglucanase.Following ethanol precipitation of undigested polymers and protein, the oligosaccharide products were treated with either CjAbf51 a-arabinofuranosidase (a1,2/3-Araf-ase) or CgGH3 b-xylosidase (b1,2-Xyl-ase).Products were subsequently derivatised with a fluorophore and separated by electrophoresis.(b) Solution nuclear magnetic resonance (NMR) spectroscopy analysis of purified XUXG oligosaccharide from endo-XGase digest of CcXBT1-expressing plants.Each strip corresponds to a reference proton (C 1 -H) from one of the seven distinct monosaccharides (A-G; stars, xylose; circles, glucose).The reducing-end glucose was not visible in our analysis.Contours represent cross-peaks detected in 1 H-1 H NOESY (blue), TOCSY (pink), and DQFCOSY (yellow) experiments; arrows indicate the assigned glycosidic linkages.Chemical shift assignments are provided in Supporting Information TableS5.(c) Monosaccharide analysis of the purified XUXG oligosaccharide following NMR.Monosaccharides were liberated using trifluoracetic acid (TFA) hydrolysis before separation and quantification with high-performance anion-exchange chromatography.Upper trace: 400 lM monosaccharide standards (representative of three technical replicates).Centre trace: monosaccharides released by TFA hydrolysis of the XUXG oligosaccharide; '?', unknown contaminant (representative of three technical replicates).Lower chart: relative quantification of glucose and xylose peak areas from XUXG hydrolysis (using standards as references).(d) Spectrum from tandem mass spectrometry (MS-MS) with collision-induced dissociation (CID) of purified XUXG oligosaccharide following reducing-end derivatisation with 2-aminobenzamide.Assigned carbohydrate fragments are labelled according to the nomenclature ofDomon & Costello (1988).
2023 The Authors New Phytologist Ó 2023 New Phytologist Foundation New Phytologist (2023) 240: 2353-2371 www.newphytologist.comXyG structure between different tissues and closely related plants, and leaves open the possibility that C. canephora tissues exhibit xyloglucans of yet different structure.

Fig. 5
Fig. 5 BoGH43A effectively hydrolyses both a-arabinosides and b-xylosides from synthetic and native substrates.(a) Schematic showing structure of xyloglucan utilisation locus (XyGUL) in the genome of Bacteroides ovatus ATCC 8483.BoGH43A and BoGH43B coding sequences are highlighted in blue and yellow, respectively.(b) Simplified schematic showing previously characterised XyGUL activities.Arrows point to the cleaved glycosidic linkage for each enzyme.(c) Activity test of BoGH43A and BoGH43B on para-nitrophenyl (pNP) glycosides.Absorbance values at A 405 are endpoints measured after overnight incubation at pH 7.0.(d, e) Activity of BoGH43A or BoGH43B, respectively, on a-arabinosylated and b-xylosylated xyloglucan oligosaccharides produced from transgenic plants polysaccharide analysis by carbohydrate electrophoresis (PACE).

Fig. 7
Fig. 7 Xyloglucan from kiwi fruit skin appears to exhibit a xyloglucan made up predominantly of XUGG units.Alkali-extracted hemicellulose from green tomato fruit or kiwi fruit skin was digested with AaXEG endo-XGase followed by ethanol precipitation.(a) Combinatorial digest of endo-XGase products from tomato with BbAfcA a1,2-fucosidase (a1,2-Fuc-ase), Fam35 b-galactosidase (b-Gal-ase), CjAbf51 a1,2/3-arabinofuranosidase (a1,2/3-Araf-ase), and/or CgGH3 b1,2-xylosidase polysaccharide analysis by carbohydrate electrophoresis (PACE).(b) Combinatorial digest of endo-XGase products from kiwi (PACE gel).See also the similar results in Supporting Information Fig. S18 from combinatorial digestions involving BoGH43A.The assignment of the band labelled XXG in a and b is based on co-migration with the XXG product seen in sequential digest experiments on XXXG-type xyloglucan (e.g.Fig. S14).Assignments of XSG and XXGG are tentative but deduced from the enzyme activities and expected band shifts (to XXG or from XUGG, respectively).Assignments of kiwi XyGOs are consistent with mass spectrometry results.(c) Matrix-assisted LASER desorption/ionisation-time-of-flight (MALDI-TOF) mass spectrometry analysis of endo-XGase products from kiwi treated with or without BoGH43A.

Fig. S1
Fig. S1 Unrooted maximum-likelihood phylogeny of GT47-A sequences from a range of asterid and nonasterid species, showing recent expansion of the GT47-A III clade in asterids.

Fig. S3
Fig. S3 Reported confidence metrics for the AlphaFold model of AtXLT2.

Fig. S8
Fig. S8 Initial time courses of BoGH43A and BoGH43B activity on para-nitrophenyl (pNP) glycosides pNP-a-Araf, pNP-b-Xyl, and pNP-a-Arap in order to screen for potential previously unreported activities.

Fig
Fig. S11 MALDI-TOF mass spectrometry analysis of endo-xyloglucanase products from olive leaf, argan leaf, and blueberry fruit skin.