Engineering of substrate speciﬁcity in a plant cell-wall modifying enzyme through alterations of carboxyl-terminal amino acid residues

SUMMARY Structural determinants of substrate recognition remain inadequately deﬁned in broad speciﬁc cell-wall modifying enzymes, termed xyloglucan xyloglucosyl transferases (XETs). Here, we investigate the Tropaeo-lum majus seed TmXET6.3 isoform, a member of the GH16_20 subfamily of the GH16 network. This enzyme recognises xyloglucan (XG)-derived donors and acceptors, and a wide spectrum of other chieﬂy saccharide substrates, although it lacks the activity with homogalacturonan (pectin) fragments. We focus on deﬁning the functionality of carboxyl-terminal residues in TmXET6.3, which extend acceptor binding regions in the GH16_20 subfamily but are absent in the related GH16_21 subfamily. Site-directed mutagenesis using double to quintuple mutants in the carboxyl-terminal region – substitutions emulated on barley XETs recognising the XG/penta-galacturonide acceptor substrate pair – demonstrated that this activity could be gained in TmXET6.3. We demonstrate the roles of semi-conserved Arg238 and Lys237 residues, introducing a net positive charge in the carboxyl-terminal region (which complements a negative charge of the acidic penta-galacturonide) for the transfer of xyloglucan fragments. Experimental data, supported by molecular modelling of TmXET6.3 with the XG oligosaccharide donor and penta-galacturonide acceptor substrates, indicated that they could be accommodated in the active site. Our ﬁndings support the conclusion on the signiﬁcance of positively charged residues at the carboxyl terminus of TmXET6.3 and suggest that a broad speciﬁcity could be engineered via modiﬁcations of an acceptor binding site. The deﬁnition of substrate speciﬁcity in XETs should prove invaluable for deﬁning the structure, dynamics, and function of plant cell walls, and their metabolism; these data could be applicable in various biotechnologies.


INTRODUCTION
Land plants classified in Embryophytes, which have evolved from Charophyta (green algae), possess characteristic polysaccharide-rich cell walls (CWs) (Cosgrove, 2005;Fry, 2004). CWs are multi-composite hydrogel structures that define the composition of their building blocks, and where variations in covalent and non-covalent linkages that adjoin polysaccharides, and their spatial distribution (Cosgrove, 2005(Cosgrove, , 2015Farrokhi et al., 2009). Besides polysaccharides, another important constituent of CWs encompasses proteins (including non-catalytic expansins) and various inorganic and organic compounds, including non-fermentable hydrophobic phenylpropanoid lignin. Despite certain shared characteristics, the composition of CWs varies in plant species (Popper & Fry, 2003, 2004, organs, and tissues (Burton et al., 2010;Fangel et al., 2012;Fincher, 2009;Kozlova et al., 2020;Sørensen et al., 2010), during growth and developmental stages (Iraki et al., 1989), adaptations to living environments, and biotic and abiotic (environmental) stresses (Kutsuno et al., 2023;Popper et al., 2011). The remarkable diversity of CWs underlies their multiple functions, which include mechanical support, diffusion of compounds, growth regulation, cell-to-cell communication (Popper et al., 2011), the definition of shape and size of a plant, formation of barriers to pathogens or resistance against turgor pressure and restrictions of the rate and direction of cell expansion (Fry et al., 2011). Other roles of CWs also include cell-to-cell adhesion in multicellular plant organisms, the sequestration of toxic metal ions, and the production of signalling compounds, such as oligosaccharins (Ochoa-Meza et al., 2021).
The fundamental feature of catalytic mechanisms of XETs and XEHs is the binding of the XG molecule (donor substrate) and breaking (1,4)-b-glycosidic linkages of the polysaccharide backbone during the first step of a reaction, while in the second step, XETs transfer an XG-fragment with an original non-reducing terminal onto a saccharide acceptor, while XEHs use an activated water molecule as an acceptor (Drula et al., 2022;Hrmova et al., 1998). Both types of reactions follow the so-called ping-pong bi-bi mechanism (Baran et al., 2000;Saura-Valls et al., 2006), which proceeds in two stages that incorporate two transition states (Johansson et al., 2004). The first step involves the deprotonation of a nucleophile (Glu amino acid residue), which attacks the anomeric carbon of a scissile glycosidic linkage adjoining glucose moieties, while another Glu acts as an acid/base, which deprotonates a glycoside acceptor; the third Asp residue controls the protonation states of both Glu residues (Drula et al., 2022;Johansson et al., 2004). The preference for a transfer reaction, instead of a hydrolytic one by XETs, leads to the re-ligation of the nascent donor end of a saccharide moiety at the nonreducing terminal of an acceptor substrate (Johansson et al., 2004;Mark et al., 2009).
The structure of the GH16_20 subfamily XETs is represented by poplar PttXET16A (Protein Data Bank -PDB 1un1 or 1umz, Johansson et al., 2004) that adopts a b-jelly-roll topology and folds into two antiparallel b-sheets, which form the b-sandwich consisting of convex and concave regions. The catalytic machinery is located in the middle of the convex region and has either the EXDXXE motif in subgroup I (GH16_1 to GH16_17 subfamilies) or an EXDXE motif in subgroup II (GH16_18 to GH16_23 subfamilies) (Viborg et al., 2019). In comparison to other GH16 subfamilies, the GH16_20 members encompass a carboxylterminal region located near the convex part of the structures (residues 208-272 in PttXET16A), that form two a-helices, and another b-strand with associated loops, on a concave side of a molecule. This addition extends the active site for the binding of acceptor substrates (Viborg et al., 2019). Another characteristic feature of this subfamily is the presence of a well-conserved loop region (residues 181-190) positioned near a catalytic site with strictly conserved Trp195 forming a hydrophobic platform at the +1 subsite, the occurrence of short loops bordering the active-site cleft at negative subsites (Johansson et al., 2004), and another aromatic platform formed by Tyr81 rather than tryptophan residue at the À1 subsite, which is observed in other GH16 members (Viborg et al., 2019). The broadening of the active-site cleft, resulting from the loop shortening located at negative subsites, is thought to be responsible for the recognition of a branched XG chain McGregor et al., 2017). Major differences between PttXET16A and TmNXG1 (XEH, PDB 2uwa; Baumann et al., 2007), are in three loops that extend the active-site cleft at positive subsites, which assure the binding of acceptor substrates. The importance of one of these loops was verified in the TmNXG1-DYNIIG mutant (PDB 2vh9; Mark et al., 2009), in which loop substitutions led to a sixfold lowering of hydrolytic and a twofold increase of the transglycosylation activities, relative to those in wild-type (WT) TmNXG1. In concert with the active-site loop changes, the fine-tuning of substrate interactive forces (through subtle variations of hydrogen bond and hydrophobic interactions, which are crucial noncovalent forces shaping and maintaining protein structures), modulates a fine ratio between hydrolytic and transglycosylation activities (Mark et al., 2009). Notable substrate specificity differences in XET mutants were observed, where protein sequence alterations were introduced near their active sites (Hrmova et al., 2007;Stratilov a et al., 2019Stratilov a et al., , 2020. In further studies, using the hetero-trans-b-glucanase (HTG) from Equisetum, the increased activity with the cellulose donor substrate was attributed to mutations at positions 10 and 34, where tryptophan and glycine residues were replaced by those of proline and serine, respectively; the latter residues directly participated in donor binding Simmons et al., 2015). The validity of data, supported by molecular modelling of HvXET5 (Hrmova et al., 2007) and AtXTH3 (Shinohara et al., 2017), which exhibit high catalytic rates with the cellulose donor, pointed to equivalent positions of proline and serine residues.
It was also reported that the R246L mutation in HTGs can be responsible for increased activity with the cellooligosaccharide (Cello-OS) acceptor , however, this has yet to be verified because TmXET6.3 (Stratilov a et al., 2019), Equisetum XTHs , and HvXET3, HvXET4, and HvXET6 (Stratilov a et al., 2020), which also catalyse transfer with Cello-OS, have an arginine residue in equivalent positions, while AtXTH3 (Shinohara et al., 2017) has the lysine residue. Similar results, i.e. activity increases with Cello-OS were obtained with the K234T/K237T mutant of TmXET6.3 (Stratilov a et al., 2019)the K237 residue localises to the vicinity of conserved R258, which corresponds to L246 in HTGs . All residues located at the carboxyl terminus of XETs, extend an acceptor binding site, which is prototypical of the GH16_20 subfamily, and thus, could interact with acceptors.
The structural differences modulating donor and acceptor substrate binding are reflected in phylogenetic relationships of the GH16 family, spanning monocots, eudicots, and a basal Angiosperm (Stratilov a et al., 2020). Here, in the Randomised Axelerated Maximum Likelihood tree of 419 GH16 sequences, constructed using the maximum likelihood phylogenetic reconstruction method, XETs of the GH16_20 subfamily (Viborg et al., 2019) created various lineages reflecting their substrate specificity. Based on these and other analyses, it was postulated that the evolution of broad specific XETs (Stratilov a et al., 2020) could have progressed through multiple gene duplication (Behar et al., 2018) and neo-functionalisation to secure their biological roles (Akdemir et al., 2022;  gain of function of ancestral GH16 from XETs by a simple loop extension (Baumann et al., 2007), additions of twothree, or more residues (McGregor et al., 2017), and coevolution of substrate-binding residues (Shinohara & Nishitani, 2021) to broaden their active site clefts (McGregor et al., 2017). Nonetheless, despite signature definitions of XET and XEH enzymes, the exhaustive understanding of their evolution requires further considerations.
We have previously observed that TmXET6.3 exhibits high positional sequence identity/similarity with barley HvXET3 (57/74%), HvXET4 (55/72%), and HvXET6 (59/74%) isoforms, and that TmXET6.3 clusters in the same lineage as HvXET6, and neighbours the lineage of HvXET3 and HvXET4 (cf. Figure 7, Stratilov a et al., 2020 therein). Based on these findings and our previous investigations (Stratilov a et al., 2019(Stratilov a et al., , 2020, in this work, we perform additional mutational and structural bioinformatics study of TmXET6.3. This study is grounded on similarities to the carboxyl-terminally located positively charged residues, specifically those of lysine residues confined near conserved Arg238 in TmXET6.3, and in HvXETs (as specified above), which underlie the recognition of negatively charged [a(1-4)GalAp] 5 . We complement these data by defining molecular interactions in individual XG/[a(1-4) GalAp] 5 complexes of WT and mutant TmXET6.3, and compare them with those of selected barley HvXETs. We discuss the significance of the broadened heterotransglycosylation activity of TmXET6.3 in the structure, dynamics, and function of plant CWs and metabolism.
It has previously been postulated that residues localised at the carboxyl termini of the XET GH16_20 subfamily extend their acceptor binding sites (Ekl€ of & Brumer, 2010;Hrmova et al., 2009;Viborg et al., 2019), compared to e.g. GH16_21 licheninases, which embody one of the shortest sequences in the GH16 family. This carboxyl-terminal extension of around 60 residues (DAFQYRRLSWVRQKY-TIYNYCTDRSRYPSMPPECKRDRD) (Figure 1a) in PttXET16A has unique structural attributes, as confirmed by PDB archive search (https://www.rcsb.org/search/advanced/ structure) of structural similarity. In XETs, this carboxylterminal extension serves as part of an acceptor binding site, mediating key interactions with acceptor substrates (Figure 1b). Here, the broadening of the active-site cleft, resulting from the loop shortening at negative subsite binding sites, is thought to be responsible for the recognition of branched XG chains McGregor et al., 2017). As for the comparisons between XETs and XEHs, one exception includes a loop that precedes the strand containing the catalytic EXDXE motif, which is specifically lengthened in XEHs (Baumann et al., 2007).
In our previous work (Stratilov a et al., 2019) we observed, that in non-specific nasturtium TmXET6.3, the H94, and Q108 residues, and in PttXET16A the Q102 and R116 residues, that locate spatially close to active sites, are essential for mediating interactions with a narrow (PttXET16A) or broad (TmXET6.3) spectrum of neutral acceptor substrates (Hrmova et al., 2022;Stratilov a et al., 2019Stratilov a et al., , 2022. This is supported by the visualisation of the molecular dynamics (MD) trajectory of the XXXG/neutral manno-tetra-oligosaccharide substrate pair in WT TmXET6.3 (broad specific) and PttXET16A (specific) (Video S1). These snapshots point out that TmXET6.3 retains both substrates in the active site, which could eventuate in covalent bond formation (Stratilov a et al., 2019), while PttXET16A is unable to do so.
With the aim to clarify the impact of residues located at the carboxyl terminus of TmXET6.3 on its heterotransglycosylation activity with the charged acceptor [a(1-4)GalAp] 5 substrate (a pectin fragment), we embarked on a mutational analysis using the XG donor and a series of acceptor substrates (Scheme S1). These investigations included conserved R238the mutation to alanine, K237mutations to threonine and tryptophan, and other residues, and converting neutral S240 and Q241 to the lysine residues, thus creating double to quintuple mutant combinations of these residues ( Figure 1a).
To investigate the effects of mutations in the carboxylterminal region of TmXET6.3, we calculated and compared the surface charge distributions in barley, nasturtium, and poplar XETs (Figure 1c). Unlike strongly negative  UniprotKB P27051) in ProMals3D (Pei et al., 2008) indicates the absence of a carboxyl-terminal region in GH_21 enzymes. Conservation indices shown on the top of the alignment are on a scale of 1-9. Grey boxes indicate residues subjected to mutagenesis in this study. Terms ss, h, and e indicate secondary structures, a-helix, and b-strands, respectively. electrostatic potentials around their catalytic sites, which are required to regulate protonation states of catalytic residues, the electrostatic potentials carried both negative and positive patch charges, where positive charges were most pronounced at the carboxyl termini of HvXET3 and TmXET6.3 ( Figure 1c). As referred to previously, the H94/Q108 residue combination in non-specific TmXET6.3 permits to catalyse transfers of XG or HEC fragments into a whole spectrum of structurally distinct neutral acceptors (Shinohara & Nishitani, 2021;Stratilov a et al., 2019Stratilov a et al., , 2022. This can be correlated with a positive charge distribution at carboxyl termini in TmXET6.3 or HvXET3, HvXET4, and HvXET6 (in HvXETs the values decrease in the given order) (Figure 1c). The importance of the H94/Q108 combination for the TmXET6.3 broad substrate specificity, compared to the Q102/R116 residue combination in PttXET16A (Shinohara & Nishitani, 2021), was confirmed by mutation analyses (Stratilov a et al., 2019) and computational chemistry tools; the latter monitors the stability of enzyme-donor-acceptor complexes ( Figure S1) (Stratilov a et al., 2022). As reported, the HvXET3, HvXET4, and HvXET6 isoforms (Stratilov a et al., 2020) transferred the XG or HEC fragments onto the charged penta-galacturonic acid acceptor [a(1-4)GalAp] 5 (Stratilov a et al., 2020), with the activity decreasing in that order; Table 1). For TmXET6.3, which carries a positive charge at the extended acceptor binding site, two mutations W75H and Y110R were introduced, based on the analysis of HvXETs. These mutations were obligatory to increase the positive charge in the acceptor site ( Figure S2) and supported the catalytic activity with [a(1-4)GalAp] 5 (Stratilov a et al., 2020). On the other hand, specific HvXET5 and PttXET16A with the Q102/R116 residue combination (Shinohara & Nishitani, 2021) exhibited negative or slightly negative electrostatic potentials at the carboxylterminal regions ( Figure 1c) and could not bind [a(1-4) GalAp] 5 (Table 1). However, these enzymes bound neutral acceptors with glucose moieties linked by (1,3)-and (1,4)glycosidic linkages (Stratilov a et al., 2022), although the activity with (1,4)-b-linked oligosaccharides was experimentally detected only in HvXET5 (Hrmova et al., 2007).
Molecular docking of the [a(1-4)GalAp] 5 acceptor in the binding site of the W75H/Y110R TmXET6.3 mutant, , and R238 at a distance of less than 4 A ( Figure 6). These interactions could be compared with those of the HvXET3, HvXET4, and HvXET6 isoforms ( Figure S3). In the quintuple, TmXET6.3 W75H/Y110R/K234T/K237W/S240K mutant (corresponding to HvXET6, which lacks K237; Table 1), the K237W mutated residue, due to a larger sidechain forming a wide platform, made close contact with [a(1-4)GalAp] 5 . Here, the interaction of the [a(1-4)GalAp] 5 acceptor with the lysine residue in position 240 (Figure 6d) was not observed in HvXET6 ( Figure S3c). It could be suggested that these lysine residues and mutations thereof could influence an electrostatic potential of a larger part of the enzyme molecule, and therefore heterotransglycosylation activities in TmXET6.3 (Figure 6d). It is also evident from Figure S1, that the disposition of carboxyl-terminal K237 is relatively stablethose data signify that altered enzyme activities might stem from local changes of positive charges at and around surfaces of acceptor binding sites. To support these conclusions, the removal of carboxyl-terminal positively charged residues in TmXET6.3 led to an activity increase with Cello-OS acceptors (cf. Figure 8, Stratilov a et al., 2019 therein), again pointing out, that these residues are not required for interactions with neutral acceptors. These and other factors (such as physical separations between key binding residues and acidic [a(1-4)GalAp] 5 that influence interactions between acceptors and binding residues), could also result from the relatively balanced dispositions of carboxylterminal K237 (Figure S1), and R238 (Figure 6), and neighbouring residues, as demonstrated by MD simulations (Stratilov a et al., 2022).
In an attempt to localise the hetero-transglycosylation XET activity in young germinated nasturtium seedlings in situ ( Figure S4a), we separately dissected their epicotyls and roots, produced crude extracts, and incubated them with SR-labelled [a(1-4)GalAp] 5 . After a 24-hour soaking of epicotyls and root extracts with fluorescently labelled [a(1-4)GalAp] 5 , we imaged these plant extracts under a UV lamp ( Figure S4b-e). Subsequently, we measured the activity of these crude extracts with the XG/SR-labelled [a(1-4)GalAp] 5 substrate pair, using size-exclusion HPLC ( Figure S4f). We detected that while the fluorescence signal was absent in crude epicotyl extracts ( Figure S4c), it was clearly discernible in root extracts, where it accumulated in root phloem/xylem components and CWs ( Figure S4d), meaning that at least some proportion of [a(1-4)GalAp] 5 was incorporated in the CW polysaccharides of roots ( Figure S4e), as previously observed in barley (Stratilov a et al., 2020). This incorporation was further confirmed by determining the enzyme activity of root extracts by sizeexclusion HPLC using the XG/ SR-labelled [a(1-4)GalAp] 5 substrate pair ( Figure S4f).
Our observations indicate that the heterotransglycosylation activity (with the XG/SR-labelled [a(1-4) GalAp] 5 substrate pair), was absent in epicotyl extracts, while it was found in root extracts ( Figure S4f). This suggests that either the seed TmXET6.3 enzyme activity (Stratilov a et al., 2019) was absent in epicotyls, although additional epicotyl isoforms present in the crude extract recognised the XG/XG-OS8 substrate pair (data not shown). In either case, it is likely that nasturtium seed, epicotyl, and root XETs might exhibit different substrate specificities or/and could be regulated differently at gene levels by corresponding transcription factors. These observations require further research and need to be examined in relation to the evolutionary history of nasturtium XET isoforms in various tissues, to uncover if these XET isoforms co-evolved, or were subjected to gene duplication and neo-functionalisation under selection pressures (Stratilov a et al., 2020). Nevertheless, the possibility remains that in nasturtium and other plants, XET gene expression profiles differ during developmental stages, as shown in grapevine (Qiao et al., 2022), allowing for XET enzymes to perform precise functions in various cellular contexts (Folkendt et al., 2021).
In this work, we focussed on the engineering of substrate specificity of nasturtium TmXET6.3 through modifications of positively charged residues at the carboxyl terminus, to reveal new avenues to broaden their carbohydrate substrate recognition. We define, how this singularly carbohydrate-based enzymatic function of TmXET6.3 underlies linking diverse and complex carbohydrates that constitute plant CWs. We construe that broad substratespecific XETs, such as TmXET6.3 could be further diversified to guarantee carbohydrate components heterogeneity, which is required for the generation of strong, yet flexible polysaccharide-based CWs. However, many questions regarding the roles of XETs in plant CW modification and re-modelling remain openthis work encourages those hypotheses that could offer new ideas and be applied to technological developments in the future.

CONCLUSIONS
Our findings demonstrate that the hetero-transglycosylation activity could be gained in the W75H/Y110R mutant and other mutants of TmXET6.3 with the XG/[a(1-4)GalAp] 5 substrate pair, that is, that this activity is directly affected by the configuration of its carboxyl-terminal acceptor binding site and consequently by the tertiary structure of an XET protein molecule. In addition to previously described roles of His75 and Arg110 residues in substrate binding, we revealed that Arg238 and Lys237, and a range of other lysine residues positioned in the carboxyl-terminal region of TmXE6.3, play key roles in donor and acceptor substrate binding. These data suggest that a broad acceptor specificity in TmXET6.3 could be diversified with respect to the chemical nature of an acceptor, and engineered by site-specific alterations of the acceptor binding site.

Substrates for enzyme reactions
The nasturtium seed XG (molecular mass >106 kDa), kindly donated by Dr Mayumi Shirakawa (Dainippon Pharmaceutical Co., Ltd., Osaka, Japan) served as a donor substrate and it was also used for the preparation of xyloglucan-derived oligosaccharides

Enzyme activity assays
Enzyme activities were assayed by size-exclusion chromatography using a fluorimetric detection method, where the increase of fluorescence was monitored due to the incorporation of fluorescent SR-labelled acceptors into newly formed transglycosylation products through the catalytic action of XET enzymes. Reaction mixtures composed of donor substrates, XG, SR-labelled oligosaccharide acceptor substrates (Scheme S1), and enzyme preparations in 0.1 M ammonium acetate/glacial acetic acid buffer, pH 5.5, were in the volumetric ratio of 5:1:4; the final concentrations of donor and SR-labelled acceptors were 0.15% (w/v) and 27.5 lM, respectively. The mixtures were incubated at 25°C for various time intervals in vials placed in an autosampler of the highperformance liquid chromatography (HPLC) Dionex UltiMateTM 3000 device (Thermo Fischer Scientific, Waltham, MA, USA) and the Dionex UltiMateTM FLD-3100 fluorescence detector using an isocratic size-exclusion chromatography method. The analyses were performed on the TSKgel G3000 SWXL column, 7.8 mm 9 300 mm (TosoHaas, Tokyo, Japan), eluted with 100 mM ammonium acetate/glacial acetic acid, pH 5.5 containing 20% (v/v) aqueous acetonitrile (φr = 0.2) at a flow-rate 0.5 ml min À1 . The detector was programmed using respective 530 nm and 575 nm excitation and emission wavelengths. The Chromeleon 6.80 software (Thermo Fischer Scientific) was used for device control and data acquisition. All assays were performed in triplicates (n = 3) and activities with standard deviations were calculated via Excel in Microsoft Office Professional 2016.

Comparative computational chemistry
Protein structural models were downloaded from the AlphaFold database (Jumper et al., 2021;Varadi et al., 2022)  LigPrep) and optimised using Jaguar (Bochevarov et al., 2013; Schr€ odinger Release 2022-4: Jaguar), and the DFT B3LYP-D3 method, including the 6-31G** basis set. The acceptor molecule was docked in optimised structures using the inner box size of 14 A 9 14 A 9 14 A (corresponding to a total box size of 30 A 9 30 A 9 30 A), and the Extra Precision protocol (Friesner et al., 2006; Glide version 6.7). Structural graphics images were generated using PyMol v.2.5.4 (Schr} odinger LLC, Portland, OR, USA), and biomolecular electrostatics or electrostatic potentials were calculated through the plugin with the Adaptive Poisson-Boltzmann Solver (Jurrus et al., 2018) embedded in PyMol.

AUTHOR CONTRIBUTIONS
S S constructed plasmids and mutants. BS, ES and KV performed biochemical characterisations. BS and SK built 3D molecular models and docked ligands. BS and MH generated structural graphics. MH conducted bioinformatics analyses. ES and MH conceived and wrote the manuscript. All co-authors reviewed and approved the manuscript.

CONFLICT OF INTEREST
The authors declare they do not have any conflicts of interest.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article.  Table S1. List of sequence-specific primers used to construct triple to quintuple mutants from TmXET6.3 W75H/Y110R for heterologous expression in P. pastoris. Table S2. Plasmids that were used for the generation of TmXET6.3 mutants. Scheme S1. Abbreviations, descriptions, and chemical structures of substrates used in this work. Figure S1. Superpositions of TmXET6.3 at 0 ns (light green), and after 500 ns (green), and 1000 ns (dark green) MD simulation runs, focused on the position of K237 (box). RMSD values of superpositions between 0 and 500 ns, and 0 and 1000 ns are 1.7 A and 2.1 A, respectively. Figure S2. Surface morphologies of binding sites of WT TmXET6.3 (left) and in W75H/Y110R (right) models, with electrostatic potentials (white, neutral; blue, +5 kT e À1 ; red, À5 kT e À1 ). Squares near W75/Y110 and W75H/Y110R residues point to differences in distributions of electrostatic potentials. Figure S3. 3D models of (a) HvXET3 (green), (b) HvXET4 (cyan), and (c) HvXET6 (orange) in complex with the XXXG donor (yellow cpk sticks)/[a(1-4)GalAp] 5 acceptor (cpk sticks) substrate pair. Separations of 2.6-3.6 A between interacting residues (cpk sticks) and XXXG in the À4 to À1 subsites, and [a(1-4)GalAp] 5 in the +1 to +5 subsites, are indicated in dashed lines. The À4 to +5 subsites are indicated at the bottom of the panels. Residues subjected to mutagenesis in this work are shown in bold blue types. Panels (a) and (b) are based on Stratilov a et al. (2020). Figure S4. A germinated nasturtium seedling under day-light (a), and (b) after incubation with the solution of fluorescently labelled [a(1-4)GalAp] 5 viewed under UV light. The in vivo incorporation of [a(1-4)GalAp] 5 into an epicotyl (c) and a root (d), with detail of CWs, shown in a root cross-section (e). The activity assay of crude extract from roots was detected by size-exclusion HPLC with the corresponding activity curve shown (f). Video S1. Visualisation of the dynamics of the XXXG/neutral manno-tetra-oligosaccharide substrate pair binding by TmXET6.3 (broad specific) and PttXET16A (specific), obtained by MD simulation. The visualisation of the MD trajectory of the XXXG donor and manno-tetra-oligosaccharide acceptor substrates of both enzymes reveals the instability of the acceptor in specific PttXET16A. After 20 ns the manno-tetra-oligosaccharide acceptor changes its position in PttXET16A and approaches the carboxylterminal, which is unfavourable for glycosidic bond formation. Conversely, the manno-tetra-oligosaccharide acceptor in TmXET6.3 remains stable during the duration of the MD simulation and could lead to a hetero-transglycosylation reaction.