Bonds broken and formed during the mixed-linkage glucan : xyloglucan endotransglucosylase reaction catalysed by Equisetum hetero-trans-β-glucanase

Mixed-linkage glucan∶xyloglucan endotransglucosylase (MXE) is one of the three activities of the recently characterised hetero-trans-β-glucanase (HTG), which among land plants is known only from Equisetum species. The biochemical details of the MXE reaction were incompletely understood — details that would promote understanding of MXE's role in vivo and enable its full technological exploitation. We investigated HTG's site of attack on one of its donor substrates, mixed-linkage (1→3),(1→4)-β-d-glucan (MLG), with radioactive oligosaccharides of xyloglucan as the acceptor substrate. Comparing three different MLG preparations, we showed that the enzyme favours those with a high content of cellotetraose blocks. The reaction products were analysed by enzymic digestion, thin-layer chromatography (TLC), high-pressure liquid chromatography (HPLC) and gel-permeation chromatography (GPC). Equisetum HTG consistently cleaved the MLG at the third consecutive β-(1→4)-bond following (towards the reducing terminus) a β-(1→3)-bond. It then formed a β-(1→4)-bond between the MLG and the non-reducing terminal glucose residue of the xyloglucan oligosaccharide, consistent with its xyloglucan endotransglucosylase/hydrolase subfamily membership. Using size-homogeneous barley MLG as the donor substrate, we showed that HTG does not favour any particular region of the MLG chain relative to the polysaccharide's reducing and non-reducing termini; rather, it selects its target cellotetraosyl unit stochastically along the MLG molecule. This work improves our understanding of how enzymes can exhibit promiscuous substrate specificities and provides the foundations to explore strategies for engineering novel substrate specificities into transglycanases.


Introduction
The cell wall, which surrounds virtually all plant cells, has a skeleton of cellulosic microfibrils tethered (and also separated) by a hydrated extraprotoplasmic matrix composed mainly of carbohydrates. Cell expansion is regulated by cell-wall extensibility, which itself is controlled in part by modification of cell-wall polymers by apoplastic enzymes. In addition, the cell wall confers strength to mature tissues, pest and pathogen defence [1], cell signalling and cell-cell adhesion; the wall also determines cell morphology [2] and serves as a source of biologically active signalling molecules [3,4]. Owing to the presence of enzymes that target the structural polysaccharides, plant cell walls are dynamic structures.
The primary cell wall is typically construed as being composed largely of three structurally and functionally distinct polysaccharide fractions: cellulose, hemicelluloses and pectins. The occurrence, prevalence and structure of different cell-wall components often exhibit phylogenic, cell-type and stimulus-dependent variation, the result of the different evolutionary pressures and the differing characteristics conferred on cell walls by each component. similarity of the GH16 enzymes exhibiting them suggests a distinct phylogenetic clade, the nomenclature recognises xyloglucan endotransglucosylase/hydrolase (XTH) as a subfamily that encompasses all XET-and XEH-active GH16 enzymes of plants [42,43].
Despite over 20 years' study, the precise roles of XTHs remain conjectural. A frequently observed negative correlation between extractable XET activity and cell age led to the suggestion that XET may be performing the hypothesised role of catalysing endotransglycosylation in controlled wall-loosening during cell expansion [34]; it was also shown that XET activity participates in cell-wall assembly [41]. More recent reports have further implicated XET activity in regulated cell expansion [44][45][46][47][48].
XTHs have also been implicated in other physiological phenomena, such as cell-wall restructuring [41], the development of vascular tissues [49] and, by allowing the integration of newly secreted xyloglucans into the pre-existing cell-wall architecture, cell-wall assembly [50]. Furthermore, it appears that some XTHs may also, in some situations, be involved in the cessation of cell expansion. Mellerowicz et al. [51] suggest that XTH proteins produced during wood secondary cell-wall development are involved in the strengthening of xylem tissues. Maris et al. [52] showed that application of AtXTH14 or AtXTH26 to onion epidermal peels during constant-load extensiometry actually decreased wall extensibility. Some evidence supports the hypothesis that XTHs may exhibit functional disparity in response to varying conditions. Nishikubo et al. [53] proposed that the ratio of newly synthesised xyloglucan to XTH within Golgi vesicles can dictate whether XET action strengthens or loosens the cell wall. Vissenberg et al. [54] showed that the localisation of XET action can vary with the mechanical properties of the cell wall.
Preferences in site of attack on the donor substrate could have significant ramifications for the roles of transglycanases. XTHs have been reported to exhibit distinct preferences with regard to the length of their substrates. For example, Vinca XTH1 only acted efficiently when using donor xyloglucans of over 10 kDa [35]. Tabuchi et al. [55] characterised an XTH from azuki bean epicotyls, which transferred 50-kDa portions from high-M r xyloglucans to labelled xyloglucan-derived oligosaccharides (XGOs) and separately hydrolysed the high-M r xyloglucans to 50-kDa products; xyloglucans of 60 kDa were not hydrolysed at all. In contrast, other work [56][57][58][59][60] has shown that (for some XTHs) large oligosaccharides can function as donor substrates, albeit at a lower rate than with polysaccharides. With regard to acceptor substrate, Vigna XTH1 acted equally regardless of differences in size [35], whereas recombinant Arabidopsis XTH22 protein had a much higher affinity for xyloglucan polysaccharides (K m = 0.3 mM) than for the nonasaccharide XLLGol [40] (K m = 73 mM). It remains unclear how XTHs might be able to 'measure' glycans of far greater size than them. Such abilities to measure xyloglucan length may have dramatic effects on XET roles in situ. Nonetheless, it appears that most XTHs exhibit a stochastic cleavage profile [61].
A kingdom-wide screen for novel transglycanases led to the surprise finding of the first predominantly hetero-transglycanase in extracts of all Equisetum species tested. This activity, MXE, catalyses a reaction identical with that of XET activity, but uses MLG as the donor substrate instead of xyloglucan [31]. Among land plants, both appreciable extractable MXE activity [31] and in-vivo MXE action [32] appear to be confined to Equisetum. A correlation of extractable MXE activity with organ age led Fry et al. [31] to suggest that MXE might function in the strengthening of ageing stems. Further studies showed MXE to be the most prevalent of three activities [MXE, XET and cellulose:xyloglucan endotransglucosylase (CXE)] that are exhibited by an Equisetum XTH subfamily member, termed hetero-trans-β-glucanase (HTG) [62]. The enzyme's preferred donor substrates are MLG and cellulose; thus, its MXE and CXE activities exceed its XET activity. By far the preferred acceptor substrates are xyloglucan-related oligosaccharides (and thus presumably also xyloglucan); HTG cannot appreciably utilise oligosaccharides of cellulose, MLG, xylan or mannan as acceptor substrates [62].
Here, by using a variety of biochemical approaches, we provide the first detailed characterisation of the MXE reaction. Since all non-Equisetum land plants tested essentially lack detectable MXE activity and action [32], and since our study of the Equisetum fluviatile transcriptome showed only a single HTG sequence [62], we can assume that when Equisetum extracts are tested for MXE activity, only a single protein is responsible. Using the native Equisetum enzyme also has an advantage over using the enzyme heterologously produced in Pichia [62] as we can guarantee correct glycosylation and any other post-translational modifications. We have characterised the site of attack and mechanism of recognition of MLG by HTG extracted from Equisetum. We identify HTG's site of attack on MLG as a β-(1→4)-bond situated exactly 3 residues following a β-(1→3) bond, though it shows little preference for the site of attack along the whole polymer chain relative to its reducing and nonreducing termini. We further present evidence that HTG forms new β-(1→4) bonds, as is consistent with its XTH subfamily membership. This work improves our understanding of how enzymes can exhibit promiscuous substrate specificity and provides the theoretical foundations to explore strategies for engineering novel substrate specificities into glycanases and for exploring the technological exploitation of MXE activity on different lignocelluloses.

Preparation of size-homogeneous barley MLG
Three 10-ml aliquots of barley MLG (6.25 mg/ml) were separately fractionated by GPC on Sepharose CL-6B (bed volume 250 ml, internal diameter 1.4 cm, from which ∼3.6-ml fractions were collected) and the fractions with K av [elution volume on GPC relative to totally excluded polymers (K av = 0.0) and totally included low-M r material ( 3 H 2 O; K av = 1.0)] 0.2-0.3 were pooled. The pool was rerun through the same column, and again the K av 0.2-0.3 region was pooled as size-homogeneous MLG; the final yield was 7.7% of the total starting MLG.

Plant material
E. fluviatile was grown outdoors in the King's Buildings pond in Edinburgh.
For lichenase digests, lichenase was dissolved at 2 U ml −1 in the same buffer. One volume of lichenase solution was added to 1 volume of substrate (MLG) solution and incubated at room temperature for 4 h.
For glucosidase digestion, the sample was incubated with β-D-glucosidase (0.125 U ml -1 in the same buffer) at 20°C for various times.
All enzyme reactions were stopped by the addition of 0.5 volumes of 50% (v/v) formic acid.

MXE reaction
To investigate the MXE reaction, we incubated E. fluviatile extracts, prepared in buffer as before (Fry et al. [31]), with a [ 3 H]XGO-ol (acceptor substrate) and MLG (donor substrate). Routinely, the reactions were stopped by the addition of 0.5 volumes of 50% (v/v) formic acid. MXE products (MLG-[ 3 H]XGO-ol) were freed from unreacted oligosaccharides by precipitation with ethanol (75%, v/v; overnight at 20°C). After benchcentrifugation, the pellet was redissolved in water at 100°C and then reprecipitated with 75% ethanol. This was repeated until the ethanolic supernatant contained no detectable radioactivity (typically four times). Details of the MXE reaction conditions for specific experiments are given in the following three paragraphs.

Hexose assay
Two volumes of 0.2% (w/v) anthrone in concentrated sulphuric acid were added to one volume of sample (aqueous solution containing <0.1 mg/ml hexose) and the mixtures were thoroughly vortexed before incubation at 100°C for 5 min. Samples were cooled in an ice bucket before being measured for absorbance at 620 nm.

Detection of radioactivity
Radioactive spots on TLC plates were localised by fluorography [66a] on pre-flashed Kodak BioMax MR film. Radioactivity in solution was assayed by scintillation counting in ScintiSafe3 (Fisher Scientific UK, Loughborough).

Defining the structure of the MXE product
Since β-(1→3) bonds act as 'hinges' between the relatively rigid cello-oligosaccharide repeat units of MLG, the bond attacked by HTG, like that attacked by other MLG-cleaving enzymes (lichenase [67] and cellulase [68]), is likely to be precisely defined relative to the neighbouring β-(1→3) bond(s). The site of attack of lichenase on MLG is known [67], so lichenase digestion of the MXE product (whose reducing terminus is presumed to be …Glc-Glc-Glc-Glc-XXXGol) should yield a product from the reducing end of the polymer whose structure is diagnostic of the site of attack of MXE on MLG. Specifically, this would tell us the position of the closest β-(1→3) bond in the non-reducing terminal direction and therefore between which negative subsites of HTG the β-(1→3) bond sits during cleavage. Therefore, to characterise the structure of the MXE product formed with [ 3 H]XXXGol as the acceptor substrate and barley MLG as the donor substrate, we digested it with lichenase. This yielded a single 3 H-oligosaccharide (assumed to be Glc n ·XXXGol), which migrated slower than [ 3 H] XXXGol on TLC and was purified by preparative TLC. Graded β-D-glucosidase treatment (Supplementary Figure S1 and Mohler et al. [32]), which releases terminal β-D-glucose residues singly from an oligosaccharide, hydrolysed the Glc n ·XXXGol in two steps to [ 3 H]XXXGol. This indicates that the lichenase product was a nonasaccharide, Glc 2 ·XXXGol. Furthermore, since there was only a single intermediate between the Glc 2 ·XXXGol and XXXGol (Supplementary Figure S1), we suggest that the two β-D-Glc residues were attached to each other, as (Glc 2 )·XXXGol. In contrast, [ 3 H]XXXGol with two separate Glc monosaccharide substituents attached at different positions would have given two different intermediary octasaccharides, for which there is no evidence in Supplementary Figure S1. The linkage within the (Glc 2 ) disaccharide was at this point still unknown. Furthermore, the site of attachment of the (Glc 2 ) moiety to the XXXGol, indicated by the dot in '(Glc 2 )·XXXGol', was also still unknown.
To test the nature of the bond within the (Glc 2 ) disaccharide moiety, we purified (Glc 2 )·XXXGol by GPC and HPLC, then digested it with Trichoderma cellulase [expected to yield a non-radioactive (Glc 2 ) disaccharide + [ 3 H]XXXGol]. Aliquots of the digest were 'spiked' with authentic disaccharides and analysed by HPLC ( Figure 2). Spiking allowed unambiguous discrimination between cellobiose and gentiobiose, and indicated that (Glc 2 )·XXXGol had been hydrolysed to XXXGol and cellobiose. This showed that (Glc 2 )·[ 3 H]XXXGol is  GG·XXXGol, where GG is cellobiose. The bond between the GG and the XXXGol moiety is also likely to be a β-(1→4)-bond because it can be cleaved by cellobiohydrolase. Such a bond identity would also be consistent with HTG's GH16 family membership.
As a test of whether the bond created during the MXE reaction is a β-(1→4) bond to the non-reducing terminal Glc residue of the XGO, we exploited the fact that, regardless of the length of an XGO, the only free Glc 4-hydroxy group is at the non-reducing terminus. In contrast, other free hydroxy groups also occur throughout the oligosaccharide. Therefore, if HTG attaches MLG to a Glc 4-OH, the new bond will always be to the non-reducing terminus of the XGO, whereas if it attaches it to any other hydroxy group, the new bond   (Figure 3a). This indicates that MXE had attached MLG to the non-reducing terminal heptasaccharide subunit alone (Figure 3b). Since the only unique feature of the non-reducing terminal heptasaccharide is its Glc 4-OH group, this observation provides further support that the linkage formed between MLG and XXXGol by HTG is indeed β-Glc-(1→4)-β-Glc. Hence, GG·XXXGol is GGXXXGol in the standard nomenclature for xyloglucan sequences [6] (Figure 3c).

Deducing the site of cleavage of MLG
Sequential MXE-lichenase treatments inform us of the position of the (1→3) bond in HTG's negative subsites during the MXE reaction. The fact that we see only a single product (GGXXXGol) means that the (1→3) bond in the negative subsites is always in the same place when the enzyme is acting on barley MLG. We reasoned that the structure of MLG substrates might affect the site of attack preferences relative to (1→3) bonds. Therefore, to further probe this apparent preference for cleavage relative to a (1→3) bond, we performed MXE-lichenase treatments on three structurally distinct MLGs: Equisetum MLG, barley MLG and Iceland moss MLG (known as lichenan), which have mainly cellotetraose units, mainly cellotriose units and almost only cellotriose units, respectively [64]. When each of these three MLGs were used as the HTG's donor substrate, the sole radioactive product of lichenase digestion was again [ 3 H]GGXXXGol, chromatographically indistinguishable from that obtained when barley MLG was the donor (Figure 4). This confirms MXE's specificity for a (1→3) bond between subsites −4 and −3.
If the target site of HTG is a cellotetraose unit, then the effectiveness of the three tested donor substrates should be Equisetum > barley > Iceland moss, since these MLGs have high, medium and low cellotetraose contents, respectively. Conversely, if the target is cellotriose, the order should be reversed. Assessed from the V max values of MXE activity on the three tested MLGs (Supplementary Figure S2), the target is the cellotetraose unit. Given this knowledge, the deduced structure of the MXE/lichenase product, GGXXXGol (Figure 3c), and the known specificity of lichenase, we can work back to show that the site of attack of HTG on the MLG chain is within a cellotetraose unit, three bonds towards the reducing terminus from a β-(1→3) bond (Figure 5a). This suggestion assumes that HTG cleaves a β-(1→4) bond in the MLG, which would agree with the action of all other known XTH family members.
Does HTG prefer to cleave its donor polysaccharide near one end or near the middle?
The above experiments identify the precise molecular site (' 0 ' in the sequence…G4G3G4G4G 0 4G3G4G…) targeted when HTG uses MLG as the donor substrate. As well as exhibiting specificity at this level, it is also possible that HTG might exhibit a preference for a general region of the MLG chain, e.g. near the non-reducing terminus, near the middle or near the reducing terminus; alternatively, the selection of G4G3G4G4G 0 4G3G4G sites for cleavage may be purely random with respect to the termini. Defining such preferences may help to define the biological role of HTG.
To test for any such preference, we adapted the method of Steele et al. [61]. With a size-homogeneous nonradioactive polysaccharide as the donor substrate ( … ) and a radioactive oligosaccharide as the acceptor ( ), the preferred regions of cleavage can be distinguished by the size of the radioactive productcleavage near the non-reducing terminus: To obtain a suitable donor substrate, we fractionated barley MLG by GPC, producing a relatively sizehomogeneous polysaccharide (Figure 6a,b). We then tested this MLG as a substrate for MXE activity, re-fractionated the products at each time-point by GPC and assayed for total hexose and 3 H. Each sample yielded a single peak of hexose residues indistinguishable from that of the initial donor substrate, indicating that only a very small proportion of the polysaccharide molecules was cleaved between 0 and 8 heither by the MXE reaction itself or by any contaminating hydrolases (Figure 6c). In contrast, there were two radioactive peaks: unchanged oligosaccharide (acceptor substrate) centred at K av ≈ 0.9, and a second, broader peak, attributable to the MXE reaction product (Figure 6d). The latter increased in height with increasing MXE incubation time and exhibited an elution profile which encompassed the maximum size of the donor all the way down to approximately the size of the acceptor. This indicates that MLG cleavage by MXE action does not discriminate strongly between either terminus or the middle of the MLG chain.
Hexose  Figure 6d) by reference to the approximate M r of the material in each fraction, as described previously [61]. As expected, the median M r of the radioactive products was lower than that of the non-radioactive donor substrate when compared on this like-for-like basis (both as molar profiles). Since the donor molecules are far larger than the acceptor molecules, random cleavage should yield radioactive transglycosylation products averaging about half the M r of the donor. Median sizes (estimated from the calibration curve of Steele et al. [61]) were estimated at 200 kDa for the donor substrate and 108-115 kDa for the 3 H-labelled product, indicating that MXE action does indeed cleave the donor substrate essentially stochastically. In this respect, MXE appears to operate a (lack of ) preference similar to that of most XTHs [61]. Our observations indicate that HTG cannot have a unique site of attack at a defined distance from either MLG terminus.

Discussion
Occupancy of HTG's negative subsites by MLG Our demonstration that HTG cleaves an MLG chain three glycosidic bonds towards the reducing terminus from a (1→3) bond distinguishes this unique Equisetum enzyme from its GH16b subfamily co-member [69] lichenase, which cleaves MLG one glycosidic bond towards the reducing terminus from a (1→3) bond. Another difference is that lichenase requires the (1→3) bond [67], whereas β-(1→3)-linkages are clearly not a requirement for the activity of HTG since it can cleave xyloglucan and cellulose [62], which lack (1→3) bonds [2].

Occupancy of HTG's positive subsites by MLG
Our results also give indirect information about the tolerance of (1→3) bonds in the positive subsites. Since HTG cleaves a (1→4) bond located three glycosidic bonds towards the reducing terminus from a (1→3) bond, and since MLGs contain very few cellopentaose or larger cello-oligosaccharide units, it would seem highly likely that a β-(1→3) bond usually exists between the glucose residues in subsites +1 and +2. We cannot say whether a β-(1→3) bond in this position is required for the MXE activity of HTG. To address this point, we would require polysaccharide substrates rich in cellopentaose or larger cello-oligosaccharide units. Nevertheless, the fact that no MLGOs, whether having a (1→3) or a (1→4) bond at the non-reducing end, are capable of acting as acceptor substrateswhile XGOs do (Simmons et al. [62])suggests that HTG shows little affinity for the glucan backbone in positive subsites; the enzyme is merely capable of accommodating the continuation of MLG's backbone into the positive subsites when binding to the … G3G4G4G sequence of the donor substrate in the negative subsites. A (1→3) bond in the positive subsite is clearly not needed by HTG when xyloglucan or cellulose serves as the donor substrate (XET and CXE activities, respectively [62]) since these polysaccharides lack (1→3) bonds [2].
Biological role of MXE activity and the selection by HTG of cleavage sites along the MLG polysaccharide as a whole Extractable MXE activity (due to HTG) peaks in late-season Equisetum plants, whereas XET activity (due to numerous XTH isozymes resembling those of flowering plants) peaks in young, rapidly growing tissue [31,32]. It may therefore be suggested that while 'standard' XTHs play a role in primary cell-wall assembly and/or loosening, the MXE activity of HTG plays a strengthening role in ageing Equisetum stemsperhaps by linking the MLG (which extends into [18], and is sometimes restricted to [21,70], the secondary cell walls, especially those of strengthening tissues) to xyloglucan, which is predominantly in the primary wall. To fulfil a role in MLG-xyloglucan bonding most efficiently, HTG might be expected to cleave the MLG donor substrate at a site (depicted as ' 0 ' below) close to the reducing terminus, so that a minimal length of the MLG chain would be lost in the form of a leaving group (depicted as ) during the interpolymeric transglycosylation reaction between MLG ( … , where each or represents a cello-oligosaccharide unit) and xyloglucan ( …, where each represents a cellotetraose-based unit such as XXXG), thus: 0 + → + However, in contrast with this suggestion, we found that HTG attacks MLG stochastically, selecting its target cellotetraose unit randomly along the polysaccharide chain. In this respect, HTG's attack on MLG resembles XTHs' attack on xyloglucan where there is no noticeable discrimination between units near the reducing or non-reducing ends or near the middle of the xyloglucan chain. It thus appears that the biological function of MXE is adequately fulfilled by randomly attacking any exposed lengths of the MLG chain.

Conclusion
In conclusion, we have defined for the first time the precise site at which HTG selects and cleaves bonds when using MLG as the donor substrate. It selects cellotetraose units, cleaving the (1→4) bond located three glycosidic bonds towards the reducing terminus from a (1→3) bond. These sites are most abundant in Equisetum MLG, which has a higher cellotetraose:cellotriose ratio than poalean and lichen MLG. It is unclear whether MLG or HTG appeared first during the evolution of the Equisetales; however, the enzyme is well adapted to its natural substrate. Nevertheless, an appreciable proportion of cellotetraose units do also occur in poalean MLG; thus, MXE activity might realistically be introduced into grasses or cereals by genetic manipulation. Any consequent effects on poalean wall mechanics could then provide new information on the role of HTG and potentially have the benefit of increasing wall strength in ageing stems in cereal crops. Furthermore, our data show that HTG exhibits no appreciable discrimination between cellotetraose units that are located near the reducing or non-reducing end, or middle, of MLG donor substrate chains.