Erwinia tasmaniensis levansucrase shows enantiomer selection for (S)-1,2,4-butanetriol

Levansucrases are useful tools in biotechnology for the synthesis of fructosyl glycosides. The structure of Erwinia tasmaniensis levansucrase in complex with (S)-1,2,4-butanetriol suggests a possible influence of polyols with defined stereocentres in the modulation of fructosylation.

LSCs have potential biotechnological applications in the transfructosylation of nonconventional acceptors (Li et al., 2015). The transglycosylation of molecules may improve their physical, chemical and bioactivity properties (solubility, stability, bioavailability and activity). Enzymatic glycosylation can efficiently produce glycosides, simplifying their in vitro synthesis. As an example, the water solubility of phlorizin (a plant metabolite with relevant pharmacological properties), and consequently its bioavailability, was enhanced via fructosylation catalyzed by Gluconacetobacter diazotrophicus LSC (Herrera-Gonzá lez et al., 2021). Phenol derivatives, such as hydroquinone and puerarin, can also be transfructosylated by LSCs from Bacillus subtilis (BsSacB) and G. diazotrophicus (Mena-Arizmendi et al., 2011;Nú ñ ez-Ló pez et al., 2019). LSCs share the same reaction mechanism (Ortiz-Soto et al., 2019;Bissaro et al., 2015) and their active sites have similar structural features (Martínez-Fleites et al., 2005): a triad of amino acids are involved in catalysis, namely two aspartates and one glutamate (Asp46, Asp203 and Glu287 in Erwinia tasmani-ensis LSC and the same residues in E. amylovora LSC). While the inner binding site is conserved, surface areas and pocket volumes vary across species due to variability in the surrounding loops (Wuerges et al., 2015;Ortiz-Soto et al., 2019).
LSCs from Gram-negative bacteria have successfully been crystallized in complex with sugars in the active site: the structure of E. amylovora LSC (EaLsc; Wuerges et al., 2015; PDB entry 4d47) has been obtained in complex with the sucrose hydrolysis products fructose and glucose, and that of Beijerinckia indica LSC (Tonozuka et al., 2020; PDB entry 6m0e) has been obtained in complex with fructose. LSCs from Gram-positive bacteria have been co-crystallized with fructose (Tonozuka et al., 2012; PDB entry 3vss), sucrose (Meng & Fü tterer, 2003; PDB entry 1pt2) and raffinose (PDB entry 3byn; G. Meng & K. Fü tterer, unpublished work) and also with oligosaccharides with a degree of polymerization up to 6 (PDB entry 6vhq). Inulosucrase from Lactobacillus johnsonii (LjInuJ) has been co-crystallized with sucrose and 1-kestose (Pijning et al., 2011;PDB entries 2yfs and 2yft). Recently, the crystal structure of a fructansucrase from a halophilic archaeal organism, namely the inulin-synthesizing InuHj from Halalkalicoccus jeotgali B3T, has been determined in the presence of sucrose and 1-kestose (Ghauri et al., 2021;PDB entries 7bjc and 7bj4).
Here, we report the crystal structure of E. tasmaniensis levansucrase (EtLsc) in complex with (S)-1,2,4-butanetriol, which was selected by the enzyme during crystallization from racemic 1,2,4-butanetriol. The preference of the enzyme for the (S)-enantiomer was confirmed by microscale thermophoresis binding assays. We analysed the interactions of (S)-1,2,4-butanetriol with conserved amino acids in the active site of EtLsc. We compared the binding mode of (S)-1,2,4butanetriol with the binding mode of the fructose moiety found in bacterial LSC and INU structures deposited in the PDB. We propose that the fructose binding modes could differ in Gram-positive and Gram-negative bacteria.

Expression and purification of recombinant EtLsc
The production of recombinant levansucrase from E. tasmaniensis (strain Et1/99) has previously been described (Polsinelli et al., 2019). In brief, the levansucrase gene lsc was amplified from E. tasmaniensis Et1/99 and the PCR product was cloned into pMCSG49 vector (Eschenfeldt et al., 2009). The protein was expressed in Escherichia coli BL21 (DE3) Star pLysS cells. The protein was purified by immobilized metal-affinity chromatography (IMAC), and the His 6 tag was cleaved with recombinant Tobacco etch virus protease at 4 C overnight and then removed by IMAC. After size-exclusion chromatography the purified EtLsc was concentrated to 10 mg ml À1 (in 20 mM HEPES pH 7.5, 150 mM NaCl) and used for crystallization (Polsinelli et al., 2019). The protein concentration was assessed using the A 280 method in 6 M urea (molecular weight 46 429.9 Da, " c = 84 800 M À1 cm À1 ).

Results and discussion
The structure of EtLsc in complex with (S)-1,2,4-butanetriol was solved to a resolution of 1.40 Å (space group P4 1 2 1 2). Data-collection and structure-refinement statistics are summarized in Table 1. Atomic coordinates and experimental structure factors were deposited in the PDB as entry 7oso.
The overall structure is comparable with other EtLsc structures available in the wwPDB (PDB entries 6frw and 6rv5; Polsinelli et al., 2019Polsinelli et al., , 2020 GH68 family members perform hydrolysis and transfructosylation through a double-displacement reaction (Bissaro et al., 2015;Ortiz-Soto et al., 2019;Homann et al., 2007). It involves hydrolysis of the glycosidic bond in a fructosyl donor. The glucose moiety is then released, while a fructosyl-enzyme intermediate coordinated by the catalytic triad is formed. Subsequently, the fructosyl moiety is transferred from the enzyme to an acceptor molecule (Raga-Carbajal et al., 2018).
To give a clearer description, the binding sites of LSCs and INUs are generally divided into subsites associated with their sugar moiety in relation to the bond hydrolysed by the enzyme (site 0). Up to five substrate-binding subsites (À1, +1, +2, +3, and +4) have been described (Raga-Carbajal et al., 2021).
Alignment of the sequences from the PDB (Fig. 2) confirms that the EtLsc residues interacting with (S)-1,2,4-butanetriol are conserved. Trp45 and Asp203 are highly conserved, as well as Asp46 and Glu287. His97 is conserved where present (a gap is present in the alignment for BsSacB and LjInuJ). His305 is conserved or substituted with another positively charged amino acid, namely an arginine.
Despite the fact that the crystallization conditions included a mixture of the enantiomers (S)-1,2,4-butanetriol and (R)-1,2,4-butanetriol, the electron-density maps (Polder omit, F obs À F calc and 2F obs À F calc ) show exclusively (S)-1,2,4butanetriol. Binding of the ligand is supported by a Polder map contoured at 5 (Fig. 1b; Liebschner et al., 2017). Fitting the R enantiomer gave strong positive and negative peaks in the F obs À F calc map (contoured at 10; Fig. 1c). Therefore, EtLsc has a lower or no affinity for the ligand with the R configuration, as the different stereochemistry at C2 would prevent C2-OH from establishing a hydrogen bond to Asp46 O 2 and give rise to unfavourable interactions with the surrounding residues. The S configuration allows the compound to partially mimic the natural substrate, with C2-OH overlapping O5 of fructose (Fig. 3a), whereas the R enantiomer cannot overlap in a similar way in relation to O5. This explains the preference of EtLsc for one of the enantiomers, highlighting the relevance of stereochemistry when screening potential nonstandard acceptors for this class of enzyme.
Although butanol has already been tested as a fructosyl acceptor in BsSacB without a relevant yield (Mena-Arizmendi et al., 2011), the three OH groups and the chiral centre in (S)-1,2,4-butanetriol mimic the fructosyl moiety in the activated complex/second transition state of the reaction. Based on the structural evidence presented here, (S)-1,2,4-butanetriol and other chiral polyols are potential candidates as a fructosyl acceptor for BsSacB and other similar LSCs or might modulate or influence transfructosylation.  To compare the binding affinity of EtLsc for (S)-1,2,4butanetriol and (R)-1,2,4-butanetriol, both compounds were tested as pure enantiomers by MST. The resulting K d for (S)-1,2,4-butanetriol is 84.9 AE 13.4 mM ( Supplementary Fig. S1). It was not possible to measure any affinity for (R)-1,2,4butanetriol in the same concentration range.

Structural comparisons
The EtLsc structure was compared with the structures of similar glycoside hydrolases from family 68 in complex with ligands present in the PDB (Supplementary Fig. S2 and Supplementary Table S1). The structure of EtLsc is very similar to that of EaLsc from E. amylovora (Wuerges et al., 2015; PDB entry 4d47), with a C r.m.s.d. of 0.396 Å and a sequence identity of 91.6%. The next most similar are BftA from B. indica (BiBftA), with an r.m.s.d. of 1.468 Å (Tonozuka et al., 2020; PDB entry 6m0e), and FFase from Microbacterium saccharophilum (ArFFase), formally known as Arthrobacter sp. K-1, with an r.m.s.d. of 1.5 Å (Tonozuka et al., 2012;PDB entry 3vss). The structures of BsSacB from B. subtilis show r.m.s.d.s in the range between 1.959 and 1.986 Å (PDB entries 6vhq, 3byn and 1pt2; Meng & Fü tterer, 2003;Raga-Carbajal et al., 2021). The alignment with the least similarity is with the inulocrase from L. johnsonii (LjInuJ), with an r.m.s.d. of 2.165 Å (Pijning et al., 2011;PDB entries 2yfs and 2yft).
We compared the binding mode of (S)-1,2,4-butanetriol with the fructose binding mode in the available structures of  (a) Interactions of (S)-1,2,4-butanetriol with EtLsc (PDB entry 7oso). The ligand C atoms are coloured green, with EtLsc C atoms in grey, O atoms in red and N atoms in blue. The ligand interacts with the catalytic triad of EtLsc (Asp46, Asp203 and Glu287). (b) Representation of a polder OMIT map calculated with exclusion of the (S)-1,2,4-butanetriol molecule. The polder map is contoured at 5. (c) 2F obs À F calc and F obs À F calc electron-density maps of (R)-1,2,4-butanetriol. The F obs À F calc map difference-peak intensities are contoured at 10.

Figure 2
Sequence alignment and residue conservation. Alignment of sequences from the structures analyzed in this paper. EtLsc residues that interact with the ligand are marked with a black star above the alignment. Note that PDB entries 6vhq, 2yfs, 3byn and 1pt2 contain mutations.  bacterial GH68 members. In the Gram-negative GH68 members EaLsc and BiBftA, C1-OH, C2-OH and C4-OH of (S)-1,2,4-butanetriol overlap with C3-OH, O5 and C6-OH of fructose, respectively (Fig. 3a). While it is possible to superimpose (S)-1,2,4-butanetriol on the fructose moiety in the Gram-negative bacterial enzymes, it is not possible to do the same in those from Gram-positive bacteria as the fructose moiety binds differently in Gram-negative and Gram-positive bacteria (Figs. 3b and 3c).
All of the LSCs and INUs belong to the same family and have similar structural features, such as the sucrose-binding pocket (Pijning et al., 2011;Wuerges et al., 2015). While subsite À1 is highly specific for a fructosyl moiety, other subsites such as subsites +1, +3 and +4 have been shown to be involved in the binding of fructooligosaccharides (FOS; Raga-Carbajal et al., 2021). Therefore, differences in the acceptor specificity and in the elongation process could be related to the outer part of the pocket. However, the lack of Gram-negative bacterial structures in complex with oligosaccharides in the active site means that this this hypothesis cannot be confirmed. Nevertheless, analysis of the fructose binding network in the available structures (Fig. 4) highlights some peculiar differences between Gram-negative and Gram-positive bacteria. The fructosyl moiety is oriented differently in  and  bacterial structures, despite the conserved pocket at subsite À1. The C6-OH of fructose in Gram-positive enzymes interacts with His97 in EaLsc and with Trp99 in BiBftA. Similarly, C6-OH of fructose interacts with tryptophan in Gram-positive enzymes (Trp85 in BsSacB, Trp111 in ArFFase and Trp271 in LjInuJ) and also with a histidine in ArFFase (His147).
Additionally, in the Gram-positive binding pocket a serine (Ser164 in BsSacB, Ser200 in ArFFase and Ser340 in LjInuJ) stabilizes C4-OH of fructose. This serine is either missing in Gram-negative structures (EaLsc and EtLsc have an alanine at this position) or is not involved in fructose binding (as in BiBftA). This serine has been shown to play a role in the BsSacB product profile, as an S164A variant of BsSacB was found to show enhanced production of blasto-FOS (Ortiz-Soto et al., 2020). Recently, the structure of a fructansucrase from H. jeotgali B3T, a halophilic archaeal organism, has been determined. This enzyme is structurally closer to the LSCs from Gram-negative bacteria than to the Gram-positive bacterial LSCs analysed in this work. However, the fructose binding mode of sucrose and 1-kestose (Ghauri et al., 2021; PDB entries 7bjc and 7bj4) is comparable with those of LSCs and INUs present in Gram-positive bacteria ( Supplementary  Fig. S3).
Despite the common structural features and similar binding sites of Gram-negative and Gram-positive bacterial LSCs and INUs, there is evidence suggesting that their biochemical behaviour could be related to a few small peculiarities; for example, the fructose binding mode and the presence/absence of the serine that binds fructose in subsite À1. These differences, together with the divergences in other subsites involved in FOS binding (Raga-Carbajal et al., 2021), might be relevant to understanding the mechanism regulating chain elongation. However, further studies are required to elucidate the reasons for the different fructose binding modes observed in Grampositive and Gram-negative family GH68 members and the consequences for the product length and the elongation process.

Conclusions
The transfructosylation of bioactive molecules may lead to glycosides relevant to pharmaceutical applications. Levansucrases and inulosucrases are promising tools for transfructosylation and there is a particular interest in discovering new nonstandard acceptors. The structure that we present here contains a polyalcohol moiety with a defined stereochemistry.
Although butanol has already been tested as a fructosyl acceptor in BsSacB without a relevant yield (Mena-Arizmendi et al., 2011), the three OH groups and the chiral centre in (S)-1,2,4-butanetriol might mimic the fructosyl moiety in the activated complex/second transition state of the reaction. The structural data, supported by the affinity study, increase the interest in the potential application of EtLsc and other levansucrases and inulosucrases in the fructosylation of nonstandard acceptors. In fact, this study suggests that the pocket of EtLsc could select a stereospecific polyalcohol motif, making the enzyme a candidate for testing the transfructosylation of molecules including polyalcohols (such as sugar alcohols) or molecules containing polyol moieties. Mutants could be designed based on these structural data to engineer enzymes with the desired preference for this type of nonstandard substrate, for example to synthesize glycosides for pharmaceutical applications. Further studies are required to elucidate the reasons for the different fructose binding modes observed in Gram-positive and Gram-negative family 68 glycoside hydrolases and the consequences for the product length, the implications for the elongation process and the differences in the product spectrum.