Crystal structure of a 117 kDa glucansucrase fragment provides insight into evolution and product specificity of GH70

Andreja Vujičić-Žagara, Tjaard Pijninga, Slavko Kraljb,2, Cesar A. Lópeza, Wieger Eeuwemab, Lubbert Dijkhuizenb, and Bauke W. Dijkstraa,1

aLaboratory of Biophysical Chemistry, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands; and bDepartment of Microbiology, University of Groningen, P.O. Box 14, 9750 AA Haren, The Netherlands

Edited by JoAnne Stubbe, Massachusetts Institute of Technology, Cambridge, MA, and approved October 20, 2010 (received for review May 30, 2010)

Glucansucrases are large enzymes belonging to glycoside hydro- inal putative -binding domain flanking the central catalytic lase family 70, which catalyze the cleavage of into domain (7). and , with the concomitant transfer of the glucose residue Because of the similarity of glucansucrases to GH13 α-amy- to a growing α-glucan . Among others, plaque-forming oral lases and GH77 amylomaltases, glucansucrases have been pro- secrete these enzymes to produce α-, which facil- posed to use an α-retaining double displacement mechanism itate the adhesion of the bacteria to the . We deter- (Fig. S1A;7,11–15). In the first step, the αð1 → 2Þ-glycosidic mined the crystal structure of a fully active, 1,031-residue fragment bond of sucrose is cleaved, fructose is released, and a glucosyl- encompassing the catalytic and C-terminal domains of GTF180 from intermediate is formed in which the glucosyl unit is Lactobacillus reuteri 180, both in the native state, and in complexes covalently attached to the catalytic nucleophile via a β-glycosidic with sucrose and maltose. These structures show that the enzyme linkage. In the second step, the covalently bound glucosyl moiety α ðβ∕αÞ has an -amylase-like 8-barrel catalytic domain that is circu- is transferred to the accepting nonreducing end sugar of a grow- larly permuted compared to the catalytic domains of members ing glucan chain, with reformation of the α-. In of families 13 and 77, which belong to the addition, the glucosyl moiety may be transferred to a low mole- BIOCHEMISTRY same GH-H superfamily. In contrast to previous suggestions, the cular mass acceptor substrate such as maltose or isomaltose, or enzyme has only one active site and one nucleophilic residue. to a water . An alternative mechanism proposes that the “ ” Surprisingly, in GTF180 the peptide chain follows a U -path, such acceptor is the reducing end of the growing polysaccharide chain that four of the five domains are made up from discontiguous (Fig. S1B). This mechanism requires either two nearby catalytic N- and C-terminal stretches of the peptide chain. Finally, the struc- sites or one catalytic site with two nucleophilic residues that both tures give insight into the factors that determine the different link- form covalent glucosyl-enzyme intermediates, and that alternat- age types in the polymeric product. ingly transfer their glucosyl moiety to the C1 atom of the other covalently bound glucose, freeing that nucleophile for a next crystal structure complexes ∣ exopolysaccharide ∣ Lactobacillus reuteri ∣ reaction cycle (16). In the absence of a three-dimensional struc- dental caries ture of a GH70 enzyme the precise catalytic mechanism has remained controversial. ental infections such as and periodontal diseases Here we describe the crystal structure of a fully active GH70 are amongst the most common bacterial infections in D glucansucrase of Lactobacillus reuteri 180, encompassing the cat- humans. The causal connection between dental caries and the alytic and C-terminal domains, but not its ∼740 residues N-term- intake of dietary sugar is well established (1). Fermentation of inal domain (17). This 117 kDa GTF180-ΔN (residues 742–1,772) carbohydrates by plaque-forming oral bacteria produces acids, produces a high molecular mass (∼36 MDa) glucose polymer, which cause the dissolution of phosphate in the tooth αð1 → 6Þ αð1 → 3Þ enamel (2). In addition, dietary sucrose serves as a substrate for with both and glucosidic linkages (18). Crystal bacterial glucansucrases, which produce α-glucan polysaccharides structures with bound sucrose (substrate) and maltose (acceptor) that facilitate formation and significantly enhance the ad- combined with site-directed mutagenesis experiments showed that hesion of bacteria to the tooth enamel (3–5). Therefore, glucan- GH70 glucansucrases possess only one active site and have only sucrases have been proposed as potential targets for anticaries one nucleophilic residue. These results provide a solid basis for drugs. Although a wide range of inhibitory substances has been structure-based inhibitor design, and could facilitate the search investigated, none has yet been discovered that is sufficiently for unique specific anticaries drugs. Surprisingly, the GTF180 specific for glucansucrases and that does not have adverse side peptide chain follows a “U”-path, such that four of its five domains effects, e.g., by inhibiting other carbohydrate hydrolyzing en- are made up from two, discontiguous N- and C-terminal stretches zymes in the oral cavity (6). of the peptide chain, respectively. Intestinal bacteria also secrete glucansucrases, and to date over 80 different glucansucrases are known (7, 8). Author contributions: A.V.-Z., S.K., L.D., and B.W.D. designed research; A.V.-Z., T.P., S.K., Glucansucrase enzymes have been classified in glycoside hydro- C.A.L., and W.E. performed research; A.V.-Z., T.P., and S.K. analyzed data; and A.V.-Z., lase family GH70 (9) on the basis of four catalytically important S.K., and B.W.D. wrote the paper. conserved sequence motifs that are similar to those of members The authors declare no conflict of interest. of glycoside hydrolase families GH13 and GH77, but that occur This article is a PNAS Direct Submission. in a different order. This observation led to the proposal that Data deposition: The atomic coordinates and structure factors have been deposited in the glucansucrases belong to the α-amylase superfamily (or GH-H Protein Data Bank, www.pdb.org (PDB ID codes 3KLK, 3HZ3, and 3KLL). ðβ∕αÞ clan), but have a circularly permuted 8-barrel as catalytic 1To whom correspondence should be addressed. E-mail: [email protected]. domain (10). However, glucansucrases are much larger enzymes 2Present address: Genencor—A Danisco Division, Archimedesweg 30, 2333 CN Leiden, (∼1;600–1;800 residues) than their GH13 and GH77 The Netherlands. ∼500–600 relatives ( amino acids), and they contain an N-terminal This article contains supporting information online at www.pnas.org/lookup/suppl/ domain of variable length and unknown function and a C-term- doi:10.1073/pnas.1007531107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1007531107 PNAS Early Edition ∣ 1of6 Downloaded by guest on September 24, 2021 Results Domains IV and V. Domain IV is positioned between domains B and Overall Structure. Lactobacillus reuteri 180 GTF180-ΔN, eluci- V (Fig. 1A). Domain IV has several interactions with the B do- dated at 1.7 Å resolution, has an elongated shape of dimensions main, but very few with domain V. Domain IV has no structural 3 150 × 60 × 60 Å and folds into five distinct, linearly arranged do- similarity to other proteins in the Protein Data Bank (DALI mains (Fig. 1A; for crystallographic details see Tables S1 and S2). Z-scores < 2.5). The function of domain IV is obscure, beyond Three of these domains structurally align with the A, B, and C providing a linker between the catalytic core of the enzyme and domains of family GH13 α-amylases and therefore were named domain V. Domain V contains the first 47 and the last 114 visible accordingly. The other two domains do not show structural simi- residues of the GTF180-ΔN enzyme. Both polypeptide segments larity to domains occurring in the GH13 family and were thus contain ∼20 amino acids long sequence repeats known as putative named IVand V.The five domains are not arranged consecutively cell wall binding repeats (19). Each repeat defines a β-hairpin fol- along the polypeptide chain. From the N to the C terminus, the lowed by a loop connecting consecutive hairpins (Fig. 1A,domain polypeptide chain contributes first to domain V, then to domain V). Two of such repeats are found at the N- and five at the C-term- IV, next to B, then to A and finally to domain C, and then returns inal end of the GTF180-ΔN sequence. Although these sequence back to domain V via domains A, B, and IV, respectively. The repeats were proposed to be involved in glucan binding (7, 8), polypeptide chain thus follows a “U”-path (Fig. 1B). soaking of GTF180 crystals with a variety of sugars (glucose, iso- maltose, panose, and nigerotriose) did not reveal any sugar bind- The A, B, and C Domains. Domain A is the catalytic domain; it ing to this domain. Domain V shows sequence and structural ðβ∕αÞ “ ” contains a 8-barrel, which is indeed circularly permuted similarity to the choline-binding domain of Streptoccocus pneumo- (10); the domain starts with the helix that is equivalent to the niae LytA [(20); 29% identity; rmsd of 2.1 Å for 106 Cα atoms], α ðβ∕αÞ 3 helix of the GH13 8-barrel, while the last strand is but choline-binding could neither be demonstrated. equivalent to strand β3 of the GH13 enzymes (α3, β4α4, …, β α β α β α β 8 8, 1 1, 2 2, 3 arrangement, Fig. S2). Domain B is inserted Sucrose Binding. A 2.2 Å resolution structure of the inactive α β ðβ∕αÞ between helix 1 and strand 8 of the 8-barrel. The domain D1025N mutant protein with bound sucrose substrate showed contributes several amino acid residues to the substrate donor that sucrose binds in the active site, with its glucosyl moiety in and acceptor binding sites (see below). Because of the equiva- subsite −1 and its fructosyl part in subsite þ1 [Fig. 2A and β α β α lence of the GH70 8/ 1 and GH13 3/ 3 secondary structure Video S1; for a definition of the subsites, see Davies et al. (21)]. elements domain B occupies the same relative position in both 2þ Six of the seven strictly conserved GH13 active-site residues (22) enzyme families (Fig. S2). A single heptacoordinated Ca ion are present in GH70 GTFs, making similar interactions with the is bound in the interface between domains A and B (Fig. 2 2þ −1 glucose as in GH13 enzymes [Fig. 2A; (14, 23)]. Among these and Video S1). Treatment of the enzyme with the Ca -chelating strictly conserved residues are the putative catalytic acid/base agent EDTA (ethylenediaminetetraacetic acid) resulted in ∼40% (E1063 in GTF180) and nucleophile (D1025; mutated to N in lower activity (see Table S3), which might be due, among others, the sucrose-bound structure), which, in agreement with their to a compromised binding of the substrate/acceptor by N1029 proposed role, are oriented towards the glycosidic oxygen and (see below). Domain C (residues 1,238–1,376) forms the bottom anomeric carbon of the −1 glucosyl moiety, respectively. Mutat- of the “U.” Even though most α-amylase superfamily enzymes have this domain its precise function is unclear. ing these residues virtually inactivates the enzyme (Table S4). The putative transition state stabilizer D1136 is at 2.8 Å from O2, and ABmay reduce the electronegativity of O2 (14, 24). The crucial role in catalysis of D1136 is affirmed by the drastic decrease in activity of the GTF180-ΔN D1136N-mutant (Table S4). The conserved H1135 contacts the glucosyl O2 and O3 hydroxyl groups, and R1023 hydrogen bonds the O2 hydroxyl group, similar to the equivalent residues in GH13 enzymes. The sixth conserved resi- due, D1504, is not located in the catalytic center, but it makes a H-bond with the hydroxyl group of the conserved Y1465, which has hydrophobic stacking interactions with the sucrose glucosyl moiety in subsite −1. The seventh conserved residue in GH13 enzymes is a histidine that hydrogen bonds to the O6 hydroxyl group of the −1 glucose (22). In GTF180 this residue is Q1509, which fulfils, however, the same function of hydrogen bonding to O6. In subsite þ1 the β-D- fructosyl moiety makes hydrogen bonds with E1063, W1065, N1029, and Q1140 (Fig. 2A). Importantly, no evidence was obtained for a second sucrose- binding site near the active site, as required for the two catalytic- site/double-nucleophile mechanism (16). A second sucrose mole- cule is bound near W1531, approximately 25 Å from the catalytic site (Fig. S3A), but lack of conservation of the binding residues suggests that sucrose binding at this distant site is not of general significance for the GH70 family. Moreover, a W1531S mutation had no effect on the product size (Fig. S4).

Maltose Binding. The disaccharide maltose is a good acceptor Δ Fig. 1. Overall structure of L. reuteri 180 GTF180- N. (A) Crystal structure of substrate (25–27). A structure of native GTF180-ΔN crystals with GTF180-ΔN, the N- and the C-terminal ends of the polypeptide chain are in- dicated, the Ca2þ ion is shown as a magenta sphere; (B) Schematic presenta- bound maltose (2.0 Å resolution) revealed four maltose-binding tion of the “U-shaped” course of the polypeptide chain. Domains A, B, C, IV, sites, M1, M2, M3, and M4 (see Tables S1 and S2 for crystallo- and V are coloured in blue, green, magenta, yellow, and red, respectively, with graphic statistics). The M2–M4 binding site residues are not dark and light colors for the N- and C-terminal stretches of the peptide chain. conserved in other GH70 family enzymes and therefore seem

2of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1007531107 Vujičić-Žagar et al. Downloaded by guest on September 24, 2021 Fig. 2. Stereo views of sucrose and maltose binding in the active site of GTF180(D1025N)-ΔN. (A) Sucrose (white stick representation) bound at subsites −1 and þ1 in the D1025N mutant. Residues of the 980–984 loop are depicted as a transparent cartoon for clarity. Strictly conserved residues in the α-amylase super- family are labeled in bold, except for D1504 which is not shown; (B) Maltose M1 (yellow stick presenta- tion) bound in subsites þ1 and þ2. Maltose M2 bound at subsites þ2 and þ3 is also shown. The glu- cosyl moiety of the sucrose from the GTF180-ΔN- sucrose complex (white stick representation) is super- imposed. Leu938 in front of the þ1 glucosyl residue was omitted for clarity. Residues from domains A and B are colored blue and green, respectively. The Ca2þ ion is shown as a magenta sphere, water

are shown as smaller red spheres. Hydrogen bonding BIOCHEMISTRY interactions are shown as black dashed lines.

of no importance to the GH70 family in general. A description of docking studies in a model of the covalent glucosyl-GTF180-ΔN the M3 and M4 binding sites can be found in the SI Text. intermediate (in subsite −1), which was made based on the Maltose M1 is bound at subsites þ1 and þ2. The þ1 glucosyl structure of the glucosyl-intermediate of Neisseria polysaccharea moiety of M1 is sandwiched between W1065 on one side and amylosucrase [PDB code 1S46; (23)]. Like maltose, isomaltose L938 and L981 on the other side (Fig. 2B and Video S2). The (Glc-αð1 → 6Þ-Glc) binds with its nonreducing glucose residue þ1 C6 hydroxyl group is at hydrogen bonding distance from in subsite þ1; its O6 hydroxyl group points towards the position the catalytic acid/base, E1063, in a productive position to attack of the C1 atom of the covalently bound sugar in subsite −1 a covalently bound enzyme-glucosyl-intermediate. The þ1 C4 (Fig. 3A). The reducing end glucose binds in subsite þ2, and hydroxyl group makes a direct, short hydrogen bond (2.6 Å) with has stacking interactions with W1065. The isomaltose binding N1029. The þ2 glucose residue has van der Waals interactions mode shows how an αð1 → 6Þ-linked glucan can be extended by with W1065 (Fig. 2B). The disaccharide shows no deviations another glucose residue via an αð1 → 6Þ-linkage. from the low energy maltose torsion angles and also the hydrogen A different mode of binding was shown by isomaltotriose bond normally present between the C2 and C3 hydroxyl groups of (Glc-αð1 → 6Þ-Glc-αð1 → 6Þ-Glc), which binds in subsites þII’, adjacent αð1 → 4Þ-linked glucose molecules is formed. Interest- þ1 and þII” (Fig. 3B). The O3 atom of the central, þ1 glucose ingly, maltose M2 binds near the active site (Fig. 2B). However, residue is directed towards the C1 atom of the glucosyl-enzyme the M2 binding mode is rather aspecific (via H-bonding interac- intermediate. This binding mode illustrates how an αð1 → 3Þ tions with the þ1 glucose, and water-mediated hydrogen bonds branch point can be introduced in the glucan chain. In principle, with the protein) and therefore unlikely to mimic a bona fide isomaltose could also bind in subsites þ1 and þII” with O3 point- acceptor substrate. ing towards the C1 atom at subsite −1 . This binding mode would allow elongation of a linear αð1 → 6Þ-glucan chain with an Analysis of Products Synthesized by GTF180-ΔN. Product analysis αð1 → 3Þ-linked glucose. of GTF180-ΔN incubated with both sucrose and maltose present Docking of the disaccharide nigerose (Glc-αð1 → 3Þ-Glc) in a 1∶1 ratio revealed that ∼42% of maltose was converted into resulted in the þ1 glucose predominantly being bound with its the trisaccharide panose (Glc-αð1 → 6Þ-Glc-αð1 → 4Þ-Glc; Glc is O6 hydroxyl group pointing towards the C1 atom of the glucosyl- glucose) and ∼11% into the tetrasaccharide αð1 → 6Þ-panose enzyme intermediate (Fig. 3C), while the reducing end glucose oc- (Glc-αð1 → 6Þ-panose) (Fig. S5 and Ta b l e S 5 , for details on cupies subsite þ2 (Fig. 3C). This binding mode shows how a glu- product analysis see SI Text). These results show that the glucosyl cosyl moiety can be added to the O6 hydroxyl group of an acceptor moiety of sucrose is transferred to the nonreducing end of maltose saccharide of which the terminal nonreducing sugar is linked via and panose. Transfer to the nonreducing end is also in agreement an αð1 → 3Þ linkage. The docking of nigerose did not result in a with other studies (13, 15) as well as with the structure of GTF180- binding mode allowing elongation at the O3 hydroxyl group. ΔN with a bound maltose, which shows that maltose binds adjacent to subsite −1 with its nonreducing end C6 hydroxyl group pointing Discussion towards the catalytic center. Thus, the O6 atom can act as the Discontiguous Domain Arrangement. Our results have revealed the nucleophile in the second reaction step, leading to the extension structural details of a GH70 glucansucrase. In contrast to the of an α-glucan at its nonreducing end by one glucose residue. previously proposed linear arrangement of domains along the polypeptide chain (7, 8), the structure of L. reuteri 180 GTF180- Binding Modes of Natural Acceptor Oligosaccharides. To analyze ΔN shows that four of the five domains are built up from discon- the acceptor specificity of the enzyme we carried out automatic tiguous parts of the polypeptide chain. Only the C-domain is

Vujičić-Žagar et al. PNAS Early Edition ∣ 3of6 Downloaded by guest on September 24, 2021 Fig. 3. Docking of isomaltose (A), isomaltotriose (B), and nigerose (C) in the active site of a modeled glucosyl-GTF180 intermediate. For modelling and docking procedure see SI text. The covalently bound glucosyl moiety is shown in white stick representation, isomaltose, nigerose, and isomaltotriose are shown in yellow stick representation. Dashed lines show hydrogen bonding distances of less than 3.5 Å, solid yellow lines show where the chain may extend; subsites are numbered with Arabic and Roman white numerals. Domains A and B of GTF180 are shown in surface presentation and are coloured blue and green, respectively.

formed from one continuous stretch of amino acids. Neverthe- the binding of longer oligosaccharides (see Fig. 2A and less, heterologous expression of the enzyme in Escherichia coli Video S1). These residues are strictly conserved in GH70 family is not problematic, showing that folding intermediates are stable enzymes, indicating that the binding of a single glucose residue in and that the N-terminal variable domain, which was absent in subsite −1 is a general property of GH70 glucansucrases. In the our construct, is not needed for correct folding of the enzyme. GH13 amylosucrase from N. polysaccharea, which produces an The structure confirms that, compared to GH13 α-amylases, αð1 → 4Þ-linked glucan polymer from sucrose, the glucose-bind- the catalytic ðβ∕αÞ8-barrel domain is circularly permuted as was ing pocket is also closed off, but by a salt-bridge between R509 previously proposed (10). and D144 (31). The “permutation per duplication model” proposes a se- Six residues that interact with the −1 glucose in GTF180-ΔN quence of gene arrangements that could have led to a circularly are strictly conserved in GH70 and GH13 enzymes, and exhibit permuted gene via gene duplication, in-frame fusion, and partial comparable interactions with the −1 glucose residue (see Results). truncation (28, 29). In this way a GH70 glucansucrase precursor In addition, Q1509 has a similar hydrogen bonding interaction consisting of a circularly permuted domain A, a discontiguous do- with the O6 hydroxyl group of the −1 glucose as the equivalent main B, and a contiguous domain C could have arisen from the His in the GH13 family. Thus, the active site of GTF180-ΔNis GH13 α-amylase fold (Fig. 4). Insertion of this precursor into the very similar to those of GH13 family members. No space is avail- gene of domain IV, followed by insertion into the gene encoding able for a glucose covalently bound to D1136, as required for the domains Vand N, could then have led to the present day protein. two-nucleophile mechanism proposed by Robyt et al. (16). Alternatively, domain IV may have inserted into the ancestor The similarity of the active sites of GH70 and GH13 enzymes α-amylase before the circular permutation occurred. In both suggests a very similar reaction mechanism for the cleavage of the cases various intermediate forms must have existed during evolu- α-glycosidic bond of sucrose and the formation of the β-glucosyl- tion. The requirement that such intermediate forms are stable enzyme intermediate. Cleavage of the glycosidic bond of sucrose and should offer some form of advantage to the organism to pre- is likely started by protonation of the glycosidic oxygen by E1063, vent removal from the gene pool, can explain why the number of resulting in bond cleavage, release of fructose, and formation of a circular permutations observed in proteins is rather small (30). partly planar oxocarbenium ion-like intermediate (12, 14, 24, 32). Conceivably, D1136 pushes the −1 O2 atom towards the inter- Active Site and Formation of the Covalent Glucosyl-Enzyme Intermedi- mediate position, while Q1509 holds the −1 glucose O6 atom. ate. The active site of GTF180-ΔN is a pocket-shaped cavity, in Similar to GH13 enzymes, the C2-O2 bond of the −1 glucose re- which the glucose moiety of sucrose binds (subsite −1). The cavity sidue is polarized by interactions of the O2 atom with H1135, is closed off by residues Q1140, N1411, and D1458, which prevent D1136, and R1023, which lowers the energy of the transition state

4of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1007531107 Vujičić-Žagar et al. Downloaded by guest on September 24, 2021 Fig. 4. Hypothetical evolutionary pathway based on the “permutation per duplication model” (28) leading to the unusual domain organization of GH70 family GTF180 starting from a putative GH13 α-amylase precursor. Sequence segments forming domains A, B, C, IV, and V are colored as before in blue, green, magenta, yellow, and red, respectively.

(14, 22, 33, 34). The reaction is completed by nucleophilic attack Conclusions of D1025 on the glucose C1 atom, which results in a covalent The crystal structures of GTF180-ΔN have revealed the molecu- intermediate with the glucosyl moiety attached to D1025 via a lar details of a GH70 enzyme and its interaction with sucrose and

β-glycosidic linkage. Because the þ1 fructosyl moiety of sucrose the acceptor maltose. The data show that, in contrast to previous BIOCHEMISTRY has fewer interactions with the enzyme than the −1 glucosyl moiety, proposals, only a single active site is present, and no space for it can easily be released from the active site in order to allow the a second covalently bound glucose residue or glucan polymer þ1 reaction to proceed by binding the acceptor molecule at subsite . is available. This finding invalidates the proposal of a double ac- tive-site/double-nucleophile insertion mechanism (16). Also the Transfer of Glucose to the Acceptor Saccharide and Product Specifi- conservation of key amino acids involved in substrate binding city. The next step in the reaction mechanism is the transfer of and catalysis at subsite −1 in both GH13 and GH70 enzymes the covalently bound glucosyl residue to the acceptor saccharide. Δ confirms that the mechanism of glycosidic bond cleavage by GTF180- N produces a branched built up from different α lengths of αð1 → 6Þ-linked glucose residues, interconnected by GH70 glucansucrases is similar to that of the -amylase super- single αð1 → 3Þ-bridges. No consecutive αð1 → 3Þ glucosidic family enzymes (14, 23). Δ bonds are present (35). The enzyme must thus be able to transfer The crystal structure of GTF180- N with bound maltose the glucosyl residue to either the O6 or the O3 hydroxyl group. revealed how an acceptor substrate binds to receive a glucosyl This requirement implies that the accepting glucose must be able group at its O6 atom. Modelling showed that transfer of the glu- to bind in two different orientations, either with its O6 or with its cosyl group to the O3 atom requires an entirely different binding O3 hydroxyl group pointing towards the catalytic site. Our dock- mode (Fig. 3B). ing studies showed that these two binding modes are indeed pos- Finally, in view of the conservation of the residues involved in sible (Fig. 3). At subsite þ2 the acceptor sugars are mostly bound subsite −1 in enzymes from families GH13, GH70, and GH77, the via aspecific stacking interactions with W1065; therefore the en- design of specific anticaries glucansucrase-inhibitors is predicted zyme is not very particular about the specific glycosidic linkage to require the exploitation of features of the enzymes more between the sugars in subsites þ1 and þ2, and can accommodate remote from subsite −1. αð1 → 3Þ, αð1 → 4Þ,orαð1 → 6Þ-linked . In all cases the accepting O6 hydroxyl group will be near the C1 atom of the Experimental Procedures covalent intermediate. Cloning, Mutagenesis, Purification, Crystallization, and Soaking In case the O3 hydroxyl group is the acceptor, a different bind- Experiments. For cloning and mutagenesis of gtf180 see SI text. ing mode of the acceptor is needed, as exemplified by the docking Purification and crystallization of native L. reuteri GTF180-ΔN results with isomaltotriose (Fig. 3B). In this case, the reducing were performed as described previously (17). Selenomethionine ” þ2 end glucose binds in subsite II rather than subsite , and (SeMet) labeled GTF180-ΔN was produced as formerly reported þ ’ the nonreducing end extends towards subsite II . In this binding (36). Both the D1025N mutant and SeMet labeled GTF180-ΔN αð1 → 3Þ mode an branch point is introduced in the glucan chain. crystallized in the same conditions as the native protein. Crystals Although our docking studies with isomaltose did not provide evi- of the D1025N mutant were soaked for 30 min in mother liquor dence, it can be envisaged that occasionally also a terminal (25% ðw∕vÞ PEG 3350, 50 mM NaCl, 50 mM 1,3-bis[tris(hydro- αð1 → 6Þ-linked glucan may bind in subsites þ1 and þII”, with O3 pointing into the active site. Such a binding mode would allow xymethyl)methylamino]propane HCl, pH 6.0, 2 mM CaCl2) linear glucan chain elongation with an αð1 → 3Þ-linked glucose. containing 25 mM sucrose, and cryoprotected in mother liquor 35% ð ∕ Þ The docking results for both isomaltose and nigerose suggest containing w v PEG 3350 and 25 mM sucrose. For mal- Δ that the most favorable glucan conformer is the one allowing tose soaking studies a crystal of SeMet-GTF180- N was trans- formation of the αð1 → 6Þ linkage. Indeed, the majority (69%) ferred first to mother liquor supplemented with 25 mM maltose of the glycosidic bonds in the α-D-glucan synthesized by (15 min), and then soaked for 2 h in mother liquor containing GTF180 from sucrose are αð1 → 6Þ-bonds and only about 31 % 250 mM maltose. Crystals were cryoprotected in 35% ðw∕vÞ are αð1 → 3Þ-bonds (18, 35). PEG 3350 and 250 mM maltose.

Vujičić-Žagar et al. PNAS Early Edition ∣ 5of6 Downloaded by guest on September 24, 2021 Data Collection. All datasets were collected at 100 K. Multiple Model building and corrections were done using the program wavelength anomalous dispersion (MAD) and “native” datasets COOT (42). The final native model consists of 1,006 amino acid were collected at beam line ID29 at the European Synchrotron residues (residues 746–1,751). The first 4 (residues 742–745) and Radiation Facility (ESRF), Grenoble, France and processed the last 21 residues (residues 1,752–1,772) are invisible in the using MOSFLM and SCALA (37). The “sucrose” and “maltose” electron density maps. Both sucrose and maltose were included datasets were collected at beam lines BM16, ESRF, Grenoble, in the model in the final stages of the refinement. The resulting France, and BW7A, European Molecular Biology Laboratory structures were validated with MolProbity (43). For refinement (EMBL) outstation at the Deutsches Elektronen-Synchrotron, statistics see Table S2. Coordinates and structure factors have Hamburg, Germany, respectively. The latter two datasets were been deposited with the Protein Data Bank (entries 3KLK, processed using the XDS package and SCALA (37, 38). All crys- 3HZ3, 3KLL). Figures were created using PyMol (44). tals belonged to space group P1, with one molecule per unit cell and 52% solvent content. The presence of one molecule per unit Product Analysis. Products synthesized by GTF180-ΔN were ana- cell is in agreement with gel filtration (Fig. S6) and dynamic light lyzed as described in SI text. scattering experiments (estimated molecular mass is 110 kDa with 13.1% polydispersity), which indicate that the protein is Docking Studies. The automatic docking calculations were done monomeric in solution. For data collection and processing statis- with the program AutoDock 4 (45). More details can be found tics see Table S1. in SI text.

Structure Determination. The GTF180-ΔN structure was solved ACKNOWLEDGMENTS. We are grateful to staff scientists and local contacts with the program SOLVE/RESOLVE (39) using MAD data col- at beam lines ID29 and BM16 at the ESRF, Grenoble, France, and BW7A at lected around the Se edge. The experimental phase information the EMBL-Hamburg outstation (DESY), Germany for their support during data collection. The authors thank M. Tietema for construction of GTF180- after RESOLVE was combined with native 1.7 Å data and used ΔN W1531S mutant and TLC analysis. This work was financially supported for further automated model building and refinement using the by Senter Innovatiegerichte Onderzoeksprogramma’s, The Netherlands programs ARP/wARP (40) and REFMAC5 (41), respectively. (IGE01021).

1. Marsh PD, Nyvad B (2008) Dental caries: the disease and its clinical management. 23. Jensen MH, et al. (2004) Crystal structure of the covalent intermediate of amylosucrase (Blackwell, Oxford), 2nd ed,, pp 163–187. from Neisseria polysaccharea. Biochemistry 43:3104–3110. 2. Featherstone J (2008) Dental caries: a dynamic disease process. Aust Dent J 53:286–291. 24. Brzozowski AM, Davies GJ (1997) Structure of the Aspergillus oryzae α-amylase 3. Colby SM, McLaughlin RE, Ferretti JJ, Russell RR (1999) Effect of inactivation of gtf complexed with the inhibitor acarbose at 2.0 Ångstrom resolution. Biochemistry genes on adherence of Streptococcus downei. Oral Microbiol Immun 14:27–32. 36:10837–10845. 4. Colby SM, Russell RRB (1997) Sugar metabolism by mutans streptococci. J Appl Micro- 25. Koepsell HJ, et al. (1953) Enzymatic synthesis of dextran—acceptor specificity and biol 83:80S–88S. chain initiation. J Biol Chem 200:793–801. 5. Rölla G (1989) Why is sucrose so cariogenic? The role of glucosyltransferase and 26. Richard GT, et al. (2003) Glucosylation of α-butyl- and α-octyl-D-glucopyranosides by polysaccharides. Scand J Dent Res 97:115–119. dextransucrase and from Leuconostoc mesenteroides. Carbohydr Res 6. Wright WG, Thelwell C, Svensson B, Russell RRB (2002) Inhibition of catalytic and 338:855–864. glucan-binding activities of a streptococcal GTF forming insoluble glucans. Caries 27. Robyt JF, Eklund SH (1983) Relative, quantitative effects of acceptors in the reaction – Res 36:353–359. of Leuconostoc mesenteroides B-512F dextransucrase. Carbohydr Res 121:279 286. 7. Monchois V, Willemot RM, Monsan P (1999) Glucansucrases: mechanism of action and 28. Peisajovich SG, Rockah L, Tawfik DS (2006) Evolution of new protein topologies – structure-function relationships. FEMS Microbiol Rev 23:131–151. through multistep gene rearrangements. Nat Genet 38:168 174. 8. van Hijum SAFT, et al. (2006) Structure-function relationships of glucansucrase and 29. Vogel C, Morea V (2006) Duplication, divergence and formation of novel protein – fructansucrase enzymes from lactic acid bacteria. Microbiol Mol Biol R 70:157–176. topologies. BioEssays 28:973 978. 9. Henrissat B, Bairoch A (1996) Updating the sequence-based classification of glycosyl 30. Uliel S, Fliess A, Unger R (2001) Naturally occurring circular permutations in proteins. – hydrolases. Biochem J 316:695–696. Protein Eng 14:533 542. α 10. MacGregor EA, Jespersen HM, Svensson B (1996) A circularly permuted α-amylase- 31. Skov LK, et al. (2001) Amylosucrase, a glucan-synthesizing enzyme from the -amylase family. J Biol Chem 276:25273–25278. type α/β-barrel structure in glucan-synthesizing glucosyltransferases. FEBS Lett 32. Mosi R, et al. (1998) Reassessment of acarbose as a transition state analogue inhibitor 378:263–266. of cyclodextrin glycosyltransferase. Biochemistry 37:17192–17198. 11. Barends TRM, et al. (2007) Three-way stabilization of the covalent intermediate in 33. Braun C, Brayer GD, Withers SG (1995) Mechanism-based inhibition of yeast α-gluco- amylomaltase, an α-amylase-like transglycosylase. J Biol Chem 282:17242–17249. sidase and human pancreatic α-amylase by a new class of inhibitors—2-deoxy-2,2-di- 12. Mosi R, et al. (1997) Trapping and characterization of the reaction intermediate in fluoro-α-glycosides. J Biol Chem 270:26778–26781. cyclodextrin glycosyltransferase by use of activated substrates and a mutant enzyme. 34. McCarter JD, Adam MJ, Withers SG (1992) Binding energy and catalysis—fluorinated Biochemistry 36:9927–9934. and deoxygenated glycosides as mechanistic probes of Escherichia coli (LacZ) β-galac- 13. Moulis C, et al. (2006) Understanding the polymerization mechanism of glycoside- tosidase. Biochem J 286:721–727. hydrolase family 70 glucansucrases. J Biol Chem 281:31254–31267. 35. van Leeuwen SS, et al. (2008) Structural analysis of the α-D-glucan (EPS180) produced 14. Uitdehaag JCM, et al. (1999) X-ray structures along the reaction pathway of cyclodex- by the Lactobacillus reuteri strain 180 glucansucrase GTF180 enzyme. Carbohydr Res trin glycosyltransferase elucidate catalysis in the α-amylase family. Nat Struct Biol 343:1237–1250. 6:432–436. 36. Doublié S (1997) Preparation of selenomethionyl proteins for phase determination. 15. Monchois V, Willemot RM, Monsan P (1999) Glucansucrases: mechanism of action and Method Enzymol 276:523–530. – structure-function relationships. FEMS Microbiol Rev 23:131 151. 37. Collaborative Computational Project Number 4 (1994) The CCP4 Suite—programs for 16. Robyt JF, Yoon SH, Mukerjea R (2008) Dextransucrase and the mechanism for dextran protein crystallography. Acta Crystallogr D 50:760–763. – biosynthesis. Carbohydr Res 343:3039 3048. 38. Kabsch W (1993) Automatic processing of rotation diffraction data from crystals of 17. Pijning T, et al. (2008) Biochemical and crystallographic characterization of a initially unknown symmetry and cell constants. J Appl Crystallogr 26:795–800. – glucansucrase from Lactobacillus reuteri 180. Biocatal Biotransform 26:12 17. 39. Terwilliger TC, Berendzen J (1999) Automated MAD and MIR structure solution. Acta 18. Kralj S, et al. (2004) Glucan synthesis in the genus Lactobacillus: isolation and Crystallogr D 55:849–861. characterization of glucansucrase genes, enzymes and glucan products from six dif- 40. Perrakis A, Morris R, Lamzin VS (1999) Automated protein model building combined ferent strains. Microbiology 150:3681–3690. with iterative structure refinement. Nat Struct Biol 6:458–463. 19. Giffard PM, Jacques NA (1994) Definition of a fundamental repeating unit in strep- 41. Murshudov GN, Vagin AA, Dodson EJ (1997) Refinement of macromolecular structures tococcal glucosyltransferase glucan-binding regions and related sequences. J Dent by the maximum-likelihood method. Acta Crystallogr D 53:240–255. Res 73:1133–1141. 42. Emsley P, Cowtan K (2004) COOT: model-building tools for molecular graphics. Acta 20. Fernández-Tornero C, et al. (2001) A novel solenoid fold in the cell wall anchoring Crystallogr D 60:2126–2132. domain of the pneumococcal virulence factor LytA. Nat Struct Biol 8:1020–1024. 43. Davis IW, et al. (2007) MolProbity: all-atom contacts and structure validation for 21. Davies GJ, Wilson KS, Henrissat B (1997) Nomenclature for sugar-binding subsites in proteins and nucleic acids. Nucleic Acids Res 35:W375–383. glycosyl hydrolases. Biochem J 321:557–559. 44. DeLano WL (2002) The PyMOL molecular graphics system (DeLano Scientific LLC, Palo 22. Uitdehaag JCM, van der Veen BA, Dijkhuizen L, Dijkstra BW (2002) Catalytic Alto, CA). mechanism and product specificity of cyclodextrin glycosyltransferase, a prototypical 45. Morris GM, et al. (1998) Automated docking using a Lamarckian genetic algorithm transglycosylase from the α-amylase family. Enzyme Microb Tech 30:295–304. and empirical binding free energy function. J Comput Chem 19:1639–1662.

6of6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1007531107 Vujičić-Žagar et al. Downloaded by guest on September 24, 2021