Báo cáo khoa học: Structural basis for substrate recognition by Erwinia chrysanthemi GH30 glucuronoxylanase potx

The cleavage of the xylan main chain by GH10 xylanases Keywords crystal structure with ligand; Erwinia chrysanthemi; GH30; glucuronoxylan-specific xylanase; substrate recognition Corresp

Structural basis for substrate recognition by

Erwinia chrysanthemi GH30 glucuronoxylanase

Lˇubica Urba´nikova´1, Ma´ria Vrsˇanska´2, Kristian B R Mørkeberg Krogh3, Tine Hoff3and Peter Biely2

1 Institute of Molecular Biology, Slovak Academy of Sciences, Bratislava, Slovakia

2 Institute of Chemistry, Center of Glycomics, Slovak Academy of Sciences, Bratislava, Slovakia

3 Novozymes A ⁄ S, Bagsvaerd, Denmark

Introduction

The important industrial enzyme endo-b-1,4-xylanase

(EC 3.2.1.8) has been placed into several glycoside

hydrolase (GH) families on the basis of hydrophobic

cluster analysis, 3D, and mode of action [1]

(carbohy-drate-active enzymes server at http://www.cazy.org)

The best-characterized xylanases belong to GH

fami-lies 10 and 11 These enzymes do not seem to be

spe-cialized for hydrolysis of a particular xylan, because

they are capable of degrading hardwood acetyl glucu-ronoxylans, cereal arabinoxylans, and even algal b-1,4-b-1,3-xylan (rhodymenan) [2–4] The activity of xylanases belonging to these two families does not appear to be dependent on the type of side chain dec-orations of the xylan main chain, but is strongly dependent on the density of substituents [2,5] The cleavage of the xylan main chain by GH10 xylanases

Keywords

crystal structure with ligand;

Erwinia chrysanthemi; GH30;

glucuronoxylan-specific xylanase; substrate

recognition

Correspondence

P Biely, Institute of Chemistry, Center of

Glycomics, Slovak Academy of Sciences,

Du´bravska´ cesta 9, SK-845 38 Bratislava,

Slovakia

Fax: +421 2 5941 0222

Tel: +421 2 5941 0275

E-mail: chempbsa@savba.sk

(Received 17 December 2010, revised 10

April 2011, accepted 13 April 2011)

doi:10.1111/j.1742-4658.2011.08127.x

Xylanase A from the phytopathogenic bacterium Erwinia chrysanthemi is classified as a glycoside hydrolase family 30 enzyme (previously in family 5) and is specialized for degradation of glucuronoxylan The recombinant enzyme was crystallized with the aldotetraouronic acid b-D -xylopyranosyl-(1fi 4)-[4-O-methyl-a-D-glucuronosyl-(1fi 2)]-b-D-xylopyranosyl-(1fi

4)-D-xylose as a ligand The crystal structure of the enzyme–ligand complex was solved at 1.39 A˚ resolution The ligand xylotriose moiety occupies sub-sites)1, )2 and )3, whereas the methyl glucuronic acid residue attached to the middle xylopyranosyl residue of xylotriose is bound to the enzyme through hydrogen bonds to five amino acids and by the ionic interaction

of the methyl glucuronic acid carboxylate with the positively charged guan-idinium group of Arg293 The interaction of the enzyme with the methyl glucuronic acid residue appears to be indispensable for proper distortion of the xylan chain and its effective hydrolysis Such a distortion does not occur with linear b-1,4-xylooligosaccharides, which are hydrolyzed by the enzyme at a negligible rate

Database Structural and experimental data are available in the Protein Data Bank database under accession number 2y24 [45].

Abbreviations

GH, glycoside hydrolase; GlcA, D -glucuronic acid; MeGlcA, 4-O-methyl- D -glucuronic acid; MeGlcA 2 Xyl2, 4-O-methyl-a- D -glucuronosyl-(1 fi 2)-b- D -xylopyranosyl-(1 fi 4)- D -xylose; MeGlcA 2 Xyl3, b- D -xylopyranosyl-(1 fi 4)-[4-O-methyl-a- D -glucuronosyl-(1 fi 2)]-b- D -xylopyranosyl-(1 fi 4)- D -xylose; MeGlcA3Xyl 3 , 4-O-methyl-a- D -glucuronosyl-(1 fi 2)-b- D -xylopyranosyl-(1 fi 4)-b- D -xylopyranosyl-(1 fi 4)- D -xylose; MeGlcA3Xyl 4 , b- D -xylopyranosyl-(1 fi 4)-[4-O-methyl-a- D -glucuronosyl-(1 fi 2)]-b- D -xylopyranosyl-(1 fi 4)-b- D -xylopyranosyl-(1 fi 4)- D -xylose; MeXyl 3 Xyl3, 4-O-methyl-a- D -glucuronosyl-(1 fi 2)-b- D -xylopyranosyl-(1 fi 4)-b- D -xylopyranosyl-(1 fi 4)- D -xylose; MeXyl 3 Xyl4, b- D -xylopyranosyl-(1 fi 4)-[4-O-methyl-a- D -glucuronosyl-(1 fi 2)]-b- D -xylopyranosyl-(1 fi 4)-b- D -xylopyranosyl-(1 fi 4)- D -xylose; VS, virtual screening; Xyl, xylose; XynA, Erwinia chrysanthemi GH30 xylanase.

requires at least two consecutive unsubstituted

xylo-pyranosyl residues, whereas hydrolysis by GH11

xylanases requires three consecutive unsubstituted

xylopyranosyl residues [2–6] Heavily substituted

xylan, such as corn fiber xylan [7], is completely

resis-tant to the action of members of these two xylanase

families (P Biely, unpublished results) An interesting

endoxylanase, classified in GH family 8, was found to

be produced by an Antarctic bacterium,

Pseudoaltero-monas haloplanktis [8] This enzyme showed the

high-est activity on rhodymenan [8], which indicates that

the enzyme might be specialized for hydrolysis of the

linear xylan present in algae

Unique xylanases are found in GH family 30 [1,9]

These enzymes were originally classified in GH

fam-ily 5 Some bacterial GH30 xylanases are specialized

for the hydrolysis of xylans that contain d-glucuronic

acid (GlcA) or 4-O-methyl-d-glucuronic acid

(MeG-lcA) side residues However, not all GH30 xylanases

show this specificity The recently described GH30

xy-lanase from the fungus Bispora sp does not show such

a requirement for these side residues [10] With

Bacil-lus subtilis GH30 xylanase and Erwinia chrysanthemi

GH30 xylanase (XynA), it was clearly demonstrated

that cleavage of the xylan main chain is dependent on

the presence of MeGlcA side residues [11–14] A

simi-lar enzyme from another Bacillus species was recently

described [15] The cleavage of the main xylan chain

takes place at the second glycosidic linkage from the

MeGlcA side group towards the reducing end of the

xylan chain The elucidation of the three-dimensional

structure of the E chrysanthemi GH30 enzyme [16],

together with its established mode of action [14],

allowed us to present a hypothesis for the basis of

sub-strate recognition in this group of so-called

‘append-age-dependent xylanases’ [11] Examination of the

structure of XynA [14] for the presence of aromatic

amino acids and positively charged amino acid groups

in the vicinity of the identified catalytic glutamic acids

(Glu253, nucleophile; Glu165, acid⁄ base) indicated

that the substituted xylopyranosyl residue should be

accommodated at the hypothetical subsite)2 Tyr290

and Trp289 near subsite )2 were considered to

consti-tute a suitable place for binding of MeGlcA However,

the space between the two aromatic amino acids was

too narrow to accommodate the uronic acid An ionic

interaction between the negatively charged MeGlcA

carboxylate and the positively charged Arg293

(pKa> 12) occurring in the vicinity of the

Tyr290⁄ Trp289 sandwich was also proposed to play an

important role in uronic acid binding [14] It became

clear that definite understanding of the recognition of

uronic acid by GH30 xylanases would require X-ray

crystallographic studies of the enzyme–ligand complex Preliminary data on the crystallization of the B subtilis GH30 xylanase have also been released, but as yet without a proper ligand [17]

Here we report an X-ray structure of the complex

of XynA with the aldotetraouronic acid b-d-xylopyr-anosyl-(1fi 4)-[4-O-methyl-a-d-glucuronosyl-(1 fi 2)]-b-d-xylopyranosyl-(1fi 4)-d-xylose (MeGlcA2Xyl3) (Fig S1) The ligand filled three of the hypothetical subsites on the glycone (subsites with negative designation) side of the substrate-binding site [18,19] Subsite)2 accommo-dates the xylopyranosyl residue substituted with MeG-lcA A detailed analysis of the enzyme–ligand complex confirmed the ionic interaction of the substrate carbox-ylate group with the enzyme Furthermore, it pointed

to a number of hydrogen bonds formed between the enzyme and its substrate

Results

Crystallization and data collection Electrophoretically homogeneous recombinant XynA was subjected to dynamic light scattering analysis before crystallization This method gives information

on the homogeneity and size distribution of particles

in solution [20] Despite a relatively high measured polydispersity (Fig S2), the protein crystallized rela-tively easily and produced high-quality crystals Attempts were made to obtain crystals of the pro-tein–MeGlcA2Xyl3 complex by diffusion of the ligand into pregrown crystals of the ligand-free enzyme or by cocrystallization Crystals of ligand-free XynA were obtained under several conditions, which were further optimized to give diffraction-quality crystals Crystals

of two distinct habits were obtained (Fig S3A,B); however, they were found to belong to the same P3221 space group

Crystals of the XynA–MeGlcA2Xyl3 complex were obtained by both methods tested; however, the best diffraction data were recorded with the crystal of the complex obtained by cocrystallization (Fig S3C) These data are reported here

The crystals of the complex belonged to the P3221 space group, with dimensions a = b = 59.578 A˚ and

c= 168.296 A˚, c = 120 Crystal symmetry, unit cell dimensions and the molecular mass of the protein gave

a Matthews coefficient of 2.05 and a 40% solvent con-tent in the crystal for one protein molecule in the asymmetric unit [21] Diffraction data statistics are shown inTable 1

For a comparison, the first structure of XynA, crys-tallized without any ligand, belonged to the monoclinic

P21 space group [16,22] The authors reported multiple

crystal forms, including hexagonal crystals with a

P6n space group and unit cell dimensions of

a= b = 60.32 A˚ and c = 165.78 A˚, which are very

close to the unit cell dimensions reported here In all

cases, only one protein molecule was found in the

asymmetric unit As expected, the crystal contacts in

the monoclinic and trigonal forms were different

Structure description

The structure of XynA in complex with MeGlcA2Xyl3

was solved by molecular replacement at resolution

1.39 A˚, with the original XynA crystal structure as a

search model (1NOF) [16] The final R-factor and Rfree

-factor were 12.2% and 16.9%, respectively The

refine-ment statistics are shown in Table 1 The model consists

of 383 amino acids (numbered 31–413 in the sequence),

a single MeGlcA2Xyl3ligand, an imidazole, three

mole-cules of poly(ethylene glycol), and 571 water molemole-cules MeGlcA2Xyl3, imidazole and poly(ethylene glycol) mol-ecules were modeled at later stages of refinement, when the electron density was unambiguous (Fig 1) [the poly(ethylene glycol) molecules are not shown]

The overall structure of XynA in the complex with MeGlcA2Xyl3 (Fig 2A,B) is nearly identical to the 1NOF structure described previously by Larson et al [16] The enzyme consists of a (b⁄ a)8-barrel catalytic domain and a b-sheet immunoglobulin-like C-terminal domain (a potential xylan-binding module) connected

by amino acids 45 and 317–322 One cis-peptide bond has been found between Val200 and Ala201

Superposition of the structure of the ligand-free enzyme with the structure of the enzyme in the complex using 378 CA atoms (CA atoms with two alternative conformations were omitted) resulted in root mean square, average and maximum xyz displacements

of 0.275 A˚, 0.241 A˚, and 0.849 A˚, respectively It is

Table 1 Data collection and refinement statistics R merge = P

hkl P

i |I i (hkl) – ÆI(hkl)æ| ⁄ P

hkl P

i I i (hkl), where I i (hkl) is the intensity measure-ment for the ith observation of reflection hkl and ÆI(hkl)æ is the average intensity for multiple measurements for this reflection.

R = P

||Fobs| ) |F calc || ⁄ P

|Fobs|, where Fobsand Fcalcare observed and calculated structure factor amplitudes A random subset (5%) of data excluded from the refinement was used for Rfreefactor calculation.

XynA–MeGlcA 2 Xyl3 Data collection

Unit cell dimensions

Refinement

Asymmetric unit content (No of molecules)

Protein ⁄ MeGlcA 2 Xyl3⁄ imidazole ⁄ poly(ethylene glycol) ⁄ water 1 ⁄ 1 ⁄ 1 ⁄ 3 ⁄ 571

B average (A˚2 )

Model quality

Ramachandran plot

)

Geometry

a Number of nonglycine and nonproline residues.

interesting that the side chain conformations of amino

acids interacting with MeGlcA2Xyl3 did not change as

a result of binding The superposition of both structures

also showed that the acetate anion observed in the first published structure (1NOF) [16] interacts with the posi-tively charged guanidinium group of Arg293 in a man-ner similar to the carboxylate group of MeGlcA2Xyl3 The protein substrate-binding site has a total area of 321.9 A˚2, and is composed of 17 amino acids; 44.4%

of the binding site surface is hydrophobic, i.e covered

by carbon atoms The remaining 55.6% is polar, cov-ered by nitrogen and oxygen atoms (Table S1; Fig 3A) Thirteen of the 17 amino acids form 174 van der Waals contacts and 10 hydrogen bonds with MeG-lcA2Xyl3 (Table 2; Fig 3B) The three xylose (Xyl) units of the ligand take part in the stacking interac-tions with the aromatic rings of Trp289, Tyr172 and Trp55 in subsites)1, )2, and )3, as shown in detail in Fig 4A,B The Xyl in subsite)1 is also coordinated with Trp113, Asn164, and the catalytic Glu165 and Glu253 (Fig 4B) The MeGlcA moiety interacts with the edges of the aromatic rings of the Trp289 and Tyr290 side chains, and also forms one hydrogen bond with the Trp289 amide nitrogen, NE1 The most important interaction for substrate recognition appears

to be an ionic interaction between the positively charged Arg293 guanidinium group and the negatively charged carboxylate of MeGlcA (Fig 4C)

An electron density found in the proximity of the catalytic amino acids Glu165 and Glu253 was ascribed

to imidazole, a component of the crystallization buffer Imidazole interacts with Tyr168 and Trp232, and is also electrostatically bound to the catalytic Glu165 (Fig 4D) Thus, imidazole appears to occupy sub-site +1, interacting with the Xyl or xylosyl residues of the enzyme-cleaved substrates Depending on the char-acter of the substrate, this Xyl becomes the product of hydrolysis or the nonreducing end of the leaving group A stereo view of the mode of binding of MeGlcA2Xyl3is shown inFig 5A The interactions of the enzyme with MeGlcA2Xyl3 and imidazole are sum-marized in Table 2

Binding energy calculations and ligand-docking studies

The energy of ligand binding was estimated with lead-finder [24] The scoring functions of leadfinder are based on a semiempirical molecular mechanical approach that explicitly accounts for various types of molecular interaction The DG-scoring is a measure of binding energy, and the virtual screening (VS) scoring corresponds to the ligand-binding potency

The experimentally determined structure of the pro-tein–MeGlcA2Xyl3 complex was used for calculating the binding energy at pH 5.5, which is the pH

MeGlcA2

Fig 1 MeGlcA 2 Xyl3and imidazole in the 2Fo) F c electron density

map (gray mesh), contoured at the 1.0 r level Atoms are shown as

sticks and colored as follows: C, green; O, red; N, blue Three Xyl

resi-dues with the MeGlcA moiety are bound in subsites )1, )2, and )3.

Fig 2 The arrangement of protein and ligand molecules in the

XynA–MeGlcA 2 Xyl3crystal asymmetric unit (A) A direct view of

the structure and (B) a view of the structure rotated 90 around the

y-axes, showing the active site of XynA The catalytic (b ⁄ a) 8 -barrel

domain is in red, the C-terminal b9-barrel domain is in blue, the

con-necting region is in orange, the catalytic amino acids Glu165 and

Glu253 are in ball-and-stick representations, and MeGlcA2Xyl 3 ,

imid-azole and three poly(ethylene glycol) molecules are in ball-and-stick

representations with bonds in bold green, yellow, and turquoise,

respectively.

optimum of the enzyme [14] Similar calculations were

performed for its three analogs For the first one,

b-d-xylopyranosyl-(1fi

4)-[4-O-methyl-a-d-xylopyranosyl-(1fi 2)]-b-d-xylopyranosyl-(1 fi 4)-d-xylose (MeXyl2

Xyl3), the carboxyl group of MeGlcA was replaced

by hydrogen; that is, MeGlcA was converted to

d-xylose For the second one,

4-O-methyl-a-d-glucuronosyl-(1fi 2)-b-d-xylopyranosyl-(1 fi

4)-d-xylose (MeGlcA2Xyl2), the nonreducing xylosyl

residue of the ligand was replaced by hydrogen The

third compound examined in this regard was Xyl3, the

core xylooligosaccharide

In addition to the above calculations, the Xyl

mono-mer was docked into the hypothetical subsite +1,

which is occupied by imidazole in the crystal structure

The program offered several different positions for Xyl bound in subsite +1 The position displayed in Fig 5B corresponds to the lowest DG and VS scores, and is also optimal from the structural point of view The Xyl O4 atom appears to be hydrogen bonded to the catalytic Glu165 and positioned in a relatively short distance (2.83 A˚) from the glycosidic oxygen (O1) of the reduc-ing-end xylosyl residue bound in subsite)1 (Fig 5B) The results of the binding energy calculations and molecular docking are summarized inTable 3 The dif-ference between the binding energies of MeGlcA2Xyl3 and its virtual analog MeXyl2Xyl3 indicates that the ionic interaction of the ligand carboxyl group with Arg293 corresponds to about 36% of the total binding energy of MeGlcA2Xyl3 ()2.29 kcalÆmol)1 versus

Table 2 Protein–MeGlcA 2 Xyl3and protein–imidazole interactions The numbers in parentheses correspond to Xyl-binding subsites.

MeGlcA2Xyl 3

Imidazole

Fig 3 Details of the interactions of XynA

with MeGlcA 2 Xyl 3 and imidazole (A) Stick

representation of MeGlcA 2 Xyl3and

imidaz-ole (atoms: green, C; red, O; blue, N).

Amino acids involved into substrate binding

are in ball-and-stick representations (atoms:

gray, C; red, O; blue, N) Hydrogen bonds

are marked by dashed lines (B) The van der

Waals surface representation of the

enzyme, showing the active site cleft filled

by MeGlcA 2 Xyl3and imidazole.

)6.22 kcalÆmol)1) The sum of both ionic and nonionic

enzyme–MeGlcA interactions corresponds to about

55% of the total binding energy ()3.47 kcalÆmol)1

ver-sus )6.22 kcalÆmol)1) The binding energy of the

non-reducing xylosyl residue in subsite)3 is only 9% of

the total DG ()0.55 kcalÆmol)1 versus )6.22

kcalÆ-mol)1) The calculated binding energies for the ligand,

its two virtual analogs and Xyl3 (Table 3) correspond

to specific enzyme activities on different substrates (see

below)

Specific activity on aldouronic acids and linear

xylooligosaccharides

Two aldotetraouronic acids differing in the presence of

the nonreducing xylopyranosyl residue filling subsite)3

were available: 4-O-methyl-a-d-glucuronosyl-(1fi

2)-b-d-xylopyranosyl-(1fi 4)-b-d-xylopyranosyl-(1 fi 4)-d-xylose (MeGlcA3Xyl3), the shortest acidic oligosac charide liberated from glucuronoxylan by

endoxylanas-es of GH10 [4], and aldopentaouronic acid, b-d-xylo pyranosyl-(1fi 4)-[4-O-methyl-a-d-glucuronosyl-(1 fi 2)]-b-d-xylopyranosyl-(1fi b-d-xylopyranosyl-(1 fi

4)-d-xylose (MeGlcA3Xyl4), the shortest acidic oligosaccha ride liberated from glucuronoxylan by endoxylanases of

GH family 11 [4] (Fig S1) XynA hydrolyzed the ald-opentaouronic acid more efficiently Specific activities at

4 mm substrate were 42 mmolÆmin)1Æmg)1for the pent-amer and 13 mmolÆmin)1Æmg)1 for the tetramer These data suggest that subsite)3 also contributes to sub-strate binding In view of the recent information that

a Bacillus GH30 xylanase shows activity on linear b-1,4-xylooligosaccharides [15], we also examined the rate of hydrolysis of xylotetraose and xylopentaose

Fig 4 Detailed view of the interaction of the enzyme with individual carbohydrate residues of MeGlcA 2 Xyl 3 derived from the enzyme–ligand complex with imidazole bound in aglycone subsite +1 Amino acids and ligands are in ball-and-stick representations, with sticks colored gold and green, respectively The atoms are colored as follows: red, O; blue, N; gold and green, C Hydrogen bonds are marked by dashed lines Stacking interactions are also highlighted as dashed lines connecting the centers of interacting groups marked by asterisks The distances are in A ˚ (A) Stacking interactions of Tyr172 and Trp55 with xylosyl residues in subsites )2 and )3 (B) Hydrogen bonds between the enzyme and xylosyl residue in subsite )1 The stacking interaction with Trp289 is also indicated (C) Coordination of the MeGlcA residue of the ligand with Tyr255, Ser258, Trp289, Arg293, and Tyr295 There is no stacking interaction of MeGlcA with the sandwich of Trp289 ⁄ Tyr290 (D) Stacking interactions of imidazole in subsite +1 with Trp168 and Tyr232, and its hydrogen bond with the catalytic Glu165 The six-membered aromatic ring of Trp168 and Leu204 might be involved in binding of Xyl in subsite +2.

At 4 mm, both oligomers served as enzyme substrates,

but with specific activities three orders of magnitude

lower than those on aldouronic acids These

observa-tions point again to a crucial role for the MeGlcA

carboxylate in enzyme substrate recognition and the role

of MeGlcA as an essential specificity determinant We

should mention that xylopentaose was hydrolyzed about

three times faster than xylotetraose, which is also in

accord with the results of the docking experiments and calculated binding energies (Table 3)

Discussion

The xylanase investigated in this work is one of the appendage-dependent endoxylanases, which are of bacterial origin and can be found in the GH30

A

B

Fig 5 Stereoview of the interactions of XynA with MeGlcA 2 Xyl3and imidazole (IMD) (A) Enzyme–MeGlcA 2 Xyl3interactions (for clarity, Ser258, forming a hydrogen bond to MeGlcA, is not shown) Ligands and amino acids involved in ligand binding are in ball-and-stick represen-tations (atoms: black, C; red, O; blue, N) with sticks colored green and light gray, respectively Hydrogen bonds and ionic interactions are marked by dashed lines Glu253 is marked by an asterisk (B) Interactions of the enzyme with Xyl docked at subsite +1 For comparison, Xyl (derived from the MeGlcA 2 Xyl3structure) in subsite )1 and imidazole are also shown The length of the hydrogen bonds is indicated in A˚.

Table 3 Summary of molecular modeling experiments and binding energy calculation.

Binding energy,

DG (kcalÆmol)1) VS score

Difference in binding energies, DG 1 ) DG 2

Functional group

of ligand

MeGlcA2Xyl 2 Crystal structure )5.67 )9.18 MeGlcA2Xyl 3 – MeGlcA2Xyl 2 )0.55 Xyl at subsite )3

a COOH group of MeGlcA was replaced by hydrogen b Nonreducing Xyl was replaced by hydrogen.

(formerly GH5) family [1,8,13–15] These enzymes are

specialized for depolymerization of xylans that contain

GlcA or MeGlcA side substituents They exhibit a

unique mode of action The cleavage of the

glucuron-oxylan main chain takes place exclusively at the second

glycosidic linkage from the branch towards the

reduc-ing end of the polysaccharide chain In other words,

the cleavage occurs during the formation of the

pro-ductive enzyme–substrate complex, in which the

substi-tuted xylopyranosyl residue is bound in the

hypothetical subsite)2 In this way, the MeGlcA or

GlcA residues determine the site of substrate cleavage,

and the content of these uronic acids determines the

xylan chain cleavage frequency In this work, we

con-firm the hydrolysis of linear xylooligosaccharides by a

GH30 xylanase [15]; however, the rate of their

hydro-lysis with XynA was negligible in comparison with the

rate of hydrolysis of aldouronic acids

After the 3D structure of the enzyme became

known [16] and the mode of GH30 xylanase action

had been elucidated [13,14], a question emerged

con-cerning the basis for the recognition of the MeGlcA

and GlcA residues by the enzyme We have

postu-lated an ionic interaction between the uronic acid

car-boxylate and the positively charged Arg293 occurring

in the vicinity of a sandwich of two aromatic amino

acids, Tyr290⁄ Trp289, that could interact with the

uronic acid [14] However, because the space between

Tyr290 and Trp289 in the published crystal structure

was too narrow to accommodate the uronic acid, it

became clear that the enzyme should be crystallized

in a complex with a suitable ligand and that the

structure of the complex could provide the required

information

We have succeeded obtaining crystals of XynA with

the aldotetraouronic acid MeGlcA2Xyl3, which is a

product of the cleavage of MeGlcA3Xyl4 by the same

enzyme In the crystal structure, the ligand was found

to be bound in a manner similar to the one that we

have predicted [14] The xylopyranosyl residue

substi-tuted by MeGlcA was bound in subsite)2, and

MeG-lcA was in a position that clearly indicates an ionic

interaction between its carboxyl group and the

posi-tively charged Arg293 However, MeGlcA was not

sandwiched between Tyr290 and Trp289, as proposed

earlier [14] Instead, in addition to the ionic interaction

with Arg293, it interacts with the side chains of the

aromatic amino acids Tyr255, Trp289, and Tyr295,

and with Ser258, through several hydrogen bonds,

which are listed in Table 2 and shown in Figs 4C and

5A

It is interesting that the ligand occurs in the complex

with XynA in the form of its a-anomer Such a

config-uration corresponds to the enzyme a-glycosyl ester intermediate with the catalytic glutamate Glu165 This

is interesting in light of the fact that the enzyme is a retaining GH [1] At this stage of our work, we do not have any explanation for this observation

An important question to be answered in connec-tion with the mode of acconnec-tion of GH30 xylanases is why the enzymes do not efficiently attack linear b-1,4-linked xylooligosaccharides The MeGlcA carbox-ylate is involved in binding by Arg293 According to the calculations of the binding energies of the ligand and its virtual analogs (Table 3), the interaction of the enzyme with MeGlcA is stronger than with the xylopyranosyl residues in the negatively numbered subsites The ionic interaction could also be impor-tant for the first contact of the enzyme with sub-strates, and also indispensable for creating a stable enzyme–substrate complex In the next steps, the enzyme–substrate complex formation could be based

on stacking interactions between aromatic amino acids covering the enzyme binding site and Xyl resi-dues of the xylan main chain The final step could be the locking of the substrate, namely MeGlcA and a xylosyl or xylobiosyl moiety, at subsites +1 and +2,

in a proper position for cleavage The importance of Xyl binding at subsite +1 is supported by calcula-tions of the binding energy of free Xyl in subsite +1 (Table 3; Fig 5B) One can envisage strong bending

of the xylan chain as a consequence of both ionic and stacking interactions This apparently cannot occur with linear oligosaccharides or a xylan main chain that is either unsubstituted or carries uncharged side substituents such as l-arabinose The strong bending could be the reason why the enzyme hardly recognizes linear xylooligosaccharides as substrates and does not attack arabinoxylan [14] To learn more about the enzyme–substrate interactions, complexes of the enzyme with larger, nonhydrolyzable ligands should be crystallized and their structure elucidated

An alternative approach could include the preparation

of inactive enzyme mutants and crystallization of these mutants with natural substrates

Conclusions

The crystal structure of XynA with MeGlcA2Xyl3 shows that the unique substrate specificity and mode

of action of bacterial GH30 xylanases on xylans with MeGlcA and GlcA side substituents is achieved mainly by recognition of the uronic acid side residue

A crucial role in this recognition is ascribed to ionic interaction of the enzyme with the uronic acid carboxylate Lack of the uronic acid renders the

xylan main chain virtually resistant to the enzyme’s

action The specific activities on unsubstituted

b-1,4-xylooligosaccharides are three orders of magnitude

lower than those on similar substrates containing

MeGlcA

Experimental procedures

Cloning and expression of XynA

Recombinant XynA was obtained by expressing its

synthetic gene in B subtilis A164delta5 [25] The synthetic

gene, based on the published gene sequence (Swiss-Prot:

Q46961), was generated by the company DNA2.0 (Menlo

Park, CA, USA) and delivered as a cloned fragment in

their standard cloning vector (kanamycin-resistant)

The synthetic gene sequence (Fig S3A) was

codon-opti-mized for expression in B subtilis following the

recommen-dations of Gustafsson et al [26] The expressed DNA

sequence can be found in Fig S3B The xylanase gene was

cloned with the signal peptide from Savinase [26] (included

in the vector), replacing the native secretion signal The

coding region without the native signal was amplified by

PCR from the plasmid containing the synthetic gene, and

cloned into the expression vector pDG268neo [25]) The

PCR primers contained an N-terminal ClaI site and a

C-terminal MluI site The PCR fragment and vector were

digested with ClaI and MluI The vector and fragment were

ligated and transformed into Escherichia coli Several

rec-ombinants were obtained A plasmid containing the correct

gene sequence was transformed into B subtilis, following

the methods in Widner et al [27] A recombinant B subtilis

clone containing the integrated expression construct was

grown in PS-1 liquid culture medium [27] The enzyme was

purified from the culture supernatant

Purification of recombinant XynA

The culture supernatant, collected by centrifugation

(17 700 g for 30 min), was filtered (0.22 lm), and the filtrate

was adjusted to pH 8.5 and subsequently loaded onto an

MEP HyperCel (Pall, East Hills, NY, USA) XK 26⁄ 20

col-umn (GE Healthcare Bio-Sciences, Piscataway, NJ, USA)

The column (60 mL) was equilibrated in 50 mm Tris⁄ HCl

buffer (pH 8.5) (buffer A) Unbound protein was washed off

with 300 mL of buffer A The proteins were eluted with

50 mm sodium acetate buffer (pH 4.5) (buffer B) Fractions

were analyzed by SDS⁄ PAGE, and fractions containing the

enzyme were combined and their pH was adjusted to pH 6.0

The combined fractions were diluted five times in 25 mm Mes

buffer (pH 6.0) (buffer C) and applied to a cation exchange

SP Sepharose Fast Flow (GE Healthcare Biosciences,

Uppsala, Sweden) XK 26⁄ 20 column (GE Healthcare

Bio-Sciences, Piscataway, NJ) The cation exchanger (20 mL)

was equilibrated in buffer C Unbound protein was washed off with 100 mL of buffer C The XynA was eluted with a linear gradient of NaCl (0–0.5 m) in buf-fer C, using five column volumes Fractions were analyzed

by SDS⁄ PAGE, and those containing XynA were combined

Other enzymes GH3 b-xylosidase was a product of a recombinant Saccharomyces cerevisiae strain expressing a plasmid-borne Aspergillus niger XlnD gene [28], GH67 a-glucuronidase was obtained from R P deVries and J Visser (Agricultural University of Wageningen, The Netherlands), and GH115 a-glucuronidase was a product of Pichia stipitis [29]

Substrates and oligosaccharide ligand The ligand used for cocrystallization with XynA was MeGlcA2Xyl3 This aldotetraouronic acid was prepared from MeGlcA2Xyl4, the shortest acidic product generated from hardwood glucuronoxylan by a family 11 endo-b-1,4-xylanase [27] by the action of recombinant XynA The enzyme catalyzed the reaction MeGlcA3Xyl4 fi MeGlcA2Xyl3+ Xyl [14] MeGlcA3Xyl4 (20 mg), isolated from glucuronoxylan-spent medium of Thermomyces la-nuginosus [30], was incubated in 2 mL of water with 0.3 mg of purified recombinant XynA at 30C After the hydrolysis was completed (examined by TLC), the prod-uct was isolated from the reaction mixture by preparative paper chromatography on Whatman No 3 (prewashed with deionized water) in the solvent system ethyl ace-tate⁄ acetic acid⁄ water (18 : 7 : 8, v⁄ v ⁄ v) for 17 h The sugars on guide strips were localized with the silver nitrate reagent The water eluate of the desired product was filtered and freeze-dried The structure of the product

as MeGlcA2Xyl3 was confirmed enzymatically (Fig S1) The compound was resistant to GH67 a-glucuronidase but served as a substrate for GH115 a-glucuronidase to yield MeGlcA and xylotriose It was hydrolyzed by the GH3 b-xylosidase [28] to Xyl and MeGlcA2Xyl2, giving MeGlcA and xylobiose with both types of a-glucuroni-dase MeGlcA3Xyl3 was isolated from glucuronoxylan hydrolysate by endoxylanase of GH10 as the shortest acidic oligosaccharide [4] Xylotetraose and xylopentaose were from Megazyme (Ireland)

Crystallization

An enzyme solution was prepared by concentrating the pro-tein in 25 mm MES buffer (pH 6.0), containing150 mm NaCl, to a concentration of 20 mgÆmL)1, using an Amicon stirred cell and a Biomax membrane with cutoff 5 kDa Fifty-microliter aliquots of the concentrated solution were

stored at )20 C until use Crystals were prepared by the

vapor diffusion method in a hanging drop, with XRL

plates and plastic coverslips (Molecular Dimensions,

Suffolk, UK) The drops were composed of the protein

stock solution and precipitant solution at a 1 : 1 ratio in a

final volume of 2 lL, and equilibrated against 500 lL of

precipitant solution In the case of the cocrystallization, the

3-lL drops were prepared by mixing the protein, ligand

and precipitant solutions at a 1 : 1 : 1 ratio An aqueous

solution of MeGlcA2Xyl3 (20 mm) was used as the ligand

solution Pact Premier I and II and Crystal Clear I

crystalli-zation kits (Molecular Dimensions) were used for

prelimin-ary crystallization screening Clusters of thin and fragile

needle crystals were obtained under 19 conditions, which

were further optimized Diffraction-quality crystals were

prepared by crystal seeding Data were collected from the

crystal of the XynA–MeGlcA2Xyl3 complex obtained by

cocrystallization with 0.1 m imidazole⁄ d,l-malic acid buffer

(pH 7.5) and 20% (w⁄ v) poly(ethylene glycol) 1500 as a

precipitant solution

Data collection and structure determination

The crystals were tested and data were collected at the X13

beamline at EMBL c⁄ o DESY, Hamburg, Germany The

crystals were mounted on the loops, soaked in a

cryopro-tectant solution, and flash cooled in a stream of cold

nitro-gen gas (100 K) directly at the goniometer head

As cryoprotectants, Paratone-N, perfluoropolyether,

paraf-fin oil and precipitant solution enriched with glycerol to a

final concentration of 20% were tested The best results

were obtained with paraffin oil Data were collected at

100 K, according to the strategy proposed by best [31], and

processed by xds [32], and scala [33], reindex and

com-bat from ccp4 suite 6.1.3 (Collaborative Computational

Project Number 4, 1994 [23]), using ccp4i Interface 2.0.6

[34] running under Windows The structure was solved by

the molecular replacement method with molrep [35] and

xylanase A (Protein Data Bank code 1NOF [16]) as a

model structure The structure was refined with

ref-mac5.5.01 [36] in combination with coot-findwaters,

and the model was visualized and rebuilt with coot [37]

All electron density maps were calculated by FFT [38] For

structure validation, procheck was used [39,40], and for

structure analyses, areaimol, contact and other programs

of the ccp4 suite were used with the default parameters

Figures were prepared with molscript [41] and pymol [42]

Molecular modeling

Docking experiments and binding energy calculations were

performed with leadfinder [23] This program was also

used for the preparation of the protein structure for

dock-ing by addition of hydrogen atoms accorddock-ing to optimal

ionization states of protein residues at a given pH The

ligand structures were prepared for molecular modeling in their optimal protonation state with ChemAxon marvin suite [43]

Specific activity on aldouronic acids and linear xylooligosaccharides

Four-millimolar solutions of the compounds in 50 mm sodium acetate buffer (pH 5.5) were incubated at 40C with XynA at various dilutions At time intervals, aliquots were taken to determine the reducing sugars by the Somo-gyi–Nelson procedure [44] The high background of the substrates reduced the accuracy of the measurements, particularly at early stages of hydrolysis

Acknowledgements

The authors are grateful to M Czisza´rova´ for excellent technical assistance This work was supported by VEGA grants 2⁄ 0001 ⁄ 10 and 2 ⁄ 0165 ⁄ 08 from the Slo-vak Academy of Sciences We acknowledge the EMBL X13 beamline at the DORIS storage ring, DESY, Hamburg for providing us with synchrotron source facilities We thank M Groves (EMBL Hamburg) for his help with data processing, and O Stroganov (BioMolTech) for technical help with leadfinder

Note added in proof

During processing of this article for publication we have learned about the appearence of the paper describing similar substrate recognition mechanism by

a GH30 xylanase from Bacillus subtilis using a crystal structure of the complex of the enzyme with different ligand (St John FJ, Hurlbert JC, Rice JD, Preston JF

& Pozharski E (2011) Ligand bound structures of a glycosyl hydrolase family 30 glucuronoxylan xylanohy-drolase J Mol Biol 407, 92–109)

References

1 Henrissat B & Davies GJ (1997) Structural and sequence-based classification of glycoside hydrolases Curr Opin Struct Biol 7, 637–644

2 Biely P (2003) Xylanolytic enzymes In Handbook of Food Enzymology(Whitaker JR, Voragen AGJ & Wond DWS eds), pp 879–915 Marcel Dekker, New York

3 Collins T, Gerday C & Feller G (2005) Xylanases, xylanase families and extremophilic xylanases FEMS Microbiol Rev 29, 3–23

4 Biely P, Vrsˇanska´ M, Tenkanen M & Kluepfel D (1997) Endo-b-1,4-xylanase families: differences in catalytic properties J Biotechnol 57, 151–166