Báo cáo khoa học: Covalent and three-dimensional structure of the cyclodextrinase from Flavobacterium sp. no. 92 pdf

Covalent and three-dimensional structure of the cyclodextrinase from Flavobacterium sp. no. 92 Hanna B. Fritzsche, Torsten Schwede and Georg E. Schulz Institut fu ¨ r Organische Chemie und Biochemie, Albert-Ludwigs-Universita ¨ t, Freiburg im Breisgau, Germany Starting with oligopeptide sequences and using PCR, the gene of the cyclodextrinase from Flavobacterium sp. no. 92 wasderivedfromthegenomicDNA.Thegenewas sequenced and expressed in Escherichia coli; the gene pro- duct was purified and crystallized. An X-ray diffraction analysis using seleno-methionines with multiwavelength anomalous diffraction techniques yielded the refined 3D structure at 2.1 A ˚ resolution. The enzyme hydrolyzes a(1,4)- glycosidic bonds of cyclodextrins and linear malto-oligo- saccharides. It belongs to the glycosylhydrolase family no. 13 and has a chain fold similar to that of a-amylases, cyclo- dextrin glycosyltransferases, and other cyclodextrinases. In contrast with most family members but in agreement with other cyclodextrinases, the enzyme contains an additional characteristic N-terminal domain of about 100 residues. This domain participates in the formation of a putative D 2 -sym- metric tetramer but not in cyclodextrin binding at the active center as observed with the other cyclodextrinases. More- over, the domain is located at a position quite different from that of the other cyclodextrinases. Whether oligomerization facilitates the cyclodextrin deformation required for hydro- lysisisdiscussed. Keywords: calcium-binding site; cyclodextrin degradation; glycosylhydrolase family no. 13; oligomerization; X-ray analysis. Cyclodextrins (CDs) are cyclic malto-oligosaccharides of at least six to generally eight glucosyl units linked via a(1,4)- glycosidic bonds. Their ability to form inclusion complexes with numerous small hydrophobic molecules is used in various applications such as microencapsulation of drugs [1] and chromatographic separation of chiral compounds [2]. The increasing application of CDs has stimulated an interest in the mechanisms of their degradation, particularly as they are resistant to hydrolysis by most a-amylases. Several CD-degrading enzymes have been isolated [3–8]. They are from various bacterial sources and generally prefer CDs but also accept other maltodextrins, converting them to a broad spectrum of products. It has been suggested that these cyclodextrinases (EC 3.2.1.54) should be combined with maltogenic amylases (EC 3.2.1.133) and neopullula- nases(EC3.2.1.135) intoasingle enzymeclass [9]because their catalytic properties differ only partially and not distinctly. One characteristic of the CD-degrading enzymes is their additional N-terminal domain, which in neopullulanase from Thermoactinomyces vulgaris (TVA-II) [10], maltogenic amylase from Thermus sp. (ThMA) [11] and cyclodextrinase from Bacillus sp. I-5 (BaCD) [9] participates in dimer formation. In these dimers, the N-terminal domain of one subunit contacts the active center of the other subunit and participates in CD binding. Moreover, it constricts the active-center pocket, affecting substrate specificity, for instance, by excluding large molecules such as starch [11–13]. Beyond the common dimerization, BaCD forms a hexamer of these dimers, i.e. a dodecamer in solution [9]. Here we investigate the cyclodextrinase from Flavobac- terium sp. no. 92 (CDase) which hydrolyzes CDs and short linear malto-oligosaccharides at comparable rates. The enzyme exhibits only minor hydrolytic activity on the a(1,4)-linkages of starch [14] and pullulan [15] but shows considerable transglycosylation activity [16]. The sequence and 3D structure of the enzyme is presented and compared with related proteins. Experimental procedures Isolation and sequencing of the gene The enzyme was purified from Flavobacterium sp. no. 92 as described [4]. The N-terminal amino-acid sequence was determined to be AAPTAIEHMEPPFW using Edman degradation in a gas phase sequencer (Applied Biosystems). Furthermore the enzyme was cleaved with CNBr, and the sequences of the six resulting peptides were analyzed. One of the fragments, with the sequence MPDRFANGDPSND, was selected because it showed  60% amino-acid sequence identity with several a-amylases and cyclodextrin glycosyl- transferases (CGTases) in the SWISSPROT Data Bank. On the basis of these two peptides the following two primers were constructed (S denotes a C/G mixture and R stands for Correspondence to G. E. Schulz, Institut fu ¨ r Organische Chemie und Biochemie, Albertstr. 21, Freiburg im Breisgau, D-79104, Germany. Fax: + 49 761 203 6161, Tel.: + 49 761 203 6058, E-mail: schulz@bio.chemie.uni-freiburg.de Abbreviations: BaCD, cyclodextrinase from Bacillus sp. I-5; CD, cyclodextrin, i.e. cyclic malto-oligosaccharide of six or more glucosyl groups; CDase, cyclodextrinase from Flavobacterium sp. no. 92; CGTase, cyclodextrin glycosyltransferase; TAKA, a-amylase from Aspergillus oryzae; ThMA, cyclodextrin-degrading maltogenic amylase from Thermus sp.; TVA-I, a-amylase 1 from Thermoactino- myces vulgaris; TVA-II, neopullulanase from Thermoactinomyces vulgaris. (Received 28 January 2003, revised 21 March 2003, accepted 2 April 2003) Eur. J. Biochem. 270, 2332–2341 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03603.x A/G): 5¢-GCSCCSACSGCSATCGAGCACATGGA-3¢ (residues 2 APTAIEHME) and 3¢-TACGGSCTRGCSAA GCGSTTG-5¢ (reverse, residues (113) MPDRFAN). Using these two primers in a PCR amplification with genomic DNA from the Flavobacterium as the template, a 350 bp DNA fragment was produced. Using the random primer method [17], this fragment was taken as a template to produce [ 32 P]dCTP-labeled probes. Genomic DNA from Flavobacterium sp. no. 92 was prepared using a slightly modified protocol of Sambrook et al. [18], partially digested with Sau3A and fractionated by sucrose gradient centrifugation. Fragments ranging from 7 to 12 kb were used to prepare a genomic library in kZAP Express DNA (Stratagene). The recombinant phages were packaged in vitro with Gigapack II Packaging Extract (Stratagene) and plated on Escherichia coli XL1-Blue MRF¢ (Stratagene) to a final concentration of 5000 pfu per plate (diameter 15 cm). As determined by blue/white selection, the library contained 55 000 independent plaques including 10% wild-type phages without inserts. Positive plaques were identified by in situ hybridization with the radiolabeled probe. They were subcloned in vivo into pBK-CMV phagemides (Stratagene) by coinfection with the helper phage M13 Exassist (Stratagene), and then analyzed with restriction enzymes. A clone with the complete gene was isolated. With the use of the dideoxy method [19], the gene sequence was determined by PAGE and by more advanced methods (SeqLab, Go ¨ ttingen, Sweden). The complete DNA sequence has been deposited in the EMBL Nucleotide Sequence Database under accession code AJ489171. Expression, purification and crystallization The CDase gene without the signal sequence was subcloned into the expression vector pET22b+ (Novagene) using restriction enzymes EcoRI and NdeI. Thereby, the first alanine of the mature enzyme was replaced by a methionine. The CDase gene was then expressed in E. coli strain BL21(DE3). Cells were grown at 25 °C in Luria–Bertani broth supplemented with 100 lgÆmL )1 ampicillin and induced at an A 600 of 0.4–0.5 by adding 0.1 m M isopropyl thio-b- D -galactoside. They were harvested 4.5 h after induction, resuspended in buffer A (50 m M Hepes, pH 6.5, 2 m M CaCl 2 ), and disrupted using a French press. After centrifugation, the supernatant was diluted 1 : 1 with buffer A and loaded on to a cation-exchange column (Source 30S; Pharmacia). The enzyme was eluted from the column at 120 m M within a 0–200 m M NaCl gradient in buffer A and was identified using SDS/PAGE. The main fractions were pooled and concentrated to 6 mgÆmL )1 protein. The yield of the purified protein was 8 mg per litre of culture medium. For the crystallization experiments, the CDase was dialyzed against deionized water. For phasing the X-ray reflections with the multiwave- length anomalous diffraction method, all methionines were replaced with seleno-methionines. For this purpose, the CDase was expressed at 25 °C in the methionine-auxo- trophic E. coli strain B834(DE3) using a culture medium containing seleno- D , L -methionine at a concentration of 50 mgÆL )1 [20,21]. The purification procedure was similar to that of the wild-type enzyme, but 3 m M dithiothreitol was added to all buffers and solutions to avoid oxidation of the incorporated seleno-methionines. The yield of Se-labeled CDase was 6 mg per litre of culture medium and thus only slightly lower than that of the wild-type enzyme. Crystallization was carried out by the hanging drop vapor diffusion method using a sparse matrix screen (Hampton Research, La Jolla, CA, USA) 1 . After optimiza- tion, the best crystal conditions for the wild-type enzyme were a 1 : 1 mixture of a 6 mgÆmL )1 protein solution and the reservoir buffer containing 50 m M Hepes, pH 7.5, and 3.8 M NaCl. The Se-labeled CDase crystallized under the same conditions, except for the addition of 3 m M dithio- threitol. Crystals appeared within 2 days and grew to final dimensions of 500 · 200 · 100 lm 3 . For cryoprotection, 15% glycerol was added just before the crystals were mounted in a cryo-loop and shock-frozen. The crystals of wild-type and Se-labeled CDase were isomorphous. Structure determination, phasing and refinement Preliminary data for wild-type CDase crystals and heavy atom derivatives were collected on a wire-frame detector (X-1000; Bruker-Nicolet, Karlsruhe, Germany) 2 using a rotating anode (RU200B; Rigaku, Tokyo, Japan) 3 . The final data, however, were collected with an Se-labeled CDase crystal at synchrotron beamline BW7A (EMBL-outstation, DESY Hamburg) at three different wavelengths, which were selected on the basis of an X-ray fluorescence spectrum taken from the same crystal. Data were processed and scaled with the program suite HKL [22] bringing Friedel pairs to the same scale. The positions of 48 selenium atoms were determined with program SHELX - D [23]. Phases were cal- culated with SHELX - E [24]andinasecondrunalsowith program SHARP / AUTOSHARP [25]. The two resulting density maps were of equal quality. The model was built by a combination of ARP / WARP [26] and manual operations using program O [27]. The complete model was refined by simulated annealing with noncrystal- lographic symmetry (NCS) restraints using program CNS [28]. Several refinement cycles with individual isotropic B-factors followed. Water molecules were either automati- cally identified by program CNS or manually introduced using program O . The final refinement was performed using the TLS approach in REFMAC [29] without NCS restraints. Program LSQMAN [30] was used for structural alignments. Figures were produced with MOLSCRIPT [31] and RASTER3D [32]. The co-ordinates and structure factors are deposited in the Protein Data Bank under accession code 1H3G. Results and discussion DNA and polypeptide sequence The CDase gene consists of 1857 bases of which the first 54 bases code for a signal sequence for protein translocation into the periplasm. The DNA sequence agreed with the independently established amino-acid sequences of seven peptides. The derived amino-acid sequence is given in Fig. 1 except for the 18-residue signal peptide. The native mature protein consists of 601 residues with an M r of 67 946. On the basis of sequence similarity, it belongs to the glyco- sylhydrolase family no. 13 [33]. The four conserved segments of family no. 13 (Fig. 1) represent the calcium Ó FEBS 2003 Cyclodextrinase structure (Eur. J. Biochem. 270) 2333 site, Ca-I, and the three invariant catalytic acids, Asp311, Glu340 and Asp418. The 102 N-terminal residues of CDase form a domain that is missing in most other members of family no. 13. However, it is also present in the other structurally established CD-degrading enzymes neopullulanase TVA-II [10], maltogenic amylase ThMA [11], cyclodextrinase BaCD [9], and a second neopullulanase that resembles TVA-II, but is not yet available from the Protein Databank [34]. Furthermore, it is present in the a-amylase TVA-I [13], a trehalohydrolase [35], and an isoamylase [36]. This N-terminal domain is not present in the CD-producing CGTases, which, however, contain about 150 additional C-terminal residues that probably mediate starch binding [37–39]. 3D structure The crystals of Se-labeled CDase belong to space group R32 with unit cell dimensions a ¼ b ¼ 181.3 A ˚ and c ¼ 231.5 A ˚ at 100 K and two CDase molecules in the asymmetric unit. They have a packing parameter of 2.6 A ˚ 3 /Da, which is 4% smaller than that of the isomor- phous wild-type crystals at 100 K. The wild-type crystals failed to show a comparable diffraction quality and were therefore not further analyzed. Data collection statistics are given in Table 1. The structural refinement yielded a crystallographic R factor of 18.8% and an R free of 22.3% with over 90% of the residues in the most favored region of a Ramachandran plot (Table 2). The resulting model is depicted in Fig. 2. The Ca backbones of the two molecules canbesuperimposed,withanrmsdof0.31A ˚ , indicating conformational homogeneity. Met1 is removed during expression in E. coli; Ala2, Glu600 and Ala601 have no density. The overall real space map correlation coefficient was 0.95 [30]. The model includes about 0.6 water molecules per residue, which is appropriate at 2.1 A ˚ resolution. The B factor plot for both molecules is shown in Fig. 3. The peaks are almost exclusively in loop regions. Following the assignments in related enzymes, CDase was divided into four domains (Fig. 2). As a member of the glycosylhydrolase family no. 13, its chain fold is similar to that of known a-amylases consisting of a central TIM barrel [40] (domain A, residues 103–516), with a 60-residue insert after the third strand of the b-barrel (domain B, 223–282) and a C-terminal domain (domain C, 517–601). In addition, CDase contains an N-terminal domain (1–102), which assumes a characteristic b-sandwich structure composed of the antiparallel strands b1tob8. The N-terminal domain is connected by an extended 10-residue linker to the TIM barrel. It contacts the bulk of the molecule at helices a6, a7 and at domain B. Domain A harbors the active center at the C-termini of the TIM barrel b-strands. The loops at the C-terminal barrel end connecting b-strands with the following a-helices are longer and more complex than the loops at the opposite barrel end. The lengths of the b-strands in the barrel vary from two (b14) to seven (b11) residues. As in other enzymes of family no. 13, the regularity of the CDase TIM barrel is broken by the a-helices after the sixth b-barrel strand where helix a9 extends in the direction of the preceding strand b14 and only the next helix a10 runs in the opposite direction (Fig. 2). For historical reasons the large loop between the third strand of the TIM barrel (b11) and helix a6 is called domain B (indicated in Fig. 2). This inserted domain participates in substrate binding and is rather variable. It is considered to play a role in determining the enzyme specificity [41]. The end of the TIM barrel domain A is connected to domain C forming two antiparallel sandwiched b-sheets. Between them, the sheet contacting the TIM barrel at helices a9, a10, a12 and a13 contains strands b17, b18, b19 and b24, whereas the solvent-exposed second b-sheet harbors strands b20, b21, b22 and b23. The b-sheet at the interface to domain A has lower B factors (Fig. 3) and is structurally much better conserved within the family than the other Fig. 1. Amino-acid sequence of native mature CDase derived from the DNA sequence. Independently established peptide sequences are underlined. The four conserved segments at the calcium-binding site Ca-I and at the three invariant catalytic residues (inverted) Asp311, Glu340 and Asp418 are given in bold letters. Residues shown in lower case are not included in the model. In the analyzed enzyme, Ala1 was replaced by Met1 which, however, was removed during protein expression in E. coli. Table 1. Data collection for phasing with multiwavelength anomalous diffraction. Data set Peak Inflection point High energy remote Wavelength (A ˚ ) 0.9795 0.9797 0.9393 Resolution a (A ˚ ) 24–2.4 (2.5–2.4) 26.7–2.4 (2.5–2.4) 25–2.1 (2.18–2.08) Number of observations 862078 829496 629678 Unique reflections a 58931 (5892) 58981 (5898) 86572 (8625) Completeness a (%) 99.8 (99.8) 99.9 (99.9) 99.9 (99.9) R sym-I a (%) 7.7 (21) 5.1 (22) 6.0 (40) Multiplicity a 14.5 (14.4) 14.1 (14.1) 7.3 (7.3) Average I/r I a 10.5 (3.4) 11.9 (3.1) 10.4 (1.9) a Values in parentheses refer to the outermost shell. 2334 H. B. Fritzsche et al.(Eur. J. Biochem. 270) Ó FEBS 2003 sheet, which supports the proposal that domain C stabilizes the TIM barrel. Like most other members of family no. 13, CDase contains Ca 2+ ions. One of the two Ca 2+ ions in CDase is at site Ca-I, which is widely conserved forming the first of the four sequence fingerprints shown in Fig. 1 [42]. The removal of Ca-I was shown to promote proteolysis [43,44]. Ca-I is co-ordinated by the side chains of Asp280 and Ser222 at the beginning and end of domain B as well as by the main-chain oxygens of Tyr315 (domain A) and Thr270 (domain B) and by two water molecules. Ser222 of the Ca-I sequence fingerprint is specific for CDase, where it replaces a highly conserved asparagine. At its position between domains A and B, Ca-I stabilizes the conformation of domain B together with residues Tyr315, Phe274 and others that are directly or indirectly involved in substrate binding. Ca-I is missing in the three CD-degrading enzymes TVA-II, ThMA and BaCD. The second and third calcium-binding sites of family no. 13 enzymes show greater variation [42,43,45]. The second calcium site of CDase is called Ca-II. It is also present in the CD-producing CGTase [38] and in the CD-degrading TVA- II [13], but not in the CD-degrading enzymes ThMA and BaCD nor in the majority of a-amylases. Ca-II is located in the loop between a1anda2 of domain A (Fig. 2) and Table 2. Refinement statistics. Values in parentheses refer to the outermost shell. Resolution range (A ˚ ) 21–2.1 (2.13–2.08) Number of reflections 78164 (4817) Protein atoms 10239 Calcium ions 4 Water molecules 695 Average B factor (A ˚ 2 )41 R cryst (%) 18.8 (21.9) R free (%) (test set of 1991 reflections) 22.3 (27.8) Rmsd bond lengths (A ˚ )/angles (°) 0.016/1.34 Ramachandran angles in: Most favored region (%) 90.2 Allowed [generally allowed] region (%) 9.5 [0.3] Fig. 2. Stereoview of a ribbon plot of CDase showing the N-terminal domain (red), the TIM-barrel domain A (blue), the inserted domain B (green) and the C-terminal domain C (orange). The two Ca 2+ ions are represented by black spheres. The active center is indicated by the three invariant catalytic residues (Fig. 1). All a-helices and b-strands are labeled. Fig. 3. B-factor plots of the main chains of the two molecules of CDase in the asymmetric unit. The averages of molecules A (solid line) and B (broken line) are 39 A ˚ 2 and 44 A ˚ 2 , respectively. The a-helices and b-strands are given for reference. Helices a1 through a14 and strands b9throughb16 comprise the TIM barrel. Domain B is inserted between b11 and a6. The seven 3 10 -helices are not indicated. Ó FEBS 2003 Cyclodextrinase structure (Eur. J. Biochem. 270) 2335 co-ordinated by the side chains of Asp125, Asp146, Asn119 and Asn124, the main-chain oxygens of Gly144 and Asp121, and by a water molecule. Ca-II stabilizes a surface region far away from the active center. The B factors of both Ca 2+ ions in both CDase molecules are similar to those of the surrounding atoms. This indicates that the sites are fully occupied in the crystal, even though the protein was dialyzed against deionized water before crystallization and the crystallization buffer lacked calcium. Oligomerization In the crystal, CDase forms putative D 2 -symmetric tetramers containing two noncrystallographic and one crystallographic twofold axes. The oligomerization was derived from the large contact areas formed within each tetramer and the small contacts between neighboring tetra- mers, making the crystal look like an assembly of tetramers. As shown in Fig. 4, the tetramer contact across the crystallographic twofold axis is 1590 A ˚ 2 in size, which is in the normal range of oligomer interfaces. The other internal contact is formed by the N-terminal domains and measures 520 A ˚ 2 , which exceeds an average crystal packing contact. The large difference between these two interface areas classifies the tetramer as a weak dimer of strong dimers. A preliminary size-exclusion chromato- graphy run (Sephacryl 300S, 100 m M Hepes, pH 7.5) at 0.15 M NaClcomparedwith3.8 M NaCl in the crystals showed a dominating dimer mixed with other oligomers. Similar runs under other conditions have yet to be performed to determine the detailed oligomerization pattern in solution. Fig. 4. D 2 -symmetric tetramer structure of CDase in the crystal together with the symmetry axes. (A) Front view placing the crystallographic twofold axis horizontally in the paper plane. The crystallographic axis runs through the large interface and the vertical noncrystallographic axis runs through the small interface between the N-terminal domains. One subunit is given in the colors and in an orientation similar to Fig. 2. A b-CD (orange) derived from a superposition with the complex between b-CD and the homologous enzyme TVA-II [47] marks the active center. (B) View from the left side of (A), which is along the crystallographic twofold axis, showing a smooth silhouette. 2336 H. B. Fritzsche et al.(Eur. J. Biochem. 270) Ó FEBS 2003 The chain fold of the N-terminal domain of CDase is similar to that of the related CD-degrading enzymes TVA-II [10], ThMA [11] and BaCD [9]. However, the positions of these domains relative to the respective TIM barrel are completely at variance as shown in Fig. 5. The N-terminal domains of TVA-II, ThMA and BaCD attach to the active center of the other subunit and participate in substrate selection [11,12]. This dimer interface is not related to either of the two interfaces in the CDase tetramer. It seems very unlikely that the deviating domain position in CDase is a packing artefact caused by domain swapping during crystallization because the interface between the N-terminal domain and the protein remainder (domains A and B) amounts to 1390 A ˚ 2 , which is much larger than a common packing contact. Comparison with related enzymes The relationships within the group of CD-degrading enzymes were evaluated by a comprehensive chain-fold comparison. The comparison was extended to the structur- ally related TVA-I [13], the CD-producing CGTases [38,39] and a-amylase from Aspergillus oryzae (TAKA) [46], which was taken as a well-known representative of family no. 13. In principle, all comparisons could have been performed with the amino-acid sequences alone, as the glycosylhydro- lase families are defined by the sequences. However, this method suffers from the rather arbitrary placing of the gaps. Therefore, we took account of the geometry and first derived the group of structurally equivalent residues in a chain-fold superposition and then counted the number of identical residues within this group. The results are given in Table 3. The most obvious result of this comparison is the close relationship between TVA-II, ThMA and BaCD, which can be almost completely structure-aligned, giving rise to about 50% identical residues. As mentioned above, these three enzymes also form similar dimers (with a further hexameric association in BaCD) and have similar catalytic activities. Therefore, we classify them as the TVA-II group. When comparing CDase with this group, only  380 of the 600 residues can be structure-aligned, and only 28% of the aligned residues are identical. This renders CDase an outlier among the structurally established CD-degrading enzymes. As for the other enzymes, a-amylase TVA-I shows considerable structural similarity to the TVA-II group, although its function differs greatly (Table 3). Moreover, the data reveal that the CD-degrading enzymes are more similar to the CD-producing CGTase than to the a-amylase TAKA. A more obvious difference between CDase and the others is the deviating spatial position of its N-terminal domain Fig. 5. Stereoview of the superposition of CDase(coloredasinFig.2)withTVA-II (black, Ca 2+ at Ca-II grey) given as Ca-backbone plots. The completely different positions of the N-terminal domains and the differences in domain B near Ca-I (right) are clearly visible. Table 3. Chain-fold comparisons within glycosylhydrolase family no. 13. The upper right triangle shows the number of Ca atoms aligned within the 3A ˚ distance criterion in superpositions of the complete polypeptide chains using program LSQMAN (30). The numbers in parentheses are the percentages of identical residues in the aligned segments. For CDase, CGTase and TAKA, only domains A, B and C could be superimposed with any of the other enzymes. The lower left triangle shows the respective numbers for a separate superposition series involving only the N-terminal domains. CDase TVA-II ThMA BaCD TVA-I CGTase TAKA CDase 387 (29) 387 (28) 376 (26) 371 (25) 380 (26) 361 (25) TVA-II 47 (11) 547 (47) 552 (47) 428 (37) 378 (25) 369 (21) ThMA 47 (13) 121 (35) 570 (54) 472 (32) 366 (24) 368 (26) BaCD 48 (6) 119 (32) 123 (49) 421 (37) 376 (22) 354 (26) TVA-I 54 (6) 98 (18) 98 (26) 102 (25) 367 (25) 377 (24) CGTase – – – – – 401 (25) Ó FEBS 2003 Cyclodextrinase structure (Eur. J. Biochem. 270) 2337 (Fig. 5). A superposition restricted to the N-terminal domains showed that those of the TVA-II group can be almost fully structure-aligned, resulting in  40% amino- acid residue identities (Table 3). TVA-I is somewhat outside the TVA-II group, but can still be well aligned. However, the N-terminal domain of CDase aligns only with about half of its residues, shows almost no sequence identity (Fig. 6 and Table 3), and clearly differs from the TVA-II group with respect to sequence, chain fold, and position. The N-terminal domains of the trehalohydrolase [35] and the isoamylase [36] have a similar chain fold to that of CDase and the TVA-II group, but they are barely related to any of them (data not shown). Interestingly, the general location of the N-terminal domains of these two outliers [35,36] corresponds to that of CDase. A superposition of the highly variable B domains, which participate in the active center, is given in Fig. 7. CDase has a very long extension, whereas CGTase and TAKA have intermediate ones. In contrast, the TVA-II group and TVA-I have a much smaller B domain. The large B domains of CDase, CGTase and TAKA are fixed by Ca-I, which is absent in the TVA-II group with their small B domains. The surprisingly large difference between CDase on one hand and the TVA-II group on the other corresponds to the different oligomeric structures. CDase uses the long exten- sion of its B domain to make an intimate contact across the strong dimer interface with domains A and C of the other subunit. In contrast, the TVA-II group dimer attaches the B domain to an N-terminal domain of the other subunit, which restricts the size of the B domain (Fig. 7). Active center The active center of CDase is depicted in Fig. 8, which includes the superimposed structure of a TVA-II dimer with bound b-CD [47]. Interestingly, the superposition causes a Fig. 6. Structural alignment of the N-terminal domain of CDase with those of TVA-II [10], ThMA [11], BaCD [9], TVA-I [13], a trehalohydrolase [35] and an isoamylase [36], which are the only structurally similar domains within glycosylhydrolase family no. 13. CD-degrading activity has been reported for the top four enzymes. The secondary structure of CDase is given, and every 10th amino acid residue is marked by a dot. Residues 87–175 of the isoamylase have been omitted (marked #). All residues that superimpose within the 3 A ˚ distance criterion of program LSQMAN [30] are underlined. For reference, strand b9 of the TIM barrel has been included, and the alignments with the TIM barrels are given in bold. Fig. 7. Superposition of the inserted B domains of CDase (green, His251 marked by a ball), TVA-II (red), CGTase (blue) and TAKA (grey) as aligned on the TIM barrels. TVA-II, ThMa and BaCD are so similar that only one of them was drawn out for clarity. As TVA-I varies only slightly from TVA-II, it was omitted. The chain direction is indicated by the N* and C* ends. 2338 H. B. Fritzsche et al.(Eur. J. Biochem. 270) Ó FEBS 2003 clash between the long B-domain extension of CDase (His251, Fig. 7) and the N-terminal domain of the other subunit of the TVA-II dimer (Tyr45¢). It has been suggested that the N-terminal domain of the TVA-II group [9–11] confers CD specificity because it covers one side of the bound CD [47]. In CDase, this role is fulfilled by the B domain of the same subunit. Therefore, it is sterically impossible for CDase to form the same dimer as the TVA-II group. As the polypeptide superposition of Fig. 8 places the three catalytic residues of CDase (Fig. 1) within less than 1A ˚ of the positions of those of TVA-II, and as the active centers closely resemble each other, the CD molecule bound to TVA-II can be expected to bind at a similar position in the CDase structure. The hydrolysis of CD should start by Glu340 protonating a bridge oxygen of the cyclic substrate. However, the distance between Glu340 and the next bridge oxygen is 6 A ˚ , which is much too long. A similar distance to a bound CD has been observed in CGTase [48], where, however, it has been demonstrated that a linear malto- oligosaccharide binds much deeper in the pocket at the required 3 A ˚ distance to the Glu340 equivalent [49]. Moreover, the conformation at the scissile bond in a CD complex with CGTase showed a substantial deviation from the circular symmetry [48]. These observations indicate that the observed CD-binding position in TVA-II is most likely displaced by about 3 A ˚ . For catalysis, the CD molecule has to be pushed 3 A ˚ deeper into the active-center pocket and deformed at its scissile bond [48]. Such a CD position has not yet been observed in any crystal structure. It would enable Phe274 of CDase (or Phe286 of TVA-II) to rotate around its Ca–Cb bond and enter the CD hollow, as has been implied for TVA-II [10] and for Tyr195 of CGTase [49]. Crystal experiments to clarify the CD position in CDase are under way. The required induced fit and deformation of the bound CD need energy, part of which may be derived from co-operative effects in the CDase tetramer (or TVA-II dimer) association. This proposal is consistent with the observation of a higher rate of CD hydrolysis for dimeric ThMA than for the monomeric ThMA [12]. Although such an energy source is conceivable for the TVA-II group in which the bound CD contacts the N-terminal domain of the other subunit, it is also possible for CDase in which the bound CD is very close to the long B-domain extension (Fig. 7) as well as to A-domain and C-domain residues of the other subunit (Fig. 4A). In fact, the interface mediating the strong dimer association would appear to explain the particularly long B-domain extension of CDase. In con- clusion, the dimer association may help to overcome the conformational activation energy barrier during CD hydrolysis. Acknowledgements We thank H. Bender for drawing our attention to the enzyme, E. Schiltz for amino-acid sequence analyses, M. Ru ¨ ckels for preparing the initializing 350-bp fragment and C. Vonrhein for help with SHARP / AUTOSHARP . Moreover, we are grateful for the contributions of S. Thorspecken, A. Dorowski, B. Phillips, S. Jelakovic, C. Schleberger and M. Mrosek at early stages of the analysis, and we thank the beamline staff of the EMBL-outstation (DESY Hamburg) for help with the data collection. The project was supported by the European Commision under BIO4-98-0022 (AGADE) and by the Deutsche Forschungsgemeinschaft under GRK-434. References 1. Loftsson, T. & Brewster, M.E. (1996) Pharmaceutical applications of cyclodextrins. 1. Drug solubilisation and stabilization. J. Pharm. Sci. 85, 1017–1025. 2. Vetter, W. & Schurig, V. (1997) Enantioselective determination of chiral organochlorine compounds in biota by gas chromato- graphy on modified cyclodextrins. J. Chromatogr. Sect. A 774, 143–175. 3. Saha, B.C. & Zeikus, J.G. (1992) Cyclodextrin degrading enzymes. Starch/Sta ¨ rke 44, 321–315. 4. Bender, H. (1993) Purification and characterization of a cyclo- dextrin-degrading enzyme from Flavobacterium sp. Appl. Micro- biol. Biotechnol. 39, 714–719. 5. Oguma, T., Matsuyama, A., Kikuchi, M. & Nakano, E. (1993) Cloning and sequence analysis of the cyclodextrinase gene from Bacillus sphaericus and expression in Escherichia coli cells. Appl. Microbiol. Biotechnol. 39, 197–203. Fig. 8. Active-center region in a superposition of CDase (blue with light green domain B) with the TVA-II dimer (grey with dark green domain B and pink N-terminal domain). The N-terminal domain of the other subunit of the TVA-II dimer is shown in red including Tyr45¢. The bound b-CD molecule is from a complex with TVA-II [47]. Active-center residues of CDase are given as ball-and-stick models. Ó FEBS 2003 Cyclodextrinase structure (Eur. J. Biochem. 270) 2339 6. Abe, J., Onitsuka, N., Nakano, T., Shibata, Y., Hizukuri, S. & Entani, E. (1994) Purification and characterization of periplasmic alpha-amylase from Xanthomonas campestris K-11151. J. Bacter- iol. 176, 3584–3588. 7. Bender, H. (1995) Purification and characterisation of a soluble, cytoplasmic decycling maltodextrinase from Lactobacillus sp. strain 26X, isolated from kitchen waste water. Appl. Microbiol. Biotechnol. 43, 838–843. 8. Feederle, R., Pajatsch, M., Kremmer, E. & Bo ¨ ck, A. (1996) Metabolism of cyclodextrins by Klebsielle oxytoca m5a1: puri- fication and characterisation of a cytoplasmatically located cyclodextrinase. Arch. Microbiol. 165, 206–212. 9. Lee, H S., Kim, M S., Cho, H S., Kim, J I., Kim, T J., Choi, J H., Park, C., Lee, H S., Oh, B H. & Park, K H. (2002) Cyclomaltodextrinase, neopullulanase and maltogenic amylase are nearly indistinguishable from each other. J. Biol. Chem. 277, 21891–21897. 10. Kamitori,S.,Kondo,S.,Okuyama,K.,Yokota,T.,Shimura,Y., Tonozuka, T. & Sakano, Y. (1999) Crystal structure of Thermo- actinomyces vulgaris R 47 a-amylase II (TVA II) hydrolyzing cyclodextrins and pullulan at 2.6 A ˚ resolution. J. Mol. Biol. 287, 907–921. 11. Kim, J S., Cha, S S., Kim, H J., Kim, T J., Ha, N C., Oh, S T., Cho, H S., Cho, M J., Kim, M J., Lee, H S., Kim, J W., Choi, K.Y., Park, K H. & Oh, B H. (1999) Crystal structure of a maltogenic amylase provides insights into a catalytic versatility. J. Biol. Chem. 274, 26279–26286. 12. Kim,T J.,Nguyen,V.D.,Lee,H S.,Kim,M J.,Cho,H Y., Kim, Y W., Moon, T W., Park, C.S., Kim, J W., Oh, B H., Lee, S B., Svensson, B. & Park, K H. (2001) Modulation of the multisubstrate specificity of Thermus maltogenic amylase by truncation of the N-terminal domain and by a salt-induced shift of the monomer/dimer equilibrium. Biochemistry 40, 14182– 14190. 13.Kamitori,S.,Abe,A.,Othaki,A.,Kaji,A.,Tonozuka,T.& Sakano, Y. (2002) Crystal structures and structural comparison of Thermoactinomyces vulgaris R 47 a-amylase1(TVAI)at1.6A ˚ resolution and a-amylase 2 (TVA II) at 2.3 A ˚ resolution. J. Mol. Biol. 318, 443–453. 14. Bender, H. (1994) Studies of the action pattern on potato starch of the decycling maltodextrinase from Flavobacterium sp, 92. Carbohydr. Res. 263, 137–147. 15. Bender, H. (1994) Studies of the degradation of pullulan by the decycling maltodextrinase of Flavobacterium sp, 92. Carbohydr. Res. 260, 119–130. 16. Bender, H. (1994) Studies of the transglycosylation reaction catalysed by the decycling maltodextrinase of Flavobacterium sp, 92 with malto-oligosaccharides and cyclodextrins. Carbohydr. Res. 263, 123–135. 17. Feinberg, A.P. & Vogelstein, B. (1983) A technique for radio- labeling DNA restriction endonuclease fragments to high specific activity. Anal. Biochem. 132, 6–13. 18. Sambrook, J., Fritsch, E.F. & Maniatis, T. (1989) Molecular Cloning: A Laboratory Manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. 19. Sanger, F. (1981) Determination of nucleotide sequences in DNA. Science 214, 1205–1210. 20. LeMaster, D.M. & Richards, F.M. (1985) 1H)15N heteronuclear NMR studies of Escherichia coli thioredoxin in samples isotopi- cally labeled by residue type. Biochemistry 24, 7263–7268. 21. Hendrickson, W.A., Horton, J.R. & Le Master, D.M. (1990) Selenomethionyl proteins produced for analysis by multi- wavelength anomalous diffraction (MAD): a vehicle for direct determination of three-dimensional structure. EMBO J. 9, 1665– 1672. 22. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac- tion data collected in oscillation mode. Methods Enzymol. 276, 307–326. 23. Uson, I. & Sheldrick, G.M. (1999) Advances in direct methods for protein crystallography. Curr. Opin. Struct. Biol. 9, 642–648. 24. Sheldrick, G.M., Hauptman, H.A., Weeks, C.M., Miller, R. & Uson, I. (2001) Ab Initio phasing. International Tables for Crystallography (Rossmann, M.G. & Arnold, E., eds), Vol. F, pp. 333–345. Kluwer Academic Publishers, Dordrecht. 25. de la Fortelle, E. & Bricogne, G. (1997) Maximum-likelihood heavy-atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods Enzymol. 276, 472–494. 26. Perrakis, A., Sixma, T.K., Wilson, K.S. & Lamzin, V.S. (1997) wARP: improvement and extension of crystallographic phases by weighted averaging of multiple refined dummy atomic models. Acta Crystallogr. Sect. D 53, 448–455. 27. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. Sect A 47, 110–119. 28. Bru ¨ nger, A.T., Adams, P.D., Clore, G.M., Delano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J S., Kuszewski, J., Nilges, N., Pannu, N.S., Read, R.J., Rice, L.M., Simonson, T. & Warren, G.L. (1998) Crystallography and NMR system: a new software suite for macromolecular structure determination. Acta Crystal- logr. Sect. D 54, 901–921. 29. Winn, M., Isupov, M. & Murshudov, G.N. (2000) Use of TLS parameters to model anisotropic displacements in macromolecular refinement. Acta Crystallogr. Sect. D 57, 122–133. 30. CCP4 Collaborative Computational Project, 4 4 (1994) The CCP4 suite: programs for protein crystallography. Acta Crystallogr. Sect. D 50, 760–763. 31. Kraulis, P.J. (1991) MOLSCRIPT : a program to produce both detailed and schematic plots of protein structures. J. Appl. Crys- tallogr. 24, 946–950. 32. Merritt, E.A. & Bacon, D.J. (1997) Raster3D photorealistic molecular graphics. Methods Enzymol. 277, 505–524. 33. Henrissat, B. & Bairoch, A. (1996) Updating the sequence-based classification of glycosyl hydrolases. Biochem. J. 316, 695–696. 34. Hondoh, H., Kuriki, T. & Matsuura, Y. (2002) Three-dimensional structure of Bacillus stearothermophilus neopullulanase. Biologia (Bratisl.), 57, 77–82. 35. Feese, M.D., Kato, Y., Tamada, T., Kato, M., Komeda, T., Miura, Y., Hirose, M., Hondo, K., Kobayashi, K. & Kuroki, R. (2000) Crystal structure of glycosyltrehalose trehalohydrolase from the hyperthermophilic archaeum Sulfolobus solfataricus. J. Mol. Biol. 301, 451–464. 36. Katsuya, Y., Mezaki, Y., Kubota, M. & Matsuura, Y. (1998) Three-dimensional structure of Pseudomonas isoamylase at 2.2 A ˚ resolution. J. Mol. Biol. 281, 885–897. 37. Hofmann, B.E., Bender, H. & Schulz, G.E. (1989) Three-dimen- sional structure of cyclodextrin glycosyltransferase from Bacillus circulans at 3.4 A ˚ resolution. J. Mol. Biol. 209, 793–800. 38. Klein, C. & Schulz, G.E. (1991) Structure of cyclodextrin glyco- syltransferase refined at 2.0 A ˚ resolution. J. Mol. Biol. 217, 737– 750. 39. Leemhuis, H., Dijkstra, B.W. & Dijkhuizen, L. (2003) Thermo- anaerobacterium thermosulfurigenes cyclodextrin glycosyltransfer- ase. Mechanism and kinetics of inhibition by acarbose and cyclodextrins. Eur. J. Biochem. 270, 155–162. 40. Banner, D.W., Bloomer, A.C., Petsko, G.A., Phillips, D.C., Pogson, C.I., Wilson, I.A., Corron, P.H., Furth, A.J., Milman, J.D., Offord, R.E., Priddle, J.D. & Waley, S.G. (1975) Structure of chicken muscle triose phosphate isomerase determined crystal- 2340 H. B. Fritzsche et al.(Eur. J. Biochem. 270) Ó FEBS 2003 lographically at 2.5 A ˚ resolution using amino acid sequence data. Nature (London) 255, 609–614. 41. MacGregor, E.A., Janecek, S. & Svensson, B. (2001) Relationship of sequence and structure to specificity in the a-amylase family of enzymes. Biochim. Biophys. Acta 1546, 1–20. 42. Boel, E., Brady, L., Brzozowski, A.M., Derewenda, Z., Dodson, G.G.,Jensen,V.J.,Petersen,S.B.,Swift,H.,Thim,L.&Wolike, H.F. (1990) Calcium binding in a-amylases: an X-ray diffraction study at 2.1 A ˚ resolution of two enzymes from Aspergillus. Biochemistry 29, 6244–6249. 43. Fujimoto,Z.,Takase,K.,Doui,N.,Momma,M.,Matsumoto,T. & Mizuno, H. (1998) Crystal structure of a catalytic-site mutant a-amylase from Bacillus subtilis complexed with maltopentaose. J. Mol. Biol. 277, 393–407. 44. Machius, M., Wiegand, G. & Huber, R. (1995) Crystal structure of calcium-depleted Bacillus licheniformis a-amylase at 2.2 A ˚ resolution. J. Mol. Biol. 246, 545–559. 45. Kadziola, A., Abe, J., Svensson, B. & Haser, R. (1994) Crystal and molecular structure of barley a-amylase. J. Mol. Biol. 239, 104–121. 46. Matsuura, Y., Kusunoki, M., Harada, W. & Kakudo, M. (1984) Structure and possible catalytic residues of Taka-amylase A. J. Biochem. (Tokyo) 95, 697–702. 47. Kondo, S., Ohtaki, A., Tonozuka, T., Sakano, Y. & Kamitori, S. (2001) Studies on the hydrolyzing mechanism for cyclodextrins of Thermoactinomyces vulgaris R 47 a-amylase 2 (TVA II). X-ray structure of the mutant E354A complexed with b-cyclodextrin, and kinetic analyses on cyclodextrins. J. Biochem. (Tokyo) 129, 423–428. 48. Schmidt,A.K.,Cottaz,S.,Driguez,H.&Schulz,G.E.(1998) Structure of cyclodextrin glycosyltransferase complexed with a derivative of its main product b-cyclodextrin. Biochemistry 37, 5909–5915. 49. Parsiegla, G., Schmidt, A.K. & Schulz, G.E. (1998) Substrate binding to a cyclodextrin glycosyltransferase and mutations increasing the cyclodextrin production. Eur. J. Biochem. 255, 710–717. Ó FEBS 2003 Cyclodextrinase structure (Eur. J. Biochem. 270) 2341 . oligopeptide sequences and using PCR, the gene of the cyclodextrinase from Flavobacterium sp. no. 92 wasderivedfromthegenomicDNA.Thegenewas sequenced and expressed. Covalent and three-dimensional structure of the cyclodextrinase from Flavobacterium sp. no. 92 Hanna B. Fritzsche, Torsten Schwede and Georg