Biologia, Bratislava, 57/Suppl. 11: 43—57, 2002

Independent folding of the A and B domainsin the α-amylase family

Gerard Pujadas* & Jaume Palau

Unitat de Biotecnologia Computacional, Departament de Bioquímica i Biotecnologia, UniversitatRovira i Virgili, Placa de la Imperial Tarraco, 1. Tarragona 43005, Catalonia (Spain); tel.: ++ 34977 55 95 65, fax: ++ 34 977 55 95 97, e-mail: pujadas@quimica.urv.es

PUJADAS, G. & PALAU, J., Independent folding of the A and B domainsin the α-amylase family. Biologia, Bratislava, 57/Suppl. 11: 43—57, 2002;ISSN 0006-3088.

Although the activities of the α-amylase family of enzymes are different, theyhave a common ancestor. Their polypeptide chain always has a multi-domainarrangement (though the number of domains usually depends on the enzymeactivity). The domains that we call A and B are always found in these struc-tures. The A domain, which is the catalytic domain, is always a TIM-barrelfold. The B domain varies in sequence length and fold, and lies between thethird β-strand and the third α-helix of the A domain. Its function has not yetbeen fully established. To determine whether the variability of the B domainaffects the folding of the A domain, we studied the geometrical characteristicsof the eight-stranded β-sheet at the core of the A domain. Our results showthat the geometry of the TIM barrel does not depend on the length or foldof the B domain and supports the idea of an independent folding pathwayfor the A and B domains in α-amylase enzymes. Unwanted mutations thatproduce a different barrel geometry may be recognised by molecular chaper-ones and discarded as functional molecules. Our results show that family 77enzymes have the same barrel geometry as family 13 enzymes. This supportsthe hypothesis that they have a common origin.

Key words: α/β barrel, TIM barrel, α-amylase family, family 13 glycosidehydrolases, family 77 glycoside hydrolases, clan GH-H, chaperones.

Abbreviations: AAMY, α-amylase; CA, carbon atoms in the α position; CG-Tase, cyclodextrin glucanotransferase; GH, glycoside hydrolases; rmsd, rootmean square deviation; SI, similarity index; TIM, triosephosphate isomerase.

Introduction

α-Amylase (AAMY) and related enzymes aregrouped in families 13 (HENRISSAT, 1991), 70and 77 of glycoside hydrolases [GH; http://afmb.cnrs-mrs.fr/CAZY/index.html; (COUTINHO &HENRISSAT, 1999)] and constitute the so-called

GH-H clan or AAMY superfamily. At present, 27different enzyme activities covering transferases(EC 2.x.x.x), hydrolases (EC 3.x.x.x) and iso-merases (EC 5.x.x.x) have been identified as mem-bers of the clan GH-H (COUTINHO & HENRISSAT,1999; JANECEK, 2000a,b; MACGREGOR et al.,2001). In all cases, these enzymes show: (i) a re-

* Corresponding author

43

taining mechanism of the substrate stereochem-istry; (ii) an AAMY-type (β/α)8-barrel fold forthe catalytic domain; and (iii) the same cat-alytic residues in equivalent locations [Asp197,Glu233 and Asp300 in β4, β5 and L7 (the loopafter β7), respectively, where the numbers cor-respond to the location of these residues in thepig pancreatic AAMY sequence (QIAN et al.,1993)]. The sequences in the segments locatedaround the β-strands β2, β3, β4, β5, β7 andβ8 in the TIM-barrel domain are well-conserved(FRIEDBERG, 1983; NAKAJIMA et al., 1986; VI-

HINEN & MÄNTSÄLÄ, 1989; JANECEK, 1994a,b)and at least four of these segments are used as sig-natures for recognizing new enzymes as membersof the superfamily (TAKATA et al., 1992). More-over, two types of mammalian proteins withoutcatalytic function have been identified as havinga sequence that is clearly similar to those of themembers of this clan (JANECEK, 2000a). One ofthese is made up of proteins involved in trans-porting dibasic and neutral amino acids across thecell membranes (BERTRAN et al., 1992; WELLS

& HEDIGER, 1992). The other one acts as the

A) B)

C) D)

Fig. 1. Schematic diagram of the structure of the A+B-domains in the Bacillus licheniformis (Figs 1A,B) andin the Bacillus stearothermophilus (Figs 1C,D) AAMYs. α-Helices are shown as spiral ribbons and β-strandsare drawn as arrows from the amino end to the carboxy end of the β-strand. The TIM-barrel fold correspondingto the A-domain is shown in light gray and the B-domain is shown in dark gray. Figures 1A,C are end viewswhose C-terminal side of the β-sheet is towards the reader. Figures 1B,D are side views whose C-terminal sideof the β-sheet is towards the top of the page. The figure was produced with MOLSCRIPT v2.1.2 (KRAULIS,1991) and the PDB files 1VJS and 1HVX.

44

4F2 heavy-chain cell surface antigen, which is atype II membrane glycoprotein involved in cellgrowth (QUACKENBUSH et al., 1987). In compari-son with the closely-related enzymes in the AAMYfamily (oligo-1,6-glucosidases and α-glucosidases),these proteins lack at least some of the amino-acid residues essential for activity, and thereforedo not contain all of the four conserved regions(NAKAJIMA et al., 1986) that are necessary forenzymatic members of clan GH-H (JANECEK,2000a,b).

Enzymes from the clan GH-H have a multi-

domain arrangement (although the number of do-mains usually depends on the enzyme activity). Asmentioned above, all the members of the AAMYsuperfamily have the TIM-barrel fold of the cat-alytic domain (the so-called A domain) in com-mon (PUJADAS & PALAU, 1999). Also, all theenzymes from the GH-H clan have a domain(the so-called B domain) intercalated between thethird β-strand and the third α-helix of the cat-alytic domain (see Figure 1). Most members ofthe AAMY superfamily also show a well-conservedsequence pattern near the C-terminus of the B

1CGT 1QHO 1HX0 1JAE

1AQM 7TAA 1BAG 1HVX

1UOK 1SMA 1BVZ 1AVA

1EH9 1GCY 1BF2 1ESW

Fig. 2. Different folds for the B domain in the clan GH-H structures. The figure was produced with MOLSCRIPTv2.1.2 (KRAULIS, 1991) and the PDB files indicated at the top of each B domain. Some B domains are represen-tative of a group of clan GH-H structures: 1CGT for CGTases, 1HX0 for mammalian AAMYs, 7TAA for fungalAAMYs and 1HVX for AAMYs from cluster BV (see Figure 3).

45

domain (JANECEK, 1992, 1995a). This conservedmotif has not been identified in the glycogenbranching and debranching enzymes (JANECEK etal., 1997), maltosyltransferase and maltooligosyl-trehalose synthase (JANECEK, 2000b). Sequencecomparison between enzymes and non-enzymaticmembers of the AAMY superfamily shows thatthe non-enzymatic members probably have a TIM-barrel domain. The same comparison shows thatthe B domain is also found in the transport pro-teins, although part of it seems to be absent inthe 4F2 heavy-chain antigens (JANECEK et al.,1997; JANECEK, 2000a). The putative B domainof the transport proteins seems to be closely re-lated in sequence and fold to the one from oligo-1,6-glucosidase (JANECEK et al., 1997; JANECEK,2000a).

The sequence length and fold of the B do-mains of enzymes with different activities can bevery different (JESPERSEN et al., 1993; JANECEK

et al., 1997). This is also true for enzymes withthe same activities (PUJADAS & PALAU, 2001)(Fig. 2). Nevertheless, different clan GH-H struc-tures have the same kind of fold for the B domain(JANECEK et al., 1997; JANECEK, 2000b). Sometypes of B domain contain several β-strands andone or two α-helices (e.g. fungi AAMYs). Others(e.g. barley AAMY) have an irregular fold withno well-defined secondary structure elements, andisoamylase has a long region in place of domainB that forms a cluster with the β4 → α4 loop ofthe TIM barrel domain (KATSUYA et al., 1998).Despite this variability, a common origin for all Bdomains has not been discarded (JANECEK et al.,1997). The B domain plays an important role in:(i) controlling the isoenzyme specificity propertiesin barley AAMY [i.e. substrate affinity/binding,catalysis, sensitivity to inhibitors and stability;(RODENBURG et al., 1994)]; (ii) binding the Ca2+

ion in AAMYs and related CGTases (BOEL etal., 1990; MACHIUS et al., 1995, 1998); (iii) thethermostability of the full protein in some Bacil-lus species (HWANG et al., 1997; DECLERCK et al.,2000); and (iv) controlling the product specificityof several Bacillus CGTases (NAKAMURA et al.,1994; SIN et al., 1994; PENNINGA et al., 1995). Thefact that the majority of the stabilizing mutationscluster in the B domain (DECLERCK et al., 2000)has led to speculation that Bacillus AAMYs areinitially inactivated by the partial unfolding of thisdomain (TOMAZIC & KLIBANOV, 1988; NIELSEN

& BORCHERT, 2000).So far, 24 sequences from the GH-H clan,

which cover 9 different enzyme activities and 2GH families, have been crystallized (Fig. 3). We

therefore have enough structural information toinvestigate how the variability of the B domainaffects the folding of the A domain. To do this, westudied the conservation of the geometrical char-acteristics of the β-sheet that forms the interiorof the TIM barrel. This is coherent with the factthat the packing of side chains within the closed β-sheet has been described as one of the most impor-tant factors for maintaining a TIM barrel struc-ture (LESK et al., 1989; WODAK et al., 1990). Ourresults suggest that the folding of the A domaindoes not depend on the characteristics of the Bdomain.

Material and methods

The PDB codes for the crystallized sequences of theGH-H clan were obtained from the CAZy databasefrom 26 July 2001 (COUTINHO & HENRISSAT, 1999).The corresponding protein structures were importedfrom the last on-line release of PDB [http://www.rcsb.org/pdb/; (BERMAN et al., 2000)]. When there wasmore than one structure for the same sequence, theone with best global quality (i.e. best resolution andlowest R-value) was selected for further calculations.In such cases, we did a structural superposition withthe “Best fit” algorithm of Swiss-PdbViewer v3.7 toensure that the β-sheet geometry of the TIM barrelin the selected PDB file was also representative of theother structures of lower quality (GUEX & PEITSCH,1997). We used the same program and algorithm forthe other superpositions needed in this paper. Onlybackbone atoms were used in all the superpositionsbetween structures with different sequences. The rootmean square deviation (rmsd) was used in all cases toquantify the degree of structural similarity. Figure 3shows the PDB code for the structures we selected forthe study.

In TIM-barrel domains, the internal β-sheetforms a closed structure that, in the case of the clanGH-H enzymes, resembles a hyperboloid. The shape ofthis closed β-sheet may be described by the geometryof the sections obtained when the hyperboloid is cutby equidistant planes -or layers- that are perpendicu-lar to the axis of the barrel. Each layer is built of fourresidues belonging to four alternate β-strands. Onlythe residues whose side chains face towards the interiorof the barrel are used to build the layers. The carbonatoms in the α position (CA) of the four residues thatform a layer are roughly located in the same plane.Figure 4 shows a schematic of the four-layered scaffoldin the interior of the TIM barrel of the human sali-vary AAMY (PDB code: 1SMD). The geometry of alayer may be described by the following parameters:(i) the distances between the two pairs of opposite CAin a layer (e.g. Gly39/Arg252 and Gly193/Phe335 forthe first layer in 1SMD); (ii) the angle between the twosegments that define the two pairs of opposite CA; and(iii) the area of the layer. The latter parameter is cal-culated under the assumption that: (i) the layers are

46

elliptical and (ii) the axes of the ellipse are definedby the segments that join the two pairs of oppositeCA in a layer. These parameters may be used to com-pare topologically identical layers from different struc-tures. A parametrization of this kind was very usefulfor studying the evolution of the packing of the inte-rior of the TIM barrel in β-amylase (PUJADAS et al.,1996) and in identifying and characterizing the con-formational change of β-strand 6 upon ligand bindingon the same enzyme (PUJADAS & PALAU, 1997). Wemade all the calculations in an Excel v98 worksheetafter identifying the layers.

The layer distribution in the β-sheet core of crys-tallized AAMYs from family 13 has recently beendescribed (PUJADAS & PALAU, 2001). All crystal-lized AAMYs have a four-layered scaffold in the in-terior of their TIM barrel (see Figure 4). In thisscaffold, the odd β-strands contribute to the secondand fourth layers, whereas the even β-strands con-tribute to the first and third layers. In general, thelayer distribution for the other clan GH-H structureshas been obtained by structural superposition withthe 1SMD structure and further visual examinationof the correctness of the superposition and layer as-signment. This was a valid method for identifying thelayer structure of cyclodextrin glucanotransferase (CG-Tase; 1CXL), maltotetraose-forming amylase (1GCY),isoamylase (1BF2), maltogenic α-amylase (1QHO and1SMA) and neopullulanase (1BVZ). Once we had char-acterized the layer scaffold of 1CXL, we superim-posed the other CGTases in the study (1CGT, 1PAM,1CYG and 1CIU) with 1CXL and identified theirlayers. The “Best fit” algorithm of Swiss-PdbViewerv3.7 (GUEX & PEITSCH, 1997) failed in the struc-tural superposition of 1ESW (4-α-glucanotransferase),1EH9 (α-amylase/glycosyltrehalose trehalohydrolase)and 1UOK (oligo-1,6-glucosidase) both with 1SMD or1CXL. For these “problematic” structures, we usedthe DALI web-server (http://www.ebi.ac.uk/dali/)to investigate the best structural counterparts. Thesewere 1BVZ, 7TAA and 1BF2 for 1ESW, 1EH9 and1UOK, respectively. Once the structural counterpartswere known, the layers for 1ESW, 1EH9 and 1UOKwere found as shown above.

We studied the similarity of the sequences cor-responding to the PDB files from Figure 3 in the fol-lowing way. First, all sequences were shortened to pro-duce informationally similar segments that correspondto the strict (β/α)8-barrel plus B domain (A+B seg-ments) structure (see Figure 1). Therefore, neither theN-terminal tail nor domain C were considered. Theamino-acid segments of reference used to define theA+B sequences were obtained after analyzing the sec-ondary structure information included in the PDBfiles. More specifically, we took the PDB sequencesfrom the first residue in βA1 (first β-strand in the Adomain) to the last residue in βA8 as our reference.Second, we used all the A+B segments to produce twosets of sequences, one corresponding to the strict do-main A and the other to domain B. The boundarybetween domains has recently been obtained for the

AAMYs (PUJADAS & PALAU, 2001). The limits forthe other AAMY superfamily members were obtainedafter visually inspecting the structural superpositionsused to define the layer distribution (see above). Do-main A sequences were obtained by joining the twosubsegments on the left and right of the correspondingdomain B. Pairwise comparisons of all the sequencesfrom the same domain were done with the Clustal Valgorithm (HIGGINS & SHARP, 1989) and the com-mercial program MEGALIGN v4.03 from the Laser-gene software package (1999, DNASTAR, Inc., Lon-don, UK) running in a Macintosh iBook. The similar-ity index (SI) between two sequences was calculatedwith the method of WILBUR & LIPMAN (1983), with agap penalty of 3, a K-tuple of 1, 5 top diagonals and awindow size of 5. The SI was calculated as the numberof exactly-matching residues in this alignment minusa “gap penalty” for every gap introduced. The resultwas then expressed as a percentage of the length of theshorter sequence. The protein weight matrix was PAM250.

Results and discussion

The characteristics of the B domain are extremelydiverseFigure 3 shows that the length of the B domain incrystallized sequences from the GH-H clan variesconsiderably [from 30 residues in the glycosyltre-halose trehalohydrolase from Sulfolobus Solfatari-cus Km1 (PDB code: 1EH9) to 105 residues in theAAMY from Bacillus licheniformis (PDB code:1VJS)]. All possible pairwise comparisons betweenthe sequences of the B domains from crystallizedclan GH-H structures are shown below the diago-nal line in Figure 5.

We can see that the sequences (SI ranges from93.8 to 71.9%) and structures (rmsd ranges from0.65 to 0.29 A; results not shown) of the B do-mains in the CGTases are very similar when theB domain of 1CGT is used as a reference for super-position with the B domains from the other CG-Tases (see Figure 2). These results within CGTasesare consistent with the fact that these structures:(i) were obtained from closely-related bacterialspecies (i.e. from the Bacillus/Clostridium taxo-nomic group); (ii) have highly similar A-domainsequences (SI ranges from 91.1 to 62.1%); and (iii)have a good level of structural superposition forthe A domain (rmsd ranges from 0.95 to 0.72 A;results not shown).

We have recently demonstrated (PUJADAS &PALAU, 2001) that if the sequences from the Adomain are used, AAMYs may be classified intotwo clusters for archaea (AI and AII), eight forbacteria (from BI to BVIII) and three for eu-karyota (EI, EII and EIII). Moreover, if we com-

47

Fig. 3. Clan GH-H structures used in this study. The different structures are sorted according to the followingpriority arrangement: (1) the GH family; (2) the EC number; and (3) the species from which the enzymes havebeen obtained. The following information is provided for the structures: the species, the PDB identification

48

pare the AAMYs from different kingdoms, someof the above-mentioned clusters group to forminter-kingdom clusters (BVIII/EIII, AII/BV/EIand EII/BIV). Figure 5 shows that the SI val-ues of other comparisons between the A domainsfrom AAMY are below the 20% threshold. In theB domain, the similarity is high if we comparesequences from the same cluster (i.e. 2AAA and7TAA; 1VJS and 1HVX; 1HNY, 1SMD, 1HX0 and1JAE). The B domains of clusters EIII and BVIIIare similar in length (71/61 and 60; see Figure 3),sequence (SI from 33.3 to 23.3%) and structure[rmsd from 1.12 to 0.87 A (results not shown);see also Figure 2] and reflect the relationship be-tween the A domain of these clusters (PUJADAS

& PALAU, 2001). We have also demonstrated theclose relationship between the A domains fromclusters BV and EI (PUJADAS & PALAU, 2001).This similarity is not found in B domains (the SIvalues for the sequence comparison are 14.8 and13.1%; Fig. 5) because they are very different inlength (105/102 residues for 1VJX/1HVX and 61for 1AVA; see Figure 3) and fold (Fig. 2). It is alsoremarkable that the B domain from 7TAA (As-pergilus oryzae) has a relatively high sequence sim-ilarity with the one from the inter-kingdom clus-ter formed by BVIII (1AQM) and EIII (1HNY,1SMD, 1HX0 and 1JAE) (see Figure 5). This sim-ilarity also extends to 2AAA (the other structurein the EII cluster) when we compare the structuresof the B domain, not their sequences (rmsd valuesrange from 1.24 to 0.92 A; results not shown). Thisis highly remarkable because AAMYs from fungi(EII) and animals (EIII) diverged a long time ago(JANECEK, 1994b; PUJADAS & PALAU, 2001), asis clearly shown by the low SI of the A domain ofthese AAMYs (15.1–12.0%; see Figure 5).

Two maltogenic AAMYs – one from Bacil-lus stearothermophilus C599 (PDB code: 1QHO)and the other from Thermus sp. IM6501 (PDBcode: 1SMA) – have also been crystallized. When

we compared the sequences for their A domains,their SI was 21.1% (Fig. 5) and their rmsd was1.37 A. The SI and rmsd in the comparison oftheir B domains were 12.7% and 1.38 A respec-tively. In fact, it has recently been suggested thatthe enzymes classed as maltogenic AAMYs underthe EC 3.2.1.133 should be re-classified into twodifferent classes, one that seems to be specific forthe α-1,4-linkage (i.e. 1QHO) and another thatis active on α-1,4 and α-1,6 bonds (i.e. 1SMA)(MACGREGOR et al., 2001). The Bacillus sequencewas very similar to CGTases (the SI ranged from39.9 to 36.0% for the A domain and from 35.9to 26.6% for the B domain). From the structuralpoint of view, the B domain of 1QHO and CG-Tases have the same fold (see Figure 2). Also,the rmsd of the structural superposition of theB domains from the CGTases onto 1QHO rangedfrom 0.96 to 0.86 A, and the maltogenic AAMYfrom Bacillus had a five-domain structure that issimilar to the one often associated with CGTases(DAUTER et al., 1999). From the sequence similar-ity between this maltogenic AAMY and CGTases,JANECEK suggested that the role of the former wasthat of an “intermediary” enzyme among “true”AAMYs and “true” CGTases (JANECEK, 1995b).On the other hand, the maltogenic AAMY fromThermus sp. (PDB code: 1SMA) was very similarto the neopullulanase from Thermoactinomycesvulgaris (PDB code: 1BVZ) in terms of sequence(54.7 and 50.9% for SI when we compared theirA or B domains) and structure (rmsd was 1.17 or0.91 A when we compared either the A or the Bdomains; see also Figure 2). The close relationshipbetween some maltogenic AAMYs and the neop-ullulanase from Thermoactinomyces vulgaris hasrecently been noted (MATZKE et al., 2000). It hasalso been suggested that these sequences – alongwith other sequences from cyclomaltodextrinases– constitute a subfamily within the GH-H clan(MATZKE et al., 2000). At present, it is suggested

(code and chain), the parameters for globally evaluating the structural quality (resolution and R-value), theaccession number in the Swiss-Prot/TrEMBL databases for the sequence that matches the PDB structure, andthe position of the B domain in the corresponding PDB file (initial and terminal residues and length of thedomain). In the case of AAMYs, the last column includes the cluster in which the corresponding A domainsequences have been classified (PUJADAS & PALAU, 2001). The references for the structures are: 1CGT (KLEIN

& SCHULZ, 1991), 1CXL (UITDEHAAG et al., 1999), 1PAM (HARATA et al., 1996), 1CIU (KNEGTEL et al.,1996), 2AAA (BOEL et al., 1990), 7TAA (BRZOZOWSKI & DAVIES, 1997), 1VJS (HWANG et al., 1997), 1HVX(SUVD et al., 2001), 1BAG (FUJIMOTO et al., 1998), 1HNY (BRAYER et al., 1995), 1SMD (RAMASUBBU et al.,1996), 1AVA (VALLEE et al., 1998), 1AQM (AGHAJARI et al., 1998), 1HX0 (QIAN et al., 2001), 1JAE (STROBL

et al., 1998), 1EH9 (FEESE et al., 2000), 1UOK (WATANABE et al., 1997), 1GCY (MEZAKI et al., 2001), 1BF2(KATSUYA et al., 1998), 1QHO (DAUTER et al., 1999), 1SMA (KIM et al., 1999), 1BVZ (KAMITORI et al., 1999)and 1ESW (PRZYLAS et al., 2000).

49

LAYER 2

LAYER 3

ILE 13

Val 14

HIS 15

GLY 39

Val 40

GLN 41

TYR 94

Val 95

ASP 96

GLY 193

Phe 194

ARG 195

TYR 231

Gln 232

GLU 233

ARG 252

Val 253

THR 254

LEU 293

Val 294

PHE 295

PHE 335

Thr 336

ARG 337

LAYER 1

LAYER 4

ββββ-strand 1 ββββ-strand 3 ββββ-strand 5 ββββ-strand 7

ββββ-strand 2 ββββ-strand 4 ββββ-strand 6 ββββ-strand 8

ββββ-stranddirection

barrelaxis

Fig. 4. Schematic of the four-layered scaffold in the interior of the TIM barrel of the human salivary AAMY(PDB code: 1SMD) produced by cutting along the hydrogen bonds between βA1 and βA8 and unrolling theβ-sheet onto a flat surface. Only segments that are three residues long are shown for each β-strand. Each layer ismade up of four residues (indicated by four black points at the same horizontal level), which belong to alternatestrands (one residue for each strand) and which face the interior of the barrel. Each β-strand contributes to twodifferent layers with two alternate residues, whereas the middle residue (in white) faces the external helices orcoils.

that the two maltogenic AAMYs should be classi-fied using the label “AM-type” (i.e. amylase-like)for the one from Bacillus and the label “N-type”(i.e. neopullulanase-like) for the one from Thermussp. (MACGREGOR et al., 2001).

The results in this section therefore supportprevious findings that, in general, the evolutionof the B domain matches that of the A domain,which therefore indicates that the insertion of theB domain in the TIM barrel is not a recent event(JANECEK et al., 1997). Some “deviations” wehave detected to this general rule are: (i) the lackof similarity and the differences in the sequencelength of the B domains in AAMYs from clustersBV and EI (PUJADAS & PALAU, 2001) do notmatch the similarities in their A domains; and (ii)the similarities between the structure and, to alesser extent, the sequence of the B domains fromAAMYs in cluster EII and those from the inter-kingdom cluster formed by BVIII and EIII.

Conservation of the barrel geometryThe packing of side chains within the closed β-sheet is one of the most important factors formaintaining a TIM barrel structure (LESK etal., 1989; WODAK et al., 1990). Any possible“external” influence on the folding of a TIM-

barrel domain can therefore be analyzed by study-ing how the geometrical characteristics of theclosed β-sheet are affected. The method used toparametrize this β-sheet is described in the Ma-terial and methods section (see above). The factthat the different clan GH-H structures share thesame geometry for the TIM-barrel β-sheet (i.e. ahyperboloid shape and four layers) makes it easierto compare the data from the structures in Fig-ure 3.

Figure 4 shows a schematic diagram of thefour-layered scaffold in the interior of the TIMbarrel of the human salivary AAMY (PDB code:1SMD). The four-layered scaffold for the otherstructures in Figure 3 can be inferred by com-bining the information in Figures 4 and 6. Figure6 shows that only a few positions have preservedthe same residue throughout evolution in all re-ported structures [i.e. Asp96, Arg195 and Glu233in β3, β4 and β5 respectively: the numbers cor-respond to the location of these residues in 1HX0(the pig pancreatic AAMY sequence); note thatGln2571CXL and Gln2081BAG in β5 correspond tonon-natural mutations]. Another position is alsowell-, though not strictly, preserved (i.e. Leu289 inthe β4 of 1ESW is different from the Gly residuefound in the other clan GH-H structures).

50

1CGT 1CXL 1PAM 1CYG 1CIU 2AAA 7TAA 1VJS 1HVX 1BAG 1HNY 1SMD 1AVA 1AQM 1HX0 1JAE 1EH9 1UOK 1GCY 1BF2 1QHO 1SMA 1BVZ 1ESW

1CGT 73,2 72,5 62,1 68,6 21,7 22,1 15,4 17,4 19,4 13,1 13,1 14,6 17,5 14,7 15,8 11,4 10,7 16,7 14,6 39,9 18,1 22,2 11,7

1CXL 79,7 91,1 68,2 70,4 18,6 24,3 17,0 18,6 19,0 14,3 14,3 15,7 17,9 14,7 16,2 12,5 14,6 17,0 13,2 37,8 20,4 23,3 9,6

1PAM 78,1 93,8 67,1 70,4 22,8 25,1 17,4 17,4 19,0 15,4 15,1 15,7 16,7 15,8 17,0 12,9 17,1 16,7 12,5 36,7 21,1 22,9 10,7

1CYG 76,6 75,0 71,9 69,3 21,7 22,4 15,0 15,9 21,1 16,6 16,6 16,1 15,4 16,6 16,2 13,2 12,9 16,7 14,3 36,0 21,9 24,8 12,1

1CIU 78,1 75,0 76,6 85,9 19,4 22,8 17,0 16,7 17,8 14,7 14,3 14,6 14,2 14,7 13,4 12,9 15,0 14,8 13,2 38,8 19,2 22,9 10,0

2AAA 12,1 12,1 12,1 13,8 13,8 71,1 15,4 17,4 14,2 12,4 12,0 14,9 13,3 13,1 14,6 16,7 13,3 14,4 14,4 20,9 19,4 16,3 9,9

7TAA 12,1 13,8 12,1 12,1 12,1 65,5 17,0 18,6 15,4 14,7 14,3 16,9 12,9 15,1 13,8 15,6 15,2 12,9 12,9 24,3 20,5 17,5 11,0

1VJS 18,8 17,2 17,2 15,6 17,2 20,7 15,5 59,7 16,6 14,6 12,6 20,6 16,7 14,2 14,2 15,8 12,3 14,6 13,8 16,6 14,6 12,6 11,1

1HVX 15,6 15,6 14,1 15,6 15,6 19,0 13,8 72,5 15,8 12,4 12,4 20,9 14,6 12,8 13,4 16,3 11,6 18,6 12,4 14,7 17,1 14,7 12,0

1BAG 18,4 18,4 18,4 18,4 18,4 16,3 16,3 18,4 16,3 17,4 17,8 17,4 18,3 19,0 17,8 13,0 10,9 15,4 17,4 16,2 13,0 14,6 11,3

1HNY 14,1 15,6 15,6 15,6 15,6 15,5 24,1 14,1 12,7 22,4 98,5 13,9 51,7 90,7 56,9 16,2 11,2 14,3 16,2 14,3 11,2 13,5 11,6

1SMD 14,1 15,6 15,6 15,6 15,6 15,5 24,1 14,1 12,7 22,4 97,2 13,9 52,5 91,1 56,9 15,8 11,2 14,3 15,8 14,7 11,2 13,1 11,6

1AVA 14,8 13,1 13,1 13,1 13,1 13,8 15,5 14,8 13,1 10,2 14,8 14,8 16,7 13,5 14,2 13,0 14,2 23,8 13,8 16,9 13,8 14,6 11,5

1AQM 15,0 15,0 15,0 15,0 15,0 19,0 24,1 11,7 15,0 20,4 23,3 23,3 11,7 55,0 42,5 15,0 10,4 15,8 15,4 14,6 11,7 12,9 10,4

1HX0 15,6 14,1 12,5 15,6 17,2 13,8 19,0 14,1 14,1 18,4 76,1 74,6 11,5 28,3 57,7 15,4 11,6 13,9 15,4 15,4 11,2 13,1 11,6

1JAE 19,7 16,4 14,8 13,1 16,4 15,5 20,7 11,5 11,5 20,4 31,1 32,8 11,5 33,3 29,5 16,2 12,3 14,6 14,6 13,8 13,0 15,0 11,5

1EH9 16,7 20,0 16,7 16,7 16,7 20,0 20,0 16,7 16,7 13,3 16,7 16,7 16,7 16,7 16,7 16,7 16,4 11,0 14,9 15,5 18,1 17,3 12,8

1UOK 12,5 15,6 15,6 14,1 12,5 12,1 12,1 12,5 15,3 16,3 12,7 12,7 11,5 16,7 12,7 13,1 23,3 10,2 12,6 16,2 17,4 18,4 11,6

1GCY 16,0 20,0 20,0 16,0 18,0 14,0 16,0 16,0 16,0 16,3 20,0 22,0 16,0 16,0 16,0 16,0 16,7 22,0 13,6 15,9 14,4 11,7 12,9

1BF2 17,0 17,0 15,1 17,0 18,9 11,3 15,1 15,1 15,1 12,2 15,1 15,1 13,2 13,2 15,1 13,2 16,7 15,1 14,0 16,2 10,9 16,9 10,1

1QHO 35,9 28,1 26,6 28,1 34,4 19,0 15,5 14,1 11,3 18,4 15,5 14,1 14,8 18,3 15,5 13,1 20,0 15,5 20,0 17,0 21,1 26,3 11,2

1SMA 10,9 10,9 10,9 12,7 9,1 12,7 18,2 12,7 10,9 10,2 14,5 14,5 14,5 9,1 12,7 10,9 16,7 16,4 14,0 13,2 12,7 54,7 14,0

1BVZ 10,7 10,7 10,7 8,9 8,9 14,3 12,5 16,1 14,3 10,2 12,5 12,5 10,7 10,7 10,7 8,9 16,7 10,7 12,0 11,3 14,3 50,9 12,4

1ESW 12,0 14,0 12,0 10,0 12,0 12,0 14,0 18,0 14,0 16,3 18,0 18,0 18,0 14,0 16,0 10,0 20,0 12,0 10,0 12,0 14,0 18,0 14,0

Fig. 5. Sequence comparison of all A-domain sequences in clan GH-H structures (above the diagonal line) and all B-domain sequences from the samestructures (below the diagonal line). SI corresponding to the sequence comparison of all A-domain sequences in the clan GH-H structures of Figure 3(above the diagonal line) and of all B domain-sequences from the same structures (below the diagonal line). Sequences from the polypeptide chains inthe PDB files correspond to those in Figure 3 and are also sorted according to this figure arrangement. SI values that correspond to sequences from thesame EC number are highlighted. A special highlight is also used for the SI values from AAMYs that belong to the same cluster or to inter-kingdomrelated clusters (i.e. BVIII/EIII and EI/BV).

51

Fig. 6. β-Strand segments involved in the packing of the TIM-barrel β-sheet in clan GH-H structures. The firstand third residues of each of the eight segments (highlighted by inversion) face the interior of the barrel andare members of two different layers (see Figure 4). The second residue of the segments faces the external helicesor coils. Residues that are in the same layer but different β-strand are in the same row and highlighted by thesame type of inversion. The combined information of this figure and Figure 4 (built with the 1SMD structure)may easily be used to infer the four-layered scaffold in all clan GH-H structures. The sequence number in thecorresponding PDB file for the residue that starts each segment is indicated before each one. PDB files are listedin the same order as in Figure 3. PDB files with the same EC number are grouped.

From the results in Figure 7 we can comparethe degrees of conservation of the β-sheet in thestructures in Figure 3. The level of variation isgenerally very low if we take into account that:(i) the residues that form the layers have —ingeneral— not been conserved throughout evolu-tion (see above); (ii) the similarity between thedomain A sequences is generally low (see Figure 5and the previous section); and (iii) the sequenceof the TIM barrel is interrupted by the presenceof the highly variable B domain. Figure 7 alsoshows that the geometry of the structure from thefamily GH-77 (PDB code: 1ESW) strongly agreeswith that from the family GH-13. This is consis-

tent with a common origin for the two families,which shows that both are part of the same GHclan (i.e. clan GH-H), and the idea that struc-tural characteristics are better preserved through-out evolution than the characteristics of the se-quence (SCHULZ & SCHIRMER, 1979).

Conclusions

Our results show that the length, fold and se-quence of the B domain only slightly affect thegeometry of the closed β-sheet that forms the coreof the TIM-barrel domain. It should be noted thatthe B domain is a long loop, not an independent

52

Fig. 7. Distances (in A) between pairs of layer-forming CA-atoms from opposite β-strands (first and secondcolumn for each layer), angles (in degrees) formed between the corresponding distance segments (third columnfor each layer) and area (in A2) if an elliptical shape is assumed (fourth column for each layer). Note: βyx, xindicates the position of the β-strand in the TIM-barrel structure and y indicates the position of the residue inthe corresponding three-residues-long β-strand segment (see Figure 6). PDB files are listed in the same order asin Figure 3. PDB files with the same EC number are grouped.

folding unit or a “true” domain. The almost nulleffect of the β3 → α3 “loop” on the geometry ofthe β-sheet where it is inserted is consistent withthe usual behavior of the other loops in TIM-barrelstructures. A key point in our discussion is there-fore to show convincing arguments for consideringthe β3→ α3 “loop” as a “true” domain. One wayto do this is to show that this “loop” is used asa module by carbohydrate-active enzymes otherthan those in the GH-H clan. GHs and glycosyl-transferases have been shown frequently to displaya plurimodular structure in which ancillary non-catalytic modules are found beside the catalyticdomain (HENRISSAT & DAVIES, 2000). Prelimi-nary FASTA results (PEARSON & LIPMAN, 1988)that use the sequence of the β3→ α3 “loop” fromthe structures in Figure 3 as the seeds in similaritysearches do not prove its possible modular use (re-

sults not shown). However, the limitations inher-ent in using in the searches one small subset of allknown GH-H sequences (the ones from crystallizedGH-H structures) does not completely discard theuse of the β3 → α3 “loop” as a module. On theother hand, the B domain may not have been usedas a building module by other carbohydrate-activeenzymes from other clans.

Another indirect way of showing that theα3→ α3 “loop” is a “true” domain involves find-ing clan GH-H members whose “loop” has a verylong sequence. Such sequences exist; for instance,the β3 → α3 “loop” reaches 250 residues in thecase of the human glycogen debranching enzyme[Swiss-Prot accession number: P35573 (YANG etal., 1992; JESPERSEN et al., 1993; JANECEK et al.,1997)]. Although such a long “loop” is expectedto constitute a “true” domain in the structure of

53

A)B)

Fig. 8. Schematic of the structure of the A+B-domains in the maltosyltransferase from Thermotoga maritima.See Figure 1 for the criteria for representing the α-helices, the β-strands, the A- and B-domains and the differentviews. The spacefilling model of the maltose ligand indicates the position of the active site (at the C-terminalside of the β-sheet in the core of the TIM-barrel fold). The figure was produced with MOLSCRIPT v2.1.2(KRAULIS, 1991) and the PDB file 1GJW.

this enzyme, we cannot prove this due to the lackof structural information. Fortunately, LIEBL andcoworkers have very recently described the struc-ture of the maltosyltransferase from Thermotogamaritima [PDB code: 1GJW; (ROUJEINIKOVA etal., 2001)], which has a β3 → α3 “loop” that is155 residues long. This is the longest α3 → α3“loop” so far crystallized (see Figure 3) and ittherefore provides an excellent base for investi-gating whether it can or cannot be considered a“true” domain. Visual inspection of the PDB fileclearly shows that the α3 → α3 “loop” consti-tutes a folded unit that is independent of the TIM-barrel domain (Fig. 8). A similar result is foundwith two other crystallized GH-H structures witha long α3→ α3 “loop” (i.e. 1VJS and 1HVX; seeFigure 1 and Figure 3). We can therefore concludethat there is enough “structural” evidence to showthat, at least in some GH-H enzymes, the α3→ α3“loop” is a “true” domain (i.e. the concept of theB domain is structurally supported).

Maintaining the TIM-barrel scaffold of clanGH-H proteins or achieving the correct fold seemsto be a matter of preserving (i) a sequence thatcan adopt the characteristic secondary-structurepattern of domain A and (ii) a set of residuesthat belong to β-strands at positions that are farapart in the sequence but which come together be-cause of medium- and long-range interactions. Un-wanted mutations leading to noncanonical or mis-folded structures may be recognized by molecular

chaperones and discarded as functional molecules(ELLIS & HARTL, 1996; WELCH & BROWN, 1996).At this point it would be interesting to ascer-tain why the B domain – with different folds andlengths and inserted in the middle of the domainA sequence – has an almost unappreciable ef-fect on the folding of the TIM barrel (see Fig-ure 7). Our findings are difficult to understand ifthe molecular chaperones do not help to fold eachdomain correctly and independently of each otherusing a step-by-step mechanism of sequential do-main folding. This conclusion is also supported bythe geometrical analysis of the TIM-barrel β-sheetin the recently-described maltosyltransferase fromThermotoga maritima, where the B domain is thelongest in crystallized clan GH-H structures (re-sults not shown). It would obviously be very in-teresting to know whether the geometry of the β-sheet barrel core also remains unchanged in thecircularly permuted members of the GH-H clan(i.e. GHs from family 70). In these GHs, the B do-main is divided to precede and succeed the TIMbarrel and, as a result of the permutation, there isan insertion of 140–150 residues in the α6 → β6loop, [equivalent respectively to α8 and β1 in non-permuted clan GH-H sequences; (MACGREGOR

et al., 1996)]. Unfortunately, no structural infor-mation is yet available for this family. We there-fore think that chaperones preserve the geometri-cal characteristics of the TIM barrel which are notaffected by the variability of domain B. Our hy-

54

pothesis about the role of chaperones in clan GH-Hevolution is consistent with CSERMELY’s hypothe-sis (CSERMELY, 1997) that holds that chaperonesare “mandatory for the evolution of our present-day catalysts”.

Acknowledgements

I thank Kevin COSTELLO of the Language Service ofour University for his help in writing the manuscript;Dr. Richard HASER and Dr. Bernard HENRISSAT fortheir comments, opinions and suggestions, which wereinvaluable for writing the conclusions; and Dr. StefanJANECEK for inviting me to attend the ALAMY Sym-posium. Finally, this work is dedicated to the memoryof my friend, Prof. Jaume PALAU, who with these re-sults made his last contribution to science.

References

AGHAJARI, N., FELLER, G., GERDAY, C. & HASER,R. 1998. Crystal structures of the psychrophilic α-amylase from Alteromonas haloplanctis in its na-tive form and complexed with an inhibitor. ProteinSci. 7: 564–572.

BERMAN, H. M., WESTBROOK, J., FENG, Z., GILLIL-AND, G., BHAT, T. N., WEISSIG, H., SHINDYALOV,I. N. & BOURNE, P. E. 2000. The Protein DataBank. Nucleic Acids Res. 28: 235–242.

BERTRAN, J., WERNER, A., MOORE, M. L., STANGE,G., MARKOVICH, D., BIBER, J., TESTAR, X.,ZORZANO, A., PALACIN, M. & MURER, H. 1992.Expression cloning of a cDNA from rabbit kidneycortex that induces a single transport system forcystine and dibasic and neutral amino acids. Proc.Natl. Acad. Sci. USA 89: 5601–5605.

BOEL, E., BRADY, L., BRZOZOWSKI, A. M., DE-REWENDA, Z., DODSON, G. G., JENSEN, V. J.,PETERSEN, S. B., SWIFT, H., THIM, L. & WOL-DIKE, H. F. 1990. Calcium binding in α-amylases:an X-ray difraction study at 2.1 A resolution oftwo enzymes from Aspergillus. Biochemistry 29:6244–6249.

BRAYER, G. D., LUO, Y. & WITHERS, S. G. 1995. Thestructure of human pancreatic α-amylase at 1.8 Aresolution and comparisons with related enzymes.Protein Sci. 4: 1730–1742.

BRZOZOWSKI, A. M. & DAVIES, G. J. 1997. Structureof the Aspergillus oryzae α-amylase complexedwith the inhibitor acarbose at 2.0 A resolution.Biochemistry 36: 10837–10845.

COUTINHO, P. M. & HENRISSAT, B. 1999. Carbohyd-rate-Active Enzymes server at URL: http://afmb.cnrs-mrs.fr/CAZY/index.html.

CSERMELY, P. 1997. Proteins, RNAs and chaperonesin enzyme evolution: a folding perspective. TrendsBiochem. Sci. 22: 147–149.

DAUTER, Z., DAUTER, M., BRZOZOWSKI, A. M.,CHRISTENSEN, S., BORCHERT, T. V., BEIER, L.,

WILSON, K. S. & DAVIES, G. J. 1999. X-ray struc-ture of Novamyl, the five-domain “maltogenic” α-amylase from Bacillus stearothermophilus: maltoseand acarbose complexes at 1.7 A resolution. Bio-chemistry 38: 8385–8392.

DECLERCK, N., MACHIUS, M., WIEGAND, G., HUBER,R. & GAILLARDIN, C. 2000. Probing structuraldeterminants specifying high thermostability inBacillus licheniformis α-amylase. J. Mol. Biol.301: 1041–1057.

ELLIS, R. J. & HARTL, F. U. 1996. Protein foldingin cell: competing models of chaperonin function.FASEB J. 10: 20–26.

FEESE, M. D., KATO, Y., TAMADA, T., KATO, M.,KOMEDA, T., MIURA, Y., HIROSE, M., HONDO, K.,KOBAYASHI, K. & KUROKI, R. 2000. Crystal struc-ture of glycosyltrehalose trehalohydrolase from thehyperthermophilic archaeum Sulfolobus solfatari-cus. J. Mol. Biol. 301: 451–464.

FRIEDBERG, F. 1983. On the primary structure of amy-lases. FEBS Lett. 152: 139–140.

FUJIMOTO, Z., TAKASE, K., DOUI, N., MOMMA, M.,MATSUMOTO, T. & MIZUNO, H. 1998. Crystalstructure of a catalytic-site mutant α-amylase fromBacillus subtilis complexed with maltopentaose. J.Mol. Biol. 277: 393–407.

GUEX, N. & PEITSCH, M. C. 1997. SWISS-MODELand the Swiss-PdbViewer: an environment forcomparative protein modeling. Electrophoresis 18:2714–2723.

HARATA, K., HAGA, K., NAKAMURA, A., AOYAGI, M.& YAMANE, K. 1996. X-ray structure of cyclodex-trin glucanotransferase from alkalophilic Bacillussp 1011. Comparison of two independent moleculesat 1.8 A resolution. Acta Crystallogr. D52: 1136–1145.

HENRISSAT, B. 1991. A classification of glycoside hy-drolases based on amino acid sequence similarity.Biochem. J. 280: 309–316.

HENRISSAT, B. & DAVIES, G. J. 2000. Glycoside hy-drolases and glycosyltransferases. Families, mod-ules, and implications for genomics. Plant Physiol.124: 1515–1519.

HIGGINS, D. G. & SHARP, P. M. 1989. Fast and sensi-tive multiple sequence alignments on a microcom-puter. Comput. Appl. Biosci. 5: 151–153.

HWANG, K. Y., SONG, H. K., CHANG, C., LEE, J., LEE,S. Y., KIM, K. K., CHOE, S., SWEET, R. M. & SUH,S. W. 1997. Crystal structure of thermostable α-amylase from Bacillus licheniformis refined at 1.7angstrom resolution. Mol. Cells 7: 251–258.

JANECEK, S. 1992. New conserved amino acid regionof α-amylases in the 3rd loop of their (β/α)8-barreldomains. Biochem. J. 288: 1069–1070.

JANECEK, S. 1994a. Parallel β/α-barrels of α-amylase,cyclodextrin glycosyltransferase and oligo-1,6-glu-cosidase versus the barrel of β-amylase: evolution-ary distance is a reflection of unrelated sequences.FEBS Lett. 353: 119–123.

55

JANECEK, S. 1994b. Sequences similarities and evolu-tionary relationships of microbial, plant and ani-mal α-amylases. Eur. J. Biochem. 224: 519–524.

JANECEK, S. 1995a. Close evolutionary relatednessamong functionally distantly related members ofthe (α/β)8-barrel glycosyl hydrolases suggested bythe similarity of their fifth conserved sequence re-gion. FEBS Lett. 377: 6–8.

JANECEK, S. 1995b. Tracing the evolutionary lineagesamong α-amylases and cyclodextrin glycosyltrans-ferases: the question of so-called “intermediary”enzymes. Biologia, Bratislava 50: 515–522.

JANECEK, S. 2000a. Proteins without enzymatic func-tion with sequence relatedness to the α-amylasefamily. Trends Glycosci. Glycotechnol. 12: 363–371.

JANECEK, S. 2000b. Structural features and evolution-ary relationships in the α-amylase family, pp. 19–54. In: OHNISHI, M., HAYAHI, T., ISHIJIMA, S. &KURIKI, T. (eds) Glycoenzymes, Japan ScientificSocieties Press, Tokyo.

JANECEK, S., SVENSSON, B. & HENRISSAT, B. 1997.Domain evolution in the α-amylase family. J. Mol.Evol. 45: 322–331.

JESPERSEN, H. M., MACGREGOR, E. A., HENRISSAT,B., SIERKS, M. R. & SVENSSON, B. 1993. Starch-and glycogen-debranching and branching enzymes:prediction of structural features of the catalytic(β/α)8-barrel domain and evolutionary relation-ship to other amylolytic enzymes. J. Protein Chem.12: 791–805.

KAMITORI, S., KONDO, S., OKUYAMA, K., YOKOTA,T., SHIMURA, Y., TONOZUKA, T. & SAKANO, Y.1999. Crystal structure of Thermoactinomyces vul-garis R-47 α-amylase II (TVAII) hydrolyzing cy-clodextrins and pullulan at 2.6 A resolution. J.Mol. Biol. 287: 907–921.

KATSUYA, Y., MEZAKI, Y., KUBOTA, M. & MAT-SUURA, Y. 1998. Three-dimensional structure ofPseudomonas isoamylase at 2.2 A resolution. J.Mol. Biol. 281: 885–897.

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 mal-togenic amylase provides insights into a catalyticversatility. J. Biol. Chem. 274: 26279–26286.

KLEIN, C. & SCHULZ, G. E. 1991. Structure of cy-clodextrin glycosyltransferase refined at 2.0 A res-olution. J. Mol. Biol. 217: 737–750.

KNEGTEL, R. M. A., WIND, R. D., ROZEBOOM, H. J.,KALK, K. H., BUITELAAR, R. M., DIJKHUIZEN, L.& DIJKSTRA, B. W. 1996. Crystal structure at 2.3A resolution and revised nucleotide sequence of thethermostable cyclodextrin glycosyltransferase fromThermoanaerobacterium thermosulfurigenes EM1.J. Mol. Biol. 256: 611–622.

KRAULIS, P. J. 1991. MOLSCRIPT: A program to pro-duce both detailed and schematic plots of proteinstructures. J. Appl. Crystallogr. 24: 946–950.

LESK, A. M., BRÄNDÉN, C. I. & CHOTHIA, C. 1989.Structural principles of α/β barrel proteins: thepacking of the interior of the sheet. Proteins 5:139–148.

MACGREGOR, E. A., JANECEK, S. & SVENSSON, B.2001. Relationship of sequence and structure tospecificity in the α-amylase family of enzymes.Biochim. Biophys. Acta 1546: 1–20.

MACGREGOR, E. A., JESPERSEN, H. M. & SVENSSON,B. 1996. A circularly permuted α-amylase-typeα/β-barrel structure in glucan-synthesizing gluco-syltransferases. FEBS Lett. 378: 263–266.

MACHIUS, M., DECLERCK, N., HUBER, R. & WIE-GAND, G. 1998. Activation of Bacillus licheni-formis α-amylase through a disorder->order tran-sition of the substrate-binding site mediated by acalcium-sodium-calcium metal triad. Structure 6:281–292.

MACHIUS, M., WIEGAND, G. & HUBER, R. 1995. Crys-tal structure of calcium-depleted Bacillus licheni-formis α-amylase at 2.2 A resolution. J. Mol. Biol.246: 545–559.

MATZKE, J., HERRMANN, A., SCHNEIDER, E. & BA-KKER, E. P. 2000. Gene cloning, nucleotide se-quence and biochemical properties of a cytoplasmiccyclomaltodextrinase (neopullulanase) from Ali-cyclobacillus acidocaldarius, reclassification of agroup of enzymes. FEMS Microbiol. Lett. 183: 55–61.

MEZAKI, Y., KATSUYA, Y., KUBOTA, M. & MATSUU-RA, Y. 2001. Crystallization and structural anal-ysis of intact maltotetraose-forming exo-amylasefrom Pseudomonas stutzeri. Biosci. Biotechnol.Biochem. 65: 222–225.

NAKAJIMA, R., IMANAKA, T. & AIBA, S. 1986. Com-parison of amino acid sequences of eleven differentα-amylases. Appl. Microbiol. Biotechnol. 23: 355–360.

NAKAMURA, A., HAGA, K. & YAMANE, K. 1994. Fouraromatic residues in the active center of cyclodex-trin glucanotransferase from alkalophilic Bacillussp. 1011: effects of replacements on substrate bind-ing and cyclization characteristics. Biochemistry33: 9929–9936.

NIELSEN, J. E. & BORCHERT, T. V. 2000. Protein engi-neering of bacterial α-amylases. Biochim. Biophys.Acta 1543: 253–274.

PEARSON, W. R. & LIPMAN, D. J. 1988. Improvedtools for biological sequence comparison. Proc.Natl. Acad. Sci. USA 85: 2444–2448.

PENNINGA, D., STROKOPYTOV, B., ROZEBOOM, H.J., LAWSON, C. L., DIJKSTRA, B. W., BERGSMA,J. & DIJKHUIZEN, L. 1995. Site-directed mutationsin tyrosine 195 of cyclodextrin glycosyltransferasefrom Bacillus circulans strain 251 affect activityand product specificity. Biochemistry 34: 3368–3376.

PRZYLAS, I., TERADA, Y., FUJII, K., TAKAHA, T.,SAENGER, W. & STRATER, N. 2000. X-ray struc-ture of acarbose bound to amylomaltase from

56

Thermus aquaticus. Implications for the synthe-sis of large cyclic glucans. Eur. J. Biochem. 267:6903–6913.

PUJADAS, G. & PALAU, J. 1997. Anatomy of a con-formational transition of β-strand 6 in soybean β-amylase caused by substrate (or inhibitor) bindingto the catalytical site. Protein Sci. 6: 2409–2417.

PUJADAS, G. & PALAU, J. 1999. TIM barrel fold:structural, functional and evolutionary character-istics in natural and designed molecules. Biologia,Bratislava 54: 231–254.

PUJADAS, G. & PALAU, J. 2001. Evolution of α-amylases: architectural features and key residues inthe stabilization of the (β/α)8 scaffold. Mol. Biol.Evol. 18: 38–54.

PUJADAS, G., RAMÍREZ, F. M., VALERO, R. & PALAU,J. 1996. Evolution of β-amylase: patterns of vari-ation and conservation in subfamily sequences inrelation to parsimony mechanisms. Proteins 25:456–472.

QIAN, M. X., HASER, R. & PAYAN, F. 1993. Structureand molecular model refinement of pig pancreaticα-amylase at 2.1 A resolution. J. Mol. Biol. 231:785–799.

QIAN, M. X., NAHOUM, V., BONICEL, J., BISCHOFF,H., HENRISSAT, B. & PAYAN, F. 2001. Enzyme-catalyzed condensation reaction in a mammalianα-amylase. High-resolution structural analysis ofan enzyme-inhibitor complex. Biochemistry 40:7700–7709.

QUACKENBUSH, E., CLABBY, M., GOTTESDIENER, K.M., BARBOSA, J., JONES, N. H., STROMINGER,J. L., SPECK, S. & LEIDEN, J. M. 1987. Molec-ular cloning of complementary DNAs encoding theheavy chain of the human 4F2 cell-surface antigen:a type II membrane glycoprotein involved in nor-mal and neoplastic cell growth. Proc. Natl. Acad.Sci. USA 84: 6526–6530.

RAMASUBBU, N., PALOTH, V., LUO, Y. G., BRAYER,G. D. & LEVINE, M. J. 1996. Structure of humansalivary α-amylase at 1.6 A resolution: implicationsfor its role in the oral cavity. Acta Crystallogr.D52: 435–446.

RODENBURG, K. W., JUGE, N., GUO, X. J., SOGAARD,M., CHAIX, J. C. & SVENSSON, B. 1994. DomainB protruding at the third β strand of the α/β bar-rel in barley α-amylase confers distinct isozyme-specific properties. Eur. J. Biochem. 221: 277–284.

ROUJEINIKOVA, A., RAASCH, C., BURKE, J., BAKER,P. J., LIEBL, W. & RICE, D. W. 2001. The crystalstructure of Thermotoga maritima maltosyltrans-ferase and its implications for the molecular basisof the novel transfer specificity. J. Mol. Biol. 312:119–131.

SCHULZ, G. E. & SCHIRMER, R. H. 1979. Principles ofProtein Structure. Springer-Verlag, New York.

SIN, K. A., NAKAMURA, A., MASAKI, H., MATSUURA,Y. & UOZUMI, T. 1994. Replacement of an aminoacid residue of cyclodextrin glucanotransferase ofBacillus ohbensis doubles the production of γ-cyclodextrin. J. Biotechnol. 32: 283–288.

STROBL, S., MASKOS, K., BETZ, M., WIEGAND, G.,HUBER, R., GOMIS RUTH, F. X. & GLOCKSHUBER,

R. 1998. Crystal structure of yellow meal worm α-amylase at 1.64 A resolution. J. Mol. Biol. 278:617–628.

SUVD, D., FUJIMOTO, Z., TAKASE, K., MATSUMURA,M. & MIZUNO, H. 2001. Crystal structure of Bacil-lus stearothermaphilus α-amylase: possible factorsdetermining the thermostability. J.Biochem. 129:461–468.

TAKATA, H., KURIKI, T., OKADA, S., TAKESADA,Y., IIZUKA, M., MINAMIURA, N. & IMANAKA,T. 1992. Action of neopullulanase. Neopullulanasecatalyzes both hydrolysis and transglycosylation atα-(1->4)- and α-(1->6)-glucosidic linkages. J. Biol.Chem. 267: 18447–18452.

TOMAZIC, S. J. & KLIBANOV, A. M. 1988. Mechanismsof irreversible thermal inactivation of Bacillus α-amylases. J. Biol. Chem. 263: 3086–3091.

UITDEHAAG, J. C. M., MOSI, R., KALK, K. H., VAN

DER VEEN, B. A., DIJKHUIZEN, L., WITHERS, S.G. & DIJKSTRA, B. W. 1999. X-ray structuresalong the reaction pathway of cyclodextrin glyco-syltransferase elucidate catalysis in the α-amylasefamily. Nat. Struct. Biol. 6: 432–436.

VALLEE, F., KADZIOLA, A., BOURNE, Y., JUY, M.,RODENBURG, K. W., SVENSSON, B. & HASER,R. 1998. Barley α-amylase bound to its endoge-nous protein inhibitor BASI: crystal structure ofthe complex at 1.9 A resolution. Structure 6: 649–659.

VIHINEN, M. & MÄNTSÄLÄ, P. 1989. Microbial amy-lolytic enzymes. Crit. Rev. Biochem. Mol. Biol. 24:329–418.

WATANABE, K., HATA, Y., KIZAKI, H., KATSUBE, Y.& SUZUKI, Y. 1997. The refined crystal structureof Bacillus cereus oligo-1,6-glucosidase at 2.0 Aresolution: structural characterization of proline-substitution sites for protein thermostabilization.J. Mol. Biol. 269: 142–153.

WELCH, W. J. & BROWN, C. R. 1996. Influence ofmolecular and chemical chaperones on protein fold-ing. Cell Stress Chaperones 1: 109–115.

WELLS, R. G. & HEDIGER, M. A. 1992. Cloning of arat kidney cDNA that stimulates dibasic and neu-tral amino acid transport and has sequence simi-larity to glucosidases. Proc. Natl. Acad. Sci. USA89: 5596–5600.

WILBUR, W. J. & LIPMAN, D. J. 1983. Rapid similaritysearches of nucleic acid and protein data banks.Proc. Natl. Acad. Sci. USA 80: 726–730.

WODAK, S. J., LASTERS, I., PIO, F. & CLAESSENS, M.1990. Basic design features of the parallel α/β bar-rel, a ubiquitous protein-folding motif. Biochem.Soc. Symp. 57: 99–121.

YANG, B.-Z., DING, J.-H., ENGHILD, J. J., BAO, Y.& CHEN, Y.-T. 1992. Molecular cloning and nu-cleotide sequence of cDNA encoding human muscleglycogen debranching enzyme. J. Biol. Chem. 267:9294–9299.

Received October 29, 2001Accepted February 12, 2002

57