(Βα) 8 Barrels: Implications for the Evolution of Metabolic Pathways

doi:10.1006/jmbi.2000.4152 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 303, 627±640

Homology among (bbbaaa)8 Barrels: Implications for the Evolution of Metabolic Pathways RichardR.Copley*andPeerBork*

Biocomputing, European We provide statistically reliable sequence evidence indicating that at least

Molecular Biology Laboratory 12 of 23 SCOP (ba)8 (TIM) barrel superfamilies share a common origin. Meyerhofstrasse 1 This includes all but one of the known and predicted TIM barrels found 69117 Heidelberg, Germany in central metabolism. The statistical evidence is complemented by an examination of the details of protein structure, with certain structural locations favouring catalytic residues even though the nature of their molecular function may change. The combined analysis of sequence, structure and function also enables us to propose a phylogeny of TIM barrels. Based on these data, we are able to examine differing theories of pathway and enzyme evolution, by mapping known TIM barrel folds to the pathways of central metabolism. The results favour widespread recruitment of enzymes between pathways, rather than a ``backwards evolution'' model, and support the idea that modern proteins may have arisen from common ancestors that bound key metabolites. # 2000 Academic Press Keywords: homology; protein evolution; pathway evolution; metabolism; *Corresponding authors TIM barrels

Introduction behomologous(Jensen,1976).Relatedtothisview is that in which enzyme evolution has been driven Horowitz(1945)proposedthatbiochemicalpath- by retention of catalytic mechanisms. There is good ways may have evolved backwards (i.e. with the evidence to suggest that this has occurred within enzyme catalyzing the Nth step of a pathway aris- manyproteinfamilies(Lawrenceetal.,1997; ingbeforetheenzymecatalyzingtheN1thstep), Babbitt&Gerlt,1997;Gerlt&Babbitt,1998; latermodifyingthistheory(Horowitz,1965)to Eklund&Fontecave,1999).Ingeneral,however, suggest that enzymes within a similar pathway the links between a protein fold and the type of could be homologous. The idea has intuitive reactioncatalyzedarenotstrong(Martinetal., appeal. As the product of one enzyme is the sub- 1998;Hegyi&Gerstein,1999). strate of the enzyme following it in a pathway, it To enquire more deeply into the nature of path- follows that binding sites need not be reinvented if way evolution, we have analyzed the distribution homologous enzymes are used within a pathway. of one particular fold in the enzymes of central There are a handful of examples where adjacent metabolism. Enzymes containing triose phosphate enzymes in a pathway are indeed homologous isomerase (TIM) (ba)8 barrel-like folds adopt a par- (e.g.seeBelfaizaetal.,1986;Bork&Rohde,1990; ticularlywiderangeoffunctions(Pujadas&Palau, Wilmannsetal.,1991;Fanietal.,1994).Analterna- 1999).Approximately10%ofenzymesmaycon- tive view of pathway evolution is that of enzyme tainsuchafold(Gerlt,2000),buttheirhistory recruitment, whereby enzymes catalyzing similar remains obscure. Although common structure is reactions in different pathways would be found to often taken as an indicator of common ancestry, the simplicity of the core fold, taken with an absence of sequence similarity, and a vast range of Abbreviations used: TIM, triose phosphate isomerase; functions, raises the possibility of convergent evol- FMN, ¯avin mononucleotide; NAD, nicotinamide- ution.FarberandPetsko(1990)havearguedfor adenine dinucleotide; PLP, pyridoxal phosphate; IMP, inosine monophosphate; IMPDH, inosine common ancestry of a limited number of TIM bar- monophosphate dehydrogenase; PEP, rels (compared to present day numbers) on the phosphoenolpyruvate. basis of similarity in the position of active sites, E-mail addresses of the corresponding authors: andotherstructuralfeatures.Janacek(Janecek, [email protected],[email protected] 1995,1996;Janecek&Balaz,1995)reportedthe

0022-2836/00/040627±14 $35.00/0 # 2000 Academic Press 628 Evolution of ( )8 Barrels and Metabolism presence of a few similar residues at approximately until no new family was found. Below, we outline similar positions within various members of TIM our searches and justify our inferences of hom- barrel-like enzymes, and used this to infer homology in more detail. ology. The signi®cance of these ®ndings, and how None of the proteins found during the searching likely they were to have arisen by chance, is dif®- stage was an obvious false positive (i.e. a signi®- cult to assess. The SCOP protein structure database cant match known to adopt a non-TIM barrel-like (Murzinetal.,1995)currentlydistinguishes23 fold). Full PSI-Blast results ®les are available for superfamilies of TIM barrel, where the members of inspectionathttp://www.embl-heidelberg.de/ a superfamily have a probable common evolution- ~copley/TIM. ary origin. Most recently, Thornton and co-workers have analyzed TIM barrels from a structural view- A superfamily of phosphate-binding barrels point, classifying nine families on the basis of hydrogen bond patterns and residue packing in Many sequence families that are readily detect- theinteriorofthebarrel(Naganoetal.,1999). able from hisA share a characteristic phosphate- In this study, we present statistically robust evi- binding motif at the end of b-strand 7 (see dence to support the hypothesis that many families Figure2).Thisincludesthesuperfamiliestriose- of TIM barrel-containing enzymes have diverged phosphate isomerase, ribulose phosphate-binding from a common ancestor. The increasing coverage barrel (including tryptophan biosynthesis enzymes, of sequence databases, coupled with increasing and ribuloase phosphate epimerase), thiamin phos- sensitivity of sequence-based search methods, has phate synthase and FMN-linked oxidoreductases, enabled the discovery of sequences that provide which in SCOP are all noted to share a common evolutionary links between major families of TIM binding site, although with no explicit statement of barrel-like structures. Taken together, the simi- homology. Additionally, we detect inosine mono- larities provide convincing evidence for a monophosphate dehydrogenase (IMPDH) from the phyletic origin of many TIM barrel families, and a NAD(P)-linked oxidoreductase superfamily, plausible route by which modern functions have although this appears to be a mis-classi®cation in arisen from ancestral ones. The ubiquity of this SCOP, as the phosphate-binding site is occupied family in the biosynthetic pathways of central by IMP rather than by the NADP characteristic of metabolism and the implications for theories of other superfamily members. The recently solved pathway evolution are analyzed. structure of orotidine 50-monophosphate decarboxylase is detected by our PSI-blast searches, and Results sharesasimilarphosphate-bindingsite(Appleby etal.,2000).TheC-terminaldomainofquinolinic In order to establish the distribution of homo- acid phosphoribosyl transferase consists of an logues over known pathways, it is ®rst necessary incomplete TIM barrel domain, with only seven to determine which proteins are homologous. b-strands(Eadsetal.,1997).Itisfoundasasigni®- Rather than simply take similarity in fold (i.e. cant hit in our searches, and contains a conserved purely structural data) as a criterion, our strategy phosphate-binding motif at a position equivalent has been to look for statistically justi®ed sequence to that in the other phosphate-binding barrels. evidence. We then analyzed the structure and There are eight equivalent b-strands in the core function of these proteins to consider the (ba)8 structure. There does not appear to be an ade- implications of the sequence evidence. quate explanation of the exact topological equival- ence (i.e. at the end of the seventh b-strand) of Homology among TIM barrels of phosphate-binding sites within TIM barrels, other known structure than homology. Related phosphate-binding sites are found in the Figure1providesanoverviewofselected TIM barrels of the SCOP PLP-binding barrel super- sequence relationships detected by PSI-Blast family, and the C-terminal domain of the large (Altschuletal.,1997)searchingwithrepresentative subunitofRubisco(Figure2),althoughnosigni®- TIM barrel sequences. Sequences and SCOP super- cant global sequence link could be found to the families(LoConteetal.,2000)aredescribedin wider family. Table1.Central(althoughnotcritical)tothe scheme is the protein encoded by the hisA gene of Aldolases Methanococcus voltae. This protein shows an internal duplication (observable in other members Aldolases are an important class of enzymes, cat- ofthisfamily(Fanietal.,1994)),whichisreadily alyzing the fusion or cleavage of two carbonyl detectable by PSI-Blast. This internal duplication compounds to form an aldol, a common strategy may assist the rapid detection of distant homo- for growing carbon chains during biosynthesis. logues, by producing a ``deeper'' sequence pro®le Many (but not all) aldolases are known to adopt a duringinitialstagesofsearching(Aravind& TIM barrel-like fold. Within the (ba)8 class of aldo- Koonin,1999)andsorepresentsanidealstarting lase folds, phosphate binding at the end of point for exploring sequence space. Hits found b-strand 7 is found in both type I and type II fruc- were used to initiate further PSI-Blast searches, tosebisphosphatealdolases(Figure2). Evolution of ( )8 Barrels and Metabolism 629

Figure 1. Overview of sequence-searching strategy used to identify homologous families using PSI-Blast. Par- ametersusedarediscussedinMethodsandinthetext.GenenamescorrespondtothoseinTable1.Identi®ersof sequences used to initiate searches are also given (SWISS-PROT identi®ers, or NCBI NR GI accession numbers). Coloured boxes around the gene name and sequence identi®er correspond to different SCOP superfamilies, with black being used for solved structures not yet in SCOP. Unboxed genes do not have a close homologue of known structure. Arrows point away from the sequence used as the seed in any particular search. The iteration number (bold face) and E-value at which the sequence was ®rst found are indicated in boxes. For clarity, many links between sequencesarenotshown.Fullsearchresultsmaybeobtainedfromhttp://www.embl-heidelberg.de/~copley/TIM/.

Type II fructose bisphosphate aldolases are However, the metal ion required for catalysis is identi®ed as putative homologues of the found in the same location as that from type II phosphate-binding superfamily by our PSI-Blast fructose bisphosphate aldolase. analysis, producing signi®cant E-values within Type I fructose bisphosphate aldolases catalyze three iterations from searches initiated with hisA. reactions via a lysine residue, which forms a Schiff Type II fructose bisphosphate aldolases operate via base with the substrate. This residue is found on a metal-dependent mechanism, as does the aldo- b-strand 6. These enzymes have been shown to be lase 3-deoxy-D-arabino-heptulosonate-7-phosphate homologous to type II aldolases and deoxyribose synthase (aroG), the enzyme catalyzing the ®rst phosphatealdolases(deoC)(Galperinetal.,2000). step of chorismate biosynthesis. This enzyme also We also detect signi®cant similarity between adoptsaTIMbarrelfold(Shumilinetal.,1999), members of the trpA family and a non-phosphory- and is detected in three iterations of PSI-Blast lated Entner-Doudoroff pathway aldolase from searching(Figure1).Althoughthisenzyme Sulfolobus solfataricus (with an E-value inclusion catalyzes the association of two phosphorylated threshold of 0.01), which is in turn homologous to substrates, neither phosphate group is found in dihydrodipicolinate synthase and N-acetylneurami- the characteristic site at the end of b-strand 7. nate lyase, both Schiff base-dependent enzymes Table 1. Guide to TIM barrel containing proteins discussed in the text

GeneProteinSCOPsuperfamilyPathwayPDBReference fba(I)Fructose-1,6-bisphosphatealdolase(Schiffbase)AldolaseGlycolysis1fbaHesteretal.(1991) fba(II)Fructose-1,6-bisphosphatealdolase(Metaldept)AldolaseGlycolysis1b57Halletal.(1999) dhnAFructose-1,6-bisphosphatealdolase-GlycolysisN/ADandekaretal.(1999) tpiTriosephosphateisomeraseTIMGlycolysis1tphZhangetal.(1994) enoEnolaseEnolaseC-terminaldomainGlycolysis1oneLarsenetal.(1996) pykPyruvatekinasePhosphoenolpyruvate/pyruvateGlycolysis1a3wJuricaetal.(1998) edaKDGPaldolase-Entner-Doudoroff1kga*Mavridisetal.(1982) rpeD-Ribulose-5-P3-phosphateepimeraseRibulosephosphatebindingPentosephosphate1rpxKoppetal.(1999) talTransaldolaseAldolasePentosephosphate1onrJiaetal.(1996) aroG2-Dehydro-3-deoxyphosphoheptanoatealdolaseAldolaseChorismate1qr7Shumilinetal.(1999) aroD3-Dehydroquinase-Chorismate1qfeGourleyetal.(1999) trpC_1PhosphoribosylanthranilateisomeraseRibulosephosphate-bindingTryptophan1nsjHennigetal.(1997) trpC_2Indole3-glycerol-PsynthaseRibulosephosphate-bindingTryptophan1a53Hennigetal.(1995) trpATrpsynthaseasubunitRibulosephosphate-bindingTryptophan1ttqRheeetal.(1997) hisAP-Ribosylformimino-AICAR-P-isomerase-Histidine1qo2Langetal.(2000) hisFImidazoleglycerolphosphatesynthase-Histidine1thfLangetal.(2000) pyrCDihydroorotaseMetallodepthydrolasePyrimidineN/AHolm&Sander(1997) pyrDDihydroorotatedehydrogenaseFMN-linkedoxidoreductasePyrimidine2dorRowlandetal.(1998) pyrFOrotidine-50-Pdecarboxylase-Pyrimidine1dbtApplebyetal.(2000) guaCGMPreductase-PurineinterconversionsN/ABorketal.(1995) guaBInosine50-PdehydrogenaseNAD(P)-linkedoxidoreductasePurineinterconversions1ak5Whitbyetal.(1997) hosc* Homocitrate synthase - Lysine (a-aminoadipic acid) N/A - dapADihydrodipicolinatesynthaseAldolaseLysine(diaminopimelicacid)1dhpMirwaldtetal.(1995) ppsAPhosphoenolpyruvatesynthasePhosphoenolpyruvate/pyruvatePyruvateconversions1dikHerzbergetal.(1996) ppcPhosphoenolpyruvatecarboxylasePhosphoenolpyruvate/pyruvatePyruvateconversions1fiyKaietal.(1999) pyc Pyruvate carboxylase - Pyruvate conversions N/A - leuA Isopropylmalate synthase - Leucine N/A - deoCDeoxyribosephosphatealdolase-NucleotidesalvageN/AGalperinetal.(2000) aceBMalatesynthaseMalatesynthaseGGlyoxylatecycle1d8cHowardetal.(2000) thiEThiaminphosphatesynthaseThiaminphosphatesynthaseThiaminsynthesis2tpsChiuetal.(1999) nadCQuinolinatephosphoribosyltransferaseQuinolinicacidPR-transferaseNADsynthesis1qpoSharmaetal.(1998) lctDLactatedehydrogenaseFMN-linkedoxidoreductaseRespiration1ltdTegoni&Cambillau(1994) folPDihydropteroatesynthaseDihyropteroatesynthaseFolicacidsynthesis1ajzAcharietal.(1997) sbmMethylmalonyl-CoAmutaseCobalmin(vit.B12)dept.Methylmalonylpathway7reqManciaetal.(1999) gltBGlutamatesynthase-GlutamatesynthesisN/ABorketal.(1995) citE Citrate lyase subunit - - N/A - deoCDeoxyribosephosphatealdolase-NucleotidesalvageN/AGalperinetal.(2000) MLE*MuconatelactonizingenzymeEnolaseC-terminaldomainPhenylalaninemetbolism1mucHelinetal.(1995) Where posssible, Escherichia coli gene name abbreviations have been used. Other abbreviations are indicated by asterisks. The reference given is either to a representative structure determination, or a paper predicting TIM barrel structure. The structure ®le for eda (KDGP aldolase, 1kga) is of extremely low quality. However, structural reports clearly show it adopts a TIM barrel-like fold, and it is known to be a Schiff base-dependent aldolase. Evolution of ( )8 Barrels and Metabolism 631

Figure 2. (a) Alignment of b-strands 6 and 7 for a variety of TIM barrels, produced using the STAMPpackage(Russell&Barton, 1992),wherestructureswereavail- able, and PSI-Blast analysis. Sequences with a conserved lysine residue in b strand 6, indicated in red, are Schiff base-dependent aldolases: b strand 7 is phosphate binding, the blue region in the alignment corresponds to the marked region in the structures. It can be seen that some sequences share both motifs, indicating that both families are likely to be related. Residues in small letters cannot be aligned accurately on the basis of structure. Yellow shading indicates conserved hydrophobic positions. The start and end residues of sequence fragments are indicated. Numbers in angle brackets represent the number of residues between the represented fragments. (b) The corresponding structural alignment, showing the distinctive bulge of the conserved residues at the end of the b-strand.

(Buchananetal.,1999).Theequivalentenzyme The enolase superfamily from the phosphorylated Entner-Doudoroff pathway eda is also a Schiff base-dependent TIM barrel The enolase superfamily, including enolase, aldolase(Mavridisetal.,1982),andshowssigni®- muconate lactonizing enzyme and mandelate cant sequence similarity to other phosphate-bind- racemase share a common mechanism of catalysis and have the same two-domain architecture ingTIMbarrels(Figure1).Comparisonofknown (Babbittetal.,1996;LoConteetal.,2000).The structures shows that the Schiff base-forming cata- C-terminal domain adopts a TIM barrel-like lytic lysine residue is found in an exactly equival- fold, although in the case of enolase itself, the ent position in type I fructose-1,6-bisphosphate arrangement of secondary structure is bbaa(ba) aldolase, N-acetylneuraminate lyase, dihydrodipi- 6 rather than the classical (ba)8. Although enolase colinatesynthase(Lawrenceetal.,1997)and catalyzes the formation of phosphoenolpyruvate, 3-dehydroquinase(seeFigure2).AfurtherSchiff 2-phospho-D-glycerate, it does not bind the phos- base-dependent enzyme, transaldolase, contains a phate group at the end of b-strand 7; rather, the similar residue believed to be related to the wider phosphate group is coordinated by a loop from family by a circular permutation of the barrel the N-terminal domain. structure(Jiaetal.,1996).Again,suchexactequiv- The link between the enolase superfamily and alence of functional sites is most easily explained the wider superfamily of phosphate-binding TIM by homology. In the cases of N-acetylneuraminate barrels is made via a sequence family of unknown lyase, dihydrodipicolinate synthase and dehydro- structure and function (Pfam name UPF0034, quinase the glycine residues of the phosphate-bind- accessionPF01207(Batemanetal.,2000)).Repre- ing loop at the end of b strand 7 are not conserved, sentative sequences of this family show signi®cant indicating this particular function has been lost. similarity to both muconate lactonizing enzyme 632 Evolution of ( )8 Barrels and Metabolism and dihydroorotate dehydrogenase. The PSI-blast They align in phase to the dihydroorotate dehydro- alignments of both dihydroorotate dehydrogenase genase family. The active-site cysteine residue of and muconate lactonizing enzyme to the UPF0034 dihydroorotate dehydrogenase is present, although protein family suggests a circular permutation-like only one glycine residue in the phosphate-binding rearrangement in the lineage leading to the enolase region of dihydroorotate dehydrogenase is superfamily, with the correct alignment mapping conserved. b-strand 3 of enolase to b-strand 5 of the super- familyofphosphate-bindingbarrels(seeFigure3). The pyruvate kinase family A study based solely on structural criteria has suggestedasimilarpermutation(Sergeev&Lee, Pyruvate kinase adopts a (ba)8 structure with an 1994).Theatypicaltopologyoftheenolasebarrel, additional inserted domain at the end of b-strand beginning with a bbaa subunit is not explained by 3, and a further domain at the C terminus of the this theory; however, our result suggests that this barrel. The enzyme binds ATP and, like enolase, region of sequence is not homologous to the abab phosphoenolpyruvate and Mg2. region that begins standard (ba)8 barrels. Within the context of our exploration of The proposed alignment equivalences metal- sequence space, we initiated a PSI-Blast search binding residues of the enolase superfamily with with the sequence of ¯avocytochrome b2. (The the glycine-rich, phosphate-binding loop of the C-terminal domain of this enzyme is a member of phosphate-binding superfamily, and a catalytic the canonical phosphate-binding set of TIM barrel lysine residue of dihydroorotate dehydrogenase is structures). With an increased inclusion threshold aligned with a further metal-binding residue of the of 0.01, we detected pyruvate kinase from Strepto- enolase family. It seems likely that the backbone myces coelicolor with reliable signi®cance (iteration conformations of these sites make them favourable 3,E-value0.001,Figure1).Thephosphate-binding locations for functional residues, even if the mol- residues of ¯avocytochrome b2 are not conserved ecular functions of the residues themselves are not (although equivalent b-strand 7 regions of the pro- conserved. teins are aligned). Inspection of available structures The UPF0034 proteins are widespread in both shows that this region does not bind phosphate in bacteria and eukaryotes, and although not exper- pyruvate kinase. However, the region is function- imentally characterized, have been predicted to ally important, containing a conserved threonine adoptTIMbarrel-likefolds(Doerksetal.,1998). residuethatmaybeinvolvedincatalysis(Larsen

Figure 3. Alignment of permuted sequences of members of the pyrD family and enolase superfamily, as implied via the link to members of the UPF0034 family (see the text). Alignment of 1mucA to 1oneA, within the enolase superfamily, produced using STAMP. Within the pyrD family, the catalytic lysine residue and the glycine-rich, phosphate- binding loop are coloured in red. Within the enolase superfamily, metal-binding residues are coloured in blue. Con- served hydrophobic positions are shown in yellow, conserved charged in magenta, and conserved small in green. The ID column shows SWISS-PROT identi®ers, or gene name and PDB identi®er. Start and end residues are indicated. Numbers in angle brackets represent the number of residues between the represented fragments. Secondary structures of pyrD and eno are shown above and below the alignment, respectively. Evolution of ( )8 Barrels and Metabolism 633 etal.,1998).Moreover,theconformationofthepro- The HMGL-like family tein backbone region of b-strand 7 and the following loop is remarkably similar to that found in the The proteins isopropylmalate synthase and characteristic phosphate-binding loops (data not homocitrate synthase, the ®rst steps in leucine shown). and lysine (a-aminoadipate pathway) biosynthesis, As we had relaxed the E-value threshold for this respectively, contain a homologous region that is PSI-blast search, we further evaluated the relation- shared with pyruvate carboxylase, oxaloacetate ship of pyruvate kinase to other (ba)8 barrels by decarboxylase and 4-hydroxy-2-oxovalerate aldo- performing structural comparisons, and calculating lase(PfamPF00682,HMGL-like(Batemanetal., the signi®cance of the sequence identity of the 2000)).Membersofthisfamilyaredetectedinthe structuralalignment(Murzin,1993).Thesequence fourth iteration of PSI-Blast searching starting from identity of the structural alignment of pyruvate the hisA seed (e.g. the b subunit of pyruvate kinase and IMPDH is 19.1 %. This corresponds to a carboxylase from Methanococcus janischii, round 4 signi®cance of 8e-04, calculated using the method E-value 1e-04), indicating that a TIM barrel-like ofMurzin(Murzin,1993;Russell&Barton,1992), fold is adopted by this family. Secondary structure suggesting that the two are likely homologues. prediction(Jones,1999)ofthissequenceclearly Further PSI-Blast searches showed signi®cant predicted eight b-strands alternating with a helices relationships between pyruvate kinase and malate (Figure4).InspectionofthealignmenttohisA synthase of the glyoxylate cycle, via the sequence shows that a characteristic HxH motif of the of citrate lyase. This latter sequence is also homolo- HMGL-like family occurs at the end of the pre- gous to phosphoenolpyruvate synthase, and pyru- dictedb-strand7(Figure4).Itthusseemslikely vatecarboxylase(seeFigure1). that this motif is involved in metal binding. As

Figure 4. Multiple sequence alignment and secondary structure prediction for the HMGL-like family of proteins. Representative proteins were chosen from PFAM, and realigned, following slight modi®cation of domain boundaries. SWISS-PROTsequenceidenti®ersareused.Thepsipred(Jones,1999)secondarystructurepredictioninitiatedwith PYCB METJA is superimposed on the alignment (two predictions of helices with fewer than four residues are not shown). The colour scheme for conserved residues is: yellow background, hydrophobic; green background, conserved glycine or proline; grey background, conserved single amino acid; white backgrounds: green, small; magenta, charged; blue, polar. Conservation is de®ned as 80 % consensus. The Figure was produced using Chroma (L. Goodstadt & C.P. Ponting, unpublished). 634 Evolution of ( )8 Barrels and Metabolism some members of this family are aldolases, it is appears to be the use of a phosphate group as a tempting to speculate that they may be related to molecular anchor, by which to bind ligands to the the metal-dependent members of the aldolase protein, and bring them into proximity with superfamily, although a precise characterization diversecatalyticresidues(Figure5).However,of must await the determination of the structure of a particular interest is the observation that transaldo- representative member. lase, the a-subunit of tryptophan synthase, triose phosphate isomerase and fructose-1,6-bisphosphate Other links aldolase all catalyze reactions where D-glyceralde- hyde3-phosphateisoneoftheproducts(Figure5). In addition to the homologies reported above, Such a common product may be indicative of a we detected signi®cant similarities from our set of common D-glyceraldehyde 3-phosphate-binding homologues to two further SCOP superfamilies, ancestor for these enzymes. An analogous situation characterized by dihydropteroate synthase and the is found with phosphoenolpyruvate, which is a N-terminal domain of methylmalonyl-CoA mutase product or substrate of enolase, pyruvate kinase, (Figure1andTable1),inadditiontofurther phosphoenolpyruvate carboxylase and phospho- sequencefamilies(seehttp://www.embl-heidel- enolpyruvate synthase. berg.de/~copley/TIM)ofproteinsnotinvolvedin The presence of the phosphate-binding site in central metabolism. the putative ancestor of TIM barrels suggests that TIM barrels homologous to this family that do not contain an equivalent phosphate-binding site are derived from phosphate-binding ancestors. Discussion Hence, we conclude that 3-dehydroquinase, A phylogeny of TIM barrel-containing enzymes N-acetylneuraminate lyase and dihydrodipicolinate synthase have all arisen from an ancestor having a The sequences of the enzymes found to be hom- phosphate-binding site and a catalytic Schiff ologous in the previous section are too divergent base-forming lysine residue on b-strand 6, such as to allow the reliable reconstruction of phylogeny fructose-1,6-bisphosphate aldolase. from a sequence alignment, and techniques for The relationships of the enolase and pyruvate phylogenetic tree inference from structures (as kinase superfamilies to the main group of enzymes opposed to simply clustering similar structures) are more problematic. Global structure alignment arenotwelldeveloped(May,1999).However,by usingtheSTAMPprogram(Russell&Barton, analyzing their sequences, structures and func- 1992)indicatesthattheclosestrelativeofthepyru- tions, it is possible to discern some likely features vate kinase barrel domain is IMPDH (guaB), which of the branching order of modern TIM barrel- is in turn closest structurally to the FMN-binding containing enzymes. Here, we concentrate only on family of barrels characterized by dihydroorotate those enzymes found in central metabolism, for dehydrogenase (pyrD). Indeed, the structural align- which a structure is known. ment equivalences the catalytic lysine residue of As has been observed, the sequence of the pro- dihyrdroorotate dehydrogenase with an essential tein encoded by hisA shows a clear internal dupli- catalyticlysineofpyruvatekinase(Bollenbachetal., cation(Fanietal.,1994).Asthisduplicationisnot 1999).WealsoplaceenolaseclosetotheFMN- easily observable in other closely related homolo- binding family, as the closest sequence link is to gous TIM barrel structures, it follows that hisA has pyrD. Solution of the structure of a representative retained ancestral features that have been obscured member of the PFAM UPF0034 family may help to in other structures, suggesting that the common resolve this position further. ancestor of modern barrels was most reminiscent Both pyruvate kinase and enolase bind phos- of the protein encoded by hisA. Moreover, the phoenolpyruvate and magnesium, and neither internal sequence duplication conserves the func- shows the standard phosphate-binding motif tionalglycine-rich,phosphate-bindingsite(Thoma characteristic of other TIM barrel-containing etal.,1998),indicatingthatitwaspresentinthe enzymes. Although the mode of binding of PEP is putative ancestral half-barrel structure and that different in both cases (after allowing for the pro- phosphate-binding at the C terminus of b-strand 7 posed permutation of enolase), these differences is likely to be an ancestral, rather than derived fea- may have been caused by the later acquisition of tureinmodernTIMbarrels(Langetal.,2000).Of additional domains. We suggest that both may the extant TIM barrels, it appears that hisA is have arisen from a common PEP and Mg2-bind- closest in appearance to the putative ancestor ing ancestor. (although not necessarily sharing the same Figure6showsasynthesisofthearguments function). above, presented as a phylogeny of TIM barrel- The protein encoded by hisA catalyzes the ring containing enzymes. We have tried to reconcile the opening of a ribose sugar moiety. An analogous nature of functional residues and details of bound reaction is catalyzed by PRA isomerase (trpC 1). ligands with a clustering of similar structures. The However, analysis of the remaining phosphate- pathway to which an enzyme belongs is shown binding proteins shows very different chemical on the phylogeny, but has not in¯uenced our activities. Rather, the function common to all reconstruction. Evolution of ( )8 Barrels and Metabolism 635

Figure 5. Diverse reactions catalyzed by canonical phosphate-binding TIM barrel enzymes. Coloured symbols as forFigure6

TIM barrel-containing proteins and the modern hisF protein. Similarly, we propose that evolution of pathways trpA had a trpC 1 or hisA-like ancestor, but not vice versa. Such ancestral proteins could have had the Known TIM barrel folds within the enzymes of ability to catalyze more than one reaction, albeit centralmetabolismareshowninFigure7.Ofthese less ef®ciently, or they could have had essentially we have shown all but dihydroorotase (pyrC)are their modern functions, with intervening reactions likely to be homologous. It is clear that homol- being catalyzed by alternative methods that have ogues are distributed widely both within and since been displaced. between pathways. When multiple homologues Recently, protein engineering followed by arti®- occur within the same pathway, they are not cial selection has been used to generate trpC 1-like necessarily found to occur in adjacent enzymes. function from a trpC 2-like scaffold (i.e. in the Such a picture points to large-scale recruitment of directionsuggestedbyHorowitz)(Altamiranoetal., enzymes between different pathways, rather than 2000).However,whilealternativeevolutionary the duplication within pathways model empha- routes are untested, the result sheds little light on sized by Horowitz. evolutionary history, besides showing the possi- TIM barrel-containing enzymes occur at adjacent bility of such functional conversion. pathway steps in tryptophan biosynthesis, histi- Intriguingly, our results suggest that both class I dine biosynthesis and glycolysis. A corollary of the and class II fructose-1,6-bisphosphate aldolases Horowitz theory of pathway evolution is that, may have arisen relatively late (compared to histi- where a set of adjacent enzymes in a pathway is dine biosynthesis). This idea is given some support homologous, the enzyme of the latest step should by the fact that there is considerable plasticity in be the deepest branching (i.e. ancestral) when a the upper half of the glycolytic pathway; alterna- phylogeny is reconstructed. Our results do not pro- tive pathways are found in some species, and vide strong evidence for this. On the evidence of non-orthologous displacement of enzymes is more the previous section, we suggest that it is more common,suggestingthatitaroselaterinevolution likely that a hisA-like protein was ancestral to the (Kooninetal.,1996;Romano&Conway,1996; 636 Evolution of ( )8 Barrels and Metabolism

Figure 6. A proposed phylogeny for the TIM barrel-containing enzymes of central metabolism. See the text for detailed discussion.

Dandekaretal.,1999).Moreover,ofthefourSchiff re¯ect the state of ancestral pathways. The scale of base-dependent TIM barrel enzymes found in replacement by analogous enzymes during the central metabolism, two (3-dehydroquinase and course of evolution is not known, but so far, fructose-1,6-bisphosphate aldolase) show examples examples in central metabolism are limited of non-orthologous enzymes performing the same (Galperinetal.,1998).Eveniftheproteinsofapar- functionindifferentspecies(Galperinetal.,1998; ticular biochemical pathway have changed over Gourleyetal.,1999),andathird(dihydrodipicoli- time, the question of how contemporary pathways nate synthase) forms part of the diaminopimelic evolved can still be addressed. acid pathway of lysine biosynthesis, which is We have found no strong evidence to support replaced by the alternative a-aminoadipic acid the idea that pathways may have evolved back- pathwayinsomespecies(Michal,1999). wards and enzymes within a pathway are likely to be homologous, even in cases where this has pre- Concluding Remarks viously been suggested (e.g. tryptophan biosynthesis). Given that the (ba)8 barrel is probably the The weight of available evidence strongly mostcommonenzymefold(Hegyi&Gerstein, favours divergent evolution for the proteins dis- 1999),itseemsunlikelythatanysuchevidencewill cussed here. Our results suggest that many very emerge in the future, at least for the pathways of different functions can evolve on a common mol- central metabolism. It is possible there is increased ecular scaffold, and that it may be possible to ®nd likelihood of homologues at certain key intermedi- extant links between very different functional ates (for instance D-glyceraldehyde-3-phosphate families. Certain backbone positions seem particu- and PEP). Equally, notwithstanding previous work larly favourable for the evolution of catalytic resi- (Gerlt&Babbitt,1998),overlongenoughtime- dues (e.g. metal-binding residues in enolase scales, the catalytic mechanism of enzymes does occurring in a position equivalent to that of the not appear to be conserved. Although members of phosphate-binding loop of other enzymes). the enolase superfamily all share similarities in Looking beyond the functions of individual mechanism, we have shown that they are likely to families, we have analyzed the homologous family be homologous to a broader superfamily that do presented here to determine the implications for not, and similarly with Schiff base-dependent TIM pathway evolution. Modern pathways may not barrel enzymes. Evolution of ( )8 Barrels and Metabolism 637

Figure 7. A schematic view showing the relative positions of TIM barrel-containing enzymes in the pathways of centralmetabolism.E.coligenenames(Karpetal.,2000)havebeenusedwiththeexceptionofhoscforhomocitrate synthase(seeTable1forfulldescriptionsofgenenames).Wheretwoenzymaticstepsareencodedbythesamegene, we have labelled them gene 1 and gene 2, in the order that they occur in the pathway as drawn. Proteins of known or predicted structure are boxed. Known TIM barrel-containing proteins are coloured blue. Predicted TIM barrel-containing proteins are yellow. Non-TIM barrel-containing proteins are white. Dotted connecting lines indicate more than one step between proteins. Arrows indicate the conventional view of the direction of the pathway. Some key compounds are shown for clarity: (1) D-glucose; (2) b-D-fructose 6-phosphate; (3) D-glyceraldehyde 3-phosphate; (4) phosphoenolpyruvate; (5) pyruvate; (6) D-ribulose 5-phosphate; (7) D-ribose 5-phosphate; (8) xyulose 5-phosphate; (9) erythrose 4-phosphate. The reaction of aroD is also catalyzed by a non-TIM barrel-containing enzyme (see the text for details).Abbreviationsofnon-TIMbarrel-containingenzymes(seeTable1forthosecontainingTIMbarrels):glk,glu- cokinase; pgi, phosphoglucose isomerase; pfkA, 6-phosphofructokinase-1; gap, glyceraldehyde 3-phosphate dehydrogenase; gpm, phosphoglycerate mutase; pgk, phosphoglycerate kinase; zwf, glucose 6-phosphate-1-dehydrogenase; pgl, 6-phosphogluconolactonase; gnd, 6-phosphogluconate dehydrogenase; tktA, transketolase; prsA, phosphoribosyl- pyrophosphate synthase; aroB, 3-dehydroquinate synthase; aroE, shikimate dehydrogenase; aroK/L, shikimate kinase I/II; aroA, 3-phosphoshikimate-1-carboxyvinyltransferase; aroC, chorismate synthase; trpE, anthranilate synthase; trpD, anthranilate phosphoribosyl transferase; trpB, tryptophan synthase b-subunit; pyrI/B, aspartate carbomyltrans- ferase; pyrE, orotate phosphoribosyl transferase; hisG, ATP phosphoribosyltransferase; hisI 1, phosphoribosyl-ATP pyrophosphatase; hisI 2, phosphoribosyl AMP cyclohydrolase; hisB 1, imidazoleglycerolphosphate dehydratase; hisC, histidine-phosphate aminotransferase; hisB 2, histidinol-phosphatase; hisD 1, histidinol dehydrogenase; hisD 2 histidinal dehydrogenase; purF, amidophosphoribosyl transferase; purD, P-ribosylamine-glycine ligase; purN, phosphoribosylglycinamide formyltransferase; purL, phosphoribosylformylglycinamide synthase; purM, phosphoribosylformylglycinamide cyclo-ligase; purK, purE, phosphoribosylaminoimidazole carboxylase; purC, phosphoribosylaminoimidazole-succinocarboxamide synthase; purB, adenylosuccinate lyase; purH 1, AICAR transformylase; purH 2, IMP cyclohydrolase; guaA, GMP synthase.

Instead, our results favour general recruitment Methods of enzymes with similar functions (either a particular catalytic activity, or speci®cally binding particu- Sequence database searching lar groups). Such recruitment need not necessarily befromthesamepathway,asdiscussedbyJensen TheSCOPdatabase(LoConteetal.,2000),inconjunc- tionwithASTRAL(Brenneretal.,2000)wasusedto (1976).Suchaviewpointstoancientenzymeshav- select representative TIM barrel sequences at the 50 % ing very broad speci®cities, but also hints that it sequence identity level. Protein families for which no may be possible to resolve the order in which path- structures were available were selected if a clear relation- ways have evolved, if it is possible to construct ship existed to a protein known to adopt a TIM barrel robust phylogenies for their constituent proteins. fold. The sequences were used to initiate PSI-Blast 638 Evolution of ( )8 Barrels and Metabolism searches of the NR protein sequence database from the tal structure of the anti-bacterial sulfonamide drug NCBI(Altschuletal.,1997).Weusethetermseedto target dihydropteroate synthase. Nature Struct. Biol. refer to these intiation sequences. PSI-Blast uses an itera- 4, 490-497. tive procedure, constructing position-dependent weight- Altamirano, M. M., Blackburn, J. M., Aguayo, C. & matrices on the basis of alignments generated by a Blast Fersht, A. R. (2000). Directed evolution of new cata- database search. Sequences are used in the construction lytic activity using the a/b-barrel scaffold. Nature, oftheweightmatrixiftheygenerateanexpectation(E) 403, 617-622. value more signi®cant than a given inclusion threshold. Altschul, S. F. L. M. T., Schaffer, A. A., Zhang, J., These weight matrices are then used to score the next Zhang, Z., Miller, W. & Lipman, D. J. (1997). iteration of database searching. The NR database was Gapped BLAST and PSI-BLAST: a new generation pre®ltered for low complexity, coiled coil and transmem- of protein database search programs. Nucl. Acids brane regions using the p®lt program from the method Res. 25, 3389-3402. ofJones(1999).PSI-Blastsearcheswereruntoconver- Appleby, T. C., Kinsland, C., Begley, T. P. & Ealick, S. E. gence, with an inclusion threshold of E < 0.001, unless (2000). The crystal structure and mechanism of oro- otherwise stated. In all cases the output was inspected tidine 50-monophosphate decarboxylase. Proc. Natl for the presence of obvious false positives, including Acad. Sci. USA, 97, 2005-2010. low-complexity regions or proteins of known non-TIM Aravind, L. & Koonin, E. V. (1999). Gleaning non-trivial barrel structure. structural, functional and evolutionary information The use of a database of 50 % identity, in conjunction about proteins by iterative database searches. J. Mol. with a sensitive algorithm such as PSI-Blast leads to con- Biol. 287, 1023-1040. siderable redundancy in the search process in many Babbitt, P. C. & Gerlt, J. A. (1997). Understanding cases. (i.e. relationships may be discovered many times enzyme superfamilies. J. Biol. Chem. 272, 30591- over). However, PSI-Blast does show sensitivity to the 30594. protein sequence chosen to seed it, and as such starting Babbitt, P. C., Hasson, M. S., Wedekind, J. E., Palmer, with a range of initial sequences can be advantageous D. R. J., Barrett, W. C., Reed, G. H., Rayment, I., (Aravind&Koonin,1999).Inourresults,wepreferen- Ringe, D., Kenyon, G. L. & Gerlt, J. A. (1996). The tially report relationships detected with few iterations. enolase superfamily: a general strategy for enzyme- In some cases, sequences that were common to the catalyzed abstraction of the a-protons of carboxylic results of two PSI-Blast runs were used to link families, acids. Biochemistry, 35, 16489-16501. inasimilarmannertotheworkreportedbyKoretkeetal. Bateman, A., Birney, E., Durbin, R., Eddy, S. R., Howe, (1999). K. L. & Sonnhammer, E. L. (2000). The PFAM protein families database. Nucl. Acids Res. 28, 263-266. Structure comparison Belfaiza, J., Parsot, C., Martel, A., Bouthier de la Tour, C., Margarita, D., Cohen, G. N. & Saint-Girons, I. Global TIM barrel structure comparisons were per- (1986). Evolution of biosynthetic pathways: two formedusingthemethodbyRussell&Barton(1992). enzymes catalyzing consecutive steps in methionine Structures were obtained from the RCSB protein data biosynthesis originate from a common ancestor and base(Bermanetal.,2000),withtheexceptionofhisAand possess a similar regulatory region. Proc. Natl Acad. hisF (M. Wilmanns, personal communication). Secondary Sci. USA, 83, 867-871. structurewaspredictedusingthemethodofJones Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., (1999). Bhat, T. N., Weissig, H., Shindyalov, I. N. & Bourne, P. E. (2000). The protein data bank. Nucl. Metabolic pathways Acids Res. 28, 235-242. Bollenbach, T. J., Mesecar, A. D. & Nowak, T. (1999). We de®ne central metabolism to include glycolysis, Role of lysine 240 in the mechanism of yeast pyru- the tricarboxylic acid (TCA) cycle, the pentose phosphate vate kinase catalysis. Biochemistry, 38, 9137-9145. pathway, amino acid biosynthesis and nucleotide bio- Bork, P. & Rohde, K. (1990). Sequence similarities synthesis. Information on metabolic pathways, enzymes between tryptophan synthase b subunit and other andsubstrateswastakenfromMichal(1999). pyridoxal-phosphate-dependent enzymes. Biochem. Biophys. Res. Commun. 171, 1319-1325. Bork, P., Gellerich, J., Groth, H., Hooft, R. & Martin, F. (1995). Divergent evolution of a ba-barrel subclass: detection of numerous phosphate-binding sites by Acknowledgements motif search. Protein Sci. 4, 268-274. Brenner, S. E., Koehl, P. & Levitt, M. (2000). The We are very grateful to Matthias Wilmanns (EMBL ASTRAL compendium for protein structure and Hamburg) for providing us with the coordinates of hisA sequence analysis. Nucl. Acids Res. 28, 254-256. and hisF, and to Reinhard Sterner for supplying us with Buchanan, C. L., Connaris, H., Danson, M. J., Reeve, preprints of his work. We thank Martijn Huynen, Berend C. D. & Hough, D. W. (1999). An extremely thermo- Snel, Warren Lathe, Steffen Schmidt and Thomas Dande- stable aldolase from Sulfolobus sofataricus with speci- kar for many helpful discussions, and JoÈrg Schultz, for ®city for non-phosphorylated substrates. Biochem. J. assistance with an early analysis which led to this work. 343, 563-570. This work was funded by the BMBF ELSA project. Chiu, H. J., Reddick, J. J., Begley, T. P. & Ealick, S. E. (1999). Crystal structure of thiamin phosphate References synthase from Bacillus subtilis at 1.25 AÊ resolution. Biochemistry, 38, 6460-6470. Achari, A., Somers, D. O., Champness, J. N., Bryant, Dandekar, T., Schuster, S., Snel, B., Huynen, M. & Bork, P. K., Rosemond, J. & Stammers, D. K. (1997). Crys- P. (1999). Pathway alignment: application to the Evolution of ( )8 Barrels and Metabolism 639

comparative analysis of glycolytic enzymes. Bio- Hester, G., Brenner-Holzach, O., Rossi, F. A., Stuck- chem. J. 343, 115-124. Donatz, M., Winterhalter, K. H., Smit, J. D. & Doerks, T., Bairoch, A. & Bork, P. (1998). Protein annota- Piontek, K. (1991). The crystal structure of fructose- tion: detective work for function prediction. Trends 1,6-bisphosphate aldolase from Drosophila melanoga- Genet. 14, 248-250. ster at 2.5 AÊ resolution. FEBS Letters, 292, 237-242. Eads, J. C., Ozturk, D., Wexler, T. B., Grubmeyer, C. & Holm, L. & Sander, C. (1997). An evolutionary treasure: Sacchettini, J. C. (1997). A new function for a com- uni®cation of a broad set of amidohydrolases mon fold: the crystal structure of quinolinic acid related to urease. Proteins: Struct. Funct. Genet. 28, phosphoribosyltransferase. Structure, 15, 47-58. 72-82. Eklund, H. & Fontecave, M. (1999). Glycyl radical Horowitz, N. H. (1945). On the evolution of biochemical enzymes: a conservative structural basis for rad- syntheses. Proc. Natl Acad. Sci. USA, 31, 153-157. icals. Structure, 7, R257-R262. Horowitz, N. H. (1965). The evolution of biochemical Fani, R., Lio, P., Chiarelli, I. & Bazzicalupo, M. (1994). syntheses - retrospect and prospect. In Evolving The evolution of the histidine biosynthetic genes in Genes and Proteins (Bryson, V. & Vogel, H. J., eds), prokaryotes: a common ancestor for the hisA and pp. 15-23, Academic Press, New York. hisF genes. J. Mol. Evol. 38, 489-495. Howard, B. R., Endrizzi, J. A. & Remington, S. J. (2000). Farber, G. K. & Petsko, G. A. (1990). The evolution of Crystal structure of Escherichia coli malate synthase a/b barrel enzymes. Trends Biochem. Sci. 15, 228- G complexed with magnesium and glyoxylate at 234. 2.0 AÊ resolution. Biochemistry, 39, 3156-3168. Galperin, M. Y., Walker, D. R. & Koonin, E. V. (1998). Janecek, S. (1995). Similarity of different b-strands Analogous enzymes: independent inventions in ¯anked in loops by glycines and prolines from dis-

enzyme evolution. Genome Res. 8, 779-790. tinct (a/b)8-barrel enzymes: chance or a homology? Galperin, M. Y., Aravind, L. & Koonin, E. V. (2000). Protein Sci. 4, 1239-1242. Aldolases of the DhnA family: a possible solution Janecek, S. (1996). Invariant glycines and prolines ¯ank-

to the problem of pentose and hexose biosynthesis ing in loops the strand b2 of various (a/b)8-barrel in archaea. FEMS Microbiol Letters, 183, 259-264. enzymes: a hidden homology? Protein Sci. 5, 1136- Gerlt, J. A. (2000). New wine from old barrels. Nature 1143. Struct. Biol. 7, 171-173. Janecek, S. & Balaz, S. (1995). Functionally essential, Gerlt, J. A. & Babbitt, P. C. (1998). Mechanistically invariant glutamate near the C terminus of strand

diverse enzyme superfamilies: the importance of b5 in various (a/b)8-barrel enzymes as a possible chemistry in the evolution of catalysis. Curr. Opin. indicator of their evolutionary relatedness. Protein Chem. Biol. 2, 607-612. Eng. 8, 809-813. Gourley, D. G., Shrive, A. K., Polikarpov, I. P., Krell, T., Jensen, R. A. (1976). Enzyme recruitment in evolution of Coggins, J. R., Hawkins, A. R., Isaacs, N. W. & new function. Annu. Rev. Microbiol. 30, 409-425. Sawyer, L. (1999). The two types of 3-dehydroqui- Jia, J., Huang, W., Schorken, U., Sahm, H., Sprenger, nase have distinct structures but catalyze the same G. A., Lindqvist, Y. & Schneider, G. (1996). Crystal overall reaction. Nature Struct. Biol. 6, 521-525. structure of transaldolase B from Escherichia coli Hall, D. R., Leonard, G. A., Reed, C. D., Watt, C. I., suggests a circular permutation of the a/b barrel Berry, A. & Hunter, W. N. (1999). The crystal within the class I aldolase family. Structure, 4, 715- structure of Escherichia coli class II fructose-1,6- 724. bisphosphate aldolase in complex with phosphogly- Jones, D. T. (1999). Protein secondary structure predic- colohydroxamate reveals details of mechainism and tion based on position-speci®c scoring matrices. speci®city. J. Mol. Biol. 287, 383-394. J. Mol. Biol. 292, 195-202. Hegyi, H. & Gerstein, M. (1999). The relationship Jurica, M. S., Mesecar, A., Heath, P. J., Shi, W., Nowak, between protein structure and function: a com- T. & Stoddard, B. L. (1998). The allosteric regulation prehensive survey with application to the yeast of pyruvate kinase by fructose-1,6-bisphosphate. genome. J. Mol. Biol. 288, 147-164. Structure, 6, 195-210. Helin, S., Kahn, P. C., Guha, B. L., Mallows, D. G. & Kai, Y., Matsumura, H., Inoue, T., Terada, K., Nagara, Goldman, A. (1995). The re®ned X-ray structure of Y., Yoshinga, T., Kihara, A., Tsumura, K. & Izui, K. muconate lactonizing enzyme from Pseudomonas (1999). Three-dimensional structure of phosphoenol- putida PRS2000 at 1.85 AÊ resolution. J. Mol. Biol. pyruvate carboxylase: a proposed mechanism for 254, 918-941. allosteric inhibition. Proc. Natl Acad. Sci. USA, 96, Hennig, M., Darimont, B., Sterner, R., Kirschner, K. & 823-828. Jansonius, J. N. (1995). 2.0 AÊ structure of indole-3- Karp, P. D., Riley, M., Saier, M., Paulsen, I. T., Paley, glycerol phosphate synthase from the hyperthermo- S. M. & Pellegrini-Toole, A. (2000). The ecocyc and phile Sulfolobus solfataricus: possible determinants of metacyc databases. Nucleic Acids Res. 28, 56-59. protein stability. Structure, 3, 1295-1306. Koonin, E. V., Mushegian, A. R. & Bork, P. (1996). Non- Hennig, M., Sterner, R., Kirschner, K. & Jansonius, J. N. orthologous gene displacement. Trends Genet. 12, (1997). Crystal structure at 2.0 AÊ resolution of phos- 334-336. phoribosyl anthranilate isomerase from the Kopp, J., Kopriva, S., Suss, K. H. & Schulz, G. E. (1999). hyperthermophile Thermotoga maritima: possible Structure and mechanism of the amphibolic enzyme determinants of protein stability. Biochemistry, 36, D-ribulose-5 3-epimerase. J. Mol. Biol. 287, 761-771. 6009-6016. Koretke, K. K., Russell, R. B., Copley, R. R. & Lupas, Herzberg, O., Chen, C. C. H., Kapadia, G., McGuire, M., A. N. (1999). Fold recognition using sequence and Carroll, L., Noh, S. J. & Dunaway-Mariano, D. secondary structure information. Proteins: Struct. (1996). Swivelling-domain mechanism for enzymatic Funct. Genet. 37 (S3), 141-148. phosphotransfer between remote reaction sites. Lang, D., Thoma, R., Henn-Sax, M., Sterner, R. & Proc. Natl Acad. Sci. USA, 93, 2652-2657. Wilmanns, M. (2000). Structural evidence for evol- 640 Evolution of ( )8 Barrels and Metabolism

ution of the b/a barrel scaffold by gene duplication natural and designed molecules. Biologia, Bratislava, and fusion. Science, 289, 1546-1550. 54, 231-253. Larsen, T. M., Wedekind, J. E., Rayment, I. & Reed, Rhee, S., Parris, K. D., Hyde, C. C., Ahmed, S. A., Miles, G. H. (1996). A carboxylate oxygen of the substrate E. W. & Davies, D. R. (1997). Crystal structures of bridges the magnesium ions at the active site of mutant (bK87T) tryptophan synthase a2b2 complex enolase: structure of the yeast enzyme complexed with ligands bound to the active sites of the a and with the equilibrium mixture of 2-phosphoglycerate b-subunits reveal ligand-induced conformational and phosphoenolpyruvate at 1.8 AÊ resolution. Bio- changes. Biochemistry, 36, 7664-7680. chemistry, 35, 4349-4358. Romano, A. H. & Conway, T. (1996). Evolution of carbo- Larsen, T. M., Benning, M. M., Rayment, I. & Reed, hydrate metabolic pathways. Res. Microbiol. 147, G. H. (1998). Structure of the bis(Mg2)-ATP oxalate 448-455. complex of the rabbit muscle pyruvate kinase at Rowland, P., Bjornberg, O., Nielsen, F. S., Jensen, K. F. 2.1 AÊ resolution: ATP binding over a barrel. Bio- & Larsen, S. (1998). The crystal structure of Lactococ- chemistry, 37, 6247-6255. cus lactis dihydroorotate dehydrogenase A com- Lawrence, M. C., Barbosa, J. A. R. G., Smith, B. J., Hall, plexed with the enzyme reaction product throws N. E., Pilling, P. A., Ooi, H. C. & Marcuccio, S. M. light on its enzymatic function. Protein Sci. 7, 1269- (1997). Structure and mechanism of a sub-family of 1279. enzymes related to N-acetylneuraminate lyase. Russell, R. B. & Barton, G. J. (1992). Multiple protein J. Mol. Biol. 266, 381-399. sequence alignment from tertiary structure compari- Lo Conte, L., Ailey, B., Hubbard, T. J., Brenner, S. E., son: assignment of global and residue con®dence Murzin, A. G. & Chothia, C. (2000). SCOP: a struc- levels. Proteins: Struct. Funct. Genet. 14, 309-323. tural classi®cation of proteins database. Nucl. Acids Sergeev, Y. & Lee, B. (1994). Alignment of beta-barrels in (b/a) proteins using hydrogen-bonding pattern. Res. 28, 257-259. 8 Mancia, F., Smith, G. A. & Evans, P. R. (1999). Crystal J. Mol. Biol. 244, 168-182. structure of substrate complexes of methylmalonyl- Sharma, V., Grubmeyer, C. & Sacchettini, J. C. (1998). CoA mutase. Biochemistry, 38, 7999-8005. Crystal structure of quinolinic acid phosphoribosyltransferase from Mycobacterium tuberculosis: a poten- Martin, A. C. R., Orengo, C. A., Hutchinson, E. G., tial TB drug target. Structure, 6, 1587-1599. Jones, S., Karmirantzou, M., Laskowski, R. A., Shumilin, I. A., Kretsinger, R. H. & Bauerle, R. H. Mitchell, J. B. O., Taroni, C. & Thornton, J. M. (1999). Crystal structure of phenylalanine- (1998). Protein folds and functions. Structure, 6, 875- regulated 3-deoxy- -arabino-heptulosonate-7-phos- 884. D phate synthase from Escherichia coli. Structure, 7, 865- Mavridis, I. M., Hatada, M. H., Tulinsky, A. & Lebioda, 875. L. (1982). Structure of 2-keto-3-deoxy-6-phosphoglu- Ê Ê Tegoni, M. & Cambillau, C. (1994). The 2.6 A re®ned conate aldolase at 2.8 A resolution. J Mol Biol. 162, structure of the Escherichia coli recombinant 419-444. Saccharomyces cerevisiae ¯avocytochrome b2-sul®te May, A. C. (1999). Towards a more meaningful hierarch- complex. Protein Sci. 3, 303-313. ical classi®cation of protein three-dimensional Thoma, R., Schwander, M., Liebl, W., Kirschner, K. & structures. Proteins: Struct. Funct. Genet. 37, 20-29. Sterner, R. (1998). A histidine gene cluster of the Michal, G. (1999). Biochemical Pathways, John Wiley & hyperthermophile Thermotoga maritima: sequence Sons, Inc. analysis and evolutionary signi®cance. Extremo- Mirwaldt, C., Korndorfer, I. & Huber, R. (1995). The philes, 2, 379-389. crystal structure of dihydrodipicolinate synthase Whitby, F. G., Luecke, H., Kuhn, P., Somoza, J. R., Ê from Escherichia coli at 2.5 A resolution. J. Mol. Biol. Huete-Perez, J. A., Phillips, J. D., Hill, C. P., 246, 227-239. Fletterick, R. J. & Wang, C. C. (1997). Crystal struc- Murzin, A. G. (1993). Sweet-tasting protein monellin is ture of Tritochomonas foetus inosine-50-monophos- related to the cystatin family of thiol proteinase phate dehydrogenase and the enzyme-product inhibitors. J. Mol. Biol. 230, 689-694. complex. Biochemistry, 36, 10666-100674. Murzin, A. G., Brenner, S. E., Hubbard, T. & Chothia, C. Wilmanns, M., Hyde, C. C., Davies, D. R., Kirschner, K. (1995). SCOP: a structural classi®cation of proteins & Jansonius, J. N. (1991). Structural conservation in database for the investigation of sequences and parallel b/a-barrel enzymes that catalyze three structures. J. Mol. Biol. 247, 536-540. sequential reactions in the pathway of tryptophan Nagano, N., Hutchinson, E. G. & Thornton, J. M. (1999). biosynthesis. Biochemistry, 30, 9161-9169. Barrel structures in proteins: automatic identi®- Zhang, Z., Sugio, S., Komives, E. A., Liu, K. D., cation and classi®cation including a sequence anal- Knowles, J. R., Petsko, G. A. & Ringe, D. (1994). ysis of TIM barrels. Protein Sci. 8, 2072-2084. Crystal structure of recombinant chicken triosepho- Pujadas, G. & Palau, J. (1999). TIM barrel fold: struc- sphate isomerase-phosphoglycohydroxamate com- tural, functional and evolutionary characteristics in plex at 1.8 AÊ resolution. Biochemistry, 33, 2830-2837.

Edited by G. Von Heijne

(Received 17 May 2000; received in revised form 6 September 2000; accepted 7 September 2000)