Structure of peptidase T from Salmonella typhimurium Kjell Ha Ê kansson* and Charles G. Miller Department of Microbiology, University of Illinois at Urbana-Champaign, Urbana, IL, USA The structure of peptidase T, or tripeptidase, was deter- mined by multiple wavelength anomalous dispersion (MAD) methodology a nd re ®ned to 2 .4 A Ê resolution. Pep- tidase T comprises two domains; a catalytic domain with an active site containing two metal ions, and a smaller domain formed through a long insertion into the catalytic domain. The two met al i ons, p resumably zinc, ar e s eparated by 3.3 A Ê , and are coordinated by ®ve carboxylate and histidine ligands. The molecular surface of the active site is negatively charged. Peptidase T has the same basic fold as carboxy- peptidase G2. When the structures of the two enzymes are superimposed, a number of homologous residues, not evi- dent from the sequence a lone, could be identi®ed. Com- parison of t he active sites of peptidase T, carboxypeptidase G2, Aeromonas p roteolytica aminopeptidase, carboxypepti- dase A and leucine aminopeptidase reveals a common structural framework with i nteresting similarities and dierences in the active sites an d in the zinc coordination. A putative b inding site for the C-terminal end of the tripeptide substrate was found at a peptidase T speci®c ®ngerprint s equence m otif. Keywords: tripeptidase; aminotripeptidase; metallopep- tidase; X-ray crys tallography; M AD. Escherichia coli and Salmonella typhimurium express s everal intracellular enzymes capable o f hydrolyzing peptides [1]. Many of these e nzymes h ave been shown to function in the degradation of intracellular proteins and in the catabolism of exogenously supplied p eptides [2]. O ne of these enzymes, peptidase T, or tripeptidase (EC 3.4.11 ) is a 409 amino- acid metalloenzyme that hydrolyzes tripeptides a t their N-termini [3,4]. The enzymatic activity of t he S. typhimurium enzyme is both speci®c and unusual; dipeptides, tetrapep- tides or t ripeptides with blocked N-termini are not cleaved. S. typhimurium peptidase T expression is re gulated b y FNR, a transcriptional a ctivator that re sponds to anaero- biosis [4,5]. The aerobic expression level of peptidase T is not suf®cient to allow this enzyme to contribute to the utilization of exogenously supplied peptides a s amino acid sources [3]. Under a naerobic c onditions, however, the pepT gene is induced, leading to levels of peptidase T that allow i t to participate in the catabolism of tripeptides [5,6]. It has been speculated t hat this pattern of regulation may contribute to the anaerobic utilization of amino acids as energy sources [1]. A 45-amino-acid region of peptidase T displays similarity to a short region in Pseudomonas sp. strain RS-16 carboxypeptidase G2 (CG2), peptidase D, and alkaline phosphatase isozyme conversion peptidase (Iap) [4]. This region of similarity contains two of the ®ve ligands that coordinate the two zinc ions in the active site of CG2, for which the three-dimensional structure is known [7]. Pepti- dase T h as therefore b een classi®ed into the M 20 family of proteases/peptidases [8]. A third zinc ligand, a histidine, can be recognized as part of a HXDT motif [9], which is conserved i n both peptidase T and CG2. While these data indicate that peptidase T is evolutionarily related to CG2, the lack of clear homology outside these regions, and the unique tripeptidase speci®city of peptidase T, suggested that the structure of the two enzymes would in part differ. We report the three-dimensional structure of S. typhimurium peptidase T solved by multiple wavelength anomalous dispersion (MAD) methodology a nd re®ned to 2.4-A Ê resolution. MATERIALS AND METHODS Crystallization and data collection Selenomethionine His-tagged peptidase T was expressed in strain TN5619, puri®ed, c rystallized from ammonium sulfate solu tions at pH 7.5 a nd ¯ash frozen in 50% sucrose as previously described [ 10]. Crystals belong to s pace group C2 with a  132.4 A Ê , b  46.0 A Ê , c  96.6 A Ê , b  116.1 A Ê . Data were collected at 100 K at NSLS beam station X4A (Brookhaven, NY, USA) at four different wavelengths, and processed with DENZO , SCALEPACK and the CCP 4 program suite [11,12]. Structure solution and re®nement The structure was solved b y MAD methodology using 15 selenium atoms, four wavelengths and data to 2.8 A Ê . Determination of selenium positions, phase and electron density calculations and model re®nement were performed with the CNS program package [13]. The model was built manually and displayed using the graphics pro gram O [14]. The model was re®ned against one of the data sets processed Correspondence to C. G. Miller, Department of Microbiology, Uni- versity of Illin o is at Urba na-Champaign, B103 CLSL, 601 S. Good- win Avenue, Urbana, Illinois 61801, USA. Fax: + 217 244 6697, Tel.: + 217 244 8418, E-mail: charlesm@life.uiuc.edu Abbreviations: MAD, m u ltiple wavelength anomalous dispersion; CG2, carboxypeptidase G2; APP, Aeromonas prot e olyti ca amino- peptidase; LAP, leucine aminopeptidase; CPA, carboxypeptidase A. *Present address: L ab oratory of C ellular and Molecular Physiology, August Kr ogh Institute, U niversitetsparken 13, DK 210 0, Kbh é , Denmark. (Received 3 September 200 1, revised 6 November 2 001, accepted 8 November 200 1) Eur. J. Biochem. 269, 443±450 (2002) Ó FEBS 2002 nonanomalously to 2.4 A Ê . Data collection and re®nement statistics are s hown i n Tables 1 and 2. T he electron density of the three N-terminal residues w as dif®cult to interpret. A putative s ulfate ion, hydrogen bonded to the main-chain amino groups of Lys3 and Leu5, was included in order to account for all of the electron d ensity in this part of the structure. R esidues His305±Pro306 a re not well de®ned due to very weak electron density signals and the C-terminal residue 409 along with the C-terminal hexahistidine tag showed no signi®cant density at a ll. Outside t hese parts, the polypeptide chain is g enerally well de®ned by the 2|Fo| ) |Fc| density maps. The side-chains of residues Arg99, Asp109, Val115 and Tyr378, however, were not de®ned beyond the Cb atom. Most of the solvent density was relatively weak and the modeled solvent molecules have high temperature factors. Coordinates and diffraction data have been deposited in the Protein D ata Bank and have the accession code 1FNO [15]. RESULTS Overall structure The ®nal model consists of residues M et1±Gly408 (see Materials and methods). All 15 methionines are built in to the s elenium atom positions determined by CNS [13]. The sulfur atoms of Cys309 and Cys343 are within 2.0 and 2.3 A Ê , respectively, of the previously determined reactive mercury sites [10]. The overall structure and the active site of peptidase T are shown in Fig. 1. The fold of the enzyme reveals a two-domain structure that is similar to that o f CG2, which we did not anticipate due to the low overall sequence similarity between the two e nzymes. T he catalytic domain contains o ne seven-stranded mixed b sheet ¯anked with ahelices, and a second, four-stranded antiparallel b sheet.InCG2,thelargerofthetwob sheetsiseight- stranded, but peptidase T residues Gly348±Glu350 (that would have formed the eighth strand) were not recognized as a b strand by the program PROCHECK [16]. The ma jor difference between the two structures is that the 20 N-terminal residues of the mature CG2 are lacking in peptidase T and that peptidase T has a 30 amino-acid insertion (residues Asn97±Gln126), which contains two additional b strands. The structure of the residues ¯anking this in sertion also differ between the two enzymes. Another difference is that the loop between the ®rst two helices (Lys19±Ser27) is larger in peptidase T. Interestingly, this loop contains a conserved proline (Pro26) with a ci s peptide bond. When the t wo enzymes are superimposed, the insertion o verlaps i n s pace with an insertion i n C G2, f ound at the position of peptidase T residue Lys55. Most of the Ser182±Ala193 a helixinCG2isbrokenupintoanirregular structure in peptid ase T. The similarities between pepti- dase T and CG2 extend beyond the catalytic domain to include the second domain [7], which in the dimeric CG2 mediates the intermolecular contacts. This domai n, which is comprised of residues Ala211±Tyr320, consists of a four- stranded antiparallel b sheet a nd two a helices. Peptidase T has an insertion after the ®rst a helix (at Pro246), while CG2 has an insertion at the turn between the second and third b strand (at p eptidase T residue Thr269). Figure 2A shows an alignment of S. typhimurium peptidase T and Pseudo- monas sp. carboxypeptidase G2 sequences, based on the three-dimensional structures. When the Ca positions of the 67 identical amino acids are superimposed, the rmsd is 2.4 A Ê (Fig. 3A). Dimerization contacts The peptidase T dimerization domains in two crystallo- graphic asymmetric units make the same contacts around the twofold axis a s i n CG2 (Fig. 1A). These consist m ainly Table 1. Crystallographic data. Data collection a nd phasing statistics. Data set k1 k2 (edge) k3 (peak) k4 k1 Anomalous statistics Yes Yes Yes Yes No Wavelength (A Ê ) 0.9879 0.9793 0.9788 0.9668 0.9879 Resolution range (A Ê ) 20.0±2.8 20.0±2.8 20.0±2.8 20.0±2.8 20.0±2.4 Completeness (®nal shell) a 97.7 (98.3) 98.7 (99.3) 98.7(99.2) 98.5 (98.8) 99.2 (99.2) Total no. of observations 62 702 78 281 78 053 72 811 99 755 Unique no. of re¯ections a 24 622 24 877 24 881 24 825 20 598 R sym (®nal shell) (%) 3.2(8.5) 4.2(11.1) 4.6(17.9) 3.8(8.2) 4.6 (24.6) I/r(I) (®nal shell) 28.8 (11.2) 23.8 (11.3) 20.3 (7.1) 26.5 (13.2) 23.4 (6.0) Phasing power, disp./anom. b 3.5/3.5 4.7/5.1 4.1/4.4 /1.9 R Cullis , disp./anom. b 0.50/0.50 0.40/0.37 0.45/0.42 /0.67 f¢obs/f¢¢obs )5.1/1.6 )7.5/5.4 )8.3/6.4 )3.8/1.7 a A Friedel pair is considered as two unique re¯ections for the anomalously processed data. b 24 409 structure factors were phased with a ®gure of merit of 0.79. Table 2 . Re®nement stat istics. Completeness of model R (%) R free (%) Rmsd B ave main/side/protein/solvent (A Ê ) Protein atoms Solvent atoms Bonds (A Ê ) Angles (°) 3119 142 22.4 26.2 0.011 1.7 39/48/43/50 444 K. Ha Ê kansson and C. G. Miller (Eur. J. Biochem. 269) Ó FEBS 2002 of an antiparallel b strand alignment involving residues Ala264±Val270 (Ala268±Ala270 in CG2), and an antipar- allel a helical coiled-coil together with the segments sur- rounding this helix (Lys228±Thr253). Thus, the two dimerization domains together make up a continuous, eight-stranded antiparallel b sheet in both peptidase T and CG2. Active site There is one strong solvent electron d ensity signal in the active site, as previously noted [10], and a second atomic site was revealed in the |Fo| ) |Fc| map. Due t o the homology of the active site residues and the similar active site structures of peptidase T , C G2 and Aeromonas p roteolytica aminopeptidase (APP) [7,17], they have both been inter- preted as zinc ions, and the positions of the two ions and the geometry of the amino-acids c oordinating them are s imilar in the three enz ymes. The weak d ensity of the second zinc ion is probably due to low occupancy, as no zinc salt was included in the crystallization solution. With full occupancy, the r e®ned temperature factors were 37 and 112 A Ê 2 , respectively (the average solvent molecule temperature factor was 50 A Ê 2 ). Despite the high temperature factor, the chemical environment and similarity with the CG2 active site suggest that they are both zinc ions. The well- de®ned, more strongly bound zinc (Zn501), is coordinated by His78, Asp140 and Glu196, while the second zinc ion (Zn502) is coordinated by Asp140, Glu174 and His379 (Fig. 1B). The av erage temperature factor for the side-chain ligands of the well-de®ned zinc ion is 26 A Ê 2 , while the values for Glu174 and His379 are h igher ( 60 A Ê 2 ). The distance Fig. 1. Ribbon re presentation and zinc l igands of peptidase T. (A) Ribbon representation [34] of peptidase T, v iewed along the crystallo- graphic two-fold symmetry axis. b Strands are showninblueiftheybelongtothecentral b sheetofthecatalyticdomain,greenifthey belong to the four-stranded bsheetofthe dimerization domain, and ligh t blue if they belong the smaller bsheets of the catalytic domain. aHelices but not 3 10 helices are shown in r ed. A second, symmetry related molecule is s hown i n gray. (B) T he zinc ligands of p eptidase T . The i nteraction between hi sti- dine and aspartate in the H XD T motif an d the cis peptide bond b etween Asp140 and Asp141 are shown. Ó FEBS 2002 Structure of peptidase T (Eur. J. Biochem. 269) 445 between the two zinc ions is 3.3 A Ê . No zinc bound water was found; the c losest zinc±water contact is 3 .1 A Ê (Zn502). The absence of a zinc bound water could be due to the l ow occupancy of the second zinc site or to the limited resolution of the data. Superimposition o f the two m etal ions and t he Ca atoms o f the ®ve ligand amino-acids of p eptidase T a nd the other two enzymes results, in either case, i n a n rmsd of 0.5 A Ê . Zinc ligand motifs The ®rst of the metal coordinating amino acids, His78, is found in an HXDT motif, where X is a h ydrophobic amino acid. In p eptidase T, residue X is a valine ( Val79), which is buried in a hydrophobic cluster. T he polypeptide makes a sharp turn at t his position, terminating a b strand (Fig. 1B). As a r esult of this turn, both the main chain carbonyl and the side chain carboxylate group of the Asp80 residue are within hydrogen bond distance to the Nd1atomofHis78 (3.0 and 3.3 A Ê , respectively). The second of the metal ion coordinating amino acids is Asp140, which interacts with both metal ions. This residue is followed by another aspartic acid residue through a cis peptide bond in peptidase T, CG2 and AP P. This cis peptide bond breaks a helix at the N-terminal end and positions the two aspartic acids closer in space than they would otherwise have been. The third metal ion ligand, Glu174, is p receded by G lu173, which has been suggested to act as a base in the catalytic mechanism of APP [18]. In both peptidase T and CG2, but not in APP, these two glutamic a cids are p receded by an aspartic acid residue, although its side chain conformation differs in the two cases. The part of the polypeptide running up to the fourth ligand, Asp196, adopts similar c onformations in the three enzymes. In CG2, the following residue, a proline, makes van der Fig. 2. Sequence alig nments and visualized electrostatic surface potential of the active s ite of p eptidase T . (A) Sequence alignment of Pseudomonas sp. strain R S-16 carboxypepti- dase G2 (CG2) a nd Salmonella typhi murium peptidase T based o n piecewise superimposi- tion of local s tructure elements. T he sequence of pe ptidase T is given i n lines of 50 a mino- acids. CG 2 residues t hat do n ot have ho mo- logues in peptidase T are written above the alignment. Peptidas e T secondary structure elements a re indicated with the same colour scheme as in Fi g. 1A. The zinc ligands are in yellow boxes, and identical residues a re marked with an asterisk (*). (B) The electro- static surface potential of the active site o f peptidase T visualized by the p rogram GRASP [35]. Re gions of positive potential are shown in blue and negative pote ntial in red. A p-iodo- D -phenylalanine hydroxamate m olecule, superimposed fr om the e xperimentally determined APP complex [18], a nd a sulfate ion are shown in g reen. The insert, viewed from the top, shows the conserved RGGTDG ®ngerprint motif i n black beneath a semitransparent s urface. 446 K. Ha Ê kansson and C. G. Miller (Eur. J. Biochem. 269) Ó FEBS 2002 Waals interactions with Ala140±Asp141 (homologous to Ala139±Asp140 in peptidase T ). The situation is similar in APP but different in S. typhimurium peptidase T, where the following three residues a re glycines. The ®rst two of these glycines are conserved in most of the available p eptidase T sequences. The ®fth ligand, His379, is m ore dif®cult to identify from the sequence alone, but is preceded by a hydrophobic residue (Tyr, Phe, Ile or Met) in all three enzymes and has a glutamate four or ®ve residues on the C-terminal side. This glutamate hydrogen bonds to the main chain amino groups of Leu137, Gly138, Ala139, and Asp140 in the loop preceding the important active site residues Asp140±141 which are connected by the cis peptide bond. This interaction probably c ontributes to the s tability of the active site framework. This HXXX(X)E motif is conserved not only in peptidase T and CG2 but also in APP and in the other more distantly related proteins, as discussed below. Fig. 3. Stereo representation of CG2 and LAP superimposed onto peptidase T. (A) Stereo representation of CG2 (thin blue) superimposed on peptidase T (thick black) using the identical ho molo gous residues indicate d i n Fig. 2A. (B) Stereo view of the superimposed active sites of LAP (thin red) and peptidase T (thic k black) s howin g the zinc ions and zinc ligand s of both p roteins as well as amastatin bound to LAP a s d iscu ssed in the text . Ó FEBS 2002 Structure of peptidase T (Eur. J. Biochem. 269) 447 Substrate binding Structural investigation of enzyme±ligand complexes is usually one of the most useful m ethods for obtaining an understanding of the active sites of enzymes. However, no peptide a nalogue inhibitor is yet known for peptidase T. Instead, a putative substrate bindin g site has been m apped based on our knowledge of the structures of leucine aminopeptidase (LAP) and APP [18,19]. It has been pointed out that, e ven though the active sites of LAP and APP appear to be d issimilar, the two zinc ions and the zinc- bound water o f L AP can be superimposed on the zinc ions and the zinc-bound water of APP [20]. We superimposed the active site of APP on LAP in this way but because the peptidase T structure h as no zinc-bound water, it was superimposed on APP using the zinc ions and the C a atoms of the amino-acid zinc ligands. Comparison w ith t he LAP± amastatin complex [19] and the APP±phenylalanine hydroxamate complex [18] indicate the location of the S1 pocket in p eptidase T. The active s ite of peptidase T and its electrostatic surface potential are shown in Fig. 2 B. The negatively charged S1 pocket is, in addition to the zinc ligands and the sequence motifs already mentioned, lined both with residues that are conserved in the peptidase T family, such as Asn370 and Thr356, and with nonconserved residues, e .g. G lu204, Gly197 and Ile352. To predict t he S1 ¢ and S2¢ subsite locations is more dif®cult, but an extended tripeptide substrate molecule would run over the cleft that separates the two domains. There are several conserved residues in this region, e.g. Arg353, Gly354 and Gly355, although the side chain of the arginine, in its present conformation, is too far a way to be able t o interact with a bound tripeptide. A solvent molecule that was interpreted as a sulfate ion due to its relatively strong density and bulk size, is hydrogen bonded to the conserved r esidues Arg280, Tyr319, Gly355 as well as to His223 from a symmetry- related molecule. The active site of LAP with a bound amastatin molecule is shown superimposed o n the pepti- dase T a ctive site in Fig. 3B. DISCUSSION Polypeptide fold and zinc ligands The overall str uctural fold of peptidase T reveals homology with the c atalytic domains of CG2, APP, LAP a nd Carboxypeptidase A (CPA) [7,17,21±23]. The hexameric LAP contains a second, N-terminal, domain involved in subunit contacts. CG2, on the other hand, is dimeric and has a 110 amino-acid insertion that forms a second domain that is responsible for the dimeric interchain contacts. This second d omain is also present i n p e ptidase T, which w e d id not anticipate from the sequence. APP and CPA, are mono- meric enzymes consisting of a s ingle domain. Although CPA has a s ingle c atalytic zinc ion, the other four enzymes have two zinc ions in their active sites. The active sites of APP, CG2 and peptidase T have homologous zinc ion ligands and are more closely related to each other than to LAP and CPA. The similarity between the structure of t he second domain in CG2 and peptidase T extends to the interchain contacts around the twofold crystallographic axis of the two structures . This suggests that peptidase T, as CG2 i s a dimer. Peptidase T from various sources has been subjected to gel ®ltration chromatography in order to determine if it forms oligomers. Lactobacillus he lveticus peptidase T was reported to be a trimer [24], Lactococcus lactis and Pediococcus pentosaceus peptidase T behave as dimers [25,26], Bacillus subtilis and S. typhimurium pepti- dase T h ave been reported to e lute as monomers [4,27], while E. coli peptidase T had an apparent molecular mass of 80 000 Da [28]. The oligomeric state of S. typhimurium peptidase T has been reinvestigated and the results indicate that it is a dimer (D. Broder & G. Miller, unpublished observations). The amino-acid sequences of LAP and CPA are not similar and the three-dimensional structures of these proteins are too different f rom that o f p eptidase T to a llow a meaningful superimposition. It is interesting, however, to compare the topological positions of the active site residues. By topological or homologous position we mean the position in the sequence with reference to the strands (i.e. before, on, or after a certain strand) that make up the central mixed b sheet found in the catalytic domain of all of these enzymes. The appearan ce of metal ligands in topologically similar positions in b sheets of the same c onnectivity clearly indicates a divergent evolutionary relationship. It has been reported that the zinc ligands of LAP and APP a re found in structurally nonequivalent position s [17]. W e ®nd, however, that two of the amino acids that coordinate the two zinc ions (His78 and Asp140) in peptidase T are indeed found in topologically similar positions in both LAP and CPA. Moreover, Glu196 in peptidase T a nd His196 in CPA are also in homologous positions. Hence, all three amino acids that coordinate the strongly bound zinc ion i n p eptidase T, CG2 and APP are homologous to the three amino acids that coordinate the single zinc i on in CPA. A similar conclusion based on extensive sequence comparisons between and within the CPA and CG2 families was recently reported [29]. In addition, one of the ligands for t he more weakly bound zinc ion in peptidase T, Glu174, is in a position that is topologically similar to t he zinc ligand Glu334 i n LAP. T he zinc ligands in the different enzymes are aligned in Fig. 4. The e volutionary relationship b etween peptidase T, CG2 and APP is also indicated by the presence of the c onserved HXDT motif [9]. T he conformation of thes e residues results in a forked h ydrogen bond between the histidine Nd1atom and the side chain carboxylate and main chain carbonyl group of the aspartate. T his probably e nsures that N d1and not N e2 i s protonated, enhancing the electronegative character of t he latter, which coordinates to the strongly bound metal atom in t he active site. Interestingly, the corresponding residue in CPA, His69, coordinates to the zinc ion via its N d1 atom, while the Ne2 atom is hydrogen bonded to a n aspartic acid residue, albeit from a different part of the structure. The role of the t hreonine residue in the HXDT motif is less obvious, but the environment of this Fig. 4. Alignment o f homologous z inc ion ligands in p eptidase T, CG2, APP, LAP and CPA. 448 K. Ha Ê kansson and C. G. Miller (Eur. J. Biochem. 269) Ó FEBS 2002 hydrophilic side chain seems t o b e a ccessible t o the solvent. In some presumably related e nzymes, e .g. p eptidase D [30], this residue is hydrophobic. The zinc ligand Asp140 in peptidase T and t he homologous residues in C G2 and A PP are linked to the subsequent amino-acid through a cis peptide bond. In terestingly, there is a lso a cis peptide bond near the active site in CPA, but the positions and orientations of these peptide bonds vis-a Á -vis thezincions as well as their locations in the amino-acid sequence are very different in the two cases. There is for example no cis peptide bond after G lu72 in CPA, which would correspond to Asp140 in peptidase T. The role of these cis peptide bonds remains obscure. Substrate binding site and speci®city The different domain organization in LAP, CPA, APP, CG2 and peptidase T may re¯ect different ways to discriminate against longer polypeptides. In A PP and CPA, the N-terminal ( APP) or the C-t erminal (CPA) end of the s ubstrate binds into a pocket and the absence of additional steric hindrance enables the enzyme to cleave polypeptides of varying size [31,32]. In peptidase T and CG2, however, t he presence o f the dimerization domain may restrict t he size of the substrate on t he C-terminal si de of the scissile bond. Interestingly, CG2, which releases C-terminal glutamic acid residues from peptides and from folic acid and folic acid analogues such as m ethotre xate, has slightly less space available and is more positively charged in this part of the cleft. In LAP ®nally, t he size of the substrate is restricted by controlled access to the active sites caused by the assembly into hexamers. It is not obvious why pepti- dase T in c ontrast to CG2 r equires a free N-ter minal amino group in its substrates. It seems possible that the negative charge around the S1 subsite may provide a favorable interaction with a free N-terminal amino group. This negative charge may also prevent dipeptides from entering the active site, as there would be electrostatic repulsion between the C-terminal carboxylate group and this part of the enzyme. APP, which also has a negatively charged active site, does not cleave peptides with a negatively charged P1 side chain, and displays lower activity towards dipeptides and peptides with a negatively charged group in P1¢ position [31,32]. This suggests that there is a penalty for having negative charge on the substrate too close t o the N-terminus. I n CG2, on the other hand, the presence of a positively charged region closer to the active site, in part caused by Arg324, creates a binding site for C-terminal glutamic acid residues in the S1 ¢ binding site. I t should be noted in this context, that the activity p ro®le of p eptidase T does vary among different species [24±26,28]. While APP, CG2 and peptidase T have the same polypeptide fold and similar active sites, a ligand complex structure has been reported only for APP. The position of this ligand and comparison of the zinc ions and the binding of am astatin t o L AP suggest the location o f t he S1 subsite. As peptidase T cleaves a variety o f tripeptides, albeit at different rates, the interactions between substrate side chains and enzyme are probably not very speci®c. However, the position of a putative sulfate ion in the peptidase T structure suggests a possible binding site for the substrate C -terminal c arboxylate g roup, within 10 A Ê of the zinc ions. The sulfate ion is hydrogen bonded to four amino-acid residues. Two of these, Arg280 and Gly355, are conserved not only within the peptidase T sequences, but in CG2 as well. In peptidase T , Gly355 is found in a highly conserved X 1 X 2 RGGTGD motif, where X 1 and X 2 in most cases are P and I, respectively. This characteristic motif distinguishes peptidase T from the other peptidases of the MH clan, and may serve as a ®ngerprint motif. The third of the sulfate binding amino acids, Tyr319, is homologous to Arg324 in CG2 (Fig. 2A). This residue has been s uggested to interact with the C-terminal glutamate residue of CG2 substrates [7]. The fourth amino acid, His223, from a symmetry-related molecule, is also conserved in CG2. However, in CG2 this residue is more than 13 A Ê away from the zinc ions and hence too remote to interact with the substrate. A catalytic mechanism has been suggested for LAP [20], in which a bicarbonate bound to Arg336 acts in the same way as Glu151 in APP [18] by promoting deprotonization of the zinc bound water. The homologous and conserved residue in peptidase T , G lu173, is indeed found in the same relative structural position, as is Glu175 in CG2, strength- ening the arguments of Stra È ter et al .[20]. The structure of S. typhimurium peptidase T not only provides a framework for understanding the unusual speci®city of this enzyme but also yields potentially useful information concerning the very l arge family of proteins related to peptidase T. The presence of three sequence motifs that might de®ne this family has been pointed out [33]. The proteins identi®ed by these motifs include peptidases (peptidase T, CG2, yeast carboxypeptidase S) and other enzymes p otentially involved in protein break- down (mammalian aminoacylases), enzymes involved in the h ydrolysis of a cylamino-acid intermediates in amino- acid biosynthesis (DapE and A rgE), and many ORFs of as yet unknown function. Representatives of the family are found in all domains of life. The p reviously proposed motifs involve the regions around the ®rst three zinc ligands. We suggest that a fourth conserved motif HXXX(X)E corresponding to the ®fth zinc ligand can be added a s a further identifying feature of t his l arge family of proteins. ACKNOWLEDGEMENTS We thank T. Knox f o r continu ous assistance, C. Ogata (NSLS X4A) for X-ray assistance, A. Wang for access to computer facilities and D. Broder for constructing strain TN5619 and for stimulating discussions. This w ork w as supported by a grant (AI10333) from the National Institute for Allergy and Infectious Diseases to C. G. M. REFERENCES 1. Miller, C.G. (1996) Protein degradation and proteolytic modi®- cation. In Escherichia coli and Salmonella typhimurium: Cellular and Molecular B iology (Neidhardt, F.C., Curtis, R. III, Ingraham, J.L., Lin, E.C.C., Low, K.B., Magasanik, B., Rezniko, W.S., Riley, M., Schaechter, M. & U mbarger, H.E., eds), pp. 938±954. American Society f or Microbiology, Washington, DC. 2. Yen, C., Green, L. & Miller, C.G. (1980) Degradation of intra- cellular protein in Salmonella typhimurium peptidase m utants. J. Mol. Biol. 143, 21±33. 3. Strauch, K.L. & Miller, C.G. (1983) Isolation and characteriza- tion Salmonella typhimurium mutants lacking a tripeptidase (peptidase T). J. Bacteriol. 154, 763±771. Ó FEBS 2002 Structure of peptidase T (Eur. J. Biochem. 269) 449 4. Miller, C.G., Miller, J.L. & Bagga, D.A. (1991) Cloning and nucleotide sequence of the anaerobically regulated pepT gene of Salmonella typhimur ium. J. Ba ct eriol. 173, 3554±3558. 5. Lombardo, M.J., Lee, A.A., Knox, T.M. & Miller, C.G. (1997) Regulation of the Salmonella typhimurium pepT gene by cyclic AMP receptor protein (CRP) and FNR acting at a hybrid CRP- FNR site. J. Bacteriol. 179 , 1909±1917. 6. Strauch, K.L., Lenk, J.B., Gamble, B.L. & Miller, C.G. (1985) Oxygen regulation in Salmonella typhimurium. J. Bacteriol. 16 1, 673±680. 7. Rowsell, S., Pauptit, R.A., Tucker, A.D., Melton, R.G., Blow, D.M. & B rick, P. ( 1997) Crys tal structure of carboxypeptidase G2, a bacterial enzyme with applications in cancer therapy. Structure 5, 337±347. 8. Barret,A.J.,Rawlings,N.D.&Woessner,J.F.(1998)Introduc- tion: clan MH containing varied c o-catalytic m etallopeptidases. In Handbook of Proteolytic Enzymes (Barret, A.J., Rawlings, N.D. & Woessner, J.F., eds), pp. 1412±1416. Academic Press, London. 9. Neuwald, A.F., Liu, J .S., Lipman, D.J. & Lawre nce, C.E. (1997) Extracting protein alignment models from the sequence database. Nucleic A cids Res. 25, 1665±1677. 10. Ha Ê kansson, K. Broder, D. W ang, A .H J., Miller & C.G. (2000) Crystallization of peptidase T from Salmonella typhimurium. Acta Crystallogr. D56, 924±926. 11. Collaborative Computational Project Number 4 (1994) The CCP4 suite: programs for protein crystallography. Acta Crystallog. D50, 760±763. 12. Otwinowski, Z. & Minor, W. (1997) P rocessing o f X-ray dir ac- tion data collection in oscillation mode. Methods Enzymol. 276, 307±326. 13. Bru È nger,A.T.,Adams,P.D.,Clore,G.M.,DeLano,W.L.,Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998) Crystallography & NMR system: a new software suite for m acromolecular structure determination. Acta Crystallogr. D54, 905±921. 14. Jones, T.A., Zou, J Y., Cowan, S.W. & Kjeldgaard, M. (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A47, 110±119. 15. Bernstein, F.C., Koetzle, T.F., Williams, G.J., Meyer Jr, E.E., Brice, M.D., Rodgers, J.R., Kennard, O., Shimanouchi, T. & Tasumi, M. (1977) The Protein Data Bank: a computer-based archival ®le for macromolecular structures. J. Mol. Biol. 112,535± 542. 16. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. (1993) PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr 26 , 283±291. 17. Chevrier, B., Schalk, C., D'Orchymont, H., Rondeau, J.M., Moras, D. & Tarnus, C. (1994) Crystal structure of Aeromonas proteolytica aminopeptidase: a prototypical member of the co-catalytic zinc enzyme family. Structure 2, 283 ±291. 18. Chevrier, B., D'Orchymont, H., Schalk, C., T arnus, C. & Moras, D. (1996) The structure of the Aeromonas proteolytica amino- peptidase c omplexed with a hyd roxamate inhibitor. Involvem ent in catalysis o f G lu151 a nd two zinc ions of the co-catalytic unit. Eur. J. Biochem. 237 , 393±398. 19. Kim, H. & Lipscomb, W.N. (1993) X-ray crystallographic d eter- mination of the structure of b ovine lens leucine aminopept idase complexed with a mastatin: formulation of a c atalytic mechanism featuring a gem-diolate transition state. Biochemistry 32, 8465± 8478. 20. Stra È ter, N., Sun, L., Kantrow itz, E.R. & Lipscomb, W.N. (1999) A b icarbonate ion a s a general base i n the mechanism of peptide hydrolysis by dizinc leucine aminopeptidase. Proc. Natl Acad. Sci. USA 96, 1115 1±11155. 21. Burley, S.K., David, P.R ., Taylor, A. & Lipscomb, W.N. (19 90) Molecular structure of leucine aminopeptidase at 2.7-A Ê resolution. Proc. N atl Acad. Sci. USA 87 , 6878 ±6882. 22. Rees, D .C., Le wis, M. & Lipscomb, W.N . (1983) Re®ned crystal structure of carboxypeptidase A at 1.54 A Ê resolution. J. Mol. Biol. 168, 367±387. 23. Artymiuk, P.J., Grindley, H.M., Park, J.E., Rice, D.W. & Willett, P. (1992) T hree-dimensional structural r esemblance b etween leu- cine am inop eptidase and c arbo xypeptidase A revealed by graph- theoretical t echniques. FEBS Lett. 30 3, 48±52. 24. Savijoki, K. & Palva, A. (2000) P uri®cation and molecular char- acterization of a tripeptidase (PepT) from Lactobacillus helveticus. Appl. E nviron. Microbiol. 66, 794±800. 25. Bosman, B.W., Tan, P.S.T. & Konings, W.N. (1990) Puri®ca- tion and characterization of a tripeptidase from Lacto coccus lactis subsp. crem oris Wg2. Appl. Environ. Microbiol. 56 , 1839± 1843. 26. Simitsopoulou, M., Vafopoulou, A., Cho li-Pap adopou lou, T. & Alichanidis, E. (1 997) Puri®c ation a nd partial characterization of a tripeptidase from Pediococcus pentosac eus K9.2. Appl. Enviro n. Microbiol. 63 , 4872±4876. 27. Park, Y.S., Cha, M.H., Yong, W M., Kim, H.J., Chung, I.Y. & Lee, Y.S. (1999) The puri®cation and characterization of Bacillus subtilis tripeptidase (PepT). J. Biochem. Mol. Biol. 32, 239±246. 28. Hermsdorf, C.L. (1978) Tripeptide-speci®c amino peptidase from Escherichia coli AJ005. Biochemistry 17 , 3370±3376. 29. Makarova, K .S. & Grishin, N.V. (1999) The Zn-peptidase super- family: fu nctional convergence after evolutionary divergence. J. Mol. Biol. 292, 1 1±17. 30. Henrich, B., Monn erjahn, U. & Plapp, R. (1990) Peptidase D gene (pep D)ofEsch erichia coli K-12: nucleotide sequence, transcript mapping, and c omp arison w ith o ther pe ptidase g en es. J. Bacter iol. 172, 4641±4651. 31. Wagner, F.W., Wilkes, S.H. & Prescott, J.M. (1972) Speci®city of Aeromonas aminopeptidase t oward amino acid amides and dipeptides. J. Biol. Chem. 247, 1208±1210. 32. Wilkes, S .H., Bayliss, M.E. & Prescott, J.M. (1973) Speci®city of Aeromonas aminopeptidase t oward oligopeptides and po lypep - tides. Eur. J. Bioc hem. 34, 4 59±466. 33. Biagini, A. & Puigserver, A. (2001) Sequence analysis of the aminoacylase-1 family. A new proposed signature for metallo- exopeptida ses. Comp Biochem. Physiol . B B iochem. M ol. Biol. 12 8 , 469±481. 34. Carson, M. (1991) Ribbons 2.0. J. Appl. Crystallogr. 24, 958±961. 35. Nicholls, A. & Honig, B.J. (1991) A rapid ®nite-dierence algo- rithm, utilizing successive over-relaxation to solve the Poisson± Boltzmann equation. J. Comput. C hem. 12, 435±445. 450 K. Ha Ê kansson and C. G. Miller (Eur. J. Biochem. 269) Ó FEBS 2002 . between the structure of t he second domain in CG2 and peptidase T extends to the interchain contacts around the twofold crystallographic axis of the two structures. program PROCHECK [16]. The ma jor difference between the two structures is that the 20 N-terminal residues of the mature CG2 are lacking in peptidase T and that peptidase T