Crystal Structure of 2-Methylisocitrate Lyase (PrpB) from Escherichia coli and Modelling of its Ligand Bound Active Centre

doi:10.1016/S0022-2836(03)00358-9

J. Mol. Biol. (2003) 328, 609–621

Crystal Structure of 2-Methylisocitrate Lyase (PrpB) from Escherichia coli and Modelling of its Ligand Bound Active Centre Clemens Grimm1, Andreas Evers1, Matthias Brock2, Claudia Maerker2 Gerhard Klebe1, Wolfgang Buckel2 and Klaus Reuter1* 1

Institut fu¨r Pharmazeutische Chemie, Philipps-Universita¨t Marburg, Marbacher Weg 6 35032 Marburg, Germany

2

Laboratorium fu¨r Mikrobiologie, Fachbereich Biologie, Philipps-Universita¨t Marburg, 35032 Marburg Germany

Following acetate, propionate is the second most abundant low molecular mass carbon compound found in soil. Many microorganisms, including most, if not all fungi, as well as several aerobic bacteria, such as Escherichia coli and Salmonella enterica oxidize propionate via the methylcitrate cycle. The enzyme 2-methylisocitrate lyase (PrpB) from Escherichia coli catalysing the last step of this cycle, the cleavage of 2-methylisocitrate to pyruvate and succinate, was crystallised and its structure determined to a resol˚ . The enzyme, which strictly depends on Mg2þ for catalysis, ution of 1.9 A belongs to the isocitrate lyase protein family. A common feature of members of this enzyme family is the movement of a so-called “active site loop” from an open into a closed conformation upon substrate binding thus shielding the reactants from the surrounding solvent. Since in the presented structure, PrpB contains, apart from a Mg2þ, no ligand, the active site loop is found in an open conformation. This conformation, however, differs significantly from the open conformation present in the so far known structures of ligand-free isocitrate lyases. A possible impact of this observation with respect to the different responses of isocitrate lyases and PrpB upon treatment with the common inhibitor 3-bromopyruvate is discussed. Based on the structure of ligand-bound isocitrate lyase from Mycobacterium tuberculosis a model of the substrate-bound PrpB enzyme in its closed conformation was created which provides hints towards the substrate specificity of this enzyme. q 2003 Elsevier Science Ltd. All rights reserved

*Corresponding author

Keywords: crystal structure; isocitrate lyase; helix swapping; methylcitrate cycle; propionate degradation

Introduction Besides acetate, propionate is the second most abundant low molecular mass carbon compound found in soil. It is generated by the degradation of several amino acids, fermentation of carbohydrates and oxidation of odd-chain fatty acids.1 Consequently, many aerobic bacteria and fungi as well Present addresses: C. Grimm, EMBL, 6 rue Jules Horowitz, BP181, 38042 Grenoble Cedex 9, France; M. Brock and C. Maerker, Institut fu¨r Mikrobiologie, Fachbereich Biologie, Universita¨t Hannover, Schneiderberg 50, 30167 Hannover, Germany. Abbreviations used: CoA, Coenzyme A; PDB, protein data bank; PEP, phosphoenolpyruvate. E-mail address of the corresponding author: [email protected]

as some anaerobes are able to use propionate as an energy and carbon source. The first and mandatory step in the degradation of propionate is its activation to propionyl-CoA. In some anaerobic bacteria and in animals propionyl-CoA is carboxylated to yield methylmalonyl-CoA which rearranges in a coenzyme B12-dependent step to succinyl-CoA, an intermediate of the citric acid cycle. In aerobic bacteria, such as Escherichia coli or most, if not all fungi, however, propionyl-CoA is passed on to an alternative propionate oxidizing pathway called the methylcitrate cycle.2 In this pathway, oxaloacetate is condensed with propionyl-CoA yielding (2S,3S)-2-methylcitrate and free CoA, a reaction that is catalysed by methylcitrate synthase.3 In the subsequent steps, methylcitrate is dehydrated to 2-methyl-cisaconitate by the action of a specific 2-methylcitrate

0022-2836/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved

610

Crystal Structure of 2-Methylisocitrate Lyase

Figure 1. Methylcitrate cycle for a-oxidation of propionate to pyruvate. Enzymes specific for this cycle are indicated.

dehydratase and rehydrated to (2R,3S)-2-methylisocitrate by aconitase B from the citric acid cycle.4 Finally 2-methylisocitrate lyase cleaves (2R,3S)-2methylisocitrate into pyruvate and succinate.5 Because oxaloacetate is easily regenerated from succinate, the methylcitrate cycle mediates the a-oxidation of propionate to pyruvate (Figure 1). The cycle closely resembles the part of the glyoxylate cycle responsible for a-oxidation of acetate to glyoxylate. Although used as an energy and carbon source by many microorganisms, addition of propionate to glucose medium leads to a concentration-dependent inhibition of fungal growth. Therefore, this short chain fatty acid serves as a preservative mainly used in food industry. The growth inhibiting effect of propionate may be due to the accumulation of propionyl-CoA. Elevated levels of propionyl-CoA lead to an unbalanced acetylCoA/propionyl-CoA ratio which might competitively inhibit other house keeping enzymes dealing with CoA-esters as substrate or product. This assumption is supported by the fact, that spore colour development is disturbed in the presence of propionate in an Aspergillus nidulans strain with a methylcitrate synthase negative genetic background. The spore colour derives from the polyketide naphtopyrone and its synthesis is strongly dependent on acetyl-CoA. In addition, the propionate concentration required for growth inhibition is decreased in such a strain by a factor of five compared to wild-type A. nidulans.3 Despite the fact

that growth of E. coli is not inhibited on glucose/ propionate, bacteria and fungi show the same stereochemistry of the intermediates during propionate degradation.5 Therefore, structural knowledge of enzymes involved in the methylcitrate cycle is desirable, since specific inhibition of any of these enzymes is likely to result in a reduced amount of propionate necessary for food preservation. We have crystallised 2-methylisocitrate lyase (the PrpB enzyme) from E. coli5 and determined its structure. Since crystals were only obtained with substrate-free PrpB, a model of this enzyme bound to its ligands was created based on the crystal structure of substrate bound isocitrate lyase from Mycobacterium tuberculosis.6 Isocitrate lyase, an enzyme involved in acetate assimilation via the glyoxylate cycle, is related to 2-methylisocitrate lyase with respect to both its structure and the catalysed reaction. However, possible differences in the substrate specificity of isocitrate lyases and methylisocitrate lyases from pro- and eukaryotic organisms are discussed.

Results and Discussion Structure determination and quality of the model The structure of E. coli 2-methylisocitrate lyase (PrpB) has been solved using a mercury derivative

611

Crystal Structure of 2-Methylisocitrate Lyase

Table 1. Crystallographic data and phasing statistics A. Data collection Crystal ˚) Resolution (A Radiation source Space group ˚ ): Unit-cell parameters (A Number of observed reflections Number of unique reflections Completenessa (%) I=sIa Rsym a,b (%) B. Phasing Number of sites per asym. unit Phasing powerc centric Phasing powerc acentric C. Refinement R-factora,d (%) Rfree a,d (%) Deviations from ideal geometry ˚) Bond lengths (A Bond angles (8) Average B-factors ˚ 2) Overall (A ˚ 2) Protein (A ˚ 2) Mg2þ (A, B) (A ˚ 2) Water (A

Native 30.0–1.90 Cu Ka P32 21 a ¼ b ¼ 82:9 c ¼ 166:4 255,945

Mercury derivative 99.0– 2.30 Cu Ka P32 21 a ¼ b ¼ 83:2 c ¼ 166:7 472,731

52,996

30,495

98.2 (98.9) 20.5 (3.0) 5.4 (37.0)

99.9 (100.0) 25.5 (8.5) 8.6 (31.6)

– – –

2 1.02 1.41

21.3 (24.5) 23.4 (27.5)

– –

0.005 1.2

– –

26.5 24.8 36.9, 43.3 31.3

– – – –

a

Number in parentheses is for highest resolution shell. Rsym ¼ SlI 2 kIll=SI; where I is the observed intensity and kIl is the average intensity for multiple measurements. c Phasing power is the root-mean-square ðlFh l=EÞ; where lFh l is the heavy-atom structure factor amplitude and E is the residual lack of closure error. d R-factor ¼ SlFobs 2 Fcalc l=SlFobs l; where Fobs and Fcalc are observed and calculated bulk solvent corrected structure factors, respectively. Rfree is the cross-validation R-factor calculated for 3% of the reflections omitted in the refinement process. b

of the PrpB crystal. This derivative allowed the structure determination without the use of synchrotron radiation by means of single isomorphous replacement including the relatively small anoma˚ in the computation lous signal of mercury at 1.54 A ˚ to an (SIRAS). The final model was refined at 1.9 A R-factor of 21.3% and an Rfree of 23.4% (Table 1). The asymmetric unit contains two PrpB monomers showing virtually identical conformations. Electron density was observed in both monomers for all amino acid residues of the enzyme except for the carboxy-terminal four residues and the aminoterminal serine. The first residue included in the model (a leucine) represents residue 3 in the PrpB numbering scheme taking into account the aminoterminal methionine which is, however, not present in the mature PrpB.5 The structure shows good stereochemistry with 92.6% of all model residues in the most favoured region of the Ramachandran plot calculated with the program PROCHECK.7 Apart from Lys121 of subunit B, which is located in a loop with poorly defined electron density, solely the q/c angle combination of

Asp87 within each subunit corresponds to a position at the border between the generously allowed and the disallowed region of the Ramachandran plot. This residue is located in the active site and involved in binding of a Mg2þ (discussed below). Overall three-dimensional architecture and comparison with other members of the isocitrate lyase family As deduced from its primary structure PrpB belongs to the isocitrate lyase protein family comprising isocitrate lyases of bacterial and eukaryotic origin as well as phosphonoenolpyruvate mutases (PEP mutases) and carboxyphosphonoenolpyruvate phophonomutases (CPEP mutases), which both catalyse a reaction related to that of isocitrate and methylisocitrate lyases.8 – 10 Like all members of this protein family that have been characterised biochemically and/or structurally so far,6,11 – 13 PrpB forms a homotetramer.5 The crystal structure shows that the PrpB subunits are arranged in an identical tetrameric fashion as observed in the crystal structures of other members of this protein family (Figure 2(a)). More accurately, this quaternary structure has to be described as a dimer of dimers. The asymmetric unit of the PrpB crystal contains one dimer which is related to the second dimer completing the tetramer by a 2-fold crystallographic axis. The PrpB subunit comprises a single domain adopting a fold closely related to the classical (ba)8 barrel. As holds true for the other members of the isocitrate lyase family, the eighth helix of the (ba)8 barrel does not contact the inner layer of the barrel formed by eight parallel b-strands but points away from the barrel and leaves a gap in its outer a-helical layer. This gap is closed by the eighth barrel a-helix of the second PrpB subunit of the same dimer that protrudes in an identical fashion from its barrel. The gap present in the barrel of the second subunit is in turn closed by the eighth barrel a-helix of the first PrpB subunit, a phenomenon referred to as “helix swapping”11 (Figure 2(b)). The crystal structures of members of the isocitrate lyase protein family that are known so far comprise two bacterial isocitrate lyases, namely from M. tuberculosis and E. coli,6,13 as well as isocitrate lyase from the filamentous fungus Aspergillus (Emericella) nidulans12 and PEP mutase from the mussel Mytilus edulis.11 E. coli PrpB consists, like mussel PEP mutase, of 295 amino acid residues. Thus, these two enzymes constitute the smallest proteins of the isocitrate lyase family. Bacterial isocitrate lyases consist of about 430 and the even larger eukaryotic isocitrate lyases of about 540 residues. Comparison of the structures of E. coli PrpB and the eukaryotic A. nidulans isocitrate lyase shows a number of insertions within isocitrate lyase not present in PrpB (Figures 3 and 4(a)). In both enzymes the amino-terminal end of the (ba)8 barrel is partially shielded from the

612

Crystal Structure of 2-Methylisocitrate Lyase

Figure 2. Oligomerization of E. coli PrpB. (a) Three-dimensional structure of the homotetramer (dimer of dimers), with each subunit coloured differently. The Mg2þ bound to the active sites of the monomers are shown as bright blue spheres. (b) “Helix swapping” between two dimer subunits. In subunit A the b-strands and a-helices contributing to the formation of the (ba)8 barrel are coloured in blue or red, respectively. Loops and secondary structure elements not contributing to the barrel are coloured ochre. Subunit B is shown in grey. The bound Mg2þ are shown as bright blue spheres. The eighth (ba)8 barrel helices are exchanged among the subunits within one dimer. The ribbon representations were created using MOLSCRIPT27 and Raster3D.28

solvent by an a-helix near the amino terminus. In A. nidulans isocitrate lyase, some 50 residues encompassing two further helices, which are not present in PrpB, precede this helix. The second a-helix of the (ba)8 barrel is extended at its carboxy terminus in A. nidulans isocitrate lyase compared to E. coli PrpB. This extension is also present but slightly shorter in the known structures of bacterial isocitrate lyases.6,13 The bulkiest insertion is present in the region connecting the fifth b-strand and the fifth a-helix of the (ba)8 barrel. Here, 123 residues comprise an additional a/b-domain in A. nidulans isocitrate lyase, which is completely missing in E. coli PrpB. In bacterial isocitrate lyases this insertion is smaller by 95 residues and essentially reduced to a parallel/antiparallel b-sheet atop of the barrel. The extra domain of the eukaryotic isocitrate lyase has been proposed to mediate targeting to peroxisomes,14 which is not required in bacteria. Finally, the regions connecting the sixth a-helix to the seventh b-strand and the seventh b-strand to the seventh a-helix of the barrel are both enlarged by five or eight amino acid residues,

respectively, compared to E. coli PrpB. In PrpB as well as in all members of the isocitrate lyase protein family the eighth helix of the (ba)8 barrel involved in helix swapping is followed by a carboxy-terminal extension comprising further a-helices and mediating contacts to neighbouring protein molecules within the oligomer. Compared to PrpB, in isocitrate lyases this extension is still prolonged by 15 – 20 residues. In the crystal structure of A. nidulans isocitrate lyase most of these residues are disordered, and therefore not visible in the model shown in Figure 3(b). PrpB active site For all members of the isocitrate lyase protein family catalysis is strictly dependent on Mg2þ. Accordingly, a Mg2þ is observed within the active site in all isocitrate lyase and PEP mutase crystal structures determined so far, even in the absence of further ligands and regardless of whether Mg2þ has been added for crystallisation or not.6,11 – 13 In addition, structure determination of M. tuberculosis

Crystal Structure of 2-Methylisocitrate Lyase

613

Figure 3. Comparison of the three-dimensional structures of (a) E. coli PrpB and (b) isocitrate lyase from the eukaryote A. nidulans. Both enzymes are shown in similar orientation. Secondary structure elements forming the (ba)8 barrel are coloured in dark blue, the remaining secondary structure elements and loops in light blue. Sections that are not present in PrpB but in isocitrate lyases of both bacterial and eukaryotic origin are coloured in ochre, those unique to eukaryotic isocitrate lyases in red. The numbering of secondary structure elements considers only b-strands and a-helices contributing to the (ba)8 barrel. Amino and carboxy termini are indicated by an N or C, respectively. The Mg2þ bound to PrpB is shown as a bright blue sphere. The arrow in (b) indicates the location of the A. nidulans isocitrate lyase bound Mg2þ. The ribbon representations were created using MOLSCRIPT27 and Raster3D.28

isocitrate lyase with bound inhibitors and in ligand-free form has shown, that for substrate binding and catalysis a drastical conformational change of a loop spanning 10– 11 amino acid residues is necessary.6 This loop contains the isocitrate lyase signature sequence Lys-Lys-Cys-Gly-His including the active cysteine. When exposed to the isocitrate lyase reaction products, namely glyoxylate or succinate, the conformational change will take place only upon binding of succinate but not of glyoxylate alone. Binding of succinate appears to trigger the movement of the Mg2þ within the ˚ . This allows the first Lys active site by some 2.5 A of the Lys-Lys-Cys-Gly-His motif to form electrostatic interactions with this region resulting in closure of the active site loop and thus in shielding of the bound substrates from the surrounding solvent. Simultaneously, the 17 carboxy-terminal residues of the second subunit within the dimer undergo a conformational change occupying the space created by the lid closure and trapping it in the catalytic conformation.6 A similar induced fit upon substrate binding is postulated for all members of the isocitrate lyase protein family. Accordingly, in the structures of substrate or

inhibitor bound PEP mutase, the active site loop adopts a closed conformation.10 – 11 In the structures of E. coli and A. nidulans isocitrate lyases, whose active sites contain glyoxylate only or no substrate at all, this loop is present in the open conformation.12 – 13 In the crystal structure of E. coli PrpB presented here, no substrate is bound to the enzyme, accordingly the active site loop is found in an open conformation. In addition, a Mg2þ is present in the active site of PrpB at a position equivalent to that observed for substrate free isocitrate lyases from E. coli, M. tuberculosis and A. nidulans. The bivalent cation is bound in a highly negatively charged region mainly formed by the side-chains of residues Asp58, Asp85, Asp87 and Glu115 (Figure 5). The side-chain carboxylate of Asp87 is directly ˚ ), involved in chelating the Mg2þ (distance 2.3 A which obviously forces its q/c angles into an unfavourable region of the Ramachandran plot. Inspection of the Ramachandran plots of the known ligand-free structures of isocitrate lyases shows that the corresponding residues (which are glutamate in case of E. coli and M. tuberculosis isocitrate lyases) display similar q/c angles.

Figure 4 (legend opposite)

615

Crystal Structure of 2-Methylisocitrate Lyase

Figure 4. (a) Structure-based sequence alignment of proteins of the isocitrate lyase family with known structure. PrpB_Ec denotes the PrpB enzyme from E. coli, ICL_MT, ICL_Ec and ICL_An denote isocitrate lyase from M. tuberculosis, E. coli or A. nidulans, respectively. PEPM_Me denotes the PEP mutase from M. edulis. The PrpB secondary structure elements were assigned according to PROCHECK.7 Only b-strands (arrows) and a-helices (bars) contributing to the (ba)8 barrel are numbered. Residues at least common to PrpB and the three isocitrate lyases within the alignment are shaded grey. The signature sequences of PrpB, KRCGH, and of isocitrate lyases, KKCGH, are shaded bright yellow. Acidic residues involved in Mg2þ complexation are shaded bright blue. The residues proposed here to be important specificity determinants towards pyruvate or glyoxylate as substrate are shaded ochre or pink, respectively. Residues observed in the crystal structure of ligand bound M. tuberculosis isocitrate lyase6 to interact with Mg2þ, glyoxylate or succinate are denoted by an p . The sequence identity of E. coli PrpB to M. tuberculosis, E. coli and A. nidulans isocitrate lyase and to M. edulis PEP-Mutase amounts to 26%, 27%, 21% or 27%, respectively. The root mean square deviations for the Ca atoms of residues common to PrpB and the remaining enzymes aligned ˚ (M. tuberculosis and E. coli isocitrate lyase), 2.7 A ˚ (A. nidulans isocitrate lyase) and 2.3 A ˚ in this Figure are 2.6 A (M. edulis PEP mutase) as determined by means of the program DALI.29 (b) Sequence alignment around residues constituting the “pyruvate/glyoxylate specificity triade” proposed here. The respective residues are shaded ochre, pink or bright green, respectively. Abbreviations are used as in (a). Further abbreviations: PrpB_St, PrpB from Salmonella typhimurium; PrpB_Pa, PrpB from Pseudomonas aeruginosa; PrpB_Nm, PrpB from Neisseria meningitis; MICL_Sc, methylisocitrate lyase from Saccharomyces cerevisiae.

Flexibility of E. coli PrpB active site loop compared to isocitrate lyase active site loop Interestingly, the open conformation adopted by the active site loop of PrpB differs completely from the open conformation observed in the three known structures of substrate-free isocitrate lyases (Figure 6). The conformation of this loop in PrpB mainly results from its interaction with the carboxy-terminal helix of the second subunit within the same dimer, although slight contacts of this loop to both a-helix 4 and the loop connecting b-strand 3 and a-helix 3 of a third subunit within the same tetramer are also observed. In contrast, in the structures of ligand-free isocitrate lyases the active site loops interact in their open form with a b-sheet which is inserted between the fifth b-strand and the fifth a-helix of the (ba)8 barrel and which is not present in PrpB. In addition, the about 17 carboxy-terminal residues of isocitrate lyase which also undergo a conformational change upon substrate binding in order to lock the active site loop in its closed conformation6 are missing in PrpB. Therefore, the conformational change taking place in PrpB upon ligand binding must essentially be restricted to the movement of the active site loop. These facts may account for the different sensitivities against proteases upon treatment with 3-bromopyruvate for isocitrate lyase and 2-methylisocitrate lyase. This inhibitor forms a covalent adduct with the active site cysteine (Cys123 in E. coli PrpB) by substitution of the bromine. Crystal structure analysis of 3-bromopyruvate treated isocitrate lyase from M. tuberculosis revealed that the

pyruvyl moiety occupies part of the succinate binding site within the active site, thus inducing the closure of the active site loop.6 3-Bromopyruvate-treated E. coli PrpB is heavily digested by trypsin, while the unalkylated enzyme is hardly affected by this protease.5 In contrast, for E. coli isocitrate lyase it was shown, that treatment with 3-bromopyruvate decreases its sensitivity against proteases, while the native enzyme is highly susceptible to protease digestion.15 Furthermore, inactivation of PrpB by 3-bromopyruvate is completed directly after addition of equimolar amounts of the alkylating substrate to the protein.5 Complete inactivation of isocitrate lyases by 3-bromopyruvate as shown for the enzymes of E. coli and Pseudomonas indigofera, however, takes This implicates a several minutes.15 – 16 strengthened shielding of the active site loop by the surrounding protein structures of isocitrate lyases compared to E. coli PrpB. Modelling of the PrpB active site loop in its closed conformation A catalytic mechanism for the enzyme reaction in isocitrate lyase has been proposed by Sharma et al.6 describing the crystal structures of both, inhibitor-bound and unliganded M. tuberculosis isocitrate lyase. Regarding the reverse reaction, in which glyoxylate and succinate are condensed to isocitrate, glyoxylate has to bind first followed by the binding of succinate. The catalytic cysteine of the Lys-Lys-Cys-Gly-His signature within the active site loop acts as a base abstracting a proton

616

Crystal Structure of 2-Methylisocitrate Lyase

Figure 5. Stereo view of the Mg2þ (shown as a blue sphere) bound in the active site of PrpB. The smaller red spheres represent well ordered water molecules. The superimposed 2Fobs 2 Fcalc electron density map was calculated from the model coordinates lacking both the Mg2þ and the neighbouring water molecules after a 1000 K simulated annealing run. The map is contoured at 2.8s.

from the Ca-atom of succinate. A particular glutamate residue is considered to protonate the carboxylate adjacent to the bond formed. It has been proposed that the negative charge of the aldehyde oxygen of glyoxylate is stabilised by the Mg2þ, as well as by two basic residues. All the residues thought to be involved in catalysis of isocitrate lyase are strictly conserved in 2-methylisocitrate lyase (PrpB), which catalyses the same reaction with the substrates differing in only one methyl group (2-methylisocitrate instead of isocitrate or rather pyruvate instead of glyoxylate). Therefore, it must be postulated, that both enzymes are following an identical catalytic mechanism. Nevertheless, a structure of ligand-bound PrpB with the active site loop in its closed conformation seemed desirable for the identification of features that account for the high substrate specificity observed for this enzyme. Since we were, despite of extensive efforts, not able to crystallise 3-bromopyruvate-treated PrpB or native PrpB in the presence of pyruvate and succinate, we created a model of this enzyme with bound ligands and the active site loop in a closed conformation. This model was created as described in Materials and Methods based on the known structure of the M. tuberculosis isocitrate lyase active site loop in the closed conformation (PDBcode 1f8i). The Mg2þ, pyruvate and succinate ligands were transferred to the model into equivalent positions as present in M. tuberculosis isocitrate lyase. The resulting model clearly demonstrates that the active centres of both enzymes share a high similarity. A total of 13 of the 18 residues that are reported by Sharma et al.6 to interact with Mg2þ, glyoxylate or succinate bound to isocitrate lyase are strictly conserved in the active site of PrpB. In particular, all residues of the flexible active site loop that directly interact with the ligands are identical in PrpB and isocitrate lyase. In contrast to the signature sequence Lys-Lys-CysGly-His present in isocitrate lyases, the active site loop of all 2-methlylisocitrate lyases known so far contains the slightly modified sequence Lys-Arg-

Cys-Gly-His around Cys123 (PrpB numbering5). Since the second basic residue of this sequence (lysine in isocitrate lyases or arginine in 2-methylisocitrate lyases, respectively) makes no contact to any substrate but points away from the catalytic pocket, it is unlikely to contribute to substrate specificity. Among those residues within the active site that are not strictly conserved there are two conservative exchanges (Glu155 in isocitrate lyase is replaced by Asp87 in PrpB and Ser315 in isocitrate lyase is replaced by Thr212 in PrpB). In addition, Trp93 is exchanged by Gly47 in PrpB. This residue, however, contacts the glyoxylate (pyruvate in PrpB) through its main-chain nitrogen with the side-chain not being involved in the formation of the active site. There are only two residues contacting substrates via their side-chains within the isocitrate lyase active site that differ significantly from the corresponding residues in PrpB: while in isocitrate lyase Ser317 makes a hydrogen bond to a succinate carboxylate via its side-chain hydroxyl, the corresponding residue in PrpB, Phe214, does not contact a substrate. In addition, the isocitrate lyase residue Thr347 hydrogen bonded to a carboxyl group of succinate, is replaced in PrpB by Pro236. This proline participates there in the formation of a hydrophobic depression accommodating the pyruvate methyl group not present in glyoxylate. This hydrophobic depression is further flanked by the side-chains of Phe186 and Leu234 that, in M. tuberculosis isocitrate lyase, are replaced by Trp283 and Phe345 (Figure 7). Phe186, Leu234 and Pro236 are apparently strictly conserved in all bacterial 2-methylisocitrate lyases similar to Trp283, Phe345 and Thr347 that are strictly conserved in all isocitrate lyases (Figure 4(b)). This suggests that these residues represent important determinants for the respective substrate specificities. Interestingly, the sequence of 2-methylisocitrate lyase from the eukaryotic organism Saccharomyces cerevisiae (encoded by the ICL2 gene) which is comparable in size to eukaryotic isocitrate lyases reveals that in this enzyme the

Crystal Structure of 2-Methylisocitrate Lyase

617

Figure 6. Comparison of the active site loop from E. coli PrpB with the active site loop from M. tuberculosis isocitrate lyase in its open and closed conformation. (a) The conformation of E. coli PrpB open active site loop (outlined in yellow) of subunit A (coloured red) mainly results from interactions with the carboxy-terminal helix of subunit B (grey) within the same dimer. (b) The open conformation of M. tuberculosis isocitrate lyase open active site loop (outlined in yellow) is caused by its interaction with an insertion between the fifth b-strand and the fifth a-helix of the (ba)8 barrel which is not present in PrpB (see also Figure 3). Isocitrate lyase subunit A is coloured in light blue. Subunit B (grey) makes in this conformation no interactions with the active loop. (c) Active site loop (outlined in yellow) of M. tuberculosis isocitrate lyase (coloured blue) in its closed conformation: the loop is locked in this conformation by the carboxy-terminal 17 residues of subunit B (grey), which also undergo a conformational change upon substrate binding. These carboxy-terminal residues have no equivalents in PrpB and are partially disordered in the structure of M. tuberculosis isocitrate lyase with the active site loop in the open conformation. (d) Superposition of (a), (b) and (c) in stereo view with the B subunits omitted for clarity. The active site loops are outlined in yellow. The wire representations were created using PyMOL (Warren L. DeLano “The PyMOL Molecular Graphics System” DeLano Scientific, San Carlos, CA, USA; http://www.pymol.org

618

Crystal Structure of 2-Methylisocitrate Lyase

Figure 7. Superposition of active site residues from M. tuberculosis isocitrate lyase in its closed conformation (Protein Data Bank access code: 1f8i) and E. coli PrpB modelled into its closed conformation (see Materials and Methods) in stereo view. Carbon atoms of isocitrate lyase are coloured in grey, while those of PrpB are coloured in yellow. The Mg2þ is shown as a bright blue sphere. The glyoxylate as present in the crystal structure of isocitrate lyase has been modified by adding a methyl group to formally yield pyruvate (carbon atoms coloured in yellow). While major parts of both substate binding pockets show very high similarity, residues Trp283, Phe 345 and Thr347 in isocitrate lyase are exchanged by the triade Phe186, Leu234 and Pro236 in PrpB, which forms a depression harbouring the methyl group of pyruvate. This methyl group comes in the model unfavourably close to the Thr347 side-chain methyl group of isocitrate lyase (indicated by a dotted line). Amino acids are labelled in the Figure using the one-letter code. The stick representation was created using MOLSCRIPT27 and Raster3D.28

Phe186-Leu234-Pro236 triade of PrpB is not conserved. In this enzyme, the first residue of the triade is a Trp as it is in isocitrate lyases, while the second residue is a Leu as it is in prokaryotic PrpB enzymes. The third residue of the triade (a Thr in isocitrate lyases and a Pro in PrpB enzymes) is a Ser in S. cerevisiae methylisocitrate lyase (Figure 4(b)). Remarkably, upon addition of a methyl group to the glyoxylate bound in M. tuberculosis isocitrate lyase, particularly the Thr347 side-chain methyl group lacking in S. cerevisiae methylisocitrate lyase comes unfavourably close to the methyl group of the resulting ˚ ). Obviously, in S. cerevisiae pyruvate (distance 2.7 A 2-methylisocitrate lyase substrate specificity is achieved by a different strategy as in bacterial PrpB. Summary and outlook Comparison of the E. coli methylisocitrate lyase (PrpB) structure with the so far determined structures of the related eukaryotic and bacterial isocitrate lyases reveals a number of regions within isocitrate lyases that are not present in the substantially smaller PrpB. These regions comprise an amino-terminal and a carboxy-terminal extension as well as specific insertions within loops connecting b-strands and a-helices of the (ba)8 barrel core common to all members of the isocitrate lyase family. As shown in this work, an insertion between the fifth b-strand and the fifth a-helix of the (ba)8 barrel core results in a different open conformation of the so-called active site loop in isocitrate lyases compared to PrpB. Furthermore, in

its closed conformation that is adopted upon substrate binding the active site loop is fixed in isocitrate lyases by a carboxy-terminal extension of the second dimer subunit, which is not present in PrpB. These structural differences may account for the different sensitivities against proteases of 3-bromopyruvate-treated isocitrate lyases and PrpB. Based on the M. tuberculosis isocitrate lyase structure, a model of PrpB with all ligands bound and the active site loop in its closed conformation was created in this study. It reveals that the active sites of isocitrate lyase and PrpB show high similarity apart from one restricted region obviously harbouring in PrpB the methyl group of pyruvate which is missing in the isocitrate lyase substrate glyoxylate. Within this region, a depression is formed in PrpB framed by the Phe186, Leu234 and Pro236 side-chains. These residues are replaced in all known isocitrate lyases by a Trp, a Phe and a Thr residue, respectively. We therefore postulate that the triade Phe186, Leu234 and Pro236 constitutes an important determinant of substrate specificity in PrpB, while the corresponding residues being Trp, Phe and Thr may confer specificity towards glyoxylate in isocitrate lyases. Inspection of the sequence of the eukaryotic S. cerevisiae 2-methylisocitrate lyase reveals that in this enzyme the first residue of the “substrate specificity triade” is a Trp (as it is in isocitrate lyases) while the second residue of the triade is a Leu (as it is in bacterial PrpB enzymes). The third residue is a Ser and thus neither corresponds to its counterpart in isocitrate lyases nor in PrpB. So it seems that substrate specificity is achieved in a

619

Crystal Structure of 2-Methylisocitrate Lyase

different fashion by PrpB and 2-methylisocitrate lyase from S. cerevisiae. Site-directed mutagenesis experiments are required to figure out exactly the substrate specificity determinants of isocitrate lyases and 2-methylisocitrate lyases. Structural knowledge about 2-methylisocitrate lyase active sites may finally serve as a basis for the rational design of specific inhibitors of these enzymes in order to increase sensitivity of microbes against propionate as a food preservative.

obtained from SOLVE by means of single isomorphous replacement including the relatively small anomalous ˚ in the computation (SIRAS) signal of mercury at 1.54 A were then improved by solvent flattening within RESOLVE. ˚ The map calculated from the improved phases at 3.5 A resolution allowed location of the overall shape of the expected (ba)8-barrel. An initial model was gained by placing two (ba)8-barrel (poly-serine) scaffolds obtained from the CPEP mutase model (Protein Data Bank access code 1pym) into the electron density map using the program O.24

Materials and Methods

Completion and refinement of the model

Crystallisation

The initial model was subjected to a rigid body refinement of the two subunits (data in the resolution range of ˚ to 3 A ˚ ) and a 3500 K simulated annealing (SA) in tor8A sion angle space and subsequent conjugate gradient minimisation refinement (data in the resolution range of ˚ to 1.9 A ˚ ) within the program CNS.21 The refinement 8A was carried out under non-crystallographic symmetry (NCS) restraints using a weight factor of 100. The progress of the refinement was monitored by the help of a free R-value; calculated from a fraction of 3% of the total number of reflections excluded from the refinement. After this first refinement, an R=Rfree -factor of 45.9%/ 48.6% was reached. The phases calculated from the model were then weighted with the SigmaA method using the program SIGMAA from the CCP4 package20 and merged with the experimental phases. The obtained phased dataset was then fed into the program DM from the CCP4 suite20 for phase improvement by 25 cycles of solvent flattening, histogram matching and 2-fold aver˚ aging together with phase extension starting from 4.5 A ˚ electron resolution. With these improved phases a 3 A density was calculated which was then inspected together with the refined model. During this first inspection, some loop regions could be fitted into the correct place of the density and 35 residues at the amino and carboxy terminus of each monomer could be built using the program O.24 At this time also the correct amino acid residues according to the PrpB sequence were introduced. The resulting model was again automatically refined by a 1500 K simulated annealing run in torsion angle space together with a conjugate gradient minimisation refinement and followed by a grouped B-factor ˚ to 1.9 A ˚ resolution. refinement using data from 8 A From the resulting improved model 2Fobs 2 Fcalc and Fobs 2 Fcalc electron density maps were calculated, which allowed the completion of all missing parts of the model. At this stage all available reflection data were included in the refinement and a bulk solvent correction was introduced. After two further cycles of manual inspection and automated refinement, the R=Rfree -factor converged at 21.3%/23.4% (Table 1).

The protein used for crystallisation was recombinant PrpB, which was isolated as described by Brock et al.5 All crystallisation experiments were carried out at 18 8C in Chryschem crystallisation plates using the sittingdrop vapour-diffusion technique. NaN3 (0.02% w/v) was added to all reservoir solutions. A drop of 1 ml protein solution (PrpB protein, 15 g/l in 10 mM Hepes pH 7.0) was mixed with 1 ml reservoir solution and sealed against 0.5 ml reservoir solution. An initial screening was carried out using the 96 solutions of an in-house crystal screen kit. Well formed crystals appeared after five days incubation with 2.0 M (NH4)2SO4 buffered with 0.1 M Mes at pH 6.5. Large diffraction-quality crystals could be grown by decreasing the (NH4)2SO4 concentration to 1.7 M and adjusting the pH within a range of 7.5 to 8.5 (Tris/HCl buffer). Data collection and processing X-ray data were collected on a Rigaku RU-300 rotating anode generator (Rigaku/MSC) operating at 50 kV and 100 mA and equipped with focussing mirrors (MSC/ Yale) and an R-AXIS IV image plate detector. Crystals were flash cooled in nylon-fibre loops in a 100 K nitrogen gas stream provided by an X-stream cryo system. As a cryo buffer the reservoir solution containing 20% glycerol (v/v) was used. Diffraction data were indexed and processed using the programs DENZO and XDISP.17 The pre-processed data were then scaled and merged using SCALEPACK.17 The statistics of the datasets are listed in Table 1. Structure determination Attempts to carry out initial phasing by the molecular replacement method18 using the coordinates of M. edulis PEP mutase (protein data bank access code 1pym) by means of the program AMoRe19 from the CCP4 suite20 and the CNS package21 failed. For the purpose of structure determination via the method of multiple wavelength anomalous diffraction (MAD), we then prepared a selenomethionyl derivative of PrpB as described by VanDuyne et al.22 This derivative, however, denatured irreversibly in the absence of high salt concentrations and was not crystallisable under the conditions used for the natural PrpB enzyme. As an alternative, a mercury derivative of the PrpB crystal was created by incubating the crystals in a solution of 1.8 M MgSO4 buffered with 0.1 M Tris – HCl (pH 8.5) and endowed with 0.5 mM methyl mercury acetate. Heavy atom sites were detected using the program SOLVE, version 2.0.23 The phases

Modelling of ligand bound PrpB with its active site loop in a closed conformation Molecular modelling was performed using MODELLER5.25 For this purpose, the major part (except for the loop region Gly119 to Lys129) of the model was generated using the crystal structure of PrpB presented here as template. The loop was modelled using the corresponding residues (residues Ala186 to Lys197) of the closed form of M. tuberculosis isocitrate lyase (Protein Data Bank access code: 1f8i) as template. Ten homology

620

models of the closed form of PrpB were generated with MODELLER5. All generated models were subjected to a default simulated annealing refinement protocol available in MODELLER5. The ten generated models were compared mainly considering the orientations of those residues that expose their side-chains into the active site. As no significant differences were apparent, the model with the lowest MODELLER5 target function ˚ to value showing root mean square deviations of 0.9 A a ˚ ˚ ˚ 1.5 A (C atoms) or rather 1.4 A to 2.0 A (all atoms) to the remaining nine models was selected as representative. The Mg2þ as well as the glyoxylate (i.e. pyruvate) and succinate ligands were then placed into the model using positions equivalent to those observed in 1f8i followed by an energy minimisation using the MAB (molecular atom bond) force-field26 in Moloc. Applying default parameters, the main portion of the protein was kept rigid, allowing only the segments comprising residues 58 – 59, 86 –88, 119 – 128 (constituting the active site loop) and 212– 216, to move. Comparing the crystal structures of M. tuberculosis isocitrate lyase in its open and in its closed form (Protein Data Bank access codes 1f61 or 1f8i, respectively) reveals that the residues within the equivalent regions of this enzyme show significant induced fit adaptations upon ligand binding, resulting ˚ to 2.5 A ˚ (not considering the in Ca-atom shifts of 0.5 A active site loop).

Crystal Structure of 2-Methylisocitrate Lyase

5.

6.

7.

8.

9.

10. Protein Data Bank accession numbers The PrpB coordinates have been deposited in the Protein Data Bank and will be released upon publication. The Protein Data Bank access codes are 1mum (PrpB crystal structure) and 1mzx (model of ligand-bound PrpB with active site loop in its closed conformation).

11.

12.

Acknowledgements We gratefully acknowledge the help of Professor Dr Milton Stubbs during data collection and thank Dr Andreas Heine and Bernhard Stengl for their help with the stereo Figures. We thank Christian Sohn for his excellent technical assistance. This work was supported by the DFG (Graduiertenkolleg: “Protein Function at the Atomic Level”).

References 1. Chin, K. & Conrad, R. (1995). Intermediary metabolism in methanogenic paddy soil and the influence of temperature. FEMS Microbiol. Ecol. 18, 85 – 102. 2. Textor, S., Wendisch, V. F., DeGraaf, A. A., Mu¨ller, U., Linder, M., Linder, D. & Buckel, W. (1997). Propionate oxidation in Escherichia coli: evidence for operation of a methylcitrate cycle in bacteria. Arch. Microbiol. 168, 428– 436. 3. Brock, M., Fischer, R., Linder, D. & Buckel, W. (2000). Methylcitrate synthase from Aspergillus nidulans: implications for propionate as an antifungal agent. Mol. Microbiol. 35, 961– 973. 4. Brock, M., Maerker, C., Schu¨tz, A., Vo¨lker, U. & Buckel, W. (2002). Oxidation of propionate to pyruvate in Escherichia coli: involvement of methylcitrate

13.

14.

15. 16.

17. 18. 19. 20.

dehydratase and aconitase. Eur. J. Biochem. 269, 6184– 6194. Brock, M., Darley, D., Textor, S. & Buckel, W. (2001). 2-Methylisocitrate lyases from the bacterium Escherichia coli and the filamentous fungus Aspergillus nidulans—characterization and comparison of both enzymes. Eur. J. Biochem. 268, 3577– 3586. Sharma, V., Sharma, S., Hoener zu Bentrup, K., McKinney, J. D., Russel, D. G., Jacobs, W. R., Jr & Sacchettini, J. C. (2000). Structure of isocitrate lyase, a persistence factor of Mycobacterium tuberculosis. Nature Struct. Biol. 7, 663– 668. Laskowski, R. A., McArthur, M. W., Moss, D. S. & Thornton, J. M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallog. 26, 283– 291. Hidaka, T., Imai, S., Hara, O., Anzai, H., Murakami, T., Nagaoka, K. & Seto, H. (1990). Carboxyphosphonoenolpyruvate phosphonomutase, a novel enzyme catalysing C – P bond formation. J. Bacteriol. 172, 3066– 3072. Hidaka, T., Hidaka, M., Uozumi, T. & Seto, H. (1992). Nucleotide sequence of a carboxyphosphonoenolpyruvate phosphonomutase gene isolated from a bialaphos-producing organism, Streptomyces hygroscopicus, and its expression in Streptomyces lividans. Mol. Gen. Genet. 233, 476– 478. Liu, S., Lu, Z., Jia, Y., Dunaway-Mariano, D. & Herzberg, O. (2002). Dissociative phosphoryl transfer in PEP mutase catalysis: structure of the enzyme/ sulfonate complex and kinetic properties of mutants. Biochemistry, 41, 10270– 10276. Huang, K., Li, Z., Jia, Y., Dunaway-Mariano, D. & Herzberg, O. (1999). Helix swapping between two a/b barrels: crystal structure of phosphoenolpyruvate mutase with bound Mg2þ-oxalate. Structure, 7, 539– 548. Britton, K. L., Langridge, S. J., Baker, P. J., Weeradechapon, K., Sedelnikova, S. E., De Lucas, J. R. et al. (2000). The crystal structure and active site location of isocitrate lyase from the fungus Aspergillus nidulans. Structure, 8, 349– 362. Britton, K. L., Abeysinghe, I. S., Baker, P. J., Barynin, V., Diehl, P., Langridge, S. J. et al. (2001). The structure and domain organization of Escherichia coli isocitrate lyase. Acta Crystallog. sect. D, 57, 1209– 1218. Matsuoka, M. & McFadden, B. A. (1988). Isolation, hyperexpression, and sequencing of the aceA gene encoding isocitrate lyase in Escherichia coli. J. Bacteriol. 170, 4528– 4536. Ko, Y. H. & McFadden, B. A. (1990). Alkylation of isocitrate lyase from Escherichia coli by 3-bromopyruvate. Arch. Biochem. Biophys. 278, 373– 380. Roche, T. E., McFadden, B. A. & Williams, J. O. (1971). Modification of the active site of isocitrate lyase from Pseudomonas indigofera. Arch. Biochem. Biophys. 147, 192– 200. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307– 326. Rossman, M. G. & Blow, D. M. (1962). The detection of sub-units within the crystallographic asymmetric unit. Acta Crystallog. sect. D, 15, 24 –31. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallog. sect. A, 50, 157– 163. Collaborative Computational Project, Number 4

Crystal Structure of 2-Methylisocitrate Lyase

21.

22.

23. 24.

(1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. sect. D, 50, 760– 763. Bru¨nger, A. T., Adams, P. D., Clore, G. M., DeLano, W., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. sect. D, 54, 905–921. VanDuyne, G. D., Standaert, R. F., Karplus, P. A., Schreiber, S. L. & Clardy, J. (1993). Atomic structures of the human immunophilin FKBP-12 complexes with FK506 and rapamycin. J. Mol. Biol. 229, 105–124. Terwilliger, T. C. & Berendzen, J. (1999). Automated MAD and MIR structure solution. Acta Crystallog. sect. D, 55, 849– 861. Jones, T. A. & Kjeldgaard, M. (1997). Electron density map interpretation. Methods Enzymol. 277, 173– 208.

621 25. Sˇali, A. & Blundell, T. L. (1993). Comparative protein modelling by satisfaction of spatial restraints. J. Mol. Biol. 234, 779–815. 26. Gerber, P. R. (1998). Charge distribution from a simple molecular orbital type calculation and nonbonding interaction terms in the force field MAB. J. Comput. Aid. Mol. Des. 12, 37 – 51. 27. Kraulis, P. J. (1991). MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallog. 24, 946–950. 28. Merrit, E. A. & Bacon, D. J. (1997). Raster3D: photorealistic molecular graphics. Methods Enzymol. 277, 505 –524. 29. Holm, L. & Sander, C. (1993). Protein structure comparison by alignment of distance matrices. J. Mol. Biol. 233, 123–138.

Edited by R. Huber (Received 16 October 2002; received in revised form 27 February 2003; accepted 4 March 2003)