Cocrystal Structures of Diaminopimelate Decarboxylase

Structure, Vol. 10, 1499–1508, November, 2002, 2002 Elsevier Science Ltd. All rights reserved.

PII S0969-2126(02)00880-8

Cocrystal Structures of Diaminopimelate Decarboxylase: Mechanism, Evolution, and Inhibition of an Antibiotic Resistance Accessory Factor Soumya S. Ray,1,5 Jeffrey B. Bonanno,1,3,5 K.R. Rajashankar,1,5 Mariana G. Pinho,2 Guoshun He,1,3 Herminia De Lencastre,2 Alexander Tomasz,2 and Stephen K. Burley1,3,4 1 Laboratory of Molecular Biophysics, 2 Laboratory of Microbiology, and 3 Howard Hughes Medical Institute The Rockefeller University 1230 York Avenue New York, New York 10021

Summary Cocrystal structures of Methanococcus jannaschii diaminopimelate decarboxylase (DAPDC) bound to a substrate analog, azelaic acid, and its L-lysine product have been determined at 2.6 A˚ and 2.0 A˚, respectively. This PLP-dependent enzyme is responsible for the final step of L-lysine biosynthesis in bacteria and plays a role in ␤-lactam antibiotic resistance in Staphylococcus aureus. Substrate specificity derives from recognition of the L-chiral center of diaminopimelate and a system of ionic “molecular rulers” that dictate substrate length. A coupled-enzyme assay system permitted measurement of kinetic parameters for recombinant DAPDCs and inhibition constants (Ki) for azelaic acid (89 ␮M) and other substrate analogs. Implications for rational design of broad-spectrum antimicrobial agents targeted against DAPDCs of drug-resistant strains of bacterial pathogens, such as Staphylococcus aureus, are discussed. Introduction Antibiotic resistance in pathogenic bacteria has become a major public health concern [1]. High-level utilization of ␤-lactams and other antimicrobial agents have placed the bacterial ecosystem under tremendous selection pressure, leading to the emergence of multidrug-resistant strains of major human pathogens, such as Staphylococcus aureus [2]. By early 1990 epidemic clones of methicillin-resistant Staphylococcus aureus (MRSA) had been detected around the world [3, 4], and strains that remain susceptible to only one therapeutically useful antimicrobial agent, vancomycin, have been described [5]. Since 1997 sporadic isolates of MRSA with reduced susceptibility to vancomycin have also been reported in several countries [6, 7, 8]. These developments raised the specter of untreatable staphylococcal disease [1], stimulating a systematic search for new antimicrobial agents against MRSA and substantial research efforts focused on mechanisms of drug resistance in MRSA. In MRSA the primary determinant of antibiotic resis4 5

Correspondence: [email protected] These authors contributed equally to this work.

tance is a low-affinity penicillin binding protein, PBP2a [9, 10, 11], encoded by the mecA gene [12]. It has been known for some time, however, that the presence of the mecA gene and its protein product cannot fully account for resistance in clinical isolates of MRSA, which demonstrate a wide range of minimal inhibitory concentrations (MICs) across isolates bearing comparable levels of PBP2a [13, 14]. Genetic analysis of transposon mutants of MRSA identified a number of auxiliary genes that work in synergy with PBP2a and appear to be essential for high-level methicillin resistance in S. aureus. Several of these auxiliary genes encode enzymes involved in cell wall metabolism, while the biochemical and cellular functions of other auxiliary resistance gene products remain unknown [15]. One of these auxiliary genes, aux239, was identified on the basis of DNA sequence homology as the S. aureus gene encoding diaminopimelate decarboxylase (DAPDC) [16]. In bacteria, L-lysine is derived from L-aspartate via a multistep biosynthetic pathway, with three slightly different intermediate pathways that operate in parallel and converge to give daminopimelate, or DAP (Figure 1A). DAPDC catalyzes the final common decarboxylation step to yield the end product L-lysine [17]. Unlike bacteria, humans obtain L-lysine from dietary sources and lack the biosynthetic machinery depicted in Figure 1A; the closest human ortholog of S. aureus DAPDC (SaDAPDC) is ornithine decarboxylase (ODC; sequence identity, 25%). In many bacteria, including S. aureus, lysine is the essential diamino acid component of the bacterial cell wall peptidoglycan, the structural stability of which critically depends on peptide crosslinks formed between the N6 amino group of lysine and the carboxyl function of D-alanine residues in neighboring muropeptide subunits of the cell wall. The inhibitory power of methicillin and other ␤-lactam antibiotics derives from blockage of peptide crosslink formation, via inhibition of the transpeptidase functions of penicillin binding proteins (PBPs). Loss of peptide crosslinks compromises the integrity of the cell wall, leading to bacterial death. A transposon insertion into the gene encoding DAPDC of the MRSA strain COL did not affect growth in rich organic media but reduced the level of methicillin resistance by nearly 100-fold [16]. Supraphysiologic concentrations of L-lysine (e.g., 10 mM) partially overcame the observed loss of methicillin resistance. DAPDC, therefore, represents a potential target for broad-spectrum inhibitors, either for use as primary antibiotics or as adjuncts to existing therapies. Extensive attempts to obtain diffraction-quality crystals for DAPDC from S. aureus failed. An ortholog isolated from Methanocococcus jannaschii (sequence identity to S. aureus enzyme, 38%) proved more amenable to crystallization. Here, we present X-ray structures of M. jannaschii DAPDC (Mj-DAPDC) bound to a subKey words: antibiotic resistance; structural genomics; X-ray crystallography; diaminopimalate decarboxylase; lysine biosynthesis

Structure 1500

Figure 1. L-Lysine Biosynthetic Pathways and DAPDC Inhibitors (A) Schematic representation of L-lysine biosynthesis pathways in bacteria. (B) Chemical structures and Ki values (where appropriate) for substrate, product, and structurally similar inhibitors.

strate analog, azelaic acid, and its product, L-lysine. Modeling of DAP in the active site permitted elucidation of the mechanism of substrate recognition. A coupledenzyme assay was used to examine the inhibitory properties of azelaic acid with DAPDCs from M. jannaschii, S. aureus, and Escherichia coli. Implications for rational design of DAPDC inhibitors are discussed. Results and Discussion Structural Overview M. jannaschii DAPDC is a 50 kDa two-domain enzyme with atypical 8-fold ␣/␤ barrel and ␤ sandwich domains (Figure 2A). DAPDC belongs to the type III class of dimeric PLP enzymes (alanine racemase family). Unlike classical TIM barrel enzymes, the N-terminal secondary

structural motif of the ␣/␤ barrel domain is an ␣ helix. The largely C-terminal ␤ sandwich domain contains an ␣ helix and a ␤ strand derived from the N terminus of the protein, which makes extensive intramolecular contacts with a C-terminal “Greek key” motif. The enzyme forms a symmetric homodimer in the crystal with an extensive intersubunit interface (Figure 2A). Both the N- and C-terminal domains participate in dimer formation. In addition, the C terminus of one monomer is draped across its partner, further stabilizing the homodimer. As expected, two PLP molecules (one per monomer) were seen in the electron density map, identifying the location of the two active sites within the dimer interface (Figure 2A). The PLP moiety makes a Schiff’s base with Lys73, a feature observed frequently in other members of this class of PLP-dependent enzymes, such

Crystal Structure of Diaminopimelate Decarboxylase 1501

Figure 2. Structure of the Mj-DAPDC/Azelaic Acid Complex (A) Ribbon diagram of the Mj-DAPDC homodimer with PLP cofactor and bound azelaic acid inhibitor depicted as color-coded spacefilled atoms (C, gray; N, blue; O, red; P, yellow). One of the monomers is uniformly colored red. The second monomer is colored to show the TIM barrel (blue), the N-terminal portion of the ␤ sandwich domain (yellow), and the C-terminal part of the ␤ sandwich domain (green). (B and C) Active site electron density features (|Fobserved| ⫺ |Fcalculated| difference Fourier synthesis contoured at 1 ␴, corresponding to bound (B) azelaic acid, depicted as a colorcoded atomic stick figure in two possible conformations, and (C) L-lysine.

as ornithine decarboxylase [18]. Bound azelaic acid appears to exist in multiple conformations, as judged from the electron density map illustrated in Figure 2B. DAPDC Sequence Comparison Figure 3A documents that about 25% of the 434 residues composing Mj-DAPDC are highly conserved among the members of this PLP enzyme subfamily. Without exception, all sites at which there are significant differences map to the surface of the enzyme (Figure 3B), where mutations would be readily tolerated, or represent conservative changes of buried residues unlikely to destabilize the hydrophobic core. Insertions and deletions in the DAPDC sequences, aligned in Figure 3A, map to random coil portions of the structure, where they would not disrupt organized secondary-structural elements. The remarkable level of sequence identity and the pat-

tern of amino acid differences across archaea and eubacteria allow us to conclude that all known DAPDCs share the same three-dimensional structure [19]. Surface electrostatic potential calculations demonstrate the presence of a conserved, highly basic surface feature within the enzyme active site, which is consistent with recognition of carboxylate groups in the substrate and Couloumb repulsion of the positively charged L-lysine product (Figures 3B and 3C). DAPDC Active Site and Substrate Recognition The active site of DAPDC is composed of residues from both dimer subunits (Lys73A, His214A, Arg297A, Tyr337A, Cys362B, Glu363B, and Ser364B; A and B denote individual protomers). Figure 4A depicts the azelaic acid-bound enzyme active site in detail. Each PLP forms a Schiff’s base with Lys73A and stacks with His214A or

Structure 1502

Figure 3. DAPDC Sequence Alignment and Surface Properties (A) Sequence alignment of DAPDCs from selected pathogens. Secondary-structural elements are shown with cylinders (␣ helices) and arrows (␤ strands). Gray dots denote poorly resolved residues in the final electron density map. Color-coding denotes sequence conservation among DAPDCs (white to green ramp, 30%–100% identity). Red asterisks denote invariant or near-invariant active site residues. (B) GRASP [42] representation of the chemical properties for the solvent-accessible surface of Mj-DAPDC, calculated with a water probe radius of 1.4 A˚. Solvent-accessible surface color-coded for (left panel) sequence conservation with the color ramp in (A) and (right panel) the calculated electrostatic potential (red and blue, respectively, represent electrostatic potentials ⬍⫺8 and ⬎⫹8 KBT, respectively, where KB is the Boltzmann constant and T is the temperature). The calculations were performed with an ionic strength of zero and dielectric constants of 80 and 2 for solvent and protein, respectively. The active site (invariant and highly basic) and the dimer interface (conserved) are labeled.

the corresponding residues in subunit B. Arg297A makes a salt bridge with the carboxylate group of azelaic acid remote from the PLP cofactor. Other interactions between the enzyme and azelaic acid include the following: His214A, NE2 → O1 ⫽ 3.6 A˚; Arg297A, closest contact, NH2 → O8 ⫽ 2.7 A˚; Tyr337A, OH-O9 ⫽ 3.3 A˚; Ser364B, N → O1 ⫽ 3.7 A˚ (Table 1; see Figure 4A for the azelaic acid atom-numbering scheme). Between the salt bridges, the aliphatic backbone of azelaic acid exhibits conformational flexibility. Similar behavior is often observed when NZ atoms of lysine side chains make salt bridges or hydrogen bonds [20]. Diaminopimelate is a meso compound with opposite chiralities at each terminus (referred to as “D-stereocenter” and “L-stereocenter,” hereafter). Productive decarboxylation must occur at the D-stereocenter to ensure synthesis of L-lysine and not D-lysine. This mechanistic

constraint requires recognition of at least one, if not both, stereochemically distinct centers within the substrate. In an attempt to understand substrate recognition by Mj-DAPDC, we modeled DAP into the active site in place of azelaic acid (Figure 4B). The L-stereocenter of the modeled substrate could be recognized by the enzyme via interactions analogous to those seen with azelaic acid, with the location of the D-stereocenter being dictated by the formation of a Schiff’s base with PLP. Following adjustment of the torsion angles of two active site residues (Arg297A and Glu363B; Figure 4B), the modeled substrate could form salt bridges and hydrogen bonds with residues from both halves of the dimer (Figure 4B). Recognition of the L-stereocenter appears to result from two salt bridges: Glu363B and Arg297A with the substrate amino (L-stereocenter) and carboxylate (L-stereocenter) groups, respectively. In ad-

Crystal Structure of Diaminopimelate Decarboxylase 1503

with the L-stereocenter making the Schiff’s base) failed because of severe steric clashes (data not shown). Inspection of DAPDC sequences from various organisms demonstrates absolute conservation of all active site residues observed in azelaic acid binding and those implicated in substrate binding in our model of the enzyme-substrate complex (Figure 3A, asterisk), with only one exception, Ser364B (serine or threonine in all available DAPDC sequences). We believe that the proposed model for substrate recognition is common to all known DAPDCs.

Figure 4. DAPDC Active Site (A) Ribbon representation of the two polypeptide chains comprising the dimeric enzyme, with monomers A and B colored gold and blue, respectively. Active site residues and azelaic acid are depicted as color-coded atomic stick figures (N, blue; C, gray; O, red; S, yellow; P, orange) with a shaded molecular surface representation emphasizing polar atoms (gray, C; red, O; blue, N). Lys64A has been omitted for the sake of clarity. The carbon atoms of the inhibitor backbone have been numbered. (B) Atomic stick figure representation of DAP modeled in the active site to permit formation of the required Schiff’s base with PLP and to optimize ionic interactions with the L- and D-stereocenters that are analogous to the observed contacts with azelaic acid. Colorcoding is as in (A), with intermolecular distances given for atoms at the L-stereocenter.

dition to the Schiff’s base between the amino group of the D-stereocenter of DAP and PLP, the carboxylate group at the D-stereocenter of the substrate could make hydrogen bonds with His214A and Ser364B. Attempts to model the substrate in the opposite orientation (i.e.,

Evolution of Type III PLP Enzymes: “Made to Measure” Active Sites Both DAPDC and its ortholog, ODC, are members of the alanine racemase (AR), or type III, family of PLPdependent enzymes. Figure 5A illustrates a superposition of AR (orange), ODC (blue), and Mj-DAPDC (green) prepared by overlaying the ␣/␤ barrel domains (rmsd ca. 1.9 A˚ for ⵑ230 equivalent ␣ carbons). The only significant structural differences between these three enzymes result from modest rotations of the ␤ sandwich domains with respect to the ␣/␤ barrels. In terms of both size and chemical complexity, alanine is the simplest of the three enzyme substrates in question. The cocrystal structure of AR bound to ethylamine phosphonate shows that the phosphonate group (analogous to the carboxylate group of alanine) makes a salt bridge with Arg257B (bold text denotes AR residues; Figure 5B). Cys311B of AR acts as the catalytic residue to effect the racemization reaction [21]. L-ornithine and DAP represent somewhat more complicated, yet chemically similar, substrates. When the active sites of ODC and DAPDC are overlayed, the two enzymes display similar spatial dispositions of active site residues (Figure 5B). His214A in Mj-DAPDC and His197A in ODC (italic text denotes ODC residues) occupy structurally equivalent positions, making ␲-␲ stacking interactions with their respective PLP cofactors. On the basis of a detailed structural comparison, Cys360B, the catalytic residue in ODC [18], corresponds to Cys362B in Mj-DAPDC, suggesting that the catalytic mechanisms of these two enzymes are probably identical (Figure 5B). The substrates of these closely related enzymes are, however, chemically distinct, and there is no substrate crossreactivity between ODC and DAPDC [22]. Asp361B in ODC makes a salt bridge with the ␣ amino group of L-ornithine. The equivalent position in Mj-DAPDC is occupied by Glu363B, which could make an analogous salt bridge with the L-stereocenter of DAP. This subtle difference almost certainly reflects the fact that L-orni-

Table 1. Distances between Atoms of Amino Acid of DAPDC and Ligands M. jannaschii DAPDC

Azelaic Acid (2.6 A˚ resolution)

L-Lysine (2.0 A˚ resolution)

PLP His214A Arg297A Try337A Cys362B Glu363B Ser364B

C4 → O1 ⫽ 3.6 A˚ NE2 → O1 ⫽ 3.7 A˚ NH2 → O8 ⫽ 2.7 A˚ OH → O9 ⫽ 3.3 A˚ SG → O1 ⫽ 6.6 A˚ OE2 → C5 ⫽ 3.5 A˚ N → O1 ⫽ 3.7 A˚

C4 → NZ ⫽ 3.4 A˚ NE2 → NZ ⫽ 5.2 A˚ O → NH2 ⫽ 3.4A˚ OH → N ⫽ 4.3 A˚ SG → NZ ⫽ 3.4 A˚ OE2 → N ⫽ 4.2 A˚ O → NZ ⫽ 4.2 A˚

Structure 1504

Figure 5. Comparison of AR, ODC, and DAPDC (A) Ribbon drawing superposition of AR (orange), ODC (blue), and Mj-DAPDC (green). (B) Comparison of the active sites of AR, ODC, and DAPDC. Arg257B in AR, Asp361B and Asp332A in ODC, and Glu363B, Tyr337A, and Arg297A in DAPDC act as molecular rulers contributing to recognition of their respective substrates. (Inset) Superposition of the active site residues of Mj-DAPDC and ODC, showing the correlation between the lengths of the substrates and the acidic residues responsible for salt bridge formation between substrate and enzyme.

thine is one carbon-carbon bond-length shorter than the DAPDC substrate. We believe that these enzymes use Asp361B/Glu363B as “molecular rulers” to “measure” the lengths of substrates bound in their respective active sites (Figure 5B). The presence of Ser364B in MjDAPDC, which is analogous to Gly362B in ODC, can be explained, at least in part, by the requirement for DAPDC to recognize the D-stereocenter of its meso substrate. ODC has no such constraint because its natural substrate exists only in the L configuration. The catalytic mechanisms of each of these type III

PLP-dependent enzymes are probably similar. In the first essential step, all three enzymes make a Schiff’s base adduct with the substrate amino group. However, the substrates for the three enzymes differ in their chemical complexity, particularly at the nonreactive ends of the substrate. AR is the simplest of the three and requires only one molecular ruler, Arg 257B, to recognize the carboxylate group of either D- or L-alanine. ODC appears to utilize at least two molecular rulers (Asp332A and Asp361B) to form salt bridges with the ␦ amino group of L-ornithine, without undesired binding to the

Crystal Structure of Diaminopimelate Decarboxylase 1505

⑀ amino group of its closest chemical analog, L-lysine. The active sites of DAPDCs recognize a substrate of even higher complexity and must also discriminate between the two chiral centers of DAP. Mj-DAPDC and, by inference, all DAPDCs appear to do so with three molecular rulers (Arg297A, Tyr337A, and Glu363B in MjDAPDC). We suggest that AR, ODC, and DAPDC have evolved increasingly complex active sites, which reflect the increasing chemical complexity of their respective substrates. It also appears likely that the precise way in which the DAPDCs recognize DAP favors decarboxylation over the other types of chemical reactions, such as transaminations and racemization, commonly catalyzed by PLP-dependent enzymes [23]. Structure of the Mj-DAPDC/Product Complex L-lysine (the product of DAP decarboxylation) is a known feedback inhibitor of the DAPDCs [24]. Determination of the structure of Mj-DAPDC from crystals grown in the presence of added L-lysine revealed an active site electron density feature that cannot be accounted for by the sequence of the protein, PLP, or water molecules. This feature could be readily explained by the presence of the bound product L-lysine (Figure 2C). There are no major differences in the conformations of the active site residues in the structures of the monoclinic (L-lysinebound) and hexagonal (azelaic acid-bound) crystal forms of Mj-DAPDC, except for the side chain of Cys362B, which points in a different direction (Table 1). If, indeed, the electron density feature adjacent to the PLP cofactor does correspond to bound product, the conformation of Cys362B is consistent with the catalytic mechanism of ODC. The sulfhydryl group of Cys362B lies closest to the ⑀ carbon of the bound L-lysine product, where it can carry out the final catalytic step by a mechanism similar to that of ODC [18]. Enzymatic Activity and Inhibition of DAPDC Mj-DAPDC follows simple Michaelis-Menten’s kinetics, with no apparent cooperativity between subunits (data not shown). The specific activity of the purified enzyme preparation used for crystallography was 2.7 ⫾ 0.4 units/ mg, and the Km is 588 ⫾ 12 ␮M. Comparable specific activities were obtained for purified, recombinant EcDAPDC and Sa-DAPDC (3.8 ⫾ 0.2 units/mg and 3.3 ⫾ 0.3 units/mg, respectively). Azelaic acid was tested for its potential to inhibit the M. jannaschii enzyme and found to have a Ki of 89 ⫾ 15 ␮M. Added substrate reverses inhibition, demonstrating that azelaic acid is a competitive inhibitor. Ten other DAP-like compounds were tested for inhibition of Mj-DAPDC, and apparent Ki values are summarized in Figure 1B. Together with azelaic acid, these compounds encompass much of the chemical complexity of DAP. Caproic acid is a simple carboxylic acid that binds weakly to the enzyme. Adipic acid is a six carbon dicarboxylic acid that is similar to azelaic acid. The apparent Ki for adipic acid is slightly higher than that of azelaic acid, which may be explained by unfavorable electrostatic interactions with the carboxylate group of Glu363B, limiting binding to DAPDC. The hydrazine derivative of adipic acid has no significant effect and may

be too bulky to fit into the enzyme active site. Suberic acid and 2,4-pentenedioic acid are very similar to azelaic acid, except that these dicarboxylic acids are unsaturated. Surprisingly, these compounds appear to have almost no inhibitory effect. It is possible that torsion angle flexibility of the fully saturated aliphatic substrate analogs contributes to inhibitor binding because azelaic acid can be modeled in at least two conformations (Figure 2B). The rigid double bonds in suberic acid and 2,4pentenedioic acid would not permit the same torsional flexibility and could thereby increase the entropic penalty of active site binding. This phenomenon has been previously seen for inhibitors of HIV protease [25]. Diamines appear to be much stronger DAPDC inhibitors than dicarboxylic acids. The addition of various diamines to Mj-DAPDC leads to a slight decrease in intensity of the UV absorption of PLP at 430 nm, suggesting that these inhibitors make Schiff’s bases with the enzyme cofactor. Moreover, ultrafiltration of the MjDAPDC/diamine complex through a porous membrane (MW cutoff, 10 kDa) led to retention of the protein and passage of PLP (data not shown), indicating that the colored cofactor had dissociated from the enzyme after diamine addition. The strong inhibitory effect of the diamines may be attributed to PLP sequestration, which precludes recycling of the enzyme after inhibitor binding. The Ki for L-lysine, the end product feedback inhibitor of DAPDC, could not be measured. We believe that the Ki for L-lysine falls outside the range of detection by our coupled-enzyme assay system, implying a lower limit of about 500–1000 ␮M. Biological Implications DAPDC has a number of features that make it an attractive drug design target. First, this enzyme is found only in bacteria, and humans obtain the amino acid L-lysine from dietary sources. Second, DAPDC catalyzes the final step common to all known L-lysine biosynthesis pathways in prokaryotes. Third, the active sites of the M. jannaschii and S. aureus enzymes appear to be identical in structure, and there is good reason to believe that DAPDC inhibitors will have a broad spectrum of action. Finally, it has been demonstrated that inactivation of the gene encoding DAPDC of S. aureus causes a 100fold reduction in the MIC for methicillin [16]. From a mechanistic standpoint, DAPDC occupies a unique position among amino acid biosynthesis enzymes. Unlike other decarboxylases, DAPDC acts on a meso substrate and is specific for the D-stereocenter. Other D amino acid oxidases use FAD as a cofactor, making cross inhibition of this class of enzymes by DAPDC inhibitors unlikely. Orthologous human enzymes responsible for decarboxylating amino acids (e.g., lysine, ornithine, and histidine decarboxylases) act only on L substrates. The substrate recognition properties of DAPDC make it an appealing target for mechanismbased drug design, because it is unlikely that an inhibitor bearing a D-chiral center will interfere with metabolism of other amino acids. The evolutionary understanding gained from structural comparison of enzymes in this family should allow design of stereoselective inhibitors

Structure 1506

Table 2. Summary of Crystallographic Statistics Data Collection and Refinement Data set

Azelaic acid

Azelaic acid (Pt)

L-lysine

Space group Unit cell parameters

P6122 a ⫽ 80.8 A˚ c ⫽ 508.5 A˚

P6122 a ⫽ 80.9 A˚ c ⫽ 508.6 A˚

P21 a ⫽ 70.2 A˚ b ⫽ 147.1 A˚ c ⫽ 89.4 A˚

Wavelength Protomers/AU Number of measurements Number of unique reflections Resolution range Completeness (%)a Rmerge (%)a,b Average intensity [I/␴(I)]a Average multiplicity Working R factor (%) Free R factor (%) Rms deviations Bond lengths (A˚) Bond angles (⬚) Overall G factorc

␭ ⫽ 1.4 A˚ 2 257,466 30,684 30–2.6 A˚

␭ ⫽ 1.07 A˚ 2 98,834 9,707 20–3.7 A˚

96.0 (88.9) 7.0 (15.6) 32.8 (10.7) 8.4 24.4 28.3

83.5 (45.9) 9.4 (9.6) 30.9 (15.8) 10.2

0.006 1.48 0.2

␤ ⫽ 93.3⬚ ␭ ⫽ 0.98 A˚ 4 654,629 119,738 30–2.0 A˚

99.1 (100) 3.5 (10.1) 35.4 (13.1) 5.5 22.3 25.9 0.007 1.99 0.0

Phasing Statistics Number of heavy atom sites Resolution range used for phasing Phasing power (acentric)d Mean figure of merit (FOM)e Resolution range used for density modification Mean FOM after density modification using NCS

7 20–3.7 A˚ 1.35 0.48 30–2.6 A˚ 0.74

Model Statistics Number of atoms Number of water molecules Overall B factor

6,955 91 33

14,871 13,80 18

a

Numbers in parentheses indicate statistics for the highest resolution shell. Rmerge ⫽ [⌺hkl ⌺i|I(hkl)i ⫺ ⬍I(hkl)⬎|/⌺hkl⌺i⬍I(hkl)i⬎] ⫻ 100, where ⬍I(hkl)⬎ denotes mean intensity. c PROCHECK statistics [40]. d Phasing power ⫽ rms(FH/E), where E is the residual lack of closure. e FOM ⫽ F(hkl)best/F(hkl), where F(hkl)best ⫽ ⌺aPFhkl/⌺aP. b

that target bacterial DAPDCs, with minimal crossinhibition of enzymes contributing to amino acid synthesis and metabolism in humans. Although azelaic acid and the diamines bind MjDAPDC with reasonable affinities, none of these highly polar molecules (with attendant poor membrane permeability) inactivate DAPDC in vivo (data not shown). We suggest that ␣-fluoromethyl analogs of DAP represent credible lead compounds for the development of DAPDC inhibitors because the proposed flourine modifications would increase membrane permeability [26]. ODC inhibitors, such as difluromethyl ornithine, have proved useful in the treatment of trypanosomasis [27]. Our structure of the Mj-DAPDC/azelaic acid complex revealed that appropriately positioned fluorine atoms could be accommodated within the active site. Vederas and coworkers have explored the possibility of using DAP analogs for inhibition of DAP decarboxylase [28]. A number of substrate analogs have been tested for inhibition against DAP epimerase and DAP dehydrogenase, which are also enzymes contributing to lysine biosynthesis [29, 30, 31, 32]. Inhibition of an essential gene product often fails to

kill a pathogenic bacterium in vivo because the host milieu supplements the missing reaction product(s). Although we think it extremely unlikely for DAPDC, given the high concentrations of L-lysine required to overcome loss of enzyme activity in vitro, this scenario remains a formal possibility. Even in this context, an inhibitor of DAPDC may be of clinical utility in combination with a ␤-lactam antibiotic. In a widely used combination drug, Augmentin, the inclusion of clavulanic acid inhibits ␤-lactamase, thereby allowing the antimicrobial component amoxycillin to reach its target. By analogy, inhibition of an auxiliary gene, such as DAPDC, could resensitize resistant strains of S. aureus to ␤-lactam antibiotics. Experimental Procedures Protein Expression and Purification Genes encoding DAPDCs from S. aureus, E. coli, and M. jannaschii were amplified from bacterial genomic DNA (M. jannaschii primers, 5⬘-TAAGGATCCCAAGTCATCATGCAACCAG-3⬘ and 5⬘-GCTGCTAG CTTAGGTAATGACACAGTAGAG-3⬘; S. aureus primers, 5⬘-GGCGCT AGCACTGTTAAATATAATCAAAATGGCG-3⬘ and 5⬘-GATGGATCCT TCTACAGTCATCTATAATGC-3⬘; E. coli primers, 5⬘-TATGCTAGCAT GCCACATTCACTGTTCAG-3⬘ and 5⬘-CGAGGATCCCAGGATTTTAG

Crystal Structure of Diaminopimelate Decarboxylase 1507

ATGGATTCC-3⬘), cloned, and expressed as N-terminal hexaHis-tag fusion proteins in E. coli BL21(DE3). The recombinant proteins were purified to homogeneity with nickel ion affinity and anion exchange chromatographies. Following proteolytic removal of the affinity tag with PreScission protease (cut site, NH2-LGVLFQLP-COOH), the expected molecular masses of the proteins were confirmed by mass spectrometry. Crystallization Initial crystallization trials with S. aureus (Sa-DAPDC) and E. coli (Ec-DAPDC) enzymes by hanging drop vapor diffusion were unsuccessful. Similar trials with Mj-DAPDC yielded promising leads from various PEG 6000 conditions, which were optimized to yield small rod-shaped crystals in the monoclinic space group P21 (unit cell, a ⫽ 70.2 A˚, b ⫽ 147.1 A˚, c ⫽ 89.4 A˚, and ␤ ⫽ 93.3⬚; four molecules per asymmetric unit). Optimization of the monoclinic crystal form required the addition of the product L-lysine. The final crystallization condition was 12% (w/v) PEG 6000, 0.1 M Tris-Cl (pH 8.0), 20 mM MgCl2, and 10 mM L-lysine. Addition of a substrate analog, azelaic acid (25 mM), replacing L-lysine in the monoclinic crystallization condition (Figure 1B), produced diffraction-quality crystals in the hexagonal space group P6122 (unit cell, a ⫽ 80.8 A˚ and c ⫽ 508.5 A˚; two molecules per asymmetric unit). The SeMet form of Mj-DAPDC did not yield crystals under a wide variety of conditions, requiring a heavy metal soak to produce an isomorphous derivative of the hexagonal crystal form (5 mM dichloro-tetrapyridyl platinum chloride, 6 hr). Data Collection, Structure Determination, and Refinement Mj-DAPDC crystals were frozen by immersion in propane with 35% (w/v) glucose (hexagonal form) or 40% ethylene glycol (v/v) (monoclinic form) combined with the respective crystal mother liquor as cryoprotectants. Diffraction data were obtained from both crystal forms under standard cryogenic conditions and processed with DENZO/SCALEPACK [33]. A summary of data processing statistics is given in Table 2. Single isomorphous replacement anomalous-scattering phases were determined at 3.7 A˚ resolution with optimized Pt anomalous and native data. Pt sites were located with SnB [34], and experimental phases were calculated with SHARP [35] (Table 2). ESSENS [36] aided in the identification of helical features in the experimental electron density map, which permitted docking of the three-dimensional structure of an orthologous enzyme (Protein Data Bank ID 2TOD; trypanosomal ornithine decarboxylase [ODC], ⵑ26% identical to Mj-DAPDC). Two copies of ODC could be placed in the asymmetric unit, permitting the calculation of the noncrystallographic symmetry (NCS) operator. Density modification with NCS averaging and phase extension [37] improved the quality of the experimental electron density map significantly, enabling de novo tracing of the Mj-DAPDC polypeptide chain. The final model, consisting of two copies of residues b–437, two PLPs, two azelaic acids, and 91 water molecules, was refined with CNS [38] to a working R factor of 25%, with an Rfree of 28% at 2.6 A˚ resolution. Diffraction data measured from a single monoclinic crystal were phased by molecular replacement. With the Mj-DAPDC dimer as a search model, AMoRe [39] gave two solutions, corresponding to the four molecules/asymmetric unit, with a final overall correlation coefficient of 62% and an R factor of 35%. The monoclinic structure (four copies of residues b–437, four PLPs, four L-lysines, and 1380 water molecules) was refined with CNS [38] to a working R factor of 23%, with an Rfree of 26% at 2.0 A˚ resolution. For both structures, PROCHECK [40] documented excellent stereochemistry (Table 2), and the electron density corresponding the polypeptide backbone was everywhere continuous in 2|Fobserved| ⫺ |Fcalculated| difference Fourier syntheses contoured at 1 ␴. DAPDC Assays and Inhibitor Characterization A coupled-enzyme DAPDC assay was established by adapting a previously reported decarboxylase assay system [41]. A degassed buffer containing 100 mM Tris (pH 8.0), 10 mM MgCl2, 15 ␮M PLP, 10 units of malate dehydrogenase, 10 units of PEP carboxylase, and 0.31 mg/mL of NADH (OD 340 nm ⫽ 1.0) and 0.01% Triton X-100 or -114 was used for all assays. DAP was added from a 100

mM stock solution. DAPDC activity was estimated by monitoring CO2 production (i.e., change in OD at 340 nm, reflecting conversion of NADH to NAD). DAPDC was incubated with varying concentrations of substrate, and velocities were obtained from rate of consumption of NADH, as previously reported [41]. Data were plotted as a Line-Weaverburk plot to obtain Vmax and Km from the y intercept and x intercept, respectively. For measurement of Ki for azelaic acid, known quantities of azelaic acid were added to the assay and incubated with varying concentrations of substrate. Residual activity was measured as described above. Data were plotted as a LineWeaverburk plot. The Ki for azelaic acid was calculated from the slope of the plot with the following equation: the slope of the inhibited reaction equals (1 ⫹ [I]/Ki) ⫻ (Km/Vmax), where [I] is concentration of azelaic acid. Equal concentrations of all other substrate analogs were incubated with substrate and enzyme in assay buffer, and residual activity was measured and converted to enzyme velocity, as described above. Apparent Ki values for the remaining substrate analogs were measured relative to that of azelaic acid, which was used in the same concentration as other inhibitors. Control assays were carried out with the coupling enzymes in sodium bicarbonate buffer in the presence of overwhelming concentrations of the inhibitors. No loss of activity of the coupling enzymes was observed under these conditions, indicating that the compounds used for DAPDC inhibition studies do not crossreact with the coupling enzymes. Acknowledgments We are most grateful to Dr. M. Becker for access to Beamline X25 at the National Synchrotron Light Source, Dr. Kevin D’ Amico at the Advanced Photon Source, and Dr. Dan Thiel at the MacCHESS facility (Cornell). We thank Ms. T. Niven for manuscript preparation and Dr. G.A. Petsko and members of the New York Structural Genomics Research Consortium (NYSGRC) for useful discussions. This work was supported by the National Institutes of Health (P50 GM62529 to S.K.B. and AI45738 to A.T.), The Rockefeller University (S.K.B.), and the Burroughs-Wellcome Fund (K.R.R.). S.K.B. is an Investigator in the Howard Hughes Medical Institute. Mj-DAPDC represents NYSGRC target T135. Received: April 22, 2002 Revised: August 7, 2002 Accepted: August 15, 2002 References 1. WHO Expert Committee (2000). The use of essential drugs. World Health Organ. Tech. Rep. Ser. 895, 1–61. 2. Tenover, F.C., and Gaynes, R.P. (2000). The epidemiology of Staphylococcus infections. In Gram-Positive Pathogens, V.A. Fischetti, R.P. Novick, J.J. Ferreri, D.A. Portnoy, and J.I. Rood, eds. (Washington, D.C.: ASM Press), pp. 414–421. 3. Oliveira, D.C., Tomasz, A., and de Lencastre, H. (2002). The secrets of success of a human pathogen: molecular evolution of pandemic clones of methicillin-resistant Staphylococcus aureus. Lancet. Infect. Dis., in press. 4. Crisostomo, M.I., Westh, H., Tomasz, A., Chung, M., Oliveira, D.C., and de Lencastre, H. (2001). The evolution of methicillin resistance in Staphylococcus aureus: similarity of genetic backgrounds in historically early methicillin-susceptible and -resistant isolates and contemporary epidemic clones. Proc. Natl. Acad. Sci. USA 98, 9865–9870. 5. Tomasz, A. (1994). Multiple-antibiotic-resistant pathogenic bacteria. A report on the Rockefeller University Workshop. N. Engl. J. Med. 330, 1247–1251. 6. Hiramatsu, K., Aritaka, N., Hanaki, H., Kawasaki, S., Hosoda, Y., Hori, S., Fukuchi, Y., and Kobayashi, I. (1997). Dissemination in Japanese hospitals of strains of Staphylococcus aureus heterogeneously resistant to vancomycin. Lancet 350, 1670–1673. 7. Sieradzki, K., Roberts, R.B., Haber, S.W., and Tomasz, A. (1999). The development of vancomycin resistance in a patient with methicillin-resistant Staphylococcus aureus infection. N. Engl. J. Med. 340, 517–523.

Structure 1508

8. Linares, J. (2001). The VISA/GISA problem: therapeutic implications. Clin. Microbiol. Infect. 7, 8–15. 9. Hartman, B.J., and Tomasz, A. (1984). Low-affinity penicillinbinding protein associated with beta-lactam resistance in Staphylococcus aureus. J. Bacteriol. 158, 513–516. 10. Reynolds, P.E., and Brown, D.F. (1985). Penicillin-binding proteins of beta-lactam-resistant strains of Staphylococcus aureus. FEBS Lett. 192, 28–32. 11. Utsui, Y., and Yokota, T. (1985). Role of an altered penicillinbinding protein in methicillin- and cephem-resistant Staphylococcus aureus. Antimicrob. Agents Chemother. 28, 397–403. 12. Song, M.D., Wachi, M., Doi, M., Ishino, F., and Matsuhashi, M. (1987). Evolution of an inducible penicillin-target protein in methicillin-resistant Staphylococcus aureus by gene fusion. FEBS Lett. 221, 167–171. 13. Matthews, P.R., Reed, K.C., and Stewart, P.R. (1987). The cloning of chromosomal DNA associated with methicillin and other resistances in Staphylococcus aureus. J. Gen. Microbiol. 133, 1919–1929. 14. Hartman, B.J., and Tomasz, A. (1986). Expression of methicillin resistance in heterogeneous strains of Staphylococcus aureus. Antimicrob. Agents Chemother. 29, 85–92. 15. De Lencastre, H., and Tomasz, A. (1994). Reassessment of the number of auxiliary genes essential for expression of high-level methicillin resistance in Staphylococcus aureus. Antimicrob. Agents Chemother. 38, 2590–2598. 16. De Lencastre, H., Wu, S.W., Pinho, M.G., Ludovice, A.M., Filipe, S., Gardete, S., Sobral, R., Gill, S., Chung, M., and Tomasz, A. (1999). Antibiotic resistance as a stress response: complete sequencing of a large number of chromosomal loci in Staphylococcus aureus strain COL that impact on the expression of resistance to methicillin. Microb. Drug Resist. 5, 163–175. 17. Born, T.L., and Blanchard, J.S. (1999). Structure/function studies on enzymes in the diaminopimelate pathway of bacterial cell wall biosynthesis. Curr. Opin. Chem. Biol. 3, 607–613. 18. Grishin, N.V., Osterman, A.L., Brooks, H.B., Phillips, M.A., and Goldsmith, E.J. (1999). X-ray structure of ornithine decarboxylase from Trypanosoma brucei: the native structure and the structure in complex with alpha-difluoromethylornithine. Biochemistry 38, 15174–15184. 19. Sander, C., and Schneider, R. (1991). Database of homologyderived protein structures and structural meaning of sequence alignment. Proteins 9, 56–68. 20. Burley, S.K., Wang, A.H., Votano, J.R., and Rich, A. (1987). Antigelling and antisickling bisphenyl oligopeptides and peptide analogues have similar structural features. Biochemistry 26, 5091–5099. 21. Morollo, A.A., Petsko, G.A., and Ringe, D. (1999). Structure of a Michaelis complex analogue: propionate binds in the substrate carboxylate site of alanine racemase. Biochemistry 38, 3293– 3301. 22. Jackson, L.K., Brooks, H.B., Osterman, A.L., Goldsmith, E.J., and Phillips, M.A. (2000). Altering the reaction specificity of eukaryotic ornithine decarboxylase. Biochemistry 39, 11247– 11257. 23. Swanson, T., Brooks, H.B., Osterman, A.L., O’Leary, M.H., and Phillips, M.A. (1998). Carbon-13 isotope effect studies of Trypanosoma brucei ornithine decarboxylase. Biochemistry 37, 14943–14947. 24. Rosner, A. (1975). Control of lysine biosynthesis in Bacillus subtilis: inhibition of diaminopimelate decarboxylase by lysine. J. Bacteriol. 121, 20–28. 25. Velazquez-Campoy, A., Todd, M.J., and Freire, E. (2000). HIV-1 protease inhibitors: enthalpic versus entropic optimization of the binding affinity. Biochemistry 39, 2201–2207. 26. Tsukamoto, T., Haile, W.H., McGuire, J.J., and Coward, J.K. (1996). Synthesis and biological evaluation of N alpha-(4-amino4-deoxy-10-methylpteroyl)-DL-4,4-difluoroornithine. J. Med. Chem. 39, 2536–2540. 27. Petru, A.M., Azimi, P.H., Cummins, S.K., and Sjoerdsma, A. (1988). African sleeping sickness in the United States. Successful treatment with eflornithine. Am. J. Dis. Child. 142, 224–228. 28. Kelland, J.G., Arnold, L.D., Palcic, M.M., Pickard, M.A., and Vederas, J.C. (1986). Analogs of diaminopimelic acid as inhibi-

29.

30.

31.

32.

33.

34. 35.

36. 37.

38.

39. 40.

41.

42.

tors of meso-diaminopimelate decarboxylase from Bacillus sphaericus and wheat germ. J. Biol. Chem. 261, 13216–13223. Cirilli, M., Scapin, G., Sutherland, A., Vederas, J.C., and Blanchard, J.S. (2000). The three-dimensional structure of the ternary complex of Corynebacterium glutamicum diaminopimelate dehydrogenase-NADPH-L-2-amino-6-methylene-pimelate. Protein Sci. 9, 2034–2037. Caplan, J.F., Zheng, R., Blanchard, J.S., and Vederas, J.C. (2000). Vinylogous amide analogues of diaminopimelic acid (DAP) as inhibitors of enzymes involved in bacterial lysine biosynthesis. Org. Lett. 2, 3857–3860. Auger, G., van Heijenoort, J., Vederas, J.C., and Blanot, D. (1996). Effect of analogues of diaminopimelic acid on the mesodiaminopimelate-adding enzyme from Escherichia coli. FEBS Lett. 391, 171–174. Scapin, G., Cirilli, M., Reddy, S.G., Gao, Y., Vederas, J.C., and Blanchard, J.S. (1998). Substrate and inhibitor binding sites in Corynebacterium glutamicum diaminopimelate dehydrogenase. Biochemistry 37, 3278–3285. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. In Methods in Enzymology, Volume 276, C.W. Carter, Jr., and R.M. Sweet, eds. (San Diego, CA: Academic Press), pp. 307–326. Weeks, C., and Miller, R. (1999). The design and implementation of SnBv2.0. J. Appl. Crystallogr. 32, 120–124. de La Fortelle, E., and Bricogne, G. (1997). Maximum-likelihood heavy atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. In Methods in Enzymology, Volume 276, C.W. Carter, Jr., and R.M. Sweet, eds. (San Diego, CA: Academic Press), pp. 472–494. Kleywegt, G.J., and Read, R.J. (1997). Not your average density. Structure 5, 1557–1569. Dodson, E.J., Winn, M., and Ralph, A. (1997). Collaborative computational project number 8: providing programs for protein crystallography. Methods Enzymol. 277, 620–633. Bru¨nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography and NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921. Navaza, J. (1994). AMoRe: an automated package for molecular replacement. Acta Crystallogr. A 50, 157–163. Laskowski, R.A., MacArthur, M.W., Moss, D.S., and Thornton, J.M. (1993). PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291. Pankuch, G.A., Jacobs, M.R., and Appelbaum, P.C. (1995). Comparative activity of ampicillin, amoxycillin, amoxycillin/clavulanate and cefotaxime against 189 penicillin-susceptible and resistant pneumococci. J. Antimicrob. Chemother. 35, 883–888. Nicholls, A., Sharp, K., and Honig, B. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins 11, 281–296.