Crystal Structure of Lipoate-Protein Ligase A from Escherichia coli DETERMINATION OF THE LIPOIC ACID-BINDING SITE*

Lipoate-protein ligase A (LplA) catalyzes the formation of lipoyl-AMP from lipoate and ATP and then transfers the lipoyl moiety to a specific lysine residue on the acyltransferase subunit of alpha-ketoacid dehydrogenase complexes and on H-protein of the glycine cleavage system. The lypoyllysine arm plays a pivotal role in the complexes by shuttling the reaction intermediate and reducing equivalents between the active sites of the components of the complexes. We have determined the X-ray crystal structures of Escherichia coli LplA alone and in a complex with lipoic acid at 2.4 and 2.9 angstroms resolution, respectively. The structure of LplA consists of a large N-terminal domain and a small C-terminal domain. The structure identifies the substrate binding pocket at the interface between the two domains. Lipoic acid is bound in a hydrophobic cavity in the N-terminal domain through hydrophobic interactions and a weak hydrogen bond between carboxyl group of lipoic acid and the Ser-72 or Arg-140 residue of LplA. No large conformational change was observed in the main chain structure upon the binding of lipoic acid.

Lipoic acid is a prosthetic group of acyltransferase (E2) subunit of the pyruvate, ␣-ketoglutarate, and branched-chain ␣-ketoacid dehydrogenase complexes and of H-protein of the glycine cleavage system (1)(2)(3)(4). It attaches to a specific lysine residue on the proteins via an amide linkage between the ⑀-amino group of the lysine residue and the carboxyl group of lipoic acid. In the reaction sequence of the complexes, the lypoyllysine arm shuttles the reaction intermediates and reducing equivalents between the active sites of the components of the complexes.
The attachment of lipoic acid to the proteins occurs by two-step reactions in which a lipoyl-AMP intermediate is formed from lipoic acid and ATP, and pyrophosphate is released in the initial activation reaction (Reaction 1).
Lipoate ϩ ATP 3 lipoyl-AMP ϩ PP i REACTION 1 The lipoyl moiety of the intermediate is then transferred to apoproteins in the second transfer reaction, yielding the lipoylated protein and AMP (Reaction 2).
Lipoyl-AMP ϩ apoprotein 3 lipoylated protein ϩ AMP REACTION 2 Lipoate-protein ligase A (LplA) 2 from Escherichia coli catalyzes the above two reactions. LplA activates and transfers not only naturally occurring R-(ϩ)-lipoate but also S-(Ϫ)-lipoate, lipoate analogues, and octanoate to the apoproteins, although S-lipoate and octanoate are used at a 44 and 16% rate of R-lipoate, respectively (5,6). LplA has a molecular mass of 37,795 Da, consisting of 337 amino acids excluding the initiating methionine residue, which is cleaved off during the biosynthesis (5). Strains with lplA null mutations have severe defects in the incorporation of exogenously supplied lipoic acid and lipoic acid analogues into apoproteins (5,7).
In E. coli, there is another enzyme, LipB, responsible for the covalent attachment of lipoic acid. LipB transfers lipoic acid/octanoic acid endogenously synthesized on the acyl carrier protein by the function of LipA to the lipoate-dependent enzymes (7)(8)(9). LipB consists of 213 amino acids, whose amino acid sequence shares only 12.7% identity with that of LplA ( Fig. 1). On the other hand, the amino acid sequence of LplA shows 31 and 35% identity with those of human and bovine lipoyltransferase, the mammalian LplA homologues, respectively (Fig. 1). However, the mammalian lipoyltransferases have no ability to activate lipoate to lipoyl-AMP and catalyze only the second half-reaction (10,11). Strong homology in the N-terminal half region suggests that the region may be responsible for the second transfer reaction, whereas some residue(s) in the C-terminal region of the E. coli enzyme may be essential for the lipoate-activating reaction, which are lacking in the mammalian sequences (10).
To elucidate the structure-function relationship, we determined the crystal structure of LplA. We present here the structures of two forms of LplA: native and lipoate complex forms. LplA consists of two domains, the N-terminal and the C-terminal domains with a substrate binding pocket at the interface between these two domains. Lipoic acid is bound on the N-terminal domain side in the pocket through hydrophobic interactions. This is the first report on the crystal structure of the enzymes responsible for the lipoic acid metabolism.
Mutation of LplA-Mutations of S72A and R140A were introduced using an In-Fusion PCR Cloning kit (BD Biosciences Clontech). For the S72A mutation, PCRs were carried out employing pET-LPLA as a template and primer 1 (5Ј-AAGGAGATATACATATGTCCACAT-TACGC-3Ј (pET-3a sequence (underlined) and then nucleotides 83-97 (5))) and primer 2 (5Ј-GCCACCGGCACTGCGCCGCGCCAG-3Ј (nucleotides 307-278; modified bases are shown in boldface letters)) to amplify the N-terminal half and primer 3 (5Ј-CGCAGTGCCGGTG-GCGGCGCGGT-3Ј (nucleotides 293-326)) and primer 4 (5Ј-GTTAG-CAGCCGGATCCTAC-3Ј (pET-3a sequence (underlined) and nucleotides 1099 -1096)) to amplify the C-terminal half. For the R140A mutation, primer 1 and primer 5 (5Ј-GGTTTCGGCATAGGC-CGAGCCTGAGA-3Ј (nucleotides 511-486)) and primer 6 (5Ј-GC-CTATGCCGAAACCAAAGATCGCGG-3Ј (nucleotides 497-522)) and primer 4 were employed to amplify the N-terminal and C-terminal half, respectively. After the recombination reaction with pET-3a digested with NdeI and BamHI and the pair of the PCR products, the resultant plasmid was transformed into E. coli BL21(DE3)pLysS. The mutant LplAs were expressed and purified up to the DEAE-Sepharose The residues conserved among four proteins, those between E. coli LplA and S. pneumonia LplA or human LT, and LipB are shown on a blue, yellow, and green background, respectively. The secondary structure elements of E. coli the LplA are assigned using the program DSSP (37). ␣-Helices, 3 10 -helices, and ␤-strands are represented with magenta rectangles, orange rectangles, and blue arrows, respectively. Amino acid numbering of E. coli LplA starts at the serine residue following the initiation methionine (5), that of S. pneumoniae LplA starts at the initiation methionine residue (30), that of human LT starts at the N-terminal amino acid of the mature proteins (11), and that of LipB starts at the methionine residue added to the N-terminal extended active LipB (9). The sequence alignment was carried out using ClustalW (38). step as described above. The purified wild-type and mutant LplAs in the DEAE-Sepharose fractions were apparently homogeneous on SDSpolyacrylamide gel electrophoresis and used for the steady state kinetic analysis.
Assay of LplA-The activity of LplA was determined in the reaction mixture of 50 l containing 40 mM potassium phosphate buffer, pH 7. 0, 0.3 mg/ml bovine serum albumin, 0. 5 mM DTT, 3 M E. coli apoHprotein as a lipoate acceptor, 30 M R-(ϩ)-lipoic acid, 2 mM ATP, 2 mM MgCl 2 , and appropriate amounts of wild-type or mutant LplA. After preincubation for 90 s, the reaction was started by the addition of ATP or lipoic acid and carried out for 5 min at 37°C. The reaction was terminated by heating the mixture at 95°C for 70 s. The amount of lipoylated H-protein in the mixture was determined by the glycine-CO 2 exchange activity using E. coli P-protein as described previously (4). Apparent K m values of wild-type and mutant LplA for lipoic acid, ATP, and apoH-protein were determined by varying the concentration of one substrate with a constant concentration of the other two substrates. The constant concentrations are the same as those in the standard assay mixture described above. The concentrations of varied substrates were 1.25-80 M for lipoic acid, 2.5-1000 M for ATP, and 0.188 -4.5 M for apoH-protein. R-(ϩ)-lipoate was generously provided by ASTA-Medica (Frankfurt am Main, Germany). The recombinant E. coli apoH-protein was prepared as described for the preparation of holoH-protein except that E. coli cells were grown in the medium without an addition of lipoic acid (15).
Crystallization-Crystallization was carried out at 20°C by the hanging drop-vapor diffusion method using a protein solution dialyzed against 5 mM HEPES-Na, pH 7.5, and 3 mM DTT. Se-LplA was crystallized in a droplet containing a 1:1 ratio of protein solution at 3 mg/ml and reservoir solutions of 12.5-14.5% ethylene glycol and 21% glycerol. Crystals of Se-LplA⅐lipoate complex was grown with a modified protein solution at 5 mg/ml and a reservoir solution containing 3 mM R-(ϩ)lipoate.
Data Collection-All X-ray data were collected at 100 K on the beamline BL44XU at SPring-8 (Hyogo, Japan). The structure of Se-LplA was solved using Se-SAD methods (16). Before data collection, wavelength was calibrated by fluorescence scan of the Se-LplA crystal, and a wavelength was selected for data collection corresponding to the white line (peak, ϭ 0.9796 Å). Data were collected in a 180°sweep in 1°oscillation steps, using an imaging plate detector, and then integrated and scaled with the programs MOSFLM (17) and SCALA (18). The data collection for crystal of Se-LplA⅐lipoate complex was carried out with a wavelength of 0.9215 Å sweeping 90°in 1°oscillation steps. The crystal containing lipoate was very weak against radiation damage; therefore, two data sets had to be merged to get enough data for the structure determination. Each data set was indexed and integrated with DENZO and then scaled and merged with SCALEPACK (19). The LplA crystals belong to space group C222 1 with unit cell dimensions of a ϭ 81.6 Å, b ϭ 112.1 Å, and c ϭ 289.2 Å for Se-LplA crystal used for structure determination.
Structure Determination-All of the 15 expected selenium sites (for three monomers containing 5 Se-Met residues/asymmetric unit) were found using the heavy atom search program in CNS (20), and the initial phases were calculated with SHARP (21). 2.8 Å resolution maps calculated after density modification by SOLOMON (22) implemented in SHARP was used for determination of molecule mask and noncrystallographic symmetry operators for further density improvement. Solvent flattening, histogram mapping, and noncrystallographic symmetry averaging were applied for phase improvement using DM (23), and the resulting map was used for model building. The structure of Se-LplA⅐lipoate was solved by the molecular replacement method with MOLREP (24) in the CCP4 suite (25) using Se-LplA structure as a search model.
Model Building and Refinement-Model building and refinement for all structures were performed using the programs O (26) and  in CCP4 suite (25), respectively. Refinement for Se-LplA was monitored with the R free value calculated for a randomly selected 5% of reflections in the data set. The R free values of Se-LplA⅐lipoate were calculated with the reflections, which were used for calculation of R free values of Se-LplA. To build an Se-LplA structure, the quality of the initial experimental electron density map and known selenium sites allowed unambiguous tracing for the protein backbone and side chains. The protein model of Se-LplA was refined against the 43.6 -2.4 Å resolution data set. The geometry library of lipoic acid was made using a monomer library sketcher in CCP4 suite (25). The Ramachandran plot was defined by PROCHECK (28).

RESULTS AND DISCUSSION
Structure Determination-Crystallization of E. coli LplA was mentioned about 10 years ago (6). The structure had never been determined, presumably because of the difficulty to obtain crystals suitable for the structural analysis. Hexagonal shape crystals of Se-LplA (0. 06 ϫ 0. 08 ϫ 0. 1 mm) without substrate were obtained in 2 weeks by the hanging drop-vapor diffusion method, employing ethylene glycol as a precipi- tant. The space group of the crystal was determined to be C222 1 . The phase problem for LplA was solved by the SAD method, and the initial model could be built into the electron density. The final structure was refined using a 2.  (29). Although the crystal structure of LplA shows dimer formation in a crystal, the gel filtration experiment shows that LplA is a monomer in solution. The dimer structure seems to be the result by the crystal packing.
The structure of LplA is similar to that of the putative lipoate-protein ligase from S. pneumoniae (Protein Data Bank accession code 1VQZ), which gives a Z-score of 29.8 for an alignment of 303/326 residue with an r.m.s. deviation of 3. 1 Å and 31% sequence identity by the DALI program (31). The high structural similarity confirmed by superposition of the ␣-carbon traces of the putative protein and LplA suggests that the putative protein is a homologue of the E. coli lipoate-protein ligase A.
The Lipoic Acid Binding Site-Because lipoic acid is a labile compound, LplA crystals complexed with R-(ϩ)-lipoic acid were obtained by seeding small crystals of the Se-LplA⅐lipoate complex into a droplet to accelerate the crystallization. Crystals suitable for analysis were obtained within 3 days after setting the droplet. The lipoic acid-bound form was determined at 66.7-2.9 Å resolution. Weak but significant electron density corresponding to lipoic acid was found in MolA and MolB (Fig. 3, A and B). Positions of the lipoic acid in MolA and MolB are slightly different. This may be caused by weak interactions between LplA and lipoic acid or insufficient information from low resolution data. Lipoic acid is located in a hydrophobic core generated by Leu-17, Phe-24, Phe-147, Ala-138, and aliphatic parts of Glu-21 and Ser-72 side chains. Hydrophobic interactions are formed between these residues and the dithiolane ring and the hydrophobic tail of lipoic acid. Leu-17, Ser-72, and Ala-138 are well conserved among LplAs and lipoyltrans-ferase (Fig. 1). The electron density for the carboxyl group is weak relative to that for the hydrophobic moiety. In MolA, the carboxyl group makes a hydrogen bond with the side chain of Ser-72, whereas in MolB, the carboxyl group hydrogen bonds to the side chain of Arg-140 (Fig. 3,  C and D). Although the actual binding mode may change somewhat when higher resolution x-ray data are eventually obtained, this complex provides insights into lipoic acid binding to the LplA molecule. Because van der Waals interactions are neither strong nor specific, and a weak hydrogen bond occurs only at the carboxyl group, it may permit LplA to bind not only R-(ϩ)-lipoic acid but also S-(Ϫ)-lipoic acid, lipoic acid analogues, and octanoic acid. The binding mode well explains why LplA activates these carboxylic acids and transfers them to apoproteins (6,7,32). The reaction rates may be influenced by the slight differences in the van der Waals interactions between LplA and these substrates.
Functional and Structural Similarities to Other Proteins-A systematic analysis using the DALI program (31) identified a number of proteins/domains with significant structural similarities to LplA. The closest of them other than the LplA homologue is BirA, biotin holoenzyme synthetase/bio repressor from E. coli (Protein Data Bank accession codes 1BIA and 1BIB) (33), which gives a Z-score of 9.4 for an alignment of 139/292 residues with an r.m.s. deviation of 3.5 Å and 12% sequence identity. BirA consists of three domains and functions as a repressor of the biotin biosynthetic operon and as an enzyme that catalyzes the activation and transfer of biotin to a specific lysine residue on the biotindependent carboxylases. The reaction mechanism is remarkably similar to that of LplA. The central domain contributes to the catalysis. Fig. 4 shows a superposition of LplA (residues 1-337) onto BirA (the central and C-terminal domains, residues 68 -317) according to the rotation and translation matrices provided by DALI. Several ␣-helices (␣ 1 , ␣ 4 , and ␣ 5 ) and ␤-strands (␤ 1 , ␤ 2 , ␤ 6 , ␤ 7 , ␤ 9 , and ␤ 10 ) of LplA are well superimposed on the corresponding secondary structure elements of BirA. In both enzymes, lipoic acid and biotin interact with ␤-strands in the N-terminal domain and the central domain, respectively, and the substrates are situated in the similar position in the respective enzymes. Functional and structural similarities of these enzymes suggest that evolutionarily they are closely related, although they share only 12% amino acid sequence identity. The absence of a DNA binding domain in the LplA structure may be related to the finding that LplA metabolizes lipoic acid that is exogenously supplied rather than that synthesized endogenously in E. coli (5,7).
Mechanistic Implication-Ser-72 and Arg-140 residues, respectively, were substituted by Ala to analyze the contribution of the residues to the lipoic acid binding (Fig. 3). Steady state kinetic studies were carried out by varying the concentration of one substrate and keeping the concentration of the other two substrates constant. Apparent kinetic constants are shown in TABLE TWO.
Whereas the S72A mutation results in an about 2-fold increase in K m(app) for lipoic acid, K m(app) for ATP is greatly increased by the mutation, indicating the reduced binding affinity for ATP. This result suggests that the hydrogen bond between lipoic acid and Ser-72 may not be so critical to fix lipoic acid at the position determined by the crystal structural analysis. However, the fixation of the carboxyl group to Ser-72 is required for the initiation of the lipoate activation reaction with ATP, or ATP may interact directly with Ser-72.
The R140A mutation results in reduced V max(app) values, although the values differ depending on the varied substrate because of the usage of the limited concentration of apoH-protein. K m(app) for lipoic acid is rather decreased. On the other hand, K m(app) for apoH-protein increases by 1 order of magnitude, indicating a reduction of the binding affinity for apoH-protein. In the protein lipoylation reaction, the presence of the Glu residue at the three residues N-terminal side from the Lys residue to be lipoylated in the lipoate acceptor protein is essential (34,35). Therefore, Arg-140 in LplA may make a salt bridge with the Glu residue in the apoH-protein in the second transfer reaction (Reaction 2). These results of the kinetic analyses support the lipoic acid binding mode shown in Fig. 3 (i.e. lipoic acid is bound at the position by the hydrophobic interactions, and the weak hydrogen bond between the carboxyl group and the Ser-72 or the Arg-140 residue is enough to accommodate lipoic acid).
We attempted to obtain a crystal of the LplA⅐ATP complex by soaking the native crystal with Mg-ATP and by co-crystallization in the presence of Mg-ATP. However, X-ray analysis of the crystals did not show electron density for ATP clear enough to determine the ATP binding site. The first step of the reaction catalyzed by LplA is thought to   be the nucleophilic attack of the carboxyl oxygen of lipoic acid on the ␣-phosphorus atom of ATP followed by the release of the pyrophosphate and the formation of a lipoyl-AMP intermediate by analogy to the reaction catalyzed by histidyl-tRNA synthetase (36). The activated carbon atom of the lipoyl-AMP intermediate is then attacked by the nonprotonated ⑀-amino group of the lysine residue of an apoprotein to be lipoylated, producing AMP and the lipoylated protein. To initiate the first reaction, ATP should be anchored near the lipoic acid in LplA, and especially the distance between the ␣-phosphorus atom of ATP and the carboxyl oxygen atom of lipoic acid should be less than 3 Å. LplA has a characteristic sequence motif RRSSGGG between positions 69 and 75, which is highly conserved among the proteins listed in Fig. 1. The motif is situated on the loop following the ␤ 4 -strand (Fig. 2C) close to the bound lipoic acid. In the native LplA structure, the side chain of Arg-70 forms a hydrogen bond network with the main-chain carbonyl groups of Gly-73 and Gly-75 and the side chain of Ser-72, stabilizing the loop structure (Fig. 5). The motif can provide positive charges and a space large enough to accommodate ATP. Therefore, the motif is presumed to be involved in the binding of ATP or the adenylate moiety of the lipoyl-AMP intermediate. Indeed, the mutation of Ser-72 affects the binding affinity for ATP (TABLE TWO). An R72G point mutation in the human lipoyltransferase has been found in a hyperglycinemia patient, but no mutation has been found in the four proteins, P-, T-, L-, and H-protein, constituting the human glycine cleavage system. 3 The Arg-72 in the human lipoyltransferase corresponds to Arg-69 in E. coli LplA. The mutation may cause the reduction of the lipoylation level of H-protein of the glycine cleavage system, resulting in a decrease in the activity of the glycine cleavage system. The expression of the recombinant R72G mutant in E. coli failed presumably because of the instability of the mutant protein. 4 These findings also suggest that the motif plays an important role for the binding of the substrate or the stabilization of the three-dimensional structure.
Conclusions-The present study presents the LplA structure and its substrate binding site for the first time. LplA consists of a large N-terminal domain and a small C-terminal domain, comprising a substrate binding pocket at the interface of the two domains. Lipoic acid is bound to the N-terminal domain via hydrophobic interactions and a weak hydrogen bond to Ser-72 or Arg-140 in the pocket. The result well explains the reason why LplA utilizes lipoate analogues and octanoic acid in addition to the natural substrate, R-(ϩ)-lipoate. The large solvent-exposed active site cleft is favorable to allow docking of the lipoate acceptor protein to initiate the second transfer reaction. In an effort to more fully define the active site geometry and to understand the reaction mechanism of LplA, co-crystallization studies of LplA with ATP and lipoyl-AMP are in progress.