Crystal Structure of Lipoate-Protein Ligase A Bound with the Activated Intermediate

Lipoic acid is the covalently attached cofactor of several multi-component enzyme complexes that catalyze key metabolic reactions. Attachment of lipoic acid to the lipoyl-dependent enzymes is catalyzed by lipoate-protein ligases (LPLs). In Escherichia coli, two distinct enzymes lipoate-protein ligase A (LplA) and lipB-encoded lipoyltransferase (LipB) catalyze independent pathways for lipoylation of the target proteins. The reaction catalyzed by LplA occurs in two steps. First, LplA activates exogenously supplied lipoic acid at the expense of ATP to lipoyl-AMP. Next, it transfers the enzyme-bound lipoyl-AMP to the ϵ-amino group of a specific lysine residue of the lipoyl domain to give an amide linkage. To gain insight into the mechanism of action by LplA, we have determined the crystal structure of Thermoplasma acidophilum LplA in three forms: (i) the apo form; (ii) the ATP complex; and (iii) the lipoyl-AMP complex. The overall fold of LplA bears some resemblance to that of the biotinyl protein ligase module of the E. coli biotin holoenzyme synthetase/bio repressor (BirA). Lipoyl-AMP is bound deeply in the bifurcated pocket of LplA and adopts a U-shaped conformation. Only the phosphate group and part of the ribose sugar of lipoyl-AMP are accessible from the bulk solvent through a tunnel-like passage, whereas the rest of the activated intermediate is completely buried inside the active site pocket. This first view of the activated intermediate bound to LplA allowed us to propose a model of the complexes between Ta LplA and lipoyl domains, thus shedding light on the target protein/lysine residue specificity of LplA.

hexahistidine-containing tag at its amino terminus was overexpressed in E. coli Rosetta(DE3) cells using Terrific broth culture medium. Protein expression was induced by 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside, and the cells were incubated for an additional 6 h at 37°C following growth to mid-log phase at 37°C. The cells were lysed by sonication in lysis buffer (50 mM Tris-HCl at pH 7.9, 500 mM NaCl, and 10% (v/v) glycerol) containing 50 mM imidazole. The crude lysate was centrifuged at ϳ36,000 ϫ g for 60 min. The supernatant was applied to an affinity chromatography column of nickel-nitrilotriacetic acid-agarose (Qiagen). The protein was eluted with lysis buffer containing 500 mM imidazole, and the eluted sample was diluted 5-fold with buffer A (50 mM Tris-HCl at pH 7.2, 5% (v/v) glycerol, and 10 mM ␤-mercaptoethanol). The diluted sample was applied to a Source 15Q ion-exchange column (Amersham Biosciences), which was previously equilibrated with buffer A. The protein was eluted with a linear gradient of 0 -1.0 M NaCl in buffer A. The next step was gel filtration on a HiLoad 16/60 Superdex-200 prep-grade column (Amersham Biosciences), employing an elution buffer of 20 mM Tris-HCl at pH 7.2 and 100 mM NaCl. For overexpressing the selenomethionine (SeMet)-substituted protein in E. coli Rosetta(DE3) cells, we used the M9 cell culture medium that contained extra amino acids, including SeMet. When the cell culture reached an A 600 of 0.6, SeMet was added at 50 mg/liter, and at the same time the synthesis of Met was repressed by the addition of Phe, Thr, and Lys at 100 mg/liter, and Leu, Ile, Val, and Pro at 50 mg/liter (9). After 15 min, expression of the SeMet-substituted protein was induced with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside, and the culture was grown for an additional 20 h at 20°C. The procedure for purifying the SeMetsubstituted protein was the same except for the presence of 10 mM dithiothreitol in all buffers used during purification steps besides buffer A.
Crystallization-Crystals were grown by the hanging-drop vapor diffusion method at 24°C by mixing equal volumes (2 l each) of the protein solution (at 18 mg ml Ϫ1 concentration in 20 mM Tris-HCl at pH 7.2 and 100 mM NaCl) and the reservoir solution. To grow crystal of the native protein in the apo form, we used a reservoir solution consisting of 100 mM sodium citrate at pH 5.5, 20% (v/v) iso-propanol, 20% (w/v) polyethylene glycol 3350, and 100 mM NaCl. Rod-shaped crystals grew to approximate dimensions of 0.1 ϫ 0.1 ϫ 0.3 mm within a few days. When we added ATP, lipoic acid, or octanoic acid to the protein solution, the enzyme aggregated rapidly and co-crystallization was not possible. Therefore, we used the soaking method to obtain crystals of the native protein bound with the ligands. To obtain crystals of the ATP complex, we soaked the apo crystals of the native protein in the reservoir solution containing ATP in 50-fold molar excess for 1 day. To obtain crystals of the lipoyl-AMP complex, we transferred the ATP-soaked crystals into the reservoir solution containing lipoic acid in 50-fold molar excess and soaked them for 1 day. Apparently the reaction took place in the crystal, as evidenced by the clear electron density for the activated intermediate, lipoyl-AMP (Fig. 1A, left). When we transferred the ATP-soaked crystals into the biotin-containing reservoir solution, the reaction did not occur in the crystal, as evidenced by the lack of electron density for biotinyl-AMP. The difference Fourier map showed only the electron density of ATP. When we soaked the native crystals in the reservoir solution containing lipoic acid or octanoic acid alone in 50-fold molar excess for 1 day, the crystal quality deteriorated and we could not collect x-ray diffraction data to sufficiently high resolution. The SeMet-substituted protein in the apo form was crystallized under conditions identical to those for the apo crystals of the native protein except for the presence of 10 mM dithiothreitol in the protein solution.
X-ray Data Collection and Structure Determination-A crystal of the SeMet-substituted protein was dipped into a cryoprotectant solution for a few seconds and was frozen in the cold nitrogen gas stream at 100 K. The cryoprotectant solution consisted of 20% (v/v) glycerol added to the reservoir solution. X-ray diffraction data was collected at 100 K on a Quantum 4R charge-coupled device detector (Area Detector Systems Corp., Poway, CA) at the BL-6A experimental station of Photon Factory, Japan. For each image, the crystal was rotated by 1°, and the crystal-todetector distance was set to 210 mm. The raw data were processed and scaled using the program suit HKL2000 (10). The SeMet-substituted crystal belongs to the space group C2, with unit cell parameters of a ϭ 109.35 Å, b ϭ 62.95 Å, c ϭ 46.86 Å, and ␤ ϭ 111.57°. There is one monomeric LplA molecule per asymmetric unit, giving a solvent fraction of 47.4%. TABLE ONE summarizes the statistics of MAD data collection. Thirteen of the fourteen expected selenium atoms in each monomer of the recombinant enzyme, except the SeMet at the amino terminus of the polypeptide chain, were located with the program SOLVE (11), and the selenium sites were used to calculate the phases with RESOLVE (12). Phasing statistics are summarized in TABLE ONE. Other x-ray diffraction data were collected essentially as above.
Model Building and Refinement-Excellent quality of the MADphased electron density map allowed automatic model building by the program RESOLVE (12), giving an initial model that accounted for ϳ80% of the backbone of the polypeptide chain with much of the sequence assigned. Subsequent manual model building was done using the program O (13). The model was refined with the program CNS (14), including the bulk solvent correction. 10% of the data were randomly set aside as the test data for the calculation of R free (15). Several rounds of model building, simulated annealing, positional refinement, and individual B-factor refinement were performed. Subsequently, this model of the apo enzyme was used to refine the ATP-bound form and the lipoyl-AMP-bound form. Refinement statistics are summarized in TABLE ONE. All of the models have excellent stereochemistry (TABLE ONE), as evaluated by the program PROCHECK (16).

RESULTS AND DISCUSSION
Model Quality and Overall Structure-We have determined the crystal structure of Ta LplA using SeMet MAD data at 2.04-Å resolution (TABLE ONE). The structure of the SeMet-substituted LplA in the ligand-free state was refined against 20 -2.04 Å data to crystallographic R work and R free values of 21.6% and 23.7%, respectively. The refined model (PDB code 2ARS) accounts for 245 residues (1-178 and 191-257). The twenty-residue amino-terminal fusion tag and the carboxylterminal five residues as well as a loop region (residues 179 -190) have no electron density. We have confirmed by SDS-PAGE that the polypeptide chains in the crystal are not cleaved. This indicates that the invisible loop is probably flexible and disordered in the crystal.
We have additionally determined the crystal structures of the native protein bound with either lipoyl-AMP or ATP. The structure of the lipoyl-AMP complex was refined against 20 -2.4 Å data to R work and R free values of 19.3% and 24.9%, respectively. The refined model (PDB code 2ART) accounts for 247 amino acid residues (1-180 and 191-257). The electron density of the bound lipoyl-AMP group clearly indicates a covalent linkage between lipoic acid and 5Ј-AMP (Fig. 1A, left). We could obtain the lipoyl-AMP intermediate complex simply by soaking the crystals serially in solutions containing each of ATP and lipoic acid, indicating that the enzyme molecules in the crystal catalyzed the activation step. The structure of the ATP complex was refined against 20 -2.5 Å data to R work and R free values of 20.4% and 25.9%, respectively. The refined model (PDB code 2ARU) accounts for 247 amino acid res-idues (1-180 and 191-257). The electron density of the bound ATP was a little weaker than that of lipoyl-AMP (Fig. 1A, right), and we arbitrarily fixed the occupancy of ATP at 0.5. We suspect that the binding affinity of Ta LplA for ATP is weaker than that for lipoyl-AMP, and incomplete occupation of the ATP binding site resulted from a partial loss of the bound ATP during quick soaking of the crystals in the cryoprotectant solution.
Ta LplA exists as a monomer in solution according to our dynamic light scattering analysis. The LplA monomer is oblate-shaped, with approximate dimensions of 40 Å ϫ 40 Å ϫ 32 Å (Fig. 1B). The central core of Ta LplA comprises two ␤-sheets: a larger eight-stranded ␤-sheet and a smaller three-stranded ␤-sheet. The larger sheet consists of the strands ␤1 (residues 2-6), ␤2 Structural Comparison-DALI structural similarity searches (17) with the apo structure of Ta LplA revealed two close relatives. The highest Z-score is obtained with a putative LplA from Streptococcus pneumoniae, a 329-residue protein (unpublished deposition; PDB code 1VQZ; a root mean square (r.m.s.) deviation of 2.5 Å for 218 equivalent C␣ positions, a Z-score of 25.6, and a sequence identity of 29%). This structure was solved without any bound ligand in the active site. S. pneumoniae LplA is longer than Ta LplA in its C terminus by 67 residues. A search of the protein sequence data base indicates that LplAs fall into two size categories; the shorter ones have ϳ260 -270 residues and the longer ones ϳ330 -340 residues. Ta LplA in the crystal catalyzed the Ϫ7. 19 is the intensity of reflection h, ⌺ h is the sum over all reflections, and ⌺ i is the sum over i measurements of reflection h. Numbers in parentheses reflect statistics for the last shells (2.11-2.04 Å or 2.05-1.98 Å). d R iso ϭ ⌺ ͉ ͉F PH ͉ Ϫ ͉F P ͉ ͉/ ⌺ ͉F P ͉, where F PH and F P are the derivative (2 or 3) and native (1) structure factors, respectively. Numbers in parentheses are for the last shells (2.11-2.04 Å or 2.05-1.98 Å). e Figure of merit ϭ ͗ ͉ ⌺P(␣)e i␣ /⌺P(␣) ͉ ͘, where ␣ is the phase angle and P(␣) is the phase probability distribution. f Numbers in parentheses refer to the last shells (2.11-2.04 Å, 2.49 -2.40 Å, and 2.59 -2.50 Å, respectively). g R ϭ ⌺ ͉ ͉F obs ͉ Ϫ ͉F calc ͉ ͉/⌺ ͉F obs ͉, where R free is calculated for a randomly chosen 10% of reflections, which were not used for structure refinement, and R work is calculated for the remaining reflections. Low resolution limit for the refinement is 20 Å for all three models.
formation of the activated intermediate (lipoyl-AMP) from ATP and lipoic acid. This suggests that the extra C-terminal region of the longer LplAs may not be required for catalyzing the activation step. The next highest Z-score is obtained with the E. coli biotin holoenzyme synthetase/bio repressor (BirA) (PDB code 1BIA; an r.m.s. deviation of 2.7 Å for 134 equivalent C␣ positions, a Z-score of 9.7, and a sequence identity of 13%) (18). The structure of BirA consists of three domains. The central domain II represents the BPL module, and houses both the biotin and the ATP binding sites. In E. coli BirA, the binding site of biotin was located using biotinyl-lysine, but there was no direct crystallographic data for the ATP binding site. A region of sequence homology to the BPL module of E. coli BirA can be identified in all biotinylating enzymes, whereas other domains are retained only in some bacterial counterparts of the E. coli enzyme (3). The equivalent of the smaller ␤-sheet of Ta LplA is missing from the domain II of E. coli BirA. Thus the ␤-sheet of the domain II of E. coli BirA is largely solventexposed on one side, where biotin was observed to bind (18). In comparison, in Ta LplA, the additional smaller ␤-sheet covers the lipoyl portion of the bound lipoyl-AMP (Fig. 1B). A careful sequence analysis showed that LplA enzymes as well as LipB enzymes bear very low but detectable homology to the BPL module of biotinylating enzymes (3), suggesting an evolutionary relationship among them.
Binding of Lipoyl-AMP and ATP at the Active Site-After submission of this report, the crystal structure of LplA from E. coli was reported (19) (PDB codes 1X2G for the apo enzyme and 1X2H for the lipoic acidcomplexed enzyme). This structure as well as those of E. coli BirA (18) and S. pneumoniae LplA does not provide detailed information about the binding modes of either the activated intermediate or ATP within the active site. To gain further insight into the ligand binding by LplA, we have determined the structures of Ta LplA in complex with either ATP or lipoyl-AMP, in addition to the apo structure. The apo, the ATPbound, and the lipoyl-AMP-bound structures of Ta LplA are highly similar to each other. The r.m.s. deviation between the apo and the ATP-bound structures is 0.23 Å for 245 C␣ atoms (residues 1-178 and 191-257), whereas that between the apo and the lipoyl-AMP-bound structures is 0.39 Å for 245 C␣ atoms (residues 1-178 and 191-257). Between the ATP-bound and the lipoyl-AMP-bound structures, the r.m.s. deviation is 0.30 Å for 247 C␣ atoms (residues 1-180 and 191-257).
The lipoyl-AMP binding pocket is bifurcated, and the conformation of the bound lipoyl-AMP is U-shaped ( Fig. 2A). The phosphate group and part of the ribose sugar are accessible from the bulk solvent; they are accessible through an ϳ10 Å-deep, tunnel-like entrance that accommodates the lysine side chain protruding from the lipoyl domain ( Fig. 2A). The rest of the activated intermediate is buried completely inside the bifurcated pocket, with the dithiolane ring of lipoyl-AMP being located roughly in the center of Ta LplA (Fig. 1B). The dithiolane ring and the aliphatic chain of lipoyl-AMP are surrounded by aliphatic side chains of four hydrophobic residues (Leu 18 , Ile 46 , Val 79 , and Ala 163 ), the side chains of two histidine residues (His 81 and His 161 ), and the aliphatic part of the side chain of an arginine residue (Arg 72 ). Four conserved glycine residues (Gly 48 , Gly 75 , Gly 76 , and Gly 77 ) also surround them (supplemental Fig. S1). The imidazole ring of His 161 rotates about 90°and moves ϳ1 Å away from the dithiolane ring of lipoyl-AMP in the lipoyl-AMP complex structure, relative to the apo structure. When we tried to incorporate biotin into the ATP-soaked crystals, the enzymes in the crystal did not catalyze the reaction and biotinyl-AMP was not formed. The present structure shows that the terminal double rings of biotin are too big to fit into the lipoate binding pocket of LplA. They are also more polar than the dithiolane ring of lipoic acid and are probably not readily accommodated in the lipoic acid binding pocket whose surface is lined largely with hydrophobic residues.
The phosphate moiety of lipoyl-AMP interacts with both Lys 135 and Lys 145 (Fig. 2, B and C). The distance from the O2A oxygen atom of lipoyl-AMP to the N atoms of Lys 135 is 2.60 Å. Lys 145 corresponds to the strictly conserved lysine residue among the members of LplA, LipB, and BPL families (3). The distance from the N atom of Lys 145 to the OAB carboxyl oxygen atom of the lipoyl part is 3.29 Å, whereas those to the O4* and O5* atoms of the ribose, and the OAF atom of the ␣-phosphate are 3.02 Å, 3.31 Å, and 3.10 Å, respectively (Fig. 2, B and C). The side chain of Lys 145 also makes a hydrogen bond with the side chain of Asp 138 (2.77 Å between N of Lys 145 and O␦1 of Asp 138 ) (Fig. 2, B and C). Asp 138 of Ta LplA is part of a somewhat variable region (Fig. 3). However, its role could possibly be played by an equivalent negatively charged residue, because there is at least one Asp or Glu near this region of other LplAs. Similar interactions were observed in E. coli BirA (18), where the conserved Lys 183 makes hydrogen bonds with the carboxyl group of biotin as well as with the strictly conserved Asp 176 . In Ta LplA, the side chain O␦1 atom of Asp 138 interacts with the N atoms of both Lys 135 (2.92 Å) and Lys 145 (3.03 Å). A water molecule (Wat 7 ) is hydrogen-bonded to the phosphate O1A oxygen atom (2.55 Å). The ribose sugar of lipoyl-AMP is hydrogen-bonded to two water molecules (Wat 34 and Wat 76 in Fig. 2B). Wat 34 is hydrogen-bonded to the ribose O3* atom (2.65 Å). Wat 76 is hydrogen-bonded to the O2* and O3* oxygen atoms of the ribose ring (2.68 And 3.05 Å); it also interacts with the side chain O␥1 atom of Thr 192 (3.02 Å). The side chain of the strictly conserved Val 196 contacts the nonpolar portion of the ribose (supplemental Fig. S1).
The adenine ring of lipoyl-AMP makes three hydrogen bonds with Tyr 80 and Asp 85 (Fig. 2C). The adenine N6 atom makes two hydrogen bonds with the main chain carbonyl oxygen atom of Tyr 80 (3.13 Å) and the side chain O␦1 atom of Asp 85 (3.03 Å), whereas the N7 atom forms a hydrogen bond with the amide nitrogen atom of Tyr 80 (3.13 Å). Two hydrogen bonds involving the N6 atom will be lost if the adenine ring is substituted with the guanine ring. The adenine ring is also surrounded by hydrophobic residues (Val 79 , Ala 150 , Ala 163 , Leu 165 , Leu 173 , and Leu 177 ). All these hydrophobic residues, except Ala 163 , as well as the above-mentioned histidine and glycine residues (His 81 , His 161 , Gly 48 , Gly 75 , Gly 76 , and Gly 77 ) are strictly conserved among bacterial LplA proteins whose sequences are aligned in Fig. 3.
ATP binds to the same site as the AMP portion of lipoyl-AMP. Its ␤and ␥-phosphate groups point inward, and the ␤-phosphate occupies roughly the same position as the carboxyl portion of lipoyl-AMP. The bent conformation of the triphosphate group is maintained though a strong interaction with the side chain of Arg 72 inside the lipoic acid binding pocket. When lipoic acid enters into its binding pocket of the ATP-bound Ta LplA, we expect that the ␤and ␥-phosphate groups of the bound ATP would be released from the lipoic acid binding pocket, because the environment surrounding the ␤and ␥-phosphate groups of the bound ATP is largely hydrophobic, and thus it would preferentially bind lipoic acid. If lipoic acid were bound prior to ATP binding, Arg 72 would neutralize the negative charge of the carboxylate of lipoic acid.
Conserved Sequence Motifs and Roles of Conserved Residues-The sequence alignment of LplAs, including the human mitochondrial lipoyltransferase, shows that there exist at least three conserved sequence motifs (Fig. 3). The sequence motif I (RRXXGGGXV(F/ Y)HD), encompassing Arg 71 -Asp 82 , is most highly conserved (boxed in red in Fig. 3; highlighted in red lines in Fig. 1D). Motif II (KhXGXA) covers Lys 145 -Ala 150 and contains the strictly conserved Lys 145 (boxed in green in Fig. 3; highlighted in green lines in Fig. 1D), whereas motif III (HXX(L/M)LXXX(D/N)LXXLXXhL) covers His 161 -Leu 177 (boxed in blue in Fig. 3; highlighted in blue lines in Fig. 1D). X stands for any amino acid, h is a hydrophobic residue, and the strictly conserved residues are in boldface. The three motifs provide key residues that are involved in  Fig. 3). The strictly conserved residues are in boldface, and the residues belonging to the conserved sequence motifs are in square brackets. Arg 71 of motif I appears to play a key role in recognizing the lipoyl domain, because our modeling of the complex between Ta LplA and the lipoyl domain of the pyruvate dehydrogenase complex from Azotobacter vinelandii (20) (PDB code 1IYU) indicates that it could interact with the strictly conserved Glu 17 and Glu 36 of the lipoyl domain (discussed below). Other highly or strictly conserved residues of Ta LplA such as Ala 19 , Leu 36 , Leu 69 , Ala 70 , and Gly 148 seem to play structural roles. Asp 82 , Thr 168 , and Asn 198 are involved in a hydrogen-bond network, whose significance is not clear. Glu 56 and Asp 169 are located on the molecular surface.
Despite low levels of overall sequence similarity among the LplA, LipB, and BPL proteins, a lysine residue is strictly conserved in all members of these protein families (3). It corresponds to Lys 145 of Ta LplA. In Ta LplA, Lys 145 makes key interactions with the phosphate moiety of lipoyl-AMP, as discussed above. Besides this role in recognizing the activated intermediate, it may also play a critical role in transferring the lipoyl group from the activated intermediate to the specific lysine residue of the lipoyl domains in the second step of the LplA-catalyzed reaction. The equivalent residues of the BPL (for example, Lys 183 of E. coli BirA) and LipB enzymes are expected to play similar roles. In the case of the reaction catalyzed by LipB, lipoyl-acyl carrier protein corresponds to the activated intermediate.
Lipoyltransferases present in human and bovine mitochondria show 31-35% sequence identity with E. coli LplA, but they cannot catalyze the initial activation of lipoic acid to form lipoyl-AMP (7,8). Instead, the medium-chain fatty acid:CoA ligase-III in bovine liver mitochondria was found to activate lipoic acid utilizing GTP (21). Further studies are needed to clarify why the human and bovine lipoyltransferases are not capable of activating lipoic acid.
Modeling of the Complexes Between Ta LplA and Lipoyl Domains-The biotinyl domains and lipoyl domains are closely related in their threedimensional structures, and there must be one or more key differences between them to ensure their correct selection. To understand how discrimination between lipoylation and biotinylation is achieved, we made an attempt to model the complexes between Ta LplA and two lipoylated proteins, the lipoyl domain of the pyruvate dehydrogenase complex from A. vinelandii (20) (PDB code 1IYU) and the apo form of the glycine cleavage system H protein from Thermus thermophilus (22) (PDB code 1ONL). The glycine cleavage system is a multienzyme complex consisting of four different components (the P-, H-, T-, and L-proteins). The present structure of the lipoyl-AMP complex allowed us to derive plausible models of the complexes between Ta LplA and the lipoylated proteins (Fig. 4). In our modeling attempts, we placed the protruding lysine of the lipoyl domains (Lys 39 or Lys 63 in each of the above-mentioned lipoylated proteins, respectively) in the tunnel-like entrance to the lipoyl-AMP pocket of Ta LplA ( Fig. 2A), and optimized the orientation and position of the lipoyl domain in such a way that the two proteins make reasonable contacts with each other. The interface solvent-accessible surface area for each of the resulting models of the complexes is ϳ1,000 Å 2 , comparable to the values for typical heterodimers (Protein-Protein Interaction Server at www.biochem.ucl.ac.uk/ bsm/PP/server/). The interface between the two proteins has also excellent charge and shape complementarities, even though the interacting partners are from different sources. Interestingly, we note that the large central ␤-sheet of Ta LplA is extended by one of the ␤-sheets of the lipoyl domain in the modeled complexes (Fig. 4A).
Furthermore, our modeling predicts interactions between highly conserved residues of LplA and its target proteins (Fig. 5). In our modeled complex between Ta LplA and the lipoyl domain of the pyruvate dehydrogenase complex from A. vinelandii, Arg 71 in the conserved sequence motif I of Ta LplA lies within a hydrogen bonding distance from the side chains of Glu 17 and Glu 36 of the lipoyl domain. The latter two residues are strictly conserved among lipoyl domains of pyruvate dehydrogenases from different sources. Similarly, the modeled complex between Ta LplA and the glycine cleavage system H protein from T. thermophilus places Arg 71 of Ta LplA within a hydrogen bonding distance from the side chains of Glu 42 and Glu 60 of the H protein. Glu 42 is highly conserved as either Asp or Glu, whereas Glu 60 is strictly conserved among the H proteins. This indicates that Glu 42 and Glu 60 of the T. thermophilus H protein are structurally equivalent to Glu 17 and Glu 36 of the A. vinelandii lipoyl domain, respectively, even though the two protein sequences are too dissimilar to be aligned. In our proposed models of the complexes, Lys 155 of Ta LplA, which is not strictly conserved but is either Lys or Arg in many other LplAs, also interacts with another negatively charged, conserved residue, i.e. Glu 43 of the A. vinelandii lipoyl domain and Asp 67 of the T. thermophilus H protein. A structural comparison of the T. thermophilus and pea H proteins revealed two negative surface regions, designated I and II, that are highly conserved between the two species (22). The negative charges of the surface region I of the T. thermophilus H-protein are attributed to Glu 42 , Glu 60 , and Asp 67 . This region was previously proposed to constitute an interaction surface with T-protein (22). Our modeling of the complex between LplA and the H-protein indicates that the interaction surfaces of the H-protein with LplA and the T-protein overlap considerably. Mutational studies of the E. coli biotinyl domain indicated that one of the key structural determinants of protein specificity for biotinylation by BPL in E. coli is a thumb-like protrusion comprising Thr 94 -Lys 100 between strands ␤2 and ␤3 of the E. coli biotinyl domain (23), which is an insertion relative to the lipoyl domains (2). When the residues Thr 94 -Lys 100 were deleted from the E. coli biotinyl domain, the mutant protein was efficiently lipoylated (23). It was concluded that the protruding thumb between strands ␤2 and ␤3 is not critical for the interaction with

Structure of Lipoate-Protein Ligase A
BPL, but its presence is sufficient to prevent the E. coli biotinyl domain from becoming lipoylated (23). This effect may be limited to E. coli, because amino acids that constitute the thumb are not present in most other biotinyl domains (23,24). When we superimpose the biotinyl domain of acetyl-coenzyme A carboxylase from E. coli (25) (PDB code 1BDO) with the lipoyl domain of the modeled complex, the protruding thumb of the E. coli biotinyl domain clashes severely with Ta LplA (mostly the carboxyl-terminal side of ␣7) and, thus, binding of the biotinyl domain to LplA would be prevented. Deletion of the residues Thr 94 -Lys 100 from the E. coli biotinyl domain would remove the severe overlap between Ta LplA and the biotinyl domain from the modeled complex. When we overlay the structure of the E. coli BirA onto that of Ta LplA, the thumb of the biotinyl domain does not block binding of the biotinyl domain to BirA. Therefore, our modeling is consistent with the previously proposed role of the thumb-like protrusion in preventing the E. coli biotinyl domain from being lipoylated (23).
Of the seven lysine residues of the A. vinelandii lipoyl domain (or six of the T. thermophilus H protein), only Lys 39 (or Lys 63 ) is lipoylated. Discrimination of the target lysine from other lysine residues may be understood from our modeling, which indicates that Lys 39 (or Lys 63 ) is most protruding and can fit into the entrance to the lipoyl-AMP binding pocket of Ta LplA without causing severe clashes between the two proteins, whereas all other lysine residues cannot.