The Crystal Structure of the Adenylation Enzyme VinN Reveals a Unique β-Amino Acid Recognition Mechanism*

Background: The β-amino acid adenylation reaction is important for biosynthesis of natural products. Results: We present the crystal structure and a mutational study of the adenylation enzyme VinN involved in vicenistatin biosynthesis. Conclusion: VinN has a characteristic substrate-binding pocket that selectively accommodates β-amino acids. Significance: This study could provide clues for β-amino acid specificity prediction and protein engineering of adenylation enzymes. Adenylation enzymes play important roles in the biosynthesis and degradation of primary and secondary metabolites. Mechanistic insights into the recognition of α-amino acid substrates have been obtained for α-amino acid adenylation enzymes. The Asp residue is invariant and is essential for the stabilization of the α-amino group of the substrate. In contrast, the β-amino acid recognition mechanism of adenylation enzymes is still unclear despite the importance of β-amino acid activation for the biosynthesis of various natural products. Herein, we report the crystal structure of the stand-alone adenylation enzyme VinN, which specifically activates (2S,3S)-3-methylaspartate (3-MeAsp) in vicenistatin biosynthesis. VinN has an overall structure similar to that of other adenylation enzymes. The structure of the complex with 3-MeAsp revealed that a conserved Asp230 residue is used in the recognition of the β-amino group of 3-MeAsp similar to α-amino acid adenylation enzymes. A mutational analysis and structural comparison with α-amino acid adenylation enzymes showed that the substrate-binding pocket of VinN has a unique architecture to accommodate 3-MeAsp as a β-amino acid substrate. Thus, the VinN structure allows the first visualization of the interaction of an adenylation enzyme with a β-amino acid and provides new mechanistic insights into the selective recognition of β-amino acids in this family of enzymes.

Adenylation enzymes, which include the nonribosomal peptide synthetase (NRPS) 2 adenylation domains, the acyl-CoA synthetases, and the luciferase enzymes, play important roles in the biosynthesis and degradation of primary and secondary metabolites (1). These enzymes first catalyze the adenylation of a carboxylate substrate using ATP to form an acyl-AMP intermediate. In most cases, the adenylation enzymes subsequently catalyze the formation of a thioester bond with the 4Ј-phosphopantetheine group of a carrier protein or CoA molecule. The carboxylate substrates of adenylation enzymes are structurally diverse and include acetate, amino acids, fatty acids, and aromatic acids. In general, each adenylation enzyme recognizes a specific carboxylate substrate.
Modular NRPS enzymes are involved in the biosynthesis of various secondary metabolites. NRPS adenylation domains, which mostly have ␣-amino acids as substrates, serve as gatekeepers of the NRPS assembly line via selective substrate recognition and activation. After the determination of the first crystal structure of an NRPS adenylation domain, PheA (see Fig. 1A) involved in gramicidin S biosynthesis (2), general rules for the assignment of substrate specificity for adenylation enzymes have been developed. These rules are based on classification of a specificity-conferring code derived from 10 amino acid residues that comprise the substrate-binding pocket (3,4). The ϳ100-amino acid stretch of the N-terminal domain (corresponding to the positions Asp 235 -Cys 331 in PheA) contains nine of the 10 residues in the substrate-binding pocket. For ␣-amino acid substrate recognition, the Asp residue at the 235 position is invariant and is essential for stabilization of the ␣amino group of the ␣-amino acid. The remaining residue is the highly conserved C-terminal Lys residue, which is involved in two key interactions with both the carboxylate group of the ␣-amino acid and the ribose moiety of adenylate. These two residues (Asp and Lys) are important to fix the position and orientation of the ␣-amino acid for the reaction. The other eight specificity-conferring code residues are involved in the recognition of the side chain of the ␣-amino acid substrate. This specificity-conferring code rule can be applied to various NRPS-type adenylation enzymes because information on the relationship between the sequence and substrate specificity of these enzymes has been accumulated (3)(4)(5)(6)(7). This rule allows us to predict the substrate specificity of many biochemically uncharacterized adenylation enzymes from their amino acid sequences and enables rational alteration of their substrate specificities (3, 8 -12). However, this specificity-conferring code is not applicable for adenylation enzymes that activate unusual substrates such as ␤-amino acids.
Various natural products containing a ␤-amino acid unit have been isolated (13). In many cases, adenylation enzymes that activate ␤-amino acids are involved in their biosynthetic pathways (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). Similar to ␣-amino acid-activating enzymes, the interaction with the ␤-amino group and the carboxylate group should be important for ␤-amino acid-activating enzymes to fix the substrate orientation. However, the recognition model for ␣-amino acid-activating enzymes cannot be applied to that for ␤-amino acid-activating enzymes because the amino group position of the ␤-amino acid is different from that of the ␣-amino acid. Therefore, it can be assumed that ␤-amino acid adenylation enzymes have a structurally different active site for recognition of the ␤-amino acid. To date, no crystal structures of ␤-amino acid-activating enzymes are available, so the ␤-amino acid recognition mechanism of adenylation enzymes is poorly understood. What determines the substrate specificity between ␣-amino acid and ␤-amino acid adenylation enzymes also remains elusive.
Macrolactam natural products are an important class of macrocyclic polyketides, and many of them contain various ␤-amino acids at the starter position for polyketide chain elongation. Vicenistatin is produced by Streptomyces halstedii HC34 and belongs to the family of macrolactam antibiotics. This enzyme possesses a unique ␤-amino acid unit, 3-amino-2methylpropionate, at the starter position of the polyketide backbone (25). Recently, we elucidated that the stand-alone adenylation enzyme VinN is involved in vicenistatin biosynthesis (14) (see Fig. 1B). VinN recognizes (2S,3S)-3-methylaspartate (3-MeAsp) as a ␤-amino acid substrate and catalyzes adenylation of the carboxyl group at the C4 position (see Fig.  1A). Then VinN transfers the ␤-amino acid group onto the stand-alone acyl carrier protein VinL to give 3-MeAsp-VinL. After decarboxylation of the 3-MeAsp moiety by the pyridoxal 5Ј-phosphate-dependent enzyme VinO, the resulting 3-amino-2-methylpropionate unit is aminoacylated with L-alanine to give dipeptidyl-VinL by another stand-alone adenylation enzyme, VinM. The dipeptide moiety is then selectively transferred onto the loading acyl carrier protein domain of polyketide synthase for polyketide chain elongation. Thus, VinN appears to play a crucial role in the selection of the ␤amino acid starter unit for polyketide synthase in vicenistatin biosynthesis. Previous biochemical studies have shown that VinN exhibits a strong preference for 3-MeAsp over other amino acids (14). VinN shows no activity against ␣-amino acids except for weak activity against L-aspartate. As VinN exhibits relatively low sequence identity with well studied adenylation enzymes such as the NRPS adenylation domain, the substrate specificity of VinN cannot be predicted from the specificityconferring code. The crystal structure of SlgN1, which catalyzes the adenylation of 3-MeAsp in the biosynthesis of streptolydigin, has been reported recently (26). SlgN1 recognizes (2S,3S)-3-MeAsp as an ␣-amino acid substrate and adenylates the carboxyl group at the C1 position (Fig. 1A). The orientation of 3-MeAsp at the active site in S1gN1 is proposed to be opposite to that in VinN. In this study, we carried out kinetic, mutational, and structural studies on VinN to clarify how VinN selectively recognizes 3-MeAsp. The VinN crystal structure allows the first visualization of the interaction between an adenylation enzyme and a ␤-amino acid and provides new mechanistic insights into the selective recognition of ␤-amino acids in this family of enzymes.

Preparation of Recombinant Wild-type and Mutant Enzymes-
The N-terminal domain region of the vinN gene (vinNN) was amplified from pCold I-vinN (14) with the primers 5Ј-acgccatatcgccgaaagg-3Ј and 5Ј-tttaagcttaccggccgccgaagtag-3Ј and then inserted between the NdeI and HindIII sites of the expression vector pCold I (Takara Biochemicals, Ohtsu, Japan), which is designed to attach a hexahistidine-tagged sequence to the N terminus of the target protein.
Escherichia coli BL21(DE3) cells (Takara Biochemicals) harboring a pCold I-vinN or pCold I-vinNN plasmid were grown at 37°C in Luria-Bertani broth containing ampicillin (50 g/ml). After the optical density at 600 nm reached 0.6, protein expression was induced by the addition of isopropyl ␤-D-1-thiogalactopyranoside (0.2 mM), and the cells were then cultured for an additional 16 h at 15°C. The recombinant protein, which was collected from cell-free extracts prepared by sonication, was purified on a TALON affinity column (Clontech). The protein was then desalted and concentrated using a PD-10 column (GE Healthcare) and an Amicon Ultra centrifugal filter (Merck Millipore, Billerica, MA), respectively. For crystallization, the recombinant protein was further purified by ResourceQ (GE Healthcare) anion-exchange chromatography with a linear gradient from 0.15 to 0.45 M NaCl in 10 mM HEPES-Na buffer (pH 7.7) containing 10% (v/v) glycerol.
Enzyme Assays and Determination of Kinetic Parameters-A continuous spectroscopic assay that measures the release of inorganic pyrophosphate with the adenylation reaction was carried out according to the method of Webb (27). The assay solution (500-l total volume) contained 50 mM Tris-HCl buffer (pH 7.5), 10% glycerol, 1 mM MgCl 2 , 1 mM ATP, 0.1 mM 2-amino-6-mercapto-7-methylpurine ribonucleoside, 1 unit/ml inorganic pyrophosphatase, 1 unit/ml purine nucleotide phosphorylase, and 1 mM amino acid (DL-threo-␤-MeAsp (Sigma-Aldrich) or L-aspartate (Kanto Chemical, Tokyo, Japan)). For kinetic assays, the amino acid concentration was varied between 0.04 and 100 mM. The reaction was initiated by addition of VinN (1-10 M) to the mixture and incubated at 28°C. The increase in absorbance at 360 nm, attributable to the formation of 2-amino-6-mercapto-7-methylpurine, per second was monitored using a UV-2450 spectrophotometer (Shimadzu, Tokyo, Japan). The initial velocity was determined from the linear portion of the optical density profile (⑀ 360 nm ϭ 11,000 M Ϫ1 cm Ϫ1 ). Steady-state parameters were determined by the Michaelis-Menten equation.
Crystallization, Data Collection, and Structure Determination-Crystals of VinNN were grown from a 1:1 (v/v) mixture of a VinNN protein solution (10 mg/ml in 10 mM HEPES-Na (pH 7.5) and 10% glycerol) and a reservoir solution (0.1 M Tris-HCl (pH 8.5), 0.2 M sodium acetate, and 30% polyethylene glycol 4000) using the sitting drop vapor diffusion method at 5°C. The 3-MeAsp complex and L-aspartate complex were prepared by soaking apocrystals in 100 mM DL-threo-␤-MeAsp for 4 h and 100 mM L-aspartate for 1 h, respectively. Prior to collection of the x-ray data, the crystals were flash frozen in a stream of liquid nitrogen. The x-ray diffraction data were collected on beamlines BL-5A and AR-NW-12A at the Photon Factory (Tsukuba, Japan) and subsequently indexed, integrated, and scaled using the iMosflm program (28). The initial phases were determined by the molecular replacement method using the Molrep pro-  gram (29) with the crystal structure of the N-terminal domain of 4-coumarate-CoA ligase from Populus tomentosa (Protein Data Bank code 3A9U) used as a search model. Rotation and translation functions were calculated using data of 50.0 -2.5-Å resolution. Two molecules were found in the asymmetric unit (the correlation coefficient of the correct solution is 0.365, which is significantly higher than those of other unrelated peaks). The ARP/wARP program (30) was used for automatic initial protein model building. Coot (31) was used for visual inspection and manual rebuilding of the model. Refmac (32) was used for refinement. The figures were prepared using PyMOL (33). The distances were measured in both molecules in the asymmetric unit and then shown as averaged values. The geometries of the final VinNN structures were evaluated using the program Rampage (34). The resulting coordinates and structure factors have been deposited in the Protein Data Bank (Protein Data Bank codes 3WV4, 3WV5, and 3WVN).

RESULTS
Kinetic Analysis of VinN-To analyze the preference of VinN for 3-MeAsp, we carried out a kinetic analysis. The K m value for DL-threo-3-MeAsp, which is a mixture of 2S,3S-and 2R,3Renantiomers, at 28°C was found to be 0.13 Ϯ 0.02 mM, which is comparable with those of the known adenylation enzymes (15-17, 35, 36) (Table 1). VinN also shows weak activity toward L-aspartate, which lacks the methyl group at the C3 position. VinN had a 35-fold larger K m value for L-aspartate (4.5 Ϯ 0.5 mM). Thus, the methyl group at the C3 position of 3-MeAsp is important for recognition by VinN. VinN showed no activity against D-aspartate, indicating that the orientation of the functional groups at the ␤-position of 3-MeAsp is important for recognition by VinN. These results suggest that VinN accepts only (2S,3S)-3-MeAsp as a substrate in the reaction with Crystallization of VinN-To investigate the structural basis of the ␤-amino acid recognition mechanism, we carried out a crystallographic analysis of VinN. We first attempted to crystallize the full-length VinN protein. However, we could not obtain any VinN crystals. Adenylation enzymes have been reported to change conformation at the C-terminal region during reaction (37). After the initial adenylation step, the C-terminal domain region is rotated by ϳ140°for the following thioester bond formation step. Crystallization of these enzymes containing flexible C-terminal parts is challenging. Hence, there are some examples of crystal structures in which only the N-terminal part has been used for crystallization (26,38,39). Therefore, we constructed the heterologous expression system of the VinNN protein containing only the N-terminal domain (Met 1 -Arg 426 ) and then attempted to crystallize the VinNN protein. Finally, we succeeded in the crystallization of VinNN and determined its crystal structure at 2.15-Å resolution ( Table 2).
Overall Structure of VinN-Two VinNN molecules are present in each crystallographic asymmetric unit. The VinNN structure has a five-layered ␣␤␣␤␣ sandwich fold as observed in other crystal structures of adenylation enzymes (2, 26, 38 -41) (Fig. 2A). The structure contains a few disordered surface loop regions (Met 1 -Lys 23 , Gly 250 -Asp 254 , and Asp 399 -Gly 403 ). A search for structurally related proteins using the Dali program (42) revealed that the closest functionally characterized protein is fatty acyl-CoA synthetase FadD13 (40)  A structural comparison of VinNN with other adenylatebound adenylation enzymes, including PheA and SlgN1 (2,26,35,41,43), showed that VinNN has a similar cleft for adenylate binding on the surface of the N-terminal domain. VinN has several conserved residues at that cleft (Fig. 3). For example, Thr 327 and Glu 328 that are supposed to interact with the ␣-phosphate moiety of adenylate, Asp 411 that is supposed to interact with ribose hydroxyl groups, and Tyr 324 that is supposed to interact with the adenine moiety are located at almost the same positions in the structure (Fig. 4A). These structural observations suggest that the manner of ATP binding of VinN is similar to that of other adenylation enzymes.
Structure of the 3-MeAsp and L-Aspartate Complexes-To understand the substrate recognition mechanism of VinN, we determined the structure of the complex with 3-MeAsp at 2.20-Å resolution. The position of the substrate-binding pocket of VinNN is similar to those of other adenylation enzymes. The entrance of the substrate-binding pocket is exposed to solvent because of the absence of the C-terminal domain. The structure shows clear electron density for (2S,3S)-3-MeAsp in the substrate-binding pocket but not for (2R,3R)-3-MeAsp (Fig. 2B). The binding of 3-MeAsp occurred with little overall structural perturbation to the VinNN polypeptide backbone (r.m.s.d. of 0.47 Å for 385 C␣ atoms of chain A). Some side-chain movements are observed in the binding pocket. The side chains of Asp 230 , Phe 231 , and Ser 299 exhibit 19°, 15°, and 140°rotations, respectively, to accommodate 3-MeAsp (Fig. 2C). In the substrate-free structure, the backbone and side chains in the region of Ser 299 and Ala 300 seem to be flexible judging from their high B-factor values (65.5 Å 2 ). In contrast, this region of the structure in the complex exhibits significantly lower B-factor values (33.0 Å 2 ), suggesting that the binding of 3-MeAsp stabilizes it. Furthermore, we determined the structure of the complex with L-aspartate at 2.20-Å resolution. Similarly, the electron density of L-aspartate was clearly observed in the substrate-binding pocket (Fig. 2D). The overall structure of the complex with L-aspartate is almost identical to that with 3-MeAsp (r.m.s.d. of 0.13 Å for 388 C␣ atoms of chain A). There are no significant displacements in the substrate-binding pocket between the structures of these two complexes (Fig. 2C). Substrate Binding in the Complex Structures-In both complex structures, the amino acid substrate is bound to the substrate-binding pocket of VinNN in the same position and orientation. The C4 carboxyl group of 3-MeAsp and L-aspartate is exposed to solvent and does not interact with any residues of the N-terminal domain.
In the structure of the complex with 3-MeAsp, two salt bridge interactions are formed between Arg 331 and the C1 carboxyl group of 3-MeAsp (2.7 and 2.8 Å), and one salt bridge interaction is observed between Lys 330 and the C1 carboxyl group of 3-MeAsp (3.0 Å) (Fig. 2B). These structural observations suggest the importance of Lys 330 and Arg 331 in 3-MeAsp recognition. In fact, the K330N mutant showed a significantly decreased activity (0.8% of the wild type). In addition, K330A and R331A mutants showed a complete loss of activity. Both Lys 330 and Arg 331 side chains are stacked between Tyr 234 and Met 323 and are located close together (3.5 Å) despite an electrostatic repulsion (Fig. 2E). Even in the substrate-free structure, these basic residues occupy the same position (Fig. 2C). Tyr 234 and Met 323 do not directly interact with the 3-MeAsp substrate but are likely to modulate the conformation of the Lys 330 and Arg 331 side chains by steric effects. To evaluate this steric effect of Tyr 234 and Met 323 , we constructed Y234A and M323G mutants. The Y234A mutant showed very weak activity (3.7% of the wild type), and the M323G mutant completely lost activity. In these mutants, the smaller side chain of the mutated residue cannot have any steric interactions with the Lys 330 or Arg 331 side chains. This probably causes displacement of these basic residues to reduce the electrostatic repulsion between them, resulting in the observed significant reduction in activity.
Ser 299 forms a hydrogen bond with the C1 carboxyl group of 3-MeAsp. The S299A mutant exhibited a 6-fold higher K m value compared with the wild type, suggesting that the sidechain hydroxyl group of Ser 299 participates in the recognition of the C1 carboxyl group of 3-MeAsp. The side-chain carboxyl group of Asp 230 interacts with the ␤-amino group of the 3-MeAsp substrate with a salt bridge interaction (2.6 Å). The main-chain carbonyl oxygen of Lys 330 also interacts with the ␤amino group through a hydrogen bond (2.8 Å). The methyl group of 3-MeAsp seems to be recognized by the side-chain aromatic group of Phe 231 (3.8 -4.3 Å) through CH-(44) and van der Waals interactions (Fig. 2, C and E). The presence of these interactions appears to give 3-MeAsp selectivity over L-aspartate. In the structure of the complex with L-aspartate, the C3 carbon of L-aspartate is relatively far from the side chain of Phe 231 (5.0 Å). F231A and F231L mutations resulted in higher K m values for both 3-MeAsp and L-aspartate. The F231L mutant showed a 7-fold higher K m value for 3-MeAsp compared with the wild type, although the Leu residue should retain the van der Waals interactions with the methyl group of 3-MeAsp. This mutant also showed a 4-fold higher K m value for  (Table 1). Furthermore, the F231A mutant exhibited very weak activity toward 3-MeAsp (1.8% of the wild type) and no detectable activity with L-aspartate. These results suggest that the mutation of Phe 231 affects substrate specificity both for 3-MeAsp and L-aspartate. The side chain of Phe 231 might be important for the recognition of not only the methyl group but also other parts of 3-MeAsp. Structural Comparison with the Crystal Structures of Other Adenylation Enzymes-We compared the structure of VinNN with those of other adenylation enzymes, including PheA, which recognizes L-phenylalanine as a substrate (2). The carboxyl group at the C4 position of 3-MeAsp that is adenylated in VinN occupies a position similar to that of the ␣-carboxyl group of L-phenylalanine in PheA (Fig. 4B), suggesting that the same interaction exists with the conserved Lys residue of the C-terminal domain (Lys 510 in VinN) as observed in other adenylation enzymes (Fig. 5). The position of the amino nitrogen atom of 3-MeAsp is similar to that of L-phenylalanine in PheA. In both structures, the conserved Asp residue forms the same interaction with the amino group of the substrate. Thus, the distance between the amino and carboxylate groups of the amino acids at the active site of both enzymes is similar, although the ␤-amino acid has one carbon atom inserted between these termini compared with the ␣-amino acid. Consequently, in the VinNN structure, the C2-C3 bond of 3-MeAsp appears to be pushed into a relatively large space of the active site. The positions of the three carbon atoms (C1, C2, and C3) of 3-MeAsp individually show 1.1-1.5-Å displacement from the corresponding atoms of L-phenylalanine in the PheA structure.
Further structural comparisons revealed that two factors likely give rise to the difference in the overall architecture of the substrate-binding pocket. First, the ␤13␤14 loop containing the two important residues Lys 330 and Arg 331 is one residue shorter in VinN than in most other adenylation enzymes, including PheA (Figs. 3 and 4C). This one-residue-shorter loop has also been reported in malonyl-CoA synthetase ScMatB (45). In the VinNN structure, this shorter loop causes a 0.8 -2.0-Å movement of the polypeptide backbone between Lys 330 and Ile 332 (Fig. 4, B and D), providing space for fixing the location and orientation of the C1-C2 bond of 3-MeAsp. The C1 carboxyl group of 3-MeAsp might be pulled in this space by interactions with Ser 299 , Lys 330 , and Arg 331 , resulting in its slight movement toward the ␤13␤14 loop. Second, VinN has bulky residues in the substrate-binding pocket. Phe 231 and Ser 299 , which are equivalent to Ala 236 and Ala 301 , respectively, in PheA, are located on the opposite side of the substrate-binding pocket from the ␤13␤14 loop (Fig. 4B). The position of the C1, C2, and C3 carbon atoms of the 3-MeAsp substrate could be affected by the presence of these two larger residues. They might be pushed toward the ␤13␤14 loop by the steric constraints of these residues. The importance of Phe 231 was confirmed by the mutational studies where mutation of Phe 231 significantly reduced the activity against both 3-MeAsp and L-aspartate (Table 1). Thus, these two factors are thought to be important for the construction of a ␤-amino acid-specific substrate-binding pocket to accommodate 3-MeAsp.

DISCUSSION
Many crystal structures of adenylation enzymes such as NRPS adenylation domains have been reported (2, 26, 35, 41, 43, 46 -48). However, no crystal structure of a ␤-amino acidactivating enzyme has been available so far. In this study, we determined the structure of the N-terminal domain of VinN, which is involved in biosynthesis of the macrolactam antibiotic vicenistatin. The structures of complexes with both 3-MeAsp and L-aspartate clearly demonstrate that VinN recognizes 3-MeAsp as a ␤-amino acid. The VinNN structure can provide important mechanistic insights into ␤-amino acid recognition as described below.
Sequence alignment and structural comparison with other adenylation enzymes revealed that the ␤-amino acid substrate specificity of VinN is also governed by the specificity-conferring code (Figs. 3 and 5). The VinNN structure and the mutational study clearly show that five of the 10 residues (Asp 230 , Phe 231 , Ser 299 , Lys 330 , and Arg 331 ) are directly involved in substrate binding (Fig. 2, B and E). Two other residues (Tyr 234 and Met 323 ) derived from the specificity-conferring code seem to be indirectly involved in substrate binding by conformational modulation of two essential basic residues, Lys 330 and Arg 331 . In addition, the conserved Lys 510 of the C-terminal domain is thought to interact with the substrate carboxyl group that is adenylated as reported for other adenylation enzymes. Importantly, the Asp residue important for recognition of the ␣amino group of the ␣-amino acid is also conserved in VinN. The VinNN structure shows that Asp 230 is used for the recognition of the ␤-amino group of 3-MeAsp. However, the overall archi-   tecture of the substrate-binding pocket of VinN is significantly different compared with those of other adenylation enzymes. The ␤-amino acid specificity of VinN seems to be dictated by the surrounding regions, including the shorter ␤13␤14 loop and the specificity-conferring code residues such as Phe 231 . These factors seem to largely contribute to control of the conformation of the ␤-amino acid substrate so that the ␤-amino group of the substrate is placed adjacent to the Asp 230 sidechain carboxyl group. A comparison of the specificity-conferring code of VinN with other adenylation enzymes revealed that VinN has a special code for the ␤-amino acid substrate. The unique feature involves the presence of one polar residue, Ser 299 , and two basic residues, Lys 330 and Arg 331 (corresponding to Ala 301 , Ile 330 , and Cys 331 , respectively, in PheA). The crystal structure shows that all of these residues are involved in the recognition of the carboxyl group at the C1 position that is not adenylated (Fig. 2B). Other ␣-amino acid-activating enzymes generally have small aliphatic residues at the positions corresponding to Ser 299 and Lys 330 in VinN (3)(4)(5)(6)(7). In addition, they contain neither Arg nor Lys at the position equivalent to Arg 331 in VinN, although they have various types of residues, including Asp and His, at this position. Thus, VinN has distinct amino acids at these posi-tions. These polar and basic residues might be specific for the recognition of dicarboxylic ␤-amino acids. Even adenylation enzymes using dicarboxylic ␣-amino acid substrates, including L-aspartate and L-glutamate, have no basic residues at corresponding positions (7). These enzymes have a basic or polar residue at a different position such as the 234 or 278 position in PheA numbering for recognition of a carboxyl group that is not adenylated. Another feature is that VinN has a bulky aromatic residue, Phe 231 , adjacent to the invariant Asp 230 . The crystal structure shows that this Phe residue likely provides a steric constraint to control the substrate conformation. Normally, ␣-amino acid-activating enzymes have a small aliphatic residue such as Ala or Val at this position. Thus, the substrate-conferring code of VinN is quite different from those of ␣-amino acid adenylation enzymes. In contrast to VinN, SlgN1 shows the above mentioned ␣-amino acid adenylation enzyme-like features. SlgN1 contains Gln 305 at the position equivalent to Thr 278 in PheA (Fig. 5). The SlgN1 structure suggests that this polar Gln 305 residue might be involved in carboxyl group recognition, although the structure of the complex with 3-MeAsp was not reported (26) (Fig. 4D). SlgN1 also has a small Ala 264 residue adjacent to the invariant Asp 263 residue. In addition, the length of the ␤13␤14 loop in SlgN1 is the same as those in   ␣-amino acid-activating enzymes (Fig. 4C). These differences could explain why SlgN1 shows the opposite orientation of 3-MeAsp binding to VinN. A few alterations in the substratebinding pocket affect the orientation of substrate in this case. This study might be important from an evolutionary point of view. Adenylation enzymes might gain ␤-amino acid specificity from ␣-amino acid specificity with a few critical alterations of the substrate-binding pocket although they maintain the overall fold and catalytic mechanism. The replacement of some substrate-binding pocket residues is likely a key evolutionary event to generate various ␤-amino acid specificities. Sequence alignment and phylogenetic analysis show that other ␤-amino acidactivating enzymes can be divided into several subfamilies (Fig.  5). These enzymes possess an Asp residue at the position corresponding to Asp 230 of VinN and are thought to use an Asp residue for the recognition of the ␤-amino group of the ␤amino acid substrate in a manner similar to that of VinN. Of these, IdnL1 (18), CmiS6 (19), BecJ (20), and Stro_2775 (21), which use ␤-amino fatty acids in macrolactam antibiotic biosynthesis, might have a similar ␤-amino acid recognition mechanism because they share some residues in common with VinN. In particular, Phe 231 and Ser 299 of VinN are conserved in these enzymes (Fig. 5), which might use both the Phe and Ser residues to control substrate conformation by steric constraints similarly to VinN. In addition, these enzymes might have a ␤13␤14 loop that is one residue shorter than that in ␣-amino acid-activating enzymes, although this is difficult to predict because of the low sequence identity (Fig. 4C). The identification of these conserved features could help the prediction of ␤-amino acid substrate specificity of biochemically uncharacterized adenyla-tion enzymes on the basis of their amino acid sequences. Conversely, other types of ␤-amino acid-activating enzymes seem to use a different amino acid residue for recognition. For example, SgcC1, which activates ␤-tyrosine in C-1027 biosynthesis, contains a Pro residue adjacent to the invariant Asp (Fig. 5). This Pro residue is thought to be involved in the catalysis and/or specificity for ␤-amino acids as evidenced from mutational studies (16). Other enzymes have different amino acids in this position. CmdD-A 7 , which activates ␤-phenylalanine in chondramide biosynthesis, has a smaller Gly residue (17), and Orf5, which activates ␤-lysine in streptothricin biosynthesis, contains a Thr residue (24). These enzymes might adopt a different strategy for ␤-amino acid recognition. The elucidation of the structures of these enzymes is necessary for the further understanding of ␤-amino acid recognition by adenylation enzymes.
In conclusion, we determined the first crystal structure of a ␤-amino acid adenylation enzyme. This structure sheds on the molecular basis for selective activation of ␤-amino acids by visualizing the orientation and conformation of the ␤-amino acid substrate. This study represents a significant contribution to the expansion of knowledge of the adenylation enzyme family. In addition, the VinN structure could provide clues useful for the protein engineering of an adenylation enzyme to enable alteration of the substrate specificity to introduce ␤-amino acids instead of ␣-amino acids.