The Catalytic Intermediate Stabilized by a “Down” Active Site Loop for Diaminopimelate Decarboxylase from Helicobacter pylori

The meso-diaminopimelate decarboxylase (DAPDC, EC 4.1.1.20) catalyzes the final step of l-lysine biosynthesis in bacteria and is regarded as a target for the discovery of antibiotics. Here we report the 2.3Å crystal structure of DAPDC from Helicobacter pylori (HpDAPDC). The structure, in which the product l-lysine forms a Schiff base with the cofactor pyridoxal 5′-phosphate, provides structural insight into the substrate specificity and catalytic mechanism of the enzyme, and implies that the carboxyl to be cleaved locates at the si face of the cofactor. To our knowledge, this might be the first reported external aldimine of DAPDC. Moreover, the active site loop of HpDAPDC is in a “down” conformation and shields the ligand from solvent. Mutations of Ile148 from the loop greatly impaired the catalytic efficiency. Combining the structural analysis of the I148L mutant, we hypothesize that HpDAPDC adopts an induced-fit catalytic mechanism in which this loop cycles through “down” and “up” conformations to stabilize intermediates and release product, respectively. Our work is expected to provide clues for designing specific inhibitors of DAPDC.

L-Lysine is an essential amino acid in the bacterial life cycle. In addition to its fundamental role in protein synthesis, L-lysine is also a basic building block of bacterial cell wall peptidoglycan. The biosynthesis of L-lysine in bacteria involves a multistep pathway (1,2) transforming L-aspartate to a common precursor meso-diaminopimelate (DAP). 4 DAP is then decarboxylated at the D-stereocenter by meso-diaminopimelate decarboxylase (DAPDC) encoded by lysA gene to give the final product L-lysine. Thus, DAPDC acts as a key enzyme in this pathway and is of great importance to bacterial growth and survival. De Lancastre et al. (3) have demonstrated that abolishing the function of DAPDC in methicillin-resistant Staphylococcus aureus by transposon insertion could reduce the level of the methicillin resistance of methicillin-resistant S. aureus by nearly 100-fold; and Gokulan et al. (13) have established that the lysA gene of Mycobacterium tuberculosis was essential for in vivo bacterial viability in immunocompromised mice. In contrast, instead of synthesizing L-lysine de novo, the human body obtains it from food, and no counterpart of DAPDC is found in mammals. Therefore, DAPDC could be a potential target for developing novel antibiotics.
DAPDC is a group IV pyridoxal 5Ј-phosphate (PLP)-dependent enzyme (also known as alanine racemase (ALR) family) (5). Except ALR, all enzymes of this family are basic amino acid decarboxylases (6 -8), including DAPDC, ornithine decarboxylase (ODC), arginine decarboxylase (ADC), decarboxylases with dual specificity for L-lysine and L-ornithine (L/ODC), and carboxynorspermidine decarboxylases. Reactions catalyzed by representative enzymes are shown in Fig. 1a. Crystal structures of these group IV PLP enzymes (except carboxynorspermidine decarboxylases) (6, 9 -12) reveal that although the family members do not share high sequence identities (only 5% between DAPDC and ALR), they adopt a similar fold consisting of an N-terminal classical ␣/␤-barrel and a C-terminal ␤-strand domain, and all exist as an obligate homodimer with two active sites lining the dimer interface.
To date, the crystal structures of DAPDC from Methanococcus jannaschii (MjDAPD, PDB codes 1TUF and 1TWI) and M. tuberculosis (MtDAPDC, PDB codes 1HKV, 1HKW, and 2O0T) (4,13) have been reported, and several other DAPDC structures (DAPDC from Aquifex aeolicus (AaDAPDC, PDB code 2P3E) and Escherichia coli (EcDAPDC, PDB codes: 1KO0 and 1KNW) (14)) were released in the Protein Data Bank. Some detailed description of cofactor and substrate binding was supplied in the reported structures. However, because of the inconsistent binding patterns of the substrate analogues (lysine or azelaic acid), the complete set of substrate-recognizing residues of DAPDC still remains uncertain.
Here we report the 2.3 Å crystal structure of DAPDC from Helicobacter pylori (HpDAPDC), a severe clinical pathogen causing peptic ulcer and gastric cancer (15)(16)(17). In the structure, the product L-lysine is covalently bonded to cofactor PLP, and the resultant external aldimine is believed to closely mimic the catalytic intermediate formed between DAP and PLP during the decarboxylation reaction. Detailed information about substrate binding is then derived and carefully compared with DAPDCs from other species as well as other group IV PLP enzymes (ALR, ODC, and ADC). In addition, structural analysis also implied the putative catalytic mechanism of DAPDC, especially the position and recognition of the carboxyl group to be cleaved. Moreover, a loop covering the active site of HpDAPDC is in a "down" conformation and shields the ligand from the solvent. Mutations of Ile 148 from this loop greatly impaired the catalytic efficiency, especially the k cat value of the enzyme. Combining the crystal structure of the I148L mutant and the results from previous studies, we propose that HpDAPDC adopts an induced-fit mechanism for catalysis in which the active site loop cycles through down and "up" conformations to stabilize intermediates and release product, respectively. Our work is expected to gain insight into the essential catalytic mechanism of HpDAPDC and provide clues for designing specific inhibitors.

EXPERIMENTAL PROCEDURES
Cloning of lysA Gene from H. pylori-Based on the genome sequences of H. pylori strains 26695 and J99 (GenBank TM accession numbers NC_000915 and NC_000921), two PCR primers, 5Ј-TGTATGTGAATGCGGGGGTGG-3Ј (sense) and 5Ј-GGGGCTTTCTTGTTTGATGAG-3Ј (antisense), were designed to amplify the corresponding region, including the lysA gene from the genome of H. pylori strain SS1 (obtained from Shanghai Institute of Digestive Disease). According to the sequencing result of the amplified DNA segment, another pair of primers, 5Ј-TGGGCATATGTTTAATTATGAAGAGC-3Ј (sense) and 5Ј-TTTCTCGAGTCAAACCCCTTTTAAGCC-3Ј (antisense), was synthesized to amplify lysA gene from the genome again. The PCR product was cloned into the NdeI and XhoI sites of pET-28a (Novagen) to produce N-terminal Histagged protein. The gene sequence has been submitted to Gen-Bank TM with accession number EU056336.
Cloning of lysA Gene from E. coli-Similar to the cloning procedures of pET28a-HpDAPDC, the lysA gene of E. coli was cloned from the genome of E. coli strain JM109 based on the sequence of strain K12 (GenBank TM accession number NC_000913). The gene was amplified by PCR with the primers 5Ј-ATATGGATCCATGCCACATTCACTGTT-CAG-3Ј (sense) and 5Ј-ATATCTCGAGTTAAAGCAATT-CCAGCG-3Ј (antisense) and cloned into the BamHI and XhoI sites of pET28a.
Expression and Purification of HpDAPDC and EcDAPDC-The recombinant plasmid pET28a-HpDAPDC was transformed into E. coli strain BL21 (DE3) for expression. The transformed cells were grown in LB media supplemented with 50 g/ml kanamycin at 37°C. When A 600 reached 0.6, the culture was induced with 0.4 mM isopropyl ␤-D-thiogalactopyranoside and incubated at 25°C for an additional 5 h. The cells were harvested by centrifugation at 6000 ϫ g for 15 min at 4°C and resuspended in buffer A (20 mM MES, pH 6.0, 500 mM NaCl, 200 M PLP, and 10 mM imidazole). The mixture was sonicated on ice for 30 min and centrifuged at 16,000 ϫ g for 45 min at 4°C. The supernatant was loaded onto a column filled with nickel-nitrilotriacetic acid resin (Qiagen) pre-equilibrated in buffer A. The column was washed with 100 ml of buffer B (20 mM MES, pH 6.0, 500 mM NaCl, and 20 mM imidazole) and then eluted with 10 ml of buffer C (20 mM MES, pH 6.0, 500 mM NaCl, and 500 mM imidazole). The eluent was dialyzed against buffer D (20 mM MES, pH 6.0, 200 M PLP, and 150 mM NaCl) for subsequent enzymatic assay and crystallization. Same protocols were used for the expression and purification of EcDAPDC and HpDAPDC mutants. Protein concentration was determined by Bradford assay using bovine serum albumin as the standard.
Enzymatic Assay-The enzymatic activities of HpDAPDC and EcDAPDC were measured by using DAPDC/saccharopine dehydrogenase (EC 1.5.1.7)-coupled spectrophotometric assay system as reported before (18). In brief, 100 l of reaction buffer containing 100 mM Tris, pH 8.0, 10 mM ␣-ketoglutarate, 0.31 mg/ml NADH (A 340 ϭ 1.0), 10 M saccharopine dehydrogenase, 10 M PLP, and DAPDC (50 -8000 nM, depending on the catalytic efficiency of the enzyme) was added to the 96-well plate, and then 100 l of DAP solution at desired concentration was supplemented to start the reaction. The initial velocity of the reaction was reflected by the rate of NADH consumption monitored by a Benchmark Plus Microplate spectrophotometer from Bio-Rad. The kinetic constants were obtained by Lineweaver-Burk double-reciprocal plots.
Crystallization-All crystallization experiments were performed by hanging drop vapor diffusion method at 4°C. Initial crystallization trials of apo-HpDAPDC failed. Thus, 10 mg/ml HpDAPDC solution was incubated with 5 mM L-lysine on ice overnight, and then 1 l of the complex was mixed with 1 l of the mother liquor containing 100 mM Bicine, pH 8.0, and 1.6 M ammonium sulfate, and equilibrated with 500 l of the same solution. Crystals grew to 0.1 ϫ 0.1 ϫ 0.2 mm after 1 week. The crystals of I148L were obtained in the same condition as the wild type.
Data Collection and Structure Determination-The crystals of HpDAPDC were cryoprotected in the mother liquor supplemented with 30% glycerol and flash-frozen in liquid nitrogen. Diffraction data were collected in-house at Ϫ180°C from a Rigaku rotating-anode x-ray generator (wavelength 1.5418 Å) operated at 100 mA and 100 kV. Images were recorded by a Rigaku R-Axis IVϩϩ image plate with an oscillation step of 1°. The data from wild-type HpDAPDC crystal was processed by CrystalClear software from Rigaku. The data from I148L crystal was indexed and integrated by Mosflm (19) and scaled by Scala (20). Statistics of the diffraction data are summarized in Table 1.
The structure of HpDAPDC was solved by molecular replacement with Molrep (21) using the monomer structure of MjDAPDC (PDB code 1TUF, chain A) as the search model. Structure refinement was carried by CNS (22) using standard protocols (simulated annealing, energy minimization, and B-factor refinement). Model building was performed in O (23) and Coot (24). The quality of the final model was checked by PROCHECK (25). The statistics of structure refinement are summarized in Table 1. The coordinates of the wild type and I148L HpDAPDC have been deposited in Protein Data Bank with accession numbers of 2QGH and 3C5Q, respectively. Structural superpositions were calculated with LSQKAB (26) from the CCP4 software suite (27). All structure figures were prepared by PyMOL (28).

RESULTS AND DISCUSSION
Overall Structure-As a classic group IV PLP enzyme, HpDAPDC consists of an N-terminal 8-fold ␣/␤-barrel and a C-terminal ␤-strand domain (Fig. 1b). A dimer of HpDAPDC could be reconstituted by crystallographic 2-fold symmetry (Fig. 1c). The two protomers adhere to each other in a "headto-tail" manner burying about 19% (3533 Å 2 ) of their surface area, confirming that the functional unit of DAPDC is an obligate dimer (13). The PLP molecule together with the covalently bonded L-lysine are positioned unambiguously in the well defined electron density map (Fig. 1d), indicating the location of the active site. Two catalytic pockets line at the dimer interface, each formed by a cavity from the N-terminal domain of one protomer and a protruded surface from the C-terminal domain of the other protomer.
The overall folding of HpDAPDC is similar to other known DAPDC structures with small r.m.s.d. values superposing either monomeric or dimeric forms (Table 2). However, the N terminus of HpDAPDC is substantially shorter and lacks the secondary structure elements usually found at this terminus (4,13). Besides, the active pocket of HpDAPDC is covered by a loop and completely shielded from the solvent (Fig. 1c). Structural superposition of HpDAPDC with other typical group IV PLP enzymes (ornithine decarboxylase from Trypanosoma brucei (TbODC) (11, 29 -31), arginine decarboxylase from Paramecium bursaria chlorella virus (cvADC) (12), and alanine racemase from Bacillus stearothermophilus (BsALR) (9, 32)) reveals that the HpDAPDC is more structurally similar to ADC and ODC than to ALR (Table 2), especially in the strand pattern and orientation of the C-terminal domain.
PLP-binding Site-The cofactor PLP binds in a deep cavity at the mouth of the N-terminal ␣/␤-barrel ( Fig. 1b) with nearly all contacting residues from the parental protomer (Fig. 2a). The pyridoxal ring is stacked in a shallow crevice, and the phosphate moiety is inserted in a highly negatively charged hole (Fig. 2c, upper panel). The si face of the pyridoxal ring is packed against His 185 , which is conserved among all group IV PLP enzymes (alignment not shown). The decarboxylation of DAP is believed to happen at this side of PLP (13). The phenolate oxygen (O-3) on the PLP ring is recognized by the highly conserved residues Arg 135 and Cys 329 . The interaction between the PLP pyridinium nitrogen and the two acidic residues Asp 65 and Glu 259 is thought to keep PLP protonated to enhance its "electron sink" nature (7). Upon the formation of the L-lysine-PLP external aldimine, Lys 46 that originally makes Schiff base with the cofactor (internal aldimine) now points away from PLP and is stabilized by Asp 65 . A similar ion pair between a broken Schiff base lysine and an asparagine residue is also observed in ALR (Lys 39 -Asp 313 ) (9,33), which is believed to be a naturally occurring conformation during catalysis.
The phosphate group of PLP makes numerous hydrogen bonds, either directly or via water molecules, with residues from conserved motifs, including HIGS (residues 185-188), the glycine-rich region (residues 224 -226), EPGRS (residues 259 -263), and GAY (residues 356 -358) motifs (Fig. 2a). Notably, where I hi and ͗I h ͘ are the i-th and mean measurement of the intensity of reflection h, respectively. two PLP-binding residues (Arg 262 and Tyr 358 ) are also involved in recognizing the L-stereocenter of the substrate.
Substrate-binding Site-The external aldimine conformation of HpDAPDC provides valuable structural information on the substrate binding and the catalytic mechanism of DAPDC.
General Description-The substrate DAP is a meso compound with two stereocenters, namely the L-stereocenter and the D-stereocenter (L-center and D-center for short). DAPDC specifically cleaves the D-center carboxyl to give the product L-lysine. This specificity comes from both the recognition of the D-center and the proper anchoring of the L-center in the enzyme. Therefore, we will dissect the substrate DAP into three parts, the aliphatic chain and the two stereocenters, and describe their interactions with the enzyme separately.
The aliphatic chain of the L-lysine is sandwiched between the side chains of Ile 148 and Tyr 358 (Fig. 2b), which constitute the "ceiling" and "floor" of the active pocket, respectively. The ceiling, Ile 148 , is in a small 3 10 helix ␣9 from the active site loop and is nonconserved among DAPDCs. The floor, Tyr 358 , is highly conserved and plays a dual role in both substrate recognition and PLP binding. Besides the sandwich structure, the alkyl moiety of the side-chain of Glu 330 B ("B" denotes the residue from another protomer, same below) also makes hydrophobic interaction with the aliphatic chain of L-lysine (Fig. 2b).
The L-center of the L-lysine is wrapped by several charged and polar residues electrostatically complementary with the carboxyl and amine groups of the substrate. In detail, Arg 262 and Arg 298 line beside the L-center carboxyl with their sidechain nitrogens salt-bridging to the carboxyl oxygens (Fig. 2b). These two basic residues constitute a positively charged inner wall at one side of the active pocket (Fig. 2c, upper panel). The other side of the pocket, however, is mainly negative (Fig. 2c,  lower panel), composed of Tyr 302 , Tyr 366 B, and Glu 330 B, which hydrogen bond to the L-center amine of the substrate (Fig. 2b). This side of the pocket is less charged than the opposite side (Fig. 2c), because part of the side-chain of Glu 330 B is hydrophobically interacting with the aliphatic chain of L-lysine. All residues mentioned above are absolutely conserved among all known DAPDC sequences, indicating their significance in substrate binding. Notably, Arg 298 and Tyr 302 come from a small helix ␣14 referred to as the "specific element" of group IV PLP-  dependent decarboxylases. The position of this helix varies across enzymes to accommodate substrates of different sizes (6,12). Besides this specificity element, we believe the two oppositely charged inner walls of the active site of the DAPDC also determine the specificity of the enzyme for DAP, because other group IV PLP-dependent decarboxylases carry only negative charge at the corresponding area. Thus, this distinct property of DAPDC might offer a chance to design specific DAPDC inhibitors. The D-center carboxyl of DAP is very likely to locate at the si face (the face toward solvent) of PLP in DAPDC, because superposition of DAP onto the bound L-lysine unambiguously places its D-center carboxyl into the only room left in the active pocket, which is above the si face of PLP (Fig. 2d). Several previous studies also support this conclusion. First, the Dunathan stereoelectronic hypothesis states that the bond to be broken by PLP-dependent enzymes lies orthogonal to the system of the cofactor (34). Second, the crystal structure of MtDAPDC complexed with L-lysine suggests that the decarboxylation occurs at the si face of PLP (13). Third, mechanistic and structural studies of TbODC indicate that the carboxyl-leaving group of L-ornithine is buried on the re face of PLP (35), whereas the CO 2 of D-ornithine locates at the si face (31). Attempt to reverse the orientation of DAP results in both electrostatic repulsion and steric hindrance, indicating that the substrate specificity of DAPDC is determined by both the shape and charge property of the active pocket. Based on these observations, the substrate DAP is modeled into the active site of HpDAPDC to provide mechanistic insights.
Putative Mechanism of the Recognition and Decarboxylation of the D-Center Carboxyl-In the modeled HpDAPDC-DAP structure, the D-center carboxyl of DAP could be recognized by several residues via water molecules, including the absolutely conserved PLP-binding residues Arg 136 (36) and His 185 and the active site loop residues Ile 148 and Thr 150 (Fig. 3a). Although the water molecule "Wx" is not included in the submitted coordinate because of its relatively weak but detectable F o Ϫ F c signal, this molecule could probably be stabilized in the presence of the D-center carboxyl). Interestingly, superposition of HpDAPDC-DAP with the BsALR-L-Ala-P (substrate analogue) complex structure (32) reveals that the substrate carboxyl groups (or the phosphonate analogue), the highly conserved arginine residues involved in carboxyl recognition, and the catalytic active lysine residues are all overlaid perfectly well (Fig. 3a), indicating that these two enzymes, although with distinctive functions and overall structures, might share some common features in substrate recognition and catalytic mechanism.
A putative decarboxylation mechanism of DAPDC could also be derived. Two residues lie within 4.5 Å from the D-center C-␣ of DAP, one is the original Schiff base residue Lys 46 , and the other is the highly conserved Cys 329 at the re side of PLP (Fig.  3a). It has been reported that in eukaryotic ornithine decarboxylases (ODC), the analogues of these two residues, Lys 69 and Cys 360 , play multiple roles in catalysis (30,37). Notably, Cys 360 is the target of ␣-difluoromethylornithine, a drug treating African sleeping sickness (11,31). Thus, it is plausible that Lys 46 and Cys 329 would have similar catalytic functions in HpDAPDC. In detail, Lys 46 might catalyze the formation and hydrolysis of the Schiff base, as well as carboxyl cleavage (37). Cys 326 , with its sulfhydryl group only 4.3 Å away from the D-center C-␣, might be involved in carbanion protonation, Schiff base formation, and decarboxylation (30).
Comparison of the Substrate-binding Sites among Group IV PLP-dependent Decarboxylases-Although the group IV PLPdependent decarboxylases all work on basic amino acids with FIGURE 2. Active site structure of HpDAPDC. a, PLP-binding site of HpDAPDC. Oxygen, nitrogen, phosphorus, and sulfur atoms are colored red, blue, orange, and dark yellow, respectively, and hydrogen bonds are presented as dotted lines (same below). The carbons of ligand and protein are colored cyan and yellow, respectively. b, substrate-binding site of HpDAPDC. The color scheme is the same as a except that the residues from another protomer (with letter "B" in their names) are colored green. c, electrostatic potential distribution on the active site inner walls, colored red to blue representing Ϫ20 e/kT to ϩ20 e/kT. The potential is calculated by Adaptive Poisson-Boltzmann Solver (42). d, active pocket viewed by looking into the interior of HpDAPDC dimer. The substrate DAP is modeled into the pocket (lower panel) by being superposed onto the bound L-lysine (upper panel). The white circles denote the space to accommodate the D-stereocenter carboxyl of DAP. similar mechanisms, they have evolved distinctive specificity determinants to recognize their own substrates.
Before comparison, we overlaid all known DAPDC structures and confirmed that the substrate-binding residues from HpDAPDC are representative of DAPDCs (Fig. 3b), which is consistent with the sequence alignment (Fig. 4). Thus, we will take HpDAPDC, TbODC (PDB code 1F3T) (30), and cvADC (PDB code 2NVA) (12) to carry out the comparison. Notably each structure contains a product-PLP external aldimine.
Because of the different length and chirality of the substrates (Fig. 1a), the inner walls of the active pockets are substantially different among the three enzymes. Accordingly, we will "cut" the active pocket of HpDAPDC into halves, namely the "DAPamine side" and the "DAP-carboxyl side," and compare them with the corresponding regions of ODC and ADC separately.
As shown in Fig. 3c, at the DAP-amine side, the arginine residues recognizing PLP and probably the D-center carboxyl (Arg 135 /DAPDC, Arg 154 /ODC, and Arg 133 /ADC), the catalytic active cysteine residues (Cys 329 B/DAPDC, Cys 360 B/ODC, and Cys 324 B/ADC), and the acidic residues lining beside the substrate (Glu 330 B/DAPDC, Asp 361 B/ODC, and Asp 325 B/ADC) are all structurally conserved, indicating the similar catalytic mechanism shared by these decarboxylases. The difference occurs at the PLP distal site. In detail, the glutamate residue (Glu 330 B) of DAPDC is substituted by aspartate residues in ODC and ADC (Asp 361 B/ODC and Asp 325 B/ADC), probably to interact with the more PLP-proximal nitrogen atoms of ornithine and arginine. Moreover, Tyr 366 B of DAPDC shift toward PLP compared with Tyr 323 B/ODC and Tyr 287 B/ADC, probably to avoid electrostatic repulsion with the L-center carboxyl of DAP. In addition, the two tyrosine residues of ODC and ADC are stabilized by forming hydrophobic interaction with nearby residues (Phe 397 B, Tyr 331 B of ODC and Phe 361 B, Leu 357 , and Phe 295 of ADC). Such stabilization effect is absent in HpDAPDC. Finally, as has been reported before (6,12), the positions of "specificity element" helices are different in these enzymes, and only the helix of DAPDC interacts with the substrate (Tyr 302 -DAP amine) at this side.
Greater difference occurs at the DAP-carboxyl side (Fig. 3d). The only constant structure element at this side is the floor tyrosine beneath the ligand (Tyr 358 /DAPDC, Tyr 389 /ODC, and Tyr 353 /ADC). The ceiling of the pocket, however, exists only in DAPDC and ADC (Ile 148 /DAPDC and Val 143 /ADC). Moreover, the charge property of this side varies greatly among the three enzymes. In detail, at the center, Arg 262 /DAPDC is extending its side chain toward DAP to make a salt bridge with the L-center carboxyl, whereas Arg 277 /ODC and Arg 255 /ADC are retracting their side chains from the substrates probably to avoid electrostatic repulsion. In the specificity element helices, the basic residue in DAPDC (Arg 298 ) is substituted by asparagine (Asn 292 /ADC) and/or acidic residues (Glu 296 /ADC and Asp 332 /ODC).
Considering the extreme diversity of the reactions catalyzed by PLP enzymes and the similar working mechanism shared by closely related PLP enzymes, the structural difference we described above might offer an opportunity to design selective DAPDC inhibitors for development of antibiotics.
Catalytic Role of the Active Site Loop-The active site loop of HpDAPDC (Asn 137 to Asn 155 ) is in a down conformation that completely shields the ligand from the bulk solvent. There is no crystal contact at this site, indicating the natural occurrence of this conformation. Structure-based sequence alignment (Fig. 4) and superposition (Fig. 5, 1st panel) both indicate that the active site loop is the most divergent element of DAPDC across species. As a consequence, the active pockets in various DAPDC structures are solvent-exposed to different extents (Fig. 5).
We speculate that HpDAPDC adopts an induced-fit catalytic mechanism, in which its active site loop cycles through down and up conformations to stabilize catalytic intermediates and release reaction products, respectively. This hypothesis is based on the following evidence. First, a highly similar loop of AaDAPDC (58% sequence identity, Fig. 4, boxed region) is flexible (crystallographically unobservable) in the absence of ligands. Second, superposition (Fig. 3d) and sequence alignment (Fig. 4) between HpDAPDC and cvADC reveal two well aligned active site loops both in down conformations, and the movement of the cvADC loop is believed to be kinetically linked with product release (12). Third, the protease-sensitive loop (residues 158 -168) of human ODC has recently been shown to undergo large conformational change and cover the active site upon the binding of an inhibitor (38). Fourth, mutation of the active site loop in TbODC (K169A) resulted in a large decrease of the catalytic efficiency (39). In addition, we also carried out some biochemical experiments to verify our hypothesis.
The attempt to delete this loop failed because the truncated protein cannot express as soluble form, indicating the contribution of this loop to the overall stability of HpDAPDC. Thus, we mutated Ile 148 , the only ligand-contacting residue of this loop, to six different types of amino acid (Leu, Phe, Ala, Lys, Asp, and Gly) and found that all mutations caused great loss of the enzymatic activity. We also determined the crystal structure of I148L mutant. Detailed descriptions are as follows. Enzymatic Assay-The kinetic constants of HpDAPDC (wild type and mutants) and EcDAPDC were determined by the reported coupled enzymatic assay approach (18). As shown in Table 3, the K m value of EcDAPDC is about 1.1 mM, comparable with the reported value of 1.7 mM measured by a different method (40), indicating the reliability of our assay system. The k cat value of HpDAPDC is 939 min Ϫ1 , similar to the 1214 min Ϫ1 of EcDAPDC. However, the K m value of HpDAPDC (0.39 mM) is 2-fold smaller than that of EcDAPDC, probably because the substrate binding pocket of HpDAPDC is more compact.
The six mutations of Ile 148 , including the conservative substitution I148L, decreased the catalytic efficiency (k cat /K m ) of HpDAPDC by 5-500-fold (Table 3). Based on the extent of activity loss, the six mutants could be classified into three groups. The first group, I148L and I148F, lost about 80% of the activity. Intriguingly, although the k cat values of these two mutants were impaired (40-and 23-fold, respectively), their K m values were enhanced (10-and 4-fold), indicating that the enzyme-substrate (ES) complex formed during reaction is more stable for the mutants. This point is supported by the structural analysis of I148L shown in the next section. The second group, I148A and I148K, suffered about 50-fold activity decrease, which is mainly caused by the reduced k cat values. The third group, I148D and I148G, retains only 1/500 and 1/200 of the original activity, respectively. The k cat values of these two mutants both decreased by more than 200fold, whereas their K m values were close to that of the wild type. The 1-fold increase of the K m value for I148D might be caused by the electrostatic repulsion between Asp 148 and the L-center carboxyl of the substrate. This is consistent with the observation that the side chain of Ile 148 in wild-type structure lies closer to the L-center carboxyl group than to the L-center amine group, which also explains why the K m value of I148K was not affected. To further understand the deactivating effect of these mutations, we determined the crystal structure of I148L mutant.
Crystal Structure Analysis of I148L Mutant-The overall structure of I148L is basically the same as that of the wild type ( Table 2). The only obvious change is the side-chain substitution of residue 148. However, combining the previous study about the catalytic mechanism of the closely related enzyme TbODC (41), we still find some explanations for the following questions.
Why Did the Mutations Cause Relatively Small Change of the K m Value?-As has been discussed, HpDAPDC is likely to adopt an induced-fit catalytic mechanism, which means the active site loop "closes" only after the proper anchoring of the substrate. In other words, this loop should have limited contri-  Why Did the Mutations Greatly Decrease the k cat Value?-The active site loop is likely to be relevant with two k cat -determining processes. One is the stabilization of catalytic intermediates; the other is the release of the reaction product. Notably, the product release has been considered to be the rate-limiting step in the reaction catalyzed by TbODC (41). Mutations on the active site loop might interfere with both processes and result in decreased k cat values.
The mutations I148L, I148F, and I148K might have reduced the efficiency of product release. As shown in Fig. 6a, in the wild-type structure, the specificity element helix ␣14 is closely opposite to the small 3 10 helix ␣9 of the active site loop, and Tyr 302 from ␣14 is hydrophobically packing against Ile 148 from ␣9 with a buried surface area of 62.7 Å 2 . Thus ␣14 acts like a "buckle" to fasten the active site loop in its down conformation, and the affinity between Tyr 302 and residue 148 is likely to determine the efficiency of loop opening and product release. In the structure of I148L mutant (Fig. 6b), however, the buried surface area between Leu 148 and Tyr 302 becomes larger (85.8 Å 2 ), which might hinder the opening of the loop and product release. This also implies that the ES complex is probably more stable in I148L, resulting in enhanced K m value. Similar explanations could be applied to I148F and I148K.
In contrast, the mutations I148A, I148G, and I148D might have compromised the ability of the enzyme to stabilize catalytic intermediates. First, no direct interaction (I148A and I148G), or even repulsion (I148D), could be made between the active site loop and the catalytic intermediates. Second, the affinity between the mutated residue (alanine, glycine, or asparagine) and Tyr 302 might be weakened, destabilizing the down conformation of the active site loop. Thus, an isoleucine is probably the most appropriate residue at this position to delicately balance the close-to-stabilize and open-to-release processes of the active site loop.
As reflected by the mutagenesis study, altering the assumed dynamic behavior of this active site loop could have profound impact on the catalytic efficiency of HpDAPDC. Thus, restricting the movement of this loop could probably be an alternative strategy for designing specific HpDAPDC inhibitors.
Conclusion-The crystal structure of HpDAPDC complexed with the cofactor PLP and the reaction product L-lysine reveals the conformation of the catalytic intermediate (external aldi-mine), which provides insight into the substrate binding and catalytic mechanism of DAPDC. The aliphatic chain, the L-stereocenter carboxyl, and amine groups of the bound L-lysine are all properly anchored in the active pocket of HpDAPDC. The D-stereocenter carboxyl of the substrate DAP is very likely to locate at the si face of PLP. The recognition and cleavage of this carboxyl group might involve some highly conserved residues. The charge property of the active pocket of DAPDC is distinct from other group IV PLP-dependent decarboxylases, which provides structural clues for designing specific DAPDC inhibitors. The active site loop of HpDAPDC (Asn 137 to Asn 155 ) is in a down conformation and completely shields the ligand from solvent. Mutagenesis study and structural analysis suggest that HpDAPDC might adopt an induced-fit catalytic mechanism, in which this loop cycles through down and up conformations to stabilize catalytic intermediates and release the reaction product, respectively.