Crystal Structure of Novel Dye-linked l-Proline Dehydrogenase from Hyperthermophilic Archaeon Aeropyrum pernix*

Background: The crystal structure of a novel type of dye-linked l-proline dehydrogenase (LPDH) from hyperthermophilic archaea was determined. Results: The C-terminal Leu428 shielded the active site cavity in which the substrate (l-proline) molecule was entrapped. Conclusion: The C-terminal Leu428 was proposed to be essential for maintaining a high affinity for the substrate. Significance: This is the first description of LPDH structure bound with l-proline. Two types of dye-linked l-proline dehydrogenase (PDH1, α4β4-type hetero-octamer, and PDH2, αβγδ-type heterotetramer) have been identified so far in hyperthermophilic archaea. Here, we report the crystal structure of a third type of l-proline dehydrogenase, found in the aerobic hyperthermophilic archaeon Aeropyrum pernix, whose structure (homodimer) is much simpler than those of previously studied l-proline dehydrogenases. The structure was determined at a resolution of 1.92 Å. The asymmetric unit contained one subunit, and a crystallographic 2-fold axis generated the functional dimer. The overall fold of the subunit showed similarity to that of the PDH1 β-subunit, which is responsible for catalyzing l-proline dehydrogenation. However, the situation at the subunit-subunit interface of the A. pernix enzyme was totally different from that in PDH1. The presence of additional surface elements in the A. pernix enzyme contributes to a unique dimer association. Moreover, the C-terminal Leu428, which is provided by a tail extending from the FAD-binding domain, shielded the active site, and an l-proline molecule was entrapped within the active site cavity. The Km value of a Leu428 deletion mutant for l-proline was about 800 times larger than the Km value of the wild-type enzyme, although the kcat values did not differ much between the two enzymes. This suggests the C-terminal Leu428 is not directly involved in catalysis, but it is essential for maintaining a high affinity for the substrate. This is the first description of an LPDH structure with l-proline bound, and it provides new insight into the substrate binding of LPDH.

Two types of dye-linked L-proline dehydrogenase (PDH1, ␣4␤4-type hetero-octamer, and PDH2, ␣␤␥␦-type heterotetramer) have been identified so far in hyperthermophilic archaea. Here, we report the crystal structure of a third type of L-proline dehydrogenase, found in the aerobic hyperthermophilic archaeon Aeropyrum pernix, whose structure (homodimer) is much simpler than those of previously studied L-proline dehydrogenases. The structure was determined at a resolution of 1.92 Å. The asymmetric unit contained one subunit, and a crystallographic 2-fold axis generated the functional dimer. The overall fold of the subunit showed similarity to that of the PDH1 ␤-subunit, which is responsible for catalyzing L-proline dehydrogenation. However, the situation at the subunit-subunit interface of the A. pernix enzyme was totally different from that in PDH1. The presence of additional surface elements in the A. pernix enzyme contributes to a unique dimer association. Moreover, the C-terminal Leu 428 , which is provided by a tail extending from the FAD-binding domain, shielded the active site, and an L-proline molecule was entrapped within the active site cavity. The K m value of a Leu 428 deletion mutant for L-proline was about 800 times larger than the K m value of the wild-type enzyme, although the k cat values did not differ much between the two enzymes. This suggests the C-terminal Leu 428 is not directly involved in catalysis, but it is essential for maintaining a high affinity for the substrate. This is the first descrip-tion of an LPDH structure with L-proline bound, and it provides new insight into the substrate binding of LPDH.
Dye-linked dehydrogenases belong to a group of oxidoreductases that catalyze the oxidation of various organic acids, amino acids, and alcohols in the presence of an artificial electron acceptor, such as 2,6-dichloroindophenol (Cl2Ind) 2 or potassium ferricyanide. In general, dye-linked dehydrogenases are flavoproteins that use FAD or FMN as a coenzyme and function as mediators of electron transfer from a reduced substrate into the electron transfer system used to produce energy for the cell (1,2). These enzymes are most often membrane-associated or membrane-bound and form complexes containing several protein components (3,4). Once solubilized from the cell membrane, the enzymes from mesophiles are not sufficiently stable to enable a high degree of purification. This is unfortunate because they could be highly useful, serving, for example, as elements in electrochemical biosensors (5). Despite their important functions in electron transport systems and their potential for practical application, information about the structure and function of dye-linked dehydrogenases remains limited because of the aforementioned instability.
In recent years, we have been studying the structure, function, and potential application of dye-linked dehydrogenases from hyperthermophiles, which are more stable than those from mesophiles or eukaryotes. So far, we have identified several dye-linked dehydrogenases from hyperthermophilic archaea, including dye-linked L-proline dehydrogenase (LPDH) (6 -8), dye-linked D-proline dehydrogenase (9), and dye-linked D-lactate dehydrogenase (10), all of which are highly stable and totally novel flavin-containing dehydrogenases. Among these enzymes, LPDH catalyzes the oxidation of L-proline to ⌬ 1 -pyrroline-5-carboxylate (P5C) with reduction of Cl2Ind (Fig. 1). Two different types of LPDH, PDH1 and PDH2, have been identified in the anaerobic hyperthermophile Pyrococcus horikoshii OT-3 (7). PDH1 is a hetero-octameric complex (␣4␤4; molecular mass, 440 kDa) that also contains FAD, FMN, Fe 3ϩ , and ATP, whereas PDH2 is a heterotetrameric complex (␣␤␥␦; molecular mass, 120 kDa) composed of an L-proline dehydrogenase, an NADH dehydrogenase, a ferredoxin-like protein, and a protein of unknown function (6,7). Using x-ray crystallography, we solved the three-dimensional structure of the PDH1 complex (11), which we found to be a unique diflavin dehydrogenase containing a novel electron transfer system that is totally different from that of the PDH2 complex. Proteins homologous to PDH1 and PDH2 are widely distributed among the anaerobic hyperthermophilic archaea that belong to the phylum Euryarchaeota (6,7).
We also recently identified LPDHs from two aerobic hyperthermophilic archaea, Pyrobaculum calidifontis (12) and Aeropyrum pernix (13), belonging to the phylum Crenarchaeota. The newly identified LPDHs are homodimers with a subunit molecular mass of 46 kDa. They are considerably less homologous (24 -28%) than the PDH1 and PDH2 ␤-subunits, which are directly responsible for catalyzing L-proline dehydrogenation. This suggests the dimeric LPDHs are inherently different from PDH1 and PDH2 and may represent a new family of LPDHs. The physiological function of the aerobic archaeal LPDHs is currently unknown. They may function as the mediators of electron transfer from a reduced substrate into the electron transfer system as supposed for PDH1. The structural analysis of these enzymes could shed light on the evolution of energy and amino acid metabolism in hyperthermophilic archaea. Here, we describe the unique homodimeric structure of the A. pernix LPDH (ApeLPDH), a novel type of archaeal LPDH.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Expression of ApeLPDH was carried out as described previously (13), except the gene encoding the enzyme (APE_1267.1) was ligated with the expression vector pET15b (Novagen, Madison, WI) to generate p15LPDH. For expression of selenomethionyl-ApeLPDH (SeApeLPDH), Escherichia coli cells (BL21 (DE3) Codon Plus RILP; Stratagene, La Jolla, CA) transformed with p15LPDH were cultivated in modified M9 medium containing selenomethionine (14). The cells were then harvested by centrifugation and washed twice with a 0.85% NaCl solution. The washed cells were suspended in a 10 mM potassium phosphate buffer (pH 7.2) containing 100 mM NaCl and then disrupted by ultrasonication. The crude extract was heated at 80°C for 10 min, and the denatured protein was removed by centrifugation (10,000 ϫ g for 10 min). The resulting supernatant was loaded into an Ni 2ϩ -charged chelating Sepharose column (2.6 ϫ 10 cm; GE Healthcare) that had been equilibrated with the 10 mM potassium phosphate buffer (pH 7.2) containing 100 mM NaCl. The column was then washed first with the same buffer and then washed again with 3 bed volumes of the same buffer supplemented with 0.1 M imidazole. The enzyme was eluted with 6 bed volumes of a linear gradient of 100 -500 mM imidazole in the same buffer. The active fractions were pooled and concentrated by ultrafiltration (Amicon Ultra 30K NMWL; Millipore, Billerica, MA). The resulting enzyme solution was applied to a Sephacryl S-300 column (2.6 ϫ 80 cm; GE Healthcare) that had been previously equilibrated with the 10 mM potassium phosphate buffer (pH 7.2) and then eluted with the same buffer. The active fractions were pooled and concentrated by ultrafiltration (Amicon Ultra 30K NMWL), and the resulting solution was used for crystallization.
Site-directed Mutagenesis-Site-directed mutagenesis was accomplished using a QuikChange XL site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions. To construct a deletion mutant lacking C-terminal Leu 428 (⌬L428), the expression vector for His-tagged ApeLPDH (p15LPDH) served as the template, and the following pair of oligonucleotide primers were used as the mutagenic primers (the mutations are underlined): 5Ј-CAGGAGAGGCTAGTCTAGACGCCGAGGATC-3Ј and 5Ј-GATCCTCGGCGTCTAGACTAGCCTCTCCTG-3Ј. Expression and purification of the ⌬L428 were performed using the same method described for His-tagged ApeLPDH.
To construct the PDH1␤ mutants, the expression vector for His-tagged PDH1␤ (pET15b/␤1) (7) served as the template, and the following pairs of oligonucleotide primers were used as the mutagenic primers: 5Ј-GGTGCGGTACTGGAATAGCACAA-CAATTCAATGATG-3Ј and 5Ј-CATCATTGAATTGTTG-TGCTATTCCAGTACCGCACC-3Ј for R52A, and 5Ј-CCTT-CAAACAAACAGGTTTCCTATTCCTACTTTACG-3Ј and 5Ј-CGTAAAGTAGGAATAGGAAACCTGTTTGTTTGAAGG-3Ј for Y87F. The E. coli BL21-CodonPlus(DE3)-RIPL cells (Stratagene) were used for the expression of the PDH1␤, R52A, and Y87F genes. All transformants were grown for 9 h at 37°C in Luria-Bertani medium (1 liter) in the presence of ampicillin (0.01%). The wild-type and mutant enzymes were purified by using the same method. The cells were collected by centrifugation and suspended in 10 mM Tris-HCl (pH 8.0) and, then disrupted by sonication. After centrifugation, imidazole and NaCl were added to the crude extract up to 50 mM and 0.5 M, respec- tively. The enzyme solution was applied on a HiTrap nickelcharged chelating column (2.6 ϫ 6 cm; GE Healthcare) equilibrated with 10 mM Tris-HCl (pH 8.0) containing 50 mM imidazole and 0.5 M NaCl. The column was washed with the same buffer, and the enzyme was eluted by a linear gradient up to 500 mM imidazole in the buffer. The enzyme-containing fractions were checked by SDS-PAGE and collected. The resulting solution was dialyzed against 10 mM Tris-HCl (pH 8.0) containing 0.2 M NaCl and was used as the purified preparation.
Determination of Enzyme Activity and Protein Concentration-LPDH activity was assayed by measuring the rate of Cl2Ind reduction at 50°C using a Shimadzu UV-120-02 spectrophotometer as described previously (12). All measurements were performed in duplicate, and if the discrepancy between the results was greater than 10%, additional reactions were carried out. Protein levels were determined using a BCA protein assay reagent kit supplied by Thermo Fisher Scientific, with bovine serum albumin serving as the standard.
Crystallization and Data Collection-Crystals of SeApeLPDH were obtained using the sitting-drop vapor diffusion method, in which a 1-l drop of 12.0 mg/ml protein solution was mixed with an equal volume of mother liquor composed of 0.2 M ammonium acetate, 0.1 M trisodium citrate dihydrate (pH 5.6), and 30% polyethylene glycol 4000. The crystals were grown for 6 days at 20°C and belong to the tetragonal space group P4 3 2 1 2 with unit cell parameters a ϭ b ϭ 60.2 and c ϭ 274.2 Å. Crystals of ⌬L428 mutant enzyme were grown in sitting drops composed of 1 l of enzyme solution (10 mg/ml) mixed with 1 l of mother liquor containing 0.1 M magnesium acetate, 0.1 M MES (pH 6.5), and 12% polyethylene glycol 8000. The crystals were grown for 30 days at 20°C and belong to the triclinic space group P1 with unit cell parameters a ϭ 48.0, b ϭ 70.6, and c ϭ 73.1 Å and ␣ ϭ 71.7, ␤ ϭ 70.8, and ␥ ϭ 70.1°. Selenium multiple-wavelength anomalous dispersion data for SeApeLPDH and single wavelength (1.0 Å) data for ⌬L428 were collected using an Area Detector Systems Corporation CCD detector system on the BL5A beamline at the Photon Factory, Tsukuba, Japan. All measurements were carried out on crystals cryoprotected with 20% (v/v) ethylene glycol and cooled to 100 K in a stream of nitrogen gas. The data were processed using HKL2000 (15).
Phasing and Refinement-SOLVE (16) was used to refine the parameters of the selenium atoms of SeApeLPDH and to calculate the phases. The initial phase was improved by solvent flattening, followed by autotracing using RESOLVE (17). The model was built using Coot (18) and Xtal View (19). Refinement of the model structure was carried out using Refmac5 (20) and CNS (21). Then after several cycles of inspection of the 2F o Ϫ F c and F o Ϫ F c density maps, the model was rebuilt. The FAD, L-proline, and ethylene glycol molecules were clearly visible in both the A -weighted 2F o Ϫ F c and F o Ϫ F c density maps and were included in the latter part of the refinement. Water molecules were incorporated using Coot (18). The model geometry was analyzed using RAMPAGE (22). The R-factor and R free values for the final model were 18.8 and 20.5%, respectively (Table  1). Hydrogen bonds were identified using the program CCP4mg (23). Ion pairs (the cutoff distance was 4.0 Å (24)) and hydrophobic interactions were identified using the WHAT IF web server (25). To determine the number of hydrophobic interactions, the interatomic contacts between the atoms of the hydrophobic side chains were calculated. A contact was defined as two atoms for which the distance between the van der Waals surfaces was less than 1.0 Å. The accessible surface area was calculated using AREAIMOL in the CCP4 program suite (26). Molecular graphics figures were created using PyMOL.
The initial phases of the ⌬L428 mutant enzyme structure were determined using the program MOLREP in the CCP4 program suite (26); the structure of chain A in SeApeLPDH served as the search model. Further model building was performed with Coot (18), and refinement to resolution 2.01 Å was carried out using Refmac5 (20). Two FAD, two acetate, and four ethylene glycol molecules were included in the refinement. The data collection and refinement statistics are listed in Table 1.
Mass Analysis by Fourier Transform Ion Cyclotron Mass Spectrometer-The purified SeApeLPDH was diluted and denatured in 90% acetonitrile and 1.0% acetic acid for 30 min in room temperature. The solution (20 M) was introduced into a solariX, Fourier transform ion cyclotron (FT-ICR) mass spectrometer (MS) shielded with 9.4-tesla superconducting magnets (Bruker Daltonics Inc., MA) by electrospray ionization under positive mode with 120 l/h flow rate, 220°C drying gas temperature, 4500 V capillary voltage, and Ϫ500 voltage spray shield. L-Proline (10 M) and DL-P5C (Sigma; 20 M) in the same acidic solvent were employed as controls for substrate and product, respectively. For data acquisition at solariX FT-ICR MS, free induction decay was set to 1 M size and mass detection range was set from 50 to 1000 m/z. One hundred mass spectra were combined to obtain an average spectrum result.

RESULTS AND DISCUSSION
Architecture of the Subunit-The structure of SeApeLPDH was determined using selenium multiple-wavelength anomalous dispersion and was refined at a resolution of 1.92 Å ( Table  1). The asymmetric unit consisted of one subunit with a solvent content of 51.5%, which corresponds to a Matthews coefficient (27) of 2.5 Å 3 Da Ϫ1 . In the present model, the amino acids extending from Se-Met Ϫ19 to His 0 were disordered and not visible in the electron density map. These residues are derived from the expression vector pET15b including His-tag sequence followed by a thrombin cleavage site.
Structural Comparison of Monomeric ApeLPDH and PDH1␤-The main chain coordinates of the ApeLPDH monomer were essentially the same as those of the PDH1 ␤-subunit (PDH1␤) (Protein Data Bank code 1Y56-B) (11). Superposition of PDH1␤ onto ApeLPDH yielded a root mean square deviation (r.m.s.d.) of 2.53 Å for the equivalent C␣ atoms from 342 resi-dues (Fig. 4). However, we found that several surface elements in ApeLPDH were markedly larger than the corresponding parts in PDH1␤. A large surface loop (L-1, residues 192-206) and an "arm" (A-1, residues 131-145) containing ␣6 in ApeLPDH were not observed in the corresponding regions of PDH1␤ (Figs. 3 and 4). Likewise, ␣9 in ApeLPDH was also absent from PDH1␤ (Figs. 3 and 4). Another noteworthy difference was observed in the structure of the C-terminal chain. In PDH1␤, the chain (residues 356 -375) contains only one helix (␣11) and the C-terminal seven residues (TAALQMG, 376 -382) are disordered because of poor electron density (Fig. 5B). By contrast, the C-terminal chain (residues 400 -428) in ApeLPDH contains ␤17, ␤18, and ␣13, and a C-terminal tail (residues 415-428) extends from the FAD-binding domain toward the substrate-binding site (Fig. 5A), providing an important component (the C-terminal Leu 428 ) to the active-site shield (see below).
Structural Comparison of Oligomeric ApeLPDH and PDH1-Although the main chain coordinates of the ApeLPDH monomer were similar to those of PDH1␤, the situation at the subunit-subunit interface of the ApeLPDH dimer was totally different from that at the interface of the PDH1 heterodimer (␣␤). The PDH1 hetero-octamer reportedly assembles as a tetramer of the ␣␤-heterodimer, within which one (␣␤) 2 block generated by an intermolecular disulfide bridge (formed between the two ␣-subunits) binds with another at the back to form the (␣␤) 4 -hetero-octamer (Fig. 6A) (11). Within the hetero-octamer, ␤-subunits interact only with ␣-subunits, and one molecule of FMN is located at the interface between the ␣and ␤-subunits (Fig. 6B). The isoalloxazine ring of FMN is sandwiched by two hydrophobic residues, ␤Trp 304 and ␣Met 444 , and the phosphate group is fixed by three basic residues, ␤Arg 46 , ␤Arg 302 , and ␣Arg 477 (11). The ␤-subunit binds to the ␣-subunit mainly via hydrophobic interactions between residues belonging to ␣8, ␣9, ␤10, and ␤14 in the ␤-subunit and  those belonging to the surface loops in the ␣-subunit, in addition to interactions between the FMN molecule and the protein (Fig. 6B) (11). By contrast, the intersubunit hydrophobic interactions within ApeLPDH are mainly between the residues around L-1, A-1, ␣8, ␣10, ␤6, ␤14, and ␤16 in both subunits (Fig. 6C). In particular, the residues belonging to A-1 and L-1, which are not present in PDH1␤, form extensive hydrophobic interactions with the surrounding residues and play a key role in the subunit-subunit interactions within the ApeLPDH dimer. In addition, superposition of PDH1␤ onto an ApeLPDH subunit showed that the C-terminal chain (residues 367-375) of PDH1␤ would sterically hinder the binding of another ApeLPDH subunit (Fig. 6D). This means that L-1 and A-1, as well as the unique structure of the C-terminal region, contribute substantially to the ApeLPDH dimer association. Structural Features Underlying Thermostability-Structural studies of hyperthermophilic proteins have suggested that it is the greater numbers of hydrophobic interactions and ion pairs that are responsible to their high thermostability (29 -32). In  addition, the common use of disulfides in thermostable proteins of hyperthermophiles has been evaluated at both the structural and genomic levels (31,(33)(34)(35). We therefore compared the numbers of hydrophobic interactions, ion pairs, and disulfide bonds within the structures of ApeLPDH and PDH1 (Table 2). When we counted the intersubunit hydrophobic interactions, we found that the number in the ApeLPDH dimer (108) is markedly larger than that in the PDH1 ␣␤-heterodimer (63), although the two have similar accessible surface areas (2000 Å 2 ) at the interface. The total number of hydrophobic interactions within the ApeLPDH monomer (699) was also larger than that within PDH1␤ (601). In addition, we found that the ApeLPDH monomer and the intersubunit interface contained 60 and 12 ion pairs (using a cutoff distance of 4.0 Å), respectively, whereas the PDH1␤ monomer and the ␣␤-heterodimer interface contained only 27 and 4 ion pairs, respectively. These results are consistent with the notion that the presence of a large number of intra-and intersubunit hydrophobic interactions, as well as extensive formation of ion pairs, are likely the main factors contributing to the high thermostability of ApeLPDH, which retains more than 90% of its full activity after incubation at 80°C (36). By comparison, we found that the total numbers of hydrophobic interactions (752) and ion pairs (73) within the PDH1␣ subunit are significantly larger than in the ApeLPDH monomer (Table 2). Moreover, the intermolecular accessible surface area of the (␣␤) 4 -hetero-octamer is a huge 14,000 Å 2 . The total number of ion pairs and hydrophobic interactions within the PDH1 subunit interfaces were estimated to be 36 and 260, respectively ( Table 2). The presence of the disulfide bonds has been shown to confer significant thermostability (31,(33)(34)(35). As described above, one (␣␤) 2 block in the PDH1 hetero-octamer is generated by an intermolecular disulfide bridge. However, no disulfide bond was present in the structure of ApeLPDH (Table 2). We therefore suggest that the huge interface area of the oligomeric structure and the presence of the intermolecular disulfide bonds, as well as the presence of a large number of hydrophobic interactions and ion pairs within the ␣-subunit, probably account for the hyperthermostability of PDH1, which showed no loss of activity during incubation at 90°C for 120 min (7).
Cofactor Binding-FAD is bound in an elongated conformation to the FAD-binding domain (Fig. 7A) such that no cofactor atoms are solvent-accessible. The N6 atom of the adenine base forms a hydrogen bond with the main chain oxygen of Val 184 , and the 3Ј-OH group of the adenine ribose interacts with the side chain of Asp 34 . The backbone phosphate groups interact Most of these interactions were conserved in PDH1␤, although the residues responsible for the interactions differ between ApeLPDH and PDH1␤. The most noteworthy differences between the two enzymes involve the residues on the si-face of the isoalloxazine ring of FAD; a stretch of CGTG (residues 47-50) in PDH1␤ is replaced by SMAA (residues 47-50) in ApeLPDH (Fig. 7A). Among these residues, Cys 47 in PDH1␤ makes parallel contact with the flavin ring; however, this residue is replaced by Ser 47 in ApeLPDH. That Cys 47 is strictly conserved among PDH1␤ homologues suggests it may be a key residue responsible for electrons passing from FAD to FMN, and it is incorporated into the electron transfer system of PDH1 (11).
superposition of ApeLPDH onto PDH1␤ also shows that residues Ile 261 , Ile 229 , and His 257 , which surround the isoalloxazine ring of FMN in PDH1␤, are replaced by Trp 300 , Arg 256 , and Glu 296 , respectively, in ApeLPDH (Fig. 7B). These substitutions result in a large structural difference between the FMN-binding sites of PDH1␤ and ApeLPDH, with the side chain of Arg 256 disrupting FMN binding in ApeLPDH (Fig. 7B).
Active Site and Insight into LPDH Reaction-The active site cavity of the ApeLPDH subunit is lined by the residues Arg 52 , Tyr 88 , Phe 90 , Arg 252 , Leu 281 , Ile 283 , Leu 289 , Ser 304 , Tyr 352 , Gly 378 , Ser 379 , and Met 382 (Fig. 8), and the re-face of the isoalloxazine ring of FAD forms the bottom of the cavity. Our initial experimental electron density map showed extra density within the cavity and, after construction and refinement of the peptide chain, an L-proline molecule could be modeled unambiguously into that density. Because L-proline was not present during crystal growth, this molecule might have originated in the E. coli cells and was then retained throughout the protein purification. To identify the bound molecule, the mass spectra of SeApeLPDH were measured by using FT-ICR MS as described under "Experimental Procedures." Both L-proline and DL-P5C were employed as the potential binding compounds in ApeLPDH, and their exact masses were measured by using the same FT-ICR MS (supplemental Fig. 1, B and C). In supplemental Fig. 1D, a mono-isotopic mass, 116.07140 m/z, is consistent with the mass of single charged L-proline with 0.00027 Da mass difference from that of control L-proline in supplemental Fig. 1B. However, the mass of DL-P5C was not detected in the ApeLPDH. These results strongly demonstrate that the bound molecule is L-proline. The map precisely defined the orientation of the L-proline (Fig. 8); the carboxyl group of L-proline hydrogen-bonded with the side chains of Arg 52 and Tyr 88 and, via a water molecule (Wat 1 ), with the side chain of Arg 52 . Interestingly, Wat 1 also forms hydrogen bonds with the main chain amide and the carboxyl group of Leu 428 (Fig. 8), which is provided by the C-terminal tail extending from the FAD-binding domain.
Leu 428 interacts with L-proline, via Wat 1 , which suggests it may be involved in the enzyme reaction. To assess the role of the Leu 428 , we constructed a ⌬L428 mutant and observed that the K m value of the mutant for L-proline was about 800 times higher than the K m value of the wild-type enzyme (230 Ϯ 22 mM (mutant) versus 0.28 Ϯ 0.07 mM (wild-type)). By contrast, the k cat (s Ϫ1 ) values for mutant and wild-type enzymes did not differ much (2.5 Ϯ 0.1 (mutant) versus 2.7 Ϯ 0.1 (wild type)). This suggests that Leu 428 is not directly involved in the enzyme catalysis but is essential for maintaining a high affinity for the substrate. Within the substrate-binding site of ApeLPDH, the carboxyl group of Leu 428 forms hydrogen bonds with the side chains of Ser 304 and Tyr 352 , and the side chain of Leu 428 forms a hydrophobic core with Phe 90 , Leu 281 , Ile 283 , and Leu 289 (Fig. 8). Through these interactions, Leu 428 is tightly held at the entrance of the cavity, where it completely shields the substrate-binding pocket (Fig. 5A).
To determine the orientation of the C-terminal chain lacking Leu 428 , we performed crystal structure analysis of ⌬L428 mutant enzyme. The model of the ⌬L428 mutant was refined at a resolution of 2.01 Å to a crystallographic R-factor of 21.0% and a free R-factor of 25.5% (Table 1). The asymmetric unit consisted of one homodimer with a solvent content of 49.5%, which corresponds to a Matthews coefficient (27) of 2.4 Å 3 Da Ϫ1 . The model contained up to 425 ordered amino acid residues in each subunit, 2 FAD coenzymes, 2 acetate molecules, 4 ethylene glycol molecules, and 213 water molecules. The two nearly identical (r.m.s.d. ϭ 0.13 Å) subunits were related by a 2-fold noncrystallographic rotation axis. In both subunits, Met Ϫ19 to Met 1 and the C-terminal residue (Val 427 ) were disordered and not visible in the electron density map. In the ⌬L428 mutant enzyme, we found no L-proline molecule in the active site. Moreover, we superimposed FAD-binding domain (residues 2-48, 160 -249 and 351-399) of the ApeLPDH onto the equivalent residues of ⌬L428 mutant (r.m.s.d. ϭ 0.43 Å), and found that substrate-binding domain of ApeLPDH was rotated about 10°relative to the corresponding domain of ⌬L428 mutant (Fig.  9A). This indicates that L-proline incorporation leads to a rotation of the substrate-binding domain. Therefore, the ApeLPDH structure assumes a closed conformation, and the ⌬L428 mutant structure assumes an open one. The C-terminal chain (residues 420 -426) of the ⌬L428 mutant was about 13°rotated clockwise compared with the corresponding residues of ApeLPDH, and this results in the active site cavity being solvent-accessible (Fig. 9B). These results strongly suggest that the function of the C-terminal Leu 428 is to provide a solvent-inaccessible environment for the enzyme reaction, as well as to hold the substrate properly within the active site.
Superposition of PDH1␤ onto ApeLPDH revealed that Arg 52 and Tyr 88 in ApeLPDH, which are responsible for the interaction between the enzyme and the carboxyl group of L-proline, are conserved in PDH1␤ as Arg 52 and Tyr 87 , respectively (data not shown). We therefore constructed R52A and Y87F mutants of PDH1␤ and observed that these mutations greatly reduced the specific activity of the enzyme (0.033 and 0.41 mol/min/ mg, respectively) compared with that of the wild-type enzyme (14.7 mol/min/mg). These results strongly suggest that PDH1␤ binds L-proline in a manner similar to ApeLPDH. As

No. of disulfide bonds
Monomer 0 0 (␤) 0 (␣) Interface 0 0 (␣␤) 2 (␣␤) 4 described above, however, the C-terminal seven residues in PDH1␤ have poor electron density, indicating they are disordered, and this results in the active site cavity being highly solvent-accessible (Fig. 5B). It is therefore likely that a conformational change in the C-terminal region of PDH1␤ occurs to accommodate the L-proline molecule within the active site during enzyme catalysis. In ApeLPDH, the rearrangement of the C-terminal tail may take place to release the reaction product after electron transport from L-proline to the electron acceptor. A structural database search using DALI (37) showed that ApeLPDH also has similarity to monomeric sarcosine oxidase (Protein Data Bank code 2GB0) (r.m.s.d ϭ 2.6 Å over 387 C␣; Z-score ϭ 36.9). The amino acid sequence identity between ApeLPDH and monomeric sarcosine oxidase is about 25% over FIGURE 7. A, stereo representation of FAD bound to ApeLPDH. The FAD is shown in magenta. The CGTG stretch in PDH1␤, which corresponds to the SMAA stretch in ApeLPDH, is shown in gray (red labels). B, stereo representation of FMN bound to PDH1␤. The structure of ApeLPDH (green and black labels) is superimposed on that of PDH1␤ (gray and red labels); FMN is shown in cyan. In both panels, the networks of hydrogen bonds are shown as dashed lines, and atoms are colored as in Fig. 2. the structurally equivalent residues. Although several different mechanisms have been proposed, including hydride transfer, single electron transfer and polar mechanisms (38 -40), the mechanism underlying the monomeric sarcosine oxidase-catalyzed amine oxidation reaction remains unclear. Based on structural and site-directed mutagenesis studies, it has been proposed that Tyr 317 and His 269 in monomeric sarcosine oxidase, which are located close to the methyl group of the substrate, are not essential for catalysis but are important for optimizing substrate binding (39,40). Tyr 317 is conserved as Tyr 352 in ApeLPDH, although His 269 is replaced by Ser 304 . Given that the Tyr 352 and Ser 304 interact with the C-terminal Leu 428 , the role of these two residues in ApeLPDH is likely to hold Leu 428 as the active-site shield.
A catalytic mechanism for hyperthermophilic LPDH was proposed by Monaghan et al. (41), who studied the PDH1 homologue from P. furiosus (PRODH) and based their proposal on its preliminary crystal structure. Initially, ␤His 225 and ␤Tyr 251 , which are located close to the re-face of the flavin ring of FAD, were selected as potential candidates for the active-site base in P. furiosus PRODH, but site-directed mutagenesis studies revealed that these residues are not essential for catalysis. In the absence of an active-site residue that acts as base, it was proposed that PRODH-catalyzed amine oxidation may occur through addition of a deprotonated L-proline at the C4 position of the FAD cofactor and abstraction of a substrate proton by the N5 atom of the flavin, a mechanism similar to the flavoproteincatalyzed amine oxidation proposed for trimethylamine dehydrogenase (42). In ApeLPDH, ␤His 225 and ␤Tyr 251 in P. furiosus PRODH are replaced by Arg 252 and Leu 289 , respectively. Thus, ApeLPDH also does not have an obvious residue that might act as a base within the active site. In this respect, the catalytic mechanism of ApeLPDH is thought to be similar to that proposed for P. furiosus PRODH.
We recently identified another dimeric LPDH from the aerobic hyperthermophilic archaeon P. calidifontis (12). In P. calidifontis LPDH, the C-terminal five residues (ERLVI) are nearly the same as in ApeLPDH, except that Leu 428 is replaced by Ile 415 . Moreover, the residues that interact with the Leu 428 in ApeLPDH are highly conserved in the P. calidifontis enzyme, except that Leu 281 and Leu 289 , which form a hydrophobic core with the side chain of Leu 428 , are replaced by Met 271 and Tyr 279 , respectively, and Ser 304 is replaced by Ala 294 . These results suggest that, at least in the case of dimeric LPDHs, a C-terminal hydrophobic residue plays an important role in the normal enzyme function. Collectively, our findings provide structural insight into the dimeric structure and substrate binding of a novel type of LPDH, whose subunit structure is much simpler than those of the other LPDHs described so far. We anticipate this will lead a better understanding of the structure-function relationships within archaeal LPDHs.