Structure-Function Relationships in l-Amino Acid Deaminase, a Flavoprotein Belonging to a Novel Class of Biotechnologically Relevant Enzymes*

l-Amino acid deaminase from Proteus myxofaciens (PmaLAAD) is a membrane flavoenzyme that catalyzes the deamination of neutral and aromatic l-amino acids into α-keto acids and ammonia. PmaLAAD does not use dioxygen to re-oxidize reduced FADH2 and thus does not produce hydrogen peroxide; instead, it uses a cytochrome b-like protein as an electron acceptor. Although the overall fold of this enzyme resembles that of known amine or amino acid oxidases, it shows the following specific structural features: an additional novel α+β subdomain placed close to the putative transmembrane α-helix and to the active-site entrance; an FAD isoalloxazine ring exposed to solvent; and a large and accessible active site suitable to bind large hydrophobic substrates. In addition, PmaLAAD requires substrate-induced conformational changes of part of the active site, particularly in Arg-316 and Phe-318, to achieve the correct geometry for catalysis. These studies are expected to pave the way for rationally improving the versatility of this flavoenzyme, which is critical for biocatalysis of enantiomerically pure amino acids.

The deracemization of a racemic amino acid to obtain the L-configuration was achieved by using a stereoselective Damino acid oxidase (DAAO, 5 EC 1.4.3.3) followed by chemical reduction. The second step iteratively converts the imino acid produced (from the D-amino acid) back into a DL-mixture to obtain the full resolution of the racemic mixture into the L-enantiomer (1). This approach requires stable recombinant DAAOs possessing wide substrate specificity as well as variants engineered to act on synthetic amino acids (3).
Amino acid oxidases with reverse stereoselectivity are also well known flavooxidases, mainly produced by snakes or by microorganisms. In particular, L-amino acid oxidases (LAAO, EC 1.4.3.2) catalyze the stereoselective oxidative deamination of L-amino acids into the corresponding ␣-keto acids and ammonia; the re-oxidation of FADH 2 by dioxygen then generates H 2 O 2 (4). These flavoenzymes catalyze an irreversible reaction (differently from aminotransferases) and do not require a specific step of cofactor regeneration, as otherwise required by the NAD-dependent dehydrogenases. However, because of the problems associated with overexpression of snake venom LAAOs in recombinant hosts and the limited substrate acceptance of the microbial counterparts, no appropriate LAAOs for biocatalysis are available (4).
L-Amino acid deaminases (LAADs) represent a suitable alternative to LAAOs. LAAD (first identified in the genera Proteus, Providencia, and Morganella) catalyzes the deamination of the L-isomer of amino acids, yielding the corresponding ␣-keto acids and ammonia without any evidence of H 2 O 2 production. LAAD from Proteus myxofaciens (PmaLAAD), expressed in the Escherichia coli K12 strain, shows a preference for L-amino acids with aliphatic, aromatic, and sulfur-containing side chains (5). Although the reaction catalyzed by LAAD is of both scientific and practical interest (5)(6)(7), no details about its structural-functional relationships and biological significance have been reported so far.
To gain more insight into the properties of LAAD, we expressed the P. myxofaciens enzyme in E. coli and demonstrated that it is a membrane flavoenzyme employing a cytochrome b-like protein as electron acceptor. By using protein engineering studies, we were able to produce a truncated, soluble form (PmaLAAD-01N), which was crystallized. Its structural and biochemical characterization clearly shows that Expression and Purification of PmaLAAD Wild-type and Variants-Recombinant wild-type and variants of PmaLAAD were expressed in E. coli BL21(DE3) strain under the following conditions: cells (from a single colony) were grown in modified TB medium containing 10 g/liter glycerol, to which 100 g/ml ampicillin was added (or 30 g/ml kanamycin for PmaLAAD-00C). Protein expression was induced by adding 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside at an OD 600 nm ϭ 0.5; after 4 h of growth at 28°C, cells were harvested by centrifugation.
E. coli recombinant cells expressing PmaLAAD variants were disrupted by sonication in lysis buffer (50 mM potassium phosphate, pH 7.0, 5 mM MgCl 2 ). The cell lysate was separated from the cell debris by centrifugation at 38,500 ϫ g for 1 h yielding the insoluble fraction of the cell lysate and the crude extract containing the soluble proteins and membrane fragments. For wild-type PmaLAAD, ultracentrifugation of the latter fraction at 150,000 ϫ g for 2 h produced a membrane fraction (pellet) and a cytoplasmic fraction (supernatant). The membrane fraction was resuspended in 2.5 ml of 50 mM potassium phosphate, pH 7.0, stored at Ϫ80°C, then thawed and diluted 5-fold in 20 mM triethanolamine-HCl, pH 7.5; centrifugation at 16,000 ϫ g for 10 min yielded two phases at different density; see Fig. 2 for details.
Recombinant PmaLAAD-00N, -00C, and -01N variants from the soluble fraction of the lysates (crude extract) were purified by using metal-affinity chromatography on a HiTrap chelating column (5 ml, GE Healthcare, Little Chalfont, UK) using 50 mM potassium phosphate buffer, pH 7.5. In the case of PmaLAAD-01N, contaminant proteins bound to the column were eluted with two steps at 25 and 80 mM imidazole, and the variant was eluted at 200 mM imidazole. For crystallographic studies, the purity of PmaLAAD-01N was increased by loading the purified protein (13.7 mg/ml) onto a second HiTrap chelating column (1 ml); elution was performed using increasing concentrations of imidazole (25,50,12, and 500 mM). PmaLAAD-00N and -01N were analyzed by size-exclusion chromatography using a Superdex 200 Increase 10/300 column (GE Healthcare).
Activity Assay-The activity on L-Phe was assayed spectrophotometrically by following the formation of the product phenylpyruvate at 321 nm, a wavelength at which the extinction coefficient of the substrate is negligible (5). The protein sample was added to 2 ml (final volume) of 25 mM L-Phe in 50 mM potassium phosphate buffer, pH 7.5, and incubated for 2 min at 25°C. Every 30 s an aliquot of 400 l was transferred into a plastic cuvette containing 400 l of 3 N NaOH to stop the reaction and allow color to develop. One enzymatic unit corresponds to the amount of enzyme that converts 1 mol of L-Phe per min. The apparent K m , k cat , and V max values for L-Phe were determined by the phenylpyruvate production assay using increasing substrate concentrations, at 25°C, and oxygen air saturation (0.253 mM).
The same assay was used to investigate the ability of commercial quinones (anthraquinone-2,6-disulfonate, anthraquinone-1,5-disulfonate, 2-hydroxy-1,4-naphthoquinone, menadione, duroquinone, and juglone), using phenazine methosulfate as mediator or artificial electron acceptors (indigotetrasulfonate, gallocyanine, 2,6-dichlorophenolindophenol/phenazine methosulfate, cytochrome c, and nitro blue tetrazolium) to stimulate the activity of PmaLAAD-00N in the absence of membranes. All compounds were used at 50 M final concentration, with the only exception being phenazine methosulfate that was used at 800 M final concentration. The activity of PmaLAAD on L-Phe in the presence of exogenous membranes prepared from E. coli, Streptomyces venezuelae, and human glioblastoma T98G cells was investigated using the spectrophotometric assay.
The deamination reaction of PmaLAAD on L-amino acids in the presence of commercially available E. coli lipid extracts (Avanti Polar Lipids, Alabaster, AL) was investigated by means of a polarographic method based on an oxygen electrode (O 2 consumption) assay (11). This assay couples the re-oxidation of the reduced flavin in PmaLAAD with an electron acceptor system that, in turn, consumes dioxygen (see under "Results" for details). This assay was also employed to investigate the substrate specificity of PmaLAAD on a number of L-amino acids.
Spectral Measurements-Absorbance spectra in the UV-visible region of PmaLAAD variants, as well as ligand-binding experiments, were recorded using a Jasco V-560 spectrophotometer (Jasco Europe, Cremello, Italy) as reported previously (13).
Anaerobic samples were prepared in anaerobic cuvettes by applying 10 cycles of evacuation and then flushing with oxygenfree argon. Flavin reduction was carried out by adding 25 mM (final concentrations) of the substrate L-Phe to samples containing Ϸ15 M enzyme; reduction of the cofactor for Pma-LAAD-00N variant was performed both in the absence and in the presence of E. coli membranes.
Photoreduction experiments were carried out at 4°C using an anaerobic cuvette containing 18.5 M PmaLAAD-00N or 13.1 M PmaLAAD-01N, 5 mM EDTA, and 0.5 M 5-deaza FAD. The solution was photoreduced using a 250-watt lamp, and the progress of the reaction was monitored spectrophotometrically (13,14). At the end of the reaction, FAD cofactor was re-oxidized anaerobically by adding 5.5 M (final concentration) benzyl viologen or by flushing O 2 . Photoreduction experiments for PmaLAAD-00N were also performed in the presence of E. coli membranes. Binding experiments of carboxylic acids and sulfite to PmaLAAD-00N and -01N variants were performed by adding small volumes (5-50 l) of concentrated stock solutions (0.1 mM to 1 M) to samples containing 0.8 ml of Ϸ15 M enzyme and following the changes in the visible absorbance spectrum of the flavin cofactor (13,14).
Far-UV CD spectra of PmaLAAD-00N and -01N variants were recorded using a Jasco J-815 spectropolarimeter equipped with a software-driven Peltier-based temperature controller; the cell path was 0.1 cm (15). All spectral measurements were recorded at 15°C in 50 mM potassium phosphate buffer, pH 7.5, except where stated otherwise.
The PmaLAAD-01N structure was solved by molecular replacement using the program Phaser (18), with the ␤-subunit of sarcosine oxidase from Pseudomonas maltophilia (PDB code 2GAG) (19) as a search model. Two protein molecules (A and B chains) were located in the crystal asymmetric unit and were subjected to rigid-body and restrained refinement using Phenix (20). The amino acid sequence of the model was modified to match the correct sequence and manually fitted to the electron density using the program Coot (21). A set of restrained refinement cycles were performed using the programs Phenix (20) and Refmac (22). The final refined PmaLAAD-01N structure (R factor ϭ 17.7%, R free ϭ 20.9%) contains 2 ϫ 447 protein residues (28 -474), 2 FAD co-factors, and 518 water molecules, with good stereochemical parameters. Poor electron density is only present at the N and C termini (residues 28 and 474 of both A and B chains).
For soaking experiments and data collection, native Pma-LAAD-01N crystals were removed from the crystallization drop with a nylon loop and soaked overnight in a mixture of their stabilizing solution (see above) supplemented with 20 mM anthranilate. The soaked crystals were quickly pulled through a drop of cryoprotectant before being flash-cooled in liquid nitrogen. A single crystal, belonging to monoclinic space group P2 1 , diffracted up to 1.75 Å resolution (data collection and processing as above).
The structure of the PmaLAAD-01N⅐anthranilate complex was determined by molecular replacement using the program Phaser (18), with the structure of PmaLAAD-01N as a search model. Two protein molecules were located in the crystal asymmetric unit (A and B chains). The structure was then remodeled with Coot (21) and subjected to restrained refinement, with anisotropic B-factors, using the program Refmac (22). After a few cycles of refinement, the 2F o Ϫ F c electron density map showed structural details that allowed unambiguous modeling of the bound substrate analogue. The final model (R factor ϭ 15.6%, R free ϭ 19.7%) contains 2 ϫ 447 protein residues (28 -474), 2 FAD cofactors, 2 anthranilate molecules, and 546 water molecules, with good stereochemical parameters. Poor electron density is present at the 28 -31 and 337-342 regions and at the C terminus (residue 474) of both A and B chains.
All data reduction and refinement statistics for both protein models (PmaLAAD-01N and PmaLAAD-01N⅐anthranilate complex) are reported in detail in Table 1. The stereochemical quality of the models was assessed with MolProbity (23). The atomic coordinates and structure factors for PmaLAAD-01N and PmaLAAD-01N⅐anthranilate complex (codes 5FJM and 5FJN, respectively) have been deposited in the Protein Data Bank.
Phylogenetic Analyses-Proteins homologous to PmaLAAD were identified with the HMMER web server (24) using the "Representative Proteomes" database (25). Sequences sharing an identity of Ͼ90% with PmaLAAD were filtered using CD-HIT Suite (26). In addition, amino acid sequences possessing an identity with PmaLAAD Ͻ28% were discarded. Alignment was performed using the MUSCLE algorithm in SeaView software (27). Maximum-likelihood (ML) phylogenetic tree of LAAD sequences was built using PhyML 3.0 server (28). Confidence of branching patterns was assessed by bootstrapping (100 bootstrap samples were used) (29). The resulting ML tree was visualized in FigTree version 1.4.2.

Results
Purification of Recombinant Wild-type PmaLAAD-Recombinant PmaLAAD (the sequence of which is reported in Fig. 1B) was expressed in E. coli BL21(DE3) cells. After cell lysis by sonication and centrifugation, the PmaLAAD activity was fully recovered in the crude extract, as assayed by following the absorbance change at 321 nm due to the conversion of L-Phe into phenylpyruvic acid. After ultracentrifugation of the crude extract, the recombinant enzyme was mainly associated with the membrane fraction (i.e. in the pellet fraction) that was resuspended in 50 mM potassium phosphate buffer, pH 7.5, obtaining a Ϸ2-fold increase in specific activity (Fig. 2).
Interestingly, thawing, diluting (5-fold) in 20 mM triethanolamine-HCl, pH 7.5, and centrifuging the membrane fraction stored at Ϫ80°C resulted in a phase separation; the lower, denser phase contained about half of the PmaLAAD of the sample (based on the enzymatic activity and SDS-PAGE analysis) and showed a further Ϸ2.5-fold increase in specific activity, reaching 2.9 units/mg of protein, with an overall yield of 33% (Fig. 2). Based on the purity degree (29%, see lane 7 in Fig. 2B, left panel), a k cat ϭ 8.5 s Ϫ1 was estimated. Because of the failure to produce a soluble membrane-free preparation of wild-type PmaLAAD, chromatographic steps were not effective in purifying enzyme.
In addition, N-and C-terminal His-tagged full-length proteins were produced (PmaLAAD-00N and -00C, respectively, Fig. 1) and expressed at a level similar to that of wild-type Pma-LAAD by using BL21(DE3) or Origami E. coli strains. When the crude extract containing PmaLAAD-00N (specific activity on L-Phe of 0.8 units/mg protein) was loaded on a HiTrap chelating column, it separated into two fractions as follows: the flowthrough containing the active enzyme (Ϸ500 units/liter fermentation broth, 90% of purification yield) associated with membrane fragments, and the fraction eluted at 0.2 M imidazole, which contained a significant part of PmaLAAD-00N (devoid of membranes). Following dialysis to eliminate imidazole, the latter fraction showed a purity degree Ն90% (as estimated by SDS-PAGE analysis, lane 5 in Fig. 2B, middle panel) with F obs being the observed and F calc the calculated structure factor amplitudes.
and was fully stable (see spectral properties, below). However, the measured LAAD activity was only marginal (5 units/liter fermentation broth); therefore, this fraction was named "quasi-inactive" PmaLAAD-00N. Similar results were obtained for the PmaLAAD-00C variant; again, after loading on the HiTrap chelating column, enzymatic activity was detected mainly in the flow-through fraction (Ϸ250 units/liter fermentation broth), although the fraction eluted at 0.5 M imidazole contained a significant part of PmaLAAD-00C protein but showed only marginal LAAD activity (Ͻ2%), and a purity of Ϸ70% (as estimated by SDS-PAGE analysis), data not shown. Biochemical Properties of Partially Purified, Membranebound PmaLAAD-The specific activity of PmaLAAD in the crude extract, which contains membranes, on 25 mM L-Phe as substrate was Ϸ0.6 units/mg protein (assayed as production of phenylpyruvic acid). The dependence of the initial rate on L-Phe concentration is shown in Fig. 3A; the K m, app ϭ 3.27 Ϯ 0.96 mM and V max, app ϭ 1.35 Ϯ 0.08 units/mg protein values are in good agreement with those reported in the literature (K m, app ϭ 2.3 mM, V max, app ϭ 2 units/mg) (5). As a result of the 12% purity degree of PmaLAAD in the crude extract, a k cat ϭ 8.3 s Ϫ1 was estimated. No substrate inhibition was evident up to 100 mM L-Phe. On the same enzyme preparation, the activity on L-Phe was also assayed as O 2 consumption (i.e. by polaro- graphic assay) (11), yielding an identical specific activity. From this evidence, we conclude that the membrane-containing PmaLAAD preparation catalyzes the O 2 -dependent oxidative deamination of L-Phe.
By employing the O 2 -consumption assay, we investigated the relative substrate specificity of PmaLAAD toward natural and synthetic amino acids; the polarographic assay was carried out using a fixed volume of membrane-associated PmaLAAD (corresponding to Ϸ0.5 mg of total membrane proteins) and 25 mM substrate concentration. As shown in Table 2, PmaLAAD is specific for aromatic, neutral L-amino acids; a low activity was assayed toward small, charged, or polar L-amino acids. Notably, a significant activity was also measured on the synthetic amino acid L-phenylalanine ethyl ester (whose ␣-carboxylic group is esterified) and on L-DOPA (33% and 56.7% in comparison to L-Phe, respectively). No significant activity was assayed with D-amino acids ( Table 2).
The pH dependence of PmaLAAD activity was evaluated on the sample from the membrane fraction by using the spectrophotometric assay in the 3-10 pH range (12). The highest activity is apparent at neutral pH (between 7 and 7.5) and is negligible below pH 5 and above pH 9 (Fig. 3B). Concerning the temperature dependence, the activity of PmaLAAD linearly increased up to 50°C and then rapidly decreased, becoming negligible at 70°C (Fig. 3C). This behavior resembles that reported for LAAD from Proteus mirabilis (optimal temperature of 45°C) (30).
Notably, PmaLAAD in the crude extract showed a fairly good stability because it showed a 50% residual activity after 24 h of incubation at 25°C. The enzyme retained Ϸ50 and Ϸ75% of the original activity when stored for 1 month in the presence of 10% glycerol at Ϫ20 and at Ϫ80°C (even without glycerol), respectively. Overall, the properties of this PmaLAAD preparation seem to satisfy the requirements for application in biocatalysis of natural and unnatural amino acids (2).
Biological Activity of PmaLAAD, Effect of Membranes-The enzymatic activity (assayed as production of phenylpyruvic acid) of purified, membrane-free, and quasi-inactive Pma-LAAD-00N is very low (see above); notably, it was significantly enhanced by adding E. coli membranes and sonication. Actually, the addition of an amount of membranes from E. coli cells (not expressing PmaLAAD) corresponding to Ϸ0.30 mg of cells for each microgram of enzyme increased the PmaLAAD-00N activity by Ϸ50-fold. No further increase in enzymatic activity was obtained using larger amounts of membranes. As a control, sonication of PmaLAAD-00N in the absence of exogenous membranes resulted in a 6-fold decrease in the original low enzymatic activity. Notably, recovery of PmaLAAD activity depended on the membrane's source; no recovery of the enzymatic activity was observed using membranes from the microorganism S. venezuelae, although a low, partial recovery was apparent (2.7-fold) using membranes from human glioblastoma T98G cells. Similarly to PmaLAAD-00N, a Ϸ32-fold increase in enzymatic activity was apparent by adding E. coli membranes prepared from 2.25 mg of cells to 75 g of purified PmaLAAD-00C.  (12). Data are expressed as percentage of maximal enzyme activity and were fitted based on the equation for two dissociations (12); estimated pK a1 was 5.8 Ϯ 0.1, and pK a2 was 8.3 Ϯ 0.3. C, activity of PmaLAAD determined at the indicated temperatures. Data are expressed as percentage of enzyme activity assayed at 50°C (the maximum value). B and C, activity was determined on 25 mM L-Phe (and at pH 7.5 in C) employing 10 l (corresponding to 25 g of enzyme) of resuspended membrane containing PmaLAAD fraction. Bars indicate Ϯ S.E. for three determinations.

TABLE 2 Relative specific activity of PmaLAAD on different substrates
The activity was determined by the polarographic assay (oxygen consumption) using a membrane-containing PmaLAAD preparation and 25 mM substrate concentration with the exception of L-Tyr (1.25 mM), ␤ 3 -Phe (12.5 mM), 1-naphthyl-Ala (0.375 mM), and Gly (0.375 mM). The reported value was the average of three determinations. 56.9 a Specific activity was expressed as percentage of the value assessed with L-Phe (100%). b The activity value for the corresponding D-amino acid was Յ4%. c The activity on this compound was confirmed by using for the assay a purified (by re-crystallization) substrate batch to avoid contamination by residual free L-Phe.

Structure-Function Relationships in L-Amino Acid Deaminase
The dependence of PmaLAAD-00N activity on a putative molecular partner associated with the bacterial membrane was investigated by mixing the purified enzyme variant with two commercially available E. coli lipid extracts (Avanti Polar Lipids). Mixing 150 g of PmaLAAD-00N with 4.5 mg of Total Lipid extract (containing 17.6% unknown components, mainly lipoproteins) resulted in a slight (4-fold) increase in activity, which is significantly lower than the activity recovery observed using the same amount of E. coli membranes (ϭ32-fold). Employing the Polar Lipid extract (with a higher degree of purity), an even lower increase in enzymatic activity was apparent (1.6-fold, Table 3). To exclude any inhibitory effect of a component of the lipid extracts, 150 g of PmaLAAD-00N were mixed with 2.25 mg of lipid extract and a corresponding amount of E. coli membranes (22.5 mg); the recovery of the enzymatic activity resembled the value observed for E. coli membranes only.
Altogether, these results suggest that a specific membrane partner(s), which is(are) available in E. coli membrane preparations, is required for the purified, quasi-inactive PmaLAAD-00N to recover the enzymatic activity. Membranes from an unrelated source or a hydrophobic environment, such as the one provided by purified lipid extracts, are not suitable for recovering PmaLAAD activity.
Spectral Characterization of Purified PmaLAAD-00N-The spectral properties were assessed using the membrane-devoid PmaLAAD-00N variant because of its higher purity compared with the form lacking the His tag and the absence of membranes affecting the measurements. PmaLAAD-00N shows the canonical absorbance spectrum of FAD-containing flavoproteins with three main peaks at 277, 380, and 456 nm (Fig. 4A). The ratio between the absorbance at 277 and 456 nm is Ϸ8.9, and the estimated molar extinction coefficient is 14,168 M Ϫ1 cm Ϫ1 at 456 nm. The absorbance spectrum of PmaLAAD-00C resembled that of PmaLAAD-00N (data not shown); this indicates that the location of the His tag does not affect the flavin environment of PmaLAAD. The far-UV circular dichroism spectrum shows that PmaLAAD-00N is folded and contains secondary structure elements (data not shown).
Amino acid oxidases are known to interact with a number of carboxylic acids, yielding specific spectral modifications and thus representing useful active-site probes (4,31,32). The binding to PmaLAAD-00N of various compounds known as DAAO and/or LAAO active-site ligands (14,33,34) was investigated in the absence and in the presence of E. coli membranes (corresponding to Ϸ0.30 mg of cells for each microgram of Pma-LAAD-00N). No spectral perturbations were evident up to Ϸ0.4 M sodium benzoate, suggesting that PmaLAAD-00N cannot bind this well known DAAO ligand (31,35). However, when the same experiment was performed using anthranilate (2-aminobenzoate), specific spectral perturbations were observed in the 400 -500-and 520 -600-nm range (Fig. 4C); the latter spectral changes are due to the formation of a charge⅐transfer complex between the ligand and the FAD cofactor, as observed for other amino acid oxidases (31,35). These absorbance changes were more evident in the absence of membranes; from the absorbance changes at 533 nm as a function of anthranilate concentration, a K d of 18.4 Ϯ 4.4 mM was calculated (n ϭ 3, Fig.  4C, inset). The same experiment was performed using kojic acid as ligand (Fig. 4E), yielding a general increase in absorbance in the visible region of the spectrum with a saturation behavior (K d ϭ 3.0 Ϯ 0.3 mM) (n ϭ 3, Fig. 4E, inset). Overall, PmaLAAD-00N binds selected carboxylic acids, resulting in specific alterations of the visible spectrum of the flavin cofactor.
We then demonstrated that the flavin cofactor bound to PmaLAAD is catalytically competent. Actually, when the Pma-LAAD-00N variant was mixed under anaerobic conditions with an excess of substrate (25 mM L-Phe), the flavin cofactor rapidly reacted with the substrate, yielding the typical absorbance spectrum of the reduced FAD (i.e. bleaching of the 450-nm peak in the dead time of mixing for the membrane-free preparation) (Fig. 4A). Notably, the time course of flavin reduction was identical under aerobic conditions. We could then establish the reductive half-reaction catalyzed by PmaLAAD and as shown in Equation 1, No spectral perturbations resembling flavin reduction were observed by adding sulfite to PmaLAAD-00N instead (up to Ϸ0.45 M Na 2 SO 3 ); this suggests divergences in FAD reactivity between PmaLAAD and the enzymes belonging to the oxidase class of flavoproteins (33).
Anaerobic photoreduction of PmaLAAD-00N in the presence of 5 mM EDTA and 0.5 M 5-deaza-FAD generated the reduced FAD cofactor form after 5 min of exposure to intense light radiation (Fig. 4G). A large amount of the anionic semiquinone flavin form was evident, as made apparent by the large increase in a peak at Ϸ380 nm, an effect due to intense light irradiation. A full re-oxidation of membrane-free, fully photoreduced PmaLAAD-00N requiring minutes to completion was observed by admitting O 2 (see below) or adding benzyl viologen (Fig. 4G).

TABLE 3 PmaLAAD-00N activity recovery with purified E. coli lipid extracts
The activity was determined by the polarographic assay (oxygen consumption) on 25 mM L-Phe as a substrate, using Ϸ150 g of purified protein. All mixtures have been sonicated before the assay, until the solution appeared completely clear.

Sample
Activity

Formation of the Reduced Form of a Cytochrome b-like Protein during PmaLAAD-00N Reaction in Membranes-
The time course of FAD reduction of PmaLAAD-00N in the presence of membranes by an excess of L-Phe under anaerobic conditions was clearly biphasic; a large, rapid decrease at 450 nm was followed by a slower and smaller decrease (taking Ϸ15 min to completion, Fig. 4B). Interestingly, three narrow absorbance peaks are evident following full FAD reduction as follows: one major peak at 429 nm and two minor peaks at 532 and 560 nm (Fig. 4B). These peaks correspond to the ones typically associated with the reduced form of cytochrome b (36), a small membrane protein usually associated with redox enzymes (37). For details concerning cytochromes in Proteus species, see Refs. 38,39. Notably, when the same reaction was performed under aerobic conditions, the spectrum of the reduced cytochrome was observed only when the L-Phe concentration exceeded the concentration of dissolved O 2 (i.e. at Ͼ0.3 mM L-Phe). These data demonstrate that in the presence of membranes, L-Phe deamination by PmaLAAD is coupled to reduction of a cytochrome b-like protein.
In the presence of membranes, 1:1 (v/v) ratio corresponding to 30 mg of cells for each microgram of enzyme, the rate of flavin photoreduction of PmaLAAD-00N was increased (3.5 min to reach complete reduction, Fig. 4H) as compared with the same experiment in the absence of membranes. Notably, the isosbestic point at 410 nm observed in the photoreduction experiment in the absence of membranes was not conserved due to the formation of a Soret band of cytochrome b at 429 nm. A large amount of the anionic semiquinone flavin form was evident also in the presence of membranes, as made apparent by the large increase in a peak at Ϸ380 nm due to intense light irradiation.
O 2 Reactivity-The L-Phe deamination activity of PmaLAAD in the presence of membranes is coupled to O 2 consumption (see above). Dioxygen is absolutely required to produce phenylpyruvic acid from L-Phe under these experimental conditions; the rate constant for ketoacid production under anaerobic conditions is Ϸ3% of the value assessed at air saturation. Indeed, the reaction rate did not depend on O 2 concentration because a similar apparent activity was determined at Ն10% O 2 saturation.
Known LAAO and DAAO flavoenzymes employ O 2 to re-oxidize the reduced FADH 2 cofactor yielding H 2 O 2 . Notably, when using wild-type PmaLAAD in the crude extract, which is associated with membranes, no halving of oxygen consumption was observed by adding a large excess of catalase, which converts H 2 O 2 into H 2 O and 1 ⁄ 2 O 2 , to the reaction mixture of the polarographic assay, thus demonstrating that no H 2 O 2 was produced during the enzymatic reaction. Lack of H 2 O 2 production was also confirmed by the classic o-dianisidine/peroxidase spectrophotometric coupled assay used for flavoprotein oxidases (11). Similarly, no superoxide was produced by the Pma-LAAD reaction because the O 2 consumption was not modified by adding up to 24 units of superoxide dismutase.
Opening the cuvette containing fully photoreduced membrane-devoid PmaLAAD-00N to air, conversion into the fully re-oxidized flavin spectrum required Ϸ20 min, showing a very limited dioxygen reactivity of reduced PmaLAAD-00N (k ox Յ0.08 s Ϫ1 ). Mixing the photoreduced membrane-devoid Pma-LAAD-00N with buffer solutions equilibrated at increasing O 2 concentrations in the stopped-flow device (5-50% oxygen saturation, final concentration) did not yield the re-oxidized form of the flavin cofactor (33) during the measurement (Ϸ2 min), as otherwise normally observed for flavo-oxidases.
Altogether, these results demonstrate that the flavin molecule bound to PmaLAAD is quickly reduced by the substrate, but it is not efficiently re-oxidized by O 2 because the rate constant is not sufficient for the observed turnover (0.08 versus Ϸ 8 s Ϫ1 , respectively). The O 2 consumption observed in the presence of membranes is clearly not due to a direct O 2 -induced re-oxidation of the flavin cofactor bound to the enzyme. We suggest that in membranes the reduced flavin form of Pma-LAAD is re-oxidized by two oxidized cytochrome b-like proteins, as shown in Equation 2, that then transfer the electrons through the respiratory chain to dioxygen yielding H 2 O. Actually, using 250 M antimycin A, an inhibitor of the respiratory chain, the O 2 consumption of L-Phe oxidation by PmaLAAD in resuspended membranes is drastically decreased (Ն60%).
Production and Properties of the Soluble PmaLAAD-01N Variant-To design a PmaLAAD variant not able to associate with membranes and thus suitable for crystallographic studies, the amino acid sequence of the wild-type protein was analyzed using the TMPred server (40). A putative transmembrane ␣-helix (residues 8 -27) was predicted in the N-terminal region (Fig.  1A). Based on this prediction, two deletion variants were designed starting from Met-28 (PmaLAAD-01N) and Ala-50 (PmaLAAD-02N), respectively. Both variants were produced with an additional His tag sequence at the N terminus to facilitate their purification (Fig. 1, A and B). The PmaLAAD-02N variant was not characterized because of its low stability.
The PmaLAAD-01N variant, lacking the putative transmembrane ␣-helix, was expressed by following the same protocol used for PmaLAAD and PmaLAAD-00N and was purified from the crude extract by HiTrap chelating affinity chromatography. PmaLAAD-01N is almost completely soluble: Ϸ25 mg of pure PmaLAAD-01N were purified from 1 liter of culture with a Ͼ90% purity (Fig. 2B, lane 5 in right panel). The enzymatic activity of the pure PmaLAAD-01N preparation was below detection even when exogenous E. coli membranes were added to the assay mixture.
PmaLAAD-01N shows an absorbance spectrum very similar to that of PmaLAAD-00N, with an estimated molar extinction coefficient at 458 nm of 12193 M Ϫ1 cm Ϫ1 (Fig. 4A). In contrast, the far-UV circular dichroism spectrum of PmaLAAD-01N differs from the one recorded for the PmaLAAD-00N counterpart, showing, as expected, a decrease in ␣-helix content (data not shown).
When analyzed by size-exclusion chromatography, purified PmaLAAD-01N eluted in a single peak, corresponding to a molecular mass of 44.6 Ϯ 0.5 kDa (n ϭ 3) kDa (slightly lower than the expected value of 49.9 kDa), thus indicating that it is monomeric in solution in the 1-5 mg/ml protein concentration range. A similar result was obtained for the soluble fraction of full-length PmaLAAD-00N (Ϸ47 kDa).
Similar to PmaLAAD-00N, PmaLAAD-01N did not bind benzoate and did not react with sulfite, but it did interact with anthranilate, yielding a spectral perturbation which resembles a charge-transfer complex (K d ϭ 36.7 Ϯ 5.6 mM, Fig. 4D). Pma-LAAD-01N was also titrated with 2-aminobenzaldehyde; binding of this compound resulted in a large spectral change in the 500-to 580-nm range (K d of 0.31 Ϯ 0.05 mM n ϭ 3, Fig. 4F).
When purified PmaLAAD-01N was mixed under anaerobic conditions with an excess of substrate (25 mM L-Phe), the FAD cofactor rapidly converted into the reduced form, thus demonstrating that it is catalytically competent. Notably, the time course of flavin reduction and the absorbance spectrum of the reduced enzyme form were identical under both aerobic and anaerobic conditions, pointing to a limited O 2 reactivity of the deleted PmaLAAD variant. Photoreduction of the FAD cofactor in PmaLAAD-01N was fully accomplished after Ϸ10 min of illumination (with a large amount of anionic semiquinone formation), and the photoreduced enzyme was reverted to the oxidized form by mixing with benzyl viologen.
Altogether, the spectral binding and oligomerization properties of the deletion variant PmaLAAD-01N resemble those of the full-length enzyme. The main difference is its inability to associate with membranes, which prevents efficient L-Phe oxidation.
The Fbd is made up of a six-stranded ␤-sheet with ␤10-␤3-␤2-␤14-␤24-␤23 topology flanked by a helix bundle formed by A1-A3-A6-A14 on one side and by a three-stranded ␤-sheet ␤11-␤12-␤13 and helix A7 on the other side. The C-terminal A15 and A16 helices (3 10 ) run orthogonally to strand ␤23 (Fig.  5, B and C). The helices A1 and A14 of the helix bundle pack against the central six-stranded ␤-sheet with their N termini directed toward the FAD molecule. Helix A3 forms an outer layer of the fold, running orthogonally behind the A1, A6, and A14 helices. On the same side, the 3 10 helix A2 flanks the pyrophosphate group of FAD. On the other side, the N terminus of helix A7 is directed toward the FAD molecule.
The Sbd is characterized by the presence of a long region of 67 amino acids nestled between strand ␤19 and helix A12; this subdomain shows an ␣ϩ␤ structure whose arrangement is unprecedented, as suggested by a DALI search (42). The ␣ region consists of four helices as follows: A8, A9, A10, and A11; the ␤-region includes strand ␤20, which forms a parallel twostranded ␤-sheet with ␤19 from the core of the Sbd, and an anti-parallel two-stranded ␤-sheet built up by strand ␤1 and ␤21. In particular, strand ␤1 belongs to the N-terminal region of PmaLAAD-01N (residues 28 -39), which crosses the subdomain on the protein surface before inserting in the Fbd (Fig.  5, A and B).
Regarding the quaternary assembly, the two PmaLAAD-01N molecules in the crystal asymmetric unit face each other in the Sbd regions corresponding to the end of A3 and the following loop, A4-␤7-A5, A6, and to ␤16-␤17. The facing residues (25 in each monomer) make six salt bridges and two H-bonds (distances 3-3.5 Å). According to the program PISA (43), these interactions make the complex stable, but the overall buried surface area is modest, Ϸ700 Å 2 for each monomer corresponding to Ϸ4% of the total accessible surface area. Thus, in agreement with size-exclusion chromatography experiments, the PmaLAAD-01N does not show any quaternary architecture and can be considered monomeric.
The structural relationship among all these proteins extends to an excellent conservation of tertiary structure, with a root mean square deviation ranging from 2.0 to 2.2 Å. The structural homology is particularly marked for the isolated Fbd, with a root mean square deviation in the 1.2-to 1.3-Å range, while for the isolated Sbd, it ranges from 1.9 to 2.2 Å (Table 4). Although the general correspondence of secondary structure elements is maintained in both domains (Fig. 5C), the orientation of the Sbd is slightly different if the Fbd is used for the structural superimposition (Fig. 6A). Most importantly, the additional ␣ϩ␤ subdomain present in PmaLAAD-01N (Fig. 5C) is absent as such in the other related proteins, with the first 20 N-terminal residues of PmaLAAD-01N either absent in 1Y56 and in 1C0I or differently oriented if compared with the N-ter-minal region of 3ADA and 2GAG (Fig. 6B). We cannot exclude that the absence of the anchorage N-terminal helix in PmaLAAD-01N, as compared with the full-length enzyme, might affect the conformation of the following region.
FAD-binding Site-FAD binds to PmaLAAD-01N in an extended conformation similar to that found in related FADbinding proteins (Fig. 7A) (41). Forty three residues are directly involved in FAD binding (distances below 4.5 Å), 16 of which establish electrostatic or polar interactions with the co-factor. Only five of the latter interactions involve side-chain atoms (residues Glu-85, Lys-86, Gln-93, Ser-94, and Thr-442), the others being due to main-chain atoms. The N-terminal ends of the helix dipole of A14 and of A1 are pointed toward the O2 position of the isoalloxazine ring, and toward the pyrophosphate group of FAD, respectively. These two dipoles are typical of GR2 family members.
Overall, the isoalloxazine ring is quite exposed to solvent, especially at the N5 reactive site, located at the interface between the Fbd and the Sbd. It establishes hydrogen bonds with the N atom of Met-441 and the N and OG1 atoms of Thr-442 (FAD O2 atom), the N atoms of Ser-99 and Gln-100 (FAD O4 atom), and the N atom of Tyr-98 (FAD N5 atom), while displaying van der Waals interactions with Gln-93, Arg-96, Ala-97, Tyr-98, Ser-99, Gln-100, Leu-279, Gln-281, Ala-410, Val-411, Val-412, Val-438, Trp-439, Gly-440, Met-441, and Thr-442 side chains. The FAD O4 atom is further hydrogen bonded to a water molecule located inside the active-site cavity (Fig. 7B). Furthermore, the model of PmaLAAD-01N in complex with N5-sulfo-FAD (generated on the basis of the isoalloxazine ring of the N5-sulfo-flavin mononucleotide, PDB code 1QCW) shows that the close vicinity of the PmaLAAD-01N helical A2 segment to the FAD N5-reactive site would provide steric clash of the bound sulfite with the A2-␤4 loop (Fig. 7C), thus preventing sulfite binding (see above). In addition, the lack of such a peculiar feature of the flavo-oxidases is also supported by the absence of positively charged groups in proximity of the N1-C2ϭO locus of the isoalloxazine ring, whose presence would inductively promote the reaction with sulfite. In Pma-LAAD, a partial positive charge may be only provided by the dipole associated with the N terminus of the nearby helix A14.
Substrate Binding Site-To gain structural insight into the substrate binding mode of the enzyme, the structure of Pma-LAAD-01N in complex with anthranilate (possessing both a carboxyl and an amino group, thus mimicking L-amino acids) was determined at 1.75 Å resolution (R factor ϭ 15.6%, R free ϭ 19.7%). Data collection and refinement statistics are reported in Table 1. The analysis of the calculated electron density map showed the substrate-analogue bond near the FAD isoalloxazine moiety (Fig. 8A).
Binding of anthranilate does not alter the PmaLAAD-01N global tertiary structure (root mean square deviation range 0.43-0.59 Å between complex and native structures, depending on the superimposed protein chains) nor promote quaternary assembly variation. Local differences in the backbone conformation are, however, evident in the ␤18-␤19 loop, in helix A9, and the following loop (Fig. 8A). In particular helix A9, which is a 3 10 helix in the protein, is reorganized as a regular ␣-helix in the PmaLAAD-01N⅐anthranilate complex. Notably, this region is part of the PmaLAAD-01N-specific ␣ϩ␤ subdomain and is located at the entrance of the substrate binding pocket.
Such an entrance is unusually wide (15-20 Å, calculated from the C␣ position of residues located on opposite sides of the entrance) and mainly negatively charged (Glu-109, Asp-145, Asp-150, Glu-154, Glu-341, Asp-417, Glu-418) (Fig. 8F). The substrate binding pocket is deep (Ϸ20 Å) and mostly hydrophobic; the anthranilate aromatic ring is stacked to the FAD isoalloxazine ring, at a distance of Ϸ4 Å, and is located in a pocket lined by Leu-279, Phe-318, Val-412, Val-438, and Trp-439. The ligand carboxyl group is H-bonded to the NE and NH 2 atoms of Arg-316 side chain (2.8 and 3.2 Å, respectively), to the NE2 atom of Gln-100 side chain (2.9 Å), and to the FAD O4 atom (3.3 Å), although the 2-amino group of the ligand is H-bonded to the carbonyl oxygen of Val-438 (3.2 Å). Upon ligand binding, Arg-316 changes its conformation dramatically and inserts its side chain into the active site and forms an H-bond, together with Gln-100, with anthranilate at the re side of the isoalloxazine ring (Fig. 8, A and B). Additionally, the Phe-318 also changes rotamer, orienting its side FIGURE 5. Crystal structure of PmaLAAD-01N, topology, and structure-based sequence alignment. A, schematic view of the domain spatial arrangement in PmaLAAD-01N. The FAD binding domain is in green (with the ␣ϩ␤ additional subdomain in blue); the substrate binding domain is in orange; ␣-helices, ␤-strands, and coils are represented by helical ribbons, arrows, and ropes, respectively. The FAD cofactor is shown in stick representation (yellow color). The N and C termini of PmaLAAD-01N are labeled. B, secondary structure topology diagram of PmaLAAD-01N; cylinders and arrows indicate helices and strands, respectively. C, structure-based sequence alignment of PmaLAAD-01N. PmaLAAD-01N is aligned with L-proline dehydrogenase from P. horikoshii (PDB code 1Y56), glycine oxidase from B. subtilis (PDB code 1C0I), and heterotetrameric sarcosine oxidases from Corynebacterium sp. U-96 (PDB code 3ADA) and from Stenotrophomonas maltophilia (PDB code 2GAG). The sequence alignment has been performed using the ClustalW program and manually corrected based on their three-dimensional structure comparisons. PmaLAAD-01N secondary structure elements are shown on the top of the alignment and shaded in gray (yellow for 3 10 helices) for all aligned proteins. Asterisks and circles indicate PmaLAAD-01N residues interacting with the FAD isoalloxazine ring and with anthranilate, respectively. chain parallel to the ligand aromatic ring, at a distance of Ϸ5 Å, and the structural changes occurring in the 316 -318 region are transmitted to the adjacent 333-343 region, which is located at the entrance of the substrate binding pocket (Fig. 8, A and B).
The PmaLAAD-01N⅐anthranilate complex is well suited to analyze the binding mode of L-Phe (Fig. 8C). The PmaLAAD-01N enzymatic activity recorded on the synthetic amino acid L-Phe ethyl ester (33% in comparison with L-Phe) finds its structural explanation in the presence of a pocket lined by Tyr-98, Ala-314, and Arg-316, where the ethyl ester moiety of the substrate fits nicely (Fig. 8D). In contrast, binding of the D-isomer of Phe would instead result in a steric clash between its aromatic side chain and the FAD isoalloxazine moiety (Fig. 8E), in keeping with the absence of activity assayed for D-amino acids.

Discussion
PmaLAAD is a FAD-containing enzyme that catalyzes the stereoselective deamination of L-amino acids into the corre-FIGURE 6. Structural comparison with related proteins. A, structural comparison is produced by superimposing the FAD binding domains of Pma-LAAD-01N (green, ␣ϩ␤ subdomain, blue) with L-proline dehydrogenase from P. horikoshii (PDB code 1Y56, orange), or glycine oxidase from B. subtilis (PDB code 1C0I, magenta), or heterotetrameric sarcosine oxidase from Corynebacterium sp. U-96 (PDB code 3ADA, cyan) and from S. maltophilia (PDB code 2GAG, yellow). B, different orientation of the N terminus in PmaLAAD-01N relative to the heterotetrameric sarcosine oxidase from Corynebacterium sp. U-96 (PDB code 3ADA, cyan) and from S. maltophilia (PDB code 2GAG, yellow) is shown. PmaLAAD-01N Leu-52 is labeled to indicate the position where the backbone of the three proteins starts to show a good superimposition. In both panels the FAD cofactor bound to the PmaLAAD-01N FAD-binding domain is shown in stick representation (green color).  (47). The map refers to the high resolution data at 1.75 Å, but a similar result can be obtained also from the 2.0 Å resolution data (see Table 1). B, FAD molecule (yellow) and the protein residues (green) located at the active-site cavity relevant for FAD binding are shown in stick representation and labeled. A water molecule is represented as red sphere (0.3 van der Waals radius). H-bonds are shown by dashed lines. C, model of the N5-sulfo-FAD molecule bound to Pma-LAAD-01N based on the isoalloxazine ring of a N5-sulfo-flavin mononucleotide molecule (PDB code 1QCW).
sponding ␣-keto acids and ammonia without forming hydrogen peroxide, thus differing from known LAAOs. PmaLAAD shows a broad substrate specificity with a preference for neutral, aromatic L-amino acids; the highest activity is apparent for L-Phe, L-Leu, L-Met, and L-Trp (Table 2). This agrees with the size and microenvironment of the substrate binding pocket, which is lined by hydrophobic residues (Fig. 8B).
PmaLAAD is a membrane-bound enzyme. Subcellular fractionation and deletion of N-terminal segments demonstrated that PmaLAAD is anchored to the membrane through a puta-  tive transmembrane ␣-helix (residues 8 -28). The cellular localization is fundamental for enzyme function; accordingly, we demonstrated that PmaLAAD is fully active in the presence of membranes only. The purified full-length and membrane-devoid PmaLAAD-00N variant is practically inactive, and the catalytic competence is acquired by adding exogenous crude E. coli membranes but is only marginally increased using synthetic membranes produced from highly pure E. coli lipid extracts (Table 3). These results indicate that PmaLAAD requires a specific membrane partner for its activity, rather than a generic hydrophobic environment of the membrane. From spectral evidence (Fig. 4B), and because cytochrome c is not considered to be part of the main electron transport chain in the Proteus genus (37), we propose that the physiological electron acceptor from PmaLAAD could be a b-type cytochrome, a membrane protein that accepts electrons from different donors.
We demonstrated that the appearance of an absorbance spectrum likely corresponding to reduced cytochrome b is observed only under anaerobic conditions, although under aerobiosis electrons are transferred to dioxygen by the respiratory chain, as observed using the polarographic assay. Indeed, Pma-LAAD-00N re-oxidized very slowly on dioxygen, although in DAAOs and LAAOs FADH 2 reacted quickly with O 2 (4,34,35); in PmaLAAD the rate constant for FADH 2 re-oxidation by O 2 is Ϸ100-fold slower than the turnover number. This limited direct O 2 re-oxidation of FADH 2 explains the inability of Pma-LAAD to produce hydrogen peroxide and classifies this enzyme outside the oxidase class of flavoenzymes (33), whose members represent the closest structural relatives (Figs. 5C and 6).
It should be noted that reactivity of flavoproteins toward dioxygen is related to the microenvironment of their isoalloxazine ring, in particular to the presence of positively charged residues in proximity of the FAD C4a position. For instance, in LAAO from snake venom, Lys-326 is present on the si face of FAD, and in glucose oxidases two His residues are located on the re face of the cofactor (48,49). In contrast, in PmaLAAD, the corresponding positions in the vicinity of the isoalloxazine moiety of FAD are occupied by the neutral Gln-281 and the hydrophobic residues Trp-439 and Val-438. Nevertheless, the isoalloxazine ring is exposed to solvent in PmaLAAD, especially at the N5 site, and its location is compatible with electron transfer to a bound acceptor protein.
The FAD cofactor is bound in an extended conformation (Fig. 7A) and establishes a high number of interactions with the PmaLAAD apoprotein, comprising the interaction of helix dipole A14 with the O 2 of the isoalloxazine ring that is expected to stabilize the anionic form of FADH 2 . The isoalloxazine ring is located at the interface between the substrate and the FAD binding domains and is exposed to solvent, especially at the N5 site; its location is compatible with electron transfer to the bound acceptor protein.
The overall fold of PmaLAAD resembles that of known amine or amino acid oxidases/dehydrogenases (Fig. 6) but with specific structural features that allow it to fulfill a different role, i.e. the deamination of L-amino acid with no hydrogen peroxide production to fuel the membrane electron-transfer (respiratory) chain. Indeed, phylogenetic analysis demonstrates that proteins highly homologous to PmaLAAD (Ͼ28% sequence identity) are present only in the proteobacteria phylum, and in particular in the subphyla of alpha-, beta-, and gamma-proteobacteria (Fig. 9). Noteworthy is that all related sequences belong to putative flavoprotein oxidases or deaminases with unknown function, although the sequence identity with known microbial DAAOs or LAAOs is quite low (i.e. between 13.9 and 16.4%). This finding suggests that LAADs diverged early from amino acid oxidases but conserved a similar active-site organization related to the use of the same substrate. In common with known amino acid oxidases, PmaLAAD shows not only the fold but also the ability to generate the anionic semiquinone form of FAD during photoreduction and to produce charge-transfer complexes following binding of carboxylic acids.
Different from other flavooxidases, PmaLAAD possesses an additional ␣ϩ␤ subdomain that is placed in a key (structural and functional) region of the protein. As a matter of fact, the putative transmembrane ␣-helix of PmaLAAD is bound to the N-terminal end of this domain (residues 29 -56), and its C-terminal end encompasses the region of the active-site entrance (helix A9 and following loop) that changes conformation upon ligand binding (Fig. 8A). Indeed, comparison of the free and anthranilate-bound structures of PmaLAAD-01N shows that part of the active site, in particular Arg-316 (which binds the carboxyl group of substrate) and Phe-318 (which forms hydrophobic interactions with the apolar moiety of the ligand), only acquires the correct geometry in the presence of the substrate (Fig. 8A). We propose that these conformational changes could favor the binding (by electrostatic interactions) and correct orientation of the substrate at the active site of the enzyme. This behavior is not observed in canonical flavoprotein oxidases such as DAAOs or LAAOs (4,34,50).
In summary, we solved the first three-dimensional structure of an L-amino acid deaminase and set up a procedure to prepare a stable and soluble preparation of the PmaLAAD form (containing membrane fragments) that can be employed in cell-free biocatalysis. Overall, these results provide the basis for pushing protein engineering studies aimed to manipulate LAAD substrate specificity to fulfill different biotechnological requests, such as the resolution of racemic mixtures of natural and unnatural amino acids or the generation of novel, synthetic biochemical pathways (1,51).
Author Contributions-P. M. and G. M. performed biochemical experiments, analyzed the experimental data, and wrote the manuscript. M. N. performed the crystallization trials, solved the threedimensional structure, and wrote the manuscript. L. P. designed the experiments, analyzed the data, and wrote the manuscript.