Molecular modeling and site-directed mutagenesis reveal essential residues for catalysis in a prokaryote-type aspartate aminotransferase.

We recently reported that aspartate (Asp) biosynthesis in plant chloroplasts is catalyzed by two different Asp aminotransferases (AAT): a previously characterized eukaryote type and a prokaryote type (PT-AAT) similar to bacterial and archaebacterial enzymes. The available molecular and kinetic data suggest that the eukaryote-type AAT is involved in the shuttling of reducing equivalents through the plastidic membrane, whereas the PT-AAT could be involved in the biosynthesis of the Asp-derived amino acids inside the organelle. In this work, a comparative modeling of the PT-AAT enzyme from Pinus pinaster (PpAAT) was performed using x-ray structures of a bacterial AAT (Thermus thermophilus; Protein Data Bank accession nos. 1BJW and 1BKG) as templates. We computed a three-dimensional folding model of this plant homodimeric enzyme that has been used to investigate the functional importance of key amino acid residues in its active center. The overall structure of the model is similar to the one described for other AAT enzymes, from eukaryotic and prokaryotic sources, with two equivalent active sites each formed by residues of both subunits of the homodimer. Moreover, PpAAT monomers folded into one large and one small domain. However, PpAAT enzyme showed unique structural and functional characteristics that have been specifically described in the AATs from the prokaryotes Phormidium lapideum and T. thermophilus, such as those involved in the recognition of the substrate side chain or the "open-to-closed" transition following substrate binding. These predicted characteristics have been substantiated by site-direct mutagenesis analyses, and several critical residues (valine-206, serine-207, glutamine-346, glutamate-210, and phenylalanine-450) were identified and functionally characterized. The reported data represent a valuable resource to understand the function of this enzyme in plant amino acid metabolism.


INTRODUCTION
Aspartate aminotransferase (aspartate: 2-oxoglutarate aminotransferase, EC 2.6.1.1; AAT) catalyzes the reversible transamination reaction between aspartate and 2-oxoglutarate to give glutamate and oxaloacetate via a ping-pong bi-bi mechanism. AAT enzymes have been classified into the aminotransferase family I and then divided into two subfamilies, Iα and Iβ, according to their amino acid sequence identities (Jensen and Gu, 1996). To date, subfamily Iα includes AATs from eubacteria and eukaryotes while subfamily Iβ includes those from bacteria and archaea. The amino acid sequence identities between members of subfamily Iα (about 40%) is slightly higher than the identities between members included in subfamily Iβ (30-35%). When Iα and Iβ sequences are compared only about 15% identity can be observed.
Many X-ray crystallographic studies have been performed on enzymes of subfamily Iα to elucidate their structure, function and catalytic mechanism. These include AATs from Escherichia coli (Okamoto et al., 1994), chicken (Malashkevich et al., 1995) or pig (Rhee et al., 1997). These studies showed that all of these enzymes have a very similar three-dimensional structure and the same homodimeric quaternary structure consisting of identical monomers with a molecular size of about 45 kDa (Kallen et al., 1985). Each polypeptide chain is folded into a large domain and a small domain on the basis of the correlated motion of the N-terminal and C-terminal parts of the polypeptide chain upon inhibitor binding. The active site pocket is located at the interface between both domains. The PLP coenzyme resides at the bottom of the active site and forms a Schiff base with a lysine residue. A large conformational change in the small domain towards the large domain (open to closed form) has been deeply described in the AAT subfamily Iα which occurs when a C4 dicarboxylic substrate (aspartate) binds close to the active site of the enzyme (McPhalen et al., 1992;Jäger et al., 1994;Okamoto et al., 1994). That means that the catalytic AAT reaction proceeds in the closed structure of the enzyme when aspartate is transaminated to 2-oxoglutarate. The closed structure is maintained by electrostatic and hydrophobic interactions between residues located in the active site and the substrate. Recently, a detailed analysis of the AAT structure for the reverse reaction using C5 dicarboxylic substrates suggested that the Michaelis complex with glutamate instead of aspartate presents the open conformation (Islam et al., 2005). Interestingly, a very similar 3D structure, based on X-ray crystallography, has been observed for AAT members of subfamily Iβ when were X-ray analyzed Kim et al., 6 2003), despite the observed divergencies in primary structure. In the AAT of Thermus thermophilus HB8 (ttAAT), a member of subfamily Iβ, a large conformational change from the open to the closed form has been described (Nakai et al., 1999).
Amino acid sequence comparison clearly shows that critical residues in the active site are conserved in all AAT enzymes belonging to subfamily Iα. When the primary structure of AAT enzymes from subfamily Iα and Iβ are compared, most of the residues that are involved in the active site and that are essential for the catalytic mechanism seem to be conserved, desspite the low identity between both AAT types (Nakai et al., 1999). Single residues are implicated in the catalytic mechanism along with other essential residues that are not conserved in subfamily Iβ. An example is the amino acid Arg292 which is involved in the recognition of the distal carboxylate of the aspartate substrate in subfamily Iα. In subfamily Iβ the same role seems to be carried out by the Lys109 residue (Nobe et al., 1998).
The AAT enzyme is present in plants as a family composed by at least five different isoenzymes associated with different subcellular compartments (cytosol, chloroplast, mitochondria and peroxisome). An increasing number of cDNA sequences encoding AAT isoenzymes have been reported in different plants such as alfalfa (Gantt et al., 1992), Arabidopsis (Schultz and Coruzzi, 1995;Wilkie et al., 1995), broomcorn millet (Taniguchi et al. 1995), carrot (Turano et al. 1992), lupine (Reynolds et al., 1992) and soybean (Wadsworth et al., 1993). Comparative analysis of amino acid sequences indicates that plant AAT isoenzymes are in the same protein family as the vertebrate AATs and bacterial AATs of subfamily Iα (Wadsworth, 1997). No X-ray crystallographic studies for any plant AAT have been performed. Furthermore, only one protein structure has been modeled (Wilkie et al., 1996), the plastidic isoform of eukaryotic-type AAT from Arabidopsis thaliana which is encoded by the At4g31990 locus.
We have recently reported the existence in plants of a novel form of AAT (PT-AAT) with a high degree of similarity to the enzymes from cyanobacteria and archaea (de la Torre et al., 2006;de la Torre et al., 2007). This finding constitutes the first evidence of AAT belonging to subfamily Iβ in eukaryotic organisms. The gene encoding PT-AAT is highly expressed in photosynthetically active tissues and the enzyme is located in the chloroplast. A putative endosymbiotic origin has been proposed for this enzyme based on its subcelullar localization and sequence similarities. The aim of our current investigation is to find out the differential role of this PT-AAT with regard to the already described members of the plant AAT gene family. Structural modeling of PpAAT and site-direct mutagenesis experiments should help in 9 Computer modeling of the structure of P. pinaster AAT The amino acid sequences of the P. pinaster PT-AAT (PpAAT) and other members of subfamily Iβ for which a 3D structure is known Kim et al., 2003) were compared. It was found that the AAT enzymes from the cyanobacteria P. lapideum (plAAT) and from T. thermophilus (ttAAT) are 45 and 42% identical to PpAAT, respectively. Since the ttAAT structure has been resolved in both open and closed conformations, it was selected as a template for the protein modeling of PpAAT. The accuracy of comparative models decreases sharply below the 30% sequence-identity cutoff, mainly as a result of a rapid increase in alignment errors (Madhusudhan et al., 2005). Since the sequence identity between ttAAT and PpAAT is 42% we are in the "safe modeling zone", and, our model should not have significant errors, assuming a standard alignment. Nevertheless, a manually created alignment was used for the initial modeling. The complete sequence for PpAAT was larger than that of the template ttAAT. Therefore, the plastid targeting peptide plus fifteen residues in the PpAAT N-terminus and ten residues in the C-terminus were not included in the model. The region under analysis consisted of the amino acid residues from Asp81 to Leu478 in the PpAAT primary sequence. Along this polypeptide fragment, PpAAT and ttAAT are 42% identical. To achieve a reliable model of the quaternary structure for both the apoenzyme and PpAAT in the internal aldimine form, we relaxed the system via energy minimization and molecular dynamics simulations over 1 ns. Based on this, we have obtained the basic structure hat will be used for further spectroscopical and mutagenesis studies.
There is no resolved structure for plant type Iβ AAT and only one plant type Iα, has been reported to date (Wilkie et al., 1996). Our model represents the first molecular model for a type Iβ AAT in eukaryotes. The modeled enzyme exists as a homodimer based on the experimental data previously described by de la Torre et al. (2006). The way the two subunits might fit together is illustrated in Figure 2. The dimer contains two equivalent active sites, each of which is formed by residues from both subunits. The overall folding of the PpAAT model resembles the known structures for other AATs belonging to both subfamily Iα and Iβ such as E. coli (Okamoto et al., 1994), yeast (Jeffery et al., 1998, pig (Rhee et al., 1997), chicken (Malashkevich et al., 1995, T. thermophilus P. lapideum (Kim et al., 2003).
Nearly all the AAT enzymes that have been studied are composed of two identical monomers. Each monomer is folded and is comprised of a large domain, a small domain and an extended N-terminal arm, the end of which interacts with the other monomer. division was first established for chicken mitochondrial AAT and for pig cytosolic AAT on the basis of the correlated motion of N-terminal and C-terminal parts of a polypeptide chain upon inhibitor binding (Malashkevich et al., 1995;Rhee et al., 1997). The AAT domain distribution described in these enzymes was later used to establish the ttAAT domain distribution (Nakai et al., 1999). According to the model developed in the present work each monomer of the PpAAT enzyme is equally composed by a small and large domain (Fig. 2B). The domains in the PpAAT monomer were identified through comparison to the primary sequence of the closely-related ttAAT polypeptide (Nakai et al., 1999). The small domain is formed by two parts of the polypeptide chains, spanning Lys92 to Asn124 and Met376 to at least Leu478.
The large domain contains the residues between Asn124 and Met376. Residues 81 to 92, which are part of the N-terminal arm, are located at the surface of the large domain of the opposite monomer (Fig. 2B). The quaternary structure representation of PpAAT, based on the model described in the present study, is fully in agreement with the previously described domain distribution as determined by comparison with the ttAAT amino acid sequence. These data support the accepted conclusion that all AAT enzymes, independent of whether they belong to subfamily Iα or Iβ, display a very similar three-dimensional structure Nakai et al., 1999). Comparison of the domain distribution in the prokaryotic and eukaryotic types of AAT (PpAAT, ttAAT and SsAAT ) clearly shows how these enzymes which differ considerably in their primary structure maintain a highly similar tertiary and quaternary structure (Fig. 3). However, an asymmetric distribution of the hydrophobic amino acids histidine, phenylalanine, tryptophan and tyrosine over the enzyme surface has been observed. This molecular feature was particularly evident in the large domain of the prokaryotic type enzyme but absent in the eukaryotic-type AAT.
The secondary structure of PpAAT was compared to that of ttAAT and a high level of correlation was found in the distribution of α-heli and ß-strand motifs. Each PpAAT monomer is composed of 16 α-helices and 11 ß-strands ( Figs. 1 and 3). The core of the large domain is formed by a wide ß-sheet structure, which is comprised of six ß-strands arranged parallel (p) and one anti-parallel (a) with a distribution of pppppap. The seven ß-strands have a tendency to twist right-handedly, which appears to be a ß-sheet with a left-handed twist when viewed along the ß-sheet normal to strands. This structure is highly conserved in other AAT enzymes, including the chloroplastic isoform of Arabidopsis which is an eukaryotic-type AAT (Wilkie et al., 1996). The small domain is comprised of four short ß-strands and five α-helices. The ßstrands are grouped pairwise in a parallel-antiparallel conformation and are arranged into two small ß-sheet regions. As described in other subfamily Iβ AATs, a high proline content was found when compared to the low content reported for mesophilic AATs. In addition, the thermolabile amino acids cysteine and asparagine are represented at a low level.

Structural analysis of the PpAAT active center and prediction of residues involved in the recognition of acidic substrates
The structure of the active center of PpAAT was deduced using information derived from two sources. The first one comprises the prediction derived from both the model and the comparison with X-ray crystal structures corresponding to other subfamily Iβ . The second source of information is composed of experimental results describing the residues involved in the catalytic mechanism reported in AATs from different organisms, such as E. coli or pig (Sus scrofa). The model developed for PpAAT follows the pattern previously described in plAAT and ttAAT with conservation of the main features of the active-site region ( Fig. 4A and 4B).
The essential residues Tyr144, Trp205, Tyr287, Lys318, and Arg457 are quite well conserved. The PLP cofactor is virtually positioned within the predicted region in the active site attached to Lys318, via a covalent imino linkage between C-4´ of the cofactor and the εamino group of the lysine residue. Lys 318 is the equivalent residue in PpAAT to the pig cytosolic enzyme's Lys258 (Metha et al., 1989). The relative positions of the other residues involved in the interaction with the cofactor were located (not shown). These interactions, which are important for the correct positioning of the cofactor within the active site, seem to be satisfied by our model.
The stabilization of the acidic substrates, glutamate and aspartate, within the AAT enzyme is mainly carried out by the interaction of its carboxylic groups (α and γ) with specific residues in the active center. Arg457 in PpAAT is conserved in the AATs from both the Iα and Iβ subfamilies. The side chain of this particular residue stabilizes the α-carboxylate group of the substrate in all known AAT enzymes. In addition, AAT enzymes of subfamily Iα have an arginine residue, which interacts with the distal carboxylate group of dicarboxylic (acidic) substrates and enhances the optimal catalytic positioning. This residue is characteristic of Iα AATs and is absent in the Iβ type where a lysine residue has a similar role, as has been reported for the ttAAT enzyme (Nobe et al., 1998). This lysine residue (Lys 181 in the PpAAT model) is essential for the recognition of acidic substrates and is highly conserved in all Iβ AATs. However, it has been suggested that Lys 181 may require assistant residues for substrate recognition since arginine is much more preferred for this purpose in many different enzymes (Nobe et al., 1998). Potential candidates for this role in the PpAAT active site are the conserved residues Val 206, Ser 207 and Gln346 (Fig. 4).

Spectrophotometric properties of PpAAT enzyme
Affinity purified preparations of the PpAAT recombinant enzyme were spectrophotometrically characterized between 250 and 550 nm (Fig. 5). The enzyme showed a maximum in the absorption at 280 nm, which is due to the aromatic residue side-chains, and another maximum at 385 nm, which is due to the bound cofactor. The PpAAT apoenzyme was separated from the PLP cofactor by incubation of the enzyme in the presence of phenylhydrazine (Fig. 5A).
The spectral curve corresponding to the apoenzyme was determined between 250 and 550 nm and a single maximum at 280 nm could be detected. When the spectral curve for PpAAT was determined in the presence of 10 mM aspartate the pyridoxamine-phosphate (PMP) form of the enzyme was detected as an absorption peak close to 329 nm (data not shown). This value is very close to the absortion peak at 327 nm observed for neutral aqueous PMP with a dipolar ionic ring and protonated 4'-amino group (Kallen et al., 1985). When the pH in the buffer containing the enzyme was reduced to acidic conditions the enzyme showed an absorption band at 280 nm and the other maximum shifted to 425 nm. The enzyme showed higher activity when the pH was neutral or slightly alkaline conditions where the band at 385 nm is the most prominent one (Fig. 5B). Similar spectral behavior related to pH changes has been extensively reported (Kim et al., 1993). These data suggest that the two forms are derived from the difference in the ionization state of the nitrogen atom of the internal Schiff base formed between the aldehyde group of PLP and the amino group of a lysyl residue of the enzyme (Braunstein, 1964). The isosbestic point was determined to be close to 406 nm, when considering the distinct spectral curves at different pHs.

Characterization of residues putatively involved in the interaction between PpAAT, acidic substrates and PLP
In order to precisely determine the topology of interactions between the protein's active center, its substrates and the cofactor PLP, we decided to introduce point mutations affecting key residues. Thus, three residues with a putative role as "assistants" in the stabilization of the substrate were selected in the PpAAT model. These were Val206, Ser207 and Gln346.
Val206 is located between two residues involved in the stabilization of the substrate but its hydrophobic side chain is predicted to be in the opposite direction. Ser207 was selected since its spatial position close to Lys181 and it has an orientation that may putatively involve it in 13 the stabilization of the distal carboxyl group of substrate and PLP. Gln346, a residue that belongs to the other subunit in the dimer (Fig. 4), could also be involved in these interactions. In addition, these three residues are all conserved in Iβ type enzymes and absent in Iα type enzymes ( Fig. 1). Site directed mutagenesis was used to study the functional significance, if any, of these selected residues.
In our model the hydrophobic side chain of Val206 is immediately close to Trp205, the amino acid residue which stabilizes the cofactor in conjunction with Tyr287. In fact, the PLP pyridine ring is nearly parallel to the indole ring of Trp205 (Fig. 4). The PpAAT Val206 was replaced by a polar and hydrophilic serine residue (V206S). The mutant protein was overexpressed and affinity purified ( Fig. 6A and 6B), and its activity was determined to be about 70% of that observed for the WT enzyme ( Fig. 6C). A significant increase in the K m values for the substrates of the AAT in the forward reaction, with aspartate and 2-oxoglutarate as substrates, was observed whereas there was no change in the affinity of substrates for the reverse reaction, with glutamate and oxaloacetate (Table I). Nevertheless, there was a considerable decrease in the absorbance of the cofactor when the absorption spectra corresponding to the V206S mutant was compared with the one corresponding to the WT enzyme (data not shown).
The oxygen of the Ser207 hydroxyl is located close (2.6 Å) to the ε-amino group of Lys181 in PpAAT. In other Iβ AAT enzymes, a similar distance which ws 2.7 Å in ttAAT and 3 Å in plAAT, was measured for the equivalent residues. Serine, which is a polar and hydrophilic residue, in position 207 was replaced by valine, which is an aliphatic and hydrophobic residue, with the goal of finding whether or not it has a role on the stabilization of the acidic substrate within the active center. Once the protein was overexpressed in E. coli and affinity purified ( Fig. 6A and 6B) with the His-tag technology, the enzyme activity was measured in both forward and reverse AAT reactions and in situ AAT activity determined on native gels. No activity was detected in any of these assays for the S207V mutant enzyme was analyzed ( Fig 6C). Furthermore spectrophotometrical characterization of the S207V mutant protein showed that there was a transition of the characteristic absorption maximum at 383 nm to 331 nm, which is close to the absorption maximum observed for the PMP form of the enzyme (Suppl. affect the catalysis. After site-directed mutagenesis and the corresponding purification of the two mutant proteins (Q346K and Q346E), no enzyme activity was detected in the enzyme preparations (Fig. 5).

Characterization of residues putatively involved in the relative domain movement from "small to large" in PpAAT
Additional single-mutant proteins were designed in order to study in depth the reported mechanism by which the small domain turns towards the large domain in the presence of the substrate. Glu210 was replaced by alanine, aspartate or lysine through site-directed mutagenesis to generate the E210A, E210D and E210K mutants. In the WT enzyme, Glu210 seems to be linked through an electrostatic interaction with the Lys92 residue in the pocket determined by surfaces corresponding to both the small and large domains (Suppl. Fig. 2). The enzyme variants E210A, E210D and E210K were overexpressed, affinity purified and functionally characterized (Fig. 6). The products of E210A and E210K mutants were almost inactive ( Fig. 6C) and, since the recovered activity was very low, no kinetic parameters could be determined (Table I). In contrast, the E210D mutant retained about 10% of the activity exhibited by the WT (Fig 6C). When the K m values for the substrates were determined a slight decrease in the affinity for substrates was observed in both the forward (aspartate and 2-oxoglutarate) and the reverse reaction (glutamate and oxaloacetate) ( Table I). All these results support the relevance of the acidic lateral side chain of Glu210 for enzyme function The Phe450 is a residue common to all subfamily Iβ ΑΑΤs and it is not present in the subfamily Iα enzymes. This hydrophobic residue is located in the same pocket described previously between the large and small domain. This pocket is a "hydrophobic-rich" region into the protein, with an internal structure that varies between the open and the closed form. The existence of a "hydrophobic patch" has been proposed in the case of E. coli AAT as a part of a mechanism involved in the "driving forces" needed to complete the transition from the Michaelis complex to the external aldimine in the reverse reaction (Jäger et al., 1994;Okamoto et al., 1994;Islam et al., 2005). Using site-directed mutagenesis the Phe450 was replaced by serine (F450S), a hydrophilic residue, in order to determine its possible role in the catalytic mechanism. This residue is located on the internal surface of the small domain in a location close to a "hydrophilic region" that corresponds to the large domain and is composed of the residues Trp205, Phe204 and Tyr208. When the activity was measured, a 50% decrease was observed with respect to the WT. The K m values for substrates in both the forward and reverse

Structural aspects of P. pinaster PT-AAT
In this paper the structure of maritime pine (Pinus pinaster) PT-AAT has been investigated in order to gain further insights on the function of this enzyme in plants. The computer modeling of the mature enzyme from P. pinaster (PpAAT) was compared with the previously reported PDB files of animal and bacterial AATs. A high level of similarity was observed in the spatial architecture of the enzyme, even when different levels were considered (Kallen et al., 1985;McPhalen et al., 1992;Jäger et al., 1994;Okamoto et al., 1994). The overall secondary structure described for PpAAT is highly conserved throughout all AAT enzymes. Some of the conserved key elements include the seven ß-strand-ß-sheet composing the core of the large domain, the amino terminal flexible α-helix, and the distribution of α-helices throughout the enzyme. The monomer is also organized into a large and a small domain that flanks the active site pocket as reported for the tertiary structure of other AAT enzymes. When the quaternary structure of PpAAT is analyzed, the docking of two identical monomers into a homodimeric enzyme is in agreement with the other AATs and with our previous results (de la Torre et al., 2006). Despite the high structural similarities, the identities between their amino acid sequences are low. These results suggest that the AAT spatial structures are highly limited by their "catalytic needs" and that only minor 3D variations are allowed.
Multiple alignments, which have been built from AATs of distant species clearly shows that most of the residues involved in the well-known catalytic mechanism described for the subfamiliy Iα (Okamoto et al., 1994;Rhee et al., 1997) are conserved in PpAAT. When the active-center in the PpAAT structure is compared with those from prokaryotic enzymes (T. thermophilus and P. lapideum) a nearly total identity is found. Similar characteristics are also found when PpAAT is compared with AATs from the subfamily Iβ for other features such as proline content (Kim et al., 1993), substrate specificity (Ura et al., 2001) or thermostability (Nobe et al., 1998). Taken together, these data suggest that the plant enzyme has retained the structural and functional properties of the prokaryotic AATs.
Previous data showed that PpAAT is a highly stable enzyme under a wide range of temperatures up to 75ºC (de la Torre et al., 2007). Its thermostability seems to be related to its amino acid composition with a high proline content (6.4 %), which results in an enzyme with a rigid and packed structure. The T. thermophilus AAT has also been reported to be a thermostable, rigid enzyme with the exception of the N-terminal α-helix (H1) that is comprised of the amino acid residues between 16 and 28. An identical helix with an equivalent position and orientation is present in the PpAAT enzyme and is comprised of the residues from 96-109. In both enzymes, this H1 α-helix is connected to the next α-helix (H2) by an almost completely conserved region. This linking region is composed mainly of small amino acids, which are AGEPDFNTP, and is common to all subfamily Iβ AATs that have the H1 motif. The binding of inhibitor maleate to ttAAT induces a conformational change from the open to the closed form, and this has also been reported for AAT subfamiliy Iα enzymes (Nakai et al., 1999).
This change does not affect the whole small domain, as only the N-terminal region that is mainly composed of the α-helix H1 approaches in order to close the active site (Nakai et al., 1999). The highly conserved region linking H1 and H2 is composed of small amino acids which allows an efficient rotation ue to the low rotation restriction around ψ and φ angles. in the PpAAT sequence that are absent in bacterial AATs. This particular region was depicted in the model as a short helix that is located outside the active center (Suppl. Fig. 3). Whether or not this structural feature has a potential role in the plant PT-AATs will require further studies.

Substrate recognition in the PpAAT enzyme
Dicarboxylate substrates are recognized, stabilized and correctly oriented into the AAT active-center through the interactions of their α and γ carboxyl groups with the side chains f residues that are located within the "catalytic pocket". The α-carboxyl group in the substrate is recognized and stabilized by an arginine residue in all AATs that have been described to date. The interaction between the distal (γ) carboxyl group and the enzyme is carried out in subfamily Iα AATs through an arginine residue. Based on previous work that described the recognition of the distal carboxyl group of acidic substrates into the T. thermophilus active center, it has been proposed that this role is mainly carried out by a lysine residue that is exclusively found in subfamily Iβ AATs (Nobe et al., 1998). The equivalent residue has been determined to be Lys181 in PpAAT. This residue has been designated as a major determinant of the acidic substrate specificity. The "Ra" type has been described as the mode of interaction between an arginine side chain and substrate´s carboxylic groups. This type of interaction describes the recognition of α carboxylic groups in all AATs and the interaction with the distal carboxylic groups in subfamily Iα. This interaction does not need additional residues to for correct charge compensation and it thought to be the most frequent type of interaction in enzymes that use carboxylic substrates. In contrast, the recognition of a carboxyl group through a lysine residue has been described as extremely rare. It has been proposed that assistant residues would be required to stabilize substrates in that case (Islam et al., 2005). Site-directed mutagenesis showed that the introduction of an Arg292 into the T. thermophilus AAT is a key step in the change of only the acidic substrate specificity in an enzyme that has dual acidic and hydrophobic substrate specificity (Ura et al., 2001).
We modeled the external aldimines for the C4 and C5 dicarboxylic substrates in the closed conformation, since it seems that the reaction proceeds when the enzyme is in its The Val206 residue is located very close to Trp205. Based on the structure described for ttAAT, the equivalent residue of Trp205 has been proposed to be involved in the stabilization by of the PLP pyridine ring by its side-chain (Nakai et al., 1999). In E. coli AAT, the N (1) of the equivalent residue of Trp205 seems to be partly involved in the binding of the distal carboxylate group of the dicarboxylic substrates (Hayashi et al., 1990). In PpAAT, however, we have observed that this interaction mainly occurs with C4 dicarboxylic substrates (Fig. 4).
The distances between the nitrogen atom from the indol group and the γ carboxylic group are 2.9 Å for the C4 substrate and 4.1 Å for the C5.
When V206 was changed to Ser, a significant decrease of the activity was observed (about 30%). The kinetic parameters for the reverse reaction were not altered, while the K m values for substrates in the forward reaction (aspartate and 2-oxoglutarate) were increased by an order of magnitude. The hydrophobic to hydrophilic (V206S) substitution seems to alter the correct orientation of the aromatic side chain of the Trp205, which should face the pyridine ring of PLP. This substitution likely altered the correct orientation of PLP within the active-center and, thus, affected the affinity for the substrates in the forward reaction. If so, the active center has been altered in a way that does not affect the reverse reaction. The Ser207 and Gln346 residues were also selected as candidates to be analyzed since they are unique residues that are close to Lys181 within the active site. Ser207 is from the same subunit and Gln346 is from the other one. Based on our PpAAT model, the distance between the Ser207 hydroxyl oxygen and the ε-amino nitrogen is 2.7 Å for the C4 substrate and 2.9 Å for the C5 substrate. Gln346 directly interacts with Ser207 and, therefore, may contribute to substrate stabilization. In order to study the importance of these two residues in the stabilization of Lys181 in the correct orientation, Ser207 was replaced by a valine whereas Gln346 was changed to either a basic (lysine) or an acidic (glutamate) residue. The replacement of Ser207 by the hydrophobic valine residue yielded a mutant protein that lacked the target bond that appears to be critical in the stabilization of Lys181. Similarly, the ability to establish hydrogen bonds at position 346 is lacking in the Gln346K and Gln346E mutants. The activity of the S207V, Gln346K and Gln346E enzymes was completely abolished in both forward and reverse AAT reactions (Fig. 6). These data indicate that S207 and Gln346 are essential "assistant" residues involved in the correct positioning of Lys181 and, therefore, involved in substrate recognition.
The negatively charged residue Glu210 was also selected for functional analysis again using site-directed mutagenesis. The model structure predicts that the negatively charged Glu210 residue can electrostatically interact with the positive charge of a Lys92 residue.

20
These two residues, Glu210 and Lys92, are located respectively in the large and small domains. Lys92 is specifically positioned in the small domain, just beside the 96-109 α-helix (H1). Therefore, it is deeply involved into the "open to close" conformational change, given that it is positioned in the H6 α-helix of the large domain. The mutant proteins E210A and E210K showed a dramatic decrease of AAT activity in relation to the WT enzyme, which suggested that the acidic lateral side chain of Glu210 is required for enzyme activity. The replacement of glutamate by aspartate further supports the above assumption. Thus, the E210D mutant enzyme was partially active although its activity was much lower than in the WT, and it exhibited decreased affinity for its substrates. Taken together, these data suggest that alteration in the electrostatic interaction between Lys92 and Glu210 afffects the ability of the enzyme to correctly develop the "open to close" catalytic step. Further functional studies of the residues in the H1 α-helix will be needed to gain a better understanding about this critical mechanism that is characteristic of the subfamiliy Iβ AATs.
Data derived from the characterization of the F450S mutant suggests that this well conserved residue has a relevant but not critical role in the enzyme. The reduced activity of the F450S mutant enzyme and the 3D location of Phe450 indicates that this hydrophobic residue seems to be a member of a "hydrophobic patch" that is located within the pocket formed between the large and small AAT domains. The existence of these "hydrophobic patches" has been previously reported in AAT enzymes (McPhalen et al., 1992, Jäger et al., 1994.

Importance of structural studies to understand the metabolic function of PT-AAT and biotechnological applications
The presence of Lys181, a residue with a shorter side chain than arginine, in the catalytic site of PpAAT possibly favours optimal interactions with the substrate glutamate. This could explain why lysine rather than arginine is conserved at this particular position in the subfamily Iβ AATs. We propose that Ser207 and Gln346 may function as assistants to fix the orientation of the distal carboxyl groups of the substrates. The structure of a more flexible catalytic site in the subfamily Iβ AAT enzymes could facilitate the recognition of glutamate. These structural features are consistent with the kinetic behavior of the enzyme. Thus, the kinetic properties of PpAAT for aspartate and 2-oxoglutarate are quite similar to those reported for the eukaryotictype AAT in plants. The affinity of PpAAT for glutamate (K m 1 mM) (de la Torre et al., 2006), however, is much higher than that reported for plastidic and other eukaryotic AATs, where the K m values range between 10 and 30 mM (Taniguchi et al. 1995;Wilkie and Warren, 1998). Under conditions of steady-state photosynthesis the concentration of glutamate in the chloroplast ranges between 15 and 50 mM (Riens et al., 1991;Winter et al. 1994), which is well above the K m value of PpAAT for this substrate. In contrast, the concentration of 2-oxoglutarate has been estimated to be in the micromolar range and even lower in the chloroplast stroma because the glutamine synthetase (GS)-glutamate synthase (GOGAT) cycle represents a strong sink (Weber and Flügge, 2002). Consequently, a high glutamate/2-oxoglutarate ratio in the stroma would drive biosynthetic transaminations that favor aspartate biosynthesis. Under these conditions the eukaryotic and prokaryotic forms of plastidic AAT would be saturated and the flux towards aspartate biosynthesis would be very high through the catalytic action of both enzymes (Fig. 7A). However, under physiological conditions where glutamate abundance is low (e.g. limited nitrogen availability and other stressful conditions), only the PT-AAT isoenzyme would ensure the biosynthesis of aspartate-derived amino acids and other N compounds in the plastids (Fig. 7B). A main pathway of aspartate utilization is the biosynthesis of the aspartate-derived amino acids lysine, threonine, isoleucine, and methionine, all of which are required for protein synthesis and produced exclusively in the plastid (Azevedo et al. 2006). All these amino acids are essential in the nutrition of animals where PT-AAT does not exist (de la Torre et al., 2006).
In cyanobacteria, inorganic nitrogen is assimilated through the GS/GOGAT cycle and turned into glutamate, which is then utilized as amino donor by AAT and other transaminases for the biosynthesis of nitrogen compounds (Muro-Pastor et al., 2005). PT-AAT could play a similar role in the chloroplasts of higher plants that utilize glutamate for amino acid biosynthesis inside that organelle. The results reported here suggest that the structure of the catalytic site is adapted for the optimal utilization of glutamate for the biosynthesis of nitrogenous

Computer modeling of the spatial structure of P. pinaster AAT
A homology model of P. pinaster AAT (residues 81-478) was built using crystal structures of AAT from T. thermophilus (ttAAT) as a template, in both open and closed conformations (PDBs 1BJW and 1BKG, respectively). The PpAAT sequence was isolated from a P. pinaster EST database of different woody tissues in a previous work (de la Torre et al., 2006). Alignments were performed with CLUSTAL W (Thompson et al., 1994) and manually edited. The whole PpAAT polypeptide fragment was modelled with Modeller 9v1 (Sali and Blundell, 1993) on a GNU/Linux box. Structure quality and minor structural conflicts were evaluated using the DOPE statistical potential within Modeller (Shen and Sali, 2006). The quaternary structure of the enzyme was built through comparison to that of ttAAT following the steps previously described for other PLP-dependent enzymes (Rodríguez-Caso et al. 2003;Moya-García et al. 2005;Moya-García et al. 2008). External aldimine was also modeled for both the C4 substrate aspartic acid and the C5 substrate glutamic acid with PyMOL (Delano, 2002a and2002b). Using the structures of maleic acid and 4'-deoxy-4'-aminopyridoxal-5'-phosphate (PMP) as scaffold, we built the PLP-glutamate and PLP-aspartate external aldimines (PLP-Glu and PLP-Asp, respectively) that were placed in the PpAAT active site by manual fitting and comparison with the positions of PMP and maleic acid in the active site of ttAAT X-ray crystal structure. We built three different enzymatic systems: PpAAT in closed conformation with PLP-Asp, PpAAT in closed conformation with PLP-Glu and PpAAT in open conformation (free enzyme). All of them were energy minimized and equilibrated using Langevin-Verlet molecular dynamics simulations (NVT ensemble) over 1 ns at a temperature of 298 K. The FORTRAN library DYNAMO (Field, 1999) and the OPLS force field were used to perform every simulation.

Site-directed mutagenesis via PCR
The PpAAT sequence used in this study was previously inserted in the NdeI-BamHI cloning site of pET-11a vector (de la Torre et al., 2006). Single mutations were introduced into the

24
Aspartate aminotransferase activity was determined for both forward and reverse reactions.
Direct reaction was determined by coupling the production of oxalacetate from aspartate and 2-oxoglutarate to the oxidation of NADH with malate dehydrogenase (Yagi et al., 1985). Reactions were developed in a mix solution containing 50 mM Tris-HCl pH 7.8, 50 mM aspartate, 10 mM 2-oxoglutarate, 0.07 mM PLP, 0.1 mM NADH, 2U malate dehydrogenase (Roche Farma S.A., Barcelona, Spain) and the appropriate amount of enzyme in a final volume of 700 µl. The reaction was started by the addition of 2-oxoglutarate. The decrease in the levels of NADH was measured at 340 nm for a period of 120 s. The reverse reaction was measured by coupling AAT activity to the reaction catalyzed by glutamate dehydrogenase in a mixture containing 15 mM glutamate, 1 mM oxaloacetate, 0.1 mM NADH, 5U of beef glutamate dehydrogenase (Roche), 0.07 mM PLP and 3.1 mM NH 4 Cl in 60 mM potassium phosphate buffer pH 7.5 at a final volume of 1 ml. The reaction was initiated by the addition of glutamate and followed by the decrease of NADH at 340 nm for a period of 120 s.
They were then placed in a bath containing 50 mL of AAT substrate solution with low shaking for 5 min. AAT activity was detected when the AAT substrate solution was supplemented with 1 mg/ml of Fast Blue (Sigma, St. Louis, MO, USA). The composition of the AAT substrate solution (pH 7.4) was 2.2 mM 2-oxoglutarate, 8.6 mM aspartate, 0.5% (w/v) polyvinyl pyrrolidone-40 (PVP-40), 1.7 mM EDTA and 100 mM Na 2 HPO 4 (Wendel and Weden, 1989).

Protein electrophoresis and western blot analysis
Proteins were separated by gel electrophoresis under denaturing conditions as described elsewhere (Cánovas et al., 1991). Both electroblotting from SDS gels to nitrocellulose membranes and immunodetection of AAT polypeptides were carried out as described elsewhere (de la Torre et al., 2002).

25
Pure PpAAT enzyme preparations were incubated at 50 mM phenylhydrazine (pH 7.4) at 37ºC for 1 h, followed by gel filtration on a Sephadex G-25 column equilibrated with 50 mM potassium phosphate buffer pH 7.4. Both phenylhydrazine-treated and -untreated protein samples were characterized by spectrophotometric absorption between 250 and 550 nm. Figure S1. Comparison of the absorption spectra of the PpAAT WT and S207V enzymes. Figure S2. Detailed view of the interaction between the large and small domains in the PpAAT structure. Figure S3. Location in the PpAAT structure of the short sequence stretch (10-11 amino acid residues) that is absent in bacterial enzymes

ACKNOWLEDGMENTS
We are grateful to Dr. Martí from the Universitat Jaume I, and to Dr. Ruiz-Pernía from the        drazine (short dashed). In both cases protein solutions were purified and concentrated to 1 mg/ ml. B) Absorption spectra of the PLP-form enzyme are represented at two different pH values. Solid black line corresponds to pH 5 and solid grey line corresponds to pH 8. The buffers used were 0.1 M sodium acetate for pH 5 and Tris hydrochloride for pH 8. In both cases, the protein solutions were purified and concentrated to 1 mg/ml. C, AAT activity was measured in the same aliquots following the method described by Karmen et al. (1955) for the forward reaction. AAT activities are shown as relative values with WT activity taken as 100% (5.6 nkats). AAT activity for the same samples was equally measured in gels following the method described by Wendel and Weeden (1989) (data not shown).