Structural Evidence of Active Site Adaptability towards Different Sized Substrates of Aromatic Amino Acid Aminotransferase from Psychrobacter Sp. B6

Aromatic amino acid aminotransferases present a special potential in the production of drugs and synthons, thanks to their ability to accommodate a wider range of substrates in their active site, in contrast to aliphatic amino acid aminotransferases. The mechanism of active site adjustment toward substrates of psychrophilic aromatic amino acid aminotransferase (PsyArAT) from Psychrobacter sp. B6 is discussed based on crystal structures of complexes with four hydroxy-analogs of substrates: phenylalanine, tyrosine, tryptophan and aspartic acid. These competitive inhibitors are bound in the active center of PsyArAT but do not undergo transamination reaction, which makes them an outstanding tool for examination of the enzyme catalytic center. The use of hydroxy-acids enabled insight into substrate binding by native PsyArAT, without mutating the catalytic lysine and modifying cofactor interactions. Thus, the binding mode of substrates and the resulting analysis of the volume of the catalytic site is close to a native condition. Observation of these inhibitors’ binding allows for explanation of the enzyme’s adaptability to process various sizes of substrates and to gain knowledge about its potential biotechnological application. Depending on the character and size of the used inhibitors, the enzyme crystallized in different space groups and showed conformational changes of the active site upon ligand binding.


Introduction
Transamination stands as a bright example of reaction, which is nowadays conducted almost exclusively by biotechnological means. [1] Enzymatic methods allow to elude multistage and complicated chemical reactions and grant an enantiomerically pure product with high efficiency. Aromatic amino acid aminotransferases (ArAT) play a crucial role in the biosynthesis and degradation of various biomolecules. They participate in several essential metabolic pathways: methionine, tyrosine and phenylalanine metabolism; phenylalanine, tyrosine and tryptophan biosynthesis; novobiocin biosynthesis and alkaloid biosynthesis. [2] Although aminotransferases have been the subject of extensive research for more than half a century, their industrial importance has gradually grown from the early 2000s. The revolution started when the main problem of enzyme-driven transamination, which is unfavorable reaction kinetics, was overcome first by using substrate excess and then by product removal and by introduction of enzymatic cascade reactions. [3][4][5][6][7] Most of ArAT's applications are related to ω-aminotransferases, as they are not limited to substrates with an amino group adjacent to the carboxyl one. However, the enantiomerically pure amino acids and their close derivatives are still in high enable its orientation to be presented in a manner mimicking the orientation of a real substrate, due to its ability to make an additional hydrogen bond with a Schiff base ( Figure  1). Due to the broad application of transamination in the chemical and pharmaceutical industry, it is important to provide comprehensive knowledge about the architecture and functions of transaminases obtained from extremophilic organisms, such as Psychrobacter sp. B6. Therefore, we found a way to show the active site of PsyArAT in complexes, with its substrate structural analogs acting as inhibitors, to observe conformational changes in the catalytic pocket in the most similar surrounding as has place upon substrate binding. The four crystal structures of PsyArAT complexed with hydroxy-analogs of enzyme substrates, i.e., phenylalanine (FOH), tyrosine (YOH), tryptophan (WOH) and aspartic acid (DOH), that are presented in this work, show great adaptability of its catalytic center towards binding various sized ligands. Moreover, a structural comparison of the investigated structures with native PsyArAT (PDB ID: 4RKC) and with that complexed with aspartate, PsyArAT/D (PDB ID: 4RKD), [20] provides an extended knowledge about interactions of the ligand within the catalytic center and the conformational changes of the enzyme active pocket.

Protein Preparation
The protein, used for obtaining PsyArAT/FOH, PsyArAT/YOH and PsyArAT/DOH complexes, was prepared as described previously [24]. To improve the homogeneity of the protein sample and crystallization, a new expression construct, expression system and protein purification protocol was designed.
The psyarat gene was optimized for expression in E. coli (Eurofins, Luxembourg). It was amplified using Q5 High-Fidelity 2X Master Mix (New England Biolabs, Ipswich, MA, USA) and primers designed to be compatible with the pMCSG7 vector (AT_lic_fw, 5′-TACTTCCAATCCAATGCCATGTTCGAGCG-3′; AT_lic_rv, 5′-TTATCCAC-TTCCAATGTTAATCCTTGAGAACATC-3′). Cloning was performed using the ligationindependent cloning (LIC) method [26][27][28]; T4 DNA Polymerase, dGTP and dCTP were provided by New England Biolabs (Ipswich, MA, USA). DH5α competent cells were Due to the broad application of transamination in the chemical and pharmaceutical industry, it is important to provide comprehensive knowledge about the architecture and functions of transaminases obtained from extremophilic organisms, such as Psychrobacter sp. B6. Therefore, we found a way to show the active site of PsyArAT in complexes, with its substrate structural analogs acting as inhibitors, to observe conformational changes in the catalytic pocket in the most similar surrounding as has place upon substrate binding. The four crystal structures of PsyArAT complexed with hydroxy-analogs of enzyme substrates, i.e., phenylalanine (FOH), tyrosine (YOH), tryptophan (WOH) and aspartic acid (DOH), that are presented in this work, show great adaptability of its catalytic center towards binding various sized ligands. Moreover, a structural comparison of the investigated structures with native PsyArAT (PDB ID: 4RKC) and with that complexed with aspartate, PsyArAT/D (PDB ID: 4RKD), [20] provides an extended knowledge about interactions of the ligand within the catalytic center and the conformational changes of the enzyme active pocket.

Protein Preparation
The protein, used for obtaining PsyArAT/FOH, PsyArAT/YOH and PsyArAT/DOH complexes, was prepared as described previously [24]. To improve the homogeneity of the protein sample and crystallization, a new expression construct, expression system and protein purification protocol was designed.
The psyarat gene was optimized for expression in E. coli (Eurofins, Luxembourg). It was amplified using Q5 High-Fidelity 2X Master Mix (New England Biolabs, Ipswich, MA, USA) and primers designed to be compatible with the pMCSG7 vector (AT_lic_fw, 5 -TACTTCCAATCCAATGCCATGTTCGAGCG-3 ; AT_lic_rv, 5 -TTATCCACTTCCAATGTT AATCCTTGAGAACATC-3 ). Cloning was performed using the ligation-independent cloning (LIC) method [26][27][28]; T4 DNA Polymerase, dGTP and dCTP were provided by New England Biolabs (Ipswich, MA, USA). DH5α competent cells were transformed with the resulting pMCSG7_psyarat vector using a basic heat shock procedure. Obtained transformants were cultivated overnight at 37 • C on LB-agar plates, with ampicillin as a selective marker. The plasmid was isolated using Promega kit and verified by sequencing (Genomed, Warsaw, Poland).
An inoculum of ArcticExpress (DE3) cells, transformed with pMCSG7_psyarat vectors, was cultivated with the addition of ampicillin (100 µg/mL) and gentamycin (20 µg/mL) overnight. TB medium was inoculated with 1% of the overnight culture, and protein overexpression was induced with 1 mM IPTG at OD~0.6 and cultivated at 12 • C overnight. Pellet was collected by centrifugation at 3000× g at 4 • C for 20 min.
Cell pellets were resuspended in buffer A (20 mM TrisHCl, 500 mM NaCl, 20 mM imidazole, pH 7.4) supplemented with DTT at a final concentration of 100 mM. Cell lysis was performed by sonication using Bandelin Sonopuls GM 3200 (Bandelin electronic GmbH & Co. KG, Berlin, Germany). The extract was centrifuged for 60 min at 4 • C and 10,000× g. The clarified supernatant was then subjected to affinity chromatography using ÄKTA Pure system (GE Healthcare, Chicago, IL, United States) and HisTrap 5 mL column (GE Healthcare) equilibrated in buffer A. The tagged protein was eluted with buffer B (20 mM TrisHCl, 500 mM NaCl, 400 mM imidazole, pH 7.4). TEV protease digestion was performed overnight, parallel with buffer exchange to buffer A via dialysis. Reverse affinity chromatography was performed using the same setup to separate PsyArAT from TEV protease and His 6 -tag. Size exclusion chromatography, with the usage of a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) and buffer C (20 mM Tris-HCl pH 7.4, 100 mM NaCl), was performed to ensure sample purity necessary for crystallization. The PsyArAT was concentrated to 12 mg/mL using Vivaspin20 (30.000 MWCO) concentrators (Sartorius, Göttingen, Germany). Each step of the purification was controlled by SDS-PAGE electrophoresis.

Crystallization
All crystallizations of the reported PsyArAT complexes were performed by the hanging drop vapor-diffusion method at 291 K, using 24-well crystallization plates (Hampton Research, Aliso Viejo, CA, USA). Crystallization drops were set manually at 1 Ml + 1 µL, protein and well solution, respectively. The protein was concentrated to 9-11 mg/mL in 10 mM Tris-HCl buffer, pH 7.5, using Vivaspin concentrators with a 10 kDa cutoff (Sartorius, Göttingen, Germany). Hexagonal crystals of the PsyArAT/DOH complex, with aspartic acid hydroxy-analog (DOH), were obtained in conditions consisting of 2.1 M DL-Malic acid, pH 7.0 and 0.1 M Tris-HCl buffer pH 7.5. Monoclinic crystals of PsyArT/YOH and PsyArAT/FOH were obtained by co-crystallization, using 0.2 M MgNO 3 , 20% PEG 2000, HEPES buffer, pH 7.5 and a 20-fold molar excess of L-tyrosine and L-phenylalanine hydroxy-analogs (YOH and FOH). C-centered monoclinic crystals of PsyArAT/WOH were obtained by soaking native crystals, grown in the presence of 0.2 M MgNO 3 and 20% PEG 3350 in HEPES buffer, pH 7.5, with a 5-fold molar excess of PLP and 10-fold molar excess of 3-indolelactic acid (hydroxy-analog of L-tryptophan-WOH), for 3 h.

Diffraction Data Collection, Structure Solution and Refinement
Diffraction measurements were carried out at 100 K temperature in the vapor stream of liquid nitrogen at HZB BESSY II synchrotron in Berlin, Germany. Crystals of PsyArAT/DOH did not need any cryoprotectant, because they were obtained in the presence of dicarboxylic acid [29]. Diffraction data of PsyArAT/DOH were processed in a hexa-gonal system to a resolution of 1.62 Å using HKL2000 [30]. The structure was solved in one of the enantiomorphic space groups, P6 5 22, using the monomer of PsyArAT (PDB ID: 4RKC) as a model in Molrep [31]. Crystals of PsyArAT/YOH, PsyArAT/FOH and PsyArAT/WOH were transferred for a moment to the drop containing a mixture of well solution and PEG 400 in a ratio of 1:1 before the diffraction experiment. For these complexes, diffraction data were processed using the XDSapp program. [32,33] PsyArAT/YOH and PsyArAT/FOH were indexed in monoclinic space group P2 1 to a resolution of 2.31 Å and 2.50 Å, respectively, and PsyArAT/WOH in C2 to a resolution 2.59 Å. The calculation of solvent content [34] showed that a monomer is present in the asymmetric unit of PsyArAT/DOH, a dimer in the asymmetric unit of PsyArAT/YOH and PsyArAT/FOH and two dimers in the case of PsyArAT/WOH. The structures of PsyArAT/YOH and PsyArAT/FOH complexes were solved by molecular replacement in Phaser [35], using the dimer of the native enzyme as a model. The structure of PsyArAT/WOH was solved by molecular replacement in Molrep [31] using the dimer of PsyArAT (PDB ID: 4RKC) as a model. For all crystal structures, the refinement was performed in REFMAC5 [36] from the CCP4I suite [37] with TLS restrains [38]. Structures were rebuilt using graphical program COOT [39] and validated using MolProbity programs [40,41]. Statistics details of the diffraction experiment, diffraction data reduction and crystal structure refinement are presented in Table 1. Table 1. Diffraction data collection, processing, and refinement statistics for crystal structures of PsyArAT complexes with substrates hydroxy-analogs.

Data Collection
PsyArAT
Å resolution, respectively. The aforementioned complexes crystallized in three space groups: hexagonal P6522 for PsyArAT/DOH, monoclinic P21 for PsyArAT/FOH and PsyArAT/YOH and C2 for PsyArAT/WOH, as a result of different crystallization conditions and the presence of ligands varying in size and chemical character. The overall architecture of the PsyArAT is typical for the type-I PLP-dependent enzymes and was previously described [24]. The dimer, functional unit of the enzyme, has an ovaloid shape with a 2-fold noncrystallographical symmetry axis (in the case of PsyArAT/DOH, it is also a crystallographic axis) laying along with its height (~45 Å ) and dividing the homodimeric structure into two identical subunits. The longest dimension of ~100 Å crosses over both monomers. Subunits are connected by salt bridges created between the carboxylic group of Glu253 and α-amino nitrogen of His289, as well as by side chains of Arg63 and Glu53 and by hydrophobic interactions between Ile101 of adjacent monomers and anchoring residues of the N-terminal coil (Met1, Phe2 and Ile5). The interface area between monomers measured 3112 Å 2 . Each subunit is divided into two domains differing in function and architecture. A flexible small domain, responsible for closing of the active site during ligand binding, is built of both N-and C-termini of the protein, involving residues Met1-Leu66 and Pro286-Asp398. Both parts are connected by short parallel β-sheets: β-strand1 (β1) Val29-Leu31 and β-strand10 (β10) Gly366-Tyr368, which allow for the simultaneous movement of the mentioned protein N-and C-terminal fragments ( Figure 3). The large domain is a conformationally stable part of the monomer; its architecture can be described as a Rossmann-like α-β-α sandwich fold, which is related to the main function: PLP binding and stabilization. This kind of arrangement is typical for nucleotide-binding proteins, and PsyArAT's interaction with pyridoxal phosphate could be explained as analogous to interactions at the active sites of FAD and NAD(P) binding proteins, e.g., phosphoglycerate dehydrogenase or glutathione reductase, where ligands are also bound by a phosphate group and nitrogen of a heterocyclic aromatic ring [42]. The overall architecture of the PsyArAT is typical for the type-I PLP-dependent enzymes and was previously described [24]. The dimer, functional unit of the enzyme, has an ovaloid shape with a 2-fold noncrystallographical symmetry axis (in the case of PsyArAT/DOH, it is also a crystallographic axis) laying along with its height (~45 Å) and dividing the homodimeric structure into two identical subunits. The longest dimension of~100 Å crosses over both monomers. Subunits are connected by salt bridges created between the carboxylic group of Glu253 and α-amino nitrogen of His289, as well as by side chains of Arg63 and Glu53 and by hydrophobic interactions between Ile101 of adjacent monomers and anchoring residues of the N-terminal coil (Met1, Phe2 and Ile5). The interface area between monomers measured 3112 Å 2 . Each subunit is divided into two domains differing in function and architecture. A flexible small domain, responsible for closing of the active site during ligand binding, is built of both N-and C-termini of the protein, involving residues Met1-Leu66 and Pro286-Asp398. Both parts are connected by short parallel β-sheets: β-strand1 (β1) Val29-Leu31 and β-strand10 (β10) Gly366-Tyr368, which allow for the simultaneous movement of the mentioned protein N-and C-terminal fragments ( Figure 3). The large domain is a conformationally stable part of the monomer; its architecture can be described as a Rossmann-like α-β-α sandwich fold, which is related to the main function: PLP binding and stabilization. This kind of arrangement is typical for nucleotide-binding proteins, and PsyArAT's interaction with pyridoxal phosphate could be explained as analogous to interactions at the active sites of FAD and NAD(P) binding proteins, e.g., phosphoglycerate dehydrogenase or glutathione reductase, where ligands are also bound by a phosphate group and nitrogen of a heterocyclic aromatic ring [42].
Depending on the crystal form of PsyArAT, the determined complexes have different asymmetric units (ASU). In the case of PsyArAT/FOH and PsyArAT/YOH, the ASU is created by a functional dimer; in PsyArAT/DOH the asymmetric unit represented only a monomer, but the functional dimer is reproduced with the symmetry-related molecule. The two-fold axis of the dimer is collinear with the crystallographic diagonal two-fold axis. In the PsyArAT/WOH complex, two functional dimers are present in the asymmetric unit ( Figure 4). Materials 2021, 14, x FOR PEER REVIEW 7 of 20 Depending on the crystal form of PsyArAT, the determined complexes have different asymmetric units (ASU). In the case of PsyArAT/FOH and PsyArAT/YOH, the ASU is created by a functional dimer; in PsyArAT/DOH the asymmetric unit represented only a monomer, but the functional dimer is reproduced with the symmetry-related molecule. The two-fold axis of the dimer is collinear with the crystallographic diagonal two-fold axis. In the PsyArAT/WOH complex, two functional dimers are present in the asymmetric unit ( Figure 4).   Depending on the crystal form of PsyArAT, the determined complexes have different asymmetric units (ASU). In the case of PsyArAT/FOH and PsyArAT/YOH, the ASU is created by a functional dimer; in PsyArAT/DOH the asymmetric unit represented only a monomer, but the functional dimer is reproduced with the symmetry-related molecule. The two-fold axis of the dimer is collinear with the crystallographic diagonal two-fold axis. In the PsyArAT/WOH complex, two functional dimers are present in the asymmetric unit ( Figure 4).

The Active Site of PsyArAT Complexes with Hydroxy-Analogs of Substrates
Entrances to the active pockets create two grooves on the opposite sites of the protein surface. Each monomer contains an equivalent active center with pyridoxal-5 -phosphate (PLP); however, depending on the reaction stage, the cofactor may have taken a different form. In PsyArAT/FOH, PsyArAT/YOH and PsyArAT/DOH complexes, fractional occupancy for PLP was observed. In PsyArAT/WOH, PLP was added to the crystallization solution, which helped to obtain very good electron density maps for the cofactor.
The PLP molecule is bound in the enzyme active center by an extensive network of hydrogen bonds. The only nonpolar interaction holding PLP is π-stacking of the pyrimidine ring with the indole ring of Trp130. Conformation of the amine group of Lys246, which is essential for transamination catalysis, depending on the stage of the reaction, can create a hydrogen bond to an entity present in position C4 of the cofactor, the formyl group of PLP, the amine group of pyridoxamine phosphate (PMP), or a covalent bond with internal aldimine (PLI) [24].
While the creation of the internal aldimine is not caused by direct contact with a substrate (or inhibitor), as this is an initial step enabling their eventual further conjugation, it seemed to be triggered by the presence of the ligand in the pocket. Interestingly, although all hydroxy-analogs of substrates used for the creation of PsyArAT/FOH, PsyArAT/YOH, PsyArAT/WOH and PsyArAT/DOH complexes acted as competitive inhibitors, the active sites should be essentially at the same pre-reaction stage; however, the form of the cofactor was not uniform among the pockets. The internal aldimine (PLI) (covalently bound PLP to Lys246) was observed in the PsyArAT/DOH, in all of the four pockets of PsyArAT/WOH and the one monomer of PsyArAT/YOH and PsyArAT/FOH homodimers. Despite the presence of an inhibitor in the pockets of chain B of PsyArAT/YOH and PsyArAT/FOH dimers, the non-bound form of PLP was present, similarly, as it was in the catalytic center at both monomers of the native enzyme structure. In the PsyArAT/FOH complex, the closest distance between the ligand and PLP (measured from the α-hydroxyl group of the inhibitor to aldehyde carbon of PLP) was 4.1 Å in monomer A and 3.3 Å in monomer B, while in the PsyArAT/YOH structure respective distances were bigger (5.7 Å and 4.6 Å). The pockets in monomers A of these two complexes were also larger in volume than the centers of chains B, the center of PsyArAT/DOH and centers of PsyArAT/WOH monomers with bound WOH. Thus, it can be concluded that strong binding of the ligand leads to a more compact structure, with the cofactor formed in a pre-reactionary stage.

PsyArAT Interactions with Hydroxy-Analogs of Substrates
The hydroxy-analogs of PsyArAT substrates used in this work ( Figure 1) were clearly visible in electron density maps ( Figure 5) of the solved structures. Ligands interact in the enzyme active center mostly by hydrogen bonds between their carboxylic groups, with the guanidine group of Arg374 belonging to the C-terminus and the amine group hydrogen of Gly34 belonging to the N-terminus of the small domain. They also created polar and hydrophobic contacts with two residues of the large domain: Asn183 and Trp130. Such a set of amino acids involved in inhibitor binding allows for the formation of the closed protein conformation by bringing together both the large and the small domains.
The hydrogen bonds network made by a FOH ligand were slightly different for both monomers of the PsyArAT/FOH complex ( Figure 6A,B). In either case, carboxyl group of FOH creates two parallel hydrogen bonds with the guanidine group of Arg374 (3.2 Å and 3.0 Å in monomer A, 2.3 Å and 2.8 Å in monomer B), one with a side-chain nitrogen of Asn183  The hydrogen bonds network made by a FOH ligand were slightly different for both monomers of the PsyArAT/FOH complex ( Figure 6A,B). In either case, carboxyl group of FOH creates two parallel hydrogen bonds with the guanidine group of Arg374 (3.  Despite the high chemical similarity to FOH, the interaction of the YOH inhibitor with the residues forming the active site of subunit A is slightly weaker, and only five hydrogen bonds were created, two with the guanidine moiety of Arg374 (2.8 Å and 3.3 Å), one with the peptide nitrogen of Gly34 (3.0 Å) and another two with the indole nitrogen of Trp130 (the first like all of those mentioned, by a carboxyl group (3.5 Å), and the second by a hydroxyl group of YOH (2.8 Å)) ( Figure 6C). In the pocket B, the ligand made a weaker network of hydrogen bonds, interacting only with one nitrogen atom of the Arg374 guanidine group (2.8 Å) and the carbonyl oxygen of Gly34 (3.2 Å), by its carboxyl group and by its hydroxyl group with Gly34 nitrogen (3.5 Å) ( Figure 6D). In the catalytic center of both complexes, the Arg280* switch was in the "down" position; however, in the case of PsyArAT/YOH, it was placed a little deeper than in the PsyArAT/FOH complex. In the subunit A, Arg280*, which was pushed down by a YOH ligand, made two hydrogen bonds between guanidine moiety nitrogen atoms and carboxyl oxygens of Asp11 (2.8 Å and 3.5 Å). In monomer B, only one contact was present with Asp11 (2.9 Å), but another bond was created with the hydroxyl group of Asn132 (3.5 Å). As stated, in PsyArAT/FOH complex, Arg280* was placed slightly further from the ligand; in the active site of monomer A, it interacted with Asp11 (2.8 Å) like in the previous case, but it also reached Tyr8 hydroxyl oxygen (3.2 Å). In the active pocket of monomer B, Arg280* made an even shorter bond with the OH group of Tyr8 (2.8 Å), but the interaction with Asn132 (2.9 Å) replaced the one with Asp11. Despite the high chemical similarity to FOH, the interaction of the YOH inhibitor with the residues forming the active site of subunit A is slightly weaker, and only five hydrogen bonds were created, two with the guanidine moiety of Arg374 (2.8 Å and 3.3 Å), one with the peptide nitrogen of Gly34 (3.0 Å ) and another two with the indole nitrogen of Trp130 (the first like all of those mentioned, by a carboxyl group (3.5 Å ), and the second by a hydroxyl group of YOH (2.8 Å )) ( Figure 6C). In the pocket B, the ligand made a weaker network of hydrogen bonds, interacting only with one nitrogen atom of the Arg374 guanidine group (2.8 Å ) and the carbonyl oxygen of Gly34 (3.2 Å ), by its carboxyl group and by its hydroxyl group with Gly34 nitrogen (3.5 Å ) ( Figure 6D). In the catalytic center of both complexes, the Arg280* switch was in the "down" position; however, in the case of PsyArAT/YOH, it was placed a little deeper than in the PsyArAT/FOH complex. In the subunit A, Arg280*, which was pushed down by a YOH ligand, made two hydrogen Tryptophan hydroxy-analog (WOH) created a network of contacts, enabling the involvement of both termini and a large domain, which resulted in closure conformation. High similarity of interactions with an inhibitor in catalytic centers of monomers A and C allowed for the description of only one of them, and we arbitrary choose monomer A. The carboxylic group of the WOH is planar to the Arg374 guanidine group, and two hydrogen bonds of 3.2 Å and 2.9 Å are made between them. The another contact in the length of 2.4 Å, of the mentioned carboxyl group, was made with the peptide bond hydrogen of Gly34, being part of the motile, long entrance loop. Additionally, side chain nitrogen atoms of Trp130 and Asn183, from the large domain, make hydrogen bonds (3.4 Å and 3.2 Å) with the carboxyl group of the ligand. Internal aldimine present in the catalytic center interacts with the carboxyl group of the WOH; the bond between the Schiff base nitrogen and oxygen atom is 3.2 Å long ( Figure 6E). Like in the case of the remaining complexes with aromatic ligands, FOH and YOH, Arg280*, from the second monomer, is in the "down" position, and one of its terminal nitrogen atoms creates two hydrogen bonds with side chains of Tyr8 and Asp11 (3.2 Å and 2.6 Å).
DL-malic acid sodium salt was used for the PsyArAT/DOH complex creation, but only L-malic acid (DOH) was bound in the active center. The ligand creates considerably more hydrogen bonds than aromatic inhibitors; 10 interactions with residues in the active site were present ( Figure 6E). Oxygen atoms of the α-carboxylic group of malic acid create strong parallel hydrogen bonds with two nitrogen atoms of the guanidine group from Arg374: (2.8 Å and 2.8 Å). Additionally, they are involved in contact with the amide group of Gly34 (2.8 Å) and Asn183 (3.0 Å). Oxygens of the β-carboxylic group of the ligand interacted, via hydrogen bonds, with Arg280* (2.9 Å and 2.9 Å) and Trp130 (2.8 Å). Hydroxyl oxygen of malic acid creates weak hydrogen bonds with Trp130 (3.2 Å) and stronger bonds with Nζ of Lys246 (2.9 Å) ( Figure 6F). Interaction with Arg280* was of special importance; in this situation, the arginine switch was in the "up" position, and the ligand is shoved further from the entrance; hence, the helix H1 is able to close the active site fully. In this position Arg280* is still interacting with Asp11 (2.8Å, 3.4 Å) from the N-termini domain.

Arginine Switch
The action of Arg280* of the adjacent monomer (*) is important for the control over volume and substrate specificity of the PsyArAT active pocket [43]. The long electrophilic side chain of this "arginine switch" can take one of two positions, "up" or "down", due to a guarding interaction with Asp11 and Asn132, responsible for limiting its conformational freedom. In the "down" position, additional contact with Tyr8 can be made. This hydrogen bond system allows for both control of the chemical environment and the volume of the active site. When an aromatic hydrophobic inhibitor or a substrate was present in the active center, Arg280* took the "down" position, allowing for the entrance of the hydrophobic ligand and creates a sterical hindrance between helices H1 and H6, preventing the closing of the side entrance and making the pocket more spacious (Figure 7). The "up" position was taken when dicarboxylic ligand, such as malic acid, entered the active pocket. A lack of sterical hindrance enabled for the full closure conformation, and additional interactions with the ligand were made due to the presence of the guanidine group of Arg280 in the vicinity of the active site ( Figure 7).
Interestingly, in the active site of monomer A and C in the PsyArAT/WOH complex, the presence of a big, hydrophobic indole group of the WOH molecule, just near the Arg280* residue, prevented its contact with the Asn132 side chain (which was pushed away from the active pocket). Seemingly, such interaction was not necessary, as the conformational freedom of Arg280* was limited by the inhibitor itself. However, the carboxyl group of Asp11 interacted with the guanidino group of Arg280* via two (2.8 Å and 3.4 Å) hydrogen bonds.

Active Pocket Volume and Area of the PsyArAT in Native and Complexes Form
Contrary to all structures of aromatic aminotransferases complexed with inhibitors which are currently published, conformational changes in PsyArAT monomers seem to be induced exclusively by a ligand presence. Unit cell dimensions for the native enzyme PsyArAT, PsyArAT/FOH and PsyArAT/YOH complexes are nearly the same, despite motions occurring in domains. Moreover, calculations done using PISA [44] showed differences in the areas of contact made by each subunit in dimeric structures. For the PsyArAT/YOH complex, the difference in the interface area between subunits A and B is only 14.5 Å 2 , in favor of domain B, while for PsyArAT/FOH, it was 87.2 Å 2 , with more contacts for monomer B as well. However, the conformational changes for these two complexes did not seem to be in relation to these values, as the PsyArAT/YOH complex showed bigger differences between monomers. Moreover, putting monomers B of these three structures in order from the most closed to the most open, which was PsyArAT/FOH, PsyArATYOH and the native enzyme, did not correlate to the order of contact area from lowest to the highest: native, PsyArAT/FOH and PsyArAT/YOH. The PsyArAT/DOH complex could not be so easily compared to the structures with analogs of phenylalanine and tyrosine, as it crystallized as a monomer in ASU in a hexagonal unit cell; however, the areas of contact (excluding the one being the dimer interface) was about 720Å, which was similar to values for PsyArAT/FOH, being much more relaxed structurally. PsyArAT/WOH was another complex with different crystallographic properties, as it crystallized in a C2 group as a double dimer with noncrystallographical symmetry. In the case of this complex, it is possible that crystal packing affected the conformation, as a difference in the area of contact with neighboring proteins in a lattice was high (~500 Å 2 ), where more contacts were made by closed domains.  Interestingly, in the active site of monomer A and C in the PsyArAT/WOH complex, the presence of a big, hydrophobic indole group of the WOH molecule, just near the Arg280* residue, prevented its contact with the Asn132 side chain (which was pushed away from the active pocket). Seemingly, such interaction was not necessary, as the conformational freedom of Arg280* was limited by the inhibitor itself. However, the carboxyl group of Asp11 interacted with the guanidino group of Arg280* via two (2.8 Å and 3.4 Å ) hydrogen bonds.

Active Pocket Volume and Area of the PsyArAT in Native and Complexes Form
Contrary to all structures of aromatic aminotransferases complexed with inhibitors which are currently published, conformational changes in PsyArAT monomers seem to be induced exclusively by a ligand presence. Unit cell dimensions for the native enzyme PsyArAT, PsyArAT/FOH and PsyArAT/YOH complexes are nearly the same, despite motions occurring in domains. Moreover, calculations done using PISA [44] showed differences in the areas of contact made by each subunit in dimeric structures. For the PsyArAT/YOH complex, the difference in the interface area between subunits A and B is only 14.5 Å 2 , in favor of domain B, while for PsyArAT/FOH, it was 87.2 Å 2 , with more contacts for monomer B as well. However, the conformational changes for these two complexes did not seem to be in relation to these values, as the PsyArAT/YOH complex showed bigger differences between monomers. Moreover, putting monomers B of these three structures in order from the most closed to the most open, which was PsyArAT/FOH, PsyArATYOH and the native enzyme, did not correlate to the order of The different position, interactions and presence or lack of the ligand in the active center of the studied complexes prompted us to investigate the volume and area of the active pockets and entrances to it in all monomers. Volumes were measured using the CASTp 3.0 tool [45], with probe size set to 1.2 Å. Interestingly, the volume of the pockets in one monomer was much smaller than in the second one of the dimer in all complexes (excluding PsyArAT/DOH), but the difference in the case of FOH bound caused a smaller disparity between them than in the case of YOH present in the active center (Table 2) ( Figure 8C-F). The PsyArAT/DOH active site is considerably smaller (Figures 8B and S1) in comparison to the two mentioned. Baffling was a lack of WOH in one monomer of each dimer in PsyArAT/WOH, which resulted in a huge difference in the volume of the active pocket, with the tryptophan hydroxy-analog present or absent in the active center ( Figure 8G,H). Even more surprising were the twice bigger values of the volume and surface of the active pocket lacking an inhibitor in comparison to the one occupied by the ligand. Analyzing these parameters with the native enzyme ( Figure 8A), we found that the fully opened pocket of the native PsyArAT was about two times bigger in comparison to pockets in PsyArAT/FOH and PsyArAT/YOH complexes, two times bigger than for PsyArAT/WOH and four times bigger than for PsyArAT/DOH (Table 2). active pockets and entrances to it in all monomers. Volumes were measured using the CASTp 3.0 tool [45], with probe size set to 1.2 Å . Interestingly, the volume of the pockets in one monomer was much smaller than in the second one of the dimer in all complexes (excluding PsyArAT/DOH), but the difference in the case of FOH bound caused a smaller disparity between them than in the case of YOH present in the active center (Table 2) ( Figure 8C-F). The PsyArAT/DOH active site is considerably smaller ( Figure 8B, Figure  S1) in comparison to the two mentioned. Baffling was a lack of WOH in one monomer of each dimer in PsyArAT/WOH, which resulted in a huge difference in the volume of the active pocket, with the tryptophan hydroxy-analog present or absent in the active center ( Figure 8G,H). Even more surprising were the twice bigger values of the volume and surface of the active pocket lacking an inhibitor in comparison to the one occupied by the ligand. Analyzing these parameters with the native enzyme ( Figure 8A), we found that the fully opened pocket of the native PsyArAT was about two times bigger in comparison to pockets in PsyArAT/FOH and PsyArAT/YOH complexes, two times bigger than for PsyArAT/WOH and four times bigger than for PsyArAT/DOH ( Table 2).  Hence, it can be concluded that the volume of the PsyArAT active center differs depending on the conformational states adopted by the enzyme, and on the size of the ligand bound in the active pocket. These open, closed and semi-closed conformations were also related to structural motions triggered by the presence of a specific ligand.

PsyArAT Motions upon the Binding of the Substrate Hydroxy-Analogs
The binding of different types of ligands caused conformational adjustment of the PsyArAT monomers/domains in relation to each other. Generally, during the pocket closing, three main motions occurred in the enzyme structure: kink on a long helix H14 and two perpendicular to each other and shear motions of the small domain main components: Nterminal H1 helix and C-terminal helices (H16, H17 and H18) towards the pocket entrance ( Figure 9). Using the open form of monomer B of the native PsyArAT as a reference, approximate lengths of these movements were measured. The biggest differences in respect to the structure with no ligands bound (except for the cofactor) were recognized for PsyArAT/DOH and PsyArAT/WOH. Namely, the kink on the helix H14 was estimated at 9 and 8.5 degrees for two complexes, respectively, measured motions of small domain Cterminal helices were equal 3. The shear motion of the helix H1 is crucial for the closing of the pocket, due to the presence of hydrophobic Ile13, Leu14 and Val17 residues being part of it. During movement of the helix H1, simultaneous rotation took place, which caused a shift of the amino acids towards the entrance from the outer region; thus, an impervious water plug was created. The side chain of Ile33 also helped to plug the entrance because of the Asn25-Val45 long loop movement, which is coordinated with the small domain C-terminus shear motion. Leu14 is responsible not only for providing the hydrophobic environment, but also acted as a "probe", sensing the size of a bound ligand. In complexes of PsyArAT/FOH, PsyArAT/YOH and PsyArAT/WOH, the leucine side chain was unable to move further into the enzyme catalytic pocket due to a sterical hindrance made with the inhibitor molecule. DOH is a ligand of smaller size than aromatic hydroxy-analogs, and due to interactions with Arg280*, it was shoved deeper in the active site; in this situation, Leu14 moved into the center, enabling proper rotation of H1 and full closure conformation (Figure 9).
In the previously determined structure of PsyArAT in the complex with aspartic acid (PDB ID: 4RKD), four dimers were present in the asymmetric unit. This structure represented a different stage of transamination reaction; the conformation of the monomers in each dimer was not identical, but more similar than in complexes studied in this work (RMSD values for Cα in a range 0.40-0.49Å). This indicated that processing of the small size substrate does not require a big conformational movement of the whole functional unit. A similar observation was noticed for two structures of complexes with ligands of medium size (FOH and YOH), where the ligand was present in the active center of both monomers (RMSD for Cα of 0.46 Å and 0.42 Å, respectively). In the apo structure of PsyArAT (PDB ID: 4RKC), the conformational changes of monomers were larger (RMSD for Cα of 0.79Å). These differences indicates that one of the active sites was more open to accommodate the potential substrate. More significant conformational differences of monomers in dimers of PsyArAT/WOH are visible, where the RMSD for Cα between monomers are around 1.07 Å and 1.30 Å (Table 1). This observation was made in a number of other structures of the PLP-dependent enzyme, which clearly indicates domain motion depending on the reaction stage.  Table 2.
The shear motion of the helix H1 is crucial for the closing of the pocket, due to the presence of hydrophobic Ile13, Leu14 and Val17 residues being part of it. During movement of the helix H1, simultaneous rotation took place, which caused a shift of the amino acids towards the entrance from the outer region; thus, an impervious water plug was created. The side chain of Ile33 also helped to plug the entrance because of the Asn25-Val45 long loop movement, which is coordinated with the small domain C-terminus shear motion. Leu14 is responsible not only for providing the hydrophobic environment, but also acted as a "probe", sensing the size of a bound ligand. In complexes of PsyArAT/FOH, PsyArAT/YOH and PsyArAT/WOH, the leucine side chain was unable to move further into the enzyme catalytic pocket due to a sterical hindrance made with the inhibitor molecule. DOH is a ligand of smaller size than aromatic hydroxy-analogs, and due to interactions with Arg280*, it was shoved deeper in the active site; in this situation, Leu14 moved into the center, enabling proper rotation of H1 and full closure conformation (Figure 9).
In the previously determined structure of PsyArAT in the complex with aspartic acid (PDB ID: 4RKD), four dimers were present in the asymmetric unit. This structure represented a different stage of transamination reaction; the conformation of the monomers in each dimer was not identical, but more similar than in complexes studied in this work (RMSD values for Cα in a range 0.40-0.49Å ). This indicated that processing of the small size substrate does not require a big conformational movement of the whole functional  Table 2.
Most of atoms, especially in the enzyme center of PsyArAT, presented a low value of the displacement parameter, but chains attending in ligand incorporation to the active pocket have considerably higher average B-factors, which indicates their motility during transamination reaction (Figure 10).

Comparing Structures of PsyArAT with PdeArAT Complexes and Other ArATs
Structural analysis of PsyArAT in relation to its bacterial homolog from Paracoccus denitrificans [20] highlights some similarities. All residues involved in cofactor binding are highly conserved among these two enzymes, as well as amino acids building the smaller compartment of the active pocket being responsible for the binding of substrates by their α-carboxylic group. The most substantial differences is observed in the large chamber of the active site being a place for the ligands' bulkier parts, responsible for their chemical character. For example, in the place of serine 285 in PsyArAT, there is a phenylalanine 297 in the mesophilic analog PdeArAT (Figure 11). Such a drastic amino acid difference impacts the depth of ligand binding, as aromatic moieties of inhibitors tend not to have enough space.
other structures of the PLP-dependent enzyme, which clearly indicates domain m depending on the reaction stage.
Most of atoms, especially in the enzyme center of PsyArAT, presented a low va the displacement parameter, but chains attending in ligand incorporation to the pocket have considerably higher average B-factors, which indicates their motility d transamination reaction (Figure 10).

Comparing Structures of PsyArAT with PdeArAT Complexes and Other ArATs
Structural analysis of PsyArAT in relation to its bacterial homolog from Para denitrificans [20] highlights some similarities. All residues involved in cofactor bindi highly conserved among these two enzymes, as well as amino acids building the s compartment of the active pocket being responsible for the binding of substrates b α-carboxylic group. The most substantial differences is observed in the large cham the active site being a place for the ligands' bulkier parts, responsible for their ch character. For example, in the place of serine 285 in PsyArAT, there is a phenylalani in the mesophilic analog PdeArAT (Figure 11). Such a drastic amino acid differen pacts the depth of ligand binding, as aromatic moieties of inhibitors tend not to enough space. Broadening the analysis by aromatic aminotransferases from a little more evolutionary distant organisms, thermophilic archaeon (Pyrococcus horikoshii (PDB id: 1DJU)) [22] and mesophilic yeast (Candida albicans (PDB id: 6HNU)) [21] confirms that many residues in the active pocket are conserved, but they also shows some interesting differences. Residues of the PLP cavity and the smaller compartment of the active pocket were generally the same. Only minor differences, such as the presence of phenylalanine instead of tryptophan in a position of a PLP π-stacking residue, for structures of archaeon and yeast aminotransferases were observed. The most outstanding difference seemed to be a lack of an arginine switch, which was proven to take part in the aromatic aminotransferases' mechanism of operation. In the case of the PhoArAT structure, the same position was occupied by a methionine residue, having a very different chemical character (Figure 11). In Figure 11. Structural comparison of the active center of bacterial aromatic aminotransferases with ligands: PsyArAT (PDB ids: 6ZUR, 6ZUP, 6ZVG; blue), PdeArAT (PDB ids: 2AY1, 2AY2, 2AY5; green) and PhoArAT (PDB id: 1DJU; salmon). All ligands bound in the active center are shown in the same color as the enzyme peptide chain. In the structure of 1DJU, a ligand was not present.
Broadening the analysis by aromatic aminotransferases from a little more evolutionary distant organisms, thermophilic archaeon (Pyrococcus horikoshii (PDB id: 1DJU)) [22] and mesophilic yeast (Candida albicans (PDB id: 6HNU)) [21] confirms that many residues in the active pocket are conserved, but they also shows some interesting differences. Residues of the PLP cavity and the smaller compartment of the active pocket were generally the same. Only minor differences, such as the presence of phenylalanine instead of tryptophan in a position of a PLP π-stacking residue, for structures of archaeon and yeast aminotransferases were observed. The most outstanding difference seemed to be a lack of an arginine switch, which was proven to take part in the aromatic aminotransferases' mechanism of operation. In the case of the PhoArAT structure, the same position was occupied by a methionine residue, having a very different chemical character (Figure 11). In the structure of CalArAT, the long sidechain of arginine was swapped for a glutamic acid having an opposite charge and a very low chance to operate in the same mechanism.

Conclusions
In this work, crystal structures of bacterial psychrophilic aromatic amino acid aminotransferases in complexes with four hydroxy-analogs of substrates were presented. Three of the ligands were derivatives of aromatic amino acids: phenylalanine (FOH), tyrosine (YOH) and tryptophan (WOH), and one was of aliphatic aspartic acid (DOH). The partial occupancy of PLP in the structures of FOH and YOH were the result of cofactor loss during the purification process. This is why a crystal of WOH was obtained by co-crystallization with a 20-molar excess of WOH and 5-molar excess of PLP. All ligands, acting as competitive inhibitors, bound in the catalytic pocket allowed for structural analysis of the enzyme active center. Observed structural changes of amino acids side chains in the enzyme catalytic center, as well as the movement of whole domains, mimicked interactions occurring in the active pocket during substrate binding. Although in both monomers of PsyArAT/FOH and PsyArAT/YOH, in the catalytic centers, ligands were bound, we observed differences between conformations of side chains of amino acids in the active pocket in monomers of homodimers. The surface of catalytic pockets in both monomers after binding YOH and FOH differs. For ligands of a similar size and chemical character, changes of the active pocket volume were comparable, but the observed differences between monomers in dimers of the PsyArAT/WOH complex were significantly bigger and the most intriguing. The volume of the pocket in monomer B (D) was twice as big as in monomer A (C) (Figure 8). It was caused by the fact that only one monomer of each dimer bound the ligand (A and C). This different action of catalytic centers in monomers of dimers may indicate that the active centers are somewhat independent while binding bulkier substrates. The small acidic inhibitor (DOH) caused different orientation of the Arg280* and forced closed the conformation of the active center. The results of the structural investigation and activity assay suggest a reverse correlation between activity and gravity of movement of the small domain. The functional dimer of the PsyArAT/DOH complex was created from two symmetry-related subunits. A lack of differences between monomers of this crystallographic dimer allowed for forming of the crystal in a high symmetry space group, with one monomer in the asymmetric unit. The crystallographic experiments performed in this work highlighted the PsyArAT active site's adaptability towards different-sized substrates.