Serine Racemase structure and dynamics − Rakyat Biologi

SR is a PLP-dependent enzyme belonging to the fold-type II structural group, containing enzymes able to catalyze b-elimination and b-replacement reactions, such as tryptophan synthase (14) and O-acetylserine sulfhydrylase (15). With respect to other PLP-dependent enzyme the highest sequence identity is found for fission yeast SR and bacterial threonine dehydratase (90% identity). A 27% identity was also found between human SR and human serine dehydratase (SDH) or cystathionine-b-synthase (12, 16). SR and SDH are both able to catalyze the b-elimination of L-serine but SDH is not capable to carry out L-serine racemization (12). The identity among SR from different species ranges between 35% and 91% with the highest values recorded among mammalian SRs and the lowest for yeast and mammalian SRs. Residues involved in the binding of PLP, and effectors, such as cations and ATP (see below) are in general highly conserved (13).

There are currently eight SR structures deposited in the PDB:

1. Human SR (hSR): 3L6B and 3L6R, both with malonate as ligand. 3L6R is selenomethionine hSR (17).

2. Rat SR (rSR): 3HMK and 3L6C, in the holo form and complexed with malonate, respectively (17).

3..Schizosaccharomyces pombe SR (SpSR): in the holo form (1V71) with 5’-adenylyl methylenediphosphate (AMP-PCP) (1WTC), modified with PDD (lysino-D-alanyl residue) in the holo form (2ZPU) and modified with PDD and complexed with serine (2ZR8) (18, 19).

All these SRs crystallized as dimers, whereas in solution SR forms dimers with traces of tetramers (12). Dimers are stabilized by the formation of hydrophobic contacts, without the involvement of any disulfide bridge (20-22). Covalently cross-linked dimers were observed in the presence of reactive oxygen and nitrogen species (23). The human and rat dimer in the presence of malonate presents a buried monomer-monomer surface area of about 840 Å² and 735 Å², respectively, while the uncomplexed rat dimer surface increases to 827 Å².

Fold type II enzymes are characterized by the presence of a larger and a smaller domain, both exhibiting the typical open a-b architecture. As represented in Figure 1a showing the structure of hSR complexed with the inhibitor malonate (17), the large domain contains the PLP cofactor covalently bound to Lys56 through a Schiff base, forming the internal aldimine. This domain is made up of residues 1-68 and 157-340 organized in a twisted six-stranded b-sheet, with all but the first strand parallel, surrounded by ten a-helices. Residues 78-155 form the small domain, also arranged in a twisted four-stranded parallel b-sheet surrounded by three a-helices, one located on the solvent side and two at the domains interface. The small and large domains are connected by a flexible hinge region formed by amino acids 69-77, not fully detected in the hSR structure. In the SpSR, the PLP cofactor is bound to Lys56 (human numeration) or to a modified lysinoalanyl residue (18, 19). The interactions with the surrounding residues resemble those typical of fold type II enzymes (24) with the involved residues conserved not only in SR orthologs but also in many other PLP-dependent enzymes (25-27). In particular, the phosphate group interacts with Gly185, Gly186, Gly187, Gly188, known as the Gly loop, Met189 and three water molecules. The 3-hydroxy group forms a hydrogen bond with Asn86 belonging to the asparagine loop (Ser83-Ser84-Gly85-Asn86-His87 (25)), while Ser313 contacts the pyridine nitrogen. Both Asn86 and Ser313 play an essential role in regulating the electronic state of the PLP-Schiff base conjugate (28-30). The pyridine ring is sandwiched by Thr285, Gly239 and Val240 on the re-face and Phe55 on the si-face, oriented towards the larger domain in the protein core.

All SR structures present a divalent cation octahedrally coordinated in the larger domain, close to the PLP binding site, i.e. Mg²⁺ in SpSR structures and Mn²⁺ in the rat and human enzymes. In hSR, Mn²⁺ is coordinated by the side-chains of Glu210 and Asp216 (conserved in rSR and mutated to Asn in SpSR), the backbone carbonyl of Ala214 (substituted by a glycine in SpSR) and three water molecules (17). Hoffman et al. recently demonstrated that other residues such as Cys217 and Lys221 are fundamental for metal binding as well as for enzymatic activity (31). The absence of Glu210, Asp216 and Ala214 in the Scapharca and Pyrobaculum SR is consistent with the absence of increased racemization in the presence of Ca²⁺ or Mg²⁺ (32, 33). Experiments performed with SpSR demonstrated that the enzyme loses its catalytic activity in the presence of EDTA and recovers it upon the addition of Mg²⁺ or Ca²⁺ (18). Also mammalian brain SR is activated by the presence of divalent cations (13, 34). Nevertheless, considering the relative distance of the ion site from the active site, cations are not considered to be directly involved into the catalytic process, rather in the stabilization of protein structure. Neither cations nor ATP appeared to affect SR oligomerization state (20, 35).

Different studies demonstrated that both racemization and a,b-elimination reactions are stimulated upon the addition of Mg•ATP (10, 34). In spite of its relevance for SR function, only an ATP analog, namely AMP-PCP, has been crystallographically detected in SpSR (PDB code 1WTC). The ATP analogue is bound into the groove formed at the domain and subunit interface (18). This binding arrangement was also supported by a hSR model that positions two molecules of AMP-PCP at the dimerization interface (36). Following the yeast numeration, the residues involved in hydrogen bonding the ATP effector are Asn25, Lys52, Met53, Ala114, Tyr119, Asn311 on one monomer and Thr31, Ser33, Thr34, Arg275, Lys277 on the other, and a significant number of water molecules. Many of these residues are conserved in human and rat SRs, while only two are present in Escherichia coli threonine dehydratase, in Salmonella typhimurium O-acetyl serine sulfhydrylase or in E. coli tryptophan synthase. A model for the hSR-ATP complex was recently reported by Jiraskova-Vanickova et al. who docked two ATP molecules with a coordinated Mg²⁺ into the hSR dimer (PDB code 3L6B), using the yeast complex as template (12). ATP is localized so that the phosphate groups are buried inside the dimerization interface while the adenosine moiety is directed towards the solution. Overall, ATP is stabilized by ten interactions, four of which are bifurcated and involve the phosphate groups. Gln89 appeared to be particularly relevant by holding in place Thr52 and Tyr121, interacting, respectively, with the phosphate groups and with the adenosine aromatic moiety. All these observations led to the conclusion that the ATP binding site resembles that of other ATP-binding proteins (37), and that only the dimer is able to provide the necessary number of interactions to stabilize the complex. It was also proposed that the last eleven C-terminus residues, not identified in the crystallographic structure, might also contact and stabilize the ATP molecule (12). The comparison with the yeast protein crystallized in the absence of AMP-PCP revealed a different organization of the dimer, thus suggesting that ATP does not affect the conformation of the monomer but the relative orientation of the two subunits. No significant change in the orientation of the cofactor or of the binding pocket side chains is observed. Even if no direct connection exists between PLP and ATP, a hydrogen bond network formed by Met53, Asn84, Gln87, Glu281, Asn311, and water molecules might be responsible for the fine tuning of the active site structure and the stimulation of catalysis. Specifically, ATP might affect the open-closed transition triggered by the binding of substrate and ligands. The structures of hSR and human selenomethionine SR complexed with malonate (PDB codes 3L6B and 3L6R, respectively), were recently solved (17). In spite of co-crystallization trials with different ligands, i.e. L-serine, glycine, malonate and other small molecules only malonate resulted in highly diffracting co-crystals solved at a resolution of 1.5 Å (17) (Figure 2). The comparison between rSR and hSR in complex with malonate with rSR and SpSR holo structures indicates that binding of the inhibitor has triggered a transition from an open to a closed conformation through a rigid body movement of the small domain (Figure 3). While, in fact, the structural alignment of the human and rat monomer led to a C_a RMSD of 0.62 Å, the superimposition of the rat holo structure with that of rSR in complex with malonate resulted in 2.19 Å RMSD. Different PLP-dependent enzymes show a significant movement of the small domain leading to the closure of the active site upon ligand binding. This is the case, for instance, of aspartate aminotransferase belonging to fold-type I (38), threonine synthase and O-acetylserine sulfhydrylase belonging to fold-type II (26, 39). In the specific case of SR, the small domain undergoes a decrease of the solvent-exposed surface and a rotation of about 20° towards the large domain, without any change of the large domain dimer interface. The helix H5 and, in particular, the N-terminal loop (Ser83-His87, human numeration) approaches the PLP-lysinoalanine Schiff base forming the binding site for a carboxylate group. The shift of the small domain is supported by a large flexibility degree of the intra-domain interface, which allows it to assume slightly different orientations in the three available crystallographic structures, symptomatic of the fact that the motion of the domain should be larger and more continuous in solution, as suggested by an unreleased human holo structure, where the small domain was found completely disordered (17). Moreover, the closure of the binding pocket should occur only after the entrance and recognition of the substrate, which, otherwise, could not access the small cleft between the large and small domains. This arrangement plays a key role in the formation of the active site and, thus, in the catalytic process (18), since the closure observed with malonate binding is the same expected for the natural substrate. As suggested by Smith et al. (17), the rearrangement of the small domain locates Ser84 in such a close proximity of the substrate that is able to donate the proton required to catalyze the isomerization reaction (see below). Again, a similar movement has been detected in the PLP-dependent bacterial aspartate decarboxylase (40), hence suggesting a possible conserved mechanism of reaction.

As already pointed out, the limited number of conformations isolated by X-ray crystallography does not likely represent the actual conformational space explored by the protein. In this perspective, Bruno et al. (41) applied Targeted Molecular Dynamics (TMD) methods with the aim of simulating the transition between the open and the closed form of SR in the presence of either the natural substrate L-serine or the orthosteric inhibitor malonate, and of identifying the key residues and structural features necessary for the protein to stably interact with ligands. The generation of different conformations along the pathway from the open to closed state of SR is also of considerable importance for the identification by in-silico screening of new inhibitors that recognize alternative conformations of the enzyme. The simulations were run starting from the “humanized” rSR open conformation, upon the placement of ligands into the binding pocket. TMD analyses showed that, for both complexes, the process of closure of the small domain towards the large domain is a rapid event, occurring in the first 150 ps of simulation. The conformations generated by the dynamics were then clustered. Six representative structures, covering the whole transition from the open to the closed malonate-complexed forms, were selected and used for docking a small library of known SR inhibitors and closely related inactive analogs. Better results were achieved when docking was applied to the TMD generated conformations rather than to the individual crystallographic structures, indicating the relevance of including protein flexibility in computational simulations (42-44). It was found that the open conformation failed in properly recognizing small active ligands, whereas the closed form highly scored small inactive molecules (41).