Control and regulation of S‐Adenosylmethionine biosynthesis by the regulatory β subunit and quinolone‐based compounds

Methylation is an underpinning process of life and provides control for biological processes such as DNA synthesis, cell growth, and apoptosis. Methionine adenosyltransferases (MAT) produce the cellular methyl donor, S‐Adenosylmethionine (SAMe). Dysregulation of SAMe level is a relevant event in many diseases, including cancers such as hepatocellular carcinoma and colon cancer. In addition, mutation of Arg264 in MATα1 causes isolated persistent hypermethioninemia, which is characterized by low activity of the enzyme in liver and high level of plasma methionine. In mammals, MATα1/α2 and MATβV1/V2 are the catalytic and the major form of regulatory subunits, respectively. A gating loop comprising residues 113–131 is located beside the active site of catalytic subunits (MATα1/α2) and provides controlled access to the active site. Here, we provide evidence of how the gating loop facilitates the catalysis and define some of the key elements that control the catalytic efficiency. Mutation of several residues of MATα2 including Gln113, Ser114, and Arg264 lead to partial or total loss of enzymatic activity, demonstrating their critical role in catalysis. The enzymatic activity of the mutated enzymes is restored to varying degrees upon complex formation with MATβV1 or MATβV2, endorsing its role as an allosteric regulator of MATα2 in response to the levels of methionine or SAMe. Finally, the protein–protein interacting surface formed in MATα2:MATβ complexes is explored to demonstrate that several quinolone‐based compounds modulate the activity of MATα2 and its mutants, providing a rational for chemical design/intervention responsive to the level of SAMe in the cellular environment. Enzymes Methionine adenosyltransferase (http://www.chem.qmul.ac.uk/iubmb/enzyme/EC2/5/1/6.html). Database Structural data are available in the RCSB PDB database under the PDB ID http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6FBN (Q113A), http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6FBP (S114A: P22121), http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6FBO (S114A: I222), http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6FCB (P115G), http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6FCD (R264A), http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6FAJ (wtMATα2: apo), http://www.rcsb.org/pdb/search/structidSearch.do?structureId=6G6R (wtMATα2: holo)


Introduction
A highly conserved family of Methionine adenosyltranferases (MATs) produces S-Adenosylmethionine (SAMe) from methionine (Met) via an ATP-driven process [1,2]. MAT catalyzes two-step reactions: (a) SAMe is formed via nucleophilic attack of sulfur atom of methionine to C-5 0 atom of ATP and the tripolyphosphate (PPPi) is generated as an intermediate of ATP cleavage; (b) PPPi is thus hydrolyzed to pyrophosphate (PPi) and orthophosphate (Pi) by tripolyphosphatase activity of MAT. SAMe donates the methyl group to a variety of biomolecules including amino acids, sugars, nucleotide bases of DNA and RNA [3][4][5]. SAMe is an intracellular regulator of cellular proliferation, differentiation, and death in liver [6][7][8] and is involved in gene expression regulation [9,10]. Decrease in SAMe results in abnormal cellular methylation and reduction in antioxidant capacity [11]. In mammalian cells, MATa1 and MATa2 are two variants of the catalytic subunit produced from MAT1A and MAT2A gene, respectively. MATa1 functions as homo-dimer and tetramer [12,13], while MATa2 forms an hetero-oligomer with the regulatory subunit MATb (which has two predominant isoforms differ in lengths, MATbV1 and MATbV2, 334 and 323 amino acids, respectively) in 2 : 1 ratio (MAT (a2) 4 (bV1) 2 or MAT(a2) 4 (bV2) 2 ) [14,15]. MATa2 is ubiquitously expressed in healthy cells, except in the liver (mainly hepatocytes) where MATa1 is primarily expressed, while colon and liver tumor cells ectopically express high amounts of MATa2 [16][17][18][19] a switch that promotes the proliferation of cancer cells [20][21][22]. Although SAMe can be produced by MATa2 alone, the production of SAMe is increased several folds when it complexes with MATb [14]. MATb interacts with MATa2 through the insertion of the C-terminal tail of the b subunit into a cavity created at the interface of the MATa2 dimer. Comparison of the MATa2 and MATa2b complex structures revealed that the catalytic site of MATa2 is unaltered upon complex formation thus the increase in SAMe production in MATa2b complex arises from factors beyond the catalytic site through an allosteric regulation [23]. The binding of MATb to MATa2 may propagate from the MATb-binding site (Gln190-Ile198) to regions connecting the active and allosteric site (Ile241-Tyr271, Gly272-Thr288), and the region (Ala109-Gly140) that contains gating loop domain of MATa2 (Gln113-Gly131). The substrate binding and product release are likely linked to and modulated by MATb binding to the MATa2 allosteric site offering the use of heterodimeric interface as a target for chemical intervention: either stabilization of defective enzymes or inhibition of the overexpressed enzymes [24].
A gating loop, residue 113-131 in both MATa1 and MATa2, has been identified in the catalytic subunits of all MAT enzymes from prokaryotes to eukaryotes [23,[25][26][27][28][29][30]. This loop is flexible and presumed to open and close to allow or block access to the active site [26,31]. Recent crystallographic studies of MATa2 and MATa2b complex have shown that the gate is open when active site is empty or occupied by PPNP ((Diphosphono)aminophosphonic acid) and is shut like a lid when the active site is occupied by SAMe or adenosine, demonstrating that the opening of the gate to release product is not driven by hydrolysis of the tripolyphosphate (PPPi) [14,23].
The majority of amino acid residues that participate in catalysis are conserved in MAT enzymes across a wide range of species, however, the enzyme kinetics of MATs strongly depends on the source of protein [31][32][33][34]. The composition of the gating loop varies across species [31]. The influence of the length and sequence of the loop on catalytic efficiency has been studied by mutations of Escherichia coli MAT gating loop region DRADPLEQ to fragment from rat liver loop sequences (HDLRNEEDV), which resulted in decrease of more than 200-fold in the maximal rate (K cat ) of SAMe formation, with little effect on the rate of PPPi hydrolysis [35]. The active site of MAT is formed at the symmetric dimer interface by mainly hydrophobic interactions except for Arg264 which is one of the few polar residues at the interface. Mutation of Arg264 in MATa1 (Arg264His/Cys) causes isolated persistent hypermethioninemia, which is characterized by low activity of the enzyme in liver and high level of plasma methionine [36]. Most genetic mutations are inherited as an autosomal recessive trait [37]; however, dominant inheritances of hypermethioninemia have been found in patients with the Arg264 (Arg264His/Cys) and Ala259 (Ala259Val) mutations of MATa1 [36,38,39]. Arg264 facilitates dimerization by providing a positive charge to form salt bridge with Glu57 of the partner subunit and interacts with Gly263 of the partner subunit via K + ion at the central region of dimer interface to form the compact geometry of the active site [40,41]. It has been hypothesized that the abrogated activity of Arg264His/Cys is a consequence of an inability to establish a functional active site [42,43]. Here, we provide direct evidence that mutation does not compromise the dimeric association but results in the deprivation of an interaction at c-site of PPPi moiety and the loss of the enzymatic activity in a manner similar to Ser114 mutation, a residue that is part of the gating loop. Ser114 coordinates the adenine ring of the substrate AMP-PNP (a nonhydrolysable analog of ATP) via a water molecule and provides a rigid helix conformation to the closed gating loop. Surprisingly the mutation of Gln113, the only residue that interacts with Met/SAMe directly, had little effect on the enzymatic activity, as loss of Gln113 does not seem to prevent substrate Met from binding to the active site and the orientation of Met is also preserved. Intriguingly, the phosphatase activity for all of the mutants remains the same as the wild-type enzyme. High-resolution crystallographic structures of these mutants together with detailed functional data establish the significance of Ser114 as a sensor of substrate and product. We also show that the loss in activity in Arg264Ala mutant is not due to any compromise in dimer formation but is related to residue's role in the interaction with tripolyphosphate. We show that the enzymatic activity of the mutated enzymes is restored upon complexation with the regulatory protein MATb and have harnessed this observation to find druggable quinolone-based compounds that may interact with the interface region utilized in the formation of MATab complex and restore the enzymatic activity.

Results and Discussion
Enzymatic and phosphatase activity Using E. coli MAT, K cat of SAMe formation has been shown to decrease by more than 200-fold with mutations in the gating loop without effecting the tripolyphosphate hydrolysis rate [35]. Eighteen variants of MATa1 mutation sites have been identified in patients with persistent hypermethioninemia [44], seven of these mutants (Arg249Trp, Ile252Thr, Gly257Arg, Asp258Gly, Ala259Val, Lys289Ash, and Gly381Arg) presented themselves with significantly lower levels of SAMe indicating decreased activity, while their phosphatase activities were unaffected. Taken together, these studies suggested that phosphatase activity is independent on SAMe formation and the gating loop has an important role in SAMe production which does not affect the PPPi hydrolysis rate.
We explored functional significance of the conserved residues (Gln113, Ser114, and Pro115) among the variable regions of the gating loop using human MATa2 and generating a number of its mutants: Gln113Ala, Ser114Ala, and Pro115Gly mutants. Pro is the only amino acid that connects its side chain to backbone, thus we mutated it to Gly which has no side chain as well, while Gln (Q113) and Ser (S114) have movable side chains and thus more suitable to mutate to Ala to see the functional role of their side chain. We also explored Arg264 residue's role of MATa2 (Arg264Ala) and MATa1 (Arg264His), the mutation of later being mainly affecting to methionine metabolism in hepatocyte of liver and linked to inherited hypermethioninemia. We have shown recently that Gln113 and Ser114 residues interact with Met/SAMe in both the enzyme on its own and when it complexes with MATb [14,23]. Gln113Ala mutation results in 75.6 AE 4.5% of SAMe production while Ser114Ala mutation results in near abolition of enzymatic activity to only 2.5 AE 0.8% (p < 0.05) when compared to wtMATa2 (Fig. 1A). In contrast, Pro115Gly mutant has activity similar to the wild-type enzyme (wtMATa2) (Fig. 1A). MATa1 (Arg264His) and MATa2 (Arg264Ala) caused major loss in enzymatic activity with SAMe production reduced to 6.1 AE 2.6% (p < 0.05) and 4.5 AE 1.4% (p < 0.05), respectively, compared to wtMATa1/2 (Fig. 1A).
Remarkably, the SAMe production is recovered in Gln113Ala and Ser114Ala mutants when they complex with either MATbV1 or MATbV2 to the extent that the production of SAMe by Gln113Ala/Ser114Ala MATa2:MATb complexes was the same as for the wtMATa2:MATb complex confirming the regulatory allosteric role of MATb (Fig. 1A). The SAMe production by Arg264Ala MATa2 mutant also recovers upon complexation but to the level of wtMATa2 on its own. Surprisingly, none of the mutants show any change in their phosphatase activities compared to the wtMATa2 (Fig. 1B) providing clear evidence that structural elements that control SAMe synthesis are different from those involved in PPPi hydrolysis. MAT enzyme was suggested to perform ATPase activity as an intrinsic property that functions independently from SAMe synthesis [29,40]. The crystal structure of MATa1 in complex with ADP (Adenosine diphosphate) was derived from the cocrystallization of ATP (Adenosine triphosphate) and methionine analog LcisAMB (L-2amino-4-methoxy-cis-but-3-enoic acid) confirming that the hydrolysis of ATP could take place without SAMe formation [25]. Complexation with MATbV1 generally results in higher activity than when MATa2 or its mutants complex with MATbV2. The amount of SAMe produced by Arg264Ala is the same as the wild-type when it complexes with either of the isoforms of MATb subunit, thus in this case the effect of MATb regulatory subunit interaction on functional recovery of Arg264Ala mutant is independent from MATb lengths. These observations provide clear evidence that interaction with MATb has a profound effect on the catalytic activity of MAT enzyme where the length of the MATb protein has a significant impact on wild-type, Gln113Ala, Ser114Ala ( [45], suggesting that the interaction region may be a candidate for intervention to control and regulate the production of SAMe. Crystallographic structures of apo, wild-type, and mutant human MATa2 enzyme Seven crystallographic structures are reported here which include apo-and ligand-bound wtMATa2, gating loop MATa2 mutants (Gln113Ala, Ser114Ala, and Pro115Gly: Q113A, S114A, and P115G, respectively) and hypermethioninemia-associated Arg264Ala: R264A mutant. All data collection and refinement statistics are shown in Tables 1 and 2. Structure of S114A mutant has been obtained in two different crystallographic forms (I222 and P22 1 2 1 ) while P115G, R264A, and ligand-bound wtMATa2 crystals had I222 symmetry and Q113A, S114A, and apo-wtMATa2 crystallized in space group P22 1 2 1 . Two structures of S114A are identical thus confirming that the detailed structural observations are independent of the crystal form and as such truly represent the molecular behavior of the enzyme providing a rigorous basis for underlying mechanism. The intermediates and products observed in the structures reported here differ (Table 3), although the enzyme, in each case, was incubated with the same substrates Met and AMP-PNP (adenylyl-imidodiphosphate, a nonhydrolyzable analog of ATP) prior to crystallization. SAMe is observed in wtMATa2, Q113A, and P115G, all of which have similar enzymatic activities (75-110% of wtMATa2). SAMe is absent in structures of S114A and R264A mutants which have less than 4% enzymatic activity fully consistent with the inability of these mutants to perform the conversion of methionine into SAMe.

Understanding hypermethioninemia and potential therapeutic intervention
A number of pathogenic mutations have been identified in human MATa1 which result in the enzyme to metabolize methionine less efficiently. This causes the accumulation of methionine in the blood and in severe cases can develop neurological problems. These sites are found to be conserved in both MATa1 and MATa2. Some missense mutations are found at the dimer interface, for example, Arg249Trp, Ile252Thr, Gly257Arg, Ala259Val, and Arg264Cys/His. In particular, Arg264 forms a salt bridge with Glu57 of the dimeric partner. Its substitution to a cysteine or histidine residue almost completely abolished the enzymatic activity, which was hypothesized to arise from an inability to dimerize [42,43]. In the homologous rat enzyme, replacement of the equivalent Arg265 by His (R265H) resulted in a monomeric species with only 0.4% of the SAMe synthetic activity [43]. In contrast, we find that human Arg264His (R264H) MATa1 and Arg264Ala (R264A) MATa2 mutants are multimeric in solution. Two fractions of R264H MATa1 eluted from size-exclusion chromatography (SEC) correspond to tetramer and dimer, respectively, while wtMATa1 and R264A MATa2 eluted primarily as a single tetramer fraction ( Fig. 2A). Equivalent mutations (R244L/R244H) carried out in E. coli MAT also showed a tetrameric assembly in solution [46]. In addition, both R264H MATa1 and R264A MATa2 are capable of forming a hetero-oligomer with MATb as shown by analytical SEC for R264A MATa2 in Fig. 2B. The complex formation has been also confirmed by native and SDS/PAGE (Fig. 2C).
To establish the oligomeric state of the R264A mutant in solution, small-angle X-ray scattering (SAXS) data of purified R264A proteins were obtained at the SEC-SAXS facility SWING at the SOLEIL synchrotron. The SAXS data (Fig. 2D) clearly showed dominant species to be a tetramer.
Although mutants of this residue in both MATa1 and MATa2 form a homo-dimer in solution, they have no ability to produce SAMe without MATbV1/ V2 subunits, which are not coexpressed with MATa1 in the adult liver. The loss of activity caused by point mutation in Arg264 residue can be recovered to wtMATa2 level when R264A forms hetero-oligomer with MATbV1/V2 (Fig. 1A). The observation that the interaction with MATbV1/V2 can modulate and restore function of the pathogenic mutant enzyme activity offers possibility of chemical intervention in finding therapeutic solutions using recently discovered the effect of compounds with triazoloquinoline cores on wtMATa2 [24]. We used the R264A MATa2 mutant as a model to study the catalytic role of Arg264, which was conserved among MATa1 and MATa2. Crystallization of MATa1 has been problematic and attempts to crystallize R264H MATa1 gave only low diffraction quality crystals (the highest resolution we obtained was 3 A, which is not sufficient to see atomic details). We thus switched to R264A in MATa2, which diffracted at high resolution (1.7 A) with enzyme substrate/product bound and well defined in the active sites. This has allowed us to use the R264A structure as the model to understand the functional role of Arg264 in MAT enzyme. The crystallographic structure of R264A MATa2 revealed the presence of ADO (occupancy, 1.0) and Pyrophosphate (PPi) (occupancy, 1.0) in the active site ( Fig. 3A,B). Neither SAMe or Met are observed in this structure. Even though the dimerization is not prevented by Arg264 mutation, the absence of an interaction between NH2 and NE atoms of Arg264 and O2G and O3G atoms of PPNP (Fig. 3C) results in Pc of PPNP not being observed in the structure, suggesting the importance of Arg264 in stabilizing triphosphate conformation for favorable nucleophilic reaction. Arg264 contributes to the stabilization of the dimer interface through many interactions, especially by forming a salt bridge with Glu57 of the partner subunit ( Fig. 3C) Inhibition of substrate turnover by a single mutation S114A in the gating loop Two states of the S114A mutant are captured in the space groups I222 and P22 1 2 1 . The crystal structure of S114A in the I222 space group has one monomer in the asymmetric unit. PPNP (1.0) and adenosine (1.0) are clearly visible in the active site (Fig. 4A). The crystal structure of S114A in the P22 1  The open resolved gating loops of Thermococcus kodakarensis MAT in apo-state showed the movement of the gating loop domain (residues 144-155) from its closed conformation (residues 142-142), which is reported in another thermophilic archaeal MAT, Sulfolobus solfataricus [27,30] (Fig. 4E,F). The presence of PPNP and ADO in the S114A structure ( Fig. 4C) indicates that the early step of reaction has taken place, but the S114A mutant fails to provide the nucleophilic attack of methionine's sulfur atom against C'5 atom of AMP-PNP, thus unable to produce SAMe. These observations are consistent with the suggestion that C-O bond breaking takes place prior to C-S bond formation at the catalytic transition state of human  wtMATa2 [47]. Previously our high-resolution structure of SAMe+ADO+MET+PPNP wtMATa2 (PDB: 5A1I) had revealed that the Ser114 is the sole residue in the gating loop that interacts with AMP-PNP via a water molecule. We propose that Ser114 cooperates with water molecule in positioning AMP-PNP via its adenine ring to the proper moiety, which eventually facilitates enzymatic reaction and SAMe production. Ser114 also interacts with the neighboring loop residues (Asp116, Ile117, and Ala118) and stabilize the helix conformation of the ordered gating loop.
The S114A structure provides information that the gating loop becomes more ordered in the closed conformation as observed by the differences in a helical initiation and termination residue of the closed (Pro115-His122) and open (Ile117-Val121) conformation (Fig. 4B): closed gating loop when active site is occupied by ligands which is opened when the active site is empty. In the wtMATa2 (PDB: 5A1I and structure reported here), two active sites are identical, occupied by substrates or products with gating loop in closed conformation, sequestration of reactive intermediates from escaping into solution. In contrast, in both S114A and Q113A mutants (see below), only one active site is occupied by substrate or product suggesting that the gating loop cannot rearrange to the closed conformation properly during the substrate binding step, so the reactive intermediates might be able to leave the enzyme. Two different loop conformations presented in S114A structure indicate that the loop has the ability to adopt different conformations when SAMe formation does not occur.

Production, control, and regulation of SAMe
Given the dramatic effect of Ser114 mutation on the enzymatic activity of MATa2, mutations of the two neighboring residues Gln113 and Pro115 were undertaken and crystallographic structures determined. As both of these mutants produce a relatively similar amount of SAMe as the wild-type enzyme, a structural comparison with wtMATa2 in its apo-and ligandbound forms were also undertaken to gain insight into the control and regulation of catalysis.
The crystal structure of Q113A in the P22 1 2 1 space group has two MATa2 monomers in the asymmetric unit (Fig. 5A). One of the monomers has closed wellordered gating loop with full occupancy of SAMe and PPNP bound in the catalytic site (Fig. 5B). The second monomer has no ligands in the active site, and its gating loop is mostly disordered and does not have any electron density for residues 117-131 (Fig. 5A). In the wtMATa2, the N and O atoms of substrate methionine's main chain form hydrogen bonds with the enzyme at OE1 atom of Glu70 and NE2 atom of Gln113 of one subunit, respectively, and its N atom also interacts with OD1 atom of Asp258 belonging to the dimeric partner subunit. Mutation Q113A does not affect the binding of Met and SAMe to the active site. N atom of SAMe forms hydrogen bonds with OE1 atom of Glu70 (2.6 A) and to OD1 atom of Asp258 (2.5 A) and O3 atom of SAMe forms hydrogen bond with OD2 atom of Asp258 (3.0 A) maintaining the position of both substrate and product in the active site (Fig. 5C). The loss of interaction with Met and ADO due to mutation to Ala is likely to be responsible for less efficient SAMe production (75% compared to wtMATa2) but the major role of Gln113 is to sense the substrate entry and stabilize the closed gating loop through its interaction with Met, while the entire orientation of the SAMe position is maintained by other residues in the active site pocket as observed from the crystal structure.
The different conformations of the gating loop observed in Q113A and S114A MATa2 mutant structures clearly shows the importance of these two residues in catalysis through regulating the gating loop conformation. These residues sense the presence of substrate Met and ATP in the active site via the interaction of Gln113 with Met and Ser114 with adenine ring of AMP-PNP, an ATP analog used in the present study (Fig. 5D). These important interactions of the gating loop with the substrates drive the rearrangement of the gating loop to its closed conformation. The crystal structure of P115G MATa2 displays the closed well-ordered conformation of the gating loop with full occupancy of SAMe and PPNP in the active site (Fig. 5E). It does not make any contacts with the substrate or the product (Fig. 5F). Despite being a neighboring residue of Ser114, the mutation has no effect on the overall arrangement of the gating loop either. SAMe interacts with Glu70 and Gln113 of one subunit and Asp258 of the partner subunit to maintain itself in the active site pocket. N and O3 atoms of SAMe form a H bond with OD1 and OD2 atoms of Asp258, respectively. The N atom of SAMe also forms a H bond with the OE1 atom of Glu70, while the O atom of SAMe interacts with the NE2 atom of Gln113.
We have also obtained the structure of apo-wtMATa2 where the gating loops are ordered and visible in the open conformation (Fig. 6A). The structure has a dimer in the asymmetric unit and belongs to P22 1 2 1 space group similar to S114A mutant. One active site is dilated because of the gating loop, adjacent loop (residues 245-264), and residues 320-328 being oriented in an outward manner to the substratebinding cleft, the other subunit features a disordered open gating loop with a less dilated binding cleft (Fig. 6B). Structural comparisons between apo-wtMATa2 and ligand bound (holo) wtMATa2, S114A, and R264A have revealed that apart from majority movement in the gating loop region itself, the adjacent loop and residues 320-328 that contribute to the substrate/product-binding pocket (Fig. 6A,B) of holo-wtMATa2 and R264A line inward to the active site pocket, while these substrate-binding clefts of S114A are oriented outwardly from the active site similarly found in apo-wtMATa2. These comparisons support the importance of Ser114 interaction with substrates in the rearrangement of the gating loop into its closed conformation and preventing the substrate-binding pocket residues from moving apart from the active site, which fails in S114A mutant. Most residues that are involved in catalysis including Ser247, Arg249 (Fig. 6A,B) have similar conformation in apo-and holo-wtMATa2, except for Phe250 and Asp258. These residues move from their original positions in the apo-wtMATa2 (Fig. 6C) A) of SAMe, respectively, while Phe250 interacts with SAMe through p-p stacking between adenine and aromatic ring of Phe250 (Fig. 6C). The positions of Glu23, His29, Lys181, Lys265, Lys285, Lys289, and Asp291 which are involved in the catalysis are preserved in both apo-and holo-wtMATa2. The dilation of the substrate-binding pocket found in the MAT (a2) 4 (bV2) 2 complex is identical to the open site found in the current apo-wtMATa2, suggesting that this dilation is not the unique property of the complex.

Interaction with MATb and open gating loop in apo-MATa2
We note that MATb interacts with the catalytic subunit residues that recognize the PPPi of the ATP in the active site [14] via the helical loop (Fig. 7A,B).  of Arg264 (MATa2) interacts with the O1G and O3G atom of PPNP (Fig. 7B). The present study has found that the phosphatase activity of the complexes is significantly higher than that of the catalytic subunits alone (Fig. 1B). Triphosphate is a product intermediate generated from SAMe formation and it has to be hydrolyzed before being released from the active site to allow MAT enzyme to regenerate for the next reaction cycle. The underlying mechanism by which the observed activities of the mutants are recovered by complex formation is in part facilitated by the profound increase in phosphatase activities of the enzyme complexes and is consistent with the protein dynamics study that showed MATa2 ATP-binding region to undergo allosteric modification upon binding of MATb [24]. Structural comparison of the wtMATa2 and MATa2b complex revealed that conformation of the residues in the catalytic site of wtMATa2 is unaltered upon complex formation with MATb; however, several movements of the solvent exposed a-helices are found in the complex formation. Residues 341-353 are observed prominently being oriented outward when compared to those in apo-and holo-oligomeric of wtMATa2. The MATa2b complex structure shows an increase in a-helix solvent-exposed area along the MAT(a2) 4 surface when compared to the homo-oligomer of the wild-type (Fig. 7A). Also, the complex stability of both MATa2bV1 (Tm = 55.2 AE 0.1°C, P < 0.001) and MATa2bV2 (Tm = 48.4 AE 0.1°C) are significantly increased when compared to catalytic subunit alone (Tm = 47.2 AE 0.2°C, P < 0.001). Overall active site conformations for both S114A and R264A mutants are similar to wild-type protein with 0.2 A movements of residues 250-286 away from the dimer interface in S114A mutant. Apo-S114A mutant structure is similar to Apo-wtMATa2, both have a wider open active cleft than in holo-wtMATa2, while in the MATa2b complex this cleft takes an intermediate position (between apo and holo). A less dilated substrate-binding pocket of MATa2b is a result of the MATb interaction. The local changes in the complex suggest that MATb insertion modulates the tightness of the dimer interface (residues 250-286) that comprises substrate-binding pocket which could result in different degrees of the activity recovery for R264A and S114A mutants.

Recovery of enzymatic activity through chemical intervention
The search for chemical compounds that effect the SAMe production of MATa2 enzyme has seen limited success. A recent study [24] showed that 1-[2-(dimethylamino) ethyl] quinoline benzodiazepine derivative (PF-9366) affects the catalysis significantly and competes with MATb protein by binding in the interaction region that is used by MATb for complex formation of MATa2b. PF-9366 binding to Apo-wtMATa2 causes a closure of the interdimer interface similar to holo-wtMATa2 structure with the biggest  in domain composed by residues 181-269 (Fig. 8A).
We have undertaken a focused screening of quinolone-based compounds (Fig. 8B) on the enzymatic activities of wtMATa1, wtMATa2, and MATa2 mutants (S114A and R264A). For the mutant enzymes where the mutation had caused near complete abolition of SAMe production, the activity is restored by these quinolone-based compounds to varying degrees (Fig. 8C). In fact for wtMATa1, the enzymatic activity doubles with a number of compounds while the SAMe production of wtMATa2 is marginally reduced by all of these compounds. In contrast, none of these quinolone compounds show any change in the phosphatase activities of wtMATa1 and wtMATa2 (Fig. 8D), while the binding of MATb to MATa2 and mutants significantly elevates their phosphatase activities. This selectivity of the quinolone compounds for modulating the SAMe production of the wild-type and disease-causing mutants offers a clear opportunity for chemical intervention for therapeutic purposes by altering SAMe levels in the cell.
To gain further insight, we used SCR0915 to study the effect of a quinolone-based compound on wtMATa2 and MATa2 R264A mutant (Fig. 9A-D). In wtMATa2, SCR0915 acts as an inhibitor in the concentration-dependent manner (Fig. 9A). The inhibition curve and IC 50 of SCR0915 on wtMATa2 are determined and shown in Fig. 9C. SCR0915 does not alter the thermal stability of wtMATa2. In contrast, SCR0915 could recover R264A activity at 0.1-100 lM (Fig. 9B) and R264A melting temperature increased to 41.48 AE 0.06°C compared with R264A alone (39.2 AE 0.07°C), indicating that SCR0915 increases the mutant stability (Fig. 9D). The compound behaves as an allosteric regulator to activate MATa2 when Met or SAMe are at a low concentration, but it inhibits enzyme activity when Met or SAMe concentrations are high. One possible mechanism is that the compound renders MATa2 mutant to be more sensitive to SAMe inhibition, which explains why it decreases the amount of SAMe in wtMATa2, and increases SAMe production of MATa2 R264A only to 41.23 AE 1.37% relatively to wtMATa2 (Fig. 8C), while MATb could recover SAMe levels of MATa2 R264A to the same level found in wild-type (Fig. 1A).

Molecular modeling of SCR0915-binding site
The binding of SCR0915 was explored using Autodock Vina and SwissDock using structures of both wtMATa2 and MATa2 (R264A). The binding modes are obtained from both local and blind docking and the most favorable energies are evaluated based on Gibbs free energy (ΔG) in Autodock Vina and FullFitness score in SwissDock. The two docking programs show the best scored positions of SCR0915-binding site at similar dimeric interface of MATa2 in both wtMATa2 and MATa2 (R264A). This dimeric interface-binding pocket has been reported to interact with MATb and PF-9366 compound [14,24]. The best binding position of SCR0915 using wtMATa2 structure as a model (PDB, 5A1I) predicted by Autodock vina (ΔG = À11.4 kcalÁmol À1 ) and SwissDock (FullFitness = À3486.08 kcalÁmol À1 , and ΔG = À7.8 kcalÁmol À1 ) are shown in Fig 10A-C. The best binding position of SCR0915 using MATa2 (R264A) structure as a model (PDB, 6FCD) predicted by Autodock vina (ΔG = À10.2 kcalÁmol À1 ) and SwissDock (FullFitness = À3111.79 kcalÁmol À1 , and ΔG = À8.1 kcalÁmol À1 ) are shown in Fig 10D-F. PF-9366-binding structure (PDB, 5UGH) is shown with SCR0915-binding site (Fig. 10G). Two molecules of PF-9366 bound to MATa2 dimer (one molecule to one MATa2 monomer), while one molecule of SCR0915 was predicted to bind to the dimer (Fig. 10G,H). We test whether SCR0915 could inhibit the formation of MATa2-MATb complexes by using SEC analysis. MATa2 was preincubated with MATa2 before MATbV1 is introduced into the solution. We find that SCR0915 partially excludes MATb from binding to MATa2 (Fig. 10I), confirming that SCR0915 interaction sites to MATa2 overlap with the interacting site of MATb.

Conclusion
The gating loop shows distinct behavior during steps of enzyme catalysis in apo-, PPNP-, SAMe-, Met-, and ADO-bound structures providing direct evidence by which the gating loop facilitates the enzyme catalysis. Functional analysis and crystallographic structures in the present study elucidate the nonputative role of the MAT gating loop where the single residue, Ser114, is involved in catalytic reaction by coordinating the adenine ring of the substrate ATP via a water molecule and also providing a rigid helix conformation of the closed gating loop. The observation of the ordered open conformation of the gating loop for the first time provides a clear understanding on how it undergoes structural changes during enzyme catalysis upon substrate binding and product release as observed in Q113A and S114A mutant structures. The interactions of the gating loop with substrates drive the rearrangement to the closed conformation. We also provide the explanation of why Arg264 mutation results in the loss of the enzymatic activity as found in patients with hypermethioninemia. The loss in activity results from the loss of interaction between Arg264 and PPPi moiety at the c-site. We also observed that the phosphatase activity and SAMe production involve different parts of the catalytic pocket of the enzyme as phosphatase activity is not affected by the pathogenic mutations where SAMe production is dramatically decreased. A focused screening of quinolone-based compound that exploits heterodimeric interface as a target for chemical intervention, shows selectivity of the compounds for modulating the SAMe production of the wild-type and disease-causing mutant interface (Fig. 11), offering a clear opportunity for therapeutic solutions. The compounds behave as an allosteric regulator to activate MATa2 when Met or SAMe are at low concentrations but inhibits enzyme activity when Met or SAMe concentrations are high.

Materials and methods
Site-directed mutagenesis

SAMe formation detection
S-adenosyl-methionine formation was analyzed by S-adenosylmethionine enzyme-linked immunosorbent assay (ELISA) kit (Cell biolabs, San Diego, CA, USA) following the manufacturer's protocol. The measurements were

Phosphatase activity
The detection of inorganic phosphate was modified from Baginski et al. [48]. One hundred and fifty microliter reaction samples were mixed with 150 lL of 2% Ascorbic acid in 10% trichloroacetic acid (A-TCA). Fifty microliters of water was added in each sample, followed by 75 lL of 2% ammonium molybdate, and 150 lL of 2% (W/V) arsenitecitrate in 10% TCA. All samples were mixed thoroughly (vortex) and incubated at room temperature for 20 min. The samples were measured at 750 nm. Pi quantification was obtained by using a calibration curve obtained from various standard KH 2 PO 4 solutions covering a Pi concentration range of 0.125-2.0 lLÁmL À1 .

Quinolone-based compounds
The synthesis of quinolone-based compounds has been described by one of the authors in Biagini et al. [49] and Charoensutthivarakul et al. [50]. For quinolone-based compound study, 10 lL of MATa2 proteins (3.2 mgÁmL À1 ) and 1.5 mM of SCR0915 are diluted in dilution buffer (10 mM HEPES pH 7.5, 500 mM NaCl, 5% (v/v) glycerol and 0.5 mM TCEP). Fluorescence intensities were measured from 25°C to 99°C with a ramp rate of 1°CÁmin À1 . Data were analyzed using MATLAB Ò executable TmTool TM following TmTool TM Quick Set-Up Guide (Life technologies). Data are represented as the average Tm of the independent experiments AE SEM (n = 3).

Crystallization and data collection
All MATa2 mutants were eluted as a single peak by sizeexclusion chromatography. Proteins were concentrated to 5.8 mgÁmL À1 and pre-equilibrated with substrates (10 mM Met and 150 lM AMP-PNP, nonhydrolyzable analog of ATP) in 50 mM HEPES buffer pH 7.5, 10 mM MgCl 2 , 50 mM KCl, and 10 mM DTT before crystallization. Crystal drops containing 1 lL of protein and 1 lL of 100 mM HEPES pH 6.5 and 30% PEG 600 were equilibrated against reservoir solution (100 mM HEPES pH 6.5 and 30% PEG 600). Crystals appeared at 25°C within 1-2 days. Prior to data collection, crystals were flash frozen in reservoir solution with additional 20% ethylene glycol. Different datasets were collected at the I24, I03, and I04-1 beamlines at Diamond Light Source (Oxford, England) and the PROXIMA-1 beamline at SOLEIL synchrotron (Saint-Aubin, France). Data were integrated by iMosflm [51,52] and scaled using Aimless [53]. The initial model of wtMAT2 (PDB: 5A1I) was from crystals with I222 space group. Molecular replacement (MOLREP software) [54] is needed for mutant structures with P22 1 2 1 space group, while mutants with I222 space group could be taken straight to restraint refinement using REFMAC [55]. Model building and restrained refinement were carried out using coot [56] and REFMAC [55]. Crystallographic data collection and refinement statistics and the summary of the enzymatic activities and ligands bound in each structure of MATa2 mutants are shown in Tables 1  and 2.

Small-angle X-ray scattering
Size-exclusion chromatography-small-angle X-ray scattering (SEC-SAXS) measurements were performed on the SWING beamline at Synchrotron SOLEIL. An Agilent BioSEC Advance 300 A, 4.6 9 300 mm column equilibrated with 25 mM HEPES pH 7.5 was used for a gel filtration step immediately prior to exposure to X-rays. Twenty microliters of protein at 10 mgÁmL À1 was injected through an Agilent 1200 HPLC system at a flow rate of 0.3 mLÁmin À1 and column temperature of 15°C. Exposures were collected prior to the column void volume, for the purposes of buffer scattering subtraction, and over the course of protein elution. Exposures were interspersed for 1 s with a 0.5-s dead-time. Frames were inspected and averaged in Foxtrot and analyzed with Primus. The scattering data obtained by SEC-SAXS are computed and compared to the dimeric and tetrameric oligomer of the crystal structures using FoXS Server [57]. FoXS computes the SAXS profile from the input crystallographic structure and fits it onto the experimental profile.

Molecular docking
Both local and blind dockings are carried out using Autodock Vina [58] and SwissDock [59,60], respectively. Many binding modes are generated either in a box (local docking) or in the vicinity of all target cavities (blind docking). SCR0915 molecule in Tripos Mol2 File Format (.mol2) is generated using Marvinsketch tool (https://chemaxon.com/ products/marvin). The crystal structure of wtMATa2 (PDB: 5A1I) and MATa2 (R264A) (PDB: 6FCD) which all ligands are excluded is used as a target protein model.
In Autodock Vina studies, the similar grid parameters (box center: x = À2. 43 A, y = 11.29 A and z = À29. determined in both wtMATa2 and MATa2 (R264A) models. All binding modes are generated in the determined box with a number of all possible clusters and the most favorable energy binding mode is evaluated based on its binding energy (ΔG binding : kcalÁmol À1 ).
In the Swissdock study, the target protein model in PDB file format and ligand in Mol2 format are uploaded onto a web browser interface (http://www.swissdock.ch/docking). Many binding modes are generated in the vicinity of all target cavities (blind docking). Once the docking processes are terminated, all possible binding clusters are available to download and visualized by Viewdock tool in UCSF Chimera suite [61]. The best binding pose is obtained by Gibbs free energy (ΔG) and FullFitness scores.
The best pose of SCR0915 is selected and combined with its model. The protein-ligand complex PDB file format is written by using UCSF Chimera tool. The protein-ligand interaction profiles are generated and visualized by DS visualizer (Dassault Syst emes BIOVIA, Discovery Studio Modeling Environment, Release 2017, San Diego: Dassault Syst emes, 2016).