Structural basis of the dominant inheritance of hypermethioninemia associated with the Arg264His mutation in the MAT1A gene

Dominant inheritance of hypermethioninemia is caused by an Arg264His mutation in methionine adenosyltransferase. The mutation lowers the affinity for dimer–dimer interaction and enzymatic activity, which can be restored by chemical intervention.


Introduction
Methionine adenosyltransferase (MAT) deficiency (OMIM 250850) is the most common cause of isolated persistent hypermethioninemia (IPH), which has been established as an inborn error of metabolism (Gaull & Tallan, 1974). MAT enzymes (EC 2.5.1.6) catalyse the formation of S-adenosyl-lmethionine (SAMe) from methionine (Met) and adenosine triphosphate (ATP). MATs are found in all living organisms, with the exception of some intracellular parasites that obtain SAMe from their host (Sá nchez-Pé rez et al., 2004). Normal cellular function and survival require SAMe as a versatile molecule with roles ranging from gene expression to membrane fluidity (Finkelstein, 1990;Friedel et al., 1989;Lu, 2000;Mato et al., 1997). In mammalian cells, MAT1 and MAT2 are two variants of the catalytic subunit encoded by the MAT1A and MAT2A genes, respectively (Chamberlin et al., 1996;De La Rosa et al., 1995). MAT1 has 84% aminoacid sequence similarity to MAT2, despite their distinct kinetic properties and expression in distinct tissues. MAT1 is a liver-specific enzyme that is mainly found in mature hepatocytes, while MAT2 is expressed in extrahepatic tissues including kidney, brain and heart and also in developing liver ISSN 2059-7983 cells (Alvarez et al., 1993;Horikawa & Tsukada, 1992;Kotb & Geller, 1993). Patients with IPH have low levels of hepatic MAT (MAT1) activity, resulting in high levels of methionine in the blood plasma. Case reports of MAT1 deficiency were identified through newborn screening programmes where the activity of MAT2 in erythrocytes, lymphocytes and fibroblasts was found to be normal (Gahl et al., 1988;Gaull et al., 1981). Hepatic MAT deficiency is usually inherited as an autosomal recessive trait (Chamberlin et al., 1996(Chamberlin et al., , 2000Mudd et al., 1995;Ubagai et al., 1995) or as the exceptional Arg264His (R264H) autosomal dominant mutation (Chien et al., 2005;Couce et al., 2008Couce et al., , 2013Martins et al., 2012;Ubagai et al., 1995). Neurological problems, including demyelination, abnormal mental development and cognitive impairment, have been reported in severe cases of autosomal recessive IPH, while the majority of patients with a mild to moderate condition present no clinical manifestations (Chamberlin et al., 1996(Chamberlin et al., , 2000Hazelwood et al., 1998;Kido et al., 2019;Tada et al., 2004;Sen et al., 2019). R264H is the most prevalent mutation found to cause IPH-associated MAT1 deficiency in newborn screening programmes, as reported in the USA (three of 13 cases), Spain (15 of 18 cases), Japan (14 of 24 cases), Portugal (all 12 cases) and Taiwan (three of 16 cases) (Couce et al., 2013;Martins et al., 2012;Nagao et al., 2013;Chadwick et al., 2014;Chien et al., 2005). IPH related to the R264H mutation is usually clinically benign in the early stages of life (Couce et al., 2013;Nagao et al., 2013); however, the discovery of a three-year-old girl with myelination abnormalities and the severe vascular diseases reported in some cases related to this mutation suggest that clinical monitoring should be performed (Martins et al., 2012).
SAMe is the principal methyl donor in the transmethylation processes of important cellular biomolecules, including DNA, RNA, proteins, neurotransmitters etc. (Chiang et al., 1996). It is also involved in the trans-sulfuration pathway related to the biosynthesis of major cellular antioxidants (for example glutathione and taurine). Furthermore, SAMe is a precursor in polyamine synthesis, which is important in biological processes including cell proliferation, cell differentiation and apoptosis. The formation of SAMe and its level are crucial in normal cellular functions that derive from its versatile roles, while its level is found to be relatively low when the methionine level is high in patients with IPH (Mudd, 2011).
The catalytic MAT isoforms are biologically active only when they assemble as a dimer or a tetramer, as the functional active sites are formed at the dimer interface. MAT1 functions as a homodimer and a homotetramer (Kotb et al., 1997), while MAT2 forms a hetero-oligomer with the regulatory subunit MAT, which has two isoforms (MATV1 and MATV2). MAT2 has been found to form complexes in a 2:1 ratio with the V1 and V2 isoforms of the regulatory subunit [MAT(2) 4 (V1) 2 and MAT(2) 4 (V2) 2 ] (LeGros et al., 2000(LeGros et al., , 2001Lu, 2000;Martínez-Chantar et al., 2003;Murray et al., 2014). In order to understand the structure-function relationship behind the dominant effect of the R264H mutation on MAT enzymes, we generated the R264H mutant of the MAT1 protein, measured its enzymatic activity and deter-mined the oligomeric form of this mutant both in solution and in three-dimensional crystallographic structures. In the central region of the wild-type MAT1 dimer the MAT subunits interact through polar contacts, creating a salt bridge between Arg264 of one subunit and Glu57 of the other. This interaction has been reported to be essential for dimerization and to also be of importance for the formation of the active site (Chamberlin et al., 1997;Mato et al., 2001). Our data show that the R264H mutant exists as a dimer and a tetramer in solution in almost equal amounts, in contrast to the wild-type enzyme, which exists primarily as a tetramer. The R264H mutant crystallized in the dimeric form, which provided a 2.85 Å resolution crystallographic structure. Our data suggest that the R264H mutation does not affect dimerization, but increases the propensity to form a dimer compared with the wild-type enzyme. The crystal structure was obtained in the apo form with no methionine or SAMe bound, even when we attempted to co-crystallize the R264H mutant with the substrates methionine and AMP-PNP, which was consistent with the defective nature of the catalytic machinery of the mutant enzyme. We show that the activity of the R264H mutant can be restored to the levels found for the wild-type enzyme when the mutant is incubated with the quinolone compound SCR0911 or the regulatory subunit MATV1. This discovery opens an opportunity for chemical intervention to alleviate this diseasecausing defect in catalysis.

Protein expression
The MAT1 plasmid (pNIC28-Bsa4) construct was generously provided by the Oxford Structural Genomics Consortium. The MATV1 and MATV2 constructs were kindly provided by the laboratory of Dr Shelly Lu, University of California Los Angeles and were subcloned into pET-28M-SUMO3 vector. All plasmids were transformed into Escherichia coli strain BL21 (DE3). Transformed cells were plated on selective LB agar supplemented with 50 mg ml À1 kanamycin and grown overnight at 37 C. A single colony was used to make a starter culture. 5 ml of the freshly grown starter culture was added to 500 ml LB medium mixed with 50 mg ml À1 kanamycin and the cultures were shaken at 240 rev min À1 and 37 C until the cell density reached an OD 600 of 0.6-0.8. Protein expression was induced by the addition of 1 mM isopropyl -d-1-thiogalactopyranoside (IPTG). The cultures were grown overnight at 240 rev min À1 and 20 C. The cells were harvested by centrifugation at 6000g for 20 min at 4 C and flash-frozen using liquid nitrogen before storage at À80 C.

Protein purification for MATa1 and MATbV1/V2
The cell pellets were resuspended in ice-cold lysis buffer [500 mM NaCl, 5%(v/v) glycerol, 5 mM imidazole, 10 mM -mercaptoethanol (ME)] mixed with 1 mM phenylmethylsulfonyl fluoride (PMSF) at a ratio of 1 g pellet to 4 ml lysis buffer. For cell disruption, 25 ml cell suspension was sonicated 15 times on ice (30 s on/off cycles). The cell suspension was centrifuged at 20 000g for 40 min to remove cell debris. The clear supernatant was collected and loaded onto a HisTrap HP column (GE Healthcare, Chicago, Illinois, USA) pre-equilibrated with lysis buffer. The column was washed with ten column volumes of lysis buffer and then with five column volumes of wash buffer [500 mM NaCl, 5%(v/v) glycerol, 30 mM imidazole, 10 mM -ME]. The proteins were eluted with elution buffer (500 mM NaCl, 250 mM imidazole, 10 mM -ME). To cleave the His tag from MAT1, the eluted fractions containing the protein were pooled together and incubated with Tobacco etch virus (TEV) protease at a 100:1 (protein:TEV protease) ratio followed by dialysis against dialysis buffer (25 mM HEPES pH 7.5, 250 mM NaCl, 10 mM -ME) overnight at 4 C. To cleave the SUMO tag from MATV1 and MATV2, the eluted proteins were incubated with sentrin-specific protease 2 (SENP2) at a 1500:1 (protein: SENP2) ratio. The proteins were incubated at room temperature for 1 h on an orbital shaker and then dialysed against dialysis buffer overnight at 4 C. The samples were centrifuged at 4000g for 30 min to remove protein precipitation. The proteins were concentrated to 3 ml and loaded onto a HiLoad 16/600 Superdex 200 gel-filtration column (GE Healthcare, Chicago, Illinois, USA) pre-equilibrated with 25 mM HEPES pH 7.5, 250 mM NaCl, 10 mM -ME. Fractions containing proteins were pooled together and mixed with storage buffer [final concentrations of 500 mM NaCl, 0.5 mM tris(2-carboxyethyl) phosphine (TCEP), 5% glycerol]. The proteins were then flash-frozen in liquid nitrogen and stored at À80 C.

MATa1-MATb complex formation
Complex formation was performed following the protocol described by Murray et al. (2014). Briefly, MAT1 was incubated with MATV1 or MATV2 in a 1:2 ratio for 1 h at 4 C in a buffer consisting of 50 mM HEPES pH 7.5, 10 mM MgCl 2 , 50 mM KCl. The complex was then loaded onto a Superdex 200 10/300 gel-filtration column pre-equilibrated with 25 mM HEPES pH 7.5 containing 200 mM NaCl, 1 mM MgCl 2 , 5 mM KCl and 1 mM TCEP.

Activity assay
The enzyme-activity assay was performed following the protocol described by Murray et al. (2014). The final concentration of the proteins was 50 nM and the final concentration of SCR0911 was 10 mM. The synthesis of SCR0911 has been described by Charoensutthivarakul et al. (2015). The SAMe synthetic activity was measured by measuring the production of SAMe. The reactions were stopped by adding 50 ml 100 mM EDTA. SAMe formation was analysed using an S-adenosylmethionine enzyme-linked immunosorbent assay (ELISA) kit (Cell Biolabs, San Diego, California, USA) following the manufacturer's protocol. The measurements were performed in triplicate. Data were presented as the mean AE the standard error of the mean (SEM). Significance was assessed using a one-way analysis of variance (ANOVA) followed by Tukey-Kramer tests using GraphPad Prism version 5. P values of less than 0.05 were considered to be significant.

Crystallization and data collection
The R264H mutant protein was concentrated to 5 mg ml À1 and pre-equilibrated with substrates [10 mM methionine and 150 mM AMP-PNP (adenylyl-imidodiphosphate), a nonhydrolysable analogue of ATP] in 50 mM HEPES buffer pH 7.5 containing 10 mM MgCl 2 , 50 mM KCl and 10 mM dithiothreitol (DTT) for 30 min before crystallization. The wild-type protein was concentrated to 5 mg ml À1 and pre-equilibrated with 50 mM HEPES buffer pH 7.5 containing 10 mM MgCl 2 , 50 mM KCl and 10 mM DTT before crystallization. Crystal drops consisting of 1 ml protein solution and 1 ml reservoir solution (200 mM NaF, 20% PEG 3350, 15% ethylene glycol pH 8.0) were equilibrated against reservoir solution. Crystals appeared at 25 C within three days. Prior to data collection, crystals were flash-cooled in reservoir solution with an additional 20% ethylene glycol. Data were collected at Diamond Light Source, Oxford, England. Data for the R264H MAT1 crystal were collected on beamline I03 at a wavelength of 0.9762 Å using an EIGER2 XE 16M detector. The data for wild-type MAT1 were collected on beamline I04 at a wavelength of 0.9795 Å using an EIGER2 XE 16M detector. Data were integrated using iMosflm (Battye et al., 2011) and scaled using AIMLESS (Evans & Murshudov, 2013) as implemented in the CCP4i interface . The crystal structures were solved by MOLREP (Vagin & Teplyakov, 2010) using human MAT1 (PDB entry 2obv; Shafqat et al., 2013) as a search model. Model building and restrained refinement were carried out using Coot (Emsley et al., 2010) and REFMAC5 . For the lower resolution R264H mutant structure, noncrystallographic symmetry (NCS) restraints were used during refinement. Crystallographic data-collection and refinement statistics are given in Table 1.

research papers
Guide (Life Technologies, Carlsbad, California, USA). Data were represented as the mean T m of the three independent experiments AE the SEM.

Molecular docking
Prediction of the SCR0911 binding site was carried out following the method described in Panmanee et al. (2019). Briefly, blind docking was performed using SwissDock (Grosdidier et al., 2011a,b). Numerous binding modes were obtained in the vicinity of all target cavities (blind docking). A Tripos Mol2 file (.mol2) for SCR0911 was created using a MarvinSketch tool (https://chemaxon.com/products/marvin). The crystal structure of the R264H mutant excluding all ligands was used as the target model. The target protein model in PDB file format and ligand in Mol2 format were uploaded using the web-browser interface (http://www.swissdock.ch/ docking). Once the docking processes had completed, all possible binding clusters could be downloaded and visualized using the ViewDock tool in the UCSF Chimera suite (Pettersen et al., 2004). The best binding pose was selected based on the lowest Gibbs free energy (ÁG) and FullFitness score. The protein-ligand complex model was written out in PDB file format using a UCSF Chimera tool. The proteinligand interaction profiles were generated and visualized using DS Visualizer (Dassault Systè mes BIOVIA, San Diego, California, USA).

MAT sequence analysis
Multiple sequence-alignment analysis of the human MAT1 protein was performed with 2963 amino-acid sequences from the MAT protein family using GREMLIN conservation analysis (Ovchinnikov et al., 2014). Sequence-conservation analysis was performed using WebLogo 3 (Crooks et al., 2004).

Results and discussion
3.1. The crystal structure of tetrameric wild-type MATa1: an apo form We report the first crystal structure of human apo MAT1, which was solved at 2.3 Å resolution, revealing a tetrameric assembly. The crystal belonged to space group C2, with 45% solvent content. The asymmetric unit contains two dimers comprised of four subunits (Fig. 1a). Two equivalent active sites are present per dimer, composed of residues from both subunits. Each MAT1 subunit consists of 395 residues, within which there are three major regions that are invisible in the electron-density maps of all subunits (chains A-D). The first region is the N-terminal region Met1-Glu15, which has been found to be missing in most reported MAT structures owing to its flexibility. The second region is the gating-loop region, Asp116-Glu127, which has also been found to be flexible in the E. coli MAT (eMAT), rat liver MAT (rMAT) and human MAT1/MAT2 structures. This loop has been suggested to regulate access to the active site by adopting an open or closed conformation. The SAMe-bound structure of MAT1 was found to have a closed conformation of the gating loop, with the binding of SAMe being partially stabilized by this loop (Shafqat et al., 2013). The gating loop is generally in the open conformation and is not present in the unliganded structure (apo form). The open conformation of the gating loop allows the active site to be solvent-accessible (Shafqat et al., 2013). The last invisible region is the Phe250-Ala259 domain, which contains methionine-binding sites located at the dimer interface. Flexibility of this region has been reported in the rMAT structure (Gonzá lez et al., 2000), but it was clearly seen in the apo structure of human MAT2 despite the similarity of its sequence to that of MAT1 (Panmanee et al., 2019). When substrates bind to the active site, the gating loop adopts the closed conformation owing to the stabilization of the Phe250-Ala259 region by the presence of SAMe (Komoto et al., 2004;Murray et al., 2014Murray et al., , 2016Shafqat et al., 2013). In addition, the Asp94-Phe99 domain is absent in chains A and D of the present apo MAT1 structure. The average temperature factors (B factors) of this region calculated for chains B and C were 82.0 and 82.2 Å 2 , respectively, while overall B factors of 33.8 and 38.6 Å 2 were found for chains B and C. The higher B factor for this domain (Asp94-Phe99) compared with the average B factor of the entire chain indicates that this domain is very flexible; it thus becomes disordered and eventually becomes invisible in chains A and D. This region is located on the solvent-exposed surface and features as a turning-loop domain connecting the -helix (Tyr79-Ile90) and -sheet (Lys53-Thr72 and Asn105-Gln112) domains.
Each subunit of the dimer largely interacts via -sheets in an inverse contact. This generates two active sites per dimer at the dimer interface (Fig. 1b). Only a few polar interactions are found at the dimer interface, mainly involving the formation of  salt bridges between Arg264 and its partner Glu57 from the interacting subunit (Fig. 1c). The previously reported SAMe-bound structure (PDB entry 2obv) contains one MAT1 monomer in the asymmetric unit in space group I222 (Shafqat et al., 2013). The biological assemblies of an identical subunit in a dimer and a tetramer were generated by twofold-symmetry axes. Structural comparison of the active-site residues of the SAMe-bound structure and the present apo structure of MAT1 revealed an overall root-mean-square deviation (r.m.s.d.) of 0.687 Å . These two structures align well, except for the active site and the gating-loop regions that adopt the conformations observed in the SAMe-bound structure. At the active site, the aromatic ring of Phe250 in the apo structure is located in the SAMebinding site, where it formsstacking with the adenine moiety of SAMe in the SAMe-bound structure ( Fig. 1d; Shafqat et al., 2013). In addition, the O atom of Ser247, the OE1 atom of Glu70 and the OD1 atom of Asp258 of the apo structure move by 3.9, 6.5 and 3.9 Å , respectively, from their positions in the SAMe-bound structure (Figs. 1d and 1e). These three residues (Ser247, Glu70 and Asp258) have been reported to interact with methionine and ATP during SAMe  formation (Murray et al., 2016;Shafqat et al., 2013). The activesite residues that interact with the substrates methionine and ATP are contributed by both of the dimeric subunits. The main chain of methionine interacts with Glu70, Gln113 and Asp258, while the positioning of ATP in the active site is stabilized by Asp179, Ser247, Arg249 and Phe250. According to the conformational change of the gating loop upon substrate binding and SAMe production, Gln113, which directly binds to methionine and SAMe and is also a part of the gating loop, shows a movement of 8.1 Å in the present apo structure (Fig. 1e).

The crystal structure of the dimeric MATa1 R264H mutant
We have determined the crystal structure of the MAT1 R264H mutant at 2.85 Å resolution. The crystal belonged to space group C2, with 42% solvent content. The asymmetric unit contains a dimer created by two subunits (chains A and B), with the R264H mutation sites clearly visible in the F o À F c OMIT map (Fig. 2a). Chain B contains three invisible regions, the N-terminus (Met1-Glu15), the gating loop (Asp116-Glu127) and the Phe250-Gly257 region, while chain A contains an ordered gating loop and the Phe250-Gly257 region shows a clear electron-density map. These two regions (Phe250-Gly257 and the gating loop) are involved in stabilizing substrate binding, so they become more flexible in the apo form. Chain A shows an overall B factor of 60.74 Å 2 , which is slightly lower than the overall B factor of chain B (61.72 Å 2 ). Chains A and B align well, with an r.m.s.d. of 0.076 Å , and are almost identical, despite some differences in regional flexibility.
The Arg265His mutant of rMAT and the corresponding R264H mutant of human MAT1 were reported to be unable to dimerize (Chamberlin et al., 1997;Pé rez Mato et al., 2001), whereas the analogous eMAT Arg244His mutant remained as a tetramer (Reczkowski et al., 1998). We have previously reported that the R264H mutant of MAT1 is found in oligomeric states (Panmanee et al., 2019); however, we were unable to produce a good diffraction-quality crystal of this mutant. Thus, we undertook a structural study of the Arg264Ala (R264A) mutant of MAT2 in order to understand the role of Arg264 in MAT catalysis. We determined its structure at 1.7 Å resolution and found that Arg264 is involved in interaction with tripolyphosphate during enzyme catalysis, while the dimer interfaces of the R264A mutant and the wild-type MAT2 structure were in a similar orientation and no significant movement was observed (Panmanee et al., 2019).
In contrast, the crystal structure of the R264H mutant of MAT1 described in the present study shows a huge difference in the dimer-interface orientation owing to the substitution of Arg264 in the central part of the dimer by His. His264 forms stronger hydrogen bonds to Glu57 (Fig. 2b) than those found in the wild type (Arg264), as the shorter side chain of histidine causes the displacement of the residues located at the dimer interface (Fig. 3). At the mutation site, the distance between His264 in each subunit is 7.53 Å in the R264H mutant, which is around 2.06 Å longer compared with the distance between the Arg264 residues in the wild type (5.47 Å ) (Figs. 2c and 3a). The residues that line the dimer interface change their orientation in order to accommodate the mutation, causing the interaction at one part (Thr262-Gly263) of the dimer interface to tighten in comparison to the other part (Arg264-Ile267) (Fig. 3b). The upper part of the mutation site becomes generally tighter (Thr262-Gly263) along the dimer interface, while the mutation site itself loosens. The displacements caused by the R264H mutation compared with wild-type MAT1 are shown in Table 2. The dimer interface was examined using PDBePISA to investigate the additional alterations derived from this mutation (Krissinel & Henrick, 2007). This analysis reveals that the solventaccessible surface area of the R264H assembly is 52 110 Å 2 , which is consistent with the value of 52 100 Å 2 found for the wild type. However, the numbers of residues that are involved in polar interactions at the dimer interface differ. The dimer interface of R264H is associated with 25 hydrogen bonds and nine salt-bridge interactions, while only 18 hydrogen bonds and four salt bridges are observed at the wild-type dimer interface (Supplementary Table S1). In addition, ÁG diss , which corresponds to the free-energy difference between the associated and dissociated states, is 7.1 and 3.1 kcal mol À1 for the R264H mutant and wild-type assemblies, respectively. These alterations at the dimer interface of the R264H mutant assembly indicate that His264 generates amino-acid displacement along the dimer interface.
Pé rez Mato et al. (2001) demonstrated that the mutation of Arg265 in the rat enzyme (equivalent to Arg264 in the human enzyme) to His (R265H) resulted in a monomeric MAT which catalysed only 0.37% of the SAMe production of the wild type. To mimic the dominant inherited form of the R264H mutant that causes hypermethioninemia, wild-type and R265H subunits of the rat enzyme were mixed and assembled as an R265H-wt hetero-oligomer. It was found that the R265H mutant could oligomerize with the wild-type subunit as a heterodimer (R265H-wt MAT; Mato et al., 2001). However, the R265H-wt dimer was unable to synthesize SAMe, suggesting its dominant effect, as was found in patients with IPH. Additionally, the tripolyphosphatase activity was found to be comparable to that of the wild type in the hybrid MAT, but could not be stimulated by methionine and ATP, indicating impairment of substrate binding (Mato et al., 2001). The  Table 2 C -C distances of the residues (Val262-Ile267) in the mutation site of each subunit in comparison to the wild type.  (Fig. 4a). All attempts to obtain a holo structure of the MAT1 R264H mutant failed and produced crystal structures of the apo form, despite the R264H mutant protein being mixed with substrates (methionine and AMP-PNP) prior to crystallization. The methionine substrate concentration giving a halfmaximum of the reaction rate (K m ) of tetrameric MAT1 has been reported to be $23 mM (Cai et al., 1996;Lombardini & Sufrin, 1983). A methionine concentration of 200 mM, which is approximately tenfold greater than the K m of the wild type, was used in our SAMe synthetic assays for the R264H mutant and a loss of SAMe production was still observed, suggesting a lower affinity of the enzyme-substrate complex. The loss of enzymatic activity caused by this mutation may thus originate from its inability to bind the substrates in the first place, resulting from the change in the positions of the residues at the dimer interface, which is further compromised by causing a lower affinity for a tetrameric assembly and a greater propensity for dimeric assembly. The structure of the apo MAT1 R264H mutant. (a) The subunits of the MAT1 R264H dimer are shown as grey and blue ribbons. The dimer is composed of two self-assemblies of the R264H subunit (chains A and B). The Arg264His mutation site is shown in the F o À F c OMIT map contoured at the 3 level (grey). The gating loop (113-131) and residues 250-259 are shown in green and pink, respectively. (b) His264 and Glu57 of each partner subunit form hydrogen bonds of 2.7 and 2.9 Å , aiding stabilization of the dimer. Hydrogen bonds are shown as black dotted lines. (c) The dimer interface of the MAT1 R264H mutant is shown. The substitution of Arg264 by His264 changes the distance between this residue of each subunit to 7.53 Å (C -C measurement), which is about 2.06 Å longer than the corresponding distance in the wild type. The Thr262-Ile266 main chains are shown as grey and blue sticks.
contrast to the wild-type enzyme, which exists primarily as a tetramer (Fig. 4b). The tetrameric form of the wild-type MAT1 structure reveals a dimer-dimer contact surface which involves a few polar interactions, including those between Thr62 (chains A and B) and Asn105 (chains C and D), and between Arg84 (chains A and B) and Glu111 (chains C and D) (Fig. 3d). Also, no disulfide bonds are observed in either the dimer or tetramer subunit interactions. The R264H mutation causes the residues at the dimer interface to be displaced from their original positions in the wild type. Accordingly, the research papers Acta Cryst. Structural comparison of the apo form of wild-type MAT1 and the R264H mutant. Wild-type MAT1 and the MAT1 R264H mutant are shown as pink and blue ribbons, respectively, with alignment of chain A of each protein. (a) The dimer interface is represented by dashed lines; the C -C distances between the Arg264 or His264 residues of each protein are illustrated. C atoms are marked by yellow circles. (b) The distance displacements caused by the R264H mutation are shown by C -C measurements (blue dots) of these residues (Val262-Ile267) in each subunit compared with the wild type. The Thr262-Ile266 main chains are shown as dark pink (wild type) and blue (R264H mutant) sticks. C atoms are marked by yellow circles. (c) The dimerdimer interface of wild-type MAT1 is shown in red (chains C and D). The areas in the dashed and solid boxes are enlarged in (d) and (e), respectively. (d) The polar interactions at the dimer-dimer interface found in the wild type are illustrated for chain A with chain C and for chain B with chain D. The residues of chains A and B are shown as pink sticks (pink labels), while the residues of chains C and D are shown as red sticks (red labels). (e) The distance displacements of Thr62 and Arg84 caused by the R264H mutation are shown compared with the wild type. Thr62 and Arg84 are presented as pink and red sticks for the wild type and as blue sticks for the R264H mutant. The Thr262-Ile266 main chains are shown as dark pink (wild type) and blue (R264H mutant) sticks to mark the mutation regions. overall protein architectures are changed despite the preservation of the secondary structure. In the mutant, Thr62 and Arg84, which participate in dimer-dimer contact interactions, are relocated by 8.6 and 9.8 Å , respectively, compared with the wild-type structure (Fig. 3e), resulting in a reduced tetramer affinity and the ability to form a tetrameric assembly of the mutant enzyme.
3.4. Can the loss of activity of R264H MATa1 be recovered?
MAT1 is a liver-specific enzyme and is found to act as a homotetramer or a homodimer, while MAT2 functions as a hetero-oligomer by forming complexes, MAT(2) 4 (V1) 2 or MAT(2) 4 (V2) 2 , with its regulatory subunits MATV1 or MATV2, respectively (Murray et al., 2014). The heterocomplex structure of MAT2 revealed that two subunits of MAT interact with the tetrameric form of MAT2 at the dimer interface. MAT has two isoforms, MATV1 and MATV2, that differ in the length and the identity of the first 20 amino acids at the N-terminus but exhibit total conservation in the rest of the primary structure (Yang et al., 2008). An in vitro study suggested that MAT1 could also form a heterocomplex with MATV1 and to a lesser extent with MATV2 (Murray et al., 2014). Here, we found that only MATV1 was able to form a hetero-complex with the R264H MAT1 mutant (Figs. 4c and 5a), providing a recovery of the enzymatic activity (SAMe production) of the mutant to the level of wild-type MAT1 (Fig. 5b). We have previously reported a series of quinolone-based compounds that could regulate the activities of MAT enzymes (Panmanee et al., 2019). We tested the effect of the compound SCR0911 on the MAT1 R264H mutant and discovered that it could also recover the activity of the mutant to the same level as wild-type MAT1 (Fig. 5b). However, pre-incubation with SCR0911 did not prevent hetero-complex formation of the R264H mutant and MATV1, suggesting that MATV1 is a competitive cognate partner of the R264H mutant. Also, the addition of SCR0911 to the MAT1 R264H-MATV1 complex did not give a synergistic effect, suggesting that SCR0911 competes with MATV1 for binding at the same dimeric interface and that MATV1 binding prevents SCR0911 from interacting with the mutant protein (Figs. 5a and 5b).
MATV1 was able to alter the activity of MAT2 without changing its catalytic transition state (Firestone & Schramm, 2017;Murray et al., 2014), and the catalytic site of MAT2 was also preserved in the presence of MAT (Murray et al., 2016), indicating that the increase in activity is modulated by allosteric regulation. The restoration of the activity of the MAT1 R264H mutant by the regulatory subunit MATV1 or the compound SCR0911 suggests a similar binding interface. The best binding position of SCR0911 in the R264H mutant was predicted by blind docking using SwissDock (FullFitness score of À3475.04 kcal mol À1 and ÁG of À7.76 kcal mol À1 ). The protein-ligand interaction profiles are shown in Figs. 5(c) and 5(d). The predicted site is in the same dimeric interface pocket where MAT and the compound PF-9366 have been reported to bind (Murray et al., 2014;Quinlan et al., 2017). We also used alternative software (PatchDock) to calculate the best binding position of SCR0911 in the R264H mutant (Duhovny et al., 2002;Schneidman-Duhovny et al., 2003). The protein-ligand interaction site predicted by PatchDock is at a similar position to that predicted by SwissDock (Supplementary Fig. S1). Among the predicted ligand-interacting residues, Arg313 and Tyr335 are known to interact with MAT subunits (Murray et al., 2014). Efforts to obtain a crystallographic structure containting the compound were not successful, suggesting some flexibility, heterogeneity and/or weaker binding compared with MAT, which is consistent with the inability of SCR0911 to compete with MAT in restoring the SAMeproduction activity of the MAT1 R264H mutant.
Neither the tetramer nor the dimer fraction from the gelfiltration column could catalyse the production of SAMe (Fig. 5b). Therefore, we presume that the impaired function of the R264H mutant is caused by distortion of the active site rather than impaired tetramerization. To test this hypothesis, 100 mM SCR0911 was incubated with R264H mutant protein for 2 h at 4 C prior to performing size-exclusion chromatography. The result shows that the compound does not change the dimer-tetramer equilibrium of the R264H mutant (Fig. 4d). In addition to hypermethioninemic conditions, the protonation state of the R264H mutant is changed at physiological pH (about 7.4) owing to the effect of mutation. Arg has a pK a of about 12.48, while His has a pK a of 6.0, and therefore Arg is fully protonated at physiological pH. Under our experimental conditions the enzyme reactions were performed at pH 7.5, so Arg264 of the wild-type enzyme is fully protonated. To test whether the R264H mutant is able to produce SAMe when the His residue is protonated, the catalytic reaction was performed at pH 6.0. We found that the lower pH could not recover SAMe production compared with The effect of regulatory subunits (MAT) and SCR0911 on the activity of MAT1 R264H. (a) Gel-filtration profiles of MAT1 R264H-MATV1 with or without SCR0911 pre-incubation are shown by the absorbance at A 280 . (b) The effects of the regulatory subunits (MATV1/V2) and SCR0911 on the enzymatic activity of MAT1 R264H are shown by measuring SAMe formation. 'aaa', 'bbb' and 'ccc' denote statistical significance at p < 0.001 compared with tetrametric R264H, dimeric R264H (dimer) and both oligomeric forms of R264H, respectively. Data are the mean AE SEM (n = 3). (c) The SCR0911 binding site of the R264H mutant (grey) obtained by molecular docking is aligned with the structure of the complex of MAT2 (yellow) and MAT (pink). SCR0911 is shown as blue sticks. (d) The SCR0911 and R264H interaction profiles obtained from SwissDock are illustrated. Protein-ligand interactions are shown in black for hydrogen bonds, blue for halogen bonds, pink forstacking and orange for electrostatic interactions. the wild-type enzyme (Fig. 4e), suggesting that the loss of catalytic efficiency is likely to be owing to the disordered active site.

Sequence-conservation analysis and insight from the structural analysis of MAT enzymes
Multiple sequence alignments of proteins homologous to MAT from all living organisms were performed using 2963 protein sequences (Fig. 6). The residues involved in MAT catalysis include five acidic amino acids (Asp31, Glu57, Glu70, Asp258 and Asp291), six basic amino acids (His29, Lys181, Arg264, Lys265, Lys285 and Lys289) and one neutral amino acid (Gln113) (Figs. 6 and 7a). All of these catalytic residues are highly conserved throughout evolution (Fig. 6). Three of these catalytic residues (Asp258, Arg264 and Lys289) play a direct role in substrate binding and have been reported to cause IPH when the MAT1A gene encodes mutations at these residues: Asp258Gly, Arg264His/Cys and Lys289Asn (Chamberlin et al., 2000;Chien et al., 2005;Ferná ndez-Irigoyen et al., 2010;Nagao & Oyanagi, 1997). All mutation points of the MAT1A gene causing IPH that have been reported to date are illustrated in Figs. 6 and 7(b). Most of the MAT1A genes in which missense mutations occur that cause IPH encode fully conserved amino-acid residues (Fig. 7a)  Amino-acid conservation of MAT1 showing catalytic residues and mutation sites causing hypermethioninemia. Different colours are used to show amino-acid hydrophobicity (blue, hydrophilic; green, neutral; black, hydrophobic). Catalytic residues are marked by green squares. Residues reported to have mutations and cause hypermethioninemia are marked with yellow squares. Red squares indicate residues that are involved in both catalysis and missense mutations. , 2000Chien et al., 2005;Ferná ndez-Irigoyen et al., 2010;Linnebank et al., 2005;Sen et al., 2019;Tada et al., 2004;Ubagai et al., 1995).
In addition to the Arg264His mutation, we attempted to establish the functional role of Arg299. Although we were successful in producing an Arg299His mutant, the majority of this mutant protein formed aggregates in solution. Arg299 is located in the middle of the solvent-inaccessible -helix domain. The side chain of Arg299 forms hydrogen bonds to Tyr141 of an underlying -sheet domain and also to Glu148, Cys149 and Glu388 of the connecting loop (Fig. 7c), suggesting its importance in protein folding and overall structural stability. Some of the IPH-causing mutations are found in highly conserved amino-acid residues: Leu42Pro, Arg177Trp, Arg199Cys, Arg249Trp, Ile252Thr, Gly257Arg, Ala259Val, Ala297Asp, Gly336Arg, Arg356Trp/Pro/Gln and Pro357Leu (Chamberlin et al., 1996;Chien et al., 2005;Ferná ndez-Irigoyen et al., 2010;Linnebank et al., 2005;Sen et al., 2019;Ubagai et al., 1995;Nashabat et al., 2018;Fig. 5a). We also attempted to obtain purified Arg356Trp mutant, but its purification exhibited problems similar to those faced with the Arg299His mutant. Arg356 is located on the surface of the solvent-accessible -helix domain and forms several hydrogen bonds to Glu128, Asp129 of the gating loop and Asp354 of the neighbouring connecting loop (Fig. 7d), suggesting an important role in the overall protein architecture. Only a few IPHcausing mutations have been found to involve poorly conserved amino acids (Leu305Pro, Glu344Ala and Val361Phe; Fig. 6; Chamberlin et al., 2000;Nashabat et al., 2018;Ubagai et al., 1995). In addition, some of these mutations (Arg356His and Glu344Ala) have been reported to cause increased susceptibility to thoracic aortic aneurysms when  mutations occur in MAT2A genes, suggesting their importance for enzyme stability and function (Guo et al., 2015).
In summary, this study provides a structural basis for the lower activity found for the R264H mutation of MAT1 at the dimer interface. This mutation changes the positions of residues that constitute the dimer interface where the active sites are located, resulting in an inability to bind the substrate and causing a loss in activity. The changes at the dimeric interface also give rise to a lower affinity for a tetrameric assembly and a greater propensity for a dimeric assembly for the R264H MAT1 mutant. Arg264 is also involved in providing enzyme stability by forming a salt bridge with Glu57; its mutation to histidine thus causes lower enzyme stability, as observed by the decreased melting temperature of the mutant. We also show that the activity of the R264H mutant can be restored, presumably by reinstating the active site, when the mutant is incubated with the quinolone compound SCR0911 or with its regulatory subunit MATV1. This observation provides an opportunity for chemical intervention to alleviate this diseasecausing defect in catalysis.