Structural Characterization of the Human Cytosolic Malate Dehydrogenase I.

The first crystal structure of the human cytosolic malate dehydrogenase I (MDH1) is described. Structure determination at a high resolution (1.65 Å) followed production, isolation, and purification of human MDH1 using a bacterial expression system. The structure is a binary complex of MDH1 with only a bound malonate molecule in the substrate binding site. Comparisons of this structure with malate dehydrogenase enzymes from other species confirm that the human enzyme adopts similar secondary, tertiary, and quaternary structures and that the enzyme retains a similar conformation even when nicotinamide adenine dinucleotide (NAD+) is not bound. A comparison to the highly homologous porcine (sus scrofa) MDH1 ternary structures leads to the conclusion that only small conformational differences are needed to accommodate binding by NAD+ or other NAD+ mimetics. Conformational differences observed in the second subunit show that the NAD+ binding elements are nevertheless quite flexible. Comparison of hMDH1 to the human mitochondrial malate dehydrogenase (hMDH2) reveals some key differences in the α7-α8 loop, which lies directly beneath the substrate binding pocket. These differences might be exploited in the structure-assisted design of selective small molecule inhibitors of hMDH1, an emerging target for the development of anticancer therapeutics.

Introduction

Malate dehydrogenases (MDH) belong to the family of nucleotide-binding proteins referred to as nicotinamide adenine dinucleotide (NAD)-dependent dehydrogenases or oxidoreductases.[1] This enzyme family includes the lactate dehydrogenases (LDHs), the liver alcohol dehydrogenases (LADHs), and the glyceraldehyde-3-phosphate dehydrogenases (GADPHs), among others.[1] Malate dehydrogenase reversibly converts malate to oxaloacetate with the use of NAD+/NADH as a cofactor in the tricarboxylic acid cycle.[2] Most cells contain two main isoforms, which differ in their cellular compartmentalization and role in cellular processes: in eukaryotes, malate dehydrogenase II (MDH2) is found in the mitochondrial matrix where it is involved in the citric acid cycle, while malate dehydrogenase I (MDH1) is localized in the cytosol where it is important to the malate/aspartate shuttle of the urea cycle.[2]

Malate dehydrogenase I is overexpressed in a variety of cancers, and MDH1 amplification in human tumors is a common aberration that correlates with poor prognosis.[3] A hallmark of cancer cells is the increased glucose consumption required for the production of macromolecules necessary for growth and division.[4] Cytosolic NAD levels are independent of mitochondrial NAD levels involved in the electron transport chain.[5] Increased cytosolic concentrations of NAD are necessary to maintain the enhanced glycolysis of proliferating cancer cells, which has largely been attributed to the production of lactate through LDH activity.[4,5] Recently, it has been shown through the use of glucose isotopomer tracing in N5 cells that MDH1 supports LDH in the replenishing of cytosolic NAD.[6] The same study also showed that Jurkat cells with MDH1 knocked out (MDH1 KO.1 and MDH1 KO.2) show slower proliferation and glucose consumption than cells with functional MDH1.[6] This observation leads to the possibility that an MDH1-selective inhibitor, used either alone or in combination with LDH inhibitors, might slow tumor growth and cancer progression in patients.

Structural studies employing crystallography have been used to study malate dehydrogenase enzymes. Structures are known for MDH2 enzymes from Escherichia coli,(7−9) plants,[10] and mammals.[11] These MDH2 enzymes all share high sequence homology to human MDH2 (55–95%) but are distinct from MDH1 enzymes that share lower homology (25–30%). Human MDH1 and MDH2 share only 26% sequence identity. Cytosolic MDH1 enzyme structures have been investigated from several species including bacteria[12,13] and plants.[14] These enzymes share good sequence homology with the human MDH1 (50–62% identity). Collectively, this work has confirmed that the structures are quite homologous across all species. All MDH enzymes share a common Rossmann fold motif characteristic of other NAD(P) binding dehydrogenases,[15] a common dimeric quaternary structure, and highly similar NAD+/NADH binding sites and mechanisms of catalysis.

Banaszak’s group at the University of Minnesota completed some of the first high-resolution crystallographic studies with porcine (sus scrofa) MDH1 over 25 years ago.[16,17] That enzyme shares a very high (95%) sequence identity with the human enzyme, but to date, no human MDH1 structure has been reported. Given the renewed interest in hMDH1 as a possible therapeutic target, we have sought to obtain a crystal structure of the human enzyme to enable the direct structure-aided design of an hMDH1 inhibitor. Here, we report the first structure of the cytosolic human malate dehydrogenase I and compare it to previously determined structures of other homologous cytosolic MDHs and also to the structure of the human mitochondrial MDH (hMDH2). The monoclinic crystal form with malonate but no NAD+/NADH bound provides a unique view of this emerging target for pharmaceutical development.

Results and Discussion

Expression, Purification, Activity, and Crystallization of hMDH1

A modified pGS-21a plasmid was engineered with a tobacco etch virus (TEV) cleavage site, and the human gene mdhI was codon-optimized mdh1 for bacterial expression and cloned into the modified plasmid. The rationale behind the engineered plasmid was to use codon optimization and a solubility tag glutathione S-transferase (GST) to increase the likelihood of soluble protein after lysis. A TEV cleavage site was included for easy cleavage of the solubility tag from the isolated protein. The combination of the solubility tag and codon optimization yielded soluble protein after cell lysis that was found to bind to the nickel column. TEV protease was able to cleave the GST-tag from hMDH1, which could then be further purified through a second nickel column. (GST and His-tagged TEV protease adhered to the column, while cleaved hMDH1 did not.) Finally, size exclusion chromatography was employed to ensure the highest possible protein purity for crystallographic and enzymatic studies.

Following isolation of the purified MDH1, the protein was tested in a spectrometric activity assay quantifying the conversion of NADH to NAD+. By varying enzyme concentration and using oxaloacetate as a substrate, the Vmax observed is 0.0012 mM/sec, and the Km observed is 0.00041 mM (Supporting Figure S2). Controls affirm that the oxidation of NADH to NAD+ was a direct result of hMDH1 activity. The oxidation of NADH in the absence of protein was not observed, showing that the conversion of NADH to NAD+ was a direct result of hMDH1. The enzyme is considerably more active than previously reported for sus scrofa MDH1 (Km = 0.036 mM).[29]

Following the confirmation of enzymatic activity, MDH1 crystallography was attempted using previously established conditions for ssMDH1 due to the high sequence similarity. Microcrystalline material grew after several days, but to generate more robust and larger crystals, alternate buffer conditions were explored. Previous researchers identified malonate (pH range 4–8) as a good buffer for the crystallization of LDHA, another member of the oxidoreductase family, so malonate was substituted for N-(2-hydroxyethyl)piperazine-N′-ethanesulfonic acid (HEPES) in crystal optimization.[19,20] Following several rounds of seeding, diffraction quality crystals were successfully generated.

Secondary, Tertiary, and Quaternary Structures of hMDH1

The hMDH1 crystal was found to have P21 symmetry with two protein molecules (the functional biological unit) in the asymmetric unit, and the structure has been solved and refined to a 1.65 Å resolution (Supporting Table S1). Each monomer adopts the Rossmann fold characteristic of other NAD(P) binding dehydrogenases,[15] composed of 9 α-helices and 11 β-strands conserved in all MDH structures (Figure A). A lengthy sequence insertion between β8 and α7 of MDH1 enzymes gives rise to an additional β pair (β8a–β8b) curled into the β8−α7 loop not present in MDH2 enzymes. This insertion is revealed in the comparison of hMDH1 and hMDH2 sequences in Figure .

Figure 1

hMDH1 monomer fold and secondary structure. (A) The monomer structure illustrated with gradient coloring to show the progression of the chain from N (blue) to C (red) terminus. Consensus secondary structure elements are identified. The bound malonate is within the CPK-colored surface. (B) An alternate view that focuses on elements of structure surrounding the predicted NAD+ binding site (shown with the transparent surface for positional reference only). Loops prominently referred to in the Results and Discussion section are identified. (C) A structure-based alignment of human MDH1 and MDH2. Secondary structure assignments in hMDH1 are identified across the top. Vertical lines joining amino acids in the two sequences denote a 1:1 correspondence in the position of residues in the structures of the two enzymes. Boxes identify cofactor (green) and ligand (orange) binding motifs. Sequence conservation marks displayed (*,:,.) are from a Clustal-Omega multiple sequence alignment of MDHs with known structures in the PDB.

The site of NAD+/NADH binding in this family of structures lies at the meeting of loops at the edge of the large parallel β sheet within this fold. Loops β1−α1 and β2−α2 support the adenosine diphosphate, β5−α5 cradles the nicotinamide nucleoside, and β4−α4 provides a sort of cap that lays over the cofactor. Substrates (malate or oxaloacetate) bind just to the side of the nicotinamide base, pinched between α helices α7 and α8, held in precise position by H-bonds to universally conserved residues (Arg92, Arg98, Asn131, and His187).

Monomers assemble into biologically relevant homodimers, forming a helical bundle with α1 and α8 replicated across a noncrystallographic twofold axis and winged with extensive contacts between α2 and α7 of opposite monomers (Supporting Figure S3A). The details of this monomer–monomer interface have been exhaustively described previously for the porcine MDH1 structure.[16] There is no direct contact between active sites in the two monomers, but previous work has shown that MDH1 functions as a dimer in solution and disruption of this interaction along the dimer interface could potentially perturb the enzymatic function.[2] While the two monomers are in distinctly different crystallographic environments in our structure, they are quite similar overall (rmsd 0.192 Å). Crystal packing does influence conformational flexibility in this crystal form. The crystallographic B-factors of α-carbons throughout the structure are illustrated in Supporting Figure S3B. Large B-factors are observed for residues 203–205 in subunit A. Amino acids in this region are modeled with lower occupancy (0.5) as positive difference peaks were observed in the 2Fo–Fc map but little electron density was observed in the Fo–Fc map. The majority of the important β4−α4 loop (residues 92–99) is disordered in subunit B and cannot be modeled; there is no interpretable electron density for this loop. The rest of the structure is well ordered, with 98% of the amino acids showing favorable torsional angles according to Ramachandran analysis with no outliers.

NAD+/NADH Binding Site

The structure reported here has neither NAD+/NADH nor a substrate bound. We have tried to form binary complexes with NADH and ternary complexes with NADH and oxaloacetate by either soaking or cocrystallization under similar conditions, but these efforts have not yet been successful. The ternary complex of Sus scrofa (porcine) MDH1 provides a good basis for predicting where cofactor and substrate might be when bound together in hMDH1 and for identifying conformational differences in the hMDH1 crystal form that might prevent cofactor binding. To create a predictive model for NAD+ binding, the highly conserved structural motifs in hMDH1 associated with NAD+ binding in all other MDHs (those amino acids identified within green and orange boxes in the sequence alignment of Figure C) were superimposed onto the analogous residues of the ssMDH1 ternary complex structure (PDBid: 5MDH), resulting in an rms difference in backbone atoms of only 0.24 Å. The tNAD and a substrate mimetic (α-ketomalonate) positions were then adopted for inclusion in a hMDH1 ternary complex (Figure ).

Figure 2

Comparison of NAD+ binding in the porcine MDH1 complex (PDBId 5MDH) and hMDH1 homology model. (A) Prominent H-bonds anchor the tetrahydroNAD (orange) in the ssMDH1 structure. Substrate mimetic α-ketomalonate (white) binds beside the nicotinamide in this ternary complex. Interactions between the adenosine ribose and Asp41 are not shown. (B) Predicted NAD+ binding in hMDH1. All of the same H-bonding should be possible in hMDH1 subunit A with no conformational change. (C) Comparison of the cofactor binding pocket of hMDH1 subunit A (green) and hMDH1 subunit B (magenta).

In the ssMDH1 complex (Figure A), hydrogen bonds from the backbone of the β1−α1 loop including Gly13 and Gly14 anchor the adenosine ribonucleotide, side chains of Ser89 and Asn131 engage the nicotinamide ribose, and His181 is H-bonded to the nicotinamide amide. The β2−α2 loop closes the adenosine side of the NAD binding pocket, with H-bonds from Asp41 to the adenylate ribose (not shown). The hMDH1 structure is poised to accept NAD+/NADH in this same position with minimal conformational adaptation (Figure B). All interactions should be preserved except the H-bond to Ser88, which is flipped to a different rotamer orientation in hMDH1 when NAD+ is absent.

The binding elements in subunit B are not well positioned for cofactor binding, however (Figure C). As mentioned above, residues 92–98 are completely disordered. While this loop makes little direct contact with NAD+, it forms the floor of the substrate binding pocket beneath the nicotinamide ring, and substrates typically H-bond to the cofactor. The Gly14–Gln15 amide bond in subunit B is rotated roughly 90 degrees, so that the carbonyl is directed away from the phosphate of NAD+ to which it should H-bond. Asp42–Met45 and Gly88–Pro91 have all shifted into the NAD+ binding space, effectively shrinking the binding pocket.

We have examined the crystal packing in the vicinity of NADH binding sites in both subunits, but there is no simple explanation for these conformational differences; no intermolecular interactions exist to prevent subunit B from adopting a conformation similar to that seen in subunit A or vice versa. It is interesting that structural flexibility exists, as it affirms the possibility that small molecules might be identified that bind and stabilize conformations of hMDH1 that cannot support cofactor binding.

Malonate Lies in Substrate Binding Pocket

Electron density present in the substrate binding pocket of subunit A can be attributed to malonate (Supporting Figure S2), a component of the buffer for crystallization. No comparable density is found in subunit B, which is less well ordered. Arg92 and Arg98, which figure prominently in positioning substrate molecules, are completely disordered in subunit B as discussed above.

Malonate has been modeled in the hMDH1 complex in two different conformations with half occupancy (conformer A and conformer B). The hydrogen bonding stabilizing each conformation is shown in Figure A. In both conformations, one carboxylate in this symmetric molecule is positioned opposite the plane of the Arg92 guanidinium to which it is H-bonded. In conformation A, the other carboxylate interacts with Arg162, Arg98, and Ser242. In conformation B, this carboxylate is rotated to hydrogen bond with His187 and the secondary amine in Arg98. These interactions collectively mimic those observed with more natural substrates and substrate analogues that have additional hydrogen bond acceptors that can make both sets of interactions, such as those observed in the porcine MDH1 complex with α-ketomalonate (PDBid: 5MDH) (Figure B). Interestingly, malonate binds in this pocket even in the absence of the NAD+ cofactor. In the α-ketomalonate complex, one oxygen of the carboxylate is able to hydrogen bond with an exocyclic oxygen of the nicotinamide ribose. We could find no other example of a malate dehydrogenase structure in the PDB that includes a bound substrate analogue in the absence of NAD+/NADH.

Figure 3

Comparison of ligand binding. (A) Hydrogen bonding to malonate by hMDH1. (B) Hydrogen bonding to α-ketomalonate by ssMDH1 (PDBId: 5MDH).

Differences between Human Cytosolic and Mitochondrial MDH

Any successful targeting hMDH1 for therapeutic purposes will require that some degree of selectivity be achieved over the mitochondrial MDH (hMDH2). While it has never been described in detail, a high-resolution (1.9 Å) structure of a ternary complex of hMDH2 was solved in 2006 by the Structural Biology Consortium and deposited in the PDB with accession code 2DFD. Its existence affords an opportunity for a detailed comparison to the hMDH1 enzyme. The structure-based alignment of residues with PDBeFold[18] results in the pairwise alignment of sequences illustrated in Figure C and an rmsd of paired backbone atoms of 2.0 Å. However, the superposition of only the NAD+ binding substructures identified in Figure B shows that those substructures are conserved with higher homology (rmsd 0.79Å) (Figure ). While there are some specific sequence differences including a nonconservative Glu94 to Pro substitutions in the β4−α4 loop, and the deletion of several residues from the β2−α2 loop in hMDH1, all of the important hydrogen bond donors and acceptors that anchor NAD+ in hMDH2 are in a position to do so in the same way in hMDH1 (Figure A). Amino acids Asp41, Asn131, Arg192, and Arg98 and the β1−α1 loop can all be expected to make comparable H-bonds to secure cofactor binding. Both enzymes also have a histidine (His182/187) positioned to interact with the bound malate.

Figure 4

Comparison of hMDH1 to hMDH2. hMDH2, NAD+, and d-malate are from the ternary hMDH2 complex (PDBid: 2dfd). (A) Conservation of structural features needed for cofactor and substrate binding. hMDH2 structural features involved in the binding of NAD+ (white with surface) include H-bonds to Asp39 of loop β2−α2, Asn124 of loop β5−α5, and extensive contacts with loop β4−α4. d-Malate (also white) of the hMDH2 complex is held by multiple H-bonds to Asn124, His182, and Arg186 and Arg192 of loop β5−α5. All of these side chains are conserved and occupy comparable positions in the hMDH1 structure, despite the overall low sequence homology and the absence of bound NAD+. Malonate bound to hMDH1 is not shown for clarity. (B) Comparison of the α7−α8 loop in hMDH1 (cyan) vs hMDH2 (salmon). The portion of this loop that should contact the bound malate is illustrated with the fragment of surface (gray).

One feature stands out as potentially relevant to prospects for selective inhibitor design in the comparison of hMDH1 and hMDH2: the insertion of two extra residues into the α7−α8 loop. This loop passes directly under the substrate binding pocket (d-malate in the hMDH2 ternary complex) and contributes the surface that underpins the bound substrate (Figure B). In addition to the insertion of two residues that results in a local shift in the registry, there are specific sequence differences (Ile235 to Val; Ser242 to Ala; A243 to Thr) that alter the shape of the substrate binding pocket. There are also likely significant differences in the flexibility and dynamics of these two loop variants in response to ligand binding. While the malate binding pocket is small, these differences may afford an opportunity for selective small molecule inhibitor design.

Conclusions

The first overexpression, isolation, and purification of human MDH1 using a bacterial expression system have been described in detail. Further, the first hMDH1 crystal structure has been determined at a 1.65 Å resolution, with only a small molecule (malonate) bound in the active site. While the structure confirms that the enzyme is very similar to the previously reported porcine structure with which it shares a 95% sequence identity, this new structure without the bound NAD+ cofactor provides some novel insights into the conformational flexibility of the enzyme. The β4−α4 loop that cradles NAD+/NADH in active enzymes adopts a conformation similar to that needed to bind NAD+ in one crystallographic environment but is largely disordered in another. While crystallization can exaggerate the importance of conformations stabilized by crystal packing, it is possible that this range of motion is accessible to the protein in solution as well. A more open conformation suggestive of extensive protein flexibility has previously been observed in an E. coli MDH2 apo structure, (3HHP[8]), but not in any MDH1 structure. This flexibility may leave the enzyme susceptible to inhibition by small molecule dinucleotide mimetics that trap the enzyme in an inactive conformation.

The comparison of the hMDH1 structure to hMDH2 is also revealing. Distinct differences in these two enzymes in the vicinity of the substrate binding pocket might be exploited in the discovery of small molecules that inhibit human MDH1 with selectivity over the mitochondrial MDH2. This selectivity will almost certainly be a desirable attribute of any agent put forward for the clinical evaluation of the therapeutic potential of MDH1 inhibitors in the treatment of cancer. The monoclinic hMDH1 crystal form reported here may be a particularly useful tool for use in future crystallographic fragment screening, specifically because of its extensively open and empty active site.

Experimental Details

Reagents

Full-length hMDH1 was a gift of Dr. Ameeta Kelekar (University of Minnesota, Department of Immunology, Minneapolis, MN). A pGS-21a plasmid containing full-length mdh1 was purchased from GenScript. All enzymes used for DNA digestion were purchased from NEB (Ipswich, MA). All DNA purification kits were purchased from Qiagen (Venlo, Netherlands). Components necessary for protein production other than IPTG were ordered from VWR (Radnor, PA). HisTrap HP column for protein purification was purchased from formerly GE Life Sciences, now Cytiva (Marlborough, MA). All other materials necessary for protein purification and isolation, crystallization, and determination of enzymatic activity were purchased from Fischer Scientific (Waltham, MA).

Cloning, Expression, and Purification of Soluble His6-GST-hMDH1 Fusion Protein

The amino acid sequence corresponding to the full-length hMDH11–334 was codon-optimized and cloned into a pGS-21a plasmid with additional TEV protease cleavage sequence using bgIII and restriction sites. The purchased plasmid was transformed into competent Rosetta2-pLysS (BL21 DE3) cells and cultured on agar plates with Amp100 and Cm30. Colonies grew overnight, and a single colony was selected and shaken at 270 rpm overnight in LB media at 37 °C. One liter of LB media was inoculated with 3 mL of overnight colony and cultured until OD600 reached 0.5–0.7. The culture was cooled down for 1 h at 4 °C prior to being induced with 1 mM IPTG. Cultures were placed in a 20 °C incubator and allowed to shake for 16 h. Cells were harvested by centrifugation at 5000g for 15 min and stored at −20 °C overnight. The cell pellet was solubilized in buffer A (25 mM HEPES pH 7.2, 150 mM NaCl, 10 mM imidazole, and 5% glycerol). To this solution were added lysozyme (final concentration: 1 mg/mL) and 1.5 μL benzonase prior to sonication for 16 min (30 s on, 30 s off) at 30% attenuation. Lysed cells were distributed among 50 mL Beckman centrifuge tubes and centrifuged for 45 min at 45000g. The resulting supernatant was syringe-filtered (0.45 μm) and loaded onto a 5 mL HisTrap HP column. The fusion protein was eluted from the column using a linear gradient of buffer B (25 mM HEPES pH 7.2, 150 mM NaCl, 400 mM imidazole, and 5% glycerol). Like fractions were pooled together and TEV protease (8% w/w) was added before adding the protein solution to a dialysis cassette in TEV cleavage buffer (25 mM HEPES pH 7.2, 150 mM NaCl, 1 mM DTT, and 5% glycerol) at 4 °C overnight. Contents of dialysis tubing were syringe-filtered (0.45 μm) and loaded onto a HisTrap HP column to separate hMDH1 from the His-tagged fusion protein. Flow-through-containing hMDH1 was pooled together and concentrated to ≤5 mL prior to being syringe-filtered (0.22 μm) and loaded onto a Sephacryl S-100 column. Peaks from the size exclusion column were analyzed via SDS-PAGE, and the resulting hMDH1 was concentrated to 5 mg/mL in buffer supplemented with 10% glycerol. Aliquots were flash-frozen before being stored in a −80 °C freezer for future use. The resulting yield from the preparation was approximately 2.5 mg/L. The sequence was confirmed by crystallography and the function was confirmed via enzymatic activity assay.

Assessment of Enzyme Activity

Recombinant hMDH1 enzymatic activity was assessed by monitoring the time-dependent conversion of substrate oxaloacetate to malate. The parallel oxidation of NADH to NAD+ was followed using an Agilent Technologies Cary Series ultraviolet–visible (UV–vis) spectrophotometer at 340 nm. NADH (100 μM) and oxaloacetic acid (240 μM) were added to cuvettes, mixed briefly, and the absorbance at 340 nm was measured at 10 s intervals over 10 min. The effect of enzyme on NADH conversion was determined by the addition of hMDH1 in a twofold dilution series (25–0.390625 nM) to prereferenced vials that were then monitored over the same time course. No change in absorbance was observed in a substrate-free control (25 nM NADH and hMDH1). Absorbance values were then converted to the concentration of NADH using Beer’s law and the extinction coefficient for NADH (6.2 mM/cm). The initial linear portion of concentration versus time curves was entered in GraphPad Prism to determine the Vmax and Km values (Supporting Figure S2). Experiments were conducted in triplicate.

Crystallization

Protocols successfully used with Sus scrofa MDH1 served as a starting point for hMDH1 crystallization.[17] Microcrystals grew in 3–4 days by hanging drop vapor diffusion using a precipitant consisting of 25–30% PEG 4000, 100 mM citrate buffer pH 6.5, and protein-concentrated to 5 mg/mL in buffer containing 25 mM HEPES pH 7.4, 150 mM NaCl, and 10% glycerol. After several rounds of seeding and optimization, plate-like crystals (75 × 25 × 150 μm3) reached full maturity within 48 h at 20 °C. Final conditions yielding monoclinic crystals were obtained by mixing equal amounts of protein with well solution containing 28% PEG 4000, 100 mM NaMalonate pH 7.2, and 0.15 mM ammonium acetate in 2 μL drops. Optimal cryo-conditions involved mother liquor plus 15% PEG 400.

X-ray Data Collection

Diffraction data was collected at IMCA-CAT beamline 17-ID at the Advanced Photon Source (APS), Argonne, Illinois. Data collection was completed at 100 K using a radiation of wavelength 1.00 Angstroms and a Dectris Eiger2 9M detector. Data was processed using autoProc and rescaled using aP_scale using R-factor (<0.4), completeness (>90%), and I/sigma (>2) as criteria, and a minimum of two of the three criteria were met in determining the proper resolution range.[21]

Structure Solution and Refinement

Due to the high sequence similarity between Sus scrofa MDH1 and hMDH1(95%), the ssMDH1 model, 5MDH (chain A) was used as the search model in molecular replacement with Phaser.[17,22] The monoclinic crystal was found to contain two protein chains in the asymmetric unit. Iterative rounds of refinement and model building were carried out using Phenix(23) and Coot.[24] Malonate was modeled into the corresponding electron density using standard geometry as in the CCP4 dictionary.[25] Refined structures were validated with MolProbity.[26]Supporting Table S1 summarizes statistics from the data collection and Phenix refinement for the assessment of the quality of the structures.[27] Molecular figures were made using the PyMOL Molecular Graphics System, Version 2.3 Schrödinger, LLC. Atomic coordinates and reflection data for hMDH1 with malonate bound in the substrate binding pocket have been deposited into the Protein Data Bank[28] (accession code: 7RM9).