Geometric Restraint Drives On- and Off-pathway Catalysis by the Escherichia coli Menaquinol:Fumarate Reductase

Abstract

Complex II superfamily members catalyze the kinetically difficult interconversion of succinate and fumarate. Due to the relative simplicity of complex II substrates and their similarity to other biologically abundant small molecules, substrate specificity presents a challenge in this system. In order to identify determinants for on-pathway catalysis, off-pathway catalysis, and enzyme inhibition, crystal structures of Escherichia coli menaquinol:fumarate reductase (QFR), a complex II superfamily member, were determined bound to the substrate, fumarate, and the inhibitors oxaloacetate, glutarate, and 3-nitropropionate. Optical difference spectroscopy and computational modeling support a model where QFR twists the dicarboxylate, activating it for catalysis. Orientation of the C2–C3 double bond of activated fumarate parallel to the C(4a)–N5 bond of FAD allows orbital overlap between the substrate and the cofactor, priming the substrate for nucleophilic attack. Off-pathway catalysis, such as the conversion of malate to oxaloacetate or the activation of the toxin 3-nitropropionate may occur when inhibitors bind with a similarly activated bond in the same position. Conversely, inhibitors that do not orient an activatable bond in this manner, such as glutarate and citrate, are excluded from catalysis and act as inhibitors of substrate binding. These results support a model where electronic interactions via geometric constraint and orbital steering underlie catalysis by QFR.

Introduction

Complex II superfamily enzymes provide a crucial link between oxidoreduction reactions in the membrane bilayer and in the soluble milieu (1). During aerobic respiration in Escherichia coli, the complex II enzyme succinate:ubiquinone oxidoreductase oxidizes succinate and reduces ubiquinone. In bacterial anaerobic respiration with fumarate as the terminal electron acceptor, the reaction proceeds in the opposite direction, and the complex II homolog menaquinol:fumarate oxidoreductase (QFR)⁴ oxidizes menaquinol and transfers the electrons to fumarate. Complex II superfamily members contain either three or four polypeptide chains (Fig. 1A): two soluble subunits (the flavoprotein, FrdA, and the iron protein, FrdB, in E. coli QFR) and either one or two integral membrane subunits (FrdC and FrdD in the E. coli QFR). Although there are significant differences in the integral membrane subunits across the family, complex II enzymes all share a high percentage of sequence identity in the soluble subunits, including the flavoprotein, where the kinetically challenging oxidoreduction of fumarate and succinate takes place (1).

FIGURE 1.

Structure of the E. coli QFR and relevant ligands. A, structure of the E. coli QFR heterotetramer with the flavoprotein subunit (FrdA) colored gray, iron-sulfur protein (FrdB) colored orange, and the transmembrane subunits colored blue (FrdC) and green (FrdD). B, structures of the substrate, fumarate, the product, succinate, and molecules that interact with the dicarboxylate binding site and are discussed in this work.

Numerous molecules, including metabolic intermediates and ingested toxins, structurally resemble succinate and fumarate. Some of these are excluded from the active site, whereas others act as inhibitors (Fig. 1B). Several of these come from normal respiratory processes. For example, oxaloacetate is a Krebs cycle intermediate and acts as a tight, slow binding inhibitor of complex II, possibly as a regulatory mechanism (2, 3). It has additionally been proposed that oxaloacetate can be formed through off-pathway catalysis by complex II when fumarate hydration to malate is followed by oxidation to oxaloacetate (3, 4). Oxaloacetate binding to the complex II active site results in a charge transfer complex between the oxaloacetate carbonyl and the FAD hydride that can be monitored spectroscopically (4, 5).

Along with physiological regulation from metabolites, complex II activity may be affected by overproduction of dicarboxylates that result from dysfunction of cellular catabolic pathways. For example, mutation of glutaryl-CoA decarboxylase can lead to elevated levels of the dicarboxylates glutarate and 3-hydroxyglutarate (Fig. 1B), and this has been associated with glutaric acidemia type I, which has symptoms that include cerebral swelling and motor defects (6). Respiratory chain deficiency correlates with glutaric acidemia type I, and complex II inhibition is observed in cell culture upon application of glutarate (7).

Ingested toxins offer a third source of complex II inhibitors, often with severe health consequences. The fungus Arthrinium sphaerospermum produces the secondary metabolite 3-nitropropionate (3-NP), an irreversible complex II inhibitor (Fig. 1B) (8–10). When ingested, 3-NP can induce striatal brain lesions similar to those observed in patients with Huntington disease, an autosomal dominant neurodegenerative disease characterized by dyskinesia, cognitive defects, and psychological pathologies (11). In fact, 3-NP is commonly used to recapitulate a subset of the pathologies of Huntington disease in animal models (11). The mechanism of 3-NP inhibition of complex II is under debate. Previously, 3-NP has been proposed to form a covalent adduct with either FAD (8), an active site cysteine residue (12), or a catalytic arginine near the dicarboxylate binding site (13). The formation of any of these adducts probably requires a single enzyme turnover but has not been determined conclusively (12). 3-NP has also been proposed to act as a non-covalent inhibitor of complex II enzymes (14).

To study the details of complex II catalysis and inhibition at the flavoprotein active site, the E. coli QFR was co-crystallized with the substrate, fumarate, and the inhibitors, oxaloacetate, glutarate, and 3-NP. Mass spectrometry and optical spectroscopy allowed unambiguous confirmation of the covalent 3-NP adduct and the proposal of a possible reaction mechanism. The implications for fumarate turnover and the mechanisms of inhibition are discussed.

EXPERIMENTAL PROCEDURES

QFR Purification

E. coli QFR was produced in E. coli strain DW35 (ΔfrdABCD, sdhC:kan) containing the pH3 (frdA⁺B⁺C⁺D⁺) plasmid. Cells were grown under microaerophilic conditions in Terrific Broth medium as described previously (15). Membranes were isolated as described (15) and were resuspended in a solution of 20 mm Tris, 0.1 mm EDTA (pH 7.4) and solubilized with C₁₂E₉ detergent (Anatrace, Maumee, OH) added to a final concentration of 2% (w/v). Purification of QFR was performed in the presence of 0.05% (w/v) C₁₂E₉ as described (15), using Q-Sepharose anion exchange (GE Healthcare), POROS anion exchange (Applied Biosystems), and size exclusion chromatography on a Superdex S-200 column (GE Healthcare).

Enzyme Assays

In order to measure full succino-quinone reductase or succino-oxidase activity of QFR, it is necessary to remove any tightly bound oxaloacetate, which co-purifies with the as-isolated enzyme (3). To activate the enzyme, QFR was diluted to 5 mg of protein/ml in 30 mm Bistris propane, pH 7.0, 0.1 mm EDTA, 0.05% C₁₂E₉, and 3 mm malonate and incubated for 20 min at 30 °C. The enzyme was then concentrated with a Centricon YM30 filtration apparatus, and the concentrated protein was passed through a PD-10 gel filtration column to remove malonate. Activated enzyme was then stored at 4 °C until used. The standard assay medium at 30 °C contained 50 mm Bistris propane, 0.1 mm EDTA, 0.006% C₁₂E₉, pH 8.0. Succino-oxidase activity of QFR was measured with the electron acceptor potassium ferricyanide as described previously (16). In order to determine apparent affinity constants for the dicarboxylates used in this study, inhibition of the ferricyanide reductase activity was determined as previously described (17).

Optical Difference Spectra

Optical properties of all of the bound ligands were measured with an Agilent 8453 diode array UV-visible spectrophotometer 1 min after the addition to isolated enzyme in 30 mm Bistris propane, 0.1 mm EDTA, 0.01% Anapoe C₁₂E₉. Optical spectra were recorded at 25 °C in 50 mm HEPES, 100 mm NaCl, 0.01% Thesit using 1.25 mg/ml isolated and activated E. coli QFR. Upon the addition of the ligand, a spectrum was recorded, and the spectrum of oxidized enzyme was subtracted. Each spectrum represents the addition of the different ligands at the concentration of, 5 mm fumarate, 50 μm oxaloacetate, 4 mm malonate 12 mm glutarate, 50 mm citrate, and 0.1 mm 3-NP, which was added from an alkaline solution. The spectra were recorded 10 min after the addition of the ligand. Inhibition of the enzyme by 3-NP was determined as described by adding a final concentration of 0.2 mm 3-NP from a pH 10.0 solution to activated QFR (pH 8.0) and measuring kinetic and optical properties at pH 8.0.

Mass Spectrometry of 3-Nitropropionate-incubated QFR

QFR at (10 mg/ml) in 20 mm glycine, pH 10.0, 0.1 mm EDTA, and 0.05% C₁₂E₉ was incubated with 1 mm 3-NP for 1 h on ice in a buffer consisting of 20 mm glycine, pH 10.0, and 0.05% (w/v) C₁₂E₉ detergent and was incubated on ice for 1 h. The QFR subunits were separated on a NuPAGE SDS gel (Invitrogen). The 66-kDa FrdA subunit was manually excised and digested with trypsin for 2 h at 37 °C. The resulting peptide mixture was separated with a microcapillary HPLC system (Eksigent 1D Plus with an AS1 autosampler) using an 11 cm × 100-μm C18 reversed phase column (Jupiter C18, 5 μm; Phenomonex) packed directly into a nanospray emitter tip. Using a nanospray source, this was interfaced with either a nominal mass resolution LTQ or high resolution LTQ orbitrap (Thermo Fisher) mass spectrometer, where data-dependent tandem (MS/MS) and MSⁿ spectra were collected throughout a 90-min separation. These spectra were searched against an E. coli protein data base considering potential amino acid mass differentials corresponding to 3-NP adducts using SEQUEST (Thermo Electron) (18). Subsequent injections targeting potentially modified peptides were also performed; this included the targeting of normal and stable isotope-labeled 3-NP adducts, using the LTQ orbitrap. Later, it was determined that adduct formation could occur at physiological pH. As a result, the analysis of ¹⁵N-labeled 3-NP adduct was performed with a modified preincubation procedure, where 10 mg/ml QFR was incubated with 1 mm ¹⁵N-labeled 3-NP in a buffer consisting of 20 mm Tris, pH 7.4, 0.1 mm EDTA, and 0.05% C₁₂E₉.

Synthesis of Isotope-labeled 3-NP Derivatives

3-Bromopropionic acid (250 mg, 1.63 mmol), Na¹⁵NO₂ (206 mg, 2.94 mmol, 98% ¹⁵N), phloroglucinol (227 mg, 1.80 mmol), and DMF (3.3 ml) were added to a flame-dried round bottom flask. The reaction mixture was stirred at room temperature for 22 h and then poured onto ice water and extracted with diethyl ether. The combined organic layers were dried over MgSO₄, filtered, and concentrated. A portion of the residue was sublimed (80 °C, 1 Torr) to yield 16 mg of the yellow crystalline product. Label incorporation, sample purity, and confirmation of structure were determined by NMR: ¹H NMR (400 MHz, CDCl₃) δ 4.66 (td, J = 6.0, 2.4 Hz, 2H), 3.06 (td, J = 6.4, 4.0 Hz, 2H); ¹³C NMR (100 MHz, CDCl₃) ppm 174.3, 69.3, 30.6; ¹⁵N NMR (60 MHz, CDCl₃) ppm 379.5.

Crystallization of the E. coli QFR with Ligands

All crystallizations of QFR used the hanging drop vapor diffusion method at 22 °C, and crystallization conditions modified from those previously described (19). QFR was co-crystallized with fumarate in 12.5% polyethylene glycol (PEG) 8000, 85 mm MgCH₃COO, 100 mm sodium citrate, pH 5.8, 0.1% DTT, 0.1 mm EDTA, and 5 mm fumarate. QFR was co-crystallized with oxaloacetate in 12% PEG 5000 monomethyl ether, 125 mm MgCH₃COO, 100 mm sodium citrate, pH 5.8, 0.1% DTT, 0.1 mm EDTA, and 0.5 mm oxaloacetate. QFR was co-crystallized with glutarate in 14.5% PEG 8000, 125 mm MgCH₃COO, 95 mm sodium citrate, pH 5.8, 0.1% DTT, 0.1 mm EDTA, and 10 mm glutarate. For co-crystallization of QFR with 3-NP, prior to crystallization, QFR was incubated with a 3-fold molar excess of 3-NP on ice for 1 h at pH 10.0. Crystals were grown in 10% PEG 5000 monomethyl ether, 250 mm MgCH₃COO, 100 mm sodium citrate, pH 5.8, 0.1% DTT, and 0.1 mm EDTA. All crystals formed in the orthorhombic space group P2₁2₁2₁ with unit cell dimensions and data collection statistics shown in Table 1.

TABLE 1.

Crystallographic data collection and refinement statistics

	With fumarate	With glutarate	With 3-NP	With oxaloacetate
Data collection
Wavelength	1.0 Å	1.03 Å	1.0 Å	1.2 Å
Beamline	SSRL 9-2	SSRL 11-1	ESRF ID-13	SSRL 9-2
Resolution	2.80 Å	3.05 Å	3.10 Å	3.35 Å
Completeness	94.9% (95.6%)a	95.0% (96.6%)	92.9% (78.2%)	88.9% (73.9%)
I/σ	23.4 (5.8)	10.7 (3.5)	16.8 (4.1)	10.0 (2.7)
Rsymb	0.089 (0.298)	0.099 (0.296)	0.098 (0.256)	0.099 (0.237)
Refinement statistics
Rcrystc	0.249	0.284	0.255	0.220
Rfree	0.272	0.284	0.279	0.261
Root mean square deviation
Bonds (Å)	0.014	0.014	0.007	0.011
Angles (degrees)	1.854	1.888	1.651	1.868

^a Values in parenthesis indicate statistics for the highest shell.

^b R_sym = Σ|I_i − 〈I〉|/Σ I_i, where I is intensity, I_i is the ith measurement, and 〈I〉 is the weighted mean of I.

^c R_cryst = Σ‖F_o| − |F_c‖/Σ F_o. R_free is the same as R_cryst for a set of data omitted from the refinement.

Data Collection and Processing

Crystallographic data were collected at the beamlines listed in Table 1 at −173 °C after the addition of 30% ethylene glycol to the crystallization conditions as a cryoprotectant. Data were processed using DENZO, SCALEPACK (20), HKL2000 (21), and the CCP4 suite of programs (22). Because crystals were isomorphous with crystals from known structures of QFR, rigid body refinement was performed with CNS (23) to determine initial phases. The starting model was QFR in complex with heptyl quinoline N-oxide (PDB entry 1KF6 (24)) for the 3-NP and glutarate co-structures, a QFR in complex with citrate (PDB entry 2B76 (17)) for the fumarate co-structure, and the finished glutarate co-structure for the oxaloacetate co-structure. Iterative rounds of model rebuilding were performed in O (25) and COOT (26). Omit maps, starting with segments of 50 amino acids in the initial stages of refinement and eventually progressing to 250 amino acid segments, were used extensively to minimize electron density map bias to the starting model. Refinement and map calculations were performed with Refmac5 (22, 27), CNS, and Phenix with loose 2-fold non-crystallographic symmetry restraints used in the fumarate structure and translation-liberation-screw (TLS) refinement (28) in all the structures. TLS refinement parameters were determined with the TLSMD server (29). Figs. 1A and 2 were created in the program PyMOL.

FIGURE 2.

Substrate and inhibitor binding to the active site. The active site of QFR is shown with nitrogen atoms colored blue, oxygen atoms colored red, carbons from the protein colored gray, carbons from the FAD colored yellow, and carbons from each bound ligand colored cyan. Relevant interatomic distances, including potential hydrogen bonds, are indicated along a dashed line. A, active site of QFR co-crystallized with fumarate. B, active site of QFR co-crystallized with oxaloacetate. C, active site of the FrdA E49Q variant of QFR co-crystallized with 3-NP. D, active site of QFR co-crystallized with glutarate.

Molecular Modeling of Dicarboxylate Binding

QFR co-structures with fumarate, glutarate, oxaloacetate, and citrate were prepared for modeling with Maestro (Schrödinger LLC) (30) by creating a file with only the flavoprotein subunit, the iron protein subunit, and covalently linked FAD. Protons were added to the QFR structure with Maestro, and the resultant structure was minimized using the molecular mechanics force field, OPLS_2005. Ligands were prepared using the Ligprep module by adding explicit hydrogen atoms. These ligands were then reinserted into the active site of the minimized protein. The Glide XP subroutine in Maestro was used for ligand docking, and the free energy of binding was estimated with the extended precision scoring function (Glide XP) (30). During this phase of the calculation, flexibility of the ligand was allowed, whereas the protein was held rigid.

RESULTS

QFR activity can be inhibited both in vivo and in vitro by dicarboxylates or carboxynitro molecules structurally similar to fumarate (Fig. 1B). In this study, we investigate the determinants for binding and catalysis by identifying the interactions between QFR and the substrate fumarate or the inhibitors oxaloacetate, glutarate, and 3-nitropropionate.

Structural Details of Fumarate Binding with the E. coli QFR

Co-crystallization of QFR with fumarate resulted in the appearance of clear electron density within the enzyme active site. This density is consistent with an alignment of the C2–C3 double bond parallel to the C(4a)–N5 bond of the isoalloxazine ring of FAD (31). Notably, a torsioned conformation of fumarate is observed (Fig. 2A), where the carboxylates are oriented 54° relative to one another, compared with the 0° orientation found in planar molecules. Torsion of fumarate is predicted to have a significant electronic effect, with the C2 carbon bearing the twisted carboxylate becoming a better hydride acceptor because electrons are better allowed to localize to the carbonyl (31, 32). Hydride transfer thus presumably occurs between N5 of FAD and C2 of fumarate, and the distance between these atoms is 4 Å. Subsequent proton transfer is predicted to occur between the side chain of Arg-A287 and the C3 of fumarate. Of note, an ordered water molecule mediates the interaction between the C3 position of fumarate and the proton shuttle, Arg-A287. Seven putative hydrogen-bonding contacts stabilize fumarate in this torsioned conformation (Fig. 2A).

The possibility of enzymatic turnover or photoreduction during data collection precludes positive identification of this dicarboxylate as fumarate with crystallography alone. To support the assignment of the ligand as fumarate (rather than citrate from the crystallization conditions or the product succinate), optical difference spectra were collected. The binding of fumarate to QFR results in the appearance of peaks at 390 and 500 nm in the FAD absorption spectrum (Fig. 3). The addition of QFR to the crystallization buffer induced optical changes that indicate replacement of oxaloacetate with citrate. Upon the addition of 5 mm fumarate, the concentration used during co-crystallization, a typical optical spectrum reflecting the replacement of citrate with fumarate was observed (data not shown).

FIGURE 3.

Optical difference absorption spectroscopy of QFR after the addition of ligands. Spectra are the difference between the spectrum of the enzyme after the addition of each ligand and the spectrum of the oxidized enzyme. The spectra are colored blue for fumarate, green for oxaloacetate, purple for glutarate, orange for citrate, black for 3-nitropropionate, and red for malonate.

Structural Details of QFR Inhibition by Oxaloacetate

The structure of QFR in complex with oxaloacetate shows that this dicarboxylate inhibitor binds to QFR with a position similar to that of fumarate (Fig. 2B). In this binding mode, the C2–C3 bond is parallel to C(4a)–N5 of the FAD isoalloxazine ring, similar to the orientation of the C2–C3 double bond of fumarate in the co-structure of fumarate with QFR. In addition, the oxaloacetate carbonyl is positioned laterally, such that C2=O carbonyl is nearly parallel with the adjacent N5–C(5a) bond. This structural feature possibly underlies the charge transfer interaction, which results in the formation of a broad peak in the optical spectrum centered at 550 nm (Fig. 3). Interestingly, although the apparent affinity of oxaloacetate to QFR is much greater than that of fumarate, only three putative hydrogen bonds stabilize oxaloacetate in this binding position (Fig. 2B). This suggests that the charge transfer interaction makes a major contribution to binding affinity.

Structural Basis of Covalent Inhibition by 3-NP

Differences in inhibition between succinate:ubiquinone oxidoreductase and QFR have mechanistic implications because each of these enzymes optimally catalyzes the reaction in one direction. As a result, we tested the effects of 3-NP addition on both wild-type QFR and the FrdA E49Q variant of QFR, which has previously been shown to have altered catalytic efficiency for fumarate reduction and succinate oxidation (17). Crystals of the FrdA E49Q variant of QFR in complex with 3-NP exhibited superior diffraction as compared with crystals of wild-type QFR in complex with 3-NP. As a result, crystals of the mutant enzyme were used for all subsequent structural analyses.

In the structure of 3-NP with the E49Q variant of QFR, electron density contiguous with Arg-A287 is evident, consistent with the formation of a covalent adduct between 3-NP and Arg-A287. This density is in the same location as the ordered solvent molecule observed in the co-structure between QFR and fumarate (Fig. 2A). Proposed adducts from the literature were used as starting models to interpret the active site density, including 3-NP as a non-covalent inhibitor (14), 3-NP forming an adduct to Arg-A287 (13), 3-NP forming a covalent adduct to the FAD (12), and mixed states combining these possibilities. A 3-NP adduct to Arg-A287 modeled as a five-membered 2,4,5-triazole structure best explains the density (Fig. 2C), in close agreement with the crystal structure of avian complex II in complex with 3-NP (13). Notably, the side chain of adducted Arg-A287 displays an extended conformation when compared with the position of Arg-A287 from citrate-bound QFR (PDB entry 2B76) (17). This results in a 2.8-Å displacement of the guanidino group toward the FAD. An electrostatic contact between the terminal carboxylate of 3-NP and the N5 of FAD stabilizes the adduct. Further, the triazole ring makes putative hydrogen bond contacts to three side chains in the active site, with likely hydrogen-bonding contacts between the NH1 of the adduct and the Oϵ1 atom on Gln-A230, between the NH₂ of the adduct and the Oϵ1 carbonyl atom on Glu-A245, and between the N13 of the adduct and the Nη2 of Arg-A390.

Although the structure of the FrdA E49Q variant of QFR in complex with 3-NP demonstrates that the binding mode is distinctly different from that of fumarate, intriguingly, the addition of 3-NP to QFR induces changes to the optical spectrum that are similar to those observed with the addition of fumarate, with peaks appearing at 390 and 500 nm (Fig. 3). This observation is consistent with previous suggestions that 3-NP could act as an alternative substrate for complex II enzymes (8, 12). Therefore, to determine whether catalytic oxidation of 3-NP takes place during adduct formation, the optical and catalytic properties of the 3-NP and E. coli QFR interaction were investigated over time. The addition of 3-NP to wild-type E. coli QFR induces changes in the optical spectrum (Fig. 4E). The time course of these optical changes matches the change in catalytic activity, suggesting that 3-NP is, indeed, a substrate of QFR and that oxidation by FAD forms the species that attacks Arg-A287 to result in the formation of the 3-NP-Arg-A287 adduct. Neither further optical changes due to FAD interaction with a ligand nor oxidase activity was observed upon the addition of a second equimolar dose of 3-NP. This indicates that a single catalytic turnover of QFR is a requirement for formation of the 3-NP-Arg adduct.

FIGURE 4.

Identification of the 3-NP adduct and a proposed mechanism for its formation. A, tandem mass spectrum (MS/MS) from analysis of fragmentation of the YMELGPR*DK peptide adducted to 3-NP. B, an MS/MS/MS (MS³) spectrum of the -CO₂ neutral loss fragment isolated from 3-NP-adducted peptide fragmentation (m/z 574.5 from A). C, tandem mass spectrum obtained from analysis of fragmentation of the YMELGPR*DK peptide adducted to ¹⁵N-labeled 3-NP. D, an MS³ spectrum of the -CO₂ neutral loss fragment isolated from the ¹⁵N-labeled 3-NP-adducted peptide fragmentation (m/z 575.1 from B). E, optical spectrum following the reduction of FAD (orange) or consumption of DCIP (dark red) following 3-NP addition. Activated QFR (∼5 μm) was added to 50 mm Bistris propane, pH 8.0, 30 °C. 3-NP (0.2 mm final concentration) was added from a pH 10.0 solution, and changes in the optical spectrum were monitored at 505–439 nm (upper trace). Lower trace, 50 μm DCIP was added to the cuvette at the same QFR concentration (5 μm) and reduction of DCIP was monitored at 600 nm upon the addition of 3-NP from the pH 10 solution. The arrow in both the upper and lower trace indicates that the addition of a second equivalent of 3-NP did not cause additional spectral changes, indicating that one equivalent of 3-NP is sufficient to inhibit the enzyme. F, possible minimal mechanism for formation of the covalent adduct between 3-NP and Arg-A287. *, site of modification (R* = R + 3-NP).

Verification of the Chemical Identity of the 3-NP Adduct by Mass Spectrometry

The location of the 3-NP adduct was verified using LC-MS/MS of trypsinized FrdA pretreated with 3-NP. The 3-NP-adducted peptide was identified as a doubly charged 596.3 m/z ion with 3-NP adduction with Arg-A287 (Fig. 4A). Further fragmentation identified the neutral loss of CO₂ in the MS³ spectrum (Fig. 4C), strongly suggesting that the adduct is oriented with the nitro group coupled to Arg-A287. Because direct coupling of a nitroalkane and amine is rare and elimination of nitrous acid (HNO₂) is much more favorable, it was important to determine unambiguously whether the nitrogen of 3-NP was incorporated into the triazole. ¹⁵N-Labeled 3-NP (98% ¹⁵N) was synthesized, and the mass spectral analysis was repeated. The MS-MS and MS³ analyses exhibited a shift of 1 Da for multiple-fragment ion peaks, demonstrating incorporation of ¹⁵N into the triazole adduct (Fig. 4, B and D).

Structural Basis for Inhibition by Glutarate

Increased systemic glutarate levels have been correlated with decreased complex II activity in vivo (7) but have not previously been shown to inhibit the enzyme directly. Activity of QFR was measured in the presence and absence of glutarate, which demonstrates that this dicarboxylate acts as a competitive inhibitor with a K_i of 1.9 mm (Table 2). The structure of QFR in complex with glutarate shows a torsioned conformation of the glutarate (Fig. 2D). The angle between the dicarboxylates is 87°, larger than in the fumarate co-structure. In this binding mode, the torsion is distributed along a greater number of bonds, and the C2 atom bends away from the rest of the ligand, causing the C2–C3 and C3–C4 bonds to buckle outward from the FAD. As a result, none of the carbon-carbon bonds of glutarate lie along the C(4a)–N5 of the FAD isoalloxazine ring. Glutarate does not induce changes in the FAD optical difference spectrum (Fig. 3), which supports the hypothesis that the alignment of a bond along the C(4a)–N5 bond of FAD underlies these spectral changes. Stabilizing glutarate in this conformation are five putative hydrogen-bonding contacts (Fig. 2D).

TABLE 2.

Comparison of the apparent affinity values of dicarboxylates to E. coli QFR

	Km	Ki
	mm	mm
Fumarate	0.02a
Succinate	0.5a
Oxaloacetate		0.0003a
Malonate		0.025a
Citrate		20
Glutarate		1.9

^a Data from Ref. 17.

Docking Calculations

Ligand docking calculations in Glide resulted in ligand poses consistent with the experimentally observed binding conformations. From these binding poses, an overall binding score was calculated with energetic contributions from internally calculated coulombic interactions, hydrogen bonding interactions, and van der Waals interactions (Table 3). Surprisingly, these simulations predict that fumarate binds with the least favorable coulombic interactions of all ligands tested and the second least favorable electrostatic interactions. However, the calculated van der Waals interactions of fumarate were more favorable than all other ligands tested. The opposite was true in the case of the larger molecule glutarate, which was calculated to have the most favorable electrostatic and hydrogen bonding interactions but the least favorable calculated van der Waals interactions. Citrate binding was also predicted to show unfavorable van der Waals contacts, probably leading to the lowest experimentally determined apparent affinity of all of the ligands tested (Table 3). Oxaloacetate binding energy was not accurately predicted by the Glide algorithm, which does not take charge transfer interactions into account.

TABLE 3.

Binding scores of ligands to the E. coli QFR as calculated in Glide XP

Ligand	Total score	Hydrogen bonding	Coulombic	van der Waals
Fumarate	−2.39	−3.9	−24.1	−7.4
Oxaloacetate	−4.37	−3.9	−21.8	−12.1
Glutarate	−10.6	−4.3	−23.6	−12.7
Citrate	−8.49	−2.8	−16.3	−11.1

DISCUSSION

Complex II enzymes must perform the challenging task of activating a chemically stable bond while discriminating between structurally and electronically similar substrates. To identify the properties of QFR important for binding and catalytic specificity, structural and spectroscopic analyses of QFR in complex with activatable and non-activatable ligands were compared. Our results demonstrate the importance of geometric constraints on ligand positioning relative to the FAD cofactor in both on-pathway and off-pathway catalysis.

On-pathway Catalysis in QFR

Previous studies demonstrate that QFR reduces fumarate to succinate in two distinct steps (33). The first is the rate-limiting hydride transfer from FAD to fumarate, and the second is proton transfer from a nearby side chain. It is hypothesized that the hydride transfer step of this α,β-dehydrogenation reaction is kinetically difficult because the relatively short fumarate lacks extensive resonance structures to delocalize developing charges during catalysis. To overcome this kinetic stability and catalyze hydride transfer, QFR has been proposed to combine multiple geometric and electrostatic mechanisms, including substrate polarization (34), active site desolvation, concerted active site rearrangement (35), and substrate torsion (31, 32).

Fumarate binds the E. coli QFR in a torsioned conformation (Fig. 2A) similar to that observed in co-structures with Wolinella succinogenes QFR (31) and soluble homologs of the QFR flavoprotein (32, 33), with interdomain rotations probably contributing to the torsioning (35). We predict that this torsioning of fumarate may facilitate catalysis by allowing improved delocalization of electrons between the C2 and the carbonyl to narrow the transition state energy barrier.

Fumarate binding induces the appearance of peaks at 390 and 500 nm in the FAD optical spectrum. Similar peaks are observed upon the addition of malonate or 3-NP but are absent with oxaloacetate, citrate, and glutarate addition (Fig. 3). Inspection of the co-structure of QFR with fumarate suggests that alignment of the chemically reactive C2–C3 bond parallel to the C(4a)–N5 bond of FAD (Fig. 5A) could be important for this spectroscopic change and may also be critical for catalysis. Although the alignment of the C2–C3 bond of oxaloacetate with the C(4a)–N5 of FAD is also observed in the co-structure (Fig. 5B), the additional interaction between the FAD and the carbonyl may modify these spectral features in that case.

FIGURE 5.

Comparison of substrate binding in flavin-containing enzymes catalyzing α,β-dehydrogenation reactions. Dashes highlight the activated substrate bond and the C(4a)–N5 bond (FAD), which are aligned. In all of the panels, nitrogen is colored blue, oxygen is colored red, FAD carbons are colored yellow, and sulfur is colored tan. A, the E. coli QFR co-crystallized with fumarate. Carbon atoms of the fumarate are colored cyan. B, the E. coli QFR co-crystallized with oxaloacetate. Carbon atoms of the oxaloacetate are colored magenta. C, d-amino acid oxidase co-crystallized with d-alanine (PDB entry 1C0L) (33). Carbon atoms of d-alanine are colored green. D, porcine mitochondrial medium-chain acyl-CoA dehydrogenase co-crystallized with acyl-CoA (PDB entry 3MDE) (32). Carbon atoms of acyl-CoA are colored orange. E, a model of the LUMO of fumarate and HOMO of FAD illustrating the predicted orbital overlap. Blue and gold represent different phases of the orbital wave function.

A similar alignment of the reactive bond of substrate and FAD is observed in co-complexes of natural substrates with evolutionarily unrelated flavoenzymes that house FAD-catalyzed α,β-dehydrogenation reactions. These include acyl-CoA dehydrogenase (Fig. 5C; PDB entry 3MDE) (36), d-amino acid oxidase (Fig. 5D; PDB entry 1C0L) (37), cholesterol oxidase (PDB entry 1COY) (39), old yellow enzyme (PDB entry 1YOB) (40), UDP-N-acetylenolpyruvoyl-glucosamine reductase (PDB entry 1MBR) (41), vanillyl-alcohol oxidase (PDB entry 2VAO) (42), and l-proline dehydrogenase (PDB entry 3E2S) (43). This conservation of substrate orientation across the isoalloxazine ring of multiple unrelated enzymes suggests that it is ideal for catalysis. Calculations show that the FAD HOMO consists largely of a dominant contribution from the N5 and C(4a) atoms in lumiflavin (44). Positioning of fumarate parallel to the C(4a)–N5 axis would be predicted to maximize the HOMO-LUMO overlap of the fumarate π* anti-bonding orbital and the FAD HOMO (Fig. 5E), the likely conduit for hydride transfer from the N5 position of FAD. A similar orbital overlap was also proposed to be important for inducing spectral changes in FAD of d-amino acid oxidase upon ligand binding to substrate (45). This observation is consistent with an orbital steering mechanism, which holds that optimal substrate orientation contributes to catalytic efficiency by maximizing overlaps of bonding orbitals and minimizing overlaps of non-bonding orbitals (46). Subtle geometric changes were shown to be important for hydride transfer upon subtle perturbations in isocitrate dehydrogenase, even after controlling for distance changes and electrostatic differences (47), and probably play a role in the catalytic mechanism of QFR.

Following hydride transfer from FAD, a proton is transferred from the side chain of an active site residue. In the Shewanella frigidimarina soluble fumarate reductase, an elegant combination of site-directed mutagenesis, x-ray crystallography, and kinetics demonstrated that an active site arginine structurally equivalent to Arg-A287 from the E. coli QFR acts as the proton shuttle (48–50). In the E. coli QFR co-structure with fumarate, a water molecule mediates the interaction between Arg-A287 and fumarate (Fig. 2A). Bound water molecules often act as proton shuttles in enzymes, but no on-pathway water molecule has been proposed as a proton donor in QFR or other complex II homologs. Although our crystal structure leaves open the possibility of water-mediated proton transfer in the E. coli enzyme, the location of this water molecule may suggest how complex II homologs perform off-pathway catalysis.

Off-pathway Catalysis 1; Oxaloacetate Formation via Enzymatic Fumarate Hydration to Malate

Oxaloacetate is a tight-binding inhibitor of QFR (3), and its formation might be catalytically induced upon fumarate hydration to malate followed by oxidation to oxaloacetate (4). In the structure of oxaloacetate bound to E. coli QFR (Fig. 5B), the oxaloacetate carbonyl is oriented parallel to the N5–C(5a) FAD similar to that observed in the oxaloacetate co-structure with avian complex II (PDB entry 1YQ4) (13) and the malate-like intermediate co-structure with a soluble fumarate reductase from S. frigidimarina (PDB entry 1QJD) (33). The planar orientation of oxaloacetate relative to FAD resembles that of fumarate, especially in relation to the similar positioning of the C2–C3 bond in fumarate and oxaloacetate (Fig. 5, A and B). Importantly, spectroscopic characterization in avian complex II has shown that fumarate can undergo a simple hydration reaction along the C2–C3 double bond to yield malate (5). This is corroborated by previous attempts to crystallize QFR homologs with fumarate, which have yielded active site density more consistent with oxaloacetate or malate than fumarate (33).

Intriguingly, only the electron density within the active site of structures of S. frigidimarina fumarate reductase (PDB entry 1QJD) (33) and avian complex II (PDB entry 2H88) (5) co-crystallized with fumarate at pH 7.2 or greater were consistent with the malate-like intermediate. Conversely, both the E. coli QFR and W. succinogenes QFR (PDB entry 1QLB) (31) were co-crystallized with fumarate at lower pH, and the electron density is consistent with a bound fumarate and a water molecule. This may reflect two separate catalytic states in the complex II homologs. Under high pH conditions, the proton shuttle may be ionized, and the active site is shifted into a state that allows for water addition, whereas lower pH leads to a deionized proton shuttle, trapping the water molecule near substrate. Although speculative, the water molecule bound between fumarate and Arg-A287 in the E. coli QFR (Fig. 2A) may identify the location of this attacking group during off-pathway malate formation from fumarate. When present, malate can be transformed to oxaloacetate via hydride transfer from FAD (4). The orientation of oxaloacetate lends support to a model where malate is first oriented correctly along the FAD ring and then activated for catalysis in a manner analogous to fumarate activation. It is easy to speculate a physiological role for this off-pathway catalysis because the formation of a tightly binding inhibitor could regulate the activity of QFR and other complex II homologs.

The higher apparent affinity of oxaloacetate compared with fumarate may potentially be rationalized by interactions with the FAD ring. The expected HOMO-LUMO overlap of the π-bonding orbital from the C2–C3=O conjugate of the enol form of oxaloacetate and the LUMO of oxidized FAD could underlie the formation of the charge transfer interaction (Fig. 5E) reflected in the band centered at 550 nm in the optical spectrum (Fig. 3) (3). This HOMO-LUMO overlap is somewhat analogous to the overlap of the C2–C3 double bond of fumarate with the FAD LUMO, except that the oxaloacetate carbonyl may also contribute to the formation of the conjugated-π system because the almost planar alignment of the C2–C3–O5 atoms of oxaloacetate probably aligns the atomic p_z orbitals of these atoms with those of the isoalloxazine ring (Fig. 5E). A similar delocalized π-conjugated system has been proposed to underlie charge transfer formation during the binding of acetoacetyl-CoA in acyl-CoA dehydrogenase (51). The Glide calculations do not include a term corresponding to the formation of a conjugated system. That computational modeling only failed to account for binding affinity of oxaloacetate is consistent with a conjugated-π interaction contributing to the affinity of oxaloacetate binding.

Off-pathway Catalysis 2; Suicide Inhibition via Covalent Adduct Formation with 3-NP

X-ray crystallography and mass spectral analysis suggest that 3-NP addition to QFR results in the formation of a 2,3,5-triazole adduct to the likely proton shuttle Arg-A287. The formation of a 2,3,5-triazole raises a number of chemical challenges, including addition to an unreactive carbon atom and reduction of the nitro functional group. The ¹⁵N-labeled 3-NP experiment (Fig. 4, A and B) confirms the presence of nitrogen from 3-NP in the adduct, ruling out loss of the nitro group through elimination, which is a common reaction for activated nitroalkanes. Instead, we considered pathways involving both an overall oxidation of 3-NP and a localized formal reduction and dehydration of nitrogen. It has previously been demonstrated that the dehydrogenation product of 3-NP, 3-nitroacrylate, shows faster inhibition kinetics than 3-NP, which suggests that 3-NP is converted to 3-nitroacrylate prior to forming a covalent adduct (12). The most parsimonious mechanism of conversion of 3-NP to 3-nitroacrylate is by the complex II enzyme itself. In a co-crystal structure of 3-NP with porcine complex II, a non-covalent binding mode for 3-NP was identified within the enzyme active site (14). Dehydrogenation of 3-NP to 3-nitroacrylate, a reaction analogous to succinate oxidation to fumarate, can be inferred by the observation that FAD oxidizes upon 3-NP addition (Fig. 4E). Furthermore, FAD oxidation exactly matches QFR inactivation, suggesting that 3-nitroacrylate is the active chemical species in adduct formation.

The 3-nitroacrylate intermediate would be expected to be a powerful electrophile that engages arginine in an addition reaction. Furthermore, an induced dipole formed by interactions with the polar active site, along with geometric torsion, may further activate 3-nitroacrylate for addition by arginine in a manner similar to fumarate activation. Optical changes in FAD similar to those seen with fumarate binding are observed upon 3-NP binding, suggesting that nonadducted 3-nitropropionate and 3-nitroacrylate probably produce the same fumarate C2–C3 bond overlap over the FAD C(4a)–N5 to further activate the unreactive C2 atom adjacent to the nitro group. Upon the addition of arginine, nitro reduction to nitroso releases a molecule of water. The reactive intermediate prior to nitroso formation could not be characterized; however, it is possible that formation of nitronic acid by resonance from a carbanion could facilitate nitro reduction. Nitroso formation provides an opportunity for N–N bond formation by the addition of the arginine nitrogen to the electrophilic nitroso nitrogen. A final dehydration and tautomerization would then provide the observed triazole (Fig. 4F).

In cases where nitroalkanes are converted to amide derivatives, the nitroalkane is first transformed to the corresponding ketone or carboxylic acid, resulting in cleavage of the carbon-nitro bond. A very recent exception to this is the reaction of an α-bromo nitroalkane with an electrophilic amine (a halamine, RNHI) (38). However, the carbon-nitro bond is cleaved during this transformation as well. The nitroalkane derivatization to a triazole without carbon-nitro bond cleavage suggests an unprecedented mechanism in the context of known nitroalkane chemistry. A mechanistic hypothesis is advanced (Fig. 4F) that is consistent with the experiments described above.

Molecular Basis for Competitive Inhibition

The alignment of an activatable bond along the C(4a)–N5 bond of the FAD, as is reflected in the optical difference spectra, is a clear requirement for either on- or off-pathway catalysis. This suggests that dicarboxylates that do not cause differences in the optical spectrum, such as glutarate and citrate, do not align similarly. This is indeed what is supported by our data. Glutarate does not show an electronic interaction with FAD but binds within the E. coli QFR active site with torsioned carboxylates and at the same location as fumarate (Fig. 2D). However, the increased length and flexibility of glutarate results in a binding mode where the alkane chain buckles outward and does not orient any bond along the C(4a)–N5 bond of FAD. A similar effect is noticed in the published QFR co-structure with citrate (17), where citrate lacks a bond aligned to the FAD ring. Additionally, both glutarate and citrate are predicted to have less favorable steric interactions, making them bind less tightly than other QFR ligands (Table 3). That neither glutarate nor citrate shows an electronic interaction with FAD in optical spectra (Fig. 3) is consistent with our hypothesis that alignment of an activatable bond parallel to the C(4a)–N5 bond of FAD underlies this spectral change and offers a possible explanation for the lack of catalytic turnover of glutarate or citrate. This model is consistent with QFR using an orbital steering mechanism where geometric constraints of the ligand dictate correct orbital overlap.

Conclusions

The geometric and electronic activation of ligands may be a key factor in determining ligand activation and behavior in flavoenzymes. This geometric activation controls reactivity in QFR and predicts that potential substrates for QFR and FAD-containing enzymes that catalyze α,β-dehydrogenation reactions may be identified by examination of optical difference spectra, which undergo changes upon orbital interaction with precisely oriented ligands.