Fumarate Reductase and Succinate Oxidase Activity of Escherichia coli Complex II Homologs Are Perturbed Differently by Mutation of the Flavin Binding Domain*

The Escherichia coli complex II homologues succinate:ubiquinone oxidoreductase (SQR, SdhCDAB) and menaquinol:fumarate oxidoreductase (QFR, FrdABCD) have remarkable structural homology at their dicarboxylate binding sites. Although both SQR and QFR can catalyze the interconversion of fumarate and succinate, QFR is a much better fumarate reductase, and SQR is a better succinate oxidase. An exception to the conservation of amino acids near the dicarboxylate binding sites of the two enzymes is that there is a Glu (FrdA Glu-49) near the covalently bound FAD cofactor in most QFRs, which is replaced with a Gln (SdhA Gln-50) in SQRs. The role of the amino acid side chain in enzymes with Glu/Gln/Ala substitutions at FrdA Glu-49 and SdhA Gln-50 has been investigated in this study. The data demonstrate that the mutant enzymes with Ala substitutions in either QFR or SQR remain functionally similar to their wild type counterparts. There were, however, dramatic changes in the catalytic properties when Glu and Gln were exchanged for each other in QFR and SQR. The data show that QFR and SQR enzymes are more efficient succinate oxidases when Gln is in the target position and a better fumarate reductase when Glu is present. Overall, structural and catalytic analyses of the FrdA E49Q and SdhA Q50E mutants suggest that coulombic effects and the electronic state of the FAD are critical in dictating the preferred directionality of the succinate/fumarate interconversions catalyzed by the complex II superfamily.

Succinate dehydrogenase (succinate:ubiquinone oxidoreductase (SQR), 5 complex II) and fumarate reductase (menaquinol:fumarate oxi-doreductase (QFR)) couple the interconversion of succinate and fumarate with quinone and quinol. Succinate dehydrogenase is part of the aerobic respiratory chain and citric acid cycle of most organisms, whereas fumarate reductase is found in anaerobic or facultative bacteria and lower eukaryotes that live a portion of their life cycle in a reduced oxygen environment (1). Based upon amino acid sequence analysis, biochemical studies, and their overall structures, it has been proposed that both enzymes arose from a common evolutionary ancestor (2)(3)(4). Succinate dehydrogenase and fumarate reductase from Escherichia coli are composed of four nonidentical subunits organized into two domains. A membrane-extrinsic domain comprises two polypeptide chains: a 64 -66-kDa flavoprotein subunit, containing a covalently bound FAD cofactor and the substrate binding site, and a 27-kDa ironsulfur subunit containing three iron-sulfur clusters ([2Fe-2S] 2ϩ,1ϩ , [4Fe-4S] 2ϩ,1ϩ , and [3Fe-4S] 1ϩ,0 ). The membrane-extrinsic domain is bound to the membrane through interactions with the hydrophobic subunits of the complex. These subunits comprise two membrane anchor polypeptides, each containing three transmembrane helices and providing a binding site(s) for quinone (for reviews, see Refs. [5][6][7][8]. In addition, the E. coli SQR hydrophobic peptides bind one b-type heme, whereas the E. coli QFR lacks heme. Comparison of the structures of complex II is possible due to the availability of x-ray crystallographic structures for both SQR and QFR of E. coli (9,10), the porcine SQR (11), the QFR from Wolinella succinogenes (12), and soluble homologs of the flavoprotein subunit that function as periplasmically localized fumarate reductases (13)(14)(15)(16). The flavoprotein subunits from the E. coli SQR and QFR are highly homologous, with 64% similarity and 44% identity of amino acid residues (3). The sequence similarity within the SQR/QFR superfamily is reflected in structural alignments of members whose structures are known (17)(18)(19). Within this group, the backbone C ␣ atoms can be superimposed with a maximal root mean square deviation of 1.5 Å (10). A detailed hydride transfer mechanism for fumarate reduction has been proposed based on structural data and enzyme assays of wild type and mutant enzymes (15, 18 -20). QFRs and SQRs all contain covalently bound FAD and are bidirectional (i.e. they will catalyze both succinate oxidation and fumarate reduction). There are, however, significant differences for the kinetics of fumarate reduction by fumarate reductase and succinate dehydrogenase. In addition to conventional steady-state solution kinetics, these differences have been measured by electro-chemical experiments (protein film voltammetry) in which the enzymes exhibit high electrocatalytic activity when adsorbed on a graphite electrode. In these studies, SQR is proficient at reducing fumarate above an electrode potential of approximately Ϫ60 mV (pH 7.0), whereas catalysis is severely constrained at potentials below this value (21,22). This electrochemical property of succinate dehydrogenase has been described as being analogous to that of a tunnel diode, which is a device displaying negative resistance in a certain potential region (21,22). Native QFRs by contrast, exhibit normal positive order kinetics (i.e. the rate of fumarate reduction increases with increasing thermodynamic driving force). The reasons for the differences in catalytic activity between the enzyme complexes remain unclear but are likely to relate to differences between the FAD and/or dicarboxylic acid binding sites of SQR and QFR (23,24).
The remarkable degree of conservation of amino acid residues involved in substrate binding and interaction with the flavin in SQR and QFR has a notable exception. In E. coli FrdA, Glu-49 is located within 5 Å of the FAD cofactor, whereas in E. coli SdhA, a Gln residue (Gln-50) is found at this position. The majority of enzymes classified as QFRs contain a Glu at this position, whereas all SQRs contain a Gln. The soluble fumarate reductase homologs are exceptions, since they contain an Ala at this position. However, their structures reveal that this location is filled by the propionate group of a heme moiety, not present in the QFR/SQR family, which conserves the negative charge found in the QFRs. It should also be noted that the soluble flavoprotein homologs contain a noncovalently bound FAD cofactor and are essentially unable to oxidize succinate (25).
In this paper, we compare the catalytic and dicarboxylate binding effects of Glu, Gln, and Ala residues at position 49 in E. coli FrdA and position 50 in E. coli SdhA. The results show that Ala substitutions do not significantly alter the catalytic properties of QFR or SQR. Substitution at this position with the residue found in the complementary enzyme has profound effects on catalytic activity, with the SdhA-Q50E SQR mutant becoming a more efficient fumarate reductase, whereas the FrdA-E49Q QFR becomes a better succinate dehydrogenase. These substitutions, however, give no detectable alteration of the kinetic properties related to the "tunnel diode" effect. The results are discussed in the context of altered substrate binding, supported by an x-ray structure of the FrdA E49Q mutant QFR, and effects on the electronic status of the FAD moiety.

Bacterial Strains and Plasmids
E. coli strain DW35 (⌬frdABCD, sdhC::kan), which was used as the host for expression of wild type and mutant forms of SQR and QFR, has been previously described (26). Plasmid pH3 (frdA ϩ B ϩ C ϩ D ϩ ) was used for expression of wild type QFR (26), and plasmid pFAS (P FRD sdhC ϩ D ϩ A ϩ B ϩ ) was used for expression of wild type SQR (27). Plasmid pFAB-HT (frdA ϩ B ϩ ) containing a C-terminal His tag has been previously described (24) and was used for expression of the two-subunit soluble fumarate reductase (FrdAB).

Mutagenesis
Plasmids pH3 and pFAS were used as the templates for mutagenesis of the frdA and sdhA genes in QFR and SQR. Plasmid pFAB-HT was used as template for mutagenesis of frdA in the soluble FrdAB enzyme. Site-directed mutagenesis was performed utilizing the QuikChange site-directed mutagenesis kit from Stratagene (La Jolla, CA). Primers for mutagenesis were obtained from Qiagen (Valencia, CA). Nucleotide changes to construct the mutants are underlined. To construct the SdhA Q50A mutant, primer 5Ј-ACCGTTTCTGCGGCAGGCG-GCATTACCG-3Ј and its reverse complement were designed. The SdhA Q50E mutant was constructed using the primer 5Ј-CCGTTTCT-GCGGAAGGCGGATTACC-3Ј and its reverse complement. Following mutagenesis, the 1390-base pair HindIII-BstXI fragment containing either the SdhA Q50E or Q50A mutation was cloned into pFAS for expression of the mutant enzyme.
For the QFR FrdA E49Q mutation, the primer 5Ј-CATACCGTT-GCTGCACAGGGGGGCTCC-3Ј and its reverse complement were designed. The 1277-base pair BstEII-ApaI fragment containing the FrdA E49Q mutation was cloned into pH3 for the expression of the four subunit QFR and into pFAB-HT for the expression of twosubunit His-tagged fumarate reductase. The QFR FrdA E49A mutant was constructed using primer 5Ј-CCGTTGCTGCAGCAGG-GGGCTCCGC-3Ј and its reverse complement. Cloning of this mutant into plasmid pH3 used a similar strategy to that used for the FrdA E49Q mutant. All mutations were confirmed by DNA sequencing using the University of California San Francisco biomolecular core facility. Plasmids encoding the mutations were subsequently transformed into E. coli DW35 for expression of mutant enzyme.

Growth Conditions
QFR, SQR, and FrdAB were expressed in DW35 cells harboring the appropriate plasmid and grown under microaerophilic conditions. Starter cultures (25 ml) in Luria-Bertani (LB) medium with 150 g/ml ampicillin were grown overnight at normal aeration. These cells were then used to inoculate 2-liter flasks filled with 1.4 liters of Terrific Broth medium containing ampicillin (150 g/ml), and the cells were grown overnight at 37°C with moderate aeration (150 rpm) on a gyrotory shaker. E. coli used to express soluble FrdAB was grown using an identical protocol for 16 h. Cells were collected by centrifugation (10 min at 5000 ϫ g) and stored at Ϫ80°C.

Enzyme Purification
Isolation of the membrane fractions enriched with wild type and mutant SQR and QFR enzymes was as previously described (28,29). Protein extraction with the detergent Thesit (Anapoe C 12 E 9 (polyoxyethylene 9-dodecyl ether; Anatrace Inc., Maumee, OH) and purification on Q-Sepharose fast flow chromatography were carried out as previously described for wild type and mutant QFR and SQR (28,29). Purified enzyme fractions were pooled, concentrated under nitrogen using an Amicon cell and YM30 membrane, and stored at Ϫ80°C. Isolation of the soluble His-tagged FrdAB suitable for protein film voltammetry was performed as previously described (24) using nickel affinity resin (Qiagen). Soluble FrdA E49Q B mutant enzyme was purified by a protocol identical to that used for wild type FrdAB.

Activation of the Enzymes
FrdAB as isolated exhibits high initial rates of succinate oxidation and does not require preactivation. SQR and QFR as isolated, however, are only partially active due to the presence of tightly bound oxaloacetate at their active sites. To activate the enzymes, SQR and QFR were diluted to 1-2 mg of protein/ml in 30 mM bis-Tris-propane buffer, pH 7.0, 0.1 mM EDTA, 0.05% Thesit, 3 mM malonate and incubated for 20 min at 30°C. Activated enzymes were stored on ice for the duration of the experiment.
Fumarate Reduction-Fumarate reduction activities of QFR and SQR with quinol analogues were determined in a reaction coupled to DT diaphorase (NADH:quinone reductase) as previously described (30) with menaquinone (MQ 1 ) and ubiquinone (UQ 1 ) for QFR and SQR, respectively (MQ 1 and UQ 1 were kindly provided by Eisai Co. Ltd., Tokyo, Japan). Fumarate reduction with a 0.2 mM concentration of the low potential electron donors benzylviologen (BV) or methylviologen was performed in 3-ml screw top cuvettes under a continuous flow of argon. Prior to initiation of the reaction, 10 mM glucose, glucose oxidase, and catalase were added to the assay medium to maintain anaerobiosis. A stoichiometric amount of sodium dithionite was added to reduce the viologens (starting absorbance ϳ1.8), and the reaction was initiated by the addition of either enzyme or fumarate. The progress of the reaction was monitored by the decrease of viologen absorbance at 602 nm (⑀ 602 ϭ 9.6 mM Ϫ1 cm Ϫ1 , pH 7.0), k cat was calculated using covalently bound FAD content.

Electrochemistry
The soluble FrdAB domain containing the FrdA E49Q mutation was used for voltammetry experiments using a mixed buffer system at 25°C as previously described (24,31). A pyrolytic graphite edge rotating disk working electrode was used in conjunction with an EG&G model electrode rotator.

Crystallization of FrdA E49Q
Crystallization of the FrdA E49Q variant of E. coli QFR was performed by the hanging drop vapor diffusion method using 0.7 l of protein solution (20 mg/ml, 20 mM Tris, pH 7.4, 0.7% Thesit) and 0.7 l of mother liquor (13.5% polyethylene glycol 5000 monomethyl ether, 200 mM magnesium acetate, 100 mM sodium citrate, pH 5.6, 100 M ethylene diamine tetraacetic acid, and 0.001% dithiothreitol) at a temperature of 20°C. Protein crystals grew after 1-2 days and were cryocooled after 3-5 days. Prior to flash cooling in liquid nitrogen, crystals were quickly dipped into cryosolution containing mother liquor and 30% ethylene glycol.

Diffraction Data Collection and Processing
Diffraction data were collected at the European Synchotron Radiation Facility (ESRF), beamline ID-13, using a wavelength of 0.9537 Å. Radiation damage was minimized by translating the crystals after every three degrees of data collection. Crystals belonged to the orthorhombic space group P2 1 2 1 2 1 with unit cell dimensions a ϭ 96.80 Å, b ϭ 139.53 Å, c ϭ 273.97 Å ( Table 1). Data were processed using DENZO and SCALEPACK (32) and the CCP4 suite of programs (33). In order to improve the completeness, data from two crystals were merged (Table 1).

Structure Solution and Refinement
Since the crystals were isomorphous with previously determined structures (9,34), the structure of QFR (Protein Data Bank code 1KF6), with the Frd Glu-49 side chain truncated to Ala, was subjected to rigid body refinement, geometric minimization, and simulated annealing using CNS (35) and then used as a starting model. Density of the side chain of Gln-49 was observed after this procedure, and the identity was changed to Gln. Model building was performed using the program O (36), and refinement was performed using both CNS (35) and REFMAC (37), whereas the quality of the stereochemical parameters was evaluated with the program PROCHECK (38). The final model of FrdA E49Q has an R cryst of 0.246 and an R free of 0.285. Geometric indicators of model quality are reasonable with root mean square deviation in bond lengths of 0.024, angles of 2.60°, and 74% of residues in the most favored region of the Ramachandran diagram (Table 1)  A, the E. coli SQR (tan) binds oxaloacetate at the active site, whereas the E. coli QFR (teal) has citrate bound at the active site. Only the main chain from the E. coli QFR is shown for clarity. Oxygen atoms are in red, nitrogen atoms are in blue, and carbon atoms follow the color of the molecule. Numbering is for the E. coli QFR with numbers for the E. coli SQR in parenthesis. Structural differences are noted at FrdA Glu-245 and FrdA Arg-287, the latter of which is the presumed proton shuttle of the reaction and is therefore expected to have a higher amount of side chain mobility. WT, wild type.
where F o and F c are the observed and calculated structure factor amplitudes of reflection h, and k is a weighting factor.
where F o and F c are the observed and calculated structure factor amplitudes of reflection h, and k is a weighting factor. T is the test set of reflections. d r.m.s.d., root mean square deviation.

Analytical Methods
FAD content was determined as previously described (42). Protein concentration was determined by the BCA method (Pierce) with bovine serum albumin as a standard in the presence of 1% (w/v) SDS.

Potentiometric Titrations and EPR Spectroscopy
In order to obtain data as representative as possible of the in vivo enzymes, all EPR data reported herein were obtained from preparations of cytoplasmic membranes that had been activated with malonate as described above. Potentiometric titrations were carried out at 25°C in a buffer containing 100 mM Tricine/KOH (pH 8.0), 5 mM EDTA, and 1 mM malonate at 25°C as previously described (43). The following redox mediators were used at a concentration of 25 M: 2,6-dichlorindophenol, 1,2-naphthoquinone, toluylene blue, phenazine methosulfate, thionine, methylene blue, resorufin, indigotrisulfonate, indigocarmine, anthraquinone-2-sulfonic acid, and neutral red. EPR spectra were recorded using a Bruker ESP300E spectrometer equipped with a Bruker liquid nitrogen evaporating cryostat (an ER4111 VT variable temperature unit). Spectra were recorded at a temperature of 150 K (Ϫ123°C) using a microwave power of 20 milliwatts at 9.438 GHz and a modulation amplitude of 2 G pp (gauss, peak-to-peak) at 100 KHz. Five scans were accumulated for each sample. E m values determined by EPR were representative of two to three independent titrations with an S.D. value of approximately Ϯ10 mV.
Potentiometric titration data were analyzed by plotting the intensity of the g ϭ 2.00 peak-trough versus E h and fitting the data to two E m values (44 -46).

RESULTS
Characterization of QFR and SQR-Using site-directed mutagenesis, FrdA Glu-49 was substituted with a Gln or Ala residue, and SdhA Gln-50 was substituted with Glu or Ala. As is the case for the wild type, each of the variants could be expressed to very high levels in the E. coli cytoplasmic membrane and be isolated and purified by standard chromatographic procedures (data not shown) (28,29,31). In addition, both the QFR and SQR mutants retained the covalently bound FAD moiety, which in the case of SQR is at a concentration of 5.5 nmol/mg of protein in an equimolar ratio to the heme b 556 .
Steady State Kinetic Analysis-In order to investigate the effects of substitutions at the FrdA 49-position and SdhA 50-position, the kinetic characteristics for fumarate reduction and succinate oxidation of the mutants were compared and are summarized in Table 2.
SQR Mutants-Alanine substitution of the target residue in SdhA (Gln-50) causes a minimal effect on the catalytic properties of the mutant enzyme. It retains high succinate oxidation and fumarate reductase activity and has no significant alteration of the kinetic parameters for inhibition by malonate or oxaloacetate. Glutamate substitution at the SdhA Gln-50 position, however, severely impairs succinate dehydrogenase activity, with the maximal observed k cat turnover number only 1% of wild type. Succinate oxidation catalyzed by this mutant still retains sensitivity to malonate and oxaloacetate inhibition ( Table 2) but with K i values increased at least 2 orders of magnitude compared with wild type SQR. These data are consistent with a major effect on the binding of dicarboxylate substrates being introduced into the SQR substrate binding site by substituting a charged residue (Glu) for the neutral Gln at SdhA position 50.
QFR Mutants-Elimination of the negatively charged side chain by introduction of Ala at FrdA Glu-49 results in enzyme that catalyzes succinate oxidation and fumarate reduction with a turnover number between 8 and 13% of wild type QFR. The K m values for succinate and fumarate and K i values for malonate and oxaloacetate are similar to wild type in the FrdA E49A mutant.

TABLE 2 Comparison of catalytic parameters for succinate-oxidase (Succ-ox) and fumarate-reductase (Fum-red) reactions catalyzed by E. coli wild type (WT) and mutant SQR enzymes and QFR enzymes
Enzymatic activities were assayed as described under "Materials and Methods" at 30°C, pH 8.0. Glutamine substitution at FrdA Glu-49 results in mutant enzyme that demonstrates typical Michaelis-Menten kinetics in the succinate oxidase reaction and retains 13% of wild type QFR activity. The K m value for succinate and the K i values for malonate and oxaloacetate decrease 5-fold in this mutant. Analysis of the quinol-fumarate reductase reaction, however, reveals a new kinetic property of the mutant enzyme (i.e. strong substrate inhibition by fumarate) (Fig. 2A). In contrast to wild type QFR, the initial rate of the reaction catalyzed by the FrdA E49Q mutant is inhibited dramatically by increasing the fumarate concentration. Maximum activity can only be estimated at low fumarate concentration (about 1 M) and is 0.25% of wild type QFR activity ( Table 2). The K m fum for wild type QFR is 20 M ( Table 2), but this concentration of the substrate strongly inhibits FrdA E49Q activity ( Fig. 2A). The strong substrate inhibition exhibited by this mutant precludes any simple calculation of its kinetic parameters. Fig. 2B shows traces of the catalytic activity of wild type QFR and FrdA E49Q in response to low fumarate concentrations (15 M) in an assay for fumarate reduction using DT diaphorase to continually recycle the quinol (30). With wild type QFR, the reaction is first-order with respect to fumarate concentration. By contrast, menaquinol-fumarate reductase activity of the FrdA E49Q mutant is initially inhibited by fumarate but increases as the substrate is utilized during the course of the assay (Fig. 2B) in the conditions when the stable reducing potential of MQ 1 H 2 is supported by continuous regeneration of menaquinol by DT diaphorase in the presence of NADH.

Succinate oxidation
pH Dependence of the Succinate Oxidase Reaction-The succinate dehydrogenase reactions catalyzed by QFR and SQR are strongly pHdependent with pH profiles similar to those of a theoretical titration curve for deprotonation of a monobasic acid. E. coli SQR and QFR exhibit simple sigmoidal dependences with estimated pK a values of 7.3 and 7.5, respectively (Fig. 3). Both QFR mutants demonstrate an acidic shift in the pK a values to 6.9 for FrdA E49A and 6.6 for FrdA E49Q. The substitution of SdhA Q50A has no detectable effect on the pH profile of the reaction, whereas introducing a negative charge in SdhA Q50E results in a bell-shaped pH profile for succinate oxidation with maximum activity observed at pH 7.0.
Protein Film Voltammetry Studies-In light of the solution assays of the mutant FrdA E49Q, it was desirable to determine if the FrdA E49Q mutant would behave like wild type SdhAB and demonstrate tunnel diode-like behavior (21,23) in protein film voltammetry (PFV) studies. To undertake the PFV studies, the two-subunit (FrdAB) soluble form of the mutant was constructed. Nonturnover voltammograms show sharp, prominent signals dominated by the cooperative two-electron oxidation and reduction of the FAD. At pH 7.0, the FAD reduction potential in FrdA E49Q B is Ϫ115 mV, whereas the value for wild type enzyme is Ϫ88 mV. The pH dependence of the FAD signal from the voltammogram was determined over the range 5.0 -8.5 (Fig. 4A) and produced a linear gradient of Ϫ38 mV/pH unit for FrdA E49Q B (errors in reduction potentials determined by PFV are Ϯ5 mV). This value is similar to the approximation of Ϫ24 mV/pH unit obtained by wild type FrdAB over the pH interval 7.0 -9.5 (24), although the data obtained by wild type FrdAB over this wider pH range reveal curvature suggesting that the "intermediate" gradient actually hides transitions between two-electron/two-proton and two-electron/one-proton reactions. Consistent  After this, the cell was washed, and the enzyme-coated electrode was placed back into the cell with fresh, fumarate-free buffer. The resulting CV is labeled trace, since some fumarate always remains in the cell. Note that between 0.2 and 10 M fumarate, there is increased activity of the enzyme; however, at higher concentrations (100 -1000 M) activity decreases. Conditions were as follows: pH 7.1, 25°C, 20 mM PIPES, 100 mM NaCl, rotation rate 1000 rpm, scan rate 10 mV s Ϫ1 . WT, wild type.
with the solution assays, which show strong fumarate inhibition, only a small catalytic wave is observed in a voltammogram of the FrdA E49Q B variant in the presence of fumarate (Fig. 4B). As the fumarate concentration is increased to 10 M, an increase in current (activity) is seen; however, at higher substrate concentrations, the current decreases, indicating substrate inhibition. After quickly washing the voltammetry cell with buffer, the trace amounts of fumarate remaining are sufficient to restore the higher catalytic activity of FrdA E49Q B. Thus, the substrate inhibition is reversible, and the decrease in current is not due to protein film loss or denaturation effects. Although the FrdA E49Q B mutant demonstrates low activities in both solution assays and PFV studies, the catalytic current is a function of electrode potential as it is for FrdAB. Typical SdhAB tunnel diode behavior, characterized by low activity at low potentials is not observed in the FrdA E49Q B mutant.
In agreement with PFV studies, the solution experiments using the benzyl viologen-fumarate reductase assay revealed that the substitution of FrdA Glu-49 by Ala and Gln does not change the reaction profile; the decrease in activity correlates with a decrease of the driving force (oxidation of reduced BV). Additionally, alteration of SdhA Gln-50 to either Glu or Ala does not eliminate the tunnel diode behavior, suggesting that other residues contribute to mediating this phenomenon (data not shown).
EPR Study of Flavin Semiquinone-In bovine SQR, the FAD semiquinone is clearly detectable by EPR at neutral or basic pH values (45). Spectra of the flavin semiquinone have not been reported for either SQR or QFR from E. coli. Fig. 5A shows representative spectra of redoxpoised EPR samples of membranes enriched in wild type and SdhA Q50E SQR. The spectra are very similar to those reported by Ohnishi et al. (45), exhibiting a line width (peak to trough) of 9 and 10 G for the wild type and SdhA-Q50E mutant, respectively (Fig. 5A). Fig. 5B shows the results of potentiometric titrations of the FAD semiquinone of the wild type and mutant enzyme, generating estimates for the E m of the FAD/FADH 2 couple of Ϫ138 mV (E 1 Ϫ E 2 ϭ 0 mV; K stab ϭ 1.0) and Ϫ190 mV (E 1 Ϫ E 2 ϭ 75 mV; K stab ϭ 18), respectively (Fig. 5B). Thus, the SdhA-Q50E mutation elicits a change in potential (⌬E m ) of Ϫ52 mV for the FAD/FADH 2 couple and increases the amount of FAD semiquinone observable from 33 to ϳ68% of its total concentration.
The FAD semiquinone of E. coli QFR has not been thoroughly investigated by EPR spectroscopy. It was not detected by this technique in studies of a menasemiquinone radical localized at the quinone-binding site (Q p ) (46). The menasemiquinone radical anion is sensitive to the inhibitor 2-n-heptyl-4-hydroxyquinoline-N-oxide and has an E m of approximately Ϫ56 mV at pH 7.2 (46), which is close enough to the reported E m7 of the FAD semiquinone (Ϫ90 mV) (31) for it to interfere with efforts to record the EPR spectrum of the latter. We prevented this interference by including 0.5 mM 2-n-heptyl-4-hydroxyquinoline-Noxide in the redox titration incubation mixture. As seen in Fig. 6A, the amplitude of the free radical signal is significantly increased in the mutant and exhibits a line width of 10 G. The amount of FAD radical in wild type QFR is almost negligible but increases in the FrdA E49Q variant enzyme, which allows determination of the E m of FAD by EPR.

FIGURE 5. EPR and potentiometric characterization of SQR and SdhA Q50E mutant.
A, EPR spectra of wild type SQR-and SdhA Q50E-enriched membranes poised at E h values of approximately Ϫ140 and Ϫ190 mV, respectively. EPR spectra were recorded under the following conditions: temperature, 150 K; microwave power, 20 milliwatts at 9.437 GHz; modulation amplitude, 2 G pp at 100 KHz. B, plots of semiquinone signal intensity versus E h for wild type SQR (triangles) and SdhA Q50E mutant (squares). Data were fit to E m and E 1 Ϫ E 2 values as described under "Results." 100% of the signal intensity corresponds to 33 and 68% as a percentage of total FAD concentration for SQR and SdhA Q50E, respectively. WT, wild type. and Ϫ195 mV, respectively. EPR spectra were recorded under the following conditions: temperature, 150 K; microwave power, 2 milliwatts at 9.437 GHz; modulation amplitude, 2 G pp at 100 KHz. B, plots of semiquinone signal intensity versus E h for the wild type (triangles) and the FrdA E49Q mutant (squares). For the FrdA E49Q mutant, data were fit to an E m of Ϫ185 mV and an E 1 Ϫ E 2 value of ϩ80 mV (K stab ϭ 21.5) as described under "Results." 100% of the signal intensity corresponds to 70% of total FAD concentration for FrdA E49Q. WT, wild type. The FAD semiquinone titrates with an E m of Ϫ185 Ϯ 10 mV at pH 8 ( Fig. 6B), comparable with the value Ϫ153 Ϯ 5 mV obtained by voltammetry under similar conditions (Table 3), and shows increased stability (E 1 Ϫ E 2 ϭ 80 mV; thus, K stab ϭ 21.5) in the FrdA E49Q variant. Interestingly, the FAD semiquinones in the resulting FrdA E49Q and SdhA Q50A proteins show strong similarities in their EPR properties (Figs. 5 and 6). We have not analyzed in detailed the pH dependence of the FAD radical; however, at pH 7.0, the intensity of the radical in the wild type SQR remained at about 70% of that determined at pH 8.0, whereas in the SdhA Q50E variant, the FAD radical concentration is negligible (data not shown), indicating that stabilization of the radical in the mutant involves a change in protonation state of the FAD environment.
Crystallography of the FrdA E49Q Mutant Enzyme-X-ray crystallographic analysis shows that this E49Q variant is folded in a stable conformation and that the fold of each individual domain is similar to that of the wild type. The crystals contain two copies of the QFR complex in each asymmetric unit. The first copy of the protein in the asymmetric unit (chains A-D) exhibits little structural change with respect to the wild type structure. The root mean square deviations between the C␣ atoms of the A-D chains of the FrdA E49Q variant of the QFR complex and wild type are close to the error values that would be expected at this resolution, suggesting that there are no major structural differences between the two proteins.
There are no significant main chain rearrangements visible in or around the active site associated with the altered binding at the FAD. Differences in the FrdA E49Q protein structure are limited to a shift in the position of the E49Q side chain itself and small corresponding rearrangements of the side chains of several active site residues (Fig. 7). During crystallization, citrate from the crystallization buffer could occupy the dicarboxylate binding site of free enzyme or replace oxalo-acetate. Previous structure determinations of E. coli QFR are likely to have citrate bound at the active site, but the density was misinterpreted as cis-oxaloacetate. This can be explained, because QFR as isolated is a mixture of enzyme containing oxaloacetate at the dicarboxylate binding site and free (uninhibited) enzyme (i.e. only some of the enzyme molecules have oxaloacetate bound) (see supplementary Fig. S1). The additional electron density in the FrdA E49Q variant also indicates that citrate is bound at the active site in this variant, and several new hydrogen bonding interactions are formed, one to the side chain of FrdA Glu-245 and one to the N-5 of the FAD (Fig. 7). In addition, a significant alteration of hydrogen bonding at the active site is caused by the side chain guanidine of Arg-287, which is the proposed proton shuttle during fumarate reduction. Both Arg-287 and the N-5 of the FAD are directly involved in catalysis. Taken together, the altered protonation pattern at the active site in the E49Q variant highlights the importance of this side chain.
More substantial are the structural alterations in the second copy of QFR in the asymmetric unit (chains M-P). The density for the soluble domain of this second copy of QFR is of significantly poorer quality than that of the first copy (chains A-D). Despite this, it is clear that the citrate bound at the active site occupies a position distinct from that observed in molecule 1 (chains A-D) and all other structures of complex II homologs determined to date. Shifts in the polypeptide backbone of the soluble subunits are large and are observed over 20 Å away from the active site, resulting in a new conformation of the soluble domain of the enzyme. Further study is needed to resolve the physiological relevance of the conformational differences in the second copy of the FrdA E49Q QFR.

Change in Catalytic Efficiency of the SQR and QFR Mutant Enzymes-
This study was conducted in order to investigate the role of the highly conserved Glu-49/Gln-50 residue found near the dicarboxylate binding site of the flavoprotein subunit of E. coli QFR and SQR. Comparison of available structures of complex II type enzymes (9 -16) does not provide an obvious suggestion for the importance of a negative charge or polar group at this position. One of the prominent differences between SQR and QFR, known as a tunnel diode effect, was observed in several SQR but not QFR enzymes and has been associated with the FAD site of the proteins (21)(22)(23)(24). We anticipated that the Glu/Gln residue ϳ5 Å from the FAD redox center may contribute to the tunnel diode phenomenology. In the present work, this hypothesis was examined by constructing mutants where the Glu/Gln residues were exchanged in the FrdA Glu-49 and SdhA Gln-50 positions. Both kinetic and protein film voltammetry analysis of the mutants exclude an important role of a charge versus polar residue in the target position for the tunnel diode effect. The mutations did, however, dramatically impair the catalytic activity for each enzyme in their respective physiological direction. As a result, the FrdA E49Q mutant becomes a better succinate dehydrogenase than fumarate reductase, and SdhA Q50E becomes more efficient in catalyzing fumarate reduction relative to succinate oxidation. The negative charge within close proximity to FAD and the dicarboxylate binding site may cause significant long range coulombic effects. This may result in changes in enzyme conformation, substrate binding parameters, and altered flavin electronic properties, including changes in reduction potential and stability of the flavinsemiquinone.
The direct comparison of catalytic efficiency of wild type and mutant E. coli SQR and QFR (Table 2) provides an opportunity to understand catalytic differences between members of the SQR/QFR superfamily. In studies carried out using conventional steady-state solution kinetics, SQR was 50-fold more proficient in oxidizing succinate than reducing fumarate (although note that voltammetry on the soluble enzyme suggests that fumarate reduction is effective only when the driving force is very small), whereas QFR was 230-fold more proficient as a menaquinol-fumarate reductase. Kinetic analysis of the alanine variants in both enzymes reveals that these substitutions decrease the catalytic efficiency in both succino-oxidase and fumarate reductase directions, but both enzymes continue to function as a succinate dehydrogenase and fumarate reductase, respectively. The FrdA E49Q and SdhA Q50E substitutions, however, cause dramatic changes in the catalytic properties of the mutant enzymes. The SdhA Q50E substitution results in a 500-fold reduction in the ability of the enzyme to oxidize succinate, but this is accompanied by only a 2-fold reduction in the ability to reduce fumarate. The k cat /K m ratio of succinate oxidase reaction catalyzed by the FrdA E49Q variant remained at 65% of that of the wild type. The major kinetic effect observed in the mutant is strong substrate inhibition by fumarate, attributed only to the fumarate reductase direction, whereas the succinate oxidation reaction retained normal Michaelis-Menten parameters. The change in catalytic efficiency reflects both a decrease of k cat and changes in the ligand binding properties. In the wild type enzymes, dicarboxylate site inhibitors generally have higher affinities for SQR than QFR. In the mutants, there is a change in the relative ligand affinities, as evidenced by changes in the K i for the inhibitors malonate and oxaloacetate ( Table 2). The SdhA Q50E mutant exhibits binding parameters that are remarkably similar to wild type QFR ( Table  2). The converse is observed for the FrdA E49Q enzyme, which has increased affinity for the dicarboxylates such that the K i values are similar to wild type SQR. The differences in affinity of the dicarboxylate site inhibitors for oxidized (succinate oxidase direction) and reduced enzymes (fumarate reductase direction) observed for enzymes from the complex II superfamily suggest structural changes at the substrate binding site upon enzyme reduction; however, no x-ray structures of the reduced form of SQR/QFR homologs are available. Fig. 7 demonstrates that there are some alternations in hydrogen bonding between the citrate dicarboxylate carboxyl group and residues thought to be involved in the proton transfer pathway (FrdA Arg-287 and Glu-245). It is possible that structural changes resulting from the FrdA E49Q substitution do not significantly affect the substrate binding site; however, they may make it somehow more compact, as reflected by decreased binding parameters for the inhibitors. More dramatic changes in the active site structure are expected upon reduction of the variant enzyme, and the second fumarate molecule could be hydrogen-bonded to FrdA Arg-287/Glu-245, thus preventing catalytic turnover and explaining the fumarate inhibition.
Properties of FAD-Both EPR titrations and PFV were used to study the effect of the mutations on the properties of FAD. PFV is very useful for investigating the properties of the redox centers of soluble FrdAB; however, these centers are not clearly resolved in soluble SdhAB. Conversely, stabilized EPR-detectable FAD radicals are observed in wild type SQR but not in QFR. As seen in Figs. 5 and 6, the reciprocal Glu/ Gln mutations in SQR and QFR resulted in increased stabilization of the FAD radical compared with their wild type counterparts. The correlation between EPR line width and optical data on neutral and anionic semiquinone has been previously established (47). The differences in line width of the flavin radicals appear to be a reflection of the fact that these two species of flavin have significantly different hyperfine interaction constants with the various magnetic nuclei of the isoalloxazine ring. Thus, all three EPR-detectable FAD radicals detected here for SQR, SdhA Q50E, and FrdA E49Q appear to be anionic at pH 8.0 based on the narrow EPR signal line width of 9 -10 G. The pK a value of N-5 of the FAD semiquinone in bovine SQR remains controversial. The pH anal-ysis of the line width of the FAD radical clearly suggests that the narrow line width of 12 G observed at both pH 6.1 and 9.1 is indicative of an anionic flavosemiquinone (48). However, in a different EPR study (45), the FAD radical is suggested to be neutral below pH 8.0 and anionic above. It should be noted that the pK a value for N-5 of the FAD semiquinone determined for E. coli QFR by PFV was also suggested to be 8.15 (24).
The reduction potential for the FAD in E. coli SQR has not been previously reported. Table 3 summarizes data on E m8 for FAD in the SQR/QFR enzymes used in this study. E m8 values for the FAD in E. coli SQR and QFR are found to be Ϫ138 and Ϫ122 mV, respectively, at pH 8.0. Interestingly, both variants demonstrate a moderate decrease in FAD potential, Ϫ52 mV in SdhA Q50E and Ϫ31 to Ϫ62 mV in FrdA E49Q. The E m8 values for FAD in FrdA E49Q determined by PFV and EPR methods are in good agreement, and the ϳ30-mV lower value determined by EPR (within experimental error) may reflect the presence of malonate, as has been previously shown for QFR and SQR (24,49). The decrease in E m for FAD in both FrdA E49Q and SdhA Q50E variants may partly explain the decrease in succinate oxidase activity for both mutant enzymes.
The two-electron reduction potential of FAD in bovine SQR shows a Ϫ60 mV/pH dependence below pH 7.7 and Ϫ30 mV/pH above (45). The observed pK a of 7.7 can be assigned to protonation of N-1 in the isoalloxazine ring in the reduced flavin (50). The mechanism of fumarate reduction/succinate oxidation involves hydride transfer from the N-5 position of the isoalloxazine ring of the FAD, implying that N-5 must be protonated in QFR upon flavin reduction to be catalytically competent (18). Cyclic voltammetry experiments with SdhAB from bovine and E. coli SQR suggest a straight line in pH dependence of the FAD midpoint potential with a slope of Ϫ60 mV (2H ϩ /2e Ϫ ) in the pH 6 -8 interval, indicating that the pK a of N-1 for reduced FAD is expected to be also near pH 8.0 (24). E. coli FrdAB, by contrast, showed a linear dependence in the pH range of 5-8 that is close to Ϫ30 mV, corresponding to a 1H ϩ /2e Ϫ ratio and suggesting that N-1 in reduced FAD should be deprotonated above pH 5.0 (24). Thus, changes in the protein environment around the isoalloxazine moiety of FAD affect the properties of the molecule that result in changes in the reduction potential and increased stabilization of the FAD radical.
Hydride transfer between N-5 of FAD and substrate is a key step in catalysis. Based on structural and mutagenesis studies, FrdA Arg-287 is in a suitable position for direct proton donation to C-3 of fumarate (9,12,13,18). The putative proton pathway to FrdA Arg-287 involves FrdA Glu-245 and FrdA Arg-248; however, the proton pathway to reprotonate N-5 of FAD is not obvious. In the W. succinogenes QFR structure determined in the presence of fumarate (12), the residue (FrdA Gln-48) equivalent to FrdA Glu-49 is hydrogen-bonded through its amide nitrogen to a water molecule, which further forms a hydrogen bond to N-5 of the flavin. This would be one potential pathway for reprotonation of the N-5 of the FAD. There is no water molecule found near the flavin N-5 position included in the models of either the E. coli QFR or SQR structures (10,34). Since these two structures were both determined in the presence of dicarboxylate-binding inhibitors and at a resolution where water molecules were added conservatively, it is difficult to evaluate if an ordered water molecule would be located in that position when substrate is bound. Replacement of FrdA Glu-49 with a Gln residue may perturb the binding of a catalytically relevant water molecule and thus affect the reprotonation of N-5. Significant stabilization of the FAD radical in SdhA Q50E at pH 8.0 may indicate altered proton uptake/ release from the reduced FAD upon ionization of Glu-50 that results in an inhibition of the succinate oxidase reaction.
Conclusions-The data presented here show that mutation of FrdA Glu-49 or SdhA Gln-50 has effects on the electronic properties of the covalently bound flavin moiety. Additionally, substrate binding is significantly changed by alteration of these amino acids such that fumarate reductase becomes more like succinate dehydrogenase with glutamine in the FrdA 49-position, and SQR with glutamate in the SdhA 50-position is more efficient as a fumarate reductase. Although redox potentials are known to control the direction of electron transfer, we believe this study also demonstrates that other factors, including coulombic effects investigated here, may be of importance in controlling the direction of the reaction.