Localization of Ubiquinone-8 in the Na+-pumping NADH:Quinone Oxidoreductase from Vibrio cholerae*

Background: The bacterial sodium-translocating NADH:ubiquinone oxidoreductase (Na+-NQR) is a redox-driven Na+ pump. Results: Its NqrA subunit provides the binding site for the final electron acceptor ubiquinone. Conclusion: Ubiquinone binding assigns a first functional role to the peripheral NqrA subunit in the enzymatic mechanism of Na+-NQR. Significance: With the ubiquinone binding site, the binding site for the second substrate has been identified in Na+-NQR. Na+ is the second major coupling ion at membranes after protons, and many pathogenic bacteria use the sodium-motive force to their advantage. A prominent example is Vibrio cholerae, which relies on the Na+-pumping NADH:quinone oxidoreductase (Na+-NQR) as the first complex in its respiratory chain. The Na+-NQR is a multisubunit, membrane-embedded NADH dehydrogenase that oxidizes NADH and reduces quinone to quinol. Existing models describing redox-driven Na+ translocation by the Na+-NQR are based on the assumption that the pump contains four flavins and one FeS cluster. Here we show that the large, peripheral NqrA subunit of the Na+-NQR binds one molecule of ubiquinone-8. Investigations of the dynamic interaction of NqrA with quinones by surface plasmon resonance and saturation transfer difference NMR reveal a high affinity, which is determined by the methoxy groups at the C-2 and C-3 positions of the quinone headgroup. Using photoactivatable quinone derivatives, it is demonstrated that ubiquinone-8 bound to NqrA occupies a functional site. A novel scheme of electron transfer in Na+-NQR is proposed that is initiated by NADH oxidation on subunit NqrF and leads to quinol formation on subunit NqrA.

Vibrio cholerae is a water-borne bacterium that causes the diarrheal disease cholera. In the marine environment, the organism relies on the generation and utilization of a sodium motive force for motility and uptake of certain substrates (1). A central player in the sodium ion cycle across the inner bacterial membrane of V. cholerae is the Na ϩ -pumping NADH:quinone oxidoreductase (Na ϩ -NQR), 4 which transports Na ϩ ions upon the oxidation of NADH by ubiquinone (Q). The Na ϩ -NQR is a respiratory NADH dehydrogenase that is not related to complex I of mitochondria or bacteria. It is composed of six subunits (NqrA-F) and harbors at least five redox carriers: a noncovalently bound FAD and a 2Fe-2S cluster in the peripheral NqrF subunit, an FMN that is covalently bound to the peripheral NqrC subunit, and a second covalently bound FMN plus a non-covalently bound riboflavin found in the membranous NqrB subunit (2)(3)(4)(5). The site for NADH oxidation has been identified in the vicinity of the FAD on NqrF (6). Upon hydride transfer from NADH to FAD, electrons are proposed to be transferred from FAD via the 2Fe-2S cluster to FMN on NqrC, FMN on NqrB, and finally, to riboflavin on NqrB (5,7,8). So far, binding site(s) for ubiquinone in the Na ϩ -NQR have not been localized, but it was proposed that the riboflavin cofactor directly reduces substrate quinone (7). We previously reported that reduction of native membranes from V. cholerae with NADH resulted in the formation of one-electron reduced quinones (ubisemiquinones) (9). The amount of ubisemiquinones observed was influenced by the concentration of the coupling cation, Na ϩ , and we proposed that enzyme-bound ubiquinone-8 (Q 8 ) participates in the overall electron transfer reaction of the Na ϩ -NQR (9). Knowledge of the Q binding site of Na ϩ -NQR is a prerequisite to follow the pathway of electrons through the complex, to describe the mode of Q binding, and to understand how the overall exergonic formation of ubiquinol is linked to the endergonic translocation of Na ϩ across the bac-terial membrane. In the present study, we identify and characterize a quinone binding site on subunit NqrA of Na ϩ -NQR and propose a novel scheme of intramolecular redox events in this respiratory enzyme.
Extraction and Analysis of Quinones-Quinones from V. cholerae membranes were extracted with chloroform/methanol (10). To determine the quinone content of proteins, 1 ml containing at least 0.1 mg of protein was vigorously mixed with 3 ml of methanol and 5 ml of n-hexane for 30 s. Phases were separated by centrifugation (1900 ϫ g, 4°C). The organic phase was collected, and the aqueous phase was extracted thrice with 5 ml of n-hexane. Organic phases were combined, and the solvent was evaporated in a SpeedVac system. The pellet containing quinones was dissolved in ethanol, filtered through a 0.22-m PVDF membrane, and analyzed by HPLC (supplemental methods and supplemental Fig. S1) and mass spectroscopy (supplemental Fig. S2).
Reconstitution of Na ϩ -NQR and NqrA with Quinones-Barquera et al. (11) reported that Na ϩ -NQR from V. cholerae solubilized with 1% lauryldimethylamine N-oxide (LDAO) did not contain Q 8 . We observed that 0.1% LDAO was sufficient to quantitatively remove Q 8 (supplemental Fig. S2). DDM-solubilized Na ϩ -NQR bound to the Ni-NTA affinity column was washed with 10 column volumes of buffer A containing 0.1% LDAO and 5 mM imidazole. Q 8 -free Na ϩ -NQR was eluted with 0.05% DDM and 50 mM imidazole in buffer A. The presence of all six subunits in the Q 8 -depleted Na ϩ -NQR was confirmed by SDS-PAGE. Incorporation of artificial quinones into Q-free Na ϩ -NQR or into NqrA was achieved by binding the proteins to the Ni-NTA column. Ten column volumes of buffer A containing 0.05% DDM, 5 mM imidazole, and 15 M of the respec-tive quinone were passed over the column followed by 10 washing steps with quinone-free buffer. The reconstituted Na ϩ -NQR or NqrA was eluted with 50 mM imidazole.
Photoaffinity Labeling of Na ϩ -NQR and NqrA with Azido-Q Derivatives-Photoreactive azido-Qs possessing a biotin at the terminus of the side chain (see Fig. 1) were recently introduced for the specific labeling of a Q binding site in the alternative, non-electrogenic NADH dehydrogenase (Ndi1) from Saccharomyces cerevisiae (12). The structures of 3-azido-2-methoxy-5-methyl-6-geranyl-1,4-benzoquinone (3-N 3 -Q 2 ), 2-azido-3methoxy-5-methyl-6-geranyl-1,4-benzoquinone (2-N 3 -Q 2 ), and biotinylated 3-N 3 -Q 2 (3-N 3 -Q-biotin) used in this work are shown in Fig. 1 (12). Na ϩ -NQR or NqrA in 10 mM HEPES/ NaOH, pH 8.0, 5% glycerol, 300 mM NaCl, 0.05% DDM was incubated with the quinone or the corresponding Q derivative for 30 -60 min on ice in the dark. In a typical experiment, Na ϩ -NQR (0.16 mg in 0.16 ml of buffer, or 4.8 M) was mixed with 1.5 l of the quinone stock solution (1 mM in ethanol; final concentration in the reaction mixture 9.4 M) in the darkroom. The sample was transferred to a quartz cuvette and illuminated with a 6-watt 365 nm UV lamp (Bachofer-Laboratoriumsgeräte) at 4°C for the indicated times. Biotinylated proteins were detected by Western blot analysis using a Strep-Tactin horseradish peroxidase conjugate according to the procedures described by the supplier (IBA GmbH). The intensity of the NqrA band was quantified with the software Quantity One (Bio-Rad).
Enzymatic Activity-NADH oxidation and quinone reduction by Na ϩ -NQR were followed simultaneously under anoxic conditions at room temperature (4). To investigate the effect of endogenous Q 8 on enzymatic activity, NADH oxidation and Q reduction rates were determined with three preparations of the complex. An aliquot from each preparation was treated with LDAO, and the activities prior to and after removal of Q 8 were compared. To study inactivation by modification with azidoquinones, Na ϩ -NQR (3 M) was incubated with a 100-fold molar excess of 3-N 3 -Q 2 , 2-N 3 -Q 2 , or Q 2 for 1 h at 0°C in the dark under anoxic conditions. Modification was initiated by illumination at 365 nm (4°C). At the indicated time points, aliquots were analyzed for enzymatic activity.
Analytical Methods-Protein quantification, SDS-PAGE and immunochemical detection of His 6 -NqrA were performed as described (5). For MALDI-MS analyses of Na ϩ -NQR subunits, the complex was precipitated by TCA and resuspended in hexafluoropropanol:formic acid (1:2). Aliquots were diluted 1:10 with matrix (saturated sinapinic acid in 0.1% trifluoroacetic acid, 50% acetonitrile) and analyzed in linear mode on a Bruker Ultraflex II TOF-TOF instrument. Circular dichroism spectra were obtained at 25°C with a JASCO 715 spectropolarimeter using quartz cuvettes (Hellma AG) with 1-mm path length. The CD spectra were recorded with 1-nm step width, 20 nm min Ϫ1 scanning speed, 2-nm excitation bandwidth, and 4-s response time. Each spectrum was averaged at least twice, corrected by subtracting the spectrum of the buffer or ubiquinone-1 (Q 1 ) in buffer, and further processed with the Savitzky-Golay filter. Q 1 exhibited a weak CD signal in the region from 185 to 200 nm, which did not interfere with the effects observed upon the addition of Q 1 to NqrA. Secondary structure fractions were calculated by deconvolution of the CD data at the DichroWeb server (13,14) using reference set 4 of the CDSSTR algorithm (15).
Surface Plasmon Resonance-SPR experiments were performed on a Biacore T100 instrument. NqrA was diluted in 10 mM acetate buffer at pH 5.4 to 1 M and coupled to the activated CM5 sensor chip surface (GE Healthcare) by standard procedures. This included blocking of the sensor surface with ethanolamine after immobilization of NqrA. The immobilization level was 16,500 RU. Q 1 and 2-n-heptyl-4-hydroxyquinoline N-oxide (HQNO) were diluted with PBS (10 mM P i , 150 mM NaCl, pH 8.0) from 20 mM stock solutions in DMSO to a concentration of 200 M, resulting in a final DMSO content of 1%. These solutions served as starting points of 1:1 dilution series leading to concentrations of 100.0, 50.0, 25.0, 12.5, 6.25, 3.13, 1.56, 0.78, 0.39, 0.20, and 0.098 M. Biosensor assays were carried out at 298 K, with 1% DMSO in PBS as running buffer, at a flow rate of 10 l/min and a data collection rate of 10 Hz. For each dilution, 85 l of Q 1 or HQNO solution was injected, leading to an association time of 500 s. Dissociation was monitored for 500 s. After every injection, one regeneration injection was applied (1 M NaCl, 1% DMSO, 60 s). Sensorgrams were analyzed with Biacore T100 Control (v1.1), Biacore T100 Evaluation (v1.1), and Origin 8. This included reference cell and buffer subtraction. Each concentration was measured in duplicates from independent dilution series.
Steady-state binding levels obtained from these dilution series were subjected to non-linear least squares fitting within Origin 8. Dissociation constants (K D ) were determined by globally fitting all data points to one-site (RU(c) ϭ RU max ϫ c/(K D ϩc)) and two-site (RU(c) ϭ RU max1 ϫ c/(K D1 ϩc) ϩ RU max2 ϫ c/(K D2 ϩc)) binding isotherms, respectively. In the latter case, RU max1 of the high affinity interaction was optimized first. Subsequently, RU max2 was set to the same value and kept fixed assuming the same maximum response when saturating one or the other binding site.
NMR Spectroscopy-Monomeric NqrA from gel filtration was transferred to D 2 O containing 10 mM potassium phosphate, pH 8.0, 150 mM NaCl, 4 mM NaN 3 (PBS buffer) by repeated (at least 5-fold) ultrafiltration with Ultrafree-4 membranes (Millipore; molecular mass cut-off, 10 kDa). To 10 M NqrA, Q 1 , HQNO, or riboflavin was added from 20 mM stock solutions in DMSO-d6 to a final concentration of 100 M, and the sample was transferred to a 5-mm NMR tube. Controls were prepared with PBS buffer without NqrA. To assign the resonances of Q 1 , HQNO, and riboflavin, 1 H, 1 H-13 C-HSQC, 1 H-13 C-HMBC, and two-dimensional-NOE spectroscopy NMR spectra were recorded at 300 K on a Bruker DRX 600-MHz spectrometer equipped with a 5-mm TXI-H/C/N-triple resonance probe. Spectra were referenced to the residual HDO signal at 4.707 ppm. Chemical shifts are given in supplemental Table S1. One-dimensional saturation transfer difference (STD) NMR spectra (16) were recorded on a Bruker AVANCE III 600-MHz spectrometer equipped with a cryogenic 5-mm TCI-H/C/N triple resonance probe with actively shielded z-gradient. Water suppression was achieved by excitation sculpting (17). Resonances of NqrA were saturated by applying a train of low power Gaussian-shaped pulses at 0.2 ppm with a total saturation time of 2 s. Off-resonance irradiation was set to 33 ppm. On-and off-resonance spectra were acquired in an interleaved manner and subtracted after processing. Up to 4096 transients were collected at a spectral width of 12 ppm. Spectra were processed and analyzed with the software TopSpin (Bruker; v2.1). The percentage of STD effects was determined within the multiple display mode by scaling the off-resonance spectrum down to superimpose with the signal of interest in the difference (off-on) spectrum.

RESULTS
Binding of Quinones to Na ϩ -NQR-The interactions of the Na ϩ -NQR with quinones and quinone derivatives ( Fig. 1) were investigated. We first analyzed the quinone content of this membrane protein complex by HPLC. Na ϩ -NQR purified in the presence of DDM contained 0.61 Ϯ 0.04 mol of Q 8 mol Ϫ1 Na ϩ -NQR (n ϭ 3). A typical HPLC chromatogram is shown in supplemental Fig. S1. The identity of Q 8 extracted from the Na ϩ -NQR was confirmed by ESI-MS analysis (supplemental Fig. S2). Detergent exchange with LDAO completely removed Q 8 from the complex (supplemental Fig. S1) without a significant effect on its Q reduction activity (23.5 Ϯ 3.8 units mg Ϫ1 in the presence of enzyme-bound Q 8 as compared with 19.1 Ϯ 0.2 units mg Ϫ1 in the absence of enzyme-bound Q 8 ; n ϭ 3). Na ϩ -NQR lacking Q 8 showed a slight decrease in NADH oxidation rates (47.9 Ϯ 3.5 units mg Ϫ1 ) as compared with complex containing Q 8 (62.6 Ϯ 0.9 units mg Ϫ1 ). This indicated that Q 1 at a concentration of 100 M in the assay replaced Q 8 , most likely by binding to a Q site of the complex. To test the latter hypothesis, the Q 8 -depleted Na ϩ -NQR, which was bound to a metal affinity column, was allowed to react with Q 1 or Q 2 . Upon elution of the enzyme with buffer containing 0.1% DDM, these short-chain quinones were found to be associated with Na ϩ -NQR at a molar ratio of 0.54, which was comparable with the initial amount of endogenous Q 8 (supplemental Fig. S3, A and B). We also investigated the interaction of Q 8 -depleted Na ϩ -NQR with Quinone Binding to Na ؉ -NQR the Q 2 derivatives 2-N 3 -Q 2 and 3-N 3 -Q 2 , which carry an azido group at the C-2 and C-3 carbon of the quinone head group, respectively, replacing the methoxy group (Fig. 1). Although 3-N 3 -Q 2 was incorporated to a molar ratio (0.51) comparable with the one observed with Q 1 and Q 2 , only 0.10 mol of 2-N 3 -Q 2 was found per mol of Na ϩ -NQR (supplemental Fig. S3C).
Quinone Binding Site on Subunit NqrA of the Na ϩ -NQR-We next used a biotinylated quinone derivative (3-N 3 -Q-biotin, Fig. 1) to identify a subunit of Na ϩ -NQR interacting with quinones. Photocross-linking in the presence of 1-2 equivalents of 3-N 3 -Q-biotin resulted in the specific labeling of a protein with an approximate mass of 50 kDa (Fig. 2). It represented subunit NqrA, as confirmed by immunostaining using an antibody specific for the N-terminal polyhistidine tag. In addition, the biotinylated ␣-subunit of the oxaloacetate decarboxylase from V. cholerae with a mass of 65 kDa was detected (18). To test whether 3-N 3 -Q-biotin competes with Q 8 for binding to NqrA, the photocross-linking of the Na ϩ -NQR complexes purified in DDM (with stoichiometric Q 8 ) and in LDAO (no Q 8 present) was compared. Both the rate of modification and the final yield of biotin-modified NqrA were decreased with Q 8 -containing Na ϩ -NQR by ϳ50% (Fig. 2). This demonstrated that 3-N 3 -Qbiotin interacted with a specific Q binding site of subunit NqrA. Please note that the yield of photolabeling is less than 10% (12). Therefore, a major fraction of Q-sites occupied by 3-N 3 -Qbiotin will not undergo covalent modification.
Mass Shift of NqrA after Modification with 3-N 3 -Q 2 -Na ϩ -NQR, which had been illuminated in the presence of 3-N 3 -Q 2 , was analyzed by MALDI-MS, and its spectrum was compared with a spectrum of Na ϩ -NQR, which was illuminated in the presence of Q 1 (supplemental Fig. S4 and Table 1). Masses assigned to NqrE, NqrD, NqrC with its bound FMN, NqrB devoid of FMN, and NqrF lacking the 2Fe-2S cluster were detected in both samples. In the MS spectra of 3-N 3 -Q 2 -treated Na ϩ -NQR, none of the signals was clearly shifted by 300 Da, as would be expected for covalent modification with azido-Q 2 (19). However, peak j at 50,793 Da observed in Q 1 -treated Na ϩ -NQR, which was assigned to NqrA (calculated mass, 50,660 Da), was not observed in 3-N 3 -Q 2 -treated Na ϩ -NQR. On the other hand, peak k at 50,904 Da, which did not correspond to a calculated mass of any of the Nqr subunits, was shifted to higher m/z (51,020 Da) upon modification with 3-N 3 -Q 2 (supplemental Fig. S4). These results suggested that some modification of NqrA by 3-N 3 -Q 2 had occurred, in accord with the observed labeling of NqrA with biotinylated 3-N 3 -Q 2 . We did not detect a protein with the calculated mass of NqrA, m/z ϭ 50,660 after the control reaction, which was performed in the presence of Q 2 . Obviously, NqrA experienced further chemical modifications either during photolabeling or during the MALDI ionization process. These modifications may very well be different for NqrAϩQ 1 as compared with NqrAϩAzido-Q 2 . Nevertheless, the signals of ϳ50,000 were the only ones affected by our photochemical labeling experiment, indicating that only NqrA was being labeled by the azido-quinone.
Inactivation of Na ϩ -NQR by Photocross-linking with Azido-Q 2 -To investigate the functional role of the quinone localized on subunit NqrA, we studied the effect of covalent modification of the Na ϩ -NQR with azido-Q 2 on enzymatic activity. The specific Q reductase activity of the Na ϩ -NQR, which was allowed to react with 3-N 3 -Q 2 for 60 min, decreased to 55% (9.1 Ϯ 0.2 mol min Ϫ1 mg Ϫ1 ) of the initial value (16.9 Ϯ 0.7 mol min Ϫ1 mg Ϫ1 ) (Fig. 3). Note that the difference in mean activity after a 60-min illumination of the 3-N 3 -Q 2 -

TABLE 1 Masses of protein species observed in Na ؉ -NQR after labeling with 3-N 3 -Q 2
Calculated m/z values are compared with the observed values. Peaks a-l denote species observed in the MALDI-MS spectra of the Na ϩ -NQR, which was incubated with Q 1 in the control reaction (see supplemental Fig. S4).

Measured mass
Predicted protein species Calculated mass a ؉Q 1 ؉ 3-N 3  treated NQR versus the untreated control was statistically significant, with p ϭ 0.0046 (Student's t test, 95% confidence interval). Some minor inhibition (10%) was observed with Na ϩ -NQR illuminated in the presence of Q 2 , which probably was caused by photolysis of flavins (20). HQNO (Fig. 1) is an inhibitor of Na ϩ -NQR that prevents the binding of Q to the enzyme (21). In a parallel reaction performed with a 100-fold molar excess of HQNO, Na ϩ -NQR was inhibited by 45%. Interestingly, 2-N 3 -Q 2 inactivated the Na ϩ -NQR by only 20%, indicating that the azido group at the C-2 position of the quinone head group obstructed the binding of this Q derivative to NqrA. This is in accord with the observation that reconstitution of Q 8 -depleted Na ϩ -NQR was significantly less efficient with 2-N 3 -Q 2 (0.10 mol mol Ϫ1 ) than with 3-N 3 -Q 2 (0.51 mol mol Ϫ1 ) (supplemental Fig. S3, A and B). We conclude that modification of NqrA with 3-N 3 -Q 2 occurs at a site that, in the native enzyme, harbors at least one molecule of Q 8 participating in electron transfer from NADH to Q 1 .
Binding of Quinones to NqrA-We next investigated the interaction of NqrA with Q in the absence of other Nqr subunits. HPLC analyses of organic extracts of purified NqrA revealed only traces of endogenous Q 8 (ϳ0.01 mol of Q 8 per mol of NqrA). This residual Q 8 was removed by treatment with LDAO, and the protein was allowed to react with Q 1 or Q 2 as described for the Na ϩ -NQR complex. Q 1 and Q 2 were incorporated into NqrA at molar ratios of 0.77 and 0.27 mol of quinone per mol of NqrA, respectively. The efficiency of reconstitution was clearly diminished with the more hydrophobic Q 2 . NqrA illuminated in the presence of 2 equivalents of 3-N 3 -Q-biotin was specifically labeled, as demonstrated by Western blot analysis.
Binding of Q 1 to NqrA was further confirmed by circular dichroism spectroscopy (Fig. 4). The addition of 15 M Q 1 to NqrA led to an increase in ellipticity of ϳ210 nm, which could indicate a decrease of the ␣-helical content from 15% in the absence of Q 1 to 7% in the presence of Q 1 . As expected, Q 1 (15 M) exhibited no features in the CD spectrum (Fig. 4). Note that the changes in the CD of ϳ210 nm would also be consistent with a change in interhelical angles in NqrA upon binding of Q 1 (22).

Affinity of Q 1 and HQNO to
NqrA-To support our results obtained by photochemical labeling and CD spectroscopy, we determined the affinity of the NqrA subunit toward ubiquinone Q 1 and to the well known inhibitor HQNO by surface plasmon resonance (Biacore). For this purpose, NqrA was immobilized on a CM5 sensor chip, and Q 1 and HQNO served as soluble analytes. This assay setup allows for characterizing the interaction of chemically unmodified small molecule ligands with NqrA; however, for sensitivity reasons, very high immobilization levels of the NqrA protein were necessary. Even with 16,500 RU of immobilized NqrA, we obtained maximum responses ϳ9 -17 RU (for one binding site). For a soluble analyte of 250-Da molecular mass, the maximum theoretical response would be 80 RU. According to this, approximately one-fifth to one-ninth of NqrA was accessible to ubiquinone and functional in our SPR assay. The sensorgrams of both Q 1 and HQNO indicate significant binding to NqrA already at concentrations below 1 M (Fig. 5, A and C). The high immobilization level of NqrA precludes kinetic analysis of the sensorgrams because of mass transport limitation and rebinding effects. We determined the affinities of Q 1 and HQNO from equilibrium binding levels at concentrations ranging from 100 nM to 50 M (Fig. 5, B and D). Sensorgrams at higher analyte concentrations did not reach equilibrium binding levels in reasonable time and were obviously affected by nonspecific interactions, not uncommon for hydrophobic analyte molecules like Q 1 and HQNO.
When fitting the full data range to a one-site binding model, rather poor regression coefficients are obtained for both Q 1 (R 2 ϭ 0.90) and HQNO (R 2 ϭ 0.74), reflecting additional, low  affinity interactions that become prominent at concentrations above 3-6.5 M (supplemental Fig. S5, A and B). We therefore decided to fit our SPR data to a two-site binding model that yielded excellent regression coefficients (Fig. 5, B and D; R 2 ϭ 0.97 and 0.94, respectively). According to this, Q 1 entertains a high affinity interaction with NqrA with a K D1 ϭ 124 Ϯ 13 nM and is involved in additional low affinity binding with a K D2 ϭ 36 Ϯ 5 M. The inhibitor HQNO has a K D1 ϭ 100 Ϯ 24 nM and a K D2 ϭ 32 Ϯ 5 M. In conclusion, surface plasmon resonance experiments indicate high affinity binding of Q 1 (and HQNO) to NqrA, in agreement with our results from photochemical labeling.
Structural Determinants of Q 1 for Binding to NqrA-To validate our results from photolabeling and SPR and to identify critical contacts between Q 1 and its binding site on NqrA, we applied STD NMR spectroscopy. To exclude artifacts brought about by misfolded or aggregated NqrA, it was important to analyze the oligomerization state of the protein subjected to NMR analysis. Size exclusion chromatography of NqrA separated NqrA aggregates with a mass Ͼ600 kDa from monomeric NqrA with an apparent mass of 49 kDa (Fig. 6). The SDS-PAGE of NqrA aggregates and monomeric NqrA is shown in supplemental Fig. S6. With monomeric NqrA, Q 1 showed STD signals between 8.5 and 3.8% (Fig. 7). The strongest signals originated from the two methoxy groups (8.5%) followed by the two terminal methyl groups of the prenyl side chain (7.6 and 6.2%, respectively). The weakest signal resulted from CH 2 -7 of the prenyl side chain (3.8%). This suggests a slightly tighter contact of the quinone headgroup and the terminus of the prenyl side chain as compared with the internal part of the prenyl group. Even stronger STD effects (maximum 17.1%) were observed with NqrA and the inhibitor HQNO (Fig. 8), indicative of a faster exchange kinetic of HQNO as compared with Q 1 . The strongest signals were attributed to the terminal methyl group of the alkyl chain (CH 3 -17, 17.1%) and several protons of the hydroxyquinoline headgroup (H5, 14.5%; H8, 13.1%; H7, 11.7%; H3, 10.4%). The internal portion of the alkyl chain showed significantly weaker STD effects (CH 2 -11 to CH 2 -15, 4%). This pattern of STD effects is consistent with a similar binding mode of HQNO and Q 1 in the same binding site of NqrA. No STD NMR effects were observed with Q 1 and HQNO in the absence of NqrA (supplemental Fig. S7). Riboflavin is a cofactor located on subunit NqrB (5). As expected, a control experiment performed with NqrA and riboflavin did not reveal STD NMR signals (supplemental Fig. S8). FIGURE 5. Interaction of Q 1 and HQNO with NqrA determined by surface plasmon resonance affinity assays. A and B, sensorgrams (A) and steadystate binding levels (B) of Q 1 binding to immobilized NqrA. C and D, sensorgrams (C) and steady-state binding levels (D) of HQNO binding to immobilized NqrA. For each analyte, sensorgrams and binding levels from two independent dilution series are shown. Equilibrium binding levels were fitted to a two-site binding isotherm that accounts for additional low affinity binding of Q 1 and HQNO to NqrA present at higher concentrations.  Upper and lower panels, reference (upper panel) and STD NMR (lower panel) spectra of Q 1 in the presence of NqrA (molar ratio ϭ 10:1). The reference spectrum was scaled to 10% of its original intensity. Signals of the same intensity in the STD spectrum corresponded to an STD effect of 10%. The signal of proton H8 (Fig. 1) was largely suppressed by the excitation sculpting scheme used to suppress the residual HDO resonance and could not be quantified reliably. Impurities and solvents signals are marked with asterisks. rel, relative.

DISCUSSION
The catalytic NADH binding site of Na ϩ -NQR is located on subunit NqrF. Here we show that the second substrate, the quinone, binds to the peripheral NqrA subunit of Na ϩ -NQR (Fig. 9). Removal of Q 8 did not destabilize the Na ϩ -NQR, which could be reconstituted with short-chain quinone analogues, suggesting that binding of Q to the complex is mainly conferred by interactions of the redox-active Q head group with NqrA. Endogenous Q 8 present in Na ϩ -NQR prevented the modification of NqrA with 3-N 3 -Q-biotin, demonstrating that this biotinylated Q derivative occupied a functionally relevant Q site in the complex. Substitution of the methoxy group at the C-3 position of Q 2 by an azido group did not affect binding to NqrA, whereas the same substitution at the C-2 position strongly reduced binding efficiency (Fig. 1). This demonstrates that the substituent at the C-2 position is crucial for recognition of the quinone. Contacts between the methoxy groups at the C-2 and the C-3 positions of Q 1 with amino acid residues of NqrA were also confirmed by STD NMR spectroscopy (Fig. 7). The precise location of the quinone binding site on NqrA remains to be identified. We should also consider that Na ϩ -NQR might contain an additional Q binding site that is distinct from the catalytic site identified in the present study (23). Our SPR assays point into this direction as the steady-state binding levels fitted best to a two-site binding isotherm. However, the K D2 values of the secondary interaction in the 30 M range must be considered as rough estimates as we could obtain acceptable sensorgrams only at concentrations below 25-50 M.
The affinity of isolated NqrA for Q 2 was clearly lower than for Q 1 , whereas Q 1 and Q 2 were equally well incorporated into Na ϩ -NQR. Thus, it has to be assumed that in the holo-complex, additional Nqr subunits contribute to the binding of Q, most likely by interacting with the hydrophobic isoprenoid tail. NqrA is a peripheral subunit of the complex that does not contain transmembrane helices. It only binds to the membrane in the presence of other Nqr subunits (5). The naturally occurring Q 8 is a hydrophobic molecule residing in the membrane bilayer.
This raises the question as to how Q 8 accesses its binding site on NqrA. The subunit might be in direct contact with the membrane via a hydrophobic surface patch, as was proposed for the alternative NADH:quinone oxidoreductase (NDH-2), which mediates membrane contact via amphipathic helices (24). We favor another scenario where additional Nqr subunits interact with the hydrophobic isoprenoid side chain of quinone. A likely candidate is the membrane-bound NqrB subunit because NqrA forms stable subcomplexes with subunits NqrBCF (4) and NqrBDE (5,25). NqrB was proposed to interact with quinone based on a mutagenesis study with Na ϩ -NQR using the Q-reductase inhibitor korormicin (26). A direct interaction of NqrA and NqrB would also be consistent with hypotheses on electron transfer routes within the complex (2,4,7,27) assuming direct transfer of electrons from the riboflavin bound to NqrB (5) to substrate quinone (Fig. 9). It remains to be investigated how the ultimate electron transfer step from riboflavin on NqrB to ubiquinone-8 on NqrA determines the transport of Na ϩ by the membrane-bound subunits of the Na ϩ -NQR.