Direct Evidence for Nitrogen Ligation to the High Stability Semiquinone Intermediate in Escherichia coli Nitrate Reductase A*

The membrane-bound heterotrimeric nitrate reductase A (NarGHI) catalyzes the oxidation of quinols in the cytoplasmic membrane of Escherichia coli and reduces nitrate to nitrite in the cytoplasm. The enzyme strongly stabilizes a menasemiquinone intermediate at a quinol oxidation site (QD) located in the vicinity of the distal heme bD. Here molecular details of the interaction between the semiquinone radical and the protein environment have been provided using advanced multifrequency pulsed EPR methods. 14N and 15N ESEEM and HYSCORE measurements carried out at X-band (∼9.7 GHz) on the wild-type enzyme or the enzyme uniformly labeled with 15N nuclei reveal an interaction between the semiquinone and a single nitrogen nucleus. The isotropic hyperfine coupling constant Aiso(14N) ∼0.8 MHz shows that it occurs via an H-bond to one of the quinone carbonyl group. Using 14N ESEEM and HYSCORE spectroscopies at a lower frequency (S-band, ∼3.4 GHz), the 14N nuclear quadrupolar parameters of the interacting nitrogen nucleus (κ = 0.49, η = 0.50) were determined and correspond to those of a histidine Nδ, assigned to the heme bD ligand His-66 residue. Moreover S-band 15N ESEEM spectra enabled us to directly measure the anisotropic part of the nitrogen hyperfine interaction (T(15N) = 0.16 MHz). A distance of ∼2.2 Åbetween the carbonyl oxygen and the nitrogen could then be calculated. Mechanistic implications of these results are discussed in the context of the peculiar properties of the menasemiquinone intermediate stabilized at the QD site of NarGHI.

In the absence of oxygen and in the presence of nitrate, Escherichia coli induces the production of two energy-converting enzymes: formate dehydrogenase-N (FdnGHI) and dissimilatory nitrate reductase A (NarGHI). 3 These two complexes cooperate in generating a proton motive force through the redox loop mechanism as originally envisaged by Peter Mitchell in his chemiosmotic hypothesis (1). The separation of positive and negative charges across the cytoplasmic membrane is achieved through electron transfer from the formate oxidation site in the periplasm to the nitrate reduction site located on the cytoplasmic space with mena-or ubiquinone/quinol cycling between them (2,3).
The heterotrimeric NarGHI complex is composed of (i) a nitrate-reducing subunit NarG containing a Mo-bis-MGD cofactor (Moco) and a [4Fe-4S] cluster FeS0 with unusual His coordination (4,5), (ii) an electron transfer subunit NarH carrying four FeS clusters (6), and (iii) a membrane anchor subunit NarI containing two b-type hemes termed b D and b P to indicate their distal and proximal positions with respect to the nitratereducing site (7)(8)(9). These prosthetic groups define an electron transfer pathway from a periplasmically oriented quinol oxidation site (Q D ) to a cytoplasmically oriented nitrate reduction site.
Although considerable research has been devoted to nitrate reductase functioning, only partial information has been gained about the number, the structure, and the specificity of quinol binding sites within NarGHI. For instance, some studies have reported on mena-and ubiquinol analogue binding (10 -20), whereas the molecular details of the interaction between natural quinols and NarGHI remain to be established.
Despite the absence of bound quinones in the available high resolution structures of NarGHI (7,20), the crystal structure of the enzyme in complex with pentachlorophenol (PCP), an inhibitor of the quinol oxidase activity, has been determined (20). Based on mutagenesis data, biochemical analyses, and molecular modeling, a model of the Q D quinol binding site has been proposed (20). In this working model, a quinone carbonyl group interacts with the protein via a hydrogen bond to a histidine residue (His-66), which is one of the axial ligands of heme b D . Molecular modeling of a menaquinone in the PCP binding site suggested that the opposite carbonyl group could form a hydrogen bond to Lys-86 (20). Additionally, an elongated hydrophobic cavity where both hemes are exposed was proposed as a possible secondary quinone reactive site (7,20).
We have shown that a semiquinone (SQ) intermediate can be stabilized in wild-type NarGHI by redox titration monitored by EPR spectroscopy (18). The radical cannot be detected in the presence of an excess of 2-n-nonyl-4-hydroxyquinoline-N-oxide (NQNO), a menaquinol analogue that binds in the vicinity of the distal heme. The radical has a midpoint potential E m,7.5 (Q/QH2) ϳϪ90 mV, characteristic of a menasemiquinone (MSQ) species. Remarkably, the MSQ stability constant measured in NarGHI (K s ϳ70) is the largest measured so far in respiratory complexes stabilizing semiquinone species (21)(22)(23). Interestingly, although this semiquinone species is present in a mutant lacking the proximal heme b P , it cannot be detected in variants lacking the distal heme or upon substitution of the surrounding Lys-86 residue by an Ala (NarGHI K86A ) (19). The latter variant exhibits a strong decrease in quinol oxidase activity whatever the tested quinol analogs are (19,20). On the basis of these observations, we concluded that the menasemiquinone radical is located in the vicinity of heme b D (hereafter referred to as MSQ D ) and that Lys-86 is required for its stabilization (19). However, the molecular factors responsible for the high MSQ D stabilization and the role of this stability during the enzyme mechanism remain to be elucidated.
Understanding the quinol oxidation mechanism by NarGHI requires identification of ligands for all three forms, Q, SQ, and QH 2 . In particular, interactions between quinone carbonyl oxygen atoms and H-bond-donating amino acid residues have been suggested to be an important factor in determining the properties of bound semiquinones (24). In this work, the Q D quinol oxidation site of NarGHI was studied by high resolution EPR methods using the unpaired electron spin of the endogeneous MSQ D intermediate as a probe of its environment. High resolution pulsed EPR techniques such as ESEEM (electron spin echo envelope modulation) and its two-dimensional variant HYSCORE (hyperfine sublevel correlation) are powerful tools for probing the local environment of protein-bound radicals at the atomic level. In particular, they have been shown to be useful for detecting weak hyperfine interactions between nuclear spins of 14 N atoms and the unpaired electron spin of protein-bound semiquinones (24 -31). Application of multifrequency techniques makes it possible to interpret the ESEEM features and to utilize them to reliably measure hyperfine and quadrupole interactions in orientationally disordered samples such as frozen protein (27,32,33). Nuclear quadrupole parameters can then be used to distinguish nitrogen originating from the peptide bond or from amino acid side chains (25, 27-31, 34 -36).
In this work, S-band (ϳ3 GHz) and X-band (ϳ9 GHz) ESEEM/HYSCORE experiments were performed directly on E. coli membrane fractions containing overexpressed NarGHI. The use of native and uniformly 15 N-labeled proteins allow us to provide the first direct experimental evidence of nitrogen ligation to the MSQ bound within the Q D site of NarGHI in its natural environment. It is identified as the N ␦ imidazole nitrogen of a histidine residue, most likely heme b D ligand His-66.

EXPERIMENTAL PROCEDURES
Sample Preparation-Purified E. coli NarGHI-enriched membrane fractions were prepared as described in Ref. 19. 15 N uniformly labeled NarGHI was obtained after growth in M9 minimal medium (37) supplemented with Mohr salt (25 mg/liter), Na 2 MoO 4 (220 ng/liter), and [ 15 N]NH 4 Cl (1 g/liter). Glycerol (0.4% v/v) was used as a carbon source instead of glucose. Thiamine (10 mg/liter) and appropriate antibiotics were added before inoculation, and all cultures were performed at 37°C under semi-anaerobic conditions. Stabilization of the semiquinone at the Q D site was achieved through redox titrations under the same conditions as those used in our previous work (18). Redox potentials are given in the text with respect to the standard hydrogen electrode.
Multifrequency Pulsed EPR Experiments-X-band (ϳ9 GHz) pulsed EPR experiments were performed using a Bruker EleXsys E580 spectrometer equipped with an Oxford Instruments CF 935 cryostat. S-band pulsed EPR experiments were carried out on a home-built pulsed S-band EPR spectrometer (27,38) at a microwave frequency of ϳ3.4 GHz, which corresponds to the resonance field ϳ122 mT for paramagnetic species with a g-factor of ϳ2.
For the two-pulse field sweep ESE (electron spin echo) spectra, the echo integral was recorded as a function of the magnetic field after a two-pulse Hahn-echo sequence (/2 -200 ns -). In the one-dimensional three-pulse ESEEM experiment (/2 --/2 -T-/2), the intensity of the stimulated echo signal after the third pulse is recorded as a function of time, t, at constant time, . 2D-HYSCORE spectra were recorded using the sequence (/2 --/2t 1 -t 2 -/2), where the inverted three-pulse echo generated at a time after the last pulse is measured as a function of t 1 and t 2 .
All ESEEM and HYSCORE spectra were recorded at a magnetic field corresponding to the maximum intensity of the MSQ D signal in the two-pulse field sweep ESE spectrum, where all orientations of the semiquinone with respect to the external magnetic field contribute, giving rise to powder ESEEM/HYSCORE spectra. Appropriate phase cycling procedures were applied in the stimulated echo and HYSCORE experiments to remove the unwanted echoes (39). Spectral processing of three-pulse and HYSCORE spectra was performed using Bruker Xepr software.
Relaxation decays were subtracted (fitting by polynomial functions) followed by zero-filling and tapering with a Hamming window, before Fourier transformation, which finally gives the spectrum in frequency domain. All spectra are shown in absolute value mode. HYSCORE spectra are presented as contour plots.
Hyperfine and Nuclear Quadrupole Interactions-A hyperfine coupling between a S ϭ 1 ⁄ 2 radical and a nucleus with nuclear spin value I consists in general of (i) the isotropic contribution A iso ϭ 2 0 g e g n ␤ e ␤ n 0 (0) 2 /3h where 0 (0) 2 is the electron spin density at the nucleus, g e and g n are electron and nuclear g-factors, respectively, ␤ e and ␤ n are Bohr and nuclear magnetons, respectively, h is Planck's constant, and (ii) the anisotropic contribution described by the dipolar coupling ten-sor T . In the point-dipole approximation T has axial symmetry with principal values (A iso -T, A iso -T, A iso ϩ 2T), T ϭ 0 g e g n ␤ e ␤ n /4hr 3 where r is the distance between the nucleus and the electron.
The hyperfine couplings of different isotopes of the same element are proportional to a very good approximation to the corresponding g n values. In this study, the parameter A iso of the nitrogen interacting with the SQ was derived from X-band 14 N and 15 N ESEEM measurements while T was directly determined from S-band 15 N ESEEM. A 14 N nucleus has a quadrupole moment that interacts with the electric field gradient (EFG) at the nucleus. The components of the EFG tensor are defined in its principal axis system and are ordered according to q ZZ Ն q YY Ն q XX . This traceless tensor can then be fully described by only two parameters: (i) the 14 N nuclear quadrupole coupling constant ϭ e 2 q ZZ Q/ h , where e is the charge of electron, Q is the nuclear electric quadrupole moment of the 14 N nucleus, (ii) the asymmetry parameter ϭ q YY -q XX /q ZZ . is a measure of the strength of the interaction between the nuclear quadrupole moment and the EFG at the 14 N nucleus site due to anisotropic charge distribution in the system, whereas is a measure of the deviation of the charge distribution from axial symmetry. and are extremely sensitive to atomic and chemical bond arrangements and are excellent probes for the identification of bonding geometry and of the chemical group housing the nucleus. They can, therefore, be used for the identification and characterization of the 14 N atom. In this study, the parameters and of the nitrogen interacting with MSQ D were derived using S-band 14 N ESEEM. 14 N and 15 N Powder ESEEM Spectra-Powder ESEEM spectra given by a frozen protein solution for S ϭ 1 ⁄2 interacting with a nucleus with I ϭ 1 ⁄2 (e.g. 15 N) and I ϭ 1 (e.g. 14 N) have been extensively analyzed (40). For I ϭ 1 ⁄2, it has been shown that the relative magnitude of the hyperfine and nuclear Zeeman interactions determines the shape of the ESEEM spectra. In the case of an axial hyperfine tensor for 4 I ϭ 2A iso ϩ T known as "matching condition" ( I is the nuclear Larmor frequency), the ESEEM spectrum has a characteristic shape; i.e. it is characterized by two frequencies with a dominant narrow feature at ␣ ϭ 3 T /4 and a broad, weak one at ␤ ϳ 2 I . A modest rhombicity of the hyperfine coupling influences the width but not the position of the intense low frequency line at ␣ . Thus the determination of T from this ESEEM frequency is not affected by the rhombicity of the hyperfine interaction (32).
For hyperfine couplings that deviate from the matching condition, the intensity and resolution of spectral features decreases significantly. For weak couplings A iso Ͻ 2 I the main peaks are centered approximately at I . A 14 N nucleus interacting with an unpaired electron spin S ϭ 1 ⁄ 2 can produce up to six lines in an ESEEM spectrum. They correspond to the 6 nuclear transition frequencies between three nuclear energy sublevels in each of the two electron spin manifolds with m s ϭ Ϯ 1 ⁄ 2. Because of their different dependence on orientation, not all transitions contribute equally to the spectra in ESEEM measurements of disordered samples (as in the case of the frozen suspensions of NarGHI-enriched membrane fractions used in our experiments). The ESEEM spectrum expected for 14 N with predominantly isotropic hyperfine coupling constant A iso is governed by the ratio between the effective nuclear frequency in each manifold efϮ ϭ I Ϯ A iso /2, and the quadrupole coupling constant, .
If efϮ / ϳ 0, i.e. A iso Х 2 I , the nuclear Zeeman and hyperfine splitting cancel in one electron spin manifold. This situation is thus called "cancellation condition." Therefore, the three nuclear frequencies from this manifold are close to the pure nuclear quadrupole resonance frequencies of 14 N transitions and are given by the relationships in Equation 2, In this case, three sharp lines at the frequencies given by Equation 2 are present in the powder ESEEM spectra with the simple additive relationshipϩ 0 ϭ ϩ . Pure quadrupole frequencies can even be observed if cancellation is not exact, as long as the condition in Equation 3, is fulfilled (33). When efϮ / Ͼ 1, only a single line without pronounced orientation dependence from each corresponding manifold is expected. It is actually a double-quantum transition line ⌬m I ϭ 2 giving maximum intensity at a frequency well described by Equation 4, The two single quantum transitions usually do not show any resolved peaks because of significant orientation dependence from quadrupole interaction. The relative strength of the hyperfine coupling and the Zeeman nuclear frequency can be varied by changing the frequency of the microwave radiation used for the ESEEM/HYSCORE measurements. In this work, this strategy was applied to reach the cancellation condition ( 14 N) or the matching condition ( 15 N) for the detected nucleus at S-band frequencies.

X-band Pulsed EPR of Native and 15 N-Labeled NarGHI-
Our previous studies revealed that an EPR-detectable menasemiquinone intermediate can be generated in the wild-type nitrate reductase A by direct chemical reduction of purified E. coli NarGHI-enriched membrane fractions (18), the maximal SQ concentration being reached at a redox potential around Ϫ100 mV at pH 7.5 (18). To specifically characterize by pulsed EPR the interaction of the semiquinone with nearby nitrogen atoms, MSQ D was generated either in unlabeled or in uniformly 15 N-labeled protein sample (see "Experimental Procedures"). In both preparations issued from bacteria grown in a modified M9 minimal medium, the semiquinone concentration relative to the [3Fe-4S] cluster (ϳ6 -8%) and the quinol: nitrate oxidoreductase activity were identical to those measured in a preparation obtained from bacteria grown in Terrific Broth medium.
Although several FeS clusters are paramagnetic at Ϫ100 mV and give signals in the spectral region where MSQ D appears (g ϳ 2.0) (Fig. 1A) (41), the radical can be specifically detected without contamination from the other centers by exploiting differences in the temperature dependence of the relaxation time of these species. Thus, at 90 K, the FeS clusters are too fast relaxing to contribute to the EPR spectrum. This is illustrated in Fig. 1B, which shows the two-pulse field-sweep ESE spectrum, measured at 90 K, of membrane fractions containing overproduced native NarGHI (with the 99.63% of 14 N natural abundance) ( Fig. 1B and inset, top) or the 15 N-labeled protein (inset, bottom). In both samples, the single line of SQ is the only one present and has identical line width ϳ0.8 mT (see inset in Fig.  1). This is expected because the major contributions to the observed line width are (i) unresolved g-tensor anisotropy and (ii) unresolved hyperfine couplings to the many protons in the vicinity of the radical (18). The line width is identical to that of the CW EPR spectrum of the native enzyme (18). In particular, magnetic interactions smaller than 0.5 mT (in field units) are not resolved on these spectra.
X-band HYSCORE-To resolve weak 14 N or 15 N magnetic interactions with the semiquinone, high resolution pulsed EPR methods were used at 90 K. Three-pulse ESEEM spectra of the radical have been recorded and have revealed low frequency modulations of the electron spin echo decay indicative of a hyperfine interaction with nuclei having low gyromagnetic ratio such as 14 N (data not shown). To facilitate the interpretation of the spectra and to increase their resolution, 14 N and 15 N HYSCORE spectra were recorded using different values to avoid peak suppression effects (i.e. blind spots) (39). A typical 14 N HYSCORE spectrum of MSQ D showing the low frequency part of the two-dimensional plot after data processing is displayed in Fig. 2A. It exhibits two intense off-diagonal crosspeaks 1 in the (ϩϩ) quadrant, correlating nuclear transition frequencies at 2.1 and 3.4 MHz.
A HYSCORE spectrum of MSQ D measured with the 15 Nlabeled enzyme is depicted in Fig. 2B. The spectrum contains a prominent peak 3 on the diagonal at the 15 N Zeeman frequency ( I ( 15 N) ϭ 1.48 MHz) and two crosspeaks 2 centered symmetrically with respect to this diagonal peak, with maxima at frequencies 0.98 and 2.05 MHz. The loss of crosspeaks 1 observed in unlabeled NarGHI is thus related with the appearance of crosspeaks 2 in the 15 N-labeled protein. No other correlation could be resolved in HYSCORE spectra measured at 90 K using other values. These observations allow an unequivocal assignment of crosspeaks 1 & 2 to a single nitrogen nucleus. Their circular contour line shape suggests that the corresponding nuclear transitions are weakly anisotropic. A 15 N nucleus has a spin 1 ⁄ 2, and so does not possess the nuclear quadrupole moment that affects a 14 N nucleus (I ϭ 1). Thus, a 15 N nitrogen hyperfine coupling of ϳ1.1 MHz can be directly estimated by reading the splitting between crosspeaks 2. The intense diagonal peak 3 is not seen in 14 N spectra. It is contributed by all the other 15 N nuclei around the SQ, which are involved only in weak interactions with the unpaired electron (27,29,34,42).
The observation of only two narrow crosspeaks 1 correlating 14 N nuclear transition frequencies belonging to opposite electron spin manifolds indicates a 14 N hyperfine interaction with both ratios efϮ / Ͼ 1. These peaks could, therefore, be assigned to the double-quantum transitions dqϩ ϭ 3.   S-band ESEEM and HYSCORE-Unambiguous determination of and is required for complete description of the nuclear quadrupole interaction tensor and the assignment of the interacting nitrogen. These values can be determined from ESEEM experiments satisfying the cancellation condition for one of the electron spin manifold. For hyperfine coupling A iso ( 14 N) ϳ0.8 MHz, the cancellation condition is reached at I ( 14 N) ϳ0.4 MHz, which requires an experiment at S-band with microwave frequency ϳ3 GHz. Fig. 3A (top) displays a representative S-band three-pulse ESEEM spectrum of MSQ D in native NarGHI. Three sharp peaks are observed at frequencies 0.49, 1.22, and 1.71 MHz. The additive relationship satisfied by these frequencies ( ϩ ϭ 0 ϩ -) as well as the typical increase in the electron spin echo envelope modulation depth measured in the time-domain spectrum (not shown) reflect the situation where the hyperfine coupling is nearly equal to twice the nuclear Larmor frequency. This allows for an immediate assignment of the narrow peaks to three nuclear quadrupole interaction frequencies from the manifold with ef-ϳ0 and thus the determination of the nuclear quadrupole interaction tensor parameters. Applications of Equation 2 using 0 ϭ 0.49 MHz, -ϭ 1.22 MHz, and ϩ ϭ 1.71 MHz lead to ϭ 0.49 MHz and ϭ 0.50. The corresponding calculated value of c ϭ 0.78 MHz 2 using Equation 5 is then fully consistent with our assignment of the nuclear frequencies measured on the X-band 14 N HYSCORE spectrum and with the analytic procedure used. The double-quantum transition of the opposite manifold is difficult to detect in the one-dimensional S-band 14 N ESEEM spectra taken at several values (Fig. 3A, top and data not shown). It is however expected to have a broader shape because of combined hyperfine and nuclear quadrupole interaction anisotropies. From Equation 4, it is predicted to appear at about ϳ2.3 MHz. We were however able to detect crosspeaks having an elongated shape in the S-band 14 N HYSCORE spectrum of MSQ D in native NarGHI, correlating the two double-quantum nuclear frequencies from each manifold at 1.71 and ϳ2.3 MHz (Fig. 3B), showing the full consistency of our assignment.
The lack of extended anisotropic contours in the crosspeaks along the line normal to the diagonal of the X-band 14 N and 15 N HYSCORE spectra limits the accuracy of the determination of the anisotropic part of the hyperfine tensor. To overcome this limitation and gain information on the location of the interacting nucleus, S-band ESEEM experiments have been carried out on 15 N-labeled NarGHI to fulfill the matching condition for the interacting nitrogen nucleus. In this case, the spectrum is not complicated by the contribution of the nuclear quadrupole interaction and allows straightforward evaluation of the anisotropic part of the hyperfine interaction. Indeed, our previous measurement of A iso ( 15 N) ϭ 1.07 MHz implies that the condition A iso ( 15 N)/2 ϳ I ( 15 N) is valid (at 122.2 mT, I ( 15 N) ϳ 0.57 MHz). Fig. 3A (bottom) shows a representative frequency domain three-pulse ESEEM spectrum of MSQ D in 15 N-labeled NarGHI recorded at 3.4 GHz. The dominating feature is a peak at 0.12 MHz that we assign to the frequency ␣ of the 15 N nucleus. A second peak is present at frequency ϳ0.56 MHz, which matches the 15 N Larmor frequency. It arises from weakly coupled distant 15 N nuclei, as previously discussed. The partner of the line ␣ is not detected in our S-band ESEEM spectra taken at several values. This is due to its expected low intensity (39) and the limited achievable signal-to-noise ratio. According to this analysis, we have 3 T( 15 N) /4 ϳ0.12 MHz which yields T( 15 N) ϳ0. 16 MHz or T( 14 N) ϳ0.11 MHz when rescaled to 14 N nucleus. This value is in full agreement with our previous analysis of 14 N and 15 N HYSCORE spectra.  parameters of this interaction. The existence of a nonzero isotropic hyperfine coupling constant A iso ( 14 N) for the interacting nitrogen indicates that unpaired electron spin density is transferred from the semiquinone onto this nucleus. The value A iso ( 14 N) ϭ 0.8 MHz, obtained from HYSCORE spectra, is determined mainly by the unpaired spin density in the nitrogen atom 2s orbital. It corresponds to a small delocalization of spin density S ϳ 0.4 ϫ 10 Ϫ3 in this orbital (43). This is indicative of the existence of an atomic bridge between the semiquinone and the 14 N nucleus. The only likely mechanism for such a transfer is via a hydrogen bond between one of the SQ oxygens and this nitrogen. This result is supported by the detection of at least one exchangeable proton strongly coupled to MSQ D by HYSCORE and ENDOR, visualized by substitution of H 2 O by D 2 O in redox-poised samples of E. coli NarGHI. 4 The dipole-dipole interaction responsible for the anisotropic hyperfine interaction of the interacting nucleus is determined by the interspin distance and by the fraction of the electron spin density, which interacts with the nuclear spin. The distance between the coupled nitrogen and the semiquinone can be roughly estimated by using Equation 1. Assuming that the interaction occurs via an H-bond, it is probable that the coupling is dominated by the oxygen-nitrogen distance. Spin densities ranging from 0.19 to 0.23 on the oxygen atoms of semiquinones have been previously measured using 17 O-labeled quinones in alcoholic solvent (44) or in bacterial reaction centers (24). Substitution of these values into Equation 1 then gives r O-N ϭ 2.23 Ϯ 0.13 Å, which is a rough estimate of the O-N distance.

Characteristics of the Interaction and
For the assignment of the nitrogen atom, we rely on comparison of the 14 N nuclear quadrupole resonance parameters deduced from the present ESEEM data ( ϭ 0.49, ϭ 0.50) with those available in the literature from previous ESEEM studies of protein-bound semiquinones (25,27-31,34 -36,45). Although experimental data are not available for all possible candidates, the parameter of the nitrogen interacting with the SQ stabilized in NarGHI is very close to the corresponding parameter measured for N ␦ histidine nitrogens which possess values ϳ0.35-0.41 (Fig. 4). Corresponding values are in the range 0.61-0.82. As the latter has been shown to be much more influenced by the geometry of the N-H bond than the parameter, the relatively small value determined in this work may reflect a slightly different H-bonding pattern compared with that occurring for Q A . in bacterial reaction centers (30,35,36) or in photosystem II (25,28). Interestingly, the quadrupole characteristics determined here for the 14 N nucleus interacting with MSQ D in NarGHI are very close to the values calculated by Fritscher for an imidazole making a strong in-plane hydrogen bond with a water molecule (r O-N ϳ2.7 Å) (46). Moreover, DFT calculations carried out on more realistic models of methylimidazole-benzosemiquinone complexes predict values of (, ) ϭ (0.48, 0.61) for short (i.e. r O-N ϭ 2.75 Å) in plane hydrogen bond between both molecules. Thus, the good agreement between our experimental quadrupole coupling parameters and these theoretical predictions provide additional support for our assignment and for the involvement of this His nitrogen in hydrogen bonding to the semiquinone at the Q D site. In a previous work, we have clearly shown that the menasemiquinone radical of E. coli nitrate reductase A binds to the Q D site, in the vicinity of the distal heme b D (19). On the other hand, the crystal structure of the enzyme in complex with pentachlorophenol has been determined to 2.0 Å resolution (20). In the three-dimensional structure, the hydroxyl group of PCP is hydrogen bonded to one of the propionate groups of b D and to N ␦ -His-66. The r O-N distance is 2.8 Å. Based on these structural data and given the fact that no other histidine residue is present in the Q D site, we assign the interacting N ␦ -His evidenced by our spectroscopic studies to His-66 (see Fig. 5). Thus, the presence of an H-bond to a histidine residue appears to be a common feature of both PCP and MSQ D binding. This contrasts with the situation observed for the menaquinol oxidation site of the polysulfide reductase from Thermus thermophilus that has been recently characterized by cocrystallization experiments using the natural substrate menaquinone or the inhibitor PCP 4 S. Grimaldi, unpublished results. The value of (, ) obtained in this work for the nucleus that interacts with MSQ D in E. coli NarGHI is shown by an arrow. N ⑀ Trp at the phylloquinone site A 1 in photosystem I (31), N ⑀ His at the ubiquinone reduction site Q i of bacterial bc 1 complex (27), N ⑀ Arg at the ubiquinone site Q H in cytochrome bo 3 (34,45), N ␦ His and backbone nitrogen at the Q A and Q B sites of bacterial reaction centers (30,35,36) and photosystem II (25,28,29). On the basis of our spectroscopic data and the structure of NarGHI co-crystallized with PCP, we propose that the N ␦ of the His-66 imidazole H-bonds to the semiquinone O 1 while the N ⑀ coordinates the heme b D iron. Whereas the nature of the ligand(s) to O4 remains unresolved, we propose that Lys-86 could be involved in hydrogen bonding to MSQ D via a water molecule (see "Discussion"). The orientation of the quinone ring is taken to be identical to that obtained from molecular modeling of a menaquinone in the PCP-binding pocket by Bertero et al. (20). (47). Although these compounds have been found in the same pocket, differences regarding the hydrogen bonding situation were evident.
Interestingly, the hyperfine coupling constants of the MSQ D electron spin with the N ␦ -His-66 nitrogen in NarGHI (A iso ( 14 N) ϭ 0.8 MHz, T( 14 N) ϳ0.11 MHz) are very similar to the corresponding values A iso ( 14 N) ϭ 0.8 MHz, T( 14 N) ϳ 0.1-0.14 MHz) recently determined for the His nitrogen nucleus magnetically coupled to the ubisemiquinone anion Q i . stabilized at the quinone reduction site Q i of the bc 1 complex from Rhodobacter sphaeroides (27,48). This may reflect a similar mechanism of spin density transfer onto these nuclei. In both systems, it is likely that the presence and the redox state of the nearby heme (b D in NarGHI, b H in the bc 1 complex) plays a significant role in this mechanism of spin density transfer, as the heme is coordinated by a histidine ligand that interacts with the semiquinone formed in the Q site (49). It is noteworthy that in the redox-poised EPR samples used in this work for studying the MSQ D properties, the b D heme (E m,7.5 Fe 2ϩ /Fe 3ϩ ϳ ϩ20 mV) is fully reduced, which mimics the catalytically relevant situation after that MQH 2 (48). Further, systematical theoretical investigations of the influence of the hydrogen bond length r O-N and of an in-or out-ofplane distortion of the bond geometry on the semiquinonenitrogen hyperfine coupling constants would be useful to get a deeper insight into the mechanism of spin density transfer onto the histidine nitrogens coupled to Q i . and to MSQ D . This will allow us to correlate the observed similar hyperfine couplings to a particular common bond geometry. Overall, the magnitude of the hyperfine interaction between MSQ D and the N ␦ -His-66 nitrogen indicates an overlap between the singly occupied molecular orbital of the radical and the His N ␦ orbitals that is favorable for rapid electron transfer to heme b D . However, the structural organization SQ-His-heme conserved in NarGHI and in the bc 1 complex does not appear critical by itself for tuning the radical stability. Indeed, the semiquinone stability constant K S differs by three orders of magnitude in these complexes (K S pH 7.5 ϳ70 in NarGHI, K S pH 7.7 ϳ10 Ϫ2 in R. sphaeroides bc 1 complex) (50). The high stabilization of MSQ D in NarGHI is directly related to the redox potentials of the redox transitions MQH 2 /MSQ (E m,7.5 ϭ Ϫ150 mV) and MSQ/MQ (E m,7.5 ϭ Ϫ40 mV), which are both thermodynamically favorable for electron transfer to the b D heme (18). However, the functional reasons of this stabilization remain to be established since, even if both one-electron redox potentials were exchanged (namely E m,7.5 (MQH2/MSQ) ϭ Ϫ40 mV and E m,7.5 (MSQ/MQ) ϭ Ϫ150 mV leading to K S pH 7.5 ϳ10 Ϫ2 ), they would be still favorable for electron transfer to heme b D . Moreover, a high MSQ D stability implies higher concentrations and possibly a longer life span for the semiquinone intermediate that however do not appear harmful for the bacterial cell as menaquinone is preferentially used during anaerobic respiration. In contrast, the lower stability of ubisemiquinone species in the bc 1 complex of aerobic organisms may have been natu-rally selected to keep radical concentrations as low as possible, minimizing superoxide generation from O 2 with deleterious physiological consequences (51). To better understand the origin of the MSQ D stability, its different possible protonation states should be taken into account. Additionally, other structural factors may drastically influence the radical stability, such as the presence of other ligands to the semiquinone: while an Asp and a Ser residue are hydrogen bonded to Q i . , a Lys residue (Lys-86) has been discussed as a potential additional hydrogen bond donor to MSQ D (19,20).
Role of the Lys-86 Residue-Using multifrequency ESEEM experiments, we did not find any evidence for the transfer of a measurable spin density on any other nucleus than that discussed above. Moreover, the 14 N ESEEM pattern is unchanged in the pH range from 6.5 to 8.5 (data not shown). This indicates that no additional direct hydrogen bond is formed between a nitrogen nucleus and MSQ D in this pH range (Fig. 5). Interestingly, we have recently shown that the conserved residue Lys-86, located at the entrance of the Q D site cavity is necessary for MSQ D stabilization (19). Its substitution into Ala strongly reduces the quinol/nitrate oxidoreductase activity measured using various electron donors (19). Additionally, the inhibitory effect of PCP measured in the wild-type enzyme is strongly attenuated in this mutant (20). Given that the N of Lys-86 lies ϳ6 Å from the Cl 4 atom of the inhibitor, this raises the question of how Lys-86 is involved in the quinol oxidation mechanism by NarGHI. Using molecular modeling of the physiological substrate menaquinone in Q D based on the PCP position, Bertero et al. (20) suggested that relatively small conformational movements of the Lys-86 side chain could occur and allow this residue to play a role in quinone stabilization, providing a direct hydrogen bond donor for the hydroxyl oxygen at position 4. Correct positioning of the quinol molecule within Q D could be an essential function of Lys-86. Indeed, in the crystal structure of NarGHI K86A /PCP, the inhibitor does not experience any polar contact with the protein, in particular with His-66, although it still binds to the enzyme. Moreover, the Q D -binding pocket is largely preserved in this mutant form, as indicated by comparison of the NarGHI K86A three-dimensional structure with that of the wild-type enzyme (20).
Although the present work excludes a direct H-bond between O4 of MSQ D and the N of Lys-86, it is possible that water molecules that are not resolved in the electron density map act as ligands to the menasemiquinone, as suggested for the ubiquinone bound to E. coli succinate dehydrogenase or for the menaquinone co-crystallized with T. thermophilus polysulfide reductase (47,52). In this context, a water-mediated hydrogen bond between the N of Lys-86 and the O4 of MSQ D could also occur (Fig. 5). Such a situation encountered in the Q i binding site of the cytochrome bc 1 complex from the yeast Saccharomyces cerevisiae led to the absence of 14 N ESEEM modulation from the N ␦ of His-202 in the intermediate ubisemiquinone state (42).
On the other hand, it should be kept in mind that several studies have reported on a dynamic behavior of quinone binding within Q sites. This could lead to different redox-dependent binding modes of the quinone during the quinol oxidation cycle. Such a phenomenon has been clearly evidenced for ubiquinone in the Q B quinone reduction site of R. sphaeroides photosynthetic reaction center: upon one-electron reduction, the ubisemiquinone radical undergoes a 4.5 Å (center to center) movement making a direct hydrogen bond to N ␦ of His-190 that is not formed in the oxidized ubiquinone state (53). Recently, Horsefield et al. (54) proposed a reaction mechanism for the reduction of ubiquinone that involves a movement of ubiquinone between two different sites in the E. coli complex II.
The critical role of the Lys-86 residue for stabilizing MSQ D and for proper NarGHI functioning could be caused by its ability to differently interact with the quinone species formed at the Q D site during the quinol oxidation mechanism. In this context, a movement of the Lys-86 side chain could occur and could be associated to protonation/deprotonation events occurring during enzyme turnover allowing protons to be released to the periplasm. These could directly influence the stability of MSQ D . A detailed study of the pH-dependence of MSQ D redox properties and of the hydrogen bonding network to the radical will enable to characterize the protonation state of the radical, to better delineate the role of Lys-86, and to assess how deprotonation and oxidation events are coordinated during quinol oxidation by NarGHI. For this purpose, additional information about the exchangeable protons in hydrogen bonds can be provided by probing interactions with local magnetically coupled protons and the changes on substitution of 2 H 2 O (18,48,55,56). Such studies are currently being performed in our laboratories using high resolution EPR methods.
Concluding Remarks-Pulsed EPR experiments performed on E. coli membrane fractions containing overexpressed NarGHI provided direct evidence for nitrogen ligation to the natural menasemiquinone stabilized in the quinol oxidation site Q D of the enzyme. S-band ESEEM/HYSCORE experiments enabled us to accurately determine the nuclear quadrupole interaction parameters of the interacting nitrogen nucleus and, thus, to assign the latter to the N ␦ of a histidine residue hydrogen bonded to one of the semiquinone carbonyl groups. From the available structural data on the Q D site, we propose that His-66, a b D heme axial ligand, is the hydrogen-bonded nitrogen donor to the C1 oxygen whereas Lys-86 appears to be indirectly involved in MSQ D stabilization, possibly via a water-mediated H-bond to the radical.