Characterization of the Semiquinone Radical Stabilized by the Cytochrome aa3-600 Menaquinol Oxidase of Bacillus subtilis*

Cytochrome aa3-600 is one of the principle respiratory oxidases from Bacillus subtilis and is a member of the heme-copper superfamily of oxygen reductases. This enzyme catalyzes the two-electron oxidation of menaquinol and the four-electron reduction of O2 to 2H2O. Cytochrome aa3-600 is of interest because it is a very close homologue of the cytochrome bo3 ubiquinol oxidase from Escherichia coli, except that it uses menaquinol instead of ubiquinol as a substrate. One question of interest is how the proteins differ in response to the differences in structure and electrochemical properties between ubiquinol and menaquinol. Cytochrome bo3 has a high affinity binding site for ubiquinol that stabilizes a ubi-semiquinone. This has permitted the use of pulsed EPR techniques to investigate the protein interaction with the ubiquinone. The current work initiates studies to characterize the equivalent site in cytochrome aa3-600. Cytochrome aa3-600 has been cloned and expressed in a His-tagged form in B. subtilis. After isolation of the enzyme in dodecylmaltoside, it is shown that the pure enzyme contains 1 eq of menaquinone-7 and that the enzyme stabilizes a mena-semiquinone. Pulsed EPR studies have shown that there are both similarities as well as significant differences in the interactions of the mena-semiquinone with cytochrome aa3-600 in comparison with the ubi-semiquinone in cytochrome bo3. Our data indicate weaker hydrogen bonds of the menaquinone in cytochrome aa3-600 in comparison with ubiquinone in cytochrome bo3. In addition, the electronic structure of the semiquinone cyt aa3-600 is more shifted toward the anionic form from the neutral state in cyt bo3.

A number of prokaryotes contain heme-copper respiratory oxygen reductases, which utilize a membrane-bound quinol as the substrate (electron donor) (1,2). These enzymes (quinol oxidases) are closely related to the cytochrome c oxidases, reduce O 2 to water, and also pump protons across the mem-brane bilayer, generating a proton motive force. The quinol oxidases lack Cu A , which is present in the cytochrome c oxidases, and the amino acid sequences of the quinol oxidases can be distinguished from those of the cytochrome c oxidases by the lack of the Cu A binding motif.
The most intensively studied heme-copper quinol oxidase is the cytochrome bo 3 ubiquinol oxidase from Escherchia coli (cyt bo 3 ) 3 (3)(4)(5)(6)(7)(8). There are currently over 400 sequences of quinol oxidases that are homologues of cyt bo 3 . The vast majority of these sequences are from proteobacteria (330 sequences) or the firmicutes (80 sequences). Bacillus subtilis, a firmicute, does not contain ubiquinone but relies on menaquinone (see Fig. 1) as a redox component in its aerobic respiratory chain (9). There is a homologue of cyt bo 3 in B. subtilis called cytochrome aa 3 -600, and as expected, this enzyme is a menaquinol oxidase (10 -16). Whereas E. coli cyt bo 3 uses only ubiquinol as a substrate, the B. subtilis cyt aa 3 -600 is strictly a menaquinol oxidase. The motivation of the current work is to decipher the differences between the protein-quinol interactions of the bo 3type ubiquinol oxidase and the aa 3 -600 menaquinol oxidase.
Cyt bo 3 has two ubiquinone binding sites, one site with high affinity (Q H ) and one with low affinity (Q L ). The Q L -site is the substrate binding site, and the quinone at this site exchanges readily with the quinone pool in the membrane (17,18). Despite significant effort (17,19,20), little is known about the location of this site within the protein (21). Cyt bo 3 , when purified using the detergent dodecylmaltoside, has 1 eq of ubiquinol-8 bound at the Q H -site, and this bound quinol does not readily exchange with the free quinol in the membrane (18,(22)(23)(24). The ubiquinone bound at the Q H -site functions as a cofactor, accepting two electrons from the quinol at the Q L -site and passing the electrons on to heme b one at a time. Reduced (ferrous) heme b then transfers an electron to the heme o 3 /Cu B active site, where O 2 is reduced to 2H 2 O (25).
The ubiquinone bound at the Q H -site of cyt bo 3 forms a stable semiquinone when the protein is partially reduced (26,27). A combination of x-ray crystallography, site-directed mutagenesis, and pulsed EPR methods have been used to define the residues at the Q H -site and details of the interac-tions between these residues and the bound semiquinone (3, 4, 6, 21, 28 -30). Four polar residues have been implicated in binding to the quinol at the Q H -site in cyt bo 3 : Arg-71, Asp-75, His-98, and Gln-101. Within the Ͼ400 sequences of quinol oxidases, Arg-71, Asp-75, and His-98 are totally conserved. Gln-101 is totally conserved in sequences from proteobacteria but is often replaced by a glutamic acid in the homologues in the Firmicutes, including the B. subtilis aa 3 -600 menaquinol oxidase.
Pulsed EPR methods have revealed several salient features of the interactions between the residues at the Q H -site of cyt bo 3 and the SQ; 1) the hydrogen bonding to the SQ is highly asymmetric, with strong hydrogen bonds to carbonyl O-1 and weaker interactions at carbonyl O-4 side (6), 2) there is one strong hydrogen bond between the ⑀-nitrogen of Arg-71 and carbonyl O-1 of the SQ, resulting in a substantial transfer of unpaired electron spin to this nitrogen (3), 3) there is a strong hydrogen bond between Asp-75 and carbonyl O-1 of the SQ (4), 4) there is a weak interaction between His-98 and carbonyl O-4 of the SQ with a small amount of spin density found on the nitrogens of His-98, 5) there is a very weak interaction between carbonyl O-4 of the SQ and the side chain of Gln-101 (3), 6) the SQ in cyt bo 3 is in the neutral, protonated state at pH 7.5 (6).
In the current work it is demonstrated that the aa 3 -600 menaquinol oxidase from B. subtilis, isolated with the detergent dodecylmaltoside, contains 1 eq of bound menaquinone-7. Partial reduction of the enzyme results in formation of a SQ, analogous to the formation of the SQ formed at the Q H -site in cyt bo 3 . In the B. subtilis aa 3 -600 menaquinol oxidase, the four residues at the putative Q H -site are Arg-70, D74, His-94, and Glu-97. The SQ stabilized by the B. subtilis aa 3 -600 was examined using continuous-wave and pulsed EPR methods. The results show a distinctly different pattern of hydrogen bonding between the protein and SQ species in the menaquinol oxidase than that observed with the E. coli cyt bo 3 ubiquinol oxidase.

EXPERIMENTAL PROCEDURES
Enzyme Preparation-The qoxABCD operon, encoding the B. subtilis aa 3 -600 menaquinol oxidase, was cloned and expressed from plasmid pLala (Cm r ) under the control of the glp promoter. Plasmid pLala replicates both in E. coli and B. subtilis. E. coli strains transformed with pLala vector were maintained on LB plates with 12 g/ml chloramphenicol. A His 6 tag was introduced at the C terminus of qoxB to faciliate protein purification by nickel-nitrilotriacetic acid. The isolated recombinant plasmid pLala was transformed into B. subtilis strain LUW143, lacking both the aa 3 -600 menaquinol oxidase and caa 3 -type cytochrome c oxidase (⌬qoxABCD::kan ⌬ctaCD::ble) (31). Liquid cultures were inoculated with B. subtilis cells grown on LB plates containing appropriate antibiotics. Cells were grown in LB medium treated with 5 g/ml chloramphenicol, 7.5 g/ml neomycin, and 1.8 g/ml zeomycin at 37°C. Enzyme expression was induced by the addition of 20 mM glycerol. To isolate cell membranes, cell pellets were resuspended in buffer containing 50 mM K 2 HPO 4 or 50 mM Tris plus 10 mM MgCl 2 at pH 7.5 and disrupted at high pressure (100 p.s.i.) by using a microfluidizer (Microfluidics Corp., Worcester, MA). Cell debris was removed by brief centrifugation at 8000 ϫ g. The supernatant was then subjected to centrifugation at 180,000 ϫ g for at least 4 h to collect membranes. The isolated membranes were dispersed in 50 mM K 2 HPO 4 , pH 7.5, and homogenized with 1% dodecylmaltoside (Anatrace) by stirring at 4°C. The solution was centrifuged at 180,000 ϫ g for 1 h to remove insoluble fragments and then loaded onto a nickel-nitrilotriacetic acid column for purification. The column was initially equilibrated with 50 mM K 2 HPO 4 , 40 mM NaCl, 0.05% dodecylmaltoside, pH 7.5, for 3-5 column volumes. At least two incremental stepwise washes were performed with buffer containing up to 15 mM imidazole. Cytochrome aa 3 -600 was eluted with buffer containing 100 mM imidazole. The protein was dialyzed overnight against 50 mM K 2 HPO 4 , 0.05% dodecylmaltoside, pH 7.5, and concentrated to ϳ200 M. For the preparation of 15 N-labeled protein sample, cells were grown in Spizizen minimal medium, where the nitrogen source was replaced with isotopically labeled 15 NH 4 Cl (Cambridge Isotope, Andover, MA).
Specific Activity-Cyt aa 3 -600 oxidase had a turnover of 61 electrons s Ϫ1 at 25°C with 2,3-dimethyl-1,4-naphthoquinone (DMN) in 50 mM Tris, 0.05%, pH 7.0. The respiratory activity was started by reducing 10 M DMN in the presence of 200 M NADH and excess amounts of purified diaphorase and by adding 0.05 M of aa 3 -600 oxidase. The steady-state activity was monitored by oxidation of NADH at 340 nm. The autooxidation of DMN at concentrations above 10 M prevented a study of oxidase activity dependence on substrate concentration. Among the various quinol-type electron donors, DMN was previously shown to have the highest enzyme activity (11).
Isolation of Bound Quinone by Reverse-phase HPLC-Quinone was extracted from the purified enzyme preparation with 3 ml of solvent containing methanol/petroleum ether (6:4, v/v) and repeated three times. The organic phase was combined and treated further after a procedure previously described (18). The quinone was isolated by reverse-phase HPLC using a Varian Microsorb-MV 100 -5 C18 column (4.6 mm ϫ 25 cm) and a Waters HPLC system. Isoprenoid quinone structure was characterized by mass spectroscopy (Mass Spectrometry Laboratory, University of Illinois at Urbana-Champaign, Urbana, IL).
EPR Measurements-To prepare the samples for EPR analysis, the purified protein (100 -200 M) was reduced anaerobically in the presence of a 3-5-fold excess of DMN and a 200fold excess of NADH and rapidly frozen in the EPR tube. The continuous-wave EPR measurements were performed on an X-band Varian EPR-E122 spectrometer and a Q-band Bruker ELEXSYS 580 equipped with a separate Q-band microwave bridge and cavity operating at a 100 kHz modulation frequency. The pulsed EPR experiments were carried out using an X-band Bruker ELEXSYS E580 spectrometer equipped with Oxford CF 935 cryostats. Several types of experiments with different pulse sequences were employed with appropriate phase-cycling schemes to eliminate unwanted features from the experimental echo envelopes. Among these experiments were one-and two-dimensional three-pulse and four-pulse sequences, which are described in detail elsewhere (3). Spectral processing of three-and four-pulse ESEEM patterns was per-formed using Bruker WIN-EPR software, including subtraction of the relaxation decay (fitting by 3-6 degree polynomials), apodization (Hamming window), zero filling, and fast Fourier transformation. Pulsed ENDOR spectra of the radicals were obtained using Davies and Mims sequences with different pulse lengths. The specifics of these experiments are described in detail elsewhere (4,32).

RESULTS
Menaquinone Co-purifies with Cytochrome aa 3 -600-The preparation of his-tagged cytochrome aa 3 -600 was assayed for the presence of quinone. It was found that 1.25 eq of menaquinone-7 ( Fig. 1) co-purifies with the enzyme, which is extracted from the membrane and purified using the detergent dodecylmaltoside.
EPR Spectra- Fig. 2 shows X-and Q-band EPR spectra of the SQ in the wild type cyt aa 3 -600. The X-band spectrum displays a single pattern with a g value of 2.0047 Ϯ 0.0001 and resolved hyperfine structure consisting of the four components with approximate relative intensities 1:3:3:1 and a splitting of ϳ0.45-0.49 mT (or 12-13 MHz). The hyperfine structure resolution is better in the sample prepared in 2 H 2 O. On the other hand, the uniform 15 N labeling of the protein does not significantly influence the EPR line-shape of the SQ. This EPR feature is tentatively assigned to three equivalent nonexchangeable protons interacting with the unpaired electron. The Q-band spectrum, measured with about 3.5-times higher microwave frequency than the X-band, resolves the g-tensor anisotropy with components g xx ϭ 2.00642 Ϯ 0.00002, g yy ϭ 2.00540 Ϯ 0.00002, g zz ϭ 2.00228 Ϯ 0.00004. Additional hyperfine structure (also better resolved in the sample prepared in 2 H 2 O) can be seen in the area around g yy ; however, its complete resolution would require experiments at microwave frequencies ϳ95 GHz or higher. The components of the g-tensor, determined from the Q-band spectra of cyt aa 3 -600 SQ prepared in H 2 O and 2 H 2 O, are within the range previously reported for various SQs in model systems and in proteins (30,33,34). A Q-band spectrum with a similar shape was previously reported for the semiquinone intermediate stabilized in the membrane-bound subunit NarI of nitrate reductase (NarGHI) from E. coli (35).
Nitrogens Detected by 14 N and 15 N ESEEM-Powder-type ESEEM spectra, obtained with frozen protein solutions, do not usually show all of the 14 N nuclei that are magnetically coupled with the SQ. This is due to the influence of the nuclear quadrupole interaction (3,4,6,36). To observe all of the nitrogens that magnetically interact with the unpaired electron spin of the SQ, it is necessary to use 15 Nlabeled protein. The 15 N nucleus is a spin 1 ⁄ 2 system and does not possess the nuclear quadrupole moment that affects the 14 N ESEEM spectra. Therefore, we performed ESEEM experiments both with the wild type cyt aa 3 -600-containing natural abundance 14 N (99.16%) and with uniformly 15 N-labeled protein. Fig. 3 shows a representative 15 N HYSCORE spectrum from a uniformly 15 N-labeled sample, measured at the maximum of the SQ EPR line and displayed in contour (A) and three-dimensional-stacked (B) presentations. The 15 N line-shape is centered around the sharp peak at a diagonal point ( 15 N , 15 N ) with a 15 N Zeeman frequency 15 N ϳ1.5 MHz. This peak is accompanied by extended shoulders with two weakly resolved maxima at (1.94, 1.08) MHz (1) and (1.73, 1.33) MHz (2) corresponding to couplings of 0.96 MHz and 0.4 MHz, respectively. The total length of the shoulders is ϳ1.5 MHz along the antidiagonal, symmetrically around ( 15 N , 15 N ). This significantly exceeds typical values of the anisotropy for protein nitrogens interacting with a SQ (3,36), suggesting that more than one nitrogen interacts with the SQ, accompanied by the transfer of unpaired spin density onto their nuclei.
Major features of the 14 N HYSCORE spectrum (Fig. 4, A and B) are the cross-peaks (1) correlating the frequencies at 3.5 and 4.5 MHz (Ϯ0.03 MHz). In addition, the spectrum resolves several other off-diagonal cross-features (2-4) of lower intensity, symmetrically located relative to the diagonal. The maxima of these cross-peaks are located at

Semiquinone Radical Stabilized in Cytochrome aa 3 -600
they could be part of extended cross-features correlating transitions with significant orientation dependence. For this reason, the one-dimensional three-pulse ESEEM spectra show only two peaks at frequencies 3.5 and 4.4 MHz from transitions possessing a low orientation dependence in the HYSCORE spectra and do not resolve any other features (supplemental Fig. S1). The 3.5-, 3.0-, and 2.0-MHz frequencies are each made up of two FIGURE 2. A, the X-band EPR spectra of the semiquinone in cytochrome aa 3 -600 in 1 H 2 O and 2 H 2 O are shown. Both the samples were measured at identical conditions. The experimental parameters used were: modulation amplitude ϭ 0.16 mT; modulation frequency, 100 kHz; time constant ϭ 32 ms; MW power ϭ 40 decibels ϭ 20 microwatts; MW frequency ϭ 9.087 GHz; temperature, 105 K. The dotted lines in the upper spectra correspond to the simulated spectra. The simulated spectra were obtained with components of the g tensor, determined from Q-band spectra (see "EPR Spectra"), and isotropic hyperfine coupling 0.49 mT with three equivalent protons. B, the Q-band EPR spectra of the semiquinone in cytochrome aa 3 -600 in 1 H 2 O and 2 H 2 O. Both the samples were measured at identical conditions. The experimental parameters used were: modulation amplitude ϭ 0.3 mT; time constant ϭ 20.48 ms; conversion time ϭ 40.96 ms; sweep time 41.94 s; MW power ϭ 5.3 microwatts; the MW frequency 34.106 GHz; temperature 90 K. 2,2-Diphenyl-1-picrylhydrazyl with a g-value of 2.00351 has been used as a field marker. The asterisk (*)-marked dip in the upper spectra seems to appear as a cavity background that is more prominent at lower MW powers. Arrows indicate magnetic field positions used for the determination of the g-tensor principal values (see "EPR Spectra").  different cross-peaks and cannot be assigned by assuming that all features are produced by the same nitrogen nucleus. A comparison of the frequencies of cross-peaks 1-4 shows that only 1 and 3 have a common frequency 3.5 MHz and, thus, could belong to the same nucleus. Cross-peaks 2 and 4 involve frequencies different from those of 1 and 3 and could be part of the extended cross-features correlating other transitions from either the same nucleus or from one or more different 14 N nuclei. In summary, the 15 N and 14 N HYSCORE spectra indicate that the SQ in cyt aa 3 -600 interacts with at least two nitrogens from the protein environment. Proton HYSCORE and ESEEM-Besides the nitrogens, the HYSCORE spectra contain information about non-exchangeable and exchangeable protons interacting with the electron spin of the SQ. Fig. 5 shows the 1 H HYSCORE spectra of the SQ in the cyt aa 3 -600 prepared in 1 H 2 O (A and B) and 2 H 2 O (C) buffer. Similar spectra for the SQ in the Q i -site of the cytochrome bc 1 complex (37) and Q H -site of cytochrome bo 3 (6) have been discussed previously. In addition to a diagonal peak with extended shoulders at the proton Zeeman frequency ( H ϳ14.75 MHz), the spectrum contains several pairs of resolved cross-features located symmetrically relative to the diagonal. They are designated 1, 2, 3, and 4. The cross-peaks labeled 1 demonstrate the largest hyperfine splitting, of the order ϳ10 MHz. Cross-ridges 2 possess the most extended anisotropic contour, with the largest deviation from the diagonal indicating a significant anisotropic hyperfine component. Cross-peaks 3 and 4 are located in a similar area of the plot, close to each other and partially overlapping, as shown in Fig. 5B. Contours 1 and 3 are approximately normal to the diagonal, suggesting a smaller anisotropy. The contours of cross-peaks 4, located above those of crosspeaks 3, indicate an anisotropic hyperfine interaction intermediate between the couplings producing cross-peaks 3 and 2.
Cross-peaks 2 and 4 ( Fig. 5C) completely disappear in the HYSCORE spectra obtained under the same conditions using the sample with 2 H 2 O, showing that they are produced by exchangeable protons. However, cross-peaks 1 and 3 as well as the diagonal peak, with its shoulders, still appear in the spectra obtained in 2 H 2 O. The 1 H/ 2 H exchange does affect the intensity of cross-peaks 3, which indicates that these peaks result from the simultaneous contribution of exchangeable and non-exchangeable protons. Additional support for this conclusion was obtained from pulsed ENDOR spectra (see below).
Quantitative analysis of the cross-peak contour line-shapes and simulations of the spectra, described in detail in Yap et al. (6) and Dikanov et al. (37) and the supplemental data (including supplemental Figs. S2 and S3 and Table S1) provides the isotropic (a) and anisotropic (T) components of the hyperfine tensors. These are obtained using an axial approximation for protons H1, H2, and H4 associated with cross-peaks 1, 2, and 4. The data are summarized in Table 1. Protons H2 and H4 are clearly exchangeable with the solvent. Cross-peaks 3 are produced by several non-exchangeable and weakly coupled exchangeable protons (see "Discussion"), and the parameters determined from this formal analysis do not correspond to any real structural characteristics.
Additional information was obtained from complementary one-dimensional four-pulse ESEEM and pulsed ENDOR exper-iments. In particular, the one-dimensional four-pulse experiments show the existence of two exchangeable protons with anisotropic couplings ͉T͉ ϳ 2.8 and 5.4 MHz (supplemental Fig. S4).
Pulsed ENDOR- Fig. 6, A and B, shows Davies pulsed ENDOR spectra for the SQ in cyt aa 3 -600 prepared in 1 H 2 O and 2 H 2 O. The spectrum in 1 H 2 O contains two pairs of peaks, 1 A and 2 A , located symmetrically relative to H with splittings of ϳ11 and ϳ 2.7-4.0 MHz, respectively. 1 H/ 2 H exchange influences the shape and relative intensity of the major peaks. In 1 H 2 O, peaks 1 A possess well resolved shoulders corresponding to a splitting of ϳ10.0 MHz between the points marked as 1 on the peaks. These shoulders (1) suggest that at least two different components contribute to the spectral features in this region. This is supported by the spectrum in 2 H 2 O solvent (peaks 1 B ) showing the signals from non-exchangeable protons, which is likely the origin of the shoulders in peaks 1 A . The difference spectrum shows peaks 1 C at the frequencies ϳ9 and 20 MHz with intensities exceeding other peaks (Fig. 6C). The splitting between these peaks of ϳ11 MHz is in agreement with the A Ќ value predicted by the analysis of the contour line-shape for exchangeable proton H2 (anisotropic component T ϭ Ϯ5.6 MHz; isotropic constant a ϭ Ϯ5.4 MHz).
The analysis of the contour line-shape in axial approximation (supplemental Fig. S2) provides the components of the hyperfine tensor in canonical orientations A Ќ ϭ 9.4 MHz and A ʈ ϭ 14.2 MHz for cross-peaks 1 in the HYSCORE spectra (Fig. 5, Table 1). The locations of peaks 1 B in the ENDOR spectra (Fig.  6B) correspond to these limits. The splitting between the points of maximum intensity at ϳ11 MHz in the ENDOR spectrum corresponds well to the splitting from three equivalent protons observed in the X-band EPR spectrum. Based on this analysis, we suggest that the major contribution to features 1 in the HYSCORE and 1 B in the ENDOR spectra is due to the three methyl protons of the menaquinone SQ. These protons would have equal hfi tensors due to rapid rotation of the methyl group.
In addition, there are changes to peaks 2 A due to the 1 H/ 2 H replacement. The difference spectrum shows the signal from exchangeable protons in the area corresponding to splittings up to ϳ5 MHz, which could be from the protons contributing to the cross-peaks 2-4 in the HYSCORE spectra.
All other non-exchangeable protons of the SQ, including the ␤-protons of the isoprenyl tail and ␣-protons of the benzoic ring (in positions 5-8, Fig. 1), possess small couplings and contribute to the cross-peaks 2 B . The frequency limits of crosspeaks 2 B suggest that the maximum principal values of the hyperfine tensors of any of these non-exchangeable protons do not exceed 6 -7 MHz.

DISCUSSION
Nitrogens Interacting with the SQ-A useful starting point for discussing the experimental data for nitrogen ( 14 N and 15 N) nuclei is the 14 N HYSCORE powder spectra of the SQ (Fig. 4). Cross-peaks (1) correlate two double-quantum transitions, dqϩ ϭ 4.5 MHz and dq-ϭ 3.5 MHz, from opposite m S manifolds of the 14 N nucleus (supplemental Fig. S5). These transitions possess low orientation dependence and produce the most intensive cross-peaks in the HYSCORE spectra (38). The frequencies of double-quantum transitions in the powder spectra are described by the equation (39), where efϮ ϭ ͉ 14 N Ϯ 14 A/2͉, 14 N and 14 A are the Zeeman frequency and hyperfine coupling of 14 N nucleus, respectively. The parameter ϭ K 2 (3 ϩ 2 ), in which K ϭ e 2 qQ/4h, the quadrupole coupling constant, and is the asymmetry parameter.
In this case, an application of Equation 1, assigning dqϩ ϭ 4.5 MHz and dqϪ ϭ 3.5 MHz, would provide an estimate of the hyperfine coupling with the 14 N nucleus, using the formula, This equation along with the 14 N Zeeman frequency ( 14 N ϭ 1.066 MHz for the field 346.3 mT) gives a hyperfine coupling 14 A ϭ 0.94 MHz. The shape of double-quantum transitions in three-pulse and HYSCORE spectra indicate that the isotropic constant provides the major contribution to the hyperfine coupling. Using this estimated value of 14 A, one can calculate the parameter ϭ K 2 (3 ϩ 2 ) from Equation 1, which is equal to 2.70 MHz 2 . Assuming that 0 Յ Յ 1, this leads to a quadrupole coupling constant K ϭ 0.82-0.95 MHz.
The quadrupole coupling constant of the 14 N atoms in different chemical groups can be used for identification of the nitrogen and characterization of the hydrogen bond to the SQ. The estimated interval of the constant K partially overlaps with the quadrupole coupling constant for a peptide nitrogen NH-CAO, whose values vary between 0.75 and 0.85 MHz in different compounds, including proteins (Ref. 6 and references therein). It is also consistent with the quadrupole coupling constant K ϳ 0.9 -1.0 MHz of nitrogens from the NH and NH 2 groups in primary aliphatic and aromatic amines and amides

TABLE 1 Hyperfine tensors of the protons H1-H4 (MHz) derived from HYSCORE spectra
The method of analysis of the 1 H HYSCORE spectra gives two possible sets of a and T with interchanged values of A Ќ and A ʈ for each proton. Arguments used for the selection of the values given in Table 1 are provided in the supplemental material (section "Proton HYSCORE") and in our previous publications (4,6).  (40). On the other hand, this coupling constant is larger than the K values reported for either the deprotonated or protonated nitrogens of the imidazole residue (36). That rules out histidine as a hydrogen bond partner to the SQ of cyt aa 3 -600. For the SQ in the Q H -site of the cyt bo 3 , the nitrogen possessing the largest hyperfine coupling was identified as the N ⑀ H from side chain of Arg-71 using selective isotope 15 N labeling (3). The hyperfine coupling, 14 A ϳ 1.8 MHz, due to this nitrogen is two times larger than the largest coupling, 14 A ϳ 0.9 MHz, in cyt aa 3 -600. The coupling in cyt bo 3 satisfies well the cancellation condition ͉ 14 N Ϫ 14 A/2͉ ϳ 0 in X-band EPR. This condition allows one to determine directly the nqi tensor from 3 resolved nqi frequencies in ESEEM spectra, yielding K ϭ 0.93 MHz and ϭ 0.51 for the arginine N ⑀ (6, 29). These values give ϭ 2.82 MHz 2 , which differs only slightly from the estimated value of ϭ 2.7 MHz 2 for the nitrogen coupled to the SQ from cyt aa 3 -600 (see above). This close coincidence of the quadrupole parameters () suggests that the N ⑀ of Arg-70 in cyt aa 3 -600 is also a primary candidate for the role of the nitrogencarrying largest unpaired spin density and, thus, involved in H-bond formation with the SQ. The 14 N HYSCORE spectra, which are calculated using the hyperfine coupling 14 A ϳ 0.9 MHz and nuclear quadrupole parameters K ϭ 0.94 MHz and ϭ 0.5, reasonably reproduce the location of cross-peaks 1 at (4.5, 3.5) MHz with intensities significantly exceeding other minor features that appear in the area of peaks 2-4 in the experimental spectra. The location of cross-peaks 1 is not influenced by an addition of an anisotropic hfi even with the perpendicular component of the tensor as large as T ϳ 0.2 MHz. The locations of other cross-features are, however, strongly influenced by the anisotropic hfi as well as by the relative orientation of the hfi and nqi tensors.

Proton a, T
The estimated coupling, 14 A ϳ 0.9 MHz, corresponds to 15 A ϳ 1.26 MHz for 15 N. The 15 N line-shape in the HYSCORE spectra shows an extended contour with the shoulders extending ϳ1.5 MHz. This is consistent with a coupling of ϳ1. 26 MHz, although even the largest resolved coupling, 15 A ϳ 0.96 MHz (peaks 1 in Fig. 5), is significantly smaller than this value. Two factors can explain this difference. First, the positions of line maxima in powder spectra for the double-quantum transitions of 14 N and single-quantum transition of 15 N, used to estimate 14 A and 15 A, are determined by different factors. This results in effectively different couplings in the presence of the anisotropic contribution to the hyperfine coupling. The difference will be greater as the anisotropic hfi gets larger. Second, because the total length of the 15 N resonance line exceeds typical values of the anisotropic hfi for protein nitrogens interacting with a SQ (3,36,41), it is likely that the extended 15 N lineshape in the HYSCORE spectra with two resolved splittings is formed by the spectra from several (probably more than two) nitrogens possessing partially overlapped intervals of nuclear frequencies and, thus, producing an extended resonance contour. Overlap of the spectra from different nuclei can produce the resolved maxima at the new frequencies, shifted from the frequency of the maximum intensity in the individual spectrum of each contributing nitrogen. Individual couplings with nitrogens from different residues in the SQ environment can be characterized using selective labeling with 15 N, which is what was done to resolve the situation for the Q H -site SQ in cyt bo 3 . The unpaired spin density producing measurable hyperfine couplings was found on at least four nitrogens from Arg-71 and His-98 in cyt bo 3

(3).
Applying Equations 1 and 2 to the frequencies of the crossfeatures 2-3, one obtains 14 A ϳ 0.4 -0.6 MHz and ϭ 0.36 -0.6 MHz 2 . The small value of is consistent with values previously reported for the protonated imidazole nitrogen from histidine residues H-bonded with the SQs (see Table 2 in Ref. 36). This estimated value of 14 A is also similar to the hyperfine couplings from the nitrogens of His-98 in cyt bo 3 (3). Thus, the available 14 N and 15 N ESEEM data can be reasonably interpreted as resulting from the presence of several nitrogens presumably from residues Arg-70 and His-94 involved in the interaction with the SQ accompanied by the transfer of unpaired spin density onto these nitrogens. Although the equivalent nitrogens from Arg-71 and His-98 interact with the SQ in cyt bo 3 , values of hyperfine couplings are not the same, reflecting differences in hydrogen bond geometry and overlap of the electronic orbitals of the SQ and protein residues.
Further work is needed to clarify the situation, utilizing 15 Nselective labeling to characterize the individual isotropic and anisotropic hyperfine couplings with the side chain and peptide nitrogens of different residues. In addition, pulsed EPR experiments at lower microwave frequencies (3-4 MHz; S-Band EPR) will better satisfy the cancellation condition for the 14 A couplings and, therefore, allow the accurate determination of the nuclear quadrupole tensors of the 14 N nuclei (36,41). Both the anisotropic hfi and nqi tensors of the nitrogens will be necessary to model the length and geometry of each of the H-bonds between the SQ and nitrogen partners.
Exchangeable Protons-The current data show two exchangeable protons with significant differences in their hyperfine couplings for the SQ in cyt aa 3 -600. These protons possess anisotropic hyperfine couplings with ͉T ͉ ϳ 5.6 and 2.9 MHz. The first value significantly exceeds and second one is comparable with value ͉T͉ ϳ 3 MHz, determined from ENDOR experiments with different SQs for in-plane hydrogen-bonded protons in alcoholic solutions (42)(43)(44). Hydrogen bonding to the quinone carbonyl groups occurs via proton donation to the two lone pairs on the sp 2 -hybridized carbonyl oxygen. Density functional theory calculations show that a hyperfine coupling of ͉T͉ ϳ 3 MHz is consistent for a proton participating in a planar hydrogen bond, forming an angle Ϯ 60°with CAO bond and with a hydrogen bond length ϳ1.8 Å (45). A similar bond length has been estimated using a point-dipole model for the O…H bond. It is reasonable to conclude that the H-bond to the SQ of cyt aa 3 -600 involving the proton with ͉T ͉ ϳ 2.9 MHz fits this geometric description.
The second H-bond to the SQ of cyt aa 3 -600 with an exchangeable proton has a much larger anisotropic hyperfine coupling, ͉T͉ ϳ 5. 6 MHz. An orientation-selected, 1 H/ 2 H Q-band ENDOR study of the Q A -site SQ in the photosynthetic reaction center has reported a hyperfine tensor with a similarly high value of T ϭ 5.2 MHz and a ϭ Ϫ1.28 MHz for one H-bonded proton (46). The point-dipole estimate gives the O…H distance equal to 1.32 Å for this proton. However, the estimated bond length based on the value of the 2 H quadru-pole coupling tensor is 1.6 Å. According to DFT calculations (45), a hydrogen bond with planar geometry with an anisotropic coupling of T ϳ 5.2 MHz would require an O….H bond length of ϳ1.4 Å. A planar H-bond with a value of T ϭ 5.6 MHz, found in the current work, would correspond to even shorter O…H distance.
It is also reasonable to consider the data in terms of an H-bond that is not in the plane of the SQ. Density functional theory calculations show that, given a constant O…H distance, deviation of the H-bond out of the SQ plane leads to a simultaneous increase of both the isotropic and anisotropic couplings of the H-bonded proton (47). This would be the result for any H-bonds that are forced by geometric constraints to be either above or below the ring plane. For the SQ in the Q A -site, described above, the H-bond containing the proton with T ϭ 5.2 MHz forms angle 40°with respect to the quinone plane and angle of 13°with the CAO bond (46). Analysis of the 1 H crosspeak contours in the HYSCORE spectrum shows that the exchangeable proton observed in the SQ of cyt aa 3 -600 with ͉T͉ ϭ 5.6 MHz has two possible values for the isotropic constant, ͉a͉ ϭ 5.4 or 0.2 MHz (supplemental Table S1). The difference between these two solutions is in the interchanged values of A Ќ and A ʈ . However, the simulations of the HYSCORE spectra (supplemental Fig. S3) and locations of the line assigned to A Ќ in the ENDOR spectrum of exchangeable protons (Fig. 6C) support the solution with the unusually large isotropic constant. An even stronger anisotropic hyperfine coupling, T ϭ 6.3 MHz, was found for one exchangeable proton in cyt bo 3 , which we interpreted as evidence for a neutral radical with substantial covalent character to the O-H bond (6), but we have not found evidence for such a large isotropic constant in cyt bo 3 . Although further work is warranted, the most reasonable interpretation of these data indicates that the H-bond involving the proton with ͉T͉ ϭ 5.6 MHz in cyt aa 3 -600 is significantly out-of-plane of the SQ ring.
Proton Couplings; Spin Density Distribution-The hyperfine couplings for both the non-exchangeable methyl protons and for the exchangeable protons obtained from HYSCORE and ENDOR experiments with the SQ of the menaquinone-7 in cyt aa 3 -600 provide information about the interaction with the protein environment, greatly aided by the results from previous studies of 2-methyl-1,4-naphthoquinone derivatives (the vitamin K group; Fig. 1) in model systems and proteins.
All members of the vitamin K group share a naphthoquinone ring structure methylated at the second position and vary in the aliphatic side chain attached at the 3-position (see Fig. 1), including menadione (called also vitamin K 3 ), phylloquinone (vitamin K 1 ), and menaquinones (vitamin K 2 ). Vitamin K 2 is a collective term for a family of menaquinones (MQs) that have side chains composed of a variable number of unsaturated isoprenoid residues. Generally they are designated as MQ-n, where n specifies the number of isoprenoids varying from 4 to 13.
When examined in liquid and solid organic solvents, the anion radicals of derivatives of 2-methyl-1,4-naphthoquinone have similar isotropic couplings and anisotropic tensors for the protons in the equivalent positions. These data, supported by density functional theory calculations, indicate a weak influence on the spin-density distribution from the side chain at the 3-position. The isotropic hyperfine constant and the average perpendicular component of the anisotropic hfi tensor for the methyl protons vary from 7.0 to 7.9 MHz and from 1.0 to 1.3 MHz, respectively (48 -52, 55; Table 2). In hydrogen-bonding solvents such as alcohols, the hfi tensors for each of the  (42)(43)(44)(45).
In contrast to these results, the isotropic constant for methyl protons of the SQ of menaquinone-7 in cyt aa 3 -600 is substantially larger, a ϭ 11 MHz. This indicates a significant redistribution of the unpaired spin density compared with the anion radical in organic solvents. A similarly large isotropic constant, ϳ10 -12 MHz, has been reported for the methyl protons for the SQ of the phylloquinone in the A 1 center of the photosystem I (49,50,54) and in the Q H -site of cyt bo 3 (55) (in this experiment native ubiquinone-8 was artificially replaced by phylloquinone). The methyl protons for the SQ of the menaquinone-9 in the Q A -site of the reaction center from Rhodopseudomonas viridis have an isotropic constant that is not quite as large, about 6.8 MHz (52).
The proton isotropic constant of the rotating methyl group is directly proportional to the -spin density on the attached carbon atom, as described by the McConnell relation, a ϭ 81 (56). Thus, the unpaired -spin density on the carbon attached to the methyl group of the SQ in cyt aa 3 -600 (and the A 1 center of photosystem I) is about ϳ1.5-fold larger (ϳ9% to ϳ13.5%) compared with the anionic SQ radicals in alcohol solvents.
Our ENDOR spectra show that other non-exchangeable protons, i.e. the ␤-protons of the isoprenyl tail and of the ␣-protons of the benzoic ring, contribute to peaks 2, with the maximum splitting of ϳ7.5 MHz. The reported values of the isotropic couplings for ␤-protons and ␣-protons in vitamin K anion-radicals in organic solvents do not exceed 4 and 2.2 MHz, respectively. The maximum components of the anisotropic hyperfine tensors of ␣-protons are smaller than 3.6 MHz in these model studies. The shape of the peaks 2 in the ENDOR spectra of the SQ from cyt aa 3 -600 (Fig. 6) also suggests an increase of the hyperfine coupling for some ␣-protons as a result of a shift in spin density compared with the model systems in organic solvents.
The differences in hyperfine couplings between anion radicals in alcohol solutions and SQ radicals in proteins result from an asymmetry of hydrogen bonding with the carbonyl oxygens of the SQ in the proteins, which leads to a redistribution of both the spin density and charges within the quinone ring. The couplings for the SQ in cyt aa 3 -600 (and for the A 1 center in photosystem I) can be explained by a stronger hydrogen bond between the protein surroundings and oxygen O4, compared with oxygen O1 (Fig. 1). The bound oxygen O4 possesses a larger negative charge to stabilize the interaction with the proton participating in this strong H-bond, and thus, the spin density is partly shifted within the SQ (33, 54). The result is an increase of the unpaired spin density at C2, which increases the isotropic coupling of the methyl group protons.
Comparison with the SQ of Center A 1 from Photosystem I-A comparison of the ENDOR-derived hyperfine tensors of the exchangeable protons can be used to estimate the relative strength of H-bonds. However, unresolved issues in the literature on the SQ of center A 1 limit the utility of this comparison. Several publications (49,50,52,53,55) report the value of A Ќ ϳ Ϫ(4.6 -5.2) MHz for the exchangeable, H-bonded proton in the SQ of center A 1 ( Table 2). One report provides the experimental value of A Ќ ϭ a Ϫ T ϭ Ϫ5.0 MHz together with A ʈ ϭ a ϩ 2T ϭ 13.4 MHz (50), which gives the value of a ϭ 1.1 MHz and T ϭ 6.  (57). The ESEEM study reports the existence of at least two protein nitrogens coupled to the SQ and carrying unpaired spin density. These nitrogens were tentatively assigned to the indole nitrogen of a tryptophan residue and a ring nitrogen of a histidine, although the side chain amide nitrogens of an asparagine or glutamine could not be ruled out. An x-ray structure reported after publication of this ESEEM study, however, shows the presence of a tryptophan at the A 1 site, but no histidine is present (58). Hence, a re-evaluation of the 14 N and 15 N ESEEM data for the A 1 . radical in photosystem I is required to unambiguously identify the nitrogens involved in the interaction with the SQ.
Model of the SQ Environment in Cytochrome aa 3 -600-Based on the similarity of the amino acid sequences of cyt bo 3 and cyt aa 3 -600, one can suggest that residues associated with the Q Hsite of cyt aa 3 -600 include Arg-70, Asp-74, His-94, and Glu-97 (B. subtilis numbering), which correspond to Arg-71, Asp-76, His-98, and Gln-101 in cyt bo 3 . Site-directed mutagenesis studies have confirmed that these four residues are functionally important in cyt bo 3 and that mutants at each position alter or eliminate the SQ that is stabilized at the Q H site (28). The glutamine (Gln-101) present at the Q H -site in cyt bo 3 , is replaced by a glutamic acid (Glu-97) in cyt aa 3 -600 (Fig. 7).
Among the four residues reasonably suggested as being at the Q H -site, only Arg-70 and His-94 have side chain nitrogens that can participate in hydrogen bonding to the SQ. Hydrogen bonds between the SQ and peptide backbone nitrogens cannot be ruled out at this point, but there is no evidence for such hydrogen bonding at the Q H -site of cyt bo 3 .
The uniform and selective 15 N isotope labeling of the nitrogens in Arg-71, His-98, and Gln-101 of cyt bo 3 has identified one nitrogen, N ⑀ , of Arg-71, that carries the largest amount of the unpaired spin density, corresponding to the coupling 1.8 MHz. All other nitrogens, including the N from Arg-71 and ring nitrogens of His-98, possess smaller couplings of the order 0.1-0.6 MHz. There is no spin density transfer to the N ⑀ of Gln-101, which is involved in only very weak anisotropic coupling with the electron spin of the SQ (3).
The 15 N spectrum of the SQ in cyt aa 3 -600 shows a poorly resolved line-shape, which suggests an interaction with several nitrogens. The largest coupling estimated from the 14 N spectrum is 0.9 MHz and can be tentatively assigned to the N ⑀ of Arg-70 based on the quadrupole parameter, K 2 (3 ϩ 2 ). In the 15 N spectrum the largest resolved coupling, 15 A, is ϳ0.95 MHz. This corresponds to a coupling 14 N of (i.e. 14 A) of only ϳ0.7 MHz, indicating an overlap of the spectra from at least two nitrogens, presumably from Arg-70 and His-94. It is possible that in cyt aa 3 -600 there is more spin density transferred to the nitrogens of His-94 than to the corresponding His-98 in cytochrome bo 3 . The mechanism of the transfer of unpaired spin density to His-98 in cyt bo 3 or to His-94 in cyt aa 3 -600 is not clear. One can propose that spin density can be transferred from the menaquinone SQ to the histidine by either hydrogen bonding or, alternatively, by -stacking of the menaquinone and the imidazole ring of His-94. If there were -stacking then of course the model shown in Fig. 1 (based on cyt bo 3 coordinates (21)) would need to be substantially revised. In either case, the nitrogen coupling data indicate a significant difference in the protein interactions with the SQs in cyt bo 3 and cyt aa 3 -600. The spin density transfer to the Q H -site arginine from oxygen O4 appears to be substantially weaker in cyt aa 3 -600, and the interaction of the histidine with oxygen O1 may be greater.
Differences in nitrogen couplings are also consistent with differences in the couplings of the exchangeable, hydrogenbonded protons. Two strongly coupled exchangeable protons with ͉T͉ ϳ 6.3 and 4.2 MHz are found near the SQ in cyt bo 3 (6). In cyt aa 3 -600 there are also two exchangeable protons, but with ͉T͉ ϭ 5.6 and 2.9 MHz. The smaller hyperfine couplings are also consistent with weaker hydrogen bonding between protein residues to the SQ in cyt aa 3 -600.
Taken together, the nitrogen and proton hfi data indicate weaker hydrogen binding of the menaquinone SQ in cyt aa 3 -600 in comparison with the ubiquinone SQ in cyt bo 3 . In addition, the electronic structure of the SQ in cyt aa 3 -600 has more anionic character compared with the neutral SQ state in cyt bo 3 . However, the asymmetry of the distribution of unpaired spin density for MQ-7 in cyt aa 3 -600 is still substantially greater than observed for MQ-9 in the Q A site of the bacterial reaction center from Rps. viridis (52), which is clearly an anionic SQ, as can be judged from the comparison of the methyl proton couplings (Table 2).
Additional studies are required to provide the necessary information to define the specifics of how the SQ of menaquinone-7 at the Q H -site of cyt aa 3 -600 interacts with the surrounding protein. Both the protein (e.g. Gln-101 versus Glu-97) and the quinone (ubiquinone versus menaquinone) are different. Yet, the high degree of amino acid sequence identity indicates that the Q H -sites must have common structural and functional properties for these two enzymes. The goal is to define the evolutionary adaptation of the protein structure to accommodate different quinones at the same location.