The Proximal Hydrogen Bond Network Modulates Bacillus subtilis Nitric-oxide Synthase Electronic and Structural Properties

Bacterial nitric-oxide synthase (NOS)-like proteins are believed to be genuine NOSs. As for cytochromes P450 (CYPs), NOS-proximal ligand is a thiolate that exerts a push effect crucial for the process of dioxygen activation. Unlike CYPs, this catalytic electron donation seems controlled by a hydrogen bond (H-bond) interaction between the thiolate ligand and a vicinal tryptophan. Variations of the strength of this H-bond could provide a direct way to tune the stability along with the electronic and structural properties of NOS. We generated five different mutations of bsNOS Trp66, which can modulate this proximal H-bond. We investigated the effects of these mutations on different NOS complexes (FeIII, FeIICO, and FeIINO), using a combination of UV-visible absorption, EPR, FTIR, and resonance Raman spectroscopies. Our results indicate that (i) the proximal H-bond modulation can selectively decrease or increase the electron donating properties of the proximal thiolate, (ii) this modulation controls the σ-competition between distal and proximal ligands, (iii) this H-bond controls the stability of various NOS intermediates, and (iv) a fine tuning of the electron donation by the proximal ligand is required to allow at the same time oxygen activation and to prevent uncoupling reactions.

Nitrogen monoxide (NO) is a well described radical molecule (1) that has been shown to exert major physiological functions in mammals, ranging from signaling processes to cytotoxic activities (2)(3)(4)(5). It is exclusively synthesized by a family of enzymes named nitric-oxide synthases (NOSs) 2 that have been cloned and characterized in the early 1990s (6 -9). With the emergence of efficient DNA sequencing techniques, the accessibility of an increasing number of genomes led to an unprece-dented quest for new NOSs in other organisms. Although no NOSs have been found so far in plants and yeasts (10,11), random BLAST analyses of several bacterial genomes led to the discovery of new NOS-like proteins, mostly in Gram-positive bacteria (12). These proteins correspond to the partially truncated oxygenase domain of mammalian NOSs (mNOSs). Because of the dominant structure-function approach, the first works on bacterial NOS-like proteins (bacNOSs) aimed at highlighting the similarities between mammalian and bacterial NOSs. Indeed, the crystallographic structures of NOSs from Bacillus subtilis (bsNOS (13)), Staphylococcus aureus (saNOS (14)), or Geobacillus stearothermophilus (15) were perfectly superimposable to the three-dimensional structure of mNOSs with the exception of a portion of the N-terminal region involved in BH 4 binding and in the formation of the zinc-tetrathiolate complex (12). In addition, the first enzymological experiments suggested that bacNOSs, similarly to mNOSs, had the capacity to catalyze oxygen activation and to metabolize L-Arg, the natural substrate of mNOSs, into citrulline and NO (16,17). However, despite such an assignment of bacNOSs as genuine NO synthases, their in vivo function has been poorly investigated and remains a matter of debate. The first investigations on the biological function of bacNOSs suggested a contribution to the biosynthesis of a phytotoxin (thaxtomin A) of Streptomyces turgidiscabies, presumably via the nitration of a Trp-like residue on a diketopiperazine precursor (18,19). This metabolic role seemed to extend to the Deinococcus radiodurans NOS, which seemed able to associate with the tryptophanyl-tRNA synthetase and to nitrate Trp amino acids (20,21). Following different hypotheses, Nudler and colleagues (22)(23)(24) proposed that bacNOSs could intervene in a series of functions related to host-pathogen interaction such as protection against oxidative stress (22), pathogen survival and virulence (23), or defense against antibiotics (24). Recently, Crane suggested that D. radiodurans NOS could intervene in the recovery processes of bacteria subject to a UV stress (25). Because NO production is primarily used by the host as a bactericidal weapon, we feel that NO is not the most appropriate molecule to elicit a concerted response against the host immune system. In fact, our recent data suggest that bsNOS is more likely to intervene in NO and peroxynitrite detoxification (26). This raises the following question about the in vivo enzymatic activity of bacNOSs. Are they genuine nitric-oxide synthases?
In comparison with mNOSs, the catalytic mechanism of bac-NOSs has not been intensely investigated, but the results obtained so far stress numerous differences between mNOSs and bacNOSs. On top of them, bacNOSs are lacking an effective electron donor. The absence of a dedicated reductase partner (27) should prevent an efficient first electron transfer, although flavodoxins have been reported to support bsNOS-mediated NO production (28). No strong evidence has been reported so far for the implication of a pterin in the second fast electron transfer. The natural pterin is presumably not synthesized by these bacteria, and although BH 4 and tetrahydrofolate increase the rate of decay of the Fe II O 2 complex of several bacNOSs (16,29,30), no pterin radical has been identified in bacNOSs catalysis so far. Another problematic issue is the NO dissociation rate of the bsNOS Fe III NO complex. Indeed, the Val 3 Ile substitution observed in the bacNOSs heme pocket seems to impede a fast NO release (30), which is mandatory for NO synthesis (31,32). Indeed, the Val 3 Ile mutation dramatically diminishes NO production by inducible NOS (iNOS) and converts iNOS into a NO dioxygenase (33). Similarly, the isomerization of peroxynitrite by bsNOS, in contrast with iNOS-mediated peroxynitrite activation, might reflect modification of the heme distal pocket that leads to a transient capture of NO or any other reactive nitrogen species within the catalytic site, which in turn could lead to additional chemical reactions. Considering these points, one could legitimately question the role of bacNOSs as authentic NO-releasing oxygenase enzymes.
In this regard, mNOSs catalysis is highly related to the oxygen activation mechanism described for cytochromes P450 (34 -37). It is characterized by two synergetic processes that allow the O-O bond cleavage, namely the "push-pull" effects (38). The pull effect engages the distal H-bond network that mediates the two proton transfers required for oxygen activation and, for NOS, L-Arg hydroxylation. The resonance Raman fingerprints of numerous saNOS (39 -41) and bsNOS (42,43) complexes reveal no significant differences in the heme distal environment of bacNOSs and mNOSs (44 -46). Some subtle differences have been observed between the x-ray structures of Fe II NO complexes of bsNOS (47) and neuronal NOS (nNOS) (48,49) in the presence of NOHA and L-Arg, but they should not significantly affect the proton transfer processes. In fact, the variability of the heme distal structure is greater within mNOSs than between bacNOSs and mNOSs.
However, we previously noticed a difference in the "push" effect. We reported that the proximal Fe-S bond was stronger for bsNOS than for mNOSs, suggesting a modulation of the electron donation exerted by the thiolate ligand (43). This regulation pattern seems to be inherent to mNOSs. Unlike P450s, the sulfur atom of NOS-proximal thiolate is naturally engaged in a strong H-bond interaction with the nitrogen proton of the indole ring of the vicinal tryptophan (Scheme 1). Modification of this H-bond alters mNOS mechanism and catalysis. Thus, the loss of this H-bond for the W409F/Y mutant of nNOS increases the rate of Fe II NO oxidation (50,51). The proposed strengthening of this H-bond, in the case of the W188H mutant of iNOS, results in a slower oxygen activation process (52). If bacNOSs are bona fide oxygenases, the same regulation pattern should prevail. This has been suggested by Couture and co-workers (53), who showed that the suppression of the Trp-Cys H-bond in saNOS seemed to modify its proximal Fe-S bond.
We report here a comprehensive investigation of the modulation of the interaction between the proximal cysteine ligand and the vicinal Trp for bsNOS. We analyzed the effects of several mutations (W66A/L/F/Y/H) on the spectral fingerprints of bsNOS key species by using a combination of resonance Raman, ATR-FTIR, and EPR spectroscopies. Our results show that the modulation of this interaction not only modifies the electron donation of the thiolate ligand but also affects the electronic structure of the Fe II NO complex. This seems to indicate that the reactivity of NOS major reactive intermediates (Fe II O 2 and Fe II NO) is controlled by the same means (the proximal H-bond network) in bacterial and mammalian NOSs.

EXPERIMENTAL PROCEDURES
Chemicals-All chemicals were purchased from Sigma-Aldrich. BH 4 and N -hydroxy-L-arginine were purchased from Enzo Life Sciences (Enzo Life Sciences Inc., Farmingdale, NY). NO and CO gases were purchased from Messer France SA (Asnières, France). NO-saturated solutions were freshly prepared by flushing NO gas through a previously degassed 100 mM potassium phosphate (KP i ) buffer at pH 7.4.
Molecular Biology-The bsNOS gene was a kind gift of Dr. Dennis J. Stuehr (16). 3 Wild-type bsNOS and mutants containing a His 6 tag attached to their N terminus were overexpressed in Escherichia coli strain BL21 (DE3) using a pET15B expression vector as described (16). Trp 66 mutations were added in the bsNOS gene using the QuikChange XL site-directed mutagenesis kit from Stratagene and synthetic mutagenic oligonucleotides. Oligonucleotides used to construct site-directed mutants in bsNOS were synthesized by Eurofins MWG. Silent mutations coding for the disappearance of the EcoRI (GAATTC) restriction site were incorporated into the oligonucleotides to aid in screening. Mutations (boldface type), the EcoRI restriction site SCHEME 1. Crystallographic structure of the active site of bsNOS highlighting the interactions between the heme, the proximal thiolate, and tryptophan 66. This structure was generated from the crystallographic structure of native bsNOS (Protein Data Bank entry 1M7V) (13) by using the Swiss-Pdb viewer Deepview and PovRay software (both available on the World Wide Web). This image illustrates the H-bond between Trp 66 and the thiolate ligand, as well as the -stacking between the tryptophan and the porphyrin.
Protein Expression and Purification-Wild-type and Trp 66 mutant bsNOS were expressed in E. coli. 400-ml cultures of terrific broth containing 125 mg/liter ampicillin were initiated with 500 l of stock glycerol bacterial culture and stirred at 250 rpm at 37°C. At A 600 ϭ 0.8 -1, starter cultures (400 ml) were used to inoculate 3.6 liter of the same medium. Protein expression was induced at A 600 ϭ 1 by adding 1 mM isopropyl-␤-Dthiogalactoside, and the cultures were supplemented with 500 M ␦-aminolevulinic acid. After 12 h of growth at 20°C, the cells were harvested by centrifugation at 6000 rpm for 20 min at 4°C and resuspended in ice-cold lysis buffer (0.1 M Tris-HCl, pH 8 -9, with 10% glycerol, 1 mM EDTA, and 0.25 M NaCl) containing 1 mg/ml lysozyme, 0.5 g/ml each leupeptin and pepstatin, 1 mM phenylmethanesulfonyl fluoride (PMSF), 50 units/ml DNase I (bovine pancreas type IV, Sigma) with or without 10 M H 4 B and 5 mM L-arginine. Cells were lysed by two cycles of French press at 700 p.s.i. The lysate was centrifuged at 16,000 rpm for 45 min at 4°C. (NH 4 ) 2 SO 4 (50 -55% w/v, final concentration) was added. After mixing for 30 min and centrifugation (30 min, 16,000 rpm at 4°C), the pellet was resuspended in binding buffer (0.1 M Tris-HCl, pH 8 -9, 10% glycerol, 0.25 M NaCl, 1 mM PMSF with or without 10 M H 4 B and 5 mM L-arginine). The supernatant was loaded on a column of Ni-ProBond resin (Invitrogen) pretreated with binding buffer containing 10 mM imidazole (buffer A). The column was washed with 10 volumes of buffer A. Bound protein was eluted with binding buffer containing 200 mM imidazole. Column fractions were pooled and were concentrated using Centriprep (10 units) (Millipore, Bedford, MA) with centrifugation at 5000 rpm. The concentrated proteins were dialyzed at 4°C against 2 liters of 0.1 M KP i buffer, pH 8 -9, containing 10% glycerol, 0.25 M NaCl, 1 mM PMSF, and 3 mM DTT in the presence or absence of 10 M H 4 B and 10 mM L-arginine for 24 h. After one change of buffer overnight, the protein was checked for heme content and purity by SDS-PAGE and stored at Ϫ80°C.
Preparation of WT and Mutant bsNOS Complexes-Samples were reconditioned in a KP i (100 mM, pH 7.4) buffer in the presence of different combinations of Arg (10 mM) and/or BH 4 (100 M to 1 mM) by three successive cycles of dilution/centrifugation in the final buffer using MicroCon membrane concentrators with a 30 kDa cut-off (Millipore). For samples containing BH 4 , a final dilution/concentration cycle with freshly prepared BH 4 buffer was performed just before the measurements.
Anerobic ferric NOS (Fe III ) was first prepared by 100 -200 cycles of alternate vacuum and argon refilling, directly in a quartz EPR tube (EPR and resonance Raman experiments) or in a quartz cuvette (ATR-FTIR and UV-visible spectrometry), both sealed with air-tight rubber septa. Ferrous samples (Fe II ) were obtained by reduction of Fe III NOS with the addition of a small volume of dithionite solution (5-100 mM) directly into the EPR tube (or the cuvette) using a gas-tight syringe (Hamilton, Reno, NV). Ferrous heme-CO (Fe II CO) samples were then obtained by flushing CO inside the EPR tube (or the cuvette) for 10 min to ensure CO saturation of the solution and complete CO binding to Fe II NOS, as verified by UV-visible absorption. Ferrous heme-NO complexes were formed by the addition with a gas-tight syringe of a small volume of a NO-saturated solution (50 -500 M final concentration) to the ferrous NOS solution. The ferric heme-NO (Fe III NO) samples were prepared by the addition of a small volume of a NO-saturated solution (50 -500 M final concentration) to the anaerobic ferric NOS solution. All UV-visible spectra were recorded at room temperature on a Uvikon XL (Secomam, Alès, France).
EPR Spectroscopy-9.4 GHz (X-band) EPR spectra were recorded on a Bruker ELEXSYS 500 spectrometer equipped with a standard TE cavity (Bruker) and an Oxford Instruments continuous flow liquid helium cryostat and a temperature control system. Simulations were performed by using the Easyspin software package (54).
Resonance Raman Spectroscopy-50-l samples of bsNOS Fe II CO complexes solutions at 70 -150 M were prepared in gas-tight quartz EPR tubes and disposed in a homemade spinning cell, at room temperature, to avoid local heating and to prevent photodissociation and degradation. Raman excitation at 441.6 nm was obtained with a helium-cadmium laser (Kimmon, Tokyo, Japan). Resonance Raman spectra were recorded using a modified single-stage spectrometer (Jobin-Yvon T64000, HORIBA Jobin Yvon S.A.S., Chilly Mazarin, France) equipped with a liquid N 2 -cooled back-thinned CCD detector. Stray scattered light was rejected using a holographic notch filter (Kaiser Optical Systems, Ann Arbor, MI). Spectra were recorded as the co-addition of 40 -60 individual spectra with CCD exposure times of 20 -30 s each. 3-6 successive sets of such spectra were then averaged. Neutral density filters were used for the Fe II CO complexes to decrease laser power (below 5 milliwatts at the sample) and avoid photodissociation and photo-oxidation. Spectral accuracy was estimated to be Ϯ1 cm Ϫ1 . Spectral resolution was about 3 cm Ϫ1 . Base-line correction was performed using GRAMS 32 software (Galactic Industries, Salem, NH).
ATR-FTIR Spectroscopy-Room temperature FTIR spectra were recorded using a Bruker IFS 66 Fourier transform infrared spectrometer (Bruker Optik GmbH, Ettlingen, Germany) coupled to a single reflection micro-ATR prism from Pike Technologies (Madison, WI). 30 l of 500 M bsNOS Fe II CO sample were prepared in a small quartz cuvette as described above. 10 l were deposited using a gas-tight syringe on the ZnSe crystal surface of the ATR unit. The crystal was sealed within a gastight in-house built chamber, which permitted the control of the atmosphere above the sample. Twenty-fold 250 co-added interferograms were averaged for each FTIR measurement. A water vapor spectrum was used for background correction. Base-line correction was achieved using the GRAMS 32 soft-ware package (Galactic Industries, Salem, NH). Each spectrum presented in this work corresponds to the averaging of 3-5 individual experiments.
Data Analysis-Identification of spectral components in Raman and/or FTIR bands was achieved as previously described by the combination of Fourier self-deconvolution and second order derivative analysis of the averaged spectra using GRAMS 32 software (Galactic Industries, Salem, NH) (55). The overlapping peaks were resolved by fitting the spectral region to Lorentzian functions using Origin 6.0 (OriginLab Corp., Northampton, MA). The NOS Raman bands were assigned following previous assignments on NOS (46,(55)(56)(57)(58)(59)(60).

Production of Trp 66 bsNOS Mutants-
The sulfur atom of NOS-proximal thiolate is naturally engaged in a strong H-bond interaction with the nitrogen proton of the indole ring of the vicinal tryptophan (see Scheme 1). In order to check the influence of this H-bond interaction on the proximal thiolate properties, we mutated the bsNOS Trp 66 residue into five different amino acids (see "Experimental Procedures"): a phenylalanine (W66F) and a tyrosine (W66Y) to maintain the aromatic ring but remove the H-bond; a histidine (W66H) that is believed to maintain and even strengthen this H-bond (52); a leucine (W66L) to remove both the H-bond interaction and the aromaticity, keeping some hydrophobic environment; and an alanine (W66A), to remove all possible interactions. We slightly modified the expression and purification protocol to take into account the potential fragility of these mutated proteins (see "Experimental Procedures"). The expression of most of the mutant proteins was generally similar to that of wild-type bsNOS and reached around 10 -20 mg of heme protein/liter of culture. However, yields of W66A were extremely weak, suggesting that the stability of the heme core was strongly altered by this mutation.

EPR Analysis of Native Wild-type and Mutated bsNOS-
Wild-type and mutated bsNOS electronic properties were analyzed by EPR spectroscopy in the absence (Fig. 1A) or presence (Fig. 1B) of L-Arg. In both cases, the dominant features correspond to a signal from the high spin pentacoordinated (HS-5c) Fe III heme moiety. This signal was simulated, taking into account the electronic Zeeman interaction (taken as isotropic) and the zero-field splitting interaction characterized by the parameters D and E. Because the zero-field splitting interaction is largely dominant over the Zeeman interaction, the important parameter obtained from the simulation is the ratio E/D that is listed in Table 1. The simulated spectra are displayed in Fig. 1 together with the experimental data. In the absence of substrate, the E/D value increases in the order WT Ͻ W66H Ͻ W66Y Ͻ W66F Ͻ W66L, with the main difference observed between the two groups WT/W66H and W66Y/F/L. The addition of L-Arg to the proteins greatly increases the structural homogeneity of the samples as indicated by the EPR spectra, which now display only a HS-5c Fe III signal with features narrower than those observed in the absence of substrate. The rhombicity of the HS signal varies continuously between the different bsNOS with the order W66H Ͻ WT Ͻ W66Y Ͻ W66F Ͻ W66L.
In the absence of substrate, all bsNOSs exhibit a fraction of low spin hexacoordinated (LS-6c) species. This is particularly true in the case of W66Y and W66F, whereas the proportion of LS-6c heme is much lower in the case of WT, W66H, and W66L. The LS-6c EPR spectra are shown in Fig. 2 for WT and W66H/Y/F along with the simulated spectra, and the parameters used for the simulations are listed in Table 2. Also for those low spin species, an effect of the mutation is observed on the EPR spectra, with the W66H mutant being similar to the WT, whereas the W66Y and W66F exhibit lower g-anisotropy and isotropic g. The Trp 66 mutation affects the rhombicity of both LS-6c species (in the absence of L-Arg) and HS-5c species (in the absence or presence of L-Arg).
These EPR results are in agreement with the UV-visible data recorded on wild-type and mutated bsNOS samples that show a predominant HS-5c for all proteins in the presence of substrate and cofactor and the presence of a fraction of LS-6c species in the absence of both cofactor and substrate (see Table 3 and supplemental Fig. S1).
UV-visible and Resonance Raman Investigation of the Fe II CO Complexes of bsNOS Proteins-We recorded the UV-visible spectrum of wild-type and mutated bsNOS in the presence of Experimental conditions were as follows: microwave frequency, 9.41 GHz; microwave power, 2 milliwatts; field modulation amplitude, 2 mT; field modulation frequency, 100 kHz; temperature, 10 K. The simulated spectra are shown as gray lines. The parameters used for the simulations are given in Table 1. a.u., arbitrary units.  (Table 3). Wild-type bsNOS exhibits a Soret wavelength around 446.5 nm. This band is blue-shifted for W66H (441.5 nm) and red-shifted for W66F and W66Y mutants (449 and 448.5 nm, respectively), suggesting again an opposite modification of the electron donation of the proximal ligand for the W66H and the W66F/Y mutants. The binding of L-Arg modifies the equilibrium between the P420 and P450 forms but also the wavelength of the Soret band, confirming that all mutants efficiently bind their natural substrate (Table 3 and supplemental Fig. S2). We also analyzed the binding of imidazole, a distal ligand of ferric NOS. We found that imidazole binds and converts bsNOS WT and W66H/F/Y mutants into a LS-6c complex with a Soret maximum around 426 nm (data not shown), as observed for other bacterial and mammalian NOSs (64 -67). However, the dissociation constant was found to vary as a function of the Trp 66 mutation. The affinity of imidazole was the highest for W66H (K d ϭ 0.4 mM), followed by WT (K d ϭ 0.95 mM), W66F (K d ϭ 3 mM), and finally W66Y (K d ϭ 6.5 mM). We recorded the resonance Raman spectra of the Fe II CO complex of the W66F, W66Y, and W66H mutants in the presence of both substrate and cofactor. They all indicate the major presence of an LS-6c Fe II CO complex, with some minor photodissociation contribution (data not shown) such as what we observed for the wild-type bsNOS (43). We analyzed the 450 -600 cm Ϫ1 spectral region (Fig. 3B). Spectra were deconvolved by a multi-Lorentzian function as described under "Experimental Procedures." The peak that corresponds to the ␦ Fe-C-O bending mode (40,43,45) was observed around 567 cm Ϫ1 for wildtype and mutated NOSs (Table 3), indicating the absence of significant alteration of the Fe II CO complex geometry. Whereas wild-type Fe II CO complex exhibits a Fe-CO frequency around 501 cm Ϫ1 (43), this frequency decreases to 500 cm Ϫ1 for W66F mutant but increases up to 504.5 cm Ϫ1 for W66H bsNOS (Table 3). These data indicate that the Fe-C bond is the strongest for W66H Fe II CO complex and then weakens for WT bsNOS and for W66F/Y mutants. Fig. 3A exhibits the ATR-FTIR spectra of the same complexes. We noticed a similar modification of the frequency of the Fe-CO stretching mode that was observed around 1920 cm Ϫ1 for W66H mutant, 1917 cm Ϫ1 for WT bsNOS, and 1913 cm Ϫ1 for W66F/Y mutants ( Table 3). The C-O bond is stronger for the W66H Fe II CO complex than for WT and weaker for the W66F/Y mutants. The concomitant decrease (respectively increase) of the Fe-CO and CO stretching frequencies reflects an increase (respectively a decrease) of FIGURE 2. EPR spectra of the low spin Fe III complex of bsNOS mutants in the absence of substrate and cofactor. Experimental conditions were as follows: microwave frequency, 9.41 GHz; microwave power, 0.25 milliwatt; field modulation amplitude, 1 mT; field modulation frequency, 100 kHz; temperature, 10 K. The simulated spectra are shown as gray lines. The parameters used for the simulations are given in Table 2. a.u., arbitrary units.

TABLE 2 Simulation parameters for the EPR spectra of the LS Fe III heme for wildtype and mutated bsNOS
Shown are the H strain values in MHz (full width at half-height line width describing broadening due to unresolved hyperfine couplings). g iso is the average of the three principal values g 1 , g 2 , and g 3 . g 1 (H strain) g 2 (H strain) g 3 (H strain) g iso ⌬g (g 3 (7) a Complexes obtained in the absence of substrate and cofactor. APRIL 8, 2011 • VOLUME 286 • NUMBER 14 the -competition between the distal and proximal ligands of the W66F/Y (respectively W66H) mutants (43,45,55).

Characterization of Heme-NO Complexes of Wild-type and Mutant bsNOS-The
Fe III NO complexes of WT and W66H/ F/Y mutants were found to be relatively stable both in the absence and presence of substrate and cofactor (see supplemental Fig. S3 for the UV-visible spectra). The mutation mostly affects the values of the Soret band maximum that varies between 433 (W66H) and 440 nm (WT; Table 3). The Fe II NO complexes of the bsNOS mutants also display spectral fingerprints similar to those of WT bsNOS in the presence of both L-Arg and BH 4 . Only the W66Y Fe II NO spectrum was found to exhibit a shoulder around 417 nm that is reminiscent of what was observed for the nNOS W409F/Y mutant (60,68). Here again, the wavelength of the Soret band maximum was observed between 434.5 nm (W66H) and 438 nm (WT; Table 3).
The Fe II NO complexes of bsNOS proteins were also analyzed by EPR spectroscopy (Fig. 4). In the presence of L-Arg and BH 4 , the observed spectra were dominated by the well known rhombic powder pattern with three different g-values and resolved hyperfine coupling with the nitrogen nucleus of the NO ligand (Fig. 4B), characteristic of a hexacoordinated structure (69). This was not the case for W66L (and a negligible fraction of W66Y) Fe II NO complex, whose spectrum corresponded to an axial powder pattern, characteristic of a pentacoordinated nitrosyl iron species. The proportion of the pentacoordinated form increased for all proteins in the absence of any substrate and cofactor (Fig. 4A). The trend in the stability/ loss of the proximal ligand is similar to the one we observed for the Fe II CO complexes; whereas WT and W66H Fe II NO complex exhibited an almost completely hexacoordinated structure, W66F/Y Fe II NO complexes corresponded to a mixture of hexa-and pentacoordinated species. W66L Fe II NO species remained entirely pentacoordinated, as expected. The stability of the proximal ligand seems to obey the following order: (WT/ H) Ͼ (F/Y) Ͼ L, which might reflect the variations in the electronic donation of the proximal ligand induced by the Trp 66 mutations.
The EPR spectra were simulated for all hexacoordinated species (WT and W66H/F/Y in the presence of L-Arg and BH 4 ; Fig.  4B) as well as for the pentacoordinated complex of W66L (Fig.  4A), taking into account the Zeeman interaction and the hyperfine coupling to the nitrogen nucleus of the NO ligand. Simulation parameters are reported in Table 4. With respect to the hexacoordinated species (WT and W66H/F/Y), the total g-anisotropy (⌬g) of the Fe II NO signal increased with the order W66H Ͻ WT Ͻ W66F Ͻ W66Y, whereas the isotropic g-value (g iso ) decreased in the same order ( Table 4). The isotropic hyperfine coupling of the nitrogen nucleus of the NO molecule also decreased in the same order, indicating that the spin density on the NO nitrogen is the greatest on W66H and decreases slightly for WT Ͼ W66F Ͼ W66Y. This indicates that Trp 66 mutations modify the electronic properties of Fe II NO complexes.

DISCUSSION
Trp 66 of bsNOS is engaged in an H-bond with the heme proximal ligand that is believed to tune the properties of the NOS Fe-S bond (53). We present here for the first time a comparative analysis of five different mutants of this residue that differentially affect the environment of the proximal ligand: (i) the F and Y mutants, already described for nNOS (70) and recently for eNOS and saNOS (53), correspond to a suppression of the H-bond; (ii) the H mutant, described for iNOS (52), is supposed to induce a strengthening of the H-bond; (iii) finally, the A and L mutants, as yet undescribed, should suppress both H-bond and -stacking interactions. A comparative analysis of   H 4 B (B). Experimental conditions were as follows: microwave frequency, 9.41 GHz, microwave power, 0.004 milliwatt; field modulation amplitude, 0.5 mT; field modulation frequency, 100 kHz; temperature, 10 K. The simulated spectra are shown as gray lines. The parameters used for the simulations are given in Table 4. a.u., arbitrary units.
the effects of these different mutations on bsNOS structural and electronic properties should allow a better understanding of the way this H-bond modulates NOS structure and function.
Control of the -Competition between NOS-distal and -proximal Ligands-The Fe II CO complex is commonly used as a probe to analyze the influence of heme environment on Fe II O 2 structure and reactivity (55,71,72). Vibrational spectroscopies, such as resonance Raman and FTIR spectroscopies, allow the characterization of Fe-CO and C-O stretching modes. The frequencies of these modes are sensitive to the electronic properties of the proximal ligand and to polar and/or steric interactions between CO and the heme distal pocket (71,72). Modifications of the electrostatic/polar distal environment of the CO ligand result in changes in the back donation from the iron d* orbital to the empty * orbital of CO. This leads to the well known inverse correlation between the frequencies of the Fe-CO and C-O (45). Additionally, there is a competition between the proximal and distal ligands for -bonding to the dz 2 orbital of the heme iron. Variations in the electronic properties of the proximal ligand will impact this competition and shift the Fe-CO / C-O correlation curve. Strong electron-donating ligands (such as the thiolate of NOS and cytochromes P450) will shift correlation lines toward low Fe-CO and C-O frequencies, whereas weak proximal ligand (in the case of globins or some peroxidases) will lead to higher Fe-CO and C-O frequencies.
Our results suggest that the mutations do not affect Fe II CO geometry (no change in the bending modes; Table 3). However, they show an increase in the Fe-CO and C-O frequencies for the W66H mutant, whereas the W66F and W66Y mutants are characterized by weaker Fe-CO and C-O bonds (Table 3). Plotted in the Fe-CO / C-O correlation graph (data not shown), these data show a change in the offset of the correlation line, univocally linked to the variation of the proximal ligand strength. This indicates that the mutation directly modifies the electron donating ability of the proximal thiolate; W66F and W66Y are characterized by stronger electron donation (in agreement with the removal of the H-bond), whereas W66H is characterized by weaker electron donation (in agreement with a strengthening of the H-bond).
In this report, using the inverse Fe-CO / C-O correlation, we are able for the first time to probe within a single NOS protein both the increase and the decrease of this -competition and to assign these variations to the modifications of the H-bonding interaction in which the proximal ligand is engaged.
Role of Tryptophan-Thiolate H-bond in bsNOS Stability-This variation in -competition has strong effects on the properties of bsNOS complexes, as reflected by the effects of the mutation on Fe II NO and Fe II CO stability ( Fig. 4 and supplemental Figs. S2 and S3). We observed a greater proportion of pentacoordinated forms of these complexes for W66F and W66Y bsNOS (due to a stronger -competition), whereas W66H hexacoordinated complex (with a weaker proximal ligand) seems as stable as (if not more stable than) WT complexes. The same trend is observed in the dissociation constants of the Fe III -Im complex. The affinity of imidazole for bsNOS heme is the greatest for W66H (weak -competition), decreases for WT, and decreases even more for W66F and W66Y (strong -competition). However, we do not observe a better stabilization of Fe II CO and Fe II NO species for W66Y; no H-bond with the tyrosine proton, such as the one observed for eNOS (53), can be deduced from our data. Our work clearly indicates that the tryptophan-thiolate H-bond interaction controls the stability of bsNOS Fe II -XO complexes, such as Fe II NO, Fe II CO, and most probably Fe II O 2 .
Additionally, our study reveals that other structural features contribute to bsNOS stability. Indeed, upon binding of a distal ligand, such as CO or NO, the W66L complex is quickly and fully converted into pentacoordinated species, indicating the loss of the proximal thiolate ligand. Compared with W66F and W66Y, this further decrease in stability might arise from the loss of specific interactions that stabilize NOS heme, such as the -stacking between the tryptophan indole ring and the porphyrin cofactor (Scheme 1). Furthermore, the W66A mutant exhibits a critical instability of the protein, as illustrated by the severe difficulties of purification. Accordingly, the EPR spectrum of the W66A ferric enzyme (data not shown) suggests a complete unfolding of the protein and a partial loss of the heme.
Effect of Trp 66 Mutations on the Electronic Structures of Fe II -XO Complexes-The variations in the electron-donating properties of the proximal thiolate seem to also exert an impact on the electronic structure of bsNOS complexes. For example, the Soret maximum of the Fe II CO spectra is red-shifted for W66F and W66Y mutants (a stronger electron-donating ligand) and blue-shifted for W66H bsNOS (weaker proximal ligand). These variations are also reflected by the shifts of the Soret maxima of Fe III NO and Fe II NO complexes (Table 3) and to a lesser extent by the modification of the low spin/high spin equilibrium. The variations of bsNOS electronic properties can also be appreciated by EPR spectroscopy. The g-anisotropy of LS-6c Fe III species, the minor fraction of native bsNOS, is greater for W66H/WT than for W66F/W66Y (Fig. 2 and Table  2). A similar distinction between W66F/W66Y on one hand and W66H/WT on the other hand is observed for the zero-field splitting rhombicity of the prevalent Fe III HS-5c species. Indeed, the E/D value is clearly greater for the W66F/W66Y

TABLE 4 Simulation parameters for the EPR spectra of Fe II NO complexes for wild-type and mutated bsNOS
The values A i correspond to the hyperfine coupling with the nitrogen nucleus of the NO ligand. g iso (respectively A iso ) is the average of the three principal values g 1 , g 2 , and g 3 (respectively A 1 , A 2 , and A 3 ). g strain is the full width at half-maximum of the Gaussian distributions of the g principal values. g 1 (g strain) g 2 (g strain) g 3 (g strain) g iso ⌬g (g 3 ؊ g 1 ) mutants than for W66H and WT ( Fig. 1 and Table 1). The same variations are observed for hexacoordinated Fe II NO complex. The anisotropy is greater for W66F/W66Y than for W66H and WT bsNOS ( Table 4). The effects of the variation of the proximal bond on the detailed electronic structure of Fe II NO complexes are more difficult to analyze. However, the comparison of the EPR fingerprints for this series of mutants indicates a significant contribution of the proximal H-bond in the distribution of the electronic density on the Fe II NO moiety. The spin density on the NO nitrogen seems maximal for W66H and decreases as the hydrogen bond strength on the proximal thiolate ligand is decreased (Table 4). These results show that changes in the electron-donating properties of the proximal ligand drastically modify the electronic structure of Fe III and Fe II NO complexes.
Influence of the Proximal H-bond on Fe II -XO Reactivity-The variations of NOS electronic structure induced by the changes in the proximal H-bond interactions are believed to modify its catalytic activity. Indeed, the rates of Fe II O 2 autoxidation and Fe II NO oxidation increase for W409F/Y nNOS (31,70). Reciprocally, the strengthening of this H-bond in the case of W188H iNOS induces a decrease of Fe II O 2 autoxidation and activation rates that eventually allow the observation of new reaction intermediates (52).
All of these results suggest that bacterial NOSs use the same regulation feature, the "push effect" (73), to finely tune its oxidative chemistry. They also underline the necessity to precisely delimit the proximal ligand electron donation. A too strong electron donation (e.g. for Phe/Tyr mutants) exacerbates the instability of NOS reaction intermediates and mostly leads to uncoupling processes and reactive nitrogen and oxygen species production. A weak electron donation (e.g. for the His mutant) leads to an insufficient reactivity of iron-oxo complexes (74). The functional analysis of our panel of mutants by the Griess assay (see supplemental Fig. S4 and Table S1) confirms this model. On one hand, the W66H mutant exhibits an extremely weak nitrite synthase activity (6 ϫ 10 Ϫ4 s Ϫ1 ), which corresponds to around 5% of WT standard activity. On the other hand, W66F activity remains comparable with that of WT bsNOS (around 50%). This difference could arise from uncoupling reactions due to a stronger electron density on the heme intermediates.

CONCLUSION
We present here a comprehensive analysis of the role of the H-bond in the control of NOS structural and electronic properties. We analyzed the effects of both suppression (W66F/Y) and strengthening (W66H) of this H-bond on Fe III , Fe II CO, and Fe II NO spectroscopic fingerprints.
We evidenced for the first time a tight correlation between the proximal H-bond network, the electron-donating properties of the proximal ligand, and the stability and reactivity of NOS. Our data indicate that the removal of the H-bond (W66F/Y) increases the electron donation properties of the proximal ligand, which in turn induces an increased -competition on the iron orbitals. This competition leads to the destabilization of Fe II -XO complexes, such as Fe II NO, Fe II CO, and putatively Fe II O 2 . This destabilization is reverted when the Phe/Tyr residues are replaced by a histidine that restores (and even strengthens) the proximal H-bond and thus diminishes the -competition. These results confirm and complete previous information obtained for the Phe/Tyr mutants of saNOS (Trp 56 ), eNOS (Trp 180 ) (53), and nNOS (Trp 409 ) (60,68) and highlight the crucial role of this proximal H-bond in the stability of NOS reactive intermediates.
Additionally, our results suggest that bacterial NOSs are genuine oxygenases that seem to utilize the electron donation from the thiolate to activate distal ligands, such as O 2 and NO, and to use the same regulation feature, namely the proximal H-bond to tune and control their catalytic activity. This pattern could also intervene in the interaction between NOS and other reactive ligands, such as peroxynitrite. A comparative analysis of the role of this H-bond in the mechanisms of dioxygen activation, of Fe II NO oxidation, or of peroxynitrite activation could help understand the biological specificities of mammalian and bacterial NOSs.