Direct evidence for nitric oxide production by a nitric-oxide synthase-like protein from Bacillus subtilis.

Nitric-oxide synthases (NOSs) are widely distributed among prokaryotes and eukaryotes and have diverse functions in physiology. Recent genome sequencing revealed NOS-like protein in bacteria, but whether these proteins generate nitric oxide is unknown. We therefore cloned, expressed, and purified a NOS-like protein from Bacillus subtilis (bsNOS) and characterized its catalytic parameters in both multiple and single turnover reactions. bsNOS was dimeric, bound l-Arg and 6R-tetrahydrobiopterin with similar affinity as mammalian NOS, and generated nitrite from l-Arg when incubated with NADPH and a mammalian NOS reductase domain. Stopped-flow analysis showed that ferrous bsNOS reacted with O(2) to form a transient heme Fe(II)O(2) species in the presence of either Arg or the reaction intermediate N-hydroxy-l-arginine. In the latter case, disappearance of the Fe(II)O(2) species was kinetically and quantitatively coupled to formation of a transient heme Fe(III)NO product, which then dissociated to form ferric bsNOS. This behavior mirrors mammalian NOS enzymes and unambiguously shows that bsNOS can generate NO. NO formation required a bound tetrahydropteridine, and the kinetic effects of this cofactor were consistent with it donating an electron to the Fe(II)O(2) intermediate during the reaction. Dissociation of the heme Fe(III)NO product was much slower in bsNOS than in mammalian NOS. This constrains allowable rates of ferric heme reduction by a protein redox partner and underscores the utility of using a tetrahydropteridine electron donor in bsNOS.

and NO, with N-hydroxy-L-arginine (NOHA) formed as an enzyme-bound intermediate (5). All mammalian NOSs are bidomain proteins comprised of an N-terminal oxygenase domain (NOSoxy) that binds protoporphyrin IX (heme), 6R-tetrahydrobiopterin (H 4 B), and Arg, and a C-terminal flavoprotein domain (NOSred), linked together by a calmodulin (CaM) binding sequence (5). NOS flavoprotein domains are similar to NADPH-cytochrome P450 reductase and related electron transfer flavoproteins (5) and function to provide NADPHderived electrons to the ferric heme for O 2 activation during NO synthesis.
Recent genome sequencing revealed that NOS-like proteins exist in many prokaryotes including Deinococcus radiodurans, Bacillus subtilis, Bacillus halodurans, Bacillus anthracis, 2 and Staphylococcus aureus 2 (6 -8). We recently sequenced, cloned, purified, and characterized D. radiodurans NOS-like protein (deiNOS) whose sequence is 34% identical and 52% conserved to the oxygenase domain of mammalian nitric-oxide synthases (NOSoxy). Purified deiNOS was dimeric, bound substrate Arg and cofactor H 4 B, and had a normal heme environment, despite its missing N-terminal structures that in NOSoxy bind Zn 2ϩ , the dihydroxypropyl side chain of H 4 B, and help form an active dimer in mammalian NOS (4). The deiNOS heme accepted electrons from a separate mammalian NOS reductase and generated nitrite from Arg or NOHA in reactions stimulated by H 4 B. However, the oxyhemoglobin assay failed to show that deiNOS synthesized NO under these circumstances. Therefore, fundamental questions remain regarding the exact nature of the nitrogen oxide product formed by prokaryotic NOS-like proteins. To help address this issue we characterized the NOSlike protein from B. subtilis (bsNOS). The results establish its Arg and H 4 B binding, product formation in multiple turnover reactions, nature and kinetics of heme transitions during Arg or NOHA oxidation under single turnover conditions, and unambiguously show that it produces NO as a product.

EXPERIMENTAL PROCEDURES
Materials-All regents and materials were obtained from Sigma or sources reported previously (9).
Molecular Biology-The NOS gene of B. subtilis (ATCC) was amplified by PCR from genomic DNA. PCR primers generated a NdeI site before the 5Ј start codon and a BamHI site after the 3Ј stop codon and the amplified fragment cloned into a pET15B expression vector. B. subtilis NOS DNA in pET15B vector transformed into Escherichia coli strain BL21 (DE3) for protein expression.
Expression and Purification of bsNOS and nNOSred-bsNOS had a His 6 tag attached to its N terminus to aid purification. Proteins were overexpressed in E. coli strain BL21 (DE3) and purified by using chromatography on Ni 2ϩ -nitrilotriacetic acid resin for bsNOS and 2Ј,5Ј-ADP-Sepharose for nNOSred as described earlier (4). * This work was supported by National Institutes of Health Grants CA53914 (to D. J. S.), grants from American Heart Fellowship (to S. A.), and American Heart Association Grant AHA01601598 (to K. S. A.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Arg Binding-Arg binding affinity was studied at 25°C by perturbation difference spectroscopy using 10 mM imidazole according to methods described previously (4).
NO Synthesis, Nitrite Formation, and Citrulline Production-The steady state activities in the reconstituted system containing bsNOS and nNOSred in a 1:1.5 molar ratio were determined at 25°C using spectroscopic and high performance liquid chromatography fluorometric assays that were described previously in detail (4).
Measurement of Apparent K m for H 4 B or Tetrahydrofolate (THF)-Apparent K m values were determined by double-reciprocal analysis of the NADPH-dependent nitrite formation against various concentrations of H 4 B or THF in a reconstitution system containing nNOSred and bsNOS.
Stopped-flow Experiments-Oxygen binding spectra were recorded in a stopped-flow instrument equipped with a rapid scanning diode array device (Hi-Tech MG-6000) designed to collect 96 complete spectra within 144 ms. Rapid scanning experiments involved mixing anaerobic solutions containing dithionite-reduced bsNOS, 40 mM EPPS buffer, pH 7.6, 0.5 mM DTT, 150 M NaCl, and 1 mM Arg or NOHA with airsaturated buffer solutions at 10°C in the presence or absence of 100 M H 4 B or THF. Formation and decay of the Fe(II)O 2 complexes were followed at 410 or 440 nm (10). Signal-to-noise ratios were improved by averaging 10 individual traces. The diode-array data were then fit to different reaction models by a Specfit program from Hi-tech Ltd. to obtain the calculated number of species, their individual spectra, the concentration of each species versus time, and rate constants for each transition.

RESULTS AND DISCUSSION
Primary Structure Analysis-We identified a 1-kb DNA segment that coded for a 369-amino acid bsNOS protein that has 35% identity and 49% conservation with deiNOS and 42% identity and 55% conservation with mouse iNOSoxy. The similarities among the primary sequences of bsNOS, deiNOS, and mouse iNOSoxy are shown in Fig. 1A. Structural elements that make up the iNOS catalytic core are well conserved in bsNOS. This includes residues that contact the heme, bind the pteridine ring of H 4 B, and position substrate Arg. Like deiNOS, a notable similarity is that bsNOS is missing an extended portion of N-terminal sequence found in the mammalian enzymes. In mammalian NOSs this region codes for an N-terminal hook, a Zn 2ϩ binding site, and contains residues that participate in forming the dimer interface and in binding the dihydroxypropyl side chain of H 4 B (Fig. 1A). Thus, bsNOS is a heme protein similar to mammalian NOSs that also contains residues likely to generate functional Arg and H 4 B binding sites.
Physical and Spectral Characteristics of bsNOS-Recombi-

FIG. 1. Comparative analysis of bsNOS, deiNOS, and iNOSoxy sequences and structural elements.
A, aligned iNOSoxy (top), bsNOS (middle), and deiNOS (bottom) sequences with mapped residue function, secondary structure, and contributions to the dimer interface as determined from iNOSoxy structures (11). NOS sequences are color-coded to highlight zinc ligands Cys 104 and Cys 109 and proximal heme ligand Cys 194 (yellow background), Arg-binding residue (blue block), and H 4 B-binding residues (red block). Above, black arrows show ␤-strands, thick boxes show ␣-helices. Key sequence stretches involved in forming the dimer interface and cofactor binding sites are boxed in magenta and denoted, N-terminal hook, switch region (zinc loop), N-terminal pterin binding, helical T, and helical lariat. B, spectrum of DTT-bound bsNOS and displacement of DTT upon H 4 B and Arg binding. Spectra were recorded after incubating bsNOS for 15 min at 25°C under the indicated conditions. nant purified bsNOS migrated in a denaturing SDS-PAGE gel at a molecular mass of ϳ40 kDa, identical to its cDNA calculated molecular mass. Its migration in a gel filtration column indicated that bsNOS was predominantly dimeric in its native form (data not shown). Thus, bsNOS is a homodimer despite its missing N-terminal elements that stabilize dimeric structures of mammalian NOS proteins (11). Apparently, differences among amino acids in the subunit interface and elsewhere must minimize the relative importance of the N-terminal elements on dimer stability. bsNOS should therefore be valuable for identifying new residues and regions of importance.
Spectral changes obtained upon Arg or H 4 B binding are shown in Fig. 1B. Ferric bsNOS showed characteristic absorbance maxima at 460 and 380 nm, indicating its heme bound DTT to form a bisthiolate species identical to mammalian NOS (12). Adding H 4 B caused a spectral transition that generated a broad Soret peak at 400 nm, indicating displacement of DTT ligand. Adding Arg caused the Soret peak to shift to 393 nm, indicating that Arg fully displaced DTT and stabilized the bsNOS heme in a five-coordinate high spin state. These spectral properties show that bsNOS binds H 4 B and Arg and has a similar heme environment compared with mammalian NOS.
We next monitored Arg displacement of heme-bound imidazole to determine the Arg binding affinity of bsNOS. Upon sequential addition of Arg to H 4 B-saturated bsNOS, there was a concentration-dependent spectral shift that indicated Arg could achieve a complete displacement of heme-bound imidazole (data not shown). The apparent k d value for Arg in presence of 10 mM imidazole and 20 M H 4 B was derived by doublereciprocal analysis and was 50 Ϯ 4 M in bsNOS, as compared with 97 Ϯ 10 M and 55 Ϯ 4 M in deiNOS and nNOSoxy, respectively (Table I). We conclude that Arg binding affinity of bsNOS is similar to mammalian NOS, consistent with its containing a conserved glutamate residue essential for high affinity Arg binding in mammalian NOS.
Catalytic Activity in Reconstituted Systems-NO synthesis by animal NOSs involves electron transfer between reductase and oxygenase domains in a NOS dimer. Because bsNOS lacks an attached reductase domain, its heme can only receive electrons from a separate donor protein. The Bacillus genome contains several electron transfer proteins, including an FAD-and FMN-containing NADPH oxidoreductase that is similar to the mammalian enzyme cytochrome P450 reductase (13). We therefore examined its catalytic ability in a reconstitution system that mixed bsNOS with purified nNOSred (as a model reductase domain) that contained a functional CaM binding site (4). Activities were assayed in the presence or absence of H 4 B and CaM, using Arg or NOHA as substrate. We measured nitrite and citrulline accumulation in an 8-min assay as reported earlier (4). Nitrite formation from Arg or NOHA was time-and enzyme concentration-dependent in reconstitution reactions that contained H 4 B and bsNOS (data not shown). Nitrite and citrulline synthesis activities of bsNOS were approximately similar to nNOSoxy at room temperature (Table I). Individual oxygenase proteins or NOSred alone had no activity under any of the assay conditions (data not shown). Our results imply that bsNOS can accept NADPH-derived electrons from free nNOSred to convert Arg or NOHA to citrulline and nitrite in a reaction.
Affinity toward H 4 B and Tetrahydrofolate-Using the reconstitution system we determined an apparent K m for H 4 B of 100 nM for bsNOS versus 30 Ϯ 10 nM for nNOSoxy (Table I). Thus, bsNOS affinity toward H 4 B is greater than deiNOS and approaches that of nNOSoxy or other mammalian NOSs whose apparent K m values range between 50 nM to 1 M. THF is a tetrahydropteridine that can also support catalytic activities of deiNOS (4). The bsNOS was found to productively bind THF with an apparent K m value of 0.4 M, which is 50 times lower than that of deiNOS (20 Ϯ 5 M). Thus, bsNOS also differs from mammalian NOS in being capable of utilizing THF in place of H 4 B to support catalysis.
The primary structure of bsNOS is entirely consistent with its behavior toward H 4 B and THF. Several residues that surround the H 4 B ring in mammalian NOSs and position it near the heme are all conserved in bsNOS. Earlier we thought that deiNOS had poor affinity toward H 4 B due to the absence of N-terminal residues that in animal NOSs bind the 6-dihydroxypropyl side chain of H 4 B (11). However, bsNOS also lacks this N-terminal region but has affinity similar to mammalian NOS. This suggests that poorer H 4 B affinity of deiNOS may be due to other structural differences, for example the unique absence of ␤4a, ␤4c, and ␤5 helix elements in deiNOS. In any case, the missing N-terminal region probably allows bsNOS to bind THF, which contains a bulky substituent in place of the 6-dihydroxypropyl side chain of H 4 B. (10,14). We utilized rapid-scanning stopped-flow spectroscopy to examine spectral and kinetic properties of the Fe(II)O 2

intermediate in bsNOS and observed whether H 4 B or THF would affect its
properties. An anaerobic solution of dithionite-reduced bsNOS containing Arg with or without H 4 B was rapid-mixed with air-saturated buffer at 10°C. The initial ferrous species displayed a Soret peak at 408 nm in both cases. This species converted within 120 ms after mixing into a transient species with Soret peak at 427 nm in both cases. The transient species converted to stable ferric enzyme that displayed a Soret peak at 393 nm and visible absorbance band at 650 nm (data not shown). Thus, bsNOS formed a transient Fe(II)O 2 intermediate that is quite similar to that of deiNOS or nNOSoxy (4, 10).
The formation and decay kinetics of the Fe(II)O 2 intermediate were determined by monitoring absorbance change at 407 or 440 nm versus time. The direction of absorbance change at these two wavelengths was reversed as expected but otherwise proceeded with identical kinetics (data not shown), as found previously for deiNOS or nNOSoxy (4,10). Spectral change during Fe(II)O 2 formation or disappearance was described by a single exponential equation in all cases, suggesting both transitions are monophasic. Rates of Fe(II)O 2 formation and decay Heme Transitions during NOHA Oxidation by Ferrous bs-NOS-To examine formation and reactivity of the Fe(II)O 2 species in the second step of NO synthesis, we utilized rapidscanning stopped-flow spectroscopy to identify consecutive heme transitions that occur during oxidation of NOHA in a single turnover reaction. A solution of ferrous bsNOS saturated with H 4 B and NOHA was rapid-mixed with air-saturated buffer at 10°C. Global analysis of the rapid-scan data showed that it best fit to a reaction model A to B to C to D. The calculated spectra for species A-D are shown in Fig. 2A. The initial ferrous enzyme (species A) displayed a Soret peak at 408 nm, consistent with the spectrum of ferrous mammalian NOS taken under identical conditions (10,15). Species B has absorbance maxima at 427 and 555 nm, consistent with the spectrum of the nNOSoxy Fe(II)O 2 intermediate at 10°C in presence of H 4 B and NOHA (10,15). Species C has absorbance maxima at 440, 547, and 585 nm, identical to the spectrum of the ferric-NO complex of deiNOS and nNOSoxy (15). Species D has an absorbance maxima at 393 and 630 nm, identifying it as ferric bsNOS. Fig. 2B shows how the concentrations of species A-D change during the reaction. The magnitude of ferric heme-NO complex accumulation indicates near quantitative NOHA oxidation occurred in the single turnover reaction (15). It also shows that newly formed NO combines with the bsNOS ferric heme in a near geminate manner before leaving the active site, just like during NO synthesis by mammalian NOS (15). The results suggest a mechanism for NOHA oxidation by bsNOS shown in Scheme 1.
Ferrous heme binds O 2 in the initial step to generate a Fe(II)O 2 transient intermediate. This intermediate then converts to a ferric heme-NO complex as an immediate product. This transformation represents k cat for the NOHA reaction. NO then dissociates from heme to generate ferric bsNOS. The fact that the heme-NO product was ferric rather than ferrous establishes that bsNOS produced NO rather than other N-oxides like nitroxyl, which would have generated a spectrally distinguishable ferrous heme-NO product (4,12).
In a replica NOHA reaction run without H 4 B we observed no heme-NO complex or citrulline production, although an identical Fe(II)O 2 intermediate was generated (data not shown). Thus, H 4 B is obligatory for NO synthesis by bsNOS in the single turnover reaction, as is true for mammalian NOSoxy enzymes (12,16). In mammalian NOSs, H 4 B is implicated as an electron donor (14,17) and was recently shown to provide an electron to the heme Fe(II)O 2 intermediate during Arg hydroxylation in a single turnover reaction (14). In that circumstance electron transfer from H 4 B is responsible for speeding the disappearance of the Fe(II)O 2 intermediate (10). The similar behavior of bsNOS regarding H 4 B binding and kinetic effects strongly suggest that a native reduced pteridine performs an identical redox function during its catalysis. Indeed, the B. subtilis genome contains all enzymes needed to synthesize H 4 B from its GTP precursor (7). Thus, the implication of our current work is that NO synthesis by bsNOS proceeds through the same mechanism as in mammalian NOS and may employ a tetrahydropteridine to provide the second electron required for O 2 activation.
Implications for bsNOS Function-During NOHA oxidation by bsNOS the heme transitions associated with O 2 binding or k cat were somewhat slower but within the range of rates we have measured in mammalian NOSoxy reactions under the same conditions (10,14,15). In contrast, the Fe(III)NO dissociation step in bsNOS was 10 -20-fold slower than in mamma-  lian NOSoxy, which has ranged from 2 to 5 s Ϫ1 at 10°C (15,18). This difference is important, because it sets the maximum rate for NO release from bsNOS during steady state NO synthesis. Moreover, it constrains the rate of ferric heme reduction to remain near or below 0.2 s Ϫ1 if the enzyme is to release NO. This is because when the rate of ferric heme reduction exceeds NO dissociation, a majority of the Fe(III)NO product becomes reduced to a ferrous heme-NO species instead of releasing NO (19,20). Dissociation of NO from the ferrous heme-NO complex is very slow (21), so instead the complex reacts with O 2 to generate higher N-oxides like peroxynitrite or nitrate (19,20,22). Thus, our kinetic analysis predicts that bsNOS will either have a relatively slow NO production during steady state due to its slow Fe(III)NO dissociation or will generate higher Noxides in place of NO if ferric heme reduction as catalyzed by its native redox partner exceeds 0.2 s Ϫ1 .
The constraint on heme reduction rate caused by slow Fe(III)NO dissociation will also prevent bsNOS from coupling its Fe(II)O 2 formation to substrate oxidation, unless it receives a second electron more quickly than the first during O 2 activation. Consider that the bsNOS Fe(II)O 2 intermediate is unreactive toward Arg or NOHA until it obtains another electron. However, its Fe(II)O 2 intermediate is inherently unstable as demonstrated by a decay rate of 0.4 s Ϫ1 in the Arg-bound, H 4 B-free enzyme (Table II). Thus, for bsNOS to perform substrate oxidation instead of superoxide production, the Fe(II)O 2 intermediate needs to receive an electron at a rate greater than 0.4 s Ϫ1 . But such high rates would compromise NO release from bsNOS due to its slow NO dissociation as explained above. bsNOS probably solves this paradox by using H 4 B, THF, or a similar donor as a source of the second electron, because they appear to reduce the Fe(II)O 2 species at rates that exceed its oxidative decay (Table I). Thus, by having a tetrahydropteridine provide the second electron during O 2 activation, bsNOS can couple its Fe(II)O 2 formation to substrate oxidation and employ a ferric heme reduction rate that is slow enough to allow for NO release.
In sum, NO synthesis by a prokaryotic NOS-like protein is now established, but a number of related issues still need to be explored. These include determining (i) the crystal structure of a prokaryotic NOS, (ii) the native redox partner(s), (iii) what N-oxide product is generated during steady state synthesis, (iv) what conditions induce expression of these proteins in the bacterium, and (v) if allosteric factors exist that modulate kinetic parameters of the enzyme, as occurs in soluble guanylate cyclase (23). Continued investigation should provide deeper understanding of NOS structure-function and the evolutionary consequence of the NOS gene.