The 42-Amino Acid Insert in the FMN Domain of Neuronal Nitric-oxide Synthase Exerts Control over Ca2+/Calmodulin-dependent Electron Transfer*

The neuronal and endothelial nitric-oxide synthases (nNOS and eNOS) differ from inducible NOS in their dependence on the intracellular Ca2+ concentration. Both nNOS and eNOS are activated by the reversible binding of calmodulin (CaM) in the presence of Ca2+, whereas inducible NOS binds CaM irreversibly. One major divergence in the close sequence similarity between the NOS isoforms is a 40–50-amino acid insert in the middle of the FMN-binding domains of nNOS and eNOS. It has previously been proposed that this insert forms an autoinhibitory domain designed to destabilize CaM binding and increase its Ca2+ dependence. To examine the importance of the insert we constructed two deletion mutants designed to remove the bulk of it from nNOS. Both mutants (Δ40 and Δ42) retained maximal NO synthesis activity at lower concentrations of free Ca2+ than the wild type enzyme. They were also found to retain 30% of their activity in the absence of Ca2+/CaM, indicating that the insert plays an important role in disabling the enzyme when the physiological Ca2+concentration is low. Reduction of nNOS heme by NADPH under rigorous anaerobic conditions was found to occur in the wild type enzyme only in the presence of Ca2+/CaM. However, reduction of heme in the Δ40 mutant occurred spontaneously on addition of NADPH in the absence of Ca2+/CaM. This suggests that the insert regulates activity by inhibiting electron transfer from FMN to heme in the absence of Ca2+/CaM and by destabilizing CaM binding at low Ca2+ concentrations, consistent with its role as an autoinhibitory domain.

The nitric-oxide synthases (NOSs) 1 are a family of dimeric enzymes found in a variety of organisms and cell types. They generate nitric oxide through the five-electron oxidation of L-arginine, consuming two equivalents of molecular oxygen and three electron equivalents derived from NADPH (1,2). All the NOSs bind heme and (6R)-5,6,7,8-tetrahydro-L-biopterin within a structurally unique N-terminal oxygenase domain (3) and bind FMN and FAD within a C-terminal reductase domain that is related to microsomal cytochrome P-450 reductase (4 -6). NOS dimerization occurs largely through contacts between oxygenase domains and is augmented by the binding of (6R)-5,6,7,8-tetrahydro-L-biopterin (7). Evidence suggests that the reductase domains may supply electrons to the heme domain of the alternate subunit, making dimerization essential for NO synthase activity (8). The many different NOS sequences currently available for mammalian enzyme variants show all to be closely related. However, they can be broadly grouped into three different isoforms. Neuronal NOS (nNOS) and endothelial NOS (eNOS) were found to be constitutively expressed in their respective cell types, whereas inducible NOS (iNOS) was first found to be induced by cytokines in macrophages (9). The three isoforms can be categorized by both amino acid sequence and functionality (10), because they exhibit different rates of catalysis and sensitivity to regulatory ligands. The constitutive NOSs are regulated by the reversible binding of calmodulin (CaM) through changes in the intracellular Ca 2ϩ concentration. iNOS, however, is expressed with CaM bound permanently and is insensitive to changes in Ca 2ϩ concentration (11). The CaM binding site lies within a linker region between the N-terminal oxygenase and C-terminal reductase domains of the NOSs and consists of 20 -25 amino acids. As Fig. 1 shows, the sequence of this region is not well conserved between the three NOS isoforms, and it has been found through the study of chimeric enzymes that it plays a major role in determining the CaM binding affinity (12,13). Regulation of NO synthase activity by CaM has been shown to occur via the control of FMN to heme electron transfer across the domain-domain interface (14). Additionally, CaM has been shown to activate the nNOS reductase domain both within the native dimer and in a mutant consisting of the reductase domain only (15). The iNOS reductase domain, on the other hand, appears to be influenced little by CaM binding (16). As well as having different CaM binding sites to iNOS, the constitutive NOSs contain an additional 40 -50-amino acid insert in the middle of a conserved region of the FMN-binding subdomain (see Fig. 1). The FMNbinding subdomain is related by amino acid sequence to the large, diverse family of flavodoxins, which generally function as small electron transfer proteins and to the FMN-binding subdomain of cytochrome P-450 reductase, which behaves similarly. However, the presence of such an insert is highly unusual. Furthermore, the insert lies between the two strands of ␤-sheet that form key binding interactions with the FMN, by analogy with the structure of cytochrome P-450 reductase (17). It has been proposed that the region acts as an autoinhibitory domain, competing with CaM for space on the interdomain linker (18,19). Salerno et al. (19) showed that peptides based on the sequence of the eNOS insert are able to inhibit CaM binding to eNOS and nNOS. However, peptides based on the nNOS insert were less effective. This indicates that in the case of eNOS at least, the insert may make contact with a sequencespecific binding site in the vicinity of the CaM-binding site. To investigate the role of the nNOS insert, we constructed two mutants in which 40 and 42 amino acids have been deleted (⌬40 and ⌬42), removing the bulk of the insert region (see Fig.  1). We also constructed a Y889A mutant of nNOS designed to destabilize the binding of FMN to examine the effect of CaM binding on the properties of the FAD-binding subdomain.
Molecular Biology-Rat nNOS cDNA was a kind gift of Dr. S. H. Snyder (Johns Hopkins School of Medicine). The ⌬40 nNOS mutant plasmid (pSD1⌬40) was generated by ligating the ScaI site at Thr 871 in the protein sequence to the BanI site at Arg 829 after blunt-ending with the Klenow fragment of DNA polymerase I. This was conducted on the plasmid pSD1 consisting of the 622-kilobase Eco0109I fragment of nNOS (Gly 721 -Arg 928 ) in the plasmid pUC19. The mutation site was transferred within the BglII-SphI fragment of nNOS (Ile 776 -His 897 ) to the corresponding region of the plasmid pBS-nNOS (21). The following mutations were generated using the oligonucleotide-directed dual amber, long and accurate polymerase chain reaction kit for site-directed mutagenesis obtained from Takara Shuzo (Kyoto, Japan). The Y889A mutation was generated using oligonucleotide 1 (5Ј-CTCGGGCGGC-CCCCCACTT-3Ј), and the plasmid pSD2 was used as a template. pSD2 was constructed from the BamHI/SphI fragment of pSD1 and the plasmid pKF18k. The mutation site was transferred as described above. The ⌬42 mutation was generated using oligonucleotide 2 (5Ј-TGAG-GCACACTGGA-GCCAATGTGAGGTT-3Ј), and the plasmid pSD3 was used as a template. pSD3 was constructed in the same way as pSD2, using the mutant plasmid pSD1⌬40 as the source of the BamHI/SphI fragment. Therefore, this mutant (⌬42) contains the 120-base pair deletion of ⌬40 and a further 6-base pair deletion comprising amino acids Pro 873 and Leu 874 . The mutation site was transferred as described above. All mutations were confirmed by sequencing over the appropriate regions. In each case, the mutant nNOS cDNA was subcloned into the XhoI site of the expression vector pAM82, and the plasmids were amplified in HB101 Escherichia coli cells and used to transform Saccharomyces cerevisiae spheroplasts. The nNOS enzyme was expressed under the acid phosphotase promoter in yeast as described previously (21,22).
Preparation of nNOS-nNOS wild type and mutant enzymes were purified using 2Ј,5Ј-ADP-Sepharose and calmodulin-Sepharose column chromatographies (Amersham Pharmacia Biotech) as described previously (23)(24)(25)(26). 5 M FMN was included in the buffers of the ⌬40 and ⌬42 mutants throughout preparation to preserve activity, and the procedure was completed within 1 day. Aliquots were flash-frozen in liquid nitrogen prior to storage at Ϫ80°C. The purified enzyme was more than 95% pure as judged by SDS-polyacrylamide gel electrophoresis. For EGTA titration assays, enzyme was used after ADP-Sepharose chromatography to avoid contamination with additional EGTA. Enzyme concentrations were determined from the [co-reduced]-[reduced] difference spectrum using ⌬⑀ 444 -467 nm ϭ 55 mM Ϫ1 cm Ϫ1 (24).
Determination of Flavin Content-Stoichiometries of FMN and FAD relative to heme (the concentration of which was determined as above) were determined as described by Louerat-Oriou et al. (27) using a modification of the procedure of Faeder and Siegel (28). NOS samples were analyzed after purification using calmodulin-Sepharose column chromatography, as above. The elution buffer consisted of 50 mM Tris-HCl (pH 7.5), 50 mM NaCl with 5 mM EGTA. Enzyme samples (of approximately 10 M) were denatured by diluting 10-fold by 6 M guanidinium chloride and incubating for 10 min at 22°C. Denatured samples (300 ml) were then diluted 10-fold into either 6 M guanidinium chloride or 0.9 M NaH 2 PO 4 containing 1 M guanidinium chloride. The fluorescence response ( ex ϭ 450 nm; em ϭ 525 nm) in these two buffers was compared with the response of FMN and FAD standards and used to calculate the flavin concentrations.
Assays of Enzyme Activity-The rate of NO formation was determined from the NO-mediated conversion of oxyhemoglobin to methemoglobin, monitored at 401 nm using a methemoglobin minus oxyhemoglobin extinction coefficient of 49 mM Ϫ1 cm Ϫ1 (29), with 10 M oxyhemoglobin, 0.1 mM NADPH, and 1 mM L-arginine. The NADPH oxidation rate was determined spectrophotometrically as an absorbance decrease at 340 nm using an extinction coefficient of 6.22 mM Ϫ1 cm Ϫ1 , with the concentration of NADPH at 0.1 mM. The ferricyanide reduction rate was determined by monitoring the decrease at 420 nm using an extinction coefficient of 1.01 mM Ϫ1 cm Ϫ1 , with concentrations of ferricyanide at 1 mM and NADPH at 0.5 mM unless otherwise indicated. Cytochrome c reduction was monitored as an absorbance increase at 550 nm using an extinction coefficient of 21 mM Ϫ1 cm Ϫ1 , with 100 M cytochrome c and 100 M NADPH. All assays were carried out at 25°C in 50 mM Tris-HCl (pH 7.5) buffer containing 10 units/ml superoxide dismutase and 100 units/ml catalase. For assays containing Ca 2ϩ and CaM, the concentrations were 1 mM and 10 g/ml, respectively.
EGTA Titrations-In the above assays for NO synthesis and ferricyanide reduction, the concentration of free Ca 2ϩ was modulated by introducing variable quantities of EGTA into the assay mixture. For these assays, the concentration of Ca 2ϩ was kept at precisely 50 M, and CaM was kept at 10 g/ml. The final concentration of enzyme in the assay mixture was 0.10 (Ϯ 0.01) M for both ferricyanide reduction and NO synthesis experiments.
Anaerobic Heme Reduction-Concentrated, purified enzyme (e.g. 2 ml x 20 M) was passed through an anaerobic Sephadex G25 column (1.5 ϫ 10 cm) contained within a Hirasawa Works (Japan) glove box under an 80% N 2 /10% H 2 /10% CO 2 atmosphere with O 2 at less than 50 ppm. Spectra were recorded on a Shimadzu 1201 scanning UV/Vis spectrophotometer contained within the anaerobic environment. Aliquots of anaerobic solutions of NADPH, L-arginine, Ca 2ϩ /CaM, and CO-saturated buffer were added as described. The temperature of the glove box was maintained at 15°C throughout. Fig. 1 compares the amino acid sequences of the FMN-binding subdomains of the three NOS isoforms and that of cytochrome P-450 reductase, the most similar protein outside the NOS family. According to the crystal structure of cytochrome P-450 reductase (17), the FMN is bound between two aromatic resi-FIG. 1. Amino acid sequence alignment of the FMN-binding subdomains of rat nNOS, human eNOS, human iNOS, and rat cytochrome P-450 reductase (redu). Sequences were taken from the SWISS-PROT data base and aligned using ClustalX (20). The nNOS calmodulin-binding site (725-745), the nNOS region deleted in the ⌬40 mutant (⌬40 deletion), and two additional amino acids deleted in the ⌬42 mutant (⌬2) are indicated. The two aromatic residues involved in FMN binding (Phe 809 and Tyr 889 for nNOS) are marked by a down arrow. Conserved residues are marked by an asterisk.

Construction of the Mutants-The alignment shown in
dues, Tyr 139 and Tyr 177 . In nNOS these align with Phe 809 and Tyr 889 , which likely perform a similar function (structures for NOS isoforms in this region are not available). The sequence similarity in the vicinity of these aromatic residues is striking. However, in both nNOS and eNOS, inserts of 42 and 45 amino acids, respectively, lie between the two homologous regions. To test the functional role of the nNOS insert, the deletion mutants ⌬40 and ⌬42 were constructed. The ⌬40 mutation was constructed by ligating the ends of nNOS DNA generated at two convenient restriction enzyme sites and removes the bulk of the insert (Pro 831 -Ser 870 ). To adjust the length of the deletion such that the number of amino acids between the FMNbinding regions of nNOS was the same as for iNOS and cytochrome P-450 reductase, a further two residues (Pro 873 and Leu 874 ) were deleted by site-directed mutagenesis, thus forming the ⌬42 mutant (Fig. 1). A third mutant, Y889A, was constructed to generate an FMN-deficient nNOS form. The change of Tyr 889 to Ala, mutates the corresponding residue to that used to generate FMN-deficient cytochrome P-450 reductase (30) by removing a key binding interaction.
All mutants were expressed and purified in the same way as the wild type enzyme. Initially, the ⌬40 and ⌬42 mutants were found to lose activity during purification; however, this was partially recovered on incubation with excess FMN. Inclusion of FMN in the purification buffer and completion of the procedure within 1 day prevented any loss of activity. The ⌬40 mutant was found to be more stable than the ⌬42 mutant and is therefore the subject of more detailed analysis. Unlike nNOS, the native form of iNOS is permanently complexed with CaM. However, despite the similarity between the nNOS deletion mutants and iNOS, addition of excess CaM did not stabilize the mutants, which were readily purified on CaM-Sepharose. The Y889A mutant was found to be inactive and lacking the stable blue semiquinone observed during the purification of the other enzymes, consistent with an absence of FMN in this mutant. This was confirmed by measuring the FMN and FAD content relative to heme using the different fluorescence response of the two flavins in different buffers. The Y889A mutant was found to contain 0.06 (Ϯ 0.10) FMN and 0.92 (Ϯ 0.15) FAD per heme, compared with 0.97 (Ϯ 0.10) FMN and 1.01 (Ϯ 0.10) FAD per heme in the case of the wild type enzyme. Following purification and removal of free flavin, the FMN content of the ⌬40 and ⌬42 mutants was found to be 0.60 (Ϯ 0.10) and 0.67 (Ϯ 0.10), respectively, per heme with 0.96 (Ϯ 0.15) and 1.13 (Ϯ 0.15) FAD, respectively.
Steady-state Kinetics-Rate constants for the steady-state turnover of wild type and mutant forms of nNOS for NO syn-thesis, aerobic oxidation of NADPH, reduction of ferricyanide, and reduction of cytochrome c are collated in Table I. For the wild type enzyme, Ca 2ϩ /CaM is required for NO synthesis or NADPH consumption to reach significant levels. In the presence of L-arginine the NADPH consumption becomes coupled tightly to NO synthesis, based on a stoichiometry of 1.5 NADPH equivalents/NO molecule generated (31). In its absence, however, NADPH is consumed faster, indicating that the supply of electrons to nNOS heme does not limit the rate of catalytic turnover.
For the ⌬40 mutant, both NO synthesis and NADPH consumption occur at significant rates even in the absence of Ca 2ϩ /CaM. In the presence of Ca 2ϩ /CaM, the rate of NO synthesis is approximately 30% of the wild type value, but the rates for NADPH consumption are similar. Unlike the wild type enzyme, the ⌬40 mutant does not efficiently couple NO synthesis to NADPH consumption, and approximately half the electron equivalents are lost to molecular oxygen during normal turnover, i.e. in the presence of L-arginine and Ca 2ϩ /CaM. In the absence of Ca 2ϩ /CaM, ⌬40 catalyzes NO synthesis at approximately 30% of the rate in its presence. The rate of NADPH consumption in the presence of L-arginine again indicates partial decoupling from NO synthesis. However, in this case (i.e. in the absence of Ca 2ϩ /CaM), NADPH oxidation is faster in the presence of L-arginine, suggesting that the rate of electron transfer to the heme is now a rate-determining factor. The ⌬42 mutant behaved in a manner similar to that of the ⌬40 mutant, retaining a similar level of NO synthase activity in the absence of Ca 2ϩ /CaM. Both deletion mutants required CaM and Ca 2ϩ to be added for maximal activity and were unaffected by the addition of Ca 2ϩ alone. The Y889A mutant had no NO synthase activity and minimal NADPH oxidase activity, indicating that electron supply to the heme has been cut by this mutation.
The cytochrome c reductase activity of the wild type enzyme is highly CaM-dependent, increasing by a factor of 10 on CaM binding. The ferricyanide reductase activity is less so, increasing by 3-fold on CaM binding. Both the cytochrome c and ferricyanide reductase activities of the ⌬40 mutant are affected less by CaM binding as the rates observed in the absence of CaM are both higher than for the wild type enzyme. For the ⌬42 mutant, the ferricyanide reductase activity is CaM-independent. The Y889A mutation results in a severe loss of cytochrome c reductase activity, indicating that in the wild type enzyme, FMN is the chief electron donor to cytochrome c. The ferricyanide reductase activity of this mutant is retained, but the effect of CaM is reversed, such that CaM binding causes a  decrease in the rate of reduction. Ferricyanide Dependence-To clarify the effect of the Y889A mutation on the ferricyanide reductase activity of nNOS, the concentration dependence of the wild type and mutant enzymes were studied in the presence and absence of CaM. Fig. 2 plots the steady-state ferricyanide reductase activity of the enzymes against the ferricyanide concentration. For the wild type enzyme two roughly parallel lines are formed, such that the effect of CaM binding is to increase the value of the y axis intercept. This situation suggests the presence of two different binding sites of high and low affinity, probably in the vicinity of FMN and FAD, respectively. In contrast, the plots derived for the Y889A mutant tend toward the origin, suggesting a single, low affinity binding site. Because Y889A lacks FMN, this binding site must be located near the FAD. For the mutant, both lines have a steeper gradient than for the wild type enzyme, indicating that the FAD is more accessible following the mutation. The FAD subdomain is nevertheless catalytically viable.
EGTA Titrations- Fig. 3 plots the normalized NO synthase activities of the wild type enzyme and deletion mutants against the concentration of the Ca 2ϩ chelator, EGTA. In each of the assays the Ca 2ϩ concentration was fixed at 50 M, so that as the EGTA concentration passes this mark it begins to compete with the enzyme-CaM complex for free Ca 2ϩ . At high concentrations of EGTA, no free Ca 2ϩ is available for stimulation of CaM binding, and the enzymes lose activity. For the wild type enzyme, the NO synthase activity rapidly drops to zero after the EGTA concentration reaches 50 M, indicating that the concentration of Ca 2ϩ has dropped below the threshold value necessary to stimulate the enzyme. Both the ⌬40 and ⌬42 mutants retain maximal activity at higher EGTA concentrations and lose activity more slowly as the EGTA concentration increases. The mutants therefore require lower concentrations of free Ca 2ϩ for stimulation. The mutants also retain 30 -40% of their maximal activity even at high EGTA concentrations and turnover at a similar rate as in the absence of CaM (Table I). Fig. 4 plots the normalized ferricyanide reductase activity against the EGTA concentration under similar conditions as those described above. Because the concentrations of Ca 2ϩ , CaM, and enzyme are the same, the plots can be compared directly with those in Fig. 3. The wild type enzyme loses activity rapidly as the concentration of free Ca 2ϩ is depleted, reaching a value consistent with the loss of bound CaM. The ⌬40 mutant retains maximal activity for longer than the wild type enzyme as with the NO synthase activity. The ferricyanide reductase activity of the ⌬42 mutant shows no Ca 2ϩ dependence, which is consistent with its lack of CaM dependence (Table I). For the Y889A mutant, the ferricyanide reductase activity increases as the free Ca 2ϩ is depleted, indicating that the FAD-binding subdomain is itself Ca 2ϩ /CaM-dependent. The resultant curve is similar in shape to that of the wild type enzyme, indicating similar Ca 2ϩ sensitivity.
Heme Reduction- Fig. 5 shows the anaerobic reduction of wild type nNOS on addition of NADPH. The first addition of NADPH (Fig. 5a) causes a broad decrease in absorbance between 400 and 500 nm consistent with the reduction of the two bound flavins of the enzyme. The position of the Soret peak is unchanged at around 398 nm. A further addition of NADPH and incubation for 20 min led to no significant change. Addition of 0.1 mM L-arginine caused a decrease in absorbance at 405-425 nm consistent with a shift in peak position to lower wavelength, which is associated with a low to high spin state change (32). Addition of Ca 2ϩ (50 M) and CaM (10 g/ml) causes a shift in the position of the Soret peak to a longer wavelength, indicating heme reduction (Fig. 5b). Addition of a saturated solution of CO up to 10% causes the formation of the reduced heme-CO complex characterized by the appearance of a peak at 444 nm. Fig. 6 shows the anaerobic reduction of the ⌬40 mutant upon addition of NADPH. As with the wild type enzyme, the first addition of NADPH causes a decrease in absorbance between 400 and 500 nm associated with the reduction of the two enzyme-bound flavins. However, in the case of the mutant, incubation over 20 min results in a shift in the position of the heme Soret peak to 409 nm, consistent with direct heme reduction (33,34)  to 10% saturation resulted in formation of the reduced heme-CO complex which was 80 -90% complete after 40 min of incubation (Fig. 6b).
Addition of NADPH to the Y889A mutant under similar conditions failed to result in heme reduction even in the presence of Ca 2ϩ /CaM and CO. The heme reduced fully, however, on addition of dithionite, producing a peak at 444 nm in the presence of CO (not shown). DISCUSSION All three NOS isoforms consist of a diflavin-binding reductase domain coupled directly to a heme-binding monooxygenase domain via a CaM-binding linker region. CaM is bound by all functional forms of NOS. However, it is bound irreversibly by iNOS, whereas in nNOS and eNOS its binding is Ca 2ϩsensitive. CaM binding enables NADPH dehydrogenation to be coupled to the monooxygenation of L-arginine via interdomain electron transfer (34). At low Ca 2ϩ concentrations, nNOS and eNOS are unable to function and have negligible activity. In neuronal and endothelial cells, Ca 2ϩ -sensitive CaM binding is used to regulate NO synthesis directly and therefore regulates important physiological functions. It may also be involved in a regulation mechanism involving isoforms of caveolin, which are reported to inhibit CaM binding and NO synthesis (35,36). Although the structure of the dimeric iNOS heme domain was recently reported (3), the nature of its interaction with the reductase domain is not known. The structure of the reductase domain itself is probably related to that of cytochrome P-450 reductase (17), which has a similar amino acid sequence. The route of electron transfer through the NOSs has therefore been assumed to be the same as in the cytochrome P-450 system, i.e. NADPH to FAD to FMN to heme. This is confirmed by our observation that the Y889A mutant of nNOS remains an effi-cient NADPH dehydrogenase-ferricyanide reductase (Table I) but is unable to transfer electrons to the nNOS heme. This is almost certainly because the mutant no longer binds FMN. The concentration dependence of the ferricyanide reductase activity of the Y889A mutant (Fig. 2) shows approximately second order behavior, indicating weak ferricyanide binding at the FAD site. In contrast, the plot for wild type in the presence of Ca 2ϩ /CaM indicates that the rate of ferricyanide reduction is largely independent of ferricyanide concentration. This suggests that ferricyanide binds much more tightly to the electron transfer site when FMN is present, resulting in the line intersecting the x axis at around 2200 min Ϫ1 , which would correspond to the rate of electron transfer from FAD to FMN. This value is similar to that reported for pre-steady-state flavin reduction by NADPH using stopped flow spectrophotometry (15). In the absence of Ca 2ϩ /CaM, the intercept value is decreased considerably (as is the pre-steady-state rate constant). If the y intercept does correspond to the rate of ferricyanide reduction at the FMN site, the slopes of the lines indicate the concentration dependence of ferricyanide reduction at the FAD site. However, the gradients of the lines plotted for the wild type enzyme are much less than for the Y889A mutant. This indicates that the mutation and consequent loss of FMN has also disrupted the FAD site, making it more accessible to ferricyanide binding. This is not surprising given the proximity of FAD to FMN reported in the x-ray crystal structure of cytochrome P-450 reductase (17) To find the structural basis for the difference between the CaM sensitivity of the inducible (iNOS) and constitutive (nNOS and eNOS) isoforms a series of chimeric enzymes have been studied. Because the amino acid sequences of the NOS isoforms consist of mostly homologous regions, it is reasonable to expect individual domains and subdomains to be interchangeable. This led Ruan et al. (12) to swap the CaM-binding sites of iNOS and nNOS. Both chimeric enzymes required lower concentrations of Ca 2ϩ than nNOS for stimulation but were not active in the absence of Ca 2ϩ , indicating that the sequence of the CaM-binding site affects the Ca 2ϩ sensitivity but does not determine it entirely. Nishida and Montellano (37) generated chimeras in which the nNOS reductase domain was coupled with the iNOS and eNOS heme domains, including the CaM-binding sites. The iNOS chimera, which was coexpressed with CaM for stability, was found to be partially Ca 2ϩ -dependent, retaining 40% of its activity in the presence of 2.5 mM EGTA. Lee and Stull (38) studied a similar chimera and found that the CaM was dissociable but retained activity in the absence of Ca 2ϩ . They also showed that a chimera consisting of the reductase domain and CaM-binding site of iNOS and the heme domain of nNOS bound CaM irreversibly and retained 60% of its activity in the absence of Ca 2ϩ . These studies all show that the Ca 2ϩ -independent activity of iNOS is brought about by a range of structural features including elements in the reductase and oxygenase domains and in the CaM-binding site.
The 40 -50-amino acid insert found only in the nNOS and eNOS isoforms has been speculatively proposed to be an autoinhibitory domain (18,19), but direct evidence for this was not presented in the previous studies. Therefore, to determine the role of the insert in nNOS, we generated two deletion mutants (⌬40 and ⌬42) lacking the bulk of this region. As expected, both of the mutants behaved similarly, with the more stable one, ⌬40, being studied in most detail. Steady-state analysis indicated that the deletion mutants supported NO synthesis in the absence of CaM at 30% of the rate in its presence. NO synthesis in the absence of CaM has not previously been demonstrated for any intact NOS form, although was thought to occur in a chimeric mutant consisting of the nNOS oxygenase domain and iNOS reductase domain (38). In our case, it is unlikely that CaM generated by the yeast could have co-purified with the enzymes following chromatography on both ADP-Sepharose in the presence of 1 mM EGTA and CaM-Sepharose. This is confirmed by the fact that the mutants were not activated by Ca 2ϩ alone but required addition of CaM to reach maximal activity. Furthermore, the EGTA titrations shown in Fig. 3 indicate that the catalytic effects of CaM binding are lost at high concentrations of EGTA, consistent with CaM dissociation. Addition of NADPH to the ⌬40 mutant under anaerobic conditions (Fig. 6) led to spontaneous heme reduction in the absence of CaM, an effect that was manifested in the bulk of the enzyme (80 -90%). This could not be repeated for the wild type. Therefore, it is unlikely that the Ca 2ϩ /CaM-independent NO synthase activity exhibited by the deletion mutants could be caused by a minority contaminant of strongly bound CaM.
The wild type enzyme and the deletion mutants all catalyze NADPH oxidation at similar rates in the presence of CaM, indicating that electron transfer to the heme is not impeded by deletion of the insert. However, the ⌬40 mutant catalyzes NO synthesis at only 30% of the rate of the wild type enzyme in the presence of CaM. For the wild type enzyme NO synthesis is closely coupled to NADPH oxidation in the presence of L-arginine, but the ⌬40 mutant is only 50% efficient. The heme active site may have been affected directly by the mutation or indirectly via the disruption of precise interactions formed between the reductase and oxygenase domains on CaM binding.
The rate of NADPH oxidation is low for ⌬40 in the absence of CaM and increases on addition of L-arginine. This is opposite to the effect of L-arginine in the presence of CaM. L-Arginine binding is known to shift the spin state of the NOS heme from mixed to high spin, which facilitates heme reduction by electron transfer. This is consistent with heme reduction being rate-determining for the ⌬40 mutant in the absence of CaM. However, in the presence of CaM and L-arginine accumulation of NO at the heme active site is reported to slow the enzyme down (39). For the wild type enzyme, CaM binding is prerequisite for electron transfer to occur to the heme; this is demonstrated by the anaerobic heme reduction experiments (Fig. 5). For the ⌬40 mutant, CaM binding is not essential but increases the rate by at least 60-fold (which is the effect on the NADPH oxidation rate). In wild type nNOS, therefore, the insert appears to ensure that the rate of electron transfer to the heme is zero in the absence of CaM, which would enable NO synthesis to be switched off completely during Ca 2ϩ /CaM-dependent regulation.
CaM binds to the interdomain linker of NOS and is thought to transform this region into an encapsulated helix, causing a change in the interface between the reductase and oxygenase domains. The insert appears to impede this process. It is possible, therefore, that by inhibiting the structural rearrangement the insert might simultaneously inhibit CaM binding, i.e. by destabilizing the active (CaM-bound) configuration.
As well as activating FMN to heme electron transfer, CaM binding increases the cytochrome c and ferricyanide reductase activities of nNOS. This has been shown to occur even in a construct lacking the heme domain (15). It was proposed that cytochrome c reduction occurred primarily at the FMN site and that CaM binding increased the rate of electron transfer from FAD to FMN. This is confirmed by the low cytochrome c reductase activity observed for the Y889A mutant (Table I), which is deficient in FMN. This mutant retains ferricyanide reductase activity, indicating that the FAD domain of the enzyme is still functioning as an effective NADPH dehydrogenase. Mutants consisting of the reductase domains of human and mouse iNOS were recently reported to show CaM-independent reductase activity and appeared to bind CaM irreversibly (16). CaM binding was also found to have less effect on the reductase activity of the deletion mutants than the wild type enzyme (Table I). This suggests that the insert, not present in iNOS and removed from the deletion mutants, also plays a role in regulating the reductase activity of nNOS. It is possible, therefore, that CaM binding induces a repositioning of the FMN with respect to both the FAD and the heme in nNOS to create an optimal electron transfer chain.
During the review of this manuscript, a similar study involving the deletion of the corresponding region of eNOS was reported (40). It was found that the eNOS loop deletion mutant was activated at a lower concentration of Ca 2ϩ than wild type eNOS but did not function as an NO synthase in the absence of Ca 2ϩ . These results are consistent with the proposal of Salerno et al. (19) that the eNOS loop competes directly with CaM binding by interacting with a sequence-specific binding site, based on the observation that peptides derived from the sequence of the eNOS loop could act as potent inhibitors. Peptides based on the nNOS loop were not effective inhibitors, however, suggesting a different mode of action for the loop of this isoform. Our results suggest that the nNOS loop not only affects the Ca 2ϩ sensitivity of the enzyme but also directly inhibits FMN to heme electron transfer in the absence of Ca 2ϩ / CaM, ensuring that the wild type enzyme can be completely deactivated at low Ca 2ϩ concentrations. Both effects help to define the response of nNOS to changes in Ca 2ϩ concentration and therefore have physiological relevance.