An Autoinhibitory Control Element Defines Calcium-regulated Isoforms of Nitric Oxide Synthase*

Nitric oxide synthases (NOSs) are classified functionally, based on whether calmodulin binding is Ca2+-dependent (cNOS) or Ca2+-independent (iNOS). This key dichotomy has not been defined at the molecular level. Here we show that cNOS isoforms contain a unique polypeptide insert in their FMN binding domains which is not shared with iNOS or other related flavoproteins. Previously identified autoinhibitory domains in calmodulin-regulated enzymes raise the possibility that the polypeptide insert is the autoinhibitory domain of cNOSs. Consistent with this possibility, three-dimensional molecular modeling suggested that the insert originates from a site immediately adjacent to the calmodulin binding sequence. Synthetic peptides derived from the 45-amino acid insert of endothelial NOS were found to potently inhibit binding of calmodulin and activation of cNOS isoforms. This inhibition was associated with peptide binding to NOS, rather than free calmodulin, and inhibition could be reversed by increasing calmodulin concentration. In contrast, insert-derived peptides did not interfere with the arginine site of cNOS, as assessed from [3H]N G-nitro-l-arginine binding, nor did they potently effect iNOS activity. Limited proteolysis studies showed that calmodulin’s ability to gate electron flow through cNOSs is associated with displacement of the insert polypeptide; this is the first specific calmodulin-induced change in NOS conformation to be identified. Together, our findings strongly suggest that the insert is an autoinhibitory control element, docking with a site on cNOSs which impedes calmodulin binding and enzymatic activation. The autoinhibitory control element molecularly defines cNOSs and offers a unique target for developing novel NOS activators and inhibitors.

by distinct genes, mammalian NO synthases (NOSs) 1 comprise a family of three calmodulin-dependent biopterohemoflavoproteins that are functionally distinguished by their modes of regulation (4). The two constitutively expressed isoforms of NOS (cNOSs), first identified in neuronal cells (nNOS) and endothelial cells (eNOS), remain dormant until calcium/calmodulin (Ca 2ϩ /CaM) binding is actuated by transient elevations in intracellular Ca 2ϩ . This Ca 2ϩ -dependent mode of regulation provides pulses of NO for moment-to-moment modulation of vascular tone and neurosignaling. In contrast, activity of the immunostimulant-induced isoform of NOS (iNOS) is Ca 2ϩ -independent, providing continuous high output NO generation for host defense. A remarkably high affinity for CaM, even at basally low levels of intracellular calcium, is responsible for the Ca 2ϩ independence of iNOS (5).
Whether a given NOS isoform binds CaM in a Ca 2ϩ -dependent or -independent manner has been assumed to be a property solely of the amino acid sequence specified by a 20 -25-amino acid CaM binding site. However, this restrictive view is challenged by findings that chimeric eNOS and nNOS, which have had their CaM binding sequences replaced with the corresponding sequence from iNOS, still require Ca 2ϩ for full activity (6,7). Because regulation of enzyme systems by Ca 2ϩ /CaM typically involves displacement of an intrinsic autoinhibitory polypeptide (8,9), we hypothesized that the binding of Ca 2ϩ / CaM to cNOSs may similarly trigger activation by displacing a control element. Here we identify a multiple amino acid insertion which serves as a control element unique to cNOSs and which molecularly defines Ca 2ϩ -dependent isoforms of NOS.

EXPERIMENTAL PROCEDURES
Protein Modeling-Molecular modeling of the FMN binding module of nitric oxide synthase isoforms was done using the Insight and Homology programs from Biosym (BIOSYM/Molecular Simulations, San Diego, CA) running on a silicon graphics Indigo2 workstation. After alignment of NOS sequences with homologous FMN binding proteins of known structure (see "Results"), structurally conserved region (SCR) boxes were created corresponding to conserved regions of secondary structure and regions involved directly in FMN binding. These regions were characterized by high positive scores as evaluated by Dayhoff's mutation matrix (10). After assignment of coordinates in the SCR regions, the loop regions between the SCR boxes were modeled by searching the Brookhaven protein data base. The crude model structure was relaxed to a sterically and energetically reasonable state using the Discover program (BIOSYM/Molecular Simulations) for molecular me-chanics and dynamics calculations. This includes splice repair to remove unrealistic structural features at SCR-loop junctions, end repair to assign reasonable structures to C-terminal and N-terminal extensions, and structural optimization to remove steric overlaps and to reduce the structure to a energetic minimum. Energy minimizations were begun using the steepest descent method; this was replaced by conjugate gradient method as convergence was approached.
Purification of NOS Isoforms-Rat neuronal cNOS (nNOS) and bovine endothelial cNOS (eNOS) were purified from Escherichia coli harboring pGroELS and pCW vector expression systems for nNOS and eNOS, as described previously (11,12). iNOS-rich cytosol was prepared from rat aortic smooth muscle cells, which were isolated from Fisher rat thoracic aortae (13) and grown in 75-cm 2 culture flasks at passage 10 -15. Cells were stimulated for 16 h in culture medium containing a combination of lipopolysaccharide (30 g/ml) and rat recombinant interferon-␥ (50 ng/ml), washed twice with 10 ml of ice-cold phosphatebuffered saline, and harvested with a Teflon cell scraper into an additional 10 ml of iced phosphate-buffered saline. Cell suspensions were centrifuged at 800 ϫ g for 10 min, resuspended in 100 l/75-cm 2 culture flask of ice-cold distilled H 2 O containing a mixture of protease inhibitors (pepstatin 10 g/ml, leupeptin 10 g/ml, and phenylmethylsulfonyl fluoride 100 M) and lysed by three cycles of freezing in liquid nitrogen and thawing in a 37°C water bath. Lysates were centrifuged at 100,000 ϫ g for 1 h, and supernatants were stored at Ϫ70°C until use.
NOS Activity Measurement-NOS activity was measured in 96-well microtiter plates at 25°C based on the kinetics of NADPH consumption or the oxidation of Fe 2ϩ -myoglobin to Fe 3ϩ -myoglobin, as described previously (14). For NADPH consumption measurements, incubation mixtures contained 50 mM Tris-HCl (pH 7.60), 100 M CaCl 2 , 10 M tetrahydrobiopterin, 500 M NADPH, 500 M L-arginine, 1 mM DTT, 1 M calmodulin, pH 7.6, and the indicated concentration of peptide in a final volume of 100 l. Reactions were initiated by the addition of 20 pmol of nNOS, 10 pmol of eNOS, or 15 g of rat iNOS-rich cytosol. NADPH consumption was determined from the rate of decrease in A 340 , measured at 15-s intervals for a period of 30 min in a kinetic microplate spectrophotometer (Molecular Devices; Menlo Park, CA). The rate of decline in A 340 measured when NOS was omitted from incubates was subtracted from all values. Samples in which iNOS activity was measured based on Fe 2ϩ -myoglobin oxidation were prepared as above, but additionally contained 15 g of rat iNOS-rich cytosol and a final concentration of 40 M Fe 2ϩ -myoglobin. Preparation of Fe 2ϩ -myoglobin and spectrophotometric measurement of its rate of oxidation to Met-myoglobin by NO, was as described earlier (14). 125 I-Calmodulin Binding Measurement-Calmodulin was labeled with 125 I to a specific activity of 25-150 Ci/g, using the Bolton-Hunter method (15). 125 I-Calmodulin binding assays were performed in triplicate using 96-well microfiltration plates with GFB filter bottoms (Millipore, Bedford, MA). Before use, filters were preincubated for 1-2 min with 100 l of buffer containing 50 mM Tris-HCl, pH 7.6, 1 mM DTT, 100 M CaCl 2 , and 0.5 mg/ml ␤-lactoglobulin; buffer was then removed by vacuum filtration. Binding reactions were comprised of: 50 mM Tris-HCl, pH 7.6, 1 mM DTT, 0.5 mg/ml ␤-lactoglobulin, 100 M CaCl 2 , 10 M tetrahydrobiopterin, 1 nM 125 I-CaM (2500 -5000 cpm) and 1-2 pmol of NOS in a 100-l total volume. Samples were incubated for 15 min at 23°C, and binding was quickly terminated by vacuum filtration. Filters were washed twice with 100 l of ice-cold buffer containing 50 mM Tris-HCl, pH 7.6, and 100 M CaCl 2 and air-dried. Scintillation mixture was added to each well (25 l; OptiPhase SuperMix, Wallac Inc., Gaithersburg, MD), and plates were counted in a Microbeta Plus liquid scintillation counter (Wallac Inc.). Nonspecific binding was determined in samples that additionally contained 10 mM EGTA. In studies of the effect of inhibitory effect of peptide inhibitors on 125 I-CaM binding to NOS, blank binding was determined in the presence of peptide and absence of NOS. Dissociation of 125 I-CaM was similarly monitored in 96-well filtration plates. 125 I-CaM⅐nNOS complexes were first produced by incubation of 2 pmol of nNOS, 1 nM of 125 I-CaM, 100 M CaCl 2 , 10 M tetrahydrobiopterin, 50 mM Tris-HCl, pH 7.6, 1 mM DTT, and 0.5 mg/ml ␤-lactoglobulin for 15 min (23°C). To initate dissociation, a 3000-fold excess of unlabeled CaM (3 M) was added at time 0, with or without simultaneous addition of the indicated concentration of bovine eNOS 607-634 . Calcineurin Activity Measurement-Activity of the calmodulin-dependent phosphatase calcineurin (protein phosphatase 2B) was moni-tored in a 96-well kinetic microplate spectrophotometer at 37°C, based on rate of hydrolysis of the artificial substrate p-nitrophenyl phosphate (17). Incubation mixtures contained 50 mM Tris-HCl, pH 7.6, 100 nM bovine brain calmodulin, 40 mM p-nitrophenyl phosphate, 400 M CaCl 2 , and 0.1% 2-mercaptoethanol in a 100-l total volume. Reactions were initiated by the addition of 20 pmol of calcineurin, and activity was monitored continually at A 405 for 60 min at 15-s intervals. Assay blanks additionally contained 10 mM EGTA, resulting in Ͼ90% inhibition of activity.
Proteolysis of NOS Isoforms-Limited proteolysis was performed on incubates containing 40 pmol of recombinant nNOS purified from stably transfected HEK-293 cells (11) or 50 pmol of recombinant eNOS purified from E. coli (12). Samples were preincubated at room temperature for 15 min in a 100-l volume containing: 50 mM Tris, pH 7.6, 1 mM DTT, 10 M CaM, 100 M CaCl 2 , with or without 10 mM EGTA. Proteolysis was initiated by the addition of 20 microunits of L-1-tosylamido-2-phenylethyl chloromethyl ketone-immobilized trypsin (Sigma) per pmol of NOS. Samples (25 l) were collected after 0, 5, 10, and 20 min, and proteolysis was terminated by boiling with an equal volume of 2 ϫ SDS gel-loading buffer. Peptide products were resolved on an 8 -16% gradient SDS-polyacrylamide gel electrophoresis and visualized by staining with Coomassie Blue. Accurate molecular mass determination of tryptic fragments was performed by mass spectrometry at the Rockefeller University Protein/DNA Biotechnology Center, using matrix-assisted laser desorption and time of flight detection (Perseptive Biosystems Inc., Framingham, MA). For N-terminal sequence analysis, tryptic digests were prepared as above and subject to SDS-polyacrylamide gel electrophoresis, but then electrotransfered to polyvinylidene difluoride membranes. Amino acid sequencing was performed on an Applied Biosystems 477A protein sequencer.
Peptides-Synthetic peptides were obtained from SynPep (Dublin, CA) and other commercial suppliers. Purity was evaluated by high performance liquid chromatography and mass spectroscopy and exceeded 80% in all cases with typical purity Ϸ90%. Predominant impurities differed from the desired products by one amino acid residue, resulting from incompletely coupled synthesis.

RESULTS AND DISCUSSION
Nitric oxide synthases are large multidomain enzymes in which a series of gene fusion events has resulted in the incorporation of modules showing significant homology to smaller ancestral proteins (18,19). NOSs can be coarsely dissected into an N-terminal oxygenase domain and a C-terminal reductase domain, bridged by a canonical binding sequence for CaM (20). Calmodulin binding initiates electron transfer between the reductase and catalytic domains, thereby activating catalysis (21). The oxygenase domains have binding sites for substrate (arginine), heme, and tetrahydrobiopterin, whereas the reductase domain has binding sites for FAD, FMN, and NADPH.
Bredt et al. (22) were first to reveal the homology between the C-terminal half of NOS and NADPH-cytochrome P450 reductase (CPR), noting conserved regions corresponding to FMN, FAD, and NADPH binding domains. The FMN-binding modules of NOS isoforms and CPR are in turn highly homologous to the flavodoxins, which are small FMN-binding proteins that function as electron carriers in bacteria (23). The FAD and NADPH binding domains are closely related to chloroplast ferredoxin-NADP ϩ reductase and other related proteins.
Sequence Alignments-Five flavodoxins have been crystallized and solved by x-ray diffraction (24). Three regions in these flavodoxins are involved in binding the FMN prosthetic group; the first of these is close to the N terminus, and is immediately proceeded by the initial ␤ strand of the structure. Although only one of these FMN binding regions was identified in nNOS by Bredt et al. (22), each of them has a corresponding homolog in NOSs. The first step in alignment of the NOS FMN binding domain with the flavodoxins was the identification of these regions in each NOS isoform. This was followed by the identification of conserved secondary structural elements in NOS, primarily by their homology to the corresponding elements in flavodoxins by mutation matrix criteria (10). Fig. 1 shows alignment of a select set of NOS, CPR, and bacterial flavodoxin sequences, illustrating the conservation of regions involved in FMN binding. It is obvious that a major insertion of Ϸ45 amino acids has occurred in mammalian cNOSs. A corresponding insert is also found in the FMN binding region of the cloned invertebrate cNOSs from Rhodnius prolixus (25) and Drosophila melanogaster (26); these inserts are somewhat larger (54 -63 amino acids), but contain regions of marked homology to mammalian cNOS inserts. Expanding this alignment to include dozens of known flavodoxins and related FMN-containing flavoproteins reveals that only cNOSs exhibit such an insertion. Moreover, the corresponding region of iNOS sequences closely resemble flavodoxin sequences but lack an insert anywhere within the FMN binding domain. Therefore, occurrence of the amino acid insertion correlates with Ca 2ϩ /CaM control. This insertion represents the single most prominent difference between cNOSs and iNOS amino acid sequences, considered over their entire length.
The cNOS FMN module insertions are notably rich in charged residues and have an excess of positive charge. This is especially true of the eNOS isoform, which contains the motif RRKRK. Considerable homology exists between the cNOS insertions, particularly toward their N termini. It is also apparent that some structural reorganization has taken place during evolution, which may allow the two or three positively charged residues (depending on species) in the nNOS equivalent of the RRKRK region to recognize a similar binding site. The pattern of conservation suggests that eNOS and nNOS insertions contain at least two motifs.
Structural Models-The availability of solved x-ray crystal structures for flavodoxins allows us to position the insertion in three dimensions relative to the calmodulin binding site. Homology-based molecular models have been constructed for the FMN binding domains of cNOSs, iNOS, and CPR, which could be relaxed to a sterically and energetically reasonable state. After relaxation of these FMN binding domain models, no steric overlaps were present, and energies were approximately Ϫ300 kcal, comparable or lower than that of reference fla-vodoxin crystal structures.
As shown in Fig. 2 (upper left), the backbone structure of iNOS and CPR are virtually superimposable on the backbone of Desulfovibrio vulgaris flavodoxin, the closest solved structural homolog of the FMN binding modules of the NOS isoforms. 2 The structure, a Rossmann fold motif (28), is a five stranded parallel ␤-sheet with the FMN binding site along one edge. Homology predicts that two aromatic residues in murine iNOS, Phe 587 and Tyr 625 , are in contact with the FMN ring system; Tyr 625 serves as a shielding residue. Fig. 2 (upper right) shows the corresponding backbone structure of eNOS; nNOS is extremely similar but not shown. Most of the eNOS backbone can be superimposed on homologs shown in Fig. 2 (upper left), with the insertion projecting from the upper edge of the sheet opposite the FMN binding site. Structurally, it corresponds to the replacement of a tight 5-10 residue ␣ 3 ␤ loop with an Ϸ50-residue structure about one-third the size of the entire FMN binding module. We are unable to propose a conformation for the insertion because we lack a solved homolog; the structure shown is merely intended to convey relative position and size.
The CaM binding site is immediately adjacent to the Nterminal edge of the FMN binding domain (painted white in Fig. 2, upper left). With CaM bound, the CaM recognition site would predictably be in a helical conformation (29,30); steric constraints suggest that it extends almost directly away from the FMN binding domain. The lower panels of Fig. 2  the CaM-binding helix of lower panels). It is notable that calmodulins (molecular mass Ϸ 17 kDa) are larger than the entire FMN-binding module. Although the insertion is midway through the sequence of the FMN-binding module, in three dimensions the model predicts it to be directly adjacent to the CaM binding site. The model predicts also that CaM binding would be sterically hindered by the insertion, suggesting that the insertion can exist in more than one physiologically relevant conformation.
Two aspects of this model strongly suggest that the insert functions as a control element, 1) the correlation between Ca 2ϩ / CaM control and the presence of the insertion and 2) the proximity of the CaM binding site to the insertion and the probable steric interactions which would ensue. An attractive potential role for the insert is that of an inhibitory polypeptide which is displaced by CaM binding. It differs from inhibitory polypeptides common to other CaM-dependent enzymes, and CaM itself, in its lack of acidic and hydrophobic amino acids; this makes direct binding of the insertion to CaM sites in NOS isoforms unlikely. Nonetheless, CaM could conceivably displace the polypeptide insert from a neighboring site by binding domain overlap or through allosteric effects.
Synthetic Polypeptide Effects on NOS Activity-Functional significance of the putative autoinhibitory insert of cNOSs was evaluated using a series of synthetic polypeptide fragments. Polypeptides corresponding to promising recognition sites such as the RRKRK motif of the eNOS insert were synthesized in lengths ranging from 6 to 33 residues, as shown in Table I. Both eNOS-and nNOS-derived peptide fragments were evaluated; effects on NOS activity are summarized in Table II. At concentrations of 50 -100 M, several polypeptide fragments of the cNOS insertions profoundly inhibited eNOS and nNOS activity. The most effective inhibitory polypeptides were from the eNOS insertion and contained the RRKRK motif (Table II and Fig. 3A). Human nNOS-derived polypeptides weakly inhibited eNOS, but were without effect on nNOS. While all peptides were less potent on iNOS, significant inhibition was obtained with eNOS 601-633 and eNOS 607-634 . Notably, inhibition of iNOS activity by these peptides was rapid and apparently complete within one minute of addition. Because CaM is very tightly bound to iNOS and has a remarkably slow off-rate, with little dissociation occurring even after boiling (5), inhibition probably occurs without CaM displacement.
Synthetic Polypeptide Effects on Calmodulin Binding-Overlap of the cNOS polypeptide insert and the CaM recognition site, suggested from molecular modeling, implies that the insert may obstruct CaM binding. If this involves "docking" of the insert within cNOSs, synthetic homologs of the insert might similarly bind and interfere with CaM binding. As shown in Table II, potent inhibition of 125 I-CaM binding to nNOS was observed with insert-derived polypeptide fragments; relative peptide potency for inhibiting CaM binding mirrored that for blocking nNOS activation. IC 50 values for eNOS-derived peptide fragments ranged from 1 to 10 M, and potency increased as the RRKRK motif was progressively lengthened to include up to 33 amino acids (Fig. 3B). Inhibition of nNOS activity and CaM binding by insert-derived peptides was fully reversed by excess CaM (see Fig. 3, C and D, for findings with eNOS 607-634 ), indicative of a competitive mode of inhibition. Thus, the greater apparent potency of peptides for inhibiting CaM binding versus activity, indicated in Table II, is explained by differences in assay conditions; lower CaM concentrations were used to assess binding (1 nM) versus activity (100 nM). Inhibition of CaM binding by peptide could not be overcome by excess Ca 2ϩ (Fig. 3E).
Conceivably, the synthetic peptides could interfere with CaM binding to NOS by interacting with either NOS or CaM itself. That NOS is the actual binding target for eNOS-derived insert peptides is indicated by several findings. First, direct binding of peptide to 125 I-CaM, quantified in the absence of NOS, was undetectable at concentrations that inhibited Ͼ90% of CaM binding to nNOS (data not shown). Second, the CaM-dependent phosphatase calcineurin, which resembles cNOSs in having a K d for CaM of 5 nM (31), was not inhibited by concentrations of insert-derived peptides that potently inhibit nNOS activity (see Fig. 4). Third, eNOS-derived peptides markedly enhanced the dissociation rate of 125 I-CaM from preformed complexes with nNOS (Fig. 3D). In this experimental setting, dissociated 125 I-CaM is prevented from reassociating with NOS by addition of a 3,000-fold molar excess of unlabeled CaM. Thus, in order for a synthetic eNOS-derived peptide to eject CaM from its binding site on nNOS, it must at least transiently form a ternary CaM-containing complex with NOS. Conceivably, this transient ternary complex could involve interactions of peptide with CaM as well as NOS. These findings suggest that the binding domain of the putative eNOS autoinhibitory element on nNOS either overlaps or allosterically perturbs the CaM binding domain.
Previously described inhibitors with demonstrated selectivity for NOS influence the arginine site in a manner that can be detected as a loss in sites or binding affinity for the arginine analog, [ 3 H]N G -nitro-L-arginine. Thus, it is notable that the cNOS insert peptides inhibit NOS activity and CaM binding with a slight increase, rather than decrease, in [ 3 H]N G -nitro-L-arginine binding (Table II). Specificity of the insert peptides is also indicated by a lack of inhibition of either NOS activity or CaM binding with each of five synthetic peptides, 10 -15 amino acids in length, derived from sites on the FMN binding domain of cNOSs, which are distinct from the insert polypeptide (not shown).
Effects of Calmodulin on Exposure of the Insert-Displacement of the insert peptide from an internalized binding site on cNOSs by CaM would conceivably enhance exposure of the insert to proteolysis. This hypothesis was tested by examining the pattern of peptide accumulation during limited trypsinolysis of both nNOS and eNOS, in the absence and presence of bound CaM (Fig. 5).
Earlier, Sheta et al. (20) showed that of 165 possible tryptic cleavage sites in rat nNOS, a single preferred cut site resides at Arg 727 within the CaM binding sequence. Cutting at this site has served as an effective means for isolation of distinct reductase and oxygenase domains. With CaM present but not bound (due to addition of the Ca 2ϩ -chelator, EDTA; see Fig. 5A), we similarly observe that tryptic cleavage of nNOS occurs almost exclusively at a single site, consistent with Arg 727 within the CaM binding site. Accordingly, we found a time-dependent accumulation of fragments with apparent molecular masses of 77 and 85 kDa, corresponding to C-terminal reductase and N-terminal oxygenase domains, respectively (Fig. 5A). When CaM was permitted to bind nNOS, by omission of EDTA, Arg 727 was protected from proteolysis, and a novel tryptic cleavage site was revealed. Cutting at this new site yielded fragments of apparent molecular masses of 63 and 93 kDa (Fig.  5A). Molecular mass refinement by matrix-assisted laser desorption ionization spectrometry indicated the smaller fragment to be 64,809 Ϯ 324 Da. This product is best explained by cleavage at Arg 855 -Lys 856 , a dibasic (RK) site within the insert peptide which predicts a C-terminal fragment of 65,071 Da. That this fragment originates from the C terminus of nNOS is indicated by our finding that it is the predominant trypsinolysis product of the bacterial-expressed C-terminal reductase domain (nNOS 721-1429 ) but is not produced by trypsinolysis of the N-terminal oxygenase domain (nNOS 1-721 ) (data not shown). Confirmation of cleavage at Lys 856 is provided by direct sequence analysis of its 10 N-terminal amino acids (KSS-GDGPDLR). In accord with our findings, a thorough analysis of nNOS trypsinolysis, in the absence of bound CaM, indicates that Lys 856 becomes a cut site following initial cleavage within the CaM binding site at Arg 727 (32). Since Lys 856 is protected from tryptic cleavage in the absence of CaM, but exposed when CaM is bound (or the CaM binding site is severed), we conclude that CaM displaces the FMN domain insert peptide of nNOS.
A similar conclusion is drawn from study of eNOS fragmentation after limited trypsinolysis. When CaM is not bound, tryptic cleavage of eNOS yields four principal peptides of nominal molecular masses of 57, 60, 68, and 77 kDa (Fig. 5B). This pattern is rationalized by cleavage at Arg 518 within the CaM  binding site (eNOS 1-518 ϭ 56,877 Da, eNOS 519 -1204 ϭ 76,308 Da) and at a second site, likely to be Lys 545 , which resides between the CaM binding site and insert peptide (eNOS 1-545 ϭ 59,916; eNOS 546 -1204 ϭ 73,270). Exposure of Lys 545 and proximity to the CaM binding site is predicted in the model shown in Fig. 2; this site appears to be within a helix-turn transition at the edge of the ␤-sheet distant from the FMN binding site. Lack of cleavage in nNOS at the site homologous to eNOS Lys 545 may be explained by the presence of a single basic residue, while eNOS contains paired basic residues (RK). In  Fig. 1), suggesting that CaM binding similarly displaces the insert peptide in each cNOS isoform.
To summarize, CaM binding not only protects the CaM binding site from degradation by trypsin, but exposes cleavage sites on both nNOS and eNOS, which are otherwise inaccessible. A preponderance of evidence points to the clusters of basic residues in the FMN domain insert as the trypsin cleavage sites which are exposed by CaM binding. Exposure of cryptic sites by CaM binding could occur by an allosteric mechanism, or by displacement through binding domain overlap. CaM-driven movement of the insert strongly suggests a switch function for activation of NO synthesis.
Mechanism of NOS Control-Herein we have shown that cNOSs possess a polypeptide insert in their FMN binding modules that is 1) unique to NOS isoforms which are regulated by transient CaM binding; 2) positioned adjacent to the CaM binding domain; 3) an impediment to CaM binding and hence, NOS activation; and 4) displaced when CaM binding does occur. Together, these results strongly imply that the insertion in cNOSs is an autoinhibitory control element. We propose that inhibition of NOS by the insert requires occupancy of key sites on cNOS. CaM binding displaces the insert, thus activating cNOS catalysis by "disinhibition." Close proximity of the inhibitory polypeptide to its cognate binding site(s) on cNOSs would result in an exceedingly high local concentration, thus favoring the bound/inhibited state in the absence of CaM. The detection of basal activity with either purified eNOS or nNOS, in the simultaneous presence of EGTA and absence of CaM (Ϸ5% of maximal), 3,4 may arise from a low steady-state concentration of the disinhibited cNOS conformer.
The control mechanism requires that CaM displace the insert upon binding to cNOS; this should translate into a reduced affinity for CaM. Reciprocally, absence of the insert from iNOS would preclude the otherwise expected steric hindrance to CaM binding, contributing to the much tighter binding of CaM at low levels of Ca 2ϩ . Studies of polypeptides, corresponding to the putative CaM binding sites on eNOS and iNOS, and of chimeras in which the putative CaM binding sequence of one NOS isoform is substituted with the corresponding portion of another, have indicated that affinity and calcium-dependence of CaM binding is provided by elements on NOS in addition to the recognized CaM binding sequence itself (6, 7). These results have been interpreted as indicating the presence of an auxiliary CaM binding region on iNOS that augments binding. An alternative explanation, raised by our findings, is that the absence of the autoinhibitory polypeptide from iNOS contributes to enhanced CaM affinity at low Ca 2ϩ levels. We hypothesize that iNOS evolved from an ancestral cNOS-like protein by loss of the inhibitory peptide; nonetheless, vestigial regulatory sites are suggested by a weak inhibition of activity in the presence of synthetic fragments of the eNOS inhibitory peptide. The CaM binding sites on iNOS and cNOS are apparently related to a similar basic region near the N terminus of CPR, and may have evolved from such a region in a common ancestral protein.
Our data suggest that binding of the inhibitory peptide may involve at least two regions. At least one recognition site binds the RRKRK motif. A second possible site might recognize sequences such as EERKSYKVRF and EQHKSYKIRF that occur in the N-terminal half of the eNOS and nNOS insertions; peptides that lack RRKRK but contain these sequences weakly inhibit NOS activity and CaM binding. Some similarity between the first and second halves of the insertion can be readily noted by comparing the sequences of peptide eNOS 628 -633 and eNOS 626 -636 with those of nNOS 835-845 and eNOS 604 -615 . The insert peptide also contains an abundance of serine and threonine residues which provide potential sites for phosphorylation (12/45 residues in the bovine eNOS insert). We speculate that phosphorylation/dephosphorylation may influence the affinity of insert peptides for binding cognate sites on cNOSs and hence, impact on parameters of NOS activation and/or deactivation. In this regard, it is notable that skeletal muscle possesses an nNOS splice variant in which the insert peptide is expanded by 36 residues (33), providing additional sites for possible cell-type specific modification.
Many important questions remain to be answered. The location and identity of the sites of interaction with the inhibitory polypeptide on the surface of the enzyme are not known. Regions of interaction could include the flanking surface loops of 3 4. CaM-stimulated nNOS activity, but not calcineurin activity, is inhibited by the FMN insert-derived peptide eNOS 607-633 . Kinetic assays were performed as described under "Experimental Procedures"; bars represent means of triplicate determinations Ϯ S.E., performed in the absence and presence of the indicated peptide concentrations. Reaction blanks were performed in the presence of 10 mM EGTA to define CaM-dependent activity of nNOS and calcineurin. In both cases, EGTA reduced activity to a level Ͼ10% of that measured in the absence of calcium chelator. the FMN domain, bound CaM, and additional more distant sites. In particular, a site on the oxygenase domain consisting of an array of acidic groups could serve as a binding site for the basic regions on the inhibitory polypeptide insert which stabilizes the inhibited conformation of cNOS.
While it seems clear that CaM binding and activation of cNOS is associated with displacement of the inhibitory polypeptide, it is not known how the presence of the polypeptide in its initial conformation inhibits electron transfer. An obvious mechanism would involve interference by the inhibitory polypeptide with interactions between the oxygenase and reductase domains or flavin subdomains, stabilizing a conformation which does not support rapid electron transfer. Such interference could involve changes in either heme/FMN or FMN/FAD distances driven by domain realignment. Intramolecular electron transfer rates are often determined by the ability of electrons to tunnel, and therefore fall off exponentially with distance at roughly an order of magnitude per bond length (34); thus, a small increase in interdomain distance could produce a large reduction in electron flux. Displacement of the inhibitory peptide may not be the only mechanism by which CaM binding stabilizes the activate conformation of cNOSs, inasmuch as CaM removal results in inactivation of iNOS.
In conclusion, we have identified a novel control element in cNOSs which will serve as a prototype for the development of potent peptide inhibitors. The exposure of tryptic cleavage sites in this element represents the first demonstration of a specific CaM-induced conformational change in NOS and may be a hallmark of the active conformer of NOS. Abu-Soud and Stuehr (21) pointed out that the use of CaM to control electron transfer is unique to cNOS. A more fundamental difference between cNOS and other CaM-regulated proteins is the lack of a CaM analog within the cNOS inhibitory peptide. Ultimately, cNOS may not be unique in this regard; it may presage the identification of other CaM regulated systems in which the CaM/ inhibitor interaction is mediated through binding domain overlap or allosteric effects, rather than competition for a common recognition site.
Acknowledgments-The encouragement and support of Joe DeAngelo and APEX Biosciences Inc. are responsible for the pilot studies that brought this investigation to fruition. , and 17 kDa correspond to uncut nNOS, eNOS, and CaM, respectively. Results indicate that a preferred tryptic cleavage site resides within the CaM binding sequence of cNOSs, however, when this site is masked by bound CaM, a novel tryptic cleavage is revealed within the FMN domain polypeptide insert. Panel A shows the progressive formation from nNOS of 75-and 85-kDa fragments in the absence of bound CaM (lanes [2][3][4][5], and the accumulation of two alternative fragments of 63 and 93 kDa when CaM is bound (lanes 6 -9). Panel B shows the corresponding experiment with eNOS giving a more complicated tryptic cleavage pattern in the absence of bound CaM (lane 6 -9) that is simplified to two major fragments of 60 and 65 kDa when CaM is bound (lanes [2][3][4][5]. See text for detailed interpretation of cleavage sites.