The function of the small insertion in the hinge subdomain in the control of constitutive mammalian nitric-oxide synthases.

Control of nitric oxide (NO) synthesis in the constitutive nitric-oxide synthases (NOS) by calcium/calmodulin is exerted through the regulation of electron transfer from NADPH through the reductase domains. This process has been shown previously to involve the calmodulin binding site, the autoinhibitory insertion in the FMN binding domain, and the C-terminal tail. Smaller sequence elements also appear to correlate with control. Although some of these elements appear well positioned to function in control, they are poorly conserved; their role in control is neither well established nor defined by available information. In this study mutations have been induced in the small insertion of the hinge subdomain, which has been shown recently to form a beta hairpin in structural studies of the neuronal NOS reductase domains adjacent to the calmodulin site and the autoinhibitory element. Modification of the small insertion in neuronal NOS tends to increase cytochrome c reduction but not NO synthetic activity; some modifications or deletions in the corresponding region in endothelial NOS modestly increase activity under some conditions. Unexpectedly, some minor changes in the sequence introduce a loss in the content of heme relative to flavin cofactors. Taken together, these results suggest that the small insertion protects the calmodulin binding site and that it may be a modulator of NOS activity.

Nitric-oxide synthases (NOS) 1 are a growing family of modular enzymes, including three mammalian isoforms, as well as a number of related eukaryotic and more distantly related prokaryotic enzymes. Because of the discovery of the function of nitric oxide (NO) as a molecular messenger in a variety of signal transduction pathways (1)(2)(3)(4), the endothelial and neu-ronal isoforms (eNOS and nNOS) have been shown to function as signal generators controlled by calcium/calmodulin (Ca 2ϩ / CaM) (5). The other mammalian isoform, iNOS, is induced during immune response (6 -9); it is constitutively active and binds CaM at basal Ca 2ϩ levels.
Ca 2ϩ /CaM control in NOS is exerted through the regulation of electron transfer between NADPH and heme (5). This regulation is primarily a function of the reductase domains (21). The canonical CaM binding site was recognized in the sequence region connecting the oxygenase and reductase regions (11). A major insertion in the FMN binding region, correlating with Ca 2ϩ /CaM control, was identified as a control element with autoinhibitory character (autoinhibitory insertion, or AI) (22,23). This has been confirmed by a number of groups (e.g. 24,25).
Recently the C-terminal region has been shown to be necessary for control (26); C-terminal truncated enzymes are uncoupled in the absence of CaM, and the region contains a regulatory phosphorylation site (27). Mutations at this site (Ser-1179 in eNOS) (28) and the nearby FAD stacking residue (Phe-1164) (29) have an altered calcium response. NADPH has been shown to be involved in producing a "conformational lock" in rapid kinetics experiments (30); displacement of NADPH by 2Ј-AMP and other putative analogs activates eNOS and nNOS. 2 Two additional regions contain sequence elements that correlate with Ca 2ϩ /CaM control. The strap connecting the FAD and FMN binding domains is slightly longer in cNOS, which may provide additional conformational flexibility. Of greater interest is the small insertion (SI) consisting of four to seven residues in the hinge subdomain contained within the FAD binding region. This sequence element has recently been shown to form a ␤ hairpin adjacent to the AI and to the CaM binding site in the first structure of nNOS reductase domains (31). There is, however, no homology at the amino acid level between the SIs of eNOS and nNOS; homology was destroyed by coupled frameshift mutations, which were tolerated despite completely changing the character of the sequence. Most recently, deletion of the SI in eNOS by substitution of the corresponding iNOS region produced an enzyme activated at 5-fold lower Ca 2ϩ , leading to a proposal that the SI functions as an additional AI element (32). The ϪSI enzyme is not more active than the wild * This work was supported by the American Diabetes Association and the American Heart Association. 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 U.S.C. Section 1734 solely to indicate this fact.
The NADPH P450 reductase nucleotide sequence reported in this paper has been submitted to the Swiss Protein Database under Swiss-Prot accession no. P16435 (37).
type enzyme under all conditions, however, and does not follow the pattern observed on AI deletion/substitution. These conflicting indications concerning the importance of the region are addressed here by a series of mutagenesis experiments.

EXPERIMENTAL PROCEDURES
Materials-All chemicals used for purification were obtained from Sigma. The genes for eNOS and nNOS were gifts from Professor B. S. S. Masters (University of Texas Health Science Center at San Antonio). The iNOS expression system was the gift of Professor Dipak Ghosh (Duke University and VA Medical Center). GroELS plasmid was provided by Dr. Anthony Gatenby (PerkinElmer Life Sciences).
Expression and Purification of Wild Type and Mutant Bovine eNOS and Rat nNOS-Expression and purification of bovine eNOS and rat nNOS were performed using procedures similar to those described previously (33). Transformed cells were broken with a French press, and after centrifugation to remove cell debris the supernatant was loaded on a 2Ј,5Ј-ADP affinity column, washed, and eluted with 2Ј-AMP as in the previous paper (33). High purity preparations can be obtained with a size exclusion step; we used a Superose 6 HR 10/30 column (Amersham Biosciences); flow rate, 0.4 ml/min; buffer composition, 50 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 1 mM ␤-mercaptoethanol, 100 mM NaCl, 10% glycerol (v/v). During the purification procedure tetrahydrobiopterin or L-arginine was added after elution from the affinity column. Enzyme sample concentrations were determined on the basis of heme concentration, except in heme-free preparations, where the flavin concentration was measured. UV-visible absorbance spectra were recorded on an Aminco DW-2000 spectrophotometer.
NO production was assayed using the Griess assay as adapted for microtiter plates (34); calcium dependence was measured using EDTA as a Ca 2ϩ buffer system. NO production by NOS isoforms was routinely assayed with different buffering systems because eNOS is much more active in MOPS, whereas iNOS and nNOS work well in Tris. NADPHdependent cytochrome c reduction was measured spectrophotometrically as described by McMillan et al. (17), adapted to 96-well plate microtiter plates. In each well, the 500-l reaction mixture contained 50 mM Na ϩ /Tris buffer, pH 7.5, 50 M EDTA, 50 M NADPH, 50 M horse heart cytochrome c, and ϳ10 nM nNOS. Cytochrome c reduction was monitored at 550 nm (⑀ ϭ 2.1 ϫ 10 4 /M). In Ca 2ϩ /CaM dependence studies, 0.75 M CaCl 2 and 10 pg/ml CaM were added to reaction mixtures. Assays were performed on a SpectraMAX plate reader (Molecular Devices).
Protein concentrations were estimated by several methods. The Bio-Rad DC protein kit, an adaptation of the method of Lowry et al. (35), provided the primary measurements; a correction for tyrosine and tryptophan content was applied. Holoenzyme and reductase domain extinction coefficients at 280 nm, given under "Results," were similar to those calculated for unfolded proteins from the sequences at the Expasy web site using the ProtParam tool.
Generation of NOS Mutants-NOS genes in pCWoriϩ were mutated using a method we devised employing the Stratagene QuikChange mutagenesis kit. As suggested by Wang and Malcolm (36), we began by separating the forward and reverse primers, but instead of a single preliminary step with separate primers followed by reversion to the Stratagene protocol, we employ 25 cycles of linear amplification with no additional steps other than to combine and anneal the samples. The removal of the "PCR" steps, which are unproductive and which actually destroy mutant strands by extension, removes the limitations on the separate linear amplification steps and greatly improves the performance of the procedure. In particular, we have obtained a consistent high yield of mutants and a very low background (Ͻ5%) of parentals. A more detailed evaluation of the method will be published elsewhere.
After mutagenesis, the products were used to transform XL10-Gold Ultracompetent Escherichia coli cells from Stratagene. Transformed cells were selected by growth on ampicillin medium; clonal colonies were cultured, and plasmid DNA was obtained using Qiagen minipreparation kits. Mutants were sequenced at the State University of New York at Albany Center for Functional Genomics sequencing facility.

RESULTS
An alignment of the SI region in eNOS, nNOS, iNOS, and NADPH P450 oxidoreductase is shown in Fig. 1. These sequences are part of the hinge subdomain within the FAD binding domain. Alignments of two dozen eukaryotic NOS sequences (not shown for reasons of space) indicate that corresponding SIs are a feature of other NOS sequences that contain AI cognates in the FMN binding domain; iNOSs lack this feature. The SI regions of constitutive NOS isoforms of primitive vertebrates and invertebrates resemble nNOS more closely than eNOS. The SI in mammalian nNOS is characterized by its hydrophobic character and by a glycine residue that marks a turning point in the path of the backbone. It can be argued that a centrally located glycine (or glycines) allowing a sharp turn is the only common SI sequence element. To examine the significance of this residue we constructed several mutants. We reasoned that the standard alanine mutant would not be particularly informative in this case because Gly and Ala both have small side chains associated with sharp turns. Tyr and Asp were selected as residues that would introduce strongly different character into the region, Tyr by virtue of its large side chain volume and Asp by introducing a negative charge into a hydrophobic region. We preferred the relatively short side chain of acidic residues to the very long positively charged side chains of Arg and Lys, both of which are capable of interacting over a considerable distance.
Mutants of nNOS-The structure of the nNOS region can be seen in Fig. 2, in which the structure of cytochrome P450 oxidoreductase (38) is overlaid with the structure of a twodomain nNOS construct, including the FAD and NADPH binding domains and the hinge subdomain (31). The SI is shown in white; it is adjacent to the edge of the FMN binding domain opposite the FMN binding site and is directly adjacent to bound CaM and the AI. In nNOS the sequence immediately preceding the SI contains the triplet LEE, corresponding to LDE in iNOS and NADPH P450 oxidoreductase and LEK in eNOS. The second acidic residue in the figure, corresponding to Glu-352 in NADPH P450 oxidoreductase and Glu-1068 in rat nNOS, is marked in deep red in both cases. Immediately after the SI is a triplet NKK in NADPH P450 reductase and NWK in nNOS. The serine residues, Ser-354 and Ser-1077 in NADPH P450 oxidoreductase and nNOS respectively, are marked in Fig. 2 in yellow. This makes the relative positions of cognate structures (e.g. the AI and CaM binding site) surrounding the SI region in the two proteins clear.
In designing deletion mutants we were influenced by the pattern observed in large multiple sequence alignments of NOS isoforms, by the structures of P450 reductase and nNOS reductase domains shown in Fig. 2, and by structural models of the NOS reductase domains we have constructed over the past 10 years using information from these and other solved homologous proteins (e.g. 22). In P450 oxidoreductase the motif LDEES (see Fig. 1) forms the turn at the end of a series of ␤ pair structures; this is probably approximated in iNOS by the cognate sequence LDESGS. The insertion of six or seven residues relative to iNOS or P450 reductase causes the corresponding sequence element to assume a different position; it now forms one side of a terminal ␤ hairpin structure. Removal of the entire insertion (corresponding to within one residue to TAL-GVIS in nNOS and PGGPPP in eNOS) should cause a reversion to the P450 reductase-like structure if this is consistent with the structural context. Partial deletions should cause shortening of the ␤ hairpin with potential disruption of the local geometry because the shorter sequence cannot make the same turn as the wild type parent. Note the large difference in structure between the flanking homologous regions in the two structures overlaid in Fig. 2; this might have less to do with the insertion than with the differences in structurally adjacent regions in the two enzymes shown and with the absence of the FMN binding domain in the nNOS structure. Fig. 3A shows optical spectra of preparations (partially purified by 2Ј,5Ј-ADP affinity chromatography) of the G1074Y and G1074D rat nNOS mutants as well as wild type enzyme. G1074Y has an abnormal absorbance spectrum reflecting a heme:flavin ratio of 1:4 rather than the 1:2 ratio of wild type enzyme, whereas the spectrum of G1074D closely resembles that of wild type nNOS. This is most obvious from the relative size of the heme Soret band near 410 nm, indicative of a mixture of high spin and low spin ferriheme, and the bands near 480 nm from FAD and FMN. We have confirmed the loss of heme by decomposition of the spectra into heme and flavin components. This unusual and somewhat counterintuitive result implies that a mutation in the reductase domains reduced the heme content while leaving the flavin content relatively unchanged. CO difference spectra (not shown) indicate that the ferrous CO complex of the remaining heme has its Soret maximum at 447 nm, indicating that the heme still has its native axial thiolate ligand; little conversion to the denatured P420Ј form has occurred. This indicates that the remaining heme is located in correctly folded oxygenase domains.
Further purification by gel filtration of nNOS constructs which express significantly as holoenzyme resulted in preparations with UV-visible spectra very similar to wild type nNOS; yields were comparable for G1074D and low (about one-quarter normal) for G1074Y. Protein concentrations were measured by several methods. The mM Ϫ1 extinction coefficients at 280 nm were 175 for nNOS holoenzyme and 70 for the reductase domain from assays for tetrahydrobiopterin-free preparations or for spectra from which the biopterin contribution had been subtracted, which agreed surprisingly well with calculations based on the content of aromatics. As shown in Table I, mutants that express significantly as holoenzyme have close to one heme/monomer. G1074Y has slightly lower heme content even in preparations that have been further purified by gel filtration. Because all of the nNOS mutants have nearly a full complement of flavin cofactors, very few oxygenase sites are associated with flavin-deficient reductase units, and the measured activities do not reflect trivial effects based on flavin content.
The effects on enzyme activity and Ca 2ϩ /CaM control of G1074Y and G1074D nNOS mutations were similar. Although the SI appears well placed to interact with the CaM binding region, the calcium dependence of activation is unchanged. However, the NO synthase activity of both mutants, based on heme content as measured at 447 nm in the CO difference spectrum, is only about two-thirds that of the wild type enzyme on a per heme basis. The activities of the mutants are summarized in Table I.
Deletion of half of the SI (TAL corresponding to 1071-1073 in nNOS) produced an effect on the spectral properties of the expressed protein analogous to the substitution of Tyr for Gly; in this case the heme:flavin ratio is reduced only to about 1:2.4 (Table I) (spectrum not shown). The calcium dependence was unaltered. Unlike the Tyr mutant, this mutant had wild type activity on a per heme basis. Deletion of the entire SI produced a spectrally similar nNOS mutant with low activity on a per heme basis.
Cytochrome c reduction by NOS isoforms has been assumed to occur primarily through FMN; recently conducted experiments with an nNOS shielding residue mutant lacking FMN support this view. 2 Cytochrome c reduction is therefore a measure of electron transfer through the reductase domains, which does not require the oxygenase component. It is activated by CaM binding but does not strictly correlate with NO production. The cytochrome c reduction experiments presented here correspond to the low salt experiments of Knudsen et al. (32). eNOS wild type and mutant activity depends on salt concentration and type and on the buffer used in assays; we hope to present additional experiments at a later date to clarify this area.
The effects of SI mutations on cytochrome c reduction rates in nNOS are shown in Fig. 4. The G1074Y mutant has a slightly reduced rate of cytochrome c reduction consistent with its lower rate of NO production. This is also true of the halfdeletion mutant. This suggests that the low rate of NO production is the result of slow electron delivery. The enhancement of cytochrome c reduction by CaM is comparable (about 5-fold) in wild type nNOS and all mutants except the full SI deletion, which has very low cytochrome c reduction activity. This is in sharp contrast to AI mutants, which tend to lose CaM sensitivity (e.g. 23). The G1074D mutant and the 1171-1173 deletion mutants are significantly more active than the wild type enzyme in cytochrome c reduction.
Mutants of eNOS-In contrast to nNOS, the SI in eNOS consists entirely of proline and glycine residues unless the flanking residue serine 838 is considered. The eNOS SI was less tolerant of mutations than the nNOS SI, probably because of rigidity imposed by the prolines. In general, 20% of wild type expression levels would produce more than enough enzyme for the purposes of this paper, but most of the mutants produced only about 10% heme-containing protein as judged by the small peak in the Soret region; we could detect no CO difference spectra in these preparations at 450 nm (cysteinyl ligand intact) or 420 nm (denatured form).
The Gly-841 deletion mutant, however, was expressed at only slightly lower levels than wild type eNOS (50 -60%). UVvisible absorbance spectra of preparations partially purified by 2Ј,5Ј affinity chromatography suggested a slightly depressed heme:flavin ratio but were otherwise normal. The Ca 2ϩ /CaM dependence of this mutant is unaltered compared with wild type eNOS. The activity/heme is slightly lower in the fully activated state (85% of wild type), which is within the variability of the activity of wild type preparations.
Neither the G841Y nor G841E mutant of bovine eNOS was significantly expressed as a heme protein, in contrast to their nNOS cognates. The lower heme:flavin ratio observed in G1074Y rat nNOS was reflected in an exaggerated form in the eNOS mutants, which could be partially purified as flavoproteins on a 2Ј,5Ј-ADP column. The optical spectra of these preparations are compared with wild type eNOS in Fig. 3B. The reductase domains of these mutants are obviously inactive in NO production, but they exhibit uncoupled NADPH oxidation (based on NADPH oxidation by O 2 during Griess assays; data

FIG. 3. UV-visible absorbance spectra of wild type NOS enzymes and mutants, showing contributions from the heme and flavin cofactors.
A, spectra isolated from wild type nNOS and the G1074Y and G1074D mutants; compare the heme Soret band near 410 nm with the FAD and FMN features in the 450 -500 nm region. B, spectra isolated from wild type eNOS and the corresponding G841Y and G841D mutants; both mutants are almost completely heme-free.

TABLE I Some properties of nNOS mutants expressed in ER2566 cells and partially purified using 2Ј,5Ј-ADP affinity chromatography
Nitrite production was measured with a Griess assay on the basis of heme content; NOS concentration in mg/ml was obtained by multiplying the spectroscopically determined heme content by the molecular mass of full-length nNOS. Heme contents were determined to an error of 0.05 assuming 100% purity of holoenzyme preparations. (a), affinity column only; (b), additional sizing column; (c), estimated from 280 and Soret bands. not shown) at two to three times the rate of wild type eNOS.
When further purified by gel filtration, the spectra of the Gly-841 deletion mutant closely resembled wild type enzyme. We calculated a minimum heme:monomer ratio of 0.5-0.6 assuming 100% purity; the extinction coefficients of ϳ150 and 75 for eNOS holoenzyme and reductase domain agree well with expected values based on aromatic contents, subject to interference by tetrahydrobiopterin. The flavin content of affinitypurified preparations was about 1.7 flavins/heme. Mutants that did not express significantly as holoenzyme could be partially purified by affinity chromatography followed by gel exclusion. Total flavin:monomer ratio ranged from 0.5 to 1 assuming 100% purity. The activities and spectra of these constructs indicate that most of the flavin is present in molecules with one FMN and one FAD.
As mentioned previously, Knudsen et al. (32) reported the effects of SI deletion in eNOS; calcium dependence was affected, but neither the cytochrome c nor the NO production activities were significantly affected, apart from a modest change in salt dependence. The eNOS mutants produced here (except for the Gly-841 deletion, which is similar to wild type) are CaM-insensitive in all assays, but this may result from exposure of the CaM binding site to proteolysis rather than loss of control because of the removal or mutation of the SI. The cytochrome c reduction data for selected eNOS mutants is summarized in Fig. 5. The pattern of activity is otherwise similar to that observed in nNOS mutants; the tyrosine mutant has low activity, whereas the G841D mutant and the Gly-841 deletion have slightly higher than wild type rates of cytochrome c reduction. It is unlikely that G841Y has a different cofactor distribution than the other mutants because it is spectrally similar, and the flavins are reducible with NADPH. Since the original submission of the manuscript, we have produced an eNOS SI deletion mutant that differs slightly from the Knudsen et al. (32) construct in that the flanking regions of our deletion are not iNOS-derived. We have thus far obtained only the reductase domains, which are highly controlled by CaM and, unlike the Knudsen et al. (32) construct, hyperactive in the reduction of cytochrome c. Clearly, small details can have large effects.
The G841D mutant was essentially the same as the Gly-841 deletion on a per flavin basis, although lack of heme suggested that it was expressed as a reductase domain preparation, and the Gly-841 deletion is expressed as holoenzyme. The absolute rates are therefore not directly comparable. The G841Y mutant, also hemeless, had low activity and was CaM-insensitive, in contrast to the corresponding nNOS construct.

Proteolysis of Mutants and iNOS-
The low heme content in some mutants suggested that the enzyme was susceptible to proteolysis; in purified eNOS and nNOS the exposed CaM binding site is the most sensitive to proteolytic cleavage. Because the 2Ј,5Ј-ADP affinity site is associated with the reductase domains, in vivo cleavage at the CaM site would produce flavin-bearing reductase domains without heme.
The iNOS holoenzyme, which lacks the SI, cannot be purified without CaM coexpression (39). When we expressed iNOS without CaM coexpression and fractionated the extract on a 2Ј,5Ј-ADP affinity column, we obtained a yellow flavoprotein preparation spectroscopically indistinguishable from the heme free eNOS mutants. The dominant proteins in 2Ј,5Ј-ADP affin- ity-purified preparations of wild type NOS isoforms are NOS holoenzyme and NOS fragments, in our preparations typically corresponding to at least 50% of the total protein; further column purification with gel filtration or anion exchange produces very pure full-length NOS. The yield of full-length enzyme in preparations of the low heme mutants is obviously low.
Comparison of SDS-PAGE of partially purified wild type iNOS, eNOS, and nNOS, and selected mutants is shown in Fig.  6. In lane 1 the major band corresponds to nNOS Y879S, an FMN-shielding residue mutant which expresses as holoenzyme, with a molecular mass of 161 kDa (upper left arrow). The nNOS mutants in lanes 2 and 4, which have normal or nearly normal heme content, also are present primarily as holoenzyme. The G1074Y mutant, which was run in lane 3, is approximately half-holoenzyme and half-flavoprotein fragment (lower left arrow, corresponding to broad bands near 70 kDa).
Corresponding wild type eNOS preparations have a major band corresponding to full-length holoenzyme as shown in lane 8 (133 kDa; upper center arrow). The eNOS mutants (lanes 9 -11) other than the Gly-841 deletion (lane 9) all have their predominant bands at ϳ 70 kDa, corresponding to flavoprotein fragments (lower center arrow). Very little full-length wild type iNOS was observed in lane 6 (upper right arrow); bands at 70 kDa corresponding to the flavoprotein reductase fragment can be seen (lower right arrow). It is not clear whether the bands visible near 50 kDa are NOS fragments or impurities.
These results indicate that intact holoenzyme is the dominant NOS species in eNOS, nNOS, and the heme-sufficient mutants. In contrast, iNOS and the low heme mutants have been reduced to reductase fragments by in vivo proteolysis. Neither iNOS nor the heme free mutants have significant bands over 100 kDa, and all have multiple bands in the 70 kDa region corresponding to proteolytic products cleaved at or near the CaM binding site. DISCUSSION Previously identified regions involved in the control of NO synthesis by Ca 2ϩ /CaM have either been CaM binding sites or elements involved in suppressing activity in the absence of Ca 2ϩ /CaM. The SI is clearly correlated with control in the evolution of NOS and is structurally positioned to interact with established control elements. The authors of the previous mutagenic study suggested that the SI functions as a secondary autoinhibitory element, but the information obtained in the studies presented here indicate that the role of the SI is more complex and may even differ among isoforms. Unlike the AI and the C-terminal tail, modification, truncation, or deletion of the SI does not consistently result in activation of the enzyme. Although AI deletion results in increased activity and reduc-tion of calcium sensitivity, SI modifications result in less straightforward changes in the activity of NO synthesis.
Typically, ϪAI mutants synthesize NO at low Ca 2ϩ concentrations and have high levels of cytochrome c reductase activity in the absence of CaM. In contrast, ϪSI mutants require CaM for optimal cytochrome c reductase activity as well as NO synthesis, except in cases in which the CaM binding site has been exposed to cleavage, producing flavoprotein expression. We point out that changes in the calcium dependence of mutants in positions adjacent to the CaM binding site can result from interactions with control elements that function as activators as well as displacement of inhibitors because these changes merely reflect the necessity of doing work on a protein structural element during CaM binding.
Intentional disruption of the structure of the SI with incompatible substitutions causes significant loss of NO synthesis activity in nNOS, whereas deletion of a major portion of the SI had a much smaller effect. This suggests that the functions of the SI are ancillary rather than essential for activity; eNOS with a SI deletion or nNOS with a reduced SI can still function, but a disrupted SI can interfere with activation. It is certainly possible that a slightly different full SI deletion in nNOS would also be fully active.
In this context it is worth considering that wild type NOS enzymes lacking a SI are active as long as they also lack an AI. When we designed these mutants we thought it possible that the AI and SI acted cooperatively and that the SI might be needed for AI-mediated inhibition (e.g. as a lock and clasp). The data presented here strongly argue against this view and suggest that instead the SI may function as an accessory element because SI mutants may either positively or negatively affect activity.
Several hypotheses can be constructed for an ancillary role in control. In a direct activation hypothesis, the SI interacts directly with CaM to shift the domain alignment equilibrium, allowing the conformational changes needed for electron transfer to the active site. The SI might also act to destabilize an inactive, conformationally restricted state. It is clear that any such function must represent an enhancement of a mechanism that is operational without the SI because eNOS and nNOS tolerate its removal and/or truncation.
Our recent proposal of a tethered shuttle mechanism for NOS electron transfer and control (40) provides a context for these studies. In this model the FMN binding domain shuttles between FAD and heme facing states, both of which bind CaM. CaM facilitates the release of the FMN domain from the reductase complex, where it is in close association with the FAD and NADPH binding domains. The release of the reduced FMN binding domain allows cytochrome c reduction but is not sufficient to allow NO production. In holoenzyme, realignment of the FMN binding domain, also CaM-facilitated, is necessary for subsequent electron transfer into the oxygenase domain to support catalysis. Since the original submission of this manuscript, a related paper has been published describing differential activation of cNOSs and iNOS by CaM chimera, which supports this interpretation of CaM activation (41).
The evolutionary ancestors of the NOS reductase domains existed as separate proteins closely related to ferredoxin NADPH reductase and flavodoxin, and in these ancestral electron transfer systems ferredoxin/flavodoxin functioned as a shuttle. The FMN binding domain is essentially a ferredoxin tethered to the two-domain reductase unit. In reductase systems in which one component acts as a shuttle, it is common to observe maximum activity at a salt concentration that allows formation of binary complexes for electron transfer (usually optimized at low salt) but does not produce complexes with such slow dissociation rates (optimized at high salt) that the dissociation rate limits the shuttle (42). Salt-inhibited shuttles (iNOS) are characterized by relatively weak interactions, whereas salt-stimulated shuttles (eNOS) are characterized by stronger interactions.
The eNOS ϪSI mutants studied by Knudsen et al. (32) were described in terms of SI inhibition which was "masked" by the AI and salt. Cytochrome c reduction by these mutants at low KCl is not significantly different from the corresponding activity of the parents; at high salt the ϪSI mutant has somewhat lower cytochrome c reductase activity, and the ϪSIϪAI mutant somewhat higher cytochrome c reductase activity, than their parents. Under the conditions that produce enhanced cytochrome c reduction, NO production is lower than or at best equal to that of their parents. Enhanced NO production in ϪSI and ϪSI ϪAI mutants with respect to the parent wild type and ϪAI mutant eNOS enzymes is only observed at low salt, which must be unrelated to enhanced cytochrome c reduction at high salt. This suggests that steady-state cytochrome c reduction by eNOS and all ϪAI and ϪSI mutants is limited at low salt by the dissociation of a tight complex which is stabilized by the AI but not the SI and at high salt by interactions with cytochrome c which can be modestly enhanced by SI removal only in the ϪAI construct.
Salt effects on the activity of NOS isoforms were studied independently by two groups (43,44); the observations were generally similar in that cNOSs were found to be stimulated by moderate salt concentrations comparable with those used by Knudsen et al. (32) and inhibited as the salt concentration was further increased. Schrammel et al. (43) provided data for eNOS and iNOS as well as nNOS and reported that for iNOS cytochrome c reduction and NO synthesis were monotonically inhibited by salt. At least part of the salt inhibition of cytochrome c reduction was attributed to their inability to saturate with cytochrome c at high salt. Schrammel et al. (43) and Nishimura et al. (44) differ with respect to nNOS cytochrome c reduction, which the former report to be monotonically inhibited and the latter initially stimulated by salt. It is likely that the modest gain in cytochrome c reductase activity in the ϪSIϪAI eNOS mutant represents a K m effect with respect to cytochrome c.
NO formation is not rate-limited by the process that limits cytochrome c reduction. In wild type eNOS it is slightly enhanced by salt, but in the ϪSI and ϪAIϪSI mutants NO formation is significantly slower at 0.2 M KCl. This suggests the participation of a second complex characterized by weaker interactions in the ϪSI mutants; clearly, this must be an internal complex involving associations between NOS domains. Interpretation of salt effects is complicated further by the effects on substrate binding and NO dissociation (44). It is of interest that a slightly different SI deletion produces a mutant hyperactive in cytochrome c reduction but, unlike AI deletions, strongly CaM-dependent.
The SI is located in the hinge subdomain, which interacts with all three reductase domains (FMN, FAD, and NADPH binding). The position of the SI indicates that it is in direct contact with bound CaM, and at the same time other residues in the subdomain are hydrogen-bonded both to residues in the other domains and directly to NADP. Although we are not confident enough in the details of models based on incomplete domain structures to assign specific interactions on the FMN binding domain or bound CaM to the SI, its displacement by CaM will clearly affect FMN domain mobility because it forms the terminus of a ␤ hairpin, which forms a three-stranded ␤ structure with the polypeptide strap linking the FMN and FAD binding domains. In this regard it may serve as an amplifier of CaM-driven conformational effects. At the same time, CaMdriven conformational effects on the hinge subdomain are likely to be transmitted to the NADPH binding site through this strap, linking CaM binding, conformation, and nucleotide binding.
The SI has at least one function indirectly related to control; eNOS and nNOS are resistant to proteolysis in cells even without bound CaM, whereas iNOS cannot survive in proteolytically deficient E. coli without CaM coexpression. The increased sensitivity of the CaM binding regions in eNOS and nNOS SI mutants implies that the proximity of the SI to the CaM binding region provides some protection to the enzyme from degradation by proteases. It is possible that this is a major function of the SI in cNOS, although it will be necessary to express these mutants in mammalian cells to determine whether the compartmentalization of activities is sufficient to protect the CaM site in SI-deficient enzymes.
It is obvious that the protective effect of CaM on iNOS is exerted largely by protecting the protease-sensitive CaM binding site. It should be possible to produce intact versions of some mutants which are otherwise produced as reductase fragments by coexpression with CaM, much as iNOS is produced in recombinant systems. In addition, the results presented here suggest the possibility of producing full-length iNOS holoenzyme without CaM coexpression by introducing an SI from eNOS or nNOS. Studies are under way to test these hypotheses.