The C Termini of Constitutive Nitric-oxide Synthases Control Electron Flow through the Flavin and Heme Domains and Affect Modulation by Calmodulin*

The sequences of nitric-oxide synthase flavin domains closely resemble that of NADPH-cytochrome P450 reductase (CPR). However, all nitric-oxide synthase (NOS) isoforms are 20–40 residues longer in the C terminus, forming a “tail” that is absent in CPR. To investigate its function, we removed the 33 and 42 residue C termini from neuronal NOS (nNOS) and endothelial NOS (eNOS), respectively. Both truncated enzymes exhibited cytochrome c reductase activities without calmodulin that were 7–21-fold higher than the nontruncated forms. With calmodulin, the truncated and wild-type enzymes reduced cytochrome c at approximately equal rates. Therefore, calmodulin functioned as a nonessential activator of the wild-type enzymes and a partial noncompetitive inhibitor of the truncated mutants. Truncated nNOS and eNOS plus calmodulin catalyzed NO formation at rates that were 45 and 33%, respectively, those of their intact forms. Without calmodulin, truncated nNOS and eNOS synthesized NO at rates 14 and 20%, respectively, those with calmodulin. By using stopped-flow spectrophotometry, we demonstrated that electron transfer into and between the two flavins is faster in the absence of the C terminus. Although both CPR and intact NOS can exist in a stable, one-electron-reduced semiquinone form, neither of the truncated enzymes do so. We propose negative modulation of FAD-FMN interaction by the C termini of both constitutive NOSs.

flavoprotein domain sharing 36% identity and 58% close homology at the amino acid level (22). Like those of CPR, the flavins of NOSs transfer electrons to artificial electron acceptors, such as cytochrome c, 2,6-dichlorophenolindophenol (DCIP), and ferricyanide (9,23); these rates of reduction also differ very greatly from CPR and between the NOS isoforms. In addition to mediating electron transfer between the flavin and heme domains, CaM also has an effect on the flavin domain itself, as its binding increases the rates of reduction of cyt c, DCIP, and ferricyanide by eNOS and nNOS and by flavoprotein constructs containing the CaM-binding sequences (24).
To account for such differences in calcium dependence, NO production, and cyt c reduction in what are otherwise very homologous structures, attention has now focused on the search for control mechanisms in several regions that exist in the flavin domains of NOSs that have no counterpart in CPR. A putative autoinhibitory domain described by Salerno et al. (25), consisting of about 45 amino acids located in the middle of the FMN-binding region, is present in the constitutive NOSs (nNOS and eNOS) but is absent in iNOS and CPR. This autoinhibitory domain was proposed to control the calmodulin binding/activation of the constitutive isoforms. Deletion of this region in nNOS resulted in an enzyme that attained maximal NO synthesis at lower levels of free calcium and even retained 30% of its activity in the absence of calcium/CaM (26), suggesting that the insert inhibits electron transfer from FMN to the heme in the absence of CaM and also destabilizes CaM binding at low calcium concentrations. The autoinhibitory peptide, however, does not limit the rates of NO synthesis and cyt c reduction; its effect is only on the calcium/CaM dependence.
Another notable difference between the NOS isoforms and CPR is that each of the NOS isoforms contains about 21-42 additional amino acids at the C terminus (42, 33, and 21 amino acids in bovine eNOS, rat nNOS, and murine iNOS, respectively), forming a tail that is not present in CPR. Xie et al. (27) showed that deletion of 22 or 23 residues from the C terminus of iNOS, which removed the tail plus 1 or 2 additional residues, reduced the rate of NO synthesis 26 or 66%, respectively. We have shown that removing the C-terminal 21 amino acids from murine iNOS results in an enzyme with significantly higher cytochrome c reduction and NO synthesis activities (10-and 1.2-fold, respectively; see Ref. 28). The kinetic consequences of this truncation were an increased rate of electron transfer to and between the flavin moieties and from the flavins to the heme. In addition, the air-stable semiquinone form of the flavin domain, observed with the wild-type enzyme, was greatly destabilized, and the heme-nitrosyl complex, formed by the wildtype enzyme during NO synthesis, was decreased in the truncated form. However, iNOS lacks the autoinhibitory domain, and experiments cannot be done in the absence of CaM due to the expression system used, which coexpresses iNOS and CaM. Thus, any effect of the C-terminal tail on the CaM dependence of the enzyme could not be examined.
To investigate the functional role of the additional 33 and 42 residues at the C terminus of nNOS and eNOS, respectively, we have removed them from nNOS and eNOS and compared the resulting enzymes to the intact proteins. The mechanistic implications of this deletion are discussed.

EXPERIMENTAL PROCEDURES
Chemicals-(6R) -5,6,7,8- Recombinant DNA Manipulations-nNOSpCW-tr1 and eNOSpCW-tr1, the plasmids for the expression of the truncated nNOS (residues 1-1397) and eNOS (residues 1-1163) in Escherichia coli, were constructed as follows. For nNOS-tr1, the initial 1397 codons of nNOSpCW (29) (from the ATG start codon to the final residue minus 33) were amplified by PCR. Primer tr1-1 (upstream primer, with NdeI site) was 5Ј-TGCTTAGGAGGTCATATG-3Ј, and primer tr1-2 (return primer, with XbaI site) was 5Ј-CTAGTCTAGATTATCCAAAGATGTCCTCGT-G-3Ј. Primers were synthesized by the Center for Advanced DNA Technologies at the University of Texas Health Science Center, San Antonio. Reaction mixtures included 50 pmol of each primer, 20 ng of nNOSpCW template, 200 M dNTPs, 1ϫ Pfu turbo polymerase buffer, and 2.5 units of Pfu turbo polymerase in 50 l total volume. The mixture was preincubated for 1 min at 94°C prior to the addition of Pfu turbo polymerase, followed by amplification for 18 cycles: 95°C for 45 s, 55°C for 60 s, and 68°C for 10 min. The PCR product was gel-purified using the Gene-Clean II kit (Bio 101, Vista, CA) and digested with NdeI and XbaI. PCW oriϩ DNA was digested with NdeI and XbaI, and the ends were dephosphorylated. The resulting nNOS cDNA and pCW plasmid were ligated, and the ensuing products were used to transform E. coli XL10gold competent cells (Stratagene) using the manufacturer's instructions. Positive colonies were identified by restriction digest analysis of plasmid DNA isolated from small (2 ml) cultures of select colonies. For eNOS-tr1, glycine in position 1163 of bovine eNOS represents the residue aligned with the terminal serine of CPR. Preparation of parental plasmid bov-eNOSpCWoriϩ was described previously (30). The sense primer 5Ј-GACATCCTGAGAAACCGAGCTG-3Ј was located between nucleotides 3276 and 3297 and the antisense primer 5-TATAT-GAATTCTCATCAGCCGAAAATGTCCTCGTGATAGC-3Ј encompassed the codon for Gly-1163 followed by two stop codons and contains an EcoRI restriction site. A DNA product amplified by PCR using the parental plasmid as the template and above-mentioned oligonucleotides was digested by XhoI and EcoRI. The longer XhoI-EcoRI fragment was gel-purified. The parental plasmid was digested by XhoI-EcoRI; the longer fragment was gel-purified and ligated with the XhoI-EcoRI fragment from PCR. DNA from the new construct was sequenced in the PCR-amplified region using automated sequencing at the Center for Advanced DNA Technologies at the University of Texas Health Science Center, San Antonio.
groELS was co-transformed with nNOSpCW-tr1 or eNOSpCW-tr1 into E. coli BL21 cells via electroporation using an Invitrogen Electroporator II (San Diego, CA) according to the manufacturer's instructions and plated on LB agar containing 50 g/ml ampicillin and 35 g/ml chloramphenicol.
Protein Expression/Purification-Protein expression and purification are as described in Roman et al. (29) for nNOS and nNOS-tr1. Protein expression and purification for eNOS and eNOS-tr1 are as described in Martá sek et al. (30). The peak fraction of dimeric protein from the gel filtration column was used for all analyses described. Both nNOS-tr1 and eNOS-tr1 were purified in tandem with their wild-type forms to reduce differences resulting from preparation to preparation.
Spectrophotometric Methods-Absolute spectra and CO difference spectra were performed as described (29). The molar protein concentrations for nNOS and nNOS-tr1 were determined based on heme content via reduced CO difference spectra, where ⑀ ϭ 100 mM Ϫ1 cm Ϫ1 for ⌬A 445-470 . The molar protein concentrations for eNOS and eNOS-tr1 were determined based on spectral determination at Soret maximum after subtraction of extrapolated base line, using ⑀ ϭ 100 mM Ϫ1 . This protein concentration determination was in good agreement with that of CO difference spectum. All spectral analyses were performed using a Shimadzu model 2101 UV-visible dual-beam spectrophotometer.
Stopped-flow Spectroscopy-Stopped-flow reactions were performed aerobically under turnover conditions at 23°C, as described in Miller et al. (23) and Roman et al. (28), using an Applied Photophysics SX.18MV diode array stopped-flow spectrophotometer, which had a dead time of 2 ms. Reactions contained 2 M enzyme, 100 M NADPH, 10 M H 4 B, and 100 M L-arginine in 50 mM HEPES, pH 7.4, and 100 mM NaCl. Where indicated, 3 M CaM was also added. Heme reduction was monitored at 397 nm, and flavin reduction was monitored at the 485 nm shoulder, rather than the absorption peak value at 455 nm, to avoid spectral interference from the heme.
Measurement of Activity-Nitric oxide formation (hemoglobin capture assay and/or [ 3 H]arginine to citrulline assay) and cytochrome c reduction were measured at 23°C as described by Sheta et al. (9) and Martá sek et al. (31), with the exception that the cytochrome c reduction assays were performed in a buffer containing 50 mM HEPES, pH 7.4, using an extinction coefficient of 21 mM Ϫ1 for reduced minus oxidized cyt c at 550 nm. NaCl was 0, 100, or 250 mM, as indicated in the figure or table legends. Ferricyanide reduction was performed the same as cyt c reduction except that 100 mM ferricyanide was used, and the extinction coefficient was 1.02 mM Ϫ1 .
Oxidation of NADPH was monitored at 340 nm in the presence of 50 mM HEPES, pH 7.4, 100 M NADPH, and 26 nM truncated or 134 nM wild-type enzyme at 23°C. In the case of NADPH oxidation under cyt c reduction conditions, 250 mM NaCl was included. The rate was determined using an extinction coefficient of 6.2 mM Ϫ1 at 340 nm for NADPH.
The reoxidation of reduced flavins was monitored at 485 nm for all enzymes in the presence of 50 mM HEPES, pH 7.4, 250 mM NaCl, 20 or 100 M NADPH, and 2 M enzyme at 23°C. Fig. 1 shows the C-terminal final 8 amino acids of CPR and the co-alignment of several NOS sequences. Depicted are the murine iNOS, rat nNOS, and the bovine eNOS along with two non-mammalian sequences for comparison. All known NOSs appear to have additional C-terminal residues as compared with CPR. Also shown in the boxed region is a highly conserved 8-amino acid sequence just prior to the NOS tail. The penultimate amino acid in CPR, tryptophan, has been shown to be important for reductase activity. Specifically, it has been proposed to be involved in modulating the hydride transfer from NADPH to FAD and, in the crystal structure of CPR (32), it appears to be a shielding or stacking residue for the FAD. The aromatic property of this residue is completely conserved in all known NOSs, and deletion of the C terminus of iNOS including this residue to the C-terminal end yields a protein with only 30% of wild-type NO synthesis activity (27), 3 whereas deletion of only the additional tail residues, i.e. the final 21 amino acids, yields a protein with 15% higher activity (28).

RESULTS
Cytochrome c Reduction-The path of electron transfer through the NOSs is from NADPH to the FAD moiety, to FMN, and finally to the NOS heme (Scheme 1). Like CPR, the NOSs also have the ability to reduce artificial electron acceptors, such as cyt c, DCIP, and ferricyanide. Cyt c and DCIP both accept electrons from the FMN, whereas ferricyanide can accept electrons directly from the FAD, as shown by its reduction by a CPR in which FMN has been removed (33,34). When the iNOS isoform was expressed without the C-terminal tail (minus the 21 residues of the C terminus), the rate of cyt c reduction was increased 10-fold (28). It was not possible to examine the effect of CaM with iNOS-tr1 since the enzyme was expressed in the presence of CaM, which remains tightly bound. The constitutive enzymes, nNOS and eNOS, are easily expressed in the absence of CaM, whose binding is modulated by the calcium concentration. The rate of cyt c reduction by wild-type nNOS is stimulated about 10-fold by CaM (345 versus 3860 min Ϫ1 , in the absence and presence of CaM, respectively; Fig. 2A) and that of eNOS is stimulated about 3-fold (173 versus 478 min Ϫ1 , in the absence and presence of CaM, respectively; Fig. 2A). Surprisingly, both nNOS-tr1-and eNOS-tr1-catalyzed cyt c reduction are inhibited by CaM. Although the rates in the presence of CaM are similar for both the wild-type and truncated enzymes (3860 versus 4604, respectively, for nNOS and nNOS-tr1, and 478 versus 665, respectively, for eNOS and eNOS-tr1; Fig. 2A), nNOS-tr1 reduces cyt c at a 56% higher rate than nNOS-tr1 in the presence of CaM and 21-fold faster than wild-type in its absence. Likewise, eNOS-tr1 reduces cyt c at a 85% higher rate than eNOS-tr1 in the presence of CaM and 7-fold faster than wild-type in its absence.
Since the most significant differences in rate between the wild-type and tr1 forms are seen in the absence of CaM, it was possible that the loss of the tail affected the binding of cyt c under these conditions. The K m and k cat values for cyt c in cyt c reduction by either wild-type or truncated enzyme were determined in the absence of CaM. There is no difference in the K m value in the absence of CaM between the wild-type and truncated forms for either nNOS or eNOS (approximately 2-3 M; Table I), and thus the C-terminal tail probably does not directly affect binding of cyt c.
To investigate further the effect of CaM on cyt c reduction, assays with both the wild-type and truncated enzymes were titrated with CaM, as shown in Fig. 3. With the wild-type enzyme, CaM is a nonessential activator of cyt c reduction; it is not required for activity, but it potentiates the formation of reduced product.
The kinetic constants for the wild-type enzymes were determined by varying the cyt c concentration in the absence of CaM and at saturating CaM concentrations, and the resulting data are shown in Table I. In the case of the wild-type nNOS, CaM at a saturating concentration increases the K m for cyt c by 1.6-fold and k cat by 13-fold. In the case of eNOS-wt, saturating CaM decreases the K m for cyt c by 25% and increases k cat about 4-fold.
As shown in Fig. 3, CaM clearly has an inhibitory effect on cyt c reduction by nNOS-tr1, even at low concentrations of CaM. CaM and cytochrome c bind independently and reversibly to different sites on NOS to produce NOS⅐CaM, NOS⅐cyt c, and NOS⅐CaM⅐cyt c complexes. Since both the NOS⅐cyt c and NOS⅐cyt c⅐CaM complexes can form product (reduced cyt c), with the NOS⅐cyt c⅐CaM complex being less effective than NOS⅐cyt c, and the activity reaches a minimum that is not zero, CaM must be a partial noncompetitive inhibitor of cyt c reduction by the NOS-tr1 enzymes.
The kinetic constants for the truncated enzymes are also shown in Table I. In the case of nNOS-tr1 and eNOS-tr1, CaM at a saturating concentration decreases the k cat of cyt c reduction by 36 and 47%, respectively. Table I also shows the enzyme efficiencies (k cat /K m ) for the wild-type and truncated enzymes in the absence and presence of CaM. Interestingly, removing the C-terminal tail from nNOS dramatically increases the ef-ficiency of cyt c reduction (26-fold) in the absence of CaM. CaM increases the efficiency of nNOS-wt, while not greatly affecting that of nNOS-tr1, although the tail-less enzyme is still 4-fold as efficient as the wild-type form. With eNOS, the catalytic efficiency changes are less dramatic; in the absence of CaM, the truncated enzyme is about 11-fold more efficient than eNOSwt, and in the presence of CaM, eNOS-tr1 is 2.6-fold more efficient than eNOS-wt.
Ferricyanide Reduction-An increased rate of cyt c reduction is caused by an alteration in the rate-determining step, which must be one of the following: binding of NADPH or cyt c, transfer of hydride ion from NADPH to FAD, transfer of electrons during catalysis from FAD to FMN, comproportionation of electrons between the flavins, or transfer of electrons from FMN to cyt c. As shown in Table I, the K m for cyt c is the same in both the intact and truncated enzymes, so interaction with cyt c is probably not altered, and it is unlikely that binding of NADPH is rate-determining. In the wild-type enzymes, the transfer of electrons from FAD to FMN is at least as fast as, if not faster than, that of transfer from NADPH to the first flavin (35). Ferricyanide is a small artificial electron acceptor that can receive electrons from the FAD moiety, bypassing the FAD to FMN transfer (Scheme 1). Measurement of ferricyanide reduction, then, will determine whether the wild-type and truncated enzymes differ with regard to introduction of the first electron to FAD. These data are shown in Fig. 2B. For nNOS-wt, the rate of FeCN reduction is increased 1.6-fold in the presence of CaM, indicating that CaM may have some effect on the introduction of electrons to FAD. For nNOS-tr1, this difference is

TABLE I Kinetic constants for Cyt c reduction
These experiments were performed in the absence of NaCl. The addition of 250 mM NaCl increases the rate of NOS-mediated cyt c reduction by about 3-fold (see rates in Fig. 2). The concentrations of enzymes used were as follows: nNOS-wt, 10 nM; nNOS-tr1, 1 nM; eNOSwt, 30 nM; eNOS-tr1, 30 nM. These concentrations were chosen so that the absolute rate of cyt c reduction was in a reasonable range. The concentration of CaM was 60 nM for nNOS-wt, 30 nM for nNOS-tr1, and 100 nM for both eNOS enzymes. Assay conditions were as described under "Experimental Procedures."

Enzyme
ϪCaM ϩCaM abolished. The rate of ferricyanide reduction is the same whether CaM is present or not. For eNOS, the data are less clear. The stimulation of FeCN reduction for eNOS-wt is about 1.3-fold in the presence of CaM. The eNOS-tr1 is inhibited about 17% in the presence of CaM, consistent with the inhibition seen with cyt c reduction in the presence of CaM, but to a much lesser extent. NO Synthesis-The ability of the truncated enzymes to synthesize NO in the presence and absence of CaM was measured, and the results are shown in Fig. 4. Although the nNOS-tr1 and eNOS-tr1 enzymes are capable of producing NO in the presence of CaM, they do so at rates that are 45 and 33%, respectively, of their intact parent enzymes. Most interesting, however, is the ability of both nNOS-tr1 and eNOS-tr1 to produce NO at measurable rates in the absence of CaM. nNOS-tr1 and eNOS-tr1 produce NO in the absence of CaM at rates 14 and 20%, respectively, as in the presence of CaM (Fig. 4). nNOS-wt has no demonstrable NO formation in the absence of CaM, whereas eNOS-wt has only negligible activity (2.5% that with CaM).
Flavin and Heme Reduction-To examine directly flavin and heme reduction of the wild-type and truncated proteins, stopped-flow spectrophotometry under turnover conditions was used. Spectral changes were monitored at 485 nm for flavin reduction and 397 nm for heme reduction following rapid mixing of NOS with NADPH. As previously reported (23), the reduction of flavin by nNOS is a biphasic process and that of heme reduction is monophasic. The initial fast phase of flavin reduction represents the conversion of the fully oxidized flavoprotein to the fully reduced and semiquinone forms, whereas the slower second phase likely represents the comproportionation of electrons, including forward and back reactions between the flavins. The rates determined are shown in Table II. For nNOS, the primary and secondary rates of flavin reduction in the presence of CaM are very similar for both the wild-type and truncated forms. In the absence of CaM, the primary and secondary rates for nNOS-tr1 are 7-and 3-fold higher, respectively, than those of nNOS-wt. These data fit extremely well with the observation that cyt c reduction activity for both enzymes is similar in the presence of CaM, but in the absence of CaM, nNOS-tr1 reduces cyt c 21-fold faster than nNOS-wt. Heme reduction by nNOS-wt in the presence of CaM is about 2-fold that of nNOS-tr1, as is NO synthesis activity, whereas in the absence of CaM that of nNOS-tr1 is 2.4-fold faster than that of nNOS-wt. Most interestingly, the rate of heme reduc-tion by nNOS-tr1 in the absence of CaM is the highest observed in these conditions, although the rate of NO formation is lowest. Presumably, under these conditions, superoxide is being formed by dissociation of the ferrous dioxygen complex from the heme domain.
The curve for flavin reduction by the eNOS-wt and tr1 proteins fit better to a single exponential for these preparations; the very slow secondary rate of flavin reduction, as shown by Miller et al. (23), could not be resolved from the initial fast rate. As with the nNOS proteins, the rates of flavin reduction for both the wild-type and truncated eNOSs are similar in the presence of CaM but that of eNOS-tr1 is 21-fold faster than that of eNOS-wt in the absence of CaM. Heme reduction rates are very low in all cases, and differences between them cannot be distinguished.
Flavin Reoxidation-Once NADPH is exhausted and catalysis stops, NOS returns eventually to its resting, fully oxidized state but only after a transient period in a partially reduced state. In studies with nNOS (23,36) and iNOS (28) the formation of an air-stable, one-electron-reduced semiquinone was demonstrated. This form of the reductase cannot reduce either its own heme or that of cytochrome c and persists for about 20 min before reverting to the fully oxidized state. Since the airstable semiquinone form of a similarly C-terminal truncated form of iNOS was shown to be unstable, the rate of flavin reoxidation was examined for intact and truncated nNOS and eNOS.
Flavin reoxidation was monitored at 485 nm and appears as an increase in flavin absorbance. The reaction begins with fully reduced flavins, and the initial, zero absorbance represents fully reduced flavins catalyzing the oxidation of NADPH. After the NADPH is exhausted, flavin reoxidation, seen as an absorbance increase, occurs. The absorbance change plateaus, representing the air-stable form of the flavins. In Fig. 5, the amplitude of the signal for nNOS-tr1, in both the presence and absence of CaM, is twice those of the wild-type enzyme, indicating that, whereas the intact enzyme initially forms the stable, one-electron-reduced semiquinone, the truncated form does not; rather, it becomes fully oxidized. The slope of the line represents the rate of flavin oxidation, and it is clear that this rate for the truncated form is at least twice that of the wildtype forms, which is consistent with both flavins being rapidly and fully oxidized, rather than just one as in the semiquinone form. Both eNOS-wt and eNOS-tr1 also follow this behavior in both the presence and absence of CaM (not shown). The circles are eNOS-wt and -tr1 and are plotted against the right axis. The open symbols are in the wild-type enzymes, and the solid symbols are the truncated enzymes. These data represent the mean Ϯ S.E. of at least three separate assays. The concentrations of enzymes used are as follows: nNOS-wt, 10 nM; nNOS-tr1, 1 nM; eNOS-wt, 30 nM; eNOS-tr1, 30 nM. These concentrations were chosen so that the absolute rate of cyt c reduction was linear and detectable. Assay conditions were as described under "Experimental Procedures," and concentration of NaCl was 0 mM. Note that the data in Fig. 2A were generated in 250 mM NaCl and thus are about 2.5-fold higher. The trends remain the same regardless of the NaCl concentration.
Superoxide Production-Although the rate of heme reduction by nNOS-tr1 in the absence of CaM is higher than in its presence or than nNOS-wt in either case, the amount of NO production is much lower. Cyt c may be reduced by superoxide in addition to direct electron transfer from NOS. To determine whether superoxide is being formed in lieu of NO or is contributing to cyt c reduction, NADPH oxidation, as well as the superoxide-mediated conversion of epinephrine to adrenochrome by intact and truncated nNOS, was measured. nNOS-wt oxidizes NADPH in the presence of cyt c at rates of 430 and 3600 min Ϫ1 in the absence and presence of CaM, respectively. nNOS-tr1 oxidizes NADPH in the presence of cyt c at rates of 7260 and 4200 min Ϫ1 in the absence and presence of CaM, respectively. These rates of NADPH oxidation are equivalent to the rates of cyt c reduction under the same conditions. Therefore, in the case of nNOS-wt and -tr1 at least, superoxide production does not contribute to cyt c reduction. In the absence of cyt c, i.e. with no electron acceptor other than oxygen present, nNOS-wt oxidizes NADPH at rates of 5 and 360 min Ϫ1 in the absence and presence of CaM, respectively, whereas nNOS-tr1 oxidizes NADPH at rates of 50 and 170 min Ϫ1 in the absence and presence of CaM, respectively. Clearly, in the absence of CaM, nNOS-tr1 is 10-fold more active than nNOS-wt in superoxide production, but in the presence of CaM, nNOS-tr1 produces superoxide at half the rate of nNOS-wt. In addition, superoxide production is affected much less by CaM during catalysis by the truncated enzyme than by the wild-type.
Superoxide production was also measured using an assay that monitors the superoxide-mediated conversion of epinephrine to adrenochrome (31,37). In this assay, the lag before the onset of redox cycling is the measure of superoxide production. Thus, comparing the lag periods between the various enzymes will give information about their relative rates of superoxide production. Consistent with NADPH oxidation, superoxide production of both nNOS and eNOS followed the trend: NOS-wt ϩ CaM Ͼ NOS-tr1 ϩ CaMϾNOS-tr1 Ϫ CaM Ͼ Ͼ NOS-wt Ϫ CaM (data not shown). DISCUSSION We have demonstrated that the 33 and 42 C-terminal residues of nNOS and eNOS, respectively, play a role in modulating electron transfer through the flavin domain and to the heme domain. We made these constructs to mimic the CPR, which ends 21-42 residues earlier than the NOSs depicted in Fig. 1. In the absence of these "tail" residues, the cyt c reductase activity is 21-and 7-fold higher (in the absence of CaM) for nNOS and eNOS, respectively, than their wild-type parents. These additional tail residues also alter the effect of CaM on the enzyme for both cyt c reduction and NO synthesis. For cyt c reduction by the wild-type enzyme, CaM is a nonessential activator of enzymatic activity. That is, it is not necessary for activity but causes an increase in the rate of cyt c reduction when present. When the tail is removed, CaM now becomes a partial noncompetitive inhibitor of cyt c reduction. Enzyme activity is inhibited but approaches a limit that is greater than zero since both the NOS⅐cyt c and NOS⅐CaM⅐cyt c complexes can form product. For NO synthesis by the wild-type enzyme, CaM is an essential activator. Product is not formed in its absence. When the tail is removed, however, CaM now becomes a nonessential activator. As with wild-type cyt c reduction, either the NOS (flavin-heme) complex or the NOS (flavinheme)⅐CaM complex is capable of forming product, in this case, NO.
Interestingly, nNOS becomes a more catalytically efficient enzyme with regard to cyt c reduction in the absence of the C-terminal tail (26-and 2-fold higher in the absence and pres- For the eNOS enzymes, the data for heme reduction best fit a two-exponential model, with the first rate being contributed by flavin reduction. Only the second rate, which represents true heme reduction is shown. ence of CaM, respectively (Table I)). eNOS-tr1 is also more efficient in the absence or presence of CaM than wild-type by 11-and 2.6-fold, respectively. Granted, the in vivo role of NOS is not to reduce cyt c, but this activity seems a reasonable gauge for the catalytic ability of the reductase domain independent of the heme domain. Therefore, why would the reductase domain need to be hobbled by the addition of the tail, and how does the presence of the tail cause this attenuation? In part, the C terminus forces the regulation of electron transfer by CaM. The wild-type and truncated enzymes are very similar in the presence of CaM, but electron transfer through the flavin domain is greatly potentiated in the truncated enzymes in the absence of CaM. The truncated enzymes are also able to form NO in the absence of CaM.
The tail is most likely not interacting directly with cyt c, as there is no change in the K m for cyt c reduction between the wild-type and truncated forms. Since the site of electron transfer from the FMN to a heme, be it that of cyt c or NOS, is probably the same regardless of the final acceptor, the tail most likely does not interact directly with the heme domain of NOS either. Other possible sites of interaction could be the autoinhibitory loop or CaM itself, but by analogy with the CPR crystal structure, these sites are probably too far away, being at the N-terminal end of the reductase domain.
Since CaM binds at the N-terminal side of the reductase domain, it is likely situated at one side of the flavin wall (re-face of the FAD isoalloxazine ring), and the C-terminal tail covers the other side of the flavin wall. Therefore, CaM and the C-terminal tail are modulating the rate of electron transfer between the two flavins by adjusting the distance and the relative orientation of the two isoalloxazine rings. It is also conceivable that these two modulators regulate electron flow from FMN to the terminal electron acceptor, i.e. cyt c, ferricyanide, or the heme domain of the NOS protein. It should be noted that the effects of CaM and the tail on the electron flow between the flavins are opposite each other, i.e. activator and inhibitor (or push and pull), respectively. The inhibitory effect of the tail, however, can be compensated by saturating amounts of CaM (Fig. 4). On the other hand, the electron transfer between FMN and the NOS heme domain (NO production) is increased by the C-terminal tail in the presence of CaM, whereas it is decreased in the absence of CaM. This suggests that, in the presence of the tail, the interaction between the reductase and the oxygenase domains is positively modulated, whereas it is negatively modulated in the absence of CaM. It is conceivable that the CaM, together with the autoinhibitory loop of the cNOSs, positively modulates the interaction between the two domains, whereas the tail blocks the interaction. Obviously the CaM effect is much greater than that of the tail.
Consistent with what was seen with iNOS, none of the truncated isoforms exhibit the long term stability of the one-electron-reduced semiquinone form of the enzyme seen with the intact forms. As proposed in the case of iNOS (28), the C-terminal tail protects the air-stable semiquinones in cNOSs by protecting the "opening" of the FAD-FMN interface from molecular oxygen in the air. Through this opening, electrons can be transferred to ferricyanide. Since the electrons flow directly to ferricyanide, bypassing FMN, the pushpull effect of the tail is no longer present in the ferricyanide reduction activity.
Fulton et al. (38) demonstrated the presence of an active protein kinase Akt phosphorylation site at serine 1179, which is in the C-terminal tail, of eNOS. In vivo phosphorylation of this site in cultured endothelial cells results in an increase in the specific activity of NO production, as measured by nitrite formation, of about 4-fold. McCabe et al. (39) mutated serine 1179 to aspartate, a negatively charged residue that mimics the negative charge contributed by phosphorylation at that site. There is also an equivalent serine in the C terminus of nNOS, which we have also mutated to aspartate, although it has not yet been shown to be physiologically relevant. The mutant eNOS proteins demonstrated a 2-fold increase in NO production and a 2-4-fold increase in cyt c reduction. The mutant nNOS protein did not show an increase in NO production but did reduce cyt c at a rate 60% greater than wild-type nNOS. Both eNOS and nNOS mutants exhibited faster rates of flavin reduction as measured by stopped-flow, particularly in the second, slow phase. 4,5 These data are consistent with a model where the C-terminal tail inhibits electron transfer through the flavin domain in the absence of CaM but is moved so as to be less inhibitory in its presence. The negatively charged phosphate or aspartate at this position may be repelled by neighboring negatively charged residues, thus changing the position of the tail.
It is interesting that the truncated iNOS enzyme does not appear to show the same inhibition by CaM as do the truncated FIG. 5. Reoxidation of reduced flavins. Enzyme concentration was 2 M; CaM concentration was 3 M, and NADPH is 50 M for nNOS-tr1 with and without CaM, 5 M for nNOS-wt without CaM, and 250 M for nNOS-wt with CaM. Assays were performed as described under "Experimental Procedures" in 50 mM HEPES, pH 7.4, and 250 mM NaCl. eNOS and nNOS isoforms (28). iNOS-tr1 reduced cyt c at a rate that was 10-fold greater than that of iNOS-wt and synthesized NO 20% faster than wild-type enzyme, all in the presence of CaM, necessarily. These experiments were not done in the absence of CaM, since it was coexpressed with the iNOS and couldn't be removed. However, for both nNOS and eNOS, the presence of CaM attenuated cyt c reduction such that it was similar to that of wild-type. Perhaps the realignment of the flavin domain by CaM binding and the autoinhibitory loop shift (which is not present in iNOS) are not relevant for iNOS, because CaM is permanently bound and not required for modulation of the enzyme. iNOS also has the shortest tail of the isoforms and lacks the protein kinase phosphorylation site in the tail region.
Although CaM and the autoinhibitory loop of NOS appear to be the primary mechanism of control of electron transfer from the flavin to the heme domain (8,25,26), the C-terminal tail also plays a role in negatively controlling electron transfer. There is a measurable conformational change in NOS, particularly in the flavin domain, upon binding of CaM, as shown by changes in intrinsic fluorescence and in the trypsin proteolysis pattern (9,25,40). It is possible that, in the absence of CaM, the C-terminal tail is located between the flavins and/or between the FAD and the NADPH, slowing down electron transfer between them. When CaM is bound and the autoinhibitory loops swings away, this change in the protein conformation realigns the flavins and the NADPHbinding site such that the C terminus is less of a barrier. This is supported by the increase in rate seen in both phases of flavin reduction as measured by stopped-flow. Also, the observation that there is little to no difference in ferricyanide reduction by the truncated NOSs in the presence or absence of CaM, while CaM stimulates ferricyanide reduction of the wild-type enzymes, indicates that the tail has an effect on NADPH to FAD transfer. Clearly, the detailed interactions between CaM, the C-terminal tail, and the autoinhibitory loop must await three-dimensional structure determination of the reductase domains of NOS isozymes.