Differential Activation of Nitric-oxide Synthase Isozymes by Calmodulin-Troponin C Chimeras*

The interactions of neuronal nitric-oxide synthase (nNOS) with calmodulin (CaM) and mutant forms of CaM, including CaM-troponin C chimeras, have been previously reported, but there has been no comparable investigation of CaM interactions with the other constitutively expressed NOS (cNOS), endothelial NOS (eNOS), or the inducible isoform (iNOS). The present study was designed to evaluate the role of the four CaM EF hands in the activation of eNOS and iNOS. To assess the role of CaM regions on aspects of enzymatic function, three distinct activities associated with NOS were measured: NADPH oxidation, cytochrome c reduction, and nitric oxide ( (cid:1) NO) generation as assessed by the oxyhemoglobin capture assay. CaM activates the cNOS enzymes by a mechanism other than stimulating electron transfer into the oxygenase domain. Interactions with the reductase moiety are dominant in cNOS activation, and EF hand 1 is critical for activation of both nNOS and eNOS. Although the activation patterns for nNOS and eNOS are clearly related, effects of the chimeras on all rived EF hand that is incapable of Ca 2 (cid:1) binding and that is unlikely to be capable of stabilizing interactions to the extent of the CaM cognate it replaces, the effectiveness of CaM 1TnC is remarkable. It suggests that the initial EF hand of CaM and the associated flanking regions are relatively unimportant in iNOS activation. On the other hand, the CaM 2TnC and CaM 3TnC results demonstrate that these regions are important determinants of iNOS Ca 2 (cid:1) -independent activity.

Ca 2ϩ represents an important signaling molecule that functions to initiate a range of diverse cellular processes such as muscle contraction, neurotransmission, memory, cell fertilization, cell proliferation, cell defense, and cell death (1). The 17-kDa cytosolic protein calmodulin (CaM) 1 is the major Ca 2ϩ sensor that functions to rapidly up-regulate intracellular metabolism through the coordinated activation of approximately 30 intracellular proteins (2). There is considerable interest in understanding the structural basis of the target protein interactions and diverse regulatory functions of CaM.
Cardiac troponin C (TnC) is a Ca 2ϩ -binding protein whose function is to activate the thin filament cardiac muscle (3). The structure of TnC is in some respects quite similar to CaM in that both have two globular domains linked by a central helix; each globular domain is composed of two EF hands. In particular, the local structures of three of the four EF hand regions in the two proteins are closely related. However, the structure of one of the EF hands (EF hand 2) of TnC differs from the corresponding region in CaM in that the final helix is redirected significantly. This produces a significant difference in the global three-dimensional structure. Despite the structural similarity, TnC has only a 70% sequence similarity with CaM. The first EF hand of TnC does not bind Ca 2ϩ because of a mutation relative to a CaM-like ancestor (5,6). Furthermore, TnC has seven more N-terminal and three more central helix residues than CaM. Significantly, TnC does not activate CaMdependent enzymes (3,4). The C-terminal domain of CaM binds Ca 2ϩ with a higher affinity (K d ϭ 10 Ϫ6 M) than the N-terminal domain (K d ϭ 10 Ϫ5 M) (7), but the Ca 2ϩ -binding site 4 of TnC binds Ca 2ϩ with 1 order of magnitude more affinity than the corresponding site in CaM (8).
Because of the overall structural similarity, domains of CaM can be exchanged for the corresponding domains of TnC to investigate the function of specific regions of the two proteins. Previous investigations utilized the exchanging of domains or elements of the EF hand motif between CaM and TnC to find regions that affect target enzyme activation (5, 6, 8 -10). Chimeras of CaM including each of the four domains from TnC have been constructed (e.g. CaM 1TnC is domain 1 of TnC and domains 2, 3, and 4 of CaM) (5,6).
Nitric-oxide synthase (NOS) enzymes are homodimers with each monomer containing a C-terminal reductase domain where NADPH, FAD, and FMN bind. The N-terminal oxygenase domain contains binding sites for protoporphyrin IX and tetrahydrobiopterin (H 4 B), as well as the substrates L-arginine and molecular oxygen (11). A CaM-binding domain separates the oxygenase and reductase domains of NOS. The binding of CaM to nNOS and endothelial NOS (eNOS) facilitates NADPH-derived electron transfer into the flavins within the reductase domain and enables the flavins to transfer electrons to the heme in the oxygenase domain (12)(13)(14). Although the investigation of nNOS interactions with CaM and mutant forms of CaM including CaM-TnC chimeras have been previously reported, there has never been a thorough investigation of the other constitutively expressed NOS enzyme, eNOS, or the inducible isoform (iNOS). A recent enzyme binding and activation study using both nNOS and eNOS has shown important differences in the effect that oxidation of CaM methionines has upon these two constitutively expressed NOS (cNOS) enzymes (15). Furthermore, very little is known about the interaction of CaM and activation of the inducible NOS (iNOS) isozyme. The present study was designed to evaluate the role of the four different CaM EF hands in the activation of the eNOS and iNOS enzymes. Five different CaM-TnC chimeras were used in the investigation. To assess the role of different CaM regions on aspects of NOS enzymatic function, three distinct activities associated with NOS were studied: 1) electron transfer from NADPH to the reductase complex (measured by NADPH oxidation); 2) electron transfer from FMN to an acceptor (measured by cytochrome c reduction); and 3) electron transfer to the heme (measured by generation of ⅐ NO). Biochemical studies coupled with molecular modeling of the different CaM-TnC chimeras show significant differences in the binding and activation of iNOS and cNOS enzymes by CaM. Our results show that domain 2 and the associated linker sequence affect iNOS domain alignment and catalytic activity at low Ca 2ϩ concentrations.

EXPERIMENTAL PROCEDURES
NOS Enzyme Expression and Purification-Rat neuronal and bovine endothelial NOS were expressed in Escherichia coli and purified as previously described (15)(16)(17). The human iNOS enzyme carrying a deletion of the first 70 amino acids and an N-terminal polyhistidine tail was coexpressed with the appropriate wild-type or mutant CaM in E. coli. The iNOS enzymes were purified using a combination of ammonium sulfate precipitation and metal chelation chromatography, followed by 2Ј,5Ј-ADP column chromatography as previously reported (16).
Calmodulin-Troponin C Chimera Subcloning, Expression, and Purification-The vectors coding for the five calmodulin-cardiac troponin C (CaM-TnC) chimeras (CaM 1TnC, CaM 2TnC, CaM 3TnC, CaM 4TnC, and CaM 3,4TnC) were a generous gift from Dr. Sam George (8). The chimera CaM 1TnC is identical to CaM except that residues 1-47 (domain 1) of CaM are replaced by amino acid residues 1-56 (domain 1) from TnC. The chimera CaM 2TnC is identical to CaM except that residues 48 -85 (domain 2 and linker region) are replaced by residues 57-97 (domain 2 and linker region) of TnC. The cloning required the conservative replacement of serine 93 by a threonine in the TnC segment found in CaM 2TnC. The chimera CaM 3TnC consists of the CaM coding region except that amino acid residues 86 -121 (domain 3) are replaced by residues 98 -134 (domain 3) of TnC. The chimera CaM 4TnC includes the entire CaM coding region except that amino acid residues 122-148 are replaced by residues 134 -161 of TnC. Finally, the chimera CaM 3,4TnC consists of residues 1-85 (domains 1 and 2 and the linker region) from CaM and residues 98 -161 (domains 3 and 4) from TnC.
These vectors were all heat-inducible and carried an ampicillinresistant marker, necessitating the construction of new expression vectors with different induction properties and antibiotic selection for co-expression with the iNOS enzyme. The coding regions for CaM 1TnC, CaM 2TnC, CaM 3TnC, CaM 4TnC, and CaM 3,4TnC were subcloned into the kanamycin-resistant pET28a vector (Novagen) using the unique NcoI and NdeI restriction sites found at the beginning and end of each of the respective genes in the coding vectors. The new expression vectors were named pCaM1TnCKan, pCaM2TnCKan, pCaM3TnCKan, pCaM4TnCKan, and pCaM3,4TnCKan. The entire coding region was sequenced to ensure that no spontaneous mutations had occurred during the construction of the expression vectors.
Overnight cultures of BL21 (DE3) E. coli transformed with the appropriate plasmid were used to inoculate 1 liter of LB media in a 4-liter flask supplemented with 100 g/ml of the appropriate antibiotic. The 1-liter cultures were grown at 37°C until the optical density measured at 600 nm reached 0.8 -1.2, at which time the cultures were induced with 500 M isopropyl-␤-D-thiogalactopyranoside and harvested after 3 h of growth at 37°C. The cells were harvested, and the cell paste was frozen and stored at Ϫ80°C. The cells were thawed on ice and resuspended in 4 volumes of 50 mM MOPS, pH 7.5, 100 mM KCl, 1 mM EDTA, 1 mM dithiothreitol (CaM lysis buffer), and then homogenized using an Avestin EmulsiFlex-C5 homogenizer (Ottawa, Canada). The protein was purified as previously described (15), then frozen in aliquots on dry ice, and stored at Ϫ80°C. Electrospray ionization-mass spectrometry was performed on the purified proteins using a quadrupole time-offlight (Micromass, Manchester, UK) using the appropriate internal standard as previously described (15).
Oxyhemoglobin Assay-The initial rate of ⅐ NO synthesis was measured using the spectrophotometric oxyhemoglobin assay as previously described (15,16,32). The assays were performed at 25°C in a Spec-traMax 190 96-well UV-visible spectrophotometer (Molecular Devices, Sunnyvale, CA) using the Soft Max Pro software (Molecular Devices). Endothelial NOS, iNOS, and nNOS were assayed at 70, 28.5, and 30 nM, respectively, in a 100-l total well volume. Unless otherwise stated, the final concentrations of the assay constituents were 1 M FAD, 1 M FMN, 1% glycerol, and 5 M H 4 B (stock H 4 B was made in 40 mM dithiothreitol, 54 mM Tris-HCl, pH 7.5 at 25°C, 0.18 mg/ml BSA, and 11% glycerol). NADPH (66.6 M), catalase (66.6 units/ml), SOD (100 units/ml), and BSA (0.11 mg/ml) were present in the buffer along with the aforementioned reagents plus enzyme before the reaction was initiated. The NADPH was added to this solution to react with any Larginine in the enzyme solution from the dialysis buffer of the NOS. The reaction was initiated by adding the following to the final concentra- NADPH Oxidase Activity-The consumption of NADPH by NOS was monitored at 340 nm (⑀ ϭ Ϫ0.0152 OD/nmol) as previously described (15). Cytochrome c Reductase Activity-The NADPH-dependent reduction of cytochrome c was monitored at 550 nm (⑀ ϭ 0.0488 OD/nmol) as described previously (18). Molecular Modeling-Models for CaM-TnC chimeric structures were constructed using the InsightII-Homology package from Accelrys. 2 Basis set protein structures included CaM and TnC complexes with Ca 2ϩ and target peptides. Where possible, EF hands derived from TnC were assigned completely from TnC structures spatially aligned with CaM. For the chimeras in which EF hand 1 or 2 was derived from TnC, part of the TnC EF hand was assigned using a homologous region of CaM to allow the structures to anneal at the splice points. The resulting chimeric structures contained no bad bond angles or distances and had no unfavorable steric overlaps, and after relaxation, their energies compared favorably to the solved structures. Fig. 1 shows a sequence alignment including the sequences of CaM, TnC, and the chimeras used in this study. The sequences labeled CaM and Trop represent the human CaM and TnC sequences obtained from SWISSPROT; the sequences labeled NIW (CaM) and A2X (troponin C) are the corresponding sequences extracted from recent crystal structures and are included to indicate which residues are ordered in the crystals. The sequences labeled CaM 1TnC, CaM 2TnC, etc. represent the chimera in which the number refers to the CaM EF hand region replaced by the cognate sequence from TnC. The positions of the eight CaM helices are indicated by the dashed lines and numbers above the alignment; the positions of the Ca 2ϩbinding EF hand regions are indicated by asterisks below the alignment.

Sequence Comparison of CaM, TnC, and Chimeras-
The sequence region derived from TnC in CaM 1TnC extends from the N terminus to the midpoint of helix 3 and includes an additional N-terminal "helix 0 " not observed in the NIW (CaM) structure. CaM 2TnC includes the TnC region corresponding to the second half of helix 3 or helix 4 in the Ca 2ϩ replete structures, the linking region, and the initial portion of helix 5. This includes the three extra troponin residues in the central region. CaM 3TnC includes the troponin cognate for the end of helix 5 and extends up to the mid region of helix seven, whereas CaM 4TnC has troponin sequence elements corresponding to the region extending from this point to the C terminus. CaM 3,4TnC includes the troponin sequence elements of CaM 3TnC and CaM 4TnC. The exact positions of these splice points is less important than might be expected because the regions including the initial portions of helices 3, 5, and 7 are well conserved in sequence and structure between TnC and CaM. The important arbitrary decision was the inclusion of the troponin linker region in CaM 2TnC; alternatively, this could have been placed in CaM 3TnC or excluded altogether.
Protein Expression and Purification-The chimeric proteins expressed on their own yielded a very high amount of protein (from 5 to 80 mg/liter depending upon the mutation). Purified CaM-TnC chimeras were analyzed by gel electrophoresis and were judged to be over 95% homogeneous (Fig. 2). The chimeras remain as single bands when smaller amounts are run. The samples were analyzed in buffers containing EDTA because the chimeras undergo typical shifts toward lower apparent molecular weights when run in a Ca 2ϩ containing buffer (results not shown). Electrospray ionization-mass spectrometry was used as a complementary tool to further assess the homogeneity of CaM samples and rule out any modification. The observed mass for each of the proteins was within 1 mass unit of the theoretical molecular weight (results not shown).
The iNOS enzyme was co-expressed with CaM and each of the individual CaM-TnC chimeras. The yield of protein varied depending upon the chimeric protein. iNOS co-expressed with wild-type CaM produced the highest yields of purified iNOS enzyme (5 mg/liter). The yields of iNOS obtained when the enzyme was co-expressed with either CaM 1TnC or CaM 2TnC (1.4 mg/liter medium) were more than twice those obtained when the same enzyme was co-expressed with CaM 3TnC, CaM 4TnC, or CaM 3,4TnC (0.3 to 0.7 mg/L). Based on the amount of iNOS produced upon co-expression with each of the chimeras, a definitive pattern emerged indicating that the chimeras expressing the C-terminal half of CaM better protect the CaMbinding region of NOS. The visible spectra of the iNOS enzymes co-expressed with each of the CaM-TnC chimeras were typical of wild-type NOS (results not shown), indicating proportionate content of heme and flavin. All of the enzymes were capable of producing ⅐ NO based upon the oxyhemoglobin assay. cNOS Activation by CaM-TnC Chimeras-Specific residues in EF hands 1, 3, and 4 of CaM are important for nNOS activation. CaM-TnC chimeras containing TnC elements corresponding to any of these elements were only able to activate ⅐ NO production from nNOS by less than 20% ( Fig. 3 and Table  I). In contrast, the chimera containing TnC EF hand 2 was able to fully activate ⅐ NO production by nNOS. The order in the rate of ⅐ NO production by nNOS activated by the CaM-TnC chimeras was wild type ϭ CaM 2TnC Ͼ Ͼ CaM 3TnC ϭ CaM 4TnC Ͼ CaM 1TnC Ͼ CaM 3,4 TnC. None of the chimeras were able to activate ⅐ NO production by nNOS in the presence of 1 mM EDTA. Our results are consistent with previous investigations of nNOS activation by CaM-TnC chimeras measured by the citrulline formation end point assay (5) and the oxyhemoglobin ⅐ NO capture assay (6).
We present here the first report of the activation of eNOS by CaM-TnC chimera. ⅐ NO production rates from eNOS activated by CaM or CaM-TnC chimera were as follows: wild type Ͼ Ͼ CaM 2TnC Ͼ CaM 4TnC Ͼ CaM 3TnC Ͼ CaM 1TnC Ͼ CaM 3,4TnC. Although the order of activation for the eNOS was similar to that of the nNOS enzyme, it is notable that CaM 2TnC was not able to fully activate the eNOS enzyme. With the were all set to 100%. Under these conditions, the activities for nNOS bound to wild-type CaM were 40 min Ϫ1 for ⅐ NO synthesis, 148 min Ϫ1 for NADPH oxidation, and 1163 min Ϫ1 for cytochrome c reduction. The activities for eNOS bound to wild-type CaM were 10 min Ϫ1 for ⅐ NO synthesis, 31 min Ϫ1 for NADPH oxidation, and 70 min Ϫ1 for cytochrome c reduction. A, nNOS (30 nM) activation by CaM-TnC chimeras. B, eNOS (70 nM) activation by CaM-TnC chimeras. The bar values represent the means Ϯ standard deviation. Each experiment was performed in triplicate and repeated three or more times. exception of CaM 2TnC, all of the chimeras were able to activate ⅐ NO production by eNOS to a significantly greater extent than nNOS (Table I).
The activation of nNOS and eNOS by wild-type CaM resulted in a NADPH consumption to ⅐ NO production ratio of more than three instead of the theoretical ratio of 1.5. Similar results for eNOS were previously reported; the higher relative NADPH consumption was associated with the redox cycling of exogenous unbound flavins added to the reaction buffer of the assay (19). The rate of NADPH oxidation by nNOS activated by either CaM or the CaM-TnC chimeras shows the same order observed for ⅐ NO production, indicating that any redox cycling by free flavins requires reduction of free flavins by the NOS reductase unit. The ratio of NADPH oxidation to ⅐ NO synthesis was the same for nNOS activated by CaM and CaM 2TnC. These studies show that CaM and the chimeras act by stimulating the nNOS enzyme to pass electrons from the flavins to the heme in the oxygenase domain. Slightly higher ratios of NADPH oxidation to ⅐ NO synthesis for the CaM 3TnC and CaM 4TnC chimeras suggest a partial uncoupling of electron transfer from ⅐ NO synthesis.
NADPH oxidation by eNOS activated by either CaM or the CaM-TnC chimeras did not show the same order as observed for the production of ⅐ NO (Table I). Endothelial NOS activation by all of the chimeras showed higher ratios of NADPH consumption to ⅐ NO production. This suggests that eNOS may be more susceptible than nNOS to uncoupling of NADPH oxidation from the production of ⅐ NO when activated by mutant CaM proteins.
The cytochrome c reduction studies showed that all of the chimeras except CaM 1TnC activated electron transfer in the reductase domain of the nNOS enzyme. The reduction rates of cytochrome c were significantly higher than the corresponding rates of ⅐ NO synthesis for the enzymes activated by the CaM 3TnC and CaM 4TnC chimeras. The rate of cytochrome c reduction by nNOS activated by CaM 2TnC was the same as when activated by CaM. The EF hands 2, 3, and 4 of TnC are sufficient for the activation of electron transfer within the reductase domain of nNOS. In contrast, specific residues in EF hand 1 of CaM appear to be important to catalytic function in the reductase domain because its replacement by its TnC cognate results in little or no electron transfer to cytochrome c.
A different trend was found when investigating the role of the four CaM domains in the activation of electron transfer within the reductase domain of eNOS enzyme. The replacement of either domain 3 or 4 in CaM only slightly diminished the rate of reduction of cytochrome c. In contrast to the results with nNOS, the CaM 1TnC chimera still allowed production of about 60% maximal cytochrome c reduction by eNOS. In further contrast to the nNOS results, the CaM 2TnC chimera allowed only about 30% maximal cytochrome c reduction rate for eNOS. The lower rate of production of ⅐ NO by eNOS bound to CaM 2TnC in comparison with nNOS may be in part ex-plained by intramolecular processes in the reductase domain as exemplified by the diminished rate of cytochrome c reduction.
iNOS Activation by CaM-TnC Chimeras-Past investigations of iNOS activation by CaM have been limited by the strong binding between the enzyme and co-factor. Coupled with the sensitivity of iNOS to proteolysis during purification, this has effectively prevented the preparation of intact CaM-free iNOS. We have overcome this problem by separately co-expressing the human iNOS enzyme with each of the five CaM-TnC chimeras used in our investigation. As previously reported, the co-expression of iNOS with wild-type CaM results in a highly stable complex that shows only a small decrease in activity when treated with EDTA (Ref. 16 and Fig. 4). The co-expression of iNOS with the different CaM-TnC chimeras provided an opportunity to determine regions of CaM that are important for activation and tight binding of CaM. The replacement of CaM domains 1, 3, or 4 with their TnC cognates resulted in reproducibly higher rates of ⅐ NO production ranging from 110 to 123% ( Fig. 4 and Table II). In contrast, the replacement of domain 2 of CaM with its cognate resulted in a protein CaM 2TnC that only activated NOS to about 60% maximal velocity. The CaM 3,4TnC chimera activated iNOS to about 70% maximal activity. The addition of excess wild-type CaM to each of these co-expressed iNOS/CaM-TnC chimera preparations did not result in any significant change in the activity of the enzyme, indicating that the CaM-binding domains were saturated with the mutant CaM proteins (results not shown).
The addition of 250 M EDTA to chelate the Ca 2ϩ resulted in a significant decrease in ⅐ NO production by all of the chimeras (Fig. 4). This was especially true for the iNOS enzyme coexpressed with CaM 2TnC, CaM 3TnC, and CaM 3,4TnC. The subsequent addition of excess Ca 2ϩ reversed the effect of the calcium chelator. In contrast, the addition of EDTA to the iNOS co-expressed with CaM 1TnC and CaM 4TnC resulted in only a small decrease in enzyme activity that was comparable with the effect on iNOS co-expressed with wild-type CaM. The absence of a strong Ca 2ϩ -dependent activation by these two chimeras is consistent with two factors: 1) the first EF hand of TnC is known not to bind Ca 2ϩ and 2) the fourth lobe of TnC binds Ca 2ϩ with 1 order of magnitude higher affinity than the corresponding site in CaM (8).
These results show that domains 2 and 3 of CaM contain important element(s) required for the tight binding of CaM to the iNOS enzyme. The addition of excess CaM in the presence of EDTA resulted in over 90% maximal activity for all of the enzymes co-expressed with the CaM-TnC chimeras except for the CaM 3,4TnC chimera that only showed 50% activation (results not shown). The mutant CaM chimeras appear to be displaced by wild-type CaM when incubated in a buffer containing EDTA. These results show for the first time that there are specific elements in CaM necessary for the tight binding of CaM to iNOS.
A comparison of the ability of the CaM-TnC chimeras to  Fig. 3 for the details of the assay conditions. NAA, no apparent activity.

Calmodulin-Troponin C Activation of Nitric-oxide Synthases
activate NADPH oxidation showed similar trends to their activation of ⅐ NO synthesis by iNOS (Table II and Fig. 4). The co-expression of CaM, CaM 1TnC, or CaM 4TnC with iNOS resulted in a stoichiometry of about 1.5 NADPH oxidized per ⅐ NO formed, even in the presence of excess EDTA. These results indicate that in the assay system used in these studies the iNOS mechanism is more tightly coupled than that found in the cNOS enzymes. Because the increased levels of NADPH consumption have been associated with the possible redox cycling of exogenous unbound flavins added to the reaction buffer (19), the difference between iNOS and cNOS enzymes may be due to a higher affinity for the FMN in the iNOS enzymes or, more likely, to a similar affinity but a much higher rate of ⅐ NO production. The iNOS co-expressed with CaM 2TnC, CaM 3TnC, or CaM 3,4TnC also resulted in the proper stoichiometry, but the addition of EDTA led to a greater than 2-fold increase in the relative rate of NADPH oxidation. The CaM domains 2 and 3 apparently contain elements necessary to maintain the Ca 2ϩ /CaM-dependent coupling of the iNOS enzyme.
The cytochrome c assays were used to monitor electron transfer from the flavins to an exogenous electron acceptor. The iNOS enzymes co-expressed with the different chimeras all showed 100% or more maximal activity in the presence of both high and low levels of Ca 2ϩ . These results are consistent with our previous study showing electron transfer from the reductase domain of iNOS enzymes to an artificial electron acceptor is CaM independent (18). Notably, the rates of cytochrome c reduction relative to wild-type CaM are over 100% for iNOS enzymes co-expressed with CaM 1TnC, CaM 2TnC, and CaM 3TnC. These increased rates may be due to the replacement of residues in CaM that are not necessary for electron transfer but may normally affect the rates of electron transfer by acting as modulators of the interactions between FAD and FMN or different segments of the reductase domain. Fig. 5 shows an alignment of the sequence region connecting the oxygenase and reductase moieties of eNOS, nNOS, and iNOS with the target peptide from the Meador et al. (21,22) structural characterization of CaM. The 1, 8, 14 pattern of hydrophobic residues in a basic background is present in all sequences as noted previously by Bredt and Snyder (37). The alignment shown is structural in that it positions sequence elements that we expect to be constrained by the helical nature of the target in the CaM complex in structurally equivalent positions. It is clear from a comparison of the gene sequences that the initial tribasic motif of the CaM-binding site in eNOS is derived from the same ancestral sequence elements as the tribasic motifs in nNOS and iNOS. But notably, these motifs have been structurally displaced by the insertion/deletion of the isoleucine/glycine or isoleucine/proline elements in nNOS and eNOS, respectively. This displacement would be expected to affect both the interaction of CaM with the binding site and "back side" interactions of CaM with the oxygenase domain.
The conservation of the region in NOS enzymes immediately following the CaM-binding site and connecting it to the FMNbinding domain is very strong. It consists of the eight residues immediately following the conserved leucine and methionine residues present in the 14-and 15-positions; the final basic residue shown is the initial residue in the first ␤ strand of the FMN-binding domain. In contrast, the connection between the CaM-binding site and the oxygenase domain is not conserved and varies in length from 5 to 9 residues. Clearly, the connection between the CaM-binding site and the reductase domain exerts far greater selective pressure.  Fig. 4 for the details of the assay conditions.

FIG. 4. Ability of co-expressed CaM-TnC chimeras to activate iNOS.
The iNOS enzyme was co-expressed separately with each of the CaM-TnC chimeras. The ⅐ NO synthesis, cytochrome c assay, and NADPH oxidation rates were measured as described in Fig. 3 except that no exogenous CaM was added to the assays. Each assay was performed in the presence of either CaCl 2 (200 M) or EDTA (250 M) as indicated. The activities obtained for iNOS co-expressed with wild-type CaM and assayed in the presence of CaCl 2 (200 M) at 25°C were all set to 100% and were 47 min Ϫ1 for ⅐ NO synthesis, 77 min Ϫ1 for NADPH oxidation, and 1395 min Ϫ1 for cytochrome c reduction. The bar values represent the means Ϯ standard deviation. Each experiment was performed in triplicate and repeated three or more times.

DISCUSSION
Structural studies of CaM and peptides derived from myosin light chain kinases and CaM-dependent kinase have provided a general mechanism of activation of CaM-regulated proteins (20 -22). The structure of the CaM-bound myosin light chain kinase peptide shows the two Ca 2ϩ -binding lobes of CaM engulfing the target peptide. In some cases, activation resulted from the CaM-dependent displacement of an autoinhibitory domain. A number of recent structural investigations on the plasma membrane Ca 2ϩ -pump (23), the Ca 2ϩ -activated K ϩchannel (24), Ca 2ϩ /CaM-dependent kinase kinase (25), and the anthrax adenylyl cyclase exotoxin (26) have revealed new ways for CaM to interact with its target proteins. The novel aspects of these interactions between CaM and its target proteins are summarized in a recent review (27). They include: 1) the binding of the Ca 2ϩ pump to only the C-terminal domain of CaM; 2) dimer formation with a segment of the K ϩ -channel; and 3) two anthrax domains wrapped around an extended CaM protein.
The binding of CaM to the cNOS enzymes has been shown to affect electron transfer at two points, whereas the effect on iNOS is less understood. The present investigation of CaM interactions with all three NOS isoforms was undertaken to further our understanding of the activation of these enzymes.
CaM-TnC chimera proteins which replace CaM regions ("domains") containing EF hands 1, 3, and 4 with their TnC cognates have significant effects on the rates of ⅐ NO production and NADPH oxidation of nNOS enzymes ( Fig. 3 and Table I).
As indicated by the cytochrome c reduction results, corresponding regions of TnC can replace domains 3 and 4 with little effect on electron transfer from NADPH into the reductase domain of nNOS. This indicates that the impaired ability of the CaM 3TnC and CaM 4TnC chimeras to activate ⅐ NO production is specifically associated with the reduction of heme by FMN. In contrast, the combined replacement of the second domain of CaM and the linker region by the corresponding domain in TnC results in a chimera (CaM 2TnC) that can fully activate nNOS. This is a remarkable result considering the much greater change in sequence produced in the CaM 2TnC construct and suggests that the less conserved, external residues in CaM 3TnC and CaM 4TnC (and possibly also CaM 1TnC) make important contacts with sequence regions outside the canonical CaM-binding region.
A similar investigation using the eNOS enzyme revealed a similar overall pattern but slightly different details. When compared with the nNOS results, the transfer of electrons into the reductase domain of eNOS as monitored by the reduction of the exogenous cytochrome c was less susceptible to the mutation of CaM. Furthermore, although the CaM 2TnC chimera was again the most potent activator of the five chimeras tested, it only activated eNOS to 50% of the maximal activity obtained with wild-type CaM. In addition, the activation of eNOS by CaM 3TnC and CaM 4TnC was significantly greater as a percentage of maximal activity with wild-type CaM.
In parallel with the nNOS results, CaM 3TnC, CaM 4TnC, and CaM 3,4TnC all activated electron transfer to cytochrome c in eNOS to a greater extent than they activated ⅐ NO production. This again indicates that the less effective activation of ⅐ NO production by these chimera proteins results specifically from a lower rate of FMN to heme electron transfer. In contrast, CaM 2TnC is less effective in the activation of cytochrome c reduction in eNOS than in the activation of ⅐ NO production. This suggests that the inability of CaM 2TnC to fully activate ⅐ NO production in eNOS could result from ratelimiting electron transfer to FMN.
The investigation of the role of different regions of CaM on the activation of iNOS required the development of separate co-expression systems for each of the CaM-TnC chimera proteins. Our studies using the proteins purified from these systems showed a significant difference in the role of CaM in the activation of iNOS when compared with the cNOS enzymes ( Fig. 4 and Table II). Co-expression with each of the five CaM-TnC chimeras in the presence or absence of a Ca 2ϩ chelator did not significantly affect electron transfer into the reductase domain as monitored using the cytochrome c assay. These results are consistent with our previous studies using only the reductase domains of both human and mouse iNOS (18).
In contrast to nNOS, CaM 2TnC activated ⅐ NO production by iNOS was over 40% less than that induced by wild-type CaM under Ca 2ϩ replete conditions. Although CaM 2TnC activates ⅐ NO production to the same extent in iNOS and eNOS when expressed as a fraction of maximum activity with wild-type CaM, it is more informative to point out that in the cNOS isoforms CaM 2TnC is the most potently activating chimera, whereas in iNOS it is the least effective. CaM 1TnC is significantly more effective than wild-type CaM in activating ⅐ NO production by iNOS, and CaM 3TnC and CaM 4TnC may also be slightly more effective; CaM 3,4TnC is more effective than CaM 2TnC but less effective than wild-type CaM. Clearly, the results for CaM 4TnC indicate that the C-terminal domain of CaM is less important for iNOS than for eNOS and nNOS, but because the TnC cognate of EF hand 4 is functional and well conserved, this region is likely to be of functional importance. CaM 3,4TnC is generally less effective in all respects than either CaM 3TnC or CaM 4TnC, even though CaM 3TnC and CaM 4TnC are as potent activators of iNOS as wild-type CaM when investigated in the presence of excess Ca 2ϩ . There are several possible explanations for this result. The potency of CaM 3TnC and CaM 4TnC when bound to Ca 2ϩ may be the result of small adjustments in the conformation of the chimera; CaM 3,4TnC might require adjustments in the 3 and 4 domains, which are mutually exclusive. Viewed in this light, these results indicate that the CaM requirement of iNOS is less stringent than the CaM requirement of eNOS and nNOS and that the most important interactions governing eNOS and nNOS activation by CaM are absent in iNOS.
The combined replacement of elements in the second domain and the linker region of CaM (CaM 2TnC) or the replacement of part of the third domain (CaM 3TnC) also affect the Ca 2ϩ dependence of CaM-iNOS interactions. The addition of excess EDTA to the assays resulted in a significant decrease in ⅐ NO production by iNOS co-expressed with CaM 2TnC, CaM 3TnC, or CaM 3,4TnC. This Ca 2ϩ -dependent decrease in ⅐ NO production correlated with a decrease in NADPH oxidation but not cytochrome c reduction, which was well supported by all chimera proteins tested. In the presence of EDTA, CaM 2TnC, CaM 3TnC, and CaM 3,4TnC all activated NADPH consumption to a greater extent than ⅐ NO production, indicating that at low Ca 2ϩ concentrations binding of these chimeras to iNOS produced uncoupled electron transfer. Because CaM 1TnC contains a troponin-de- FIG. 5. Alignment of CaM binding sequences. The CaM-binding sequences from iNOS, eNOS, nNOS, and the target peptide 3 in the original CaM structure (22) were aligned. The conserved amino acids in the 1-8-14 motif are in bold and numbered above the alignment. The initial tribasic motif is shown displaced in nNOS and iNOS with respect to eNOS to represent its position in the structure, although these elements were derived from a common ancestor. The final basic residue in the NOS sequences corresponds to the initial residue in the first strand of the FMN-binding domain.
rived EF hand that is incapable of Ca 2ϩ binding and that is unlikely to be capable of stabilizing interactions to the extent of the CaM cognate it replaces, the effectiveness of CaM 1TnC is remarkable. It suggests that the initial EF hand of CaM and the associated flanking regions are relatively unimportant in iNOS activation. On the other hand, the CaM 2TnC and CaM 3TnC results demonstrate that these regions are important determinants of iNOS Ca 2ϩ -independent activity.
All of the single domain replacement chimeras were at least as effective as wild-type CaM in activating electron transfer to cytochrome c; CaM 1TnC, CaM 2TnC, and CaM 3TnC were significantly more effective than wild-type CaM. Taken together, the results for iNOS, eNOS, and nNOS indicate that, in general, the requirements for the activation of cytochrome c reduction are less stringent than the requirements for the activation of ⅐ NO production. The iNOS results suggest that cytochrome c reduction can benefit from a loosening of the conformational constraints imposed by CaM binding. Specific isoform requirements are reflected in the details shown in Tables I and II; notably, CaM 1TnC is ineffective in activating cytochrome c reduction by nNOS, and CaM 2TnC is less effective than the other chimeras at activating cytochrome c reduction in eNOS. Our results for both cNOS enzymes support the previous studies showing that CaM can activate reductase domain catalysis by a mechanism other than stimulating electron transfer into the oxygenase domain (6). Importantly, we also show that although a similar result is found for the eNOS enzyme, effects of the chimeras on all the reactions are not equivalent; notable differences still exist between nNOS and eNOS (see results with CaM 2TnC in Table I and above). Significant differences between cNOS enzymes have also been reported for the effect of CaM oxidation on enzyme activity and electron transfer (15).
The differences in the structural composition of all five CaM-TnC chimeras used in our investigation are summarized in Table III. Although limited by the fact that more than one region of NOS enzymes interact with CaM (41), the recent crystal structure of CaM bound to the CaM-binding peptide of eNOS provides useful information for the interpretation of our results (28). The differences between the CaM 1TnC chimera and CaM include two variant amino acid contacts (L18I,E14A), the replacement of two N-terminal residues in CaM ( 1 AD 2 ) by the 9 residues in TnC ( 1 MDDIYKAVE 9 ), and the loss of the D24 Ca 2ϩ ligand, with the result that the first EF hand of TnC does not bind Ca 2ϩ . Any combination of these three changes could account for the diminished activity of ⅐ NO production by the cNOS enzymes. The lack of a significant effect on iNOS activity by the co-expression of CaM 1TnC is consistent with the fact that iNOS is fully active at basal levels of Ca 2ϩ when the N-terminal of CaM may not be Ca 2ϩ replete.
A similar comparison for the CaM 2TnC chimera shows only a single variation (K75C) in the residues in contact with the eNOS peptide. Of possibly greater importance is the elongation of the central helix in CaM 2TnC by 3 additional amino acids ( 91 KGK 93 ) not found in CaM. It has been proposed that the flexibility of the linker region of CaM allows the protein to adjust the orientation of the two globular domains with respect to each other and to the target peptide (30,31). The three additional residues in the linker region of the CaM 2TnC chimera elongate the central helix by about 4.5 angstroms and would affect the relative position of contact residues in the N and C globular domains of CaM unless compensated by conformational adjustments as shown in Fig. 7. These adjustments are feasible, but because they are expected to require free energy, they will inevitably be reflected in the binding equilibrium. The effects of CaM 2TnC on the Ca 2ϩ /CaM dependence of ⅐ NO production by iNOS could be due to a change in the interaction of the linker region of CaM with the enzyme or subtle changes in the relative positions of the globular domains of CaM when bound to iNOS.
The CaM 3TnC chimera has five variant residue contacts with the eNOS peptide when compared with CaM (Table III). Three of these substitutions (A88L, M109L, and L112T) reduce the exposed hydrophobic surface of this region of TnC and have been proposed to be a major cause for the inability of TnC to bind to certain CaM-binding proteins (29). These residues may be important in the activation of cNOS enzymes as well as the Ca 2ϩ /CaM dependence of ⅐ NO production by iNOS bound to CaM 3TnC. A previous study using CaM-TnC chimeras and mutant proteins indicated that the inability of the CaM 3TnC chimera to activate nNOS is attributable largely on the substitution of helix 6 and especially to residue Leu 122 in CaM (5).
The fourth domain of TnC binds Ca 2ϩ with 1 order of magnitude more affinity than the corresponding site in CaM (8). The CaM 4TnC chimera has 5 variant residues, but 2 of them are conservative (Table III). The relative ⅐ NO production by nNOS is less than half that obtained by eNOS when activated by CaM 4TnC (Fig. 3 and Table I). Different effects on the relative activity of these two cNOS enzymes were previously observed when the methionines in domain 4 of CaM were chemically oxidized (15). In the case of CaM 4TnC, the CaM Met 145 is replaced by a lysine, and the crystal structure indicates that this residue is in contact with the eNOS peptide (28). In contrast to the results with the cNOS enzymes, the activa-tion profile for iNOS when co-expressed with CaM 4TnC is identical to the wild-type CaM. The differences in the binding and activation of the three NOS isoforms by CaM are apparently due to elements in different regions of the CaM protein. The CaM-eNOS structure is shown in purple in each case, the TnC structure is in red, and the chimera is shown in yellow. The region shown in yellow ribbon represents the sequence derived from TnC in each case. Bound Ca 2ϩ atoms are rendered in CPK mode. A, CaM 1TnC structure. The red ribbon represents the trace of the TnC sequence region that was assigned using CaM coordinates because it was incompatible with the rest of the structure (note overlap with white target peptide). This results in the redirection of the final helix coming out of the first EF hand loop so that it points straight down the right-hand side of the panel. The structure in the EF hand 1 region is feasible but is unlikely to contribute effectively to the stability of the complex or to replace CaM EF hand 1 effectively. No Ca 2ϩ binding is possible. B, CaM 2TnC structure. The EF hand and most of the sequence derived from TnC has been assigned from the TnC structure, but the final helix has been redirected to connect to the CaM derived EF hands 3 and 4. Note the deviation of the TnC backbone in red ribbon and the path of the CaM backbone in purple ribbon. Only the end of the linking structure can be CaM-like because of the insertion of three extra TnC-derived residues. This structure is feasible but deviated significantly from wild-type CaM. TnC EF hand 2 did not bind Ca 2ϩ in the crystal structure, but all the ligands are present, and Ca 2ϩ may be bound in the chimera. C, CaM 3TnC; the structure can be readily derived from TnC coordinates alone and annealed into the structure of CaM. The effects are expected to be primarily the result of the replacement of specific residues. D, CaM 4TnC; the structure can be readily derived from TnC coordinates alone and annealed into the structure of CaM. The effects are expected to be primarily the result of the replacement of specific residues. Fig. 6 (A-C) shows the structures of a series of molecular models of CaM-TnC chimeras superimposed on the structures of CaM and TnC from which they were derived. In each case, the structure of the chimera (yellow ␣ trace) follows the structure of CaM for most of the sequence, whereas most of the structure of the TnC derived EF hand is taken from a TnC crystal structure. These regions are emphasized by the yellow ribbon. For the structure of CaM 2TnC, the final helix was derived primarily from CaM.
The structure of CaM bound to a target peptide derived from the eNOS CaM-binding sequence, as reported by Aoyagi et al. (28), is represented by the purple trace; this structure is in most respects similar to other structures representing "classical" CaM binding modes. This suggests strongly that in line with previous CD and NMR work (38 -40), the canonical CaM-binding sites of the cNOS isoforms are able to interact with CaM in situ in the classic central single helix mode, with binding determinants represented by the 1-8-14,15 placement of hydrophobic residues and the presence of positively charged residues.
CaM consists of four repeating elements, each composed of a single EF hand and associated flanking regions. All four EF hands engage in Ca 2ϩ binding and interact with the peptide target. The strongest interactions are with helices 1, 2, 5, and 6; the ends of helices 2 and 6 are slightly farther apart in the complex with the nNOS target peptide than in some structures with other targets, but this need not represent the situation in situ, where interactions with other regions of NOS are known to be important (41). The orientation of the central helix, in line with other classical CaM-binding structures, places EF hand 4 closest to the oxygenase domain and EF hand 2 closest to the FMN-binding domain and the adjacent region of the subdomain that interrupts the sequence of the FAD-binding domain. EF hands 1 and 3 are placed so that these structures and the adjacent sequence regions could interact with structures extending from either the oxygenase or the reductase end.
The structure of the troponin C complex with a target peptide is represented by a red trace. In this structure, only EF hands 3 and 4 interact with the target, and only these regions bind Ca 2ϩ . EF hand 1 of TnC is incapable of binding Ca 2ϩ with any reasonable affinity because it lacks a cognate for aspartic acid 24 in CaM; cognates of the residues involved in Ca 2ϩ binding are present in the other three EF hands.
Although the first EF hand of the CaM 1TnC structure lacks the capacity to bind Ca 2ϩ , it could readily form a rough approximation of the CaM cognate. It is reasonable to expect that such a structure, homologous to CaM but lacking a key residue, would produce a CaM-like structure when inserted into the CaM context, but that this structure would lack precision and stability. The model shown in Fig. 6A was produced using the coordinates for TnC from the available crystal structure from the N-terminal until residue 34; the troponin C sequence elements between residues 35 and 56 were assigned using available CaM structures, because otherwise the backbone position would interfere with the position of the target peptide. This is reasonable because the sequence regions involved are 65% identical and differ structurally only in the direction of the terminal helix, which in CaM is imposed by the bound Ca 2ϩ and the position of the other EF hands. The resulting structure had no unfavorable steric interactions internally or with its target peptide and could be readily relaxed to an internal free energy of approximately Ϫ1000 kcal/mol (equivalent to the gas state e.g. without intermolecular terms).
The model structure of the CaM 2TnC complex is shown in Fig. 6B. In this case residues 57-87, including all of EF hand 2, were assigned from the TnC structure; residues 88 -97, corre-sponding to the start of helix five, were assigned using a CaM structure to align the structure with the positions of EF hands 3 and 4 on the target peptide. This altered the structure of three residues in the flexible connecting region between helices four and five. The resulting structure again had no unfavorable steric interactions internally or with its target peptide and could be readily relaxed to a gas state free energy of approximately Ϫ1000 kcal/mol.
A side view of this model for CaM 2TnC, shown in Fig. 7, provides a better picture of the conformational adjustments in the linker region needed to position all four EF hands on the target peptide. The red ribbon indicates the path of the TnC backbone after it forms EF hand 2; the purple ribbon shows the path of the CaM backbone. The extra residues force a bulge in the turn formed by the linker.
The CaM 3TnC, CaM 4TnC, and CaM 3,4TnC model structures were generated using the coordinates of the troponin structure for the troponin derived sequence regions. These could be readily incorporated into the CaM structure with minimal relaxation of the splice points marking the transition between TnC-and CaM-derived regions. None of the resulting structures had unfavorable steric interactions internally or with the target peptides, and each could be readily relaxed to a gas state free energy of approximately Ϫ1000 kcal/mol.
The vast majority of residues in contact with the internal target helix are conserved or conservatively substituted; notable exceptions are E14A in CaM 1TnC, K75C in CaM 2TnC, L112T in CaM 3TnC, and M145K in CaM 4TnC. Clearly, if each of the four repeats were equally important in the binding of CaM to NOS and the resulting activation of ⅐ NO production, we should expect the substitution of EF hands 1 and 2 from TnC to be more important than the substitution of EF hands 3 and 4, because the structures of the cognate regions in TnC are more divergent, and their function in Ca 2ϩ and target peptide binding is diminished or lost. The results described here do not follow this simple pattern and instead reveal isoform-specific differences in the interaction of CaM regions with NOS.
Regarding the NOS elements involved in activation, it is clear that interactions with the conserved canonical CaM-binding site alone are insufficient to account for the observations presented here. The outside (N-terminal with respect to strands) edge of the FMN-binding domain and the hinge subdomain that positions the FMN-and FAD-binding domains are in contact or proximity with the outside of bound CaM; these regions contain cNOSspecific insertions that are associated with CaM control and are well positioned to interact directly with the CaM exterior (33). The FMN-binding domain insertion is ϳ42 residues long and has autoinhibitory character (17,33,34). Removal of the small insertion in the hinge subdomain (35) has been reported to increase the activity of eNOS to a lesser degree, which was taken as evidence for autoinhibitory character (36). Although it is unlikely that every mutant that disrupts cNOS control reveals an additional "autoinhibitory element, " it is likely that these adjacent sequence regions work together as part of the CaM control mechanism. 3 The C-terminal tail of NOS, which has also been implicated in maintaining control of electron transfer (35), is not positioned to interact directly with CaM and is likely to be less important here.
Additional adjacent structures include less well defined areas associated with the oxygenase domain. We have already pointed out that the connector linking the CaM-binding site with the oxygenase domain is not well conserved in NOS isoforms, indicating that the interactions between bound CaM and the oxygenase domain within a monomer may not be preserved. However, electron transfer from FMN to heme in a NOS dimer is believed to involve the oxygenase domain of the other monomer. Direct interaction of bound CaM with residues adjacent to the face of the oxygenase domain through which electron transfer occurs is potentially important.
We recently proposed that electron transfer in NOS occurs by a tethered shuttle mechanism in which the FMN-binding domain releases from the FAD domain, to which it remains covalently tethered by a connecting polypeptide linker and rebinds to the oxygenase unit for electron transfer to heme (11). CaM control in such a model involves enabling of the shuttle, probably through the release of the FMN domain. Viewed from this perspective, the differences between activation of cytochrome c reduction and ⅐ NO production can be readily rationalized. Release of the FMN-binding domain is necessary for the activation of cytochrome c, but ⅐ NO production also requires efficient rebinding of the FMN domain to the oxygenase domain. Chimeras that are able to efficiently promote release of the FMN domain are not necessarily competent to promote reassociation of the FMN and oxygenase domains and may even inhibit the association of the FMN and oxygenase units.
Cytochrome c reduction by iNOS is readily activated by each of the chimeras examined here and may be constitutive. The disparity between cytochrome c reduction and ⅐ NO production at low Ca 2ϩ can be attributed to poor association of heme and FMN domains when the bound CaM constructs are depleted of Ca 2ϩ . Interactions with the reductase moiety are dominant in cNOS activation, and EF hand 1 is critical for activation of both nNOS and eNOS. In general, cNOS enzymes are much more difficult to activate than iNOS, which can be attributed to the extra sequence elements they possess, which are adjacent to the CaM-binding region and associated with CaM control.