Domain Swapping in Inducible Nitric-oxide Synthase

Cytokine-inducible nitric-oxide (NO) synthase (iNOS) contains an oxygenase domain that binds heme, tetrahydrobiopterin, and l-arginine, and a reductase domain that binds FAD, FMN, calmodulin, and NADPH. Dimerization of two oxygenase domains allows electrons to transfer from the flavins to the heme irons, which enables O2 binding and NO synthesis froml-arginine. In an iNOS heterodimer comprised of one full-length subunit and an oxygenase domain partner, the single reductase domain transfers electrons to only one of two hemes (Siddhanta, U., Wu, C., Abu-Soud, H. M., Zhang, J., Ghosh, D. K., and Stuehr, D. J. (1996) J. Biol. Chem. 271, 7309–7312). Here, we characterize a pair of heterodimers that contain an l-Arg binding mutation (E371A) in either the full-length or oxygenase domain subunit to identify which heme iron becomes reduced. The E371A mutation prevented l-Arg binding to one oxygenase domain in each heterodimer but did not affect thel-Arg affinity of its oxygenase domain partner and did not prevent heme iron reduction in any case. The mutation prevented NO synthesis when it was located in the oxygenase domain of the adjacent subunit but had no effect when in the oxygenase domain in the same subunit as the reductase domain. Resonance Raman characterization of the heme-l-Arg interaction confirmed that E371A only prevents l-Arg binding in the mutated oxygenase domain. Thus, flavin-to-heme electron transfer proceeds exclusively between adjacent subunits in the heterodimer. This implies that domain swapping occurs in an iNOS dimer to properly align reductase and oxygenase domains for NO synthesis.

Nitric oxide (NO) 1 acts as a signal and cytotoxic molecule in biology (1)(2)(3) and is synthesized from L-arginine (L-Arg) by enzymes termed NO synthases (NOS). The NOS exhibit a bidomain structure in which a N-terminal oxygenase domain that contains binding sites for iron protoporphyrin IX (heme), tetrahydrobiopterin (H 4 B), and L-Arg is fused to a C-terminal reductase domain that contains binding sites for calmodulin (CaM), FMN, FAD, and NADPH (4,5). To synthesize NO, NADPH-derived electrons must transfer from the reductase domain flavins to the oxygenase domain heme irons, which are bound to the protein via cysteine thiolate axial ligation as in the cytochromes P450 (6 -10). The flavin-to-heme electron transfer is thought to be critical for catalysis because it enables each heme iron to bind and activate oxygen at two steps in the reaction sequence, resulting in oxygen insertion into L-Arg to form N -hydroxy-L-Arg, and subsequent oxygenation to generate NO and citrulline as products (11)(12)(13).
The NOS are only active as homodimers (4,14), and understanding how dimerization relates to NOS catalysis is a topic of current interest. Studies with the cytokine-inducible NOS (iNOS) indicate its dimer assembly occurs with stable incorporation of one heme and one H 4 B into each subunit (15,16). The dimeric interaction only requires the oxygenase domains of each subunit with the reductase domains apparently not interacting with one another (17). The NOS oxygenase and reductase domains can be expressed separately and fold and function independently of one another (17)(18)(19)(20)(21). This has facilitated spectroscopic (17,20), mutagenic (22)(23)(24), and crystallographic (25,26) characterization of the iNOS oxygenase domain (iNOSox).
Both full-length iNOS and iNOSox dimers can reversibly dissociate into folded monomers in the presence of urea (27,28). Although the full-length iNOS monomer still binds its flavin and heme groups and can transfer electrons from NADPH into its flavins, it can no longer catalyze reduction of its heme iron (29), indicating that dimeric structure in some way is critical for the flavin-to-heme electron transfer step. 2 To investigate how iNOS dimerization, electron transfer, and catalytic function are related, we characterized an iNOS "heterodimer" comprised of one full-length and one iNOSox subunit. The single reductase domain of this "wild-type" heterodimer transferred NADPH-derived electrons to only one of the two heme irons located in the dimeric oxygenase core, but OXY WT , a heterodimer comprised of a full-length iNOS subunit containing the E371A mutation and a wild-type oxygenase domain subunit; FL WT /OXY E371A , a heterodimer comprised of a full-length wild-type iNOS subunit and an oxygenase domain subunit containing the E371A mutation; FL WT /OXY WT , a heterodimer comprised of wild-type fulllength and oxygenase domain subunits. 2 Dimeric structure does not influence the electron transfer properties of the NOS reductase domain, which can reduce electron acceptors like cytochrome c with equal activity whether present in a NOS dimer or monomer (26). this was sufficient to support a normal rate of NO synthesis by that heme (29). Although these results showed that dimerization enables the flavin-to-heme electron transfer, they did not identify which of the two hemes accepts electrons from the single reductase domain and thus did not distinguish whether electron transfer occurs between flavin and heme groups located in the same or adjacent subunits of a dimer.
In the current report, we address this question by constructing a complimentary pair of iNOS heterodimers, each comprised of one wild-type subunit and one subunit containing a point mutation (E371A) that causes an exclusive and absolute defect in L-Arg binding (22). The catalytic, spectroscopic, and electron transfer properties of these complimentary heterodimers and of a heterodimer selectively enriched in 54 Fe establishes the electron transfer pathway in iNOS, and helps explain why dimer formation is essential to complete electron transfer to the heme.
Generation of Full-length E371A iNOS-A cDNA fragment containing the mutation E371A was removed from a pCWori plasmid containing E371A iNOSox (22) by restricting with BlnI (AvrII) and Eco47III (Boehringer Mannheim). This fragment was ligated into similarly restricted pCWori containing full-length iNOS (31). The ligation product was transformed into competent Escherichia coli JM109, and the miniprep DNA prepared from a single colony was sequenced in the core facility of the Cleveland Clinic to confirm the mutation was transferred. The pCWori plasmid containing E371A full-length iNOS and a plasmid containing human CaM (31) were sequentially transformed into competent E. coli BL21 (DE3) for protein expression.
Purification of Wild-type iNOSox and E371A iNOSox-E. coli (BL21) expressing C-terminal His 6 -tagged iNOSox (amino acids 1-498) (23) or the E371A iNOSox (22) were grown in Terrific broth containing 100 g/ml ampicillin at 37°C with shaking at 250 rpm. The cultures were induced at A 600 ϭ 0.8 -1.2 with 1 mM isopropyl-␤-D-thiogalactopyranoside plus 400 M ␦-aminolevulinic acid, and continuously shaken at 25°C for another 36 h. Cells were harvested by centrifugation at 5000 rpm at 4°C, and resuspended in one-fifth of the culture volume in lysis buffer (40 mM EPPS, pH 7.6, containing 1 mM phenylmethylsulfonyl fluoride, 100 g/ml aprotinin, 1 g/ml leupeptin, 4 M H 4 B and 1 mM L-Arg). The cells were lysed by freeze-thawing, followed by sonication on ice. Cell debris was removed by centrifugation at 10,000 rpm for 30 min at 4°C. Proteins were purified as described previously (22,23) using Ni 2ϩ -chelate affinity resin equilibrated with 40 mM EPPS, pH 7.6, containing 0.25 M NaCl, 4 M H 4 B, and 1 mM L-Arg. After initial adsorption the column was washed with equilibration buffer followed by the same buffer containing 60 mM imidazole. The protein was eluted with 150 mM imidazole. The eluted protein was concentrated using Centricon 30 concentrators and dialyzed against 40 mM EPPS, pH 7.6, supplemented with 10% glycerol and 0.2 mM DTT.
Preparation of 54 Fe-enriched iNOSox-iNOSox containing heme enriched in 54 Fe was prepared essentially as detailed in Ref. 32 by growing E. coli expressing this protein in semisynthetic medium (33). The growth medium contained 54 Fe metal that had been dissolved in acid, along with a small amount of yeast extract (0.06% w/v) and ampicillin. The cells were grown at 37°C in this media to an A 600 of 0.6 -1.2, induced with isopropyl-␤-D-thiogalactopyranoside and ␦-aminolevulinic acid as described above, and then harvested after an additional 14 -18-h growth period at 30°C. Purification was done as for wild-type iNOSox.
Expression and Purification of Wild-type and E371A Full-length iNOS-Expression of N-terminal His 6 -tagged full-length iNOS and the E371A mutant in E. coli, cell harvest, and lysis were identical to that described for the oxygenase domains. Proteins were purified as described previously (31). The lysates were subjected to Ni 2ϩ -chelate affinity chromatography as described above, except the 60 mM imidazole wash was eliminated, and the protein was eluted with 175 mM imidazole. The eluted protein was concentrated and loaded onto a 2Ј,5Ј-ADP-Sepharose column equilibrated with 40 mM EPPS, pH 7.5, containing 0.25 M NaCl, 4 M FAD, 4 M FMN, 1 mM DTT, 4 M H 4 B, and 1 mM L-Arg. After adsorption the column was washed with the equilibration buffer followed by a 10 mM NADP ϩ wash. The protein was then eluted with 10 mM NADPH. Selected fractions were pooled, con-centrated, and dialyzed against 40 mM EPPS, pH 7.6, containing 10% glycerol and 0.2 mM DTT.
Preparation and Purification of iNOS Heterodimers-Heterodimers were constructed and purified as described previously (29). Briefly, the iNOSox and full-length iNOS dimers were dissociated into monomers in the presence of 5 M and 2 M urea, respectively. Ni 2ϩ -affinity resin was then saturated with the oxygenase domain monomer, and the resin was washed to remove unbound protein. The full-length iNOS monomer was added and dimerization was promoted by addition of 10 M H 4 B and 5 mM L-Arg at 25°C. After 60 min of gentle shaking, the resin was washed with 40 mM EPPS, pH 7.6, containing 0.25 M NaCl, 4 M H 4 B, and 1 mM L-Arg, and then the bound protein was eluted with the same buffer containing 125 mM imidazole. The eluate was concentrated and loaded onto a 2Ј,5Ј-ADP-Sepharose column. Contaminating iNOSox dimer was removed by washing with ADP column buffer, and the heterodimer was eluted with 10 mM NADPH.
Gel Filtration Chromatography-A Superdex 200 HR column (Amersham Pharmacia Biotech) of ϳ25 ml bed volume was used to size fractionate heterodimer preparations as noted in the text. The column was equilibrated with 40 mM EPPS, pH 7.6, containing 10% glycerol, 0.2 M NaCl, and 4 M H 4 B. The flow rate was maintained at 0.5 ml/min using a Pharmacia FPLC. Protein in the eluate was detected using a flow-through detector set at 280 nm.
NO Synthesis Activity-The initial rate of NO synthesis by fulllength iNOS and the heterodimers was quantitated at 37°C using the oxyhemoglobin assay for NO (34). The iNOS (10 -50 nM with respect to heme iron concentration) was added to a cuvette containing 40 mM EPPS, pH 7.6, supplemented with 10% glycerol, 0.3 mM DTT, 5 mM L-Arg, 4 M each of FAD, FMN, and H 4 B, 100 units/ml catalase, 10 units/ml superoxide dismutase, 0.1 mg/ml bovine serum albumin, and 10 M oxyhemoglobin to give a final volume of 0.3 ml. The reaction was started by adding NADPH to give 0.1 mM. The NO-mediated conversion of oxyhemoglobin to methemoglobin was monitored over time as an absorbance increase at 401 nm and quantitated using the extinction coefficient of 38 mM Ϫ1 cm Ϫ1 .
NADPH Oxidation-The initial rate of NADPH oxidation at 25°C was quantitated spectrophotometrically at 340 nm using an extinction coefficient of 6.2 mM Ϫ1 cm Ϫ1 . The composition of the assay mixture was similar to that of the NO synthesis measurement except that oxyhemoglobin was not added. In some cases, L-Arg was omitted or replaced by 1 mM thiocitrulline.
Calculation of Specific Activity-The nanomoles of NO produced (or NADPH oxidized) per min per nmol of heme iron was calculated based on the concentration of heme iron, determined from the absorbance difference between 444 and 500 nm for the ferrous-CO species and using an extinction coefficient of 74 mM Ϫ1 cm Ϫ1 for ferrous CO-bound iNOS (34).
Resonance Raman Experiments-The instrumentation for resonance Raman experiments has been described in detail previously (35). Excitation was provided by a He-Cd laser at 441.6 nm with a power at the sample of 1 milliwatt. For all experiments, 40 mM BisTris buffer, pH 7.6, containing 1 mM DTT was used. Heme concentration in the sample was between 20 and 50 M. Ferrous-CO iNOS samples were prepared by adding a small amount of dithionite solution to the anaerobic sample cell under a CO atmosphere, and conversion to the CO-bound form was confirmed by UV-visible spectroscopy. In some cases, an anaerobic solution of L-Arg was added to give a final concentration of 3 mM.

RESULTS
Characterization of Full-length E371A iNOS-UV-visible spectra of purified E371A iNOS are shown in Fig. 1. The broad Soret absorbance centered at 405 nm in the absence of H 4 B and L-Arg indicates that the ferric heme iron is a mixture of low and high spin states, as observed for wild-type full-length iNOS purified under identical conditions (36). The absorbance shoulders at 450 -485 nm indicate the presence of bound flavins and are unusually prominent with respect to the Soret absorbance. Addition of L-Arg alone failed to bring about any change in the spectrum, whereas addition of H 4 B alone converted the heme iron to a mostly high spin state. Addition of L-Arg to the pterin-bound protein did not change the spectrum (not shown). Chemical reduction by dithionite in the presence of CO enabled formation of the thiol-ligated ferrous-CO complex absorbing at 444 nm. Thus, the E371A mutation rendered full-length iNOS incapable of binding L-Arg but did not appear to affect its heme ligation or interaction with H 4 B, consistent with previous mutagenic and crystallographic results (22,25).
The gel filtration profile of E371A iNOS isolated in the absence of H 4 B and L-Arg (Fig. 2, panel A) indicates that the protein is a mixture of approximately 75% dimer and 25% monomer that elute at 9.5 and 10.9 ml, respectively. The protein that elutes at 13 ml was identified as the iNOS reductase domain derived from nonspecific proteolysis of full-length E371A iNOS (data not shown). The presence of the reductase domain fragment likely explains why the flavin absorbance bands are so prominent in the visible spectrum of E371A iNOS (Fig. 1). Like wild-type iNOS, the E371A dimer can be dissociated into folded monomers in 2 M urea (Fig. 2, panel B), and the monomers can reassociate into dimer after being diluted 10fold and incubated with H 4 B (Fig. 2, panel C).
Unlike native iNOS, the E371A iNOS dimer was unable to synthesize NO from L-Arg and did not increase its rate of NADPH oxidation in response to L-Arg (data not shown). Adding NADPH to H 4 B-saturated E371 iNOS led to reduction of ϳ90% of its heme iron as determined by formation of the ferrous-CO species absorbing maximally at 444 nm (Fig. 3), similar to wild-type iNOS (6). L-Thiocitrulline, a substrate analog that acts as a competitive inhibitor versus L-Arg and blocks NADPH-dependent heme iron reduction (37), did not do so for E371A iNOS (Fig. 3), consistent with a defect in substrate binding. In sum, the E371A mutation prevented iNOS from binding or responding to L-Arg, but did not appear to affect iNOS dimeric interaction, H 4 B binding, heme environment, or electron transfer from the reductase domain to the heme. This confirms and extends results obtained with E371A iNOSox (22) and with an analogous mutant of endothelial NOS (38).
Generation and Purification of Complimentary E371/Wildtype Heterodimers-We next utilized the E371A full-length and oxygenase domain mutants to prepare complimentary heterodimers with wild-type partners. As shown in Fig. 4, these were a full-length wild-type iNOS subunit paired with an E371A oxygenase domain subunit (FL WT /OXY E371A ); and a full-length E371A subunit paired with a wild-type oxygenase domain subunit (FL E371A /OXY WT ). A completely wild-type heterodimer was also prepared as a control (FL WT /OXY WT ). To generate heterodimers, the original full-length and oxygenase domain homodimers were first dissociated into monomers using urea, mixed, and then induced to dimerize with H 4 B and  L-Arg. Each heterodimer was purified by sequential chromatography on Ni 2ϩ -chelate, 2Ј,5Ј-ADP-Sepharose, and Superdex gel filtration columns. A typical gel filtration profile for a heterodimer is shown in Fig. 5, along with gel filtration profiles for the two principle contaminants (full-length and oxygenase domain homodimers). After gel filtration chromatography the heterodimer fractions were greater than 80% pure and were biochemically characterized at this level of purity.
L-Arg Binding-We first determined how the E371A mutation affected L-Arg binding by each heterodimer. We utilized a spectroscopic perturbation method that is based on the ability of L-Arg to displace heme-bound imidazole and convert the iNOS heme iron from a low spin to high spin state (28,39). As shown in Fig. 6, panels A-C, addition of 0.4 mM imidazole to each of the three H 4 B-saturated heterodimers converted their predominantly high spin heme iron (Soret absorbance at 400 nm) to a fully low spin, imidazole-bound form with Soret absorbance at 426 nm. L-Arg was then added in graded amounts, and a spectrum was recorded after each addition until no further spectral change was observed. For the FL wt /OXY wt heterodimer control (panel A), adding a saturating concentration of L-Arg (5 mM) caused an almost complete displacement of bound imidazole and generated high spin heme, which absorbs maximally at 390 nm. In contrast, a saturating concentration of L-Arg only displaced imidazole from about half of the heme iron in either mutant heterodimer (panels B and C). The difference spectra for the three heterodimers are shown in Fig. 6, panel D, and confirm that L-Arg saturation displaced approximately half of the imidazole in each E371A mutant heterodimer as compared with the wild-type control. Thus, the E371A mutation only prevented L-Arg binding to one of two oxygenase subunits present in the mixed heterodimers.
To determine if the E371A mutation affected the affinity of the oxygenase partner that still could bind L-Arg, we examined plots of absorbance change (390 -430 nm) versus L-Arg concentration (Fig. 7). A similar concentration range of L-Arg ap-  peared to saturate all three heterodimers. Calculated Ks values for L-Arg binding were 32, 40, and 41 M for FL WT /OXY WT , FL WT /OXY E371A , FL E371A /OXY WT , respectively. This indicates the E371A mutation did not affect L-Arg affinity in one of the two subunits of each mixed heterodimer. Thus, our spectral analysis shows that the mutation preserves its phenotype in one subunit of a mixed heterodimer without affecting the other subunit's affinity toward L-Arg.
NADPH-dependent Heme Iron Reduction-We next compared the heterodimers regarding NADPH-dependent heme iron reduction under anaerobic conditions and in the presence of CO (Fig. 8). Under these conditions the wild-type heterodimer had previously been shown to reduce only one of its two heme irons (29). This was confirmed in Fig. 8, panel A, where addition of excess NADPH caused reduction of approximately 50% of the heme iron in the FL WT /OXY WT heterodimer as judged by the NADPH-dependent buildup of a ferrous-CO peak 444 nm relative to the peak intensity obtained upon reducing all of the heme iron with dithionite. Similar results were obtained with the FL WT /OXY E371A and FL E371A /OXY WT mutant heterodimers, indicating that only half of their heme iron was reduced by NADPH (Fig. 8, panels B and C, respectively). 3 Thus, electron transfer between the single reductase domain and one oxygenase domain heme iron was not disabled when the E371A mutation was present in the same or adjacent subunit relative to the reductase domain.
Catalytic Properties- Table I compares the NO synthesis and NADPH oxidation activities of the three heterodimers. The FL WT /OXY WT control heterodimer synthesized NO at a rate that was approximately half that of a full-length iNOS homodimer on a per heme basis, consistent with our previous FIG. 6. L-Arg binding to wild-type and E371A heterodimers. Binding of L-Arg to each heterodimer was assessed spectroscopically as the ability to displace imidazole as a sixth ligand to the heme iron. Panels A-C contain spectra for wild-type, FL WT /OXY E371A , and FL E371A /OXY WT heterodimers, respectively. Each heterodimer was diluted to ϳ1.75 M in 40 mM EPPS, pH 7.7, containing 10% glycerol, 0.05 mM DTT, and 4 M H 4 B. After recording an initial spectrum (-), 0.4 mM imidazole was added, and another spectrum was recorded (⅐ ⅐ ⅐ ⅐ ⅐). The imidazole displacement was then attempted by incrementally adding L-Arg (3 M to 6 mM) and recording a spectrum after each addition. In all cases, no further change was observed after 6 mM L-Arg had been added. Panels A-C show the final spectrum obtained after addition of 6 mM L-Arg (---). Panel D contains the three difference spectra obtained by subtracting the initial imidazole-bound spectrum from the imidazole-bound, L-Arg-saturated spectrum for wild-type (-), FL WT /OXY E371A (---), and FL E371A /OXY WT (⅐ ⅐ ⅐) heterodimers. The experiment shown is representative of two done with different heterodimer preparations.
report (29). L-Arg addition also increased its rate of NADPH oxidation 4.5-fold relative to L-Arg-free conditions, whereas L-thiocitrulline reduced the rate of NADPH oxidation 4-fold below the substrate-free value. The FL E371A /OXY WT heterodimer synthesized NO at 75% of the wild-type control rate, and its rate of NADPH oxidation was also increased by L-Arg and decreased by L-thiocitrulline in a manner similar to the control. In contrast, the FL WT /OXY E371A heterodimer only exhibited residual NO synthesis compared with the control (9%), 4 and its rate of NADPH oxidation was affected only slightly by L-Arg and not at all by L-thiocitrulline. Thus, the E371A mutation only affected heterodimer NO synthesis and responses to L-Arg or L-thiocitrulline when it was present in the oxygenase domain located opposite to the reductase domain.
Heme-L-Arg Interaction-At this point, our heterodimer data is consistent with two possibilities. (1) Electron transfer occurs between reductase and oxygenase domains located on adjacent subunits if the E371A mutation affects L-Arg binding in the subunit in which the mutation resides. (2) Electron transfer occurs between reductase and oxygenase domains in the same subunit if the E371A mutation affects L-Arg binding by the adjacent wild-type subunit. To differentiate between possibilities 1 and 2, we examined the nature of the heme-L-Arg interaction using resonance Raman spectroscopy and a FL E371A / OXY WT heterodimer whose OXY WT heme was selectively enriched in 54 Fe (Fig. 9). 5 This mixed heterodimer contains only one functional L-Arg binding site and two nonequivalent heme irons, and thus enabled us to view the interaction that occurs between a single L-Arg and a single reduced, CO-bound heme iron. The method relies on the fact that the lighter 54 Fe isotope causes the Fe-CO stretching vibration, which is sensitive to L-Arg 6 (Fig. 10, compare traces a-d) to shift 3 cm Ϫ1 toward higher frequency compared with natural abundance Fe (designated 56 Fe in Fig. 10, compare traces d and e).
The resonance Raman spectrum of the reduced, CO-bound, 54 Fe-enriched FL E371A /OXY WT heterodimer (Fig. 10, trace f) contains both a broad Fe-CO band centered at 486 cm Ϫ1 and a sharp Fe-CO band centered at 514 cm Ϫ1 . The former is attributable to a 56 Fe-CO species that is not interacting with bound L-Arg, whereas the latter is attributable to a 54 Fe-CO species that is interacting with L-Arg. Because the E371A mutation and the 54 Fe heme are located in opposite subunits of the FL E371A /OXY WT heterodimer, we conclude that the E371A mutation must block L-Arg binding in the same subunit in which the mutation is present. Thus, interactions between L-Arg and the 54 Fe heme-CO complex must occur within a single subunit, eliminating "possibility 2" above.
FIG. 8. NADPH-dependent heme iron reduction in the wild type and mutant heterodimers. The degree of heme iron reduction was judged under anaerobic conditions by formation of the ferrous-CO complex. Panels A-C contain spectra for wild-type, FL WT /OXY E371A , and FL E371A /OXY WT heterodimers, respectively. An initial spectrum of each heterodimer (ϳ2.5 M) was taken in CO-saturated 40 mM EPPS, pH 7.7, containing 10% glycerol, 0.05 mM DTT, 4 M H 4 B, and 1 mM L-Arg (-). NADPH (0.1 mM) was then added anaerobically, allowed to interact for 30 min at 25°C, and a spectrum recorded (---). Dithionite solution was then added and a final spectrum that represents full heme reduction was recorded (-⅐ -⅐). Data shown is representative of that obtained for three separate heterodimer preparations.

TABLE I
Catalytic activities of the iNOS heterodimers NO synthesis from L-Arg and NADPH oxidation in the presence or absence of L-Arg or L-thiocitrulline were determined using separate heterodimer preparations under assay conditions detailed under "Experimental Procedures." Under these conditions the NO synthesis activity for the parent full-length iNOS homodimer was 181 mol/min per mol of heme. The values represent the mean and standard error for three measurements each.  . 9. The 54 Fe-enriched heterodimer constructed to examine the L-Arg-heme iron interaction. The 56 Fe-heme subunit contains the E371A mutation (*) and natural abundance iron isotopes. The 54 Fe-heme subunit was generated by bacteria grown in 54 Fe-enriched medium.

DISCUSSION
Because NOS heme iron reduction is critical for NO synthesis, it is important to understand the path of electron transfer and how it is controlled. During NO synthesis the reductase domain flavins accept electrons from NADPH and transfer them to the heme irons located in the oxygenase domains. However, the homodimeric structure of iNOS introduces complexity to this system and creates three possibilities regarding electron transfer from a single reductase domain. Electrons could flow to both hemes, only to the heme located on the same subunit, or only to the heme located on the opposite subunit. Our previous work with an iNOS heterodimer revealed that its single reductase domain transfers electrons to only one of the two oxygenase domain heme irons. There was no subsequent sharing of electrons between hemes, and this system supported a normal rate of NO synthesis (29). Here, we generated a complimentary pair of heterodimers that contained an L-Arg binding mutation (E371A) in either oxygenase domain to make the oxygenase domains dissimilar and identify which of the two heme irons becomes reduced. Our data is consistent with electron transfer from the single reductase domain proceeding exclusively to the oxygenase domain that represents the adjacent subunit.
Our conclusion derives from experiments that detailed the basis of each mutant heterodimer's catalytic profile. Such anal-ysis is essential when using mutagenesis to investigate structure-function aspects of any NOS dimer, because in principle a single oxygenase point mutation could affect the functioning of the domain in which the mutation is present, the adjacent wild-type domain, or both domains. The E371A mutation completely prevents L-Arg binding when present in both oxygenase domains of an iNOSox (22) or full-length iNOS homodimer. When the E371A mutation was present in only one of two oxygenase domains of a heterodimer, it completely prevented L-Arg binding to one oxygenase domain in each heterodimer, but did not affect the L-Arg affinity of the oxygenase domain partner and did not prevent heme iron reduction in any case. The E371A mutation prevented NO synthesis when it was located in the oxygenase domain opposite to the reductase domain but had no effect on NO synthesis when it was located in the same subunit as the reductase domain. Our 54 Fe resonance Raman study showed that the E371A mutation only affected L-Arg binding within the same oxygenase domain in which the mutation was present, and consequently revealed that a bound L-Arg molecule interacts only with the heme that is located in the same oxygenase component to which L-Arg is bound. 7 Together, this led us to conclude that heme iron reduction in the heterodimer must occur entirely in the oxygenase domain adjacent to the full-length subunit, and this enables the oxygenase domain to generate NO by oxidizing the L-Arg molecule bound within it.
Our findings establish a critical link between iNOS dimeric structure and heme reduction that helps explain how dimerization activates iNOS. Previous studies had shown that only the oxygenase domains of iNOS participate in the dimeric interaction (17,28). We now propose that the oxygenase-oxygenase interaction enables the enzyme to engage in a type of "domain swapping" (40) shown in Fig. 11, which in turn allows NADPH-derived electrons to transfer between reductase and oxygenase domains that are located on adjacent subunits. This arrangement possibly circumvents a physical barrier for electron transfer between reductase and oxygenase domains located on the same subunit (29).
Dimerization also enables the oxygenase domains to stably incorporate H 4 B and to bind L-Arg with high affinity (23,27). This impacts on iNOS electron transfer in a different way 7 The resonance Raman data is consistent with recent crystallographic structures of the iNOSox monomer and dimer that show Glu-371 is located near the heme that is bound within the same subunit, and binds to the guanidino nitrogens of L-Arg such that they are held above that same heme iron (25,26). through an effect of H 4 B and L-Arg on the heme. In the absence of bound L-Arg and H 4 B, the heme irons in an iNOS dimer are solvent-exposed (23) and are poised at such a low reduction potential (Ϫ350 mV) that they appear to be thermodynamically unable to accept electrons from the reductase domain flavins (41). 8 Binding either H 4 B or L-Arg to the dimer alters the heme environment and increases the heme iron reduction potential by 50 or 100 mV, respectively (41). Thus, dimerization properly orients iNOS reductase and oxygenase domains for electron transfer, and creates an environment where substrate and pterin can positively modulate the thermodynamics of heme iron reduction.
Given our iNOS structure-function model in Fig. 11, it is important to consider the model's limitations and the implications of the domain swapping interaction. Although our data support an exclusive interaction between adjacent reductase and oxygenase domain pairs in an iNOS homodimer, it is still possible that electrons could exchange between the two reductase domains in the homodimer prior to transfer to an individual oxygenase domain. Such "short circuiting" of electron transfer would result in both hemes receiving electrons that were originally provided to a single reductase domain by NADPH. Present evidence suggests that short circuiting may not occur. For example, during titrations with NADPH, electron transfer between individual reductase domains in solution occurs only slowly. 9 Also, the crystal structure of the iNOSox dimer shows that the two reductase domains would be attached on opposite sides of the dimeric oxygenase core with each one positioned to interact with an exposed heme edge located on the backside of the adjacent subunit's oxygenase domain (26).
Which protein residues facilitate electron transfer between reductase and oxygenase domains are still unknown. However, putative domain contact regions have already been suggested based on the crystal structures of iNOSox (25,26) and of NADPH-cytochrome P-450 reductase (42), which is a protein homologous to the reductase domain of iNOS. Interestingly, the proposed contact region on the surface of iNOSox may only become exposed upon dimerization (26), implying another key role for this process. Molecules that antagonize the reductaseoxygenase interaction probably exist both in nature and in the laboratory. For example, cytochrome c (43), caveolin (44 -46), and certain oxygenase domain peptides (47) may bind to the oxygenase or reductase contact sites and inhibit NO synthesis by blocking domain interactions required for electron transfer to the heme.
Electron transfer as depicted in Fig. 11 takes place between noncovalently linked reductase and oxygenase domains of the dimer. However, their productive interaction probably requires that the reductase domains still remain covalently linked to the dimeric core, because mixing detached reductase monomers with iNOSox dimers promotes only inefficient heme reduction and slow NO synthesis (48). The reductase-oxygenase interaction does not endow the full-length iNOS dimer with greater stability than an iNOSox dimer (27,28) and, unlike the oxygenase-oxygenase domain interaction, does not remain intact after the reductase and oxygenase domains are separated by limited proteolysis (17). CaM is likely to have an important role in the reductase-oxygenase domain interaction of iNOS, because it controls electron transfer between these domains in the neuronal and endothelial NOS isoforms (6,49). Interaction between neuronal NOS and the "latch region" of CaM appears to be essential for electron transfer to its heme (51,52). Whether a similar interaction controls iNOS heme reduction is an interesting possibility. CaM binding to iNOS is essentially irreversible (50) and may therefore promote a more stable reductase-oxygenase domain interaction compared with neuronal and endothelial NOS, which bind CaM reversibly depending on the local Ca 2ϩ concentration (1,2). The domain swapping model provides a new perspective to all of the issues noted above.
Domain swapping as envisioned for iNOS is a relatively common feature among oligomeric proteins, and often provides advantages in function and control that are unavailable to the protein monomer (40). Domain swapping provides a way for iNOS to prevent uncoupled NADPH oxidation and may provide a means for H 4 B function or for CaM control of heme iron reduction. In general, interactions between domains from adjacent subunits arise only after oligomerization has occurred in evolution of the protein's structure. Over time, intersubunit interactions often replace an original intrasubunit interaction between domains linked together on the polypeptide monomer, because after oligomerization the intrasubunit interaction is not under selective pressure to be maintained (40). Accordingly, the iNOS polypeptide probably acquired its bi-domain structure first (linked reductase and oxygenase domains), followed by its dimeric structure, which eventually led to reliance on intersubunit interactions between adjacent reductase and oxygenase domains for heme iron reduction (and NO synthesis). Evidence that some bacteria express NOS activity and contain genes encoding an oxygenase domain-like protein (56) suggest that more primitive forms of the enzyme exist and may help trace the evolution of this complex enzyme.