Characterization of Mouse nNOS2, a Natural Variant of Neuronal Nitric-oxide Synthase Produced in the Central Nervous System by Selective Alternative Splicing*

Mouse neuronal nitric-oxide synthase 2 (nNOS2) is a unique natural variant of constitutive neuronal nitric-oxide synthase (nNOS) specifically expressed in the central nervous system having a 105-amino acid deletion in the heme-binding domain as a result of in-frame mutation by specific alternative splicing. The mouse nNOS2 cDNA gene was heterologously expressed in Escherichia coli, and the resultant product was characterized spectroscopically in detail. Purified recombinant nNOS2 contained heme but showed no l-arginine- and NADPH-dependent citrulline-forming activity in the presence of Ca2+-promoted calmodulin, elicited a sharp electron paramagnetic resonance (EPR) signal at g = 6.0 indicating the presence of a high spin ferriheme as isolated and showed a peak at around 420 nm in the CO difference spectrum, instead of a 443-nm peak detected with the recombinant wild-type nNOS1 enzyme. Thus, although the heme domain of nNOS2 is capable of binding heme, the heme coordination geometry is highly abnormal in that it probably has a proximal non-cysteine thiolate ligand both in the ferric and ferrous states. Moreover, negligible spectral perturbation of the nNOS2 ferriheme was detected upon addition of either l-arginine or imidazole. These provide a possible rational explanation for the inability of nNOS2 to catalyze the cytochrome P450-type monooxygenase reaction.

duction of nitric oxide (NO) (1)(2)(3)(4)(5)(6)(7)(8). It is a bifunctional enzyme that is comprised of an N-terminal heme-binding domain and the C-terminal flavin-containing cytochrome P450 reductaselike domain. Ca 2ϩ -promoted calmodulin binding activates electron transfer from the flavin site to the heme site (9), where NO is produced at the heme center in the presence of oxygen, H 4 BP, and L-arginine (6,8). The heme-binding domain of NOS shows negligible structural homology to regular cytochrome P450 (6,8,10,11), although a wide range of spectroscopic evidence supports coordination of an endogenous thiolate sulfur donor ligand to the central heme iron (12)(13)(14)(15). Recent high resolution x-ray crystal structure determinations of the monomeric and dimeric forms of the heme-binding domain fragment of inducible NOS (iNOS) have proven that the overall protein topology is very different from that of either regular cytochrome P450 or chloroperoxidase, although the proximal ligand to the central ferriheme iron is a conserved cysteine residue (10,11). The endogenous thiolate sulfur donor ligand to the central heme iron provides a rational basis for the proposed two-step sequential catalytic mechanism of NOS, where the first step involving the formation of the intermediate N -hydroxyarginine may follow a cytochrome P450-type monooxygenase mechanism (1, 16 -20). The substrate (L-arginine) and pterin cofactor (H 4 BP) binding sites have been defined by the crystal structure of the dimeric iNOS heme domain (11), and the former site (conserved glutamate residue) at the C-terminal part of the NOS heme domain has also been subjected to site-directed mutagenesis studies (21,22).
Aside from the protein chemistry, the in vivo regulation of NOS, from the viewpoint of control of NO synthesis, has been the subject of extensive investigation because of diverse physiological functions of NO in cells. Thus, NO produced by NOS isoforms induces vascular smooth muscle relaxation, serves as a messenger in the central and peripheral nervous systems, and also acts as a cytotoxic agent in the immune system to kill tumor cells and intracellular parasites (2,(23)(24)(25)(26)(27). In the case of nNOS, the genes are either constitutively or stage-and tissuespecifically expressed in different neuronal cell types and in skeletal muscle (5,28). Although the coding region of the nNOS gene encodes a 160-kDa protein, the mRNA is considerably larger (ϳ9.5 kilobase pairs), and significant molecular diversity is found, mostly in the 5Ј-noncoding region. This diversity is produced by specific alternative splicing at different splice sites and may affect the stability of individual mRNA species (5).
It has been reported that several different nNOS are produced in different cell types by selective alternative splicing in the coding region (2, 7, 28 -35). The presence of at least six different variants of alternatively spliced nNOS mRNA species, which are expressed in a tissue-specific and developmentally regulated manner, has been reported (28). Among them, a natural mutant of mouse nNOS created by specific alternative splicing and designated nNOS2 by Ogura et al. (29) is one of the most unusual and interesting. First, it has been shown that the primary structure of mouse nNOS2 lacks the C-terminal half of the highly conserved "dihydrofolate reductase module" region in the heme-binding domain (corresponding to residues 504 -608 in mouse or rat nNOS1) due to in-frame mutation with skipping of exons 9 and 10 by alternative splicing (29) (see Fig.  1). Second, the missing 105-amino acid residue region contains Glu-597, which is involved in L-arginine binding (11,21,22). Third, despite the difference of the heme domain sequence of nNOS2 from that of nNOS1, their calmodulin-binding and cytochrome P450 reductase domains remain intact at the primary structure level (29 -31). Fourth, the alternatively spliced product is specifically expressed in mouse central nervous system cells (29,30), and its homolog has also been identified in rat and human central nervous system and Drosophila head cells (36), implying an unknown conserved function in the central nervous system. Nevertheless, none of these natural variants of NOS isoforms has been characterized spectroscopically in detail. It is therefore of particular interest to investigate whether or not the alternatively spliced product nNOS2 has a native heme coordination geometry and retains the ability to function as a true "nitric-oxide synthase." In this paper, we report the heterologous overexpression of mouse nNOS2 cDNA gene in Escherichia coli for the first time and the molecular and spectroscopic characterization of the recombinant enzyme.
DNA Manipulations-The baculovirus-Spodoptera frugiperda (Sf9) insect cell expression system (37) and E. coli pCWoriϩ expression system (38 -40) were employed for the expression of mouse full-length wild-type nNOS (nNOS1) and the alternatively spliced form (nNOS2). Unless otherwise stated, vectors were constructed by using E. coli HB101 as the host strain. The site-directed mutagenesis was done by using a Muta-Gene Phagemid in vitro mutagenesis kit (Bio-Rad). All of the altered DNA sequences were analyzed by using a Sequenase version 2.0 DNA sequencing kit (U. S. Biochemical Co).
Expression of nNOS1 and nNOS2 Using the Baculovirus-Sf9 Insect Cell System-The plasmid vectors, pTZ19RNOS1 carrying the fulllength cDNA for mouse nNOS and pTZ19RNOS2 carrying the alternatively spliced cDNA reported previously (29), and the baculovirus transfer vector pJVP10Z (41) (kindly provided by Dr. S. Kawamoto, Yokohama City University) were used. The transfer vectors carrying the cDNA for nNOS1 and nNOS2 were individually constructed for recombination with AcNPV as follows: an NheI site was newly generated upstream of the first Met codon of the cDNA for nNOS1 with a phosphorylated DNA oligomer (NOSNhe I: 5Ј-GTA AAA CGA CGG CCA GTG AGC TAG CAT GGA AGA GCA CAC G-3Ј). The DNA fragment coding the altered 5Ј-leader sequence and N-terminal region of nNOS was replaced by the corresponding region of nNOS2 using two unique restriction enzyme sites, ScaI and AflII sites. The obtained plasmids (pTZ19RNOS1/NheI, and pTZ19RNOS2/NheI) were excised by NheI and XbaI digestion, and the DNA fragments encoding nNOS genes were individually ligated into the NheI site of pJVP10Z using the compatibility between XbaI and NheI sites. The directions of the cDNA were identified by DNA sequencing.
The coinfection with AcNPV DNA and the constructed transfer vectors was conducted by using a linear transfection module (Invitrogen), and the screening of the recombinant virus was carried out according to the manufacturer's manual, Max Bac baculovirus expression system manual (Invitrogen) and the literature (37). The recombinant enzymes were expressed as described in the literature (42)(43)(44).
Expression of nNOS1 and nNOS2 Using the E. coli Expression System-A heterologous expression system for nNOS1 and nNOS2 in E. coli was constructed with pCWoriϩ vector (38 -40) and the chaperonin expression vector pKY206 (pACYC184) (Nippon Gene, Toyama, Japan) carrying the E. coli chaperonin groELS genes (kindly provided by Dr. K. Ito, Kyoto University) as reported (39,40), with the following minor modifications. First, an NdeI site was newly generated to include the first Met codon of the cDNA for the wild-type nNOS in pTZ19RNOS1 by using the mutation primer (NOSNde I: 5Ј-GAC GGC CAG TGA GAA CAT ATG GAA GAG CAC ACG-3Ј). Because the obtained plasmid (pTZ19RNOS1/NdeI) has two NdeI sites, the plasmid was partially digested with NdeI, followed by complete digestion with XbaI. The NdeI-XbaI fragment containing the full-length nNOS1 cDNA was then inserted between the NdeI and XbaI sites of the multicloning linker of pCWoriϩ. The pCWori vectors for nNOS2 were prepared by exchanges of the AflII-XbaI fragment encoding the alternatively spliced region. The resultant pCWori vectors and pKY206 were introduced by repeated transformation into the host strain, E. coli strain BL21 (Takara Biomedicals), which lacks the two proteases lon and ompT. The recombinant enzymes were expressed as described in the literature (39,40).
Activity Measurement-NOS activity was measured by monitoring the conversion of L-[ 14 C]arginine to L-[ 14 C]citrulline as described previously (45). The standard assay was performed at 25°C in assay mixture containing 16. The NADPH-dependent diaphorase activity was measured using dichlorophenolindophenol as an artificial electron acceptor by monitoring the NADPH-dependent reduction of dichlorophenolindophenol at A 600 nm, essentially as described previously (9,46). The standard assay was performed at 25°C in an assay mixture containing 50 mM Tris-HCl buffer, pH 7.4, 100 M dithiothreitol, 85 M NADPH, 50 M dichlorophenolindophenol , 100 g of bovine serum albumin, 5 M FAD, 5 M FMN, and the enzyme in a total volume of 1 ml. The effect of Ca 2ϩcalmodulin complex on NADPH diaphorase activity was measured in the same assay mixture, except for the presence of 1 mM CaCl 2 and 10 g of calmodulin.
Purification of Recombinant nNOS-Recombinant nNOS1 and nNOS2 produced using the baculovirus-insect cell expression system were partially purified on a 2Ј,5Ј-ADP-Sepharose 4B column (Amersham Pharmacia Biotech), followed by a fast desalting column, Ampure SA (Amersham Pharmacia Biotech).
Purification of recombinant nNOS1 and nNOS2 produced in E. coli strain BL21 was performed on ice or at 4°C essentially as described in the literature (40), except that purification was conducted using 2Ј,5Ј-ADP-Sepharose 4B column chromatography (Amersham Pharmacia Biotech), followed by Sephacryl S200HR and DEAE-Sepharose Fast Flow column chromatography (Amersham Pharmacia Biotech) (47). H 4 BP (10 M) was supplied in the ultrasonification step. The catalytic activity and the purity of the purified wild-type enzyme, nNOS1, were comparable with those previously reported for recombinant nNOS1 by others (40).
Analytical Procedures-Absorption spectra were recorded using a Hitachi U3210 spectrophotometer or a Beckman DU-7400 spectrophotometer. EPR measurements were carried out using a JEOL JEX-RE1X spectrometer equipped with an Air Products model LTR-3 Heli-Tran cryostat system, in which the temperature was monitored with a Scientific Instruments series 5500 temperature indicator/controller as reported previously (47,48). EPR spectra of several different batches of recombinant nNOS samples were also measured at JEOL Ltd. (Tokyo, Japan), using a JEOL JES-TE200 spectrometer equipped with an ES-CT470 Heli-Tran cryostat system, in which the temperature was monitored with a Scientific Instruments digital temperature indicator/controller model 9650 and the magnetic field was monitored with a JEOL NMR field meter ES-FC5. All spectral data were processed using Ka-leidaGraph software version 3.05 (Abelbeck Software).
Purified nNOS was estimated by using the Coomassie protein assay reagent (Pierce) with bovine serum albumin as a standard. The homology search against data bases was performed with the BEAUTY and BLAST network service (49). The multiple sequence alignments were performed using the CLUSTAL X graphical interface (50) with small manual adjustments.

RESULTS AND DISCUSSION
The alternatively spliced product of mouse nNOS variant, nNOS2, is specifically expressed in the central nervous system and has a 105-amino acid deletion in the C-terminal highly conserved region of the heme-binding domain (29) (Fig. 1). We first produced the nNOS2 cDNA gene product in a baculovirus-Sf9 insect cell expression system and partially purified nNOS2 as described under "Experimental Procedures." In contrast to the case of the recombinant wild-type enzyme nNOS1, partially purified nNOS2 lacked citrulline-forming activity at 25-37°C (data not shown). Because essentially the same result has been obtained with recombinant mouse nNOS2 expressed in cell culture (29,30,51), we constructed the pCWoriϩ vector harboring the cDNA of the full-length nNOS2 gene, and produced the recombinant mouse nNOS2 in an E. coli expression system co-producing E. coli GroELS for detailed spectroscopic characterization. The recombinant enzyme was purified as described under "Experimental Procedures" and was used for further characterization (see below).
Characterization of Recombinant nNOS2- Fig. 2 shows typical optical absorption spectra of purified mouse nNOS2 pro-duced in E. coli strain BL21. The ferric form of nNOS2 purified either in the presence or absence of L-arginine contained bound protoheme IX, whose content varied considerably from preparation to preparation. Thus, a broad Soret peak centered at 416 nm was clearly visible in most preparations (represented by preparation 1) of purified nNOS2, 2 whereas a small shoulder around 420 nm could be detected in a few instances in which the spectral contribution was primarily from the bound diflavin centers in the reductase domain (represented by preparation 2) ( Fig. 2A). In either case, the occupancy of bound heme in purified nNOS2 was consistently much lower than that of recombinant nNOS1 purified in the presence of L-arginine and H 4 BP, which exhibited a broad Soret peak centered around 395-400 nm ( Fig. 2A). This is presumably because of a lower efficiency of heme incorporation into the nNOS2 heme domain than in the case of recombinant nNOS1. Interestingly, the spectral properties of resting nNOS2 are somewhat similar to FIG. 1. Multiple amino acid sequence alignments of the heme-binding domain of selected NOS isoforms that highlight the region deleted in the alternatively spliced product nNOS2. The sequence data were retrieved from data bases using the BEAUTY and BLAST network service (49), and the multiple sequence alignments were performed using a CLUSTAL X graphical interface (50) with minor manual adjustments. The region specifically deleted in the alternatively spliced product nNOS2 (corresponding to exons 9 and 10 of the mouse nNOS gene) is highlighted in the nNOS sequences (darkly shaded). The position of the heme ligand, the secondary structure elements of the monomeric form of the iNOS heme domain structure (10), and the substrate and cofactor binding residues defined in the crystal structures of dimeric iNOS heme domain fragments (11) are also shown at the top of each column. The putative helical T and helical lariat regions detected in the dimeric iNOS heme domain fragment structures (11), and the putative calmodulin-binding site (CaM Binding) of NOS isoforms are mentioned near the bottom of the figure. those of the C415H mutant of recombinant nNOS1 (42,52,53) and the C194H and C194S mutants of recombinant iNOS (54), all of which exhibited a very weak shoulder around 420 nm. Because the visible absorption spectrum of the purified diflavin reductase domain of NOS exhibited no shoulder around 415-420 nm (46), this indicates the presence of a small or trace amount of bound heme in purified nNOS2 preparations as well as in these mutant enzymes.
The CO-reduced minus reduced difference spectrum of purified nNOS2 (preparation 1) typically exhibits a weak peak at 422 nm, 2 instead of the 443-nm peak seen in the case of the wild-type enzyme nNOS1 (Fig. 2B, solid traces). No 450-nm peak could be detected in the difference spectrum of recombinant nNOS2 in the crude cell lysate, either, indicating that this is not purification artifact (Fig. 2B, dashed traces). These data suggest that the nNOS2 heme domain is capable of binding heme, and that the heme coordination geometry of nNOS2 is markedly different from that of the wild-type enzyme nNOS1.
Preparations 1 and 2 of purified nNOS2 produced in E. coli BL21 both showed no detectable citrulline-forming activity, as opposed to the case of the wild-type enzyme (nNOS1) purified in the presence of H 4 BP and assayed without preincubation with L-arginine or pterin cofactor (ϳ220 -360 nmol/min/mg at 25°C, which is comparable with the values reported by others) (40). This clearly shows that the occupancy of heme does not correlate with the absence of the citrulline-forming activity in purified nNOS2. Thus, the heterologous overexpression of mouse nNOS2 cDNA gene with several different expression systems consistently resulted in a recombinant protein with no NOS activity, supporting the previous result obtained with a crude enzyme preparation (29,30).
Gel filtration analysis with a calibrated Tosoh G-3000SW XL column connected to the high performance liquid chromatography system suggested that purified recombinant nNOS2 is a multioligomeric form, but not a dimeric form, even in the presence of added 10 M H 4 BP (data not shown), as was found recently for the alternatively spliced product of human iNOS (55). This suggests that the formation of the dimeric form is impaired in the alternatively spliced NOS isoforms.
To test if electron transfer occurs from the reductase domain to the heme domain of nNOS2, the purified enzyme (preparation 1) was reduced with NADPH in the presence of Ca 2ϩ , calmodulin, L-arginine, and H 4 BP under anaerobic conditions in the presence of CO. Fig. 3 shows the resultant spectral change of nNOS2, indicating the formation of ferrous-CO complex at 421 nm in 10 min under anaerobic conditions. The formation of the ferrous-CO complex was not completed within 5 min (data not shown), and hence is catalytically insignificant. Thus, the heme reduction of nNOS2 by NADPH was very slow. It is probably attributable to intermolecular electron transfer under the anaerobic conditions. In another complex metallo-  flavoprotein, xanthine oxidase, it has been reported that the anaerobic reduction of the enzyme by substrates proceeds in two phases, and that the slower phase is due to an intermolecular electron transfer rather than to a reduction brought about by direct electron transfer from the substrate to the chromophores undergoing reduction in this slow phase (56).
Nevertheless, our results shown in Fig. 3 are in line with previous reports showing the calmodulin-binding and cytochrome P450 reductase domains of nNOS2 being intact at the primary structure level and having a certain level of NADPH diaphorase activity and an ability to bind Ca 2ϩ -calmodulin complex (29 -31, 55), 3 and with the EPR detection of a stable radical feature in purified nNOS2 indicative of a putative flavin semiquinone (see below), as observed for the wild-type enzyme nNOS1 (12,(57)(58)(59).
EPR Properties-To investigate the nature of the heme coordination environment of nNOS2, the EPR spectra were recorded at 7-30 K. The EPR properties of the wild-type enzyme nNOS1 have been reported by several other groups (12,(57)(58)(59). Fig. 4 shows the comparative X-band EPR spectra at 7 K of purified nNOS1 and nNOS2 preparation 1 in the presence of excess L-arginine and H 4 BP. The substrate-bound form of purified nNOS1 exhibited a predominant high spin ferriheme species at g ϭ 7.59, 4.08, and 1.81 (trace A), which was clearly detected at temperatures below 10 K and is consistent with the endogenous thiolate sulfur donor ligand coordinated to the central heme iron (12, 47, 57, 59 -62). In sharp contrast, nNOS2 (preparation 1) prepared under the same conditions exhibited a high spin ferriheme species at g ϭ 6.02, 5.89, and ϳ2.0 (trace B), but no detectable amount of the g ϭ ϳ7.6 species. Notably, the g ϭ 6.02 species of nNOS2 is very similar to that observed for a predominant high spin ferric form of horse metmyoglobin at g ϭ 5.98, 5.87, and 2.0, which is known to have a proximal histidine ligand coordinated to the central heme iron (trace C).
The resting nNOS1 prepared in the presence of 10 M H 4 BP exhibited the high spin ferriheme at g ϭ 7.68, 4.07, and ϳ1.8 (1) as a predominant ferriheme species (Fig. 5, trace A). It also exhibited at least two overlapping low spin ferriheme species at g ϭ 2.45-2.41, 2.28, and ϳ1.91 with slightly different temperature dependences as minor ferriheme species (typically ϳ20 -40% of the total ferriheme, depending on the preparation) at 15-30 K, in addition to a stable radical feature at g ϭ 2.0 (presumably flavin semiquinone) (Fig. 5, trace A). Addition of L-arginine caused conversion of the remaining "g ϭ 2.41" low spin species to the high spin species at g ϭ 7.59, 4.08, and 1.81, such that the rhombicity (defined in terms of the ratio of the rhombic and axial zero field splitting parameters, E/D) of the high spin ferriheme decreased from 0.075 to 0.073 upon binding of L-arginine, as previously reported by others (57,59,60,62) (Fig. 5, trace B). The low to high spin state conversion was also detected in the visible absorption spectral change of the Soret band of the purified enzyme (data not shown). Thus, it is confirmed that the catalytic site geometry and the shape of the substrate-binding site of the resting nNOS1 are perturbed upon binding of L-arginine, even though it does not directly coordinate to the central heme iron. On the other hand, incubation of the resting nNOS1 with excess imidazole (7 mM) caused a decrease of a high spin ferriheme species and led to formation of a hexacoordinated low spin species at g ϭ 2.58, 2.30, and 1.83 as a predominant ferriheme species (Fig. 5, trace  C), which is different from the minor low spin species found in the resting enzyme (Fig. 5, trace A). The high to low spin state conversion was also detected in the visible absorption spectral change of the Soret band of the purified enzyme (data not shown). Comparison of the crystal field parameters of the low spin ferriheme species of the resting and imidazole-bound nNOS1 (traces A and C, respectively, in Fig. 5) suggests the decrease of both tetragonality (⌬/; from 4.44 to 4.14) and rhombicity (V/⌬; from 1.12 to 0.83) upon imidazole coordination. This is similar to the case of cytochrome P450cam upon imidazole binding (18,63), although a comparison of the crystal field parameters of the low spin ferriheme species suggests that the distal ligand of nNOS1 has a more hydrophilic environment than that in regular cytochrome P450.
In sharp contrast, no clear resonance could be detected for the resting nNOS2 in the g ϭ ϳ2 region in the temperature range examined (7-22 K), except for a stable radical feature at g ϭ 2.0 (presumably flavin semiquinone), as depicted in Fig. 5 (trace D). The addition of either H 4 BP (trace DЈ), L-arginine (trace E) or imidazole (trace F) to the resting nNOS2 caused only a negligible change in the EPR lineshape of the high spin ferriheme, and no new low spin ferriheme species appeared; none of these compounds caused significant perturbation of the heme coordination environment of nNOS2, as opposed to the case observed with nNOS1 (Fig. 5). This was also confirmed by the visible absorption spectra, showing negligible perturbation of the ferriheme center of nNOS2 (data not shown). These data suggest that L-arginine and imidazole do not have ready access to and/or bind only weakly to the putative distal substratebinding pocket of nNOS2 under the conditions used.
Taken together, the present EPR analysis clearly demonstrates that the coordination geometry of the ferriheme center and the distal substrate-binding pocket of purified nNOS2 are abnormal in that they are markedly different from those of the wild-type enzyme nNOS1. Because a high spin ferriheme center with an endogenous thiolate sulfur ligand would not elicit the "g ϭ 6" EPR signal (18,64,65), we suggest that the ferriheme center of the resting nNOS2 is predominantly high spin with a proximal non-sulfur (possibly N-donor such as histidine imidazole) ligand, as has been reported for the high spin ferriheme of metmyoglobin. In other words, the heme cofactor is not correctly attached to a cysteine residue in nNOS2.
Correspondence with the X-ray Crystal Structure of iNOS Heme Domain Fragments-The spectroscopic data reported herein suggest that the proximal side of the heme center of recombinant nNOS2 is coordinated by a non-sulfur ligand (presumably a histidine imidazole ligand) in both the ferric and ferrous states. The abnormal heme coordination geometry and distal substrate-binding pocket of purified nNOS2 are in line with the absence of any detectable citrulline-forming activity and the negligible spectral change of the high spin ferriheme center upon addition of L-arginine or imidazole, supporting the previous proposal by Ogura et al. (29,30) that nNOS2 probably does not serve as a true nitric-oxide synthase in vivo. Because Cys-415, which serves as the proximal thiolate sulfur donor ligand to the heme center in nNOS1, is not deleted by the alternative splicing (corresponding to residues 504 -608 in mouse nNOS1, see Fig. 1), some other amino acid residue (presumably histidine) should serve as the alternate, non-sulfur ligand to the heme center in nNOS2. It is noteworthy in this connection that the H 4 BP-deficient, recombinant iNOS has . The substrate-binding helix ␣7 (also termed helix ␣7a) (11) of the iNOS heme domain fragment supplies residues that interact with both L-arginine and pterin cofactor and also contributes to the dimer interface upon dimerization (10,11). been reported to be unstable and partially converted in vitro to a cytochrome P420-like form, which also has a non-sulfur proximal ligand (presumably a histidine ligand), as indicated by resonance Raman studies (66).
Recently, Crane et al. (10,11) have reported the x-ray crystal structure of the H 4 BP-free, monomeric iNOS heme domain fragment ⌬114 complexed with imidazole (residues 115-498, corresponding to residues 342-724 in mouse nNOS1) and the dimeric iNOS heme domain fragment ⌬66 complexed with H 4 BP and L-arginine (residues 67-498, corresponding to residues 299 -724 in mouse nNOS1). Because of the high amino acid sequence homology among the corresponding regions of the NOS isoforms (see Fig. 1), the structural data reported for the monomeric iNOS heme domain fragment provide possible structural insight into the putative protein folding topology of the corresponding region of nNOS2, which is multioligomeric (not dimeric). Fig. 6 compares the putative protein folding topology diagrams of the heme domain fragment of monomeric nNOS1 and nNOS2. In this tentative diagram of nNOS1 (left panel), the structural elements probably depleted in nNOS2 due to alternative splicing are shaded; they include all structural elements from the C-terminal part of helix ␣5 to the N-terminal part of helix ␣7 (nomenclature based on Crane et al., see Ref. 10). Interestingly, the missing regions involve several main ␤-sheets and strands (␤8b, ␤8c, ␤9b, and ␤9c) and the substrate-binding helix ␣7. The substrate-binding helix ␣7 of the monomeric form of iNOS heme domain fragment is also termed helix ␣7a in the dimeric form, and contributes to the dimer interface (11). The most notable changes at the distal substrate-binding pocket of the nNOS2 heme domain, as predicted from the iNOS heme domain fragment structures, are the inherent lack of the five conserved residues probably involved in binding of L-arginine (Trp-587, Tyr-588, Glu-592, Asp-597, Arg-603, which correspond to Trp-366, Tyr-367, Glu-371, Asp-376, Arg-382 of the iNOS heme domain (11), respectively), and the conserved residue involved in binding of the pterin cofactor (Arg-596, which corresponds to Arg-375 of the iNOS heme domain (11)) of nNOS1 ( Figs. 1 and 6). Moreover, several important residues corresponding to the distal Phe-363 of the iNOS heme domain fragment, which stacks with the porphyrin ring (10,11), and the arc of strictly conserved hydrophobic residues in the C-terminal conserved region (Val-346, Pro-344, Tyr-367, Ile-372, and Met-368 of the iNOS heme domain), which curves around the heme coordination site (10,11), are inherently missing in the distal pocket of the nNOS2 heme domain. These predictions provide complementary explanations for the marked modifications of the proximal ligand to the central heme iron and the distal pocket of nNOS2 detected spectroscopically, which are reflected in the abnormal coordination geometry of the bound heme, the absence of the L-argininedependent spectral perturbation, the impairment of formation of the dimeric form even in the presence of added H 4 BP, and the absence of citrulline-forming activity.
The molecular diversity of nNOS mRNA generated by selective alternative splicing has been a subject of extensive studies, mainly directed at the stage-and tissue-specific transcriptional regulation of nNOS (2, 7, 28 -35), but the products have been rather poorly characterized at the protein level. The mRNA of the natural alternatively spliced variant nNOS2 has been shown to be specifically and either constitutively or stage-and tissue-specifically expressed in mouse, rat, and human central nervous system and Drosophila head cells (29 -31, 36, 51, 67), implying a conserved function. Moreover, an analogous alternatively spliced variant was also identified in human iNOS very recently (55,68). The combined biochemical and spectroscopic evidence reported herein clearly shows that recombinant nNOS2 cannot serve as a true nitric-oxide synthase in vitro because of the abnormal heme coordination environment (e.g., see Refs. [17][18][19][20], the marked modification of the distal substrate-binding pocket, and the impairment of the formation of the active dimer. Thus, some developmental regulatory function and/or a novel physiological role (other than L-arginine-dependent NO production) should be expected for the gene product of nNOS2 in the central nervous system, and will be the subject of future investigation.