Zinc Content of Escherichia coli-expressed Constitutive Isoforms of Nitric-oxide Synthase

Recently, we obtained x-ray crystallographic data showing the presence of a ZnS4 center in the structure of Escherichia coli-expressed bovine endothelial nitric-oxide synthase (eNOS) and rat neuronal nitric-oxide synthase (nNOS). The zinc atom is coordinated by two CXXXXC motifs, one motif being contributed by each NOS monomer (cysteine 326 through cysteine 331 in rat nNOS). Mutation of the nNOS cysteine 331 to alanine (C331A) results in the loss of NO⋅ synthetic activity and also results in an inability to bind zinc efficiently. Although prolonged incubation of the C331A mutant of nNOS with high concentrations ofl-arginine results in a catalytically active enzyme, zinc binding is not restored. In this study, we investigate the zinc stoichiometry in wild-type nNOS and eNOS, as well as in the C331A-mutated nNOS, using a chelation assay and electrothermal vaporization-inductively coupled plasma-mass spectrometry. The data reveal an approximate 2:1 stoichiometry of heme to zinc in (6R)-5,6,7,8-tetrahydro-l-biopterin-replete, wild-type nNOS and eNOS and show that the reactivated C331A mutant of nNOS has a limited ability to bind zinc. The present study substantiates that the zinc in NOS is structural rather than catalytic and is important for maintaining optimally functional, enzymatically active, constitutive NOSs.

The constitutively expressed isoforms of nitric-oxide synthase (NOS), 1 endothelial nitric-oxide synthase (eNOS; NOS III) and neuronal nitric-oxide synthase (nNOS; NOS I), catalyze the NADPH-dependent conversion of L-arginine to L-cit-rulline with the concomitant formation of nitric oxide (NO ⅐ ) (1)(2)(3)(4)(5). NOS isoforms are bidomain global structures in nature, being composed of a flavin-containing C-terminal reductase domain and an N-terminal oxygenase domain (6). The oxygenase domain contains iron protoporphyrin IX (7-10) and binding sites for tetrahydrobiopterin (BH 4 ) and the substrate, L-arginine (11). All of the aforementioned cofactors are required for full enzymatic activity; thus cofactor binding would be expected to alter the enzyme function.
Site-directed mutagenesis studies have indicated that cysteine 99 of human eNOS is involved in BH 4 binding to that isoform (12) and mutation of iNOS cysteine 109 leads to diminished BH 4 binding (13). Additionally, deletion of the entire CXXXXC motif (Fig. 1) in bovine eNOS causes a dramatic loss of enzyme stability (14). To test the possibility of altered cofactor binding in nNOS, we mutated cysteine 331 (the homologous residue to cysteine 99 of human eNOS) to alanine and have shown that this mutation affects L-arginine binding and reductase-to-heme electron transfer (15,16). Prolonged incubation of the nNOS C331A mutant with high concentrations of L-arginine is required for substrate binding to this mutant. However, once L-arginine binding is restored, BH 4 can then bind, resulting in efficient flavoprotein-to-heme electron transfer. This reactivated mutant, with BH 4 and L-arginine bound, possesses the same electron transfer properties and enzymatic activity as wild-type nNOS (16).
In addition to cysteine 331, cysteine 326 of nNOS is also phylogenetically conserved (Fig. 1). Recently, we have obtained x-ray crystallographic data showing the presence of a tetrahedral tetrathiolate zinc (ZnS 4 ) center in the trypsin-cleaved heme domain of eNOS (17) and also nNOS, 2 which is formed by the CXXXXC motifs from two NOS subunits. Because full enzymatic activity of the C331A mutant can be achieved by prolonged incubation with L-arginine (16), we were interested in what role zinc plays in NOS structure/function. Therefore, considering that mutations within the CXXXXC motif should disrupt the ZnS 4 center, we have determined the amount of zinc bound to wild-type forms of nNOS and eNOS, as well as to various constructs of each and the C331A-mutated nNOS. Furthermore, we have investigated the time course required for L-arginine activation of the C331A-mutated nNOS.
nNOS, eNOS, and Constructs-nNOS and the C331A-mutated nNOS were prepared as described by Roman et al. (18) and Martá sek et al. (16), respectively, and portions of the preparations were incubated overnight either in the presence of 10 mM L-arginine or 10 mM Larginine and 0.5 mM BH 4 after affinity chromatography and before FPLC purification. Wild-type eNOS was prepared as described by Martá sek et al. (19). Enzymes and constructs were purified by FPLC in Chelex® 100-treated buffer consisting of 50 mM Tris, 100 mM NaCl, 1 mM 2-mercaptoethanol, 0.5 mM L-arginine, and 10% glycerol, pH 7.8. * This work was supported by National Institutes of Health Grants GM52419 and NHLBI 30050 and Grant AQ1192 from the Robert A. Welch Foundation (to B. S. S. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by National Heart and Lung Cardiovascular Training Grant HL07350- 19.
ETV-ICP-MS Analysis-All labware used for handling samples for metal determination was soaked overnight in 4 M nitric acid and rinsed with Chelex®-treated water before use. Enzymes were purified by FPLC in EDTA-free Chelex®-treated buffer, concentrated, and subsequently passed over a 10-cm column of Chelex®-100 and reconcentrated. ETV-ICP-MS analysis was performed by Delony Langer and Dr. James Holcombe (Department of Chemistry and Biochemistry, University of Texas at Austin). Standard deviations were calculated from the uncertainties in the slopes of the calibration curves. All results were derived from three replicates, plotted, and analyzed using a least squares fit. Metal concentrations were corrected for the metal contribution of the buffer, which, as a percentage of signal, was 29% for zinc and 18% for copper.
Chelation Assay-Zinc content of the constitutive NOS isoforms and constructs was also measured by the PAR assay essentially as described by Crow et al. (20) with modifications. PAR has a low absorbance at 500 nm in the absence of Zn 2ϩ (Fig. 2). However, in the presence of Zn 2ϩ , the absorbance at 500 nm increases dramatically as the PAR 2 Zn 2ϩ complex is formed. Enzyme (1 nmol in a volume of 20 l) was added to a rapidly stirred 3-ml cuvette containing 100 M PAR in Chelex® 100-treated 50 mM Tris, 100 mM NaCl, pH 7.8, and, for some experiments, including Chelex® 100-treated 7 M guanidine HCl. In some assays, oxidants were added to facilitate Zn 2ϩ release. Assays were run at 23°C in rapidly stirred cuvettes, and the total assay volume was 1.5 ml. Near the end of the assay, 1 mM NTA was added. Under these assay conditions, NTA selectively chelates Zn 2ϩ from the PAR 2 Zn 2ϩ complex and causes a decrease in absorbance at 500 nm, which is used to quantitate the amount of zinc that has been released from the protein. Following the NTA addition, 1 mM EDTA was added to confirm that no other metal was being chelated in addition to zinc. The extinction coefficient for PAR 2 Zn 2ϩ was calculated, using a standard solution of ZnCl 2 , to be 87.7 mM Ϫ1 cm Ϫ1 at pH 7.8. The results reported in Table I represent the means Ϯ S.D. of three or four determinations except for those obtained with nNOS heme domain (nNOSwt-HD with BH 4 ).
Activity Assay-NOS activity was measured using the hemoglobin capture assay in 50 mM Tris, 100 mM NaCl, pH 7.8. The rate of methemoglobin formation from oxyhemoglobin was measured as the difference in absorbance change between 401 and 411 nm/min as described by Sheta et al. (6) except that an extinction coefficient of 60 mM Ϫ1 was used. All spectrophotometric assays were conducted on a Shimadzu 2101-PC dual-beam spectrophotometer.
Protein Determination-The protein concentration for the enzymes and constructs was determined on the basis of heme content by reduced CO difference spectra using an extinction coefficient of 100 mM Ϫ1 cm Ϫ1 for a ⌬⑀ of 444 -475 (21). This method of protein determination probably underestimates the actual protein concentration because apoprotein (minus heme) exists in all preparations to some extent. However, the average heme content of the NOS preparations used in this laboratory is ϳ80%. 3

RESULTS AND DISCUSSION
With nNOS-wt holoenzyme, 90% of the zinc (based on 1 zinc atom/2 hemes) was released in the presence of 7 M guanidine HCl and 1 mM H 2 O 2 ( Table I). Results of the chelation assay were corroborated by ETV-ICP-MS analysis (Table II) and confirmed that zinc and not copper was bound to nNOS. Likewise, the trypsin-cleaved, dimeric heme domain of nNOS-wt contained a similar zinc content (89 -92%). The zinc content (89 -92%) of the trypsin-cleaved heme domain of nNOS, derived from the full-length nNOS holoenzyme, contrasts sharply to the zinc content of the heme domain construct (residues 1-714) of nNOS (Ϸ 20%, data not shown) expressed in Escherichia coli. These observations are consistent with the results of Crane et al. (22), who crystallized the E. coli-expressed heme domain of iNOS (residues 66 -498) and did not observe enzyme-bound zinc but rather identified a disulfide bond located in the analogous position where Raman et al. (17) found the ZnS 4 center in the eNOS heme domain. When the zinc content of the BH 4 -free

TABLE II ETV-ICP-MS analysis of copper and zinc content of eNOS and nNOS
The method of standard additions was used with ETV-ICP-MS for analysis. The concentrations of copper and zinc match well with isotopic ratios for the two metals (zinc 1.4 and copper 2.2 versus the theoretical values for zinc and copper, 1.5 and 2.3, respectively), thus confirming the accuracy of the analyses. A zinc/heme ratio of 0.45 represents 90% repletion based on a stoichiometry of 1 zinc/2 hemes. Heme content was measured by dithionite-reduced CO difference spectroscopy. Inductively coupled plasma-mass spectrometry data are presented as the means Ϯ S.D. of three replicate determinations. nNOS-wt holoenzyme was examined, a zinc content of only 66% was observed, suggesting that in the absence of bound BH 4 , disruption or destabilization of the ZnS 4 center occurs. In contrast to the zinc contents observed with nNOS-wt holoenzyme and the trypsin-cleaved nNOS-wt heme domain, only 5 and 19% of the theoretical zinc content was obtained with the C331A nNOS holoenzyme mutant preincubated with L-arginine or preincubated with L-arginine and BH 4 , respectively. These data strongly suggest that mutation of cysteine 331 of nNOS results in disruption of the ZnS 4 center and prevents stoichiometric zinc binding. However, because prolonged incubation of the nNOS C331A mutant as isolated with high concentrations of L-arginine is sufficient to restore enzymatic activity (16), these findings strongly implicate zinc in a structural rather than a catalytic role. eNOS-wt holoenzyme behaved in a similar manner to nNOS with respect to zinc content. In the BH 4 -repleted state, eNOS was 94% replete with zinc versus only 76% repletion in the absence of bound BH 4 . Again, data from the ETV-ICP-MS analysis of eNOS-wt holoenzyme were used to confirm that zinc and not copper was present in eNOS (Table II). Additionally, x-ray crystallographic data of both the BH 4 -free and BH 4 -bound eNOS heme domain show the presence of zinc (17).
The nNOS C331A mutant is inactive as isolated, but activity can be restored by prolonged incubation with L-arginine. Fig. 3 shows the time course for activation of the nNOS C331A mutant upon exposure to 10 mM L-arginine. The mutant enzyme progressively became more active with time. However, maximal activity (255 nmol/min/mg) was not reached for 6 -8 h following exposure to L-arginine.
In contrast to the recent findings of Perry and Marletta (23), which showed that copper is bound to E. coli-expressed nNOS-wt holoenzyme as isolated, we found no significant amount of copper in our preparations (copper to heme ratios: nNOS 0.01 and eNOS 0.03; Table II) and feel that zinc is the naturally occurring metal in the E. coli-expressed constitutive isoforms of NOS. This finding was confirmed by both the chelation assay performed in this laboratory, by x-ray crystallographic structure data (17), 2 and by ETV-ICP-MS analysis performed by an independent laboratory. Perry and Marletta (23) concluded that copper and zinc are the relevant metals in nNOS and iNOS, respectively. Our crystallographic work on eNOS (17) and nNOS 2 unambiguously establishes that a ZnS 4 center is present at the bottom of the dimer interface in both proteins. It has also been shown that zinc can be successfully incorporated into human iNOS. 4 Our biochemical characterization lends support to the x-ray crystallographic evidence for a structural zinc in all NOS isoforms. Our earlier work (24) has already provided evidence for a zinc-mediated inhibition of nNOS (K i ϭ 30 M). It is conceivable that a metal binding site in intact NOS, rather than the ZnS 4 site, may be responsible for this effect. The high affinity of zinc for the tetrathiolate ligands (K d ϭ nanomolar range) further supports a view that the zincmediated inhibition previously observed (24) arose from a site different from that of the ZnS 4 center. It is conceivable that the work by Perry and Marletta (23) has also implicated a metal site different from the zinc site (ZnS 4 ) discussed in the present study. Culture conditions are important and can play a critical role in which metal is incorporated into the protein. However, the quantity of copper present in the culture medium is not germane to this issue because copper cannot substitute for zinc and utilize a tetrathiolate liganding geometry. A survey of all the mononuclear copper sites in the Brookhaven Protein Data Bank reveals that the preferred geometry for copper sites in proteins is tetragonal or trigonal (25). Tetrahedral geometry at the copper site is very rare, and there is no report in the literature, to the best of our knowledge, involving CuS 4 , either in model compounds or in proteins. It should be noted that the only other metal that can substitute for zinc in the metal tetrathiolate is iron. Both physiological relevance and coordination geometry criteria would be satisfied if iron were to substitute for zinc. Having narrowed the number of metals to two, the issue of culture conditions that would favor one metal over the other can now be addressed. There are numerous examples in the literature addressing the problem of metal homeostasis and heterologous expression in E. coli. There are three examples worth noting. First is rubredoxin, containing a FeS 4 center, which when expressed in E. coli leads to a mixture of 70% ZnS 4 and 30% FeS 4 containing protein (26). No zinccontaining rubredoxin is isolated from the original host, Clostridium pasteurianum. Crystal structures of both the natural FeS 4 rubredoxin and the ZnS 4 rubredoxin show the ability of the thiolates to coordinate either metal with similar geometry (27). Second, expression of azurin, a copper protein in E. coli, leads to a mixture of copper-azurin and zinc-azurin (28). In this protein, however, the native copper is coordinated via two histidines and a methionine. Zinc is able to substitute for copper in a trigonal geometry. Once again, crystal structures of both copper-and zinc-azurin reveal the possible need for a copper chaperone in vivo for maintaining metal homeostasis. Third, ribonucleotide reductase R2 subunit, when expressed in E. coli, is produced as an apoprotein and exogenous Fe 2ϩ needs to be added to obtain a catalytically active protein (29). Thus, it is conceivable that NOS, when expressed in E. coli, incorporates zinc, whereas the endogenous source may contain iron. We can, however, narrow down the possibility to two metals, zinc or iron (see above). Also, as has been established for metallothionein, there remains the possibility that the ZnS 4 site in NOS may be thermodynamically stable but kinetically labile, thus facilitating exchange with iron. However, in our crystallographic experiments, we did not find any evidence for the ability of iron to displace zinc (17). Therefore, we do not feel that iron is the endogenous metal in E. coli-expressed NOS. We are currently investigating the identity of the metal bound to NOS expressed in mammalian cells.
The C331A-mutated nNOS is inactive (as isolated), and its lack of activity is reflected in its inability to bind cofactors efficiently and stabilize an active conformation. L-Arginine activation of C331A nNOS is an interesting observation, but it is just that. The work of Chen et al. Therefore, it is clear that zinc-mediated stabilization of the bottom of the dimer interface is key for catalytic activity, albeit an indirect effect. The comparison of other ZnS 4 proteins with the NOS isoforms suggests that the ZnS 4 center facilitates protein-protein interaction(s). In NOS, the ZnS 4 center stabilizes the dimer interface and/or the flavoprotein-heme domain interface (17). Zinc is not needed for folding because L-arginine is sufficient to restore enzymatically active protein. If zinc were only essential for folding, then the enzyme should function perfectly in its absence. Without preincubation in the presence of high concentrations of L-arginine, this is definitely not the case. Preincubation with L-arginine must stabilize an active conformation. However, it does not restore zinc binding. Therefore, zinc facilitates but is not absolutely required for activity. The present study substantiates that zinc content is influenced profoundly by pterin and confirms a structural, rather than catalytic, role for zinc in maintaining enzymatically active constitutive nitric-oxide synthases. Furthermore, we demonstrate the one zinc/two heme stoichiometry in the dimers of the constitutive NOS isoforms isolated from E. coli expression systems.
Note Added in Proof-In support of our finding of zinc in the E. coli-expressed nitric-oxide synthases, Fischmann and colleagues (30) have reported the crystal structures of human inducible and endothelial nitric-oxide synthase domains in which they find a zinc tetrathiolate center at the dimer interface.