Reversal of Cyanide Inhibition of Cytochrome c Oxidase by the Auxiliary Substrate Nitric Oxide

Nitric oxide (NO) is shown to overcome the cyanide inhibition of cytochrome c oxidase in the presence of excess ferrocytochrome c and oxygen. Addition of NO to the partially reduced cyanide-inhibited form of the bovine enzyme is shown by electron paramagnetic resonance spectroscopy to result in substitution of cyanide at ferriheme a3 by NO with reduction of the heme. The resulting nitrosylferroheme a3 is a 5-coordinate structure, the proximal bond to histidine having been broken. NO does not simply act as a reversibly bound competitive inhibitor but is an auxiliary substrate consumed in a catalytic cycle along with ferrocytochrome c and oxygen. The implications of this observation with regard to estimates of steady-state NO levels in vivo is discussed. Given the multiple sources of NO available to mitochondria, the present results appear to explain in part some of the curious biomedical observations reported by other laboratories; for example, the kidneys of cyanide poisoning victims surprisingly exhibit no significant irreversible damage, and lethal doses of potassium cyanide are able to inhibit cytochrome c oxidase activity by only ∼50% in brain mitochondria.

Nitric oxide (NO) is shown to overcome the cyanide inhibition of cytochrome c oxidase in the presence of excess ferrocytochrome c and oxygen. Addition of NO to the partially reduced cyanide-inhibited form of the bovine enzyme is shown by electron paramagnetic resonance spectroscopy to result in substitution of cyanide at ferriheme a 3 by NO with reduction of the heme. The resulting nitrosylferroheme a 3 is a 5-coordinate structure, the proximal bond to histidine having been broken. NO does not simply act as a reversibly bound competitive inhibitor but is an auxiliary substrate consumed in a catalytic cycle along with ferrocytochrome c and oxygen. The implications of this observation with regard to estimates of steady-state NO levels in vivo is discussed. Given the multiple sources of NO available to mitochondria, the present results appear to explain in part some of the curious biomedical observations reported by other laboratories; for example, the kidneys of cyanide poisoning victims surprisingly exhibit no significant irreversible damage, and lethal doses of potassium cyanide are able to inhibit cytochrome c oxidase activity by only ϳ50% in brain mitochondria.
The reactions of cytochrome c oxidase with nitric oxide (NO) have recently been the subject of renewed scrutiny (1-6) because of their emerging physiological significance (7)(8)(9)(10)(11)(12). Furthermore, the discovery and functional characterization of a mitochondrial NO synthase (9,13,14) raises a series of questions regarding the possible regulation of the electron transport chain by NO. Several catalytic cycles have been proposed (4,6,(15)(16)(17) to explain the consumption of NO by the oxidase while turning over in the presence of cytochrome c and oxygen. Some of these cycles implicitly treat the NO as an auxiliary substrate rather than simply a competitive inhibitor, to account for the observed production of nitrite in addition to the slowing of electron transfer rate. In some cases an oxyferryl intermediate of heme a 3 is proposed and interaction of incoming NO with Cu B assumed (15,16), whereas other schemes invoke the formation of a nitrosylheme a 3 intermediate (4,6,17). Consequently, clarification of which (if either) of the two centers of the binuclear pair, heme a 3 or Cu B , preferentially reacts with NO to form detectable derivatives should be of some value in assessing the relative likelihood of one proposed scheme over another as conditions are varied.
When metalloenzymes have multiple binding sites for small molecular substrates and inhibitors, especially within the same active site domain, studies with mixed-ligand adducts can often reveal useful details of the coordination chemistry. Here we report the results of an investigation into the preferred site of NO reaction with bovine heart cytochrome c oxidase in the presence of the potent competitive inhibitor, cyanide, both under conditions of turnover and where the enzyme has been allowed to equilibrate in the presence of NO, cyanide, and reducing equivalents. The findings we report appear to be surprising in two respects. First, the inhibitory effects of NO plus cyanide toward the enzyme are not additive; NO actually eliminates cyanide inhibition. Second, contrary to the observations of earlier authors (18 -21), we find that cytochrome c oxidase displays a marked tendency to form NO adducts of heme a 3 that are clearly 5-coordinate. The results have bearing on two related issues addressed under "Discussion." These are the likely mechanism of NO oxidation by the enzyme under conditions of high electron flux and the apparent existence of a cyanide-insensitive "pool" of cytochrome c oxidase in vivo.

MATERIALS AND METHODS
Enzymes and Reagents-Cytochrome c oxidase (EC 1.9.3.1, complex IV) was isolated from beef heart pericardium without the preparation of Keilin-Hartree particles (requiring acidification to pH Ͻ 6) common to most procedures. Using intact mitochondria, complex IV was separated from the other components of the electron transport chain by deoxycholate extraction as described by Ragan et al. (22) and then finally purified by ammonium sulfate/cholate fractionation according to the procedure of Yonetani (23). This method (in which the pH is maintained in the range of 7.8 to 8.0 throughout) yields good activity preparations of the oxidase that exhibit a 424 nm Soret absorption maximum, do not exhibit a g ϭ 12 EPR signal, rapidly bind cyanide, and rapidly react with NO, which are characteristics of the pulsed form of the enzyme (24). Moreover, the preparations usually persist in the pulsed form when stored overnight at 5-6°C or for months (at least) in frozen (glassing) solution at 77 K. The enzyme was determined to be spectroscopically pure if the 444 -424 nm ratio for the reduced enzyme was 2.2 or higher (25). Derivatives were prepared in 50 mM potassium phosphate, 1 mM in sodium EDTA, and 0.1% in lauryl maltoside, pH 7.4 -7.8, to concentrations of 10 -80 M (in enzyme). Enzyme concentrations were determined as total heme a using the differential (absorption) extinction coefficient of ⌬⑀ 604 ϭ 12 mM Ϫ1 cm Ϫ1 for the reduced minus oxidized spectrum of the enzyme (26). Concentrations throughout are given on a per enzyme concentration basis (not per [heme a]). Ferrocytochrome c:O 2 oxidoreductase activity was determined employing the high ionic strength method of Sinjorgo et al. (27). Using this assay, we obtain a turnover number with respect to cytochrome c of 340 (Ϯ30) s Ϫ1 (0.1 M sodium phosphate, 0.1% lauryl maltoside, pH 7.4, 22°C) similar to that of the bovine enzyme isolated from a variety of tissues by others (27).
A typical heme a extraction, maintaining nominally anaerobic conditions and under subdued light, was as follows. Approximately 2 ml of 0.26 mM (in total heme a) cytochrome c oxidase solution was taken, and 8 drops of concentrated sulfuric acid were added to precipitate the protein. After centrifugation (7000 ϫ g, 22°C, 2 min) the supernatant was discarded and heme a extracted from the pellet (22°C, 20 min, frequent vortexing) into 1 ml of dimethylformamide. The concentration of heme a (80 -100 M) was determined using the extinction coefficient ⑀ 558 ϭ 24 mM Ϫ1 cm Ϫ1 reported for the pyridine hemochrome (28). The procedure also results in the extraction of copper from the enzyme which appears in the final dimethylformamide solution as a stable Cu(II) species that is not reducible by ascorbic acid. EPR signals arising from this species have been removed from the nitrosylferroheme a spectra of Fig. 2, C and D, by subtracting the spectra of control samples prepared without adding NO.
All reagents were ACS grade or better, used without further purification, and unless stated to the contrary, were purchased from Aldrich or Sigma. Sodium dithionite, 87% minimum assay (ϩH 2 O), was obtained from EM Science. 13 C sodium cyanide, 99%, was obtained from Cambridge Isotope Laboratories. Argon and nitric oxide gases were obtained from Matheson Incorporated. Nitric oxide was scrubbed with water and KOH pellets prior to use and added to enzyme samples volumetrically with gas-tight syringes. Buffered solutions never exhibited any significant change of pH (i.e. Ͻ0.05 pH units) following NO and/or sodium dithionite additions.
Instrumental Methods-Electronic absorption spectra were measured and photometric determinations made using a Shimadzu UV-2501PC spectrophotometer. X-band EPR spectra were obtained using an IBM ESP 300 spectrometer equipped with a Bruker B-E 25 electromagnet and Bruker ER4116DM resonant cavity. Cryogenic temperatures were maintained using an Oxford Instruments ESR 910 liquid helium flow cryostat in conjunction with a VC30 controller. Frequency calibration was with reference to diphenylpicrylhydrazyl.

RESULTS
Enzyme Kinetics-The enzymatic activity ( Fig. 1) of cytochrome c oxidase (50 nM) was assayed photometrically by measuring the rate of oxidation of reduced cytochrome c in the presence of excess oxygen (control, q) as described previously (17). The presence of a small excess of cyanide (100 nM) was observed to inhibit the measured activity greatly (f), and the presence of NO (1 M) was also observed to inhibit the activity, but less so (OE). However, when NO was added to the assay mixture in the presence of cyanide (ࡗ), a net increase in the enzyme activity was attained compared with the experiment with cyanide alone. That is, with both ''inhibitors'' present, the effect of NO was phenomenologically opposite to that of cyanide, reversing the cyanide inhibition. This is a very surprising result because the effects of two inhibitors competitive with oxygen binding are expected to be additive and would be predicted to yield results as indicated by the broken line in Fig. 1.
In fact, we made the effort to verify, at least in qualitative terms, that the net inhibition observed with 100 nM cyanide plus 1 M carbon monoxide did indeed follow this simple expectation (not shown), although obtaining good quality linear data at the resulting very high levels of inhibition was challenging.
It should be noted additionally that under these particular reaction conditions, biphasic kinetics were observed in the case of the cyanide plus NO experiments (ࡗ), with the initial rates comparable with those obtained in the presence of NO only and the later phase (not shown) identical to that observed with only cyanide present. The results of these turnover experiments strongly suggest that NO displaces cyanide during reaction cycles with oxygen. Also, it is clear that the NO is not simply acting as a reversibly bound competitive inhibitor, because if it were, the inhibitory effects of NO and cyanide would be additive. Rather, NO seems to be consumed during the reaction, in keeping with the earlier findings of several groups (2,17,29) who have shown that the reaction between cytochrome c oxidase and NO leads to production of nitrite under a variety of conditions. We have suggested previously (17) one way in which this process can also support oxygen consumption (NO oxidase activity, see ''Discussion''). Cyanide is not consumed during turnover, merely temporarily displaced, with the system returning to cyanide-inhibited kinetics once the available NO has been oxidized.
EPR Spectroscopy-A series of EPR experiments was undertaken to further examine the ability of NO to displace cyanide from the binuclear, oxygen-binding site of the enzyme. It has been observed previously that NO undergoes a variety of reactions with the resting oxidized and reduced forms of ''dimeric'' cytochrome c oxidase (i.e. preparations containing enzyme aggregates with Ͼ20 subunits, rather than the 13 minimally required). The nitrosylferrocytochrome a 3 (g ϭ 2.09, 2.00, 1.97, and 9-line hyperfine structure on g ϭ 2.00, A NO ϭ 21 G, A His ϭ 7 G) resulting from the reduction of Hartzell-Beinert enzyme preparations in Tween 80 detergent incubated with NO have been investigated intensely (19,20). Studies using 15 NO and the yeast enzyme from cells grown in 15 N-labeled histidine-rich media (18,21) demonstrated conclusively that (i) the NO was bound to ferrocytochrome a 3 and (ii) the EPR signal was further split by the presence of the proximal histidine ligand of the enzyme.
The EPR spectrum obtained immediately following the addition of NO (1 atm) to a reduced sample of the current pulsed oxidase preparation is shown in Fig. 2A. Unlike the previously reported EPR spectrum of nitrosylferrocytochrome a 3 derived from other preparations dispersed in Tween 80, the present data consist of overlapping spectra containing both 3-line (A ϭ 18 G) and 9-line (A ϳ 20 G and A ϳ 7 G) hyperfine features. The two sets of signals clearly indicate a mixture of at least two products; one is the 6-coordinate species described above and the other (minority species) a new 5-coordinate derivative (see below).
It has been shown repeatedly by magnetic circular dichroism spectroscopy (30 -33) that in the presence of excess reductant, cyanide-bound heme a 3 is 6-coordinate and remains in the ferric form, whereas the other metal ion centers become reduced. Almost certainly, this partially reduced cyanide adduct is the major inhibited form in turnover experiments where cyanide is present. The EPR spectrum of the partially reduced cyanide adduct (10 mM in sodium cyanide plus excess sodium dithionite) is broad but readily detectable at 15 K (Fig. 2B,  broken line). Treatment of this inhibited species with NO (1 atm) results in the quantitative formation of a new derivative exhibiting a characteristic EPR signal containing a distinct 3-line hyperfine pattern (A ϭ 18 G) (Fig. 2B, solid line). This spectrum is predicted in the case of a single electron spin coupled to the three allowed orientations (M I ϭ Ϫ1, 0, ϩ1) of a nucleus like 14 N with nuclear spin I N ϭ 1 and is sometimes referred to as ''superhyperfine'' structure if it is known that the electron and nuclear spins are not mainly localized on the same atom. We have also obtained qualitatively similar results starting from preparations of the dimeric resting enzyme dispersed in Tween 80 detergent (not shown); but in this case, the derivative exhibiting the 3-line hyperfine signals appeared to be only one of at least two species formed. The nitrosyl complexes of ferrohemes have been shown to have 5-coordinate structures in the absence of other strong-field ligands by elemental analysis of pure crystalline model compounds (34 -36) and a structure determination by x-ray crystallography (34). Moreover, complexes of this type (NO-bound, d 7 , 5-coordinate hemes) have been demonstrated unequivocally to exhibit distinct 3-line hyperfine EPR spectra like those shown in Fig. 2 (37,38). To better illustrate the point, the EPR spectrum of the nitrosyl adduct of extracted ferroheme a is shown in Fig. 2C (solid line), and also the spectrum of the analogous enzyme derivative has been superimposed (broken line) for comparison. It is to be stressed that the 5-coordinate nature of the complex is unambiguously demonstrated by the nitrosylferroheme a EPR spectrum of Fig. 2C. If two axial nitric oxide ligands were bound, they would either couple antiferromagnetically, resulting in no EPR signal, or couple ferromagnetically, leading to a more complicated signature than the observed 3-line hyperfine. The effect of introducing a sixth nitrogenous ligand (e.g. pyridine) on the EPR spectrum of the extracted nitrosylferroheme a EPR spectrum is shown in Fig. 2D. Clearly, the spectrum of the nitrosyl enzyme derivative formed in the presence of cyanide does not look at all like the 6-coordinate case.
In summary, the EPR data convincingly demonstrate that substitution of cyanide by NO at heme a 3 in the partially reduced cyanide adduct (starting with monomeric pulsed enzyme) renders the heme a 3 site 5-coordinate (Fig. 2B). To verify that the enzyme spectrum of Fig. 2B (solid line) represents a species in which cyanide is no longer bound to heme a 3 after NO addition, samples made with 13 C-labeled cyanide were examined and found to contain no additional hyperfine features (not shown), confirming the displacement of the cyanide ion. The overall similarity of the 3-line hyperfine features of Fig. 2, A and B, indicates that, contrary to the earlier findings of others, a fraction of the heme a 3 can also be pentacoordinate in the fully reduced derivative prepared in the absence of cyanide. These results suggest that the histidine ligand proximal to heme a 3 is more labile than previously thought, indicating some conformational flexibility at the binuclear pair. In support of this observation, it is noteworthy that the proximal ligand-to-heme a 3 bond length in the crystal structure of the Thermus thermophilus enzyme is significantly longer than in other reported structures (39).
Electronic Absorption Spectroscopy-In Fig. 3A the electronic absorption spectra of two NO adducts of cytochrome c oxidase are shown. The first is formed by the addition of NO to the dithionite-reduced enzyme (dashed line) and the second by the addition of NO to the partially reduced cyanide adduct (solid line). In the case of the second species, the order in which cyanide, NO, and reducing agent are added is not important, as the same derivative was always obtained (result confirmed by EPR spectroscopy; data not shown). These data appear to be in agreement with the EPR spectra of Fig. 2, A and B; that is, in the presence and absence of cyanide somewhat different NO adducts are obtained.
Because NO is converted to nitrite by cytochrome c oxidase (2,17) it was important to establish whether any of the species formed in the presence of excess NO were nitrite adducts. As shown in Fig. 3B, the Soret maximum shifts from 424 nm in the pulsed enzyme (solid trace) to 419 nm (dotted trace) following the addition of excess sodium nitrite. Note that neither of these is the same Soret maximum found associated with the monomeric resting enzyme (421 nm, dashed line). Furthermore, we did not observe the appearance of a g ϭ 12 EPR signal (not shown) upon formation of the nitrite adduct. In fact, when nitrite is added to the pulsed enzyme, some reduction of nitrite to NO occurs, and a very small amount of nitrosylferrocytochrome a 3 appears in the EPR spectrum. Superficially, the absorption spectrum of the nitrite adduct resembles that of the dimeric resting cytochrome c oxidase in Tween 80 (Soret maximum 417-418 nm, not shown), but these are clearly not the same species. More importantly, there is no evidence in the spectra of Fig. 3A, where the Soret features are at 431 and 442 nm, for a shoulder in the vicinity of 419 nm. Consequently, none of the species formed in the presence of excess NO seem to be simple nitrite adducts.
It is important to bear in mind that the Soret maxima of cytochrome c oxidase derivatives are subject to distortion by the presence of heme contaminants. Specifically, complex III of the electron transport chain can be an especially difficult im- purity to eliminate (40). The pyridine hemochrome spectrum (28) of heme a extracted from the current enzyme preparations is shown in Fig. 3C. The presence of the visible region band at 588 nm establishes that the heme a macrocycle was not derivatized during the extraction procedure. More importantly, the absence of any peaks at 557 and 550 nm confirms that the preparations contained no significant amounts of heme b or heme c, respectively (we estimate Ͻ3% of these relative to heme a). The Soret maxima we document herein have, of course, all been determined employing preparations essentially free of other heme-containing impurities.

DISCUSSION
Mechanism of NO Turnover-In conjunction with the findings of others, the present results have implications concerning the likely mechanism(s) of NO turnover by cytochrome c oxidase. Sarti et al. (4,6) have recently proposed the existence of two competing mechanisms for the reaction of NO with the enzyme, one operating at low electron flux and the other at high electron flux. The experiments we report here all correspond to (or exceed in terms of excess reductant) the high electron flux conditions of the earlier authors (i.e. 50 nM enzyme and 18 M cytochrome c). We have no basis in the current data for any opinion regarding the proposed low electron flux reaction, and so this need not concern us further.
Sarti et al. (4,6) have demonstrated unequivocally that the rate-determining step in the high electron flux reaction of NO with cytochrome c oxidase is photosensitive; that is, dissociation of NO from ferroheme a 3 is rate-limiting. However, the same authors assume (apparently) that the dissociated NO then diffuses out of the active-site pocket to be replaced by O 2 , and the uninhibited reaction proceeds. They did provide additional data showing that the presence of an NO scavenger further overcame NO inhibition of the enzyme, but this merely demonstrates the removal of free NO from solution, not its elimination from the active-site pocket. On the other hand, the present data (Fig. 1) show that at high electron flux and in the presence of cyanide, NO is not simply acting as a reversibly bound competitive inhibitor, because if it were, the inhibitory effects of NO and cyanide would be additive. Moreover, we have previously shown that cytochrome c oxidase out-competes oxymyoglobin for available NO (17), consistent with the rapid formation of nitrosylferroheme a 3 (4, 6) but leading to essentially quantitative production of nitrite (17).
In Fig. 4 we present an alternate reaction scheme for the high electron flux process that we think satisfactorily accounts for the observations from all laboratories. The present work and that of others clearly support a marked preference of NO for the heme site in variously prepared derivatives, and in transient kinetics experiments the inhibited state is thought to be associated with the formation of nitrosylferroheme a 3 (5,16). Therefore, in the chain of events where NO enters the active site pocket before O 2 , the available evidence strongly suggests that it will bind to heme a 3 (I). We have previously shown that the nitrosylferroheme a 3 derivative reacts with dioxygen in the presence of an additional electron donor to produce nitrite (17), perhaps via a peroxynitrite intermediate (41). Accordingly, any O 2 entering the active site can reasonably be proposed to acquire an electron from Cu B to form superoxide ion (O 2 Ϫ ) (II). Following rate-limiting dissociation of the NO (but not its loss from the pocket (III)), the very rapid reaction between NO and O 2 Ϫ to form peroxynitrite (ONO 2 Ϫ ) can take place (IV). The ONO 2 Ϫ formed at the active site will then quickly abstract two additional electrons from the enzyme, ejecting NO 2 Ϫ and water. Under appropriately controlled conditions in vitro, NO may certainly react with several of the intermediates in the normal catalytic cycle of cytochrome c oxidase (16). However, as the enzyme exhibits a marked tendency to react preferentially with NO compared with O 2 (5), otherwise plausible inhibitory reaction sequences at high electron flux where O 2 enters the activesite pocket before NO are likely minority reactions. We have previously shown in single cultured cardiomyocytes, the very rapid and essentially quantitative catabolic conversion of en- dogenously generated NO to nitrite (17). Moreover, at endogenously generated levels of NO, there is no evidence for scavenging of NO by oxymyoglobin in cardiac tissue (42). These two observations are in keeping with a transient (not reversible) inhibition of cytochrome c oxidase, leading to quantitative production of nitrite in situ, with essentially zero release of NO from the active site of the enzyme. Also, as the oxygen concentration in mitochondria in vivo is a few micromolar at most (43), experiments at high electron flux (a large excess of reductant) are likely relevant to the in vivo situation. Therefore, on the basis of the presently available information, it is our opinion that the putative reaction scheme of Fig. 4 is a plausible candidate for that which operates in vivo when cytochrome c oxidase turns over in the presence of endogenously generated levels of NO.
NO Binding to Cu B ?-The present data provide only evidence for NO bound to heme a 3 and not Cu B under reducing conditions (Fig. 2B). This requires some further comment, as NO is generally thought to be a good ligand for electron-rich metal ions such as both Fe 2ϩ and Cu 1ϩ rather than their common oxidized forms, Fe 3ϩ and Cu 2ϩ . The fact that we were unable to readily make NO adducts of Cu B in its reduced state is, therefore, intriguing. The tenacity with which NO seeks to displace cyanide from heme a 3 in its Fe 3ϩ state rather than bind Cu B 1ϩ both under turnover conditions (Fig. 1) and when NO is added to the partially reduced cyanide adduct (Fig. 2B) is even more remarkable. It seems that there is either some feature of the Cu B site that blocks NO binding, or the strength of the interaction between heme a 3 and NO results in Cu B losing any competition for the ligand. Experiments with reduced forms of the dimeric bovine NO adduct (44,45) have demonstrated that NO can be photodissociated from heme a 3 and, in one case (45), transferred to Cu B . However, there appears to be a consensual view among all authors that complexes of NO with reduced Cu B do not form under non-photolytic ambient conditions. It has been suggested (46, 47) that reduced Cu B may transfer an electron to a modified tyrosine close by in the active site. This would be expected to leave the nominally reduced Cu B with considerable cupric character and might afford an explanation for the lack of reactivity with NO. Such a suggestion is not necessarily in conflict with the observation that in the oxidized enzyme Cu B can seemingly exchange couple with NO (19,20); because this simply requires the two spins to be in close proximity, there need be no bond between NO and the Cu B 2ϩ . Although the available data indicate a marked preference for the addition of NO to reduced heme a 3 , this does not mean that electron transfer from Cu B 1ϩ to either NO, or O 2 in the active site need necessarily be slow, as outer-sphere processes can be very fast over short distances.
Cyanide Resistance and Toxicity-The presence in cells of an endogenous antidote to cyanide poisoning is probably of some ancient survival value. Inhalation of smoke from burning wood, other plant-derived fuels, and tobacco leads to the absorption of two particularly effective respiratory poisons, carbon monoxide and hydrogen cyanide. However, neither of these toxins inhibits cerebral O 2 consumption when delivered at low levels (48). The bloodstream provides a significant barrier to carbon monoxide diffusing to the tissues by formation of carbonylhemoglobin. However, less than 1% of the total hemoglobin in blood is normally present as the cyanide-binding methemoglobin. Consequently, the bloodstream is less able to protect the tissues from cyanide exposure. The present observation of facile cyanide displacement from cytochrome c oxidase by NO would seem to have at least two additional practical consequences, the first of some possible experimental significance and the second relating to clinical issues.
Inhibition of mitochondrial electron transport by the addition of cyanide is a standard protocol for establishing the basal O 2 consumption in studies with respirometers (49). In practice, there is usually some residual O 2 reduction measurable under these conditions. This observation can now be understood to be due, in part, to the endogenous NO production present in many samples. Also, there is a cyanide-resistant terminal oxidase known to be present in the mitochondria of all plants, many fungi, and some protozoa, but it is seemingly absent from metazoa (50). This alternate oxidase is certainly not a cyto- For simplicity, only the reactions at the binuclear pair are explicitly considered. As NO is consumed and converted to nitrite, it should properly be thought of as an auxiliary substrate. In the absence of NO, turnover of the enzyme is limited by the rate of internal electron transfer to the binuclear pair. However, in the above scheme, NO dissociation from ferroheme a 3 (II 3 III) is suggested to be the rate-limiting step in keeping with the findings of others (4). The NO-inhibited structure (I) could be 5-coordinate, 6-coordinate, or a mixture, as indicated by the EPR spectra of Fig. 2. Conversion of peroxynitrite to nitrite and water (IV3) is a known facile reaction of the enzyme that appears to sustain the electron transport chain (41). chrome and, based upon sequencing data, is thought to contain a di-iron active site (51), although crystallographic or spectroscopic verification of this assertion has proven elusive. In light of the activity data shown in Fig. 1 and the recently verified existence of mitochondrial nitric-oxide synthase (9), there is scope for erroneously ascribing any cyanide-resistant oxidase activity measured in mammalian mitochondria to the presence of an alternate di-iron oxidase if steps are not taken to inhibit NO production.
The standard treatment for cyanide poisoning involves the intravenous injection of sodium nitrite and sodium thiosulfate (52). The presumption has been that the nitrite oxidizes hemoglobin to methemoglobin, which may then strongly bind cyanide, whereas the thiosulfate promotes enzymatic conversion of cyanide to thiocyanate ion, a much less toxic compound, in the mitochondria. However, these arguments do not really explain how inhibition of cytochrome c oxidase by cyanide, an extremely facile reaction in vitro, is substantially avoided during acute poisoning. A study of murine brain mitochondria showed that lethal doses of potassium cyanide only inhibited cytochrome c oxidase activity by ϳ50%, and it was suggested that a large proportion of the oxidase activity may be a ''functional reserve'' (53). Moreover, the kidneys of cyanide poisoning victims have been shown to exhibit no significant irreversible damage and, therefore, have been suggested to be suitable for use as allografts (54,55). The results shown in Figs. 1 and 2B strongly suggest that mitochondrial NO could function to protect cytochrome c oxidase activity, displacing any bound cyanide, which may then be relatively slowly converted to thiocyanate or may diffuse to the vasculature to be complexed by methemoglobin. Furthermore, it may also be the case that production of additional NO derived from the administered nitrite ion, boosting the availability of the auxiliary substrate for the terminal oxidase, is a previously unrecognized aspect of the standard therapy. These ideas would seem to clarify the otherwise puzzling observations that the benefits of nitrites in treating cyanide poisonings cannot be due solely to generation of methemoglobin and that vasodilation is somehow involved in the mechanism (56).
Estimation of in Vivo NO Levels-Under physiologically relevant conditions, regulation of oxygen consumption by the normally functioning mitochondrial electron transport chain depends upon the prevailing [O 2 ]/[NO] ratio (57). It is therefore necessary to know (or have reliable estimates of) the steadystate levels of both O 2 and NO in tissue if reasonable assessments are to be made of the rate of mitochondrial respiration and the extent of secondary reactions producing reactive oxygen and/or nitrogen species in vivo. In the case of isolated rat heart and liver mitochondria, Boveris et al. (57)  Consequently, although many mitochondria-rich cultured cell lines survive well at 100 nM NO, the same types of cell may very well not survive in vivo at such high NO concentrations, and this probably contributes to the necrosis evident in inflamed tissue.
In the absence of any inducible NO synthase up-regulation, neither cultured cardiomyocytes (or other cell lines) nor intact endothelium produce measurable levels of NO until a stimulus has been applied (9). The detection limit of the electrodes used for such measurements is ϳ1 nM, and consequently, basal production of NO in all individual cell types studied thus far leads to NO concentrations of Յ1 nM. If these conditions hold in vivo, [O 2 ]/[NO] ϳ 10 3 , and mitochondrial respiration will not be significantly inhibited. Following stimulation, NO levels can rapidly, but transiently, rise to several hundred nanomolar (9), where [O 2 ]/[NO] ϳ10, and mitochondrial respiration will be more or less fully inhibited. It follows that the system appears to have evolved in such a way as to allow respiration to be turned off (or substantially inhibited) for ϳ1 s in a single cell and then recover. Alternately, at a reduced level of stimulation, NO release can probably be sustained for many seconds before a recovery period, but this will result in NO concentrations much closer to the basal level than the ϳ1 s duration transient maximum. The latter situation likely exists most of the time in vivo under normal physiological conditions. Accordingly, in cells/tissues of high mitochondrial content, where NO is efficiently catabolized, one can reasonably anticipate the steady state [O 2 ]/[NO] ratio to be ϳ10 3 .
The idea that cytochrome c oxidase catabolizes NO, which is strongly supported by the current findings, suggests that the functioning electron transport chain serves to limit nitrosative stress in mitochondria-rich tissues. It has been argued (43) that about 70% of the NO flux in liver mitochondria is converted to ''free'' peroxynitrite, a potentially damaging oxidizing and nitrating agent. However, a steady-state NO level of 50 nM was assumed, and the conversion of NO to nitrite by cytochrome c oxidase was not included in the analysis. In short, the amount of NO available for conversion to peroxynitrite was probably overestimated. Furthermore, we have previously shown (41) that cytochrome c oxidase is a reasonable peroxynitrite reductase (k 2 ϳ10 6 M Ϫ1 s Ϫ1 ) in the absence of excess O 2 and/or NO. As the in vivo O 2 consumption of mitochondria corresponds to about half-maximal activity of the terminal oxidase (58), peroxynitrite should be scavenged efficiently. In our opinion, if these additional observations are taken into account, it can plausibly be argued that the peroxynitrite output of functioning mitochondria under normal physiological conditions is virtually zero.