Kinetics of NO Ligation with Nitric-oxide Synthase by Flash Photolysis and Stopped-flow Spectrophotometry*

Nitric-oxide synthase (NOS) catalyzes conversion of l-arginine to nitric oxide, which subsequently stimulates a host of physiological processes. Prior work suggests that NOS is inhibited by NO, providing opportunities for autoregulation. This contribution reports that NO reacts rapidly (ka ≅ 2 × 107 m −1 s−1) with neuronal NOS in both its ferric and ferrous oxidation states. Association kinetics are almost unaffected by l -arginine or the cofactor tetrahydrobiopterin. There is no evidence for the distinct two phases previously reported for association kinetics of CO. Small amounts of geminate recombination of NO trapped in a protein pocket can be observed over nanoseconds, and a much larger amount is inferred to take place at picosecond time scales. Dissociation rates are also very fast from the ferric form, in the neighborhood of 50 s−1, when measured by extrapolating association rates to the zero NO concentration limit. Scavenging experiments give dissociation rate constants more than an order of magnitude slower: still quite fast. For the ferrous species, extrapolation is not distinguishable from zero, while scavenging experiments give a dissociation rate constant near 10−4 s−1. Implications of these results for interactions near the heme binding site are discussed.

Nitric-oxide synthases (NOS) 1 constitute a family of heme proteins that catalyze conversion of L-arginine to citrulline and nitric oxide (1). The production of nitric oxide in specific cell types fulfills certain physiological roles for which each isoform is suited according to its structure and regulation (2). Neuronal NOS ϩ (nNOS) (1) isoforms are localized in brain and in skeletal myotendinous junctions for the production of NO as a neurotransmitter. Inducible NOS (iNOS), first found in macrophages and induced by cytokines, produces NO for cytotoxic action. Endothelial NOS (eNOS) produces NO as a vasodilator. It was reported some years ago that NOS activity is inhibited by diatomic ligands such as carbon monoxide (3)(4)(5), cyanide (3), and nitric oxide (3). Although CO inhibition of NOS would be due to ligand binding with the ferrous heme intermediate, the situation with NO is more complex. Griscavage et al. demonstrated that NO inhibits and probably provides negative feedback for all three isoforms: iNOS (6), eNOS (7), and nNOS (8).
The last of those studies made a particularly convincing case that NO is a powerful inhibitor and acts by interacting directly with the ferric heme intermediate (8).
We previously reported the kinetics of CO association with nNOS and its expressed heme domain as well as dissociation from the nNOS-CO complex (9). From the rates, we estimated an equilibrium constant suggesting that in vivo inhibition of NOS by CO is unlikely for plausible physiological concentrations, unless there is great local enhancement of CO concentration, as could possibly be created by heme oxygenase activity. Nevertheless, in vitro studies of inhibition of NOS activity by CO are important for elucidating the reaction mechanism of NOS activity. Reported here are similar studies of the reaction of NO with both ferric and ferrous nNOS. These provide additional insights into the mechanism of reactivity in general, bear on the suggestions of Griscavage et al. (8) that NO autoregulates NOS, and have direct consequences for in vivo NOS activity as affected by continuous changes in local NO concentrations.
It is relevant that eNOS is a membrane-bound enzyme, whose activity is considerably increased by shear stress due to blood flow. The functional significance of eNOS association with plasma membrane is poorly understood. Perhaps blood flow modulates eNOS activity by, among other mechanisms, continuously depleting local NO concentrations. An optimum regulatory role for eNOS, or any isoform of NOS, would seem to require finely tuned affinity for NO.
In the much studied hemoglobins and myoglobins, there is a coordinate covalent bond between the heme iron and a proximal histidine; and the iron is exclusively Fe(II). In NOS, the bond is with the thiolate function of a cysteine residue; and the heme iron cycles between its ferric (ϩ3) and ferrous (ϩ2) oxidation states during catalytic cycles (10). One might conjecture that changing the proximal ligand would have significant effects. Proximal Cys favors the Fe(III) state and seems to make the ligation kinetics of ferric and ferrous species more similar, but heme proteins with either proximal base exhibit wide variation in rate constants. Tuning of the reactivity in heme proteins is achieved by varying the electrostatic and steric environment in the neighborhood of the iron, both at the proximal bond and on the opposite, or distal, side.
Association rate constants of NO with heme proteins range from ϳ2 ϫ 10 7 to ϳ200 M Ϫ1 s Ϫ1 , and the dissociation rate constants range from 100 to 0.03 s Ϫ1 for the ferric species and at least 1000-fold slower yet for the ferrous form (11,12). The fastest association rates, which are similar for a large number of heme proteins, are limited by the rate of diffusion of the ligand through solvent to the protein and entry into the protein, after which bonding is very efficient. Apparently, entry of ligands into the protein and access to the iron is quite similar for many proteins. The lower association rate constants apply for proteins having substantial steric hindrance to bonding or when there is a need to displace water or some other ligand. The slow association in horse heart Cyt c is attributed to the need to displace a methionine sulfur (11). Like other heme proteins, NOS has charged or polar amino acid side chains in the distal pocket, which might stabilize or destabilize coordination of a ligand, such as H 2 O (13). Ligand dissociation rate constants are also affected by factors that stabilize ligand bonding. For example, NO dissociates very slowly in many Mb and Hb, because it is stabilized by a particular distal His. In contrast, NO dissociates is much more rapidly from Mb(EL) (Asian elephant), which lacks the distal histidine (14). Effects of distal histidines have been examined recently in great detail using site-directed mutations in Mb (15,16). The overall mechanisms proposed (10,17,18) for NOS-catalyzed conversion of L-Arg to citrulline specify both an interaction of L-Arg or N-hydroxyarginine bound to the enzyme in the distal heme pocket with a ligand (a reduced oxygen species, in this case) coordinated to the iron, as well as formation of hydroxyheme (NOS ϩ -H 2 O) just before release of NO in the last step. Furthermore, it has been speculated that a coordinated water molecule in the hydroxy intermediate of NOS may be responsible for slowing down NO ligation long enough for the NO to diffuse away from the binding site and, consequently, minimizes self-inactivation of NOS.
For ferrous heme proteins, affinity of NO is usually extremely high, so that NO binds more strongly than other ligands, such as O 2 , and often acts as a "poison." One significant exception is Cyt c, in which the affinity of the ferrous species for NO is quite low and only 20-fold greater than that of its ferric counterpart (11). This is due to an exceptionally small association rate, only 8.3 M Ϫ1 s Ϫ1 (11). In a different, recently described case, that of soluble guanylate cyclase in the presence of substrate GTP, low affinity is achieved in the opposite manner, i.e. primarily by unusually rapid dissociation (19). Since both guanylate cyclase and NOS are involved in the NO regulatory cycle, it is appropriate that they both be regulated by NO but not poisoned by very high affinity for NO. So, questions for NOS are as follows. Is affinity for NO reasonably low? Does affinity differ between ferric and ferrous oxidation states? Is low affinity achieved by slow association or by fast dissociation? Are the kinetics or the equilibria affected by cofactors or substrate? In the studies reported here, we answer those questions by determining the kinetics of NO ligation with both ferric and ferrous derivatives of nNOS holoenzyme and its heme domain (residues 1-714). We investigated the effects of tetrahydrobiopterin (BH 4 ) and L-arginine. From the bimolecular rate constants, we could derive an equilibrium constant for NO binding to nNOS.

MATERIALS AND METHODS
Nitric oxide, high purity argon, and premixed NO diluted in argon were all obtained from Matheson. Nitric oxide was further purified by passage through aqueous KOH solution. Buffers (20 mM Tris-HCl, pH 7.8, 100 M EDTA, 100 mM NaCl) were prepared in a gas-tight syringe and deoxygenated by bubbling with argon for at least 40 min. When indicated, buffers were prepared with 250 M BH 4 and/or 10 mM Arg.
Rat nNOS holoenzyme was expressed in Escherichia coli and purified and reconstituted as described previously (20). The amino-terminal heme-binding domain (residues 1-714) was also rat nNOS expressed in E. coli and purified as reported previously (21). Assay for activity and other characterization and controls were as reported previously (9). Proteins were degassed very gently by blowing argon over the solution in a cuvette with a large surface-to-volume ratio. For studies of the ferrous oxidation state, protein solutions were reduced by adding small amounts of 100 M sodium dithionite. Sample solutions were prepared by vigorously bubbling premixed NO diluted in argon through degassed buffer to obtain the desired [NO] and then adding small amounts of degassed concentrated protein solution. Measurements were made at 23 Ϯ 0.5°C.
Kinetic measurements were carried out as described previously (9). For the laser flash photolysis, 3 mJ in 4 ns at 550 nm was used for excitation over an area of 0.1 cm 2 collinear with the probe light. Protein concentrations were 10 M for the heme domain and 5 M for holoenzyme, and the path length was 1 cm. The probe light bandwidth was 8 nm. Digitization was by a Lecroy 9361 oscilloscope. Typically, 200 -1000 laser "shots" were averaged for each run. Kinetics were always characterized at two different wavelengths, as a check that they gave similar results, one near 445 nm corresponding to bleaching of the original solution and the other near 405 nm corresponding to maximum transient absorption. Preliminary measurements were made from 390 nm to 450 nm to search for any anomalous behavior, but none was found. The isosbestic region near 420 nm was examined to confirm that only two states were involved, at least within available resolution. Steady state absorption was monitored before and after laser irradiation in order to verify that there was very little photodegradation. Statistical noise was reduced to ␦⌬A rms ϭ 0.00002, typically. This was necessary because the low quantum yield for escape of dissociated NO from the protein resulted in small transient absorbance changes (at best, ⌬A(t ϭ 0) Ϸ 0.01). A more extensive discussion of signal-to-noise issues has been given elsewhere (9).
Kinetics of NO dissociation from ferric nNOS-NO were determined using a Durrum stopped-flow instrument. NO dissociation was monitored at 412 nm during reaction with oxymyoglobin, which is converted to metmyoglobin by reaction with NO, a method we used recently for another protein (22). Trapping of free NO by MbO 2 is fast and irreversible. Four different concentrations of scavenger were used in order to establish that the absorption changes were independent of scavenger concentration and, therefore, rate-limited by NO dissociation. Dissociation from ferrous nNOS-NO is so slow that it could be measured in an ordinary spectrophotometer, utilizing sodium dithionite (SDT) as a scavenger, in the presence of CO, and monitoring absorbance change at 420 nm (23). Concentrations of some components are slightly different for the dissociation measurements than for photolysis; they are listed in figure captions.

RESULTS
Definition of the Rate Constants-Rate constants for bimolecular association are defined in Reaction 1.
Photolysis perturbs the equilibrium and creates a temporary excess population on the right. Flooding the sample with excess NO ensures pseudo-first-order conditions under which relaxation back to equilibrium is described by Equation 1.
When k obs is plotted versus [NO], the slope and intercept determine k a and k d . We used 100% NO (2000 M) and certified mixtures of 10%, 0.5%, and 0.1% NO in Ar in order to ensure known [NO]. In all cases, there was good linear behavior extrapolating close to the origin. Since association was quite fast, we needed 0.5% and 0.1% NO mixtures to establish the intercept. Repeated measurements verified that the y intercept was measurably above the origin for ferric nNOS ϩ -NO. For ferrous nNOS-NO, the intercept was indistinguishable from the origin and could be measured only by a mixing experiment. The overall bimolecular rate constants k a and k d for a variety of conditions are collected in Tables I and II. An observation that is only semiquantitative, but nonetheless important, is that the net photodissociation yield for ligands to leave the protein was quite low, roughly 1%. Even this low value is significantly higher than for Mb-NO, but small compared with photodissociation yields for Mb with most other ligands. Consistent with the small photodissociation yield, we also observed geminate recombination. It was identified by the fact that the process was independent of NO concentrations and occurred on much faster time scales than the bimolecular association process. The largest geminate recombination efficiency seemed to occur for the conditions that yield the fastest bimolecular association rates in the tables, as should be expected (24).
Heterogeneous Behavior?-An important question is whether the kinetics are well described by single exponentials, as prescribed by Reaction 1 and Equation 2. Fig. 1 shows that they are, unless a very good measurement is made. In the past, a noisier version of Fig. 1 would have been fit to a single exponential, at least over two or three half-lives. With today's technology, capable of better signal-to-noise ratios, it is not unusual to discover that multiple processes over a narrow range of k a are present, as is the case in Fig. 1. Data there can be fit over many half-lives with two exponentials differing in rate by a small factor of only 2-4. Since the rates are so close and the heterogeneity remains similar over all conditions reported here, we characterize the bimolecular association by a single, "effective" rate. It is calculated as the harmonic mean (reciprocal of the weighted mean of the reciprocals) of the two rates used in fitting (9). This gives a datum that is very close to what is obtained for the best fit of a single exponential to the same record, as long as the entire record is fit in both cases and not truncated by instrumental limitations or human bias.
The possibility that systematic distortions could be present in the detection apparatus is rejected because we do not see the same sort of heterogeneity for Mb. It could well be that the heme enzymes tolerate or even require a range of different protein conformations that the simple oxygen carriers do not, and this leads to a narrow distribution of k a ; however, at this point in time, it is difficult to rule out absolutely the possibility that the more complex isolation and purification procedures required for the enzymes might lead to slightly damaged protein that is avoided in the much simpler procedures used for the globins. The main point is that with precision capable of detecting such inconsequential heterogeneity even in the very difficult case of NO photolysis, with its low photodissociation yield, we can be very confident that no major feature is obscured by inadequate sensitivity.
Variation of k a -All the association rates for both ferrous and ferric nNOS are quite fast, in the neighborhood of 10 M Ϫ1 s Ϫ1 . In ferric nNOS ϩ , the NO association rate for the heme domain alone is consistently 2-3 times faster than that for the holoenzyme. Ferrous nNOS does not show this variation, and its k a have an intermediate value. For ferric nNOS ϩ , we were able to prepare protein with and without substrate, L-Arg, and cofactor, BH 4 . Any differences are very small. All of the holoenzyme cases lie within statistical scatter. For the heme domain, the difference with and without L-Arg is probably real, but just barely measurable. For ferrous nNOS-NO, solutions without arginine are unstable, a point of considerable significance in itself, and could not be characterized. Instability in nNOS has been discussed by others at some length (25), although that study put more emphasis on the role of BH 4 .
Once BH 4 is present, it is difficult to remove completely, so it is important that we started with protein prepared (21) so that BH 4 was never present, by employing recombination in E. coli that itself does not make BH 4 . It is also difficult to exclude arginine completely. The cases listed as lacking it may have trace amounts.
For comparison, we note that the common Hb and Mb proteins have k a that differ substantially between ferrous and ferric forms. Ferrous globins usually have k a roughly similar to what we find for nNOS; but the ferric globins have much smaller association rate constants.
Variation of k d -For ferric nNOS, the values derived from the intercept of k obs in the flash photolysis experiments are   near 50 s Ϫ1 , with a significantly higher rate constant only for the holoenzyme in the presence of L-Arg. There was scatter in measurements made during several repetitions on different days, as much as a factor of 2, but no observations anywhere near 10 s Ϫ1 , which would have been easy to detect. This presented a conundrum, because such a fast k d in conjunction with the k a already determined implies that very little NO should be bound at 0.1% (2 M) NO. Nonetheless, we were able to prepare and characterize such samples. Titration of nNOS by NO to determine K a predicted a k d that was a factor of 10 smaller than that determined in the photolysis experiments. Stoppedflow mixing experiments were then undertaken. They gave good signals, as illustrated in Figs. 2 and 3, that yielded values for k d that are smaller by a factor of 10 -20 than the flash photolysis results, but consistent with estimates from equilibrium titration.
For ferrous nNOS the flash data extrapolated close to the origin, implying slower k d and larger K a . The equilibrium association constant K a was, in fact, very large, beyond our ability to measure by titration, since we worry that NO is sufficiently reactive to cast some doubt on efforts to prepare very low [NO] with accuracy. We determined k d by trapping spontaneously released NO and found the small k d reported in the table. We have no way to guess whether the intercept in flash photolysis might actually be a factor of 10 higher than the k d obtained by scavenging, as there is no way to measure by extrapolation such a small departure from the origin. DISCUSSION A small photodissociation yield, as NOS-NO has, entails very efficient geminate recombination. The corollary is that almost every ligand that enters the protein will bond to the iron, so that k a is controlled by the time needed to reach the protein and pass through whatever "gate" exists at entry. This means that steric and electronic effects exercised at the heme iron binding site are not very effective in modulating bimolecular association k a . In contrast, the net ligand dissociation rate k d remains directly proportional to the rate of breaking the Fe-NO bond, albeit reduced by a constant factor dependent on the geminate recombination probability. Protein control of ligand affinity in the case of NO can be exerted effectively through the dissociation constant, utilizing steric and electronic variation on either proximal or distal sides of the iron. The situation is different for CO and any other situation with large photodissociation yield, in which variation near the iron immediately affects both k a and k d and, consequently, K a . Tables  I and II, rate constants for ligand association of NO are very fast for both ferric and ferrous nNOS derivatives and quite similar in magnitude. All of these values for k a are among the fastest association rate constants known for binding to heme iron in a protein, equal to or faster than rates constants common to many ferrous heme proteins and faster than all but a few known for ferric species. The fast association is a reflection of efficient geminate recombination, as discussed above. It also suggests that there is no water molecule nor any other ligand able to coordinate to the distal side of heme prior to NO addition after flash photolysis, which would have to be displaced before NO could return. The values of k a are close to those for Hb opossum ␣-chains or Mb elephant, neither of which have distal histidine nor a water molecule at the ligand binding site (14). The lack of interfering water is not due to lack of access for an H 2 O, which is smaller than O 2 or NO. Elephant Mb seems to allow access to the iron. The issue, rather, is stabilization of H 2 O by hydrogen bonding, which is not available in elephant Mb, which has the double substitution H64Q/L29F. Effects of distal histidines have been explored recently in great detail using site-directed mutations in Mb (15,16). The observed NO association and dissociation rate constants for H64A, H64V, H64T, and H64F metMb are all near 100 M Ϫ1 s Ϫ1 and 100 s Ϫ1 , respectively. These values are close to those for the ferrous forms of the same mutants. Eich et al. (15) also showed that the single H64Q mutation increases k a for NO binding to metMb from 0.07 to 8.2 M Ϫ1 s Ϫ1 . The fact that stabilization is also lacking in nNOS is interesting in that there are a number of charged residues in the substrate binding pocket (26,27), which might possibly have stabilized a water ligand, but apparently do not. All this may seem to be inconsistent with some indications of low spin hexacoordinated heme in the equilibrium deliganded state, which suggests a sixth coordinated ligand (28). A likely resolution could be slow incorporation of water, requiring times longer than the few seconds or less needed for ligand association after photolysis. Perhaps a slow conformational change is needed to stabilize water bonding. This would explain indications of hexacoordination at equilib-  Table II. rium. It could also explain the different results we obtained for two different methods of measuring k d .

Bimolecular Rate Constants k a and k d -According to
Rate constants in Table I for k a of the heme domain are consistently higher than for the holoenzyme by a factor of 3-4, which demonstrates that the active site is influenced in some manner by more distant parts of the protein.
There is a striking qualitative difference between association of NOS with CO and with NO, aside from any quantitative differences. Two independent studies (9,29) demonstrated that CO association exhibits two distinct phases separated in rate by a factor of about 100. Reaction with NO either has none of the slower phase or so little that we have not been able to resolve it. The major effect of cofactor, BH 4 , or substrate, L-Arg, on CO reactivity is to shift the relative contributions of the two phases. Those two agents have very little effect on NO reactivity, which is reasonable since there is only one phase present.
Several considerations: the proposed overall reaction mechanism, the recently determined (26) crystal structure for part of iNOS (residues 66 -498 of the oxygenase domain), as well as ENDOR studies of intact nNOS (30), all entail interaction between activated oxygen coordinated to heme and the guanidino group of L-Arg or N-hydroxy-Arg. The NO association rates do not reflect such interactions. The lack of sensitivity of association rates to L-Arg and BH 4 has consequences given our finding that the rate-limiting step for k a is diffusion to and entry into the protein. Diffusion should not be affected, and protein entry apparently is not affected by L-Arg or BH 4 .
Ligand dissociation rate constants are also unaffected by L-Arg or BH 4 . The k d for ferric nNOS ϩ , as determined by extrapolation of flash photolysis transient kinetics, all lie within a factor of 3 of each other and are very fast, suggesting that NO coordination has little stabilization by hydrogen bonding with distal residues. This was discussed above in the context of k a for different species and mutations of Mb. It has also been observed previously for catalase (11) and cytochrome c peroxidase (11), as confirmed in our own studies (12). However, k d for the ferric species are lower by a factor of 10 or more when determined by extrapolation by stopped-flow mixing with a scavenger. The latter values are closer to the value for ferric Hb A ϩ -NO, in which coordinated NO is stabilized somewhat by hydrogen bonding with distal histidine. A plausible explanation of the difference between flash photolysis and scavenging is the same one invoked above for another purpose: that following bimolecular addition of NO there is a structural change on a slow time scale that stabilizes NO binding. The hydrophilic residue nearest to the heme is Glu 592 . The analogous residue in iNOS, Glu 371 , although a little removed from the ligand binding site, is known to interact with L-Arg in the heme pocket (26). It has been found that L-Arg and the mechanismbased inactivator aminoguanidine bind in the distal pocket adjacent to the heme and interact directly with Glu 592 . We speculate that either substrate arginine or, in its absence, Glu 592 may be stabilizing NO binding via a hydrogen-bonded water molecule drawn into the pocket after initial NO ligation.
In the case of ferrous nNOS, extrapolation of observed association rates to zero nitric oxide concentration did not yield an intercept significantly different from zero, but that only means Ͻ10 s Ϫ1 . The k d measured by scavenging NO from ferrous nNOS-NO is near 10 Ϫ4 s Ϫ1 , and comparable to that in Mb-NO. It too is minimally affected by Arg or BH 4 . This suggests that there is no serious steric interaction between coordinated NO and arginine. The fact that electron transfer occurs between these centers does not present a contradiction. Tuning donors and acceptors to facilitate efficient electron transfer is not just a matter of having them as close as possible; rather, they are optimized carefully to control both forward and reverse trans-fer, and that frequently involves less than closest packing, in order to reduce back transfer.
Hurshman and Marletta (28) have reported that L-Arg decreases affinity of NO for ferric iNOS ϩ but not for ferrous iNOS. They also measured the inverse dependence, that the apparent K d for arginine binding to ferric iNOS ϩ varies over the range 60 -200 M, depending on NO concentration. For the reactant concentrations used in the present study, the kinetics of NO ligation to ferric nNOS ϩ showed no dependence on L-Arg. The affinity of NO for ferrous nNOS is so high that any affect of arginine is inconsequential.
Equilibrium K a -From the tables, we conclude that K a (NO-nNOS ϩ ) Х 10 5 M Ϫ1 s Ϫ1 within seconds after photolysis; K a (NO-nNOS ϩ ) Х 2 ϫ 10 6 for "real" equilibrium after some relaxation; K a (NO-nNOS) Х 4 ϫ 10 10 for "real" equilibrium after any possible relaxation. The fourth case, if it exists, was not measurable.
The answer to a question posed in the introduction is now clear. The affinity of NO for ferric nNOS is modest, due exclusively to rapid dissociation able to offset rapid association, which is actually quite fast. This keeps nNOS from being poisoned, but also allows rapid kinetic changes. The remaining question is how the enzyme escapes being inactivated by irreversible binding of NO to ferrous nNOS. The explanation probably lies in the observation that the ferrous nNOS-NO complex is stable only in the presence of L-Arg. In a sense, the protein shifts to its ferric form to offload NO. Due to the anionic nature of the proximal ligand (Cys 415 ), the higher oxidation state is favored in NOS. Although the presence of the positively charged guanidino group would tend to stabilize the ferrous oxidation state, if the rate of heme oxidation is faster than the second order combination of NO with ferrous NOS, then only very small amounts of ferrous NO will be formed during the enzyme turnover.
The affinity of NO with ferric nNOS ϩ is sufficiently high that in unstirred solutions high local concentrations of NO may cause self-inactivation, but not so high as to render the enzyme unresponsive to changes in physiologically plausible local concentrations of nitric oxide. Since the ferrous form of the enzyme is present only in very small amounts, and NO is produced while the enzyme is in the ferric state, one may appreciate the observation by Griscavage et al. that NO inactivates by binding to the ferric state of NOS (8).
Kinetics of NO (and CO) binding to iNOS were reported very recently by Abu-Soud et al. (31), after the experiments described here were complete. In some ways, their study was similar to ours: both compared the Fe(II) and Fe(III) species, and both investigated the effects of L-Arg and BH 4 . In other ways, the two studies were complementary: we studied nNOS, they studied iNOS; we compare the heme domain with the holoenzyme, they studied only the heme (oxygenase) domain; we worked at 23°C, they at 10°C; we used flash photolysis, they used stopped-flow mixing; we used both extrapolation and scavenging to measure dissociation, they report only association measurements; we began a characterization of geminate recombination, their mixing experiments will never address geminate recombination; we merely mention protein instability for certain conditions, while they pursue that issue at considerable length. The differences in methods and conditions suggest that a number of general conclusions reached in both studies are robust and probably apply to all NOS isoforms; our association rate constants are generally about an order of magnitude faster, but after a plausible correction for different temperatures, this implies that the association rate constants are fairly similar in nNOS and iNOS. Both studies agree that NO association is similar for Fe(II) and Fe(III) and that the effects of L-Arg and BH 4 are minimal. Both studies agree that the dissociation rate for Fe(III) as measured by extrapolating association kinetics is very high. The k d are numerically very close in the two studies, but this means either that they will diverge slightly after a temperature correction is applied (which is entirely possible since nNOS and iNOS are, after all, different), or that the activation energy for dissociation is surprisingly small. Abu-Soud et al. (31) were able to extrapolate association data for ferrous iNOS-NO and determine intercepts that were as fast for Fe(II) as for Fe(III). Our best estimates place our intercepts lower and close to zero, and our scavenging experiments determined a very small k d for ferrous nNOS-NO.
The general similarity between our nNOS and their iNOS is particularly intriguing in light of a study by Raman spectroscopy (32), whose authors found similarities between iNOS and eNOS but differences between the two constitutive isoforms eNOS and nNOS, which might have been expected to be the more similar pair. They explain the difference by postulating a more open pocket in nNOS. That should have implications for ligand binding kinetics, although they might be subtle. Since k a is rate-limited by protein entry, one might expect an "open pocket" to affect k a , but it does not. The situation with k d is unclear because of the lack of scavenging measurements on iNOS, but the only difference that seems like it may exist (faster k d in iNOS) is in the wrong direction; it should be nNOS that is faster, if it has the open pocket.
To summarize, we have shown that fast ligand association in nNOS ϩ combines with fast dissociation to keep affinities reasonable in the ferric form, but an additional mechanism (cycling between oxidation states) plays an important role in avoiding inactivation by strong NO binding to the ferrous state of nNOS.