FMN fluorescence in inducible NOS constructs reveals a series of conformational states involved in the reductase catalytic cycle


J. C. Salerno and D. K. Ghosh, Department of Biology, Kennesaw State University, MB #1202, 1000 Chastain Road, Kennesaw, GA 30144, USA
Fax: +770 423 6625
Tel: +770 423 6177


Nitric oxide synthases (NOSs) produce NO as a molecular signal in the nervous and cardiovascular systems and as a cytotoxin in the immune response. NO production in the constitutive isoforms is controlled by calmodulin regulation of electron transfer. In the tethered shuttle model for NOS reductase function, the FMN domain moves between NADPH dehydrogenase and oxygenase catalytic centers. Crystal structures of neuronal NOS reductase domain and homologs correspond to an ‘input state’, with FMN in close contact with FAD. We recently produced two domain ‘output state’ (oxyFMN) constructs showing calmodulin dependent FMN domain association with the oxygenase domain. FMN fluorescence is sensitive to enzyme conformation and calmodulin binding. The inducible NOS (iNOS) oxyFMN construct is more fluorescent than iNOS holoenzyme. The difference in steady state fluorescence is rationalized by the observation of a series of characteristic states in the two constructs, which we assign to FMN in different environments. OxyFMN and holoenzyme share open conformations with an average lifetime of ∼ 4.3 ns. The majority state in holoenzyme has a short lifetime of ∼ 90 ps, probably because of FAD–FMN interactions. In oxyFMN about 25–30% of the FMN is in a state with a lifetime of 0.9 ns, which we attribute to quenching by heme in the output state. Occupancy of the output state together with our previous kinetic results yields a heme edge to FMN distance estimate of 12–15 Å. These results indicate that FMN fluorescence is a valuable tool to study conformational states involved in the NOS reductase catalytic cycle.





endothelial NOS


inducible NOS


neuronal NOS


nitric oxide synthase


two-domain NOS construct in which only FMN and heme domains are present


Nitric oxide synthases (NOSs) function as signal generators in homeostasis in the cardiovascular system [1,2] and neurotransmission in the central nervous system [3–5], and produce NO as a cytotoxin in the immune response [6–8]. Synthesis of NO from l-arginine and two molecules of O2 requires three electrons derived from NADPH. Calmodulin (CaM) regulates NO synthesis by the endothelial (eNOS) and neuronal (nNOS) isoforms; CaM stimulates electron transfer through flavin cofactors to the heme catalytic site [9]. Inducible NOS (iNOS) is synthesized in response to cytokines, binds CaM at all physiological Ca2+ concentrations and is not regulated by Ca2 [10].

Reducing equivalents are supplied through reductase domains homologous to P450 reductase; NADPH, FAD and FMN binding domains are involved in electron transfer [9,11–16]. In addition to a canonical CaM binding site located between the N-terminal oxygenase domain and the FMN binding domain, control of electron transfer requires an auto-inhibitory element present as an insertion in the FMN binding domain and a C-terminal element that restricts electron transfer [17–20]. These elements are absent or reduced in iNOS.

It has become clear that electron transfer requires a concerted series of conformational changes, and we previously proposed a formal tethered shuttle mechanism in which the FMN binding domain dissociates from a reductase complex ‘input state’ and reorients to supply electrons to the oxygenase domain [11,21–24]. Recent work from other laboratories also supported this shuttle concept [e.g. 12,24–26].

CaM binding to nNOS [26–30] and eNOS holoprotein (unpublished observation) triggers increase in flavin fluorescence and in addition increases cytochrome c reductase activity in both isoforms [12,27–30]. We interpret this in terms of CaM activating the release of the FMN binding domain from the reductase complex and making FMN available as an electron donor for the oxygenase domain or cytochrome c [11,12,20–26]. On the other hand, since iNOS holoenzyme has CaM permanently bound, it is presumed to have FMN always available for electron donation to any electron acceptor including the oxygenase catalytic center or external electron acceptors like cytochrome c.

The recent characterization and availability of two-domain NOS constructs in which only FMN and heme domains are present (oxyFMN constructs) makes FMN fluorescence potentially a powerful and flexible probe of heme–isoalloxazine interaction in constructs without FAD–FMN interactions [21–26]. The response of the system to CaM is fundamentally different, because the majority state FAD–FMN dimer of holoenzyme is absent [31,32]. Because the high spin ferriheme of the oxygenase domain is capable of quenching the fluorescence of the isoalloxazine ring of the FMN binding domain when it is within electron tunneling distance, fluorescence spectroscopy provides a flexible method of probing domain interactions in NOSoxyFMN and can provide information that is inaccessible by EPR and kinetic approaches. A complete description of the system from a fluorescence perspective requires quantum yield and fluorescence lifetime information, but good preliminary data can be obtained from fluorescence intensity information. In the present study we use a murine iNOSoxyFMN construct as a representative for putative CaM activated NOS output state to study the CaM regulated interaction between oxygenase and FMN domains by measuring both lifetime and steady state FMN fluorescence.

The current paper describes a series of experiments with iNOS using flavin fluorescence as a probe of conformational states associated with the catalytic cycle. Obligatory conformational intermediates can be resolved by using appropriate constructs and logically assigning species with different lifetimes and their correlation to NOS catalytic cycles. The approach described here should allow us to advance the knowledge of catalysis and control of NOS enzymes by providing information about important conformational states. This will open new avenues for biophysical and structural analysis of this important enzyme, and in addition advance our understanding of other enzymes containing homologous reductase catalytic units.


It has been established for more than a decade that FMN fluorescence in NOS is increased by CaM binding or addition of chaotropes that weaken protein–protein interactions, indicating a conformational change associated with activation of electron transfer [25,32–34]. The results presented here use iNOS constructs to relate these fluorescence effects to transition between specific conformational states that are an obligatory part of the reductase catalytic cycle.

As shown in Fig. 1A, the fluorescence emission spectra of iNOS holoprotein and the iNOSoxyFMN construct are both dominated by a peak at 530 nm from FMN. The fluorescence intensity of the holoenzyme is only about 15% of that of the oxyFMN construct (based on total FMN), suggesting that the holoenzyme fluorescence is much more highly quenched. Figure 1B shows a Stern–Volmer plot for iNOS holoenzyme and oxyFMN flavin fluorescence. The plots are indistinguishable, suggesting that the dominant species in the steady state fluorescence of both systems has similar exposure to solvent.

Figure 1.

 (A) Flavin fluorescence emission spectra for iNOS holoenzyme and iNOSoxyFMN showing high intensity from the two-domain construct. (B) Stern–Volmer plot for iNOS holoenzyme and iNOSoxyFMN showing equivalent exposure of the dominant fluorescent species. The Stern–Volmer equation was used to analyze the quenching of the flavin fluorescence by KI as described in [48]: F0/= 1 + KSV[Q] in which F0 and F are the fluorescence intensities in the absence and presence of the quencher, [Q] is the concentration of the quencher and KSV is the Stern–Volmer quenching constant [48].

Although this seems paradoxical, it will be shown to be reasonable in light of the lifetime results. The dominant species in the fluorescence of both systems is a long lifetime state in which FMN is exposed to solvent. The lower fluorescence of holoenzyme is due to a smaller fractional occupation of this high fluorescence yield state.

Addition of highly concentrated oxyFMN to buffer, corresponding to a dilution from 100 μm to a few micromoles, caused a gradual fluorescence increase on a timescale of 10–20 min. This suggests that in concentrated solution oxyFMN forms aggregates that have reduced fluorescence. The fluorescence intensity at 530 nm doubles between 1 and 20 min, and the change is 90% complete at 10 min (data not shown), following an ∼ 4 min exponential.

EDTA, a powerful Ca2+ chelator, reduces the effectiveness of CaM constructs in promoting association of the FMN binding domain and oxygenase domain in iNOSoxyFMN (Fig. 2). After EDTA treatment of iNOSoxyFMN, the flavin fluorescence initially decreased by ∼ 10%, but stabilized at a higher intensity value (∼ 50% greater than baseline). The conversion of the low fluorescence state to the high fluorescence state again takes place on the timescale of minutes, following an ∼ 10 min exponential. This indicates that the effects of Ca2+ removal from the construct are multiphasic and slow, perhaps because of slow off rates for Ca2+ and/or CaM in the iNOS system. Because increases in flavin fluorescence suggest separation of the FMN binding domain from the quenching heme group of the oxygenase domain, these results together indicate that the Ca2+ chelating effect of EDTA disrupts the ability of Ca2+/CaM to stimulate the output state in iNOSoxyFMN where the association of the FMN binding domain to the oxygenase domain would lead to a decrease in flavin fluorescence intensity.

Figure 2.

 Effect of Ca2+ and EDTA treatment on FMN fluorescence in iNOSoxyFMN. The sample contains 2.5 μm iNOSoxyFMN in 40 mm BisTris propane containing 1 mm dithiothreitol and 10% glycerol. Dilution of concentrated enzyme into buffer requires ∼ 10 min incubation before full intensity is attained. The observed decrease immediately after EDTA treatment is rapid and is followed by a slow increase to a level significantly higher than observed without EDTA. A back titration with increasing concentrations of CaCl2 after 0.6 mm EDTA and 1.0 mm Ca2+ addition restored the initial level of fluorescence. These data are representative of three replicate experiments.

Addition of Ca2+ rapidly reverses the effect of EDTA on the iNOSoxyFMN system. Figure 2 shows the effect of EDTA addition followed by a titration with Ca2+ on the emission spectra of the iNOSoxyFMN system. After a transient decrease, EDTA causes a rise in fluorescence intensity. Addition of Ca2+ in excess of the free EDTA concentration returns the fluorescence to the level of the original preparation.

Fluorescence intensity data are suggestive but do not contain enough information to allow characterization of individual states. In 1989, Bastiaens et al. [35] published a paper on the FMN fluorescence of P450 reductase, a homolog of the NOS reductase domains, reporting a series of states with fluorescence lifetimes ranging from ∼ 2–3 or 5–6 ns (similar to free FMN or FMN in flavodoxins) to around 100 ps. The radiative lifetime of FMN, calculated from the absorbance and emission spectra, is around 15–18 ns [35,36]. The majority state in P450 reductase had a lifetime of ∼ 100 ps. From the Forster equation the rate of exciton transfer (in s−1) is


Qf is the fluorescence quantum yield of the donor in the absence of an acceptor, κ2 = (a.d − 3(a.r)(d.r))2 is the dipole orientation factor, τD0 is the lifetime of the donor in the absence of acceptor, n is the refractive index of the fluorophore surroundings, J(λ) is the spectral overlap integral and R is the distance in Ångströms.

In order to gain insight into the conformational changes responsible for the changes in intensity, fluorescence lifetime measurements were carried out. Table 1 shows the data for lifetime experiments with iNOS holoprotein and iNOSoxyFMN complex with Ca2+ replete CaM. The excitation wavelength was 440 nm and data were collected with a 500 nm long-pass filter; similar results have been obtained using 375 and 478 nm excitation. In all cases the intensity decays were fitted with a multiexponential function. Table 1 shows the parameters for the two-component fits for iNOS holoenzyme and iNOSoxyFMN. In each case there are at least two well-resolved components. Holoenzyme iNOS has a relatively long lifetime (4.4 ns) component. This component accounts for the majority of the steady state intensity but only about 30% of the enzyme population. The short lifetime (90 ps) component has an amplitude contribution of 65% for the intensity–time decay. These states correspond very well to components observed previously in P450 reductase [35]. The long lifetime component is reported as a single species, but it is clear that it is heterogeneous, with component substrates ranging from 2–3 ns to 5–6 ns. Examples of the raw decay data are shown in Fig. S1.

Table 1.   Fluorescence lifetime data for iNOSoxyFMN and iNOS holoenzyme.
SampleA1τ1 (ns)A2τ2 (ns)Average lifetime (ns)χ2
  1. The 4.2 ns component represents the aggregate of two reported long lifetime components with 2.3 and 5.3 ns lifetimes. The 0.56 ns minority component present in same sample does not correspond to the 0.9 ns component in NOS because it is not heme dependent.

iNOS holoprotein0.354.40.650.091.6530.99
CaM bound nNOS FMN domain0.684.30.3223.5721.096
Free FMN [35] 
P450 reductase [35]a0.244.20.620.0841.14 
P450 reductase [35]b0.140.56    

The spectral overlap integral for FMN and FAD has been reported as ∼ 4.6 × 10−15 cm3·m−1 [35]. Surprisingly, the overlap integral between FMN and high spin heme is even larger, because the Stokes shifted flavin emission spectrum overlaps the alpha and beta features of the heme completely, as shown in Fig. 3, but only overlaps the tail of the FAD absorbance. We estimate the overlap integral as about 0.9 × 10−13 cm3·m−1. Clearly, heme quenching can readily account for short lifetime states in NOS (but not in P450 reductase, which is heme free).

Figure 3.

 Overlap of flavin emission and iNOS high spin ferriheme absorbance spectrum showing overlap due to heme α and β bands (solid line). The riboflavin emission spectrum was recorded on a PTI quantum master with 475 nm excitation. The NOS absorbance spectrum was recorded on an Aminco DW2000 spectrometer. Heme concentration was 1 μm.

The results for the iNOSoxyFMN complex are illuminating. A long lifetime (4.3 ns) state accounts for about 72% of the population. This state appears to be very similar to the long lifetime state in the holoenzyme, differing only in that it is slightly more homogeneous. No 90 ps state is observable, but a new state with a lifetime of 0.9 ns is present which accounts for 28% of the population. Brunner et al. [34] observed similar states in a study of nNOS, and in addition reported a minor 30 ps component, which would not be resolved in our experiments. The assignment of long lifetime states to FMN and short lifetime states to FAD was not unreasonable but cannot be maintained because we observe the 0.9 ns state in preparations with no FAD and because much more than half the observable flavin in some holoenzyme preparations is in the 90 ps state.

The slow component lifetime is similar to FMN in highly fluorescent flavoproteins, and we therefore attribute it to FMN in the open conformation, distant from FAD. We will use the term ‘open’ to designate FMN states with relatively long lifetimes that are unquenched by FAD or heme. Crystal structures of homologous proteins show FAD and FMN separated by only ∼ 4 Å (see Fig. 4); this is likely to be the majority state, and suggests that the low fluorescence of the holoenzyme is due to FAD–FMN interactions. The short lifetime component, associated with the input state FMN/FAD pair, contributes relatively little to the steady state fluorescence, which is dominated by open states because of their higher quantum yield. We can assume that most of the steady state fluorescence is due to FMN, but in holoenzyme there is likely to be some contribution from FAD to short lifetime states. The proximity of the isoalloxazines in the input state crystal structures makes it unlikely that FMN and FAD could function as independent fluorophores.

Figure 4.

 Flavin orientation from crystal structures of (A) P450 reductase (1AMO) or (B) the nNOS reductase domain (1TLL). The relative positions of FMN and FAD isoalloxazines are slightly different owing to different positions of the FMN domain in the two structures. Aromatic residues stacking against isoalloxazine rings are shown. Major hydrogen bonds between flavins and protein are depicted as dashed lines.

Independently expressed nNOS FMN domain has FMN fluorescent lifetimes similar to flavodoxin and to the slow component of holoNOS. The majority state has a lifetime of 4.3 ns, and a secondary component has a lifetime of ∼ 2 ns. This directly confirms the assignment of the long lifetime states to a ‘free’ FMN binding domain in the open conformation. Lifetimes are summarized in Table 1, which includes results from previous work [35] for comparison. The iNOSoxyFMN construct conformational distribution is sensitive to Ca2+; EDTA treatment causes the loss of most of the 0.9 ns output state, which is reversible by addition of Ca2+. This is probably not due to release of CaM but instead is likely to reflect a change in conformation due to the release of Ca2+ by the low affinity EF hand. Figure 5 shows the lifetime results graphically, along with a schematic that shows the visible wavelength chromaphores in the constructs used in each experiment.

Figure 5.

 Schematic diagram showing the domains of constructs used in the experiments and the lifetime distribution. Holoenzyme contains the oxygenase domain linked at its C terminal end to the N terminal end of the FMN binding domain via the CaM binding site, followed by the FAD and NADPH binding domains. OxyFMN constructs are terminated after the FMN binding domain and lack the FAD and NADPH binding domains. In consequence, they do not display a short lifetime (∼ 100 ps) state. Independently expressed FMN binding domain has lifetime states resembling flavodoxin and free flavin, and lacks the ∼ 1 ns state we attribute to coupling to heme.

Bastiaens et al. [35] had no structural information and used an averaged fluorescence lifetime to estimate a characteristic distance from kDA = 1/τD − 1/τD0. Their average lifetime was ∼ 0.7 ns, and with a rough estimate of the orientation term obtained from anisotropy experiments they estimated a donor–acceptor separation of 1.6–2 nm. Crystal structures are now available for both nNOS reductase domains and P450 reductase (see Fig. 4 and [37,38]). This provides additional orientation information; although uncertainties in the orientation of the transition dipole moments within the isoalloxazine ring systems make this approximate, it is clear that their orientation term was not unreasonable. However, it is clear from our work that the multiple lifetimes result from different states that have very different donor–acceptor interactions and should be treated separately. The lifetime of the majority input state, with FMN and FAD in close association, is about six times shorter than the averaged lifetime used by Bastiaens et al. [35]. Assuming dipolar mechanisms predominate and using the Forster equation, distances in the 1–1.2 nm regime are compatible with possible orientation factors. Since crystal structure information for the input state is available, extraction of structural information from the fluorescence data is unnecessary. However, the fluorescence data identify the holoenzyme majority state of P450 reductase and NOS with the crystal structures.


The very low flavin fluorescence intensity of iNOS can be attributed to the formation of FMN–FAD input state pairs in most of the holoenzyme molecules. In the Results section most of the holoenzyme fluorescence was attributed to a minority open state in which the FMN domain was dissociated from the rest of the reductase complex. There is no reversible effect of EDTA and Ca2+ on steady state flavin fluorescence of iNOS holoenzyme. The irreversible increase in fluorescence does not correlate with changes in NOS or cytochrome c reductase activity, both of which are only slightly EDTA and Ca2+ sensitive [31,33]. Flavin fluorescence in iNOS holoenzyme is highly quenched, and it is likely that the irreversible fluorescence increase is associated with an inactive population of enzyme molecules since it does not correlate with activity.

In contrast, iNOSoxyFMN construct coexpressed with wild-type CaM is highly fluorescent, with a steady state intensity about seven times greater than that of the holoenzyme. We attribute this to the lack of the two C-terminal cofactor binding domains, so that FMN–FAD pairs are not possible in this construct. The 50–70% EDTA induced rise in fluorescence is completely reversible by titration with Ca2+. This suggests the presence of several conformations with different contributions to fluorescence.

The iNOSoxyFMN construct has an ∼ 4.3 ns component, representing FMN which is not strongly coupled to other prosthetic groups and hence is in an ‘open’ conformation. There is a new rapid component with a lifetime slightly < 1 ns. The ratio of lifetime components is 3 : 7. The majority long lifetime component dominates steady state fluorescence by more than 10 : 1. We assign the shorter lifetime component to an output state in which the heme interacts with the FMN. Because heme itself is paramagnetically quenched, it does not contribute significantly to fluorescence. Stern–Volmer plots of iNOSoxyFMN and holoenzyme are similar because the steady state fluorescence is in both cases dominated by open states with similar access to solvent.

Clearly, iNOSoxyFMN is much more fluorescent than holoenzyme. If components observed in lifetime experiments are the only contributing states, we can account for this by noting that the dominant ‘open’ long lifetime state is 5–6-fold larger in iNOSoxyFMN because the input state is tighter than the output state.

It is clear that the timescale of the fluorescence lifetime experiments (about 4 ns) is rapid enough to allow resolution of the conformational states. Any quenching of the long lifetime states by transition to the short lifetime states must occur on a timescale slower than 6–8 ns; an exchange effect on this timescale could account for the microheterogeneity of the long lifetime states, but other quenching effects are possible and the exchange rate could be much slower. On the other hand, it is clear that in iNOSoxyFMN conformational equilibration between open states and the output state must be fast compared with electron transfer. Only a single electron transfer component is observed in flash experiments [22–24], and it is clear from results presented here that both the open conformation and the output state are well represented in the conformational equilibrium. The time frame from 10−3 to 10−8 s is consistent with the observed results; the rate must be faster than milliseconds to average the electron transfer components but slower than the fluorescence lifetimes because the lifetime components are discrete.

Recently viscosity experiments using our iNOSoxyFMN construct have been used to support models in which the output state is present as a very small fraction, and transitions between the open state and the output state are rate limiting [32]. Because only one component in the electron transfer kinetics was observed, this kind of model requires a small minority output state. A single component was always observed in the previous flash experiments on NOS [22–24]. The evidence does not support interpretation of these results as a viscosity effect on the kinetics of output state formation. The conformation and aggregation state of NOS is very sensitive to the presence of modifiers such as glycerol, sucrose and other proteins, and solution conditions have a significant effect on the conformational distribution in NOS holoenzymes and constructs; this is a complex subject that will be addressed in subsequent papers. The steady state fluorescence data presented in support of their interpretation shows instead that the conformational equilibrium is affected by glycerol and sucrose. Finally, the whole model requires that the population of the output state is too low to detect as a kinetics component. Many lines of evidence now show that this is not true.

In the holoenzyme most of the enzyme is tied up in the input state, so rate limitation by FMN domain release would produce only a small (∼ 10%) component of comparatively rapid electron transfer, which would probably be composed of a distribution of open state components, difficult to observe. However, cytochrome c reduction is an order of magnitude more rapid than NO synthesis or FMN/heme electron transfer in holoenzyme. This indicates that release of the FMN binding domain is much more rapid than electron transfer between FMN and the catalytic site, and any rate limited conformational step would have to be between a series of open states unresolved by fluorescence lifetime experiments. In CaM activated NOS, electron transfer is limited during turnover by the fraction of the enzyme in the output state, but this in turn is determined by the details of conformational equilibration.

Both the rate of NADPH–cytochrome c reduction by holoNOS and the rate of electron transfer in oxyFMN constructs are too rapid to account for the slow rate of NADPH–oxygenase electron transfer, which limits NO synthesis. The critical point here is that there is no single ‘open’ conformation, but instead a manifold of open states in which the FMN is more or less isolated from heme and FAD but exposed to solvent. Some of these states are in rapid equilibrium with the output state, and others are in rapid equilibrium with the input state. The rate limiting step for electron transfer is the passage of the FMN binding domain through the conformational manifold, which may require passage through a conformational bottleneck; the other steps are too rapid to account for the slow rate. The effects of CaM binding, which must include effects on all stages of the conformational cycle of the FMN binding domain, will be explored in the next paper in this series, which provides data on CaM effects in eNOS and nNOS.

The data presented here account very well for the difference in the electron transfer rate between iNOS constructs. Electron transfer occurs in the ∼ 1 ns state if our assignment of this state to the output state is correct, and we would expect the measured electron transfer rate to be proportional to the fraction of enzyme molecules in this state multiplied by a rate constant that is a function of output state geometry and thermodynamics. However, in holoenzyme many factors make it unlikely that all the conformations are in rapid equilibrium.

We have measured a much slower electron transfer rate in holoenzyme, about 20 times slower than iNOSoxyFMN. All other things being equal, this leads us to expect < 2% of the enzyme in the output state. A component this small would not be detected in the holoenzyme lifetime experiments, and a rapid electron transfer component would also probably be undetectable. Hence, the single turnover rate measured in holoenzyme is probably not the result of a set of rapidly equilibrating conformations, as in oxyFMN, but the result of the passage of the FMN domain from a source population (the input state plus associated open conformations in rapid equilibrium with it) through a conformational manifold to the open states in the vicinity of the output state. The ratio of the output state to the open conformations appears to be smaller in iNOS holoenzyme than in iNOSoxyFMN, probably due to steric effects from the massive di-flavin reductase complex.

This is illustrated in Fig. 6. The input state is on the left, and the output state is on the right. After release from the input state reductase complex, the FMN domain moves through many possible ‘open’ conformations before reaching the output state. The many open conformations in rapid equilibrium with the input state account for the rapid reduction of cytochrome c. A second conformational manifold is in rapid equilibrium with the output state, but these are separated by a conformational bottleneck that limits the electron transfer rate in holoenzyme. No such bottleneck exists for oxyFMN, which is much less conformationally constrained. The effects of the control elements of the signal generating isoforms, on the conformational manifold, and the mechanism of their activation by CaM will be addressed subsequently.

Figure 6.

 Illustration of the conformational cycle of the FMN domain in NOS. Dissociation of the FMN domain from the input state initially produces a manifold of rapidly equilibrating open states that are competent to reduce cytochrome c, accounting for rapid NADPH–cytochrome c reduction. To reduce the oxygenase domain, the FMN domain must traverse a conformational bottleneck. Hence, NADPH–heme electron transfer is slow although both NADPH–cytochome c and heme–FMN electron transfer are rapid. The open state manifold consists of a heterogeneous population of states that have relatively long FMN fluorescence lifetimes; the input and output states are well defined.

The measured rate of electron transfer in the iNOSoxyFMN construct is 850 s−1, which would correspond to a unidirectional electron transfer rate of 425 s−1 if the heme and FMNH.FMNH2 couple were isopotential [22]. If the effective potential difference on the experimental timescale is 18 mV (as in nNOS [23]), the downhill rate will be 566 s−1. Assigning the competent state to the 1 ns fluorescence lifetime state, the fractional occupancy of the competent state is 30%, and the electron transfer rate in the competent state is 1650 ± 240 s−1. This calculation assumes that the conformational equilibrium does not depend strongly on the redox state of the flavin. This assumption is supported by comparison of EPR and fluorescence results, which suggest that a quarter to a third of the system is in the output state in both the semiquinone and oxidized FMN redox state.

Marcus’s theory [39] indicates that the electron transfer rate in this state will be an exponential function of distance between the electron carriers. Moser et al. [40] have studied numerous cases and provided a semi-empirical description of tunneling in electron transfer proteins as a function of distance. Numerous recent papers [41–43] have provided additional depth of understanding to the field, but the simple semi-empirical expression provides a convenient estimate of the distance between FMN and heme in the distance regime addressed here. The ‘Dutton’s ruler’ expression is


where ket is the electron transfer rate, R is the edge to edge distance between donor and acceptor in ångströms, ΔG is the driving force in kilocalories per mole and I is the reorganization energy. Because electron transfer is between two domains, we expect I∼ 1 eV, and from the results of flash experiments we estimate ΔG ∼ 0.018 eV [23]. This is a small value of ΔG and has only a small effect on the distance estimate, almost exactly canceled by the partitioning of the experimental electron transfer rate [23].

With these assumptions, the edge to edge distance would be 15 Å; the range of likely reorganization energies would only alter this distance from 14 to 16 Å. A more important source of uncertainty lies in the protonation state of the FMN semiquinone. The majority state of the FMN semiquinone is the blue neutral form between pH 6 and 10. Heme–flavin electron transfer is thus linked to FMN protonation/deprotonation. We expect protonation to be rapid compared with interdomain electron transfer and, while this superficially runs counter to usual Frank–Condon assumptions, over interdomain distances the measured electron transfer rates are indeed relatively slow. Protonation then does not become rate limiting, but it would introduce significant increases in ΔG and I. These effects are likely to be of the order of 0.1–0.2 meV and could decrease the estimated distance to ∼ 12 Å, close to the closest approach between FMN and heme possible without reorientation of the aromatics covering the heme edge (W366 in human iNOS) and FMN (Y631 in human iNOS [11,13,16,17]).

This distance is consistent with the partial quenching of FMN fluorescence by heme. The weak line broadening results we previously observed in EPR experiments with iNOSoxyFMN constructs are consistent with dipolar line broadening at heme iron to FMN distances of ∼ 16 Å, with an FMN–heme edge distance of 12–15 Å [21]. A recent double resonance experiment using oxyFMN complex resolved a Pake lineshape for this interaction, and determined a more accurate distance of ∼ 18 Å by simulation [44]. Docking the structure of the FMN binding domain to the structure of the oxygenase domain suggests that the edge to edge distance must be 3–5 Å less than the heme iron to FMN distance, depending on whether FMN approaches W366 directly or from an angle; there is at least one feasible alternative to a head-on approach. The alternative shown in [44] has an FMN to heme edge distance of 13.3 Å. Factors limiting the accuracy of the calculations include delocalization of spin on the isoalloxazine and the likelihood of strain broadening from a frozen-in distribution of dipolar coupling constants [45].

In conclusion, these results indicate that the FMN binding domain in iNOS exists in three major distinct states. The shortest lifetime state, identified with the input state, is the majority state in the holoenzyme. The longest lifetime state, observed in both holoenzyme and iNOSoxyFMN constructs, is the open state in which FMN is isolated from the other prosthetic groups. The ∼ 1 ns state, observed only in iNOSoxyFMN, is the output state. The results suggest that electron transfer in the output state occurs on a sub-microsecond timescale requiring close docking of FMN to a site near the heme edge, and that the electron transfer rate is limited by the fraction of the enzyme in the output conformation.

Experimental procedures

Protein expression

iNOS was coexpressed with CaM in Escherichia coli BL21 and purified as previously described [46]. iNOSoxyFMN was coexpressed with wild-type CaM and purified as previously described [21]; it consists of the oxygenase and FMN binding domains and is truncated directly after the negatively charged tail of the final α-helix of this domain. OxyFMN constructs lack the FAD and NADPH binding domains. Independently expressed CaM bound nNOS FMN binding domain was a gift from T. Poulos; its purification and characterization have been described previously [25]. This protein was used here as a reference for FMN domain free of heme and FAD interaction in fluorescence lifetime experiments.

Flavin and heme contents were measured spectrophotometrically. The activity of purified iNOS holoenzyme was measured using the oxyhemoglobin method as described previously [46].

Fluorescence spectroscopy

Emission spectra were recorded on a Cary Eclipse fluorescence spectrophotometer at 23 °C. Excitation and emission slits were both set at 5 nm, and a photomultiplier voltage of 800 V was used. Samples were measured in a 1.4 mL quartz cuvette with a path length of 1 cm. The spectral data were smoothed with the boxcar method. The fluorescence of iNOSoxyFMN was measured at a concentration of 2.5 μm in 40 mm BisTris propane containing 1 mm dithiothreitol (pH 7.4). Flavin fluorescence emission spectra were measured by exciting samples at 450 nm, and fluorescence intensity was measured from 470 to 650 nm. To study Ca2+ dependence, 0.6 mm EDTA was added directly into the cuvette to sequester Ca2+ and emission spectra were measured every 2 min until fluorescence intensity stabilized. After signal stabilization Ca2+ aliquots were introduced into the system to study reversibility of the EDTA chelation effect. All spectra were corrected for instrumental artifacts by subtracting the baseline emission spectrum of the buffer.

Similar experiments with iNOSoxyFMN were performed utilizing a kinetics-based approach in order to ensure that features of the emission spectra were not artifacts of experimental time constraints. Flavin fluorescence at 530 nm was monitored as a function of time under the same conditions as those used to measure emission spectra. Measurements were taken every 0.0015 s, and the data were smoothed with the boxcar method. Readings were zeroed by subtracting the average measured fluorescence intensity of the buffer from the fluorescence intensity of the samples.

Fluorescence lifetime measurement

Time-resolved intensity decays were recorded using a PicoQuant Fluotime 100 time-correlated single-photon counting fluorescence lifetime spectrometer as described earlier [47]. The excitation at ∼ 440 nm was obtained using a pulsed laser diode (PicoQuant PDL800-B) with 20 MHz repetition rate. The excitation was vertically polarized and the emission was recorded through a polarizer oriented at 54.7°, the magic angle. Appropriate long-pass filters from Chroma Technology Group (Rockingham, VT, USA) were used for collection, thus eliminating the scattered excitation light and collecting the fluorescence from the samples in the region of interest.

The fluorescence intensity decays were analyzed in terms of the multiexponential model as the sum of individual single exponential decays:


In this expression, τi is the decay time and αi is the amplitude. The fractional contribution of each component to the steady state intensity is described by


The mean (intensity-weighted) lifetime is represented by


and the amplitude-weighted lifetime is given by


The values of αi and τi were determined using the picoquant fluofit 4.1 software (PicoQuant GmbH, Berlin, Germany) with deconvolution of the instrument response function and nonlinear least squares fitting. The goodness-of-fit criterion was determined by the χ2 value. (See also [48].)


This work was supported by NIH 1R15GM083317. We acknowledge access to the facilities at the Center for Fluorescence Spectroscopy at University of Maryland School of Medicine in obtaining fluorescence lifetime measurements. We thank Thomas Poulos and Huiying Li for productive discussions. We thank Lesa Hall, Veterans Affair Medical Center, Durham, NC, USA, for graphics.