Allostery in the nitric oxide dioxygenase mechanism of flavohemoglobin

The substrates O2 and NO cooperatively activate the NO dioxygenase function of Escherichia coli flavohemoglobin. Steady-state and transient kinetic measurements support a structure-based mechanistic model in which O2 and NO movements and conserved amino acids at the E11, G8, E2, E7, B10, and F7 positions within the globin domain control activation. In the cooperative and allosteric mechanism, O2 migrates to the catalytic heme site via a long hydrophobic tunnel and displaces LeuE11 away from the ferric iron, which forces open a short tunnel to the catalytic site gated by the ValG8/IleE15 pair and LeuE11. NO permeates this tunnel and leverages upon the gating side chains triggering the CD loop to furl, which moves the E and F-helices and switches an electron transfer gate formed by LysF7, GlnE7, and water. This allows FADH2 to reduce the ferric iron, which forms the stable ferric–superoxide–TyrB10/GlnE7 complex. This complex reacts with internalized NO with a bimolecular rate constant of 1010 M−1 s−1 forming nitrate, which migrates to the CD loop and unfurls the spring-like structure. To restart the cycle, LeuE11 toggles back to the ferric iron. Actuating electron transfer with O2 and NO movements averts irreversible NO poisoning and reductive inactivation of the enzyme. Together, structure snapshots and kinetic constants provide glimpses of intermediate conformational states, time scales for motion, and associated energies.

The substrates O 2 and NO cooperatively activate the NO dioxygenase function of Escherichia coli flavohemoglobin. Steady-state and transient kinetic measurements support a structure-based mechanistic model in which O 2 and NO movements and conserved amino acids at the E11, G8, E2, E7, B10, and F7 positions within the globin domain control activation. In the cooperative and allosteric mechanism, O 2 migrates to the catalytic heme site via a long hydrophobic tunnel and displaces LeuE11 away from the ferric iron, which forces open a short tunnel to the catalytic site gated by the ValG8/ IleE15 pair and LeuE11. NO permeates this tunnel and leverages upon the gating side chains trigger the CD loop to furl, which moves the E and F-helices and switches an electron transfer gate formed by LysF7, GlnE7, and water. This allows FADH 2 to reduce the ferric iron, which forms the stable ferricsuperoxide-TyrB10/GlnE7 complex. This complex reacts with internalized NO with a bimolecular rate constant of 10 10 M −1 s −1 forming nitrate, which migrates to the CD loop and unfurls the spring-like structure. To restart the cycle, LeuE11 toggles back to the ferric iron. Actuating electron transfer with O 2 and NO movements averts irreversible NO poisoning and reductive inactivation of the enzyme. Together, structure snapshots and kinetic constants provide glimpses of intermediate conformational states, time scales for motion, and associated energies.
Distantly related members of the Hb superfamily, including flavoHb, HbN, phytoglobin, Mb, and Cygb, function as NO dioxygenases (NODs) (EC 1.14.12.17) to limit NO toxicity and control NO signaling in cells (1,2). A greater understanding of the enzyme mechanism promises insights for the design of novel antibiotics, vasorelaxants, and antitumor agents. We previously suggested an ordered ping-pong mechanism in which O 2 enters the distal pocket and binds the ferrous heme forming a stable ferric superoxide intermediate that reacts with NO to form NO 3 − and Fe 3+ , and in which Fe 3+ is rapidly reduced to repeat the catalytic cycle (1)(2)(3)(4)(5)(6). In support of the mechanism, the conserved distal TyrB10 hydroxyl group in flavoHb and HbN confers high O 2 affinity by stabilizing the proposed intermediate Fe 3+ O 2 − thus allowing rapid turnover (4,(7)(8)(9). Also, the ferric heme was shown to be reduced by NADH, in the absence of O 2 or NO, at rates approaching NOD turnover rates suggesting an unregulated reduction step (4). However, the mechanism has failed to fully account for weak and reversible NO inhibition and a large O 2 requirement relative to NO and O 2 affinities measured for the ferrous heme (1,4,5,10). Indeed, the relatively high affinity of ferrous flavoHbs for NO (4,5) led Hausladen et al. (11) to suggest a reversed order ping-pong mechanism in which a ferric nitroxyl intermediate reacts with O 2 to form NO 3 − , a reaction with an intrinsically small rate constant that is inconsistent with the large NO dioxygenation rate constants of (flavo)Hbs and the critical role of O 2 affinity in high turnover rates (1,2). In addition, the 100fold discrepancy between the NO migration rate to the ferrous heme as measured by transient kinetics and the calculated V max / K m (NO) or k' NOD values for flavoHb-NODs (4, 5) has demanded clarification. Deficiencies in both of the proposed mechanisms has spurred greater efforts toward a resolution (2,(12)(13)(14)(15)(16). Nevertheless, the mechanism has remained enigmatic, and controversial, partly because of the physical and chemical similarities of the substrates O 2 and NO, and also because of insufficient knowledge of the unique structures and dynamics supporting the mechanism, and an apparent underdetermination of kinetic data (4). A simple ping-pong mechanism also fails to fully explain the dioxygenase activity of Cygb (10,17) or the alpha subunit of HbA in complex with its chaperone (18). Both NOD activities would also be expected to be exquisitely sensitive to irreversible poisoning by NO. In fact, the well-known rapid and tight binding of NO with the ferrous heme of erythrocyte HbA or muscle Mb, with comparable O 2 association and NO oxidation (dioxygenation) rate constants (19,20), has long diminished enthusiasm for an enzymatic NOD function for Mb and HbA, or its subunits (2,21) or, for that matter, any globin.
A greater understanding of the NOD mechanism, and a solution to the paradox, is slowly emerging from a wealth of structural models, molecular dynamics simulations, and spectroscopic and kinetic studies of globins. O 2 and NO tunnels and a nitrate egress pathway have been suggested to play roles in the NOD function of the truncated HbN from This article contains supporting information. * For correspondence: Paul R. Gardner Mycobacterium tuberculosis (22,23). A long hydrophobic tunnel that runs parallel to the H helix and perpendicular to the heme plane is thought to allow NO (22,24,25), or O 2 (16,24,26), to access the distal heme reaction chamber, while a short tunnel formed between the G and H helices at the AlaG5 and LeuH8 residues allows O 2 (22), or NO (24,26,27), to access the distal heme pocket. Moreover, movement of PheE15 caused by O 2 binding to the ferrous heme and TyrB10-GlnE11 hydrogen-bonding interactions (28) may function as a gate, in cooperation with LeuG8, that controls NO access to the heme (22,25,(29)(30)(31). In support, a TrpE15 variant showed two-to threefold less NOD activity when reductively coupled and assayed within E. coli (30) but inexplicably showed no effect on NO dioxygenation by HbNFe 3+ (34). An important mechanistic clue has also been provided by studies of the mammalian Cygb-NOD. We and others have reported 10-to 100-fold larger rate constants for the ascorbate and cytochrome b 5 -mediated reduction of the ferric Cygb during NOD catalysis (10,17). The data are explained if O 2 and NO movements or reaction within Cygb serve as a trigger for electron transfer during catalysis (10). Evidence of an electron transfer trigger and switch would also explain how the reduction step limits the NOD function (4,35). Furthermore, Cygb structures reveal potential tunnels and gates (36,37) that appear functionally similar to those found in HbN.
We have observed, and now report, cooperative activation of the E. coli flavoHb-NOD activity by O 2 and NO. To understand the apparent allostery, we have analyzed structures of flavoHbs, and related globins, formulated a structural and kinetic model for the NOD mechanism, and evaluated the model for its compatibility with the steady-state and transient kinetic behavior of wild-type (WT) and variant flavoHbs. The data and model allow us to derive kinetic constants and energies for a flavoHb-NOD mechanism that employs O 2 (38) and NOspecific supply tunnels, gates, or channels for regulating NO entry to and nitrate efflux from the heme reaction center, a structure-sensitive electron transfer switch (38)(39)(40), an actuating lever to control the gates and switch, and an elastic torsion spring to store energy released by the NO dioxygenation reaction to drive and control motions. A lever and switch that together synchronize heme reduction with O 2 and NO positioning, and thus prevent Fe 2+ NO formation, may explain the puzzling resistance of the NOD activity to irreversible NO inhibition, the observed reversible NO inhibition, and the higher-than-expected O 2 dependence in the steady state (4,5).

O 2 and NO dependence of flavoHb-NOD activity
The E. coli flavoHb-NOD activity (4) (Fig. 1A) supporting a ping-pong mechanism. At higher NO concentrations, NO reversibly inhibits the activity, and NO inhibition appears competitive with respect to O 2 with roughly 50% inhibition observed at a 1:100 ratio of NO to O 2 as previously reported (4,5). O 2 shows no competitive inhibition with respect to NO at O 2 :NO ratios exceeding 10,000:1. However, flavoHb shows less activity than expected with low [NO] and [O 2 ] thus revealing activation by NO and O 2 . The activating effects of NO and O 2 are more apparent at 20 C (Fig. 1B). The data suggest a more complex ping-pong mechanism in which O 2 and NO cooperatively activate NO dioxygenation.

Structural features and a proposed NOD mechanism
Structural dynamics may explain the cooperative activation of NODs by O 2 and NO, the weak reversible competitive inhibition by NO, and the larger-than-expected requirements for O 2 . Similar to the M. tuberculosis HbN-NOD, the flavoHb-NOD appears to utilize a hydrophobic long tunnel (LT) for Note the 6-fold y-axis scale difference between panels A and B. C and D, log-log plots of the data from panels A and B, respectively. The kinetic constants used to fit the kinetic model with the data (dashed lines) are given in Table 1.
O 2 entry (38), a hydrophobic short tunnel (ST) with an amphipathic entry for NO, and a gate for NO 3 − efflux ( Fig. 2A).
Six residues from the G and H helices form the outer wall of the 20-Å LT with three residues from the A helix and single residues from the B, E, and H helices forming the inner wall ( Fig. 2B) (38). The passageway follows along the G-helix dipole in the positive direction. The side chains lining the LT are highly conserved among flavoHbs and single-domain Hbs (Table S1). Furthermore, the residues form a strictured or segmented tunnel with the capacity to hold up to three O 2 molecules (38). The LT strictures show dimensions compatible with the passage of the nonpolar 3.2 Å by 4.3 Å van der Waals' dimension O 2 molecule as previously demonstrated using molecular dynamics simulations with an estimated maximum energy barrier of 6 kcal/mol (38). However, although the static dimensions and clearances appear similar among the flavoHb and single domain Hb structures, it is noteworthy that the strictures differ with ligand states (Table S2). For example, with nitrite bound in the Saccharomyces cerevisiae structure (41) the clearance between the E15 and H12 residues is 3.1 Å wider and the restrictive distance between residues A16 and G16 is 1.6 Å greater, suggesting an active mechanism for controlling O 2 movement during turnover. The LeuE11 side chain is in van der Waal's contact with the ferric iron in the E. coli flavoHb structure, thus apparently hindering O 2 access to the iron, but LeuE11 does not form part of the LT. The putative ST shows a highly conserved outer entry defined by GlyG9, GlyH9, LeuG13, and TyrH12 (Table S3) and appears to be gated in the interior by ValG8-IleE15 and LeuE11-heme interactions (Fig. 2C, top). The S. cerevisiae ferric-nitrite enzyme structure shows a widening of the G8-E15 gate by 4 Å and a shift of the entire E-helix by 2 Å toward the putative NO entry site (compare red arrows in Fig. 2C, top and bottom, and Table S4). In the open state (bottom), LeuE11 is displaced toward the NO entry site to form a part of the ST interior presumably guiding NO for reaction  The ferric E. coli flavoHb structure shows a CD loop in an unusual random coil with the GlnE2 side chain extended forming contact with the adenine ring of FAD in the reductase domain (Fig. 2D, top). A chloride ion is found bound in an anion hole in the E helix and CD loop corner (Fig. S1). Moreover, the ArgE3 guanidinium group, the E7 amide nitrogen, and a bound water bounding the anion hole show dimensions compatible with binding the trigonal planar nitrate anion in an E2dependent efflux mechanism, whereas, with nitrite bound to the ferric heme in the S. cerevisiae flavoHb structure, the GlnE2 residue is found oriented inward with the D3-E4 segment in an energetically more stable alpha helix with the ArgE3 side chain oriented outward (Fig. 2D, bottom). Displacement of the LeuE11 side chain from the iron by nitrite, and presumably by nitrate, appears to twist the E-helix and exert a right-handed torsion to furl the CD loop (Fig. S1). In the unoccupied state, and presumably with nitrate release in the NOD mechanism, large phi torsion angle changes of 180 occur for the D4 and E3 residues in the conformers with the unfurled E2 and E3 carbonyl Oatoms interfering and repelling each other (Fig. S1), possibly storing mechanical energy from the NOD reaction, and providing spring-like right-handed elastic torque with refurling to assist LeuE11 displacement and E and F helix movements. Both CD loop types are also observed in the dimeric Ralstonia eutropha flavoHb-econazole structure (Fig. S2) (42) thus arguing for CD loop conformational changes rather than sequence-specific static structures.
The ferric E. coli and S. cerevisiae ferric-nitrite flavoHb structures also show differences in the positioning of the proximal LysF7 forming the bridge for electron transfer (ET) (39,40), suggesting a switching mechanism that is activated by O 2 -and NO-leveraged motions. In the unoccupied ferric structure (Fig. 2D, top), the LysF7 epsilon ammonium group and GlnE7 amide oxygen atom hydrogen bond a water molecule near the middle of the 6-Å-long line-of-flight for ET between the heme carboxylate and FAD C8 methyl group (43) that is open to solvent exchange (38). In this model, the positioning of the electron-rich O atom of water hinders ET. The NO 2 − occupied structure shows a 10 torsional rotation of the F-helix relative to the heme plane (Fig. 2D, bottom) in which the GlnE7 amide is positioned with TyrB10 hydrogen bonding and stabilizing the Fe 3+ O 2 − intermediate and the LysF7 side chain positioned with the straight line-of-flight for ET passing 0.3 Å beyond the van der Waals' radius of the epsilon ammonium nitrogen, which presumably moves to the position of balance between repulsion by the core and attraction by the annular shell for optimal electron tunneling (44). Electron tunneling paths involving the bridging water molecule and LysF7 have been previously considered (2,(38)(39)(40). The structural differences suggest a LysF7-H 2 O-GlnE7 ET switch mechanism that is activated by the permeation of NO through the G8-E15 gate and by the further upward leveraging LeuE11 with a right-handed torsional twist of the E and F helices that is coupled to CD loop furling.

Reaction steps and a kinetic model
The structural features of flavoHb described above suggest a catalytic cycle, and testable kinetic model, to explain the cooperative O 2 and NO activation observed in Figure 1, A-B. In the kinetic model depicted in Figure 3, nonpolar O 2 molecules are driven from the aqueous solvent into the LT with an equilibrium constant of K 1 , and the leading O 2 displaces LeuE11 forcing the movement of the E-helix, docks with the ferric iron, the LeuE11 isobutyl group and the TyrB10 hydroxyl, and weakens the G8-E15 gate with an equilibrium constant of K 2 . NO then enters and passes through the ST further widening the G8-E15 gate with equilibrium constant K 3 and leveraging the movement of LeuE11, the E-helix, GlnE7, the F-helix, and the LysF7 side chain while triggering furling of the CD loop with the first-order rate constant k' T and equilibrium constant (K T ), ET (k ET ), and nitrate release (k p (45), which readily dissociates (4) and resists reduction to Fe 2+ NO. The observed ET rate depends upon the LysF7-H 2 O-GlnE7 gate positioning and the FAD/ FADH 2 reduction potential, the driving force. The reaction scheme also shows pathways for O 2 − and H 2 O 2 production as well as the motifs and residues most critical for each step.
We have formulated a steady-state velocity equation for the catalytic cycle with the activating trigger step (K T ) (Equa- We have also factored in NO acting as a competitive inhibitor with respect to O 2 (Equation 2) (46).

Allosteric nitric oxide dioxygenase mechanism
And, we have allowed NO and O 2 to each act as uncompetitive inhibitors of catalysis (Equations 3 and 4, respectively).
where d ¼ 1=ð1 þ ð½O 2 =K iðO2Þ ÞÞ These additional independent statistical factors and the equilibrium constants K ii(NO) and K i(O2) have been included for situations in which NO and O 2 do not compete for LT or ST entry, or iron binding, but nevertheless inhibit catalysis. In addition, we have introduced a factor and equilibrium constant K a(O2) for rare situations in which O 2 causes an additional NOindependent activation of catalysis possibly by occupying unique binding sites or cavities (Equation 5). Equations 3-5 assume a normal hyperbolic binding and inhibition or activation response.

Fit of the mechanistic kinetic model with steady-state kinetic data
The kinetic data for the WT flavoHb in Figure 1, A-B are transformed to the respective log-log plots in Figure 1, C-D. Logarithmic plots were preferred since they display large ranges of substrate concentrations and velocities with a much greater resolution (46) and allow for same-scale visual comparisons. We have estimated the kinetic constants in Table 1 by fitting the velocity equation to the steady-state kinetic data with the resulting fits shown as dashed lines in Figure 1, C-D. For fitting, we assumed K 1 << K 2 . K 1 was taken to be 2 μM, the value estimated from O 2 association kinetics where V max /k' O2 was the theoretical K m (O 2 ) of 2 μM at 20 C (4). It should be noted that the theoretical v is most affected by the K 1 value at low O 2 and NO concentrations. When K 1 is assigned a 10-fold larger value and K 2 = 120 μM, v is <2-fold smaller with the largest effects at low [O 2 ] and [NO]. And, smaller K 1 values relative to a large K 2 diminish any effects of K 1 on v. A range of values for K 3 and K 4 were required to model the data at 20 C as shown by the black and red dashed lines in Figure 1D and provided in the data in Table 1. A loss of harmonics and frustration in enzyme motion (47) may explain the deviations from a more ideal kinetic behavior at the subphysiological 20 C. A k' NOD value of 1 to 2 × 10 10 M −1 s −1 can Allosteric nitric oxide dioxygenase mechanism  (20) 200 (50) 80 (30) 400 (225) 100 (10) Values in parentheses are for 20 C.   Table 1.  Table 1.
be estimated from the V max /K 4 ratio at 37 C, a rate constant approaching the value of 6.9 × 10 10 M −1 s −1 for the bimolecular reaction of NO and free O 2 − (48).
Similar steady-state kinetic data for G8 and E11 substitution mutants are shown in Figures 4 and 5, respectively, and kinetic constants used to fit the data are also given in Table 1. Substituting ValG8 with the larger Leu residue increases K 3 , the apparent K m (NO), whereas substituting ValG8 with a smaller Ala decreases K 3 as expected for a role for ValG8 in gating NO entry. The critical function of ValG8 in the allosteric mechanism is also suggested by the large changes in the K T values in the mutants, and the apparently large frustration of AlaG8 at 20 C. The data also reveal a role for ValG8 in determining the susceptibility to NO inhibition. AlaG8 shows smaller K i(NO) and K ii(NO) values, whereas the bulkier LeuG8 shows larger values. The effect is more discernible at 37 than 20 C, again suggesting greater frustration at the lower temperature. A greater resistance of LeuG8 to NO inhibition also confers a higher V max at both temperatures; however, both the NO sensitivity, as indicated by small K i(NO) and K ii(NO) values, and the impaired allosteric activation, as indicated by large K T values, lower the V max for AlaG8. The data support the roles for G8 shown in the kinetic model in Figure 3 and structure model in Figure 2C. G8 substitutions also cause relatively smaller effects on the O 2 dependence (K 2 ) suggesting effects on E-helix and LeuE11 movement.
Substituting LeuE11 with the larger Trp caused large increases in K 3 and K 2 and a large decrease in K T (Table 1).
Consistent with the structure-based kinetic model ( Fig. 2C and Fig. 3), TrpE11 hinders NO access, increases the O 2 requirement for activity, but nevertheless favors the activation by O 2 and NO. In contrast, the smaller Ala substitution shows the opposite effects. Larger K 2 values for the TrpE11 mutant indicate larger O 2 requirements for displacement and E-helix movement. The small K T suggests a strong, and effective, response to NO movements. The data support roles for LeuE11 as the O 2 and NO activated trigger for the allosteric mechanism and as a part of the NO access tunnel. However, the data for the AlaE11 variant were unusual in that the kinetic model could only be fitted to the data by allowing for O 2 inhibition at 37 C and O 2 activation at 20 C.
GlnE2 within the CD loop was tested for its proposed role in the allosteric mechanism. Data for E2 substitution flavoHbs are shown in Figure 6, and kinetic constants used to fit the data are given in Table 2. Substitution of the highly conserved GlnE2 with Asn, Leu, His, and Glu causes larger K T values representing an impaired activation of ET by NO and O 2 , thus producing a two-to threefold decrease in the V max values and proportionally smaller K 3 and K 4 values. K 2 values increase two-to threefold except for the His substitution at 37 C. The data demonstrate an important role for GlnE2 in the NOD mechanism. The larger V max values and lesser effects of oppositely charged His and Glu substitutions on K T suggest that Gln allows optimal furling-unfurling of the CD loop rather than interacting directly with nitrate. Furthermore, the consistently smaller K ii(NO) values of the E2 variants suggest a  Table 2.
Allosteric nitric oxide dioxygenase mechanism CD loop structure dysfunction that opens a migration pathway for NO.
TyrB10 and GlnE7 are important for stabilization of the Fe 3+ O 2 − intermediate in (flavo)Hbs (1,4,49). In the proposed kinetic model (Fig. 3 Figure 7, and kinetic constants used to fit the data are given in Table 3 In the kinetic model (Fig. 3), the velocity (v) also serves as a measure of the rate of ET as expressed as a first-order timeaveraged rate constant for ET, and the minimal k ET value (4,5). Furthermore, changes in the NADH dependence of the velocities of variant NODs reveal alterations in the ET gating efficiency, where the v approximates the k ET value owing to the gating mechanism, provided that hydride transfer, and the k' H value, remains constant for FAD reduction (Fig. 3). Thus, [NADH]k' H determines the steady-state FAD/FADH 2 ratio, and the reduction potential that directly influences the ET rate through a tunneling matrix (44), and the absolute k ET value. In the steady state, Δv/Δ[NADH] values thus provide a sensitive, albeit indirect, measure of the putative LysF7-H 2 O-GlnE7 ET gate function during catalysis (Fig. 2D) where a smaller Δv/ Δ[NADH] value indicates a decreased rate of ET, an inefficient tunneling matrix, and dysfunctional gating mechanism. As shown in Figure 8, the TrpE11 mutant, all E2 substitutions, and the HisE7 mutant show increased slopes in 1/v versus 1/    Table 3.

Table 3
Kinetic constants for the B10 and E7 flavoHb-NOD variants Allosteric nitric oxide dioxygenase mechanism [NADH] plots at 37 and 20 C and smaller Δv/Δ[NADH] values (Table 4). In addition, the AlaE11 substitution shows an increased slope and smaller Δv/Δ[NADH] value at 20 C. Dysfunction of ET in the TrpE11, LeuE2, LeuE7, and HisE7 substitutions is observed at 37 C, and even greater effects of all E11 and E2 substitutions and the HisE7 substitution are observed at 20 C. LeuE7 shows a small stimulatory effect at 20 C compared with GlnE7. The G8 substitutions show only minor effects, and those are observed at 20 C with Ala decreasing and Leu increasing the efficiency. The PheB10 substitution also shows only minor effects. The results demonstrate roles for GlnE7, LeuE11, and GlnE2 in ET consistent with the proposed mechanism and kinetic model. The analysis also indicates that higher NADH concentrations, and thus more negative flavin redox potentials, can drive catalysis and ET in a defective gate mechanism. The low-temperature dependence of Δv/Δ[NADH] also indicates a relatively low energy barrier for the hydride and electron transfer steps.

Kinetics of hydride and ET
As previously reported (4), NADH reduces FAD to FADH 2 with a large bimolecular rate constant for hydride transfer (k 0 H ) of 1.5 × 10 7 M −1 s −1 as detected by the transient decrease in FAD absorbance at 460 nm (Fig. 9A, line 1). Similar FAD reduction rates are observed for the PheB10, LeuE7, and HisE7 variants (lines 2-4, respectively) with the HisE7 mutant showing a greater extent of FAD reduction. FADH 2 rapidly reduces the WT flavoHb ferric heme to the ferrous state with a maximum k ET value of 120 s −1 at 20 C as measured by the transient increase in absorbance at 430 nm (Fig. 9B, line 1), similar to the 136 s −1 value previously reported (4). In comparison, the apparent k ET values for the PheB10, LeuE7 and HisE7 variants are 160, 140 and a biphasic 1.7 s −1 or 50 s −1 , respectively (lines 2-4). It is noteworthy that the HisE7 variant showed a biphasic behavior with an initial slow reduction   Allosteric nitric oxide dioxygenase mechanism followed by a more rapid phase with incomplete reduction (line 4) despite achieving a more negative FAD/FADH 2 reduction potential (Fig. 9A, line 4). The HisE7 variant appears to favor the off conformation (see Discussion). In contrast, WT, LeuE7, and PheB10 flavoHbs appear to favor an on conformation in the resting state. Both the k 0 H and k ET values of the WT, LeuE7, and PheB10 flavoHbs are within the range of values required to support the NOD turnover rates measured at 20 C (Tables 1 and 3). The transient kinetic data further support a role for GlnE7 movements in the ET gate function. ET rates, and k ET values, depend upon the restingstate structure and the FAD/FADH 2 potential.

O 2 , NO, and CO binding
The kinetic model predicts specificity and dynamic changes in ligand migration pathways during catalysis. Indeed, a heterogeneity in ligand migration pathways was evident in previous biphasic ligand binding kinetics (4,5). However, biphasic kinetics was attributed to a structural heterogeneity in the isolated flavoHbs, rather than to the existence of multiple and dynamic pathways, as recently argued for CO migration in Cygb (50).
A critical reanalysis of the ligand binding kinetics of flavoHb shows evidence of dynamic changes and multiple pathways. Following the high-energy laser photolysis perturbation, fla-voHb rebinds O 2 , NO, and CO with biphasic kinetics (Fig. S3) with transition times (τ T ) for the fast and slow phases in the 10-to 200-μs range (  (1,49), and together with the kinetic model in Figure 3, help to explain the steady-state kinetic data in Figure 7 and Table 3. Moreover, the 29-fold larger k(O 2 ) value measured for the HisE7 variant relative to that for GlnE7 ( Table 5) further suggests that the HisE7 variant, in its resting state, favors the off conformation with positively charged HisE7 in proximity to the negatively charged heme propionate as modeled for GlnE7 in Figure 2D (top). In addition, the comparable k 0 slow (NO) and k 0 slow (O 2 ) values and the large k(NO) values for the ferric flavoHbs (Table 5) may account, in part, for the reversible NO inhibition reported in Figure 1 and modeled in Figure 3. Conversely, comparable k 0 slow values for O 2 and NO and extremely small k(NO) and K d (NO) values for the ferrous flavoHb (Table 5), and an absence of irreversible NO inhibition, largely preclude the previously suggested pingpong mechanism (3,4). Applying the k(NO) values, we can calculate 90% activity recovery following NO binding in 3.2 h for the ferrous NO species and a mere 0.6 ms for the ferric NO form. With NO elimination via reduction (≤0.1 s −1 ) (4, 5), 90% reversal in the steady state would require 23 s. However, NO inhibition is competitive with O 2 and seamlessly transient under the steady-state conditions of Figures 1 and 4-7 indicating a predominant competition for the LT entry and ferric iron as modeled in Figure 3. NO and O 2 show remarkably similar k 0 slow values for Fe 3+ and Fe 2+ , respectively, suggesting a similar entry and migration through the LT and interactions with LeuE11.

Role of LysF7 in ET control
LysF7 was investigated for its role in gating ET with an AsnF7 substitution that removes the proposed toggling epsilon ammonium group in the gate structure (Fig. 2D). The starting activity of the AsnF7 enzyme is 6-fold less than the LysF7 (WT) enzyme and rapidly inactivated. Ninety percent of the starting activity is lost in less than 1000 turnovers or 30 s (Fig. 10A, compare line 1 with line 2). In the absence of NO, the AsnF7 enzyme loses 85% of its NOD activity in 90 s with apparent zero order kinetics (Fig. 10B, line 1). Thus, NO is not essential for the inactivation, but NO greatly increases the inactivation. Catalase provides modest protection (Fig. 10B, compare lines 1 and 2) suggesting formation of Fe 3+ -OOH, and H 2 O 2 , and oxidative damage to the heme. Indeed, flavoHb-NOD activity is sensitive to H 2 O 2 , and the heme ligand miconazole (51) and the peroxidase substrate ascorbate both protect in the resting state (Fig. S5). The AsnF7 variant shows other indications of changes in ET fidelity or control including a 2.8-fold greater NADH oxidase activity (0.76 versus 0.27 s −1 ), a 2.4-fold greater NO reductase activity (0.45 versus 0.19 s −1 ), a fourfold higher methyl viologen reductase activity (1.45 versus 0.36 s −1 ), a 4.6-fold higher cytochrome b 5 reductase activity (0.65 versus 0.14 s −1 ), a 4.5-fold higher Ngb reductase activity (0.90 versus 0.20 s −1 ), and a twofold higher trimethylamine-N-oxide (TMAO) reductase activity (0.41 versus 0.21 s −1 ). CO (50 μM) inhibits TMAO reduction by 90% suggesting an increased ET and reduction by the heme. Other reductase activities are relatively resistant to 50 μM CO. Increased rates of reduction demonstrate an increased ET from the flavin and a loss of the electron tunneling path directionality (44) and coupling control in the mutant.
Together these data support a role for the conserved LysF7 epsilon ammonium group in forming part of a switchable gate, or dynamic matrix, for ET that controls the rate, timing, and tunneling pathway for ET. Mistiming of heme reduction may also cause irreversible NO inhibition and a rapid inactivation; i.e., a mistiming of ET creates the opportunity for NO to bind Fe 2+ irreversibly during turnover.

Discussion
Cooperative activation of the NOD function of the E. coli flavoHb by the substrates O 2 and NO is explained by motions of the protein triggered by the gases and the product nitrate. Motions, although not directly observed, are inferred from the available x-ray crystal structures and from the effects of conserved amino acid substitutions on the steady-state and transient kinetics. Deficiencies and paradoxes of the rudimentary ping-pong model (3,4) are resolved by an allosterically controlled ping-pong mechanism in which ET is activated by O 2 and NO positioning (Fig. 3). A critical reevaluation of NADH-driven enzyme velocities (Table 4) and heme reduction rates demonstrates activation of ET by O 2 and NO. Comparable changes in the rates of heme reduction in the absence of O 2 and NO require approximately fourfold greater NADH concentrations (4), and thus more negative flavin redox potentials and a greater driving force. Furthermore, ET transfer rates measured by stopped flow (Fig. 9) reveal on and off states in resting state structures representative of steps in the dynamic switching mechanism. From the data in Tables 1-4, we can begin to estimate times and activation energies for the steps in catalysis and compute the effects of various mutations. For example, from the V max values, we can calculate that the rate-limiting step of the WT enzyme requires 1.7 and 6.3 ms at 37 and 20 C, respectively. The near diffusion-limited V max /K 3 value of 2.3 × 10 9 M −1 s −1 and the four-to eightfold larger V max /K 4 value at 37 C indicates that multiple NO collisions and the wiggling of G8, E15, E11 side chains within the ST are required for, and limit, ET activation and the NOD reaction. Applying the Arrhenius equation for V max at the two Allosteric nitric oxide dioxygenase mechanism temperatures, we can approximate an overall activation energy (E a ) equal to 13.7 kcal/mol, an amount equivalent to that required to break three to four hydrogen bonds. The largest effect on the E a is observed with the AlaE11 substitution with a calculated E a of 24.6 kcal/mol and a 5-to 15-fold larger temperature-dependent K T value further suggesting that LeuE11 movements due to NO collisions, and concerted motions of the CD loop, E and F-helices, and LysF7, form the largest energetic barrier, and also the rate-limiting step in the NOD mechanism. Following this analysis, we can also see that the forward rate constant for the trigger and switching mechanism, k 0 T , limits and equals the V max value. ET in flavoHb (k ET ), on the other hand, has been reported to be 40-fold faster by time-resolved pulse radiolysis (6800 s −1 ) (39) than the maximal NADH-driven ET rate measured by stopped flow (160 s −1 ) (Fig. 9B) (4) indicating that ET per se is not rate limiting and suggesting experimental differences in the matrix, transfer distance, or FAD/FADH 2 potential (E 0 = < −0.15 V) (52). Furthermore, the HisE7 substitution may impair ET in the steady state ( Fig. 8 and Table 4) and resting state (Fig. 9B, line 4) by strengthening bonding to the heme propionate in the ET gate structure and stabilizing the off state (Fig. 2D, top) and, moreover, by significantly decreasing the nominal −0.125 V heme mid-point potential (52). The E7 position is too distant for the histidine to coordinate with the iron. Nevertheless, the structures (Fig. 2C) together with spectral analysis (unpublished results) suggest a relaxed to tense motion in the heme corresponding to a high spin (S = 5/2) to low spin (S = 1/2) ferric iron transition (53) with a large calculable heme mid-point potential drop of 0.09 V that may account for decreased extents of heme reduction despite lower FAD/FADH 2 potentials and a greater driving force (Fig. 9B). Furthermore, thermally fluctuating spin state mixtures in the resting state may account for the observed intermediate ET rates (Fig. 9B). Accordingly, the proposed conformational changes of the F-helix in the trigger step (K T ) force a low to high spin transition and associated potential increase, besides switching the ET gate on (Figs. 2D and 3). Although matrix perturbations in outer sphere ET processes, as modeled for flavoHb in an on state (Fig. 2D, bottom) (38,39), show greater effects on k ET values than small differences in ΔE 0 values with generally large 1-V reorganization energies (44), corrections to the nominal heme midpoint potential due to a mixed resting ferric iron spin state will also influence k ET values following the Marcus electrontransfer equation (38,39,44).
The free energy released in the reaction of NO and O 2 − forming peroxynitrite is 29 kcal/mol (54), and the reaction of NO with MbFe 3+ O 2 − to form nitrate releases more than 30 kcal/mol (55). The Mb-catalyzed isomerization of peroxynitrite to nitrate occurs in the submillisecond time range (6,56) indicating that k is would not be rate limiting. In the model (Fig. 3), NO 3 − release (k p ) and efflux are coupled with CD loop unfurling-furling, and the rate constants thus equal V max . We suppose that the initial step in the exergonic NO dioxygenation reaction, as represented by k' 4 , where k' 4 equals V max /K 4 and k' NOD , ultimately provides the energy required to drive protein motions and support the catalytic cycle. Furthermore, by analogy, a 577-nm photon, bearing 49.6 kcal/mol, and absorbed by the flavoHb during photolysis of O 2 or other bound ligands, would suffice to trigger large protein motions via LeuE11, TyrB10, and GlnE7 interactions and contribute to the microsecond-scale heterogeneity of ligand migration evidenced by the data in Tables 5 and S5; and Figs. S3 and S4, and elsewhere (4,5). In the photolysis of MbFe 2+ CO, most motion occurs in the nanosecond range, but CO-induced structural changes and hydration shell fluctuations, nevertheless, continue for microseconds (57)(58)(59)(60).
In the model, K 2 , not K 1 , and the apparent K m (O 2 ) is only roughly approximated by V max /k' slow (O 2 ). K 1 may also be further broken down to movement of O 2 through the strictures between the docking sites S3, S2, and S1 with a relatively large 6 kcal/mol energy barrier estimated for the narrow stricture 2 controlling passage from S2 to S1 (38). Displacement of LeuE11 by nitrite, and presumably nitrate also, widens the clearance of A16-G16 in stricture 2 by 1.6 Å, and the clearance of E15-H12 in stricture 1 by 3.1 Å (Table S2). Indeed, the K i(NO) values of 0.3 to 1.2 μM determined for NO competition with O 2 for LT entry (Table 1), and also comparable K i(CO) values of 1 μM (5) A surprising outcome or prediction of the allosteric model is that a paramagnetic O 2 molecule, with an induced dipole, interacts magnetically with the paramagnetic ferric iron atom in displacing LeuE11 and, moreover, that TyrB10 and GlnE7, as well as the G-helix dipole, affect this interaction. Indeed, the Fe 3+ and O 2 interactions are relatively weak given the large micromolar K 2 values (Tables 1 and 3) relative to the corresponding nanomolar K d values for Fe 2+ O 2 ( Table 5). The strictured and segmented LT further suggests potential mechanisms for the controlled storage and influx of O 2 in the NOD function of flavoHb and roles for similar tunnels in other globin-NODs including HbN, the alpha subunit of HbA (18,61), and Cygb (10,17,36,37,50). The failure of O 2 to compete with NO suggests strong discrimination against O 2 entry by the ST. In the S. cerevisiae flavoHb ferric-nitrite structure, water molecules hydrogen bonded to GlyH9 and ProG6 peptide carbonyls guard the ST entryway. Nonpolar O 2 does not readily penetrate hydration shells on protein surfaces (62), and some protein sites bind water more tightly than others (63). Polar NO may displace the polar water molecule(s) and enter, whereas O 2 may not. In the case of thermoglobin (64) (Table S3), a putative NOD, the unique charged LysG9 and GluH9 pair may better serve the NO gating function at the hydrogen bond-breaking 95 C of the archebacterial existence. The failure of ubiquitous and abundant N 2 , the carrier gas, to compete with O 2 for LT entry and inhibit the NOD function also requires explanation. The greater than 80-fold selectivity for O 2 over N 2 apparent in the steady-state data (Fig. 1A) may be achieved by a discriminator for the inducible lateral O 2 dipole. The diametric PheG20 phenyl group and ThrA12 Oatoms within stricture 3 ( Fig. 2B and Table S1) may serve this function by conjugating and repelling π*/π molecular orbitals thus allowing O 2 passage, but excluding entry of the less polarizable N 2 molecule.
The model also predicts that motions in the LysE7 epsilon ammonium group, triggered by NO motions and passage by the LeuE11 barrier in the ST, ultimately control the on and off states and tunneling path for ET, and, furthermore, that GlnE7 and GlnE2 movements participate in the switching mechanism as evidenced by the strong inhibitory effect of the HisE7 and E2 substitutions on ET (Table 4 and Figs. 8 and 9B). We suppose that a small K T value, and a strong trigger, will favor the on position and that larger K T values, like that shown by the HisE7 variant, will favor the off position in the resting state. Further exploration of the effects of mutations, and tunneling matrix modifiers or dopants, on ET rates in flavoHbs should be illuminating provided that the fine structure of the switch mechanism can be mirrored, and the FAD/FADH 2 potential controlled, in the resting state(s). Furthermore, the model and The data and model can also be used to determine the physiological limits of the allosteric NOD mechanism. The data in Figure 1C demonstrate kinetics consistent with the mechanism at a physiologically relevant 8 μM O 2 with a velocity of 8 s −1 at 2 μM NO. Applying Equation 3 and the constants in Table 1, we can calculate that at 0.4 μM O 2 and 1 μM NO, the velocity of the allosteric NOD activity approximates the V max of 0.1 s −1 for Fe 2+ NO reduction (4,5). A similar comparative analysis can be done to evaluate the functional significance of the bimolecular reaction of O 2 with Fe 2+ NO/Fe 3+ NO − (11), but the dependence of the reaction on [O 2 ], and a small rate constant, make the reaction inconsequential (2). Consistent with this analysis, flavoHb fully protects E. coli growth under an atmosphere containing 5 μM O 2 and 0.5 μM NO (67) and shows insignificant NO metabolic activity in anaerobic E. coli (68).
Finally, the aggregate data and model suggest that all globins functioning as NODs have evolved allosteric mechanisms for controlling gas movements (69), timing ET (10), expelling nitrate, and averting irreversible NO inhibition and a reductive inactivation. Further investigations of ET in flavoHb promise to increase our understanding of conformationally gated ET, the NOD mechanism of globins, and the evolutionary origins of the relaxed-tense heme transitions for allosteric cooperativity in O 2 binding by O 2 -transporting Hbs (53).

Reagents
Reagents were purchased from Sigma-Aldrich Chemical Co (St Louis, MO, USA) unless otherwise stated. DNA primers were synthesized at the University of Cincinnati Medical Center DNA facility. Praxair (Bethlehem, PA, USA) was the supplier of 99.993% O 2 , 99.999% CO, and 99.98% N 2 . Bovine liver catalase (260,000 U/ml) was obtained from Roche Applied Science. Recombinant rat cytochrome b 5 (soluble form) and murine Ngb were expressed in E. coli and purified essentially as described (10,70). NADH purity (≥95%) was assessed by the ratio of the 340 to 260 nm absorbance with a value of 0.437 equal to 100% reduced.

Expression and purification of NODs
Cultures of AG103 bearing pUC19hmp, and engineered variants thereof, were grown microaerobically in a phosphatebuffered medium containing yeast extract, tryptone, glucose, nitrate, hemin, and ampicillin and were harvested, purified, and analyzed as previously described (4). Batches of 1-l cultures were grown microaerobically in 2-l flasks by a slow gyrorotatory motion and were harvested after 16 h. Isolated flavoHbs contained between 0.6 and 0.9 mol fraction of heme and between 0.3 and 0.9 mol fraction of FAD. FlavoHbs were reconstituted with heme and FAD (4,72).

Model for NOD kinetics and fitting
Rate constants were estimated from the model kinetic equation using Python programming with successive approximations to optimize fits with varying NO and O 2 concentrations. V max , K 1 , K 2 , K T , and K ii(NO) values were treated as independent variables for a given flavoHb while allowing K 3 , K 4 , and K i(NO) to vary over limited ranges with varying O 2 concentration. K ii(NO) was estimated from NO inhibition at saturating O 2 concentrations. K i(O2) and K a(O2) were included as independent variables in order to model the AlaE11 mutant data.

Reduction and ligand binding kinetics
FAD and heme reduction were monitored by following the changes in flavoHb absorbance at 460 and 430 nm, respectively, with stopped-flow rapid mixing of 50 μM NADH and 10 μM flavoHb in N 2 -purged buffer (4). O 2 , NO, and CO association rate constants were determined by photolyzing 50 μM of the respective ferrous flavoHb species with a 300-ns pulse from a 577-nm laser and by spectrophotometrically following ligand reassociation in O 2 -, NO-, or CO-saturated buffer unless otherwise indicated (4,5). Binding of O 2 or CO was measured by the change of absorbance at 436 nm by applying the bound versus unbound difference extinction coefficients of 0.046 and 0.053 μM −1 cm −1 , respectively. Binding of NO was measured by the change of absorbance at 422 nm by applying a difference extinction coefficient of 0.040 μM −1 cm −1 . O 2 and CO dissociation rate constants were determined by stopped-flow rapid mixing with either CO or NO displacement (4,5). The cuvette path length was 1.0 cm. All reactions were at 20 C in filtered 100 mM sodium phosphate buffer, pH 7.0, containing 0.3 mM EDTA. Catalase (1000 U/ml) was included in all reactions containing O 2 to scavenge H 2 O 2 produced by the NADH oxidase activity of flavoHb and to regenerate O 2 .

Data analysis
Data are representative of two or more trials. Structures were modeled and analyzed using Jmol version 14.2.15.

Data availability
All data are contained within the manuscript.