Conformational plasticity surrounding the active site of NADH oxidase from Thermus thermophilus

Abstract

Biotechnological applications of enzymes can involve the use of these molecules under nonphysiological conditions. Thus, it is of interest to understand how environmental variables affect protein structure and dynamics and how this ultimately modulates enzyme function. NADH oxidase (NOX) from Thermus thermophilus exemplifies how enzyme activity can be tuned by reaction conditions, such as temperature, cofactor substitution, and the addition of cosolutes. This enzyme catalyzes the oxidation of reduced NAD(P)H to NAD(P)⁺ with the concurrent reduction of O₂ to H₂O₂, with relevance to biosensing applications. It is thermophilic, with an optimum temperature of approximately 65°C and sevenfold lower activity at 25°C. Moderate concentrations (≈1M) of urea and other chaotropes increase NOX activity by up to a factor of 2.5 at room temperature. Furthermore, it is a flavoprotein that accepts either FMN or the much larger FAD as cofactor. We have used nuclear magnetic resonance (NMR) titration and ¹⁵N spin relaxation experiments together with isothermal titration calorimetry to study how NOX structure and dynamics are affected by changes in temperature, the addition of urea and the substitution of the FMN cofactor with FAD. The majority of signals from NOX are quite insensitive to changes in temperature, cosolute addition, and cofactor substitution. However, a small cluster of residues surrounding the active site shows significant changes. These residues are implicated in coupling changes in the solution conditions of the enzyme to changes in catalytic activity.

Introduction

Enzymes have many industrial applications, including biocatalysis, bioremediation, and biosensing.^1–6 In many instances, this involves employing enzymes under conditions that are quite different from those found physiologically in terms of temperature and cosolutes. An important challenge in the rational use and optimization of enzymes in biotechnology is to understand at a molecular level how changes in an enzyme's environment affect its structure and dynamics and how this governs catalysis. For example, moderate concentrations of denaturants such as urea or chaotropic anions in the Hofmeister series increase the catalytic activity of some enzymes.^7–11 This effect has been linked to changes in the structure of the active site⁷ and modulation of protein dynamics,^8,11 yet many questions remain regarding the molecular mechanisms underlying this phenomenon. Conversely, thermophilic enzymes generally have lower activity than mesophilic ones at ambient temperatures, and only regain activity as the temperature is raised.¹² Nevertheless thermophilic enzymes are associated with several practical advantages; they can perform at higher temperatures, are less prone to microbial contamination, have lower viscosity, and heat treatment simplifies purification.^13,14 Furthermore, the substitution of one cofactor for another and the engineering of non-natural cofactors can improve the cost efficiency of biocatalysis and expand the repertoire of available enzymatic reactions.^15–17 Thus, a better understanding of active site adaptation to cofactor substitution and thermal and denaturant activation of enzymes is desirable in order to elucidate design principles for improved biocatalysts.

NADH oxidase (NOX) from the thermophilic bacterium Thermus thermophilus is a dimer of 27 kDa monomers that catalyzes the oxidation of reduced nicotinamide adenine dinucleotide (NAD(P)H) with the concurrent reduction of oxygen to hydrogen peroxide in a two electron transfer mechanism. It is functional with either FMN or FAD cofactors bound in a 1:1 monomer:cofactor ratio.^18,19 FAD is significantly larger than FMN, as illustrated in Figure 1, differing by the addition of an adenosine monophosphate group. The physiological role of NOX is still not well understood although some homologous NOX enzymes may be engaged in the defense against oxidative stress.²⁰ NOX has high thermal and chemical stability, relatively low molecular mass, broad pH activity, and the ability to accept NADH and NADPH as substrates, making it attractive for biosensor applications.²¹ In addition, it belongs to a protein family that performs valuable chemical modifications of prodrugs²² and explosive contaminants such as 2,4,6 trinitrotoluene (TNT).⁶ NOX is most active at approximately 65°C, near the physiological temperature of T. thermophilus, and exhibits approximately sevenfold lower activity at 20°C.¹¹ The activity of NOX is enhanced in the presence of low concentrations of denaturants, such as urea and chaotropic anions of the Hofmeister series.¹⁹ Conversely, NOX activity is diminished by kosomotropic agents that promote hydrophobic interactions. For example, the maximum turnover rate is about 2.5-fold higher with the addition of 1M urea and threefold lower with 1M Na₂SO₄ at 20°C.¹¹

Figure 1.

NOX cofactor structures. flavin adenine dinucleotide (FAD) (a) and flavin mononucleotide (FMN) (b).

X-ray crystallographic structures have been determined for NOX in complex with both FMN and FAD²¹ however there are no structural data for the apo form of the enzyme. Previous studies have also examined the temperature and denaturation activation of NOX by molecular dynamics simulations and fluorescence spectroscopy,^11,23–25 but high-resolution experimental information is lacking. We have investigated the effects of temperature and urea on NOX as well as its interactions with FMN and FAD at atomic resolution using a combination of nuclear magnetic resonance (NMR) titration experiments, ¹⁵N spin relaxation NMR experiments, and isothermal titration calorimetry (ITC). We find that FMN binds NOX approximately 1250-fold more tightly than FAD, suggesting that FMN is the preferred cofactor. The NMR spectrum of the apoenzyme is quite similar to those of the FMN-bound and FAD-bound holo forms, suggesting that the enzyme retains its fold in the absence of cofactors. The ¹⁵N spin relaxation measurements report the amplitudes of NH bond vector motions on the ns to ps timescale. We find that the addition of 1.2M urea at 50°C results in some redistribution of flexibility, including several residues in the active site. However, the overall level of dynamics remains relatively constant for the protein as a whole. Series of ¹⁵N/¹H NMR correlation experiments were collected as the temperature and concentrations of FAD, FMN, and urea were varied. These measurements identify the regions of the protein whose conformations and interactions are most affected by changing these environmental variables. Interestingly, most NMR signals from NOX are relatively insensitive to these perturbations. However, the peaks from a few small clusters of residues surrounding the active site shift significantly in response to multiple environmental factors. These structurally malleable sites of the protein are likely involved in the temperature and/or denaturant activation of the enzyme, as well as its ability to accommodate different cofactors. Interestingly, the side chain indole ¹⁵N/¹H signal of W47 shows among the largest peak movements in response to urea addition and cofactor substitution. This side chain has been previously suggested to play a key role in gating substrate access to the active site and mediating denaturant activation of the enzyme.²³

Results

Tryptophan side chain assignments

Previously, we assigned 165 of the 190 backbone peaks expected in ¹H–¹⁵N HSQC correlation spectra for nonproline residues of NOX, using standard triple-resonance experiments.²⁶ For this study, we had particular interested in the NH signals from the indole side chains of the tryptophan residues in NOX, since the side chain of W47 is believed to play a key role in catalysis.^11,23 NOX contains four tryptophans, two buried, W47 and W52, and two solvent exposed, W131 and W204. Only three ¹H–¹⁵N signals are observed in the characteristic tryptophan indole region of ¹H–¹⁵N correlation spectra obtained for NOX (downfield in both ¹H and ¹⁵N). Two of these were tentatively assigned to the side chains of W131 and W204 based on NOE correlations between indole and backbone protons, but the remaining indole peak was unassigned. In order to confirm these two assignments and to determine the identity of the third indole peak, we employed site-directed mutagenesis. The conservative Trp to Phe substitution was made for each tryptophan residue (W47F, W52F, W131F, W204F). Indole regions of ¹H–¹⁵N HSQC correlation spectra of the apo forms of the WT and three of the mutants are shown in Figure 2. The Phe side chain is aromatic and has similar size to that to that of Trp but produces no ¹H/¹⁵N correlation peak, therefore the expected outcome of a Trp to Phe mutation is the disappearance of a single indole resonance from the ¹H/¹⁵N correlation spectrum with minimal additional spectral perturbation. Indeed, one of the three wild-type Trp NH indole side chain peaks is clearly absent in each of the HSQC spectra collected for the W47F, W131F, W204F mutants. These results agreed with our tentative assignments for the W204 and W131 side chains and indicated that the third indole peak arises from the W47 side chain. Since the backbone signal for W47 is absent, this assignment is consistent with the fact that no NOE connectivity was found between this third side chain peak and a corresponding backbone signal. In the case of the W52F mutant, the HSQC of the apo form resembles that of an unstructured protein,²⁷ with many proton shifts clustered around 8.5 ppm. (Supporting Information Fig. S4). This is consistent with the observation that W52F has diminished flavin binding affinity.²⁸ W52 is deeply embedded in NOX structure at the beginning of the first β-strand of the four-stranded antiparallel β-sheet located in the core of the protein and adjacent to R53, which is involved in an intermolecular salt-bridge network according to molecular dynamics (MD) simulations.²³ These interactions provide some rationalization for the intolerance of the protein to substitutions at this position.

Figure 2.

Indole side-chain regions of NOX ¹H/¹⁵N NMR spectra. Wild-type NOX (a), and the mutants W47F (b), W131F (c), and W204F (d) recorded at 50°C and 18.8 T (800 MHz ¹H. Larmor frequency) in the absence of cofactor.

FMN and FAD binding

NOX accepts either flavin mononucleotide (FMN) or flavin adenine dinucleotide (FAD), which bind in a 1:1 cofactor:monomer ratio.²⁹ The structural difference between these two molecules is the addition of an adenosine monophosphate group in FAD (Fig. 1). We characterized the binding of FMN and FAD to NOX using isothermal titration calorimetry (ITC), which yields equilibrium dissociation constants, K_D, binding enthalpies, ΔH, and binding entropies, ΔS, of the interaction. The ITC data are well fit by the one-set-of-sites binding model (Fig. 3), which implies that cofactors bind to the two subunits of NOX essentially independently. FMN binds NOX much more tightly than does FAD, with K_D values of 0.4 ± 0.2 μM and 0.54 ± 0.03 mM for FMN and FAD, respectively. These results support the idea that FMN is the preferred, or physiological, NOX cofactor.²¹ Interestingly, although FAD binds much less tightly than does FMN, the reaction enthalpies for the two ligands are very similar with ΔH values of –10.1 ± 0.4 and –9.4 ± 0.7 kcal/mol for FAD and FMN, respectively. The weaker binding in the case of FAD is due to a strongly unfavourable entropy term, ΔS = −16 ± 1 cal mol⁻¹ K⁻¹, compared to a negligible binding entropy for FMN, ΔS = 0.2 ± 3 cal/(mol K).

Figure 3.

NOX cofactor titrations with FMN (a, b) and FAD (c–f). Raw differential power ITC isotherms for NOX with FMN (a) and FAD (c). Integrated ITC injection heats plotted as a function of the cofactor:monomer molar ratio for FMN (b) and FAD (d). Data were fit to a 1 set of identical sites binding model. FAD injections into buffer alone are shown in the inset to panel (e). NOX ¹H (e) and ¹⁵N (f) chemical shift displacements plotted as a function of the ligand:protein molar ratio. Lines correspond to fits according to Eq. 2.

We also studied the interaction of the cofactors with NOX by performing NMR titrations, collecting series of ¹H/¹⁵N HSQC protein spectra as FMN or FAD was incrementally added to the sample. Many peaks shifted gradually as cofactors were added and exhibited some broadening at intermediate points in the titration, characteristic of binding kinetics that are on the intermediate to fast timescale regimes. Under these conditions, the shifts in peak position as a function of ligand concentration can be analyzed to yield binding constants.³⁰ In the case of FMN, the affinity is too high to reliably extract a binding constant from the NMR titration data, as the protein concentrations required for NMR are much greater than the K_D. In the case of FAD, the shifts in peak position were analyzed to yield an estimate of the equilibrium binding constant, K_D = 0.6 ± 0.3 mM, that matches the ITC-derived value well, as shown in Figure 3.

Interestingly several peaks that are present in the spectra of NOX bound to either FMN or FAD are missing in the absence of cofactors. These peaks reappear gradually as cofactor is added. The affected residues cluster around the active site and include residues A46, and Q105 on the re face and G72, A74, and G159 on the si face of the cofactor flavin ring as well as S193, H194, and L196 in the C-terminal tail of NOX which is also involved in cofactor binding [Fig. 4(a)]. There are two likely causes of these peak disappearances. Firstly, the NMR experiments were performed at 50°C and pH 7.2, in order to match previous studies of NOX catalysis.¹¹ Under these conditions, an unprotected amide hydrogen is expected to exchange with those of water with a rate on the order of 300 s⁻¹, and is therefore not expected to be detected in ¹H NMR experiments.^31,32 Amide groups that are strongly hydrogen bonded or buried deeply in the protein interior are typically protected from solvent exchange, with rates reduced by factors on the order of 10⁴ to 10⁸.³³ Even in unfolded proteins, hydrogen exchange can occur more than tenfold more slowly for amides whose surroundings partially exclude water molecules, for instance due to hydrophobic collapse.^33,34 Removal of the cofactor could enhance the hydrogen exchange rates of the residues listed above, leading to their disappearance from NMR spectra. Notably the amide hydrogens of A46 and G72 participate in intramolecular hydrogen bonds in the bound form, according to the X-ray crystal structure.²¹ Enhanced hydrogen exchange rates for these residues would imply that removing the cofactor results in conformational changes that weaken or break these intramolecular hydrogen bonds. For the others, removal of the cofactor could lead to greater access to water molecules. Alternatively, the disappearance of some or all of these eight resonances could be due to dynamical broadening caused by ms-μs timescale dynamics that are present in the NOX active site when cofactor is absent and are quenched when cofactor is present. In either case, A46, G72, and possibly other residues experience a change in conformation and/or flexibility when cofactor is removed.

Figure 4.

NOX chemical shift perturbations by environmental variables. Backbone ribbon diagrams of NOX²¹ (PDB 1NOX), with the two subunits colored in blue and white and the FMN cofactor illustrated in green balls and sticks. In (a), dark purple spheres correspond to residues whose ¹H/¹⁵N peaks are missing in spectra of apo NOX but not in the presence of FMN or FAD. In (b–d) spheres correspond to residues with larger than average ¹H/¹⁵N chemical shift perturbations (ppm) in response to changing environmental variables: FMN vs. FAD [>0.25 dark purple; 0.17–0.25 purple; 0.11–0.17 dark pink; 0.10–0.11 light pink] (b) 25–50°C [|∂δ_N/∂T| > 17 ppb/K dark pink; ∂δ_H/∂T < −4.5 ppb/K light pink; both [|∂δ_N/∂T| > 17 ppb/K, and ∂δ_H/∂T < −4.5 ppb/K dark purple] (c) and 0 vs. 2M urea [>0.55 dark purple; 0.45–0.45 purple; 0.35–0.45 dark pink; 0.32–0.35 light pink] (d). Small white spheres on the white monomer indicate the locations of residues lacking crosspeaks in ¹H/¹⁵N correlation NMR spectra. Malleable residues listed in Table2 are labeled on the structures. Image generated with UCSF chimera.⁶⁹

We then compared the HSQC spectra of NOX saturated with FMN and FAD (Supporting Information Fig. S5) in order to pinpoint regions of the protein that are most affected by cofactor substitution. The ¹H and ¹⁵N chemical shift differences between corresponding peaks in the two spectra were used to calculate Δδ, where

1

The positions of most peaks do not change significantly with the substitution of FAD for FMN. The average Δδ value is only 0.082 ppm with a standard deviation of 0.009 ppm. Some of the largest shifts are exhibited by residues G190, Y191, and S193 with Δδ = 0.56, 0.20, and 0.20 ppm, respectively [Fig. 4(b)]. A second cluster of residues consisting of R109, E110, Q112, and K113, in helix F, and Q105, in the loop preceding helix F also have high Δδ values (0.19, 0.18, 0.24, and 0.12 ppm, respectively). Both of these groups of residues abut the additional adenine that distinguishes FAD from FMN cofactor (Supporting Information Fig. S9). Ring currents in the aromatic adenine are expected to alter the chemical shifts of nearby ¹H nuclei in an orientation-specific manner.³⁵ Therefore, these changes in peak position likely reflect the formation of new contacts with the larger FAD cofactor and/or conformational rearrangements. Signals for A46 and N48 and the W47 side chain (scW47) also exhibit large backbone chemical shift variations with Δδ values of 0.20, 0.19, and 0.17 ppm, respectively, as do A70, F71, G72, Q73, A74, and G159, F160, D161. These clusters of residues surround the flavin ring of the cofactor, whose position is conserved in the FMN- and FAD-containing structures (Supporting Information Fig. S9). These changes in chemical shift therefore are likely to reflect rearrangement of active site rather than the formation of new protein-cofactor contacts.

Temperature variation

In order to gain some insight into how raising the temperature increases the activity of NOX, we have examined temperature-dependent changes in spectra of the protein. ¹H–¹⁵N HSQC correlation spectra were recorded for FMN-saturated NOX at temperatures ranging from 25 to 50°C in 5°C intervals [Fig. 5(a)]. Both ¹H and ¹⁵N chemical shifts exhibited linear dependences on temperature [Fig. 5(a)]. The temperature coefficients (slopes of δ vs. T plots) range from −13.85 to 0.66 ppb/K for ¹H and −41.20 to 19.07 ppb/K for ¹⁵N with means values of −2.86 and −5.48 ppb/K respectively. No significant correlation is evident between the ¹H and ¹⁵N temperature coefficients of the same peaks (r = 0.18). Amide groups involved in stable secondary structure typically have proton temperature coefficients less negative than −4.5 ppb/K. Amides with ¹H temperature factors more negative than −4.5 ppb/K are most likely not hydrogen bonded³⁶ and are indicated with black bars in Figure 5(d). As expected, most residues with ¹H temperature factors less than –4.5 ppb/K lie in loops and at the N- and C-termini of the protein. However, several residues with quite negative ¹H temperature factors are located within elements of secondary structure, for instance in helices D, E, F, and G. One of these, A74, lies in the active site, on si side of the cofactor flavin ring. In these cases, the highly negative coefficients likely indicate some degree of temperature-dependent structural rearrangement. Interestingly, at pH 7.2 and 50°C, unprotected amide hydrogens are expected to exchange with those of water sufficiently rapidly (≈300 s⁻¹) so that their NMR signals are predicted to be extremely weak or invisible.^31,32 The fact that many of the amide ¹⁵N/¹H groups lack hydrogen bonds in the X-ray crystal structure and yet produce strong signals in NMR spectra suggests that they are sterically protected from solvent hydrogen exchange by their surroundings in the molecule.^33,34

Figure 5.

Effect of temperature on NOX NMR spectra. Overlay of subregion of NOX ¹H–¹⁵N HSQC correlation spectra obtained over temperatures spanning 25°C (blue) to 50°C (red) (a) representative temperature-dependent changes in ¹H (b) and ¹⁵N (c) chemical shifts obtained for the backbone amide resonance of Val 7. Temperature coefficients for ¹H (d) and ¹⁵N (e) plotted as a function of residue number. ¹H temperature coefficients more negative than −4.5 ppb/K and ¹⁵N temperature coefficients greater in magnitude than 177 ppb/K are indicated by black bars. Secondary structure elements²¹ were taken as αA (10–16), αB (30–42), β1 (52–57), αC (60–69), αD (74–78), β2 (81–87), αE (89–98), αF (107–121), αG (125–149), β3 (153–157), αH (162–169), β4 (179–184), αJ (198–201).

In addition, we have used the temperature dependence of ¹⁵N chemical shifts reported for ubiquitin³⁷ as an indication of what is to be expected for a stably folded protein. For ubiquitin, all ¹⁵N temperature factors are smaller in magnitude than 17 ppb/K. Many ¹⁵N temperature factors in NOX exceed this value [black bars in Fig. 5(e)], which likely indicates that these residues experience temperature-dependent conformational changes. Interestingly, about half of these lie near the ends of elements of secondary structure or at the ends of loops, closely following or preceding secondary structure. Studies of β-turn and coiled-coil α-helical peptides have found that termini of the structured regions partially unfold or fray as the temperature is raised.^38,39 Thus, it is possible that the elevated ¹⁵N temperature coefficients for residues near the ends of secondary structure regions similarly correspond to some degree of fraying. Several peaks in the active site show unusually large ¹⁵N temperature coefficients. For example, G159, on the si face of the cofactor flavin ring and S193, which is located in the C-terminal region of the protein and interacts directly with the FMN phosphate have a large negative ¹⁵N temperature coefficients of –41.2 and –28.3 ppb/K, indicative of some degree of conformational adjustment in the active site as the temperature is raised.

Urea titrations

It was previously shown that adding low concentrations of denaturants increases NOX activity. This is believed to result from local conformational changes and/or increased dynamics in the active site.¹⁹ NMR titrations were carried out to explore the impact of urea on NOX at both 20 and 50°C. The extent of urea activation is much greater at 20°C than at 50°C. For example, the maximum catalytic rate increases approximately 2.5-fold with the addition of 1M urea at 20°C, while the increase at 50°C is roughly 5%.¹¹ However, the spectral properties of NOX are far more favourable at 50°C, as described below. At both 20 and 50°C, ¹H–¹⁵N HSQC spectra were collected for FMN-saturated NOX with increasing concentrations of urea up to 2M. In both cases, peaks throughout the spectra shift with increasing urea concentrations. However, there are distinct differences between the two datasets. At 50°C, the spectra remain sharp and well-resolved throughout the titration series. In contrast at 20°C, there was pronounced broadening and weakening of signals, with approximately 15% of peaks disappearing by 1.2M urea and the majority of signals gone or unresolvable by 2M urea (Supporting Information Fig. S6). This is unlikely to be due to increasing tumbling correlation time as the viscosity of 1.2 molar urea is only 5% greater than that of pure water.⁴⁰ It is also unlikely to be due to global unfolding of the protein, as the enzyme is equally active at 0 and 4M urea,¹¹ and global unfolding would be expected to be more prevalent at the higher temperature. Also unlikely is increased exchange of amide hydrogens with those of solvent, as the intrinsic rate of exchange is expected to be roughly 100-fold slower at 20°C compared to 50°C.³¹ Furthermore, it is not likely to be caused by protein aggregation, as denaturants typically reduce rather than increase protein self-association.⁴¹ The most likely explanation, therefore, is that the addition of low concentrations of urea induces pervasive ms–μs timescale dynamics throughout the enzyme. These motions are either absent at 50°C, or the timescale has been shifted to a faster regime in which spectral broadening is no longer evident.

Comparing spectra obtained with 0 and 2M urea at 50°C, Δδ values range from 0.028 ppm for A132 to 1.74 ppm for scW47, with a mean of 0.3 ppm and a standard deviation of 0.3 ppm. 75% of peaks shift by less than 0.30 ppm, so any Δδ value greater than 0.30 can be considered a relatively large shift. These peak movements may reflect either urea-induced conformational rearrangements of the enzyme or direct interactions between the protein and urea molecules. As urea is not aromatic, it will not produce long-range ring-current effects per se. Only nuclei directly interacting with urea molecules, for instance by hydrogen bonding, or those that experience changes in local protein conformation are expected to exhibit chemical shift perturbations. According to the X-ray crystal structure, 25 of the 33 ¹⁵N/¹H pairs with Δδ > 0.3 are physically occluded from interacting directly with urea (see “Methods”). Thus, chemical shift perturbations for these residues probably reflect protein conformational changes rather than direct binding of urea, assuming that the steric accessibility is similar in the solution and crystal forms of the enzyme. The large Δδ value for the side chain of W47 is notable as it has previously been implicated in urea-modulated gating of the active site. Many of the residues exhibiting the largest changes in chemical shift are located near the active site of the protein [Fig. 4(d)]. For instance, in helix F, which is also implicated in gating substrate access to the active site, E110, K113, A115, Q117, and R118 exhibit Δδ values of 0.49, 0.41, 0.36, 0.52, and 0.38 ppm, respectively. In addition, residues L158, G159, F160 which are located at the upper edge of the cofactor exhibit large Δδ values of 0.36, 1.60, and 0.37 ppm.

The peak shifts observed at 50°C are largely recapitulated at 20°C, although pervasive broadening is the most pronounced spectral change with urea addition at the lower temperature. Of the 37 peaks that shift furthest with the addition of urea at 50°C, 12 are absent from the spectrum at 20°C and 0 M urea. A further 4 disappear at 1.2 M urea. 17 are within the top 25% of Δδ values obtained at 20°C. Notably, only 4 out of 37 peaks with the largest shifts at 50°C are present at 20°C and do not show large urea-dependent changes.

Fast timescale NOX dynamics with 2M urea

We investigated the effects of urea on the ns–ps timescale motions of NOX using NMR ¹⁵N spin relaxation experiments. We recorded ¹⁵N R₁, R_1ρ, and steady-state {¹H}–¹⁵N NOE NMR measurements on an FMN-saturated NOX sample at 50°C in the presence of 1.2M urea. The relaxation data were analyzed according to the Lipari–Szabo model-free formalism⁴² which yields an order parameter, S², for each residue. The order parameters reflect the mobility of individual NH bond vectors and can take values between 0 and 1. NH bond vectors that are completely rigid in the molecular frame lead to S² values of 1, while bond vectors that undergo completely unrestricted dynamics on the ns–ps timescale in the molecular frame lead to S² values of 0. Values of S² obtained in the presence of 1.2M urea were compared with those previously obtained in absence of urea (Fig. 6).²⁶ Overall, the same trends in flexibility are observed in presence and absence of urea. The more flexible regions of the protein include residues V28, located in a loop connecting helices A and B, a cluster of residues, E188, G190, Y191, and S193, located in a long loop that connects the fourth β-strand and the short C-terminal helix J and forms part of the cofactor binding site, and W204, located at the extreme C-terminus of the protein. The root-mean-squared deviation between the two datasets is just 0.036. Nevertheless, some redistribution of flexibility is apparent when NOX is in presence of urea. Among others, NH bonds of residues A40, G141, G184 experience more mobility (ΔS² = –0.12 ± 0.05, –0.09 ± 0.07, –0.17 ± 0.08) while those of I100 and A121 are more rigid (ΔS² = 0.10 ± 0.04, 0.11 ± 0.05) with the addition of urea. Although the urea-induced changes in fast-timescale dynamics are limited to a small number of residues, it is it noteworthy that I100 and A121 lie in the active-site cleft of the enzyme, and their rigidification could be related to urea-induced changes in activity occurring there. The more distal residues are less likely to be directly involved, although in special cases, changes in fast-timescale dynamics can have long-range effects on enzyme catalysis.⁴³

Figure 6.

Effect of urea on fast timescale motions in NOX. Lipari-Szabo order parameters, S², reflecting motions on the ns–ps timescale, obtained for NOX in the presence (black) and absence (gray) of 1.2M urea, plotted as a function of residue number.

Discussion

NOX from the thermophilic bacterium Thermus thermophilus, exhibits several characteristics that are of interest from the point of view of understanding how enzyme function can be tailored through solution conditions and the selection of cofactors and cosolutes in biotechnological applications. NOX is capable of accepting either FMN or the much larger FAD molecule as cofactors.²¹ Similarly to many other thermozymes, it exhibits significantly greater activity at 65°C than it does at 20°C.¹¹ Its activity can be partially rescued at ambient temperatures through the addition of moderate concentrations of denaturants such as urea,^11,19 a useful characteristic it shares with several other enzymes.^7–10 We have used solution NMR spectroscopy and ITC to study the interactions of NOX with FMN and FAD, the denaturant urea, and to probe the effects of temperature on this enzyme.

We find that the spectra of apo- and both FMN- and FAD-saturated NOX are similar, suggesting that the protein maintains its overall fold even in the absence of cofactor. This places it in contrast to some other proteins,⁴⁴ such as some c-type cytochromes,⁴⁵ iron–sulfur containing proteins,⁴⁶ and acetyl transferases,⁴⁷ which require cofactor binding to adopt the holo-configuration. We found evidence for some enhanced protein mobility in the absence of cofactor, as signals for some residues in the NOX active site are absent in the apo-spectrum, possibly due to weakened hydrogen bonding that enhances solvent exchange or ms–μs timescale motions that broadened the signals beyond detection. Interestingly, the FMN-binding protein flavodoxin also retains its structure but exhibits enhanced dynamics in the absence of cofactor.⁴⁸ In addition, a comparison of the spectra obtained for FMN- and FAD-saturated NOX provides an indication of how the protein adjusts conformationally to the larger cofactor. X-ray crystallographic structures of the two forms indicated that the only structural changes accompanying cofactor substitution occur in the C-terminal region of the protein, comprising residues E189 to H194 (Supporting Information Fig. S9).²¹ To some extent the NMR data agree with this observation as large chemical shift changes are observed for G190, Y191, and S193, although these perturbations could also be due to aromatic ring current shifts from the adenine moiety of FAD. Interestingly, comparable chemical shift perturbations are also observed for other residues in and around the active site, including those facing both sides of the flavin ring. This suggests that there may be structural adjustments in NOX accompanying cofactor substitution in solution at 50°C that are not evident in the crystal lattice at 12°C.

Our ITC binding data for FMN and FAD are consistent with cofactors binding independently to the two subunits of NOX, which agrees with the observation that subunits undergo only minor changes in structure or dynamics when binding cofactor. Our ITC results indicate that FAD binds about 3 orders of magnitude less tightly than FMN. The heats of binding are essentially identical, with ΔH values of –10.1 ± 0.4 and –9.4 ± 0.7 kcal/mol, for FAD and FMN, respectively. The reduction in affinity is due to a large entropic penalty for FAD binding [ΔS = −16 ± 1 cal/(mol K)], while the binding entropy for FMN binding is close to zero [ΔS = 0.2 ± 3 cal/(mol K)]. This unfavorable binding entropy of FAD compared to FMN can be largely rationalized in terms of conformational restriction of the cofactor. The X-ray crystal of the FAD-bound enzyme shows that the additional adenine moiety is sandwiched in between the helix E/helix F loop and the C-terminal region of the protein (Supporting Information Fig. S9). Thus, binding of FAD is associated with the restriction of up to seven additional rotatable bonds not present in FMN. In fact, the difference in binding entropies for FMN and FAD determined by ITC closely approaches the entropic penalty of restricting 7 independently rotatable bonds from 3 rotamers to 1 (assuming an entropic cost of – R × ln{3} cal/(mol K) per bond where R is the ideal gas constant). This is clearly an oversimplification, as it ignores the enthalpies and entropies of additional FAD/protein contacts and protein conformational rearrangements. Nevertheless, the difference in binding entropies between FMN and FAD is of a magnitude than can be explained largely in terms of differences in the sizes and number of degrees of conformational freedom for the two cofactors. In other words, FAD binds less tightly than FMN because the additional adenine moiety is essentially trapped in the active site cleft. It suffers an entropic penalty upon binding while forming few, if any, favorable compensatory interactions with the protein.

NOX activity is enhanced approximately 2.5-fold in 1M urea and sevenfold at 65°C, compared to the denaturant-free enzyme at 20°C. The NMR data shed light on the mechanisms of urea and temperature activation of NOX. Of particular interest, the side-chain indole NMR signal of W47 shows the largest urea-induced chemical shift perturbations of all NOX ¹⁵N/¹H resonances at both 20 and 50°C. This residue has been previously implicated in the denaturant activation of the enzyme based on tryptophan fluorescence experiments.¹¹ Furthermore, molecular dynamics (MD) simulations have suggested that W47 controls substrate access to the active site through a “gating” mechanism.²¹ The residue alternates between an “open” conformation in which its indole side chain contacts helix F and a water molecule bridges its backbone amide and the cofactor, and a “closed” conformation in which its side chain packs more closely against the cofactor and the bridging water molecule is displaced (see Supporting Information Fig. S10 and Fig. 4 of Hritz et al.²³). The crystal structure corresponds to the “open” conformation, while the MD simulations indicated that the “closed” conformation is more stable in solution. Urea was found to shift the equilibrium towards the “open” state in MD simulations. Thus urea activation of the enzyme was proposed to be mediated through a shift in the “gating” equilibrium of W47 that facilitates substrate access to the active site. The large urea-dependent changes in the chemical shift we observe for the indole side chain of W47 provide strong, site-specific, experimental support for this mechanism.

The temperature dependence of NOX dynamics has also been addressed by MD simulations.^24,28 These studies suggested that urea and temperature activation of the enzyme proceed via different mechanisms, as the “open”/“closed” equilibrium of W47 is largely unaffected by temperature. Our NMR results agree with these findings as the W47 indole resonance does not exhibit large changes in chemical shift as the temperature is varied. The nitrogen temperature factor, dδ_N/dT = −9.84 ppb/K, is typical for a stably folded protein,³⁷ while the temperature factor of the proton, dδ_H/dT = −6.24 ppb/K, is characteristic of those lacking hydrogen bonds, as expected.³⁶ NMR chemical shifts are consistent with additional aspects of the molecular dynamics simulations. NOX contains two networks of salt bridges involving D88, R174, E90, R53, D91, E98, and R200, which are believed to help conformationally stabilize helix F. Only interactions of E90 and R128 were found to be disrupted with increasing temperature in the simulations.²⁴ Interestingly, E90 and R128 are the only two residues out of the seven participating in the salt bridge network to show large temperature dependent changes in chemical shift (E90 dδ_N/dT = 19.1 ppb/K; R128 dδ_N/dT = −22.0 ppb/K). The NMR data provide further insight into differences in the mechanism of temperature and urea activation of this enzyme. Addition of urea at 20°C is accompanied by pervasive spectral broadening characteristic of motions on the millisecond to microsecond timescales, while increasing the temperature does not produce this effect. Motions on these timescales can play significant roles in the catalytic cycles of enzymes^49–51 and are not captured in standard molecular dynamics simulations which typically extend for only tens of nanoseconds. Thus, the NMR results suggest that urea activation may involve extensive millisecond timescale motions in the enzyme, while temperature activation does not.

Finally, an interesting pattern emerges from a comparison of spectra obtained for NOX in the apo form and in the presence of FMN, FAD, urea, and at different temperatures. The majority of ¹⁵N/¹H NMR cross peaks are not sensitive to changes in the enzyme's environment, or shift in response just one of the variables tested. However, the signals for a small subset of residues (listed in Table2) are sensitive to multiple (two or more) co-solutes or variations in temperature. Backbone ¹⁵N/¹H chemical shifts are largely determined by aromatic ring-current shifts, amide proton hydrogen bonding and by the dihedral angles of the side chain and backbone of adjacent residues (1–2 residues apart in the primary sequence).^52–54 Thus a few specific sites in the protein exhibit a far greater tendency to structurally reconfigure in response to changing environmental conditions than does the rest of the protein. We shall refer to the tendency to undergo local restructuring in response to changing environmental conditions malleability. The first malleable site, containing A46, scW47, and N48 is located in the loop between helix αB and the first beta sheet, adjacent to the re face of the cofactor. All three residues are sensitive to replacing FMN with FAD. Since they are located near the conserved flavin moiety of the cofactor, this is likely due to reorganization of the active site rather than the formation of new contacts with the additional adenine moiety of FAD. Additionally, A46 experiences a conformational and/or dynamical transition upon cofactor removal, N48 exhibits a large ¹⁵N temperature factor, while the signals for scW47 and A46 are perturbed by the addition of urea. In the case of scW47, this is consistent with a urea-dependent conformational transition seen in molecular dynamics simulations.²³ In the case of A46, the amide ¹⁵N and ¹H are inaccessible to urea in the X-ray crystal structure, implicating urea-dependent conformational rearrangements at this location. A second malleable site, containing F71, G72, A74, G159, and F160, abuts the oxygen-rich side of the cofactor. Signals for all five residues are displaced upon FMN/FAD substitution. Since they are not adjacent to the additional adenine moiety of FAD, these changes in chemical shift likely reflect conformational rearrangements in the active site rather than the formation of new protein/cofactor contacts. The signals for three of these residues are also absent in the apo-protein, reflecting either ms–μs timescale motions or elevated hydrogen exchange in this state. The amide of G72 is involved in intramolecular hydrogen bonding in the cofactor-bound protein, thus it is clear that this residue experiences a change in conformation and/or flexibility when cofactor is removed. As well, 3 of these cross-peaks shift significantly with urea addition (2 of which correspond to urea-inaccessible residues), and 2 are temperature sensitive. Finally, S193 is sensitive to cofactor addition and temperature variation. All of these residues are within 15 Å of the cofactor which is not surprising since many of the environmental perturbations we tested are expected to target the active site of the enzyme. Nevertheless, there are over 100 nonproline backbone amide groups within 15 Å of the cofactor, therefore we are confident that the hypersensitivity of the 8 residues listed in Table2 is not simply a function of being located near the active site and instead reflects an unusual proclivity for remodeling of local protein structure and interacting with cosolutes. The peaks for these residues shift in different directions depending on the environmental condition being varied. This implies that structural plasticity surrounding the active site is more complicated than a dynamic equilibrium between two different conformations whose populations are shifted by changes in the environment, as all peaks for a given residue would fall on a line in this case.⁵⁵ Instead, each environmental variable produces a different conformational change, with the largest perturbations recurring at the hypersensitive residues identified here. Interestingly, the majority of malleable residues do not exhibit greater than average ns–ps timescale mobility when evaluated under individual sets of conditions (FMN-saturated, 50°C, 0, and 2M urea). Their conformational plasticity is revealed only when environmental conditions are varied. In this sense, malleability can be distinguished from flexibility in terms of the energy landscape view of protein folding and function.^56,57 When a residue is flexible, the corresponding minimum in the energy landscape is relatively broad in the corresponding dimensions, or multiple minima of similar energy are present so that a plurality of conformations is well populated. When a residue is malleable, the minimum in the energy landscape may be either broad or sharp; however, the width or position of the energy minimum shifts significantly in response to environmental variables. Thus flexibility is a property of the energy landscape under a single set of conditions, whereas malleability is a property of how the energy landscape changes as conditions are varied. It has been previously shown that dynamically flexible residues can play key roles in catalysis, and that mutations of these residues can profoundly affect enzyme efficiency.^58,43,59 Malleable residues may have a similar relationship to the activation of enzymes (by temperature, denaturant, cofactor substitution, etc.). Targeting these residues for mutagenesis combined with fine-tuning of the environmental conditions may be an avenue for optimizing the efficiency of enzymes in industrial applications.

Table 2.

Residues with NMR Signals Sensitive to Environmental Conditions

	Apo proteina	Cofactor substitutionb	Temperature variationc	Urea titrationd	Relative reagent accessible surface area (%)e
A46	X	X		X	0
scW47		X		X	67f
N48		X	X		22
F71		X		X	0
G72	X	X			0
A74	X	X	X		0
G159	X	X	X	X	79
F160		X		X	0
S193	X	X	X		0

^aBackbone ¹⁵N/¹H resonance absent in spectra of apo-NOX and present in holo-NOX.

^b > 0.1 ppm, comparing FMN- and FAD-saturated NOX spectra.

^c|∂δ_N/∂T| > 17 ppb/K or ∂δ_H/∂T < −4.5 ppb/K.

^dΔδ > 0.3, comparing NOX spectra obtained with 0 and 2M urea.

^eAmide NH accessibility based on a 2.5 Å probe.

^fIndole NH accessibility based on a 2.5 Å probe.

Materials and Methods

Construction of mutants

NOX W47F, W52F, W131F, and W204F mutants were constructed using two-step PCR directed mutagenesis (primers listed in Table1). Tryptophan residues at positions 47, 52, 131, and 204 were mutated to phenylalanine by replacing the codon TGG with TTC. In the first step, two separate PCR reactions were performed using the pTNADOX plasmid containing the Nox gene⁶⁰ as the template and either primers (Nox_fwd and WxF_rev) or (Nox_rev and WxF_fwd), giving rise to two amplicons. The amplicons were mixed and subjected to a second PCR with only primers Nox_fwd and Nox_rev. Primers Nox_fwd and Nox_rev contained EcoRI and HindIII sites, respectively. The amplicons were purified, digested with EcoRI and HindIII and ligated into the pTNADOX plasmid which had been treated with the same restriction enzymes. High fidelity Phusion polymerase (NEB, New England Biolabs, CA) was used for amplification and the mutants were verified by DNA sequencing. Only one step PCR was done for mutant 204, using primers Nox_fwd and W204F_rev, since the mutation is located close to the 3′ end of Nox gene.

Table 1.

Primers Used for Mutagenesis

Primer	Sequence
NOX_fwd	5′-ACTCCCgaattcATGGAGGCGACCCTT-3′
W47_fwd	5′-GCCCTCGGCCTTCAACCTCCAGC-3′
W52_fwd	5′-CCTCCAGCCCTTCCGGATCGTGG-3′
W131_fwd	5′-GCGAAAGGCCTTCGCCTCCGGG-3′
NOX_rev	5′-TCCTCaagcttCCCTAGCGCCAGAGGA-3′
W47_rev	5′-GCTGGAGGTTGAAGGCCGAGGGC-3′
W52_rev	5′-CCACGATCCGGAAGGGCTGGAGG-3′
W131_rev	5′-CCCGGAGGCGAAGGCCTTTCGC-3′
W204_rev	5′-TCCTCaagcttCCCTAGCGGAAGAGGA-3′

Mutated codons are underlined in bold. Lowercase letters indicate restriction endonuclease sites (fwd, EcoRI; rev, HindIII).

Protein expression and purification

The protein was overexpressed in Escherichia coli bacteria¹⁹ (NovaBlue, EMD Biosciences, San Diego CA), previously transformed with the pTNADOX plasmid encoding Thermus thermophilus NOX under the control of a tac promoter.⁶⁰ Cells were grown at 37°C in Luria–Bertani broth up to an OD600 of 0.7 a.u., pelleted by centrifugation (4500g, 10 min, RT), washed once in M9 salts, and resuspended in M9 minimal media containing ammonium chloride as the exclusive nitrogen source. Unlabelled ammonium chloride was used to produce samples for ITC experiments and ¹⁵NH₄Cl [200 mg/L (Sigma-Aldrich, St. Louis, MO, USA)] was used for NMR samples. Protein expression was induced with 200 mg/mL isopropyl β-d-thiogalactoside (IPTG, EMD Biosciences, San Diego, CA, USA) 1 h after the transfer. Bacteria were harvested by centrifugation (4500g, 10 min, RT) after 6 h of incubation.¹³ Cells, resuspended in lysis buffer (50 mM Tris–HCl, 5 mM benzamidine, 2 mM ethylenediaminetetraacetic acid (EDTA) at pH 7.5), were disrupted by sonication, centrifuged to remove cell debris (43,000g), incubated for 5 min at 80°C then separated by centrifugation (17,000g). The heat treatment causes mesophilic native E. coli proteins to precipitate while the thermophilic NADH oxidase remains in solution.²⁹ The sample was further purified using DEAE anion-exchange chromatography, in which impurities, but not NOX, bound to the column. The protein solution was further purified by adding 50 mM acetate buffer, reducing the pH to 5 loading on an SP–sepharose cation-exchange column and applying a 0–2M NaCl gradient, with elution occurring at 50 mM NaCl. Size-exclusion chromatography was performed using a HiLoad 16/60 Superdex 75 FPLC (GE Healthcare, Little Chalfont, UK) column as a final purification step. The sample was dialyzed into a 50 mM potassium phosphate, pH 7.2, consistent with sample conditions used previously for enzymatic activity assays.¹¹ The NMR samples contained 50 mM potassium phosphate pH 7.2, between 0.8 and 1 mM NOX (monomer concentration) with 2.4 mM riboflavin 5′-monophosphate sodium salt (FMN, Sigma-Aldrich, St. Louis MO, USA) unless otherwise specified, 5% D₂O and 10 μM 2,2-dimethyl-2-silapentane-5-sulfonate (DSS, Sigma-Aldrich, St. Louis, MO).

NMR spectroscopy

All NMR experiments were recorded at 50°C, except for the temperature titration and low-temperature urea titration, using Varian INOVA spectrometers equipped with high-sensitivity cold probes operating at ¹H Larmor frequencies of 800 MHz (18.8 T) and 500 MHz (11.7 T). This temperature was found to be optimal for sample stability and spectral quality. The temperature settings of both instruments were calibrated using a methanol reference.⁶¹ The data were processed and analyzed using the NMRPipe/NMRDraw suite of programs.⁶²

NMR titrations

¹⁵N/¹H HSQC correlation spectra⁶³ were collected at 18.8 T with 102 increments in the indirect dimension, 8 scans and 64 steady-state scans, for a total acquisition time of 37 min per spectrum. The urea titrations consisted of spectra collected for FMN-saturated NOX in the presence of 0–2M urea in 0.2M increments repeated at 20 and 50°C (Supporting Information Fig. S1). For the temperature titration, a series of spectra was collected from 25 to 50°C in 5°C intervals with 0M urea (Supporting Information Fig. S2). Amide proton and nitrogen temperature coefficients of each residue, indicative of hydrogen bonding and flexibility patterns, were derived from the slope (ppb/K) of chemical shift changes with increasing temperature.⁶⁴ The FMN and FAD titrations consisted of NOX spectra collected with 0.00, 0.05, 0.15, 0.20, 0.25, 0.30, 0.50, 1.00, 2.00, 2.40, 3.00, and 5.00 mM FMN and 0.0, 0.2, 0.4, 1.0, 2.0, 2.5, 3.0, 3.5, 14.9, and 39.9 mM FAD, respectively (Supporting Information Fig. S3). Titrant stock solutions were prepared in NMR protein buffer so that the final concentrations of all buffer components were identical to those of the protein NMR sample. Titrant was added by removing the protein sample from the NMR tube using a Pasteur pipette and placing it in an 1.5 mL disposable test tube, adding the required amount of the stock solution, mixing thoroughly, and replacing it in the NMR tube, using the same Pasteur pipette and test tube for each addition. The maximum changes in pH due to titrant addition (2M urea, 5.0 mM FMN, 39.9 mM FAD) were measured to be negligible (less than 0.1 pH units) All ¹H and ¹⁵N chemical shifts were referenced using the ¹H methyl signal of DSS for all temperature and cosolute titrations.⁶⁵

The shifts in ¹⁵N and ¹H resonance frequencies were analyzed as a function of FAD concentration according to:³⁰

2

where

[L]₀ and [P]₀ are the total concentrations of FAD and NOX (monomer), δ is the observed chemical shift, δ_P and δ_C are the chemical shifts of the free protein and the complex, respectively, and Δδ = δ_P − δ_C. The titration data for 20 residues were analyzed simultaneously by nonlinear least-squares optimization using inhouse MATLAB scripts to yield a global equilibrium dissociation constant, K_D, and an individual value of Δδ for each set of ¹⁵N and ¹H chemical shifts used. Uncertainties in the global K_D were estimated by a bootstrap method⁶⁶ where the fitting was repeated 100 times on Monte Carlo datasets comprising data from randomly-selected sets of residues. The error in the global K_D value was calculated as the standard deviation of the Monte Carlo K_D estimates. The FMN titration was not analyzed in this manner as the affinity of the interaction is too great to yield a reliable K_D value by this method.

¹⁵N spin relaxation experiments

¹⁵N R₁, R_1ρ,⁶⁷ relaxation rates and ¹H/¹⁵N steady-state NOE (ssNOE) values were recorded for FMN-saturated NOX in the presence of 1.2M urea. The R₁ experiments employed relaxation delays of 0.0111, 0.0999, 0.200, 0.322, 0.433, 0.566, 0.699, 0.855, 1.03, 1.22, 1.42, 1.67 s (11.7 T) and 0.0109, 0.0978, 0.196, 0.315, 0.424, 0.554, 0.685, 0.837, 1.01, 1.20, 1.39, 1.63 s (18.8 T). The R_1ρ experiments employed relaxation delays of 0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.07, 0.08, 0.09, 0.1 s and spin–lock field strengths of 1.5 kHz at both 11.7 and 18.8 T. NOE experiments were performed with 5 s of proton presaturation and a 7 s interscan delay, and with no ¹H saturation and 12 s interscan delay at 18.8 T. ssNOEs were calculated as the ratios of peak intensities in the saturated and unsaturated spectra while uncertainties were obtained from noise level estimates in NMRDraw. R₁ and R_1ρ relaxation rates were determined by nonlinear least-squares fitting of the peak intensities, I(t), to a monoexponential decay function as a function of the relaxation delay, t:

4

Error estimates were obtained from the deviations between experimental peak intensities and their back-calculated values. R₂ relaxation rates were calculated according to:

5

where ν₁ is the strength of the spin–lock pulse (1.5 kHz) and Δν is the offset of ¹⁵N resonance frequency from the carrier frequency. The R₁, R₂, and ssNOE values were analyzed according to the Lipari–Szabo model-free formalism⁴² as described previously.⁶ Our approach simultaneously optimizes the order parameters, S², and global tumbling parameters.⁶⁸ In this case, an axially symmetric model of rotational diffusion in solution with D_|| = 1.04 × 10⁷ s⁻¹ and D_⊥ = 1.2 × 10⁷ s⁻¹ gave significantly better agreement with the relaxation data than an isotropic tumbling model. A fully anisotropic tumbling model did not significantly improve the fit and the axially symmetric rotational diffusion model was used. All calculations were performed using inhouse MATLAB programs.

Isothermal titration calorimetry (ITC)

ITC experiments were performed on a MicroCal VP-ITC instrument (MicroCal, North Hampton, MA) at 50°C using buffer conditions identical to the NMR experiments. Three replicates of each titration were collected as well as three blank titrations of FAD into buffer. For FMN experiments, the syringe contained 122 µM FMN and the sample cell contained 8.7 µM NOX. Most titrations employed an initial delay of 60 s, a 2 μL initial injection followed by a 150 s delay and 27 × 10 μL injections each followed by a 300 s delay. Two replicates of the FMN experiments used different settings: an initial delay of 120 s, a 2 μL initial injection volume followed by a 150 s delay, 11 × 5 μL injections followed by 180 s delays and 44 × 10 μL injections followed by 240 s delays. For FAD experiments, the syringe contained 10 mM FAD and the sample cell contained 104 µM NOX or buffer alone. Titrations employed an initial delay of 60 s, a 1 μL initial injection volume followed by a 300 s delay and 10 × 5.6 μL injections followed by a 300 s delay. The FAD blank titration injection heats were subtracted from those of the NOX titration prior to analysis.

Relative reagent accessible surface area calculations

All surface area calculations were performed using UCSF Chimera and the 1NOX PDB atomic coordinates.⁶⁹ The relative reagent accessible surface areas (column 6 in Table2) were calculated as the sum of the nitrogen and hydrogen solvent accessible surface areas (SASA) for each residue, normalized by the sum of the nitrogen and hydrogen surface areas for the second glycine residue in an extended Gly-Gly-Gly tripeptide. For the side-chain NH group of W47, the normalization was taken relative to the sum of exposed nitrogen and hydrogen surface areas in an individual tryptophan residue. All SASA calculations were performed using a 2.5 Å probe. This is slightly smaller than the Van der Waals radius of a urea molecule relative to its geometric centre, r_VDW = 3.6 Å.⁷⁰

Glossary

2D

two-dimensional

Δδ

chemical shift displacement

DEAE

diethylaminoethyl

FAD

flavin adenine dinucleotide

FMN

flavin mononucleotide

FPLC

fast protein liquid chromatography

HSQC

heteronuclear single quantum correlation

IPTG

isopropyl β-d-thiogalactoside

ITC

isothermal titration calorimetry

K_D

dissociation constant

MD

molecular dynamic

¹⁵NH₄Cl

ammonium chloride

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

NOX

NADH oxidase

OD600

optical density 600

PCR

polymerase chain reaction

RMSD

root-mean-square deviation

RT

room temperature

S²

order parameter.

Supporting Information

Additional Supporting Information may be found in the online version of this article