Characterization of the enzymatic activity of the rhythmic human NADP(H) phosphatase, Nocturnin

Abstract

NAD(P)(H) is an essential cofactor for metabolic processes and a protector against oxidative stress. Phosphorylation of NAD(H) is completed by NAD kinase while dephosphorylation is performed by Nocturnin (NOCT). Because the phosphatase activity of NOCT is a rather new discovery, its enzymatic mechanism has not been reported. In this study, we use classical steady-state kinetics and pH-rate profiles of site-directed mutants in the NOCT active-site to deduce the essential residues for catalytic function for dephosphorylation of NADP(H). The pH-rate profile is bell shaped which supports that NOCT performs acid-base catalysis. Site-directed mutagenesis showed N149 is important for coordinating a deprotonated nearby residue, which is likely D324, and activates a nearby water to act as the nucleophile. After hydrolysis is performed, protonated H286 donates the proton to the leaving group. Finally, NAD(H) and inorganic phosphate is released. Since NOCT has been implicated in many important metabolic and stress response-related processes, understanding the mechanism is essential for the development of pharmacological modulators of NOCT activity.

Introduction

Nicotinamide adenine dinucleotide phosphate (NADP⁺) and its reduced form, NADPH, are essential co-substrates for enzymes which protect against reactive oxygen species (ROS) and are involved in numerous metabolic processes across all kingdoms of life.^{1, 2} NAD⁺ can be synthesized de novo from dietary amino acids such as tryptophan, and it can also be recovered in a salvage pathway.^{3, 4} NAD Kinase (NADK) phosphorylates NAD(H) to make NADP(H), a process that can be reversed by phosphatases.^{5, 6} One such phosphatase is the clock-controlled enzyme, Nocturnin (NOCT, UniProt Q9UK39).^{7, 8} NOCT was first discovered in Xenopus laevis retina⁹ and the gene was found to be rhythmically expressed and directly regulated by the core clock transcription/translation feedback loop.^{10, 11} Although its function was at first unknown, early studies noted that the secondary structure of Xenopus NOCT closely resembled that of CCR4, a yeast protein involved in mRNA processing whose deadenylase activity is responsible for cleaving the poly-A tail of mRNA.^{12, 13} However, despite the striking similarities in predicted structure, NOCT displayed modest deadenylase activity.^{14, 15} The determination of the crystal structure of human NOCT^{16, 17} further complicated its functional characterization, as the high structural homology yet low sequence homology with other endonucleases such as APE1¹⁸ suggested a broader substrate specificity within this family (Figure 1A). The active site shares high sequence homology with other nucleases, including key residues involved with the stabilization of the metal ion such as E195 (Figure 1B). These discoveries solidified NOCT as a member of the endonuclease/exonuclease/phosphatase (EEP) family which act on a large variety of substrates. Recently, it was determined that NADP(H) is NOCT’s physiological substrate with NOCT-dependent depletion of NADP(H) in cell culture demonstrating that NOCT is one of only two bona fide NADP(H) phosphatases in the cell.^{7, 8, 19} The other phosphatase, MESH1, has also been recently reported and is localized primarily to the cytosol and its activity has been thoroughly characterized.²⁰ NOCT has also been implicated in a number of stress phenotypes such as high fat diet-induced obesity²¹ and H₂O₂-induced cell viability⁸. In both phenotypes, loss of NOCT is protective and it is hypothesized this is due to an increase in the availability of NADPH.⁸

Figure 1.

A) Crystal structures of NOCT (6NF0) overlayed with CNOT6L (3NGO, pink) and APE1 (7TC3, blue). Sequence homology and RMSD were determined by the matchmaker function in UCSF Chimera. B) Over-layed active-site residues of CNOT6L (pink) and NOCT (tan). C) Steady-state kinetics of NOCT as a function of NADPH (black) and NADP⁺ (purple). D) Competition assay for NOCT activity using 100 μM NADPH as the substrate and varying [NAD(H)]. E) Double-reciprocal plot of the steady state kinetic parameters in the absence and presence of 1 mM NAD⁺. F) Double-reciprocal plot of the steady state kinetic parameters in the absence and presence of 1 mM NADH. All error bars are SD.

The enzymatic mechanism of phosphoryl transfer has been described extensively.^{22, 23, 24} These reactions can occur via three mechanisms: a concerted mechanism of direct transfer through a water activated metal ion, via a two-step acid-base mechanism which includes the formation of a transient phospho-enzyme intermediate²⁵, or a general acid-base mechanism where the nucleophile is indeed an activated water molecule which then forms a intermediate pentavalent phosphorane with the leaving phosphate group²⁶. Because NOCT has only been recently identified as a phosphatase, the enzymatic mechanism has not yet been described. Additionally, the structural and functional divergence in the EEP family suggests fine-tuned substrate specificity. Therefore, we sought to elucidate the enzymatic mechanism of human NOCT to uncover the catalytic residues and dissect its substrate specificity.

Results

NOCT prefers NADPH as a substrate and does not experience product inhibition

It has been described that NOCT is unable to utilize nucleotides or other phospho-containing metabolites other than the nicotinamide-containing substrates.^{7, 8} Therefore, to assess whether NOCT exhibits any abnormal kinetics or substrate-mediated regulation, we first examined its enzymatic behavior focusing on NADP(H). NOCT utilizes NADP⁺ and NADPH with equal k_cat values and displays classical Michaelis-Menten kinetics with a k_cat of 53 and 52 s⁻¹, respectively (Figure 1C). NOCT prefers NADPH slightly with a k_cat/K_m of 265 compared to 78 for NADP⁺. This suggests that small modifications of the nicotinamide-containing metabolites influence the substrate affinity.

Because the reverse reaction of phosphorylation of NAD(H) performed by NADK is subject to product inhibition⁵, we investigated whether NOCT is similarly regulated by its products NAD⁺ and NADH. We performed a competition assay where the concentration of substrate (NADPH) was kept constant while varying [NAD⁺] or [NADH] (Figure 1D). The results showed that NOCT activity was not affected by either competitor at one concentration of NADPH however to determine whether the products affect kinetic parameters for NOCT, we used Lineweaver-Burke graphical representation in the absence and presence of 1 mM NAD or NADH (Figures E and F). The slopes and intercepts are unchanged for either metabolite suggesting NAD and NADH are not inhibitors of NOCT activity.

NOCT activity is optimal with Mg²⁺ but Mn²⁺ and Fe²⁺ can substitute

NADK activity has been shown to be optimal with Mn²⁺ and Zn²⁺, while Mg²⁺ is also efficient albeit not preferred.²⁷ Minimal activity was observed using Co²⁺ and Ca²⁺, while no activity was detected with Cu²⁺ or Ni²⁺. To assess whether NOCT activity was similarly regulated by metal ion identity, we measured its phosphatase activity with various divalent and one trivalent metal ion. Mg²⁺ produced optimal activity, but the enzyme also displayed considerable activity with Mn²⁺ and Fe²⁺ (Figure 2A). NOCT showed no detectable activity in the presence of Fe³⁺, suggesting that its activity can be regulated by the oxidation state of the metal ion. We then examined whether differences in atomic radius could explain the variation in activity but found no correlation (Figure 2B). Molecular dynamics simulations of NOCT with bound NADPH and either Mg²⁺, Ca²⁺ or Zn²⁺ present in the active site were carried out, followed by MM-PBSA calculations to estimate the binding energy of NADPH to NOCT with either of the three ions. Mn²⁺, Fe²⁺, Cu²⁺, Ni²⁺ and Co²⁺ are not available in the forcefield used for the simulations. The results revealed NADPH binding was reduced in the presence of Zn²⁺ or Ca²⁺, relative to Mg2+ indicating that lack of activity with these metals likely results from decreased binding compared to Mg2+ (Figure 2C). The most recent structure of NOCT bound by NADPH¹⁶ contains a Ca²⁺ in the active site, which was used to prevent turnover and stabilize the substrate bound state. While this approach was successful, crystallization procedures utilize over-saturated concentrations of components which is likely to lead to substrate binding despite low affinity.

Figure 2. NOCT activity is optimal with Mg²⁺ but Mn²⁺ and Fe²⁺ can substitute.

A) Phosphatase activity of NOCT in the presence of varying divalent cations (1 mM final concentration). B) Atomic radii as a function of phosphatase activity. C) MM-PBSA estimation of the binding affinity of NADPH to NOCT in the presence of Mg²⁺, Zn²⁺, or Ca²⁺. D) Competition assay of NOCT activity with NADPH as the substrate as a function of increasing [TRIS-HCl]. Error bars denoted as SD.

Another noteworthy observation is the effect of Tris-based buffers on NOCT activity. Tris acts as a inhibitor for NOCT as demonstrated by a competition assay measuring NADPH hydrolysis as a function of increasing amounts of Tris-HCl (Figure 2D). Tris-based buffers have been reported to act as an inhibitor for many enzymes, specifically metalloenzymes, as the amine group has the propensity to bind metals when deprotonated.^{28, 29, 30} Instead, switching the buffer to HEPES results in a k_cat of 53 s⁻¹ (Figure 1C) in activity compared to currently published data reporting a k_cat of 15 s⁻¹ when using a Tris-based reaction buffer^{8, 31}.

The pH-activity profile of NOCT indicates general acid-base catalysis

Protein phosphatase and phosphoryl transfer enzymatic mechanisms have been studied extensively.^{24, 32} To understand the mechanism for catalysis and the protonation states of NOCT active-site residues, we performed pH-dependent steady-state kinetic analysis on NOCT using NADPH as the substrate. The concentration of metal in the form of magnesium was saturating at 1 mM. K_cat and k_cat/K_m as a function of pH display a bell-shaped curve indicative of general acid-base catalysis with pK₁= 5.9 or 5.7 and pK₂= 8.5 or 9.0, respectively, when fit with equation 1 (Figure 3A and 3B, Table 1). In equations 1–3, C is the maximal activity, K₁ is the ionization constant (pK) of unprotonated group and K₂ is the pK for the protonated group and H is the hydrogen ion concentration. The crystal structure of the catalytic domain shows the active site of NOCT contains the divalent cation (in this case Ca²⁺) and residues N149, E195, H286, D324, and H414 (Figure 3C). E195 is well studied and the mutant enzyme, E195A NOCT, has been utilized as a catalytically “dead” mutant as a control since the crystal structure of NOCT (PDB 6NF0) shows E195 stabilizes the divalent cation and the mutant protein shows no phosphatase activity. To fully understand their role in the catalysis, we performed site-directed mutagenesis on these active site residues.

Figure 3. The pH-activity profile of NOCT indicates general acid-base catalysis.

A) WT NOCT k_cat as a function of pH. The fit was performed using a user-defined equation in GraphPad Prism. B) WT NOCT k_cat/K_m as a function of pH. C) Active-site of NOCT (PDB 6NF0) highlighting the residues which contact the leaving phosphate. The green sphere denotes the Ca²⁺ while red spheres indicated waters. The dashed contacts are hydrogen bonding included in the PDB while the blue lines indicate hydrogen bonding calculated using UCSF chimera. pH profiles of D) H286N fit with equation 2 and E) N149A NOCT was fit with equation 3. The fit was performed using user-defined functions on GraphPad Prism. F) Fluorescence spectra of WT NOCT with and without 2 mM NADPH when excited at 340 nm. G-L) Fluorescence chnage as a function of [NADPH]. The fit is the Hill equation performed on GraphPad Prism and the errors denote SEM.

Table 1.

Kinetic parameters for NOCT and NOCT mutants.

Enzyme	kcat/Km (mM−1 s−1)	pK1	pK2
WT NOCT	281	6.0 ± 0.1	8.5 ± 0.2
H286N	0.0079	6.5 ± 0.1	--
N149A	0.02	--	9.9 ± 0.07
D324S	n.d.	--	--
H141N	n.d.	--	--

$$\mathit{log}Y=\mathit{log}[C/(1+H/K_1+K_2/H)]$$ (1)

$$\mathit{log}Y=\mathit{log}[C/(1+H/K_1)]$$ (2)

$$\mathit{log}Y=\mathit{log}[C/(1+K2/H)]$$ (3)

H286 is the general acid and N149 stabilizes an adjacent ionizable residue

Beginning with the two histidines located in the active site pocket, H414 and H286, we mutated these residues to asparagine to preserve side-chain length while altering catalytic function, and we assessed their steady-state phosphatase activity. H414N NOCT showed no detectable activity, even when enzyme concentration and reaction time were increased. Since H414 is within hydrogen bonding distance (2.6 Å) of the leaving phosphate group, this suggests that it plays a crucial role in substrate stabilization rather than direct catalysis. H286N exhibited drastically reduced activity with an apparent k_cat decrease of 5 ×10³-fold compared to wild-type (WT) NOCT. Due to the severe loss of activity, k_cat could not be determined, as the reaction failed to reach saturation even at 2 mM NADPH. Therefore, k_cat/K_m was determined by calculating the slope of V₀ as a function of substrate concentration. The pH profile of H286N revealed a break in the acidic limb, with a calculated pKa of 6.5 when fit with equation 2 (Figure 3D, Table 1). This loss of the basic limb suggests the protonated form of H286 is required for catalysis. This strongly suggests that the protonated H286 acts as the general acid, donating the proton required for stabilization of the leaving phosphate group.

Since a water molecule and the residues surrounding the substrate must be precisely coordinated for efficient catalysis, we investigated N149, a polar residue adjacent to H286 and D324 forming a triad around the leaving phosphate group. N149 is within hydrogen bonding distance to a nearby water (2.5Å), suggesting it also plays a structural role in the active site. Mutagenesis of N149 to alanine (N149A) resulted in a drastic decrease in activity by 1.6 × 10³-fold, similar to H286N, making V_max unattainable. Thus, k_cat/K_m was calculated by the slope of the initial rate as a function of the substrate concentration. The pH profile of N149A showed a break in the basic limb with a pKa of 9.9 when fit with equation 3 (Figure 3E, Table 1). Asparagine is not considered an ionizable amino acid as the amide side chain does not readily accept or donate protons. This suggests N149 is paired with an ionizable residue which is deprotonated likely coordinating the active site phosphate group and stabilizing charge distribution, and/or acting as a general base. N149 is also in close proximity to E195 (3.3 Å) and D324 (2.8 Å), forming a potential hydrogen bonding network. Previous work has established E195 is coordinating the active site divalent cation.^{8, 16} We hypothesized that its hydrogen bonding to D324 is stabilizing the residue for it to act as the general base, activating a nearby water to become the nucleophilic hydroxide. To further probe this interaction, D324 was mutated to serine (D324S). D324S NOCT showed no measurable activity even at increased enzyme concentrations, substrate levels, or reaction times (data not shown).

To the assess the differences in binding affinity of NADPH to these mutant NOCT proteins, we utilized the intrinsic fluorescence of NADPH and quantified the change in fluorescence when incubated with WT NOCT, H286N, D324S, N149A and H414N NOCT. Upon the addition of NOCT, we see an increase in fluorescence when excited at 340 nm with a maximal difference at 450 nm (Figure 3F). We quantified the change in fluorescence (ΔF) at an emission wavelength of 450 nm as a function of the NADPH concentration for each NOCT construct (Figures 3G-K). We find that WT NOCT has a K_d for NADPH of 308 μM while H286N has a K_d of 219 μM (Figure 3L). The two dissociation constants are within error of each other are therefore indistinguishable. N149A and H414N have a K_d of 473 and 338, respectively. The error in the fits are due to the sensitivity of the fluorescence change and these two mutant NOCTs result in smaller ΔF than the other constructs. However, the ΔF indicates quantifiable NADPH binding for WT NOCT and these mutants suggesting these mutations do not inhibit NADPH binding in the active site. D324S NOCT did not have any quantifiable fluorescence change and therefore is unable to bind NADPH suggesting that it is essential in the active site coordination of the substrate. Another notable finding is that NADPH binding to NOCT is cooperative due to the Hill Coefficients of 1.4–1.7 in WT, H286N, and N149A NOCT. This cooperativity is lost in the H414N mutation.

Together, these data support a mechanism where H286 acts as the catalytic general acid and its protonation is essential for proper NOCT activity. N149 itself is not an ionizable residue but likely interacts with a deprotonated ionizable nearby residue which stabilizes the catalytic site via a hydrogen bonding network. Because N149 is hydrogen bonded to D324 according to the crystal structure of the active site (Figure 3C) we propose that D324 is in fact the deprotonated residue that is stabilized by N149.

Discussion

In this study, we determined that human NOCT performs its catalytic function using general acid-base chemistry, defined the essential residues, and further characterized NOCT’s substrate and metal specificity. Our kinetic analyses showed that NOCT utilizes nicotinamide-containing substrates readily, albeit with varying affinity, but prefers NADPH. Further studies on how NOCT’s preference in substrate affects the individual pool sizes of NADPH and NADP in cell culture would be valuable in understanding redox dynamics. MESH1, the other eukaryotic NADPH phosphatase, has been reported to also display a preference for NADPH compared to NADP.²⁰ We were unable to detect significant product inhibition by NAD⁺ or NADH at 1 mM concentration which is above the reported physiological levels in the cell.³ This is contrast to NADK which performs the reverse reaction to NOCT. We found that the Kd of NADPH for WT NOCT is 300 μM which is significantly higher than NAD’s affinity for NADK which is reported to be 10 nM.³³ However, NADK is also bound by its product with high affinity with a K_d of 16 μM.³³

The specificity of the divalent cation is likely due to the size of the catalytic pocket, residue interactions, and its ability to activate an adjacent water molecule. Mg²⁺ has a high charge density and is effective at stabilizing negatively charged groups like phosphate. Additionally, Mg²⁺ forms stable but flexible octahedral complexes which allows dynamic and reversible binding.³⁴ Mn²⁺ and Fe²⁺ can also stabilize the phosphate but are slightly larger than Mg²⁺. Ca²⁺ is too large and has a lower charge density. Zn²⁺ is a “soft” divalent cation that prefers nitrogen and sulfuric ligands.³⁵ Ni²⁺ is not found in phosphate-acting eznymes as it does not meet the geometric needs for phosphate binding but is more coordinated for redox chemistry.³⁶ Many EEP family enzmyes use Mg²⁺, such as Ape1, DNase I-like endonucleases and exonuclease III (ExoIII).³⁷ In vitro experiments have shown Mn²⁺ can substitue for Mg²⁺ in many of these enzymes but usually alters activity.³⁸ It is likely that Mg²⁺ preference stems from its abundant bioavailability in both the cytosol and mitochondria.³⁹ Mn²⁺ is present at low levels while free Zn²⁺ concentrations are very low and buffered by metallothioneins.⁴⁰ Additionally, we showed Tris-based buffer acts as a inhibitor of NOCT activity similar to reported inhibition of a variety of enzymes.^{28, 29} Tris-HCl has been shown to affect metal-containing enzymes particularly, leading to speculation that it acts as a metal chelator. Therefore, we extend caution to other metal-containing enzyme assays to avoid using Tris as the buffering agent in these cases.

The pH-profile of the steady-state kinetics of NOCT displayed a bell-shaped curve, indicating acid-base catalysis. We found N149’s coordination is essential for optimal activity and since it is not an ionizable residue, it coordinates an adjacent deprotonated residue via hydrogen bonding (possibly D324 based on the reported crystal structure (PDB: 6NF0). The deprotonated residue would abstract a proton from an adjacent water generating a hydroxide ion that acts as the nucleophile. We propose that the activated OH- attacks the phosphorus center of the NADP(H) phosphate group. This could then proceed in a concerted mechanism or form a pentacovalent intermediate where protonated H286 donates a proton to the leaving group which releases the inorganic phosphate and regenerates NOCT (Figure 4). Future studies would be required to determine if a covalent intermediate is formed. We also report the affinity of NADPH to WT NOCT and these active site mutants and found that H286N does not affect the affinity but N149A and H414N decrease the affinity of NADPH to NOCT. Additionally, D324S has no detectable substrate binding and is essential for coordination of NADPH. The binding of NADPH showed positive cooperativity to NOCT which is curious considering we did not observe cooperativity in the rate of cataylsis. Since NOCT is a monomeric protein with one reported substrate binding site³¹, this suggests possible allosteric coopertivity in the form of structural changes in the enzyme but further structural analyses would need to be conducted to confirm this hypothesis.

Figure 4. Proposed catalytic mechanism for NOCT activity.

1) After NADP(H) binds the active site, the 2’-phosphate group coordinates with Mg2+ and additional electrostatic and hydrogen bonding with nearby residues stabilizes the substrate. 2) The deprotonated D324 abstracts a proton from the bound water molecule to generate the hydroxide ion which is the nucleophile. 3) Nucleophilic attack on the phosphorous center. The protonated H286 donates a proton to the leaving group. 4)This is finalized with product release and the regeneration of NOCT. Created with Moldraw.com

In conclusion, our study sheds light on the unique enzymatic properties of NOCT and its potential mechanisms of action. NOCT utilizes a general acid-base mechanism for catalysis, with H286 playing a central role as the general acid. The enzyme’s substrate specificity is limited to nicotinamide containing metabolites yet the broad specificity of the EEP family does not rule out the possibility of other phosphate-containing substrates. NAD(P)(H) pools have been shown to be affected by NOCT’s temporal expression and cellular compartmentalization ^{7, 8}. Our data adds mechanistic information to the already fine-tuned regulation of NOCT’s enzymatic activity. The findings presented here contribute to a deeper understanding of NOCT’s biochemical properties and its potential physiological roles in regulating NADP(H) levels, as well as its broader implications in circadian biology and cellular metabolism. The chemical mechanism presented here is essential for guiding the production of therapeutics for NOCT-related morbidities.

Methods

Expression and purification of NOCT

Recombinant human NOCT was expressed and purified using a modified protocol similar to that described previously.³¹ Briefly, full length NOCT is cloned in frame with an N-terminal His-tagged NusA protein separated by a TEV cleavage site in the pET22B vector. Two liters of LB supplemented with ampicillin were inoculated with 1.5 mL of an overnight starter culture and incubated at 37°C while shaking at 250 rpm until the OD at A600 reached 0.6–0.8, at which point the flasks were removed from the incubator and cooled down on ice. The flasks were then returned to a 20°C incubator and induced with 100 μg/mL IPTG followed by an additional 18 hrs of growth, after which the cultures were harvested at 5422×g in a JLA 8.1 rotor for 15 minutes at 4°C. The cell pellets were resuspended\in 40 mL lysis buffer per liter of culture containing 50 mM HEPES pH 7.5, 150 mM KCl, 3 mM MgCl₂, 10% glycerol, 1 mM DTT, 1% tween-20 and 1x Roche complete protease inhibitor (Sigma #11697498001). The cells were homogenized using a 15 mL glass homogenizer until smooth then the lysate was sonicated for 4x 30s with 30s breaks. The sonicated lysate was then ultracentrifuged for 50 minutes at 236,510×g at 4°C. The supernatant was then applied to an equilibrated 1 mL HisTrap column (Cytiva #17524701) with 50 mM HEPES pH 7.5, 150 mM KCl, 3 mM MgCl₂, 10% glycerol, 1 mM DTT and 20 mM imidazole (Buffer A). After loading the sample, the column was washed with Buffer A until the A₂₈₀ was below 0.1. The protein was eluted with Buffer A containing 300 mM imidazole. The peak A₂₈₀ fractions were pooled and desalted using Cytiva PD 10 columns (Sigma #GE17–0851-01) after which the A₂₈₀ was measured, and TEV protease was added to the pool at a ratio of 1 mg of TEV to 50 mg of NOCT and incubated at 4°C overnight. The digested sample was reapplied to a HisTrap column using the same buffers as described above. The high A₂₈₀ fractions during the wash step were collected for NOCT as the NusA chaperone contains the N-terminal His-tag and remained bound. These wash fractions were concentrated to 500 μL and applied to a Superdex S200 10/300 GL column that was equilibrated with 50 mM HEPES pH 7.5, 150 mM KCl, 3 mM MgCl₂, 10% glycerol and 1 mM DTT (SEC buffer). The resulting fractions containing the predominant peak were collected and pooled. The protein concentration was determined by its extinction coefficient at A₂₈₀ (ε₂₈₀=41.7 mM⁻¹ cm⁻¹) and purity was confirmed by SDS-PAGE.

Site-directed mutagenesis

The active site mutations were introduced using the pET22B vector containing the NusA-NOCT gene using the QuikChange II (Agilent #200521) site-directed mutagenesis kit with oligonucleotides from Sigma. BL21(DE3) competent cells transformed with the resulting constructs. The coding regions of all expression plasmids were Sanger sequenced at Eurofins Genomics to verify the mutation and to ensure no other mutations were present.

Steady-state kinetic activity

NOCT activity was determined using the Malachite Green Phosphate Assay kit (Sigma #MAK307). The reaction buffer contained 100 mM HEPES pH 7.5, 10 mM KCl, 1 mM MgCl₂, 1 mM DTT and 0.005 % triton-x. For pH dependent experiments, HEPES was substituted with the following buffers: MES (pH 6.0, 6.5), HEPES (pH 7.0, 7.5, 8.0), CHES (pH 9.0, 10). NOCT (0.2 μM) was mixed with 0–4 mM NADPH for a final concentration of 0.1 μM and 0–2 mM, respectively. For steady state experiments, the reaction time was 0, 30, 60 seconds after which the reaction was quenched with 4 μL of 0.5 M EDTA. After a 30-minute incubation of the assays with the working reagent, the reactions were read using a plate reader at 620 nm. The phosphate concentrations were quantified using a standard curve for 0–40 μM phosphate. The data were fit with the Michaelis-Menten equation or linear regression to determine the steady-state kinetic parameters using GraphPad Prism. All activity determinations were completed in triplicate.

Lineweaver-Burke Analysis

The steady-state kinetic parameters in the presence of either 1 mM NAD or NADH were conducted as described above with the modification of the addition of the products directly to the enzyme prior to the initiation of the reaction. NAD or NADH was allowed to incubate with NOCT at room temperature for 10 minutes before the reaction began. The data was processed to produce the double reciprocal plots using GraphPad Prism and fit with a linear regression.

Fluorescence spectrometry

The intrinsic fluorescence values of 0–2 mM NADPH were obtained using an ex 340 nm/em 450 nm prior to the addition of NOCT. Afterwards, NOCT (or NOCT mutant) was added to NADPH to produce a final concentration of 2.5 μM of protein and the fluorescence was obtained. Each assay consisted of 1 μL of enzyme in a final volume of 50 μL. The readings were obtained in a 96 black-walled plate using an Agilent BioTek plate reader. ΔF (F-Fo) was plotted as a function of the NADPH concentration and fitted with the Hill equation using GraphPad Prism.

In silico protein preparation

The crystal structure of NOCT (PDB ID: 6NF0) was prepared for simulations using the charmm-gui input-generator ^{41, 42}, keeping the protein and CA ion only. The protonation state at pH 7 was predicted using PDB2PQR ^{43, 44}, and the missing amino acids (133–140) were modeled using the charm-gui implemented galaxyFill ⁴⁵. The resulting protein structure was used to generate a topology and parameter file with gromacs pdb2gmx, using the amber 99SB-ILDN forcefield ⁴⁶. Separately, the NADPH ligand was parameterized using the GAFF2 forcefield ⁴⁷ in the Amber software suite ⁴⁸. The protein:ligand complex was solvated in a cubic water box with a distance of 1 nm to the protein and the charges were neutralized with Na Cl ions. Single point amino acid mutations were done using the Pymol software ⁴⁹.

Molecular dynamics

Molecular dynamics simulations were carried out using the GROMACS software package ^{50, 51}. Simulations were done in 4 stages: I) energy minimization for 5000 steps using the steepest descent approach, II) equilibration of the water and ions in the NVT ensemble for 5000 steps III) equilibration in the NPT ensemble for 50000 steps IIII) production run in the NPT ensemble for 50 ns. For each simulation, the temperature was controlled at 300 K using the modified Berendsen thermostat ⁵², while the pressure was held at 1 bar using the Parrinello-Rahman barostat ⁵³. A timestep of 2 fs was employed and all hydrogen bonds were restrained using the LINCS ⁵⁴ algorithm. Long range electrostatic interactions were calculated using PME ⁵⁵ and a short-range cut-off of 1 nm was employed with the Verlet cutoff scheme. To compare the effects of the active site ions on NADPH binding to the protein, simulations with Ca, Zn and Mg were carried out. Restraints were placed on the NADPH ligand, to more accurately compare the binding energies between the three different ions.

The structure obtained from charmm-gui was used to generate the mutations D324S and H286N, N149A and H414N, with the exception of the Ca²⁺ being replaced with Mg²⁺ to remove steric hinderances caused by Ca²⁺. The mutant structures were simulated similarly to the WT structure

Simulations used for the MM-PBSA estimates were run using a similar approach; the equilibrated structure from the 50 ns production run (WT and H286N) was used as the input and 15 independent simulations were carried out, each with a production length of 1 ns.