A Dynamic Core in Human NQO1 Controls the Functional and Stability Effects of Ligand Binding and Their Communication across the Enzyme Dimer

biomolecules Article A Dynamic Core in Human NQO1 Controls the Functional and Stability Effects of Ligand Binding and Their Communication across the Enzyme Dimer Pavla Vankova 1,2, Eduardo Salido 3, David J. Timson 4, Petr Man 1,* and Angel L. Pey 5,* 1Institute of Microbiology, Academy of Sciences of the Czech Republic, Videnska 1083, 142 20 Prague 4, Czech Republic; [email protected] 2Department of Biochemistry, Faculty of Science, Charles University, Hlavova 2030/8, 128 43 Prague 2, Czech Republic 3Center for Rare Diseases (CIBERER), Hospital Universitario de Canarias, Universidad de La Laguna, 38320 Tenerife, Spain; [email protected] 4School of Pharmacy and Biomolecular Sciences, University of Brighton, Huxley Building, Lewes Road, Brighton BN2 4GJ, UK; D.T[email protected] 5Department of Physical Chemistry and Unit of Excellence in Chemistry, University of Granada, Av. Fuentenueva s/n, E-18071 Granada, Spain *Correspondence: [email protected] (P.M.); [email protected] (A.L.P.) Received: 23 October 2019; Accepted: 10 November 2019; Published: 12 November 2019   Abstract: Human NAD(P)H:quinone oxidoreductase 1 (NQO1) is a multi-functional protein whose alteration is associated with cancer, Parkinson’s and Alzheimer ´ s diseases. NQO1 displays a remarkable functional chemistry, capable of binding different functional ligands that modulate its activity, stability and interaction with proteins and nucleic acids. Our understanding of this functional chemistry is limited by the difficulty of obtaining structural and dynamic information on many of these states. Herein, we have used hydrogen/deuterium exchange monitored by mass spectrometry (HDXMS) to investigate the structural dynamics of NQO1 in three ligation states: without ligands (NQO1 apo ), with FAD (NQO1 holo ) and with FAD and the inhibitor dicoumarol (NQO1 dic ). We show that NQO1 apo has a minimally stable folded core holding the protein dimer, with FAD and dicoumarol binding sites populating binding non-competent conformations. Binding of FAD significantly decreases protein dynamics and stabilizes the FAD and dicoumarol binding sites as well as the monomer:monomer interface. Dicoumarol binding further stabilizes all three functional sites, a result not previously anticipated by available crystallographic models. Our work provides an experimental perspective into the communication of stability effects through the NQO1 dimer, which is valuable for understanding at the molecular level the effects of disease-associated variants, post-translational modifications and ligand binding cooperativity in NQO1. Keywords: protein structural dynamics; NQO1; ligand binding; protein stability; allostery; protein degradation 1. Introduction Human NAD(P)H:quinone oxidoreductase 1 (NQO1; EC 1.6.5.2) is a multifunctional stress protein mostly localized in the cellular cytosol [ 1 ]. NQO1 expression is upregulated as a response to different types of cellular stress and through several mechanisms, including the antioxidant response through Nrf2-mediated and Ah2 signaling pathways [1–4]. NQO1 displays multiple enzymatic and non-enzymatic functions [ 1 – 5 ]. NQO1 catalyzes different reactions with cytoprotective and metabolic roles, such as the two-electron reduction of quinones Biomolecules 2019,9, 728; doi:10.3390/biom9110728 www.mdpi.com/journal/biomolecules Biomolecules 2019,9, 728 2 of 17 to form hydroquinones [ 2 , 6 ] (Figure 1), redox cycling of quinones [ 2 ], reduction of coenzyme Q 10 and vitamin E to their antioxidant form [ 2 ], scavenging reactive oxygen species [ 1 – 3 , 7 ], reduction of catecholamines and vitamin K [ 1 , 2 ] and maintenance of the NADH/NAD + redox balance [ 3 , 8 ]. The main features of these biochemical reactions involving NQO1 have been investigated in detail mainly through enzymological and structural analyses. Structurally, the enzyme forms obligate functional homodimers, with two active sites located in the monomer:monomer interface (MMI), and each monomer consists of two different domains: (i) an N-terminal domain spanning residues 1–224 that contains part of the active site and is involved in the tight binding of one FAD molecule per monomer and protein dimerization; (ii) a C-terminal domain (residues 225–274) that contributes to stabilizing the protein dimer and to the binding of the NAD(P)H coenzyme and the substrates [ 1 , 4 , 9 – 14 ]. The functional cycle of NQO1 generally involves two steps according to a ping-pong mechanism: first, in the reductive half-reaction, an NAD(P)H molecule binds to the enzyme and rapidly reduces the FAD to FADH 2 (with a second-order rate constant of ~10 6 M −1· s −1 ) with the subsequent release of the oxidized nicotinamide dinucleotide; and second, in a very fast oxidative half-reaction (with a second-order rate constant >10 9 M −1· s −1 ), the substrate binds and is reduced by the FADH 2 , thus regenerating the flavin in oxidized form and releasing the reduced product [ 1 , 10 ]. This catalytic cycle is known to be inhibited by different coumarin-based molecules (the best characterized is the biscoumarin, dicoumarol) that act as competitive inhibitors by blocking the NAD(P)H access to the active site and partially occupying the NAD(P)H binding site [ 1 , 15 ]. Importantly, comparison of the crystal structures of NQO1 with FAD bound (NQO1 holo ) with that containing also dicoumarol bound (NQO1 dic ) have revealed that inhibitor binding causes only minor structural rearrangements in the conformation, which localize at the surface of the catalytic site [ 15 ]. Among non-enzymatic functions, we must highlight the ability of NQO1 to develop protein: protein and protein: RNA interactions [ 1 , 2 , 16 – 20 ]. In particular, protein:protein interactions involving NQO1 are relevant to understand its multiple roles in physiological and pathological processes. NQO1 interacts with key transcription factors associated with cancer (e.g., p53, p73 α and HIF-1 α ) [ 17 , 18 ] and proteins involved in HIV infection (e.g., Tat protein) [ 19 ], and these interactions increase the intracellular stability of these protein partners by preventing their degradation by the proteasome. These protein:protein interactions presumably depend on the functional state of NQO1: binding of NADH may increase the strength of these interactions, while dicoumarol binding has the opposite effect [ 9 , 17 , 19 ]. In addition, NQO1 holo binds to the 20S particle of the proteasome and inhibits its proteolytic activity, while FAD withdrawal (i.e., NQO1 apo ) renders NQO1 susceptible to degradation by this mechanism [16]. Biomolecules 2019, 9, x FOR PEER REVIEW 2 of 17 quinones to form hydroquinones [2,6] (Figure 1), redox cycling of quinones [2], reduction of coenzyme Q10 and vitamin E to their antioxidant form [2], scavenging reactive oxygen species [1–3,7], reduction of catecholamines and vitamin K [1,2] and maintenance of the NADH/NAD+ redox balance [3,8]. The main features of these biochemical reactions involving NQO1 have been investigated in detail mainly through enzymological and structural analyses. Structurally, the enzyme forms obligate functional homodimers, with two active sites located in the monomer:monomer interface (MMI), and each monomer consists of two different domains: i) an N-terminal domain spanning residues 1–224 that contains part of the active site and is involved in the tight binding of one FAD molecule per monomer and protein dimerization; ii) a C-terminal domain (residues 225–274) that contributes to stabilizing the protein dimer and to the binding of the NAD(P)H coenzyme and the substrates [1,4,9– 14]. The functional cycle of NQO1 generally involves two steps according to a ping-pong mechanism: first, in the reductive half-reaction, an NAD(P)H molecule binds to the enzyme and rapidly reduces the FAD to FADH2 (with a second-order rate constant of ~106 M−1·s−1) with the subsequent release of the oxidized nicotinamide dinucleotide; and second, in a very fast oxidative half-reaction (with a second-order rate constant >109 M−1·s−1), the substrate binds and is reduced by the FADH2, thus regenerating the flavin in oxidized form and releasing the reduced product [1,10]. This catalytic cycle is known to be inhibited by different coumarin-based molecules (the best characterized is the biscoumarin, dicoumarol) that act as competitive inhibitors by blocking the NAD(P)H access to the active site and partially occupying the NAD(P)H binding site [1,15]. Importantly, comparison of the crystal structures of NQO1 with FAD bound (NQO1holo) with that containing also dicoumarol bound (NQO1dic) have revealed that inhibitor binding causes only minor structural rearrangements in the conformation, which localize at the surface of the catalytic site [15]. Among non-enzymatic functions, we must highlight the ability of NQO1 to develop protein: protein and protein: RNA interactions [1,2,16–20]. In particular, protein:protein interactions involving NQO1 are relevant to understand its multiple roles in physiological and pathological processes. NQO1 interacts with key transcription factors associated with cancer (e.g., p53, p73α and HIF-1α) [17,18] and proteins involved in HIV infection (e.g., Tat protein) [19], and these interactions increase the intracellular stability of these protein partners by preventing their degradation by the proteasome. These protein:protein interactions presumably depend on the functional state of NQO1: binding of NADH may increase the strength of these interactions, while dicoumarol binding has the opposite effect [9,17,19]. In addition, NQO1holo binds to the 20S particle of the proteasome and inhibits its proteolytic activity, while FAD withdrawal (i.e., NQO1apo) renders NQO1 susceptible to degradation by this mechanism [16]. Figure 1. NQO1 catalyzes the two-electron reduction of a wide range of quinones. This enzyme also reduces other organic and inorganic species, such as superoxide radicals and iron(III) ions. Alterations in NQO1 stability and function are associated to different extents with a variety of human diseases, including cancer, neurological disorders (such as Parkinson´s and Alzheimer´s diseases, multiple sclerosis and schizophrenia) and cardiovascular diseases [1,21]. In these cases, either the wild-type (WT) NQO1 protein and/or a common polymorphic variant (rs1800566, causing a Pro187Ser amino acid exchange) have been found to be associated with increased disease predisposition. The Pro187Ser variant decreases NQO1 activity due to a large defect in FAD binding Figure 1. NQO1 catalyzes the two-electron reduction of a wide range of quinones. This enzyme also reduces other organic and inorganic species, such as superoxide radicals and iron(III) ions. Alterations in NQO1 stability and function are associated to different extents with a variety of human diseases, including cancer, neurological disorders (such as Parkinson ´ s and Alzheimer ´ s diseases, multiple sclerosis and schizophrenia) and cardiovascular diseases [ 1 , 21 ]. In these cases, either the wild-type (WT) NQO1 protein and/or a common polymorphic variant (rs1800566, causing a Pro187Ser amino acid exchange) have been found to be associated with increased disease predisposition. Biomolecules 2019,9, 728 3 of 17 The Pro187Ser variant decreases NQO1 activity due to a large defect in FAD binding (10to 40-fold lower affinity than that of WT) and in conformational stability, leading to its rapid intracellular degradation by the proteasome [ 1 , 10 , 14 , 16 , 22 – 25 ]. In general, reduced NQO1 activity or protein levels are commonly observed under these pathological conditions [1,26], although for the particular case of cancer, overexpression of NQO1 is also associated with cancer progression, which makes pharmacological inhibition of NQO1 (e.g., by dicoumarol or related compounds) a potential therapeutic strategy to treat this disease if it selectively targets cancer cells [ 27 – 29 ]. Linked to some of these pathological conditions, the intracellular stability of NQO1 WT is controlled by the population of the NQO1 apo state, which is efficiently targeted to the ubiquitin-dependent proteasomal degradation pathway [ 14 , 25 , 30 ]. Recent works also demonstrated that alterations in the phosphorylation pattern of NQO1 WT at different sites might be associated with these pathological states, likely through effects on the FAD binding affinity and consequently on the intracellular stability of NQO1 WT [26,30,31]. NQO1 is an excellent model to decipher the role of protein dynamics in the function and stability of flavin-dependent enzymes, the role of ligand binding in disease-associated protein stability, and the molecularmechanismsbywhichmutationscauseloss-of-functiongeneticdiseases[ 1 , 5 , 9 , 10 , 14 , 25 , 30 – 34 ]. FAD binding to NQO1 WT triggers a large conformational change, which can be observed by some biophysical techniques (circular dichroism, infrared and NMR spectroscopies or small-angle X-ray scattering), and increases the kinetic stability of the protein dimer, although high-resolution structural information is only available for the NQO1 holo state [ 9 – 11 , 14 , 22 , 34 , 35 ]. This structural change is accompanied by significant changes in overall protein flexibility (evidence provided by proteolysis experiments and structure-based analyses of FAD binding energetics) [ 10 , 14 , 34 ], presumably linked to the fast degradation of NQO1 apo vs. NQO1 holo in the cell [ 14 , 25 ], although no high-resolution experimental information on these dynamic changes is available [ 14 ]. Regarding dicoumarol binding, the comparison of the X-ray crystallographic structures of NQO1 holo and NQO1 dic has revealed only local changes in protein structure at the active site [ 15 ], and thus, this did not provide details on the remarkable stabilizing effect of dicoumarol binding on the overall conformational stability and the dynamics of the C-terminal domain [ 14 ]. A critical role of protein dynamics in the mechanisms causing alterations in NQO1 function due to the Pro187Ser polymorphism and other rare cancer-associated mutations, phosphorylation at specific sites as well as the effect of suppressor mutations of the Pro187Ser phenotype have been put forward from experimental and computational studies [ 5 , 9 , 10 , 14 , 30 – 34 , 36 ]. Thus, they also await high-resolution information on the changes in protein dynamics due to these site-specific changes in different ligation states (NQO1apo, NQO1holo and NQO1dic). We report herein a detailed experimental analysis on the structural dynamics of human NQO1 in three functionally relevant ligation states (NQO1 apo , NQO1 holo and NQO1 dic ). Our results uncover the existence of a dynamic network within the NQO1 dimer that readily responds to the binding of functional ligands and help to explain their effects on NQO1 function and stability in vitro and in vivo . Our work also provides an experimental benchmark to understand the allosteric effects of disease-associated variants, post-translational modifications and ligand binding in NQO1. 2. Materials and Methods 2.1. Protein Expression and Purification Protein expression and purification was carried out as described [ 30 ]. E. coli BL21(DE3) cells were transformed with pET46 Ek/LIC vector containing the cDNA of human WT NQO1 [ 22 ] and grown for 16 h in LBA medium (LB containing 0.1 mg · ml −1 ampicillin at 37 ◦ C). This culture was diluted 40-fold in fresh LBA and grown at 37 ◦ C for 3 h. Expression was then triggered by the addition of IPTG (isopropyl β -D-1-thiogalactopyranoside) at a final concentration of 0.5 mM. Induced cells were incubated for 6 h at 25 ◦ C, harvested by centrifugation, washed with binding buffer (BB, 20 mM sodium phosphate, 300 mM NaCl and 50 mM imidazole at pH 7.4) and frozen overnight at − 80 ◦ C. Then, cells were thawed and resuspended in BB containing 1 mM PMSF (phenylmethylsulfonyl Biomolecules 2019,9, 728 4 of 17 fluoride) and lysed by sonication. Crude extracts were clarified by centrifugation (20 min at 20,000 × g and 4 ◦ C), and supernatants were loaded into immobilized metal affinity chromatography (IMAC) columns (GE Healthcare) equilibrated in BB. Columns were washed with BB, and the protein was eluted with elution buffer (BB containing 500 mM imidazole). The eluate was exchanged to 50 mM K-HEPES (2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid, potassium salt) pH 7.4 using PD-10 columns (GE Healthcare), centrifuged for 30 min at 20,000 × gand 4 ◦ C, and the UV–visible spectra of the supernatants were registered in a HP8453 UV–Visible spectrophotomer (Agilent). This purified holo-protein containing high levels of FAD bound (70%–80% based on the absorbance ratio at 450 nm and 280 nm; see also [ 30 , 33 , 36 ]) was stored at − 80 ◦ C upon flash freezing in liquid N 2 . Further purification of the NQO1 dimer was carried out by size-exclusion chromatography using a HiLoad ® 16/600 Superdex ® 200 prep grade (GE Healthcare) and using 20 mM K-HEPES 200 mM NaCl at pH 7.4 as the mobile phase. This purified protein was subsequently used to obtain apo-protein upon treatment with BB containing 2 M urea and 2 M KBr, 1 mM DTT (1,4-dithiothreitol) and 1 mM PMSF at 4 ◦ C, and separation of the apo-protein and the FAD released was carried out by IMAC at 4 ◦ C. Apo-proteins were finally exchanged to 50 mM K-HEPES at pH 7.4 using PD-10 columns at 4 ◦ C, concentrated using VIVASPIN 6 30,000 MWCO PES devices (Sartorius) and stored at − 80 ◦ C after flash freezing in liquid N 2 . The purity and integrity of the holoand apo-protein was checked by polyacrylamide gel electrophoresis in the presence of sodium dodecylsulphate (SDS-PAGE; Figure S1). The holo-protein was also found to display enzyme kinetic parameters (Figure S1) consistent with those previously reported using the same protein construct and purification procedure [9,22]. Purified proteins were also verified by mass spectrometry. The intact mass was analyzed through direct infusion on ESI-FT-ICR MS (Figure S2), showing that the protein was expressed intact and lacking the N-terminal Met. The dimeric state of NQO1 apo and NQO1 holo was verified by native ESI-MS (Figure S3). 2.2. Hydrogen/Deuterium Exchange Mass Spectrometry (HDXMS) Amide hydrogen/deuterium exchange (HDX) of NQO1 was followed for its apo (NQO1 apo ) and holo (NQO1 holo ) forms, and the holo form was also analyzed in the presence of dicoumarol (NQO1 dic ). Prior to the exchange, the NQO1 holo and NQO1 dic at 20 µ M concentration were pre-incubated with 10 molar excess of FAD for 5 min. NQO1dic was then further mixed with 10 molar excess of dicoumarol and incubated for another 5 min. The exchange reaction was initiated by a 10-fold dilution into a D 2 O-based 50 mM K-HEPES, pD 7.4, 1 mM TCEP (tris(2-carboxyethyl)phosphine). The exchange was thus followed at 2 µ M protein concentration. Deuterium labelling was quenched by 0.5 M Glycine-HCl, pH 2.3, which was added at 1:1 ratio. The samples were then frozen in liquid N 2 . Exchange was followed for 10 s, 30 s, 2 min, 5 min, 20 min, 1 h and 3 h, where 10 s, 5 min and 3 h samples were done in replicate. Each sample was quickly thawed and injected into a cooled LC system. Here, the protein was digested on custom-made nepenthesin-2 (Nep-2) and pepsin columns coupled in tandem (each having a bed volume of 66 µ L), and the resulting peptides were trapped on a VanGuard Pre-column (ACQUITY UPLC BEH C18, 130Å, 1.7 µ m, 2.1 mm × 5 mm, Waters, Milford, MA, USA), where they were desalted. The solvent used for digestion and desalting (0.4% formic acid (FA) in water) was pumped by a 1260 Infinity II Quaternary pump (Agilent Technologies, Waldbronn, Germany) at a flow rate of 200 µ L · min −1 . After three minutes of digestion and desalting, the peptides were separated on an analytical column (ACQUITY UPLC BEH C18, 130 Å, 1.7 µ m, 1 mm × 100 mm, Waters, Milford, MA, USA) using linear gradient (5%–45% B in 7 min) followed by a quick step to 99% B lasting 5 min. Solvent A was 0.1% FA/2% acetonitrile (ACN) in water, B was 0.1% FA/98% ACN in water. The gradient was delivered by the 1290 Infinity II LC System (Agilent Technologies, Waldbronn, Germany) at a flow of 40 µ L · min −1 . Digestion, desalting and separation were done at 0 ◦ C and pH 2.3 to minimize deuterium loss. The LC system was connected directly to an electrospray ionization source of a 15T FT-ICR mass spectrometer (solarix XR, Bruker Daltonics, Bremen, Germany) operating in broad-band MS mode. Data were peak-picked and exported using DataAnalysis v. 5.0 (Bruker Daltonics, Bremen, Biomolecules 2019,9, 728 5 of 17 Germany) and then processed by the in-house developed program Deutex (unpublished). Peptides arising from the digestion were identified through separate data-dependent LC-MS/MS analyses (using the same setup as described above) and database searching by the MASCOT algorithm v. 2.4 (Matrix Science, London, United Kingdom) against a custom-build database containing sequences of pepsin, nepenthesin-2 and NQO1. Fully deuterated samples were prepared and used for back-exchange correction as described previously [ 37 , 38 ]. Differences in deuteration can be considered as significant if they exceed 0.25 Da (calculated as 3×the average standard deviation). The optimization of digestion conditions, including numerous proteolytical setups, showed that the serial combination of nepenthesin-2 with pepsin, operated at 200 µ L · min −1 , provided the best results in terms of sequence coverage (98.9%, missing the last three amino acids), number of peptides (140), average peptide length (8.3) and redundancy (4.1) (Figure S4). At this point, it should be also noted that the region between 100 and 110 yielded peptides of considerable hydrophobicity for which the signal intensity/quality was just at the threshold level, and thus conclusions derived from their analyses must be made with caution. The redundant peptide set was used to calculate deuteration in the shortest possible segments using the overlapping peptides. This analysis provided more detailed non-redundant information (high-resolution set). Data in this high-resolution set are mainly described in the manuscript (Figures 2–7, Figure S6 and Table S1), while those using experimental peptides (low resolution set) are found in Figures S7–S11 and Table S2. Note that the high-resolution set essentially leads to the same key conclusions as the low resolution one but, in principle, the former narrows the region for which HDX kinetics is assessed. To report data on NQO1 segments from HDXMS, we did not consider the His-tag and used the native sequence from Met1 to Lys274. Therefore, the numbering used throughout the manuscript differs from that reported in some crystal structures of NQO1, which did not include Met1 and thus numbered the residues from Val1 to Lys273 (Val2 to Lys274 in the native sequence). 3. Results and Discussion 3.1. A Stable Folded Core in NQO1apo with Highly Dynamic Functional Sites The results of HDX kinetics for NQO1 apo are shown in Figure 2A. Results are presented as the % of the maximal (theoretical) deuterium incorporation for each segment (%D). Virtually all peptides characterized in this work for NQO1 complied with EX2 behavior (the presence of a tiny contribution from the EX1 regime can be detected in a few NQO1 peptides; see Figure S5). In the EX2 mechanism, the intrinsic exchange rate constant (k int ) is much lower than the rate constant (k cl ) for the conversion between non-exchanging (NE-NH) and exchanging (E-NH) states, according to the Linderstrøm-Lang model (Scheme 1): Biomolecules 2019, 9, x FOR PEER REVIEW 5 of 17 by the MASCOT algorithm v. 2.4 (Matrix Science, London, United Kingdom) against a custom-build database containing sequences of pepsin, nepenthesin-2 and NQO1. Fully deuterated samples were prepared and used for back-exchange correction as described previously [37,38]. Differences in deuteration can be considered as significant if they exceed 0.25 Da (calculated as 3× the average standard deviation). The optimization of digestion conditions, including numerous proteolytical setups, showed that the serial combination of nepenthesin-2 with pepsin, operated at 200 μL·min−1, provided the best results in terms of sequence coverage (98.9%, missing the last three amino acids), number of peptides (140), average peptide length (8.3) and redundancy (4.1) (Figure S4). At this point, it should be also noted that the region between 100 and 110 yielded peptides of considerable hydrophobicity for which the signal intensity/quality was just at the threshold level, and thus conclusions derived from their analyses must be made with caution. The redundant peptide set was used to calculate deuteration in the shortest possible segments using the overlapping peptides. This analysis provided more detailed non-redundant information (high-resolution set). Data in this high-resolution set are mainly described in the manuscript (Figures 2–7, Figure S6 and Table S1), while those using experimental peptides (low resolution set) are found in Figures S7–S11 and Table S2. Note that the high-resolution set essentially leads to the same key conclusions as the low resolution one but, in principle, the former narrows the region for which HDX kinetics is assessed. To report data on NQO1 segments from HDXMS, we did not consider the His-tag and used the native sequence from Met1 to Lys274. Therefore, the numbering used throughout the manuscript differs from that reported in some crystal structures of NQO1, which did not include Met1 and thus numbered the residues from Val1 to Lys273 (Val2 to Lys274 in the native sequence). 3. Results and Discussion 3.1.A Stable Folded Core in NQO1apo with Highly Dynamic Functional Sites The results of HDX kinetics for NQO1apo are shown in Figure 2A. Results are presented as the % of the maximal (theoretical) deuterium incorporation for each segment (%D). Virtually all peptides characterized in this work for NQO1 complied with EX2 behavior (the presence of a tiny contribution from the EX1 regime can be detected in a few NQO1 peptides; see Figure S5). In the EX2 mechanism, the intrinsic exchange rate constant (kint) is much lower than the rate constant (kcl) for the conversion between non-exchanging (NE-NH) and exchanging (E-NH) states, according to the LinderstrømLang model (Scheme: 1): kop kint 𝑁𝐸−𝑁𝐻 ↔ 𝐸−𝑁𝐻 → 𝑒𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑑 𝑁𝐻 kcl Scheme 1. Linderstrøm-Lang model for HDX kinetics. Assuming a pure EX2 behavior, the rate constant for exchange of individual backbone amides would be equal to the product of the equilibrium constant between NE-NH and E-NH (Kop = kop/kcl) and kint [39]. Therefore, for this very simple mechanism (note that the conformational equilibrium is simply two-state and experimental HDX is rarely pure EX2), the experimental rate constant for exchange reflects to some extent the local stability (e.g., due to hydrogen bonding and burial in the structure) of the secondary structure. Overall, the HDX kinetics was very heterogeneous among different protein segments of NQO1apo (Figure 2A). Most of the segments showed fast HDX kinetics (typically exchanging more than 20%D in the seconds-minutes time scales) while only a few peptides showed essentially no exchange after 3 h (%D < 20). Thus, in a first approach, we simply discerned between exchanging and non-exchanging segments considering the %D after 3 h (%D < 20 vs. %D ≥ 20, respectively). The functional implications of this simple analysis were considered regarding those residues in different Scheme 1. Linderstrøm-Lang model for HDX kinetics. Assuming a pure EX2 behavior, the rate constant for exchange of individual backbone amides would be equal to the product of the equilibrium constant between NE-NH and E-NH (K op =k op /k cl ) and k int [ 39 ]. Therefore, for this very simple mechanism (note that the conformational equilibrium is simply two-state and experimental HDX is rarely pure EX2), the experimental rate constant for exchange reflects to some extent the local stability (e.g., due to hydrogen bonding and burial in the structure) of the secondary structure. Overall, the HDX kinetics was very heterogeneous among different protein segments of NQO1 apo (Figure 2A). Most of the segments showed fast HDX kinetics (typically exchanging more than 20%D in the seconds-minutes time scales) while only a few peptides showed essentially no exchange after 3 h Biomolecules 2019,9, 728 6 of 17 (%D <20). Thus, in a first approach, we simply discerned between exchanging and non-exchanging segments considering the %D after 3 h (%D <20 vs. %D ≥ 20, respectively). The functional implications of this simple analysis were considered regarding those residues in different segments belonging to three functional sites: the FAD binding site (FBS), the dicoumarol binding site (DBS) and the monomer: monomer interface (MMI)(see Figure 2A), as provided by analysis of an X-ray crystallographic structure (PDB 2F1O [ 15 ]). It is worth noting that in this structure, the FBS and DBS are located adjacent to each other in the NQO1 monomer (actually, FAD is structurally part of the DBS), and both sites are close to the MMI (Figure 2B). Importantly, most of the residues that belong to the FBS, DBS and MMI are classified as exchanging in NQO1apo (%D ≥20; Figure 2A,B). Biomolecules 2019, 9, x FOR PEER REVIEW 6 of 17 segments belonging to three functional sites: the FAD binding site (FBS), the dicoumarol binding site (DBS) and the monomer: monomer interface (MMI)(see Figure 2A), as provided by analysis of an Xray crystallographic structure (PDB 2F1O [15]). It is worth noting that in this structure, the FBS and DBS are located adjacent to each other in the NQO1 monomer (actually, FAD is structurally part of the DBS), and both sites are close to the MMI (Figure 2B). Importantly, most of the residues that belong to the FBS, DBS and MMI are classified as exchanging in NQO1apo (%D ≥ 20; Figure 2A,B). Figure 2. Overall HDX kinetics of NQO1apo. (A) HDX kinetics for segments of NQO1apo. Residues belonging to the monomer: monomer interface (MMI), FAD binding site (FBS) and dicoumarol binding site (DBS) are indicated as colored circles. These sites were retrieved from the analysis of the NQO1 structure (PDB 2F1O; [15]) using the PISA server (https://www.ebi.ac.uk/pdbe/pisa/); (B) Structural representation of HDX after 3 h [%D (3 h)] in NQO1apo. For visual aid, the left panel shows the location of bound FAD and dicoumarol, the middle panel shows those residues belonging to the MMI, FBS and DBS, and the right panel displays the %D after 3 h for residues belonging to these functional sites. Within non-exchanging segments, we found that these contained mostly residues buried in the crystallographic structure of NQO1 (in a ternary complex with FAD and dicoumarol bound, PDB 2F1O; NQO1dic) (Figure 3A). Thus, these sequences likely represent regions that are critical for the acquisition and maintenance of a minimally stable dimeric fold in NQO1apo (note that NQO1apo is dimeric in solution but more expanded and flexible, and with lower conformational stability, than NQO1holo [9,14]). This minimal core involves helices α1, α3 and α4 and sheets β1 and β3–β5 (Figures 3A and S8). This core may also contribute to the acquisition of a minimally folded monomeric state that becomes stabilized in the dimeric state by the interactions between helices α3 and α4 across the monomers (i.e. the MMI) (Figures 3B and S8). Figure 2. Overall HDX kinetics of NQO1 apo . ( A ) HDX kinetics for segments of NQO1 apo . Residues belonging to the monomer: monomer interface (MMI), FAD binding site (FBS) and dicoumarol binding site (DBS) are indicated as colored circles. These sites were retrieved from the analysis of the NQO1 structure (PDB 2F1O; [ 15 ]) using the PISA server (https://www.ebi.ac.uk/pdbe/pisa/); ( B ) Structural representation of HDX after 3 h [%D (3 h)] in NQO1 apo . For visual aid, the left panel shows the location of bound FAD and dicoumarol, the middle panel shows those residues belonging to the MMI, FBS and DBS, and the right panel displays the %D after 3 h for residues belonging to these functional sites. Within non-exchanging segments, we found that these contained mostly residues buried in the crystallographic structure of NQO1 (in a ternary complex with FAD and dicoumarol bound, PDB 2F1O; NQO1 dic ) (Figure 3A). Thus, these sequences likely represent regions that are critical for the acquisition and maintenance of a minimally stable dimeric fold in NQO1 apo (note that NQO1 apo is dimeric in solution but more expanded and flexible, and with lower conformational stability, than NQO1 holo [ 9 , 14 ]). This minimal core involves helices α 1, α 3 and α 4 and sheets β 1 and β 3– β 5 (Figure 3A and Figure S8). This core may also contribute to the acquisition of a minimally folded monomeric state that becomes stabilized in the dimeric state by the interactions between helices α 3 and α4 across the monomers (i.e., the MMI) (Figure 3B and Figure S8). Biomolecules 2019,9, 728 7 of 17 Biomolecules 2019, 9, x FOR PEER REVIEW 7 of 17 Figure 3. Non-exchanging segments define a minimally stable core in NQO1apo. (A) Plot of the % SASA (solvent accessible surface area) for individual residues (considering backbone and side-chain) calculated from the structure of NQO1dic (PDB 2F1O [15]) using GETAREA (http://curie.utmb.edu/getarea.html; this algorithm does not consider the ligands in the calculation). Secondary structure elements are depicted according to [11]; residues belonging to non-exchanging segments are displayed as green circles; (B) Structural representation of non-exchanging segments (using PDB 2F1O [15]). The left panel shows a surface representation highlighting the burial of the minimally stable core. The middle panel shows segments belonging to this core plotted onto secondary structure elements. The right panel shows that this core may contribute to the stable folding of the individual monomers as well as their assembly into the dimer, with only a few stable contacts with the FAD (in orange ball representation) and the dicoumarol (in yellow ball representation). Importantly, the FBS and DBS in NQO1apo are overall exchanging (Figures 2B, 3B and S8B), with the main exceptions being some marginal contacts in helices α1 and α3 (FBS) and α4 (DBS) (Figures 3B and S8B). Thus, our HDX analyses support that the conformational ensemble of NQO1apo is essentially populated by states non-competent for FAD or dicoumarol binding due to the high structural dynamics of their binding sites [9,14,34]. 3.2. Complex HDX Kinetics To provide deeper insight into the structural dynamics of NQO1apo, we carefully analyzed the HDX kinetics for all protein segments (Figure S6). It should be noted that these HDX kinetics were very consistent with those obtained directly from peptides experimentally characterized (Figure S7 and S11). For the majority of the cases, HDX kinetics was described very well by a simple function with two kinetic phases (see Figures S6 and S7 for fittings, and Tables S1 and S2 for the best-fit values): a burst-phase corresponding to HDX mostly occurring within the experimental dead time (i.e., very few seconds), and thus characterized by a single parameter: its amplitude Aburst; and a slow phase that occurred typically in a scale of several seconds to minutes, characterized by two parameters: its amplitude Aslow and an apparent first-order rate constant (kslow), following this equation: % 𝐷(𝑡)= 𝐴 +𝐴  ·(1−𝑒𝑥𝑝 ·) Figure 3. Non-exchanging segments define a minimally stable core in NQO1 apo . ( A ) Plot of the % SASA (solventaccessiblesurfacearea)forindividualresidues(consideringbackboneandside-chain)calculated from the structure of NQO1 dic (PDB 2F1O [ 15 ]) using GETAREA (http://curie.utmb.edu/getarea.html; this algorithm does not consider the ligands in the calculation). Secondary structure elements are depicted according to [ 11 ]; residues belonging to non-exchanging segments are displayed as green circles; ( B ) Structural representation of non-exchanging segments (using PDB 2F1O [ 15 ]). The left panel shows a surface representation highlighting the burial of the minimally stable core. The middle panel shows segments belonging to this core plotted onto secondary structure elements. The right panel shows that this core may contribute to the stable folding of the individual monomers as well as their assembly into the dimer, with only a few stable contacts with the FAD (in orange ball representation) and the dicoumarol (in yellow ball representation). Importantly, the FBS and DBS in NQO1 apo are overall exchanging (Figure 2B, Figure 3B and Figure S8B), with the main exceptions being some marginal contacts in helices α 1 and α 3 (FBS) and α 4 (DBS) (Figure 3B and Figure S8B). Thus, our HDX analyses support that the conformational ensemble of NQO1 apo is essentially populated by states non-competent for FAD or dicoumarol binding due to the high structural dynamics of their binding sites [9,14,34]. 3.2. Complex HDX Kinetics To provide deeper insight into the structural dynamics of NQO1 apo , we carefully analyzed the HDX kinetics for all protein segments (Figure S6). It should be noted that these HDX kinetics were very consistent with those obtained directly from peptides experimentally characterized (Figure S7 and S11). For the majority of the cases, HDX kinetics was described very well by a simple function with two kinetic phases (see Figures S6 and S7 for fittings, and Tables S1 and S2 for the best-fit values): a burst-phase corresponding to HDX mostly occurring within the experimental dead time (i.e., very few seconds), and thus characterized by a single parameter: its amplitude A burst ; and a slow phase that occurred typically in a scale of several seconds to minutes, characterized by two parameters: its amplitude Aslow and an apparent first-order rate constant (kslow), following this equation: %D(t) = Aburst +Aslow·(1−exp−kslow·t) Biomolecules 2019,9, 728 8 of 17 We chose to use this phenomenological description of HDX kinetics mainly for two reasons. First, it provided a simple scenario from which, using three characteristic parameters (A burst , A slow and k slow ), we could compare the HDX kinetics of different segments of NQO1 apo (see Figure 4; note that this approach also worked very well with the HDX kinetics of NQO1 holo and NQO1 dic ; see Figures S6 and S7 and Tables S1 and S2). Second, although HDX kinetics analyzed using more complex functions (e.g., the two kinetic phases each containing a characteristic rate constant) may provide in some cases better fits, this would put the analyses at two intertwined risks: increasing the fitting parameters would make comparisons between behaviors more difficult, and importantly, in many cases these fittings show evident signs of overparametrization. Kinetic analyses of HDX for protein segments considered as exchanging (>20%D after 3 h) revealed certain interesting behaviors. First, for many protein segments (of different lengths), we observed a significant contribution to the HDX kinetics from both the burst and slow phases (Figure 4and Figure S9, and Tables S1 and S2). As indicated above, the HDX kinetics of NQO1 in all three ligation states is vastly consistent with EX2 kinetics, and thus, the observed kinetics depends to some extent on the equilibrium constant between non-exchanging and exchanging states [ 40 ]. Therefore, the presence of two clearly differentiated kinetic phases suggests the existence of complexity (i.e., heterogeneity) in the conformational ensemble of NQO1 apo , and plausibly, the significant population of at least two conformational substates with different HDX behavior, which may or may not significantly reequilibrate upon the intrinsic HDX step. Interestingly, although ligand binding affects these two kinetic phases (NQO1 holo and NQO1 dic , see Figures S6 and S7 and Tables S1 and S2), both these phases still contribute to the HDX kinetics in these ligation states, suggesting that a certain degree of conformational heterogeneity remains upon ligand binding. Second, although in the EX2 scenario the overall kinetics depends on the intrinsic HDX rate constant, and therefore, on the individual backbone amides and their vicinity [ 40 ], some sort of correlated behavior at larger scales than small protein segments (e.g., secondary structure elements) is observed (Figure 4B and Figure S9). Consistent with the above-mentioned proposal of a stable core of NQO1 apo with highly dynamic FBS and DBS (simply made by analysis of %D after 3 h, Figure 3), these kinetic analyses suggest that secondary structure elements outside the stable core typically exchanged quite fast (i.e., with large burst phases and with kslow often in the range of 10−1to 10−2s−1; Figure 4B and Figure S9). 3.3. FAD and Dicoumarol Binding Cause Large-Scale Changes in Protein Structural Dynamics FAD binding to NQO1 apo is known to cause significant overall changes in protein structure and dynamics: it increases the content in the ordered secondary structure, reduces the protein hydrodynamic volume, and substantially enhances protein stability and resistance towards proteolytic attack [ 9 , 14 , 22 , 34 , 35 ]. In addition, structural and biophysical analyses have shown that NQO1 must contain bound FAD in order to bind dicoumarol with high affinity [ 9 , 14 ]. We first compared the %D incorporated to NQO1 holo and NQO1 apo after 3 h of reaction (Figure 5A,B), observing some interesting changes upon FAD binding. Particularly large differences were observed in loop L1 (involved in the MMI and the FBS), loop L4 (involved in the MMI, the FBS and the DBS) and helix α 5 (involved in the FBS). The stabilization observed for the MMI thus explains the increased thermostability of the NQO1 dimer upon FAD binding. The much lower structural dynamics of the FBS upon FAD binding is also consistent with an induced-fit mechanism, in which binding competent states (with high structural stability) are marginally populated in the absence of FAD, according to a recent proposal based on binding structure-thermodynamic relationships [ 34 ]. Interestingly, these results also imply that FAD is not only required for dicoumarol binding as a part of the DBS, but also that FAD binding modifies the dynamics of protein structural elements involved in the binding of the inhibitor (Figure 5C). It is worth noting that FAD binding also significantly slows down (3to 5-fold) HDX of other regions such as sheet β 6, helix α 7 and loop L3 (Figure 6and Figure S6, and Table S1), some of them not directly involved in the MMI, FBS or DBS. Biomolecules 2019,9, 728 9 of 17 Biomolecules 2019, 9, x FOR PEER REVIEW 9 of 17 Figure 4. Segment-specific HDX kinetics of NQO1apo. (A) Fittings of HDX kinetics for selected segments typically showing at least 50%D incorporation after 3 h and considered part of the functional sites (MMI, FBS and DBS). (B) Plots of the amplitudes for the burstand slow-phase in HDX for segments (upper panel) and rate constants for the slow phase (lower panel) for segments with at least 20%D after 3 h. The elements of the secondary structure along the protein sequence are also indicated. 3.3. FAD and Dicoumarol Binding Cause Large-Scale Changes in Protein Structural Dynamics FAD binding to NQO1apo is known to cause significant overall changes in protein structure and dynamics: it increases the content in the ordered secondary structure, reduces the protein hydrodynamic volume, and substantially enhances protein stability and resistance towards proteolytic attack [9,14,22,34,35]. In addition, structural and biophysical analyses have shown that NQO1 must contain bound FAD in order to bind dicoumarol with high affinity [9,14]. We first compared the %D incorporated to NQO1holo and NQO1apo after 3 h of reaction (Figure 5A–B), Figure 4. Segment-specific HDX kinetics of NQO1 apo . ( A ) Fittings of HDX kinetics for selected segments typically showing at least 50%D incorporation after 3 h and considered part of the functional sites (MMI, FBS and DBS). ( B ) Plots of the amplitudes for the burstand slow-phase in HDX for segments (upper panel) and rate constants for the slow phase (lower panel) for segments with at least 20%D after 3 h. The elements of the secondary structure along the protein sequence are also indicated. Biomolecules 2019,9, 728 16 of 17 20. DiFrancesco, A.; DiGermanio, C.; Panda, A.C.; Huynh, P.; Peaden, R.; Navas-Enamorado, I.; Bastian, P.; Lehrmann, E.; Diaz-Ruiz, A.; Ross, D.; et al. Novel RNA-binding activity of NQO1 promotes SERPINA1 mRNA translation. Free Radic. Biol. Med. 2016,99, 225–233. [CrossRef] [PubMed] 21. Betancor-Fernandez, I.; Timson, D.J.; Salido, E.; Pey, A.L. Natural (and unnatural) small molecules as pharmacological chaperones and inhibitors in cancer. Handb. Exp. Pharmacol. 2018,45, 345–383. 22. Pey, A.L.; Megarity, C.F.; Timson, D.J. FAD binding overcomes defects in activity and stability displayed by cancer-associated variants of human NQO1. Biochim. Biophys. Acta 2014 ,1842, 2163–2173. [CrossRef] [PubMed] 23. Claveria-Gimeno, R.; Velazquez-Campoy, A.; Pey, A.L. Thermodynamics of cooperative binding of FAD to human NQO1: Implications to understanding cofactor-dependent function and stability of the flavoproteome. Arch. Biochem. Biophys. 2017,636, 17–27. [CrossRef] [PubMed] 24. Siegel, D.; Anwar, A.; Winski, S.L.; Kepa, J.K.; Zolman, K.L.; Ross, D. Rapid polyubiquitination and proteasomal degradation of a mutant form of NAD (P) H: Quinone oxidoreductase 1. Mol. Pharmacol. 2001 , 59, 263–268. [CrossRef] [PubMed] 25. Martinez-Limon, A.; Alriquet, M.; Lang, W.H.; Calloni, G.; Wittig, I.; Vabulas, R.M. Recognition of enzymes lacking bound cofactor by protein quality control. Proc. Natl. Acad. Sci. USA 2016 ,113, 12156–12161. [CrossRef] [PubMed] 26. Luo, S.; SuKang, S.; Wang, Z.H.; Liu, X.; Day, J.X.; Wu, Z.; Peng, J.; Xiang, D.; Springer, W.; Ye, K. Akt Phosphorylates NQO1 and Triggers its Degradation, Abolishing Its Antioxidative Activities in Parkinson’s Disease. J. Neurosci. 2019,39, 7291–7305. [CrossRef] [PubMed] 27. Nolan, K.A.; Scott, K.A.; Barnes, J.; Doncaster, J.; Whitehead, R.C.; Stratford, I.J. Pharmacological inhibitors of NAD (P) H quinone oxidoreductase, NQO1: Structure/activity relationships and functional activity in tumour cells. Biochem. Pharmacol. 2010,80, 977–981. [CrossRef] [PubMed] 28. Nolan, K.A.; Zhao, H.; Faulder, P.F.; Frenkel, A.D.; Timson, D.J.; Siegel, D.; Ross, D.; Burke, T.R., Jr.; Stratford, I.J.; Bryce, R.A. Coumarin-based inhibitors of human NAD (P) H: Quinone oxidoreductase-1. Identification, structure–activity, off-target effects and in vitro human pancreatic cancer toxicity. J. Med. Chem. 2007,50, 6316–6325. [CrossRef] [PubMed] 29. Scott, K.A.; Barnes, J.; Whitehead, R.C.; Stratford, I.J.; Nolan, K.A. Inhibitors of NQO1: Identification of compounds more potent than dicoumarol without associated off-target effects. Biochem. Pharmacol. 2011 ,81, 355–363. [CrossRef] [PubMed] 30. Medina-Carmona, E.; Rizzuti, B.; Martin-Escolano, R.; Pacheco-Garcia, J.L.; Mesa-Torres, N.; Neira, J.L.; Guzzi, R.; Pey, A.L. Phosphorylation compromises FAD binding and intracellular stability of wild-type and cancer-associated NQO1: Insights into flavo-proteome stability. Int. J. Biol. Macromol. 2019 ,125, 1275–1288. [CrossRef] [PubMed] 31. Mesa-Torres, N.; Betancor-Fern á ndez, I.; Oppici, E.; Cellini, B.; Salido, E.; Pey, A.L. Evolutionary divergent suppressor mutations in conformational diseases. Genes 2018,9, 352. [CrossRef] [PubMed] 32. Medina-Carmona, E.; Betancor-Fern á ndez, I.; Santos, J.; Mesa-Torres, N.; Grottelli, S.; Batlle, C.; Naganathan, A.N.; Oppici, O.; Cellini, B.; Ventura, S.; et al. Insight into the specificity and severity of pathogenic mechanisms associated with missense mutations through experimental and structural perturbation analyses. Hum. Mol. Genet. 2019,28, 1–15. [CrossRef] [PubMed] 33. Pey, A.L. Biophysical and functional perturbation analyses at cancer-associated P187 and K240 sites of the multifunctional NADP (H): Quinone oxidoreductase 1. Int. J. Biol. Macromol. 2018 ,118, 1912–1923. [CrossRef] [PubMed] 34. Munoz, I.G.; Morel, B.; Medina-Carmona, E.; Pey, A.L. A mechanism for cancer-associated inactivation of NQO1 due to P187S and its reactivation by the consensus mutation H80R. FEBS Lett. 2017 ,591, 2826–2835. [CrossRef] [PubMed] 35. Lienhart, W.D.; Strandback, E.; Gudipati, V.; Koch, K.; Binter, A.; Uhl, M.K.; Rantasa, D.M.; Bourgeois, B.; Madl, T.; Zangger, K.; et al. Catalytic competence, structure and stability of the cancer-associated R139W variant of the human NAD (P) H: Quinone oxidoreductase 1 (NQO 1). FEBS J. 2017 ,284, 1233–1245. [CrossRef] [PubMed] 36. Medina-Carmona, E.; Fuchs, J.E.; Gavira, J.A.; Mesa-Torres, N.; Neira, J.L.; Salido, E.; Palomino-Morales, R.; Burgos, M.; Timson, D.J.; Pey, A.L. Enhanced vulnerability of human proteins towards disease-associated inactivation through divergent evolution. Hum. Mol. Genet. 2017,26, 3531–3544. [CrossRef] [PubMed] Biomolecules 2019,9, 728 17 of 17 37. Zhang, Z.; Smith, D.L. Determination of amide hydrogen exchange by mass spectrometry: A new tool for protein structure elucidation. Protein Sci. 1993,2, 522–531. [CrossRef] [PubMed] 38. Trcka,F.; Durech,M.; Vankova,P.; Chmelik, J.; Martinkova,V.; Hausner,J.; Kadek, A.; Marcoux,J.; Klumpler,T.; Vojtesek, B.; et al. Human stress-inducible Hsp70 has a high propensity to form ATP-dependent antiparallel dimers that are differentially regulated by Cochaperone binding. Mol. Cell Proteom. 2019 ,18, 320–337. [CrossRef] [PubMed] 39. Bai, Y. Hydrogen Exchange Experiments: Detection and Characterization of Protein Folding Intermediates. In Protein Folding, Misfolding and Aggregation; Muñoz, V., Ed.; Royal Society of Chemistry: Cambridge, UK, 2008; pp. 70–83. 40. Konermann, L.; Pan, J.; Liu, Y.H. Hydrogen exchange mass spectrometry for studying protein structure and dynamics. Chem. Soc. Rev. 2011,40, 1224–1234. [CrossRef] [PubMed] 41. Megarity, C.F.; Abdel-Bettley, H.; Caraher, M.C.; Scott, K.A.; RA, W.; Jowitt, T.A.; Gutierrez, A.; Bryce, R.A.; Nolan, K.A.; Stratford, I.J.; et al. Negative cooperativity in NAD (P) H quinone oxidoreductase 1 (NQO1). ChemBioChem 2019. [CrossRef] [PubMed] 42. Wyman, J.; Gill, S.J. Binding and Linkage: Functional Chemistry of Biological Macromolecules; University Science Books: Mill Valley, CA, USA, 1990. © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).