Quinones are redox active biological molecules that are found in all domains of life and can be broadly sorted into three groups based on their core structure: benzoquinones, naphthoquinones, and anthraquinones (Fig. 1). Ubiquinone, also known as coenzyme Q, is among the most common quinones in higher organisms. It is a 1,4-benzoquinone with an isoprenoid tail of varying lengths. The aliphatic tail results in ostensibly all ubiquinones being localized to the inner hydrophobic environment of membranes and it is found in most, if not all, biological membranes [1]. Another notable 1,4-benzoquinone is plastoquinone, which is found primarily in chloroplasts [2]. Naphthoquinones also occur naturally in plants and some bacteria in the form of phylloquinone and menaquinone, known as vitamins K1 and K2 respectively [3], while humans can also utilize synthetic vitamin K3 (menadione) [4]. Anthraquinones constitute the most diverse family of quinones and are found primarily in fungi and higher plants, although examples have been identi昀椀ed in all domains of life [5]. 1,2-benzoquinones are also found naturally, although they are relatively less stable and are typically found in the reduced catechol form, such as dopamine and adrenaline.
The structural diversity of quinones mirrors their diverse physiological functions. Among the most well documented functions of quinones are their role in respiratory electron transport. Their ubiquity in membranes allows them to transport electrons between membrane associated respiratory complexes, such as those in the mitochondrial electron transport chain or photosynthesis [6,7]. They can also be used in processes which shuttle electrons outside of the cell, such as is the case for Fe 3+ reducing bacteria [8]. Quinones are found in several anabolic pathways including heme [9] and pyrimidine biosynthesis [10]. Reduced quinols are potent antioxidants and can counteract oxidative stress [11]. This is far from a comprehensive list of the biological functions of quinones, and assembling such a list would be a nearly insurmountable task due to the wide range of processes they are involved in. However, it is clear that quinones are either integral to or indirectly involved in most cellular processes.
Quinone reducing enzymes are crucial to the antioxidant and metabolic roles of quinones, which rely on the regeneration of reduced quinols. As will be demonstrated here, these enzymes vary greatly in their reductant substrate, subcellular localization, and structure. Flavins are common cofactors for mediating redox chemistry and are utilized by a number of quinone reducing enzymes, offering the advantage of mediating both one and two electron chemistry and thus widening the range of reactions they can catalyze. The two-electron reduction of substrates by hydride transfer is well documented in 昀氀avoproteins [12,13] and proceeds via a direct hydride transfer from N5 of the 昀氀avin to an acceptor site on the substrate, requiring these two loci to be in close proximity. A survey of the 昀氀avoprotein quinone oxidoreductases in the PDB reveals that quinones typically bind on the periphery of the 昀氀avin isoalloxazine, nearby the xylene or pyrimidine ends, precluding a hydride transfer mechanism. Confoundingly, these enzymes generate primarily quinols, and the semiquinone products are somewhat less common. As such, the peculiar binding modes of quinones in 昀氀avoprotein oxidoreductases and obligatory one electron reduction steps likely serve a purpose other than generating the single electron reduced quinones.
Here we present a cursory review of quinone utilizing 昀氀avoprotein oxidoreductases and their quinone binding sites and discuss plausible rationalizations for this binding mode. The landscape of quinone utilizing enzymes is broad, and we have selected a subset for review using several criteria. We have selected only enzymes which utilize 昀氀avins to directly reduce quinones, without intermediate electron carriers. This excludes well known enzymes, such as complex I, which use iron-sulfur clusters to transmit electrons between the 昀氀avin and quinone, but not enzymes which use other cofactors for internal electron transfer, such as NQR. Additionally, we have selected enzymes that have structures in the Protein Data Bank with quinones or quinone analogues bound and which have been studied suf昀椀ciently to draw conclusions about their mechanisms. Using the UniProt database and these criteria, we have narrowed down the scope to six well documented enzymes that cover the breadth of the quinone utilizing 昀氀avoprotein oxidoreductases. This is not intended to be a comprehensive review of each enzyme, nor is it a comprehensive review of all quinone utilizing enzymes. Rather, we aim to establish a framework from which to discuss the binding modes of quinones in 昀氀avoprotein oxidoreductases.
1.1. Sul昀椀de-quinone oxidoreductase (SQOR) 1.1.1. Background Sul昀椀de-quinone oxidoreductase (SQR or SQOR) is a 昀氀avoenzyme belonging to the 昀氀avin-disul昀椀de reductase (FDR) class of enzymes [14]. SQOR reduces quinone substrates using electrons derived from reduced sulfur compounds, most commonly H 2 S, resulting in the detoxi昀椀cation of these compounds as well as the production of reducing equivalents that can be used metabolically in the form of reduced quinones [15]. The physiological function of this activity is highly context dependent. The 昀椀rst SQOR described in the literature is from O. limnetica, an extremophilic cyanobacterium that utilizes hydrogen sul昀椀de as its primary anoxygenic energy source [15]. In this case, SQOR catalyzes the 昀椀rst reaction of anoxygenic photosynthesis. In higher eukaryotes, SQOR is in the mitochondria and primarily detoxi昀椀es hydrogen sul昀椀de from endogenous and exogenous sources [16,17]. Additionally, it has recently been implicated in primary metabolism in the biosynthesis of cysteine [18]. The development of SQOR inhibitors as a treatment for heart failure has begun and primarily targets the quinone binding site [19]. The reductive half reaction of SQORs has been studied extensively [17,[20][21][22][23][24][25] and produces polysulfur or persul昀椀de products (Fig. 2) and the mechanisms of sul昀椀de oxidation in SQOR has been comprehensively reviewed [26].
The substrate speci昀椀city of SQORs varies from species to species. Most SQORs that have been examined show a broad substrate speci昀椀city for benzoquinone substrates [15,23,27,28], and rarely naphthoquinones [23]. Unsurprisingly, most homologues show the greatest catalytic ef-昀椀ciency with the quinone that is most abundant in the host organism [15,27,28]. The sul昀椀de oxidation reaction requires both a donor sulfur compound, which is oxidized in the reaction cycle, and an acceptor compound, which accepts the zero valent sulfur product (Fig. 2). All SQORs preferentially utilize H 2 S as the sulfur donor, however there is signi昀椀cant divergence between the acceptor molecule in prokaryotic and eukaryotic SQORs (Fig. 2). Both will utilize a range of small molecule nucleophiles as the acceptor, obscuring the identi昀椀cation of the native substrate. Prokaryotic SQORs were hypothesized to use hydrogen sul昀椀de, cyanide, sul昀椀te, or thioredoxin at different times [29,30]. This was largely settled by crystal structures solved independently by Brito et al., Marcia et al., and Cherney et al., all of whom observed polysul昀椀de chain formation between the catalytic cysteines, implicating hydrogen sul昀椀de as the physiological acceptor and polysul昀椀de as the product [28,31,32]. This agrees with an observation made years earlier by Griesbeck et al., who observed elemental sulfur deposits on R. capsulatus colonies which were not observed in strains lacking SQOR activity [29]. For eukaryotic SQORs, glutathione and sul昀椀te have both been proposed as the physiological acceptor, and a consensus has not been reached. Sul昀椀te was proposed 昀椀rst on the basis of increased H 2 S toxicity in patients with diminished sul昀椀te production and because thiosulfate is a primary product of H 2 S metabolism, which would be produced directly by SQOR with sul昀椀te as the acceptor [17]. Glutathione was later hypothesized as the acceptor molecule, as it would generate glutathione persul昀椀de, a substrate for thiosulfate sulfurtransferase (rhodanese), feeding into known H 2 S metabolic pathways [33]. Most of the arguments made for each substrate have been based on kinetic simulations using known rate constants and concentrations of substrates, however these have been largely undermined by uncertainty in the cellular concentration of sul-昀椀te [20,33,34]. Because sul昀椀te oxidase is also present in the mitochondria, it is likely that attempts to quantify cellular sul昀椀te concentrations still overestimate the effective concentration experienced by SQOR, and thus that glutathione is the physiological acceptor [20].
Three structures of SQOR have been solved in complex with quinones. Each is expected to be membrane bound; however, they show subtle structural differences, most notably in their quinone binding pockets. Overall, SQORs show signi昀椀cant structural similarity to both FDR and 昀氀avocytochrome C (FCC) enzymes [14,35]. While there are crystal structures of SQOR from more than three organisms, here we focus solely on the available quinone-bound structures.
A. aeolicus crystallizes as a homotrimer with relatively little interface between the monomers. The surface electrostatics are remarkably asymmetric (Fig. 3), with one face showing a net negative charge across the surface and the other showing a net positive charge across the surface. The positively charged side is believed to be membrane-facing due to its ability to associate with the negatively charged head-groups of membrane lipids, and is referred to as the membrane association domain [31]. Based on the charge distribution as well as sulfate ions bound in the crystal structure, it is believed that the enzyme inserts ~12 Å into the membrane [31]. The negatively charged face contains two Rossmann fold domains, as is commonly seen in members of the FDR family. One Rossmann fold contains a covalently bound FAD, which is tethered via the C8 methyl group to the sulfur of Cys124. The re-face of the 昀氀avin faces a cavity ~4 Å deep and ~15 Å wide, which is solvent accessible through a positively charged channel. This channel leads to the second Rossmann fold that would typically bind a second nucleotide, but this is presumably prevented by the relatively small channel between the solvent and 昀氀avin. This cavity contains the catalytic cysteine pair of Cys156 and Cys347, which are proposed to be bridged by additional sulfur atoms derived from H 2 S [31]. The si-face of the 昀氀avin is largely blocked by the side chains of Thr42 and Pro43, precluding substrate binding stacked on that face. These residues are located on a helix that has been proposed to prevent NAD(P)H binding to the second Rossmann fold [28]. As the si-face is inaccessible, the quinone (Q 10 ) instead binds roughly parallel to the 昀氀avin ~3.5 Å distant from the C2 oxygen. The benzoquinone moiety occupies a slightly positively charged cavity, while the isoprenoid tail occupies an ostensibly neutral channel that leads to the membrane association domain (Fig. 3). Presumably, this orientation allows for direct stochastic translocation of the quinone from the membrane into the active site. The quinone moiety is stacked between Ile346 and Phe385 (Fig. 4), providing nonpolar contacts for the relatively hydrophobic ring in the otherwise positively charged cavity.
A. ferrooxidans SQOR crystallizes as a dimer, however the crystallographic dimer is not expected to be the physiologically relevant form the enzyme. It has been proposed that the SQOR dimer seen in the crystal structure is an artifact caused by packing of the hydrophobic surface typically used for membrane association, and the orientation of an alternative native dimer was proposed [32]. Fig. 3 shows only the monomer, as there is no evidence that the oligomerization plays a functional role [32]. Moreover, there is speculation that SQORs do not oligomerize at all in vivo [22]. Membrane association is mediated by a set of amphipathic helices in the C-terminal domain, which insert ~20 Å into the membrane. In addition to the C-terminal domains, A. ferrooxidans SQOR contains two Rossmann folds, similar to A. aeolicus SQOR. One Rossmann fold contains a non-covalent FAD, unlike other homologues [28,31,32]. The second Rossmann fold contains an additional loop which blocks association of an additional nucleotide, as was seen with A. ambivalens SQOR [28,32]. In other FDR enzymes, this domain would bind NAD(P)H. Like A. aeolicus SQOR, the re-face of the FAD faces a cavity that contains the catalytic cysteine pair: Cys160 and Cys356, with Cys160 being located on the loop that blocks nucleotide association to the second Rossmann fold domain [32]. The density seen between the thiol groups of the catalytic cysteine pair suggest the formation of a sulfur chain containing anywhere from 2 to 5 sulfur atoms in the active site [28,31]. Despite the overall similarities to A. aeolicus SQOR, the quinone binding site differs slightly. The quinone remains bound parallel to the 昀氀avin, ~3 Å off the pyrimidine end of the isoalloxazine rather than stacked with it (Fig. 3B). But rather than being stacked between a Phe and Ile, it is stacked between two aromatic side chains, stacking directly over Phe394 and staggered with Phe357 (Fig. 4B). This stacking was found to be essential for SQOR activity [23]. The isoprenoid tail extends through a relatively uncharged channel towards the membrane associated face of the protein, through which it makes a number of hydrophobic contacts (Fig. 3B). Comparison of the unliganded structure to the quinone bound structure reveals that Met418 and Leu415 normally occupy the position of the benzoquinone and isoprenoid tail, respectively, indicating a slight conformational change upon quinone binding [32].
H. sapiens SQOR, while similar to the prokaryotic SQORs, has differences. Like A. ferrooxidans SQOR, it is believed to be a monomer in vivo, and the crystallographic dimerization observed is thought to be an artifact, which was supported by computational results [22]. It contains the characteristic dual Rossmann folds, one of which harbors a non-covalently bound FAD. The C-terminal domain contains amphipathic helices that extend away from the body of the protein and are thought to mediate membrane association [22] (Fig. 3C). The re-face of the FAD is solvent accessible and contains the catalytic disul昀椀de, Cys201 and Cys379, which is bridged by an additional sulfur atom as was Fig. 2. The reactions catalyzed by prokaryotic and eukaryotic SQORs. SQORs from prokaryotes oxidize hydrogen sul昀椀de, progressively lengthening the sulfane chain until it reaches eight atoms and is released as cyclooctasulfane. Eukaryotic SQORs oxidize hydrogen sul昀椀de and transfer the sulfane sulfur to an acceptor molecule, which is likely glutathione or sul昀椀te in vivo, although a variety of small molecule nucleophiles can accept the sulfur atom.
previously proposed [17,22], with Cys201 residing on a loop that blocks access of nucleotides to the second Rossmann fold. The cavity on the re-face of the 昀氀avin is relatively small compared to prokaryotic SQORs, likely attributable to the difference in products [22]. The largest difference between human SQOR and the prokaryotic homologues is found in the quinone binding pocket. The benzoquinone makes no hydrophobic contacts with the protein, and instead hydrogen bonds to Trp345 and Ser378 via its carbonyl oxygens [21,22] (Fig. 4C). The quinone still binds roughly parallel to the 昀氀avin, ~3.5 Å off the pyrimidine end of the isoalloxazine with the isoprenoid tail extending through a hydrophobic channel towards the membrane association face of the protein (Fig. 3C).
The observation that the quinone binds offset from the 昀氀avin raises questions about the mechanism of reduction by the 昀氀avin hydroquinone. A two electron reduction, as is seen for the reduction of the quinone, is generally achieved via hydride transfer in 昀氀avoenzyme oxidoredcutases [36]. This requires a transfer directly from N5 of the FAD to the recipient atom of the oxidant and would not be possible when the quinone is offset. This necessitates two successive single electron transfers to the quinone, a curiosity which has been acknowledged for two SQOR homologues [22,31]. Most work on SQOR has been focused on the reductive half-reaction (sul昀椀de oxidation), and little has been done regarding the quinone reduction. Generally, it is accepted as fast relative to sul昀椀de oxidation and thus not limiting in steady state turnover or single turnover transient state experiments.
The only evidence concerning the quinone reduction step is computational, in which it was suggested that electron transfer happens with consecutive single electron steps in human SQOR [22]. It was predicted that the electrons travel through space from O2 of the FAD to the methoxy carbon of the ubiquinone. With each reduction, a proton is abstracted from either Trp345 or Ser378, the residues shown to form hydrogen bonds with the carbonyl carbons of the quinone [21,22]. This may suggest that transient negative charges are generated on the carbonyl oxygens upon reduction. While little de昀椀nitive information about the oxidative half reaction has been put forward, from a positional standpoint quinone reduction must proceed via successive single electron transfers rather than canonical hydride transfer.
Dihydroorotate dehydrogenase (DHOD) catalyzes the only redox reaction in de novo pyrimidine biosynthesis, oxidizing dihydroorotate to orotate and installing a 5,6-vinylic bond. While most cells can recycle pyrimidines, rapidly dividing cells, such as T-cells, cancer cells and many pathogenic unicellular organisms rely on de novo biosynthesis to provide adequate quantities of pyrimidine bases. This has led to the identi昀椀cation of DHOD as a drug target for arthritis, bacterial infections, and cancer [37][38][39][40][41][42]. DHOD is broadly divided into two classes based on structure, function, and subcellular localization [43]. Class 1 DHODs are soluble, cytosolic proteins that reduce NAD + or fumarate and are further subdivided into class 1A and 1B based on quaternary structure, oxidative substrate, and cofactor content [43]. Class 2 DHODs (DHOD2) exhibit only 20 % homology with Class 1 [44], are associated with the inner mitochondrial membrane, and reduce quinones (Fig. 5) [45]. This section will focus solely on the quinone reducing DHOD2 class of enzymes.
DHOD2s have been studied from a number of organisms, with the most mechanistic detail being elucidated for the human and E. coli homologues. All known DHOD2s have been shown to contain a single FMN cofactor, despite initial assertions that the enzyme contained a catalytic zinc, iron, an iron-sulfur cluster, or no cofactors at all [46][47][48][49]. The speci昀椀city for the quinone substrate is broad and activity has been reported with 1,4-benzoquinone, ubiquinones of varying isoprenoid tail lengths, decylubiquinone, menadione, menaquinone, and plastoquinones [44,48,50], with a direct correlation between increased isoprenoid tail length and increased catalytic ef昀椀ciency [48,50]. It is unclear whether this lack of speci昀椀city serves a purpose physiologically or if it is simply inconsequential and thus cannot be selected against evolutionarily. The reduced quinones are funneled into the electron transport chain, coupling dihydroorotate oxidation to ATP production [48]. Complex III (the entry point for reduced quinones) can utilize both benzoquinones and naphthoquinones [51], so it is conceivable that the broad substrate speci昀椀city of DHOD has evolved to mirror this.
Signi昀椀cant effort has gone into characterizing the reductive half reaction of DHOD2. Dihydroorotate (DHO) binds parallel to the FMN and stacked ~3.5 Å above the si-face [52], which results in a signi昀椀cant red-shifting of the 昀氀avin absorbance [53,54]. This orientation allows for ef昀椀cient hydride transfer from C6 of DHO to N5 of the FMN, resulting in a long wavelength charge transfer absorbance, characteristic of the FMN red "orotate complex, which decays with reoxidation of the enzyme [54]. A proton must also be abstracted from DHO to yield orotate, which is accomplished by a serine located on a disordered loop that is believed to form a closed active site upon orotate binding [54,55]. It was found that class 1 DHODs achieve oxidation and deprotonation in a concerted step [56], however a similar analysis in DHOD2 revealed a stepwise mechanism [57] (Scheme 1). Mutagenesis of residues that may stabilize charge buildup in the active site was performed and revealed a critical function for two asparagines: one in the correct orientation to stabilize the iminium intermediate and one in the correct orientation to stabilize the enolate intermediate [58]. This result does not differentiate between the two pathways, however it does corroborate a stepwise mechanism, as a similar mutation in class 1 DHOD had a much weaker effect [56]. Fig. 5. The reaction of class 2 dihydroorotate dehydrogenase. Class 2 DHOD catalyzes the oxidation of the 5,6-vinylic bond of dihydroorotate and the reduction of a variety of quinone substrates, yielding orotate and the corresponding quinol product.
DHOD2 enzymes have highly similar quaternary structures between homologues, with deviations occurring primarily on the periphery of the enzyme. DHOD2 is a membrane-associated monomeric enzyme that is composed of two domains: a large C-terminal domain and a smaller Nterminal domain [52,59]. The C-terminal domain contains a Rossmann fold that harbors the FMN cofactor and serves as the binding site for orotate, as well as a mobile loop which contains the catalytic base. This loop, which is frequently disordered in crystal structures, moves toward the FMN upon dihydroorotate binding and caps the binding site, ostensibly isolating orotate from solvent access [52]. This observation has led to the identi昀椀cation of a proton relay that moves protons from orotate to the bulk solvent upon deprotonation and is believed to promote the basicity of the catalytic serine [54,57,60,61], which would typically not be suf昀椀ciently basic to deprotonate C5 of orotate. The N-terminal domain is linked to the C-terminal domain via an extended loop and comprises two α-helices that pack close to the FMN binding site. The N-terminal domain is the de昀椀ning characteristic of class 2 DHODs, as it is missing entirely from class 1 DHODs [52]. This domain is believed to facilitate membrane association, and the net positive charge on the surface likely interacts with the negatively charged lipid head groups on membrane surface (Fig. 5). This would orient the putative quinone binding site towards the membrane and allow for stochastic desorption of quinones from the membrane.
The quinone binding site is believed to form between the two helices of the N-terminal domain, which forms a 10 × 20 Å channel leading from the membrane associated surface of the protein to the FMN active site (Fig. 6). This binding site has not been observed crystallographically and has largely been inferred from kinetic observations and crystal structures with bound quinone analogues. Steady-state kinetic experiments suggest that DHOD utilizes a two-site ping-pong mechanism, with separate binding sites for dihydroorotate and the quinone [44,62]. However, transient-state kinetic experiments have provided greater insight showing that orotate release, as observed by decay of the FMNH 2 "orotate charge transfer decay, is too slow to support the turnover number, indicating that the quinone must be able to bind independently and oxidize the orotate-bound, reduced enzyme [54]. This precludes quinone binding stacked with the FMN, as the si-face is capped by orotate, and the re-face is blocked by a loop ~3 Å away from N5. Over 100 structures of DHOD2 with inhibitors have been solved, and these ligands resemble the native quinone substrates to varying degrees. The inhibitors that most closely resemble biological quinones and have been crystallized bound to DHOD2 are HQNO, with the E. coli enzyme, and ferulenol, with the human enzyme (Fig. 7). Both are known inhibitors of quinone utilizing enzymes [63,64] and are expected to occupy the native binding site. Quinone analogues have been shown to competitively inhibit DHOD2 with respect to the quinone substrate, but noncompetitively with respect to dihydroorotate, suggesting that they occupy a separate quinone binding site [65,66]. Noncompetitive inhibition for these inhibitors with respect to the quinone substrate has been reported, however this was demonstrated to be an artifact of the DCIP coupled assay that is commonly used when studying DHOD2 [65]. Discussion of the quinone binding site will focus on the structures with these molecules bound, as they are believed to provide the best approximation of native quinone binding.
The naphthoquinone-like head-groups of HQNO and ferulenol are bound ~4 Å from and roughly perpendicular to the xylene end of the FMN. They occupy largely hydrophobic pockets, making several hydrophobic contacts with nonpolar side chains. The head-groups of the two analogues also have similar hydrogen bonding interactions with their respective proteins. The enol oxygen of HQNO forms a 3 Å long hydrogen bond with His19 of E. coli DHOD2, and the N-oxide oxygen forms a 3 Å long hydrogen bond with Arg102 (Fig. 8A). Similarly, the enol oxygen of ferulenol forms 3 Å long hydrogen bonds with Arg136 and Gln47 of human DHOD2 (Fig. 8B). The histidine that provides a hydrogen bond in E. coli DHOD2 is conserved in the human enzyme and presumably forms a similar hydrogen bond with the native substrate, however ferulenol does not contain a group that can accept a hydrogen bond at this position. The aliphatic tails of each analogue extend through a hydrophobic channel towards the putative membrane association Scheme 1. Stepwise and concerted oxidation of dihydroorotate by DHOD2. surface of the protein (Fig. 6). Both the head-group pocket and the channel leading to the membrane are large, which likely contributes to the observed broad oxidant substrate speci昀椀city, as there are no structural features that tightly constrain binding. This is evidenced by the binding of several inhibitors that are signi昀椀cantly bulkier than biological quinones [44,52,[67][68][69]. Despite this, inhibitors can be designed to be homologue-speci昀椀c based on differences in the contacts made in the hydrophobic channel [68][69][70].
The oxidative half-reaction of class 2 DHOD is, from a mechanistic point of view, poorly understood. It was found that the oxidative halfreaction is ~4-fold faster than the reductive half reaction at pH 8.5 [54]. As most kinetic studies performed with DHOD2 have relied on Fig. 7. Quinone analogues used for structural determination of the quinone binding site. DHOD2
has not yet been co-crystallized with a native quinone substrate. In lieu of this, quinone analogues that are commonly used as inhibitors have been used to identify the likely quinone binding site. E. coli DHOD2 has been crystallized with HQNO bound (PDB 7T5K), and human DHOD2 has been crystallized with ferulenol bound (PDB 6IDJ). Both analogues most closely resemble menaquinone, which has been found to be a substrate for both homologues. steady-state analyses [44,62,65,67], they contain no information about steps that are not rate-limiting and thus provide no insight into the mechanisms of quinone reduction. The putative binding site for quinone substrates precludes hydride transfer and dictates that the quinone must be reduced by consecutive one electron transfers. Hydride transfer can also be ruled out kinetically. The dissociation of orotate from the reduced enzyme occurs too slowly to support the turnover number, however dissociation from the oxidized enzyme is much more rapid, dictating that reoxidation of the 昀氀avin occurs while orotate is still bound. Because orotate caps the si-face of the 昀氀avin, and the re-face is blocked by a loop, it is not kinetically feasible for quinones to bind in a manner that permits hydride transfer. Palfey et al. demonstrated that DHOD2 is capable of participating in single electron transfer chemistry using ferricyanide as an obligate one electron acceptor [54], supporting the hypothesis that the quinone is reduced one electron at a time. They found no detectable buildup of 昀氀avin semiquinones in the oxidative half reaction, indicating that the second electron transfer is appreciably faster than the 昀椀rst, such that the entire process is rate-limited by the 昀椀rst electron transfer [54]. Additional work is needed to verify the quinone binding site in class 2 DHODs, as well as to clarify the mechanism of quinone reduction.
Proline dehydrogenase (ProDH) catalyzes the 昀椀rst step in proline catabolism, converting proline into Δ 1 -pyrroline-5-carboxylate (P5C). P5C is subsequently hydrolyzed to L-glutamate-5-semialdehyde (GSAL), which is oxidized to glutamate via P5C dehydrogenase (P5CDH or GSALDH) (Fig. 9). ProDH activity is present in both prokaryotes and Fig. 8. Putative quinone binding sites of E. coli and human class 2 DHOD. Crystal structures of DHOD2 in complex with menaquinone analogues have been solved for the E. coli and human enzymes. In each panel FMN is green, orotate is burgundy, and menaquinone analogues are blue. A. E. coli DHOD2 in complex with HQNO (PDB 7T5K). B. Human DHOD2 in complex with ferulenol (PDB 6IDJ). (For interpretation of the references to colour in this 昀椀gure legend, the reader is referred to the Web version of this article.) Fig. 9. Reactions catalyzed by PutA, ProDH, and P5CDH. ProDH catalyzes the concomitant oxidation of L-proline and reduction of membrane-associated quinones. The product, Δ 1 -pyrroline-5-carboxylate (P5C), is hydrolyzed nonenzymatically to L-glutamate-5-semialdehyde, the substrate for P5CDH, also known as GSALDH. The multifunctional proline utilization A enzyme combines both functions into a single protein.
eukaryotes. In prokaryotes, ProDH confers the ability to use proline as a primary energy source when it is abundant [71], which is of particular importance for pathogenic bacteria that reside in the upper gastrointestinal tract where the concentration of proline is relatively high [72]. In gram-negative bacteria, ProDH and P5CDH are fused into the multifunctional proline utilization A (PutA) protein which ensures commitment to both oxidation steps, emphasizing its role in energy metabolism [73,74]. In higher organisms ProDH is a monofunctional enzyme that plays a few roles including detoxifying excess proline, generating reactive oxygen species, providing precursors for the GABA biosynthetic pathway, and regulating mitochondrial respiration [75][76][77][78]. These functions have led to the identi昀椀cation of ProDH as a drug target for bacterial infections, cancers, and schizophrenia [79][80][81][82][83].
Most research into ProDH has focused on bacterial forms of the enzyme, which vary signi昀椀cantly between organisms. Broadly, bacterial ProDHs can be categorized as trifunctional PutAs, bifunctional PutAs, or monofunctional ProDHs. Trifunctional PutAs were 昀椀rst reported from E. coli and catalyze the oxidation of proline (ProDH activity) and P5C (P5CDH activity) at different active sites [71,84,85]. The oxidation of proline is coupled to the reduction of membrane-associated quinones and the oxidation of P5C is coupled to the reduction of NAD + , both of which directly link PutA to the electron transport chain. Additionally, trifunctional PutAs serve as transcriptional regulators for the put operon. This operon encodes for PutA as well as PutP, a proline transporter, conferring the ability to uptake and catabolize proline [71]. The transcriptional regulation activity of PutA is controlled via a redox sensing mechanism in which the oxidized state of the enzyme binds tightly to the put operon, preventing transcription [71,86,87]. Upon reduction by proline, PutA dissociates from the DNA and localizes to the membrane, which permits transcription of the operon and facilitates electron transfer from the enzyme into the quinone pool [71]. Thus, the concentration of proline directly regulates expression of the put operon in vivo. Bifunctional PutAs do not act as transcriptional regulators but retain both catalytic activities, while monofunctional ProDHs only catalyze the oxidation of proline. Organisms which utilize a monofunctional ProDH have a separate P5CDH enzyme, allowing for regulation of each activity individually [88].
PutA enzymes have been proposed to utilize substrate channeling, translocating the P5C intermediate from the ProDH active site directly into the P5CDH active site without releasing it into the bulk medium. This was 昀椀rst suggested for S. typhimurium PutA on the basis of kinetic observations [89]. If P5C was released into the bulk solvent, NADH production would be expected to reach maximum velocity after a lag phase, as the concentration of P5C would have to build up gradually. No lag was observed in the steady state, suggesting that P5C is not released. This was corroborated when a 70 Å long curved channel spanning the 45 Å between active sites was identi昀椀ed in a crystal structure of G. sulfurreducens PutA [90]. Transient-state kinetic approaches provided more insight into the mechanism of substrate channeling. A series of experiments, both transient and steady state, were 昀椀t to a comprehensive model to demonstrate that substrate channeling was needed to account for all observations [91]. Curiously, a transient lag phase was observed in the P5CDH reaction that was not discernible in steady state data. The reaction accelerated 38-fold after 15 turnovers, indicative of hysteretic channeling behavior [92]. The authors suggest that this may be a result of the channel populating fully with P5C and note that the channel is expected to accommodate 12-15 P5C molecules, consistent with the number of turnovers necessary for full activation [92]. Similar to PutA, monofunctional ProDH and P5CDH enzymes were found to form transient interactions that allow for ef昀椀cient substrate channeling even in the absence of a covalent tethering, however the two enzymes exist separately in solution and are not observed to oligomerize to any signi昀椀cant extent [88].
Crystal structures have been solved for bifunctional PutAs and monofunctional ProDHs. Only partial structures of trifunctional PutAs have been solved, however the entire ProDH domain is included in this partial structure. The ProDH domains of all characterized PutA and ProDH enzymes have the same (ɑβ) 8 -barrel fold. As such, the quinone binding site is expected to be homologous across the classes of PutA and ProDH, despite differences in oligomerization state and domain arrangement. This is highlighted by the GsPutA and BjPutA homologues, which have 27 % sequence similarity but an RMSD of only 2.0 Å between their aligned structures [90]. This review will focus primarily on the structure of bifunctional G. sulfurreducens PutA as a representative for the ProDH domain structure, as it is the only homologue that has been characterized in complex with a quinone analogue. As GsPutA is a bifunctional PutA, it lacks the signi昀椀cant conformational rearrangement upon 昀氀avin reduction that has been identi昀椀ed in trifunctional PutAs [93][94][95]. Because this conformational change has not been fully characterized, it is unclear how this would affect the quinone binding site, and as such the conclusions drawn here may not apply to all PutA enzymes. The structural differences between the classes of PutA have been reviewed extensively elsewhere [96].
GsPutA forms a homodimer, mediated by a C-terminal oligomerization domain which is reminiscent of aldehyde dehydrogenases [90]. This dimerization places the P5CDH domains adjacent to each other, while the ProDH domains are spatially separated and extend at a roughly 45 ç angle away from the body of the protein (Fig. 10). The two active sites of each protomer are connected by a curved channel that spans ~70 Å and is believed to be the site of hydrolysis for the conversion of P5C to GSA [90,97,98]. The channel is lined with primarily hydrophilic residues and is 昀椀lled with water in the crystal structures, both of which should promote hydrolysis. Molecular dynamics simulations have identi昀椀ed six putative tunnels connecting this cavity to the bulk solvent, which may allow water into the channel to facilitate hydrolysis or allow water out to facilitate diffusion of P5C/GSA through the channel [90,91]. Three of these tunnels are located near the ProDH active site and connect the si-face of the FAD to the bulk solvent, making them of particular interest to substrate acquisition. It is unclear which tunnel proline enters through, however upon proline binding it forms hydrogen bonds with Arg421, which in turn forms hydrogen bonds with Glu149, shifting an ɑ-helix 3 Å towards the 昀氀avin and closing off one of the tunnels, redirecting the hydrolysis cavity towards the adenine of the FAD. This ion pair has been heavily implicated in gating the egress of products into the hydrolysis cavity [90,97,99].
While GsPutA does not undergo major conformational rearrangement upon 昀氀avin reduction like trifunctional PutAs, crystal structures of the dithionite reduced enzyme have revealed minor local rearrangements. The reduced 昀氀avin is bent 23 ç , convex on the si-face, pushing N5 toward the proline binding site [90,98,99]. This results in a steric clash between the reduced 昀氀avin and the P5C product, which may aid in driving the product into the substrate channel [98]. Because the ProDH domain utilizes a ping-pong mechanism [91,100,101], only quinone binding to this form of the enzyme is catalytically relevant. N-propargylglycine (NPPG) was found to covalently modify the 昀氀avin, trapping it in a con昀椀guration that mimics the reduced state while rendering it unreactive to quinones [90,102]. As such, the NPPG inactivated enzyme can be used as a surrogate for the reduced enzyme in crystallographic studies. NPPG inactivated GsPutA was crystallized with menadione bisul昀椀te (MB), a soluble derivative of menaquinone, which is the primary quinone found in Geobacter [103]. MB occupies a site similar to proline, binding on the si-face of the FAD ~3 Å over N5. It forms a 40 ç angle with the isoalloxazine and one of its carbonyl oxygens is pointed towards N10 of the 昀氀avin. It makes contacts with the side chains of three tyrosines (Tyr309, Tyr406, and Tyr418), Leu385, and, most notably, Arg421 (Fig. 11). The sul昀椀te moiety of MB forms two hydrogen bonds with the guanidinium of Arg421. In the NPPG inactivated enzyme, Lys203 is covalently bound to N5 of the FAD and presumably in the native form of the enzyme is available to hydrogen bond with the quinone. This conformation of the quinone would allow the isoprenoid tail to occupy a channel that runs adjacent to the FAD ribityl and exits the protein near the adenine (Fig. 10) [90].
The assignment of this quinone binding site comes with several caveats. The tunnel running near the ribityl of the FAD, which was proposed to accommodate the isoprenoid tail, is the only observable tunnel in the crystal structure, despite MD simulations identifying more [90]. In the crystal structure, MB is in the correct orientation for the isoprenoid tail to extend through this tunnel, however this could be an artifact induced by the sul昀椀te moiety of the analogue. The sul昀椀te forms multiple hydrogen bonds with Arg421, interactions which would be absent with the native substrate, which may bias the quinone towards a non-native orientation in the otherwise spacious active site. Indeed, multiple orientations of the quinone analogue were observed crystallographically [90], indicating that the active site permitted multiple binding modes. Additionally, this orientation suggests that the membrane association region of the enzyme is located on the inner surface of the outstretched ProDH domain. However, surface charge calculations using APBS Electrostatics show that this surface is predominantly negatively charged, making it less likely that this region interacts directly with the membrane (Fig. 10). Two other potential tunnels that were identi昀椀ed face the outer portion of the outstretched ProDH domain and lead to a surface that is largely hydrophobic, providing a more probable surface for membrane association. Given that PutA has not been crystallized with any prenylated quinones, it is purely conjecture at this point as to which tunnel provides access to the active site.
While the current evidence is far from conclusive, it is reasonable to conclude that the quinone binds on the si-face of the FAD. Visual analysis of the structure does not reveal any pockets on either end or on the re- Fig. 11. Putative quinone binding site in GsPutA. GsPutA was crystallized with menadione bisul昀椀te (MB) to identify the quinone binding site (PDB 4NMF). In this binding site, MB makes contacts with three tyrosines and a leucine, as well as two hydrogen bonds between its sul昀椀te moiety and an arginine. These hydrogen bonds are not expected to be physiologically relevant, as biological quinones do not contain a sul昀椀te. Quinones are shown in blue, 昀氀avins are shown in green. The FAD has been modi昀椀ed by reaction with Npropargylglycine. (For interpretation of the references to colour in this 昀椀gure legend, the reader is referred to the Web version of this article.)
face of the isoalloxazine that could accommodate a quinone, necessitating it's binding over the si-face. This provides a clear deviation from the trend of quinones binding on the end of the 昀氀avin. Because PutA exists both cytosolically and membrane-associated, it is possible that it does not require separate binding sites, as it can utilize the same point of ingress for each substrate. PutA displays a contrasting mode of quinone binding that provides important context for understanding the peculiarity of the end-on binding that is typically observed.
Quinones were 昀椀rst identi昀椀ed as substrates for the S. typhimurium homologue of PutA [94], and the speci昀椀city for quinone substrates is broad [91,104]. Minimal work has gone into understanding the mechanism of quinone reduction. The proximity of the quinone to the isoalloxazine may indicate that reduction proceeds via a typical hydride transfer mechanism. However, the 40 ç angle of the quinone away from N5 would be counterproductive to hydride transfer. It is possible that this angled binding is an artifact, as the covalently bound lysine in the NPPG inactivated enzyme results in a steric clash with substrates that would bind stacked over N5. Perhaps native binding occurs with parallel stacking, facilitating hydride transfer. This is corroborated by the lack of semiquinone states of the 昀氀avin observed during the transient-state oxidative half reaction [100], however this could also indicate two single electron transfers in rapid succession. More work is needed to clarify the mechanism of quinone reduction in ProDH enzymes.
Some pathogenic bacteria have evolved to utilize a Na + translocating NADH:quinone oxidoreductase (Na-NQR or NQR) in place of Complex I (H + translocating NADH:quinone oxidoreductase) in the respiratory electron transport chain. This confers the ability to generate a sodium gradient, rather than a proton gradient, for use in respiration and thus permits the organisms to survive in high salt environments, such as blood [105,106]. Moreover, pumping sodium would appear to be a pathogenic adaptation that avoids growth limitations in the frequently proton limiting physiological environment. Despite NQR being functionally homologous to Complex I, the two enzymes have negligible structural similarity [107,108]. Complex I comprises 45 subunits, while NQR comprises 6, all of which are coded for in the nqr operon [109]. NQR is most homologous to ferredoxin:NAD + oxidoreductase (RNF), which also translocates H + coupled to the oxidation of ferredoxin and reduction of NAD + [110]. The low homology to human respiratory enzymes and critical role in pathogen survival has led to NQR being identi昀椀ed as a drug target, most notably for cholera [111]. Despite signi昀椀cant efforts to characterize NQR by several groups, many aspects of the enzyme remain contentious. For the sake of brevity, here we summarize primarily what we believe to be the most well supported assertions relating to NQR, other than those pertaining to quinone binding which are discussed more thoroughly in accord with the topic of this review. More extensive reviews of other facets of NQR have been published elsewhere [111][112][113][114][115][116].
The reductive half reaction of NQR is considerably more complex than is typical for 昀氀avoprotein oxidoreductases. NQR's six subunits, NqrA-F, house six cofactors: FAD, two FMNs, ribo昀氀avin, a Fe 2 S 2 clusters, and an additional Fe containing cofactor that has been reported as both a Fe 2 S 2 cluster and a 4Cys mononuclear iron center [108,[117][118][119]. This is the 昀椀rst and only example of a redox active ribo昀氀avin cofactor in vivo, although 昀氀avodoxins have been shown to be functional with ribo昀氀avin as a cofactor when the native FMN is removed in vitro [120]. EPR results indicate that the ribo昀氀avin puri昀椀es in the neutral semiquinone state [121][122][123], which has signi昀椀cant implications for the mechanism of electron transfer through the protein. Initial attempts to characterize the reductive half reaction did not account for the second Fe containing cofactor, which has since been observed in crystal structures, EPR, and cryo-EM structures [108,117,118]. Moreover, the authors posited that the ribo昀氀avin is the electron donor to the quinone substrate based on the reverse reaction with quinol as the reductant. However, this experiment was undertaken in the presence of millimolar quantities of dithionite [124], which has been demonstrated by the same group to readily reduce NQR [123]. The combination of these factors brings into question the conclusions drawn from this study. Some years later, a more thorough study was carried out in which the reductive half reaction of the NQR complex was observed and the spectral components derived from the global 昀椀t were compared to standard spectra generated for each subunit individually (and thus sets of 1-2 cofactors) to assign the phases [119]. This analysis also accounted for the additional Fe containing cofactor, which was assigned as 4Cys mononuclear iron center but was later shown from cryo-EM structures to be a second Fe 2 S 2 center (referred to as the distal Fe 2 S 2 below) [117]. Global 昀椀tting of this data revealed six phases. In phases one and two, the electrons are transferred from NADH to the FAD and subsequently equilibrate with the Fe 2 S 2 cluster, generating roughly equal amounts of FAD neutral semiquinone and hydroquinone. In phase three, two more electrons are acquired, resulting in formation of the FAD hydroquinone along with reduction of both Fe 2 S 2 clusters. The distal iron-sulfur center is reoxidized to reduce the ribo昀氀avin anionic semiquinone to the corresponding hydroquinone in the fourth phase. In the 昀椀fth phase, two more electrons from NADH establish the anionic semiquinone states of both FMNs, one of which is further reduced to the hydroquinone in the sixth phase [119]. This complex mechanism is summarized in Fig. 12. Notably, the authors do not specify an oxidation state for the distal Fe 2 S 2 center after the 昀椀fth phase, due to the lack of signal corresponding to this transition, but it can be deduced from the oxidation state of the other cofactors and is annotated as such in Fig. 12. The authors also suggest that phases 昀椀ve and six are catalytically irrelevant, as the two-electron reduced ribo昀氀avin that is established in phase four is expected to reduce the quinone [119].
Five of the six NQR subunits contain transmembrane helices, with NqrA being the only fully cytoplasmic subunit. Subunits, NqrB, NqrD, and NqrE provide the majority of the membrane interactions as they contain 10, 6, and 6 transmembrane helices, respectively, while NqrC and NqrF each contain only one. NqrA and NqrF reside in the cytosol while NqrC faces the periplasm (Fig. 13). The cofactor arrangement of NQR is complex: NqrB contains a ribo昀氀avin and an FMN, NqrC contains an FMN, NqrD and NqrE both contribute residues to a Fe 2 S 2 cluster at their interface, and NqrF contains an FAD and 2 Fe 2 S 2 cluster. Notably, NqrF has a ferredoxin:NAD + oxidoreductase-like fold fused to a Fe 2 S 2 ferredoxin [125], supporting a modular role in electron ingress into the enzyme, much as is the case for other oxidoreductases, such as DHOD1B [126,127]. Two channels have been proposed for sodium translocation through NQR, one of which passes directly through NqrB [108] and another which involves a number of acidic residues originating from NqrB, NqrD, and NqrE [128]. Hau et al. have recently published a series of cryo-EM structures of NQR in different conformations that convincingly show that Na + is translocated through NqrB [129]. This same series of cryo-EM structures demonstrates that large-scale conformational changes, which are coupled to redox events, allow NqrC to act as a "redox switch" and move the cofactors into positions that are amenable to electron transfer [129]. It had been previously acknowledged that some of the cofactors, as positioned in the oxidized NQR structure, were too distant to promote electron transfer [130] (Fig. 12).
The quinone binding site in NQR has long eluded identi昀椀cation and has been a primary area of investigation. Early reports placed the quinone binding site in the NqrA subunit on the basis of surface plasmon resonance (SPR) and NMR studies using photoactivatable quinone analogues. However, the SPR assays suggested multiple binding sites and isolated NqrA showed different binding characteristics from the NQR complex, suggesting more complexity [131]. Subsequent studies, using interligand nuclear Overhauser effects in NMR, con昀椀rmed that quinones bind to NqrA and demonstrated that two quinones, or a quinone and an inhibitor, can bind near this site [132]. This provides a possible explanation for the multiple binding sites observed in SPR assays and the noncompetitive inhibition that is typical of NQR inhibitors [131][132][133]. Two more studies demonstrated that photoactivatable inhibitors could interact with both NqrA and NqrB and the interaction was dependent on the inhibitor's structure, leading the authors to conclude that the NqrA site is the catalytic site while inhibitor binding in NqrB prevents the structural rearrangements necessary to close the distance between the ribo昀氀avin and NqrA [134,135].
A contrasting view has the catalytic quinone binding site in the NqrB subunit. A korormicin resistant NQR was puri昀椀ed from V. alginolyticus and a mutation at NqrB-Gly140 was identi昀椀ed, suggesting that this site is catalytically relevant [136]. However, korormicin is a noncompetitive inhibitor towards quinones and inhibition by HQNO, a quinone analogue, was unaffected by this mutation, suggesting that this is not the catalytic quinone binding site. Nonetheless, subsequent studies using mutagenesis found that Gly140 and Gly141 in NqrB are crucial for quinone reduction [133] and a follow up study by the same group elaborated on this signi昀椀cantly. The authors identi昀椀ed two quinone binding sites, one in NqrA and one in NqrB, however they reported that the NqrA was unable to bind quinones in the presence of detergent and concluded that quinone binding to this site is nonspeci昀椀c [137]. Physiologically, NqrA is entirely cytosolic, so the addition of detergent may perturb quinone binding in a manner that is completely irrelevant to the native enzyme. Additionally, the authors observed that mutations of NqrB-Gly140 and Gly141 did not alter the af昀椀nity of NQR for the quinones, but impaired quinone reduction and the conformational changes associated with quinone binding [137]. This observation is consistent with the assignment above that NqrB is important for conformational changes associated with quinone binding but is not the catalytic binding site. However, the authors conclude that these residues control the accessibility of the quinone binding site [137], diametric to the observation that quinones still bind to these variants with ostensibly the same af昀椀nity. The same group later identi昀椀ed NqrB-Phe211 and Phe213 as a potential quinone binding site and showed that mutating these residues abolished quinone binding and reduction [138]. All crystal and cryo-EM structures to date with bound ubiquinones or HQNO show the ligands in a site on NqrB, close to Gly140 and Gly141 (Fig. 13).
The two potential quinone binding sites are shown in Fig. 13. Quinones and inhibitory ligands have only been found in the NqrB binding site, which is shown in the solid box. No quinone has been observed in the NqrA binding site, which is indicated using a dashed box. Even so, there is not a straightforward answer as to which is the catalytic quinone binding site. In the quinone-bound structures, the substrate is bound ~15 Å distant from ribo昀氀avin and separated by a helix, making it unlikely that this conformation is catalytically relevant. Similarly, quinone binding in NqrA would place the quinone too far from any redox cofactors to be directly reduced. It is noteworthy that the putative NqrB binding site is formed by transmembrane helices, indicating that it would be buried in the membrane in vivo. This would provide signi昀椀cantly more opportunity for quinone acquisition; however it could also indicate that the quinones bound in crystal structures are adventitiously binding to a hydrophobic patch that would typically be unavailable. If NqrB has the catalytic quinone binding site, it calls into question why the NqrA subunit is present. NqrA has minimal sequence similarity to any known proteins, but has structural similarities to Nqo1, a component of bacterial respiratory complex I [108], and has been proposed to have evolved from RnfC, a 昀氀avin and FeS cluster containing subunit of RNF [139]. That NqrA does not contain these cofactors may suggest that it is entirely vestigial, and that quinone binding to the large hydrophobic cavity is artifact. However, the preponderance of kinetic, SPR, and NMR evidence points towards the catalytic quinone binding site residing in NqrA. Given that large scale conformational changes must be invoked for either putative binding site, we posit that static structures are particularly prone to non-native binding of substrates and that the NqrB quinone binding site may also be an artifact. More information is needed before the catalytic quinone binding site can be con昀椀dently assigned to either subunit.
Doubts about the quinone binding site have prevented signi昀椀cant probing of the mechanism of quinone reduction. It has been established that electrons exit the protein through the ribo昀氀avin [123,140], which has been shown to purify in the one electron reduced state [121][122][123]. In fact, the ribo昀氀avin has never been observed in the fully oxidized state, which may indicate that it operates by a mechanism in which the ribo昀氀avin hydroquinone donates one electron to the quinone substrate and is back昀椀lled by the other redox cofactors to then donate a second electron. Alternatively, it has been proposed that NQR may generate ubisemiquinones rather than ubiquinols [141,142]. In either case, it appears valid to assume that NQR reduces quinones via single electron transfers given its high propensity for one electron chemistry and the signi昀椀cant structural rearrangement that would have to occur to allow for ring stacking to facilitate canonical hydride transfer.
Cytosolic quinone reductase activity is largely mediated by two enzymes, NAD(P)H:quinone oxidoreductase (NQO1) and NRH:quinone oxidoreductase (NQO2). The early literature record of NAD(P)H: quinone oxidoreductases is fraught with ambiguity due to the wide range of oxidant substrates that these enzymes will utilize, as well as the overlap in catalytic activity between distinct but functionally homologous enzymes. As such, NQO1 has gone by several aliases including DTdiaphorase and vitamin K reductase, which have since been identi昀椀ed as a single enzyme [143]. NQO1 catalyzes the obligate two electron reduction of quinone and quinoid-like molecules using electrons derived from NAD(P)H without the generation of ubisemiquinone intermediates. This prevents detrimental redox cycling that arises from quinone radicals, which can indiscriminately react with cellular components [144], and permits the conjugation and further detoxi昀椀cation to the quinol products [145,146]. As such, it has been suggested that the primary function of NQO1 in vivo is to detoxify dietary quinones [143]. Indeed, suppression of NQO1 activity greatly decreases cell's resistance to quinone toxicity [147] and compounds that are protective against quinone toxicity have been shown to upregulate NQO1 [148]. NQO1 also plays a role in vitamin K cycling, which is critical for post-translational modi昀椀cation of proteins involved in blood coagulation [149], making it the target of anticoagulant drugs such as warfarin [143]. Evidently, the broad substrate speci昀椀city permits NQO1 to serve as a non-speci昀椀c quinone detoxi昀椀er. This has been leveraged therapeutically, as NQO1 has been utilized to activate quinoid-like prodrugs in vivo [150,151].
NQO2 is believed to be an isoform of NQO1 based on their ~50 % sequence similarity [152]. Curiously, NQO2 cannot oxidize NAD(P)H and instead uses mononucleoside derivates of dihydronicotinamide ) [153]. The dihydronicotinamide-benzyl, methyl, and ribosyl derivatives have been used most commonly as substrates as the unsubstituted dihydronicotinamide is not catalytically competent [153]. It is unclear, however, which of these is the physiologically relevant substrate. Viable explanations for the generation of methyldihydronicotinamide and ribosyldihydronicotinamide in vivo have been put forward, but neither molecule is expected to be present in signi昀椀cant quantities under normal cellular conditions [154]. The physiological role of NQO2 is thus less well de昀椀ned than NQO1, largely owing to the lack of a consensus substrate pair for the enzyme. One potential role is in the detoxi昀椀cation of 1, 2-quinones to the corresponding catechol. Oxidized catecholamines, which result from the oxidation/metabolism of several neurotransmitters, are highly electrophilic and require detoxi昀椀cation [155]. NQO2 detoxi昀椀es adrenochrome with near diffusion-limited ef昀椀ciency, whereas NQO1 is completely inactive towards it [155], and mutations in the NQO2 gene result in neurological disease [156][157][158]. As such, it is conceivable that NQO2 has evolved to detoxify the orthoquinones that cannot be reduced by NQO1.
Non-enzymatic roles have been suggested for both NQO1 and NQO2. NQO1 binds p53, p73, and ornithine decarboxylase to protect them from proteolysis. All three of these are tumor suppressors, indicating that NQO1 has an anti-neoplastic role [159]. Small molecule inhibitors of NQO1 enzymatic activity also inhibit its protein binding functions, which may indicate that these two activities are linked [159]. It has been suggested that NQO2 is MT3, a melatonin receptor with distinctly different binding kinetics from the well characterized melatonin g-protein coupled receptors [160]. NQO2 does bind melatonin with kinetics mirroring those reported for MT 3 [161], however it has been shown that the binding af昀椀nity is too low to be physiologically relevant outside of locally high concentrations of melatonin in certain tissues or under conditions of high dietary melatonin [162]. Alternatively, NQO2 may serve as a homeostatic regulator. Reduction of NQO2 activity via small molecule inhibition or gene knockout results in a paradoxical decrease in susceptibility to quinone toxicity, despite NQO2 directly detoxifying quinones [148,163]. This has been attributed to an upregulation of other metabolic enzymes when NQO2 is inhibited [148], suggesting a biological sensor role for NQO2. Resveratrol, 昀氀avones, and other polycyclic aromatic hydrocarbons are strong inhibitors of NQO2, which may explain their observed antioxidant properties [148,152,164].
The structures of NQO1 and NQO2 are remarkably similar given that they have only ~50 % sequence similarity (Fig. 14). NQO comprises a homodimer in which each subunit contains a solvent accessible FAD near the dimer interface. Each subunit comprises two domains: the large catalytic domain and the smaller C-terminal domain. The largest difference in the structures of NQO1 and NQO2 is the 43 C-terminal domain that is present in NQO1 but is truncated in NQO2. This domain interacts with the AMP of NAD(P) in NQO1 and was hypothesized to confer the ability for NQO1 to utilize dinucleotides, whereas NQO2 lacks this domain and can only utilize mononucleosides [154]. To test this, a chimeric NQO2 with the appended C-terminal domain from NQO1 was generated. This chimeric enzyme could not utilize NAD(P)H, suggesting that the C-terminal domain does not serve this function [152]. Additionally, the truncated C-terminal domain of NQO2 contains a metal ion, usually zinc, the physiological role of which is unclear [154]. The area of highest homology between NQO1 and NQO2 spans residues 94-114 with only residue 104 not being conserved, a residue which has been shown to be important for chemotherapeutic prodrug activation [152]. The Gln104 present in NQO2 confers a much higher rate of activation than the Tyr104 in NQO1, and each isoform can be made to have similar activation kinetics to the other via mutagenesis at this position (i.e., Q104Y-NQO2 behaves like NQO1 and vice versa) [152].
The active sites of the two isoforms of NQO are similar, however there are some notable differences. The active site cleft of each is primarily hydrophobic with hydrogen bond donors/acceptors on either side, consistent with the observation that polyaromatic hydrocarbons with polar functional groups on the periphery of the molecule bind most ef昀椀ciently [148,153]. The si-face of the FAD is remarkably solvent accessible and forms the back wall of the spacious active site. Three structural differences between the isoforms have been suggested to confer the different substrate speci昀椀cities. Histidine 161 in NQO1 was hypothesized to stabilize a negative charge on the quinone during reduction, while the homologous position in NQO2 is an asparagine and cannot perform this function [165,166]. To test this, the N161H-NQO2 mutant was generated and no difference in menadione reduction activity was seen, indicating that this residue is not necessary for quinone reduction [166]. This conclusion does, however, presuppose that both enzymes follow the same mechanism of quinone reduction, which is discussed below. Interestingly, the N161H-NQO2 mutant was no longer capable of activating the CB 1954 prodrug, demonstrating that this residue is indeed important for substrate recognition, despite not interacting with naphthoquinone substrates [166]. The second difference that has been identi昀椀ed between the two active sites is the orientation of phenylalanine 106. In NQO1, Phe106 interacts with Phe178, forming a continuous hydrophobic surface ~4 Å above O4' of the FAD. In NQO2, Phe106 is rotated slightly and allows substrates to interact with the backbone of Gly174, which is a crucial interaction between polyaromatic catechols and NQO2 but is not relevant to quinone binding [162,166]. The 昀椀nal difference between the active sites is tyrosine 128 in NQO1, which is located near the active site entrance. It has been proposed to stack with the adenine of NAD(P) and may account for the difference in reductant speci昀椀city [167]. The homologous position in NQO2 is an isoleucine, which could not partake in π-stacking with the adenine and may sterically hinder dinucleotide binding.
Both isoforms of NQO have been crystallized in complex with quinones. NQO1 was crystallized with duroquinone (tetramethyl-1,4-benzoquinone) and NQO2 was crystallized with menadione. As can be seen in Fig. 15, both quinones bind stacked over the si-face of the FAD and occupy only a fraction of the relatively spacious active site, making minimal contacts with the protein. The duroquinone found in NQO1 has hydrogen bonds with His161 and Tyr126' while the menadione found in NQO2 has no hydrogen bonds directly to the protein (Fig. 16). In each case, the active site can accommodate much larger substrates. It is worth noting that the orientation of duroquinone in the NQO1 active site that is shown in Fig. 16 is one of three observed orientations, with the other two orientations placing the quinone further from the FAD and angled away from N5. While these binding modes are likely catalytically irrelevant, they highlight the lack of speci昀椀city of the NQO active site.
The mechanism of quinone reduction in NQO is of signi昀椀cant interest, as the mandatory two electron reduction of quinones is a primary characteristic of the enzyme. Two mechanisms are conceivable to accomplish this. Hydride transfer directly from N5 of the 昀氀avin to an acceptor site on the quinone would generate no intermediate ubisemiquinone, however the same could ostensibly be accomplished by two successive single electron transfers, in which the second transfer is considerably faster than both the 昀椀rst electron transfer and the ubisemiquinone intermediate's dissociation from the active site. Structural data supports a hydride transfer mechanism, as the quinones stack parallel to the 昀氀avin with one of the carbonyl oxygens placed within 3.5 Å of N5 of the FAD (Fig. 16), however this positioning does not preclude single electron transfers.
Tedeschi et al. demonstrated that photoreduction of NQO1 resulted in fractional accumulation of the anionic semiquinone, and that ferricyanide could reoxidize the two-electron reduced enzyme [168]. Taken together, these results demonstrate that NQO1 can stabilize one electron reduced states and may point towards a single electron transfer mechanism. However, the same group also demonstrated that NQO1 reconstituted with deaza昀氀avin, a 昀氀avin analogue that can only participate in two electron chemistry, was still catalytically competent [169], verifying that the binding mode of the quinones in NQO1 is conducive to hydride transfer. It was later demonstrated that NQO1 reduces the arti昀椀cial electron acceptor DCPIP via hydride transfer [170]. The preponderance of evidence suggests that NQO1 utilizes a hydride transfer mechanism to reduce quinones but is not suf昀椀cient to exclude a single electron transfer mechanism. Likewise, it was shown that NQO2 will not reoxidize with single electron acceptors [152], which ostensibly establishes that the enzyme must utilizes a hydride transfer mechanism.
1.6. Alternative NADH:Quinone oxidoreductases (NDH-2) 1. 6
The alternative NADH:quinone oxidoreductase (NDH-2) family of enzymes typically serves as a secondary point of entry for the electron transport chain [171], although some homologues have alternative functions [172]. NDH-2s are non-electrogenic NADH:quinone reductases that are found in all domains of life and generate quinols that can participate in the electron transport chain, replacing complex I entirely in some organisms [173]. Despite the functional similarities to complex I, NDH-2 enzymes are considerably less complex and comprise only a single subunit, utilize a single 昀氀avin cofactor, and are peripherally membrane associated [174]. NDH-2s are a large family of enzymes and have been reviewed extensively elsewhere [175,176]. While members of the family have low sequence homology (<26 %), their folds are similar and their catalytic mechanism is expected to be consistent based on the conservation of key residues around the 昀氀avin [177]. Here we highlight the animal NDH-2 homologue, ferroptosis suppressor protein 1 (FSP1), as a representative of the family. The signi昀椀cant structural homology and putative conserved mechanism across the family suggests that a discussion of any individual homologue should provide a basis for understanding the entire family, even if minor differences do exist.
Early reports of FSP1 used several different names for the protein, with each conveying a supposed function or denoting homology to another proteins: PRG3 (p53-responsive gene 3) [178], AMID (apoptosis-inducing factor-homologous mitochondrion-associated inducer of death) [179], and AIFM2 (apoptosis-inducing factor mitochondria-associated 2) [180]. This nomenclature re昀氀ects the purported role of FSP1 as a pro-apoptotic marker that is upregulated with p53 signaling [178]. It was shown to have NAD(P)H oxidase activity, bind double stranded DNA nonspeci昀椀cally, and induce caspase-independent apoptosis [179]. Some years later, it was renamed FSP1 on the observation that it is an anti-ferroptotic (a form of iron-mediated regulated cell death) NDH-2 enzyme [172]. This is in direct contrast to previous assertion that FSP1 is pro-apoptotic, however initial studies used N-terminally tagged FSP1 to induce apoptosis.
Natively, FSP1 has an N-myristoyl tag that mediates membrane association [172], and the tags may have induced non-native activity via interference with the myristoylation. Nonetheless, the anti-ferroptotic role of FSP1 has become well accepted. FSP1 suppresses ferroptosis by generating quinols in the membrane which intercept radicals to prevent lipid peroxidation and can compensate for the loss of glutathione peroxidase 4 activity [172]. In addition to membrane-associated UQ 10 , FSP1 can reduce cytosolic vitamin K, contributing to the coagulation cascade and undermining the ef昀椀cacy of anti-coagulants that target NQO1 [181]. As such, FSP1 has been identi昀椀ed as a drug target for cancer [182,183] and clotting disorders [184].
Two homologues of FSP1 have been characterized structurally, the human HsFSP1 and the chicken GgFSP1. They share 70 % sequence homology [177] and are extremely similar structurally (Fig. 17). HsFSP1 has been reported to exist as a monomer in solution, while GgFSP1 exists as a dimer [177]. The dimerization is hypothesized to increase membrane association, as each dimer would have two myristoyl tags, however it is conceivable that the dimerization is an artifact resulting from packing of hydrophobic surfaces that would typically interact with the membrane. FSP1 comprises three domains: two nucleotide binding domains and a C-terminal domain [177]. The two nucleotide binding domains make up a glutathione reductase like fold, with one housing the FAD in an extended conformation and the other serving as the NAD(P) binding site. The C-terminal domain, sometimes referred to as the substrate binding domain [185], sits adjacent to the FAD binding domain and is the proposed site of oligomerization in the GgFSP1 dimer. The FAD isoalloxazine sits at the interface of the three subunits, with the re-face oriented towards the NAD(P) binding domain and the pyrimidine end perpendicular to the C-terminal domain. The re-face is solvent accessible through a large cavity leading through the NAD(P) binding domain, allowing NAD(P) association on this face of the isoalloxazine. A smaller channel exists to the pyrimidine end of the isoalloxazine, through the C-terminal domain. This channel sits on the same face of the protein as the N-terminus, which is the site of myristoylation, and presumably would be oriented towards the membrane, allowing for 1DXO). The quinone makes hydrogen bonds with His161 and Tyr126'. Tyr128′ stacks with the adenine of NAD(P) (not shown) and is missing from NQO2. Phe106 blocks access to Gly174, which is important for catechol binding in NQO2. B. NQO2 in complex with menadione (PDB 2QR2). The quinone makes no direct contacts with the protein. Asn161 and Gly174 are important for substrate recognition with bulkier substrates. Phe106 is shifted relative to its position in NQO1, allowing access to Gly174. (For interpretation of the references to colour in this 昀椀gure legend, the reader is referred to the Web version of this article.) quinone desorption from the membrane. This face also has a slight positive charge (Fig. 17), which would be conducive to peripheral membrane association. A third, smaller channel leads through the FAD binding domain to C6 of the FAD, which may explain the observed hydroxylation at this position.
The quinone binding site of GgFSP1 was identi昀椀ed crystallographically (Fig. 18) [177]. The benzoquinone head group binds ~2.5 Å away and roughly perpendicular to the pyrimidine end of the isoalloxazine. Fig. 18. FSP1 quinone binding site. Quinones are shown in blue and 昀氀avins are shown in green. A. GgFSP1 in complex with UQ1 (PDB 7YTL). The quinone has only a single hydrogen bond with N3 of the isoalloxazine. B. HsFSP1 (PDB 8JSC) with a predicted quinone binding site based on homology with GgFSP1. The quinone is predicted to make nearly identical contacts to those in GgFSP1. (For interpretation of the references to colour in this 昀椀gure legend, the reader is referred to the Web version of this article.)
One of the carbonyl oxygens forms a hydrogen bond with N3 of the 昀氀avin, constituting the only hydrogen bond between the substrate and enzyme in the structure. Lys362 is ~4.5 Å away from the other carbonyl oxygen and may contribute a hydrogen bond with alternative substrates or in a different binding pose. Alternatively, it may protonate this oxygen with reduction to the quinol. The benzoquinone lies between Ala294 and Phe359, approximately 4 Å from each, with Tyr295 3.5 Å away at a ~60 ç angle to the quinone ring. This likely serves to select for planar quinone substrates, as bulkier molecules would clash with these side chains. The quinone site is otherwise spacious, putting minimal constraints on substrate binding. The isoprenoid tail, in the binding orientation seen crystallographically, would extend through the spacious cavity towards the putative membrane association face of the protein, making minimal hydrophobic contacts with residues 372-329. Notably, this binding pose does not preclude dimerization, as the channel that the isoprenoid tail would occupy exits the protein at a site distinct form the dimerization interface.
HsFSP1 has not been crystallized in complex with a quinone substrate, however the structural similarity to GgFSP1 permits predictions of the quinone binding site to be made by comparing the two structures. Placing the quinone in the homologous site on HsFSP1 by aligning the structures reveals a similar binding site. The quinone is ~4 Å away from both Ala295 and Phe360, and Tyr296 stacks ~3.5 Å away at a ~45 ç angle to the quinone ring. The most notable difference is that the residue homologous to Lys362 is Thr363. This may indicate that Lys362 in GgFSP1 is inconsequential to the quinone reduction, contrary to previous assertions [177]. The HsFSP1 putative quinone binding site places similar constraints, or lack thereof, on substrate binding, as it is spacious enough to accommodate a range of planar molecules. This is borne out in kinetic data, which show that the enzyme can reduce benzoquinones [186], naphthoquinones [181], and anthraquinones (unpublished data, Moran research group) effectively.
Observational limitations have impeded studies on the mechanism of quinone reduction by HsFSP1. Normal turnover of HsFSP1 is completely rate limited by the release of prebound NADP + , and the rate of quinone reduction is signi昀椀cantly faster than product release [187]. As such, steady-state and single turnover observations return only the rate of NADP + release from the enzyme [187] and contain no information about the quinone reduction step. When HsFSP1 was pre-photoreduced and mixed with stoichiometric concentrations of quinones, the reoxidation of the enzyme was complete in the 1.2 ms deadtime of the stopped-昀氀ow instrument, dictating that quinone reduction occurs at a rate >2000 s -1 [187]. While kinetic details of the process remain unknown, the binding mode clearly dictates two single electron transfers to the quinone. This has been proposed to prevent the back昀氀ow of electrons and inject electrons into the UQ pool despite the relatively high ratio of quinol to quinone in the membrane [187].
Of the six enzymes discussed here, four of them bind quinones in orientations that necessitate successive single electron transfer steps, and two of them bind quinones in orientations that are conducive to a hydride transfer. This demonstrates that 昀氀avoprotein quinone reductases can indeed use either mechanism, indicating that the evolutionary selection of one over the other may be consequential. Two explanations are conceivable, and they are not mutually exclusive. The 昀椀rst is that enzymes use different binding sites for substrates that originate from different cellular environments to avoid the free energy cost of desorbing substrates. Cytosolic reductants are typically obligate two electron donors that require hydride or carbanion chemistry and position near to or stack over N5 of the 昀氀avin isoalloxazine. Quinone reducing enzymes could leverage the ability of 昀氀avins and quinones to do single electron chemistry to allow for separate binding sites. The second explanation is that precluding two electron transfers disfavors the reverse reaction and helps to establish unidirectional transfer of electrons into the quinone pool despite the relatively high concentration of quinol in the membrane.
There is an obvious kinetic limitation to the reaction of molecules that reside in different cellular compartments. Besides NQO, all the enzymes presented here utilize cytosolic reductant substrates and membrane-associated oxidant substrates. Several strategies have been identi昀椀ed to overcome these limitations [188] however we posit that the chemical properties of 昀氀avins and quinones permits a speci昀椀c strategy that does not require either substrate to be entirely removed from its cellular environment. Using the 昀氀avin isoalloxazine as both a redox mediator and partition between the two environments, these enzymes associate with the membrane and use separate binding sites for cytosolic substrates and quinones, ostensibly overcoming the cross-solubility limitations imposed by each molecule. Of the enzymes discussed, two are cytosolic: NQO and ProDH. While NQO exists exclusively cytosolically, ProDH is suggested to exist cytosolically for only the reductive half reaction and translocate to the membrane for the oxidative half reaction [71]. Interestingly, these enzymes are also the only ones predicted to use a hydride transfer mechanism for the quinone. This supports the hypothesis that the quinone binding mode observed in most quinone reductases is a strategy to overcome kinetic limitations imposed by the substrate cellular localizations. NQO has no need for separate binding sites as all of its substrates originate from the same cellular compartment. Similarly, ProDH can utilize a single binding site as it can be reduced in the cytosol prior to membrane association, orienting the same binding site towards the membrane and negating the need for complete desorption of the quinone from the membrane.
An alternative and/or additive explanation is that single electron reduction steps offer a thermodynamic advantage by hindering back-昀氀ow of electrons out of the quinol pool. The two-electron reduction potential of ubiquinone to ubiquinol is +0.045 V, while the single electron reduction potential is -0.163 V (Fig. 19) [189]. The second single electron reduction potential is thus +0.208 V. The forward reaction for two successive one electron reductions is somewhat less favorable given the endergonic 昀椀rst step (in either direction), however it is presumably pulled forward by the highly exergonic second reduction step. This energetic pro昀椀le dictates that accumulation of the ubisemiquinone and 昀氀avin-semiquinone forms is unlikely and this has been borne out by observation in a few instances [189]. Contrastingly, the two-electron reduction is slightly exergonic. This suggests that, when the binding mode permits, the two-electron transfer is more favorable and corroborates that NQO and ProDH operate by hydride transfer mechanisms. However, for enzymes that utilize an end-on binding mode that dictates two single-electron transfers there is an initial energetic price in the forward reaction but a contextual advantage for the reverse reaction. This is a characteristic of the quinone substrate as the 昀氀avin reduction energies for single and two-electron transfers are similar (though admittedly this is a generality as the actual 昀氀avin potentials are in most cases not known and highly in昀氀uenced by the environment within the enzyme). While the equilibria will be the same, positionally dictating single electron-transfers serves as a kinetic barrier to back昀氀ow of the electrons from the membrane quinol pool, as has been suggested as a key catalytic strategy for FSP1 [187].
Besides only acting in a single cellular compartment, NQO also acts on an entirely separate oxidant substrate pool relative to the membrane associated quinone reductases. The reduction potential of the orthoquinone substrates of NQO2 are considerably higher than for paraquinones, such as the +0.6 V reduction potential for the dopamine/ dopamine quinone pair [190]. This makes the reverse reaction extremely unfavorable, even for a hydride transfer mechanism. Additionally, the cytosolic quinone pool is considerably smaller than the membrane associated quinone pool, acquisition speci昀椀c concentration limitations notwithstanding. The physiological concentration of vitamin K is ~6 orders of magnitude lower than that of ubiquinone [191,192] and both are kept ~90 % reduced under normal conditions [193,194]. Together, these correlations suggest that the reverse reaction for cytosolic quinones (producing NAD(P)H from cytosolic quinols) is relatively inconsequential to the redox status of the cell. The obligate two electron reduction also prevents single electron redox cycling, which has been proposed as the primary catalytic strategy of NQO [144]. As such, it was likely more evolutionarily advantageous to commit to a low barrier to reversibility to avoid radical redox cycling.
Other than the orientation relative to the 昀氀avin, there are no notable trends in the quinone binding pockets, even amongst homologues of the same enzyme. The best example of this is SQOR. In the prokaryotic homologues presented here, the quinone primarily interacts with the protein through hydrophobic interactions. However, the eukaryotic homologue is oriented by hydrogen bonds to each of its carbonyl oxygens. Similarly, DHOD2 binds quinones primarily through hydrogen bonds while FSP1 and ProDH show primarily hydrophobic interactions. NQO places ostensibly no constraints on quinone binding, making only minor direct contacts between the enzyme and quinone. This demonstrates why a canonical quinone binding motif has not been identi昀椀ed and has no doubt obscured unambiguous identi昀椀cation of the quinone binding site in enzymes such as NQR. The minimal constraints placed on binding in quinone reductases may be bene昀椀cial, allowing these enzymes to reduce both benzoquinones and naphthoquinones, as has been observed in the majority of the enzymes reported here. It should be noted that the development of inhibitors against these enzymes is complicated signi昀椀cantly by the lack of speci昀椀city in each enzyme, as off target effects would be expected for most quinone analogues.
The peculiar binding modes of quinones in 昀氀avoprotein oxidoreductases evidently serve multiple functions related to the physiological roles of these enzymes. The end-on binding mode that is observed in membrane-associated quinone reductases precludes the facile hydride mechanism of reduction but allows the enzyme to maintain the partition between cytosolic and membrane-embedded substrates. The relatively high redox potential of the substrate semiquinone intermediate, relative to the quinone and quinol, also serves as a kinetic gate, preventing the back昀氀ow of electrons from the quinol pool. Cytosolic quinone reductases operate on a separate quinone pool and have no apparent need for either of these adaptations. Instead, they perform an obligate two-electron reduction to prevent radical redox cycling that would damage cells.