Xanthine oxidoreductase (XOR) catalyzes two hydroxylation steps in the metabolic pathway of purine degradation, i.e., hypoxanthine to xanthine and xanthine to uric acid, utilizing either NAD⁺ or O₂[1–3] (Figure 1). In higher animals, XOR exists as a homodimer of 150 kDa subunits [4]. Each subunit contains one molybdenum center (molybdenum cofactor; Moco), one flavin adenine dinucleotide (FAD) cofactor and two distinct iron sulfur centers ([2Fe-2S] type) [1–3]. The purine hydroxylation reaction occurs at the molybdenum center. Electrons, which are transferred to molybdenum during the hydroxylation reaction, are further transferred to FAD via the two iron sulfur centers [5,6]. Finally, NAD⁺ or oxygen molecule, which is the final electron acceptor, is reduced at the FAD center.

XOR has two forms: xanthine dehydrogenase (XDH), which prefers NAD⁺ as the substrate and xanthine oxidase (XO), which prefers O₂[1]. Historically, XDH and XO have been studied as distinct enzymes. XOR has been isolated only as the XO form from mammalian sources, whereas it has always been purified in the XDH form from other organisms [2]. It is becoming clear, however, that mammalian XORs exist in the XDH form under normal conditions in the cell, but are converted to the XO form during extraction or purification, either irreversibly by proteolysis or reversibly by oxidation of cysteine residues to disulfide bridges. In some particular cases, XDH can be converted to the XO form [2]. The mechanism of conversion from XDH to XO has been thoroughly elucidated in the past decade by means of a range of techniques, including X-ray crystal structure analysis of various mutants, and it has become clear that the protein environment influences the reactivity of the FAD cofactor towards different substrates through substantial conformational changes triggered by modifications located far from the cofactor [5,6].

The enzyme is a target of drugs to treat hyperuricemia, gout or reactive oxygen-related diseases [7,8]. It is distributed in various organs including liver, mammary gland and endothelial cells of vascular vessels [9,10]. The enzyme was proposed to be localized in peroxisomes of rat liver [11], but was found to be present in cytosol [12]. As XOR inhibitors significantly lower uric acid production and concentration in the blood, they can be used to treat gout. Allopurinol, which was introduced by Elion et al.[13], has been on the market for over 40 years [14]. In recent years, however, several companies have developed very effective inhibitors [14–16], of which one example is febuxostat [17]. Clinical trials indicate that febuxostat is superior to allopurinol in lowering uric acid production, although the mechanism of inhibition is different [18,19]. By means of enzymatic, spectroscopic and structural-biological analyses of the inhibition mechanism, it has been shown that these recently developed inhibitors bind tightly to both the oxidized and reduced forms of XOR in a highly structure-specific manner [15], whereas allopurinol, a substrate analogue, binds covalently to the reduced molybdenum atom (MoIV) after having been converted to the hydroxylated product (oxipurinol: alloxanthine) [20], mimicking the reaction intermediate formed during the hydroxylation reaction with xanthine as a substrate [21]. Although oxipurinol binds very tightly to the enzyme, it can be dissociated from the molybdenum (VI) by spontaneous reoxidation due to electron transfer to other centers with a half-time of 300 min at 25 °C [20]. Potent inhibition seems to be essential to lower the uric acid level in blood or tissue, since XOR is a house-keeping enzyme that exists abundantly in various organs [10]. However, it has been suggested that lowering uric acid levels may cause side effects in humans, since uric acid acts as a radical scavenger in the body [22,23]. Further, it is proposed that NO formed by XOR via reduction of NO₂ (with any electron donor) may induce vasodilatation under ischemic conditions [24–26]. On the other hand, XOR has the potential to generate oxygen radical species (H₂O₂ and O₂⁻) after conversion from XDH to XO [1–6]. O₂⁻ would rapidly react with NO to form ONOO⁻[27]. This reaction may serve to eliminate NO, at least in part, but the ONOO⁻ produced is highly toxic [28]. As to the question of potential NO formation by XOR, the activity for NO formation from NO₂ is extremely low, even under anaerobic conditions, although from a chemical point of view it is possible that the water-exchangeable hydroxyl group at OH-Mo(IV) can be replaced by NO₂ to produce NO, since various compounds, such as uric acid (which reacts very slowly to form xanthine), can behave similarly, as discussed by Okamoto [29]. The reported k_cat value of NO formation is 0.17 s⁻¹ at 37 °C with NADH as an electron donor under anaerobic conditions [30], i.e., less than 1% of k_cat for xanthine oxidizing activity (k_cat value 15–20 s⁻¹ at 25 °C) [31,32]. It is questionable whether such a weak activity can have any physiological significance, even under ischemic conditions. The present review focuses mainly on mutational studies of XOR and mutations associated with hereditary dysfunction of XOR in humans, since these are useful for understanding the enzyme reaction mechanism and also as models to estimate the possible side effects of using XOR inhibitors as drugs.

Human diseases associated with genetic dysfunction of XOR are termed xanthinuria, because xanthine is excreted in the urine [33]. Although the enzyme catalyzes two steps of reaction, as described above, so that XOR dysfunction might be expected to be associated with tissue accumulation of hypoxanthine due to inhibition of the first step (conversion of hypoxanthine to xanthine), in fact hypoxanthine is not normally significantly excreted in urine [34,35]. Instead, hypoxanthine is converted to inosine monophosphate (IMP) owing to activation of the salvage pathway (Figure 1) [36]. Patients typically have low levels of uric acid (less than 1 mg/dL) in blood, so XOR-deficient patients are frequently identified based on measurement of uric acid in blood. Various diseases or disorders other than xanthinuria may lead to hypouricemia (Table 1). Renal hypouricemia, which can be caused by decreased re-absorption due to impaired function of urate transporter in the nephrons, is also clinically asymptomatic in most cases.

Xanthinuria is classified into two subtypes, type I and type II (Table 1) [35]. The type I is due to a genetic defect of XOR, whereas the type II is due to a genetic defect in molybdenum cofactor sulfurase [37,38]. Aldehyde oxidase (AO), also a molybdoflavo enzyme, is similar to XOR. A terminal sulfide group is necessary as the third ligand in the active center of XOR and AO for enzymatic activation of these enzymes after biosynthesis of the molybdenum cofactor. Molybdenum cofactor sulfurase catalyzes this final maturation step by generating a protein-bound persulfide, which is the source of the terminal sulfur ligand of the molybdenum cofactor. Thus, lack of sulfurase results in type II xanthinuria. Type I and II xanthinuria are not clinically distinguishable. In order to differentiate them, allopurinol loading test and gene analysis are performed, because a measurement method for molybdenum cofactor sulfurase activity has not yet been established [39,40]. In the allopurinol loading test, oxipurinol is detected in serum and urine of type I xanthinuria patients after administration of allopurinol, as conversion of allopurinol to oxipurinol is catalyzed by XOR and AO, while oxipurinol is not detected in the case of type II xanthinuria. AO has broad substrate specificity, oxidizing different types of aldehydes and heterocyclic rings [41,42]. No clinical symptom or abnormal laboratory examination result due to lack of AO has yet been identified. However, it has recently been reported that AO plays an important role in the metabolism of numerous compounds. Thus, classification of type I and II xanthinuria might be indispensable for optimum medical treatment of patients with xanthinuria in the future.

In higher animals other than primates, xanthinuria is lethal due to kidney damage resulting from xanthine stones in the urinary tract [43–46]. Although primates have lost uricase during evolution and seem to have acquired tolerance to oxipurines, e.g., through downregulation of XOR gene expression, other animals convert uric acid to more soluble allantoin, catalyzed by peroxisomal uricase, and do not seem to have such tolerance [47]. Urolithiasis is sometimes accompanied with xanthinuria due to xanthine deposition, and rarely this may lead to acute renal failure [48–53]. In addition to its role in uric acid production, XOR has bactericidal activity via ROS generation under certain conditions, particularly in mammalian mammary gland [54]. In addition to the NO₂ reduction as described in the previous section, XOR has been proposed to play a role in lactation, though the mechanism of the role in lactation remains unclear [55,56]. XOR has also been suggested to be implicated in hypertension, cardiovascular disorder, and adipogenesis [57,58]. Although the situation is not simple, clinical observations in xanthinuria patients with extremely low serum uric acid levels, who show no symptoms, suggest that administration of XOR inhibitors may not cause severe side effects if the inhibitor has no other effect than inhibition of XOR, except possibly in special cases, such as cancer, pregnant or breast-feeding patients. Many purine analogue cancer drugs, such as mercaptopurine, are known to be catabolized by XOR [59]. The third type of XOR deficiency, type III XOR deficiency, involves the molybdenum cofactor. Molybdenum cofactor deficiency involves triple deficiency of XOR, AO and SO (sulfite oxidase), due to a defect in the synthesis of molybdopterin, which is a precursor of molybdenum cofactor for all three enzymes. Symptoms of molybdenum cofactor deficiency include severe neurological disorder, lens dislocation and dysmorphism, and the outcome is poor [60].

The primary structure of human XOR was first reported by Ichida et al.[61], who isolated cDNA clones encoding human XOR by cross hybridization with rat cDNA, the structure of which was reported by Amaya in 1990 [62]. The XOR gene has 36 exons, and is located in chromosome 2p23.1 [63,64]. The primary structure of human XOR has 90% homology with the rat enzyme over the entire length. Although cloning of human XOR was subsequently reported by several groups [64–66], the sequences were all very similar, except for one reported by Wright et al that was later found to encode AO, not XOR. Although mammalian XORs and AOs from various sources have similar molecular weights and cofactors [2,6,67], their substrate specificities are different. AO exclusively utilizes O₂ as the oxidizing substrate rather than NAD⁺, which is used by dehydrogenases. The specificities of the two enzymes for reducing substrates partially overlap, and each is capable of hydroxylating a distinct subset of a wide range of aldehydes and aromatic heterocycles. Purine bases are good substrates of XORs, but are not good substrates of AOs. The physiological substrates of mammalian AOs are not known, although the involvement of AOs in drug metabolism is well-established [42]. Structure-based sequence comparisons have identified residues in the vicinity of the active site molybdenum center of AO that differ from those in XOR, and these are most probably the determinants of the substrate preferences exhibited by the family members [68]. However, a glutamic acid residue, thought to represent an essential catalytic base, is strictly conserved in AOs and XORs, pointing to a common catalytic mechanism for all family members [69].

The crystal structures of human XOR from natural milk at 3.6 Å resolution (PDB: 2CKJ) and recombinantly produced XDH at 2.6 Å resolution (PDB: 2E1Q) [69] are available. Higher resolution structures of mammalian XORs are available for native bovine XDH and XO, as well as recombinantly produced rat XDH and XO, including various mutants [69,70]. The subunits in the crystal structures of all these mammalian XORs are arranged as identical dimers that display a distinct butterfly shape [4,69,70]. The dimensions of the whole enzyme molecule are about 155 Å × 90 Å × 70 Å (Figure 2). Each monomer is composed of three subdomains. The small N-terminal domain (residues 1 to 165 in the human enzyme) contains both of the iron-sulfur centers (Fe/S I and Fe/S II) and is connected to the second, FAD-containing domain (residues 226 to 532, colored light green in Figure 1) via a long, partially disordered segment consisting of residues 166 to 225. The FAD domain, in turn, is connected to the third, C-terminal domain via another extended segment (residues 533 to 590), which is also partially disordered. The third and largest domain (residues 591 to 1317, colored light blue in Figure 2) binds Moco close to the interface of the Fe/S- and FAD-binding domains, connected with a C-terminal loop (residues 1318–1333, colored blue in Figure 2) [4,69,70].

In order to elucidate the mechanisms of hydroxylation at the molybdenum center, electron transfer within the redox centers and reoxidation of the reduced FAD by the natural substrate, NAD⁺ or molecular oxygen, various chemical modification and mutation studies have been performed during the last two decades. The amino acid residues of human XOR corresponding to those that have so far been found to be crucial for enzyme function, either by chemical modification or by mutagenesis studies with bovine or rat XORs, are summarized in Table 2.

Although inherited XOR deficiency was first reported in 1954 [33], detailed analysis of mutation sites of XOR was first reported in 1997 [37], and subsequently there have been several reports on XOR protein mutations associated with xanthinuria, as summarized in Table 3, including recent work on SNPs not necessarily associated with xanthinuria. The incidence of XOR deficiency, including type II, has been reported to be 1/69,000, but SNP analysis suggested a higher frequency of mutation in XOR, possibly because most mutations not cause dysfunction, being asymptomatic or merely producing a lower level of uric acid in blood.

Any mutation that causes nonsense substitution [92,94–96,98,99] can be expected to cause loss of activity, since the active site of xanthine hydroxylation lies in the C-terminal domain and therefore truncated proteins should be inactive for hydroxylation. Arg881X is the longest peptide among the reported mutants having a stop codon (Table 3), and as the stop codon site is just at the active site region, as described above, it seems very likely that an active site cavity cannot be formed.

The mutation of Arg149Cys at the Fe/S I cluster motif [93] may influence the formation of the cluster, resulting in loss of electron transfer, even if the protein is completely processed and folded. Thr910 is located at a distance of 7.3 Å from Mo=S in the molybdenum center. Mutation of this residue to a bulky methionine or lysine residue seems likely to result in the loss of Moco or its sulfur atom, which is essential for the activity. Alternatively, insertion of the lysine residue may change the electrostatic environment in the active center cavity.

SNP analysis suggests that mutations of XOR may be quite frequent [97]. Although the conditions of activity determination, such as XDH/XO ratio and content of the desulfo-form of each mutant may have varied, it was reported that mutation of some residues not directly involved in the catalysis may result in partial loss of activity, possibly through effects on the protein conformation. It is intriguing to note that mutants Ile703Val and His1221Arg show increased activity due to an increase of V_max. Those residues are located not in the active site cavity, but rather at the surface of the C-terminal Moco domain. As stopped-flow studies with XDH showed that the rate-limiting step of the overall reaction is release of urate, such mutation might increase the rate of release of urate. It has been reported that the k_cat value of bacterial XDH is 10 times higher [87] and the enzyme inhibition pattern is very different from that of mammalian enzyme, i.e., bacterial XDH was not efficiently inhibited by febuxostat, a potent inhibitor of the mammalian enzyme [100]. Molecular dynamic simulation indicated that the bacterial enzyme molecule is much more mobile due to different mobility of surface amino acid residues, suggesting that the release rate of urate may be slower than that of the mammalian enzyme. This may be consistent with the finding that the modification of surface amino acid residues with FDNB caused slower release of urate, as described in the previous section.

As described above, the molybdenum atom of XOR and AO is coordinated by 5 atoms, of which one is a sulfide atom (Mo=S). During hydroxylation, two electrons from the substrate are transferred as hydride to Mo=S to form Mo-SH. The natural preparation is known to contain a significant amount of inactive form in which the sulfide atom (S) is replaced by an oxygen atom (O) [101]. The ratio Mo=O/Mo=S varies from batch to batch. The enzyme can be inactivated spontaneously by loss of sulfide, and the sulfide can also be removed by CN treatment to give SCN [101,102]. Fully active enzyme (Mo=S) can be separated using affinity chromatography [103,104]. The amount of desulfo-form seems to be regulated by the sulfur-donating activity in various organisms, including fly [102] or chicken [105,106], suggesting the existence of a sulfur-donating enzyme, Moco sulfurase. In Drosophila melanogaster, some mutations at maroon-like locus (ma-l) are known to cause inactivation of both XDH and XO [107], and combined deficiency of XOR and AO in humans was reported [108]. In 1995, it was proposed that type II xanthinuria might be due to a defect in sulfur donation, resulting in combined deficiency of XOR and AO [35]. Subsequently the ma-l gene, bovine Moco sulfurase gene and finally human Moco sulfurase gene were cloned and sequenced; all of them are members of a superfamily having a NifS-like domain in the N-terminal followed by a possible Moco-binding domain with a total of 888 amino acids [38,109,110]. Two independent xanthinuria patients were found to having a mutation that converts codon 419 to a nonsense codon [38]. Subsequently, other mutants, Ala156 to Pro [111] and Arg776 to Cys [112], were reported to cause type II xanthinuria. As human Moco sulfurase has not yet been successfully expressed as a soluble protein and its three-dimensional structure is not available, we can only speculate that the mutations cause some conformational change or folding error that affects Moco binding. Further studies can be expected on this interesting protein and on the mechanism of sulfur incorporation, including the question of whether the sulfur atom is incorporated before or after Moco is incorporated into XOR or AO protein.