Xanthine oxidoreductase and cardiovascular disease: molecular mechanisms and pathophysiological implications

Gene structure

The gene encoding human XOR is >60 kb, comprises 36 exons (Xu et al. 1996), and is located on the short arm of chromosome 2 (Ichida et al. 1993; Xu et al. 1994b). The mRNA transcript contains an open reading frame of 3999 bp encoding 1333 amino acids (Ichida et al. 1993; Xu et al. 1994a, 1995). The amino acid sequence is 91% homologous with rat and mouse XOR (Xu et al. 1994a, 1995), with binding sites for XOR cofactors well conserved amongst human, chicken, rat and mouse XOR (Ichida et al. 1993). The Mo–Co binding site is the most conserved region of XOR with 94% homology between human, rat and mouse amino acid sequences (Xu et al. 1994a, 1995).

The gene encoding the related enzyme AO exhibits striking similarity to XOR. The intron–exon organization is almost identical between XOR and AO genes (Terao et al. 1998), and their amino acid sequences are 50% homologous as determined from bovine cDNA (Terao et al. 1997). Their similarities in primary structure extend to their tertiary structures, for they share the same cofactors and redox centre distribution (Fig. 1A), as well as many of the same substrates (Terao et al. 1998). The genes encoding AO and XOR are closely spaced on chromosome 2 (Terao et al. 1998), suggesting they arose from an ancestral gene via tandem duplication. The functional role of AO remains unknown but is likely to be different from XOR as it lacks dehydrogenase activity (Turner et al. 1995).

Several mutations have been identified in the XOR gene that cause classical xanthinuria type I (Ichida et al. 1997; Sakamoto et al. 2001), whereas mutations responsible for xanthinuria type II have been located in the gene for human molybdenum cofactor sulphurase, which is necessary for both XOR and AO activity (Ichida et al. 2001). More than 50% of patients with xanthinuria are asymptomatic, while the remainder present with xanthine calculi, renal failure and myopathy (Simmonds, 1994). In contrast, nearly all patients lacking SO activity due to congenital Mo–Co deficiency suffer from severe neurological disease that is often fatal (Simmonds, 1994). Interestingly, XOR knockout mice are runted and rarely survive up to 6 weeks, suggesting greater importance of this enzyme in lower mammals (Vorbach et al. 2002).

Gene expression and regulation

Basal expression of the XOR gene is mediated by several transcription factors. For example, various sequence motifs have been located upstream of the XOR coding sequence that bind C/EBP, ETS-1, AP-1, AP-2 and TF-IID (Xu et al. 1996, 2000). Additionally, nuclear factor Y has been shown to activate XOR transcription (Martelin et al. 2000). The basal activity of the human XOR promoter is nominal when compared with other mammals (Xu et al. 2000), probably due to putative repressor elements identified in non-coding regions of XOR. For example, the human promoter contains a unique TATA-like element not found in rodents, mutation of which enhances promoter activity (Xu et al. 2000). An E-box has also been located upstream of the human gene that may restrict transcriptional activity (Xu et al. 2000). Promoter suppression involving a constitutively expressed repressor protein or labile protein with RNase activity has been suspected, for inhibition of protein synthesis with cycloheximide results in elevated levels of XOR mRNA (Dupont et al. 1992; Pfeffer et al. 1994; Kurosaki et al. 1996).

Although the basal expression of human XOR is low, a variety of factors up-regulate transcription (Table 1). Many potential cytokine-responsive elements have been identified in the regulatory region upstream from the gene (Xu et al. 1996), along with putative glucocorticoid or hypoxia-responsive elements (Hoidal et al. 1997). In support of their postulated regulatory role in vivo, these entities activate XOR gene transcription in vitro. For example, Pfeffer et al. (1994) demonstrated that bovine renal epithelial cells up-regulate XOR transcription when treated with TNF-α, IFN-γ, IL-1, IL-6, or dexamethasone. In contrast, Dupont et al. (1992) found that only IFN-γ activated transcription in rat pulmonary vascular endothelial cells, suggesting regulation of XOR gene expression is cell-specific. In this regard, Hoidal et al. (1997) showed that endothelial and epithelial cells utilize different regions of the XOR promoter for basal transcription. Furthermore, cell location also influences promoter activity. For example, depending on their glandular location, murine mammary epithelial cells exhibit a differential response to glucocorticoids, which activate XOR expression in alveolar epithelial cells but not in epithelial cells of the lactiferous ducts (Kurosaki et al. 1996). Several studies have also demonstrated additive or synergistic effects using various cytokines and glucocorticoids (Pfeffer et al. 1994; Kurosaki et al. 1996; Hassoun et al. 1998; Page et al. 1998).

In addition to these regulatory factors, oxygen tension also influences XOR gene expression (Lanzillo et al. 1996). Terada et al. (1997) demonstrated that hypoxia activates and hyperoxia inhibits transcription in endothelial cells. Similar results were obtained by Hassoun et al. (1998) when they exposed various rat and bovine cells to hypoxic conditions. These effects on XOR gene expression may be mediated by hypoxia-inducible factor-1. Interestingly, Terada et al. (1997) observed increasing XOR activity shortly after exposure to hypoxia, before there was a measurable change in XOR transcription. Likewise, in hyperoxic conditions, XOR activity declined at a faster rate than could be achieved with cycloheximide. Page et al. (1998) similarly found that increased XOR activity could not be accounted for by enhanced gene expression alone in cytokine-stimulated epithelial cells. These findings implicate oxygen tension and other transcriptional modifiers in both pretranslational and post-translational regulation of XOR expression.

Protein structure

The XOR enzyme is a homodimer composed of catalytically independent subunits with an approximate molecular mass of 150 kDa each (Krenitsky et al. 1986). The interface between the subunits occurs such that the overall complex has a butterfly shape (Fig. 1C). Each subunit is organized into three domains associated with a specific cofactor (Fig. 1B). The N-terminal domain (amino acids 1–165) is composed of two subdomains, each with one Fe₂–S₂ centre coordinated to four cysteine residues. A linker peptide connects it to the intermediate domain (amino acids 226–531), which holds a deep binding pocket for FAD that positions the flavin ring in close proximity to an Fe₂–S₂ centre. Another linker peptide joins the FAD domain with the C-terminal domain (amino acids 590–1332), which is the largest domain and the location of Mo–Co binding (Enroth et al. 2000).

The crystal structure of aldehyde oxidoreductase, a molybdoenzyme from the organism Desulfovibrio gigas, demonstrates penta-coordination of the molybdenum ion with one oxo ligand, one water molecule, one sulphido group, and two dithiolene sulphur atoms from molybdopterin (Romão et al. 1995; Huber et al. 1996). In XOR, the sulphido group is necessary for maintaining enzyme activity, as the desulpho-form is unable to react with substrate (Kisker et al. 1997). Similarities between the crystal structures of aldehyde oxidoreductase and bovine XOR suggest the coordination of the molybdenum ion in aldehyde oxidoreductase may apply to XOR as well (Romão et al. 1995; Huber et al. 1996; Enroth et al. 2000). A more detailed discussion of XOR structure can be found in the review by Harrison (2002).

Conversion from XDH to XO

When the different forms of XOR were first discovered, the purification process converted variable amounts of enzyme to XO, implicating it as the predominant form. However, it has since been demonstrated that most XOR is in the XDH form in vivo (Della Corte et al. 1969; Stirpe & Della Corte, 1969; Waud & Rajagopalan, 1976a). Both forms catalyse reactions involving similar substrates, but they prefer different electron acceptors. Flavin active site probe studies suggested and crystal structure elucidation confirmed that changes at the FAD site dictate electron acceptor preferences (Massey et al. 1989; Saito et al. 1989; Enroth et al. 2000). Site-directed mutagenesis identified a group of amino acids (Phe549, Arg335, Trp336, Arg427) that is disrupted during conversion, resulting in loss of interaction between Phe549 and Trp336. This significantly alters the flavin active site structure, making it inaccessible to NAD⁺. However, the structural change also creates a channel that facilitates the approach of molecular oxygen (Kuwabara et al. 2003).

XOR conversion occurs via two different routes which determine its reversibility. When XDH is treated with proteases like trypsin, chymotrypsin, or pancreatin, it is irreversibly transformed into XO (Della Corte et al. 1969; Stirpe & Della Corte, 1969). Trypsinization of rat enzyme (Fig. 1A) yields three digestion products (20, 40 and 85 kDa) that associate strongly to maintain enzyme activity unless subjected to denaturing conditions (Amaya et al. 1990). Reversible conversion occurs due to conditions that oxidize thiol groups of Cys535 and Cys992 (Fig. 1A) (Nishino & Nishino, 1997), including storage at –20°C (Della Corte et al. 1969), incubation at 37°C (Della Corte & Stirpe, 1972), exposure to sulphydryl agents (Waud & Rajagopalan, 1976b), and exposure to anaerobic conditions (Della Corte et al. 1969). This mechanism of conversion is preventable and reversible by treatment with dithioerythritol or other thiol compounds, unlike proteolytic conversion (Waud & Rajagopalan, 1976a).

Enzymatic properties

Although XOR reacts with many substrates including purines, purine ribonucleosides, and 2′-deoxyribonucleosides (Krenitsky et al. 1974), it is best known as the rate-limiting enzyme of the purine degradation pathway (Fig. 2) (Parks & Granger, 1986). XOR converts hypoxanthine to xanthine and xanthine to UA, the end product of purine catabolism in humans. In lower mammals, urate oxidase further metabolizes UA to allantoin, but this enzyme is inactivated in most primates (Usuda et al. 1988).

Although aspects of the mechanism of XOR reaction with xanthine remain uncertain, proposed reaction schemes share some common features (Xia et al. 1999). The reductive half-reaction occurs at Mo–Co where XOR accepts two electrons from xanthine, reducing Mo(VI) to Mo(IV) (Fig. 3A) (Xia et al. 1999). From the xanthine-C8 position, a hydrogen is transferred to the sulphido ligand of molybdenum, such that Mo = S becomes Mo–SH (Bray et al. 1979). At the same time, nucleophilic attack by a hydroxyl group occurs at C8, thus forming UA. This hydroxyl group is ultimately derived from water (Murray et al. 1966), but it is unknown whether it reacts independently or as a Mo ligand. Several water molecules that may be involved have been identified in the active site of XOR by crystallography (Enroth et al. 2000). Crystal structure studies have also suggested that a basic amino acid from XOR assists in nucleophilic attack by playing a role in proton transfer. A glutamate residue (Glu869) potentially responsible for this was first designated in the crystal structure of D. gigas aldehyde oxidoreductase (Romão et al. 1995), and an analogous residue (Glu1261) was subsequently identified in XOR (Enroth et al. 2000). Of note, the hypoxanthine analogue allopurinol reacts with XOR at Mo–Co to yield alloxanthine (also known as oxypurinol), which binds to XOR via direct coordination to Mo, thereby inhibiting enzyme interaction with substrate (Massey et al. 1970; Truglio et al. 2002).

In contrast to the reductive half-reaction, the oxidative half-reaction takes place at FAD (Fig. 3B) (Hille & Nishino, 1995). Intramolecular electron transfer between Mo–Co and FAD is mediated by the Fe₂–S₂ centres (Hille & Anderson, 1991), which serve as an electron reservoir or ‘sink’ to maintain Mo–Co as Mo(VI) and flavin as FADH₂ for efficient reaction catalysis (Olson et al. 1974a). Electrons are subsequently transferred from FAD to NAD⁺ or O₂ (Olson et al. 1974b; Hille & Massey, 1981). In the reoxidation of fully reduced XO, the first two steps each involve transfer of two electrons to O₂, generating hydrogen peroxide (H₂O₂) (Olson et al. 1974b; Hille & Massey, 1981). XO then transfers its remaining electrons in separate steps, with each electron independently reducing O₂ to produce superoxide (·O₂⁻). Thus, the overall reoxidation of fully reduced XO yields two H₂O₂ and two ·O₂⁻ species (Hille & Massey, 1981). Interestingly, XDH actually generates more ·O₂⁻ per mole O₂ (Saito & Nishino, 1989) due to greater thermodynamic stability of the FAD form that is reactive with oxygen (Hunt et al. 1993). However, despite greater efficiency in ·O₂⁻ production, XDH reacts slowly with oxygen, with a maximal rate of ·O₂⁻ generation that is 25% of XO V_max (Saito & Nishino, 1989). In addition to this slow reaction rate, NAD⁺ exhibits tight, rapid binding to XDH (Harris & Massey, 1997a), so O₂ is a very poor competitor. The mechanism of electron transfer from XDH to NAD⁺ is not as well elucidated as that for oxygen, but evidence suggests that XDH cycles between two and four electron-reduced states. NAD⁺ binds to XDH when it is reduced with four electrons, with subsequent transfer of two electrons to produce NADH (Harris & Massey, 1997a). Hille & Nishino (1995) have reviewed the kinetics and mechanism of reaction of XOR in greater detail.

Since NAD⁺ is the preferred electron acceptor for XDH, generation of ·O₂⁻ by XDH is limited when the supply of NAD⁺ is ample (Sanders et al. 1997). However, once NAD⁺ is converted to NADH, XDH can operate as an NADH oxidase (Harris & Massey, 1997b; Sanders et al. 1997; Zhang et al. 1998a). NADH directly transfers electrons to FAD, and subsequent reduction of oxygen generates ·O₂⁻. This NADH oxidase activity is significant, reaching 40% of the XDH activity with xanthine (Harris & Massey, 1997b). In contrast to XDH, XO exhibits little reactivity with NADH (Waud & Rajagopalan, 1976a). Interestingly, NADH oxidase activity is insensitive to allopurinol binding (Sanders et al. 1997), since Mo–Co is not involved. In contrast, diphenyleneiodonium (DPI), a non-specific FAD site inhibitor, effectively blocks the NADH oxidase activity of XDH (Sanders et al. 1997; Zhang et al. 1998a).

Superoxide radical produced by either XO or XDH can combine with nitric oxide (NO) to form peroxynitrite (ONOO⁻), a potent non-radical oxidant species (Godber et al. 2000a). NO is usually generated by nitric oxide synthase (NOS). However, XOR may also be a contributory source of NO (Millar et al. 1998), especially in hypoxic conditions when NOS cannot generate NO. Li et al. (2003) recently demonstrated that NO generation occurs in hypoxic rat myocardium in the presence of NOS inhibitors but not allopurinol, implicating XOR in the production of NO. XOR has both inorganic nitrate reductase (Li et al. 2003) and nitrite reductase (Zhang et al. 1998b; Godber et al. 2000b; Li et al. 2001) activities at its Mo–Co site (Godber et al. 2000b; Li et al. 2001; Li et al. 2003). NO also may be generated by XOR via the reduction of organic nitrates like isosorbide dinitrate and glyceryl trinitrate (Doel et al. 2001) and organic nitrites like isoamyl nitrite (Doel et al. 2000). Doel et al. (2001) demonstrated that reduction of organic nitrates occurs at the FAD site, in contrast to the reduction of inorganic nitrate at Mo–Co. Harrison (2002) has reviewed the generation of NO by XOR-mediated metabolism of nitrovasodilators at great length.

In humans, there is a wide range of XOR activity, with greater than threefold variation in enzymatic activity amongst individuals (Guercionni et al. 1991). However, when compared with other mammals, the magnitude of enzyme activity in humans is relatively low (Parks & Granger, 1986), and suppressed gene expression cannot solely account for this. For example, although similar yields of XOR can be obtained from human and bovine milk, human XOR activity is variably 1–6% of bovine activity when xanthine is the substrate (Abadeh et al. 1992). When all substrates are considered, the difference between bovine and human XOR activity spans three orders of magnitude (Abadeh et al. 1992). Some of this disparity is due to the presence of demolybdo-XOR in human milk, which can account for 5% (Abadeh et al. 1992) to >95% (Godber et al. 1997) of total inactive XOR, depending on the sample. Other inactive forms of the enzyme, such as desulpho-XOR, may also be found in milk (Abadeh et al. 1992).

Post-translational regulation

Oxygen tension, which affects XOR activity by regulating gene expression, also regulates XOR activity at a post-translational level (Terada et al. 1997). Poss et al. (1996) demonstrated that XOR activity in bovine aortic endothelial cells doubles after exposure to hypoxia without any increase in mRNA expression for 24 h. Terada et al. (1997) observed a similar response to hypoxia in fibroblasts. Interestingly, XOR is inactivated by high oxygen tension, also via post-translational modification (Terada et al. 1988; Panus et al. 1992). Terada et al. (1997) detected a relatively fast decline in XOR activity in association with hyperoxia, more rapid than would be expected if only de novo synthesis was affected. The mechanism by which oxygen tension regulates enzyme activity may involve phosphorylation of XOR, for Kayyali et al. (2001) demonstrated that XOR phosphorylation in hypoxia is associated with enhanced enzyme activity. Interestingly, XOR phosphorylation is partially blocked by p38 kinase inhibitors, and p38 is up-regulated in hypoxia, suggesting a possible role in regulation of XOR activity (Kayyali et al. 2001).

In addition to oxygen tension, NO also may regulate XOR activity. Several studies have established that XOR is progressively inactivated during reduction of nitrates and nitrites (Doel et al. 2000, 2001; Godber et al. 2000b), and inactivation of XOR by exogenous NO has also been demonstrated (Ichimori et al. 1999). Doel et al. (2001) hypothesized that progressive inactivation of XOR may be responsible for the clinical phenomenon of tolerance that occurs in patients with continuous exposure to organic nitrates. Sulphuration of inactivated XOR has been shown to restore activity, indicating the mechanism of inactivation involves conversion to desulpho-XOR (Ichimori et al. 1999; Doel et al. 2000, 2001; Godber et al. 2000b). In contrast to these results, other investigators have found that NO does not affect XOR activity but peroxynitrite does, suggesting that observations of XOR inhibition by NO are artifacts of peroxynitrite production and subsequent XOR inactivation (Houston et al. 1998; Lee et al. 2000).

Other products and byproducts of XOR reactions have also been implicated as negative regulators of XOR activity, including ·O₂⁻ (Terada et al. 1988), H₂O₂ (Terada et al. 1991a) and hydroxyl radical (·OH) (Terada et al. 1991a). Such product-mediated regulation may play a role in XOR inactivation in hyperoxia.

Tissue distribution and subcellular localization

The determination of XOR tissue distribution has been complicated by inconsistent results obtained using different methods of XOR detection. Antibody-dependent methods like immunohistochemistry recognize active and inactive forms of the enzyme in intact tissues, whereas biochemical techniques rely on detection of enzyme activity in tissue homogenates. This distinction becomes especially important in human tissue, since a large proportion of human XOR is inactive (Abadeh et al. 1992).

In mammals, the highest levels of XOR activity are found in the liver and small intestine (Parks & Granger, 1986). XOR has also been identified in other mammalian tissues, including bovine heart (Jarasch et al. 1986; de Jong et al. 1990) and rat heart (Muxfeldt & Schaper, 1987; de Jong et al. 1990). The abundance of XOR is relatively low in human tissue, but it has been consistently identified in liver, small intestine and mammary gland (Linder et al. 1999). Furthermore, several (Muxfeldt & Schaper, 1987; Abadeh et al. 1993; Vickers et al. 1998; Cappola et al. 2001) but not all (Eddy et al. 1987; Linder et al. 1999) studies have identified XOR protein and/or activity in the human heart. Importantly, Jarasch et al. (1981) consistently located XOR in bovine capillary endothelial cells using immunohistochemistry, and similar experiments established the presence of XOR in human endothelial cells (Jarasch et al. 1981; Bruder et al. 1984). This same group of investigators successfully detected bovine and human XOR activity in various tissue samples using an ultrasensitive radioimmunoassay (Bruder et al. 1983; Jarasch et al. 1986). However, the XOR activity in these tissue fractions was 10- to 1000-fold lower than activity in fractions from liver or lactating mammary gland, suggesting XOR activity may have been below the limit of detection of less sensitive methods used in other studies (Bruder et al. 1983).

Subcellular localization methods have demonstrated the presence of XOR both in the cytoplasm (Rouquette et al. 1998; Frederiks & Vreeling-Sindelarova, 2002) and on cell membranes (Rouquette et al. 1998), with cell surface binding likely mediated by glycosaminoglycans (Adachi et al. 1993; Radi et al. 1997; Rouquette et al. 1998). XOR affinity for glycosaminoglycans, especially heparin, has been implicated by in vitro and in vivo studies. For example, heparin is very effective at displacing XOR in affinity chromatography (Radi et al. 1997), and the addition of heparin or heparinase to endothelial cell cultures increases free XOR in the culture media (Adachi et al. 1993). Similarly, injection of human subjects with heparin rapidly results in a two- to threefold increase in plasma XOR (Adachi et al. 1993). In the past XOR has been reported to be solely cytosolic (Ichikawa et al. 1992), but recently it has been identified by electron microscopy in the peroxisomes of hepatocytes and in various organelles of Kupffer and sinusoidal cells, including rough endoplasmic reticulum, lysosomes, and endocytic vesicles (Frederiks & Vreeling-Sindelarova, 2002). Thus, the subcellular localization of XOR remains controversial and incompletely described.

Physiological functions of XOR

For many decades, the sole purpose of XOR was presumed to be purine catabolism (Fig. 2), but growing evidence suggests a much broader biological role for this enzyme. For example, XOR has antimicrobial properties, as it inhibits the growth of bacteria in vitro in an NO-dependent manner (Hancock et al. 2002). There is also evidence that XOR plays an antimicrobial role in vivo, for infants who receive breast milk rich in XOR are less likely to develop gastroenteritis than those who are fed formula (Stevens et al. 2000). Furthermore, in mice with chronic granulomatous disease, allopurinol decreases clearance of pathogens and reduces killing efficiency in vivo, suggesting XOR may contribute to host defence against oxidant-sensitive organisms (Segal et al. 2000). Although XOR clearly participates in antimicrobial defense, patients with xanthinuria are not immunocompromised, indicating the role of XOR in host defense is non-essential (see review by Harrison, 2002).

UA production by XOR may itself have broader biological consequences beyond purine degradation, for it possesses antioxidant properties (Becker, 1993). Importantly, humans have higher UA concentrations compared with other mammals, as urate oxidase is inactivated in primates (Usuda et al. 1988). Watanabe et al. (2002) hypothesized that this provided a survival advantage for humans because hyperuricaemia maintains blood pressure in the face of low dietary salt. Furthermore, Ames et al. (1981) have speculated that UA contributes to increased life span in humans by providing protection against oxidative stress-provoked ageing and cancer.

A role for XOR in other physiological processes including oxidant-mediated signal transduction has been suggested (see review by Meneshian & Bulkley, 2002).

Pathophysiological role of XOR

XOR-generated ROS are implicated in both tissue structural damage and cell signalling interference. Classically, ROS can cause lipid peroxidation, resulting in disruption of membrane architecture and lysosomal enzyme release (Weiss, 1986), and DNA and amino acid oxidation, causing genetic mutations and enzyme dysfunction or proteolysis (Kehrer, 2000). In regard to XOR, oxidative injury is often achieved via the byproducts of ·O₂⁻ and H₂O₂ generation. For example, in the Haber–Weiss reaction, iron(III) reacts with ·O₂⁻ to yield iron(II) and molecular oxygen; iron(II) then reacts with H₂O₂ to generate highly reactive ·OH (Kehrer, 2000). Hydroxyl radicals have a half-life of 1 ns and thus indiscriminately oxidize their closest target (Weiss, 1986; Kehrer, 2000). In another example of byproduct-mediated damage, peroxynitrite, which is generated from ·O₂⁻ and NO (Houston et al. 1998), is a powerful oxidant involved in nitration of tyrosine residues (Sawa et al. 2000). Dysfunction of proteins due to nitration has been implicated in the pathophysiology of several cardiovascular diseases, including autoimmune myocarditis, hypertension, and heart failure (Turko & Murad, 2002), although this remains controversial.

In 1981 Granger and colleagues hypothesized that XOR-generated ROS cause ischaemic bowel injury due to ATP catabolism during hypoxia and increased electron acceptor availability on reperfusion (Fig. 4). In their study, ischaemic damage only occurred on reperfusion and was attenuated by superoxide dismutase (SOD; Granger et al. 1981). Since this concept of ischaemia–reperfusion (I-R) injury was introduced (Granger et al. 1981; McCord, 1985), several lines of reasoning have supported XOR involvement. First, XOR products increase after I-R, as demonstrated in human aortic endothelial cells using electron paramagnetic resonance (EPR) spectroscopy (Zweier et al. 1994). EPR studies have also identified bursts of ·O₂⁻, ·OH, and ONOO⁻ from reperfused rabbit and rat hearts (Zweier, 1988; Wang & Zweier, 1996). Second, XOR products are associated with cardiac tissue damage, further supporting XOR's role in I-R injury. For example, reperfusion of ischaemic rat hearts results in elevated H₂O₂ levels and ventricular dysfunction, which can be partially attenuated with radical scavengers or allopurinol (Brown et al. 1988). Similarly, ischaemic rabbit hearts pretreated with allopurinol experience less ventricular dysfunction on reperfusion compared with untreated hearts (Terada et al. 1991b).

In regard to peroxynitrite-mediated damage, Ferdinandy and colleagues showed that myocardial contractile function declines with increasing XOR activity and ONOO⁻ generation (Ferdinandy et al. 1999). Furthermore, myocardial reperfusion damage can be attenuated with SOD or NOS inhibitors, which prevent ONOO⁻ formation (Zweier et al. 2001). Peroxynitrite also causes myocardial injury via activation of matrix metalloproteinases (MMP) (Wang et al. 2002a) like MMP-2, which degrades troponin I (Wang et al. 2002b). Activated MMPs are associated with depressed functional recovery of ischaemic rat hearts after reperfusion (Cheung et al. 2000). Other MMPs, including type I collagenase, are also activated by oxidative stress, suggesting a potential role in myocardial remodelling (Siwik et al. 2001).

In addition to tissue damage, XOR-generated ROS can also exert cardiotoxicity via interference with cell signalling (Chiamvimonvat et al. 1995; Thomas et al. 1995). For instance, ROS modify thiol-containing proteins involved in calcium cycling, such as the ryanodine receptor (Campbell et al. 1996) and Ca²⁺-ATPase (Xu et al. 1998), which may result in diminished myocardial performance. Pérez et al. (1998) demonstrated XOR may be responsible for such oxidant-induced effects, showing that allopurinol and oxypurinol improve force generated by stunned cardiac myofilaments while at the same time reducing systolic Ca²⁺ transients, thereby enhancing myofilament Ca²⁺ sensitivity. Moreover, subsequent studies using rat models of heart failure have suggested that changes in calcium cycling due to XOR inhibition may be an adapative response to the depression of myofilament force-generating capacity (Pérez et al. 1999).

In corroboration with this evidence supporting XOR involvement in I-R injury, XOR activity is increased after I-R (Brown et al. 1988; Ferdinandy et al. 1999). Weinbroum et al. (1995) found that XOR activity in postischaemic hepatic effluent and lung lavage fluid increases over 1000-fold and 100-fold, respectively. Elevated XOR activity also occurs on reperfusion of human aortic endothelial cells, with an eightfold elevation in UA production (Zweier et al. 1994). Elevations in circulating XOR after I–R (Weinbroum et al. 1995) can lead to tissue damage at distant sites that are low in endogenous XOR activity (Nielsen et al. 1994). One reason that XOR activity increases after I-R is that substrate availability increases significantly during ischaemia because of ATP degradation. High performance liquid chromatography (HPLC) of rat heart extracts demonstrated that ATP is most abundant under normoxic conditions. However, after 30 min of ischaemia, AMP predominates and hypoxanthine and xanthine also appear. After 5 min of reperfusion, hypoxanthine/xanthine levels are almost undetectable, whereas ATP levels have nearly recovered (Xia & Zweier, 1995). This increase in substrate availability also occurs in vivo, for urinary hypoxanthine/xanthine levels double in acute coronary syndrome patients (Turgan et al. 1999) and serum xanthine increases in patients with acute myocardial infarction (Kock et al. 1994).

Despite the supporting evidence, the role of XOR in I-R injury remains controversial. First, as discussed above, some studies of tissue distribution have found that XOR is absent from major tissues known to be adversely affected by I-R, including cardiac and brain tissue. However, many of these studies utilized functional assays that were not sensitive enough to detect the small proportion of XOR that is active in human tissues. Moreover, as described earlier, XOR has been localized to the endothelial cells of all tissues using immunohistochemistry. In addition to this, circulating XOR has been demonstrated in human, rat and rabbit serum (Nielsen et al. 1994; Weinbroum et al. 1995; Mathru et al. 1996). Elevation of circulating XOR occurs after I-R, so tissues with low endogenous XOR activity may be damaged by circulating and/or endothelium-bound XOR in the vasculature (Nielsen et al. 1994). Another argument refuting the role of XOR in I-R injury calls attention to the predominance of the XDH form in vivo, assuming that conversion from XDH to XO is essential for XOR to produce sufficient ROS to cause I-R damage. However, the time required for XOR conversion has been reported with inconsistent results varying from seconds (McCord, 1985) to hours (Engerson et al. 1987). In support of conversion of XOR in I-R injury, several studies have demonstrated an increase in the percentage of XO relative to total XOR activity during ischaemia (Chambers et al. 1985; Parks et al. 1988), but this may be due to inhibition of XDH by accumulating NADH rather than conversion from XDH to XO. Even the mechanism of conversion is controversial, with studies implicating both reversible (McKelvey et al. 1988; Brass et al. 1991) and proteolytic conversion (McCord, 1985; Engerson et al. 1987).

The significance of XOR involvement in I-R injury has also been debated because of many studies suggesting neutrophils may be the primary source of ROS. Both the depletion of neutrophils (Romson et al. 1983; Hernandez et al. 1987) and prevention of neutrophil adherence to endothelial cells with neutralizing antibody to CD18 (Hernandez et al. 1987; Duilio et al. 2001) attenuate the injury associated with tissue I-R. However, tissue damage after reperfusion has also been demonstrated in experimental systems devoid of neutrophils (Ratych et al. 1987; Brown et al. 1988). A complimentary role is likely, with ROS generated from both XOR and neutrophils causing I-R injury. This is supported by evidence that the pretreatment of feline intestine with allopurinol attenuates neutrophil infiltration after reperfusion (Grisham et al. 1986), and allopurinol and tungsten treatment inhibit neutrophil adhesion to endothelial cells (Terada et al. 1992). Interestingly, neutrophil chemotaxis may be mediated by XOR products, for tissue injection of superoxide prompts infiltration of neutrophils without additional signs of inflammation, and the infiltration is inhibited by SOD (McCord & Roy, 1982).

In response to some of these concerns, Harrison (1997) expanded the I-R hypothesis to include the NADH oxidase activity of XOR, suggesting it too may contribute to I-R oxidative stress. XDH actually exhibits preference for oxidation of NADH rather than xanthine when physiological pH is decreased to 6.5, as may occur locally in tissue ischaemia (Harrison, 1997). Furthermore, NADH levels increase two- to sixfold in guinea-pig and rat hearts during ischaemia (Williamson, 1966; Varadarajan et al. 2001), thereby increasing substrate availability for NADH oxidase activity. Interestingly, the NADH oxidase activity of XDH has a maximal rate of ·O₂⁻ production fourfold greater than XO, and it can still maintain rates of ·O₂⁻ production comparable to XO in the presence of 20 μm NAD⁺ (Sanders et al. 1997), suggesting conversion from XDH to XO is not necessary for superoxide-mediated I-R injury. Harrison's hypothesis is especially interesting because NADH oxidase activity is maintained in demolybdo- and desulpho-XOR, so traditional XOR assays may not detect this activity. Moreover, it provides a rationale consistent with the apparently low activity of XOR in human tissues. Importantly, NADH oxidase activity is not inhibited by allopurinol, since it is independent of the Mo–Co site (Sanders et al. 1997; Zhang et al. 1998a). Therefore, in many studies where allopurinol was unable to completely attenuate injury due to ·O₂⁻ (Brown et al. 1988; Thompson-Gorman & Zweier, 1990; Terada et al. 1992), NADH oxidase activity may have been responsible. The NADH oxidase activity of XOR is effectively inhibited by DPI (Sanders et al. 1997), which also inhibits neutrophil NADPH oxidase, a fact not considered in some investigations of the relative contributions of XOR and neutrophils to I-R injury. Recently, Zima et al. (2003) demonstrated that NADH inhibits cardiac ryanodine receptor function, and it is plausible to speculate that the NADH oxidase activity of XOR may be involved. As with Granger's original hypothesis, Harrison's proposal does not exclude other means of ROS generation and I-R injury, and it may be that superoxide production by both XO and XDH-NADH oxidase contribute to I-R injury in vivo.

XOR in cardiovascular disease

The success of allopurinol in attenuation of myocardial I-R in animal studies prompted clinical trials to determine if it might clinically benefit patients. In aortocoronary bypass surgery, allopurinol administration to patients attenuated histological I-R damage to the myocardium and resulted in a threefold reduction in the need for inotropic support postoperatively (Gimpel et al. 1995). Similar benefits were demonstrated in patients with acute myocardial infarction undergoing primary coronary angioplasty; those who received allopurinol before reperfusion had significantly better left ventricular function both immediately after the procedure and 6 months later (Guan et al. 2003).

Oxidative stress contributes significantly to endothelial dysfunction in cardiovascular disease, as superoxide radicals readily inactivate endothelial NO, thereby impairing vasorelaxation. In fact, NO and .O₂⁻ react at a rate threefold greater than the rate at which SOD can eliminate .O₂⁻ (Cai & Harrison, 2000), and thus oxidative stress may overpower antioxidant protection in many disease states, resulting in endothelial dysfunction. Houston et al. (1999) established that XOR contributes to impairment of vasorelaxation by showing that endothelium-bound XOR inhibits NO-dependent cGMP production in smooth muscle cells. Furthermore, many clinical studies have established that XOR inhibition attenuates endothelial dysfunction. For example, allopurinol increases NO-stimulated blood flow in smokers (Guthikonda et al. 2003) and diabetics (Butler et al. 2000), reversing endothelial dysfunction. XOR may also contribute to endothelial dysfunction in hypercholesterolaemia. In cholesterol-fed rabbits ·O₂⁻ production is threefold greater in diseased vessels than normal vessels (Ohara et al. 1993), and oxypurinol improves endothelial function in the diseased vessels (Ohara et al. 1993; White et al. 1996). Further studies in rabbits have indicated ·O₂⁻ production can be attenuated by injection of heparin or heparin-binding SOD, since heparin can out-compete XOR for binding to endothelial cell surface glycosaminoglycans. Interestingly, heparin administration is also associated with improved endothelial function (White et al. 1996). The role of XOR in endothelial dysfunction in diseased coronary arteries has been directly demonstrated by measurement of endothelium-bound XOR activity, which is elevated >200% in diseased vessels and exhibits a negative relationship with endothelium-mediated vasodilatation (Landmesser et al. 2002). Furthermore, cholesterol and XOR have been colocalized in atherosclerotic plaques from carotid endarterectomy specimens. Local concentrations of UA are elevated five- to sixfold in such plaques, indicating up-regulation of XOR activity (Patetsios et al. 2001).

XOR has also been implicated in the pathogenesis of hypertension. For example, tungsten-rich diets, which inactivate XOR, successfully lower blood pressure in Dahl salt-sensitive hypertensive rats but have no effect on Dahl salt-resistant rats, indicating a potential role for XOR-generated ROS in salt-induced hypertension (Swei et al. 1999). Furthermore, treatment with tungsten (Suzuki et al. 1998), oxypurinol (Nakazono et al. 2003), or heparin-binding SOD (Nakazono et al. 2003) lowers blood pressure in spontaneously hypertensive rats (SHR). Laakso et al. (1998) measured renal XOR activity in both of these rat models and demonstrated that increased sodium intake elevates activity in a dose-dependent manner in Dahl salt-sensitive rats. XOR activity in SHR kidneys also increases in response to increased salt intake, but in contrast, renal XOR activity is already elevated at baseline in SHR when compared with normotensive rats (Laakso et al. 1998). In contrast to other studies (Nakazono et al. 2003), Laakso et al. (1998) found that XOR inhibition did not affect systolic blood pressure in SHR, suggesting induction of XOR renal activity may occur secondary to increased salt intake or salt-induced hypertension.

There is evidence that UA levels correlate with blood pressure (Mazzali et al. 2001), independent of XOR activity. When mild hyperuricaemia is induced in rats with oxonic acid (urate oxidase inhibitor), hypertension develops that is preventable by treatment with allopurinol or a uricosuric agent (Mazzali et al. 2001). Once elevated, blood pressure can be decreased by allopurinol, withdrawal of oxonic acid, or treatment with enalapril or l-arginine. Oxonic acid-treated rats also develop renal injury attenuated by enalapril or l-arginine (Mazzali et al. 2001). In rats fed low-salt diets, oxonic acid results in systemic and glomerular hypertension preventable with allopurinol (Sánchez-Lozada et al. 2002). Renal biopsies from oxonic acid-treated rats show afferent arteriolar thickening that correlates with blood pressure and uric acid levels and is preventable with allopurinol (Sánchez-Lozada et al. 2002). The importance of UA is further supported by the finding that elevated serum UA is associated with increased cardiovascular risk in hypertensive patients, regardless of whether their hypertension is treated (Alderman et al. 1999). Recently, UA has also been shown to have prognostic value in patients with heart failure (Anker et al. 2003).

Heart failure (HF) is often the ultimate result of pathogenetic mechanisms that cause endothelial dysfunction and hypertension. Thus, studies of XOR in these disease states have prompted investigation of its role in HF. Patients with HF have significantly reduced endothelium-associated SOD activity when compared with controls, whereas their endothelium-bound XOR activity is enhanced >200% (Landmesser et al. 2002). Randomized, placebo-controlled trials of allopurinol have demonstrated attenuation of endothelial dysfunction in HF patients (Doehner et al. 2002; Farquharson et al. 2002), with improved endothelial-dependent vasodilatation and enhanced peak blood flow (Doehner et al. 2002).

In HF patients, XOR also adversely affects myocardial function, specifically myocardial energetics. In both experimental animals and humans, allopurinol is effective in attenuating the mechanoenergetic uncoupling that occurs in HF. For example, in dogs with pacing-induced HF, allopurinol decreases myocardial oxygen consumption and increases myocardial contractility (Ekelund et al. 1999). In addition to improving left ventricular function at rest in HF dogs, allopurinol also enhances contractile performance of the myocardium in response to exercise- and dobutamine-induced β-adrenergic stimulation (Ukai et al. 2001). We have demonstrated that high dose ascorbate mimicks these effects of allopurinol. Interestingly, NOS inhibition prevents the benefits of allopurinol and ascorbate in HF dogs, suggesting XOR-generated ·O₂⁻ may interfere with NO regulation of myocardial energetics. NOS inhibition also decreases myocardial efficiency in healthy dogs, further substantiating the role of XOR and NO cross-talk in HF (Saavedra et al. 2002). Normal interaction between XOR and NO may be disrupted in HF because XOR abundance in failing myocardium is elevated (Fig. 5) (Ekelund et al. 1999; de Jong et al. 2000; Cappola et al. 2001). Recently Kögler et al. (2003) found that rat myofilament twitch tension is enhanced by oxypurinol. However, the effect is much more pronounced in failing myocardium, where total XOR activity is elevated relative to normal myocardium, with a fractional increase in twitch tension that is threefold higher in failing myocardium (Kögler et al. 2003). Thus, the inotropic consequences of oxypurinol are in effect relatively selective for heart failure, where XOR activity is elevated. In regard to the potential therapeutic use of oxypurinol in HF patients, oxypurinol does not affect resting tension or intracellular Ca²⁺ transients; thus diastolic myocardial function is not impaired (Kögler et al. 2003). Most importantly, in corroboration with these animal studies, we have demonstrated that allopurinol infusion into the coronary circulation of HF patients improves myocardial efficiency by diminishing oxygen consumption (Cappola et al. 2001), suggesting XOR inhibition is a potential strategy for the treatment of human heart failure.

Conclusion

Over a century after its discovery, our knowledge of the biological activity of XOR continues to grow. In the past 20 years, greater understanding of its genetic and biochemical properties has enhanced appreciation of XOR participation in human physiology and pathophysiology. Both laboratory and clinical investigations have implicated XOR in oxidative stress-related diseases, most recently in cardiovascular disease. The precise mechanisms by which XOR inflicts tissue damage and disrupts cell signalling in the cardiovascular system remain to be elucidated, but many studies suggest XOR-oriented therapy holds promise for patients with heart failure and other cardiovascular diseases. The new appreciation of the role of XOR in cardiovascular pathophysiology has led to a wave of clinical trials. The largest of these studies, entitled ‘A Phase II–III Prospective, Randomized, Double-Blind, Placebo-Controlled Efficacy and Safety Study of Oxypurinol Added to Standard Therapy in Patients with NYHA Class III–IV Congestive Heart Failure’ (OPT-CHF), is a 400 patient study currently enrolling participants in the United States and Canada. If positive, the results of this trial have the potential to substantially influence heart failure management in the future.

Disclosure statement

J.M.H. is a paid consultant at Cardiome Pharma Corporation. The terms of this arrangement are being managed by The John Hopkins University in accordance with its conflict of interest policies.