Introduction

The metalloflavoenzymes, xanthine oxidase (XO, EC 1.1.3.22) and xanthine dehydrogenase (XDH, EC 1.1.1.204), are products of a single gene and are expressed in most tissues including the liver, intestine, brain, plasma and capillary endothelial cells [1,2,3]. Both these enzymes catalyse the oxidative hydroxylation of hypoxanthine to xanthine, and xanthine to ultimately yield uric acid (Fig. 1) [1]. These represent the final steps in the catabolism of adenine nucleotides. During these catalytic processes molecular oxygen is reduced to yield the superoxide anion radical and hydrogen peroxide. XDH may be irreversible converted to XO by proteolysis, or reversible by the oxidation of cysteine residues. In this respect healthy tissues display predominantly XDH activity while conversion to XO occurs under pathological conditions [4, 5].

Fig. 1
figure 1

The metabolism of hypoxanthine and xanthine catalysed by XO and XDH

Full size image

XO is of medicinal importance since inhibitors of this enzyme are used for the treatment of gout, the most common cause of inflammatory arthritis [6, 7]. The excessive production of uric acid may lead to hyperuricemia and the deposition of monosodium urate crystals in joints, tendons and surrounding tissues, which causes the characteristic inflammation observed in gout [8]. In this respect, the enzyme uricase that converts uric acid to the more soluble allantoin is not expressed in humans [9]. Furthermore, the oxygen species that are produced by the XO catalytic cycle may be converted into highly reactive hydroxyl radicals, which contribute to gout by causing oxidative damage to tissue [10]. Reactive oxygen species generated by XO may be involved in a variety of pathological conditions [6, 10, 11], most notably ischemia-reperfusion injury where the activation of proteases irreversibly converts XDH to XO, while the cell energy crisis leads to enhanced adenosine triphosphate (ATP) degradation and subsequently hypoxanthine formation [2, 12]. In certain pathological conditions such as ischemia-reperfusion injuries, the activity of XO in the systemic circulation may be increased, which could lead to enhanced levels of reactive oxygen species in the blood [13]. As a result of tissue damage during liver injury, XO activity in the systemic circulation is significantly increased [14].

XO inhibitors are best known for the treatment and prevention of hyperuricemia and gout, while they may find future application in various other disease states that are associated with XO-induced production of reactive oxygen species. Allopurinol has been used clinically for decades and is the first-line choice for the chronic treatment of hyperuricemia and gout (Fig. 2). Allopurinol is first converted to oxipurinol by XO, which binds via a reversible covalent bond to the molybdopterin cofactor [15]. The thiazole derivative, febuxostat, has also been approved for the treatment of gout and is a potent non-competitive inhibitor of XO [16, 17]. Both these compounds however have significant side effects which led to the development of topiroxostat, a compound that firstly acts as a competitive inhibitor and then forms a tight-binding complex with the enzyme that involves the formation of a covalent bond with the molybdopterin cofactor [17, 18]. Other noteworthy XO inhibitors include salicylate, quercetin, chrysin and curcumin, which bind to the substrate binding site in a competitive manner [19,20,21,22]. These compounds are phenol derivatives with quercetin and chrysin being plant flavonoids. IC50 values for the inhibition of XO were reported as follows: salicylate (580 µM), quercetin (2.62 µM) and chrysin (0.84 µM) [23, 24], while conflicting reports regarding the inhibition of XO by curcumin exists [22, 25, 26].

Fig. 2
figure 2

The structures of known XO inhibitors

Full size image

The present study attempted to discover new lead inhibitors of XO and investigated 55 diverse compounds from an in-house library (Figs. 3 and 4). Many of these compounds have been designed or discovered to inhibit the flavoenzymes, monoamine oxidase (MAO) A and B, and since XO also is a flavoenzyme, some overlap in inhibitor specificities between these enzymes may exist. In support of this notion is the observation that quercetin and apigenin are inhibitors of both XO and MAO [27,28,29]. When considering the numerous structures that are present in the in-house library, no attempt was made to filter them for the potential to inhibit XO, by for example comparing the structures and binding groups to known XO inhibitors. Instead, compounds from the different classes of inhibitors were selected in an attempt to preserve the structural diversity of the series that was to be evaluated. The primary aim of this strategy was to discover new classes of XO inhibitors that have not previously been considered. Such compounds would be suitable leads for future structure optimization.

Fig. 3
figure 3

The structures of the compounds 1–23 that were investigated as potential XO inhibitors in this study

Full size image
Fig. 4
figure 4

The structures of the compounds 24–55 that were investigated as potential XO inhibitors in this study [50]

Full size image

The aim of this study was therefore to discover novel compounds that inhibit XO by evaluating a library of compounds with diverse structures. It was hypothesized that among the selected compounds certain derivatives may possess XO inhibition activity and could be considered as leads in future design studies. Alternatively, this study may contribute to knowledge of the biochemical properties of compounds such as the naphthoquinones and tricyclic derivatives that were among the library compounds.

Results and discussion

Inhibition of XO

The inhibition potencies of the test compounds and the reference inhibitor, chrysin, were measured with XO from bovine milk as enzyme source and xanthine as substrate. The catalytic cycle of XO generates hydrogen peroxide when molecular oxygen is reduced in the oxidative half-reaction of the enzyme. For each mole of xanthine that is oxidised, one mole of hydrogen peroxide is produced. To measure the catalytic rate of XO, the formation of hydrogen peroxide was monitored using the peroxidase-linked spectrophotometric assay system described previously [30]. In this assay system, a quinoneimine compound that has a maximal absorbance at 498 nm is generated when 4-aminoantipyrine is oxidised by peroxidase, and subsequently condenses with vanillic acid. The enzyme reaction can thus be monitored continuously.

By measuring XO activity in the absence and presence of the test inhibitors, sigmoidal plots of enzyme catalytic rate versus the logarithm of the inhibitor concentration were constructed from which the IC50 values were estimated. Examples of such plots are given in Fig. 5. The results are presented in Table 1 and show that most compounds did not act as XO inhibitors. Interestingly the nitrocatechol compounds (13 and 14) exhibited some inhibition with the chalcone 14 possessing an IC50 of 45.3 µM, while pyrazoline 13 was less active (IC50 = 226 µM). Five additional compounds were found to inhibit XO, all of which are naphthoquinones, Lawsone (19), juglone (20), menadione (21), 2-methoxy-1,4-naphthoquinone (22) and 1,4-naphthoquinone (23). Among these juglone and menadione were the most potent with IC50 values of 2.45 and 4.38 µM, respectively. Considering the structures of these compounds, it may be speculated that the hydrogen bond donor substituent (OH) of Lawsone on the C2 position was less optimal for inhibition compared to the more lipophilic methyl of menadione and to a lesser extent the OCH3 of 22. On the other hand, an OH group on C5 as with juglone resulted in good XO inhibition. Interestingly, the OH-substituted naphthoquinone compound, lapachol (18) was not a XO inhibitor although the larger side chain may have contributed to the loss of activity. The finding that 1,4-naphthoquinone (23, IC50 = 29.3 µM) was a significantly weaker XO inhibitor compared to juglone and menadione, demonstrated the requirement for the hydroxyl and methyl substituents, respectively. It was interesting to note that other carbonyl containing compounds that bear some structural similarity and similar size to naphthoquinone were not inhibitors. These include most notably isatin (16), but also 1, 5 and 11. Since the relatively large compounds 13 and 14 showed some XO inhibition it may be concluded that the presence of the correct functional groups at the appropriate positions of the compounds play a more important role than the size of the structure.

Fig. 5
figure 5

Sigmoidal plots for the inhibition of XO by selected inhibitors and the reference inhibitor, chrysin. Each value is given as the mean ± SD of triplicate determinations. The absence of error bars at some data points are due to small deviations of the values

Full size image
Table 1 The IC50 values for the inhibition of bovine milk XO by the active compounds of this study, and reference inhibitor, chrysin
Full size table

Among the tricyclic derivatives, five compounds inhibited bovine milk XO with IC50 < 10 µM. 9,10-Phenanthrenequinone (48) was the most potent inhibitor with an IC50 value of 0.726 µM. Other noteworthy inhibitors were benz(g)isoquinoline-5,10-dione (30), 2-chloro-7-methoxy-10H-phenothiazine (32), cinnabarinic acid (34) and quinalizarin (50). As mentioned, quinalizarin is a known inhibitor of XO. These compounds represent good leads for the future development of XO inhibitors. It is interesting to note that the phenothiazine compound, 2-chloro-7-methoxy-10H-phenothiazine, was among the XO inhibitors. This suggested that the phenothiazine moiety may be a suitable scaffold for the design of XO inhibitors and with the appropriate structure modification, potent XO inhibitors may be obtained. While there was no apparent structural similarity between the active XO inhibitors of this study, it was clear that small structural modification may abolish XO activity. This was apparent when comparing the structures of benz(g)isoquinoline-5,10-dione (30) with anthraquinone (27), and quinalizarin (50) with 1,8-dihydroxyanthraquinone (39) and quinizarin (51).

Among the compounds that were evaluated in the present study are coloured compounds and dyes which have the potential to interfere with the absorbance measurements that were taken during the XO inhibition studies. In addition, by interfering with the peroxidase-linked spectrophotometric assay system, test compounds could suppress the absorbance signal, which would wrongly be attributed as inhibitory activity against the XO enzyme. The potential for this was evaluated by measuring the effect of the most active inhibitors on the absorbance signal that was produced by the reaction of H2O2 and the peroxidase-linked assay system used in this study. The absorbance recorded without inhibitor present (100%) was compared to the absorbance values recorded in the presence of the active inhibitors (Table 2). The results showed that none of the inhibitors suppressed the absorbance signal which indicated that the compounds did not interfere with the assay system and are thus true XO inhibitors. However, at a concentration of 100 µM, all inhibitors increased the absorbance signal which could be attributed to intrinsic absorbance of the compounds at the working wavelength of 498 nm. For certain inhibitors (e.g., 32 and 50), an increase in absorbance signal was also observed at 10 μM. This increase in the absorbance signal rather than suppression thereof further supported the conclusion that the compounds are indeed XO inhibitors.

Table 2 Percentage of the absorbance signal that was produced by the reaction of H2O2 with the peroxidase-linked spectrophotometric assay system in the presence of the active XO inhibitors (1, 10, 100 µM) compared to the absorbance signal that was produced when the inhibitor was absent (100%)
Full size table

Molecular docking

To obtain more insight into the interactions between the inhibitors and the enzyme, the structures of juglone, menadione and 9,10-phenanthrenequinone were docked into the XO active site [31,32,33]. The Discovery Studio 3.1 software package (Accelrys) was used for the docking simulations and the X-ray crystal structure of bovine milk XO co-crystallised with quercetin (PDB entry: 3NVY) served as protein model [20]. The protein model was firstly protonated and the pKa values and protonation states of the amino acid residues were subsequently calculated. The model was finally energy minimised with the protein backbone constrained. The structures of the ligands were drawn in Discovery Studio, and after docking with CDOCKER, the docked orientations were refined using in situ ligand minimisation. For each ligand, the highest ranked orientation was selected among the ten solutions generated.

To evaluate the validity of the docking protocol, the co-crystallized ligand (quercetin) was redocked into the active site of the XO model. The root mean square deviation (RMSD) of the docked orientation from the position of the co-crystallised ligand was calculated for the highest ranked orientation, and was found to be 1.52 Å (Fig. 6). RMSD values < 1.5 Å are considered to be successful, and this protocol was thus deemed appropriate for docking experiments with XO [34].

Fig. 6
figure 6

The orientation and interactions of quercetin (yellow) in the active site of XO as determined by X-ray crystallography (PDB code: 3NVY). The docked orientation of quercetin (green) is shown for comparison (panel a). The predicted orientations and interactions of juglone (panel b), menadione (panel c) and 9,10-phenanthrenequinone (panel d)

Full size image

In the crystal structure, the aromatic rings of quercetin are coplanar and sandwiched between the conserved residues, Phe-914 and Phe-1009. The bicyclic rings of quercetin undergo pi stacking with these two residues. The 7-OH group of quercetin is hydrogen bonded to two nitrogen atoms of the guanidinium group of Arg-880 and also to the hydroxyl group of Thr-1010. The C4 carbonyl is hydrogen bonded to both oxygens of Glu-802, while the 5-OH group is hydrogen bonded to a molybdenum (Mo) OH group. These are the principal interactions that determine the binding orientation of quercetin. Additionally, the 3-OH group may be hydrogen bonded to the carboxylate of Glu-802, while the 3’-OH group may be hydrogen bonded to Asn-768 via a bridging water molecule [20].

Both juglone and menadione bind in the space occupied by the chromone ring system of quercetin. Juglone binds with the phenyl moiety in proximity to the molybdopterin cofactor, while for menadione, the quinone ring is in proximity to the molybdopterin. The OH group of juglone undergoes hydrogen bonding to two nitrogen atoms of the guanidinium group of Arg-880, as well as to the hydroxyl group of Thr-1010. One of the quinone carbonyl groups is hydrogen bonded to the peptide nitrogens of Thr-1010 and Val-1011. For menadione, the hydrogen bond network is similar to juglone with hydrogen bonding between a quinone carbonyl and two nitrogen atoms of the guanidinium group of Arg-880, as well as to the peptide nitrogen of Thr-1010. For juglone and menadione, pi-interactions between both aromatic rings and Phe-914 and Phe-1009 may occur. With the exception of the interaction with Glu-802 and the Mo-OH group, juglone and menadione thus undergoes the same principal interactions that determine the binding orientation of quercetin. These may, in part, explain the XO inhibition properties of these two naphthoquinone compounds.

9,10-Phenanthrenequinone also form an extensive hydrogen bonding network. The two carbonyl oxygens act as hydrogen bond acceptors for these interactions, which shows their importance for inhibitor binding. 9,10-Phenanthrenequinone forms the following principal interactions with XO: pi-stacking of the aromatic rings with Phe-914; pi-T-shape interaction with Phe-1009; hydrogen bonding with Arg-880, Thr-1010 and Val-1011; Pi-sulphur interaction with the molybdenum (Mo). These numerous interactions underscore the potential of 9,10-phenanthrenequinone as an inhibitor of XO, and its potential as a lead for the future design of XO inhibitors.

Conclusion

This study investigated the XO inhibition properties of diverse compounds from an in-house library. While most compounds did not inhibit XO, the naphthoquinones, juglone and menadione, proved to be good potency inhibitors. The naphthoquinones thus represent a new class of XO inhibitors and although juglone and menadione are interesting new XO inhibitors, little structural modification is tolerated. To derive structure-activity relationships (SARs) systematic modification of the menadione structure is required. Among the tricyclic derivatives, five compounds were found to inhibit XO with IC50 < 10 µM. 9,10-Phenanthrenequinone proved to be the most potent inhibitor with an IC50 value of 0.726 µM. This compound represents a good lead for the future development of XO inhibitors.

The inhibition of XO by the study compounds have not yet been reported before. These findings may pave the way to further investigate the compounds as potential XO inhibitors, while they may be considered as leads for future development of XO inhibitors. It is noteworthy that some of the potent XO inhibitors discovered in this study are natural products. For example, juglone is present in plants of the Juglandaceae family while cinnabarinic acid is a trace metabolite of kynurenine metabolism [35]. Juglone, in particular, has various biological activities such as antimicrobial, antifungal and anti-proliferative actions and is a privileged structure for medicinal chemistry [36]. XO inhibition by these compounds should be considered in future studies where their biochemical properties are studied. The present study is an example of the advancement of our understanding of the pharmacological activities of these compounds.

Experimental section

Materials and methods

XO from bovine milk (X1875) was purchased from Sigma-Aldrich. Xanthine, horse radish peroxidase (P8250), vanillic acid and 4-aminoantipyrine were also obtained from Sigma-Aldrich.

Compounds 1–15 were from an in-house compound library for which the compounds have been reported [37,38,39,40,41,42,43,44,45,46,47,48,49]. Isatin (16), (+)-kavain (17), lapachol (18), Lawsone (19), juglone (20), menadione (21), 2-methoxy-1,4-naphthoquinone (22) and 1,4-naphthoquinone (23) as well as the reference inhibitor, chrysin, were obtained from Sigma-Aldrich. The tricyclic compounds that were used in this study (24–55) were obtained from Sigma-Aldrich [50].

Instrumentation

A SpectraMax® iD3 multi-mode microplate reader (Molecular Devices) was used to record changes in absorbance during the peroxidase-coupled spectrophotometric assay system used to measure hydrogen peroxide formation by XO.

Determination of IC50 values for the inhibition of XO

To determine the catalytic activity of XO, the enzyme from bovine milk served as enzyme source, while xanthine was used as substrate. Bovine XO was selected for this study based on commercial availability of the enzyme, while several inhibition studies have reported the use of bovine XO as enzyme source [51, 52]. The initial rate of hydrogen peroxide formation by XO was measured with a peroxidase-linked spectrophotometric assay system described previously [30]. This approach was used to investigate the 55 test compounds as potential XO inhibitors, while chrysin was used as reference inhibitor. The enzyme reactions were carried out to a final volume of 200 μL in potassium phosphate buffer (100 mM, pH 7.4) in clear 96-well microtiter plates. The enzyme reactions contained 50 µL substrate, 40 µL of the chromogenic solution, 8 µL of the inhibitor stock solutions, 77 µL reaction buffer and 25 µL enzyme. The xanthine stock solution (4 mM) was prepared in buffer and a minimum volume of sodium hydroxide (2 N) was added to dissolve the xanthine. The stock solution was then diluted to 500 µM in buffer. The chromogenic stock solution was prepared in potassium phosphate buffer (100 mM, pH 7.4) and contained horse radish peroxidase (20 units/mL), vanillic acid (1 mM) and 4-aminoantipyrine (500 µM) [30]. The inhibitors were dissolved in DMSO and added to the reactions to yield a final concentration of DMSO of 4%. The enzyme, provided as an ammonium sulphate suspension, was firstly centrifuged for 5 min at 5000 × g. The supernatant was discarded and buffer was added to the pellet to obtain a 1.2 units/mL solution.

The final concentrations of the reaction constituents were: substrate (125 µM), peroxidase (4 units/mL), vanillic acid (0.2 mM), 4-aminoantipyrine (100 µM), inhibitor (0.003–100 µM) and enzyme (0.15 unit/mL). Control incubations carried out in the absence of inhibitors were also included. The reactions were initiated with the addition of enzyme and the change in absorbance (at 498 nm) was measured kinetically at room temperature for 20 min, with measurements taken at 51 s intervals. After discarding a lag period of approximately 100 s, the initial linear rate of absorbance change was noted and plotted versus the logarithm of the inhibitor concentration. These data were fitted to the one site competition model incorporated in the Prism® version 5.0 software package (GraphPad Software). This yielded sigmoidal plots from which the IC50 values were determined and expressed as mean ± SD. The following equation was used to determine IC50 values for the sigmoidal plots: Y = Bottom + [Top – Bottom)/(1 + 10^(X – LogIC50)], where Top and Bottom are plateaus of the Y axis, and LogIC50 is the logarithm of the concentration of competitive inhibitor that results in a response half-way between Bottom and Top.

Suppression of the absorbance signal by the test compounds

The possibility exists that the test compounds might suppress the absorbance signal generated by the reaction of H2O2 with the peroxidase-linked assay system described above. To evaluate this, the absorbance was measured after incubation of H2O2 and the chromogenic stock solution (horse radish peroxidase, vanillic acid and 4-aminoantipyrine) in the presence of the test inhibitors (at concentrations of 1, 10 and 100 μM). The absorbances of the resulting reactions were compared to the corresponding value of the control, which was conducted in the absence of inhibitor.

The reactions were carried out to a final volume of 200 μL in potassium phosphate buffer (100 mM, pH 7.4) in clear 96-well microtiter plates, and contained 102 µL buffer, 8 μL of the test compounds, 50 μL H2O2 and 40 μL of the chromogenic stock solution. The final concentration of H2O2 in the reactions was 25 μM while the test compounds were evaluated at 1, 10 and 100 μM. The chromogenic solution was prepared as described above, the H2O2 stock solution (100 µM) was prepared in reaction buffer while the test compounds were dissolved in DMSO and added to the reactions to yield a final concentration of 4%. The control reactions did not contain test compound but had a final DMSO concentration of 4%. The reactions were preincubated for 10 min at room temperature prior to the addition of H2O2. After the addition of H2O2, the reactions were incubated for a further 10 min and the absorbance was measured at endpoint at 498 nm.

Molecular docking studies

The Discovery Studio 3.1 suite of software (Accelrys) was used for all molecular modelling simulations. The structure of bovine XO complexed with quercetin (PDB code: 3NVY) was obtained from the Protein Data Bank and was prepared for docking in Discovery Studio [20]. The pKa values and protonation states of the ionisable amino acids were firstly calculated, and hydrogen atoms were added to the models at pH 7.4. The Momany and Rone CHARMm forcefield was used to type the protein models and a fixed atom constraint was applied to the backbone. The protein models were energy minimised with the Smart Minimizer algorithm using the implicit generalised Born solvation model with molecular volume, and a maximum amount of steps of 50,000. The co-crystallised ligand, waters and backbone constraint were removed from the model and the active site was identified from the position of the co-crystallized ligand. Discovery studio was used to construct the structures of the ligands and a Dreiding-like forcefield (5000 iterations) was used to optimise their geometries. The ligands were submitted to the Prepare Ligands protocol, and atom potential types and partial charges were assigned with the Momany and Rone CHARMm forcefield. The CDOCKER algorithm was used to dock the ligands, allowing for the generation of ten random ligand conformations, while the heating target temperature was set to 700 K and full potential mode was employed. Finally, the Smart Minimizer algorithm was used to refine the docking solutions by means of in situ ligand minimisation. Illustrations were prepared with the PyMOL molecular graphics system [53].