Materials and Methods

Androsterone, UDPGA, and saccharic acid-1,4-lactone were purchased from Sigma-Aldrich Chemie GmbH (Steinheim, Germany). Androsterone glucuronide and etiocholanolone glucuronide sodium salt were from National Analytical Reference Laboratory (Pymble, NSW, Australia). Etiocholanolone was from Makor Chemicals Ltd. (Jerusalem, Israel). The diols 5α-diol and 5β-diol were from Steraloids Inc. Ltd. (London, UK). Magnesium chloride hexahydrate, disodium hydrogen phosphate, potassium dihydrogen phosphate, and perchloric acid were from Merck (Darmstadt, Germany). Dimethyl sulfoxide (DMSO) and ammonium acetate were from Riedel-de Haën (Seelze, Germany). All the other reagents were analytical grade, and solvents were high-performance liquid chromatography grade; eluents were filtered through a 0.22-μm filter for ultraperformance liquid chromatography (UPLC) analysis.

Recombinant UGTs.

Recombinant human UGTs of subfamilies UGT1A and UGT2B were expressed in baculovirus-infected insect cells as described previously (Kurkela et al., 2007 and references therein). The recombinant human UGTs of subfamily UGT2A were expressed in the same system, and the details of their cloning will be published shortly (N. Sneitz, M. H. Court, X. Zhang, K. Laajanen, K. K. Yee, P. Dalton, X. Ding, and M. Finel, manuscript in preparation). Protein concentrations were determined by the bicinchoninic acid method (Pierce Biotechnology Inc., Rockford, IL). The relative expression level of the recombinant UGTs was determined using a monoclonal antibody, tetra-His (QIAGEN, Hilden, Germany) directed to the C-terminal His-tag that they carry, as detailed earlier (Kurkela et al., 2007). The relative expression level was used to normalize the glucuronidation rate of the different UGTs, with the exception of the commercial UGT2B15* (BD Gentest, Woburn, MA), which was used in parallel to our UGT2B15 (see also Itäaho et al., 2008). The expression level of the commercial UGT2B15 could not be compared with our enzymes because of the lack of suitable antibodies.

Analytical Methods.

We have developed a liquid chromatography/mass spectrometry (LC/MS) method for studying androsterone and etiocholanolone glucuronidation in which the LC separation part was carried out on a UPLC system. Waters Acquity UPLC (Waters, Milford, MA) with MassLynx 4.1 software was used for the analyses. The conditions were as follows: eluents A, 0.1% ammonium acetate (aq.) and B, acetonitrile with gradient elution of 0 to 3 min B, 1 to 55%; 3 to 3.5 min B, 55 to 90%; 3.5 to 3.6 min B, 90 to 10%, followed by equilibrium time of 2.4 min. Flow rate was 0.2 ml/min, and column temperature was 40°C. Acquity BEH Shield RP18, 100 × 0.1 mm i.d. with 1.7-μm particle size (Waters) was used for the separation. The injection volume was 2 μl. The first eluting buffer salts were directed to waste by a six-port switching valve (Rheodyne LLC, Rohnert Park, CA). Quadrupole time-of-flight Micro (Waters/Micromass, Manchester, UK) with electrospray ionization source was applied under negative ion LC/MS mode. Spectra over the 50 to 800 m/z mass range were recorded in the centroid mode. The capillary, sample cone, and extraction cone voltages were 3500, 45, and 2.0 V, respectively. Desolvation temperature was 250°C, and source temperature was 100°C. Nitrogen produced by a high-purity nitrogen generator (Peak Scientific, Inchinnan, Renfrewshire, UK) was used as desolvation and cone gas at 700 and 200 l/h, respectively.

Under the above chromatographic conditions, androsterone and etiocholanolone glucuronide were eluted at retention times of 2.34 and 2.32 min, respectively (Fig. 2). The corresponding aglycones were eluted at 3.78 and 3.63 min, respectively. The same method was used for studying α-diol and β-diol glucuronidation, where two glucuronides were anticipated from each substrate. The two glucuronides generated from 5α-diol were eluted at 2.16 and 2.30 min, and the two 5β-diol glucuronides were eluted at 2.15 and 2.34 min. The standard curves of androsterone and etiocholanolone glucuronide were linear in the concentration range from 0.17 μM (limit of quantification) to 21 μM, and the limit of detection was 0.08 μM for both analytes.

Analytical Method Calibration and Validation.

Stock solutions of androsterone and etiocholanolone glucuronide in 70% acetonitrile (aq.) at a concentration of 1.0 mM were stored at −20°C. Standard calibration working solutions containing individual analytes were prepared in acetonitrile. The samples for the calibration curves were prepared by evaporating the solvent from the standard solution and reconstituting the solid with blank incubation matrix. Testosterone glucuronide, 5 μM final concentration, was added as an internal standard before sample pretreatment. The ions at m/z 465 (androsterone and etiocholanolone glucuronide [M-H]⁻) and m/z 463 (testosterone glucuronide [M-H⁻]) were measured, and analyte signal area was corrected according to the internal standard signal. The analytical method was validated with respect to accuracy, precision, linearity, limits of detection, and quantification as recommended by Shah et al. (2000). The limits of detection and quantification were determined as signal-to-noise ratios of 3 and 10, respectively. The between-assay coefficients of variation for androsterone and etiocholanolone glucuronide were less than 9%, except at the limit of quantification when they were <16%. Because of lack of authentic reference standards, the diol glucuronides were quantified using androsterone glucuronide standard curve.

The 5α-diol and 5β-diol metabolites were identified as glucuronides by their mass in combination with their absence in the control samples that were incubated without UDPGA. The glucuronidation position of the diols was deduced based on previous knowledge of 5α-diol glucuronidation by UGT recombinant enzymes (see under Results). Because authentic standards for the glucuronides of androstane diols were not available, they were quantified using the androsterone glucuronide standard curve, assuming equal electrospray ionization response for the glucuronides. The detection limit and quantification limit were defined as analyte peak (m/z 467, [M-H]⁻) versus internal standard (m/z 463, [M-H]⁻) area, and the linearity of the response versus sample concentration was verified by sample dilution.

Incubation Conditions for UGT Screening.

The assays to identify the human UGT isoforms active in steroid glucuronidation were performed in the presence of 5 mM UDPGA, 5 mM d-saccharic acid 1,4-lactone, 50 mM Na-K-phosphate buffer, pH 7.4, and 5 mM MgCl₂. The protein concentration was 0.5 mg/ml, and the total reaction volume was 50 μl. The aglycone was added as DMSO solution so that the final aglycone concentration was 50 μM and the DMSO concentration was 5% in all the assays. Control incubations were carried out in the absence of UDPGA or in the absence of aglycone substrate and the presence of UDPGA. The reactions were started by the addition of UDPGA, and the samples were incubated at 37°C in a shaking incubator (Vortemp; UniEquip Laborgerätebau und Vertriebs GmbH, Munich, Germany) for 120 min. The reactions were terminated by the addition of ice-cold 4 M perchloric acid, 10% of the reaction volume, and transfer to ice. After internal standard addition, the mixtures were subsequently centrifuged for 10 min at 16,000g (Eppendorf 5415 D; Eppendorf AG, Hamburg, Germany), and aliquots of the supernatants were subjected to UPLC/MS analyses. The reactions were carried out in triplicate (androsterone and etiocholanolone) or duplicate (diols, less than 10% difference between measurements) except for the negative controls, which were single samples.

Kinetic Analyses.

Kinetic assays of androsterone and etiocholanolone glucuronidation by UGT2B17 and UGT2B7, as well as UGT2A1 in etiocholanolone glucuronidation, were carried out. The reaction conditions were similar to the screening assays, except that the protein concentration and incubation time were adjusted so that product formation was linear with respect to both parameters, as well as yielded quantitative amounts of the assayed glucuronides (see legends to the relevant figures). The activity data were fitted to Michaelis-Menten, substrate inhibition, and Hill equations using GraphPad Prism version 4.02 for Windows (GraphPad Software Inc., San Diego, CA). The best-fit kinetic model was chosen based on the estimated S.E., 95% confidence intervals, and R². If no clear preference was found, the simplest model was chosen. The kinetic parameters K_m and V_max were calculated according to the best-fit model.

Results

We have examined the glucuronidation rate of androsterone and etiocholanolone by the 19 human UGTs of subfamilies UGT1A, UGT2A, and UGT2B using an improved LC/MS method in which the chromatography was carried out on a UPLC system (Fig. 2, see under Materials and Methods). Among the nine UGTs of subfamily UGT1A, only UGT1A3 and UGT1A4 exhibited detectable activity (Table 1). The rate of androsterone glucuronidation by UGT1A4 could be quantified, but the generation of androsterone and etiocholanolone glucuronides by UGT1A3, as well as etiocholanolone glucuronidation by UGT1A4, was below the quantification limit. Among the three enzymes of subfamily UGT2A, the extrahepatic UGT2A1 and UGT2A2 exhibited activity toward both androsterone and etiocholanolone. We were surprised to find that the rate of etiocholanolone glucuronidation by UGT2A1 was very high, the second highest among the human UGTs, and nearly 20 times higher than the rate of androsterone glucuronidation by this enzyme (Table 1). The most active enzymes of subfamily UGT2B in androsterone and etiocholanolone glucuronidation were UGT2B7 and UGT2B17. These enzymes exhibited reverse preference to these substrates; UGT2B7 glucuronidated androsterone approximately three times faster than etiocholanolone, whereas UGT2B17 catalyzed etiocholanolone glucuronidation approximately twice the rate of its androsterone glucuronidation activity (Table 1). In addition, UGT2B4 glucuronidated both substrates at detectable rates.

Kinetic assays were carried out to study the characteristics of androsterone and etiocholanolone glucuronidation by the most active enzymes, namely, UGT2B7 and UGT2B17 in both reactions and UGT2A1 in etiocholanolone glucuronidation (Figs. 3 and 4). Substrate inhibition was observed in androsterone glucuronidation by UGT2B7, whereas the reaction catalyzed by UGT2B17 followed Michaelis-Menten kinetics (Fig. 3). In the case of etiocholanolone glucuronidation, on the other hand, the reaction catalyzed by UGT2B7 followed Michaelis-Menten kinetics, whereas UGT2B17 revealed minor substrate inhibition (Fig. 4). UGT2B17 exhibited low and similar K_m values for both androsterone and etiocholanolone, approximately 1 μM (Table 2). The K_m values of UGT2B7 for both substrates were also similar to each other but approximately 5 times higher than the respective values in UGT2B17.

The kinetics of etiocholanolone glucuronidation by UGT2A1 was unusual and significantly different from both UGT2B7 and UGT2B17. The affinity of UGT2A1 for etiocholanolone was exceptionally low, and it was practically impossible to derive an accurate K_m value for this reaction because up to a substrate concentration of 120 μM, the velocity of etiocholanolone conjugation by UGT2A1 increased almost linearly (Fig. 4). However, application of higher substrate concentration was limited by solubility, making it very difficult to conclude which type of kinetics this reaction follows. In Table 2, we have calculated the K_m and V_max values for UGT2A1 assuming Michaelis-Menten kinetics.

The 19 human UGTs were subsequently screened for 5α-diol and 5β-diol glucuronidation (same enzyme batches as used for androsterone and etiocholanolone glucuronidation screens). The two expected glucuronides from each of these substrates were well separated by the UPLC (Fig. 2, see under Materials and Methods for details), and we have examined the regioselectivity of glucuronidation by different UGTs. Because no reference standards were available, we have used the androsterone glucuronide standard curve for quantification, assuming similar response for all four structural and regioisomeric glucuronides. Several recombinant UGTs revealed activity toward one or both of the diols, ranging from barely detectable rates of glucuronidation, e.g., UGT1A3, to high rates and either strict regio-selectivity, as in the case of UGT2B7 and 5α-diol, or high rates and less strict regioselectivity, as with UGT2B17 (Table 3).

The glucuronidation site in the 5α-diol glucuronides, either 3-OH or 17-OH (Fig. 1), was assigned based on a previous report where authentic standards for the two 5α-diol glucuronides were used (Chouinard et al., 2006). In the latter work it was found that only 3-O-glucuronide of 5α-diol was formed by UGT2B7, and the single product generated by UGT2B7 after incubation in the presence of 5α-diol in our study was, therefore, deduced to be the 3-O-glucuronide (Fig. 2). In another earlier study it was found that UGT2B15 is strictly regioselective toward the 17-OH of androgens (Green et al., 1994). In agreement with that study, we have detected a single glucuronide in the incubation of UGT2B15 with either 5α-diol or 5β-diol and recognized it as the respective 17-O-glucuronide (Fig. 2). Both UGT2B15 samples, the commercial and our lower activity preparation, exhibited the same regioselectivity regardless of considerable differences in the glucuronidation rate (Table 3). The results with UGT2B15 helped us to assign the glucuronides generated from 5β-diol, even if no authentic standards were available, either for us or in previous reports. Hence, the single glucuronide generated by UGT2B15 from either 5α-diol or 5β-diol, both assumed to be diol-17-O-glucuronide, was different from the respective glucuronides generated by UGT2B7 from these diols (Fig. 2), leading us to conclude that UGT2B7 is strictly specific for the 3-OH also in the case of 5β-diol.

Examining the results from the point of view of different subfamilies of the human UGTs, it seems that the enzymes of subfamily UGT1A do not contribute much to androgen glucuronidation. Nevertheless, it is of some interest that UGT1A4 exhibited activity in our assays, particularly toward the 17-OH of 5α-diol (Table 3). UGT1A4 was also able to glucuronidate 5α-diol at the 3-OH but at low rates. It even catalyzed 5β-diol glucuronidation, but in this case the rates were below the limit of quantification. It is interesting to note that UGT1A8 and UGT1A10, two enzymes that readily glucuronidate the aromatic 3-OH in estradiol (Lépine et al., 2004; Itäaho et al., 2008), did not yield any 3-O-glucuronides in the current study. Both enzymes revealed a detectable glucuronidation rate toward 5α-diol, but it was low and restricted to the 17-OH (Table 3).

The UGT2A subfamily members, UGT2A1 and UGT2A2, exhibited interesting activity with unique regio- and stereoselectivity. In the case of 5α-diol, both UGT2A1 and UGT2A2 had a clear preference, or even absolute specificity (UGT2A1), toward the 17-OH, namely, a similar regioselectivity to the few UGTs of subfamily UGT1A that catalyzed these reactions. On the other hand, both UGT2A1 and UGT2A2 were strictly specific for the 3-OH in 5β-diol glucuronidation, exhibiting the same regioselectivity as UGT2B4 and UGT2B7 (Table 3).

In subfamily UGT2B, UGT2B4 exhibited the same regioselectivity as UGT2B7 in both 5α-diol and 5β-diol glucuronidation, but the normalized rates of UGT2B4 activity were much lower than UGT2B7 (Table 3). The regioselectivity of UGT2B17, on the other hand, was very different from UGT2B7. UGT2B17 exhibited high activity toward the 17-OH of both diols, whereas UGT2B7 was highly selective for the 3-OH in these diols. UGT2B17 also catalyzed glucuronidation of both diols at the 3-OH but at much lower rates than at the 17-OH of either diol (Table 3). Unlike UGT2B17, UGT2B15 was strictly selective for the 17-OH in both diols. In addition, UGT2B15 exhibited high preference for 5α-diol over 5β-diol, whereas in UGT2B17 this preference was less pronounced (Table 3).

Discussion

This study on 5α- and 5β-androstane steroid glucuronidation by recombinant human UGTs is an extension of our recent work on testosterone and epitestosterone glucuronidation (Sten et al., 2009). We have mainly focused on structure-activity relationships in steroid glucuronidation, and in this respect the diols can be seen as structural analogs of androsterone (5α-diol) and etiocholanolone (5β-diol), in which the keto group at C17 is replaced by a hydroxyl in the β-configuration (Fig. 1). Hence the current set of compounds allows us to examine the effect of a flat versus bent A-B-ring steroid scaffold, a result of the α- versus β-configuration of C5, on the glucuronidation at the 3-OH of androsterone and etiocholanolone. In addition, we can now study the influence of both the configuration of C5 and the presence of a hydroxyl at C3 on the glucuronidation of the 17β-OH of the diols.

The screening results for each of the 19 human UGTs with the four androgens are listed above (see under Results, Tables 1 and 3). This work provides many new findings, not least the activities of UGT2A1 and UGT2A2 toward the different androgens. However, the results obtained with UGT2B7, UGT2B15, and UGT2B17 are probably the most interesting from the point of view of structure-function relationships and substrate selectivity of the human UGTs. UGT2B7 and UGT2B17 exhibited the highest activity rates in the glucuronidation of androsterone and etiocholanolone (Tables 1 and 2). The results suggest that UGT2B7 and UGT2B17 were affected by the configuration of C5 in an opposite manner. UGT2B7 glucuronidated an-drosterone at much higher rate than etiocholanolone (Table 1) and similarly yielded more 5α-diol-3-glucuronide than 5β-diol-3-glucuronide (Table 3). UGT2B17, on the other hand, glucuronidated etiocholanolone at a higher rate than androsterone, and its low rate of activity toward the 3-OH of 5β-diol was higher than toward the same position in 5α-diol (Table 3). Interestingly, the kinetic analyses suggest that the differences in glucuronidation rates between androsterone and etiocholanolone by UGT2B7 and UGT2B17 do not stem from the K_m. The K_m values of UGT2B7 for both the planar and the bent androgens were similar, and the same situation was observed in UGT2B17 (Table 2). This result may suggest that a difference in the geometry or the exact positioning of the functional hydroxyl group in the substrate with respect to the catalytic amino acid residue(s) in the enzyme, rather than the affinity of the enzyme for the substrate, leads to the rate differences, particularly in UGT2B7.

The inclusion of 5α-diol and 5β-diol in this glucuronidation study added a new dimension to the differences between UGT2B7 and UGT2B17. UGT2B7 appeared to bind 5α-diol in the same manner as androsterone and 5β-diol like etiocholanolone, leading to a strict regioselectivity toward the 3-OH in both diols (Table 3). UGT2B17 revealed a different and “flexible” regioselectivity. Once a 17-(β)OH was available on the substrate androgen, almost no glucuronidation at the 3-(α)OH was catalyzed by UGT2B17 (Table 3). This regioselectivity flexibility in UGT2B17 is even more unusual when compared with UGT2B15 because these two are 92.7% identical in their amino acid sequence, differing by merely 30 of the 507 residues of the mature proteins. Like UGT2B17, UGT2B15 exhibited a strong preference for glucuronidation at the 17-OH, particularly in 5α-diol (Table 3). However, unlike UGT2B17, UGT2B15 did not glucuronidate any of the tested androgens at the 3-OH (Tables 1 and 3). Hence UGT2B7 (and UGT2B4) and UGT2B15 revealed opposite regioselectivity, but in both enzymes it was strict and rigid. UGT2B17, on the other hand, glucuronidated androsterone and etiocholanolone at the 3-OH at high rates and high affinity (Tables 1 and 2), but in the diols it exhibited a strong preference for the 17-OH (Table 3).

UGT2A1 probably do not contribute significantly to the type and level of androgens glucuronides in the human plasma and urine because it is mainly expressed in the nasal epithelium (Jedlitschky et al., 1999). Nevertheless, its activities could unveil interesting features in the substrate selectivity of other UGTs, particularly when trying to identify individual amino acids that are directly involved in the binding and glucuronidation of specific substrates. In the current study, UGT2A1 exhibited a highly unusual activity toward etiocholanolone, namely, a capacity for a higher turnover in the glucuronidation of this compound than any other human UGT, but in combination with a much large K_m value than in UGT2B17, the second fastest enzyme in etiocholanolone glucuronidation (Fig. 4; Table 2). UGT2A1 also resembled UGT2B17 in exhibiting a flexible regio-selectivity, even though in UGT2A1 this was restricted to androsterone and 5α-diol. UGT2A1 catalyzed androsterone glucuronidation (at the 3-OH) at a moderate rate (Table 1), but in 5α-diol it was inactive toward the 3-OH, while conjugating the steroid at the 17-OH at a rather high rate (Table 3). In 5β-diol glucuronidation, on the other hand, UGT2A1 only glucuronidated the 3-OH, the equivalent of the hydroxyl in etiocholanolone (Fig. 1). Hence if we consider UGT2B7 and UGT2B15 as having have a “rigid” regioselectivity and UGT2B17 as having a flexible one, then the regioselectivity of UGT2A1 toward the current set of four steroids could be termed semirigid.

The current results on steroid glucuronidation also have implications for the fight against illicit doping in sport. Testosterone is a precursor in the biosynthesis of the compounds examined in this study (Fig. 1), and abusing exogenous testosterone might affect not only the levels of testosterone glucuronide in the urine but also the level of some of these compounds or their glucuronides. The latter is particularly relevant in homozygous carriers of the UGT2B17 deletion polymorphism because UGT2B17 is the main player in testosterone glucuronidation (Jakobsson et al., 2006; Schulze et al., 2008; Sten et al., 2009). Thus, it may be useful to identify possible urinary androgen metabolites, the level of which is unlikely to be directly affected by the presence or absence of UGT2B17. The results of this study provide such targets. Although UGT2B17 can contribute significantly to the glucuronidation of each of the four androgens that were tested in this work, there is glucuronide regioisomer whose urinary level UGT2B17 does not significantly affect, namely, 5α-diol-3-O-glucuronide (Table 3). Because the 3-glucuronide and the 17-glucuronide of each of these diols can be separated by high-performance liquid chromatography (Chouinard et al., 2006) and UPLC (current study, see under Materials and Methods), it may be possible to use 5α-diol-3-O-glucuronide as a compound, the urinary level of which is independent of the presence or absence of a functional UG2B17.

Finally, in this work we have studied the glucuronidation of four physiologically relevant androgens by all the 19 human UGTs of subfamilies UGT1A, UGT2A, and UGT2B. The results provide new insights into the complex and challenging area of substrate specificity among these enzymes, particularly with respect to steroids. On the other hand, our finding reveals that the overlap in substrate selectivity of the UGTs, particularly when one is looking at the individual metabolites that are produced by them, is not as broad as often assumed.