Synthetic anabolic androgenic steroids are a group of testosterone derivatives that are widely misused in sport to improve the physical performance of skeletal muscle and to balance the catabolic condition after stress (Celotti and Negri-Cesi, 1992). These compounds are extensively modified in the body by phase I and phase II metabolic reactions before excretion in the urine (Gower et al., 1995), and their phase I metabolism has been described previously (Schänzer and Donike, 1993; Schänzer, 1996; Rendic et al., 1999). Glucuronidation, a typical reaction of phase II metabolism, is probably the main pathway of conjugation with anabolic androgenic steroid metabolites (Schänzer, 1996), but detailed information about the factors that determine and affect this biotransformation is still very limited. More is known, however, on glucuronidation of clinically important endogenous steroids, which have been taken as an estimate for the total androgenic pool in men (Labrie et al., 1997), or for the production of dihydrotestosterone in extrahepatic tissues (Bélanger et al., 1998).

UDP-glucuronosyltransferases (UGTs¹; EC 2.4.1.17) are a family of membrane-bound enzymes of the endoplasmic reticulum. They catalyze the glucuronidation of various endogenous and exogenous compounds, including steroids, thereby converting the substrate molecules (the aglycones) into a less toxic and more polar β-d-glucuronides (Dutton, 1980; Radominska-Pandya et al., 1999; Tukey and Strassburg, 2000). The human genome encodes at least 16 different UGTs, and they are divided into families (1 and 2) and subfamilies (2A and 2B) according to the degree of sequence identities and genomic organization (Burchell et al., 1991; Mackenzie et al., 1997; Tukey and Strassburg, 2000, 2001). Most of the UGTs are expressed in the liver, the organ that is considered to be the major site of glucuronidation. However, some UGTs are extrahepatic enzymes, and many of the liver UGTs are also found in other tissues (Tukey and Strassburg, 2000, 2001).

The involvement of UGTs of the 2B subfamily in steroid glucuronidation is well documented (Chen et al., 1993; Jin et al., 1993; Bélanger et al., 1999; Turgeon et al., 2001), and evidence of regio- and stereoselective conjugation of endogenous androgens and pregnanes for these enzymes has been presented (Jin et al., 1997). The activity of UGT1A isoforms toward steroids was also studied, particularly for aglycones having a C18 structure (Hum et al., 1999). In addition, conjugation capabilities toward C19 steroids were described for UGT1A3 (Mojarrabi et al., 1996), UGT1A4 (Green and Tephly, 1996), and UGT1A10 (Strassburg et al., 1998). Nevertheless, a systematic approach that will help to underline the structure-function relationships in this activity, both with respect to the steroid and the individual human UGT, is still missing.

In the present study we have examined the activity of recombinant human UGTs in glucuronidation of a set of 11 exogenous anabolic steroids and their phase I metabolites to gain insight into the structural factors that affect the enzyme-aglycone interactions. The analyses were performed using liquid chromatography-tandem mass spectrometry (LC-MS/MS) with electrospray ionization, which allowed direct determination of steroid glucuronides. The results reveal interesting differences in substrate specificity among the human UGTs, particularly those of the 1A family.

Materials and Methods

Substrates and Reagents. The steroid aglycones that were used in this study are listed in Table 1. Testosterone, nandrolone, and methyltestosterone were received as a generous gift from United Laboratories Ltd. (Helsinki, Finland). Nandrolone metabolites 5α-estran-3-ol-17-one and 5β-estran-3-ol-17-one were obtained from Steraloids (Wilton, NH). All other steroid metabolites were prepared via chemical syntheses at the German Sport University (Deutsche Sporthochschule, Cologne, Germany), according to synthesis routes described earlier (Schänzer et al., 1991, 1992; Schänzer and Donike, 1993). Steroid aglycone structural confirmations were carried out by NMR and GC-MS analyses (Thevis, 2002). The deuterium-labeled (17α-CD₃) analog of 17α-methyl-5β-androst-3α,17β-diol was prepared at the Department of Pharmacy, University of Helsinki, according to the previously published procedure (Shinohara et al., 1984), glucuronidated in enzyme-assisted synthesis described elsewhere in detail (Kuuranne et al., 2002), and used as internal standard in LC-MS/MS analyses. Uridine-5′-diphosphoglucuronic acid (UDPGA, disodium salt) and saccharic acid-1,4-lactone were purchased from Sigma-Aldrich (St. Louis, MO). High-performance liquid chromatography grade solvents and analytical grade reagents were used throughout the study. Human genomic DNA from healthy individuals was a generous gift from Prof. Ismo Ulmanen (Dept. of Molecular Medicine, National Public Health Institute, Biomedicum, Helsinki, Finland).

Recombinant Human UGTs. The expression of human UGTs 1A1, 1A3, 1A4, 1A6, 1A9, 2B4, 2B7, and 2B15 in baculovirus-infected insect cells was published recently (Kurkela et al., 2003). In addition, we have now cloned and expressed the extrahepatic UGTs 1A7, 1A8, and 1A10 as follows. The first exons of UGT1A7, 1A8, and 1A10 were amplified from genomic DNA in two rounds using Vent DNA polymerase (New England Biolabs, Beverly, MA) and the oligonucleotides listed in Table 2. Oligonucleotides that fully match the genomic DNA sequences were used for the first round of amplification, and those that carry additional restriction sites were used in the second round. A HindIII restriction site was introduced, as a silent mutation, into the beginning of exon 2 of the cloned UGT1A9 using primer 1 (Table 2), and the same site was included in primer 10 (Table 2) that was used in the second amplification rounds of all the three UGTs. New restriction sites, either BamH1 (1A7 and 1A8) or BglII (1A10), were inserted upstream of the first ATG of the amplified segments (primers 7–9, Table 2). Subsequently, full-length UGTs 1A7, 1A8, and 1A10 were generated by combining the respective amplified exon 1 with exons 2 to 5 from UGT1A9, using the newly generated HindIII sites. The constructs were verified by DNA sequencing and then transferred into the modified pFastBac derivative that included a C-terminal extension (pFBXHC; Kurkela et al., 2003). Membrane fractions from baculovirus-infected insect cells expressing individual recombinant human UGTs were prepared as previously described (Kurkela et al., 2003). The expression level of individual recombinant UGTs was estimated by Western blot analyses using monoclonal antibodies (Tetra-His antibodies; Qiagen, Germany) against the His-tag that all of them carry (Kurkela et al., 2003). For activity comparison between individual UGTs, the enzyme level was normalized with respect to UGT2B15, which exhibited the lowest expression level in the membrane batches used in this study.

Human and Rat Liver Microsomes. Human liver microsomes were purchased from BD Gentest (Woburn, MA). Rat liver microsomes were isolated from Aroclor 1254-induced male Wistar rats as previously described (Luukkanen et al., 1997) and stored at -70°C before use. The preparation of rat liver microsomes was approved by the local Ethical Committee of the Department of Occupational Health (Helsinki, Finland).

Activity Assays. After checking the activity of enzymes with known substrates, the assays of enzyme activity were carried in a total volume of 100 μl containing 5 mM MgCl₂, 5 mM saccharic acid lactone, and 50 mM phosphate buffer, pH 7.4. Steroid aglycones were added as dimethyl sulfoxide (DMSO) solution, to a final concentration of 50 μM in the reaction mixtures. The protein concentration in the assays, either recombinant UGTs or microsomal preparations, was 0.5 mg/ml. The reactions were initiated by the addition of UDPGA to a final concentration of 5 mM and carried out at 37°C for 120 min. The reactions were terminated by the addition of 10 μl of 4 M perchloric acid and transferred to ice. The mixtures were then centrifuged (14,000g × 10 min), and the supernatants were subjected to cleanup by solid phase extraction (SPE).

Solid Phase Extraction. The nonvolatile salts from the reaction mixtures were removed before LC-MS/MS analyses by a simple SPE purification, using non-endcapped C18 cartridges, according to a modified version of a published method (Borts and Bowers, 2000). Buffer A (Table 3) was added to the supernatant fractions from the centrifuged reaction mixtures, together with the internal standard, and the samples were mixed thoroughly. The cartridges were preconditioned with 1 ml of methanol, 1 ml of water, and 1 ml of buffer B (Table 3). After loading the samples, the cartridges were rinsed with 1 ml of buffer B followed by 1 ml water. The samples were eluted by 500 μl of 100% methanol, which was then evaporated to dryness in a dry bath at 60°C under nitrogen. Finally, the dry residues were dissolved in 50 μl of the LC-MS/MS eluent mixture A/B (9:1; Table 3).

Analytical Methods. Protein concentrations were determined by the BCA Protein Assay Kit (Pierce Chemical, Rockford, IL), using bovine serum albumin as standard. Glucuronide conjugates were detected by LC-MS/MS based on the monitoring of two structure-specific precursor-product ion pairs (multiple reaction monitoring, MRM) per each steroid glucuronide. The liquid chromatographic set-up consisted of a Hewlett-Packard 1100 binary pump equipped with an autosampler, and the mass spectrometer was an API3000 triple quadrupole instrument (MDS Sciex, Concord, ON, Canada) with a turbo ionspray source (Table 3). Collision gas and lens offset voltages were optimized using either [M + NH₄]⁺ or [M + H]⁺ as precursor ions to yield maximum intensity of two characteristic product ions (dwell time 350 ms) per analyte. Formation of characteristic ions, mainly those resulting from the cleavage of glucuronide moiety with further losses of one or two water molecule(s), was in good agreement with the steroid glucuronides discussed earlier (Kuuranne et al., 2000). The data were collected and processed with Analyst 1.1 software (Applied Biosystems/MDS Sciex).

Results

Detection of Glucuronides by LC-MS/MS. The glucuronidation of anabolic androgenic steroids and their phase I metabolites (Table 1) was assayed in this study using LC-MS/MS (Kuuranne et al., 2003). Steroid glucuronides eluted within 2 to 7 min, and each was monitored with two precursor ion-product ion transitions (Fig. 1). Due to a lack of reference material, relative glucuronidation rates were determined as LC-MS/MS peak area ratios in comparison to an isotope-labeled internal standard (5β-LMTG, 17α-CD₃-labeled, and glucuronide-conjugated analog of 5β-MT). The estimation of steroid glucuronide formation, and the activity of individual UGTs, was semiquantitative. It was performed by calculating the ratio of the analyte peak area to the peak area of the internal standard (r = A_A/A_IST). The r values were transformed to symbols, from + to +++++, in which an additional + means about a 10-fold increase in A_A/A_IST. Our previous work with reference steroid glucuronides (T. Kuuranne and R. Kostiainen, unpublished observations) revealed that the variations of the MS response factors between structurally closely similar steroid glucuronides is 2.2- to 4.8-fold. Hence, the effects of possible inaccuracies caused by different MS responses are relatively small in comparison to the steps in this +/+++++ scale.

Glucuronidation of Steroid Aglycones by Individual UGTs. Recombinant human UGTs were expressed in baculovirus-infected insect cells, as described recently (Kurkela et al., 2003). The membranes fraction of the cells was used in the assays described here, 50 μg of protein per 100 μl of reaction mixture. The recombinant UGTs carried a short extension, including a His-tag, at their C terminus, i.e., the end of the cytoplasmic tail. It may be noted here that despite the different membranes and the presence of that C-terminal extension, recombinant UGT1A9 exhibited the same kinetics of entacapone glucuronidation as its counterpart in human liver microsomes (Kurkela et al., 2003). Furthermore, the presence of such a tag in all the enzymes enabled estimation of the expression level of each enzyme using anti-His-tag antibodies, something that cannot be done when the anti-UGT1A and anti-UGT2B7 antibodies are used for such a purpose.

The lipophilicity of the aglycones requires organic solvents to be included in the assay mixtures. However, solvents may inhibit enzyme activity and lead to underestimation of the activity of particular UGTs. Due to previous studies in which 10% methanol was used as the solvent, we carried out the first set of activity test trials under such an experimental set-up. However, the poor activity UGT2B15 under those conditions suggested that this isoform may be sensitive to methanol, and we have examined this directly. The results presented in Fig. 2 indicate that 10% DMSO is much more suitable for such analyses than either 5 or 10% methanol or similar concentrations of ethanol (not shown). Subsequently, the entire set of experiments was carried out in the presence of 10% DMSO, and these results are given in Table 4. It may be added, however, that we have noticed significant differences among the recombinant UGTs in their solvent sensitivity and thus far did not find a single solvent at a given concentration that is best for all the UGTs.

Eleven recombinant human UGTs were included in this study, namely UGT1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, and 2B15. All the enzymes, except UGT1A6 and 1A7, glucuronidated several of the tested aglycones (Table 4). It should be noted here that in our laboratory UGTs 1A6 and 1A7 were active toward suitable substrates such as α-naphthol and scopoletin (UGT1A6; Kurkela et al., 2003) or scopoletin and entacapone (1A7; M. Kurkela and M. Finel, unpublished results). For example, in the presence of 500 μM entacapone, the activity of recombinant UGT1A7 was 120 nmol of entacapone glucuronide per minute per milligram of membrane protein. Hence, the current lack of detectable activity of UGTs 1A6 and 1A7 toward any of the anabolic androgenic steroids and their phase I metabolites is meaningful.

Inspection of the substrate specificity and the relative activity rates of the recombinant enzymes (Table 4) suggests that, with the exception of UGT1A10, they can be divided into two main groups. Group A members, UGTs 1A1, 1A8, 1A9, and 2B15, were more selective, and mostly exhibited lower relative activity toward the tested compounds in comparison to group B enzymes, namely UGTs 1A3, 1A4, 2B4, and 2B7. The selectivity difference is particularly high when the compounds on the left-hand side and the center of Table 4 are examined. In addition to these two groups, the substrate preference of UGT1A10 suggests that it constitutes a group of its own, as discussed below. The group A enzymes, among which UGT2B15 was the sole member of the 2B subfamily, mainly glucuronidated aglycones carrying a 17β-hydroxyl group [testosterone (TES), nandrolone (NAN), 5α-1-ME, and 5α-A]. It was somewhat surprising that the substrate profile of 2B15 in this study (Table 4) did not resemble the two other representatives of the 2B subfamily, but rather those of UGT1A8 and 1A9, which are highly homologous to each other (Tukey and Strassburg, 2001). Further studies are required to test whether or not the substrate specificity of UGT2B15 was negatively affected by both its high sensitivity to solvents (Fig. 2) and its relatively low expression level (see Materials and Methods).

In addition to 17β-hydroxyl moiety, the UGTs of group B also glucuronidated the 3α-hydroxyl group, and thus most of the compounds within the current set of aglycones, at relatively high rates. With few exceptions, the conjugation of anabolic steroids and their derivatives by these enzymes, UGTs 1A3, 1A4, 2B4, and 2B7, was largely unaffected by the 3α-, 17α-, or 17β-position of the hydroxyl group of the substrate (Tables 1 and 4). A clear difference between UGT1A3 and 1A4 on one hand, and 2B4 and 2B7 on the other hand, was observed when 17β-methyl-5β-androst-4-ene-3α,17α-diol (5β-EPIM) was used as the aglycone. Within group B both enzymes of the 2B subfamily, 2B4 and 2B7, glucuronidated this substrate, whereas only UGT1A3 of the 1A subfamily produced detectable amounts of glucuronide-conjugated 5β-EPIM (Table 4). UGTs 1A3 and 1A4 are rather highly homologous to each other in their primary structure, and their specific activity toward nandrolone was similar (Table 4). In light of the latter similarity, it is interesting to note that UGT1A3 was more efficient than UGT1A4 in testosterone glucuronidation (Table 4). The apparent high specificity of UGT1A3 toward testosterone was further supported by the observation that UGT1A4 exhibited higher activity than UGT1A3 toward 5α-MT, 5β-MT, and 5α-1-ME, thereby indicating that the enhanced testosterone glucuronidation by UGT1A3 was not merely the result of a higher expression level. In the case of nandrolone derivatives, the results of this study show that UGTs 1A3, 1A4, and 2B4 readily glucuronidated 5β-estran-3α-ol-17-one (5β-N), while their activity toward 5α-estran-3α-ol-17-one (5α-N) was significantly lower (Table 4). In this respect, UGT2B7 differed from the other members of group B enzymes since it glucuronidated both nandrolone metabolites at about the same rate (Table 4).

Perhaps the most interesting human UGT, as far as the glucuronidation of anabolic steroids is concerned, is 1A10. This extrahepatic enzyme is the only one that exhibited both high activity rates and high substrate selectivity (Table 4). In particular, all the aglycones without a sterically hindered 17β-OH group (TES, NAN, 5α-1-ME, and 5α-A) were readily conjugated by this enzyme. In addition, UGT1A10 could transfer glucuronic acid to the 3α-hydroxyl group of 5β-N, and most probably to that position in 5β-EPIM as well.

Steroid Glucuronidation with Human and Rat Liver Microsomes. Methyltestosterone (MT) was the only compound in the current study that was not glucuronidated by any of the recombinant human UGTs (Table 4). It may be relevant to note that nicotine and cotinine are glucuronidated by human liver microsomes, but not by individual recombinant human UGTs (Ghosheh and Hawes, 2002; Nakajima et al., 2002). To examine whether the same is true for methyltestosterone, we have examined the glucuronidation of the current set of aglycones by two different liver microsome preparations. Induced rat liver microsomes were compared to human liver microsomes, and the results are presented in Fig. 3. Methyltestosterone glucuronidation activity could be detected in human liver microsomes, but it was barely above the detection limit. The specific glucuronidation activities in microsomes toward most of the substrates examined in this study were clearly higher than in individual recombinant UGTs. In line with this observation, the very low methyltestosterone glucuronidation activity of human liver microsomes may indicate that such an activity in the recombinant sample, if present, would be below the detection limit of the assay method used in this work. Interestingly, methyltestosterone glucuronidation activity in rat liver microsomes was significantly higher than in the human counterpart, but also in this case the presence of a methyl group in the vicinity of the hydroxyl group at position C-17 of the steroid backbone drastically reduces glucuronidation rate (Fig. 3 and Table 4).

The glucuronidation assays using liver microsomes revealed that quantitatively, per milligram of membrane protein in the assay, the recombinant enzymes are significantly less active than the native ones. There may be several reasons for this, ranging from expression level of specific UGTs to effect of specific membrane lipids, and others. One way to exclude the effect of expression level on the apparent rates is to compare the relative activity of a given enzyme preparation toward different aglycones. To this end, 5α-A glucuronidation activity by the examined enzyme was taken as a reference (=1.00), and their activities toward other aglycones were calculated as the ratio between the glucuronidation rates of the specific aglycone and 5α-A. UGTs 1A3, 1A4, 2B4, 2B7, 1A10, and the human liver microsomes were subjected to such an analysis, and the results are presented in Table 5. The substrates in this table are arranged according to the relative rates of their glucuronidation by human liver microsomes. Inspection of the data reveals that the relative aglycone preference of the liver UGTs 1A3, 1A4, 2B4, and 2B7 generally, although not fully, resembles that of human liver microsomes. However, the substrate profile of the extrahepatic enzyme UGT1A10 is largely different, particularly with respect to glucuronidation of nandrolone and 5α-1-ME (Table 5).

Discussion

In this study we have examined the glucuronidation activity of 11 recombinant human UGTs, human liver microsomes, and rat liver microsomes toward a selection of anabolic androgenic steroids. The substrates represent the parent compounds and metabolites found in human urine after dosing of metandienone, metenolone, methyltestosterone, nandrolone, and testosterone, all of which are often misused by sportsmen.

LC-MS/MS analysis with electrospray ionization offers a rapid and straightforward procedure for direct measurement of steroid glucuronides, and they were measured with two characteristic fragment ions per each analyte. Despite the lack of commercially available reference material, the use of a deuterium-labeled internal standard enabled semiquantitative comparisons of the results for individual aglycone among the various UGTs. In our experience, the minor differences between mass spectrometric response factors of structurally similar steroid glucuronides allow a reliable comparison between them.

Three of the 11 compounds used in this study, namely 5α-N, 5β-N, and 5α-ME, had a single potential glucuronidation site, the α-oriented hydroxyl at C-3 (Table 1). It has previously been reported that reduction at C-3 predominantly yields the 3α-isomers that are excreted as glucuronides regardless of the 5α- or 5β-configuration (Gower et al., 1995). Four of the tested compounds, MT, 5α-MT, 5β-MT, and 5β-EPIM, were alkylated at C-17. In MT, the only site for glucuronidation is the 17β-oriented hydroxyl group adjacent to the 17α-methyl substitution that was previously proposed to act as a sterical hindrance for glucuronidation (Schänzer, 1996). The current study also included TES, NAN, and 5α-1-ME, in which the only potential glucuronidation site is the 17β-oriented hydroxyl group. The last compound among the 11 aglycones was 5α-A, and it was unique in having both a 3α-hydroxyl group and a 17β-hydroxyl group without a sterical hindrance, i.e., two potential glucuronidation sites.

Among the individual recombinant UGTs, 1A8, 1A9, and 2B15 appeared to be specific for 17β-O-glucuronidation, while most of the other isoforms did not exhibit a clear preference for either the 3α- or the 17β-position of the hydroxyl group (regioselectivity; Table 4). In addition, each of the tested UGTs that were active in the formation of 3α-O-glucuronides also catalyzed the production of 17β-O-glucuronides (Table 4). Another case for regioselectivity is the endogenous testosterone metabolite, 5α-A, which in our hands was intensively conjugated by most UGTs (Table 4). It was previously reported that in the body 5α-A is mainly conjugated at the 17β-OH position (Rittmaster et al., 1988; Beaulieu et al., 1996), but studies with recombinant enzymes also revealed 3α-O-glucuronides (Jin et al., 1997). In this study, the close similarity of glucuronidation behavior of 5α-A with TES, NAN, and 5α-1-ME is in good agreement with 17β-O-glucuronidation. We hope that in the future we will have sufficient amounts of 5α-A-glucuronide for NMR studies that can resolve this question.

Methyltestosterone was not glucuronidated by any recombinant human UGT, and only scarcely by human liver microsomes (Fig. 3 and Table 4). This observation strongly suggests that the adjacent alkyl substitution sterically hinders conjugation of the 17β-hydroxyl group, and the same probably holds for the epimeric structure of 5β-EPIM. Another potential explanation may be the inductive effect of electropositive methyl substitution that pushes the electrons toward the steroid ring, thus making the glucuronidation reaction unfavorable.

Stereoselective glucuronidation of the endogenous 5α/β-diastereomeric pair, androsterone and etiocholanolone (5α-H- and 5β-H-structure, respectively), by UGT2B7 has been reported previously (Jin et al., 1997). The orientation of the proton has a dramatic effect on the steroid ring structure, as the A/B-cis junction of 5β-steroids changes the spatial ring geometry into a sharply bent form. Furthermore, the 3α-bond is equatorial in 5β-steroids, but axial in 5α-steroids (Kirk and Marples, 1995). The set of aglycones that was used in the current study included two diastereomer pairs, the nandrolone metabolites 5α-N and 5β-N and the methyltestosterone metabolites 5α-MT and 5β-MT (Table 1). The results demonstrated that glucuronidation of 5β-N was mostly performed at higher rates (UGT1A1, 1A3, 1A4, 2B4), and by more UGTs (also by UGT1A10), than glucuronidation of 5α-N (Table 4). However, glucuronidation of the diastereomer pair of methyltestosterone metabolites 5α-MT and 5β-MT did not show a similar pattern of stereoselectivity.

Liver is often considered the main site of steroid glucuronidation (Hum et al., 1999)—reactions that are catalyzed primarily by members of the UGT2B subfamily. We have observed that several of the hepatic UGTs among our collection of recombinant enzymes, including UGTs 1A3, 1A4, 2B4, and 2B7, provide a qualitatively reliable model for the human liver, at least as far as androgen glucuronidation is concerned (Table 5). Nevertheless, glucuronidation is also performed in other tissues (Tukey and Strassburg, 2000), and in this respect, the results about the activity of UGT1A8, and particularly UGT1A10 (Table 4), are of special interest since they shed new light on the possible involvement of extrahepatic UGTs in the metabolism of anabolic androgenic steroids. Furthermore, these results open new avenues for the study of structure-function relationships among UGTs, because isoforms 1A7, 1A8, 1A9, and 1A10 are highly homologous to each other (Tukey and Strassburg, 2001), although their activities toward anabolic androgenic steroids differ remarkably (Table 4). It may be noted here that the inactivity of UGT1A6 toward steroids is not surprising since this enzyme is considered rather specific for small phenols (Green et al., 1998). However, the lack of detectable activity in the current assays for UGT1A7 was unexpected. In retrospect, this observation may provide some indications about the possible role of specific amino acids within the N-terminal half of the protein in binding different steroids.