Systematic Error in Gas Chromatography-Mass Spectrometry–Based Quantification of Hydrolyzed Urinary Steroids

There are many naturally occurring steroids that are excreted mainly through the urine by their water-soluble conjugates formed by the substitution of 3- or 17-hydroxyl groups with either sulfate or β-glucuronide (1), and their direct measurements have been introduced to clinical studies (2-5). Gas chromatography-mass spectrometry (GC-MS)–based profiling is a proven technique in steroid analysis, whereas immunoassays have limited applicability due to cross-reactions (6, 7). However, GC-MS–based techniques mainly require the hydrolysis of steroid conjugates due to their low volatility before instrumental analysis (8-11), and the need to treat samples with one of two enzyme solutions, β-glucuronidase and a mixture of β-glucuronidase/arylsulfatase, which are extracted from Escherichia coli and Helix pomatia, respectively. The enzyme solution of Helix pomatia is used widely to produce deconjugated steroids because sulfate conjugates are not hydrolyzed by E. coli (12-16).

However, unexpected transformations of steroids during hydrolysis could obstruct the analysis. The conversion of 3β-hydroxy-5-ene steroids leads to both 3-oxo-4-enes as the major products and 6-oxy metabolites as the minor ones with Helix pomatia (14, 16-21). These two actions suggest that they are caused by the presence of 3β-hydroxysteroid dehydrogenase/Δ_5-4-ene steroid isomerase and 6-hydroxylase as additional enzymes (17, 18, 20) in the Helix pomatia extracts. The generation of 3-oxo-4-ene steroids might be converted by cholesterol oxidase because 3β-hydroxysteroid dehydrogenase does not require oxygen (16, 17, 21-24), but this has not been proven. The variability of the selectivity and reactivity of Helix pomatia is also affected by the reaction temperature, incubation time pH, and amount of enzyme added (14, 25-27).

Metabolite profiling in biological fluids can help understand the metabolic perturbation of biological systems with comprehensive insight by comparing many metabolites of individual or populations simultaneously (28-31). However, the unwanted transformation of steroids during hydrolysis can result in an underestimation or overestimation of their concentrations. Because most of the cancer progression is correlated with either the inhibition or promotion of targets in specific molecular pathways, the tests of multiple biomarkers would be focused on metabolic pathways and not on a single molecule (32). Although the analytic conditions for this particular study were optimized, the relative amounts of biomarkers are affected by the presence of unrelated and noncancer cell type from the samples. In turn, this would ultimately provide a means to establish a threshold concentration in absolute terms, beyond which a test sample would be deemed to contain biomarker levels indicative of the disease state.

Understanding the actions of steroid hormones in mammary carcinogenesis is critical for developing methods of diagnosing, preventing, and treating breast, thyroid, and prostate cancers (33-39). In addition, the exact quantification is at the center of clinical applications; without reliable methods to accurately quantify differentially expressed biomolecules, it would not be possible to identify disease biomarkers. Here, we describe the unexpected transformation phenomena of urinary steroids, including androgens, estrogens, corticoids, progestins, and sterol, in the presence of two difference enzyme systems to establish experimental protocols in GC-MS–based quantitative steroid profiling.

The 84 steroids examined in this study (Table 1) were obtained from Sigma-Aldrich, Steraloids, and NARL. The internal standards (IS), 16,16,17-d₃-testosterone and methyltestosterone for 25 androgens, 2,4,16,16-d₄-estradiol for 17 estrogens, 9,11,12,12-d₄-cortisol for 23 corticoids, 2,2,4,6,6,17α,21,21,21-d₉-progesterone and 2,2,4,6,6,21,21,21-d₈-17α-hydroxyprogesterone for 14 progestins, and 2,2,3,4,4,6-d₆-cholesterol for 5 sterols were purchased from NARL and C/D/N isotopes.

Sodium phosphate monobasic (reagent grade), sodium phosphate dibasic (reagent grade), sodium acetate (reagent grade, anhydrous), acetic acid (glacial, 99.99+%), and L-ascorbic acid (reagent grade) was obtained from Sigma-Aldrich. A solution of β-glucuronidase/arylsulfatase from Helix pomatia [aqueous solution stabilized with 0.01% thiomerosal: β-glucuronidase (100,000 Fishman U/mL) and sulfatase (800,000 Roy U/mL)] and a 50% glycerol solution of β-glucuronidase extracted from E. coli (140 U/mL) were obtained from Roche Diagnostics GmbH. The trimethylsilylating agents, N-methyl-N-trifluorotrimethylsilyl acetamide (MSTFA), ammonium iodide, and dithioerythritol, were purchased from Sigma. All organic solvents used in analytic and high performance liquid chromatography grade were purchased from Burdick & Jackson. The deionized water was prepared using a Milli-Q purification system.

The quantitative metabolite profiling of urinary steroids was achieved based on previous reports (11, 40). Briefly, the urine samples (2 mL) were added to 20 μL of the 7 ISs (1 μg/mL d₃-testosterone and d₄-estradiol; 5 μg/mL d₄-cortisol and d₈-17α-hydroxyprogesterone; and 10 μg/mL methyltestosterone, d₉-progesterone, and d₆-cholesterol). The samples were extracted with Oasis HLB SPE cartridges placed in a device fitted with a small peristaltic pump and operated at a low flow rate (<1 mL/min). After loading the sample onto the cartridge, it was washed with 2 mL water and eluted twice with 2 mL of methanol. The combined eluate was evaporated under a nitrogen stream for the next two different enzymatic hydrolysis steps. (a) For the hydrolysis of both glucuronide and sulfate conjugates, the dried eluate was added to 1 mL of 0.2 mol/L acetate buffer (pH 5.2), 100 μL of 0.2% ascorbic acid, and 50 μL of β-glucuronidase/arylsulfatase solution. The resulting mixture was then incubated at 55°C for 3 h. (b) To hydrolyze glucuronide conjugates only, the dried eluate was added to 1 mL of 0.2 mol/L acetate buffer (pH 7.2), 100 μL of 0.2% ascorbic acid, and 50 μL of β-glucuronidase and then incubated at 55°C for 1 h. After enzymatic hydrolysis, the solution was extracted twice with 2.5 mL of ethyl acetate/n-hexane (2:3, v/v). The organic solvent was evaporated in an N₂ evaporator at 40°C and further dried in a vacuum desiccator over P₂O₅-KOH for at least 30 min. Finally, the dried residue was derivatized with N-methyl-N-trifluorotrimethylsilyl acetamide/ammonium iodide/dithioerythritol (40 μL; 500:4:2, v/w/w) at 60°C for 20 min, and 2 μL of the resulting mixture were subjected to GC-MS in selected-ion monitoring (SIM) mode.

Each stock solution of the reference standards was prepared at a concentration of 1,000 μg/mL in methanol and the working solutions were made up with methanol at various concentrations ranging from 0.1 to 10 μg/mL. l-ascorbic acid (1 mg/mL) was used to prevent the oxidation of catechol estrogens. All standard solutions were stored at − 20°C until needed and they were stable for a minimum of 3 mo. The urine samples for the calibration and quality control were prepared in house as steroid-free urine (41).

The pure steroid standards were examined individually to confirm the conversion phenomena of steroids in the presence of β-glucuronidase/arylsulfatase and β-glucuronidase solutions only. After evaporating the standard solution added to known amounts, the dried standard was incubated, extracted, and derivatized using the methods described above. Acquisition was done in scan mode (m/z 100-650) to detect the by-products, and their peak identification was achieved by comparing the retention times and matching the mass spectra with those of the reference standards.

Each working solution of steroids was prepared at six different concentrations (1, 5, 20, 50, 100, and 200 ng/mL) to examine the calibration linearity of the deconjugated steroids treated with 50 μL of the enzyme solutions. For the within-day repeatability, triplicates were analyzed, whereas the reproducibility was measured from the samples run over 4 different days. In addition, the unexpected transformation of steroids was evaluated using a steroid-profiling procedure in the same urine samples obtained from two healthy male and female volunteers (ages 21 and 20 y, respectively) in triplicate. The differences in the steroid concentrations derived from the two different enzymes are represented as fold units by dividing the concentrations of β-glucuronidase/arylsulfatase by those of β-glucuronidase. Data processing and illustration were carried out using Microsoft Office Excel 2007 (Microsoft Corp.) and SigmaPlot (version 10.0, Systat Software, Inc.).

GC-MS analysis was carried out using an Agilent 6890 Plus gas chromatograph interfaced with a single-quadrupole Agilent 5975 MSD. The electron energy was 70 eV and the ion source temperature was 230°C. Each sample (2 μL) was injected in split mode (10:1) at an injector temperature of 280°C and was separated through an Ultra-1 capillary column (25 m × 0.2 mm i.d., 0.33 μm, film thickness; Agilent Technologies). The oven temperature was initially 215°C, which was ramped to 260°C at 1°C/min and then finally increased to 320°C (hold for 1 min) using a 15°C/min ramping program. Ultrahigh purity helium was used as the carrier gas with a column head pressure of 210.3 kPa (column flow, 1.0 mL/min at an oven temperature of 215°C). For quantitative analysis, the characteristic ions of each steroid were determined as their trimethylsilyl (TMS) derivatives. Peak identification was achieved by comparing the retention times and by matching the peak height ratios of the characteristic ions (Table 1).

During enzymatic hydrolysis using β-glucuronidase/arylsulfatase of Helix pomatia, many transformation phenomena were observed and the calibration linearity of some steroids was in a narrow dynamic range. This is in contrast to β-glucuronidase of E. coli, which did not lead to by-products, and the signal of the substrate at concentrations <1 ng/mL was also detectable. To confirm these unwanted transformation phenomena, the pure reference standards were examined individually with β-glucuronidase/arylsulfatase and were analyzed to identify the major by-products. Because β-glucuronidase from E. coli does not lead to by-products (12-14, 16), the individual experiment with this enzyme was excluded. Among the 84 steroids monitored, 18 compounds were transformed into substrate-derived by-products. The total ion chromatograms and mass spectra of the TMS derivates of the substrates lost and by-products generated by Helix pomatia were compared (Supplementary Fig. S1). In addition to the previously reported conversion of DHEA (A-1) to A-dione, A-diol (A-2) to T, and Preg (A-3) to Prog (14, 16-20), additional transformation was found as follows: the major product of 17α-OH-Preg (A-4) was clearly consistent with a molecular ion at m/z 546 corresponding to 17α-OH-Prog. The by-products of 5α-androstane-3β,17β-diol (B-1), 5α-androstane-3β,17α-diol (B-2), and 5β-androstane-3β,17β-diol (C-1) were also observed as 5α-DHT, Epi-DHT, and 5β-DHT, respectively. These three products showed a molecular ion and major fragment at m/z 434 and m/z 405 and 419. In addition, the major product of Epi-An (B-3) showed a molecular ion of m/z 432 and fragment ions of m/z 417 and 327. According to the general scheme of the relationship between the substrate and by-products with Helix pomatia, the conversion of 3β-hydroxy-5-ene steroids to 3-oxo-4-ene steroids was previously reported (16-19). However, unexpected transformations of 3β-hydroxy-5α–reduced and 3β-hydroxy-5β–reduced steroids into 3-oxo-5α–reduced and 3-oxo-5β–reduced steroids, respectively, were newly defined in this study (Fig. 1). However, no steroids with 3α-hydroxy-5-ene, 3α-hydroxy-5α or 5β–reduced, and 3-oxo-5α or 5β–reduced structure produced any by-products, which are accordance with a previous report (16). These results suggest that Helix pomatia also contains cholesterol oxidase in addition to 3β-hydroxysteroid oxidoreductase/3-oxosteroid-4,5-ene isomerase and 6-hydroxylase, 6-hydroxysteroid as additional enzymes (16-18, 20-24).

As 3β-hydroxy-5-ene steroids, 16α-OH-DHEA (A-5) and 21-OH-Preg (A-6) were converted to 3-oxo-4-ene steroids of 16α-OH-A-dione and 21-OH-Prog, respectively. The sterol compounds, such as Chol, 24S-OH-Chol, 20α-OH-Chol, desmolsterol, and lanosterol also produced by-products, but these compounds were assumed to be less affected than the other 3β-hydroxy-5-ene steroids or 3β-hydroxy-5α–reduced steroids. (see Supplementary Fig. S1). In particular, both desmolsterol (A-10) and lanosterol (B-5), which have a double bond at C-24 in contrast to 24S-OH-Chol (A-8) or 20α-OH-Chol (A-9), showed a small amount of by-product that might be affected by steric hindrance. Among the five sterols, cholestenone as a by-product of Chol (A-7) was confirmed using the reference standard and the other four corresponding by-products were identified from their mass spectra. In addition, Iso-P-one (B-4), as a 3β-hydroxy-5α-reduced steroid, transformed into 5α-DHP, as a 3-oxo-5α–reduced steroid, resulting in two chromatographic peaks (27.25 and 28.72 minutes) because of the nonselective derivatization. Pregnane derivatives with a 20-keto-21-methyl side chain without a 17α-hydroxy or 21-hydroxy group never led to single reaction products due to the two most stable side chain conformations (42, 43). The retention times (28.13 and 29.35 minutes, respectively) and mass spectra of the two peaks were compared with those of the reference standards. As 3β-hydroxy-5β–reduced steroids were also believed to have been converted to 3-oxo-5β–reduced steroids, EpiP-one (C-2, 19.38 and 20.90 minutes) was transformed into 5β-DHP (22.71 and 24.16 minutes). Although it was not directly identified with the reference standards, the product of ββα-diol (C-3) from incubation with Helix pomatia was assumed to be a 3-oxo-5β–reduced steroid (5β-androstan-17α-ol-3-one) because the significant ion at m/z 434 was obtained in the mass spectrum of the product. In the case of estrogens, no transformation was observed, and they could be evaluated in any enzymatic hydrolysis to quantify either the glucuronide or sulfate conjugates, or both.

The effects of incubating β-glucuronidase/arylsulfatase in an acetate buffer along with increasing substrate concentrations were examined by comparing the peak height ratios of the analyte to that of the IS (Fig. 2). By plotting the analyte to IS ratio, the semiquantitative results showed the extent of the unwanted transformations described above in the presence of increasing amounts of substrate. The 3β-hydroxy-5-ene (Fig. 2A-D), 3β-hydroxy-5α–reduced (Fig. 2E-H), and 3β-hydroxy-5β–reduced (Fig. 2I-K) steroids could not be detected at concentrations as low as 1 to 20 ng/mL, whereas the corresponding by-products were generated. All the by-products derived from the substrates tended to saturate in the concentration range of 50 to 100 ng/mL (Fig. 2). This suggests that the loss of substrate and the formation of by-products from Helix pomatia are dependent on the substrate concentration becoming saturated with 50 μL of the enzyme in the present conditions. In addition, the loss of steroid and the generation of by-products could reach a saturation point, which also decreased in the presence of a competing substrate (16).

Both enzyme systems with β-glucuronidase and β-glucuronidase/arylsulfatase were applied to real urine samples obtained from two healthy male and female volunteers. The resulting concentrations of the 84 urinary steroids were compared (Supplementary Table S1), and the extraction yield of each steroid in Helix pomatia was generally higher than that of E. coli. Some urinary steroids detected in the β-glucuronidase system could not be detected in the β-glucuronidase/arylsulfatase system, which might be decomposed into 3β-hydroxy steroids, because ββα-diol, αβα-diol, Iso-P-one, and lanosterol were found in the male samples, and 20α-OH-Chol was found in female samples. Therefore, the use of β-glucuronidase/arylsulfatase can result in a lower yield of 3β-hydroxy steroids, whereas the amount of 3-oxo steroids can be overestimated. This indicates that the use of a Helix pomatia extract in steroid analysis can affect the accuracy of the assay. Although this experiment has a limitation on the small number of samples, there was a >2-fold difference in the concentrations of the 24 steroids obtained from β-glucuronidase/arylsulfatase and β-glucuronidase only. In the cases of DHEA, Epi-An, A-diol, αββ-diol, A-dione, 16α-OH-DHEA, Preg, 20α-DHP, and B (corticosterone), a >15-fold change was obtained (Fig. 3). It should be noted that these steroids may be more prominent in the sulfate conjugates than free and glucuronic conjugates.

In summary, the transformation of 3β-hydroxy-5-ene steroids into 3-oxo-4-ene steroids has been observed previously, whereas the 3α-hydroxy and 3-oxo-steroids did not produce any by-products during enzymatic hydrolysis with Helix Pomatia (Table 2). This study dealt with 84 urinary steroids, including 3β-hydroxy-5-ene steroids, which can be analyzed by GC-MS combined with hydrolysis procedures. The 3β-hydroxy-5α-reduced and 3β-hydroxy-5β–reduced steroids showed a transformation to 3-oxo-5α–reduced and 3-oxo-5β–reduced steroids, respectively. Although the use of antioxidant improves yield in some urinary steroids, it is not easy to suggest the best condition of enzymatic hydrolysis for experimental purposes. However, these results could indicate variability in different enzymatic hydrolyses combined with GC-MS–based steroid profiling in clinical applications.

No authors declared any potential conflicts of interest.

Grant Support: Intramural grant from the Korean Institute of Science and Technology, and by grants from the National R&D Program of the Korean Ministry of Education, Science and Technology and the Korean Science and Engineering Foundation (KOSEF).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.