Simultaneous quantification of cholesterol sulfate, androgen sulfates, and progestagen sulfates in human serum by LC-MS/MS.

Steroids are primarily present in human fluids in their sulfated forms. Profiling of these compounds is important from both diagnostic and physiological points of view. Here, we present a novel method for the quantification of 11 intact steroid sulfates in human serum by LC-MS/MS. The compounds analyzed in our method, some of which are quantified for the first time in blood, include cholesterol sulfate, pregnenolone sulfate, 17-hydroxy-pregnenolone sulfate, 16-α-hydroxy-dehydroepiandrosterone sulfate, dehydroepiandrosterone sulfate, androstenediol sulfate, androsterone sulfate, epiandrosterone sulfate, testosterone sulfate, epitestosterone sulfate, and dihydrotestosterone sulfate. The assay was conceived to quantify sulfated steroids in a broad range of concentrations, requiring only 300 μl of serum. The method has been validated and its performance was studied at three quality controls, selected for each compound according to its physiological concentration. The assay showed good linearity (R2 > 0.99) and recovery for all the compounds, with limits of quantification ranging between 1 and 80 ng/ml. Averaged intra-day and between-day precisions (coefficient of variation) and accuracies (relative errors) were below 10%. The method has been successfully applied to study the sulfated steroidome in diseases such as steroid sulfatase deficiency, proving its diagnostic value. This is, to our best knowledge, the most comprehensive method available for the quantification of sulfated steroids in human blood.


MS
MS/MS was applied to the detection and quantifi cation of all the sulfated steroids. The experiments were performed with a it is well-known that humans can convert intact sulfated steroids directly into other sulfated steroids without cleaving the sulfate group ( 3 ). Interestingly, lack of STS in humans, a condition known as STS defi ciency or recessive X-linked ichthyosis (RXLI), is compatible with life. Of note, the only compounds elevated in RXLI patients, independent of their age, are 5-cholesten-3 ␤ -ol-3-sulfate [cholesterol sulfate (CS)] and oxysterol sulfates ( 4 ). High levels of CS seem to be related to the skin scaling present in most of the RXLI males ( 5 ).
There is a direct relationship between blood concentrations of sulfated steroids, age, and sex. The sulfated steroid studied with more detail is DHEAS, which reaches its higher concentrations in blood at around 20-30 years of age in males and females and decreases markedly in the years thereafter ( 6,7 ).
Factors like the broad range of concentrations of sulfated steroids within a sample, or between different patients, and the chemical differences among some of them, from polar steroids like 5-androsten-3 ␤ ,16 ␣ -diol-17-one-3-sulfate [16-␣ -hydroxy-dehydroepiandrosterone sulfate (16OHDHEAS)] to more lipophilic molecules like CS, has traditionally limited the number of these compounds quantifi ed in a single analysis.
The application of LC-MS to the steroidomics fi eld allowed for the study of intact sulfated steroids without any modifi cation, replacing immunoassays. These compounds exhibit good ionization properties, especially under ESI. Additionally, their specifi c fragmentation pattern permits the use of MS/MS for detection purposes. GC-MS can be used to quantify sulfated steroids too, but only after chemical or enzymatic cleavage of the sulfated group and a further derivatization step. An important limitation of GC-MS, when applied to the analysis of sulfated steroids, is that other conjugates, particularly steroid glucuronides, can contribute to the signal of sulfated steroids. Most commercial sulfatase enzymes have glucuronidase activity too, and chemical acidifi cation cannot distinguish among sulfates and glucuronides.
In 1990, Shackleton et al. ( 8 ) reported the fi rst method for the quantifi cation of intact DHEAS in human blood by LC-MS. Yet then, they described several signals of sulfated steroids with a m/z of 369 in negative detection mode. The authors described how the analysis of a specifi c androgen sulfate is infl uenced by the presence of other structurally related compounds, and their baseline separation is therefore highly desirable. Additionally, internal standards (ISs) should be carefully selected.
Ever since then, several methods have been developed for the analysis of sulfated steroids by LC-MS. Most of these assays focus on the analysis of unconjugated steroids and Cholesterol is conjugated by the action of a sulfotransferase, which requires 3 ′ -phosphoadenosine-5 ′phosphosulfate (PAPS) as cosubstrate. CS can be desulfated by the action of STS. The lack of STS in humans dramatically increases the concentration of CS in blood.
Linearity, calibration curves, limits of detection, and limits of quantification. Linearity, calibration curves, limits of detection (LODs), and limits of quantifi cation (LOQs) were studied for each sulfated steroid. The calibration curves were obtained by spiking increasing concentrations of standard solutions prepared in a mixture of MeOH and water [1:1 (v/v)] into 300 l of in-house prepared charcoal-stripped serum and IS, as described in the previous section. The parameters for CS were studied at the following levels: 20, 40, 80, 200, 800, 2,000, 4,000, 8,000, 20,000, 40,000, 80,000, 160,000, and 480,000 ng/ml. For the other sulfated steroids, the parameters were tested for a range which included the following points: 0.25, 0.5, 1, 2.5, 10, 25, 50, 100, 250, 500, 1,000, 2,000, and 6,000 ng/ml. The calibration curves were constructed plotting the area ratio (peak area of the analyte divided by the peak area of the IS) of each calibration point against its corresponding concentration in nanograms per milliliter. LOD was determined at a signalto-noise ratio higher than 3. LOQ was the lowest concentration with a signal-to-noise ratio of 10 or more and acceptable precision and accuracy.
Quality controls. Quality control samples (QCs) containing different concentrations of each analyte were prepared and analyzed. The selected values for each sulfated steroid were in accordance with the expected concentrations reported in the bibliography ( 11,13,14 ) and with the preliminary results found in real samples prior to method evaluation . QCs were always prepared in stripped human serum, because a real serum sample lacking all the sulfated steroids to analyze could not be obtained.
Recovery. The recovery of the method evaluates the loss of sulfated steroids due to the use of C18 cartridges. It was calculated by dividing the area ratios obtained for the QCs for each analyte by the area ratio measured after spiking the same amounts of the QC and IS postextraction. The postextraction addition consisted of spiking the standards and ISs onto the methanolic fraction obtained from the workup of charcoal-treated serum, which was afterwards evaporated and reconstituted as described above .
Matrix effects. Matrix effects were studied by plotting response ratios of spike experiments prepared in charcoal-treated serum after sample preparation and reconstitution (Y axis) against the same concentrations of standards prepared in the fi nal aqueous reconstitution solution without sample workup (X axis). A line slope of one corresponds to absence of matrix effects. Slopes above one indicate that there is an ion enhancement, and slopes below one are found whenever there is ion suppression. triple quadrupole mass spectrometer (TSQ, Quantum Ultra; Thermo Fisher Scientifi c, Dreieich, Germany), operated in negative mode detection with a HESI probe.
Monitoring reactions, collision energies, and tube lens voltages are shown in Table 1 . Mass spectrometer conditions were as reported before ( 16 ). Briefl y, the capillary temperature and the vaporizer temperatures were set to 270 and 350°C, respectively. The sheath gas was 50 arbitrary units and the auxiliary gas was 20 arbitrary units. The voltage applied was Ϫ 3,500 V. The collision gas pressure was 1.5 mTorr for all experiments.

LC
Chromatographic separation of the analytes was achieved with a column Accucore Phenyl -X (100 × 2.1 mm, 2.6 m) from Thermo Fisher Scientifi c, connected to a HPLC system (Agilent 1200SL, Waldbronn, Germany).
Solvents used for separation of sulfated steroids were buffer solution A with ammonium acetate [10 mM (pH 7) dissolved in 85% water and 15% ACN] and solution B with organic solvents composed of 70% MeOH and 30% ACN (v/v). The fl ows and solvent compositions applied can be found in Table 2 . The method had a total duration of 11.8 min.

Sample preparation
Sample preparation was modifi ed from a previous protocol ( 4 ). Three hundred microliters of each serum or plasma sample were incubated with 50 l of IS mix during 15 min. Each IS in the mix had a concentration of 1 g/ml, with the exception of d 7 CS, which was 6 g/ml. Proteins were precipitated with 1 ml of ACN-ZnSO 4 [89 g/l, 4:1 (v/v)]. After protein precipitation, the samples were incubated again for 15 min and then centrifuged for 10 min at 14,500 g . The supernatant, free of proteins, was then mixed with 3 ml of water in a glass tube. For each sample, a SepPak C18 cartridge was conditioned with 2 ml of MeOH and 2 ml of water. Next, the sample, diluted in water, was loaded onto the cartridge and washed with 3 ml of water, 3 ml of hexane, and 4 ml of chloroform. Sulfated steroids were eluted from the cartridge with 4 ml of MeOH. After evaporation of the methanolic fraction with nitrogen at 40°C, the samples were reconstituted with 250 l of a solution of 79.75% water, 10% MeOH, 10% ACN, and 0.25% ammonium hydroxide. Samples were centrifuged after reconstitution and then injected in the LC-MS/MS system (10 l). Each sample was injected at least three times.

Evaluation of the method
The method was validated in accordance with the US Food and Drug Administration and the EU European Medicines Agency guidelines for bioanalytical evaluation . important considering the high concentrations of DHEAS, AnDiolS, AnS, and epiAnS in blood.

Analytical performance
The complete set of data can be found in Table 3 . The table presents the information obtained at three different QCs for each sulfated steroid (33 QCs), which are within the physiological range. Linearity (R 2 ) was always higher than 0.99. Most of the sulfated steroids showed good linearity, even at higher concentrations (data not shown). Those concentrations are not refl ected in Table  3 because they are far from physiological. Recovery was good for all the QCs, ranging between 85.5 and 111.6% with an average value of 97.9%. Most of the compounds showed no matrix effects in our assay. Nevertheless, three compounds exhibited important ionization problems due to the components of the matrix. Both PregS and 17OHPregS suffered ion suppression, which in the case of 17OHPregS was 40%. On the other hand, DHTS signal was enhanced by 20%. To minimize the problems associated with matrix effect, all the calibrations were prepared with a surrogate of the original matrix; charcoal-stripped serum. Regarding accuracy and precision, the method showed good precision and accuracy at all QCs (always below 20%), with the exception of the fi rst QC for 17OHPregS (5 ng/ml). At this concentration, and especially for inter-day experiments, the precision expressed as percent CV was close to a 24%. Precision at the LOQ for 17OHPregS was 31.1%.

Stability of the samples
Freeze-thaw cycles, tested with the same sample during three consecutive months (C9S1, C9S2, C9S3 and C10S1, C10S2, C10S3 in the supplementary table), did not change the measured concentrations of the sulfated steroids.

Study of plasma samples
Half of the blood extracted from control donors C9 and C10 was processed to obtain serum, and the other half as plasma. Plasma samples were measured following the same sample preparation as for those of serum, and their concentrations were calculated by using calibrations performed with stripped serum. The concentrations found in plasma (C9P and C10P, respectively, in the supplementary table) demonstrate that the sample preparation and the analytical method are effective to profi le sulfated steroids in plasma too.

DISCUSSION
The method described in this work shows a reliable quantitative performance for the profi ling of sulfated steroids in human serum. Most of the compounds analyzed with our assay met the standards of the FDA guidelines . The compounds with precision and accuracy below 15% for all QC levels were CS, PregS, 16OHDHEAS, DHEAS, AnDiolS, AnS, and EpiAnS. In the case of 17OHPregS, the results were higher than 15% just for precision and only at the lower level, which could be related to the Precision and accuracy. Intra-day precision [percent coefficient of variation (CV)] and accuracy [percent relative error (RE)] studies were performed by analyzing fi ve spiked replicates of each QC (Q1, Q2, and Q3) in one batch during the same day. Between-day precision and accuracy were measured in fi ve different batches during different days with the same concentration in validation samples. The concentrations were calculated by calibration curves.
Stability of the samples. Each real sample chosen for the stability test (serum samples C9 and C10 in the supplementary table) was analyzed three times to evaluate the stability of the sulfated steroids during freeze-thaw cycles. The time difference between re-analyses was 1 month.
Analysis of the samples. We analyzed 11 samples from patients with STS defi ciency, and a sample from a baby with cholestasis. All STS-defi cient patients showed the classical RXLI phenotype with scales on their skin. One of the RXLI patients was undergoing a long-term treatment with prednisolone (10 mg/day) and was especially selected because his steroid levels were expected to be low. Among the control samples, there were patients with ichthyosis vulgaris, a skin condition with a similar phenotype as RXLI, but in which steroid sulfates are normal. Other controls were healthy volunteers and the mothers of two RXLI patients.

Data analysis
Calibration curves, linearity, LODs, and LOQs were studied with the program Xcalibur 2.1, applying a 1/x or equal weighting regression. Box plots were generated with BoxPlotR ( 19 ).

MS
The ESI heated probe was tested with three different temperatures, 325°C (capillary temperature 250°C, sheath and auxiliary gas 30 arbitrary units, Ϫ 4,000 V), 350°C (capillary temperature 270°C, sheath gas 50 arbitrary units and auxiliary gas 20 arbitrary units, Ϫ 3,500 V), and 400°C (capillary temperature 300°C, sheath gas 60 arbitrary units and auxiliary gas 40 arbitrary units, Ϫ 3,500 V). Tests were performed during the same day using standards spiked in stripped serum and a sample from a healthy volunteer after sample workup. The selected parameters ( 16 ) provided the best area ratios (peak area of the sulfated steroid divided by the peak area of the IS) for the compounds which were present in lower concentrations in blood (PregS, 17OHPregS, DHTS, TS, and eTS) and for CS. For the remaining androgen sulfates, the area ratios were slightly higher at 325°C. The differences were not exception of 16OHDHEAS, all sulfated androgens exhibit a comparable polarity. A C18 reversed phase column was tested (Hypersil Gold column, 50 × 2.1 mm, 5 m; Thermo Fisher Scientifi c), showing that resolution of all the mentioned compounds was not possible with such chromatography (data not shown). The introduction of ACN in the mobile phase and the selection of a different reversed-phase chromatography with phenyl groups allowed for effective separation of all the compounds of interest ( Fig. 2 ). To ensure the analysis was always correct, we added deuterated ISs to qualitatively identify some of the sulfated steroids never studied before in blood by LC-MS/MS (d 3 eTS, d 3 TS, d 3 DHTS, and d317OHPregS). The chromatograms obtained for each analyte in the sample from a healthy volunteer (control C10) are depicted in Fig. 2A . This was chosen as a representative chromatogram for an adult male, where the sulfated steroids are in high concentrations. There it can be observed how the m/z 367 from DHEAS contributes to the transition m/z 369 → 97, because DHEAS is normally present in high concentrations in blood. Interestingly, a recent method analyzed DHEAS tracing transition m/z 369 → 97 instead of m/z 367 → 97 ( 17 ). Some of the expected sulfated steroids, TS, eTS and DHTS, were not found or they were below the LOQ of important matrix effects found for this analyte. Still at Q1, precision remained below 25%. The precision and accuracy values for DHTS, TS, and eTS were always below 20%.
Previously reported methods achieved a good separation for most important androgen sulfates ( 8,15 ), showing the complexity of the signal m/z 369 in human blood. Interestingly, compounds like DHTS or TS were never analyzed before by LC-MS in human blood, probably because they are expected to be present in low concentrations ( 20 ). Additionally, there are two LC-MS assays focused on the analysis of CS in serum ( 11,12 ). Other LC-MSbased methods performed the analysis of androgen sulfates ( 13,14,17 ). Our method is the fi rst to quantify CS in combination with androgen sulfates, PregS, and 17OHPregS in a single run.
One of the goals of our work was to achieve a realistic profi le of the androgen sulfates present in human blood within a runtime that permitted the analysis of a large number of samples. To achieve this, the method should provide a good separation of all isobaric compounds: DHTS, AnDiolS, AnS, and epiAnS ( m/z 369); and TS, eTS, and DHEAS ( m/z 367). Additionally, DHEAS should be always baseline resolved, because its high concentration might affect the m/z 369 signal as well . With the Of note, for the transition m/z 369 → 97, other unidentifi ed peaks were always found with our method ( Fig. 2A ), in consonance with ( 11 ). For comparison, the retention times and peaks of commercial standards are shown in Fig. 2B .

Analysis of estrogen sulfates
Our assay showed a good separation of estrogen sulfates too, including estrone sulfate (E1S), estradiol sulfate (E2S), and estriol sulfate (E3S). These compounds appeared in our method at retention times 0.78 (E3S), 3.12 (E2S), and 4.06 (E1S), when they were spiked and measured after sample preparation. Estrogen sulfates were fi nally not included in the assay because of their poor performance. This could be attributed to the sample preparation and the physiological concentration of these compounds. The importance of sample workup for the analysis of estrogen sulfates was recently studied by Dury et al. ( 17 ). They published an LC-MS/MS method for the analysis of fi ve sulfated steroids, including E1S. In their work, they tested the importance of several factors affecting the quantifi cation of E1S, coming to the conclusion that protein precipitation with organic solvents during sample preparation modifi es the concentration of estrogen sulfates, but does not affect the rest of compounds. This is probably due to the binding of their phenolic A-ring with proteins, and in order to include E1S in their assay they used solid phase extraction.

Measurement of sulfated steroids in real samples
The method was applied to calculate the concentrations of the analytes in different serum samples. The values obtained are summarized in the supplementary table. As mentioned before, the compounds DHTS, TS, and eTS were not found in any of the samples. The sample from the RXLI patient treated with prednisolone (P2, supplementary table) showed an important decrease in the concentrations of all the sulfated steroids with the exception of CS. The CS level was similar to the rest of RXLI patients. This suggests that CS, unlike the rest of sulfated steroids, is not linked to the production of steroids by the adrenal glands.
The sample from the patient with cholestasis (P1) refl ected a big increase in most of the sulfated steroids. The signal for AnDiolS could not be resolved because several interferences were found at the retention time of the peak. The rest of sulfated androgens were low or not found. This same sample was measured before to profi le its levels of oxysterol sulfates, resulting in a high elevation of hydroxycholesterol sulfates ( 4 ). Figure 3 shows a comparison of the levels of CS between the samples analyzed. The levels of CS were about 32 times higher in RXLI patients when compared with healthy donors, in good consonance with the method of Shackleton and Reid ( 11 ). The concentration of CS in the serum sample from the baby with cholestasis was three times lower than those of the RXLI patients .
We studied different samples to test our assay. The analysis of serum from patients with STS defi ciency has proven its utility for the diagnosis of this condition based on the levels of CS, even when compared with cholestasis. The profi ling of sulfated steroids in plasma samples revealed that our method can be successfully applied to their quantifi cation in that matrix as well.