Multi-steroid profiling by uPLC-MS/MS with post-column infusion of ammonium fluoride ==================================================================================== * Lina Schiffer * Fozia Shaheen * Lorna C. Gilligan * Karl-Heinz Storbeck * James Hawley * Brian G. Keevil * Wiebke Arlt * Angela E. Taylor ## Abstract **Background** Multi-steroid profiling is a powerful analytical tool that simultaneously quantifies steroids from different biosynthetic pathways. Here we present an ultra-high performance liquid chromatography-tandem mass spectrometry (uPLC-MS/MS) assay for the profiling of 25 steroids using post-column infusion of ammonium fluoride. **Methods** Following liquid-liquid extraction, steroids were chromatographically separated over 5 minutes using a Phenomenex Luna Omega C18 column and a water (0.1 % formic acid) methanol gradient. Quantification was performed on a Waters Acquity uPLC and Xevo® TQ-XS mass spectrometer. Ammonium fluoride (6 mmol/L, post-column infusion) and formic acid (0.1 % (vol/vol), mobile phase additive) were compared as additives to aid ionisation. **Results** Post-column infusion (PCI) of ammonium fluoride (NH4F) enhanced ionisation in a steroid structure-dependent fashion compared to formic acid (122-140% for 3βOH-Δ5 steroids and 477-1274% for 3-keto-Δ4 steroids). Therefore, we fully analytically validated PCI with NH4F. Lower limits of quantification ranged from 0.28 to 3.42 nmol/L; 23 of 25 analytes were quantifiable with acceptable accuracy (bias range −14% to 11.9%). Average recovery ranged from 91.6% to 113.6% and average matrix effects from −29.9% to 19.9%. Imprecision ranged from 2.3% to 23.9% for all analytes and was <15% for 18/25 analytes. The serum multi-steroid profile of 10 healthy men and 10 healthy women was measured. **Conclusions** uPLC-MS/MS with post-column infusion of ammonium fluoride enables comprehensive multi-steroid profiling through enhanced ionisation particularly benefiting the detection of 3-keto-Δ4 steroids. **Highlights** * This multi-steroid profiling assay quantifies 25 steroids in 5.5 minutes * Post-column infusion of NH4F enhances the ionisation of 3-keto-Δ4 steroids * The assay simultaneously quantifies steroids from several biosynthetic pathways * We present analytical data validated for serum steroid profiling Key words * steroid quantification * 11-oxygenated androgens * liquid chromatography-tandem mass spectrometry * androgens * mineralocorticoids * glucocorticoids ## 1. Introduction Steroid hormones are biosynthesised in the adrenal cortex and gonads via cascade-like, interlinked enzymatic pathways and undergo extensive metabolism with both activation and inactivation in peripheral tissues (Figure 1) leading to a complex circulating steroid metabolome [1]. Steroid flux through the different pathways can be severely dysregulated in various conditions, requiring comprehensive steroid assessment to develop a mechanistic understanding of the condition and for diagnosis and treatment monitoring. Disorders of steroidogenesis with distinct steroid metabolome profiles include inborn enzymatic deficiencies such as congenital adrenal hyperplasia, autonomous adrenal steroid production such as Cushing syndrome or primary aldosteronism [2], adrenocortical carcinoma [3, 4], polycystic ovary syndrome and idiopathic intracranial hypertension [5-7]. Steroid-dependent cancers, such as prostate cancer, can locally activate steroids and are treated with various pharmacological or surgical approaches to deplete the relevant steroids and their precursors in circulation [8]. ![Figure 1:](http://medrxiv.org/http://medrxiv.stage.highwire.org/content/medrxiv/early/2022/03/09/2022.03.08.22271681/F1.medium.gif) [Figure 1:](http://medrxiv.org/content/early/2022/03/09/2022.03.08.22271681/F1) Figure 1: Pathways of adrenal steroidogenesis (A) and androgen biosynthesis (B). General steroid precursors are shown in yellow, mineralocorticoids in green, glucocorticoids in orange, androgens in blue. Dark shades of each colour indicate active steroids, light shades inactive precursors. Abbreviations for steroids included in the uPLC-MS/MS assay are shown in bold. Traditionally, individual steroids have been used for the assessment of a suspected underlying condition (e.g. testosterone (T) for female androgen excess [9]; 17α-hydroxyprogesterone (17OHP) for congenital adrenal hyperplasia [10]). Mechanistic in-vitro studies often rely on the measurement of the end products of a biosynthesis pathway only. However, the determination of a multi-steroid profile has several advantages over the use of selected individual markers. First, steroid precursor/product ratios can be calculated to identify distinct effects on individual enzymes [11] and to assess the degree of steroid precursor activation in conditions of steroid excess. For example, in serum the ratio of T to its less active precursor androstenedione (A4) and the ratio of T to the more potent androgen 5α-dihydrotestosterone (DHT) have been shown to be excellent markers for androgen excess and the associated adverse metabolic phenotype in polycystic ovary syndrome [5, 12, 13]. Secondly, multi-steroid profiling enables different pathways of steroidogenesis to be assessed simultaneously, which is essential due to the interlinked nature of steroidogenesis and the contribution of individual enzymes to several pathways (Figure 1). Thirdly, multi-steroid profiling can be used to investigate off-target effects of inhibitors of steroidogenesis. Finally, multi-steroid profiling can be combined with machine learning approaches to generate powerful, unbiased and automated diagnostic steroid metabolomics tools [3, 4]. In contrast to immunoassays, mass spectrometry allows for multiplexing of analytes and liquid chromatography for the high-throughput that is required to effectively use multi-steroid profiling in clinical and research laboratories. Steroid analysis by mass spectrometry is often limited by poor sensitivity due to low analyte concentrations in biological samples and low ionisation efficiency of the analytes. Mobile phase additives can improve the chromatography (peak separation and shape), enhance the signal and, thereby, sensitivity. Formic (methanoic) acid and acetic acid are common additives for corticosteroids and androgen analysis in the positive ionisation mode [5, 14, 15] as they can promote the formation of protonated molecular ions (M+H)+. Ammonium fluoride (NH4F) can aid the ionisation of steroids in electrospray ionisation (ESI) in negative mode [16] and hence improve the sensitivity of oestrogen measurements [17, 18]. Additionally, NH4F has been reported to augment the ionisation of steroids using ESI in positive ion mode in a structure-dependent manner, when coupled to supercritical fluid chromatography [19]. Here, we present an ultra-high performance liquid chromatography-tandem mass spectrometry (uPLC-MS/MS) assay to measure 25 steroids (Table 1) from the mineralocorticoid, glucocorticoid, androgen, alternative and 11-oxygenated steroidogenic pathways. The separation was achieved in 5 minutes using post-column infusion of NH4F in positive ionisation mode to enhance ionisation. View this table: [Table 1:](http://medrxiv.org/content/early/2022/03/09/2022.03.08.22271681/T1) Table 1: Nomenclature of the 25 steroids analysed in this uPLC-MS/MS assay ## 2. Material and methods ### 2.1 Preparation of external standards and quality controls Pure standards of all analytes were purchased as powders from Sigma-Aldrich (Gillingham, UK) and Steraloids (Newport, USA); for details see Supplemental Table 1. The purity of all standards was confirmed by gas chromatography-mass spectrometry. Individual stock solutions at 1 mg/mL were prepared in uPLC grade methanol (Biosolve, Dieuze, France) and stored at −80 °C. Using the stock solutions, combined calibrators for all analytes were prepared by spiking phosphate buffered saline (PBS) pH 7.4 supplemented with 0.1% (wt/vol) bovine serum albumin (BSA) (Sigma-Aldrich) yielding at least 10 different final concentrations from 0 up to 250 ng/mL (Supplemental Table 2). A combined internal standard stock solution containing 0.5 μg/mL of each internal standard was prepared in deuterated methanol (Sigma-Aldrich). Quality control (QC) samples were prepared by spiking PBS 0.1% BSA from independent 1 mg/mL stock solutions. Four different concentrations covering the expected concentration range human serum (Supplemental Table 2). Additionally, pooled human serum (Sigma-Aldrich) was aliquoted, stored at −80 °C and used as biological QC in each analytical run. All calibrators, internal standards and QCs were stored at −20 °C. ### 2.2 Collection of serum samples The collection of blood samples was approved by the authors’ Institutional Review Board (Science, Technology, Engineering and Mathematics Ethical Review Committee of the University of Birmingham, ERN_17-0494). Informed, written consent was obtained from all individuals included in this study. Following venous puncture, blood was collected into gold top SST Vacutainers® (Becton, Dickinson, Wokingham, UK). Samples were spun and the serum removed and stored at −80 °C until analysis. The method was applied to 20 serum samples from healthy volunteers 10 female and 10 males (20-40 years). ### 2.3 Sample preparation Steroids were extracted by liquid-liquid extraction. A set of calibrators, QCs and biological QCs were prepared with each batch of samples. Sample, calibrator, or QC (200 μL) were transferred into a hexamethyldisilazane-treated glass tube and 10 µL of the internal standard mixture containing all stable isotope labelled internal standards listed in Table S1. 50 μL of acetonitrile (Biosolve, Dieuze, FR) was added to precipitate the proteins and the samples were vortexed. 1-mL of tert-butyl methyl ether (Acros Organics, Fisher Scientific UK Ltd, Loughborough, UK) was added to each sample and the samples were vortexed at 1000 rpm for 10 minutes on a multi-vortex. The samples were incubated at room temperature for at least 30 minutes to aid phase separation. The organic phase was subsequently transferred into the wells of a 2-mL square well 96-well plate (Porvair Sciences Ltd, Wrexham, UK) containing 700 µL glass inserts (Randox, Crumlin, UK) and dried under a nitrogen stream at 45 °C. The dried extract was reconstituted in 100 µL of 50% (vol/vol) uPLC grade methanol (Biosolve) in uPLC grade water (Biosolve) prior to analysis. ### 2.4 Ultra-high performance liquid chromatography Chromatography was performed on an Acquity ultra performance liquid chromatography system (uPLC; Waters Ltd, Wilmslow, UK) using a Phenomenex Luna Omega column, 1.6 µm, polar C18, 100 Å, 2.1 × 50 mm; (Phenomenex, Macclesfield, UK) at 60 °C. 10 µL of the reconstituted sample was injected. Mobile phase A consisted of uPLC grade water (Biosolve) and mobile phase B of uPLC grade methanol (Biosolve). An optimised method with a linear gradient from 45% to 75% of mobile phase B was applied over five minutes at a flow rate of 0.6 mL/min to separate the analytes followed by a 98% wash and equilibration at starting conditions prior to the injection of the next sample. The auto sampler was maintained at 10 °C. To investigate ionisation 0.1% (vol/vol) formic acid (added to mobile phase A or both A and B) was compared to PCI of NH4F. ### 2.5 Tandem mass spectrometry The uPLC eluate was injected into a XEVO® TQ-XS mass spectrometer (Waters Ltd) using ESI in positive ion mode. Post-column infusion 6 mmol/L NH4F in 50% (vol/vol) was combined using the fluidics system on the mass spectrometer under full software control. The capillary voltage was maintained at 1.5 kV, the source temperature was 150 °C, desolvation temperature and gas flow were 600 °C and 1200 L/h, cone gas was 150 L/h. MassLynx 4.2 software (Waters Ltd) was used for systems control. Qualifier and quantifier mass transitions, cone voltages and collision energies are summarised in Table 2. TargetLynx software was used for data processing and quantification. Peak area ratios of analyte to internal standard were plotted against the nominal concentrations of the calibrators and 1/x weighting and linear least square regression were used to produce the standard curve. View this table: [Table 2:](http://medrxiv.org/content/early/2022/03/09/2022.03.08.22271681/T2) Table 2: Retention times, quantifier and qualifier mass transitions, collision energies and cone voltages of target analytes and internal standards. ### 2.6 Validation Validation was performed following protocols from published guidelines [20-22]. For each steroid, validation has been performed at different concentrations representing low, medium and high concentrations of that steroid in serum. While this method has utility for many applications, we opted to validate for a serum application. We selected serum because it is a complex mixture of steroids, proteins, salts and many other constituents. These constituents can vary between individuals and can each affect steroid extraction and ionisation efficiencies. In addition, serum contains steroids at a range of concentrations, for example low level poor ionising steroids such as DHT (<0.3 ng/mL (1 nM)) and high concentration good ionisers like cortisol (∼18 to 300 ng/mL (50-800 nM)). Optimisation using such a complex matrix means adaptation of this method to other ‘cleaner’ bio-fluids or cell culture supernatants should be simple. #### 2.6.1 Recovery and matrix effects Matrix effects and recovery were determined as previously described [23, 24]. Six different human serum samples (four male and two female) were spiked at 5 ng/mL before extraction (pre-extract) and after extraction and reconstitution (post extract). Additionally, the reconstitution solvent (50% (vol/vol) uPLC grade methanol in uPLC grade water) was spiked with the same amounts of analytes (no extract). Percentage matrix effects and recovery were calculated from the concentrations quantified in the samples: ![Formula][1] For matrix effects mean values between −15 and 15% were considered ideal and −20% and 20% were acceptable. For recovery mean values between 80% and 120% were defined acceptable. #### 2.6.2 Linearity Three calibration series were prepared by different scientists. For each analyte the ratio of analyte peak area to internal standard peak area was plotted against the nominal concentrations of the standard as described above. Calibration curves were accepted as linear if the correlation coefficient of the linear regression (R2) was ≥0.99. #### 2.6.3 Lower limit of quantification The lower limit of quantification (LLOQ) was defined as the lowest concentration for which 10 replicate samples of spiked surrogate matrix could be run with a CV <20% and a bias not greater than ±20%. #### 2.6.4 Imprecision Pooled surrogate matrix samples spiked with all analytes at four different concentrations (0.3, 1, 3 and 30 ng/mL) were run 10 times in the same batch to assess intra-assay imprecision. Pooled serum samples and pooled serum samples spiked with 5 ng/mL for all analytes were also run 10 times in the same batch to assess intra-assay imprecision. These samples were extracted and run-in different batches to assess inter-assay imprecision (n =20). A CV ≤15% was considered optimal. #### 2.6.5 Accuracy 10 samples of surrogate matrix were spiked individually at four different concentrations (0.3, 1, 3 and 30 ng/mL). A bias of the observed concentration and the nominal concentration between −15% and +15% was considered optimal. ## 3. Results ### 3.1 Chromatographic separation Analytes were separated using a linear gradient over five minutes followed by 0.5 minutes at starting conditions for equilibration resulting in a total run time injection-to-injection of 5.5 minutes. All analytes eluted as distinct, identifiable peaks (Supplemental Figure 1). Two pairs of analytes co-eluted: (cortisol (*m/z* 363) and 11KA4 (*m/z* 301); 17OHP (*m/z* 330) and 17Preg (*m/z* 332)). No interference was observed due to the differences in *m/z* (Table 2). ### 3.2 Comparison of additives to improve ionisation The effect of NH4F on signal intensity was assessed in comparison to our previously published method using 0.1% (vol/vol) formic acid as mobile phase additive in both the methanol and water phases [5, 7]. We chose to add NH4F by post-column infusion as preliminary experiments with NH4F as an additive led to increases in system pressure indicative of damage to the column material (data not shown). The intra-assay imprecision using this modification was acceptable with a mean CV of the peak area of 7.0% for all analytes (CV range 3.9 to 13.7%; n=60 injections of the same sample (data not shown)). Peak areas of all analytes in serum (n=73) were compared between the assay with formic acid in both mobile phases and the assay using post-column infusion of NH4F. NH4F increased peak area in a structure-dependent fashion (Figure 2, Supplemental Figure 2). In comparison to formic acid, NH4F induced significant increases in the peak area of steroids with 3-keto-Δ4 structure (Prog, DOC, B, Aldo, 17OHP, S, E, F, A4, 11OHA4, 11KA4, T, 11KT, 11OHT), with mean increases varying from 477% (17OHP) to 1274% (Aldo). NH4F had a lower impact on the peak area of the majority of A-ring reduced steroids with changes varying between 100% and 381% (DHT, 5α-dione, 3α-diol, An, 5α-DHP, 3α5α17HP), with the exceptions of alloP (841%) and 17OHDHP (1280%). NH4F post-column infusion had only a very minor impact on the peak areas of 3β-OH-Δ5 steroids (Preg, 17Preg, DHEA). Our final optimised method employed 6 mmol/L NH4F introduced via post-column infusion at a flow rate of 5 µL/min, with 0.1% (vol/vol) formic acid in the water mobile phase to limit the risk of microbial contamination. ![Figure 2:](http://medrxiv.org/http://medrxiv.stage.highwire.org/content/medrxiv/early/2022/03/09/2022.03.08.22271681/F2.medium.gif) [Figure 2:](http://medrxiv.org/content/early/2022/03/09/2022.03.08.22271681/F2) Figure 2: Ammonium fluoride (NH4F) post-column infusion enhances ionisation of steroids in a structure-dependent fashion as compared to mobile phase formic acid. The percentage increase in peak area of the quantifier transition when comparing post-column infusion of NH4F (6 mmol/L, 5 μL/min) to the use of 0.1% (vol/vol) formic acid in both the methanol and water mobile phase. Bars represent the mean percentage peak area and whiskers the relative standard deviation for serum samples from different individuals. Samples from all individuals with detectable peaks were included (Prog, DOC, B, 17Preg, 17OHP, S, F, E, DHEA, A4, T, DHT, An, 11KA4, 11KT, 11OHA4, 11OHT n=74; Preg n=73; Aldo n=71; 3α-adiol n=56; 5α-DHP n=48; 17OHDHP n=39; 3α5α17HP n=18; alloP n=15; 5α-dione n=14). The red line indicates a relative peak area of 100% (identical ionisation with both methods). The colour of the bars indicates the functional class of each analyte; yellow, general steroid precursor; light green, mineralocorticoid precursor; dark green, mineralocorticoid; orange, glucocorticoid precursor; red, glucocorticoid; turquoise, androgen precursor; dark blue, androgen; light blue, alternative DHT biosynthesis pathway intermediate. ### 3.3 Validation of the analytical performance The analytical performance of the optimised assay with post-column infusion of NH4F and 0.1% (vol/vol) formic acid in mobile phase A was validated. Matrix effects and recovery were assessed for six different serum samples (Table 3). Mean matrix effects ranged from −19.5% to 19.9% for all analytes and were hence in the desired range from −20% to 20%, except for 5αDHP (−29.9%), 17OHDHP (−22.6%) and 3α5α17HP (−62.7%). Mean recovery ranged from 90.0% to 113.6% for all analytes except for 3α5α17HP (122.1%), which was just outside the acceptance range of 80-120%. Calibration curves were linear with an R2 ≥0.99 for all analytes (Supplemental Table 2). Lower limits of quantification ranged between 0.1 ng/mL (∼0.3 nmol/L) and 0.5 ng/mL (∼1.5 nmol/L) for all analytes except for 17Preg and 3α-adiol, which had a limit of quantification of 1 ng/mL (corresponding to 3.01 and 3.42 nmol/L, respectively) (Table 3). The accuracy (bias) and imprecision of the assay were assessed with spiked PBS 0.1% BSA samples at four different concentrations. The bias between the observed and nominal concentrations was calculated as a measure of accuracy and the coefficient of variation (CV) as a measure of imprecision (Table 4). Bias was within acceptable limits for the majority of analytes at all concentrations above the LLOQ ranging from −14.0 to 11.9%, with the exception of 5αDHP at 0.3 ng/mL (−21%) and 17OHPreg at 1 ng/mL (−17.0%). Imprecision was within acceptable limits for the majority of analytes at concentrations above the LLOQ ranging from 2.3 to 14.7% with the exception of An (15.2% at 0.3 ng/mL), 3α-adiol (17.8% at 1 ng/mL), 17OHPreg (16.5% at 3 ng/mL) and alloP (at all concentrations imprecision ranged from 15.1 to 20.5%). View this table: [Table 3:](http://medrxiv.org/content/early/2022/03/09/2022.03.08.22271681/T3) Table 3: Validation summary of matrix effects, recovery and lower limits of quantification for all steroids. Matrix effects and recovery were assessed at 5 ng/mL (approximately 17 nmol/L) in serum samples from six different donors. The lower limit of quantification (LLOQ) was defined as the lowest concentration that can be assessed with appropriate accuracy (bias within ± 20%) and imprecision (CV < 20%). View this table: [Table 4:](http://medrxiv.org/content/early/2022/03/09/2022.03.08.22271681/T4) Table 4: Accuracy and intra-assay imprecision determined at four concentrations spiked into surrogate matrix (n=10). n/a, not applicable as concentration