Simultaneous quantitation of four androgens and 17‐hydroxyprogesterone in polycystic ovarian syndrome patients by LC‐MS/MS

Abstract Background Due to the low concentration of androgens in women and the limitation of immunoassays, it remains a challenge to accurately determine the levels of serum androgens in polycystic ovary syndrome (PCOS) patients for clinical laboratories. In this report, a liquid chromatography‐tandem mass spectrometry (LC‐MS/MS) method was developed and validated for simultaneous quantitation of testosterone (T), androstenedione (A4), dehydroepiandrosterone sulfate (DHEAS), dihydrotestosterone (DHT), and 17‐hydroxyprogesterone (17‐OHP) that are associated with PCOS. Methods The serum samples were processed by protein precipitation and solid phase extraction before analysis with the in‐house developed LC‐MS/MS. The chromatographic separation was achieved with a C18 column, using a linear gradient elution with two mobile phases: 0.02% formic acid in water (phase A) and 0.1% formic acid in methanol (phase B). The separated analytes were detected by positive or negative electrospray ionization mode under multiple reaction monitoring (MRM). Results The assay for all the five analytes was linear, stable, with imprecision less than 9% and recoveries within ±10%. The lower limits of quantification were 0.05, 0.05, 5, 0.025, and 0.025 ng/mL for T, A4, DHEAS, DHT, and 17‐OHP, respectively. In the receiver operating characteristic curve (ROC) analyses with the PCOS (n = 63) and healthy (n = 161) subjects, the AUC of the four‐androgen combined was greater than that of any single androgen tested in PCOS diagnosis. Conclusions The LC‐MS/MS method for the four androgens and 17‐OHP showed good performance for clinical implementation. More importantly, simultaneous quantitation of the four androgens provided better diagnostic power for PCOS.


| INTRODUC TI ON
Polycystic ovary syndrome (PCOS) is recognized as one of the most common endocrine disorders in child-bearing aged women around the world 1 and is complicated with reproductive, metabolic, and psychological features. 2 However, the etiology of this disease remains largely unknown. Because of the heterogeneity in its clinical presentations, the diagnostic criteria of polycystic ovary syndrome have been debatable, 3 which poses a huge challenge for its clinical diagnosis. 4 Three significant diagnostic features of PCOS-chronic anovulation, hyperandrogenism, and polycystic ovaries 1,5 -had been proposed and gained wide acceptance. 1,5,6 Hyperandrogenism plays a prominent role in the pathological process of PCOS, 7 and it's considered as the most constant and important diagnostic component of this syndrome. 6,8,9 However, which androgens should be measured for the diagnosis of PCOS is still controversial. Ideally, the serum levels of testosterone (T), androstenedione (A4), dehydroepiandrosterone sulfate (DHEAS), dihydrotestosterone (DHT), and 17-hydroxyprogesterone (17-OHP) were suggested to assess the origins and the extent of excessive androgens of women. 10,11 On the other hand, direct immunoassays (IAs), such as radioimmunoassay (RIA), chemiluminescence, and enzyme immunoassays, are the most widely used methods for measuring androgens in clinical laboratories. 6,12 The IAs were found to be susceptible to temperature, pH value ionic strength, and other factors, 13 resulting in relatively poor sensitivity and specificity. 12,13 More importantly, IAs are prone to generate erroneous results and overestimate the androgens levels due to the unavoidable antibody-antigen cross-reactions. As a result, the IAs have been advised against being applied in the androgen measurement in women and children. 12 Recently, the liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based methods, with high sensitivity and "gold-standard" specificity, 14,15 have been employed to measure steroids in human serum. 14 Many studies have confirmed that the LC-MS/MS had good performance in the determination of T, DHEAS, A4 or other steroids, [16][17][18] offering an ideal alternative methodology in the determination of hyperandrogenism in PCOS. 19 Some previous studies have shown that the simultaneous determination of T and A4, or T and DHT was achievable with excellent accuracy and little matrix interference. 20,21 As part of the continual efforts to improve the laboratory diagnostic accuracy of hyperandrogenism in PCOS patients, we established an efficient LC-MS/MS method for simultaneous measurement of serum T, A4, DHEAS, DHT, and 17-OHP and evaluated the clinical utility of this androgen testing panel with the PCOS patients.

| Sample preparation
For each sample preparation, 50 μL patient serum + 5 μL methanol or 55 μL plasma-diluted QC/calibrator, 10 μL internal standard mixture, 145 μL methanol and 150 μL water were mixed in a 96-well protein precipitation plate. The mixture was shaked for 1 minute. Then, the supernatant was transferred to the Agela Cleanert PEP 96 Well Microplates for solid phase extraction, followed by a washing step with 200 μL of 10% acetonitrile in hexane. Lastly, the analytes of interest were eluted by 50 μL methanol/acetonitrile (1:9, v/v) and collected by a 96-well collection plate. The flow-through was further diluted with 50 μL water and 10 μL was injected for the LC-MS/MS analysis.

| Instrumentation and conditions
The LC-MS/MS analysis was performed using an AB Sciex 5500 mass spectrometer coupled with a Shimadzu Nexera X2 high-performance liquid chromatography (HPLC) system. A Venusil MP C18 column (VA93050, 3.0 × 50 mm, 3 μm, Agela Technologies) was used and maintained at a constant temperature of 40°C during operation. The mobile phase A was composed of 0.02% formic acid in water, and the mobile phase B was composed of 0.1% formic acid in methanol. The chromatography gradient conditions (%B) were set as follows (for a total run time of 6.5 minutes) with a flow rate of 0.6 mL/min: 55% B for 0-0.5 minutes, 55%-75% B for 0.5-3.0 minutes, 75%-90% B for 3.0-3.5 minutes, 90% B for 3.5-4.5 minutes, 90%-55% B for 4.5-4.6 minutes, 55% B for 4.6-6.5 minutes.
The analytes were detected by the mass spectrometer with

Based on the recommendations of the Clinical and Laboratory
Standards Institute (CLSI) 62-A guideline, 22 the method was validated for linearity, lowest limit of quantitation (LLOQ), precision, accuracy, matrix effect, serum sample stability, selectivity, carry over, and product ion cross-talk as described in the following paragraphs.

| Linearity and LLOQ
The linearity was evaluated by measuring the ratio of analyte peak area to the IS area against nominal concentrations of calibrators with linear regression and 1/X 2 weighing. The linearity validation was performed in three times of the same experimental day and the average slope, intercept, and correlation coefficient R of the three repeats were reported. The acceptance criterion for a calibration curve was a correlation coefficient R of 0.990 or better. The LLOQs were calculated by analyzing the serially diluted QC specimens spiked with IS over 5 days. The LLOQ was defined as the average concentration at which the S/N ratio > 10 and CV < 20% and bias was within ±20%.

| Precision, accuracy, stability and matrix effect
The intra-assay imprecision was estimated by analyzing the QCs for five times in the same run. The inter-assay imprecision was estimated by analyzing the QCs twice a day for a total of 10 days. The accuracy was evaluated by the recovery studies, in which the recovery of each androgen was calculated at high-, medium-, and low-level QCs by comparing the IS peak area ratio of extracted QC samples to the IS peak area ratio of non-extracted standard solutions at the same concentration. The stability of the analytes in serum was assessed by evaluating serum samples kept at 4 and 21°C (room temperature) for 6 days. The matrix effect was assessed according to the study by Matuszewski et al. 23 Briefly, the sample A was basically QC-L, QC-M, or QC-H prepared in 50% methanol in water (as neat sample), and it reflected 100% recovery with no matrix effects. The sample B was essentially the extracted blank plasma or serum pool of 20 healthy women. Then, the sample B extraction was split into two parts: One part was spiked with standard chemicals (prepared in methanol), with the final concentrations equivalent of QC-L, QC-M, or QC-H (sample B1); the other part was spiked with pure methanol (sample B2). With the MS responses of each compound from samples A, B1, and B2, the matrix effect was calculated with the following formula:

| Selectivity, carry over, and product ion crosstalk
According to the CLSI 62-A guideline, 22 it is important to confirm within the testing system (including sample processing and LC-MS) that the method has low background noise allowable for the assay.
For the selectivity validation, the human hormone-free plasma with no addition of standards was prepared and analyzed as the doubleblank control with the LC-MS/MS method.
Carry over was assessed by running a human hormone-free plasma sample immediately after injecting a calibrator 6 (upper limit of calibration curve) sample to verify the minimal sample carry over.
The calculated carry over in the blank sample should be less than 25% of LLOQ to be acceptable. 22 To check whether any product ion cross-talk exists for the analytes that have identical product ions, 22 we simply monitored all the ion transitions pairs listed in Table 1 at the chromatographic retention time of each analyte of interest (T, A4, DHEAS, and 17-OHP sharing the primary product ions with the m/z of 97.0).

| LC-MS/MS method optimization
The mass spectrometry instrumentation and conditions, includ-

| Assay validation summary
The linearity of the assay was tested by regression analysis, and the  Table 2). The 95% confidence intervals (CIs) of linearity, precision, and sensitivity studies were provided in the Tables S2 and S3. The intra-assay CV for QC-L ranged 5.2%-8.7%, for QC-M ranged 3.2%-5.0%, and for QC-H ranged 3.7%-6.3%; the interassay CV for QC-L ranged 6.5%-8.2%, for QC-M ranged 4.2%-6.0%, and for QC-H ranged 4.1%-5.9%. The accuracy for each analyte was evaluated by the recovery studies as described in the Methods. The accuracy measured by % bias was within ±20% acceptance criteria at QC-L, (Table 3 and Table S3). The analytes were found to be stable in serum for at least 6 days, when stored at 4 or 21°C (Table 2).

QC-M, and QC-H levels from intra-or inter-assay experiments
In the matrix effect assessment (Table 3), the matrix effects with blank plasma or patient serum pool were within 100 ± 20%, showing insignificant ion suppression or signal enhancement. More specifically, the matrix effects with QC-H were 104.3%-117.2% in blank plasma and healthy women serum pool; the matrix effects with QC-L were 92.6%-120.0%, with QC-M were 99.4%-118.4% in blank plasma and were not evaluated with healthy women serum pool due to relatively high background of endogenous hormones.
As seen in Figure S1, the signal from a double blank sample (no analyte, no IS) reflected low background noise in the LC-MS system. Moreover, the carry over for all the five analytes were acceptable (<25% LLOQ in the blank sample) (data not shown).
According to the CLSI C62-A, 22 product ion cross-talk is only of real concern when compounds enter the mass spectrometer at the same time. In Figure 1, no essential peak "overlaps" were observed between any of the analytes of interest. Nevertheless, we still performed the product ion cross-talk by monitoring all the ion transitions listed in Table 1. No product ion cross-talk was seen across the entire chromatographic running time (data not shown).

| Reference intervals and PCOS evaluation
The reference intervals determined with the healthy controls for   Table 4, and Table S1). At the same time, with the IDI evaluation of the ROC analyses, the four androgen testing panel showed better performance than any of the single andren or the T + DHT combination in discriminating PCOS patients, suggesting the clinical utility and significance of simultaneous quantifying these androgens (P < .05 for AUC comparison and IDI calculation in Table S1). The raw data of the four androgen measurements in both the PCOS patients and the healthy controls were available in Table S4.
As shown in Table 4 Nevertheless, it has been reported that the combination of an effective serum processing method such as optimized SPE and a sensitive LC-MS/MS instrument increased the assay performance, allowing a low quantification limit well below that reported for the androgens (T, A4, and DHEA) in female serum. 27,28 In this study, we employed a relatively simple serum preparation method and were able to completely separate and accurately quantitate the T, A4, DHEAS, DHT, and 17-OHP with a LC-MS/MS procedure in 6.5 minutes.
Impressively, the imprecision of this assay for all the five analytes was <9%, along with ideal LLOQ, linearity, and stability, suggesting that this was a reliable method with high suitability for wide clinical implementation.
In the same study, we also evaluated the potential clinical unitality of the four-androgen plus 17-OHP panel in the PCOS diagnosis. As part of the natures of PCOS, complicated, and dynamic sex hormone changes were observed in several studies with T elevation being the TA B L E 3 The recoveries, matrix effects, and imprecisions in assay validation

| CON CLUS ION
In summary, a laboratory-developed assay for simultaneous quanti-