ARTICLE pubs.acs.org/crt
Quantification of 4-Beta-Hydroxycholesterol in Human Plasma Using Automated Sample Preparation and LC-ESI-MS/MS Analysis Angela K. Goodenough,*,†,# Joelle M. Onorato,†,# Zheng Ouyang,† Shu Chang,‡ A. David Rodrigues,‡ Sreeneeranj Kasichayanula,§ Shu-Pang Huang,|| Wesley Turley,^ Richard Burrell,^ Marc Bifano,§ Mohammed Jemal,† Frank LaCreta,§ Adrienne Tymiak,† and David Wang-Iverson† )
Departments of †Bioanalytical and Discovery Analytical Sciences, ‡Metabolism and Pharmacokinetics, §Discovery Medicine and Clinical Pharmacology, Global Biometric Sciences, and ^Department of Chemical Synthesis, Research and Development, Bristol-Myers Squibb, Princeton, New Jersey 08543-4000, United States ABSTRACT: It has recently been proposed that plasma levels of 4β-hydroxycholesterol (4βHC) may be indicative of cytochrome P450 3A4 (P450 3A) activity and therefore could be used to probe for P450 3A-mediated drug drug interactions. With this in mind, we describe a highly sensitive and precise liquid chromatography electrospray ionization tandem mass spectrometry method for the measurement of 4βHC in human plasma with a lower limit of quantification established at 2 ng/ mL using 50 μL of plasma. The entire sample preparation scheme including saponification and derivatization of 4βHC to the corresponding dipicolinyl ester (DPE) was completed in less than 8 h using an automated sample preparation scheme enabling higher-throughput capabilities. Chromatographic resolution of 4βHC from 4R-hydroxycholesterol and other endogenous isobaric species was achieved in 11-min using an isocratic gradient on a C18 column. Because of endogenous concentrations of 4βHC in plasma, a stable isotope labeled (SIL) analogue, d7-4βHC, was used as a surrogate analyte and measured in the standard curve and quality control samples prepared in plasma. A second SIL analogue, d4-4βHC, was used as the internal standard. The intraday and interday accuracy for the assay was within 6% of nominal concentrations, and the precision for these measurements was less than 5% relative standard deviation. Rigorous stability assessments demonstrated adequate stability of endogenous 4βHC in plasma and the corresponding DPE derivative for the analysis of clinical study samples. The results from clinical samples following treatment with a potent P450 3A inducer (rifampin) or inhibitor (ketoconazole) are reported and demonstrate the potential future application for this highly precise and robust analytical assay.
’ INTRODUCTION 4βHC is produced from the enzymatic conversion of cholesterol by members of the P450 3A subfamily.1 P450 3A enzymes (e.g., P450 3A4 and P450 3A5) are also involved in the metabolism of approximately half of currently prescribed therapeutics on the market,2,3 making them key targets in the assessment of drug drug interactions (DDIs). Several recent reports describe the potential use of 4βHC as an endogenous marker of P450 3A activity in clinical samples using GC/MS for detection and quantification.1,4 7 Basal levels of 4βHC concentrations are reported at 14 59 ng/mL by this method.7 9 Upon treatment with a P450 3A inducer, an increase in 4βHC over basal levels would be expected, while a decrease in 4βHC levels would be expected following dosing with a P450 3A inhibitor. Elevated plasma levels of 4βHC have been observed following P450 3A induction by rifampin4,8 and antiepileptics1,5 and the magnitude of the changes in 4βHC was dose and time dependent. For example, treatment with rifampin for one week (20, 100, or 500 mg/day) increased plasma levels of 4βHC by 31%, 97%, and 176%, respectively.8 There are fewer reports on the effect of P450 3A inhibitors on circulating 4βHC plasma r 2011 American Chemical Society
levels; modest decreases were measured upon treatment with different antiviral regimens such as atazanavir + ritonavir ( 18%) and lopinavir + ritonavir ( 11%) after 4 weeks of dosing7 and following treatment with 400 mg of itraconazole for approximately one week ( 26%).10 There are currently several markers used to monitor for P450 3A activity. One approach involves dosing with a probe drug such as midazolam11 or erythromycin,12 but such studies are complex and pose challenges in specific patient populations such as pediatric and geriatric subjects, transplant recipients, and cancer patients where the administration of probe drugs may be detrimental to patient safety. Furthermore, poor correlation between midazolam clearance and erythromycin breath test results have been observed.13 The use of an endogenous marker such as 4βHC for the assessment of P450 3A-mediated DDIs would greatly simplify the study design because a probe drug is not required, which may be of particular utility in specific patient populations where the Received: May 5, 2011 Published: July 05, 2011 1575
dx.doi.org/10.1021/tx2001898 | Chem. Res. Toxicol. 2011, 24, 1575–1585
Chemical Research in Toxicology
Figure 1. Pathways to the formation of 4RHC and 4βHC.
dosing of probe drugs is more challenging. This has recently been demonstrated using a GC/MS method for the measurement of 4βHC in mothers and neonates.14 In addition, circulating levels of 4βHC in plasma have been shown to be relatively stable with a half-life estimated between 62 h and 17 days8,15 suggesting that plasma is a suitable matrix for 4βHC measurements, affording the possibility of using the same plasma sample for both pharmacokinetic (PK) analysis and 4βHC measurements. It may also be possible to assess the time course and dose response of the DDI. Analysis of the predose sample enables each subject to serve as his/her own control providing greater accuracy in the determination of the magnitude of change in 4βHC as a result of treatment. Studies also suggest that 4βHC levels in plasma may accurately reflect P450 3A activity and are not influenced by the activity of other major drug metabolizing enzymes in the human liver (e.g., P450s 1A2, 2C9, and 2B6)1 or from the oxidation of cholesterol either in vivo or during sample handling16 (Figure 1). While the urinary 6β-hydroxycortisol-to-cortisol ratio (UCR) is another example of endogenous markers used in the assessment of enzyme induction or inhibition, the UCR suffers from pronounced diurnal variation of cortisol and high interindividual variation, and there is often poor correlation with the results obtained from a probe drug.17 An isotope-dilution GC/MS method has been reported for the measurement of 4βHC in clinical samples following derivatization1,18,19 and more recently an LC/MS method using atmospheric pressure photoionization (APPI) that does not require derivatization prior to analysis has been described.20 The GC/ MS method suffered from both relatively large sample volume requirements and overnight sample preparation including saponification and derivatization to the trimethylsilyl derivatives, potentially limiting sample throughput. In contrast, the liquid chromatography atmospheric pressure photoionization tandem mass spectrometry (LC-APPI-MS/MS) method did not require derivatization and used solid-phase extraction (SPE) for sample cleanup following saponification. However, this method also required a relatively large sample volume for 4βHC measurements. In addition, sample throughput was still fairly low even without the need for derivatization. There are several other reports that use APPI or atmospheric pressure chemical ionization (APCI) for the measurement of oxysterols21 24 or electrospray ionization (ESI) for the measurement of 4βHC (in addition to other oxysterols);25 however, they all share the same disadvantages of lower sensitivity (requiring larger plasma sample volumes), or
ARTICLE
the analyses were performed in nonplasma matrixes making direct sensitivity comparisons to plasma difficult. We chose to develop a liquid chromatography electrospray ionization/tandem mass spectrometry (LC-ESI-MS/MS) assay because it is a more widely used analytical platform in the pharmaceutical industry in conjunction with 4βHC derivatization to meet sensitivity requirements. Improved mass spectral sensitivity in ESI has been reported following oxysterol derivatization to picolinyl esters,26,27 Girard P hydrazones,28 30 and dimethylglycine derivatives,31 and more recently with dansylated derivatives.32 In addition, oxysterols exist in plasma as both free forms and esterified to fatty esters making saponification prior to derivatization necessary to cleave the cholesterol esters to free sterols for the analysis of clinical samples.33,34 Preparation of the picolinyl ester derivative(s) was used in these studies because both sample saponification and derivatization required fairly short reaction times enabling sample preparations to be completed in a single day. In addition, derivatization to the DPE is the only method with demonstrated success for the measurement of 4βHC in human plasma. Here, we describe a robust, sensitive, and precise LC-ESI-MS/ MS based method for the quantification of 4βHC in 50 μL of human plasma. An automated sample preparation procedure was developed for the saponification and derivatization of clinical samples. Chromatographic resolution between 4βHC, isomeric 4R-hydroxycholesterol (4RHC), and additional endogenous isobaric species was achieved. Assay sensitivity was sufficient to measure any decreases in 4βHC concentrations following dosing with a P450 3A inhibitor and also afforded measurement of 4RHC in most samples (basal levels of 4RHC reported between 4 and 10 ng/mL1,8). Studies suggest that 4RHC is produced from the autoxidation of cholesterol and is not a product of P450 3A activity1,10 (Figure 1) and quantification of 4RHC was of secondary interest for potential use as a normalization factor in the analysis and interpretation of clinical 4βHC results. Since 4βHC (and 4RHC) is endogenous, a SIL surrogate analyte, d7-4βHC, was used for the generation of standard curve and quality control (QC) samples generated in human plasma. A second SIL analogue of 4βHC, d4-4βHC, was used as the internal standard (IS) in all samples. High assay precision was crucial, especially for the assessment of P450 3A inhibitors, and both intraday and interday accuracy and precision measurements clearly demonstrated the potential utility of this assay for measuring small changes in 4βHC levels. Rigorous stability assessments on both endogenous and derivatized 4βHC were also performed. The measurement of 4βHC in clinical samples following treatment with either rifampin (P450 3A inducer) or ketoconazole (P450 3A inhibitor) was successfully achieved using this method and demonstrated measurable changes in 4βHC levels.
’ EXPERIMENTAL PROCEDURES Chemicals and Reagents. 4βHC was purchased from Avanti Polar Lipids (Alabaster, AL). 4RHC, d7-4βHC, and d4-4βHC were synthesized by Radiosynthesis Group at Bristol-Myers Squibb (Lawrenceville, NJ and Wallingford, CT). Untreated human plasma (K2EDTA) (both pooled and individual) was obtained from Bioreclamation (Hicksville, NY). All other chemicals and solvents used were of the highest purity available and purchased through commercial vendors. Preparation of the 4rHC, d7-4βHC, and d4-4βHC Standards. 4RHC was prepared from cholesteryl benzoate in a 4-step synthesis.35,36 Cholesteryl benzoate was hydroxylated at C4 using 1576
dx.doi.org/10.1021/tx2001898 |Chem. Res. Toxicol. 2011, 24, 1575–1585
Chemical Research in Toxicology selenium dioxide and formic acid. The 4β-hydroxyl intermediate was oxidized using triphenylbismuth carbonate and sodium carbonate to produce the corresponding ketone. The ketone was reduced with sodium cyanoborohydride to provide an alcohol with the desired R stereochemistry at C4. The benzoate protected alcohol at C3 was deprotected using sodium carbonate to provide the desired product 4RHC. Isotopically labeled d7- and d4-4βHC standards were each prepared from the appropriately labeled form of cholesterol (either d4or d7-) in two steps.36,37 Stable isotope labeled cholesterol (either d4- or d7-) was reacted with bromine in chloroform followed by silver acetate in pyridine to install an acetate protected alcohol at C4. The acetate protected intermediate was hydrolyzed with methanolic KOH to provide the desired stable isotope labeled (either d4- or d7-) 4βHC. The chemical purity of d7-4βHC (surrogate analyte) was found to be 93.3% by LC/UV with an isotopic purity of 93.8% determined by LC/MS. In addition, there was no d4-4βHC contamination in the d7-4βHC product. An overall purity of 87.5% was used for the preparation of standard curve and QC stocks. For d4-4βHC, a chemical purity of 94.4% was determined by LC/UV, and an isotopic purity of 97.5% was found by LC/MS. An overall purity of 92.0% was used for the preparation of the IS stocks. Standard Solutions. All standard stock solutions were prepared by dissolving an accurately weighed amount of analyte (taking into account isotopic purity when relevant) in EtOH at an initial concentration of 1 mg/mL. Subsequent dilutions of d4-4βHC (IS) to 50 ng/mL EtOH were prepared in Corning 50-mL conical polypropylene centrifuge tubes (Corning, NY) and used as the IS working solution in all sample analyses. An intermediate diluted stock of d7-4βHC (surrogate analyte) was prepared at 10 μg/mL EtOH. This stock was used to generate standard curve working solutions in EtOH at 2, 10, 20, 40, 60, 80, 160, and 500 ng/mL and QC working standard solutions at 15, 25, 30, and 375 ng/mL. The standard curve and QC working solutions were prepared and stored in Kimble Chase 20-mL borosilicate glass scintillation vials with caps (Vineland, NJ). All standard solutions were prepared at room temperature and stored at 20 C Calibration Curve and Quality Control Samples. Standard curve and QC samples were freshly prepared with each analytical set in pooled, untreated human plasma by the addition of 50 μL of the appropriate standard working solution of d7-4βHC to 50 μL of control plasma. Because the study samples were diluted with ethanol in the same proportion (see Sample Preparation below), the final standard curve concentrations were designated as 2, 10, 20, 40, 60, 80, 160, and 500 ng/mL. The final QC concentrations were designated as 15, 25, 30, and 375 ng/mL. Samples were also spiked with 50 μL of the IS working solution for a final concentration of 50 ng/mL d4-4βHC. Clinical Samples. Rifampin Studies. Healthy human subjects (N = 14; all subjects were Asian males; mean age 25.1 years; average body mass index (BMI) 23.5) were treated with rifampin (Rifampin) 600-mg QD by oral administration on days 3 9 (total study duration was 11 days). Plasma samples collected at time 0 h on days 3, 5, and 7 were analyzed for 4βHC concentrations; 4RHC concentrations were also determined. For each subject, a predose sample was also collected for measurement. Blood was collected in plastic tubes containing K2EDTA as the anticoagulant. Plasma was prepared by centrifugation (1000g for 15 min) within an hour of sample collection and was transferred to a polypropylene tube for shipment and storage. The blood/ plasma samples were maintained at 4 C throughout the collection and preparation procedure. Collected plasma was stored at 80 C prior to sample analysis. This study was approved by Institutional Review Board at Inje University Busan Paik Hospital. Ketoconazole Studies. Healthy human subjects (N = 13; 12 white/1 Asian American; 10 Hispanic/3 non-Hispanic; 12 males/1 female; mean age 36 years; average BMI 26.4) were treated with ketoconazole (Nizoral) 400-mg QD by oral administration on days 5 8 (total study duration
ARTICLE
was 14 days). Plasma samples collected at time 0 h on days 5 and 7 were analyzed for 4βHC concentrations; 4RHC concentrations were also determined. For each subject, a predose sample was also collected for measurement. Blood was collected in plastic tubes containing K2EDTA as the anticoagulant. Plasma was prepared by centrifugation (1000g for 15 min) within an hour of sample collection and was transferred to a polypropylene tube for shipment and storage. The blood/plasma samples were maintained at 4 C throughout the collection and preparation procedure. Collected plasma was stored at 80 C prior to sample analysis. This study was approved by MDS Pharma Services (US) Inc. Institutional Review Board. Sample Preparation. Standards, QCs, and study samples were analyzed in 96-well plates. The calibration curves were analyzed in duplicate and the QCs in quadruplicate. To match the dilution of the plasma with the working standard solution in the standard curve and QC samples, all study samples and double blank (db) samples were diluted 1:1 with ethanol prior to sample preparation. Tecan Genesis RSP 150 Automated Liquid Handling System (Tecan, M€annedorf, Switzerland) employing a customized program (Gemini version 4.2) was used to dispense the samples and for all liquid handling steps in the saponification and derivatization procedures. Study samples (50 μL) were diluted 1:1 with ethanol, followed by the addition of a 50 μL volume of the IS (d4-4βHC). The plate was vortexed at high speed for 1 min. Samples were next saponified in the presence of butylated hydroxytoluene (25 μL of a 1 mg/mL EtOH stock stored at 20 C) and 1 M ethanolic KOH (250 μL) at 37 C for 1 h. Following saponification, the samples were extracted with hexanes (0.5 mL) and water (0.14 mL). The samples were centrifuged (4,000 rpm for 5 min at 4 C) and the supernatant transferred to clean 1.4-mL Micronic tubes and dried under N2 at 50 C. To the dried residues was added 170 μL of freshly prepared derivatization mixture: 1244.5 mg of 2-methyl-6-nitrobenzoic anhydride, 373.3 mg of 4-dimethylaminopyridine, 995.5 mg of picolinic acid, 2.5 mL of triethylamine (stored over molecular sieves), and 18.7 mL pyridine (stored over molecular sieves). Samples were capped, vortexed, and incubated for 30 min at room temperature. The DPEs were extracted from the reaction mixture with hexanes (0.5 mL). Samples were centrifuged (4,000 rpm for 5 min at 4 C), and the supernatant was transferred to clean 1.4-mL polypropylene tubes and dried under N2 at 50 C. The dried samples were resuspended in CH3CN (200 μL), vortexed, and centrifuged, and then 125 μL of the supernatant was transferred to Axygen 96-well polypropylene microplates (Union City, CA) using a CyBio CyBi-Well 96-channel simultaneous pipettor (Boston, MA). The plate was capped with Axygen silicone capmats for LC-ESI-MS/MS analysis. Chromatography and Mass Spectrometry. Chromatography was performed using a Waters Acquity UPLC (Milford, MA) equipped with an Acquity UPLC BEH C18 (100 2.1 mm, 1.7 μm) column from Waters (Milford, MA) maintained at 35 C. An isocratic elution (11 min total run time) at a solvent composition of 90% B with solvent A = 98:2 MeOH/H2O containing 0.1% formic acid and B = 1:1 CH3CN/MeOH containing 0.1% formic acid was used for sample analyses at a flow rate of 0.5 mL/min. The autosampler was maintained at 4 C throughout the analyses. Sample injection volume was 10 μL. The Waters Acquity UPLC was coupled to an AB Sciex API 5000 triple quadrupole mass spectrometer (Foster City, CA). The mass spectrometer was equipped with an electrospray ionization source operated in positive ionization mode at unit resolution in Q1 and Q3. Nitrogen was used for the curtain and collision gas. Data were collected in selected reaction monitoring (SRM) mode using transitions of m/z 613 f 490 (4β/4RHC-DPE), m/z 617 f 494 (d4-4βHC-DPE), and 620 f 497 (d7-4βHC-DPE). The dwell time used was 150 ms per channel with a 5 ms pause between mass transitions. Optimized MS acquisition parameters were as follows: collision gas (CAD) 5, curtain 1577
dx.doi.org/10.1021/tx2001898 |Chem. Res. Toxicol. 2011, 24, 1575–1585
Chemical Research in Toxicology gas (CUR) 30, ion source gas 1 (GS1) 25, ion source gas 2 (GS2) 40, ion spray voltage (IS) 5500, source temperature (TEM) 450, interface heater (ihe) ON, declustering potential (DP) 20, entrance potential (EP) 10, collision energy (CE) 20, and collision cell exit potential (CXP) 10. The data were acquired and processed using AB Sciex Analyst 1.5 Software. Data Analysis. Calibration curves were generated using a weighted (1/x) quadratic regression of peak area ratios (PAR) of d7-4βHC (surrogate analyte) to d4-4βHC (IS) vs the nominal concentrations of d7-4βHC. Analyst software version 1.5 was used to generate curves and to calculate the concentration and accuracy of d7-4βHC in calibration standards and QCs. Because Analyst does not allow the calibration for one analyte (d7-4βHC) to be applied to other analytes (4R- and 4βHC), GraphPad Prism 4 (La Jolla, CA) was used to generate d74βHC calibration curves and to generate concentration values for 4-R/ βHC. Identical curve fittings were used in both Analyst and GraphPad (quadratic fit with 1/x weighting). Accuracy and Precision. The intraday and interday accuracy and precision for the assay were evaluated by measuring QC samples at four different concentrations of d7-4βHC (15, 25, 30, and 375 ng/mL) (N = 8 at each concentration) over six days with two sets of calibration curve samples generated in parallel on each day. Acceptance criteria for the accuracy and precision determinations were specified to be within 15% of the nominal value and 15% relative standard deviation (RSD). An additional accuracy and precision assessment was also performed to evaluate the performance of the method over a limited dynamic range. In these experiments, pooled control human plasma was spiked with d74βHC at concentrations of 18, 22, 25, 30, 35, 50, 125, and 250 ng/mL (N = 8 at each concentration). Stability. Autosampler Stability. Autosampler stability of the derivatized analytes was evaluated by making replicate injections from the same set of processed plasma samples: the first injection was made immediately after sample processing (t = 0), and the second injection was made 24 h later. The processed samples were maintained at 4 C in the autosampler during the 24 h period between injections. Freeze Thaw Stability. The freeze thaw stability of endogenous levels of 4βHC and 4RHC was evaluated. The commercially obtained pooled control human plasma was thawed, and 50 μL of plasma was transferred to Micronic 1.4 mL polypropylene tubes with thermoplastic elastomer pushcaps and immediately frozen on dry ice; this was designated cycle 1. For each freeze thaw cycle, the plasma samples were thawed, maintained at room temperature for 1 h, and then refrozen on dry ice with the process repeated four times for a total of five freeze thaw cycles (including the initial thaw cycle when the samples were transferred to the Micronic polypropylene tubes used for the subsequent saponification and derivatization steps). Following each freeze thaw cycle, samples (N = 8) were prepared and analyzed as described above. Long-Term Stability. Long-term storage stability (at 80 C) determinations were based on comparison of the calculated levels of endogenous 4R- and 4βHC in the stored samples to the concentration determined in fresh control human plasma. Control pooled human plasma samples were transferred to Micronic 1.4 mL polypropylene tubes with thermoplastic elastomer pushcaps in the same manner as that for the freeze thaw stability studies and were immediately frozen and stored at 80 C. Plasma samples were stored at 80 C and analyzed at 1-month and 4-month time points as described above.
Inter-Individual Differences in Endogenous 4rHC and 4βHC Levels. Human plasma from 10 healthy, untreated individuals (5 males and 5 females) was obtained from a commercial source. Duplicate analyses from a single plasma sample collected from each individual were performed as described above. Matrix Effects. QC samples at concentrations of 25, 30, and 375 ng/mL (N = 6 samples at each QC concentration) and standard curve samples (N = 1 standard curve at final concentrations of 2, 10, 20,
ARTICLE
40, 60, 80, 160, and 500 ng/mL) were prepared in EtOH following the same saponification and derivatization protocol used for the plasma samples. Matrix effects were determined by comparing the signals of the analytes prepared in pure solvent to the signals in plasma containing samples.
’ RESULTS AND DISCUSSION Method Development. Sample Saponification and Derivatization. It has been reported that the majority of 4βHC in plasma
exists bound as cholesterol esters;6 therefore, it was necessary to first cleave these hydroxycholesterol esters by alkaline hydrolysis to ensure accurate measurement of 4βHC. Butylated hydroxytoluene was added as an antioxidant to the sample prior to saponification to minimize any artifactual oxidation of cholesterol to oxysterols during this step. A time course for the saponification was performed, and an increase in the levels of 4βHC (and 4RHC) was observed between time 0 and 30 min. A plateau in the 4βHC concentrations was observed between 30 min and 24 h (the longest time point investigated) (results not shown) suggesting that both 4R- and 4βHC were stable under the hydrolysis conditions used. We chose to move forward with a 60 min saponification for the analysis of plasma samples. The majority of current literature reports on the measurement of 4βHC in clinical samples use an isotope-dilution GC/MS method.1,18,19 Because of the need for relatively large sample volumes (250 μL to 1 mL plasma) and overnight sample preparation including saponification and derivatization to the trimethylsilyl derivatives,1,6 we sought the development of an LC-ESI-MS/MS based method for the measurement of 4βHC in human plasma, which is a more commonly used analytical platform within the pharmaceutical industry. Since using the same plasma sample for both PK and 4βHC measurements was of interest, the effort was made to keep sample volume requirements to a minimum during method development. When these studies were initiated, there were limited literature reports describing the analysis of oxysterols by LC/MS without prior derivatization. Several reports described the use of APCI21,23 or ESI25 for the measurement of 4βHC and/or other oxysterols, but all of these studies were performed in biological matrixes other than plasma, making direct sensitivity comparisons inconclusive. We investigated the use of both ESI and APCI on several MS platforms but found the ionization of underivatized 4βHC to be quite poor, and insufficient levels of detection were obtained. Several derivatization procedures for oxysterols with the demonstrated ability to improve mass spectral sensitivity in ESI were also reported in the literature, including derivatization to picolinyl esters,26,27 Girard P hydrazones,28 30 and dimethylglycine derivatives.31 Of these, only derivatization to the picolinyl esters was reported to be complete within an hour, making it more amenable to higher-throughput sample analyses. In addition, it was the only derivatization method that was successfully used for the measurement of 4βHC in small volumes of human serum (5 μL).26 Derivatization of the two free-hydroxyl groups present in 4R- and 4βHC (as well as the deuterated analogues) to the corresponding DPE dramatically improved assay sensitivity, providing a lower limit of quantification (LLOQ) of 2 ng/mL (defined as the lowest standard concentration where the predicted concentrations are within 20% of theoretical), which was sufficient for our purposes. We required 50 μL of human plasma to reach this LLOQ. This is a larger volume than the 5 μL 1578
dx.doi.org/10.1021/tx2001898 |Chem. Res. Toxicol. 2011, 24, 1575–1585
Chemical Research in Toxicology reportedly used for 4βHC measurements in human serum, but this could be due to differences between the use of plasma vs serum as the biological matrix. In addition, because we were also interested in the measurement of 4RHC, which is present at lower basal levels than 4βHC, greater assay sensitivity was required (e.g., using larger sample volumes). Regardless, 50 μL sample volumes for the analysis of human plasma samples is sufficiently low to enable multiple analyses from the same sample (e.g., PK and 4βHC measurements). Automation of the multistep sample processing protocol was essential to enable practical processing of large analytical sets associated with clinical studies. More recently, derivatization with dansyl chloride has also been shown to be complete within an hour and to improve LCESI-MS/MS sensitivity for several P450 oxidation products in tissue extracts;32 however, this derivatization scheme was published after the method development and clinical sample analyses using derivatization to the DPE’s described here was completed. To date, this method has also been described for the measurement of 4βHC in microsomal incubations38 but not in human plasma. A more recent report using LC-APPI-MS/MS for the measurement of 4βHC in human plasma samples without the need for sample derivatization was also published after this work was completed.20 Relatively large sample volumes (400 μL) were needed to measure 4βHC (reported LLOQ 10 nM = 4 ng/mL), reducing the possibility that the same sample could be used for both PK and 4βHC measurements by this method. In addition, a sample throughout of approximately 80 samples per day was reported (ca. 60 study samples plus standard curve and QC samples). Even with the need for derivatization to the corresponding DPEs in the approach we have developed, sample throughput g96 samples per day was easily obtained when using automation. Mass Spectrometry. In addition to shorter sample preparation times, data interpretation of the DPE derivative was also simplified relative to the conversion to the Girard P hydrazones, which produced a variety of different derivatized forms (e.g., syn and anti, mono- and bis-GP, etc.),28,29 while conversion to the dimethyl glycine esters used the doubly protonated ion for detection, all of which complicate data analysis.31 Previous reports used the sodiated adduct as the precursor ion in SRM for 4βHC.26 We observed the [M + H]+ ion as the dominant ion in our mass spectrometric analyses and used this as the precursor ion in the SRM method. The product ion used in the final SRM method was also different from that reported in the literature and corresponded to the loss of one picolinyl ester moiety (Figure 2). Assay sensitivity requirements were based on literature reports for endogenous levels of 4RHC, which were reported at 4 10 ng/mL. The LLOQ for 4βHC (and 4RHC) using the above-described method was 2 ng/mL, which should be sufficient for accurate quantification of both isomers at the reported endogenous levels. Chromatography. Baseline resolution between 4RHC-DPE and 4βHC-DPE was achieved in an 11 min LC run time using ultra-high performance liquid chromatography (UHPLC) with an isocratic gradient and C18 column. Using these chromatographic conditions, the derivatized analytes elute in the following order (retention times in min): 4RHC-DPE (7.95), d7-4βHCDPE (8.47), d4-4βHC-DPE (8.54), and 4βHC-DPE (8.59) (Figure 3, panels A C). This is currently the shortest run time reported for the different 4βHC-specific analytical methods:
ARTICLE
Figure 2. Structure of the 4βHC-DPE and the masses and fragmentation used for quantification in the SRM method.
GC/MS (g18 min),1 underivatized 4βHC by LC-APPI-MS/MS (15 16 min),20 or the dansylated derivative (13 min).38 Several additional peaks were also observed in the same SRM transition used to detect 4R- and 4βHC and are presumed to be other isomeric dihydroxycholesterols, but these species are well resolved from 4R- and 4βHC and do not appear to impact their accurate quantification. Additional experiments have been performed with several commercially available oxysterols, and none has been found to coelute with either 4βHC or 4RHC. Calibration Curve and QC Samples. The endogenous presence of 4βHC in plasma required a different approach for the preparation of standard curve and QC samples. The use of charcoal stripped plasma has been successfully used as a suitable matrix in the analysis of other steroids39 but was not a viable alternative for this assay because it was found to contain levels of 4R- and 4βHC comparable to untreated plasma. For this reason, two different SIL versions of 4βHC were synthesized and used in this assay: d7-4βHC was used as a surrogate analyte in standard curve and QC samples prepared in pooled control human plasma, and d4-4βHC was used as the IS. No chromatographic peaks were observed in human plasma double blanks in the SRM transitions used for d7-4βHC-DPE and d4-4βHC-DPE (Figure 3, panels D and E), indicating there were no endogenous species in human plasma that would contribute to the overall signal of d7-4βHC and d4-4βHC in clinical samples. In addition, equivalency in the ionization efficiency among d0-, d4-, and d74βHC was demonstrated, further supporting the use of d74βHC as a surrogate analyte for d0 4βHC in the QC and standard curve samples (results not shown). The use of 2-propanol for the generation of standard curve and QC samples has been reported, especially at the lower 4βHC concentrations,20 but we felt it was preferable to maintain consistency with the matrix for the clinical samples. We elected to monitor four QC concentrations covering the expected low, medium, and high concentration ranges for 4βHC. Modeling results suggested treatment over a relatively short period with P450 3A inhibitor ketoconazole would result in minor decreases in 4βHC.40 On the basis of the modeling, two closely grouped midrange QC concentrations were selected around the reported mean value of endogenous levels of 4βHC to ensure that the assay was sufficiently precise to measure small changes in 4βHC levels . Equivalency of 4βHC and 4RHC Calibration Curves. In an effort to simplify sample analyses, a single calibration curve measuring d7-4βHC as the surrogate analyte and d4-4βHC as the IS was used for the quantification of both 4R- and 4βHC. While d7-4βHC was shown to be an appropriate surrogate 1579
dx.doi.org/10.1021/tx2001898 |Chem. Res. Toxicol. 2011, 24, 1575–1585
Chemical Research in Toxicology
ARTICLE
Figure 3. Representative chromatograms of (A) 4RHC-DPE and 4βHC-DPE; (B) d7-4βHC-DPE; and (C) d4-4βHC-DPE from human plasma and of the SRM transitions for (D) d7-4βHC-DPE and (E) d4-4βHC-DPE from double blank samples.
Figure 4. Equivalency in the calibration curves for 4RHC-DPE and 4βHC-DPE.
analyte for the measurement of 4βHC, its suitability for the measurement of 4RHC also needed to be demonstrated. Equivalency in ionization response between 4RHC and 4βHC was shown by preparing calibration curves of each analyte in water using authentic standards, which were shown to be identical within experimental error (Figure 4). The accuracy and precision of each standard curve was within (15% of the expected value at all concentrations. While some minor divergence of the curves outside of the error bars is visually observed at the highest standard curve concentration (500 ng/mL), this is somewhat misleading as the average calculated concentrations of 4RHC and 4βHC are the same (499.7 for 4βHC and 500.0 ng/mL for
4RHC), and the average peak area ratios (the ratio of either 4RHC-DPE or 4βHC-DPE to d4-4βHC-DPE) are within 10% of each other. With equivalency in the calibration curves for 4R- and 4βHC demonstrated, all reported measurements for both 4RHC and 4βHC were calculated from a single calibration curve using d7-4βHC as the surrogate analyte and d4-4βHC as the IS. Accuracy and Precision. The intra- and interday accuracy and precision of the method was determined by preparing and analyzing QC samples along with calibration curve samples on six different days. Calibration curve samples consisted of eight different concentrations of d7-4βHC (surrogate analyte) ranging from 2 to 500 ng/mL prepared in pooled control human plasma 1580
dx.doi.org/10.1021/tx2001898 |Chem. Res. Toxicol. 2011, 24, 1575–1585
Chemical Research in Toxicology
ARTICLE
Table 1. Intraday and Interday Precision and Accuracy [d7-4βHC] day 1
day 2
theoretical
calculated
accuracy
[d7-4βHC]
mean [4βHC]
(mean % deviation
precision
(ng/mL)
(ng/mL)
from theoretical)
((% RSD)
6.6%
18
19.1
6.9
10.7
407
22
24.2
9.8
13.1
18
25
27.4
9.6
12.0
8.5%
30
31.8
6.0
10.2
4.3% 395
35
36.5
4.4
6.3
50
51.3
2.6
5.1
13
125
128
2.1
5.3%
250
244
31.5
420
SD
0.3
0.5
0.6
28
Δ%
10.8%
5.8%
4.9%
12%
% RSD
2.0%
2.1%
1.8%
mean
% RSD mean
15.8 0.4 5.0% 2.3% 15.1 0.8
25.9 0.7 3.5% 2.8% 25.0 1.0
31.8 1.1 6.0% 3.5% 30.8 1.8
Δ%
0.9%
% RSD
5.0%
4.0%
5.7%
3.3%
mean
14.8
25.7
30.7
400
SD
0.3
0.9
1.0
14
3.0%
2.4%
6.6%
3.2% 30.4
3.6% 369
1.0
15
Δ%
day 6
375 ng/mL
26.5
SD
day 5
30 ng/mL
16.6
Δ%
day 4
25 ng/mL
mean
SD
day 3
15 ng/mL
1.6%
0.2%
2.6%
% RSD mean
1.8% 15.2
3.4% 24.8
SD
0.3
0.5
Δ%
1.5%
% RSD
1.8%
2.0%
3.2%
4.0%
mean
15.2
24.7
30.0
380
SD
0.9
0.9
1.2
11
Δ%
1.4%
% RSD grand mean
5.8% 15.5
0.9%
1.2% 3.6% 25.4
1.2%
0.1% 3.9% 30.9
Table 2. Assay Accuracy and Precision Over a Limited Dynamic Range
2.8% 395
Δ%
3.0%
1.7%
2.9%
5.4%
interday % RSD
4.1%
2.5%
1.8%
4.4%
intraday % RSD
3.5%
3.1%
3.7%
4.4%
total % RSD
5.4%
4.0%
4.1%
6.2%
with curves prepared in duplicate on each day. Four concentrations within the range of the calibration curve were selected for QC samples: low, two mid, and high (all relative to the expected 4βHC concentrations). A total of eight replicates were performed for each QC concentration on each of the six different days. A summary of the one-way ANOVA intraday and interday accuracy and precision measurements is shown in Table 1. The assay performed extremely well with an overall interday % deviation e5.4% (reported as % change from the nominal/calculated concentration) for all four QC concentrations, and the results did not exceed (12.0% on any single day. More importantly, both intraday and interday precision were e4.4% RSD, and the total % RSD was e6.2% for all four QC concentrations: as indicated above, assay precision is of paramount importance in determining small changes in 4βHC concentrations between treatment groups. Because of the long half-life of 4βHC, modeling has suggested that only modest decreases in the plasma 4βHC concentrations would be expected upon treatment with a P450 3A inhibitor without prolonged dosing.40 To address this concern, the accuracy and precision of the assay over a limited dynamic range was evaluated by spiking d7-4βHC into human plasma at concentrations of 18, 22, 25, 30, 35, 50, 125, and 250 ng/mL (N = 8 samples at each concentration) (Table 2). The closely grouped concentrations were selected to cover the range of reported 4βHC basal levels (14 59 ng/mL). The accuracy and precision
1.4
Table 3. Autosampler Stability (4 C Overnight) nominal
1.6%
1.4%
1.4
2.2
time = 0 h
time = 0 h
time = 24 h
[d7-4βHC] % accuracy
precision
% accuracy
time = 24 h precision
(% RSD) (N = 8)
(N = 8)
(% RSD) (N = 8)
ng/mL
(N = 8)
15
101
5.1
101
4.9
25
99.9
4.1
102
2.8
30
103
5.7
103
2.7
375
105
3.2
105
3.5
determined at each QC concentration was within the (15% accuracy and % RSD acceptance criteria. Statistical analyses (students one-way t test) showed the calculated mean results at each concentration were significantly different from each other. These results show the analytical method is both sufficiently accurate and precise to detect very modest changes in the levels of 4βHC. Stability. Autosampler Stability. QC samples were analyzed by LC-ESI-MS/MS immediately following sample processing and then again following storage at 4 C in the autosampler for approximately 24 h. Both the accuracy and precision measurements were consistent between time 0 and 1 day and fell within the (15% accuracy and % RSD criteria established for this assay (Table 3). Freeze Thaw stability. To ensure endogenous levels of 4βHC and 4RHC were stable during multiple freeze thaw cycles, pooled control human plasma was analyzed after multiple freeze thaw cycles. Samples were maintained at room temperature for 1 h during each thaw cycle, refrozen on dry ice, and kept frozen for approximately 30 min during each freeze cycle; this process was repeated for a total of 5 freeze thaw cycles. There was no statistically significant difference in the relative mean concentration of 4βHC (
