Neurosteroids in Rat Brain: Extraction, Isolation, and Analysis by

The neurosteroids allopregnanolone and dehydroepiandrosterone modulate resting-state amygdala connectivity. Rebecca K. Sripada , Robert C. Welsh ...
15 downloads 0 Views 331KB Size
Download PDF
Anal. Chem. 2003, 75, 5835-5846

Neurosteroids in Rat Brain: Extraction, Isolation, and Analysis by Nanoscale Liquid Chromatography-Electrospray Mass Spectrometry Suya Liu,† Jan Sjo 1 vall,† and William J. Griffiths*,†,‡

Department of Medical Biochemistry and Biophysics, Karolinska Institutet, SE-171 77 Stockholm, Sweden, and Department of Pharmaceutical & Biological Chemistry, The School of Pharmacy, University of London, 29-39 Brunswick Square, London WC1N 1AX, U.K.

A method designed for the analysis of sulfated neurosteroids and unconjugated ketonic neurosteroids in rat brain using nanoscale liquid chromatography-electrospray (nano-LC-ES) mass spectrometry is described. Neurosteroids in rat brain tissue were extracted, purified, and separated into two groups, neutral unconjugated steroids and steroid sulfates, by employing solid-phase partition, cation- and anion-exchange chromatography. The steroid sulfate fraction was analyzed by nano-LC-ES mass spectrometry. Contrary to expectations, the sulfates of pregnenolone and dehydroepiandrosterone (DHEA) were not detected. Internal standards, including pregnenolone sulfate, were recovered and the detection limit of the method was 0.3 ng/g of wet brain. Cholesterol sulfate was detected at a level of 1.2 µg/g of wet brain. The neutral unconjugated steroid fraction was derivatized with hydroxylamine hydrochloride to convert oxosteroids into their oximes. The oximes were isolated using cationexchange chromatography and were analyzed by nanoLC-ES tandem mass spectrometry. The analyses of the neutral unconjugated steroid fraction confirmed the presence in rat brain of pregnenolone, pregnanolone isomers, progesterone, testosterone, and DHEA, which were characterized by their retention times, the mass of the protonated molecules, and characteristic fragment ions. The levels were estimated by addition of [3,4-13C2]-progesterone as an internal standard and found to be in a range of 0.04-20 ng/g. Neurosteroids are synthesized de novo in the nervous system or are metabolites of steroids derived from a peripheral source.1 Their occurrence in the central and peripheral nervous system is independent, at least partially, from the endocrine secretion of steroids. Known neurosteroids include pregnenolone (3β-hydroxypregn-5-en-20-one) and dehydroepiandrosterone (DHEA, 3β-hydroxyandrost-5-en-17-one), progesterone (pregn-4-en-3,20dione), and its reduced metabolites. It has been demonstrated that neurosteroids modulate neurotransmission by binding to * To whom correspondence should be addressed. E-mail: william.griffiths@ ulsop.ac.uk. Tel.: +44 (0)207753 5876. Fax: +44 (0)207753 5964. † Karolinska Institutet. ‡ University of London. (1) Baulieu, E. E. Recent Prog. Horm. Res. 1997, 52, 1-32. 10.1021/ac0346297 CCC: $25.00 Published on Web 09/20/2003

© 2003 American Chemical Society

neurotransmitter receptors and exert physiological functions that are clearly different from those of endocrine steroids.1 The effects of neurosteroids on improving the memory of cognitively impaired aged rats, on the inhibition of aggressiveness in castrated male mice, and on trophic effects on neuronal regeneration and remyelination have been documented.1 It has also been reported recently that pregnenolone binds to microtubule-associated protein 2 and stimulates microtuble assembly.2 The local synthesis, selective interaction with neurotransmitter receptors, and behavioral effects of neurosteroids strongly suggest that they may have important physiological or pathophysiological roles and have a potential as pharmaceuticals. There is an increasing need to develop methods to analyze these important hormones with high sensitivity and high specificity. Furthermore, the establishment of a method for the analysis of the neurosteroid profile in specific tissue areas is required. In most studies of neurosteroids to date, the measurement of steroids has been accomplished by radioimmunoassay.3,4 While such assays are highly sensitive, they lack specificity due to unknown nonspecific reactions and cross-reactions. Gas chromatography/mass spectrometry (GC/MS) has also been employed for the characterization3,4 and quantitative analysis of neurosteroids.5-9 High sensitivity can be achieved by the use of chemical ionization with specific derivatization. However, only volatile and thermally stable steroids can be analyzed by GC/MS, which precludes the direct analysis of steroid conjugates, e.g., sulfates. As steroid sulfates are believed to be an important class of neurosteroids, there is now a considerable incentive to develop direct mass spectrometric methods for their analysis. The utility of electrospray (ES) mass spectrometry for the analysis of steroid sul(2) Murakami, K.; Fellous, A.; Baulieu, E. E.; Robel, P. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 3579-3584. (3) Corpechot, C.; Robel, P.; Axelson, M.; Sjo ¨vall, J.; Baulieu, E. E. Proc. Natl. Acad. Sci. U.S.A. 1981, 78, 4704-4707. (4) Corpechot, C.; Synguelakis, M.; Talha, S.; Axelson, M.; Sjo¨vall, J.; Vihko, R.; Baulieu, E. E.; Robel P. Brain Res. 1983, 270, 119-125. (5) Cheney, D. L.; Uzunov, D.; Costa, E.; Guidotti, A. J. Neurosci. 1995, 15, 4641-4650. (6) Liere, P.; Akwa, Y.; Weill-Engerer, S.; Eychenne, B.; Pianos, A.; Robel, P.; Sjo ¨vall, J.; Schumacher, M.; Baulieu, E. E. J. Chromatogr., B 2000, 739, 301-312. (7) Valle´e, M.; Rivera, J. D.; Koob, G. F.; Purdy, R. H.; Fitzgerald, R. L. Anal. Biochem. 2000, 287, 153-166. (8) Kim, Y. S.; Zhang, H.; Kim, H. Y. Anal. Biochem. 2000, 277, 187-195. (9) Shimada, K.; Yago, K. J. Chromatogr. Sci. 2000, 38, 6-10.

Analytical Chemistry, Vol. 75, No. 21, November 1, 2003 5835

fates10,11 and neutral steroid derivatives12-14 has been demonstrated by several groups. The combination of liquid chromatography (LC) and ES mass spectrometry has been widely employed for the analysis of steroids from biological samples.15-19 Shimada et al. have published a series of papers on the analysis of neurosteroid sulfates and neutral neurosteroids found in rat brain using LC in combination with mass spectrometry.20-22 However, as yet no convincing evidence for the presence of intact steroid sulfates in rat brain has been presented. Because of the low levels of neurosteroids in brain, a combination of nanoscale liquid chromatography (nano-LC) and microES mass spectrometry is the method of choice for the analysis of neurosteroids. Nano-LC-ES mass spectrometry provides higher sensitivity than conventional LC-ES mass spectrometry, because of the higher analyte concentration in the eluting peaks when a nano-LC column is used and the inherent gain in ionization efficiency when low-flow rate ES is used. The current paper describes a method for the extraction and isolation of steroids from rat brain tissue and their analysis by nano-LC-ES mass spectrometry. The method was designed to permit multicomponent analysis of both neutral unconjugated steroids and sulfated steroids with a wide range of polarities. Neutral unconjugated steroids and steroid sulfates were separated into two fractions by anion-exchange chromatography. Oxosteroids in the neutral unconjugated steroid fraction were converted to their oximes13 and isolated by cation-exchange chromatography. Then the oxosteroid oximes and the steroid sulfates were subjected to nanoLC-ES mass spectrometry for detection and characterization. EXPERIMENTAL SECTION Materials. Solvents were of analytical grade. Water was from a Milli-Q water system (Millipore, Molsheim, France). Reversedphase packing material (Genesis C18, 3 µm) from Jones Chromatography Ltd. (Mid Glamorgan, U.K.) was used to pack nanoLC analytical columns and precolumns (100-µm i.d. 375-µm o.d., length of analytical column 350 mm, precolumn 50 mm).23 Bondesil C18 (30-40 µm) was from the former Analytichem International (presently Varian, Palo Alto, CA). The cation exchanger SP-LH-20 (sulfohydroxypropyl Sephadex LH-20) was previously synthesized as described by Axelson and Sjo¨vall.24 The (10) Chatman, K.; Hollenbeck, T.; Hagey, L.; Vallee, M.; Purdy, R.; Weiss, F.; Siuzdak, G. Anal. Chem. 1999, 71, 2358-2363. (11) Griffiths, W. J.; Liu, S.; Yang, Y.; Purdy, R.; Sjo ¨vall, J. Rapid Commun. Mass Spectrom. 1999, 13, 1595-1610. (12) Shackleton, C. H. L.; Chuang, H.; Kim, J.; de la Torre, X.; Segura, J. Steroids 1997, 62, 523-529. (13) Liu, S.; Griffiths, W. J.; Sjo ¨vall, J. Rapid Commun. Mass Spectrom. 2000, 14, 390-400. (14) Johnson, D. W.; ten Brink, H. J.; Jakobs, C. J. Lipid Res. 2001, 42, 16991705. (15) Komatsu, F.; Morioka, M.; Fujita, Y.; Sugahara, K.; Kodama, H. J. Mass Spectrom. 1995, 30, 698-702. (16) Bowers, L. D.; Sanaollah J. Chromatogr., B 1996, 678, 61-68. (17) Yang, Y.; Griffiths, W. J.; Nazer, H.; Sjo ¨vall, J. Biomed. Chromatogr. 1997, 11, 240-255. (18) Ma, Y. C.; Kim, H. Y. J. Am. Soc. Mass Spectrom. 1997, 8, 1010-1020. (19) Bean, K. A.; Henion, J. D. J. Chromatogr., B 1997, 690, 65-75. (20) Shimada, K.; Mukai, Y.; Yago, K. J. Liq. Chromatogr. Relat. Technol. 1998, 21, 765-775. (21) Mitamura, K.; Yatera, M.; Shimada, K. Anal. Sci. 1999, 15, 951-955. (22) Nakajima, M.; Yamato, S.; Shimada, K. Biomed. Chromatogr. 1998, 12, 211216. (23) Liu, S.; Griffiths, W. J.; Sjo ¨vall, J. Anal. Chem. 2003, 75, 791-797. (24) Axelson, M.; Sjo ¨vall, J. J. Chromatogr. 1979, 186, 725-732.

5836

Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

anion exchanger Lipidex DEAP (diethylaminohydroxypropyl Sephadex LH-20) was from Packard Instrument Co. (Downers Grove, IL). Steroids were from previous studies.11,13 [3β,11,11-2H3]Allopregnanolone (3R-hydroxy-5R-pregnan-20-one) sulfate was prepared in this laborotary,25 and it was quantitated against standard allopregnanolone sulfate. [3,4-13C2]-Progesterone was from EURISO-TOP (Saint-Aubin Cedex, France). [19,19,19-2H3]Testosterone (17β-hydroxyandrost-4-en-3-one) was a kind gift from Prof. Kasuya and Dr. Shinohara, Tokyo University of Pharmacy and Life Science, Japan. Radioactive compounds, 1,2-di[1-14C]-palmitoyl phosphatidylcholine (4.14 GBq/mmol) and [4-14C]-cholesterol (1.92 GBq/ mmol) were from Amersham Life Science (Buckinghamshire, U.K.). [1,2,6,7-3H4]-DHEA sulfate (2220 GBq/mmol), [1,2,6,7-3H4]DHEA (3400 GBq/mmol), [1,2,6,7-3H4]-progesterone (3600 GBq/ mmol), and [7-3H]-pregnenolone (780 GBq/mmol) were from NEN Life Science Products, Inc. (Boston MA). Rat brain samples used for analysis were kindly provided by Prof. Tomas Ho¨kfelt, Department of Neuroscience, Karolinska Institutet. The sampling was approved by the ethics committee of Karolinska Institutet. Male and female Sprague-Dawley rats, body weight about 250 and 275 g, respectively, were housed separately for one week after arrival at the laboratory. Animals under anesthesia with sodium pentobarbital were decapitated. The entire brain was removed and collected in a glass tube containing ethanol. In some cases, hippocampus and amygdala were isolated from the brain for separate analysis. Extraction and Group Isolation of Neutral Steroids and Steroid Sulfates. The sample preparation procedure for the analysis of neutral oxosteroids and sulfated steroids from rat brain is outlined in Figure 1. Brain samples ranging from 50 to 300 mg of wet weight were used. Either the entire brain or isolated amygdala or hippocampus regions were homogenized and aliquots analyzed. For the analysis of steroid sulfates, 1.7 ng of [3β,11,112H ]-allopregnanolone sulfate and 100 000 cpm of [1,2,6,7-3H ]3 4 DHEA sulfate were added as internal standard and radioactivity tracer, respectively, to the brain pieces prior to homogenization. For the analysis of neutral steroids, 1.0 ng of [3,4-13C2]-progesterone and 64 000 cpm of [1,2,6,7-3H4]-DHEA were added as internal standards and radioactivity tracer, respectively, to the homogenized brain. Homogenization was performed in ethanol using a glass homogenizer, followed by sonication for 1 min using a Soniprep 150 disintegrator set at 20-µm amplitude (Sanyo Gallenkamp PLC, Loughborough, U.K.), with subsequent ultrasonication for 10 min in an ultrasonic bath. Then water was added to dilute the ethanol content of the homogenate to 70%, and the sample was ultrasonicated for a further 5 min. The mixture was then centrifuged, and the residue was extracted with 1 mL of 70% ethanol and again centrifuged. The supernatants were combined and applied to a bed of Bondesil C18 (100 mg) packed in a Pasteur pipet followed by a bed of the cation exchanger (SP-LH-20, 5 cm × 0.4 cm, in H+ form) packed in a glass column. The effluent was collected, and the two beds were washed with 2 mL of 70% methanol. The combined effluent and wash were then applied to a 4 × 0.4 cm column of Lipidex-DEAP in the acetate form.26 The effluent from the Lipidex-DEAP column and a wash with 3 mL of (25) Anderson, R. A.; Baillie, T. A.; Axelson, M.; Cronholm, T.; Sjo ¨vall, K.; Sjo ¨vall, J. Steroids 1990, 55, 443-457. (26) Sjo ¨vall, J.; Axelson, M. Vitam. Horm. 1982, 39, 31-143.

Figure 1. Scheme for the extraction and isolation of steroids from brain tissue.

70% methanol constituted the neutral steroid fraction. The LipidexDEAP column was further washed with 2 mL of 0.25 M formic acid in 70% methanol, and this wash was discarded. Then steroid sulfates were eluted in 4 mL of 0.3 M ammonium acetate buffer, pH 6.5, in 70% methanol. One hundred milligrams of hydroxyammonium chloride (Fluka, Buchs, Switzland) was added to the neutral fraction and allowed to react for 3 h at 70 °C. After this time, the reaction solution was evaporated to almost dryness under a stream of nitrogen and redissolved in 2 mL of 20% methanol. The resulting solution was applied to a 30-mg bed of Bondesil C18. After a wash of the bed with 2 mL of 20% methanol, nitrogen was passed through the bed to remove the interstitial water. Steroid oximes were then eluted with 1 mL of methanol. This eluate was then applied to an 8 cm × 0.4 cm column of SP-LH-20 in the H+ form.24 Following a wash of the SP-LH-20 column with 5 mL of methanol to remove unretarded compounds, e.g., neutral nonoxosteroids, steroid oximes were eluted with 4 mL of 0.3 M ammonium hydroxide in 70% methanol.13,24 This solution was evaporated to dryness and reconstituted in 100 µL of 20% methanol, ready for injection into the nano-LC-ES mass spectrometry system. The steroid sulfate fraction eluted from the Lipidex-DEAP anion-exchange column (Figure 1), was evaporated to almost dryness under a stream of nitrogen, dissolved in 20% methanol, and applied to a bed of Bondesil C18, 10 mg, packed in a Pasteur pipet. Following a wash with 1 mL of water, steroid sulfates were eluted with 100 µL of methanol. This solution was evaporated to dryness under a nitrogen stream and dissolved in 100 µL of 10% methanol. Aliquots of this solution were injected into the nanoLC-ES mass spectrometry system. The recovery of steroids during the sample preparation procedure was evaluated by the addition of 3H-labeled steroids at different stages of the procedure. [3H4]-DHEA sulfate, [3H4]-

progesterone, [3H4]-DHEA, and [3H]-pregnenolone were used as markers for steroid sulfates, 3-oxosteroids, 17-oxosteroids, and 20-oxosteroids, respectively. Radioactivity was determined in an LKB 1211 Minibeta liquid scintillation counter (Wallac Oy, Turku, Finland) using OptiPhase Hisafe 3 (Wallac) as scintillation liquid. For the analysis of a human plasma sample, the plasma was first diluted 10 times with saline. Then 50 µL of the diluted plasma was added to a tube containing 2 mL of 70% ethanol, to which 100 µL of a solution of [3H4]-DHEA sulfate (108 100 cpm) in 50% methanol had been added as internal standard. This solution contained about 42, 117, 335, 91, and 10 pg of the 3H0, 3H1, 3H2, 3H , and 3H species of DHEA sulfate, respectively, as estimated 3 4 from the cluster of deprotonated molecules in the ES mass spectrum. After centrifugation, the sample was treated as described above for the brain samples with the exception of the homogenization and ultrasonication. Nano-LC-ES Mass Spectrometry. Nano-LC-ES mass spectrometry23 was performed on an AutoSpec-OATOFFPD (Micromass, Manchester, U.K.) hybrid double-focusing magnetic sectororthogonal acceleration (OA) time-of-flight (TOF) tandem spectrometer and Quattro Ultima and Quattro Micro tandem-quadrupole instruments (Micromass). The nano-LC-ES system, which is described fully in ref 23, consists of two syringe pumps (A and B), an injector, a T-splitter (splitter A), a precolumn, another T-splitter (splitter B), the analytical column, and a zero dead volume union to which the ES emitter (PicoTip, 15 µm, New Objective Inc., Cambridge, MA) is attached.23 The nano-LC columns were packed with 3-µm Genesis C18 particles. The ES emitter was directly inserted into the ES interface of the sectorOATOF instrument. When the tandem quadrupole instruments were used, a transfer capillary (30 cm, 25-µm i.d.) was used to connect the column end to the ES emitter. For the analysis of steroid sulfates from brain, the mobile phases were methanol/water (1:9, v/v) in pump A, and methanol/ water (8:2, v/v) in pump B, both containing 10 mM ammonium acetate. When cholesterol sulfate was to be analyzed, the solvent in pump B consisted of methanol/2-propanol/water (75:20:5, by volume) containing 10 mM ammonium acetate. For the analysis of steroid sulfates in plasma, and for the analysis of steroid oximes, the solvent in pump B was methanol/water (19:1, v/v) containing 10 mM ammonium acetate. Twenty-microliter solutions of brain samples, or 2-20 µL of a reference mixture, were injected from a 20-µL loop when mobile phase A was being pumped through the precolumn at a pressure of ∼160 bar, with splitter A closed and splitter B open. After 20 min, the time needed for the sample to be transferred to and be sorbed on the precolumn, splitter B was closed and splitter A opened. Then pump A was stopped and a gradient program for separation of analytes initiated. During the gradient elution, the pumps were operated at a total flow rate of 20-30 µL/min, the flow rate through the columns being 0.2-0.3 µL/min. Data acquisition was started when the gradient was started or after a suitable delay. Details of the gradients employed are given in the captions to Figure 2, 3, and 5. Mass Spectrometry of Steroid Sulfates. The sector-TOF instrument was used for the analysis of steroid sulfates. Negativeion ES spectra were recorded. The needle (emitter) voltage and skimmer voltage were approximately -5.2 and -4.5 kV, respecAnalytical Chemistry, Vol. 75, No. 21, November 1, 2003

5837

tively. The accelerating potential was -4 kV. A countercurrent of nitrogen gas (100 L/h) was employed to aid droplet desolvation. The instrument was tuned to 3000 resolution (10% valley definition) and the m/z range 416-360 Th (thomson) scanned at a rate of 10 s/decade. When cholesterol sulfate was analyzed, the m/z range scanned was 360-500 Th. The tandem quadrupole instruments were also used for the analysis of steroid sulfates. The needle (emitter) voltage was optimized at ∼-2.0 kV. The cone voltage was -90 V, and a cone gas flow of 50 L/h was used to aid evaporation of solvent. Unit mass resolution was used in all experiments. A scan rate of 100 Th/s was used for both mass scans and product ion scans. A collision energy of 30 eV was used in tandem mass spectrometry (MS/MS) experiments. Argon was the collision gas at a pressure of ∼3 × 10-3 mbar. Mass Spectrometry of Steroid Oximes. The tandem quadrupole instruments were used for the analysis of steroid oximes. The capillary (emitter) voltage was optimized at ∼1.8 kV. A cone voltage of 40 V was used. A cone gas flow of 50 L/h was used to aid evaporation of solvent. Unit mass resolution was used in all experiments. A scan rate of 100 Th/s was used for both mass scans and product ion scans. A collision energy of 30 eV was used for all MS/MS experiments. Argon was used as the collision gas at a pressure of 3 × 10-3 mbar. Multiple-reaction monitoring (MRM) experiments were carried out with a dwell time of 0.5 s for each transition and an interscan delay of 0.05 s. Mixtures of reference compounds were used to determine the retention times on the nano-LC column of different steroid sulfates and steroid oximes. One mixture of reference compounds consisted of the sulfates of 7-oxopregnenolone (3β-hydroxypregn-5ene-7, 20-dione), DHEA, pregnenolone, epiallopregnanolone (3βhydroxy-5R-pregnan-20-one), and allopregnanolone, which were dissolved in 10% methanol at concentrations of 50, 100, 100, 100, and 100 pg/µL, respectively. A second mixture of reference compounds was obtained by derivatization of a mixture of oxosteroids containing DHEA, pregnenolone, progesterone, allopregnanolone, and [2H3]-testosterone with hydroxylamine hydrochloride. A 10-ng sample of each steroid was used, and the products were purified by C18 solid-phase extraction. Serial dilution with 20% methanol generated working solutions at different concentrations. A third reference mixture containing oximes of four pregnanolone isomers was made by derivatizing the isomers separately and then mixing them in 20% methanol. RESULTS AND DISCUSSION Extraction and Group Separation. Many procedures have been described for the extraction of steroids from biological samples. Extraction of steroids from brain with acetone/ethanol3 and from homogenates in saline or water with organic solvents4 has been reported. An ideal procedure should aim at good recoveries of the analytes with the removal of interfering compounds. However, it is not a simple matter to evaluate recoveries of endogenous steroids in extracts from tissues, because the added tracers or standards may not reflect the true state of the endogenous compounds in the tissue (tracers are added to a homogenate or a solution, not into undisturbed tissue). In this work, we have chosen ethanol as the solvent for homogenization and extraction since it is a good solvent for neutral as well as conjugated steroids and readily penetrates cell membranes. 5838

Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

Following homogenization, the initial ethanol extract was diluted with water to give a solvent of 70% ethanol. This results in precipitation of some nonpolar lipids in the extract. Passage of the 70% ethanol extract solution through a bed of Bondesil C18 removes additional nonpolar lipids by sorption. About 95% each of added 14C-labeled cholesterol and 14C-labeled phosphatidylcholine was removed in this step. It should be noted that in this analytical method solid-phase extraction is used for the removal of contaminants, rather than for extraction of analytes. Passage of the extract through a bed of the lipophilic cation exchanger SP-LH-20 has the function to remove cationic compounds that may interfere with the later analysis of steroid oximes. Because neutral unconjugated steroids and steroid sulfates have very different physical and chemical properties that affect their chromatographic and ionization behavior, they cannot be analyzed with sufficient sensitivity by a single mass spectrometric method. In the current study, the lipophilic anion exchanger Lipidex-DEAP was used to separate neutral unconjugated steroids from steroid sulfates.26 Neutral unconjugated steroids are not retained by the anion exchanger and appear in the effluent and a 70% methanol wash fraction. A subsequent wash of the anion exchanger with 0.25 M formic acid in 70% methanol yields a fraction containing acids weaker than sulfates, e.g., glucuronides. This fraction was not analyzed in the present study. Finally, steroid sulfates are eluted with 4 mL of 0.3 M ammonium acetate buffer, pH 6.5, in 70% methanol. Different solvents were tested for this separation procedure in order to achieve optimal purification of steroid sulfates. Since most interfering compounds are lipophilic, attempts were made to emphasize the reversed-phase partitioning of such compounds into the lipophilic gel using solvents with a lower content of methanol. However, when the methanol content of the eluting solvent was decreased below 50%, the elution of added 3H-labeled steroid sulfates tailed. Column dimensions were studied in order to minimize solvent volumes. At column heights below 2 cm (0.4-cm i.d.), [3H]-DHEA sulfate began to appear in the neutral steroid fraction, indicating insufficient sorption or solvent channelling through the column bed, so 4-cm columns were used for all experiments. Up to this stage, the recovery of 3H-labeled DHEA sulfate added to the tissue extract was 85% (n ) 5); the recoveries of 3H-labeled DHEA, progesterone, and pregnenolone added to the tissue extract were 90, 87, and 87%, respectively (n ) 5). Appearance of 3H-labeled DHEA sulfate in the neutral fraction and appearance of 3H-labeled DHEA, progesterone, and pregnenolone in the sulfate fraction were not observed, in agreement with previous results.26 Since many important unconjugated neurosteroids contain an oxo group, this functional group was used as a handle for selective isolation and for derivatization. The oxo group was converted to an oxime, and then the steroid oximes were isolated by sorption on a lipophilic ion exchanger in a nonaqueous solvent.13,24 Recoveries of 3H-labeled progesterone, DHEA, and pregnenolone added to the neutral steroid fraction were above 90% through the derivatization and isolation procedure. The recovery of 3H-labeled DHEA was 73% (n ) 4) through the whole procedure, starting with the extraction. The steroid sulfates were eluted from the Lipidex DEAP column in 4 mL of 70% methanol containing 0.3 M ammonium acetate at pH 6.5. To permit injection of most of this fraction into

the nano-LC-ES system, it had to be concentrated to ∼50 µL and the salts removed. A last step of solid-phase extraction was designed for this purpose and for final purification. The average recovery of added [3H]-DHEA sulfate through the whole procedure was 80% (n ) 5). Nano-LC-ES Mass Spectrometry of Steroid Sulfates. Although tandem mass spectrometry makes it possible to distinguish synthetic pregnanolone sulfate isomers,11 analysis of mixtures in biological samples requires separation of isomers prior to mass spectrometry. The present on-line system separates the sulfates of three pregnanolone isomers from each other and from the sulfates of DHEA, pregnenolone, and 7-oxopregnenolone.23 The limit of detection in reconstructed ion chromatograms (RIC) was (S/N ) 10) 3 pg injected for the pregnanolone sulfate isomers, when the m/z range 416-360 was scanned. Detection of Steroid Sulfates in Brain. As pregnenolone sulfate and DHEA sulfate have previously been analyzed by indirect methods and appear to be present in rat brain,1 their detection in the intact form was our initial aim. However, neither pregnenolone sulfate nor DHEA sulfate was detected given the detection limit of our method, either in whole brain or in isolated areas of brain (amygdala or hippocampus), when brain samples of 50-300 mg were extracted. Figure 2 shows the results from an analysis of brain samples, in which a sample solution corresponding to 60 mg of brain was injected on to the nano-LC column interfaced to the sector-OATOF mass spectrometer. The RIC for the [M - H]- ion of pregnenolone sulfate (m/z 395) (Figure 2a) showed no peak above the noise level at a retention time of 38 min, the time at which reference pregnenolone sulfate eluted (Figure 2b). There was a peak at a retention time of 61.5 min, i.e., much later than reference pregnenolone sulfate. When a brain sample, to which 2 ng of pregnenolone sulfate had been added to 300 mg of brain, was analyzed an intense peak was detected (signal/noise ratio ∼100) at 38.1 min in the RIC as shown in Figure 2b. In both cases, the added internal standard, [2H3]allopregnanolone sulfate (1.7 ng in 300 mg of brain) was detected with a retention time of 44.4 min in the RIC of its [M - H]- ion at m/z 400 as shown in Figure 2c. The same brain samples were also analyzed by nano-LC interfaced to a tandem quadrupole instrument making use of the precursor ion scan facility. A precursor ion scan for m/z 97 was performed, and the acquired spectra were interrogated for fragment ions of m/z 97 originating from precursor ions of m/z 395 or 400. In a precursor ion scan for m/z 97, the first quadrupole is scanned to allow successive ions of increasing m/z to be transmitted to the collision cell. The final quadrupole is set to transmit ions of m/z 97. When a precursor ion containing a sulfate ester group is transmitted through the first quadrupole and fragments in the collision cell to give a product ion of m/z 97, this ion (m/z 97) will then be transmitted through the final quadrupole and generate a signal. A precursor ion scan for m/z 97 will only detect the compounds that generate this specific ion, so this scan mode is more specific, and also gives a better S/N ratio than a simple mass scan. Again there were no peaks observed at the retention times expected for pregnenolone sulfate and DHEA sulfate (data not shown) in unadulterated brain samples. The peak corresponding to the [M - H]- ion of the internal standard [2H3]-allopregnanolone sulfate, present at a level of 1.7 ng in 300 mg of brain, was observed at a

Figure 2. (a) RIC for pregnenolone sulfate (m/z 395) from a brain sample. The arrow indicates where pregnenolone sulfate is expected to elute. (b) RIC for pregnenolone sulfate (m/z 395) from a brain sample to which 2 ng of the steroid had been added to 300 mg of brain. (c) RIC for [2H3]-allopregnanolone sulfate (m/z 400) added as an internal standard to a brain sample (1.7 ng in 300 mg of brain). Spectra were recorded on an sector-OATOF instrument. Steroid sulfates were extracted and purified from 300 mg of brain. Twentymicroliter sample solutions (corresponding to 60 mg of brain) were injected onto the capillary column. The gradient program was started at 30% B solvent and kept constant for 20 min, then increased to 50% B in 5 min, then increased to 100% B in 70 min, and kept at 100% B for 60 min. The A and B solvents were 10 and 80% aqueous methanol, respectively, both containing 10 mM ammonium acetate. The total flow rate from the pumps was 30 µL/min and the postinjector split ∼100:1. Data acquisition was started 30 min after the start of the gradient. The X axis represents the time (min) from the start of the acquisition.

retention time of ∼44.7 min. According to these results, we estimated that the levels of pregnenolone sulfate and DHEA sulfate in rat brain were below 0.3 ng/g as estimated from the signal given by the added pregnenolone sulfate (Figure 2b). These results are consistent with those recently published by Higashi et al.27,28 using radioimmunoassay and show that the sulfate ester is not a major form of DHEA and pregnenolone in rat brain. The reasons for the discrepancy between our results and those published earlier3,4 are probably methodological. The earlier (27) Higashi, T.; Daifu, Y.; Shimada, K. Steroids 2001, 66, 865-874. (28) Higashi, T.; Daifu, Y.; Ikeshima, T.; Yagi, T.; Shimada, K. J. Pharm. Biomed. Anal. 2003, 30, 1907-1917.

Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

5839

results3,4 were obtained by the analysis of free steroids liberated by solvolysis of fractions considered to contain steroid sulfates, whereas our method and those of Higashi et al. measured intact steroid sulfates directly. It would thus appear that the steroids analyzed by Corpechot et al.3,4 were not derived from the solvolysis of steroid sulfates, but from some other solvolyzable form of conjugated steroid present in the steroid sulfate fraction isolated from brain. Cholesterol sulfate has been reported by Iwamori et al.29 to be present in rat brain at a level of 15-50 µg/g, but their analytical methods were also indirect and involved solvolysis. To detect cholesterol sulfate in our experiments, the m/z range scanned was extended to 280-500 Th. The results from an analysis of a brain sample are shown in Figure 3. Two peaks were detected in the RIC for the [M - H]- ion of cholesterol sulfate (m/z 465.3)

(Figure 3a). Retention time data indicated that the peak at 88.1 min was cholesterol sulfate. The same brain sample was also analyzed on the tandem quadrupole instrument using a precursor ion scan for the fragment ion of m/z 97. There was now only one peak detected in the RIC for m/z 465 at 87.3 min as shown in Figure 3b. The level of cholesterol sulfate in the samples analyzed in the current study was estimated to be ∼1.2 µg/g, using [2H3]allopregnanolone sulfate as an internal standard and after correction for any response difference in nano-LC-ES. This level is much lower than the reported levels determined by GC/MS analysis of free cholesterol liberated by solvolysis.29 The reason for this discrepancy may be that, as in the case of pregnenolone and DHEA, some other form of solvolyzable steroid conjugate is present in the sulfate fraction from brain. Interestingly, many compounds were detected with retention times in the range of 60-80 min. The mass spectra showed many peaks at the same nominal mass as [M - H]- ions of steroid sulfates, e.g., m/z 367, 381, 395, 397, and 409. Lysophosphatidic acids (CnH2n+1CO2CH2CH(OH)CH2OPO3H2) or their plasmalogen (ether) analogues can have the same nominal mass as steroid sulfates. Lysophosphatidic acids behave in a manner similar to steroid sulfates in our sample preparation procedure, including reversed-phase and ion-exchange chromatography. Numerous lysophosphatidic acid species were found in the steroid sulfate fraction. Significantly, these lysophosphatidic acids also produce a fragment ion at m/z 97 (H2PO4-) upon collision-induced dissociation (CID). To distinguish between a steroid sulfate and a lysophosphatidic acid of the same nominal mass in a mass spectrum a resolution of 13 000 would be needed. Alternatively, a precursor ion scan for fragment ions of m/z 153 characteristic of the glycerol phosphate group would allow the identification of lysophosphatidic acids.30,31 In our method, the steroid sulfates are separated from lysophosphatidic acids by the nano-LC so that they can be distinguished by retention time. It should be noted that the injection of a large volume of biological sample onto the nano-LC system results in an initial decrease in the flow rate through the nano-LC column. This is probably due to formation of precipitates in the precolumn, causing an increase of the column back pressure. In our experiments, variation of retention time was observed, and the relative variation could be more than 3.5%. This is a common problem in nano-LC and makes the direct comparison of retention times of pure reference compounds to those of samples difficult. The addition of an internal standard to the sample circumvents this problem to some extent, as the compound can be located in the RIC. Two or three isotope-labeled internal standards added to the sample may serve to calibrate the retention time scale and correct for the variation of retention times more effectively. Washing the precolumn and the analytical column with a strong solvent after each injection of biological sample is important to maximize the lifetime of the column. It also helps to reduce the variation of retention times between different injections by removing any remaining phospholipids and other nonpolar compounds. This wash step was carried out easily by an injection of 20 µL (9 times the column volume) of a mixture of methanol and 2-propanol

(29) Iwamori, M.; Moser, H. W.; Kishimoto, Y. Biochim. Biophys. Acta 1976, 441, 268-79.

(30) Bru ¨ gger, B.; Erben, G.; Sandhoff, R.; Wieland, F. T.; Lehmann, W. D. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 2339-2344. (31) Griffiths, W. J. Mass Spectrom. Rev. 2003, 22, 81-152

Figure 3. (a) RIC for cholesterol sulfate [M - H]- ions (m/z 465) from m/z scans (280-500 Th) recorded on the sector-OATOF instrument. (b) RIC for m/z 465 from precursor ion scans for m/z 97 recorded on the tandem quadrupole instrument. Steroid sulfates were extracted and purified from 300 mg of brain. Twenty microliters of sample solution (corresponding to 60 mg of brain) was injected onto the nano-LC column. Nano-LC conditions were as in Figure 2 except that the B solvent was a mixture of methanol/2-propanol/water (75/ 20/5) containing 10 mM ammonium acetate. Data acquisition was started 40 min after the start of the gradient. The X axis represents the time (min) from the start of the acquisition.

5840 Analytical Chemistry, Vol. 75, No. 21, November 1, 2003

Figure 4. CID spectra of the protonated oximes of (a) testosterone, (b) pregnenolone, (c) DHEA, and (d, e) progesterone. The solvent was methanol for (a-d) and 95% methanol containing 10 mM ammonium acetate for (e). The spectra were recorded on a tandem quadrupole instrument. Argon was used as collision gas, and the collision energy of 30 eV.

(1:1) with the two splitters closed. By using this wash step, deterioration of the column was prevented during the period of this study. The average standard variations of the retention times of the steroid sulfate references and the steroid oxime references were