Ultrasensitive Quantification of Serum Vitamin D Metabolites Using

Feb 12, 2010 - DOI: 10.1002/mas.21353. Christopher A. Jackson, Neelu Yadav, Sangwon Min, Jun Li, Eric J. Milliman, Jun Qu, Yin-Chu Chen, Michael C. Yu...
5 downloads 0 Views 1MB Size
Download PDF
Anal. Chem. 2010, 82, 2488–2497

Ultrasensitive Quantification of Serum Vitamin D Metabolites Using Selective Solid-Phase Extraction Coupled to Microflow Liquid Chromatography and Isotope-Dilution Mass Spectrometry Xiaotao Duan,†,‡ Bianca Weinstock-Guttman,§ Hao Wang,†,‡ Eunjin Bang,† Jun Li,†,‡ Murali Ramanathan,† and Jun Qu*,†,‡ Department of Pharmaceutical Sciences, University at Buffalo, State University of New York, Amherst, New York 14260-1200, New York State Center of Excellence in Bioinformatics and Life Sciences, 701 Ellicott Sreet, Buffalo, New York 14203, and Baird Multiple Sclerosis Center and Department of Neurology, University at Buffalo, State University of New York, Amherst, New York 14260-1200 The capacity for quantification of active metabolites of vitamin D (VitD) is highly valuable to evaluate the risks and therapies for numerous diseases such as multiple sclerosis. However, the extremely low circulating levels and poor detectability of some dihydroxyl metabolites such as the 1r,25-dihydroxy-VitD3 constitute a daunting challenge. Based on the combination of a selective solid-phase extraction (SPE) and a microflow liquid chromatography tandem mass spectrometry (µLC-MS/ MS), we developed an ultrasensitive method for the robust, selective, and accurate quantification of four key VitD metabolites, including 25-hydroxy-VitD2, 25hydroxy-VitD3, 24(R),25-dihydroxy-VitD3, and 1r,25dihydroxy-VitD3, in serum samples. A one-step derivatization was employed to improve the ionization efficiency of the metabolites. The SPE procedure was optimized so that the analytes were selectively extracted from serum, while the sample matrix was substantially simplified. By eliminating majority of undesirable compounds from the matrix, the selective SPE enabled a high sample loading volume on the µLC column without causing overcapacity of the µLC column and thus helped to achieve ultralow detect limits in serum. An on-column sample focusing approach was employed to prevent band-broadening, and a sufficient µLC separation was achieved to eliminate endogenous interferences and to minimize ion suppression effect. Detect limits of the four metabolites ranged from 0.5-1 pg/mL, and the linearity was excellent for all compounds. The method showed high quantitative accuracy (error < 13.8%) and precision (CV < 14.1%). * To whom correspondence should be addressed. Phone: (716)645-2844 x283. Fax: (716)645-3693. E-mail: [email protected]. Corresponding author address: Department of Pharmaceutical Sciences, 537 Cooke Hall, University at Buffalo, State University of New York, Buffalo, NY 14260-1200. † Department of Pharmaceutical Sciences, University at Buffalo, SUNY. ‡ New York State Center of Excellence in Bioinformatics and Life Sciences. § Baird Multiple Sclerosis Center and Department of Neurology, University at Buffalo, SUNY.

2488

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

For 1r,25-dihydroxy-VitD3, a lower limit of quantification (LLOQ) of 5 pg/mL was validated. This high level of sensitivity, for the first time, enabled the robust and consistent LC/MS/MS-based analysis of the four metabolites in a large-scale clinical investigation. Serum samples from 281 multiple sclerosis patients and 22 healthy subjects were analyzed, and it was discovered that the levels of both 24(R),25-dihydroxy-VitD3 and 1r,25-dihydroxy-VitD3 were significantly lower in patients than healthy subjects (P < 0.05). This novel observation may imply that the incidence of multiple sclerosis is inversely associated with the levels of the two metabolites. Moreover, the method was highly robust and reproducible as evaluated extensively in the clinical analysis; therefore, it could serve as a more selective and accurate alternative to immunoassay for large-scale clinical studies. Vitamin D (VitD, in the forms of VitD2 and VitD3) plays a critical role in many important biological processes, including maintenance of calcium homeostasis, immunomodulation, and cell differentiation.1 VitD itself is not biologically active and requires further metabolism that generates the active metabolites.2 Studying the physiological actions of VitD metabolites would contribute greatly to the mechanism research, diagnosis, staging, and therapy of numerous diseases, such as multiple sclerosis, diabetes, cancer, osteoporosis, microbial infections, and cardiovascular diseases.1,3 For an example, the 25-hydroxy VitD [25(OH)VitD] metabolite is clinically used as marker for VitD deficiency.3 One of its metabolite, the 1R, 25-dihydroxy-VitD [1,25(OH)2VitD], is considered the primary biologically active form of VitD and responsible for stimulating intestinal calcium absorption, modulating immune response, and maintaining (1) Smolders, J.; Damoiseaux, J.; Menheere, P.; Hupperts, R. J. Neuroimmunol. 2008, 194, 7–17. (2) Houghton, L. A.; Vieth, R. Am. J. Clin. Nutr. 2006, 84, 694–697. (3) Zerwekh, J. E. Am. J. Clin. Nutr. 2008, 87, 1087S–1091S. 10.1021/ac902869y  2010 American Chemical Society Published on Web 02/12/2010

calcium homeostasis.4 Another dihydroxyl metabolite, the 24(R), 25-dihydroxy-VitD [24,25(OH)2VitD], has been demonstrated as a critical component for healing processes in tissues and bones.5 Therefore, the ability to quantify these circulating VitD metabolites would be highly valuable for the clinical studies of diseases that are associated with the deficiency and/ or metabolic dysfunction of VitD. A wide variety of methodologies have been developed for the quantification of VitD metabolites in serum/plasma samples. Among them, immunoassays including the radioimmunoassay (RIA) and enzyme immunoassay (EIA) are the most prevalent.3 Nevertheless, most of the RIA and EIA methods fall short in that only one metabolite can be measured per assay and that the selectivity, accuracy, and interbatch/lab reproducibility could be problematic.3,6,7 Other methods include competitive proteinbinding assay,8 automated chemiluminescent protein-binding assay,8 and HPLC-UV.9 More recently, liquid chromatography tandem mass spectrometry (LC/MS/MS) has emerged as a promising alternative, in that these methods provide superior selectivity, accuracy, and sensitivity and does not require laborious sample preparation that are necessary for many other techniques. 10-13 Although LC/MS/MS-based methods could readily measure 25(OH)VitD, which presents at relatively high levels in serum, quantification of the dihydroxyl metabolites can be highly challenging. For example, the 1,25(OH)2VitD3, which is highly active and of primary interest for the research of many diseases,3,4 is present at extremely low levels (low pg/mL) in human serum that are significantly lower than the detection limits of conventional LC/MS/MS methods. The relatively poor ionization efficiency of this metabolite, further compounds the problem.14 In order to improve the sensitivity for 1,25(OH)2VitD3, a recent study employed a derivatization procedure and achieved a lower limit of quantification (LLOQ) of 25 pg/mL using 0.5 mL of human serum.14 Though the derivatization approach enabled a more sensitive LC/MS/MS analysis than possible previously, the LLOQ achieved is not sufficiently sensitive for the measurement of 1,25(OH)2VitD3 in all serum samples, especially for these from patients of certain diseases such as multiple sclerosis, where the serum levels of this metabolite are expected to be lower than those in healthy subjects (discussed below). Recently, we developed a novel analytical strategy for ultrasensitive quantification of drugs in highly complex biological (4) Norman, A. W.; Okamura, W. H.; Farach-Carson, M. C.; Allewaert, K.; Branisteanu, D.; Nemere, I.; Muralidharan, K. R.; Bouillon, R. J. Biol. Chem. 1993, 268, 13811–13819. (5) Seo, E. G.; Norman, A. W. J. Bone Miner. Res. 1997, 12, 598–606. (6) Carter, G. D. Clin. Chem. 2009, 55, 1300–1302. (7) Roth, H. J.; Schmidt-Gayk, H.; Weber, H.; Niederau, C. Ann. Clin. Biochem. 2008, 45, 153–159. (8) Haddad, J. G.; Chyu, K. J. J. Clin. Endocrinol. Metab. 1971, 33, 992–995. (9) Jones, G. Clin. Chem. 1978, 24, 287–298. (10) Tsugawa, N.; Suhara, Y.; Kamao, M.; Okano, T. Anal. Chem. 2005, 77, 3001–3007. (11) Maunsell, Z.; Wright, D. J.; Rainbow, S. J. Clin. Chem. 2005, 51, 1683– 1690. (12) Higashi, T.; Awada, D.; Shimada, K. Biol. Pharm. Bull. 2001, 24, 738– 743. (13) Vogeser, M.; Kyriatsoulis, A.; Huber, E.; Kobold, U. Clin. Chem. 2004, 50, 1415–1417. (14) Aronov, P. A.; Hall, L. M.; Dettmer, K.; Stephensen, C. B.; Hammock, B. D. Anal. Bioanal. Chem. 2008, 391, 1917–1930.

samples.15-17 Because ESI-MS/MS is a concentration-dependent detector, we employed a microflow capillary LC (µLC), which provides much lower sample dilution than does conventional LC, and achieves high absolute sensitivity for target analyte. However, the µLC suffers from shortcomings associated with its small loading capacity, which tends to counteract the sensitivity gain, especially when quantification in complex biological matrices in required.15 To overcome this limit, we devised a solid-phase extraction (SPE) procedure that selectively concentrated the analytes, while removing most matrix components and thus enabling loading of a relatively large injection volume (I.V.) on µLC columns. This selective-SPE/µLC/MS/MS strategy has achieved subpg/mL detection limits and provided accurate and robust quantification for drugs in plasma or cellular samples.15-17 In this study, based on a similar rationale, we attempted to develop an approach for the ultrasensitive quantification of key VitD metabolites in serum samples and then thoroughly evaluated its robustness and applicability in a large-scale clinical analysis. The quantification of VitD metabolites in multiple sclerosis patients was used as a model system for method development. As the most prevalent neurological disorder affecting young adults in the U.S.,18 multiple sclerosis is a neurodegenerative, chronic inflammatory disease that causes demyelination, sclerotic plaque formation, and central nervous system atrophy and eventually leads to chronic paralysis and cognitive deficits.19 In recent years, the association of low circulating levels of VitD metabolites with multiple sclerosis susceptibility and pathogenesis has attracted increasing interest. Much evidence suggests that low levels of the VitD metabolites are potential risk factors for developing multiple sclerosis, probably owing to their immune modulating effects.1 For instance, some epidemiological studies showed that low serum concentrations of 25(OH)VitD in adolescence are associated with an increased risk of developing multiple sclerosis20 and VitD supplementation and high serum concentrations of 25(OH)VitD are protective.21 The association of VitD metabolites and multiple sclerosis has been also supported by the correlations between multiple sclerosis disease activity and concentrations of 25(OH)VitD.22 Furthermore, treatment with 1,25(OH)2VitD was found to inhibit the development of autoimmune diseases.23 Nevertheless, the precise mechanisms for the association of VitD metabolites and the development of multiple sclerosis remain unclear.24 In order to study this association, an ultrasensitive and robust analytical method that is capable of quantifying accurately the key circulating VitD metabolites in clinical samples is essential. Such a method will also help to direct the therapeutic efforts. (15) Qu, J.; Qu, Y.; Straubinger, R. M. Anal. Chem. 2007, 79, 3786–3793. (16) Yu, H.; Straubinger, R. M.; Cao, J.; Wang, H.; Qu, J. J. Chromatogr., A 2008, 1210, 160–167. (17) Gaspar, J. R.; Qu, J.; Straubinger, N. L.; Straubinger, R. M. Analyst 2008, 133, 1742–1748. (18) Khatri, B. O. Ther. Apheris 2000, 4, 263–270. (19) Weinstock-Guttman, B.; Bakshi, R. CNS Drugs 2004, 18, 777–792. (20) Ozgocmen, S.; Bulut, S.; Ilhan, N.; Gulkesen, A.; Ardicoglu, O.; Ozkan, Y. J. Bone Miner. Metab. 2005, 23, 309–313. (21) Munger, K. L.; Levin, L. I.; Hollis, B. W.; Howard, N. S.; Ascherio, A. JAMA 2006, 296, 2832–2838. (22) Embry, A. F.; Snowdon, L. R.; Vieth, R. Ann. Neurol. 2000, 48, 271–272. (23) Cantorna, M. T.; Hayes, C. E.; DeLuca, H. F. Proc. Natl. Acad. Sci. U. S. A. 1996, 93, 7861–7864. (24) Mahon, B. D.; Gordon, S. A.; Cruz, J.; Cosman, F.; Cantorna, M. T. J. Neuroimmunol. 2003, 134, 128–132.

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2489

In this study, four VitD metabolites that are of high clinical significances were selected. These include 25(OH)VitD2 and 25(OH)VitD3, which are indicators of VitD repletion, and 1,25(OH)2VitD3 and 24,25(OH)2VitD3, which are biologically active metabolites.3 EXPERIMENTAL SECTION Chemicals and Reagents. The 25(OH)VitD3 and 1,25(OH)2VitD3 were purchased from EMD Biosciences (La Jolla, CA). 25(OH)2VitD2 was from Sigma-Aldrich (St. Louis, MO), and 24,25(OH)2VitD3 was from MP Biomedicals (Solon, OH). Isotopecoded I.S. were obtained from Medical Isotopes (Pelham, NH). The chemical and isotopic purities of the above standards are greater than 99%, according to the manufacturers. The 4-phenyl1,2,4-triazoline-3,5-dione (PTAD, purity g 99%) was purchased from Tokyo Chemical Industry Co. (Tokyo, Japan). HPLC-grade water, acetonitrile (ACN), and methanol were from Fisher Scientific (Fairlawn, NJ). LC/MS grade formic acid, acetic acid, and ammonium formate were from Sigma-Aldrich (St. Louis, MO). Oasis HLB 1-mL (30 mg sorbent) solid-phase extraction cartridges were purchased through Waters Corporation (Milford, MA). Standards for interference test were acquired from different commercial sources (listed in SI Table 1) with purities g97%. Study Population. A pathological group of subjects (n ) 281, age 45.5 ± 10.6 yr, 222 females, 59 males) diagnosed with multiple sclerosis according to the McDonald criteria25 and a control group with a similar age and gender distribution (n ) 22, age 40.9 ± 11.5 yr, 15 females, 7 males) were enrolled in this study. All participants were from Buffalo, NY. None of the patients used steroids or had a relapse 1 month prior to study. Disability ascertainment was based on the Kurtzke Extended Disability Status Scale (EDSS).26 This work was approved by the Human Subjects Institutional Review Board of the University at Buffalo. Informed consent was obtained from all study participants. The demographic and clinical characteristics are described in SI Table 2. Sample Preparation. Blood was obtained by venipuncture, allowed to stand for 15 min, and then centrifuged at 3000 g for 15 min. Serum aliquots were transferred into polypropylene tubes and stored at -80 °C until analysis. Once thawed, the serum samples were vortexed, and 200 µL portions were taken. The samples were spiked with 20 µL of working I.S. solution, vortex-mixed, and allowed to equilibrate for 25 min at 4 °C. Proteins were precipitated by adding 1 mL of methanol/ACN (80/20), followed by 10 min centrifugation at 10,000 g. The supernatant was evaporated to complete dryness under nitrogen stream, and then 200 µL of PTAD solution (1 mg/ mL in ACN) was added. The reaction was carried out at room temperature for 2 h and terminated with 0.8 mL of water. Selective SPE was performed on Oasis cartridges via a Supelco vacuum manifold (Park Bellefonte, PA) under the optimized conditions. The cartridges were conditioned by sequential washing with 1 mL of ACN for twice, 1 mL of water, and 25% ACN in 0.1% formic acid. After sample loading, the cartridge was washed subsequently (25) McDonald, W. I.; Compston, A.; Edan, G.; Goodkin, D.; Hartung, H. P.; Lublin, F. D.; McFarland, H. F.; Paty, D. W.; Polman, C. H.; Reingold, S. C.; Sandberg-Wollheim, M.; Sibley, W.; Thompson, A.; van den Noort, S.; Weinshenker, B. Y.; Wolinsky, J. S. Ann. Neurol. 2001, 50, 121–127. (26) Kurtzke, J. F. Neurology 1983, 33, 1444–1452.

2490

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

with 1 mL of 30% ACN and 1 mL of 75% ACN (both containing 0.1% formic acid). The analytes were eluted with 1 mL of elution solvent (95% ACN/IPA 4:1 with 0.1% formic acid). The eluate was evaporated, reconstituted in 100 µL of mobile phase A, and injected into the µLC/MS/MS system. SPE Optimization. For the purpose of achieving a selective SPE extraction that eliminates matrix components to the greatest extent possible without loss of analytes, an extensive optimization of the wash/elute conditions was conducted. Briefly, the four analytes were spiked at 1 ng/mL each to a pooled serum and subjected to protein precipitation and chemical derivatization procedure as described above. Cartridges were conditioned as described above. Aliquots of 200 µL of the samples were then loaded onto 16 cartridges at a rate of approximately 1 mL/min, followed by washing with 1 mL 30% ACN. The cartridges were divided into two groups: for Group 1, individual cartridges were washed with 0.1% formic acid in one of the following concentrations of ACN: 50%, 55%, 60%, 65%, 70%, 75%, 80%, or 85%. The absorbed analytes were eluted from all nine cartridges with 1 mL of ACN/IPA (80/20) containing 0.1% formic acid. For Group 2, all eight cartridges were washed with 1 mL of 30% ACN containing 0.1% formic acid, and the absorbed analytes were then eluted with 1 mL of 0.1% formic acid in one of the following concentrations of ACN/IPA (80/20): 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 100%. The eluates were evaporated to dryness, reconstituted, and analyzed by µLC/MS/MS. This optimization procedure was performed in duplicate, and the recoveries of the derivatized Vit D metabolites were determined by µLC/MS/ MS using an I.V. of 0.6 µL. µLC/MS/MS. µLC-MS/MS was performed on an Eksigent direct-flow capillary/nano LC system (Dublin, CA) coupled to a Thermo Scientific Quantum Ultra EMR triple-quadrupole mass spectrometer via an ESI interface (San Jose, CA). The LC pumps are powered by pressurized nitrogen (100 psi) and employs Microfluidic Flow Control technology to generate precise LC gradients at low flow-rate without flow splitting. A sequential 3-cycles transport liquid wash with mobile phases A and then 95% methanol in 0.1% formic acid was employed to minimize carryover by the autosampler. The separation was performed on a Zorbax C18 Stablebond capillary column (150 mm × 0.5 mm innerdiameter (ID)) with a particle size of 3.5 µm and a pore size of 100 Å (Agilent, Santa Clara, CA). The injection volume was 9.5 µL, and the needle ejection rate was 40 µL/min. The µLC flow rate was set at 10 µL/ min. The mobile phase consisted of (A) 2 mM ammonium formate in 60% methanol (pH 3.0) and (B) 2 mM ammonium formate in 97% methanol (pH 3.0). The percentage of mobile phase B was held at 0% for the first 3 min for sample focusing and then linearly increased to 85% over 10 min and kept for 2 min. The column was then washed with 98% B at 20 µL/min for 4 min and re-equilibrated with 0% B for 5 min. The entire analysis cycle was estimated to be 27 min for each sample from injection to column washing. To obtain optimal microspray performance, a 34-gauge microbore stainless steel spray needle was used. The ionization voltage, skimmer offset, and capillary transduction tube temperature were set at 3.2 kV, 8 V, and 250 °C, respectively. The pressure of the sheath gas (N2) was set at 15 (arbitrary units), and the argon collision gas pressure was set at 1.5 mTorr (0.2 Pa). SRM conditions for analytes and the I.S.,

including m/z of SRM pairs, collision energy, spray voltage, and tube lens voltage, are shown in SI Table 3. The dwell time of each SRM transition was 50 ms. The Q1 and Q3 resolutions were set at 0.7 mass units fwhm (full width at half-maximum). Calibration, Validation, and Evaluation of Assay Performance. Calibration standards were prepared at levels of 0.1/0.1/ 0.005/0.05 (25(OH)VitD2/25(OH)VitD3/1,25(OH)2VitD3/24,25 (OH)2VitD3, same below), 0.2/0.2/0.01/0.1, 0.5/0.5/0.025/0.25, 1.0/1.0/0.05/0.5, 2.0/2.0/0.10/1.0, 5.0/5.0/0.25/2.5, 10/10/ 0.50/5.0, 20/20/1.0/10, and 50/50/2.5/25 ng/mL with phosphate-buffered saline containing 30 mg/mL human serum albumin (HSA, Sigma-Aldrich). An I.S. working solution was prepared in ACN containing 26,26,26,27,27,27-d6-25(OH)VitD2, 26,26,26,27,27,27-d6-25(OH)VitD3, and 26,26,26,27,27,27-d61R,25(OH)2VitD3 at concentrations of 20, 20, and 2 ng/mL, respectively. Calibration curves were constructed by plotting the peak area ratios of each VitD metabolite and its I.S. versus the corresponding concentrations and fitting the data using linear regression with a 1/x2 weighting factor. Method accuracy and precision were evaluated using quality control (QC) samples prepared by spiking standard compounds at three levels into both blank matrix (referred as the “blank QC”) and a pooled human serum sample (referred as the “fortified QC”) for which the endogenous levels of the analytes were measured on a daily basis. For assessment of method specificity, 25 compounds (SI Table 3) having structures and molecular weights close to those of the target analytes, including VitD3, 7-dehydrocholesterol, and some endogenous or synthetic steroids, were tested at a concentration of 100 ng/mL for potential interference. These compounds were individually subjected to derivatization and µLC/MS/MS analysis described above, and the chromatograms were scrutinized for interference. Ion suppression was evaluated by comparing the peak areas of I.S. obtained from neat standard solutions with that obtained from patient samples (n ) 6). For stability investigation, 20 mL of pooled serum from healthy subjects was supplemented with 1,25(OH)2VitD3 at 1 ng/mL and 10 ng/mL, respectively. Aliquots of samples were subjected to various evaluations including multiple freeze-thaw cycles (from -80 to 22 °C), storage at ambient temperature, storage at 4 °C, and brief exposure to direct sunlight. At each designated intervals, isotope-labeled I.S. was added, and then the sera were stored at -80 °C until analysis. During the quantification of the large numbers of clinical samples, the method robustness was evaluated for the reproducibility of calculated concentrations, retention times, S/N, and peak shapes, by analyzing the four randomly selected patient samples (i.e., system control samples) immediately after every 48 injections, for 10 times in total (n ) 10) . RESULTS AND DISCUSSION 1. Serum Sample Treatment and Derivatization. A protein precipitation with ACN/methanol prior to SPE extraction was found necessary to achieve high efficiencies of extraction for all metabolites, probably because the precipitation is effective in releasing the analytes from serum proteins such as the VitD

binding protein (Ka ) 5 × 108 M-1 and 4 × 107 M-1 for 25(OH)VitD3 and 1,25(OH)2VitD3, respectively27). Derivatization by PTAD was chosen to improve the detectability of the VitD metabolites under ESI/MS/MS. Based on approaches described previously,14,28 the derivatization procedure was thoroughly optimized to achieve a straightforward procedure with high efficiency and reproducibility and to yield an eluate compatible with the following selective SPE procedure. The optimized conditions are shown in the Experimental Section. It is important to note that prior to the derivatization, the complete removal of water from the serum samples was found critical to obtain a high level of reaction completeness and reproducibility. The ionization efficiency and stability of the PTAD-VitDmetabolites were improved appreciably over the unmodified compounds (data not shown). Some extents of isomerization were found for 3 of the tagged metabolites (except for the 1,25(OH)2VitD3), which agrees with previous observations.14,28 Further evaluation showed the isomerization did not compromise quantitative accuracy and selectivity, because baseline separation was achieved for these isomeric peaks by adequate chromatographic resolution (discussed below), and that the ratios of the peak areas between these peaks were found highly constant. Oxidation of the metabolites during reaction, which was one of the primary concerns in this study, was not observed. 2. DEVELOPMENT OF THE SELECTIVE-SPE/µLC/ MS/MS Strategy In our pilot studies, conventional LC/MS/MS analysis after PTAD derivatization had achieved a LLOQ of approximate 75-100 pg/mL for the four metabolites using 0.2 mL of serum (data not shown). However, preliminary assessment suggested such LLOQ was not sufficiently low for the dihydroxyl metabolites. For an example, it was found a LLOQ far below was necessary to enable a robust quantification of 1,25(OH)2VitD3 in the patient samples. To further increase the sensitivity to the level desired, we employed a low-flow-rate µLC/MS/MS. A µLC provides higher peak concentrations than a conventional LC that runs at higher flow rates and thus could result in drastically increased “absolute sensitivity” (i.e., sensitivity expressed in terms of mass of the analyte on column) when connected to a concentrationdependent detector, such as an ESI/MS/MS.15 Nonetheless, in order to achieve a high “relative sensitivity” (i.e., the sensitivity expressed in terms of concentration of the analyte in sample), which is critical for this study, it is essential to be able to load a relatively high volume of sample on the µLC column without compromising the analytical performance.15,16 Due to the low loading capacity of µLC columns, however, injecting a high volume of serum samples is not feasible if nonselective sample preparation procedures such as protein precipitation or generic SPE were used. This is because nonselective procedures may retrieve numerous matrix compounds from a highly complex biological sample, which can readily exceed the capacity of a µLC column unless a small injection volume (I.V.) is used.15,16 To address this limitation, we attempted to develop a selective SPE approach in order to simplify the sample matrix (27) White, P.; Cooke, N. Trends Endocrinol. Metab. 2000, 11, 320–327. (28) Eyles, D.; Anderson, C.; Ko, P.; Jones, A.; Thomas, A.; Burne, T.; Mortensen, P. B.; Norgaard-Pedersen, B.; Hougaard, D. M.; McGrath, J. Clin. Chim. Acta 2009, 403, 145–151.

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2491

Figure 1. Optimization of strong-wash and elution conditions to achieve a selective SPE extraction of the four VitD metabolites. (A) Optimization of the strong-wash solvent: SPE cartridges were washed with a range of 50-85% ACN and eluted with 100% ACN-IPA (4:1). (B) Optimization of the elution solvent: SPE cartridges were washed with 30% ACN and eluted with a range of 60-100% ACN-IPA. Based on the data, the optimal conditions were determined as 75% ACN for strong-wash and 95% ACN-IPA for elution.

while concentrating the analytes, thereby enabling a large I.V. on the µLC column. With the emphasis on 1,25(OH)2VitD3, the strategy has been developed by optimizing each component of the analytical flow path: (i) the selective SPE extraction procedure, (ii) the ionization and SRM method for MS/MS, and (iii) the µLC loading and separation method. We also evaluated the robustness of this method and its applicability in large-scale clinical analysis. 2.1. Optimization of Selective SPE. In our preliminary experiments, the µLC column was found liable to overcapacity when serum samples were analyzed following a nonselective SPE approach. Deterioration of chromatographic performance and high chemical noise were observed with injection volumes greater than 0.6 µL (SI Figure 1). In order to enable a high I.V. without causing overcapacity of the column, a selective SPE method has been optimized thoroughly to achieve three goals: i) the overall highest and most consistent absolute recovery of the VitD metabolites, ii) a washing condition that would remove polar matrix components to the greatest extent possible without eluting the VitD metabolites, and iii) an elution condition that would subsequently recover the VitD metabolites efficiently while minimizing the elution of less polar matrix components. Oasis HLB cartridges were selected after evaluation of the analyte recovery and reproducibility of cartridges produced by several manufacturers. Aliquots of a pooled serum sample, which were spiked with the same levels of the four analytes, precipitated, and derivatized as described in the Experimental Section, were used for the optimization. After the SPE cartridges were loaded with derivatized serum samples, they were washed in two steps. First, 30% ACN was used to remove residual serum proteins (“weak-wash” step) and thus avoid column blockage that could otherwise occur during subsequent strong-wash using high concentrations of ACN. Second, a higher concentration of ACN, identified through detailed optimization, was used to remove relatively polar matrix components without loss of the analyte 2492

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

(“strong-wash” step). Finally, the target analytes were eluted using an optimized solvent composition, which efficiently recover the analytes while leaving more hydrophobic matrix components on the cartridge. In order to identify the optimal conditions for strong-wash and elution, absolute recovery of each analyte through the SPE procedure was investigated using i) strong-wash solvent compositions ranging from 50-85% ACN and 100% ACN-IPA (4:1, v/v) for elution and ii) weak-wash solvent of 30% ACN and elution solvent compositions of 65-100% ACN-IPA. The results are shown in Figure 1. It appeared all analytes remained on the cartridge when the percentage of ACN up to 75% was used for the strongwash step (Figure 1A). In our pilot experiments, it was found that complete elution of the four metabolites was difficult even with 100% ACN. Therefore, we added isopropyl alcohol (IPA) into the elution solvent (ACN:IPA)4:1) to increase its strength and obtained a satisfactory elution of all four metabolites when g95% ACN-IPA was used (Figure 1B). Based on the results, 75% ACN was selected as optimal for the strong-wash step, and 95% ACNIPA was selected as optimal for the elution solvent. The selective SPE procedure substantially simplified the sample matrix and thus avoided the deterioration of chromatographic performance by a large I.V. on the µLC column; moreover, chemical noises were drastically reduced when compared to the use of a generic SPE procedure (SI Figure 1). 2.2. Optimization of MS/MS Conditions. For all PTADtagged VitD metabolites, the full scan spectra under positive ESI were dominated by dehydrated precursors, which likely resulted from gas-phase dehydration process involving the hydroxyl group at position 25.14 The dehydrated precursors were selected for SRM of the derivatized metabolites for two reasons. First, the dehydrated precursors were the most abundant form of the ionized analytes; second, these precursors were found pretty stable under ESI, perhaps because they carry a stable tertiary carbocation moiety.

Figure 2. Typical chromatograms showing injection volume (I.V.) optimization for pooled serum spiked with 100 pg/mL of 1,25(OH)2VitD3 and extracted by selective SPE. The I.V. was (A) 0.2 µL; (B) 0.6 µL; (C) 3 µL; and (D) 9.5 µL. A 15 cm × 0.5 mm ID capillary column (manufacturer suggestion I.V. was 0.2-0.4 µL) was used for separation. Using the selective SPE and a sample focusing approach before the separation, a large I.V. was enabled without compromising chromatographic performance and with substantially increased sensitivity. Int ) intensity.

The optimal conditions for the productions of the dehydrated precursors were identified by investigating the effects of factors such as MS parameters and mobile phases. An in-source fragmentation voltage at 8 V was employed to obtain the maximal intensity of the precursors. Additionally, it was found that sodiumadducted ions were prominent for both of the dihydroxyl VitD metabolites, which markedly compromised assay sensitivity. By adding pH-buffered ammonium formate (2 mM, pH ) 3.0) as a mobile phase additive, the sodium-adducted ions were efficiently suppressed. For each PTAD-tagged metabolite, the CID fragmentation yielded a predominant product ion, which was postulated bearing moieties from both PTAD and the A-ring of the metabolites (SI Figure 2). Because this product carries high sensitivity and selectivity, it was chosen for SRM of the analytes. Optimized SRM conditions are shown in SI Table 3. 2.3. Optimization of µLC Strategy. Because of the low flow rates employed on the µLC system, excessive void volumes in flow path may result in undesirable consequences such as severe band broadening and lag for mobile phase change. To minimize the void volume both pre- and postcolumn, we employed a set of splitless, low-volume microflow pumps, small-ID (50 µm) tubing, zero-dead-volume connectors, and valves. As a result, neither significant delay of gradient changes nor deterioration of peak resolution caused by void volume was observed.

A 0.5-mm-ID capillary column was chosen, after balancing considerations of sensitivity gain, loading capacity, and operational robustness. The manufacturer-recommended I.V. for the µLC column is 0.2-0.4 µL. Nonetheless, we hypothesized that the serum matrix has been reduced substantially by the selective SPE, and therefore a markedly higher I.V. of the extracted sample can be tolerated without compromising chromatographic performance, provided an efficient sample focusing strategy was employed. To test this hypothesis, we evaluated the µLC/MS/MS performance as a function of I.V. using serum samples fortified with the four analytes. In order to focus a larger I.V. of the selectively extracted samples on the µLC column and thus to prevent peak broadening, 100% mobile phase A (60% methanol in 2 mM ammonium formate, pH ) 3) were held for 3 min during the sample loading. Typical chromatograms are shown in Figure 2. The retention times of the analyte increased only slightly as the I.V. increased, but no appreciable change in peak width and peak shape was observed over the range of I.V. from 0.2 to 9.5 µL, while both the intensity and S/N increased roughly in proportion with I.V. This indicated the selective SPE procedure, in conjunction with the sample focusing approach, permits an I.V. that is more than 20-fold higher than manufacturer’s recommendation, without causing column overcapacity. An I.V. of 9.5 µL was selected for quantification of serum samples. Such a large I.V. did not cause column fouling or Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2493

Figure 3. Representative µLC/MS/MS chromatograms of VitD metabolites from a patient serum sample, after the selective SPE procedure. Injection volume was 9.5 µL. For the first 3 min, 100% mobile phase A (60% methanol in 2 mM ammonium formate) was held to focus the injected sample on the µLC column, and then the separation gradient started. Int ) intensity.

compromise robustness of the method, as demonstrated in the following evaluations in the large-scale clinical analysis. Sufficient chromatographic separation is critical in this work, for three reasons: first, the peak of target compounds must be separated from isomeric peaks, interfering signals and endogenous compounds that cause ion suppression. Second, given the close similarity in the structure and m/z among the analytes and I.S., it is necessary to resolve these compounds chromatographically to avoid various “cross-talk” effects. Third, a sufficient µLC separation was found further helped to suppress the sodium-adducted precursors, a phenomenon also observed previously.15,16 Other considerations for developing the µLC method included the following: i) upon complete elution of all analytes, it is necessary to perform an extensive cleaning of the column and spray tip, to minimize the memory effect29 and ii) a relatively high throughput is desirable for clinical analysis. The gradient conditions were optimized carefully to achieve the above objectives, and the optimal µLC conditions are described in the Experimental Section. Under the optimized conditions, all chromatographic peaks of the target compounds were well-defined and symmetric, without appreciable interference. Typical chromatograms of a patient serum sample are shown in Figure 3. 3. METHOD VALIDATION AND EVALUATION 3.1. Linearity, Sensitivity, Accuracy, and Precision. Excellent linearity was obtained for the 9-point calibration curves over the concentration range of 0.005-2.50 ng/mL for 1,25(OH)2VitD3, 0.050-25.0 ng/mL for 24,25(OH)2VitD3, and 0.100-50.0 ng/mL for both 25(OH)VitD2 and 25(OH)VitD3 (r2 >0.99, Table 1). The limit of detections (LOD), defined as the lowest concentration that gave a signal-to-noise ratio (S/N) of 3, were in the (29) Duan, X.; Young, R.; Straubinger, R. M.; Page, B.; Cao, J.; Wang, H.; Yu, H.; Canty, J. M.; Qu, J. J. Proteome Res. 2009, 8, 2838–2850.

2494

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

Table 1. Sensitivity, Quantification Range, and Linearity for the Four Metabolites VitD metabolites

sensitivity quantification linearity (LOD, pg/mL)a range (ng/mL) (n = 6) (r2)

25-hydroxy vitamin D2 25-hydroxy vitamin D3 1R,25-dihydroxy vitamin D3 24(R),25-dihydroxy vitamin D3 a

0.5

0.100-50.0

g0.9924

1.0

0.100-50.0

g0.9967

0.5

0.005-2.50

g0.9913

1.0

0.050-25.0

g0.9905

Defined as S/N ) 3.

range of 0.5-1 pg/mL for the four metabolites (Table 1). The LLOQ of the four metabolites were validated based on their expected concentrations in the sera of healthy subjects and multiple sclerosis patients. Validated LLOQ were respectively 5 pg/mL and 50 pg/mL for 1,25(OH)2VitD3 and 24,25(OH)2VitD3 and 100 pg/mL for both 25(OH)VitD2 and 25(OH)VitD3. Method accuracy and precision were evaluated using quality control (QC) samples prepared by spiking standard compounds at three levels into both blank matrix (referred as the “blank QC”) and a pooled human serum sample (referred as the “fortified QC”) for which the endogenous levels of the analytes were previously measured. Good accuracies with bias 500 injections in total). Each of the four “system control” (SC) samples, which were randomly selected from sera of multiple sclerosis patients, were analyzed once for every 48 injections, for totally ten times (n ) 10); the reproducibility of calculated concentrations, retention times, S/N, and peak shapes of all metabolites was assessed. Some representative data are shown in SI Table 4. The calculated concentrations (RSD% 2.4-12%) and retention times (RSD% 0.5-2%) of all analytes were highly reproducible, and no decline in sensitivity (RSD% 2-12%) and symmetry of peaks (as indicated by the tailing factors, RSD% 6-13%) was observed. Additionally, no noticeable deterioration in chromatographic performance, increase in back pressure, or fouling of the µLC was observed after more than 500 injections of serum samples. The high degree of robustness is attributable to the use of a selective SPE and a sufficient and reproducible µLC separation and that an effective cleaning procedure was employed for each (30) Lewis, J. G.; Elder, P. A. Clin. Chem. 2008, 54, 1931–1932.

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2495

Figure 4. The distribution of the serum concentrations of 1,25(OH)2VitD3 by both McDonald standard and EDSS. Totally 303 subjects were investigated, including 281 multiple sclerosis patients and 22 healthy subjects. The LOD/LLOQ for the most sensitive LC/MS/MS method reported previously was 25 pg/mL, by which the metabolite would not be quantifiable in >40% of the multiple sclerosis samples and >27% of healthy subjects. By comparison, with the selective-SPE/µLC/MS/MS method, the metabolites were detected in all samples, and the serum levels in 96% of patient samples and 100% of healthy subjects were above the LLOQ (5 pg/mL). Abbreviations: CIS ) clinically isolated syndrome; RRMS ) relapsing remitting multiple sclerosis; SPMS ) secondary progressive multiple sclerosis; PPMS ) primary progressive multiple sclerosis; EDSS ) expanded disability status scale. Table 3. Levels of VitD Metabolites in Sera from Multiple Sclerosis Patients and Healthy Subjects multiple sclerosis patients (n ) 281) VitD metabolites 25-hydroxy vitamin D2 25-hydroxy vitamin D3 25-hydroxy vitamin Da 1R,25-dihydroxy vitamin D3 24 (R),25-dihydroxy vitamin D3 c

EIA/RIA results for multiple sclerosis patients [Mean (SD), pmol/mL]

concentration range (pmol/mL)

mean (SD) (pmol/mL)

control (n ) 22) mean (SD) (pmol/mL)

p

by Barnes et al.32 (n ) 29)

by Smolders et al.33 (n ) 267)

ND-150.1b 4.32-148.5 4.82-165.6 ND-0.389d 0.420-55.3

10.8 (21) 51.9 (26) 62.6 (27) 0.067 (0.05) 8.64 (7)

3.40 (5) 54.9 (29) 58.3 (29) 0.100 (0.06) 14.0 (16)

0.12 0.53 0.47 0.004 0.004

NAc NAc 69.1 (40) 0.092 (0.04) NAc

NAc NAc 62.5 (32) 0.108 (0.04) NAc

a Sum of 25-hydroxy vitamin D2 and D3. b As VitD2 is primarily from dietary sources, its metabolite may not be detected for some subjects. Data not available. d For less than 4% of the samples, the serum levels of this metabolite were lower than the LLOQ (5 pg/mL).

run. In addition, the LLOQ adopted for each analyte is at least 10 times higher than the LOD; the sufficiently high S/N at LLOQ contributes to the high robustness, accuracy, and reproducibility of the method, in that it renders the method well-tolerable to fluctuations in analytical sensitivity, which could often be caused by such occasions as the higher chemical noises in certain samples and decrease in instrumental sensitivity during operation. 4. LARGE-SCALE QUANTIFICATION IN MULTIPLE SCLEROSIS PATIENTS Using the developed method, serum samples from 281 multiple sclerosis patients and 22 healthy subjects were analyzed. The ultrahigh sensitivity of the method enabled a comprehensive investigation of 1,25(OH)2VitD3 levels in multiple sclerosis patients, which are expected to be lower than these in healthy subjects. The distribution of the serum levels of 1, 25(OH)2VitD3 by disease course is illustrated in Figure 4. Previous LC/MS/MS-based methodologies are clearly not adequately sensitive for such a clinical investigation, as indicated by the fact that for >40% multiple sclerosis patients and >27% control samples, the serum levels of 1,25(OH)2VitD3 are lower than 25 pg/mL, which is the lowest reported LOD/LLOQ by LC/MS/MS (using 0.5 mL of serum).14 By comparison, the levels of 1,25(OH)2VitD3 in all samples were well above the LOD achieved in this study (0.5 pg/mL), using only 0.2 mL of 2496

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

serum; moreover, with an LLOQ of 5 pg/mL, the 1,25(OH)2VitD3 has been quantified with ease in 100% of the healthy subjects and >96% of the patient samples (Figure 4). The levels of the four VitD metabolites are summarized in Table 3. For the healthy subjects, the mean serum concentration of the total 25(OH)VitD (i.e., the sum of 25(OH)VitD2 and 25(OH)VitD3, which is the a well-accepted indicator for VitD deficiency3) was 58.3 ± 28.5 pmol/mL. The average concentrations of 1,25(OH)2VitD3 and 24,25(OH)2VitD3 were respectively 0.100 ± 0.057 and 14.0 ± 16.2 pmol/mL. These values are consistent with the data acquired previously for healthy subjects.31 As for the multiple sclerosis patients, the mean serum concentration of total 25(OH)VitD was 62.6 ± 26.8 pmol/ mL, which agreed well with the results obtained recently by immunoassay32,33 the mean concentration of 1,25(OH)2VitD3 (0.067 ± 0.045 pmol/mL) determined here was moderately lower than that acquired formerly by immunoassay in smaller(31) Hollis, B. W. Detection of vitamin D and its major metabolites. In Vitamin D, 2nd ed,; Feldman, D., Pike, J. W., Glorieux, F. H., Eds.; Elsevier/ Academic Press: New York, 2005. (32) Barnes, M. S.; Bonham, M. P.; Robson, P. J.; Strain, J. J.; Lowe-Strong, A. S.; Eaton-Evans, J.; Ginty, F.; Wallace, J. M. Mult. Scler. 2007, 13, 670– 672. (33) Smolders, J.; Menheere, P.; Kessels, A.; Damoiseaux, J.; Hupperts, R. Mult. Scler. 2008, 14, 1220–1224.

scale investigations32,33(Table 3). Mean concentration of 24,25(OH)2VitD3 was 8.64 ± 6.8 pmol/mL, for which the previous data in multiple sclerosis patients were not available. Independent t-test revealed no significant difference between the healthy subjects and multiple sclerosis patients for the serums levels of 25(OH)VitD3 or total 25(OH)VitD. However, the levels of both 1,25(OH)2VitD3 and 24,25(OH)2VitD3 were significantly lower in the patient group compared to the healthy group (p < 0.05), which is a novel observation. As a potent immunoregulator, the 1,25(OH)2VitD3 may suppress the incidence of autoimmune diseases such as the multiple sclerosis via regulations of cytokines, as suggested in previous studies.23 Therefore, the deficiency of 1,25(OH)2VitD3 could contribute to an increased susceptibility of multiple sclerosis. This hypothesis is supported by the discovery of this study that significantly lower levels of 1,25(OH)2VitD3 were found in the multiple sclerosis patients than in healthy subjects. The levels of 24,25(OH)2VitD3 appear to correlate well with those of 1,25(OH)2VitD3 in vivo, perhaps via the regulation of 25hydroxyvitamin-D-24-hydroxylase.34 This probably explained our observation that the serum levels of 24,25(OH)2VitD3 were also significantly lower in multiple sclerosis patients. CONCLUSIONS The ability to quantify circulating VitD metabolites in clinical samples is highly critical to study the mechanisms and to direct the therapeutic efforts for many diseases. However, daunting challenges lie in the analysis of low-abundance dihydroxyl metabolites such as 1,25(OH)2VitD3, which are highly active albeit presenting at serum levels too low to be measured with sufficient sensitivity by most analytical methods, including conventional LC/MS/MS techniques. As a result, currently the immunoassay-based methods remain the dominant means for quantifying dihydroxyl VitD metabolites, though the selectivity, accuracy, and interbatch/lab reproducibility of these techniques could be problematic. Based on a strategy that combines selective-SPE and µLC/MS/ MS, we developed a highly sensitive, accurate, and selective method that permits the robust quantification of four key VitD metabolites in human serum. The strategy has been optimized with the emphasis on 1,25(OH)2VitD3, the analyte of the lowest abundance. A derivatization procedure was optimized to achieve a one-step procedure with high efficiency and reproducibility. The derivatization markedly improved the ionization efficiency of the analytes, and stabilized the dihydroxyl metabolites. To achieve an ultrahigh sensitivity that enables a robust quantification of all analytes in serum samples, a selective-SPE/µLC/MS/MS strategy was developed. SPE washing and elution conditions were optimized to extract the analytes from serum selectively with a high recovery, while eliminating majority of the matrix components. Because of the selective SPE procedure, the acceptable I.V. of serum samples on the µLC column was increased significantly. As a result, sensitivity was increased without compromising chromatographic performance or operational robustness. During the sample loading onto the µLC column, a sample focusing step was employed to prevent the peak broadening that would otherwise result from a large I.V. A sufficient µLC separation has been developed to (34) Akeno, N.; Saikatsu, S.; Kawane, T.; Horiuchi, N. Endocrinology 1997, 138, 2233–2240.

separate the analyte peaks from interfering signals and to minimize ion suppression effects. The ionization and SRM conditions were optimized to attain maximal sensitivity and selectivity. The quantitative method was validated and evaluated extensively for specificity and robustness. The method developed here holds several advantages that are important for clinical analysis of VitD metabolites. First, ultrahigh sensitivities, with LOD ranging from 0.5-1 pg/mL, were achieved for the four analytes. Such a high level of sensitivity, which is comparable or superior to that of immunoassays, enabled a robust quantification of all analytes in human serum samples. Second, due to the use of selective SPE extraction and adequate µLC separation, the method is highly specific, as examined using structurally similar compounds. The excellent specificity helped to achieve a high quantitative accuracy. Third, the method is very robust and reproducible, as revealed by comprehensive evaluations during a large-scale clinical analysis. Additionally, the method is superior to immunoassay methods, in that it offers significantly better selectivity, accuracy, and the capacity of simultaneous quantification of multiple VitD metabolites. As a proof-of-concept, this method was applied to the analysis of serum samples from 281 multiple sclerosis patients and 22 healthy subjects. Because of the ultrahigh sensitivity achieved, the 1,25(OH)2VitD3 were detected in all samples, and for 100% of healthy subjects and 96% of the patients, the serum levels were above a LOQ of 5 pg/mL. Apparently, conventional LC/ MS/MS is not sufficiently sensitive for this task. The quantitative results revealed that the serum levels of both 1,25(OH)2VitD3 and 24,25(OH)2VitD3 are significantly lower in multiple sclerosis patients than in healthy subjects. However, with a cognizance that this study may be limited in the sample sizes, this finding may need to be examined in further clinical investigation. More mechanism-related studies are desired to elucidate how this altered VitD metabolism relates to multiple sclerosis development. The method developed here permitted, for the first time, the consistent and robust quantification of low-abundance dihydroxyl VitD metabolites such as the 1,25(OH)2VitD3 in clinical samples by LC/MS/MS-based methods. This work also demonstrated the applicability of the selective-SPE/µLC/MS/MS strategy in largescale clinical analysis. Therefore, the method is well suited for the analysis of VitD metabolites in large-scale clinical studies, especially for these involving pathological conditions where the levels of the dihydroxyl VitD metabolites could be low. ACKNOWLEDGMENT This work is financially supported by the National Multiple Sclerosis Society (RG3743), a grant by the Center of Protein Therapeutics, and a NIH grant (1R21DA027528). X.D. and B.W.G. contributed equally to this work. SUPPORTING INFORMATION AVAILABLE SI Tables 1-4 and SI Figures 1-4. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review December 16, 2009. Accepted January 30, 2010. AC902869Y Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2497