Anal. Chem. 2005, 77, 2985-2991
Technical Notes
Determination of 16 Phthalate Metabolites in Urine Using Automated Sample Preparation and On-line Preconcentration/High-Performance Liquid Chromatography/Tandem Mass Spectrometry Kayoko Kato, Manori J. Silva, Larry L. Needham, and Antonia M. Calafat*
Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention, 4770 Buford Hwy, Mailstop F17, Atlanta, Georgia 30341
We developed an on-line solid-phase extraction (SPE) method, coupled with isotope dilution high-performance liquid chromatography/tandem mass spectrometry (HPLC/ MS/MS) and with automated sample preparation, to simultaneously quantify 16 phthalate metabolites in human urine. The method requires a silica-based monolithic column for the initial preconcentration of the phthalate metabolites from the urine and a silica-based conventional analytical column for the chromatographic separation of the analytes of interest. It uses small amounts of urine (100 µL), is sensitive (limits of detection range from 0.11 to 0.90 ng/mL), accurate (spiked recoveries are ∼100%), and precise (the inter- and intraday coefficients of variation are 99.9%), their 13Clabeled internal standards (>99.9%), 13C4-4-methylumbelliferone, D4-dimethyl phthalate (D4-DMP), D4-diethyl phthalate (D4-DEP), D4-dibutylphthalate (D4-DBP), and D4-di(2-ethylhexyl) phthalate (D4-DEHP) were purchased from Cambridge Isotope Laboratories, Inc (Andover, MA). MCPP and 13C4-MCPP were obtained from Los Alamos National Laboratory (Los Alamos, NM) and from Cambridge Isotope Laboratories, Inc. and 13C2-PA, 4-methylumbelliferone and its glucuronide, ammonium acetate (>98%), and acetic acid (glacial) were purchased from Sigma Aldrich Laboratories, Inc. (St. Louis, MO). MiBP, D4-MiBP, MECPP, and D4MECPP were gifts from Prof. Ju¨rgen Angerer (University of ErlangensNuremberg, Germany). Acetonitrile (ACN), methanol (MeOH), and water (HPLC grade) were purchased from Tedia (Fairfield, OH), and ammonium hydroxide (30%) was purchased from J. T. Baker (Phillipsburg, NJ). β-Glucuronidase (Escherichia coli-K12) was purchased from Roche Biomedical (Mannheim, Germany). Preparation of Standards and Quality Control (QC) Materials. Reagent solutions were prepared in ACN and water using standard laboratory procedures. Stock solutions of phthalate monoester metabolites, 4-methylumbelliferone and isotopically labeled phthalates, and 4-methylumbelliferone internal standards were prepared in ACN and stored at -20 °C in Teflon-capped amber glass bottles until use.18 The working standards were prepared in 1:9 ACN:water from serial dilutions of the stock solutions to create 11 standard solutions, containing phthalate metabolites, 4-methylumbelliferone, and their isotopically labeled internal standards, whose concentrations encompassed the entire linear range of the method. The working standards were stored (33) Cabrera, K.; Lubda, D.; Eggenweiler, H. M.; Minakuchi, H.; Nakanishi, K. HRC-J. High Resolut. Chromatogr. 2000, 23, 93-99. (34) Tanaka, N.; Nagayama, H.; Kobayashi, H.; Ikegami, T.; Hosoya, K.; Ishizuka, N.; Minakuchi, H.; Nakanishi, K.; Cabrera, K.; Lubda, D. HRC-J. High Resolut. Chromatogr. 2000, 23, 111-116. (35) Tanaka, N.; Kobayashi, H.; Ishizuka, N.; Minakuchi, H.; Nakanishi, K.; Hosoya, K.; Ikegami, T. J. Chromatogr. A 2002, 965, 35-49. (36) Miller, S. Anal. Chem. 2004, 76, 99A-101A. (37) Tanaka, N.; Kobayashi, H.; Nakanishi, K.; Minakuchi, H.; Ishizuka, N. Anal. Chem. 2001, 73, 420A-429A. (38) Quirino, J. P.; Dulay, M. T.; Zare, R. N. Anal. Chem. 2001, 73, 55575563.
at 4 °C in Teflon-capped glass vials until use.18 Calibration standards were prepared in 1.5-mL autosampler vials by automatically mixing the working standard solution (20 µL) with 0.1% acetic acid:water (480 µL) on the autosampler of a ThermoFinnigan Surveyor liquid chromatograph (ThermoFinnigan, San Jose, CA) right before injecting the standard on the SPE/HPLC/MS/MS system. The enzyme solution was prepared daily by adding β-glucuronidase (30 µL, 200 units/mL) to 1.5 mL of ammonium acetate buffer (1 M, pH 6.5). An aqueous standard solution of 4-methylumbelliferone glucuronide (0.16 ng/mL) was added to all samples. The 4-methylumbelliferone/13C4-methylumbelliferone peak area ratio was monitored to check the extent of the deconjugation reaction with β-glucuronidase.18 QC materials were prepared from a base urine pool obtained from multiple anonymous donors. The pool was divided in two subpools that were enriched with native phthalate metabolites to create low-concentration (QCL, 5-90 ng/mL) and high-concentration (QCH, 17-390 ng/mL) QC materials. The two pools were uniformly mixed, dispensed in 2-mL portions in polypropylene cryovials (Nalge Nunc International, Rochester, NY), and stored at -40 °C. Each QC material was characterized by 60 repeated measurements in a 3-week period to define the mean concentrations and the 95% and 99% control limits of each phthalate metabolite. Sample Pretreatment. Human urine samples were thawed, sonicated for 5 min, and vortex mixed. Then, 100 µL of urine was aliquoted in a 1.5-mL autosampler vial (Sun-Sri). Next, the vial was placed on a Surveyor autosampler, operated using the Xcalibur software (ThermoFinnigan, San Jose, CA), and the rest of sample pretreatment steps were performed automatically (Table S-1). An Xcalibur program was designed to perform the following tasks: spike the urine samples with solutions of isotopically labeled phthalate metabolites (100 µL), 4-methylumbelliferone glucuronide (25 µL), and enzyme solution (25 µL) and mix the urine sample with the spiked solutions using the syringe autosampler. The sample tray temperature was set at 37 °C for the duration of these spiking and mixing steps. The spiked urine samples were incubated at 37 °C for at least 90 min. Then, the deglucuronidation reaction was stopped by adding 50 µL of acetic acid (glacial) and 200 µL of 5% ACN/water. After quenching the enzyme activity, the samples were mixed using the syringe autosampler. Next, the sample tray temperature was set at 0 °C, and the samples were kept at this temperature until analysis. For analysis, the deconjugated urine samples were removed from the autosampler used for sample preparation, gently mixed, and transferred to the autosampler of the on-line SPE/HPLC/MS/MS system. On-Line SPE/HPLC/MS/MS: The on-line SPE/HPLC/MS/ MS system was built from several Agilent 1100 modules (Agilent Technologies, Wilmington, DE) and one ThermoFinnigan Surveyor liquid chromatograph coupled with a ThermoFinnigan TSQ Quantum triple quadrupole mass spectrometer equipped with an electrospray ionization (ESI) interface and a six-port switching valve (Figure 1). The Agilent modules included a handheld control module, a binary pump with a degasser, and a column compartment with a six-port switching valve. The switching valve in the Agilent column compartment was programmed using the Agilent handheld control module. The Agilent pump, Surveyor system,
Figure 1. Connection diagram for the on-line SPE/HPLC/MS/MS system.
and TSQ quantum, including its switching valve, were controlled by the Xcalibur software. A CAN (closed area network) cable connected the Agilent column compartment and binary pump and allowed synchronization of the switching valve with the solvent program of the pump. An in-house-built communication cable connected the pump remote output to the Surveyor autosampler time function 1 (TF1) and ground (GND) outputs and allowed synchronization of the Agilent pump with the Surveyor system. With the Agilent switching valve in the sample loading position, 500 µL of the deconjugated urine sample, injected using the Surveyor autosampler, was directly loaded in 1 min on a Chromolith Flash RP-18e precolumn (4.6 mm × 25 mm, Merck KGaA, Germany) using 0.1% acetic acid in water:0.1% acetic acid in ACN (85:15) at 1.5 mL/min provided by the Agilent binary pump. After 1 min, the Agilent valve automatically switched to its alternate position, allowing the analytes to be transferred from the monolithic precolumn onto a BETASIL Phenyl analytical column (3 µm, 2.1 mm × 150 mm; ThermoElectron Corp., Bellefonte, PA), preceded by inline filters (2 µm and 0.5 µm, Upchurch Scientific, Oak Harbor, WA), in backflush mode, by the Surveyor MS quaternary pump using a nonlinear solvent gradient (Table 1) of 0.1% acetic acid in water (mobile phase A) and 0.1% acetic acid in ACN (mobile phase B). After 6 min, the Agilent valve switched back to the loading position, and the precolumn was rinsed with 4 mL of 0.1% acetic acid in ACN (0.5 mL/min for 8 min) and 2.5 mL of 85:15 0.1% acetic acid in water:0.1% acetic acid in ACN (0.25 mL/min for 10 min). The Surveyor autosampler was configured with syringe washes between injections to eliminate potential carryover. After injection, the TSQ Quantum switching valve directed the post-HPLC column flow to waste for 4.0 min. Then, the valve switched to direct the flow to the mass spectrometer. ESI in the negative ion mode was used to form negatively charged analyte ions at the interface under the following fixed instrument settings: spray ion voltage, -3800 V; sheath gas pressure, 35 mTorr; auxiliary gas pressure, 4 mTorr; capillary temperature, 280 °C; tube lens voltage, -140 V; collision gas pressure, 1.5 mTorr. Ionization parameters and collision cell parameters were optimized for each analyte (Table 2). The source collision-induced dissociation voltage was set to 10 V to break down acetate clusters. Unit resolution was used for both Q1 and Q3 quadrupoles. The mass Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
2987
Table 1. On-Line SPE and HPLC Solvent Gradient Programsa Agilent SPE pump time (min)
A (%)
B (%)
flow (µL/min)
Agilent valve position
0 0.4 0.5 1 1.1 3 5 6 7 19.75 23.1 25.6 26.8 26.9
100 100 85 85 100 100 100 100
0 0 15 15 0 0 0 0
1800 1800 1800 1800 100 100 100 100
SPE/waste SPE/waste SPE/waste SPE/HPLC SPE/HPLC SPE/HPLC SPE/HPLC SPE/waste
a
0 0 100 100 100
100 100 0 0 0
500 500 250 250 250
SPE/waste SPE/waste SPE/waste SPE/waste SPE/waste
Surveyor HPLC pump A (%)
B (%)
flow (µL/min)
77
23
300
75 75
25 25
300 300
70 60 50 0 0 77
30 40 50 100 100 23
325 350 350 350 350 350
Quantum valve position HPLC/waste HPLC/waste HPLC/waste HPLC/waste HPLC/MS/MS
HPLC/MS/MS HPLC/MS/MS HPLC/MS/MS HPLC/MS/MS HPLC/MS/MS HPLC/MS/MS
Mobile phase A is 0.1% acetic acid in water, and mobile phase B is 0.1% acetic acid in acetonitrile.
Table 2. Phthalate Metabolites and Their Native and Labeled Precursor and Product Ion Transitions, Collision Energies, and Retention Times analyte PA 13C
tR, min
precursor/product ion (confirmation ion)
collision energy,a V
4.64
165/77 (121) 167/77 251/103 (121) 255/103 179/77 (121) 183/79 193/77 (121) 197/79 179/135 (121) 185/141 221/77 (121) 225/81 221/77 (121) 225/79 293/121 (145) 297/124 307/159 (121) 311/159 291/121 (143) 295/124 255/183 (77) 259/186 247/77 (121) 251/79 277/134 (77) 281/137 277/125 (77) 281/127 291/121 (77) 295/125 305/261 (77) 309/264
25 (16)
2-PA
MCPP 4-MCPP MMP 13C -MMP 4 MEP 13C -MEP 4 MMiP 13C -MMiP 6 MiBP D4-MiBP MBP 13C -MBP 4 MEHHP 13C -MEHHP 4 MECPP D4-MECPP MEOHP 13C -MEOHP 4 MBzP 13C -MBzP 4 MCHP 13C -MCHP 4 MEHP 13C -MEHP 4 MOP 13C -MOP 4 MNP 13C -MNP 4 MDP 13C -MDP 4
6.65
13C
6.76 8.38 8.88 16.36 17.16 16.75 17.40 18.76 21.94 22.03 26.23 26.39 26.44 26.80
10 (27) 24 (28) 25 (18) 18 (28) 26 (20) 22 (20) 27 (19) 27 (29) 22 (20) 16 (21) 27 (20) 21 (29) 23 (29) 17 (32) 16 (18)
a The collision energy of the precursor/confirmation ion transition is shown in parentheses.
spectrometer was set in selective reaction monitoring mode. We monitored two precursor/product ion combinations (one for quantitation and one for confirmation) for each analyte, and one precursor/product ion combination for the quantitation of each isotopically labeled analogue. 2988
Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
Data Analysis. Data acquisition and analysis were performed by a program created using the Xcalibur software on a PC-based data system. The data analysis program automatically selected and integrated each ion of interest in the chromatogram. The identity of the phthalate metabolites was confirmed by matching retention times with the isotopically labeled internal standard. The peak integrations were corrected manually, if necessary. A response factor (RF), calculated as the peak area of each analyte ion divided by the peak area of its isotope-labeled standard, was used for quantification. For increased selectivity, the ratio of the peak area of the quantitation ion to the confirmation ion for each analyte was determined for each sample. We used 11 standard analyte concentrations encompassing the entire linear range of the method to construct daily calibration curves, weighted by the reciprocal of the standard amount (1/x), of RF versus standard amount. The full set of 11 standards was injected twicesbefore and after all samples (e.g., QCs, blanks, unknowns)sto monitor sensitivity changes. The calibration curves were linear over 3 orders of magnitude and had correlation coefficients exceeding 0.99. The calibration data, along with the integrated peak areas for each analyte and retention times, saved in a Microsoft Excel file, were exported to a Microsoft Access database, and the data were statistically analyzed using SAS statistical software (SAS Institute, Cary, NC). RESULTS AND DISCUSSION First, we evaluated two precolumnssChromolith Flash RP-18e (4.6 mm × 25 mm) and ALexa (2.1 mm × 25 mm, 30 µm particle size, Varian Inc., Walnut Creek, CA)sfor the on-line preconcentration of the phthalate metabolites from the urine. ALexa columns contain a proprietary alkyl-bonded phase on ultrapure silica designed to extract a broad range of polar, nonpolar, acidic, neutral, and basic analytes. Because of its relatively large particle size, the ALexa on-line SPE column displays size exclusion characteristics. Macromolecules, such as proteins, are not able to penetrate the pores and elute with the dead volume. In turn, the skeleton of the silica-based Flash RP-18e monolithic column contains an array of large (1 µm) and small (13 nm) pores. The small size of the silica skeleton and adequate size of the mesopores
Figure 2. HPLC/ESI-MS/MS chromatogram of (A) a standard mixture (concentrations are between 30 and 35 ng/mL) and (B) a nonspiked human urine extract. Concentrations (in ng/mL) in the extract were as follows: PA, 11.8; MCPP, 1.7; MEP, 24.7; MiBP, 1.3; MBP, 3.6; MECPP, 26.8; MEOHP, 20.4; MEHHP, 31.6; MBzP, 2.8; MEHP, 1.4; MMP, MMiP, MCHP, MOP, MNP, and MDP were not detected.
result in better mass transfer than conventional distinct particles. Therefore, monolithic columns can be used to preconcentrate dilute samples.38 On-line sample enrichment using monolithic precolumns in microcolumn liquid chromatography has been reported recently.39 Using MS/MS, we monitored the retention of the matrix components and of the analytes in the Flash RP-18e and ALexa (39) Lim, L. W.; Hirose, K.; Tatsumi, S.; Uzu, H.; Mizukami, M.; Takeuchi, T. J. Chromatogr. A 2004, 1033, 205-212.
columns. To select the optimal composition of the loading solvent, we determined the retention of the analytes at different acetic acid: organic solvent content ratios. Without the acetic acid, the phthalate metabolites did not bind effectively to the precolumn sorbents. The matrix band, corresponding to the unretained urine components, eluted completely in 1 min from the monolithic column, using 0.1% acetic acid:MeOH (90:10) or 0.1% acetic acid/ water:0.1% acetic acid/ACN (85:15) at 1.5 mL/min, and from the ALexa column, using 0.1% acetic acid:MeOH (95:5) at 0.8 mL/ Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
2989
Table 3. Precision of Replicate Measurements of Phthalate Metabolites in Spiked Quality Control (QC) Pools (N ) 60), Limit of Detection (LOD), and Accuracy of the Spiked Recoveries in Urine (N ) 5) precision QCL
a
QCH
accuracy
analyte
mean (ng/mL)
CV (%)
mean (ng/mL)
CV (%)
LOD (ng/mL)
concn (ng/mL)
recovery (%)
CVa (%)
PA MCPP MMP MMiP MEP MiBP MBP MEHHP MECPP MEOHP MCHP MBzP MEHP MOP MNP MDP
25.5 5.8 36.6 7.2 88.4 25.5 25.5 13.8 22.0 6.6 7.6 28.4 12.3 8.7 11.7 22.5
6.5 8.4 14.4 5.9 4.5 9.4 6.9 7.7 5.4 10.1 4.7 5.3 6.9 5.1 10.0 6.4
200.1 17.5 215.6 133.3 390.0 142.3 32.8 85.3 87.5 45.5 42.9 208.9 68.9 27.8 71.8 136.3
7.2 6.3 8.2 5.4 5.7 6.7 9.2 4.9 5.2 6.9 5.6 5.3 10.3 13.5 11.5 9.4
0.42 0.16 0.15 0.36 0.40 0.26 0.40 0.32 0.25 0.45 0.20 0.11 0.90 0.67 0.36 4.30
8.0 8.0 8.0 8.0 8.0 8.0 8.0 8.0 2.0 8.0 4.0 12.0 6.0 8.0 8.0 8.0
88.5 96.4 103.3 87.9 100.9 103.8 95.2 93.6 94.2 98.9 97.8 96.2 94.5 95.8 94.4 90.8
2.3 2.5 2.9 3.7 5.2 3.7 4.2 2.8 3.8 4.2 4.7 4.9 6.7 5.1 3.4 3.5
CV is the coefficient of variation.
min. Under these experimental conditions, phthalic acidsthe least retained analytesseparated from the matrix band in both columns. The unwanted urine components eluted from both ALexa and Flash RP-18e columns in a short time (1 min), while the phthalate metabolites were retained. We chose to use the monolithic column, however, because it provided the best resolution and resulted in the best peak shape of the analytes’ signals. Under our experimental conditions, the Flash RP-18e monolithic precolumn could be used for more than 600 sample injections, with no measurable loss in column performance or carryover. Next, we compared the retention time, resolution, and peak shape performance of several HPLC analytical columns, including Zorbax SB-CN (2.1 mm × 150 mm, 3.5 µm; Agilent Technologies), YMCODS-AQ (2.0 mm × 150 mm, 3 µm; Waters, Milford, MA), BETASIL Phenyl (2.1 mm × 150 mm, 3 µm), and Chromolith Performance RP-18e (4.6 mm × 100 mm; Merck KGaA, Germany) columns. We used the same preconcentration parameters, HPLC flow rate (0.35 mL/min), and mobile phase solvents (0.1% acetic acid in water and in ACN) and optimized the gradient for each column. The Zorbax SB-CN, YMCODS-AQ, and Chromolith Performance RP-18e columns provided a good separation between the isomeric MBP and MiBP peaks. However, the MEOHP and MEHHP peaks were broad and tailing. Furthermore, the MEHHP signal split into two peaks, corresponding to the two diastereomers of MEHHPsthe only phthalate metabolite of the 16 measured with two chiral centers. With the BETASIL Phenyl column, the MBP and MiBP signals were adequately resolved and the peak shapes of the MEHHP and MEOHP signals were also adequate (Figure 2). To validate the on-line SPE method, we analyzed 20 QCH and 20 QCL urine samples using both the on-line method and our automated off-line SPE method.24 The agreement between the two methods was excellent (Figure S-1). The on-line SPE method required less urine (100 µL vs 1 mL), solvents, reagents, laboratory waste (e.g., disposables), and sample handling than did the automated off-line SPE method.24 The on-line SPE method also 2990
Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
had a higher throughput (HPLC run time, including the preconcentration step, is 27 min) than the automated off-line SPE method (HPLC run time, excluding the SPE step, is 25 min). In addition, the on-line SPE method did not involve the evaporation and reconstitution of the urine extract; thus, the potential evaporation losses of analytes were eliminated. More importantly, the reduced use of solvents and reagents, along with the minimal sample preparation procedures, potentially minimized the exposure to hazardous chemicals. Spiked urine was analyzed repeatedly to determine the limits of detection (LODs) and the accuracy and precision of the method (Table 3 and Table S-2). The LOD was calculated as 3S0, where S0 is the standard deviation as the concentration approaches zero.40 S0 was determined from five to seven repeated measurements of low-level standards. The LOD values, which reflect the good sensitivity of the method, are comparable to the detection limits previously achieved using SPE.20,22,24,25 Specifically, the on-line SPE method was more sensitive for most analytes than our automated off-line SPE method.24 More importantly, our on-line SPE method allows the quantification of more phthalate metabolites in urine than have previous on-line SPE methods.14,20-22,25 The method accuracy was assessed through five replicate analyses of urine spiked at three different concentrations (8-16, 25-64, and 50192 ng/mL) and expressed as a percentage of the expected value. The intraday variability, reflected in the method accuracy, was very good (i.e., between 87.9% and 108.6%) for all 16 analytes at all spike levels (Table 3 and Table S-2). We determined the method precision by calculating the coefficients of variation (CVs) of 60 repeated measurements of the QCL and QCH materials over a period of 3 weeks (Table 3). The intraday CVs (
