Characterizing and Compensating for Matrix Effects Using

Feb 22, 2008 - Worsfold Water Quality Centre, Trent University, Peterborough, Ontario, Canada. Matrix effects are a great challenge for the quantitati...
0 downloads 0 Views 749KB Size
Anal. Chem. 2008, 80, 2010-2017

Characterizing and Compensating for Matrix Effects Using Atmospheric Pressure Chemical Ionization Liquid Chromatography-Tandem Mass Spectrometry: Analysis of Neutral Pharmaceuticals in Municipal Wastewater Xiaoming Zhao† and Chris D. Metcalfe*

Worsfold Water Quality Centre, Trent University, Peterborough, Ontario, Canada

Matrix effects are a great challenge for the quantitative analysis of environmental samples by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Signal suppression or enhancement can compromise the accuracy of analytical results. While matrix effects have been relatively well studied for applications of LC-MS/MS instrumentation with electrospray ionization, there have been relatively few studies to evaluate matrix effects when using atmospheric pressure chemical ionization (APCI) as the ion source. In this study, we determined the effects of sample matrix on the analysis of six neutral pharmaceuticals (i.e., caffeine, cotinine, carbamazepine and its major metabolite, carbamazepine-10,11-dihydrodiol, trimethoprim, and fluoxetine) in samples of municipal wastewater using LC-APCI-MS/MS and evaluated whether isotope-labeled internal standards can be used to compensate for matrix effects. The matrix effects were measured using postextraction spikes and postcolumn direct infusion, respectively. The results showed that the matrix in the extracts prepared from municipal wastewater enhanced the signals for four of the six analytes when using an APCI source. Without correction for signal enhancement, apparent recoveries of the analytes from wastewater samples were overestimated to levels as high as 178% of the spiked amount. Isotope-labeled compounds corrected for these overestimates that occurred as a result of interferences from the sample matrix. Hyphenated mass spectrometric techniques, such as liquid chromatography-mass spectrometry (LC-MS) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) have been widely used for many analytical applications, such as the analysis of pharmaceuticals in body fluids, environmental samples, and foods.1-6 Liquid chromatography-tandem mass spectrometry * To whom correspondence should be addressed. Phone: 705-748-1011, ext. 7272. Fax: 705-748-1569. E-mail: [email protected]. † Present address. Ontario Ministry of the Environment, Etobicoke, Ontario, Canada. (1) Reemtsma, T. Trends Anal. Chem. 2001, 20, 500-517. (2) Reemtsma, T. Trends Anal. Chem. 2001, 20, 533-542. (3) Petrovic´, M.; Hernando, M. D.; Dı´az-Cruz, M. S.; Barcelo´, D. J. Chromatogr., A 2005, 1067, 1-14.

2010 Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

has gained great popularity because of its high selectivity and sensitivity for the quantitative determination of trace levels of analytes of medium to high polarity. A major milestone in the evolution of this analytical instrumentation was the development of atmospheric pressure ionization sources, including electrospray (ESI)) and atmospheric pressure chemical ionization (APCI), which are “soft” ionization techniques. However, depending on the complexity of the samples, the matrix coextracted with the analytes can modify the signal, causing either ion suppression or enhancement when using these ionization techniques.7,8 These “matrix effects” result in poor analytical accuracy and reproducibility. There have been extensive reports on the ion suppression observed when ESI is used as the ionization technique with either LC-MS or LC-MS/MS.9-14 In these cases, signal suppression is believed to result from competition between ions generated from the analyte and matrix components for the access to the droplet surface for gas-phase emission. In environmental analysis, ESI is currently the most widely employed ionization technique. Our previous studies with LC-ESI-MS/MS showed that the degree of ion suppression increased with the complexity of the sample matrix when analyzing pharmaceuticals in municipal wastewater.15 It has been reported that APCI is less sensitive to matrix effects (4) Dı´az-Cruz, M. S.; Barcelo´, D. Anal. Bioanal. Chem. 2006, 386, 973-985. (5) Gros, M.; Petrovic´, M.; Barcelo´, D. Anal. Bioanal. Chem. 2006, 386, 941952. (6) Kennedy, D. G.; McCracken, R. J.; Cannavan, A.; Hewitt, S. A. J. Chromatogr., A 1998, 812, 77-98. (7) Liang, H. R.; Foltz, R. L.; Meng, M.; Bennett, P. Rapid Commun. Mass Spectrom. 2003, 17, 2815-2821. (8) Schuhmacher, J.; Zimmer, D.; Tesche, F.; Pickard, V. Rapid Commun. Mass Spectrom. 2003, 17, 1950-1957. (9) Fu, I.; Woolf, E. J.; Matuszewski, B. K. J. Pharm. Biomed. Anal. 1998, 18, 347-357. (10) Choi, B. K.; Hercules, D. M.; Gusev, A. I. J. Chromatogr., A 2001, 907, 337-342. (11) Mallet, C. R.; Lu, Z.; Mazzeo, J. R. Rapid Commun. Mass Spectrom. 2004, 18, 49-58. (12) Hernando, M. D.; Petrovic´, M.; Ferna´ndez-Alba, A. R.; Barcelo´, D. J. Chromatogr., A 2004, 1046, 133-140. (13) Go´mez, M. J.; Petrovic´, M.; Ferna´ndez-Alba, A. R.; Barcelo´, D. J. Chromatogr., A 2006, 1114, 224-233. (14) Vieno, N. M.; Tuhkanen, T.; Kronberg, L. J. Chromatogr., A 2006, 1134, 101-111. (15) Miao, X.-S.; Metcalfe, C. D. Anal. Chem. 2003, 75, 3731-3738. 10.1021/ac701633m CCC: $40.75

© 2008 American Chemical Society Published on Web 02/22/2008

than ESI.16,17 However, both ion suppression and enhancement have been observed for LC-MS/MS using APCI.7,18 Gas-phase neutralization processes and coprecipitation of the analytes with nonvolatile materials may be responsible for the matrix effects observed with APCI. Regardless, suitable strategies need to be developed to compensate for the matrix effects in both ESI and APCI. Several operational strategies can be employed to compensate for the interference caused by matrix components. Extensive cleanup procedures prior to LC-MS/MS analysis could help to reduce the introduction of matrix components into the analytical system.19,20 However, these are sometimes laborious. Improved chromatographic separation can often reduce the interference of the matrix components with the target analytes.21-23 Another approach is to dilute as much as possible the final sample extract to be injected onto the analytical column. In some instance, this method is effective for eliminating signal suppression, while achieving acceptable sensitivity.12,13 In most cases, the matrix effects cannot be completely eliminated, despite the use of the aforementioned operational strategies. Standard additions of each analyte into the sample may also be used to compensate for matrix effects.24 However, the approach is time-consuming and laborious and does not improve analytical sensitivity. One of the most effective options to compensate for signal suppression or enhancement is the use of isotope-labeled analogues for quantitative analysis19,20 as they are chemically and structurally the same as their target analytes but differ in molecular mass. In LC-MS/MS analysis, it is expected that ionization of the target analytes and the isotope-labeled analogues will be suppressed or enhanced to the same extent, as the analogues should have the same chromatographic retention times and ionization characteristics as the analytes. These analogues should be labeled at a sufficient number of sites to achieve adequate mass resolution from the native compounds (i.e., > 2 amu). In this study, we evaluated the effects of the sample matrix on the LC-APCI-MS/MS analysis of six neutral pharmaceuticals (i.e., caffeine, cotinine, carbamazepine and its major metabolite, carbamazepine-10,11-dihydrodiol (CBZ-DiOH), trimethoprim, and fluoxetine). The sample matrixes were extracts prepared from wastewater samples collected at a municipal wastewater treatment plant (WWTP). WWTPs are the primary source of pharmaceutically active compounds (PhACs) that are discharged into the aquatic environment. Within WWTPs, there is incomplete removal of many classes of PhACs and nanogram per liter to microgram (16) Vanderford, B. J.; Pearson, R. A.; Rexing, D. J.; Snyder, S. A. Anal. Chem. 2003, 75, 6265-6274. (17) Zuehlke, S.; Duennbier, U.; Heberer, T. Anal. Chem. 2004, 76, 6548-6554. (18) van Hout, M. W. J.; Niederla¨nder, H. A. G.; de Zeeuw, R. A.; de Jong, G. J. Rapid Cummun. Mass Spectrom. 2003, 17, 245-250. (19) Benijts, T.; Dams, R.; Lambert, W.; De Leenheer, A. J. Chromatogr., A 2004, 1029, 153-159. (20) Van De Steene, J. C.; Mortier, K. A.; Lambert, W. E. J. Chromatogr., A 2006, 1123, 71-81. (21) Choi, B. K.; Hercules, D. M.; Gusev, A. I. Anal. Chem. 2001, 369, 370377. (22) Dijkman, E.; Mooibroek, D.; Hoogerbrugge, R.; Hogendoorn, E.; Sancho, J.-V.; Pozo, O.; Herna´ndez, F. J. Chromatogr., A 2001, 926, 113-125. (23) Pascoe, R.; Foley, J. P.; Gusev, A. I. Anal. Chem. 2001, 73, 6014-6023. (24) Miao, X.-S.; Metcalfe, C. D. In Analysis, Fate and Removal of Pharmaceuticals in the Water Cycle; Petrovic, M., Barcelo, D., Eds.; Elsevier: Amsterdam, The Netherlands, 2007.

per liter concentrations of these compounds have been detected in treated wastewater.15,25,26 Because of the trace concentrations of these compounds in environmental samples, preconcentration needs to be performed prior to instrumental analysis, and this is generally done by extraction onto solid-phase extraction (SPE) cartridges.24 Coextracted matrix components are inevitably eluted from the SPE cartridges into the extract. In this study, matrix effects were compared in different matrixes, including Milli-Q water, untreated wastewater (i.e., influent), and treated wastewater (i.e., effluent) using APCI as the ionization technique for LC-MS/MS analysis. Isotope-labeled compounds were utilized as the internal standards to compensate for variations in the matrix effects observed for extracts of complex environmental samples. EXPERIMENTAL SECTION Reagents and Chemicals. Caffeine was purchased from Aldrich (Milwaukee, WI). Carbamazepine, (-)-cotinine, fluoxetine hydrochloride, and trimethoprim were supplied by Sigma (St Louis, MO). Carbamazepine-10,11-dihydrodiol (CBZ-DiOH), a metabolite of carbamazepine, was provided by Novartis Pharma AG (Basel, Switzerland). 10,11-Dihydrocarbamazepine (CBZ-DiH) was kindly provided by Dr. Thomas Ternes (Bfg, Germany). Fluoxetine-d5 HCl (phenyl-d5, 99.67 atom % D) and DL-cotinine-d3 (N-methyl-d3, 99.9 atom % D) were purchased from CDN Isotopes (Pointe-Claire, QC, Canada). Caffeine-13C3 and carbamazepine-d10 (98 atom % D) were purchased from Cambridge Isotope Laboratories, Inc. (Andover, MA). Stock solutions of the analytes were prepared at concentrations of 1 mg/mL in 50:50 methanol-water. The working solutions were then prepared by serial dilution of stock solutions. Methanol and acetone were supplied by Caledon Laboratories (Georgetown, ON, Canada). Acetonitrile (Optima) was purchased from Fisher Scientific (Mississauga, ON, Canada). Ultrapure water was produced by a Milli-Q water purifying system (Millipore, Toronto, ON, Canada). Ammonium acetate was supplied by Sigma. All chemicals were used as provided, without further purification. LC-MS/MS Analysis. Chromatographic separations were conducted with an Agilent 1100 series HPLC system consisting of a binary pump, an autosampler, and an in-line solvent degasser. The analytes were separated at room temperature on a Genesis C18 column (150 mm × 3 mm, 4 µm; Jones Chromatography, Hengoed, U.K.) using a mobile phase flow rate of 0.5 mL/min. The guard column used was also of C18 material (4.0 mm × 2.0 mm; Phenomenex, Torrance, CA). The mobile phases A and B were an aqueous solution of 10 mM ammonium acetate and acetonitrile, respectively. The following linear elution gradient was used to achieve the chromatographic separation of the target analytes: mobile phase B was increased from the initial 5% to 20% over 4 min and reached 95% at 12 min; it remained at 95% for the next 3 min and then was ramped back to 5% within 2 min. The sample injection volume was 20 µL. Mass spectrometric determination was performed using an MDS/Sciex (Toronto, ON, Canada) QTrap mass spectrometer equipped with an APCI source and controlled by Analyst 1.4 (25) Metcalfe, C. D.; Miao, X.-S.; Koenig, B. G.; Struger, J. Environ. Toxicol. Chem. 2003, 22, 2881-2889. (26) Metcalfe, C. D.; Koenig, B. G.; Bennie, D. T.; Servos, M.; Ternes, T. A.; Hirsch, R. Environ. Toxicol. Chem. 2003, 22, 2872-2880.

Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

2011

Table 1. Compound Name, CAS Number, Formula, Molecular Weight, Chromatographic Retention Times, and Optimized APCI-MS/MS Parameters compound

CAS no.

caffeine

58-08-2

cotinine

486-56-6

carbamazepine

298-46-4

CBZ-DiOHb

35079-97-1

trimethoprim

738-70-5

fluoxetine

54910-89-3

CBZ-DiHc

3564-73-6

caffeine-13C3d cotinine-d3d carbamazepine-d10d fluoxetine-d5d

110952-70-0 132183-78-9

formula and MWa

retention time (min)

MRM transition

declustering potential (V)

collision energy (eV)

C8H10N4O2 194.08 C10H12N2O 176.09 C15H12N2O 236.09 C15H14N2O3 270.10 C14H18N4O3 290.14 C17H18F3NO 309.13 C15H14N2O 238.11

7.42

195 f 138

47

25

7.16

177 f 80

40

40

10.89

237 f 194

40

28

9.01

271 f 236

15

20

8.98

291 f 230

40

28

12.63

310 f 44

15

17

10.96

239 f 194

40

28

7.43 7.13 10.84 12.59

198 f 140 180 f 80 247 f 204 315 f 44

47 40 25 15

25 40 30 17

a MW, molecular weight calculated for the lowest monoisotopes. b CBZ-DiOH, carbamazepine-10,11-dihydrodiol. c CBZ-DiH, 10,11-dihydrocarbamazepine, performance standard. d Internal standards.

software. Multiple reaction monitoring (MRM) in the positive ion mode with unit resolution on both the first and second analyzer was employed for data acquisition. MRM provided sufficient specificity for the identification of the analytes, while achieving sufficient sensitivity for detection of the analytes at nanogram per liter concentrations in water and wastewater. Nitrogen was used as the curtain, nebulizer, auxiliary, and collision gases. Optimization of the mass spectrometric parameters was accomplished in studies with continuous infusion of a solution containing 1.0 µg/ mL concentrations of the analytes. A built-in syringe pump was used for infusion at a flow rate of 20 µL/min, which was then connected by a tee into the LC flow of 50:50 10 mM aqueous ammonium acetate-acetonitrile at 0.5 mL/min. The operating conditions of the mass spectrometer were curtain gas 20 psi, nebulizer gas 60 psi, auxiliary gas 15 psi, corona discharge needle current 2.0 µA, probe temperature 470 °C, interface heater on, CAD gas 5 (corresponding to vacuum gauge operating pressure of 3.1 × 10-5 Torr). Collision energy and declustering potential for each analyte were optimized to the parameters listed in Table 1. The dwell time for each MRM transition was set at 200 ms. Wastewater Collection. Grab samples of untreated (influent) and treated (effluent) wastewater were collected from the WWTP for the city of Peterborough, ON, Canada. The WWTP currently serves a population of about 75 000 people according to the municipal utility. Its designed average flow capacity is 60 000 m3/ day, and its average handling flow is presently 46 000 m3/day (77% of its design capacity). The influent flow rate consists of about 25% industrial wastewater and 75% domestic sewage. The treatment process includes grit removal and screening, primary treatment, and secondary treatment, followed by UV disinfection. The wastewater samples were collected in solvent washed amber glass bottles, which were stored in a cold room at 4 °C and processed within a week of collection. Because of the possible presence of microbial pathogens in the influent wastewater, caution must be taken to avoid direct contact with wastewater and protective gloves must be worn. 2012 Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

Sample Preparation. Protective masks and gloves are required to avoid direct contact with domestic wastewater that contains human pathogens. In addition, the operation of the solid-phase extraction device and the use of all solvents should be performed in the fume hood. To remove suspended solids, wastewater samples were filtered under vacuum through Whatman 1.5 µm glass microfiber filters. No filtration was required for samples prepared from Milli-Q water. The pH of the samples was adjusted to 7.5 with 3.5 M H2SO4 and 1 M NaOH. Note that fluoxetine, which is a weak base, is in the neutral form at this pH. Volumes of 100 mL of Milli-Q water, effluent, and influent wastewater were applied onto Waters (Milford, MA) Oasis HLB cartridges (6 mL/500 mg), which had been preconditioned sequentially with 6 mL of acetone, 6 mL of methanol, and 6 mL of water of pH 7.5 on a vacuum assembly. The flow rate was controlled at about 5 mL/min. At the end of the extraction, the sample bottles were rinsed with 10 mL of Milli-Q water at pH 7.5, and the rinses were passed through the SPE cartridges. Thereafter, the cartridges were dried under the vacuum for 10 min before elution with 3 × 2 mL of methanol. Each aliquot of methanol was eluted through the cartridges for a minimum of 10 min. After elution, the extracts were concentrated on a UVS 400 (Savant Instruments, Farmingdale, NY) vacuum centrifuge to almost dryness and then reconstituted in 1 mL of 50:50 methanol-water in preparation for LC-MS/MS analysis. First, two sets of samples were prepared to investigate matrix effects and the analyte recoveries, respectively. In one set, the sample extracts after SPE extraction were spiked with the target analytes (i.e., caffeine, cotinine, carbamazepine, CBZ-DiOH, trimethoprim, fluoxetine) at two concentrations, together with fixed concentrations of the isotope-labeled internal standards (i.e., caffeine-13C3, cotinine-d3, carbamazepine-d10, and fluoxetine-d5) and a fixed amount of performance surrogate, CBZ-DiH. In another set of samples, the same target analytes, plus the isotope-labeled internal standards and performance surrogate were added before

Figure 1. Matrix effects evaluated for neutral drugs spiked into Milli-Q water and effluent and influent wastewater extracts at lower concentrations (0.5 µg/L), before (A) and after (B) correction with corresponding internal standards. The results are the average of three replicate spikes and the error bar represents the relative standard deviation (RSD).

the SPE procedure was performed. Triplicate samples were prepared for each spiking level. The third set of experiments was performed using postcolumn infusion. The SPE sample extracts from MilliQ water, effluent and influent wastewater that were prepared without adding any target analytes, performance surrogate, or isotope-labeled internal standards were injected onto the analytical column, and the gradient elution conditions described above in the LC-MS/MS analysis section was employed to chromatographically separate the analytes. As a reference, a solvent blank of 50:50 methanol-water was also injected using these chromatography conditions. At the same time, a neat standard solution containing the target analytes, performance surrogate, and isotope-labeled internal standards was infused into the postcolumn LC flow through a tee-connection using a syringe pump at a constant flow of 10 µL/min. The continuous signal detection throughout the LC process was achieved by monitoring the MRM transitions for each compound, with the optimized LC-APCI-MS/MS parameters (Table 1). RESULTS AND DISCUSSION The target analytes are commonly detected in wastewater samples. Therefore, in order to measure the matrix effects or recovery of the extraction procedures, the amounts of the analytes in nonspiked samples must be subtracted from the amounts in the spiked samples. For this purpose, both nonspiked and spiked samples were simultaneously extracted using SPE and analyzed by LC-APCI-MS/MS.

Evaluation of Matrix Effects Using Post-Extraction Spikes. To evaluate the matrix effects, MilliQ water and effluent and influent wastewaters were extracted and concentrated on HLB cartridges. The SPE extracts were then spiked with two levels (0.5 and 5 µg/L) of the target analytes, a performance surrogate (CBZ-DiH) at a constant concentration (0.5 µg/L), and isotopelabeled internal standards at constant concentrations, respectively, followed by LC-MS/MS analysis. The matrix effects (i.e., signal suppression or enhancement) were assessed using methods described by Matuszewski et al.,27 which involve relating the peak area obtained from a neat standard (A) to the peak area of the corresponding analyte in a matrix sample (B). The matrix effects (ME) were calculated as follows:

ME (%) ) B/A × 100 An ME value either smaller or greater than 100% indicates the presence of signal suppression or enhancement, respectively, while a value equal to 100% shows that the system is free of any interference from matrix components. The analytical results that reflect the effects of the matrix components on the target analytes are shown for samples spiked at the lower concentration in Figure 1 and for sample spiked at the higher concentration in Figure 2. Before correction with the isotope-labeled internal standards, no (27) Matuszewski, B. K.; Constanzer, M. L.; Chavez-Eng, C. M. Anal. Chem. 2003, 75, 3019-3030.

Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

2013

Figure 2. Matrix effects evaluated for neutral drugs spiked into Milli-Q water and effluent and influent wastewater extracts at higher concentration (5 µg/L), before (A) and after (B) correction with corresponding internal standards. The results are the average of three replicate spikes, and the error bar represents the relative standard deviation (RSD).

significant matrix effects were observed in extracts from MilliQ water extracts, while carbamazepine, CBZ-DiOH, trimethoprim, and CBZ-DiH showed considerable signal enhancement in extracts from wastewater effluent and influent spiked at both the high and low concentrations. Little effect of the sample matrix was observed for the ionization of caffeine and cotinine, and the responses for fluoxetine were variable across the two spiking concentrations (Figures 1 and 2). Where signal enhancement was observed, the matrix effects were clearly greater for extracts from untreated wastewater (i.e., influent) than for extracts from treated wastewater (i.e., effluent), reflecting the greater complexity of the matrix extracted from the influent sample. The signal enhancement using APCI as the ionization source can be attributed to the presence of coextracted components in the final extracts, which coeluted with the analytes of interest during chromatographic separation. Because the isotope-labeled analogues have the same retention times and ionization characteristics, they can be used to compensate for the signal enhancement observed in these LC-APCI-MS/ MS analyses. Therefore, the matrix effects were calculated by correcting for the response with the corresponding internal standards, which turned out to give matrix effects between 80 and 120%. It appears that since CBZ-DiOH and CBZ-DiH are structurally similar to carbamazepine, carbamazepine-d10 could be used to at least partially compensate for the matrix effects for all three analytes. Although trimethoprim is structurally and pharmacologically different from carbamazepine, carbamazepine-d10 2014

Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

was still effective at compensating for the effects caused by the interfering matrixes. In selecting an appropriate isotopelabeled internal standard, two factors need to be taken into account. If available, an isotope-labeled analogue should be selected. In cases where these analogues are not available, an isotope-labeled compound that has a similar chromatographic retention time and responds similarly to the matrix components can be chosen. Evaluation of Matrix Effects by Postcolumn Infusion. In addition to measuring the matrix effects using postextraction spikes, the effects of matrix components on the target analytes and corresponding internal standards were further demonstrated by postcolumn direct infusion of neat standard solution into the mobile phase flow of the extracts prepared from MilliQ water and effluent and influent wastewaters. As a reference, an injection of a solvent blank consisting of 50:50 methanol-water was also performed, accompanied by postcolumn direct infusion of a neat standard solution into the mobile phase flow. This allows the matrix effects to be investigated throughout the whole chromatographic run. These studies showed that caffeine and continine, and their corresponding isotope-labeled compounds, did not show significant signal suppression and enhancement at their respective retention times (Figure 3). By comparison, the signals for carbamazepine, carbamazepine-d10, CBZ-DiOH, trimethoprim, and CBZ-DiH were greatly enhanced in extracts from the effluent and influent (Figure 3). These results are consistent with those

Figure 3. Matrix effects evaluated using postcolumn infusion of target analytes, isotope-labeled internal standards, and performance surrogate into the flow of the injected solvent blank, Milli-Q water extract, and effluent and influent wastewater extracts (concentration level: 2.5 µg/L). A, caffeine and caffeine-13C3; B, cotinine and cotinine-d3; C, carbamazepine and carbamazepine-d10; D, CBZ-DiOH and trimethoprim; E, fluoxetine and fluoxetine-d5; F, CBZ-DiH. The arrows indicate the retention times for the chromatographic separation.

obtained when the analytes were spiked into the extracts after SPE. Noteworthy is the trend for fluoxetine in the postcolumn experiment in which the signal was somehow enhanced in the

effluent and influent wastewaters compared to that in the solvent blank and Milli-Q water. This phenomenon was also observed in the study of matrix effects using postextraction spikes at the Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

2015

Table 2. Overall Recoveries of Target Analytes in Different Matrixes before and after Correction with Isotope-Labeled Internal Standards spiked concentration (0.5 µg/L)

spiked concentration (5 µg/L)

recovery ( RSD (%)

recovery ( RSD (%)

compound

before correction

after correction

before correction

after correction

caffeine cotinine carbamazepine CBZ-DiOH trimethoprim fluoxetine CBZ-DiHa

87 ( 4 86 ( 4 81 ( 2 77 ( 1 87 ( 1 35 ( 8 85 ( 4

MilliQ Water 100 ( 7 103 ( 2 94 ( 8 89 ( 6 101 ( 6 100 ( 7 99 ( 8

71 ( 8 90 ( 6 81 ( 8 88 ( 8 83 ( 6 37 ( 9 85 ( 3

106 ( 6 102 ( 5 97 ( 6 106 ( 6 100 ( 5 94 ( 3 103 ( 9

caffeine cotinine carbamazepine CBZ-DiOH trimethoprim fluoxetine CBZ-DiH

75 ( 11 85 ( 2 124 ( 3 178 ( 4 145 ( 2 84 ( 6 100 ( 4

Effluent Wastewater 91 ( 6 94 ( 2 95 ( 2 136 ( 4 111 ( 3 92 ( 2 76 ( 5

88 ( 7 92 ( 3 114 ( 5 148 ( 2 105 ( 1 97 ( 2 117 ( 4

107 ( 2 106 ( 6 100 ( 1 130 ( 4 92 ( 5 100 ( 2 102 ( 2

88 ( 4 78 ( 1 139 ( 1 152 ( 5 124 ( 3 96 ( 9 137 ( 3

98 ( 3 91 ( 2 95 ( 2 104 ( 4 85 ( 2 90 ( 7 94 ( 1

Influent Wastewater caffeineb cotinine carbamazepine CBZ-DiOH trimethoprim fluoxetine CBZ-DiHa

79 ( 5 150 ( 1 155 ( 3 161 ( 5 86 ( 6 129 ( 3

88 ( 6 95 ( 4 105 ( 3 109 ( 8 84 ( 4 86 ( 6

a The spiking concentration of CBZ-DiH was constant throughout the experiment (0.5 µg/L). b Because of the high concentration of caffeine natively present in the influent wastewater, only 10 mL was used and a higher concentration level was spiked (5 µg/L).

Figure 4. Extracted ion chromatograms of target analytes, isotope labeled internal standards, and performance surrogate in neat solvent (A) and in an extract from an influent wastewater sample (B). Note the shift in retention time between chromatograms A and B for both fluoxetine and fluoxetine-d5, which was probably caused by the sample matrix. The presence of fluoxetine in the wastewater was confirmed by spiking the analyte into the sample.

higher spiking level of fluoxetine (see Figure 2A). Thus, the behavior of fluoxetine in the postcolumn infusion experiments is comparable to that in the higher concentration postextraction spikes. 2016 Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

The higher spiking level for fluoxetine in the postextraction experiment was very similar to the concentration used in the postcolumn infusion, which might explain the signal enhancement observed in these treatments. Regarding the elution profile of the

Table 3. Mean (n ) 3) Measured Concentrations ((SD) in Influent and Effluent Wastewater Samples (ng/L) compound

influent

effluent

caffeine cotinine carbamazepine CBZ-DiOH trimethoprim fluoxetine

70,066.7 ( 5154.0 2370.0 ( 112.7 539.3 ( 18.5 1203.3 ( 89.2 210.0 ( 14.2 37.6 ( 6.8

46.7 ( 10.2 54.1 ( 4.6 453.3 ( 14.0 1343.3 ( 81.4 192.3 ( 11.0 ND

postcolumn infused target compounds and their isotope-labeled analogues, it can be seen that they were affected by the matrix components to a similar extent in Milli-Q water and the effluent and influent wastewater extracts. In reverse-phase chromatographic separations, it is believed that the majority of matrix effects occur in the solvent front or early in the chromatographic separation.24 However, as seen from Figure 3, the matrix effects observed in effluent and influent wastewater samples spanned over a large range of the chromatographic run. Normally, the most polar matrix components can be separated from the target analytes by selecting a mobile phase that will elute these components at the void volume. However, it appears from this work that later eluting matrix components may interfere with the target analytes, thus affecting their signal intensity. It is obvious from this study that signal enhancement was observed for some of the analytes and not for others. Thus, the chemical properties, retention times, and the elution profile of the sample matrix all contributed to the degrees of matrix effects observed during LC-APCI-MS/MS analysis. Overall Performance of SPE LC-MS/MS Procedure. The overall recoveries of target analytes during the SPE LC-MS/MS process, which is one measure of the performance of the sample extraction procedure, are listed in Table 2. Recoveries are listed for the target analytes at two concentrations when spiked into Milli-Q water and effluent and influent wastewaters. The overall recoveries resulting from this process were calculated by comparing the peak areas of the anlaytes of interest spiked into water samples before SPE to those in neat solvents. As shown in Table 2, the uncorrected recovery data show higher apparent recoveries for some of the analytes when they were extracted from wastewater samples relative to the lower apparent recoveries when they were extracted from Milli-Q water (i.e., carbamazepine, CBZDiOH, and trimethoprim). Signal enhancement produces biased results that overestimate the actual concentrations. Therefore, isotope-labeled compounds are essential as internal standards to correct for overestimates of the concentrations of the target analytes (Table 2). Note that the recoveries for fluoxetine are lower than for the other neutral compounds. Fluoxetine is actually a weak base, with a pKa of 8.7. A pH of 7.5 for SPE sample preparation is not optimal for extraction of fluoxetine on an HLB cartridge. It has been suggested that the signal enhancements observed with APCI are due to gas-phase neutralization processes and/or

coprecipitation of the analytes with nonvolatile materials,7,18 but any attempt to explain the results observed in this study would be speculation. The matrix effects are mainly caused by components that are coextracted during sample preparation and are coeluted from the chromatographic column. In municipal wastewater, the matrix components could consist of organic compounds excreted by humans, such as carbohydrates, proteins, and fats, or anthropogenic chemicals of domestic origin (e.g., surfactants). As native target compounds and their isotope-labeled internal standards are eluted very closely during chromatographic separation and possess the same or very similar physicochemical properties, their signals are enhanced or suppressed to the same extent during ionization, despite the presence of other interfering sample components. The subtle differences in mass between the native compounds and their isotope-labeled analogues should not result in different degrees of signal suppression or enhancement. Therefore, the use of isotope-labeled compounds can compensate for matrix effects. Application to Environmental Samples. On the basis of the validated SPE and LC-MS/MS methods described previously, the influent and effluent samples collected from wastewater treatment plant were analyzed using the selected isotope-labeled compounds as internal standards. As the main goal of this study was to determine the matrix effects and ways to circumvent this problem, we just reported the numerical values of the measured concentrations of the target analytes without giving the method detection limits. With isotope-labeled internal standards, reliable and accurate analytical results could be obtained, which are shown in Table 3. The extracted ion chromatograms are shown in Figure 4 for target analytes, isotope-labeled internal standards, and performance surrogate in both neat solvent and wastewater samples. The concentrations reported in Table 3 are consistent with the concentrations of these analytes previously reported in treated and untreated wastewater samples from Canada.15,25,26 Comparison between the concentrations in influent and effluent also show that caffeine and cotinine are effectively removed from wastewater by the treatment process, while the other analytes are removed less efficiently, with the possible exception of fluoxetine. CONCLUSIONS Signal enhancement was observed in wastewater extracts analyzed for neutral pharmaceuticals using LC-APCI-MS/MS as a result of interferences from the sample matrix. Without correction for these matrix effects by using isotope-labeled internal standards, the concentrations of several analytes would have been overestimated in wastewater samples. Elimination and/or compensation for matrix effects is an important component of method development and validation for the analysis of pharmaceuticals in the environment; especially when analyzing extracts prepared from complex sample matrixes, such as municipal wastewater. Received for review August 1, 2007. Accepted January 4, 2008. AC701633M

Analytical Chemistry, Vol. 80, No. 6, March 15, 2008

2017