2882 Anna Laura Capriotti Chiara Cavaliere Susy Piovesana Roberto Samperi Serena Stampachiacchiere Salvatore Ventura Aldo Lagana` Dipartimento di Chimica, Universita` di Roma “La Sapienza”, Piazzale Aldo Moro 5, Rome, Italy Received June 27, 2014 Revised July 26, 2014 Accepted July 28, 2014
J. Sep. Sci. 2014, 37, 2882–2891
Research Article
Multiresidue determination of UV filters in water samples by solid-phase extraction and liquid chromatography with tandem mass spectrometry analysis UV filters, contained in sunscreens and other cosmetic products, as well as in some plastics and industrial products, are nowadays considered contaminants of emerging concern because their widespread and increasing use has lead to their presence in the environment. Furthermore, some UV filters are suspected to have endocrine disruption activity. In the present work, we developed an analytical method based on liquid chromatography with tandem mass spectrometry for the determination of UV filters in tap and lake waters. Sixteen UV filters were extracted from water samples by solid-phase extraction employing graphitized carbon black as adsorbent material. Handling 200 mL of water sample, satisfactory recoveries were obtained for almost all the analytes. The limits of detection and quantification of the method were comparable to those reported in other works, and ranged between 0.7–3.5 and 1.9–11.8 ng/L, respectively; however in our case the number of investigated compounds was larger. The major encountered problem in method development was to identify the background contamination sources and reduce their contribution. UV filters were not detected in tap water samples, whereas the analyses conducted on samples collected from three different lakes showed that the swimming areas are most subject to UV filter contamination. Keywords: Emerging contaminants / Lake waters / Liquid chromatography with mass spectrometry / Solid-phase extraction / UV filters DOI 10.1002/jssc.201400708
Additional supporting information may be found in the online version of this article at the publisher’s web-site
1 Introduction In the last years, there has been an increasing attention of both public institutions and researchers toward the “contaminants of emerging concern,” including not only new substances appeared recently, but also known substances for which the environmental contamination issues were not fully realized and concerns have been raised much more recently [1]. Among these contaminants there is the commercial category of UV filters, which include various and heterogeneous chemical classes. UV filters are present as ingredients in personal care products, such as sunscreen products, beauty creams, cosmetics, lipsticks, hair sprays, shampoos etc., to protect Correspondence: Dr. Chiara Cavaliere, Dipartimento di Chimica, Universita` di Roma “La Sapienza”, Piazzale Aldo Moro 5, 00185 Rome, Italy E-mail: [email protected] Fax: +39-06-490631
Abbreviations: GCB, graphitized carbon black; IS, internal standard; ME, matrix effect; SRM, selected reaction monitoring
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
skin and hair from the negative effects of sunlight. These compounds can enter the aquatic environment both directly from recreational activities (e.g. sunbathing and swimming in seas, lakes, and rivers) and industrial wastewater discharges and indirectly via wastewater treatment plants [2–4]. Furthermore, some UV-absorbing species (UV stabilizers) [5] are also present in several plastic products and packaging, paints, glasses, textiles, to prevent yellowing and degradation of polymers and pigments [6]. As a consequence of their widespread employment in various personal care and industrial products, UV filters are continuously released in the environment, and they have been found in lakes, rivers, wastewater, drinking water, soil, sludge, and fish [2]. From the toxicological point of view, besides possible allergenic effects, some UV filters have shown significant estrogenic and/or anti-androgenic activity and potential developmental toxicity in experimental animals [3, 7]. The levels of UV filters detected in the environment are not so far below the doses that cause toxic effects in animals [8,9]. UV filters can be both inorganic (e.g. titanium dioxide, zinc oxide) that work by reflecting and scattering UV light, and organic that work by absorbing UV light in the
www.jss-journal.com
Liquid Chromatography
J. Sep. Sci. 2014, 37, 2882–2891
wavelength range of 280−315 (UVB) and/or 315−400 nm (UVA) [8]. Organic UV filters include benzophenones, camphor derivatives, cinnamates, crylenes, benzimidazole derivatives, p-aminobenzoic acid, and derivatives, dibenzoyl methane derivatives, salicylates, and triazines, as well as their primary transformation products [6, 8]. Currently, the European Regulation on cosmetic products [10] permits the employment of 26 compounds, within certain limits, in cosmetic (sunscreen) products, and 25 of these are organic compounds [3]. Although several analytical methods based on GC–MS for UV filter determination in waters are published in the literature [11–18], methods based on LC–MS/MS [6, 19–29] are often preferred for the simultaneous analysis of many compounds and their environmental transformation products within a wide range of physicochemical properties [24]. Indeed, the UV filters selected in the published works vary considerably, from the very polar sulfonates to the less polar salycilates [22]. Therefore, for their analysis different API sources, i.e. the most widespread ESI [19, 21, 24, 27], APCI [22, 23, 28] and atmospheric pressure photoionization [20] were used in both polarity ion modes. ESI is the ionization source of choice when LC–MS/MS multiclass methods for the determination of personal care products and pharmaceuticals in aquatic environment are proposed [25, 26, 28, 29]. CE–ESI-MS has also been used for UV filter determination in river water [30]. In most cases, before LC–MS analysis UV filters were extracted from water by SPE using as sorbent material a styrenedivinylbenzene polymer [24] or a hydrophilic–lipophilic polymer such as Oasis HLB [19, 21, 25, 26, 28, 29] and Bond Elut Plexa [25]. More recently, microextraction techniques aimed at the minimization of solvent consumption [31] have been used, such as stir bar sorptive extraction [22, 23, 27] and nonporous membrane-assisted LLE [20]; the very promising ionic liquid-based dispersive liquid–liquid microextraction was also applied to extract benzophenones from water [32, 33], however ionic liquids are not compatible with MS detection. The lowest LODs were achieved by on-line SPE–LC–MS/MS [24]. Most of the works focus the attention only on a few UV filters, being benzophenones the most investigated; also analytical methods for determination of various categories of water contaminants [25,26,28,29] consider a limited number of UV filters. In the present work, 16 UV filters (one compound prohibited and ten compounds allowed by the European Regulation, together with five of their degradation products) were selected between the most occurring in water samples. The analytes were determined by UHPLC–ESI-MS/MS analysis with ion polarity switching; the extraction was performed by SPE using graphitized carbon black (GCB) as adsorbent material. The method performances were comparable to those of other published methods, however the number of target analytes was larger. The major encountered problem in the method development was to reduce the background contamination of UV filters to obtain reliable recoveries also at low spiking levels. C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
2883
2 Materials and methods 2.1 Chemical and reagents Standards of 2,4-dihydroxybenzophenone (BP1), 2,2’,4,4’tetrahydroxybenzophenone (BP2), 4-hydroxybenzophenone (4-HB), 2-hydroxy-4-methoxybenzophenone (BP3), 4-hydroxy-2-methoxybenzophenone-5-sulfonic acid (BP4), 3-(4methylbenzylidene)bornan-2-one (4-MBC), 2,2-dihydroxy4-methoxybenzophenone (DHMB), 2-phenylbenzimidazole5-sulfonic acid (PBSA), 2-ethylhexyl 4-(dimethylamino) benzoate (OD-PABA), 2-ethylhexyl-4-methoxycinnamate (EHMC), diethylamino hydroxybenzoyl hexyl benzoate (DHHB), 1-(4-methoxyphenyl)-3-(4-tert-butylphenyl)pro pane-1,3-dione (BDM), 2-ethylhexyl 2-cyano-3,3-diphenyl2-propenoate (OC), ethyl 4-aminobenzoate (Et-PABA), p-aminobenzoic acid (PABA), 4,4 -dihydroxybenzophenone (4-DHB) were obtained from Sigma–Aldrich (Steinheim, Germany). The isotopically labeled compounds BP3-d5 and 4-MBC-d4 , used as internal standards (ISs), were obtained from CHEMICAL RESEARCH 2000 (Rome, Italy). All chemicals had a purity grade greater than 99%. The chemical structure of the analytes is shown in Fig. 1; names and acronyms are reported in Supporting Information Table S1. Stock standard solutions of each analyte and the ISs were prepared by dissolving the adequate amount of standard in methanol to obtain a concentration of 200 g/mL. From these individual solutions, a working mixture containing the 16 analytes was prepared and diluted at a concentration adequate for the experiments performed. This mixture was prepared weekly to avoid the possible degradation of the analytes. All the solutions were stored at −20⬚. To prevent photodegradation of compounds, stock standard solutions and samples were always covered with aluminium foil and stored in the dark. Formic acid, ultrapure water (LC–MS grade, resistivity 18.2 M⍀ cm), methanol (LC–MS grade), ammonium formate, ammonium acetate, dichloromethane, and hydrochloric acid were obtained from Sigma–Aldrich. Glass fiber filters (1 m) were obtained from Whatman International (Maidstone, UK).
2.2 Samples Drinking water was collected from a laboratory tap, and Na2 S2 O3 ·5H2 O at concentration of 0.5 g/L was added to the sample to eliminate hypochlorite. Surface water samples were collected in the area of Lazio (Italy) from January to May 2014, from the Lakes Bracciano (Rome), Martignano (Rome), and Posta Fibreno (Frosinone). The water samples were taken using amber glass bottles, previously washed with acetone, methanol, and ultrapure water. The lake water samples were filtered through 1 m glass fiber filters and stored at 4⬚C until analyzed (normally within 48 h). www.jss-journal.com
2884
A. L. Capriotti et al.
J. Sep. Sci. 2014, 37, 2882–2891
Figure 1. Chemical structures of the UV filters investigated (labeled with their acronym, see text and Supporting Information Table S1).
2.3 Extraction and clean-up SPE of samples was carried out with a Visiprep SPE manifold (Supelco, Bellefonte, PA). GCB of Carbograph-4 type (particle size range of 120–400 mesh; specific surface area 130 m2 /g) was purchased from LARA (Rome, Italy); Oasis HLB cartridges (packed with 60 and 200 mg) were obtained from Waters (Mildford, MA, USA). GCB cartridges were prepared by placing 250 mg of the adsorbent material inside 6 mL polypropylene tubes between two polyethylene frits. Both tubes and frits were purchased from Supelco.
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Water samples of 200 mL were extracted and cleanedup by a preconditioned GCB SPE cartridge. Before processing samples, the cartridges were attached to a vacuum manifold apparatus and washed sequentially with 5 mL of dichloromethane/methanol (80:20, v/v) containing 10 mmol/L ammonium formate, 3 mL of methanol, 10 mL of 10 mmol/L hydrochloric acid, and 5 mL of ultrapure water. Once conditioned, the 200 mL sample was passed through the cartridge at a flow rate of about 15–20 mL/min; then, 10 mL of ultrapure water were used to clean-up the cartridge. Most of the water remaining in the cartridge was expelled under vacuum for 1 min; the vacuum pump was disconnected,
www.jss-journal.com
J. Sep. Sci. 2014, 37, 2882–2891
and 200 L of methanol was passed by gravity through the sorbent to further decrease the residual water content without eluting the analytes. After that, the cartridge was turned upside down, and the analytes of interest were eluted in backflushing mode [34, 35] with 1 mL of methanol and 15 mL of a solution dichloromethane/methanol (80:20, v/v) containing 10 mmol/L ammonium formate. The eluate was collected into a 1.4 cm id round-bottomed glass vial, spiked with 100 L of 0.1 ng/L of IS solution, and solvents were removed to dryness in a water bath at 37⬚C under a gentle nitrogen stream. Finally, the residue was reconstituted with 500 L of water/methanol (70:30, v/v) containing 0.1% formic acid; 5 L of this final solution was analyzed by LC–ESI-MS/MS.
2.4 LC–MS/MS analysis LC was performed by using an Ultimate 3000 UHPLC system (Thermo Fisher Scientific, Bremen, Germany) consisting of a binary pump, equipped with a degasser, a thermostatted microwell-plate autosampler, and a thermostatted column oven. A 5 L aliquot of sample was injected on a Hypersil Gold C18 column (50 × 2.1 mm id, 1.9 m particle size) equipped with a Hypersil Gold C18 guard column (10 × 2.1 mm id, 3 m particle size), both from Thermo Fisher Scientific. The analytical column was maintained at 40⬚C. A short Hypersil Gold C18 column (50 × 2.1 mm id, 5 m particle size) was placed between the mixer and the sample valve. The mobile phase was water (A) and methanol (B), both containing 5 mmol/L ammonium acetate; flow rate was 200 L/min. A mixture of isopropanol/water (80:20, v/v) was used as washing solution of the autosampler injection system. The gradient profile was as follows (t in min): t0 , B = 10%; t1 , B = 10%; t8 , B = 80%; t9.5 , B = 99.5%; t14.5 , B = 99.5%, t15.5 , B = 10%; t20.5 , B = 10%. MS/MS spectra were acquired with a TSQ VantageTM triple-stage quadrupole mass spectrometer (Thermo Fisher Scientific) connected to the LC system via a heated ESI source. Mass calibrations and resolution adjustments on the resolving lens and quadrupoles were automatically performed using the manufacturer solution introduced via infusion pump at 5 L/min flow rate. The whole LC–MS system was managed by Xcalibur software (v.2.1, Thermo Fisher Scientific). The ionization of the analytes was performed both in positive and negative ionization mode by continuous positive/negative polarity switching. Eleven UV filters were analyzed under positive ESI mode, while four compounds (BP4, BP2, BP1, and DHMB) were analyzed under negative ESI mode; PBSA was analyzed in both polarity modes for identification purpose, however, for quantitation only the positive selected reaction monitoring (SRM) transition was considered. Spray voltages of 3.2 and –3.0 kV were used for positive and negative ionization mode, respectively. The vaporizer temperature was set at 250⬚C and the capillary temperature at 220⬚C. Sheath gas pressure, ion sweep gas pressure, and C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Liquid Chromatography
2885
auxiliary gas pressure were fixed at 50, 1, and 25 (arbitrary units), respectively. In order to optimize the tuning parameters for each compound, 1 ng/L standard solutions were infused at 5 L/min. The [M+H]+ or [M–H]− precursor ions were selected by the first quadrupole and fragmented in the collision cell. From the MS/MS full-scan spectra, for each compound two suitable transitions with the appropriate collision energy were selected for acquisition in SRM mode. The optimized parameters of the analyzed UV filters are reported in Table 1.
2.5 Evaluation of method performance 2.5.1 Recovery experiments Recovery experiments were carried out in both tap water and lake water samples at 20 and 50 ng/L spiking levels. Five replicates for each concentration level were performed. Recovery of each UV filter added to the considered water sample typology at any given concentration was assessed by measuring the peak area, calculating the peak area ratio relative to that of the corresponding IS, and comparing this result with that obtained for standard solution containing the same nominal analyte quantities and the IS. For naturally contaminated water samples, the contribution due to this contamination (measured in a blank sample) was subtracted from the value obtained in the recovery experiments, according to the following equation: R(%) = area analyte area analyte (spiked sample) − (blank sample) area IS area IS × 100 area analyte (solvent standard) area IS
2.5.2 Calibration graphs Two calibration graphs, named “standard” and “matrixmatched,” respectively, were constructed. Standard solutions were prepared by adding appropriate volumes of the working standard solution to the SPE elution solvents, i.e. 1 mL of methanol and 15 mL of a mixture dichloromethane/methanol (80:20, v/v) containing 10 mmol/L ammonium formate, and adding the same amount of the two ISs all times. After that, the samples were handled as previously described: solvents were removed under a nitrogen stream at 37⬚C, and the residue was reconstituted with water/methanol (70:30, v/v). Matrix-matched solutions were prepared by fortifying blank water sample extracts just before solvent removal with known and appropriate amounts of the working standard and the IS solutions, and following the rest of the procedure. The same amount of the two ISs was added all the times. Both standard and matrix matched solutions were prepared at five concentration levels, in the ranges of 5–100 pg www.jss-journal.com
2886
J. Sep. Sci. 2014, 37, 2882–2891
A. L. Capriotti et al.
Table 1. Main instrumental acquisition parameters: retention time, precursor ion (protonated or deprotonated molecule), product ions chosen for the selected reaction monitoring (SRM) acquisition mode with the optimized collision energy (CE) values
Compound
Retention time (min)
Precursor ion [M-H]– (m/z)
Precursor ion [M+H]+ (m/z)
Product ionsa) (m/z) (CE, V)
PBSA BP4 4-DHB PABA Et-PABA BP2 4-HB BP1 DHMB BP3 4-MBC DHHB OC OD-PABA BDM EHMC
6.0 7.6 7.7 8.0 8.0 8.1 9.2 9.7 10.1 10.8 11.9 12.1 12.3 12.4 12.4 12.5
273 307
275b)
194 (32); 80 (48) 227 (24); 211 (36) 93 (33); 121 (19) 92 (24); 120 (14) 120 (17); 138 (11) 135 (20); 91 (33) 105 (17); 121 (17) 91 (30); 135 (21) 93 (23); 123 (19) 105 (20); 151 (20) 105 (32); 212 (20) 296 (20); 149 (18) 232 (25); 250 (11) 151 (28); 166.1 (20) 135 (24); 161.1 (23) 133 (33); 161 (16)
215 138 166 245 199 213 243 229 255 398 362 278 311 291
a) For each analyte, the product ions are listed in order of SRM transition intensity. b) For quantification, only the positive mode SRM transition was considered, whereas the negative mode SRM transition was used for identification confirmation.
injected (corresponding to 2.5–50 ng/L in matrix); 5 L of all solutions were injected. The software XcaliburTM Quan Browser (Thermo Fisher Scientific) was used for quantitation. For each analyte, the combined ion current profile for all transitions was automatically extracted from the LC–SRM dataset, and the plot of the ratio between the peak area of the analyte and that of the IS versus injected amount or spiking level was measured. All samples were run in triplicate and results averaged. Unweighted regression lines for standard and matrix-matched calibration were calculated. 2.5.3 Matrix effect Signal suppression or enhancement on ESI-MS/MS response due to matrix coeluting components, i.e. matrix effect (ME) was assessed, for each analyte, by comparing the peak area obtained in a matrix-matched solution with the peak area obtained in the standard solution spiked at the same concentration level, according to the following equation: ME(%) =
area analyte (matrix matched standard) × 100 area analyte (solvent standard)
2.6 Blank contamination To reduce blank contamination, cautions suggested by other authors [19, 24] were followed: gloves were worn during all the sample preparation steps; only dedicated glassware, solvents, chemicals, and other supplies were used; all glassware was previously washed and dried in oven at 80⬚C and, as well as SPE tubing, further sequentially rinsed with acetone, C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
methanol, and LC–MS-grade water immediately prior of use. When possible, the use of plastic material was avoided. Polyethylene frits employed in SPE cartridges were previously extracted for 5 h sequentially with acetone and then hexane in a Soxhlet apparatus, whereas SPE tubes were only subject to washing and conditioning of the packed material. All the solvents utilized were carefully checked to be analyte-free, performing blank procedure analysis. Also guard columns required extensive washing and blank solvent injections before use to remove residual UV filter contamination. A blank sample was processed together with each batch of samples.
3 Results and discussion 3.1 Blank contamination A common problem in the determination of ubiquitous contaminants at trace levels in environmental samples is the background contamination, as also reported in the analysis of UV filters by several authors [6, 19, 24]. However, even following their suggestions to reduce the contamination, it resulted still significant and not reproducible in the blank samples. On the basis of a previous experience in the analysis of the ubiquitous perfluorinated substances [36], we hypothesized that contamination could originate from the instrumentation, very likely from the chromatographic system tubes and from the degasser membrane. Therefore, also in this work, a short C18 column was placed between the mixer and the sample loop for retaining the compounds coming from the chromatographic apparatus [37]. In this way, retention times www.jss-journal.com
J. Sep. Sci. 2014, 37, 2882–2891
Liquid Chromatography
2887
Figure 2. Extracted SRM mass chromatograms relative to DHHB, OC, OD-PABA, and BDM, obtained from the lowest point (5 pg injected) of the solvent standard calibration. For OC, OD-PABA, and BDM it is possible to distinguish two peaks, the one with the lower retention time deriving from the sample and the other one with the higher retention time deriving from the system.
of the compounds released from the system, namely OC, OD-PABA, and BDM, were higher than those of the injected analytes. In fact the short column, packed with the same stationary phase of the analytical column, delayed the elution time of compounds deriving from the LC system, thus separating them from the analytes contained in the sample. As an example, in Fig. 2 are shown the extracted SRM mass chromatograms relative to DHHB, OC, OD-PABA, and BDM: DHHB does not show any peak deriving from the system, whereas the other three analytes show two distinguishable peaks, the one with the lower retention time deriving from the sample and the other one (broader) with the higher retention time deriving from the system. C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
We found out that the residue contamination still observed in the blank samples derived from the polyethylene frits used in the SPE cartridges, which very likely contain some of the selected analytes, namely 4-DHB, BP3, 4-MBC, OC, OD-PABA, BDM, and EHMC. Therefore, to eliminate these interferences the frits were preliminarily extracted with a Soxhlet apparatus, as described in Section 2.6. Extraction with dichloromethane allowed to reduce contamination in the blank samples up to ten times for certain analytes, whereas for other analytes (e.g. 4-DHB and BP3) residue contamination was still high. Although blank subtraction was applied, it was not possible to obtain reliable recoveries at low spiking levels, both because contamination contribution www.jss-journal.com
2888
J. Sep. Sci. 2014, 37, 2882–2891
A. L. Capriotti et al.
was much higher than the added amount and because contamination was quantitatively variable from a sample to another. Thus, two sequential Soxhlet extractions were carried out to remove in a satisfying manner the blank contamination: acetone was able to remove more effectively the most polar compounds and only partially the apolar ones, whereas hexane was able to remove more effectively the most lipophilic UV filters. After the two Soxhlet extractions, residual contamination deriving from SPE frits in blank extract was negligible. Carryover in the chromatographic system was checked by running solvent blanks [38], however, it was negligible.
3.2 Sample extraction For SPE, two different adsorbent materials were tested and in different amount: Oasis HLB (60 and 200 mg) and Carbograph-4 (150 and 250 mg). Oasis HLB cartridges were employed following the methodology proposed by Rodil et al. [19], however the recoveries were >70% only for few analytes common to the two works (i.e. PBSA, BP3, 4-MBC), around 40–50% for other analytes common to the works (i.e. OC, BP4, OD-PABA, and BDM), but were
J. Sep. Sci. 2014, 37, 2882–2891
Research Article
Multiresidue determination of UV filters in water samples by solid-phase extraction and liquid chromatography with tandem mass spectrometry analysis UV filters, contained in sunscreens and other cosmetic products, as well as in some plastics and industrial products, are nowadays considered contaminants of emerging concern because their widespread and increasing use has lead to their presence in the environment. Furthermore, some UV filters are suspected to have endocrine disruption activity. In the present work, we developed an analytical method based on liquid chromatography with tandem mass spectrometry for the determination of UV filters in tap and lake waters. Sixteen UV filters were extracted from water samples by solid-phase extraction employing graphitized carbon black as adsorbent material. Handling 200 mL of water sample, satisfactory recoveries were obtained for almost all the analytes. The limits of detection and quantification of the method were comparable to those reported in other works, and ranged between 0.7–3.5 and 1.9–11.8 ng/L, respectively; however in our case the number of investigated compounds was larger. The major encountered problem in method development was to identify the background contamination sources and reduce their contribution. UV filters were not detected in tap water samples, whereas the analyses conducted on samples collected from three different lakes showed that the swimming areas are most subject to UV filter contamination. Keywords: Emerging contaminants / Lake waters / Liquid chromatography with mass spectrometry / Solid-phase extraction / UV filters DOI 10.1002/jssc.201400708
Additional supporting information may be found in the online version of this article at the publisher’s web-site
1 Introduction In the last years, there has been an increasing attention of both public institutions and researchers toward the “contaminants of emerging concern,” including not only new substances appeared recently, but also known substances for which the environmental contamination issues were not fully realized and concerns have been raised much more recently [1]. Among these contaminants there is the commercial category of UV filters, which include various and heterogeneous chemical classes. UV filters are present as ingredients in personal care products, such as sunscreen products, beauty creams, cosmetics, lipsticks, hair sprays, shampoos etc., to protect Correspondence: Dr. Chiara Cavaliere, Dipartimento di Chimica, Universita` di Roma “La Sapienza”, Piazzale Aldo Moro 5, 00185 Rome, Italy E-mail: [email protected] Fax: +39-06-490631
Abbreviations: GCB, graphitized carbon black; IS, internal standard; ME, matrix effect; SRM, selected reaction monitoring
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
skin and hair from the negative effects of sunlight. These compounds can enter the aquatic environment both directly from recreational activities (e.g. sunbathing and swimming in seas, lakes, and rivers) and industrial wastewater discharges and indirectly via wastewater treatment plants [2–4]. Furthermore, some UV-absorbing species (UV stabilizers) [5] are also present in several plastic products and packaging, paints, glasses, textiles, to prevent yellowing and degradation of polymers and pigments [6]. As a consequence of their widespread employment in various personal care and industrial products, UV filters are continuously released in the environment, and they have been found in lakes, rivers, wastewater, drinking water, soil, sludge, and fish [2]. From the toxicological point of view, besides possible allergenic effects, some UV filters have shown significant estrogenic and/or anti-androgenic activity and potential developmental toxicity in experimental animals [3, 7]. The levels of UV filters detected in the environment are not so far below the doses that cause toxic effects in animals [8,9]. UV filters can be both inorganic (e.g. titanium dioxide, zinc oxide) that work by reflecting and scattering UV light, and organic that work by absorbing UV light in the
www.jss-journal.com
Liquid Chromatography
J. Sep. Sci. 2014, 37, 2882–2891
wavelength range of 280−315 (UVB) and/or 315−400 nm (UVA) [8]. Organic UV filters include benzophenones, camphor derivatives, cinnamates, crylenes, benzimidazole derivatives, p-aminobenzoic acid, and derivatives, dibenzoyl methane derivatives, salicylates, and triazines, as well as their primary transformation products [6, 8]. Currently, the European Regulation on cosmetic products [10] permits the employment of 26 compounds, within certain limits, in cosmetic (sunscreen) products, and 25 of these are organic compounds [3]. Although several analytical methods based on GC–MS for UV filter determination in waters are published in the literature [11–18], methods based on LC–MS/MS [6, 19–29] are often preferred for the simultaneous analysis of many compounds and their environmental transformation products within a wide range of physicochemical properties [24]. Indeed, the UV filters selected in the published works vary considerably, from the very polar sulfonates to the less polar salycilates [22]. Therefore, for their analysis different API sources, i.e. the most widespread ESI [19, 21, 24, 27], APCI [22, 23, 28] and atmospheric pressure photoionization [20] were used in both polarity ion modes. ESI is the ionization source of choice when LC–MS/MS multiclass methods for the determination of personal care products and pharmaceuticals in aquatic environment are proposed [25, 26, 28, 29]. CE–ESI-MS has also been used for UV filter determination in river water [30]. In most cases, before LC–MS analysis UV filters were extracted from water by SPE using as sorbent material a styrenedivinylbenzene polymer [24] or a hydrophilic–lipophilic polymer such as Oasis HLB [19, 21, 25, 26, 28, 29] and Bond Elut Plexa [25]. More recently, microextraction techniques aimed at the minimization of solvent consumption [31] have been used, such as stir bar sorptive extraction [22, 23, 27] and nonporous membrane-assisted LLE [20]; the very promising ionic liquid-based dispersive liquid–liquid microextraction was also applied to extract benzophenones from water [32, 33], however ionic liquids are not compatible with MS detection. The lowest LODs were achieved by on-line SPE–LC–MS/MS [24]. Most of the works focus the attention only on a few UV filters, being benzophenones the most investigated; also analytical methods for determination of various categories of water contaminants [25,26,28,29] consider a limited number of UV filters. In the present work, 16 UV filters (one compound prohibited and ten compounds allowed by the European Regulation, together with five of their degradation products) were selected between the most occurring in water samples. The analytes were determined by UHPLC–ESI-MS/MS analysis with ion polarity switching; the extraction was performed by SPE using graphitized carbon black (GCB) as adsorbent material. The method performances were comparable to those of other published methods, however the number of target analytes was larger. The major encountered problem in the method development was to reduce the background contamination of UV filters to obtain reliable recoveries also at low spiking levels. C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
2883
2 Materials and methods 2.1 Chemical and reagents Standards of 2,4-dihydroxybenzophenone (BP1), 2,2’,4,4’tetrahydroxybenzophenone (BP2), 4-hydroxybenzophenone (4-HB), 2-hydroxy-4-methoxybenzophenone (BP3), 4-hydroxy-2-methoxybenzophenone-5-sulfonic acid (BP4), 3-(4methylbenzylidene)bornan-2-one (4-MBC), 2,2-dihydroxy4-methoxybenzophenone (DHMB), 2-phenylbenzimidazole5-sulfonic acid (PBSA), 2-ethylhexyl 4-(dimethylamino) benzoate (OD-PABA), 2-ethylhexyl-4-methoxycinnamate (EHMC), diethylamino hydroxybenzoyl hexyl benzoate (DHHB), 1-(4-methoxyphenyl)-3-(4-tert-butylphenyl)pro pane-1,3-dione (BDM), 2-ethylhexyl 2-cyano-3,3-diphenyl2-propenoate (OC), ethyl 4-aminobenzoate (Et-PABA), p-aminobenzoic acid (PABA), 4,4 -dihydroxybenzophenone (4-DHB) were obtained from Sigma–Aldrich (Steinheim, Germany). The isotopically labeled compounds BP3-d5 and 4-MBC-d4 , used as internal standards (ISs), were obtained from CHEMICAL RESEARCH 2000 (Rome, Italy). All chemicals had a purity grade greater than 99%. The chemical structure of the analytes is shown in Fig. 1; names and acronyms are reported in Supporting Information Table S1. Stock standard solutions of each analyte and the ISs were prepared by dissolving the adequate amount of standard in methanol to obtain a concentration of 200 g/mL. From these individual solutions, a working mixture containing the 16 analytes was prepared and diluted at a concentration adequate for the experiments performed. This mixture was prepared weekly to avoid the possible degradation of the analytes. All the solutions were stored at −20⬚. To prevent photodegradation of compounds, stock standard solutions and samples were always covered with aluminium foil and stored in the dark. Formic acid, ultrapure water (LC–MS grade, resistivity 18.2 M⍀ cm), methanol (LC–MS grade), ammonium formate, ammonium acetate, dichloromethane, and hydrochloric acid were obtained from Sigma–Aldrich. Glass fiber filters (1 m) were obtained from Whatman International (Maidstone, UK).
2.2 Samples Drinking water was collected from a laboratory tap, and Na2 S2 O3 ·5H2 O at concentration of 0.5 g/L was added to the sample to eliminate hypochlorite. Surface water samples were collected in the area of Lazio (Italy) from January to May 2014, from the Lakes Bracciano (Rome), Martignano (Rome), and Posta Fibreno (Frosinone). The water samples were taken using amber glass bottles, previously washed with acetone, methanol, and ultrapure water. The lake water samples were filtered through 1 m glass fiber filters and stored at 4⬚C until analyzed (normally within 48 h). www.jss-journal.com
2884
A. L. Capriotti et al.
J. Sep. Sci. 2014, 37, 2882–2891
Figure 1. Chemical structures of the UV filters investigated (labeled with their acronym, see text and Supporting Information Table S1).
2.3 Extraction and clean-up SPE of samples was carried out with a Visiprep SPE manifold (Supelco, Bellefonte, PA). GCB of Carbograph-4 type (particle size range of 120–400 mesh; specific surface area 130 m2 /g) was purchased from LARA (Rome, Italy); Oasis HLB cartridges (packed with 60 and 200 mg) were obtained from Waters (Mildford, MA, USA). GCB cartridges were prepared by placing 250 mg of the adsorbent material inside 6 mL polypropylene tubes between two polyethylene frits. Both tubes and frits were purchased from Supelco.
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Water samples of 200 mL were extracted and cleanedup by a preconditioned GCB SPE cartridge. Before processing samples, the cartridges were attached to a vacuum manifold apparatus and washed sequentially with 5 mL of dichloromethane/methanol (80:20, v/v) containing 10 mmol/L ammonium formate, 3 mL of methanol, 10 mL of 10 mmol/L hydrochloric acid, and 5 mL of ultrapure water. Once conditioned, the 200 mL sample was passed through the cartridge at a flow rate of about 15–20 mL/min; then, 10 mL of ultrapure water were used to clean-up the cartridge. Most of the water remaining in the cartridge was expelled under vacuum for 1 min; the vacuum pump was disconnected,
www.jss-journal.com
J. Sep. Sci. 2014, 37, 2882–2891
and 200 L of methanol was passed by gravity through the sorbent to further decrease the residual water content without eluting the analytes. After that, the cartridge was turned upside down, and the analytes of interest were eluted in backflushing mode [34, 35] with 1 mL of methanol and 15 mL of a solution dichloromethane/methanol (80:20, v/v) containing 10 mmol/L ammonium formate. The eluate was collected into a 1.4 cm id round-bottomed glass vial, spiked with 100 L of 0.1 ng/L of IS solution, and solvents were removed to dryness in a water bath at 37⬚C under a gentle nitrogen stream. Finally, the residue was reconstituted with 500 L of water/methanol (70:30, v/v) containing 0.1% formic acid; 5 L of this final solution was analyzed by LC–ESI-MS/MS.
2.4 LC–MS/MS analysis LC was performed by using an Ultimate 3000 UHPLC system (Thermo Fisher Scientific, Bremen, Germany) consisting of a binary pump, equipped with a degasser, a thermostatted microwell-plate autosampler, and a thermostatted column oven. A 5 L aliquot of sample was injected on a Hypersil Gold C18 column (50 × 2.1 mm id, 1.9 m particle size) equipped with a Hypersil Gold C18 guard column (10 × 2.1 mm id, 3 m particle size), both from Thermo Fisher Scientific. The analytical column was maintained at 40⬚C. A short Hypersil Gold C18 column (50 × 2.1 mm id, 5 m particle size) was placed between the mixer and the sample valve. The mobile phase was water (A) and methanol (B), both containing 5 mmol/L ammonium acetate; flow rate was 200 L/min. A mixture of isopropanol/water (80:20, v/v) was used as washing solution of the autosampler injection system. The gradient profile was as follows (t in min): t0 , B = 10%; t1 , B = 10%; t8 , B = 80%; t9.5 , B = 99.5%; t14.5 , B = 99.5%, t15.5 , B = 10%; t20.5 , B = 10%. MS/MS spectra were acquired with a TSQ VantageTM triple-stage quadrupole mass spectrometer (Thermo Fisher Scientific) connected to the LC system via a heated ESI source. Mass calibrations and resolution adjustments on the resolving lens and quadrupoles were automatically performed using the manufacturer solution introduced via infusion pump at 5 L/min flow rate. The whole LC–MS system was managed by Xcalibur software (v.2.1, Thermo Fisher Scientific). The ionization of the analytes was performed both in positive and negative ionization mode by continuous positive/negative polarity switching. Eleven UV filters were analyzed under positive ESI mode, while four compounds (BP4, BP2, BP1, and DHMB) were analyzed under negative ESI mode; PBSA was analyzed in both polarity modes for identification purpose, however, for quantitation only the positive selected reaction monitoring (SRM) transition was considered. Spray voltages of 3.2 and –3.0 kV were used for positive and negative ionization mode, respectively. The vaporizer temperature was set at 250⬚C and the capillary temperature at 220⬚C. Sheath gas pressure, ion sweep gas pressure, and C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Liquid Chromatography
2885
auxiliary gas pressure were fixed at 50, 1, and 25 (arbitrary units), respectively. In order to optimize the tuning parameters for each compound, 1 ng/L standard solutions were infused at 5 L/min. The [M+H]+ or [M–H]− precursor ions were selected by the first quadrupole and fragmented in the collision cell. From the MS/MS full-scan spectra, for each compound two suitable transitions with the appropriate collision energy were selected for acquisition in SRM mode. The optimized parameters of the analyzed UV filters are reported in Table 1.
2.5 Evaluation of method performance 2.5.1 Recovery experiments Recovery experiments were carried out in both tap water and lake water samples at 20 and 50 ng/L spiking levels. Five replicates for each concentration level were performed. Recovery of each UV filter added to the considered water sample typology at any given concentration was assessed by measuring the peak area, calculating the peak area ratio relative to that of the corresponding IS, and comparing this result with that obtained for standard solution containing the same nominal analyte quantities and the IS. For naturally contaminated water samples, the contribution due to this contamination (measured in a blank sample) was subtracted from the value obtained in the recovery experiments, according to the following equation: R(%) = area analyte area analyte (spiked sample) − (blank sample) area IS area IS × 100 area analyte (solvent standard) area IS
2.5.2 Calibration graphs Two calibration graphs, named “standard” and “matrixmatched,” respectively, were constructed. Standard solutions were prepared by adding appropriate volumes of the working standard solution to the SPE elution solvents, i.e. 1 mL of methanol and 15 mL of a mixture dichloromethane/methanol (80:20, v/v) containing 10 mmol/L ammonium formate, and adding the same amount of the two ISs all times. After that, the samples were handled as previously described: solvents were removed under a nitrogen stream at 37⬚C, and the residue was reconstituted with water/methanol (70:30, v/v). Matrix-matched solutions were prepared by fortifying blank water sample extracts just before solvent removal with known and appropriate amounts of the working standard and the IS solutions, and following the rest of the procedure. The same amount of the two ISs was added all the times. Both standard and matrix matched solutions were prepared at five concentration levels, in the ranges of 5–100 pg www.jss-journal.com
2886
J. Sep. Sci. 2014, 37, 2882–2891
A. L. Capriotti et al.
Table 1. Main instrumental acquisition parameters: retention time, precursor ion (protonated or deprotonated molecule), product ions chosen for the selected reaction monitoring (SRM) acquisition mode with the optimized collision energy (CE) values
Compound
Retention time (min)
Precursor ion [M-H]– (m/z)
Precursor ion [M+H]+ (m/z)
Product ionsa) (m/z) (CE, V)
PBSA BP4 4-DHB PABA Et-PABA BP2 4-HB BP1 DHMB BP3 4-MBC DHHB OC OD-PABA BDM EHMC
6.0 7.6 7.7 8.0 8.0 8.1 9.2 9.7 10.1 10.8 11.9 12.1 12.3 12.4 12.4 12.5
273 307
275b)
194 (32); 80 (48) 227 (24); 211 (36) 93 (33); 121 (19) 92 (24); 120 (14) 120 (17); 138 (11) 135 (20); 91 (33) 105 (17); 121 (17) 91 (30); 135 (21) 93 (23); 123 (19) 105 (20); 151 (20) 105 (32); 212 (20) 296 (20); 149 (18) 232 (25); 250 (11) 151 (28); 166.1 (20) 135 (24); 161.1 (23) 133 (33); 161 (16)
215 138 166 245 199 213 243 229 255 398 362 278 311 291
a) For each analyte, the product ions are listed in order of SRM transition intensity. b) For quantification, only the positive mode SRM transition was considered, whereas the negative mode SRM transition was used for identification confirmation.
injected (corresponding to 2.5–50 ng/L in matrix); 5 L of all solutions were injected. The software XcaliburTM Quan Browser (Thermo Fisher Scientific) was used for quantitation. For each analyte, the combined ion current profile for all transitions was automatically extracted from the LC–SRM dataset, and the plot of the ratio between the peak area of the analyte and that of the IS versus injected amount or spiking level was measured. All samples were run in triplicate and results averaged. Unweighted regression lines for standard and matrix-matched calibration were calculated. 2.5.3 Matrix effect Signal suppression or enhancement on ESI-MS/MS response due to matrix coeluting components, i.e. matrix effect (ME) was assessed, for each analyte, by comparing the peak area obtained in a matrix-matched solution with the peak area obtained in the standard solution spiked at the same concentration level, according to the following equation: ME(%) =
area analyte (matrix matched standard) × 100 area analyte (solvent standard)
2.6 Blank contamination To reduce blank contamination, cautions suggested by other authors [19, 24] were followed: gloves were worn during all the sample preparation steps; only dedicated glassware, solvents, chemicals, and other supplies were used; all glassware was previously washed and dried in oven at 80⬚C and, as well as SPE tubing, further sequentially rinsed with acetone, C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
methanol, and LC–MS-grade water immediately prior of use. When possible, the use of plastic material was avoided. Polyethylene frits employed in SPE cartridges were previously extracted for 5 h sequentially with acetone and then hexane in a Soxhlet apparatus, whereas SPE tubes were only subject to washing and conditioning of the packed material. All the solvents utilized were carefully checked to be analyte-free, performing blank procedure analysis. Also guard columns required extensive washing and blank solvent injections before use to remove residual UV filter contamination. A blank sample was processed together with each batch of samples.
3 Results and discussion 3.1 Blank contamination A common problem in the determination of ubiquitous contaminants at trace levels in environmental samples is the background contamination, as also reported in the analysis of UV filters by several authors [6, 19, 24]. However, even following their suggestions to reduce the contamination, it resulted still significant and not reproducible in the blank samples. On the basis of a previous experience in the analysis of the ubiquitous perfluorinated substances [36], we hypothesized that contamination could originate from the instrumentation, very likely from the chromatographic system tubes and from the degasser membrane. Therefore, also in this work, a short C18 column was placed between the mixer and the sample loop for retaining the compounds coming from the chromatographic apparatus [37]. In this way, retention times www.jss-journal.com
J. Sep. Sci. 2014, 37, 2882–2891
Liquid Chromatography
2887
Figure 2. Extracted SRM mass chromatograms relative to DHHB, OC, OD-PABA, and BDM, obtained from the lowest point (5 pg injected) of the solvent standard calibration. For OC, OD-PABA, and BDM it is possible to distinguish two peaks, the one with the lower retention time deriving from the sample and the other one with the higher retention time deriving from the system.
of the compounds released from the system, namely OC, OD-PABA, and BDM, were higher than those of the injected analytes. In fact the short column, packed with the same stationary phase of the analytical column, delayed the elution time of compounds deriving from the LC system, thus separating them from the analytes contained in the sample. As an example, in Fig. 2 are shown the extracted SRM mass chromatograms relative to DHHB, OC, OD-PABA, and BDM: DHHB does not show any peak deriving from the system, whereas the other three analytes show two distinguishable peaks, the one with the lower retention time deriving from the sample and the other one (broader) with the higher retention time deriving from the system. C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
We found out that the residue contamination still observed in the blank samples derived from the polyethylene frits used in the SPE cartridges, which very likely contain some of the selected analytes, namely 4-DHB, BP3, 4-MBC, OC, OD-PABA, BDM, and EHMC. Therefore, to eliminate these interferences the frits were preliminarily extracted with a Soxhlet apparatus, as described in Section 2.6. Extraction with dichloromethane allowed to reduce contamination in the blank samples up to ten times for certain analytes, whereas for other analytes (e.g. 4-DHB and BP3) residue contamination was still high. Although blank subtraction was applied, it was not possible to obtain reliable recoveries at low spiking levels, both because contamination contribution www.jss-journal.com
2888
J. Sep. Sci. 2014, 37, 2882–2891
A. L. Capriotti et al.
was much higher than the added amount and because contamination was quantitatively variable from a sample to another. Thus, two sequential Soxhlet extractions were carried out to remove in a satisfying manner the blank contamination: acetone was able to remove more effectively the most polar compounds and only partially the apolar ones, whereas hexane was able to remove more effectively the most lipophilic UV filters. After the two Soxhlet extractions, residual contamination deriving from SPE frits in blank extract was negligible. Carryover in the chromatographic system was checked by running solvent blanks [38], however, it was negligible.
3.2 Sample extraction For SPE, two different adsorbent materials were tested and in different amount: Oasis HLB (60 and 200 mg) and Carbograph-4 (150 and 250 mg). Oasis HLB cartridges were employed following the methodology proposed by Rodil et al. [19], however the recoveries were >70% only for few analytes common to the two works (i.e. PBSA, BP3, 4-MBC), around 40–50% for other analytes common to the works (i.e. OC, BP4, OD-PABA, and BDM), but were
