Liquid chromatography mass spectrometry

Journal of Chromatography A, 1453 (2016) 62–70

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Liquid chromatography mass spectrometry determination of perfluoroalkyl acids in environmental solid extracts after phospholipid removal and on-line turbulent flow chromatography purification M. Mazzoni, S. Polesello, M. Rusconi, S. Valsecchi ∗ IRSA-CNR, Water Research Institute, Via del Mulino 19, 20861 Brugherio, Italy

a r t i c l e

i n f o

Article history: Received 16 March 2016 Received in revised form 10 May 2016 Accepted 12 May 2016 Available online 13 May 2016 Keywords: Turbulent flow chromatography PFOA PFOS Perfluorinated compounds Phospholipid removal Environmental solid matrices

a b s t r a c t An on-line TFC (Turbulent Flow Chromatography) clean up procedures coupled with UHPLC–MS/MS (Ultra High Performance Liquid Chromatography Mass Spectrometry) multi-residue method was developed for the simultaneous determination of 8 perfluroalkyl carboxylic acids (PFCA, from 5 to 12 carbon atoms) and 3 perfluoroalkyl sulfonic acids (PFSA, from 4 to 8 carbon atoms) in environmental solid matrices. Fast sample preparation procedure was based on a sonication-assisted extraction with acetonitrile. Phospholipids in biological samples were fully removed by an off-line SPE purification before injection, using HybridSPE® Phospholipid Ultra cartridges. The development of the on-line TFC clean-up procedure regarded the choice of the stationary phase, the optimization of the mobile phase composition, flow rate and injected volume. The validation of the optimized method included the evaluation of matrix effects, accuracy and reproducibility. Signal suppression in the analysis of fortified extracts ranged from 1 to 60%, and this problem was overcome by using isotopic dilution. Since no certified reference materials were available for PFAS in these matrices, accuracy was evaluated by recoveries on spiked clam samples which were 98–133% for PFCAs and 40–60% for PFSAs. MLDs and MLQs ranged from 0.03 to 0.3 ng g−1 wet weight and from 0.1 to 0.9 ng g−1 wet weight respectively. Repeatability (intra-day precision) and reproducibility (inter-day precision) showed RSD from 3 to 13% and from 4 to 27% respectively. Validated on-line TFC/UHPLC–MS/MS method has been applied for the determination of perfluoroalkyl acids in different solid matrices (sediment, fish, bivalves and bird yolk). © 2016 Elsevier B.V. All rights reserved.

1. Introduction Per- and polyfluoroalkyl substances (PFASs) constitute a class of man-made substances that have been produced and used in numerous industrial and commercial applications, mainly as fluorinated surfactants and fluoropolymer processing aids, since the 1950s [1]. PFASs include thousands of chemicals but the most prominent families are perfluoroalkyl carboxylic acids (PFCA), which include perfluorooctanoic acid (PFOA), and perfluoroalkyl sulfonic acids (PFSA), which include perfluorooctane sulfonic acid (PFOS) [2]. Perfluoroalkyl acids (PFAA) have a fully fluorinated carbon chain of variable length and a terminal carboxylate or sulfonate group. Because of their widespread distribution, high persistence

∗ Corresponding author. E-mail address: [email protected] (S. Valsecchi). http://dx.doi.org/10.1016/j.chroma.2016.05.047 0021-9673/© 2016 Elsevier B.V. All rights reserved.

in the environment and bioaccumulation capability [3], the European Commission included PFOS in the list of priority hazardous substances which must be monitored in biota living in the EU water bodies, setting an Environmental Quality Standard (EQS) in fish of 9.1 ng g−1 wet weight (ww) (Directive 2013/39/EU). PFAA determination in biota by LC–MS is markedly affected by the influence of the biological matrix components on the analytes ionization in electrospray (ESI) [4], causing ion-suppression or –enhancement effects which were reviewed by Trufelli et al. [5]. Endogenous compounds and target analytes can compete for the available charge and space on the droplet surface causing an inhibition of ion ejection. Specific effects have been described for phospholipids [6] and fatty materials [7] which are thought to form a film on the droplet surface that inhibits ion evaporation. Despite these effects could be less evident for PFAA which themselves act as surfactants, ion suppression or enhancement in ESI were reported by several authors [8]. In particular, phospholipids

M. Mazzoni et al. / J. Chromatogr. A 1453 (2016) 62–70

are a source of uncertainty in the quantification of PFAS in LC–MS in biological samples, because they are polar lipids and their highly ionic and surfactant nature influences the ionization and the desolvation of the droplets in ESI sources [9,10]. The biological matrix influence can be overcome by improving the preparative steps, e.g. by introducing clean-up after the extraction without adding excessive manipulation of the sample. Turbulent Flow Chromatography (TFC) is a technique introduced in the late 1990s for the analysis of biological fluids that combines high-throughput, high reproducibility and reduced timeconsuming sample clean-up [11]. TFC is effective in excluding molecules larger than 8000–10,000 Da, such as particulate and proteins. The sample is injected at high flow rate, higher than 1 mL min−1 , into 0.5–1.0 mm internal diameter columns packed with large particles (30–60 ␮m) whose pores are functionalized with different chemistries. Under the turbulent flow conditions the improved mass transfer across the bulk mobile phase allows all molecules to improve their radial distribution, but around the stationary phase particles a laminar zone persists, where diffusional forces still dominate the mass transfer process [12]. The smaller molecules, which diffuse faster than larger molecules, have time to interact with stationary phase and bind to pores, while the larger molecules are quickly flushed to waste. Because the resolution capability of the TFC column is low, the analytical separation is carried out on a coupled conventional analytical column under laminar flow. In recent years TFC was applied to environmental waters and sediment samples as an automated clean-up step in the determination of emerging pollutants such as perfluoroalkyl compound [13], pharmaceuticals [14] and endocrine disrupters [15]. It has some advantages like minimum sample manipulation, low error introduction, very efficient extraction and good reliability, but exhibits also some limitations [16]. Matrix molecules are difficult to remove because physical mechanism of exclusion probably is not very efficient. In fact, in some cases target compounds and matrix molecules have the same dimensions and, consequently, same diffusion capability in the stationary phase and are not washed out by flow. Besides, if the analytes molecules have a wide spectrum of polarities and acidities, it is necessary to use different serially connected TFC columns to cover the whole range of analyte properties, but losing in selectivity [14]. The aim of our work is to optimize an on-line TFC purification and phospholipid removal procedure for the detection of PFAA in animal tissues and sediments. The optimized and validated method has then been applied to soft tissue of bivalve, egg, fish fillets and sediment samples collected during field research in Northern Italian and Swiss water bodies.

63

HybridSPE® Phospholipid Ultra cartridges (30 mg, 1 mL SPE Tubes) were obtained by Sigma Aldrich (St. Louis, Missouri, USA). Phosphatidylcholine solution (10 mg L−1 ) was prepared by dissolution in acetonitrile of purified phosphatidylcholine of soybean origin (Epikuron 200 purchased from Cargill Inc. Minneapolis, MN, USA). All reagents were analytical reagent grade. LC–MS grade Chromasolv methanol, LC–MS grade Chromasolv acetonitrile, ammonium acetate (99%), and concentrated formic acid were purchased from Sigma-Aldrich. Water (99.98%) was used as the collision gas (1.5 mTorr). The mass spectrometer operated at a resolution of 0.7 Da in negative multiple reaction monitoring (MRM) mode. Table SM2 lists the MS/MS transitions, tube lens offset, and collision energies applied for the different target analytes and isotope labelled standards. The Xcalibur 2.1 (Thermo Scientific) was used for instrument control, data acquisition and processing. 2.6. Confirmation and quantification Analytes were identified by comparing their retention times (RT) with those of the labelled or non-labelled standards (deviation ≤0.25%). For all the analytes, one precursor and two product ions were monitored. Calibration curves were prepared using mixed standard solutions in acetonitrile, which were acidified to pH 3 and spiked with SIL-IS by adding 50 ␮L of concentrated formic acid and 100 ␮L of the diluted SIL-IS solution (40 ␮g L−1 ) before injection. Quantification was performed by using the isotopic dilution method and calibration curves were acquired before each analytical run. Blank samples (acidified acetonitrile) were injected at the beginning and at the end of each analytical sequence and every ten samples. No carryover and contamination were detected. 3. Results and discussion 3.1. TFC optimization Optimization of the most important TFC conditions (choice of the phase, matrix extract modification, composition and flow of the loading solution, loop volume) was carried out by injecting the 20 ␮g L−1 multi-standard PFAA solution in acetonitrile, at an initial injection volume of 20 ␮L. Responses of PFAA standards after injection in four different TFC columns (Thermo Fluoro XL, Thermo Cyclone, Thermo Cyclone P

e Thermo Cyclone MAX), characterised by the same dimensions (0.5 × 50 mm), were compared (Fig. 1). Runs with Cyclone MAX and Cyclone columns showed no or small peak areas. The Cyclone P column showed the best chromatographic efficiency and symmetry but was less sensitive for the earlier eluting PFAAs (e.g. PFPeA, PFHxA). Use of the Fluoro XL column resulted in the best responses for all the PFAAs considered, but with a significant peak tailing. The definitive configurations was obtained by serially connecting Cyclone P and Fluoro XL which showed the best responses without giving up satisfactory chromatographic peak shapes (Fig. 1). Secondly, the different composition of the loading solution (0.1% formic acid or 2 mM CH3 CO2 NH4 + 5% MeOH) and loading flows (from 0.3 to 2.0 mL min−1 ) were tested in order to optimize the trapping the PFAA analytes in the TFC columns during the loading phase (Step B), when the matrix components are washed out. As predictable on the basis of the TFC principles, higher loading flow (2 mL min−1 ) improved the retention of all PFAAs while the retention of the coupled TFC columns was higher with formic acid as loading solution than with ammonium acetate solution (Fig. SM2). Thereby 0.1% formic acid at 2.0 mL min−1 was chosen as loading conditions. Applying the optimized conditions, retention on TFC columns ranged from 85% for the shortest alkyl-chain perfluorinated compounds (PFPeA and PFBS) to 100% for the longest alkyl-chain ones (PFUnDA and PFDoDA) (Fig. SM2). Finally the effect of the matrix extract acidification and the injected volume on analyte responses was investigated. Since all target analytes are acid compounds, their retention on the TFC phase was promoted by sample acidification at pH 3 by adding 50 ␮L of concentrated formic acid in 1 mL as shown in Fig. SM3. Different injection volumes ranging from 20 to 100 ␮L were compared and the response curve vs injection volume was linear only up to 50 ␮L which was selected as the injection volume. 3.2. Removal of phospholipids The presence of phospholipids in the biological tissue extracts (e.g. yolk or soft tissue of bivalves) was monitored with the same chromatographic conditions optimized for PFAA analysis, applying the mass spectrometric experiments developed by Xia and Jemal [20]. The technique allows a qualitative evaluation of different classes of phospholipids in a single injection by acquiring the following MS acquisition modes: (a) negative precursor ion scan of m/z 153; (b) positive neutral loss scan of the fragment m/z 141; (c) positive precursor ion scan of m/z 184. Fig. 2 shows the phospholipids profiles obtained injecting 20 ␮L of mussel extract (I) and yolk extract (II) without any on-line TFC clean-up step (by valve V0). In the same figure the MS/MS chromatogram of the multi-standard solution of PFAA is overlapped, but with a different response scale, to the phospholipids tracks. Phospholipids were detected, in high quantity, in the positive mode both in mussel (Fig. 2Ib, Ic) and yolk extracts (Fig. 2IIb, IIc), whereas phospholipids were detected in the negative mode only in mussel extract (Fig. 2Ia, IIa). The identification and quantification of phospholipids is beyond the scope of this study but it is evident that phospholipids extracted by acetonitrile co-elute with the more retained PFAAs (Fig. 2I and II). The removal of phospholipids was evaluated preparing a phosphatidylcholine standard at 10 mg L−1 with a commercial soy lecithin reagent (Epikuron 200). This product consists of phosphatidylcholine and a small amount of accompanying phospholipids and thus produces a large peak during positive precursor ion scan of m/z 184 mass spectrometric experiment but only negligible peaks in the other two MS acquisition modes (Fig. 2III). The phosphatidylcholine standard was also injected in the online TFC system: a reduction of phosphatidylcholine was detected but about 10% of the initial amount was not removed by TFC

M. Mazzoni et al. / J. Chromatogr. A 1453 (2016) 62–70

65

Fig. 1. Chromatograms of a not acidified PFAA standard solution (20 ␮g L−1 ) with the four TFC columns tested (injection volume of 20 ␮L and other chromatographic conditions as in Section 2).

clean-up (Fig. 2III-TFC). Therefore another method, based on an off-line SPE clean-up step with HybridSPE® Phospholipid Ultra cartridges before the acidification, was tested for the complete removal of the phospholipids. These cartridges have a packed bed of zirconia-coated silica particles, designed for the gross level removal of endogenous phospholipids from biological plasma and serum prior to LC–MS analysis [18]. Injection of the phosphatidylcholine standard purified on HybridSPE® cartridge showed the complete removal of the phosphatidylcholine peak (Fig. 2III-Hybrid). Complete removal of other phospholipids by HybridSPE® in biological tissue extracts (e.g. yolk and mussel) was also verified (data not showed). The assessment of any loss of analytes during the HybridSPE® clean-up procedure was carried out calculating the PFAA recoveries of a multi-standard solution of PFAA at 20 ␮g L−1 containing phosphatidylcholine at 10 mg L−1 . Purification by HybridSPE® cartridge did not affect the PFAA recoveries which ranged from 96 to 99% for PFSAs to 100–125% for PFCAs. In order to obtain an efficient matrix removal and preserve the analytical column and MS analyser, we decided to filter all biological extracts on the HybridSPE® Phospholipid Ultra cartridges before the injection in the TFC system. 3.3. Effectiveness of TFC purification The acetonitrile extraction is not selective and many biomolecules of the biological matrix or organic matter of sediment samples are co-extracted with the analytes and the resulting extraction solutions were turbid and colored. In order to assess the effectiveness of the on-line TFC purification, the UV spectra (190–1100 nm) of the fractions, wasted by the TFC during the loading STEP B, were acquired. The UV-spectra could not identify the matrix composition but fingerprinted the wasted fraction of every extracted matrices (Fig. SM4). The TFC wastes of sediment, clam and mussel extracts showed generally a higher absorbance than the acetonitrile standard solution (Fig. SM4),

proving that part of the organic matrix in the extracts is removed by TFC purification. Spectrum of mussel extract showed two absorption peaks at 630 and 670 nm that match the peak absorption wavelengths of chlorophyll (Fig. SM4), linked to the presence of phytoplankton which is the main mussel food. Furthermore, in all cases the TFC waste is colored as the initial extracts, suggesting that also some pigments are removed by TFC clean-up. 3.4. Method validation The optimized method was subjected to the validation procedure. 3.4.1. Matrix effect The matrix effect (ME) on ionization of target compounds, caused by the presence of coeluting matrix components, was evaluated by comparing clam extracts spiked at two concentrations (2 and 15 ␮g L−1 ), with a standard solution in acetonitrile, spiked at the same concentrations. Percentages below 100% mean ionization suppression, while percentages above 100% mean ionization enhancement. Generally, we observed reduction of the signal response of native compounds (Table 1). Native PFOA, PFDoDA, PFHxS and PFOS showed low suppression ranging between 1 and 25%; on the contrary the ionization suppression was higher (between 34 and 60%) for PFPeA, PFHxA, PFHpA, PFNA, PFDA, PFUnDA and PFBS. Therefore we tested the isotope dilution with SIL-IS to correct matrix effects on ionization as well as compensate for injection, extraction and instrumental parameter variabilities. The SIL-IS solution at 4 ␮g L−1 was added both to the clam extracts and to the standard solutions. The matrix effects on SIL-IS were similar to those of the native target compounds (Table 1), thereby the isotope dilution can be used to compensate for ionization suppression of the analytes. In the case of PFOS the difference was probably due to the lack of precision associated with low concentrations. For those target compounds for which SIL-IS were not available (i.e. PFPeA,

66

M. Mazzoni et al. / J. Chromatogr. A 1453 (2016) 62–70

Fig. 2. UHPLC–MS/MS chromatograms obtained, unless otherwise indicated, by direct injection, by valve V0 without any on-line TFC clean-up step, of 20 ␮L of standard or extract using the gradient elution described in Section 2. Mass spectrometric acquisition modes for phospholipids: (a) negative precursor ion scan of m/z 153; (b) positive neutral loss scan of the fragment m/z 141; (c) positive precursor ion scan of m/z 184. (I) extract of 10 g of soft tissue of mussel; (II) extract of 1 g of gull’s yolk. The chromatogram of a multi-standard solution of PFAA at 20 ␮g L−1 is overlapped. (III) phosphatidylcholine solution (10 mg L−1 ); (III-TFC): positive precursor ion scan of m/z 184 of on-line TFC injection of phosphatidylcholine solution (10 mg L−1 ); (III-Hybrid): positive precursor ion scan of m/z 184 of direct injection of phosphatidylcholine solution (10 mg L−1 ) purified on HybridSPE® Phospholipid Ultra cartridge.

M. Mazzoni et al. / J. Chromatogr. A 1453 (2016) 62–70 Table 1 Matrix effects (ME) in clam samples. ME (%) = (Peak Areaclam /Peak AreaACN ) × 100; (N = 3). Analyte

PFPeA PFHxA PFHpA PFOA PFNA PFDA PFUnDA PFDoDA PFBS PFHxS PFOS

Matrix Effect (%)

Table 2 Validation parameters: linearity and sensitivity. Analytes

Matrix Effect (%)

(␮g L

SIL-IS 4 ␮g L−1

Native 15 ␮g L−1

SIL-IS 4 ␮g L−1

n.d. 40 43 75 66 57 66 99 n.d. 85 n.d.

n.a. 42 n.a. 109 51 52 54 87 n.a. 81 135

41 40 46 98 57 57 55 83 49 81 99

n.a. 46 n.a. 94 55 57 54 84 n.a. 93 149

PFPeAa PFHxA PFHpAa PFOA PFNA PFDA PFUnDA PFDoDA PFBSa PFHxS PFOS a

Sensitivity (ng g−1 ww)

Linearity −1

Native 2 ␮g L−1

67

1–100 1–100 1–100 1–100 1–100 1–100 0.5–50 0.5–50 1–100 1–100 1–100

)

2

R

ILOD

ILOQ

MLOD

MLOQ

0.997 0.994 0.998 0.997 0.992 0.995 0.997 0.998 0.999 0.989 0.987

0.2 0.1 0.04 0.03 0.04 0.03 0.02 0.03 0.2 0.03 0.04

0.7 0.2 0.1 0.1 0.1 0.07 0.04 0.08 0.4 0.07 0.1

0.3 0.06 0.1 0.1 0.07 0.07 0.04 0.03 0.3 0.1 0.2

0.9 0.2 0.4 0.4 0.2 0.2 0.1 0.1 0.8 0.2 0.5

Values corrected by SIL-IS PFHxA.

n.d.: not determined because response 0.98 for all target compounds over two orders of magnitude (Table 2). Limits of detection (LOD) and limits of quantification (LOQ) were estimated, according to the ISO 6107-2: 2006 standard, as threefold and tenfold, respectively, the standard deviation of the lowest multicomponent standard (ILOD and ILOQ) and of an extract of clam (MLD and MLQ) fortified at 1 ␮g L−1 . ILODs and ILOQs ranged from 0.02 to 0.2 and from 0.04 to 0.7 ng g−1 ww while MLDs and MLQs ranged from 0.03 to 0.3 from 0.1 to 0.9 ng g−1 ww, respectively (Table 2). 3.4.3. Accuracy Method precision was evaluated as the relative standard deviation of four replicate injections of multicomponent standard solutions or fortified extracts of clam (Table 3) at low and high concentrations. Generally repeatabilities (intraday precision) and reproducibilities (inter day precision, on 3 nonconsecutive days) were better for PFCA than for PFSA. Standard solutions and extracts of clam did not show any significant difference in precision. At low concentration repeatabilities and reproducibilities were similar (RSD: 1–40%). At high concentration repeatability ranged from 3 to 8% for PFCA and from 5 to 13% for PFSA while reproducibility ranged from 4 to 17% and from 12 to 27% for PFCA and PFSA respectively. Since no biota certified reference material (CRM) was available for PFAA, the trueness was evaluated by recoveries in biological samples (10 g ww of soft clam tissue) fortified at two concentrations (0.4 and 3 ng g−1 ww) compared to the procedural blank fortified at 1.5 ng g−1 ww level (Table 3). Recoveries in clam were similar to those calculated for the blank sample. Recoveries of PFCA were satisfactory (98–133%), while those of PFSA were lower (40–60%), because the method is ten-fold less sensitive for PFSA respect to PFCA. The analysis of a candidate reference material, IRMM-427, a fish fillet certified for the mass fraction of perfluoroalkyl substances [21], allowed a further assessment of the accuracy of our method (Table 4). IRMM-427 was also used in the IMEP-42 study on the

determination of PFASs in fish which assessed the performance of control laboratories by calculating Z-scores according to ISO/IEC 17043 [22]. Z-scores of our method were very satisfactory because they were