Article pubs.acs.org/ac
Alleviating the Reference Standard Dilemma Using a Systematic Exact Mass Suspect Screening Approach with Liquid Chromatography-High Resolution Mass Spectrometry Christoph Moschet,§,∥ Alessandro Piazzoli,§,∥ Heinz Singer,*,§ and Juliane Hollender§,∥ Eawag, Swiss Federal Institute of Aquatic Science and Technology, Ü berlandstrasse 133, 8600 Dübendorf, Switzerland Institute of Biogeochemistry and Pollutant Dynamics, ETH Zürich, 8092 Zürich, Switzerland
§
∥
S Supporting Information *
ABSTRACT: In this study, the efficiency of a suspect screening strategy using liquid chromatography-high resolution mass spectrometry (LC-HRMS) without the prior purchase of reference standards was systematically optimized and evaluated for assessing the exposure of rarely investigated pesticides and their transformation products (TPs) in 76 surface water samples. Water-soluble and readily ionizable (electrospray ionization) substances, 185 in total, were selected from a list of all insecticides and fungicides registered in Switzerland and their major TPs. Initially, a solid phase extraction-LC-HRMS method was established using 45 known, persistent, and high sales volume pesticides. Seventy percent of these target substances had limit of quantitation (LOQ) < 5 ng L−1. This compound set was then used to develop and optimize a HRMS suspect screening method using only the exact mass as a priori information. Thresholds for blank subtraction, peak area, peak shape, signal-to-noise, and isotopic pattern were applied to automatically filter the initially picked peaks. The success rate was 70%; false negatives mainly resulted from low intense peaks. The optimized approach was applied to the remaining 140 substances. Nineteen additional substances were detected in environmental samples, two TPs for the first time in the environment. Sixteen substances were confirmed with reference standards purchased subsequently, while three TP standards could be obtained from industry or other laboratories. Overall, this screening approach was fast and very successful and can easily be expanded to other micropollutant classes for which reference standards are not readily accessible such as TPs of household chemicals.
T
quantitative target analysis approach (using reference standards), a qualitative suspect screening (exact mass as a priori information) approach or nontarget screening (no previous information of the chemical available) can be pursued with high resolution mass spectrometers. In many cases, an automated qualitative target screening is performed using a customized database (with information of the exact mass, retention time, and fragments of tandem mass spectra) containing up to several hundred substances.10−15 Such databases are powerful for practical screening of compounds where reference standards are commercially available,16 mainly because of the knowledge of the retention time under the specific chromatographic conditions. However, the construction of such databases is a laborious and costly task, as all substances must be purchased and measured individually. New reference standards need to be purchased and integrated into the database as soon as the focus of the study changes. In addition, substances where reference standards are not easily accessible (e.g., many TPs) cannot be
he use of pesticides in agricultural practices may lead to release of both parent compounds and transformation products (TPs) into surface waters,1,2 which can then threaten the health of aquatic organisms even at low concentrations.3,4 Therefore, appropriate analytical tools are necessary to detect residues of current-use pesticides and TPs in the low ng L−1 range. Multiresidue methods using solid phase extraction (SPE) and LC-MS/MS5−7 can provide fast and powerful target analysis for more than 100 chemicals at sensitivities in the low ng L−1 range. However, as the exposure of surface waters to pesticides is heavily dependent on local conditions (e.g., land use, application procedures), a single-run measurement of all pesticides potentially occurring in a sample is desirable to provide a holistic exposure estimate. The use of high resolution mass spectrometers (HRMS) such as orbitrap and time-of-flight (TOF) instruments provide both high mass accuracy and resolution in full scan mode, enabling accurate mass screening of a theoretically unlimited number of polar organic pollutants.8 High-resolution mass spectrometers are not yet widespread in use due to higher costs but seem to be one of the future MS trends.5 Krauss et al.9 introduced the three screening methods target analysis, suspect screening, and nontarget screening. Besides the classical © 2013 American Chemical Society
Received: July 15, 2013 Accepted: October 4, 2013 Published: October 4, 2013 10312
dx.doi.org/10.1021/ac4021598 | Anal. Chem. 2013, 85, 10312−10320
Analytical Chemistry
Article
from the analysis (see step 1 in Figure 1). Log Kow values were taken from the Footprint database27 or calculated using Jchem
integrated in such a database. On the other hand, comprehensive nontarget screening in its current state is very time-consuming because many picked peaks at all masses and retention times (usually several thousands) must be evaluated irrespective of the focus in a study. Suspect screening without reference standards, using only the information of the chemical structure a priori, is a very promising approach. The number of substances that can be screened qualitatively is theoretically unlimited but can be limited intentionally depending on the focus of the study and on the substances expected to be present in the sample. As there are currently approximately 70 million chemicals registered in the Chemical Abstracts Service (CAS),17 a screening based only on the “exact mass” could result in an unmanageable number of results with a large number of false positives. However, as opposed to nontarget screening, expert knowledge regarding substances likely to occur in the surface water is necessary and should be used as a prefilter for the analysis. In addition, compound-specific information such as molecular formula and structure is available for suspects.9 Therefore, it is hypothesized that this concept is feasible for substance groups such as current-use pesticides and household chemicals, which usually contain polar functional groups (often electrospray-ionizable heteroatoms) and distinguishable isotopic patterns (e.g., Cl- or Br-containing molecular formula). If such a screening strategy is feasible, it alleviates the dilemma of requiring reference standards a priori and opens the door for the fast detection of compound classes for which reference standards are not easily accessible (e.g., TPs). Segura et al.18 validated a suspect screening method with drinking water and surface water samples spiked with 17 substances. In a few other publications,19−21 suspects (pharmaceuticals, pesticides, or industrial products) were screened in this manner for a limited set of environmental samples. However, these approaches involved relatively laborious manual evaluation. A comprehensive strategy for an automated suspect screening for a whole substance class, systematically evaluated on a large number of environmental samples, is critically needed. To test this hypothesis and to bring the proof of concept, a semiautomated, suspect screening approach based on liquid chromatography-high resolution mass spectrometry (LCHRMS) was developed and validated for 218 insecticides, fungicides, and their major TPs. A list containing the exact mass of each suspect compound was thereby the only information required a priori. This brute force method, which includes a large number of known TPs that were not yet investigated in surface waters, was tested for the first time in composite surface water samples under realistic concentrations (ng L−1 range).
Figure 1. Scheme of the established suspect screening method.
for Excel (Version 5.11.5.906) for the 68 substances where no data were available (SI-1, Supporting Information). Then, the ionization efficiency of all remaining substances was estimated. On the basis of expert knowledge for over 400 diverse compounds analyzed with ESI, only substances containing the following functional groups were set as readily ionizable in the spray: a nitrogen atom (except cyano or nitrate functional groups); an oxygen or sulfur atom in a carboxylic/sulfonic acid or if the neighbors are very electron donating or withdrawing (e.g., phenolic group with halogen substitution on the ring); a phosphorus atom if in the form of an organophosphate. Substances that did not meet these criteria were excluded from the suspect screening (see Figure 1). From the “ionizable” compounds, 45 substances (SI-1, Supporting Information) were selected in order to establish a broad SPE-LC-HRMS method. This list included 19 insecticides (5 carbamates, 4 organophosphates, 4 neonicotinoids, 2 diacyl hydracines, 2 pyridines, 2 others), 18 fungicides (7 azole fungicides, 2 anilino pyrimidines, 2 carbamates, 2 morpholines, 5 others), and 8 TPs (7 insecticide TPs, 1 fungicide TP). The selection was based on the theoretical
■
MATERIAL AND METHODS Substance Selection. 218 substances (55 insecticides, 72 fungicides, 14 various pesticides (e.g., acaricides), and 77 TPs) were chosen for the suspect screening (see SI-1, Supporting Information). This includes the complete list of synthetic organic insecticides and fungicides registered in Switzerland between 2007 and 2013.22 Major TPs for the most commonly used insecticides and fungicides in Switzerland were also selected, as well as minor TPs for neonicotinoid insecticides, as they are an important insecticide class in terms of use quantity and known impacts on aquatic organisms23,24 and bees.25,26 To begin, all substances were evaluated for their relevance to occur in water. All substances with log Kow > 5 were excluded 10313
dx.doi.org/10.1021/ac4021598 | Anal. Chem. 2013, 85, 10312−10320
Analytical Chemistry
Article
resolution: 17,500; microscan: 1; underfill ratio: 0.1%; isolation window: 1 m/z; dynamic exclusion: 8 s. The mass accuracy was determined to be 5 × 105 (negative mode), (ii) peak score > 0.5, and (iii) signal-to-noise (S/N) > 100, isotope score > 50. ExactFinder V 2.0 fits peak shape (peak symmetry) and calculates a resulting peak score. For the calculation of the S/N, the signal height is baseline corrected whereas the noise height is the peak-to-peak height of the baseline noise. The isotope score is a vector sum approach taking into account the deviation from the predicted intensity and exact mass of each expected isotope (ExactFinder V 2.0). The remaining peaks were checked manually for appropriate S/ N and isotope pattern. The positive findings from the suspect screening obtained for the 45 substances in the 76 composite water samples were compared to the manually evaluated target screening results. Success rates and number of false negatives and false positives were estimated. Application of Suspect Screening for Real Suspects. The developed and validated suspect screening procedure was
likeliness to be found in the water (sales amounts in Switzerland (internal document), physicochemical properties,27 and predicted fate behavior27). In addition, the selected substances covered a wide range of physicochemical properties and structures (log Kow, pKa, functional groups; see SI-1, Supporting Information). Field Study. The method was tested on a large field study, where 5 agriculturally influenced streams were sampled over an entire application season (March−July 2012) (see SI-2, Supporting Information). The catchments were selected on the basis of different land use characteristics, so that all crops that are either present in high densities in Switzerland (cereals, corn, sugar beet) or have an intense pesticide application (oilseed rape, potatoes, vegetables, apple orchards, vineyards) are present at high density in at least one catchment. Twentyeight biweekly (March, April, and July) and 40 weekly (May and June) time-proportional composite samples and 8 grab samples (opportunistic samples during high-flow conditions) were collected. The composite samples were obtained by automatic sampling devices (Isco sampler) with a 60 min subsampling interval. Subsamples were cooled during sampling; weekly or biweekly samples were stored at −20 °C until preparation. Sample Preparation. In order to enrich the analytes from all water samples, an offline solid phase extraction (SPE) method as described by Kern et al.20 was used. Briefly, the pH was set to 6.5−6.7 (using formic acid or ammonia solutions); the sample was filtered (GF/F; 0.7 μm, Whatman, Amesham Place, UK), and 1 L of surface water was measured into a prerinsed glass bottle. Then, 200 ng of internal standard mix was added (see SI-3, Supporting Information). The samples were passed over a multilayered cartridge containing Oasis HLB (Waters, Massachusetts, USA), Strata XAW, Strata XCW (both Phenomenex, Munich, Germany), and Isolute ENV+ (Biotage, Uppsala, Sweden), in order to enrich neutral, cationic, and anionic species of a broad range of Kow values. The elution of the analytes was achieved with 6 mL of ethyl acetate/ methanol 50:50 v/v with 0.5% ammonia and 3 mL of ethyl acetate/methanol 50:50 v/v with 1.7% formic acid. The sample extracts were evaporated to 100 μL using a gentle nitrogen stream and reconstituted to 1 mL using nanopure water to give a final water/methanol ratio of 90:10 in the aliquot. The aliquots were stored at 4 °C in a glass vial until analysis. LC-HRMS/MS. Liquid Chromatography. Twenty μL of the aliquot was injected and separated on a reversed phase column (XBridge C18 column, 3.5 μm, 2.1 × 50 mm; Waters, Ireland). The following gradient of water (solvent A) and methanol (solvent B) both acidified with 0.1% formic acid (Merck KGaA, Darmstadt Germany) at a flow rate of 200 μL/min (Rheos 2200 pump, Flux Instruments, Switzerland) was used: 0 min 10% B, 0−4 min linear gradient to 50% B, 4−17 min linear gradient to 95% B, 17−25 min kept at 95% B, 25−25.1 min switch to 10% B, and 25.1−29 min kept at 10% B. High Resolution Mass Spectrometry. Mass spectrometric analysis was performed on a high resolution mass spectrometer (QExactive, Thermo Fisher Scientific Corporation) with ESI. Each sample was run once each in positive and negative mode, with the following parameters. Ion source: HESI-II; spray voltage: 4000 V (+)/3000 V (−); sheath gas flow: 40 AU; capillary temperature: 350 °C; heater temperature: 40 °C. Orbitrap f ull scan. Mass range: 100−1,000 m/z; mass resolution: 140,000; AGC target: 500,000; maximal injection time: 250 ms. Data dependent MS/MS (Top 5). Mass 10314
dx.doi.org/10.1021/ac4021598 | Anal. Chem. 2013, 85, 10312−10320
Analytical Chemistry
Article
Table 1. Quality Parameters for the Target Analytesa
then applied to the remaining 140 substances. As an additional step, for further and/or unequivocal confirmation, the samples with positive findings were remeasured with a targeted MS/MS approach (HCD collision energies 15, 30, 45). If a library spectrum existed (Thermo Fisher Library Manager (Version 2.0) or MassBank28), the MS/MS spectra of the suspects in the environmental samples were compared with those in the library. When the spectrum matched, a reference standard was purchased to unequivocally confirm the identity with MS/MS and retention time. If the spectrum did not match, no reference standard was purchased and the substance was considered a false positive. If no library spectra existed but a reference standard was commercially available, the reference standard was purchased in order to unequivocally confirm or exclude the identity of the substance. If neither a library spectrum nor a reference standard was commercially available, the MS/MS fragments were checked for their plausibility using the fragmentation prediction software MassFrontier (Version 6.0), MetFrag29 (settings: search ppm: 5, PubChem and ChemSpider search), or the Peak Search tool from MassBank28 (search settings: relative intensity: 10, tolerance: 0.01, Instrument Type: ESI). For substances with a plausible MS/MS spectrum, a request was sent to the manufacturer or to other analytical laboratories.
parameter SPE recovery
ion suppression
spike recovery
precision (RSD %) LOQ (ng/L)
a
■
criteria
fungicides (18)
insecticides (19)
TPs (8)
75−125% 125% 25% 0.3−1 1−5 5−20 >20
15 3 0 0 1 17 0 16 2 0 15 3 0 6 10 0 2
19 0 0 2 2 10 5 16 2 1 13 5 1 3 10 2 4
7 1 0 0 2 5 1 7 1 0 5 3 0 1 2 3 2
TPs: transformation products; RSD: relative standard deviation.
between 75% and 125%, independent of the substance class, showing the broad enrichment efficiency of the multilayered cartridge. For only four substances (fenpropidin, propamocarb, spiroxamine, and 2-isopropyl-6-methyl-4-pyrimidinol), the recovery was between 65% and 75%. Ion suppression in river water was less than 50% for 87% of all substances and was between 50% and 75% for only six substances (aldicarb, clothianidin, flonicamide, methoxyfenozide, thiacloprid, 3,5,6trichloro-2-pyridinol). Good spike recoveries between 75% and 125% were achieved for 87% of all compounds, showing that the method is also able to accurately quantify the substance concentrations. For cyproconazole, myclobutanil, methoxyfenozide, tebufenozide, and fipronil-sulfone, spike recovery was 35−57%, and for flonicamide, it was 150%. A good agreement in triplicate concentration (precision) of