Non-Targeted Analyses for Pesticides Using Deconvolution, Accurate Masses, and Databases – Screening and Confirmation

The book offers a professional look on the recent achievements and emerging trends in pesticides analysis, including pesticides identification and characterization. The 20 chapters are organized in three sections. The first book section addresses issues associated with pesticides classification, pesticides properties and environmental risks, and pesticides safe management, and provides a general overview on the advanced chromatographic and sensors- and biosensors-based methods for pesticides determination. The second book section is specially devoted to the chromatographic pesticides quantification, including sample preparation. The basic principles of the modern extraction techniques, such as: accelerated solvent extraction, supercritical fluid extraction, microwave assisted extraction, solid phase extraction, solid phase microextraction, matrix solid phase dispersion extraction, cloud point extraction, and QuEChERS are comprehensively described and critically evaluated. The third book section describes some alternative analytical approaches to the conventional methods of pesticides determination. These include voltammetric techniques making use of electrochemical sensors and biosensors, and solid-phase spectrometry combined with flow-injection analysis applying flow-based optosensors.


Introduction
Technical innovations in crop protection have been a key component in the globalization of food production and distribution over the past several decades. The ease of access to foodstuffs from distant growing regions has depended, to a large extent, on the new pesticides that combat the historic foes of food sufficiency: fungi, insects, and weeds. Yet the same public that has come to expect and demand the ready availability of food products also has high expectations of regulators to ensure the safety of their food supply. Regulators in the United States, Europe, and Asia have, therefore, had to grapple with the task of ensuring that exposure levels to agricultural chemicals in the food supply remain within strictly determined parameters. In each region, regulatory agencies have, during the past decade, produced pesticide regulations that are increasingly stringent both in terms of the number of pesticides tracked and the allowable tolerances of those pesticides in the food supply. Because over a thousand pesticides are used globally to protect crops and improve food production, regulatory stringency translates into new challenges for the refinement of residue data generation and analysis. Because pesticide contamination of foodstuffs and environmental matrices could provide significant risk to consumers, vigilant monitoring is required; but the number of targeted and non-targeted pesticides that must be monitored in a given global region has risen significantly. Therefore, simple, rapid methods for screening hundreds of pesticides at trace levels in various matrices must be established. The widely divergent chemical properties (polarity, thermal stability, etc.) of these non-targeted pesticides call for the application of different analytical methods. Both GC and LC approaches are available, with each using a different technique for data analysis.

Gas phase analyses
For samples that are amenable to GC analysis, i.e., nonpolar and thermally stable, retention time locking (RTL) can be used. RTL is a technique developed by Agilent Technologies (Santa Clara, CA) that allows analysts to match analyte retention times (RTs) on any Agilent GC instrument in any laboratory in the world, provided that the same nominal GC method and capillary column are used (Giarocco et al., 2000). Using RTL, Agilent has developed several retention time-locked databases for GC and GC/MS that include the locked retention time, compound name, CAS number, molecular weight, and mass spectrum (RTL URL). The Agilent RTL Pesticide Library contains this information for 927 compounds, including pesticides, metabolites, and endocrine disruptors along with important polychlorinated biphenyls, polybrominated biphenyls, polycyclic aromatic hydrocarbons, synthetic musk compounds, Sudan dyes, and organophosphorus fire retardants. Another database contains all of the analytes specified for GC/MS analysis in the new Japanese "Positive List" regulations. While data analysis based on retention time is extremely productive and reliable, additional power and speed of analysis is obtained by deconvoluting the spectra. In GC/MS, deconvolution is a mathematical technique that "separates" overlapping mass spectra into "cleaned" spectra of the individual components. As seen at the top left of Figure 1, a total ion chromatogram (TIC) might consist of several overlapping components. The deconvolution software utilized in the Agilent Deconvolution Reporting Software (DRS) and discussed throughout this article is the Automated Mass Spectral Deconvolution and Identification System (AMDIS) developed by National Institute of Standards and Technology (NIST) (AMDIS URL). In addition, Agilent's RTL Pesticide Library also includes the AMDIS format for use with DRS. A filter can be set in AMDIS that requires the analyte's RT to fall within a user-specified time window at the expected RT. The spectra at any particular retention time can be deconvoluted (cleaned) by reporting only those ions whose chromatographic apexes are reached at that particular retention time; all other ions that may elute slightly earlier or later, may be present as general background and are discarded. In addition to the chromatographic apex at a specific RT, peak shape is an additional factor considered in deconvolution. That is, ions with both the same apex location www.intechopen.com and similar peak shape are clustered into a component for searching. In this way, the deconvolution process finds those ions whose individual abundances rise and fall together within the spectrum. As illustrated in Figure 2, deconvolution produces "clean" spectra that are the composite of only those ions at the same apex location and with similar peak shape. Deconvolution finds the components (a component is a group of related ions) from a complex TIC. Each component is searched against a retention time locking (RTL) library in AMDIS format. In addition to spectral matching, the locked RT can also be used as a criterion for hits. Depending on the match factor from the search, target compounds can be identified or flagged in a complex TIC.

Fig. 2. Graphical representation of the deconvolution process
Because RTL is used to reproduce the RTs with high precision, this window can be quite small-typically 10 s or less. This "RT qualifier" is very useful for blind study/screening (Sandy, 2004). The components identified by deconvolution and Agilent's Pesticide Library can be further confirmed against the NIST library containing pesticides and numerous other compounds. Note that the NIST reverse search is based solely on the ion pattern for the compound and does not incorporate retention time matching.

Liquid phase analyses
For samples that are more amenable to LC analysis, i.e., polar and thermally labile, a triple quadrupole (QQQ) is often used for trace level target analysis of complex matrixes. The multiple reaction monitoring process removes chemical background from the sample matrix, thereby providing superior selectivity and sensitivity for pesticide analysis. But LCsingle quadrupole and QQQ instruments cannot readily be used for nontarget identifications for two reasons. First, due to lack of selectivity and sensitivity, LC-single quadrupole and QQQ instruments do not usually operate in the full-scan mode for pesticide screening. Second, common LC/MS spectral libraries are unavailable due to difficulties in standardizing and reproducing fragmentation energies among instruments from different vendors. Therefore, when screening for a broad range of pesticides, a TOF (time-of-flight) or Q-TOF is the instrument of choice (Mezcua et al, 2009). A high-resolution TOF instrument always acquires full spectra and gives accurate masses. Recent experience has demonstrated that LC/TOF-MS has the capability and sensitivity to obtain full spectra with a mass accuracy of less than 1 ppm . This level of accuracy in mass allows the analyst to distinguish between compounds whose molecular weights are extremely similar. Table 1 lists several pesticides whose exact mass varies by a fraction of an amu. Exact Mass Database of hundreds of pesticides, the normally tedious manual process of matching can be done rapidly and reliably  using mass accuracy and RT as the criteria. After identification, the sample can be reanalyzed in the MS/MS mode using a QQQ or Q-TOF instrument to confirm any hits from the database search.

GC: Analysis of surface water
The California Department of Food and Agriculture (CDFA) prepared and analyzed surface water samples using an Agilent 6890N gas chromatograph equipped with an Agilent 5973 inert mass selective detector (MSD) (Siegel et al, 2004). The sample collection and preparation procedure is the following: A 1 L water sample is delivered to the laboratory in an amber glass bottle. Samples are stored under normal refrigerated conditions (approximately 4°C) until extraction within 7 days.

Sample preparation
a. Weigh and record the 1 L water sample, including sediments. b. Pour the water sample, including sediments, into a 2 L separatory funnel. Do not filter since pesticides would stick to humic materials. c. Spike with 1 mL surrogate spiking solution: 0.5 ng/mL Chlorpyrifos-methyl (0.5 ng).
Shake the separatory funnel gently to mix. d. Add 10-15 g granular Sodium Chloride (NaCl) for salting-out purposes. Shake gently to dissolve salt. e. Rinse water sample container with 60 mL Methylene chloride and add to the separatory funnel. Weigh and record the empty water sample container. Subtract and record the water sample weight. f. Shake and release pressure several times. Shake well for 3 min. Let settle until the lower Methylene chloride layer is completely separated from the above water layer. If there is too much emulsion in the funnel, use a sonicator to break up the emulsion. g. Filter the bottom organic layer through a bed of granular anhydrous sodium sulfate (approximately 20 g) into a 250 mL round-bottom flask. The sodium sulfate is supported on glass wool and is prewashed with 30-40 mL Methylene chloride. h. Add 60 mL Methylene chloride into the funnel and repeat steps (f) and (g) two more times. i. The round-bottom flask should now contain about 180 mL methylene chloride. Place the round-bottom flask on a Rotavapor evaporator (at about 100 rpm) and evaporate down to 5-7 mL at 40°C. j.
Transfer the contents of the round-bottom flask to a 15 mL collection tube. Rinse the round-bottom flask with 5 mL methylene chloride and add to the collection tube. k. Place the 15 mL collection tube on an N 2 -Evaporator with water temperature set at 40°C. Evaporate the sample to near dryness. l. Remove the tube from the N 2 -Evaporator and carefully add 1.0 mL methylene chloride and 10 mL 0.5 pg/mL internal standard (ISTD) solution into the collection tube. m. Vortex and transfer the solution into an autosampler vial. n. Cap and store the vial in a -5°C freezer until analysis.

LC/Q-TOF: Analysis of grape sample
An acetonitrile extract of a grape sample was prepared by the QuEChERS extraction method and analyzed by 1200 LC and 6510 Q-TOF (Agilent Technologies; 10). The QuEChERS sample extraction and cleanup procedure for both GC and LC is the following:

Extraction a.
Chop samples into small pieces and freeze in a bag overnight before grinding. Dry ice should be added during grinding. b. Weigh 10 g homogenized sample into a 50 mL Teflon centrifuge tube. c. Add 10 mL acetonitrile (and ISTD solution, if used). d. Add 4 g anhydrous magnesium sulfate, 1 g sodium chloride, 1 g trisodium citrate dehydrate, and 0.5 g disodium hydrogen citrate sesquihydrate to the tube. e. Adjust the pH to 5-5.5 using 5 M sodium hydroxide (NaOH).
f. Shake the sample vigorously for 1 min using a vortex mixer at maximum speed or by hand shaking. g. Centrifuge for 5 min at 3000 rpm.

Cleanup
a. Transfer 6 mL supernatant into a 12 mL polypropylene centrifuge tube that contains 150 mg primary-secondary amine adsorbent and 900 mg Magnesium Sulfate Anhydrous (MgSO 4 ). b. Shake for 30 s. c. Centrifuge for 5 min at 3000 rpm. d. Adjust the pH of the cleaned extract to 5.0 for analysis, if necessary.

Results and discussion
In a GC/MS scan analysis, it is always very difficult to identify compounds from high matrix background because the matrix ions overwhelm the compound signal. To be certain www.intechopen.com of the results, spectral averaging and background subtraction are often practiced. It is therefore a very time-consuming process to confirm compounds in a complex matrix.

Analysis of GC surface water sample data with DRS
Data files acquired by the CDFA on 17 surface water extracts were compared using two approaches (Wylie et al, 2004). Three example TICs are shown in Figure 3. The TICs show that the extracts were very complex. Approximately 8 h of a skilled analyst's time were required to process and confirm the results generated by ChemStation and searching the NIST library without the benefit of deconvolution. This workflow resulted in the identification of 37 pesticides plus one false positive. The same data files for the 17 TICs were processed using DRS without re-running the samples. The DRS reports were rapidly generated on all 17 TICs using batch processing. The display screen depicting this option is shown in Figure 4.    Fig. 6. MSD Deconvolution Report for the chromatogram in Figure 5 www.intechopen.com associated to the specific DRS library used (commercially available or user-built). Using DRS in conjunction with the 927 compound Agilent Pesticide Database, the same 37 pesticides were identified along with an additional 99 new identifications and no false positives. In addition to the improved results, the speed of the analysis was reduced from 8 h to 32 min, thus representing a 15-fold gain.

Analysis of pear data with DRS: DRS following the GC/MS analysis
The power of deconvolution is appreciated while comparing the top two spectra in Figure 7. The raw scan or original nondeconvoluted scan is shown on top. The clean scan, that is the deconvoluted component, is shown in the middle. The bottom scan is the identified compound in the AMDIS library. Without deconvolution, the analyst would visually compare the background subtracted raw scan and library scans for confirmation. It would be very difficult, if not impossible, to say that Kresoxim-methyl, the target compound in this example, is present using that type of comparison. The top mass spectrum is the raw scan from the TIC at 24.747 min and is clearly unsuited for library-searching purposes. In contrast, the middle spectrum (dashed lines are uncertain ions) presents the deconvoluted scan extracted from the raw scan. This deconvoluted (cleaned) scan is now easily and confidently matched against the library spectrum for Kresoxim-methyl presented in the bottom panel. For discussion on deconvolution parameter settings and additional advantages of deconvolution, reference the Agilent Application Note (Meng & Szelewski, 2010)

Analysis of peach data with DRS: DRS following the GC/MS analysis
An example DRS report for a peach extract is shown in Figure 8. As shown, Carbaryl appears with an AMDIS match factor of 74 versus the theoretical best match score of 100. Fig. 8. DRS report for analysis of peach extract in full scan mode.
In general, an AMDIS match factor of 80 or higher is enough to confirm a hit. The RT difference of 1.1 s between the actual and expected RTs also indicates excellent agreement, i.e., high confidence in the results. RTL and a suitable time window (e.g., ±10 s) provide this added qualification function in screening. Some missing quantitation results in the ChemStation column were due to some out-of-range qualifier ion ratios. Qualifiers out-ofrange are very typical for complex matrixes. The last two columns of the DRS report show the results from searching all the AMDIS hits against the NIST08 mass spectral library, which contains information on mass spectra only (not on RT). When the NIST library search finds a compound in the top 100 matches (a user-settable value) that agrees with the AMDIS results, its match factor is listed in the Reverse Match column. The Hit Number is shown in the last column, with 1 being the compound with the best match (highest match factor) in the NIST database. The compounds listed in the figure are all in the top 20 hits, with many as the number one hit. The further confirmation of hits found by DRS can easily be done with another injection in selected ion monitoring (SIM) mode or using GC detectors. In Figure 8, four compounds of interest with AMDIS match factors close to or less than 80 (Carbaryl, Captan, Propiconazole-II, and Fenbuconazole) were highlighted and further confirmed using SIM and GC detectors. Figure 9 is a DRS report of the same peach extract analyzed in SIM mode.  All of the compounds of interest in Figure 8 were found with AMDIS match factors over 90. Noticeably, the AMDIS quantitation results in Figures 8 (full scan) and 9 (SIM) were very comparable. These quantitation results are actually "semi-quant" results based on one average response factor for all of the 927 compounds in the DRS quantitation method. Figure 10 shows another approach for compound confirmation. The figure shows simultaneously collected MSD TIC and GC element-selective detector signals [µECD (micro electron capture detector) and FPD (flame photometric detector) in the phosphorus (P) mode] for peach extract. The results were obtained from a single injection with column flow split into three detectors (Meng & Szelewski, 2007). All compounds of interest were shown in the µECD chromatogram at the expected RT. Both SIM and GC element-selective detector analyses are easy and useful approaches to confirm hits found in the DRS screening process. Figure 11 shows the raw TIC of grape sample obtained from the LC/Q-TOF. Fig. 11. Raw full spectrum TIC obtained from LC/Q-TOF analysis of grape sample.

Analysis of grape sample data with MFE (Molecular Feature Extractor) and exact mass search (Meng et al, 2009)
The MFE was used to find the masses of interest within the TIC based on user-defined settings (see Table 2 for settings). The result was 510 potential compounds. The accurate mass measured for each of the 510 compounds was then searched against an Exact Mass Database, using search parameters described in Table 3. The RT, an optional entry in the database, can be used as a qualifier. Out of the 510 potential compounds, 15 had a match of exact mass with a compound in the database within 3 ppm accuracy. As an example, in the Compound List shown in Figure 12, compound 82 was identified as Spiroxamine. The hits from the database search can easily be confirmed using MS/MS analysis available on the Q-TOF. The three highlighted compounds were selected for targeted-MS/MS analysis. The "DB Diff (ppm)" column in the figure presents the difference in ppm (not amu) between the experimental mass found and the database-listed mass, and shows the excellent mass accuracy of the instrument. As shown in Figure 13, the grape sample was analyzed again in the LC/Q-TOF-MS/MS mode where both the MS1 full spectrum and the MS/MS full spectrum were acquired.  Table  4. This additional injection in the targeted MS/MS mode confirmed the hits from the database search.

Analysis of strawberry sample data with MFE (Molecular Feature Extractor) and exact mass search (Meng et al, 2009)
A strawberry sample was also analyzed similar to the grape sample above. The MFE produced 822 potential compounds. Figure 14 shows the TIC, the hyperlinked extracted compound chromatogram (ECC), and the mass spectrum for one of these compounds.  Table 3. Twenty-six of the 822 compounds had mass matches (3 ppm tolerance) with pesticides in the database. Three plausible exact-mass match compounds, cyprodinil, azoxystrobin, and boscalid were then selected for further confirmation using MS/MS (Q-TOF) analysis with the same instrument. The ECC and mass spectrum for each of the three are provided in Figure 15 along with the database search results showing a difference of less than 1 ppm in experimental and database masses.  Table 4. Using the accurate MS/MS masses for the fragment ions, formulas were generated for each compound found in this step. The confirmation process for azoxystrobin will be further discussed as an example. In Figure  16, the best-fit (with mass accuracy of 0.26 ppm and isotopes) formula generated from the Targeted MS/MS analysis for one of the compounds was C 22 H 17 N 3 O 5 , the formula for azoxystrobin. The two associated fragment masses for this peak had less than 1 ppm difference in mass (0.31 and 0.2 ppm) when compared to the database masses for fragments expected from the C 22 H 17 N 3 O 5 parent formula. In addition, the three isotope masses for the molecular ion all differed by less than 1 ppm. The table outlined in Figure 16 shows that the experimental isotope abundances of the three isotopes match well with the calculated (theoretical) abundances. The boxes in Figure 17 surrounding the isotopes represent the theoretical isotope abundances. www.intechopen.com

www.intechopen.com
Final confirmation of the structure is obtained by comparing the experimental fragment masses to likely theoretical fragment ion masses. Analysis of the structural formula, as seen in Figure 18, shows two likely fragments, with masses of 344.10351 and 372.09843 that match closely in mass to the two fragment ions found in the MS/MS data: 344.10286 and 372.09781 amu. The difference between experimental and calculated is 0.31 and 0.20 ppm, respectively. The accuracy of this comparison, along with the MS data identification and MS/MS formula generation results all strongly suggest the presence of Azoxystrobin in the analyzed strawberry extract. A diagram depicting this "screen and confirm" workflow is shown in Figure 19.  The book offers a professional look on the recent achievements and emerging trends in pesticides analysis, including pesticides identification and characterization. The 20 chapters are organized in three sections. The first book section addresses issues associated with pesticides classification, pesticides properties and environmental risks, and pesticides safe management, and provides a general overview on the advanced chromatographic and sensors-and biosensors-based methods for pesticides determination. The second book section is specially devoted to the chromatographic pesticides quantification, including sample preparation. The basic principles of the modern extraction techniques, such as: accelerated solvent extraction, supercritical fluid extraction, microwave assisted extraction, solid phase extraction, solid phase microextraction, matrix solid phase dispersion extraction, cloud point extraction, and QuEChERS are comprehensively described and critically evaluated. The third book section describes some alternative analytical approaches to the conventional methods of pesticides determination. These include voltammetric techniques making use of electrochemical sensors and biosensors, and solid-phase spectrometry combined with flow-injection analysis applying flow-based optosensors.