Because of the complexity of soil matrices, careful sample preparation was required to obtain a repeatable and sensitive multiresidue analysis. As mentioned earlier, the extraction efficiency of ethyl acetate was equivalent in comparison to other water-miscible solvents in different matrices.
Sample size optimisation
The sample size was optimised to ensure satisfactory recoveries. In practice, it is common to add a volume of water to dry samples prior to the extraction step (Rutkowska et al. 2018; Zhou et al. 2012). Next, the effect of sample size on method performance was evaluated. Almost all of the analytes provided an RSD of < 20% for a 2 g sample. The ME for 85% of the test analytes was less than 20%. In addition, the initial residue was diluted, which lowered the quantification limit. This made the analysis of the incurred residues of the compounds difficult. For sample sizes of 5 g and 10 g, the precision-RSDs were satisfactory (<20%) for most of the analytes. But, when compared to a 5 g sample size, the ME was higher for approximately 22% of the compounds (including dichlorvos, atrazine, fenitrothion, profenofos, propargite, p,p’-DDT and pretilachlor) for a 10 g sample. As a result, 5 g was determined to be the best sample size for further research. The results of the sample size optimisation of selected pesticides (including early, mid and late eluting ones from different chemical classes) are presented in Figure 1a and 1b.
Shaking versus sonication
When compared to sonication, shaking provided a better recovery percentage. As shown, 20 min of shaking time offered a better recovery (10-15%) (Figure 2) in comparison to the shaking time of 10 min. But, when this time was increased from 20 to 30 min, no significant changes in the recovery were observed. Thus, a shaking time of 20 min was chosen.
Optimisation of cleanup
Without cleanup, 13% of the analytes showed recoveries higher than 130%, while 26% of them exhibited a significant ME (above 50%). This indicated that the cleaning up of the soil extract was necessary to minimise the ME.
Optimisation of freezing temperature
The main fraction of the soil matrix is humus, which mainly comprises organic acids and their conjugate bases, and 20% of soil humas occurs in the form of lipids (Kononova 1966; Morrison 1969). Because of their high solubility in ethyl acetate, these lipids might get co-extracted with the target pesticide residues. By putting the extracts at -80 oC for 20 min, the GC-MS full scan mode analysis revealed that a large extent of matrix-derived compounds were reduced (Supplementary Figure 1). A longer span of freezing (20 min, 30 min and 1 hour, at -20 oC) affected the recoveries of some of the organochlorine pesticides in BS. The successful recovery of 86.7% of the tested pesticides demonstrated the effectiveness of this cleanup step. The ion ratio (the qualitative to quantitative MRMs, %) deviations (with the matrix matched standards) of some compounds (e.g. dichlorvos and parathion) did not comply with the accepted criteria (<30%) due to higher matrix effects, resulting in their false negative detection. Such false detection issues were resolved by performing the freezing step (Supplementary Figure 2a, 2b). Additionally, the peak shape of a number of compounds (e.g. chlorothalonil, oxyfluorfen, chlorfenapyr and fipronil) was improved with the freezing step (Supplementary Figures 3a, 3b, 3c, 3d).
dSPE cleanup optimisation
A separate cleanup experiment with different dSPE sorbents was also evaluated. The Cleanup 1 strategy included only PSA, which increased the pH of the BS extract (pH-7.8). As seen, satisfactory results were obtained for only 71% of the target analytes. The recoveries were unsatisfactory for the remaining analytes (mostly the base sensitive compounds), keeping their recoveries between 60-70% and 120-140%. The precision-RSDs were mostly < 20%. A lower recovery for a few herbicides (e.g. fluzifop-p-butyl, clomazone, propanil) might be due to their reaction with the dSPE sorbents (Lehotay 2007; Niell et al. 2009). Hence, for Cleanup 2, the amount of PSA was reduced and combined with C18. This provided satisfactory results for 84.8% of the analytes. The Cleanup 3, which had an enhanced amount of C18, was effective in binding and removing the starch and sugar co-extractives, also reported earlier by Caldas et al. 2011. The results were comparable to Cleanup 2 and Cleanup 3. Therefore, we chose Cleanup 2 as the optimum sorbent combination for the dSPE cleanup.
Combined effect of freezing and dSPE cleanup
When the effectiveness of Cleanup 2 was estimated, almost 83.3% of the analytes demonstrated a recovery range of 70-120%, which was much lower than the percentage of the analytes obtained only in the case of the freezing step without dSPE cleanup. For example, a significant number of compounds, for example, chlorfenapyr, tetrachlorvinphos, triazophos and tricyclazole, suffered from poor recoveries of <30% when the cleanup sorbents were used. The results showed that freezing at -80 oC for 20 min (without a dSPE cleanup) recovered the highest number of compounds (Figure 3a) within the acceptable range of 70-120%. The use of freezing plus dSPE cleanup did not have a significant effect on the recovery of pesticides from the extracts. The absence of a dSPE cleanup step is also observed in the study by Rutkowska et al. 2018 for dry herb samples, although Caldas et al. obtained better results for multiclass pesticides in the soil samples under paddy cultivation (Caldas et al. 2011). In another study on herbicides in soil, Zhou et al. noted even a better recovery for pyrazosulfuran ethyl with a normal extraction over PSA or C18 (Zhou et al. 2012). The recovery data of some compounds including allidochlor, dichlobenil, chlorthalonil, biphenyl, ethion, mevinphos, paclobutrazole, phorate, profenofos, tau- fluvalinate, trans-chlordane, tricyclazole and 2,3,5,6 tetrachloro aniline supporting this fact are presented in Figure 3b. Compared to the extraction method using the dSPE cleanup step, this method was more cost-effective as it involved fewer solvents and reagents (excluding expensive phase PSA and C18). Besides, the current study was less laborious as it did not require a partitioning method or the weighing of cleanup reagents.
All the tested pesticides could be selectively estimated in a single GC run of 21.31 min. Dichlorvos was the earliest compound (tR = 4.44 min), while fenvalerate (tR = 21.04 min) was the last one in the chromatogram. The GC programme provided chromatographic separation of the pyrethroid isomers and facilitated their accurate quantification as observed for cyfluthrin [alpha-HCH (tR = 9.303 min), beta-HCH (tR = 9.918 min), gamma-HCH (tR = 9.985 min) and delta-HCH (tR = 10.584 min), cyfluthrin-I (tR = 19.117 min), cyfluthrin-II (tR = 19.234 min), cyfluthrin-III (tR = 19.307 min) and cyfluthrin-IV (tR = 19.357 min)] and cypermethrin [cypermethrin-I (tR = 19.501min), cypermethrin-II (tR = 19.633 min), cypermethrin-III (tR = 19.705 min) and cypermethrin-IV (tR = 19.762 min)]. The quantitative results in three soil matrices are presented in Table 2. The validation of this optimised analytical method was evaluated by different parameters including linearity (expressed as R2), LOQ, ME, accuracy (expressed as recovery) and precision (expressed as RSD) (SANTE/12682/2019).
A good linearity (R2≥ 0.99) was obtained in the calibration curve for all compounds over a concentration range of 2-40 ng/g. A slightly lower R2 value (~0.98) was noted for fipronil (BS), hexachlorobenzene (NAS), cypermethrin (RS) and permethrin (RS). In the case of BS, a considerable ME was mainly observed for phenylureas (e.g. linuron), aromatic ethers (e.g. oxyfluorfen), polychlorobenzene (e.g. hexachlorobenzene), monosubstituted with nitro- (e.g. quintozene) and dinitrile (e.g. chlorothalonil) group, the mid-eluting organophosphates (e.g. phoratesulfide, tolclofos methyl chlorfenvinphos etc.) and late-eluting synthetic pyrethroids (e.g. cypermethrin etc.) compounds (Supplementary Figure 4).
According to the findings, 69% of the compounds had a lower ME (±20%) in BS, while 59 pesticides showed a medium ME (ranged from 21 to 49%). Among all, only 9 pesticides including tolcofos methyl, phorate sulfide, linuron, 4th peak of cypermethrin eluting at 19.762 min, delta-HCH, endosulfan sulfate, etridazole, oxyfluorfen and parathion showed a stronger ME of ≥ 50%. Tolcofos methyl showed the greatest ME value of 78.39%. In the RS soil type, 83% of pesticides showed a lower ME; 33 pesticides had values ranging from ±21 to ±49%. Four pesticides including alpha-endosulfan, chlorothalonil, fenvalerate and tatradifon exhibited signal enhancements of >50%. Tolcofos methyl showed the greatest ME value of 78.4% in RS. A signal suppression was observed only for (E)-chlorfenvinfos in BS as well as RS. Whereas in NAS, 90% of compounds showed no ME. A moderate ME was observed for 17 compounds excluding chlorpyrifos, alpha-endosulfan and oxyfluorfen, which had a higher ME. The ME of some selected pesticides is presented in Supplementary Figure 4.
The ME was lowered by applying the freezing step. A considerable ME of the same or different compounds was also observed previously in soil matrices. For instance, Fernandes et al. 2013 reported high MEs for 12 pesticides, which included alpha- and beta-HCH, HCB, o,p-DDT, bupirimate, chlorpyrifos, fludioxonil, malathion, methiocarb and pendimetaline. In another study, Asensio-Ramoset et al. reported a significant ME for 11 pesticides, namely buprofezin, chlorpyrifos, chlorpyrifos-methyl, diazinon, dimethoate, ethoprofos, fenirothion, malaoxon, malathion and phosmet (Asensio-Ramoset et al. 2010). According to Łozowicka et al. positive MEs (signal enhancement) were reported for alpha-endosulfan, beta-endosulfan, endosulfan sulfate, dichlorvos, methamidophos, oxyflurofen and oxamyl, whereas negative MEs (signal suppression) were observed in the cases of bupirimate, dichlobenil, etaconazole, propham and trifloxystrobin (Łozowicka et al. 2017). Therefore, to correct the recoveries, in our study, the matrix-matched calibration graphs (soil type-specific) were used for quantification. This approach adjusted the accuracies within the acceptable range of 70-120%.
At the lowest fortification level of 10 ng/g, 85% of the analytes demonstrated satisfactory method performance (recoveries ranged within 70-120% with a <20% RSD). The LOQ for these compounds was set at 10 ng/g. At this LOQ level of 10 ng/g, 11 chemicals, including p,p'-DDT, p,p'-DDD, p,p'-DDE, o,p'-DDE, o,p'-DDT, hexachloro-exo-epoxide, cypermethrin-1, aldrin and others, showed a recovery of between 60 and 69% and a reasonable precision RSD of 10%. For example, transfluthrin, trans nonachlor, trans permethrin, linuron, fluchloralin and the third isomeric peak of cypermethrin (eluting at 19.703) showed around 60% of recoveries. The precision RSD in each case was < 20%. The only exception was trans-nonachlor, which had a poor precision-RSD at 10 ng/g (above 20%). At 20 ng/g, however, a satisfactory result was achieved for all these pesticides. Only a handful of compounds, for instance, fenitrothion, propoxur, tau-fluvalinate and azinphos ethyl, showed recoveries of more than 120%, although their precision-RSDs were <10%. At the fortification level of 20 ng/g, all these analytes complied with the method performance criteria (Table 2).
In a previous study, certain compounds, for example, aldrin, tolyfluanid, dichlofluanid, dieldrin, fipronil, propargite, profenofos, chlorothalonil, p,p'-DDT, lambda-cyhalothrin, pyriproxyfen, diazinon, oxadiazon and edifenphos, demonstrated poor recoveries (Walorczyk et al. 2012). In our earlier study conducted on animal feed, a poor recovery was also experienced for these compounds (Kumar et al. 2020). However, the recoveries and precision of these compounds were satisfactory in this study (highlighted in bold, Table 2). Our results showed non-detection of allethrin, flucythrinate, phenothrin, tetramethrin, cycloate, endrin and paraoxon methyl in BS. Some compounds did not meet the identification criteria of the ion ratio matching (within ±30%), especially at lower concentrations where isobaric interferences from co-eluting matrix compounds might alter the value and result in non-detection (Walorczyk et al. 2012) (Figure 4a). Another reason for false negatives in the case of some compounds (e.g. paraoxon methyl, cycloate and endrin) might be because of their recovery loss in the sample preparation stage. The degradation of cycloate, a thiocarbamate ester, might have occurred due to alkaline hydrolysis. For paraoxon methyl, the breakdown could have been due to the oxidation of mixed mineral oxides present in the soil matrix (Zimmermann et al. 2013).
The method performance was evaluated in the remaining soils, namely NAS and RS. As illustrated in Figure 4b and Figure 4c, a few compounds, for instance, boromophos ethyl, chlorobenzilate, hexachlorobenzene, propanil, myclobutanil and nitrofen in NAS and bromofenvinphos, cis-permethrin, etofenprox, phenothrin and tetramethrin in RS, were not detected. In the NAS matrix, 89% of pesticides satisfied the method validation criteria laid out in the SANTE/12682/2019 guidelines, with the associated recoveries within 70-120% and precision RSD ≤20% at the LOQ of 10 ng/g. Among all, only prochloraz and ethalfluralin suffered from a poor recovery of <60%, with an associated precision RSD ≤20%. However, at a 20 ng/g level, their recoveries were satisfactory.
In RS, which is also an acidic matrix, this proportion was decreased to ~88.63%. Both beta-endosulfan (a sulfite ester) and EPN (a phosphonic ester) suffered from poor recoveries even in this acidic soil at 10 and 20 ng/g. Nevertheless, the recoveries were obtained in the range of 70-120% for the majority of pesticide-matrix combinations. In all matrices, a similar group of problematic compounds was observed with recovery values ranging from 60 to 130% with a consistent ≤20% RSD (Table 2, highlighted in bold, Supplementary Figure 5).
When calculating against different calibration curves, agreeable quantification results were obtained, which clearly indicated the satisfactory reproducibility and ruggedness of this method (Supplementary Table 2). The repeatability (intra-day precision) and reproducibility (inter-day precision) RSD values were <20% for all the target analytes. Besides, 30 continuous injections of the post-extraction spiked samples at 50 ng/g provided the RSD values <7% for all the target compounds, indicating good instrument repeatability (Supplementary Table 2).
Following GC-MS/MS analysis, four batches of the extracted samples (48 in number) could be acquired in a 24-hour cycle. This facilitated a high throughput residue analysis. For all the target compounds in all matrices, the output of this method was also economically viable with regard to time and productivity.
Applications in real-life situations
In the real soil samples, six different pesticides viz, dichlorvos, biphenyl, diphenylamine, anthraquinone, tricylazole and chlorpyrifos were detected (Supplementary Table 3). A repeat-extraction did not result in any additional recovery of residues. In all cases, the precision-RSDs were highly satisfactory (RSD, < 10%, n=6).
Comparison with other methods
As previously stated, the ethyl acetate-based method is widely used in pesticide residue analysis, particularly for vegetables and fruits. However, no previous applications of this method have been reported on soil. So, in this study, 220 pesticides were analysed in three different types of soil using a GC-MS/MS instrument in a single chromatographic run time of 21.31 minutes. The previous works required a longer run time (e.g. 41.88 min or 25 min), had a limited number of pesticides (e.g. 34, 58, 123 and 216 compounds) in scope or matrices (specific soil, no wider pH range). In comparison to the QuEChERS method, the ethyl acetate extraction provided superior recoveries for the traditionally low-recovery compounds, as described in section 3.4. However, in our study, the absence of a dSPE cleanup step might have resulted in better recoveries. When some compounds were extracted with acetonitrile as done in the QuEChERS method, the results pertaining to chlorobenzenes (RSD ranged between 26 and 41%), chlorophenols (RSD ranged between 21-35%) and HCH isomers (RSDs for alpha, beta, delta, gamma were 25, 31, 37 and 29% respectively) were less repeatable. These observations were also supported by an earlier study by Rouviere et al. (Rouviere et al. 2012). However, the ethyl acetate-based method here showed satisfactory and consistent recovery of the above-mentioned compounds with satisfactory precision RSDs of ≤ 20% (Table 2) in all the three soil matrices.