Natural deep eutectic solvents as a sustainable alternative for multi-class pesticide extraction in food safety analysis

Abstract

Pesticide residue analysis in food is frequently carried out worldwide, often requiring a large volume of organic solvents. To improve sustainability, new generation solvents such as natural deep eutectic solvents (NADES) have recently emerged as a promising alternative to conventional solvents. This study demonstrates the applicability of NADES as green extraction solvents for pesticides in food samples prior to analysis by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). The present research goes beyond previous studies by covering an extended set of pesticide residues. For that, five hydrophilic and hydrophobic NADES were evaluated as extraction solvents in a solid-liquid extraction (SLE). Initially, eleven representative pesticides covering a broad range of chemical classes and physicochemical properties were selected, while food matrices were chosen to cover different commodity groups according to SANTE/11312/2021v2, including orange, wheat, and spinach. NADES were tailored in order to cover a wide range of physicochemical properties such as polarity, pH, and viscosity. Thymol: menthol (molar ratio, 1:1) was found to be the most effective, and key extraction parameters were optimized. The greenness and transferability of the NADES-based SLE were evaluated using the AGREEprep and BAGI metrics. Finally, the method was successfully validated for the extraction of a wide range of pesticides commonly analyzed in food safety assessments.

Introduction

The development of efficient analytical methods in food safety analysis, fulfilling the legislation requirements, is crucial to control them and minimize their impacts on human health. Nevertheless, most official analytical procedures rely on toxic reagents and hazardous solvents¹, posing risks to operators and generating significant waste streams with serious environmental implications.

Regarding the human impact, the International Labour Organization (ILO) has identified solvents as a leading cause of work-related illness in industries dealing with chemicals, such as laboratories². The European Chemicals Agency (ECHA) promotes the substitution of hazardous chemicals with safer alternatives, especially those that pose significant risks to human health and the environment³. This approach aligns with the REACH Regulation (Registration, Evaluation, Authorisation, and Restriction of Chemicals) of the European Union, which establishes procedures for collecting and assessing information about the properties and hazards of chemical substances¹. However, organic solvents are widely used in routine extraction methods for solid matrices, particularly in the analysis of pesticides, mycotoxins, and veterinary drugs. Acetonitrile, in particular, is a key component of the QuEChERS method, which is commonly employed in routine analysis of pesticide residues in food and agricultural samples. However, it is considered ‘problematic’ in terms of SH&E (Safety, Health, and Environment). Prolonged exposure to acetonitrile can lead to adverse health effects, including damage to the digestive, cardiovascular, and central nervous systems, primarily due to the formation of cyanide in the body⁴. Another potential issue with acetonitrile is its availability. It is mostly a by-product from the manufacture of acrylonitrile, and hence its availability depends on market demands of that bulk chemical, and with that essentially on the world economy. An acute shortage occurred following the financial crisis of 2008⁵.

In the last decade, trends in green analytical chemistry (GAC) have been increasingly focused on the employment of alternative solvents, also known as “green solvents,” to promote sustainability in the sample treatment procedures⁶. These green solvents included ion liquids (ILs), deep eutectic solvents (DES), supercritical fluids, and bio-based solvents, and are utilized for the extraction of both contaminants and bioactive compounds from diverse matrices^7–9. NADES (natural DES) are composed of primary metabolites such as sugars, alcohols, amino acids, organic acids, and choline derivatives, and, at a specific molar ratio composition, they present a significant melting point depression, becoming liquid at room termperature¹⁰. In this stable and homogeneous eutectic mixture, the components function as hydrogen bond donors (HBDs) and acceptors (HBAs), leading to a lower melting point than their individual components, mainly due to the generation of intermolecular hydrogen bonds. NADES are considered eco-friendly, non-toxic, and biodegradable with a low cost, being easy to produce in the laboratory¹¹, which are suitable for the development of new sample treatments aligned with the green sample preparation (GSP) principles¹². While NADES have been widely used for extracting bioactive compounds from diverse food and plant matrices^13,14, their application for residue and contaminant extraction has hardly been exploited so far¹⁵. This is due to their very low volatility, which prevents evaporative pre-concentration steps that are often employed in sample preparation procedures for trace-level analysis. Consequently, their use is primarily limited to dispersive solvents in contaminant analysis, and therefore, their application in food safety methodologies remains scarce¹⁶.

In order to evaluate the sustainability of the analytical methodologies, different green metrics have been proposed in the last decade^17,18. Among these, AGREEprep has been used to assess standard methods widely applied in environmental, food, and pharmaceutical analysis, highlighting the widespread lack of alignment of GAC and GSP principles in current standard methods, including those for pesticide extraction¹.

In this regard, we have proposed the use of NADES as potential extraction solvents in multi-residue analysis of pesticides from various types of food matrices, including orange, spinach, and wheat. Both hydrophilic and hydrophobic NADES were evaluated as extraction solvents within a simple solid-liquid extraction (SLE) procedure. Key parameters affecting the extraction efficiency (e.g., volume of NADES, sample pH, and extraction time) were optimized to maximize performance in extracting 11 representative model pesticides. Next, the proposed NADES-based SLE was applied to extract a total of 313 pesticides commonly analyzed in routine food safety monitoring. Finally, the greenness and applicability of the NADES-based SLE were assessed and compared with other protocols developed for the same purpose.

Results and discussion

Selection of target samples, pesticides, and NADES

Three different matrices, namely orange, spinach, and wheat, were selected to represent foods belonging to different commodity groups: spinach as a representative matrix for group 1 (high-water content); orange for group 2 (high acid and water content), and wheat for group 5 (high starch and/or protein, low water and fat content).

The initially selected pesticides, which include fungicides, herbicides, and insecticides, belong to different chemical classes and present a wide spectrum of physicochemical properties. These pesticides include omethoate, acetamiprid, propamocarb, metoxuron, azoxystrobin, cyprodinil, tebuconazole, spinosyn A, spinosyn D, chlorpyrifos, and deltamethrin, reflecting the diverse array of pesticides available on the market. Their main characteristics are summarized in Table 1, and their chemical structures are shown in Supplementary Fig. 1. The distribution of the 11 compounds across the chromatographic space (retention time, x-axis) in relation to their LogP values (y-axis), with the size of each label indicating the molecular weight of each compound, is illustrated in Fig. 1.

Table 1.

Characteristics and physicochemical properties of target pesticides

Pesticide	Chemical group	Type	logKow	pKa	Compound property	Acid/base
Omethoate	Organophosphate	Insecticides	−0.74		Neutral very polar	-
Acetamiprid	Neonicotinoid	Insecticides	0.8	0.7	Potentially cationic	Basic
Propamocarb	Carbamate	Fungicides	1.12	9.48	Potentially cationic	Basic
Metoxuron	Urea	Herbicides	2.11		Neutral interm. polar	-
Azoxystrobin	Strobilurin	Fungicides	2.5		Neutral interm. polar	Non-ionized
Cyprodinil	Anilino-pyrimidine	Fungicides	3.59	4.44	Neutral interm. polar	Basic
Tebuconazole	Triazole	Fungicides	3.7	2.3	Neutral interm. polar	Non-ionized
Spinosyn A	Macrocyclic lactone	Insecticides	4	8.1	Neutral nonpolar	-
Spinosyn D	Macrocyclic lactone	Insecticides	4.5	7.87	Neutral nonpolar	-
Chlorpyrifos	Organophosphorous	Insecticides	5		Neutral nonpolar	Non-ionized
Deltamethrin	Pyrethroid	Insecticides	6.2		Neutral very nonpolar	Non-ionized

Fig. 1.

Visualization of the distribution of 11 representative pesticides by retention time (x-axis), LogP (y-axis), and molecular weight (Label Size).

For assessment of applicability to multi-residue analysis, a mixture of 313 pesticides, representative of those commonly investigated in routine analyses and including the 11 selected compounds described above, was considered for method validation. Detailed information about the pesticides involved in the study, including CAS, molecular formula, SMILE, as well as chemical identifiers (InchiKey and planar InchiKey), is available in Supplementary Data 1. A chemical space map was created for the 11 selected standards and the additional 302 pesticides (full set) to confirm that the selected compounds effectively represent the broader dataset, both in physicochemical diversity and overall chemical space coverage (Supplementary Fig. 2).

The selection of studied NADES was based on several criteria. First, these NADES have been previously molecularly characterized, ensuring the formation of the supramolecular structure and confirming their eutectic nature^2,3. Second, their physicochemical properties (see Table 1) were considered to provide a wide range of pH and polarity, facilitating an evaluation of how these factors influence extraction efficiency. Finally, a maximum viscosity of 50 mPa·s was selected to ensure efficient mixing and extraction with the food matrix. In addition, choline chloride (ChCl)-based NADES was avoided due to recent studies reporting its potential eco- and cytotoxicity, particularly when combined with organic acids¹⁹.

Sample treatment optimization

The key variables affecting the extraction efficiency of the representative target pesticides from the selected matrices were carefully evaluated using univariate optimization. Thus, type and volume of NADES, addition of water, and pH adjustment of the sample, as well as extraction time, were carefully investigated.

Selection of the extraction NADES

For this experiment, the three commodities were initially fortified with the target pesticides and extracted using the five different NADES to evaluate their extraction effectiveness in such matrices. 1 g of orange and spinach was directly mixed with 3 mL of each NADES under mechanical agitation. In case of wheat 1 mL of water was added to 0.5 g of the samples prior to the addition of 3 mL of each NADES. After centrifugation, 500 µL of the supernatant was diluted with 500 µL water or MeOH, depending on the NADES, to avoid phase separation in the chromatographic vial.

As shown in Supplementary Fig. 3, all the tested NADES were capable of extracting some of the target compounds. In general, NADES 2 and 3 showed lower recoveries, particularly for chlorpyrifos, spinosyn A, spinosyn D, and deltamethrin. These compounds were better extracted using NADES 4 and 5; however, NADES 4 was unable to extract propamocarb. While NADES 5 showed higher recoveries overall, the results were still unsatisfactory for propamocarb and spinosyn A and spinosyn D, especially in orange samples. NADES 1 generally yielded good recoveries, except for chlorpyrifos and deltamethrin (both with high LogKw values). However, other considerations apart from recoveries should also be taken into account. For instance, the shape of chromatographic peaks. The use of NADES 4 resulted in poor peak shapes for most compounds, complicating their integration. A similar effect was observed with NADES 1, particularly for polar compounds eluting early, which is likely related to its polarity. To further investigate this effect, the matrix effect of the NADES was assessed. For this, a work mix solution of the target pesticides was prepared using the corresponding NADES and MeOH (50:50, v/v) as injection solvent, and compared with the same mix prepared and injected in the absence of the NADES; i.e., MeOH:water (50:50, v/v). The ME (%) of the NADES, considering peak areas, was calculated following Eq. 1:

$$\begin{array}{c}{\mathrm{ME}}_{\mathrm{NADES}}(\%)=\left(\right(\mathrm{area}\mathrm{of}\mathrm{standard}\mathrm{solution}\mathrm{in}\mathrm{NADES}-\mathrm{area}\mathrm{of}\mathrm{standard}\mathrm{solution}\mathrm{in}\hfill \\ \mathrm{solvent})/\mathrm{area}\mathrm{of}\mathrm{standard}\mathrm{solution}\mathrm{in}\mathrm{solvent})*100\hfill \end{array}$$ 1

As observed in Supplementary Fig. 4, NADES 1 exhibited a strong signal suppression for more polar compounds, as expected. Considering higher recoveries, improved peak shapes, and lower intrinsic ME, NADES 5, composed of thymol and menthol in a 1:1 molar ratio, was selected for further experiments. Therefore, all subsequent experiments were carried out using NADES TM.

The ME for each matrix type was assessed by fortifying the food sample (n = 3) prior to extraction with TM and comparing the peak areas obtained with those from the NADES extract of a blank sample spiked just before injection. In both cases, NADES TM was in equal proportion in the vial, e.g., TM:MeOH (50:50, v/v). Figure 2 presents the recovery and matrix effects obtained for all target compounds when using TM as the extraction solvent. As expected, the first peaks exhibit matrix effects, with signal enhancement observed in this case. Overall, the matrix effects were below 20% for the tested food matrices.

Fig. 2. Performance of the selected NADES as extraction solvent in different food matrices.

a Recoveries (%) and b matrix effects (ME, %) obtained using thymol–menthol NADES (molar ratio 1:1) for the extraction of wheat, spinach, and orange samples. Error bars represent relative standard errors (n = 4).

Addition of water

The addition of water to the solid food matrix was considered to improve its contact and interaction with the NADES solvent, which could also allow for a reduction in the volume of NADES used without compromising the mass transfer of the analytes to the extraction solvent. Reducing the volume of NADES is useful to avoid the dilution of the analytes in the resulting extract. Each sample was mixed with 1, 2, and 5 mL of water prior to the addition of TM (Supplementary Fig. 5). It is important to notice that the employment of a hydrophobic NADES such as TM allowed phase separation formation without the addition of salts. When water was added, a significant decrease in the recovery for propamocarb was observed in orange and spinach samples. Nevertheless, the addition of water favored the extraction of other pesticides such as acetamiprid and metoxuron in spinach, while no improvement was observed in orange samples. As indicated before, the addition of water to the wheat samples was necessary to facilitate extraction (wet/swell the matrix). In this case, the addition of water had little impact on the recovery of the compounds. However, when 5 mL of water was used, a decrease in the recovery of omethoate was observed, with the recovery dropping to 60%. Considering the effect of water addition in all tested matrices, 2 mL was selected for further experiments.

Volume of NADES

The volume of TM was evaluated in the range of 1, 2, and 3 mL to determine if a lower volume of NADES could maintain the same recovery efficiency. With 1 mL, extracting 0.5 mL from the supernatant was challenging due to the viscosity of the NADES and the interface of the matrix. This could lead to reduced reproducibility among replicates. Between 2 and 3 mL no substantial differences were observed, so 2 mL was considered enough as the extraction volume. The obtained recoveries using 2 mL of NADES were similar to those illustrated in Supplementary Fig. 5 when using 3 mL.

Effect of the sample pH

Considering the poor recoveries of propamocarb in both orange and spinach samples, as well as spinosyn A and D in orange, and the pKas of these compounds, we decided to adjust the pH of the sample to a basic level. The pH adjustment was based on a previous protocol where the use of phosphate buffer and NaOH was needed for acidic food matrices. The phosphate buffer (0.1 M) and NaOH (1 M) were prepared by dissolving the required amounts of each in ultrapure water under magnetic stirring.

Spinach and orange were mixed with 2 mL of water, buffer, or buffer containing NaOH (130 µL). Initially, the pH of the spinach was 5.5, and the pH of the orange was 3.4. After adding the buffer, the pH increased to 6.5 for both samples. Upon adding NaOH, the pH rose to 8–8.5 in orange and 9–10 in spinach. While the addition of the buffer improved the recoveries of propamocarb, the most significant improvement was seen after adding NaOH. This enhancement was observed not only for propamocarb, spinosyn A, and spinosyn D, but also for the other pesticides (Fig. 3). Therefore, the use of phosphate buffer and NaOH instead of water was selected for spinach and orange samples.

Fig. 3. Effect of pH adjustment on analyte recoveries in vegetable and fruit matrices.

Recoveries obtained for a spinach and b orange after pH adjustment using phosphate buffer (0.1 M) and NaOH (0.1 M). Error bars represent relative standard errors (n = 4).

Time of extraction

The rotatory shaker was chosen to ensure proper mixing of the matrices and NADES. The extraction time was evaluated for 2, 5, 10, and 15 min. It was observed that 2 min were insufficient for achieving good extraction efficiency for most of the target compounds. The recoveries increased up to 10 min, which was selected as the optimum extraction time (Supplementary Fig. 6).

Supramolecular structure confirmation of the TM extract

The selected NADES TM (1:1) has been previously characterized in terms of its physicochemical properties and the supramolecular structure formation⁴, as well as its long-term stability³. However, the stability of the NADES post-extraction has not been assessed. Therefore, after the extraction of the pesticides, the resulting NADES from each matrix were collected and re-analyzed by NMR. In the ¹H-NMR spectra after the extraction of wheat, spinach, and orange (Fig. 4), it can be observed how the signal belonging to menthol + thymol-OH combined appeared in all cases (between 6 and 6.5 ppm). The slight differences in the chemical shifts of orange and spinach with respect to wheat can be due to slight variations in the sample pH. In addition, the NOESY experiment was conducted to confirm the interaction between thymol and menthol. After irradiating a signal from thymol at a chemical shift of 3.28 ppm, the corresponding signals of the menthol were observed (Supplementary Fig. 7). In addition, the combined OH groups of menthol and thymol were observed at 6-6.5 ppm. This analysis demonstrated that, even after the extraction procedure, the NADES maintained its supramolecular structure, despite being exposed to complex food matrices, aqueous solution, and intense agitation. This experiment provides additional evidence supporting the stability of the NADES throughout the process.

Fig. 4. ¹H NMR characterization of thymol–menthol NADES after extraction of food matrices.

¹H NMR spectra of thymol–menthol (TM, molar ratio 1:1) obtained after extraction of a wheat, b orange, and c spinach samples, recorded in CDCl₃ at 30 °C. For each sample, 100 μL of TM supernatant was dissolved in 500 μL of CDCl₃.

Greenness and applicability assessment

The greenness of and applicability of the proposed NADES-based SLE for the extraction of multi-residues of pesticides in several matrices were assessed by AGREEprep and BAGI tools, respectively, and compared with the AOAC version of the QuEChERS method (AOAC Official Method 2007.01 2007)²⁰, and a recently proposed miniaturized version of this protocol²¹. Table 2 summarizes the main parameters involved in each sample treatment procedure as well as the AGREEprep and BAGI scores obtained in each case.

Table 2.

Summary of the selected sample treatment procedures for greenness and applicability assessment, and comparison

Extraction method	Sample size (g)	Extraction solvent (volume, mL)	Salting-out (salts, g)	TET/Tª (min/°C) *	AGREEprep score	BAGI score
AOAC QuEchERS20	15	MeCN (15)	MgSO4 (6) NaOAc (1.5)	25/−80
Miniaturized QuEchERS21	0.5	MeCN (0.5)	MgSO4 (0.2) NaOAc (0.05)	18/−80
SLE (this work)	1	NADES TM (2)	-	20/room Tª

^*TET/Tª: Approximated total extraction time and extreme temperature

AGREEprep is an open-access software based on the 10 principles of GSP that allows the evaluation of the sustainability of a sample procedure, providing a final score on a unified 0–1 scale. Regarding the scores obtained for all evaluated methods, the proposed NADES-based SLE obtained the highest score (0.65), which highlights the greenness of the sample treatment even compared to a scaled-down version of the commonly used QuEChERS protocol. This underscores that the reduction of solvents and reagents is an effective approach to lowering the environmental impact of analytical methods; however, further improvements may be difficult to achieve unless green solvents are employed. In addition, both QuEChERS procedures are multi-step procedures with freeze-out (−80 °C, 10 min), which also increases the energy consumption and sample throughput, reflected in criteria 8 and 6, respectively. The use of safer and bio-based reagents, such as NADES (criteria 2 and 3), and the reduction of waste and solvents (criteria 4 and 5) considerably improve the output of the proposed sample treatment. In addition, due to the simplicity of the procedure, since no salting-out or clean-up steps were required, we could maximize sample throughput (criterion 6).

BAGI can be considered complementary to the well-established green metrics, and it is mainly focused on the practical aspects of white analytical chemistry (WAC)²². This metric considers 10 criteria to produce a pictogram and a score that depicts the practicability and functionality of an analytical method. The BAGI score provides insight into how easily a procedure can be transferred to another laboratory, which is related to the applicability of the sample treatment in routine analysis. As illustrated by BAGI pictograms (Table 2), the scores obtained in all cases exceed the 60-point threshold recommended by the author to consider a method “transferable” to other laboratories. In our case, the fact that NADES are in-house prepared is penalized in the assessment. Currently, NADES are not commercially available as pre-made products. While this may be considered a limitation, it is also the current state of the art in the field. Commercially available solvents are preferred in this evaluation since they are easier to obtain. On the other hand, multi-analyte analysis and simultaneous sample preparation are preferred, which is fully accomplished with the proposed method. In this regard, the simplicity of the NADES-based SLE, which involves fewer steps than other protocols, results in an improvement that leads to a similar overall score when compared with the scaled-down proposal.

Validation of the proposed method

Linearity and matrix effects

Matrix-matched calibration curves were prepared by spiking blank extracts of the corresponding matrix with the working solution of 313 pesticides at five concentration levels (corresponding to 1, 2.5, 5, 25, and 50 µg/kg) for spinach and orange, and at four concentration levels (1, 5, 10, and 20 µg/kg) for wheat. Matrix-matched calibration curves were prepared in triplicate for each matrix, and RSD (%) for the three injections at each concentration level. The back-calculated concentration (BBC, %), which represents the average percentage deviation of the experimental concentration from the theoretically predicted value, was also checked. In most cases, an acceptable linearity (RSD within ±20%) was obtained across the matrices. On the contrary, some compounds were not detected in a specific matrix type at 25 µg/kg, although spiked after extraction. This concerned eight pesticides in orange (aminopyralid, asulam, carbosulfan, dodine, fludioxonil, imazapic, pyridaat-metabolite, tembotrione); 12 pesticides were not detected in spinach (aminopyralid, asulam, bitertanol, carbosulfan, dodine, fludioxonil, imazapic, phospmet-oxon, pyridate – metabolite, pyridanyl, spirotetramat-keto-hydroxy, tembotrione); 20 pesticides were not detected in wheat (acephate, aminopyralid, asulam, bitertanol, carbosulfan, disulfoton, dodemorph, fluazuron, halofenozide, imazalil, imazapic, matrine, mecarbam, pyrethrinen, pyridalyl, spirotetramat, teflubenzuron, thiofanate-methyl, triadimefon, triadimenol). Matrix effect (ME) can be attributed to many factors, affecting analyte ionization in the interface between the LC and the MS and, therefore, resulting in ion suppression or signal enhancement. ME can be estimated by comparing the analytical response or calculated concentration of blank extracts spiked after the sample treatment with those results from a standard in solvent solution at the same concentration. In this case, the blank extracts were in NADES:MeOH (50:50, v/v) while the standard in solvent solution was a mixture of water:MeOH (50:50, v/v), both at 25 µg/kg. This way, suppression or enhancement effects from both the food matrix and the nature of the NADES were considered. The following equation (Eq. 2) was used for this comparison:

$$\begin{array}{c}\mathrm{ME}(\%)=\left(\right(\mathrm{concentration}\mathrm{of}\mathrm{the}\mathrm{spiked}\mathrm{sample}\mathrm{extract}-\mathrm{concentration}\mathrm{in}\mathrm{solvent})/\hfill \\ \mathrm{concentration}\mathrm{in}\mathrm{solvent})*100\hfill \end{array}$$ 2

A ME of 0% indicates the absence of the matrix effect, a ME below 0% involves signal suppression, while a ME above 0% reveals signal enhancement. ME below │20%│ (signal suppression or enhancement) is considered here as acceptable here. For those analytes with ME of │20–50%│ (medium ME) and above │50%│ (strong ME), it is necessary to employ matrix-matched calibration curves. Figure 5 summarizes the ME (%) for all pesticides investigated across all matrices. The total number of compounds represented for each matrix depends on the detection of these compounds in the corresponding matrix. Consequently, a lower number of compounds were evaluated in wheat, as fewer compounds were detected in this matrix with acceptable linearity.

Fig. 5. ME (%) for pesticides in each evaluated matrix.

a Spinach, b orange, and c wheat samples, separated by ranges.

ME values were classified in serious suppression (below −50%), high suppression (in the range −50 to 20%), insignificant ME (between −20 and 20%), high enhancement (from 20 to 50%), and serious enhancement (above 50%). It can be observed that spinach presented insignificant ME for the highest number of pesticides (256), followed by orange (236) and wheat (210). In general, signal suppression was observed for most of the other analytes, presenting medium ME. In the case of wheat, a higher number of analytes presented strong matrix suppression. Nevertheless, these effects were compensated for by using matrix-matched calibration curves.

Recovery experiments

Performance criteria for the optimized SLE-LC-MS/MS method were evaluated for 313 pesticides at 10, 20, and 100 µg/kg (six replicates each) for the three matrices. A bar graph for each matrix, summarizing the recovery results for all pesticides, is shown in Fig. 6.

Fig. 6. Recovery statistics for each matrix at three concentration levels.

Each panel represent the recoveries (%) in a spinach, b orange, and c wheat samples.

In spinach, 189 pesticides were recovered within the range 70–120% and RSD below 20% at all three levels, 20 and 100 µg/kg. The number of pesticides recovered increased to 252 and 256, respectively. In orange, the number of pesticides recovered at 10 µg/kg was similar to that in spinach, and the recovery improved as the spiking concentration increased. In wheat, pesticide recovery was less efficient, with 144 pesticides recovered at all three concentration levels, possibly due to the reduced sample size or increased signal suppression. Although recovery rates were lower compared to spinach and oranges, the RSD remained below 20% for around 100 pesticides that were recovered in a 30–70% or in a >120–<140%, indicating good precision. Recoveries within these ranges can be considered acceptable when the RSD does not exceed 20%. According to SANTE guidelines, and considering this criterion, the LOQ for the majority of the pesticides was established at 10 µg/kg.

Recovery values, as well as repeatability data (expressed as RSDr), obtained for the representative target pesticides at each concentration level and matrix are presented in Supplementary Data 3.

Proficiency test analysis

The proposed and validated method was further evaluated using a FAPAS (Food Analysis Performance Assessment Scheme) proficiency test material (19297, 2020) still available in the laboratory, which is an international external quality assurance program designed to assess laboratory performance in food analysis. Therefore, a FAPAS spinach sample was also analyzed using the proposed method. In this case, the FAPAS included 12 pesticides of which 9 were part of our multi-residue validation (e.g., ametoctradin, azoxystrobin, benalaxyl, dimethoate, diniconazole, fluopyram, linuron, spiroxamine, terbufos-sulfone). Information about FAPAS used in this study is provided in Supporting Information (Supplementary Table 1). This FAPAS was submitted to the proposed NADES-based SLE procedure in triplicate and analyzed by UHPLC-MS/MS. The concentrations of the target pesticides were determined using a matrix-matched calibration curve prepared in organic spinach. The obtained results, along with a comparison to the assigned values provided by FAPAS, are presented in Table 3. For this purpose, z-scores (Eq. 3) were calculated to indicate the deviation of the experimental results from the assigned values.

$$\mathrm{Z}=\mathrm{X}-\mu /\sigma$$ 3

Where X = measured value, μ = assigned value reported by FAPAS, and σ = standard deviation for proficiency assessment (often FFP-RSD × μ)

Table 3.

Comparison of experimental results with assigned values and z-scores in the FAPAS spinach proficiency test

Pesticide	Experimental X, µg/kg (%RSD, n = 3)	FAPAS, µg/kg	Z-scores, | z |
Ametoctradin	30.9 (3.9)	89.6	2.3
Azoxystrobin	131.8 (1.0)	155	1.1
Benalaxyl	28.4 (9.5)	54.8	1.2
Dimethoate	58.3 (6.3)	59.9	0.1
Diniconazole	43.3 (1.6)	73.0	1.4
Fluopyram	126.9 (2.4)	129	0.1
Linuron	45.9 (2.6)	43.1	0.1
Spiroxamine	13.28 (1.5)	52.7	1.8
Terbufos-sulfone	89.2 (1.3)	90.0	0.0

For FAPAS, the FFP-RSD is 22% when X is 120 µg/kg or lower. When X is higher than 120 µg/kg (i.e., azoxystrobin and fluopyram), then the FFP-RSD decreases with X with the modified following the modified Horwitz equation (Eq. 4):

$$\mathit{FFP}-\mathit{RSD}\left(\%\right)=2^{(1-0.5log10\left(X\right))}$$ 4

Where X is expressed in kg/kg.

In all cases, satisfactory z-scores (|z| ≤ 2) were obtained except for ametoctradin, for which a questionable performance (2 < |z| < 3) was observed. Nevertheless, the analytical performance for this compound was acceptable, showing good linearity, recoveries above 70%, RSD below 9% and ME (%) close to 0. These differences could be attributed to potential degradation of ametoctradin in the FAPAS material, as the sample had been stored in the freezer for 4 years before analysis.

Overall, the use of the FAPAS proficiency test confirmed that the proposed NADES-based SLE can be considered a suitable alternative for the multi-residue pesticide analysis in food samples such as spinach.

In conclusion, the key novelty of this study is the application of NADES as environmentally friendly alternatives to conventional organic solvents for multi-pesticide residue extraction, highlighting their potential to make LC-MS/MS-based food analysis greener and more sustainable. Among the NADES evaluated, thymol:menthol (1:1, molar ratio) showed the best overall performance, providing satisfactory recoveries, acceptable matrix effects, and good compatibility with the chromatographic separation and mass detection. The optimized NADES-based SLE method proved robust and was successfully validated for a large number of pesticides in different food commodities, including spinach, orange, and wheat, in accordance with SANTE guidelines. To the best of our knowledge, this is the first time a NADES-based extraction has been validated for such a large number of compounds across different food matrices. Importantly, NMR analyses confirmed that the supramolecular structure of NADES remained stable after extraction, underscoring their robustness in complex food matrices.

In addition, the proposed approach achieved higher greenness scores compared with conventional and miniaturized QuEChERS procedures, as confirmed by AGREEprep and BAGI assessments, while also demonstrating good applicability in a FAPAS proficiency test.

Overall, the results highlight the potential of NADES to serve as greener, efficient, and transferable extraction solvents in routine food safety monitoring, thereby advancing the alignment of pesticide residue analysis with the principles of GAC. Future work should further explore the large-scale applicability of NADES-based SLE and investigate its potential integration into standardized protocols.

Methods

Chemicals and reagents

All reagents used throughout this work were of analytical grade, and solvents were of HPLC grade, unless otherwise specified. Methanol (MeOH) and acetonitrile (MeCN) were obtained from Actu-All Chemicals (Oss, The Netherlands). Phosphate buffer and NaOH were purchased from Sigma Aldrich Chemie B.V. (Zwijndrecht, The Netherlands). Water was obtained from a Milli-Q water purification system from Millipore (Burlington, MA, USA).

Standards of the target pesticides such as omethoate, acetamiprid, propamocarb, metoxuron, azoxystrobin, and spinosad (mixture of spinosyn A and spinosyn D 74.2 & 22.3%) were purchased from Dr Ehrenstorfer, while cyprodinil, tebuconazole, chlorpyrifos, and deltamethrin were obtained from Sigma Aldrich Chemie B.V. The individual stock solutions were prepared in MeCN at a concentration of 5 µg/mL and stored in the dark at −18 °C. Subsequently, an intermediate mix solution with a concentration of 1 µg/mL was prepared from these individual solutions and used for fortification purposes.

The mixture of 313 pesticides was prepared in MeOH + 0.04% of acetic acid at a concentration of 1 µg/mL.

The primary components for NADES preparation such as betaine anhydrous (97%, CAS 107-43-7) was obtained from TCI (Tokyo, Japan) while glycerol (≥99.5% CAS 56-81-5), urea (CAS 57-13-6), menthol (CAS 1490-04-6), thymol (CAS 89-83-8), dodecanoic acid (CAS 143-07-9) and lactic acid (CAS 50-21-5) were purchased from Sigma Aldrich Chemie B.V. These reagents were kept at room temperature in their original packaging until NADES preparation.

Deuterated chloroform was obtained from CortecNet (Voisins-Le-Bretonneux, France).

Preparation of NADES

In this study, three hydrophilic and two hydrophobic NADES were considered. The studied NADES alongside their composition and main physicochemical properties are listed in Table 4. These NADES were prepared by mixing different HBDs and HBAs at a predetermined molar ratio as indicated in Table 4.

Table 4.

NADES composition and physicochemical properties

Code	Name	Component 1	Component 2	Component 3	Molar ratio	Polarity ENRN (kcal/mol)	pH	Viscosity (mPa*s)
NADES 1	LGLH	Lactic acid	Glycerol	water	1:1:3	-	1.4	26.0
NADES 2	BLH	Betaine	Lactic acid	water	1:1:3.5	1.1	3.3	12.3
NADES 3	UGLH	Urea	Glycerol	water	1:3:2.5	1.2	8.0	17.3
NADES 4	DAM	Dodecanoic acid	Menthol	-	1:4	0.3	4.2	14.7
NADES 5	TM	Thymol	Menthol	-	1:1	0.5	6.1	47.9

E_NR^N Normalized Nile Red transition energy.

The assigned names correspond to the abbreviations of the two main components of the mixture. The hydrophilic nature of NADES, LGLH, BLH, and UGLH, and the addition of water to the mixture, was represented by adding “H” at the end of the nomenclature.

NADES were prepared using the heating-stirring method as indicated by Dai et al²³. The heating and stirring method consists of the agitation of the mixture using a stirring bar at a temperature of 50–90 °C until a clear liquid is formed (approximately 30–60 min). The corresponding natural components of each NADES were placed in a dark bottle, which was heated at 40 °C for 1 h under magnetic stirring (IKA™ RET Basic, Fisher Scientific, Landsmeer, The Netherlands) to obtain a homogeneous transparent liquid. The absence of precipitation or crystal formation was checked 24 h after their preparation. The in-house prepared NADES were stored in the dark at room temperature.

Samples and sample treatment

Blank samples of orange, spinach, and wheat purchased from a local store in the Netherlands were used for the optimization of the method. The absence of the target compounds was confirmed before the fortification of the samples at the desired concentrations. In all cases, the sample was homogenized and ground considering the whole food using two different mechanical grinders, depending on the commodity, namely Stephan Chopper Model CUT 05 (Stephan Machinery GmbH, Hameln, Germany) for Spinach and Orange, while a Retsch GM200 was used for grinding wheat. Samples were stored at 4 °C until analysis.

The sample treatment involved a NADES-based solid-liquid extraction (SLE). For the procedure, 1 g of orange or spinach and 0.5 g of wheat were weighed and placed in a 15-mL Falcon tube. Subsequently, for the spinach and orange samples, 1.87 mL of 0.1 M phosphate buffer and 0.13 mL of 1 M NaOH were added to adjust the pH. For the wheat sample, 2 mL of water was added, and, due to the dry nature of this sample, vortexing for 10 s was required to ensure homogenization. Next, 2 mL of the selected NADES TM (molar ratio, 1:1) was added to each tube. The tubes were subjected to rotary mechanical agitation for 10 min (Heidolph REAX 2), followed by centrifugation at 4500 rpm and 4 °C for 10 min (MSE, Rotixa 50 RS). A 0.5 mL aliquot of the NADES supernatant was then transferred to a chromatographic vial and diluted with 0.5 mL of MeOH before injection into the UHPLC-MS/MS system. A schematic overview of the proposed NADES-based SLE is illustrated in Fig. 7.

Fig. 7. NADES-based SLE procedure for the extraction of multi-class pesticides from several food commodities.

a represents the first step (pH adjustment) for orange and spinach samples while b represents the just the addition of water in the case of wheat samples.

LC-MS/MS analysis

The analyses were carried out using an ExionLC system (SCIEX, Framingham, MA, USA) coupled to a TripleQuad 6500+ mass spectrometer (SCIEX). An electrospray ionization (ESI) interface was used for the ionization of the analytes. The chromatographic separation was achieved on an ACQUITY LC BEH C8 (2.1 × 100 mm, 1.7 µm particle size, 130 Å pore size, Milford, MA, USA) connected to an ACQUITY UPLC BEH C18 VanGuard Pre-Column (1.7 μm, 2.1 mm × 5 mm; Waters).

The mobile phase consisted of 5 mM ammonium formate in water (v/v) (A) and MeOH (B). The flow rate was 0.45 mL/min. The following gradient was used: 0–1 min, linear gradient from 5 to 20% B; 1 − 11 min, linear gradient from 20 to 90% B; 11 − 13 min, 90% B; 13–13.5 from 90 to 5% B; 13.5–16 re-equilibration at 5% B. The total run time was 16 min, including the re-equilibration period. The column temperature was set at 40 °C, and the injection volume was 2 µL. The autosampler was kept at 10 °C during the analysis.

Concerning the MS, measurements were performed in ESI positive (ESI + ) ionization mode, and the acquisition was performed in multiple reaction monitoring (MRM) mode. For each target compound, two product ions were recorded. The precursor and product ions, as well as the retention time, for all the investigated pesticides are shown in Supplementary Data 1, together with the CE (collision energy), DP (declustering potential), EP (entrance potential), and CXP (Collision Cell Exit Potential).

The generic MS parameters were as follows: Curtain Gas flow, 20 psi; CAD Gas, 9; IonSpray Voltage 4500 V; Voltage; Temperature, 400 °C; Ion Source Gas 1, 70 psi; Ion Source Gas 2, 40 psi.

Data were acquired by Analyst software (version 1.7.2, SCIEX) and processed by SCIEX OS software (version 2.2.0.5738, SCIEX).

NMR spectroscopy

The formation of the supramolecular structure of the NADES studied in this work has already been reported^24,25. After the preparation of each NADES, the H-bonding was confirmed by NMR, particularly by the appearance of nuclear Overhauser effect (NOE) through the NOESY experiment. In addition, in this study, we aimed to confirm the stability of the eutectic nature of the NADES post-extraction. Therefore, after the extraction of the pesticides, the resulting NADES from each matrix were collected and re-analyzed by NMR.

¹H-NMR and one-dimensional NOESY (1D-NOESY) experiments were carried out using a Bruker Avance III HD 400 MHz spectrometer (proton frequency: 399.59 MHz; magnetic field strength: 9.38 tesla) equipped with a PI HR-BB0400S1 5 mm probe.

100 µL of the NADES phase obtained after the extraction of each sample was transferred into a 2-mL Eppendorf tube, followed by the addition of 500 µL of deuterated chloroform. The mixture was vortexed until fully homogenized, and the resulting solution was placed in a 5-mm NMR tube for direct analysis on the NMR system.

The temperature during all experiments was maintained at a stable 25 °C. For the acquisition of ¹H-NMR and NOESY spectra, 64 K data points were collected across a spectral width of 20 ppm (10,000 Hz). A standard one-pulse sequence with a 30° flip angle and an inter-scan delay of 1.0 s was employed for excitation. Sixteen transients were recorded for the 1H-NMR spectra, while 32 transients were acquired for the NOESY spectra. The NOESY spectra involved selective irradiation of the target proton signal with a mixing time of 0.5 s.

Each sample was shimmed prior to data acquisition using the TopShim automatic shimming method available in the Bruker BioSpin software. The resulting spectra were manually phased, baseline-corrected, and integrated using TOPSPIN (version 3.6.4). The spectral data were subsequently imported and processed with MestReNova software version 14.2 (Mestrelab Research SL, Santiago de Compostela, Spain).

Method validation

The optimized method was validated following the SANTE guideline for pesticide residues in food (SANTE/11312/2021v2). Trueness (mean recovery), precision, limit of quantification (LOQs), linearity range, and matrix effects were assessed. For that, a representative organic sample for each matrix type, previously confirmed to be free of the target compounds, was selected. Six sample test portions of each matrix were spiked at 10, 20, and 100 µg/kg to determine the recovery and precision (repeatability) of the extraction procedure. The linearity was assessed using matrix-matched calibration, at five concentration levels: 1, 2.5, 5, 25, and 50 ng/mL (in vial). The LOQ was defined as the lowest spiked level that can be quantified with acceptable accuracy (recovery between 70 and 120%) and precision (RSD < 20%). Matrix effects were assessed by comparison of the pesticide response in matrix-matched calibration curves and the corresponding responses using calibration curves of standards in solvent.

The NADES-based SLE was also applied to a FAPAS proficiency test sample (spinach matrix) to evaluate its performance under real proficiency testing conditions.

Author contributions

L.C.R. contributed toward experimental design, investigation, data analysis, data interpretation, and writing the original draft. I.A. contributed toward experimental design, data analysis, data interpretation, and supervision. A.G.F., L.R., and H.M. contributed toward data interpretation, conceptualization, and supervision. L.R., I.A., and H.M. contributed to funding acquisition. All authors contributed to reviewing and editing the writing.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Competing interests

L.C.R. is Guest Editor for the Green sample preparation in food science Collection in npj Science of Food. L.C.R. was not involved in the journal’s review of, or decisions related to, this manuscript.

Supplementary information

The online version contains supplementary material available at 10.1038/s41538-026-00717-7.