Accurate determination of pesticides, hormones and endocrine disruptor compounds in complex environmental samples using matrix dilution and matrix matching with dispersive liquid–liquid microextraction

Reagents

4-n-Octylphenol, diazinon, estrone, aldrin, bisphenol-A, 4-n-nonylphenol, 17-β-estradiol, cis-chlordane, heptachlor, endosulfan α-β and dieldrin standards were all obtained from Dr. Ehrenstorfer (Augsberg, Germany). A mixed standard stock solution (80 mg L⁻¹) of these analytes was prepared in acetonitrile and appropriate aliquots diluted to prepare aqueous working standard solutions. Methanol, chloroform, potassium iodide and acetonitrile all obtained from Merck (Germany) were used in the extraction processes.

Instrument

Separation, identification and quantitative determination of analytes was achieved with an Agilent 6890 gas chromatography system (HP-5MS column – 30 m; 250 μm; 0.25 μm) coupled to a mass selective detector. Chromatographic data was obtained in the scan mode and peak area integration of each analyte was done with the most prominent ion (m/z). All injections (1.0 μL) were carried out in the splitless mode and helium gas was kept at a constant flow rate of 1.8 mL min⁻¹. Temperatures of the injection port and detector transfer line were 250°C and 280°C, respectively. The oven was programmed to increase from an initial temperature of 70°C to 180°C, 210°C and 290°C at rates of 60, 4.0 and 40°C min⁻¹, respectively. Other apparatus used in this study were vortex (ISOLAB – M1010002), centrifuge (Hettich Zentrifugen – EBA 20) and mechanical shaker (Kermanlab – 51).

Procedure

Optimized extraction conditions of a previous study was adopted in this work [22]. Potassium iodide was added to 8.0 mL sample solutions and vortexed for about 10 s to facilitate dissolution. A mixture of chloroform (200 μL) and methanol (2.0 mL) was injected into the sample solution and hand shaken for 30 s. The solution was then centrifuged for 2.0 min at 3461 g and a sufficient amount withdrawn from the bottom chloroform phase into insert vials for GC-MS determination. For the soil sample, 1.0 g was weighed and washed into a 15 mL centrifuge tube with 10 mL acetonitrile and then agitated for 10 min on a mechanical shaker. Soil and solvent phase separation was quickened by centrifugation at 3461 g for 2.0 min and then filtered to obtain a clean soil extract. Matrix matching calibration standards were prepared by diluting appropriate stock standard solutions with samples of a similar matrix composition after blank determinations showed analytes were not present. Matrix dilution of soil sample extracts was performed to reduce matrix effects and allow application of dispersive liquid–liquid microextraction.

Samples

Surface soil was sampled in the university campus and crushed into small particles using a laboratory scale ceramic mortar and pestle. The crushed soil was then passed through a 100 μm sieve to obtain very fine particles. Prior to each analysis, weighed soil samples were kept in an oven below 70°C for 1.0 h to ensure complete dryness. In the sampling of water samples, plastic bottles were thoroughly rinsed with the samples before being filled to the brim. The water samples included two sea water (Turkey), two well water (Turkey), three tap water (one from North America and two from Turkey), and three lake water (Turkey) samples. All samples were successively filtered through 125 mm paper and 0.45 μm RC filter to remove particulate materials. The water samples were stored in cabinets under ambient laboratory conditions and analyzed within 1 month of sampling.

Effect of experimental process on analytes

Sample preparation is a very crucial step in the determination of analytes because it can alter the accuracy, precision and sensitivity of the method being used. Errors recorded in analytical analysis mainly originate from sample preparation processes [16]. The loss of analytes during sample preparation is a major concern and as such, the effect of all processes used in this study was tested. The processes tested were filtration through 125 mm and 0.45 μm pores, vortexing, mechanical shaking, and centrifugation. A 2.0 ng mL⁻¹ standard solution was taken through 10 s of vortexing, 10 min of mechanical shaking, 2.0 min of centrifugation at 3461 g and successively filtered through 125 mm and 0.45 μm pore sized filter papers into GC vials. In comparison to this, a 2.0 ng mL⁻¹ standard solution was directly taken into a GC vial for triplicate measurements. The standard solution taken through the processes above recorded slightly higher peak area values. This could be a result of minimal solvent evaporation and a subsequent increase in concentration. The percentage difference between treated and untreated standard solutions (Fig. 1) was less than 4.0% for all samples, except for bisphenol A which was about 6.4% based on the mean value. These percentage differences were not statistically significant to affect the overall results of the method.

Fig. 1:

A chart of average peak area values showing closeness in results for treated and untreated standards, n=3 standard deviation (* and ** represent 2 times scaled down and 2–5 times scaled up results, respectively).

Tap water

Contaminants are rarely transmitted from water treatment facilities to households but leakages in the distribution lines near contaminated land sites can result in contaminated tap water. A more common mode of contamination is the leaching of pipe materials over time into the water stream [23]. Bisphenol A, nonylphenol and octylphenol can be found in plastic and polymer compounds used in pipes and other materials. Old pipelines are prone to corrosion and leaching of several contaminants into water. The tap water samples, representing three geographical regions as Europe, Asia and North America were sampled and analyzed under the optimum experimental conditions. None of the 12 analytes were detected in the tap water samples, and they were therefore spiked at the three different concentrations. Percent recoveries obtained ranged between 76 and 93% for all spiked concentrations, with the lowest recoveries being recorded for 10 ng mL⁻¹ spiked samples. The presence of some interferants could have slightly hindered the extraction of analytes from the tap water samples. Figure 2 is an overlay chromatogram of 50 ng mL⁻¹ aqueous standard and tap water samples spiked at 50 ng mL⁻¹ for diazinon. The signals of the three samples can be seen to be lower than the signal of the standard solution. The tap water sample from North America recorded peak area values that were between the values of the other two samples and was therefore used to prepare matrix matched calibration standards. The percent recoveries calculated using the matched standard ranged between 96–104, 94–106 and 95–106% for 10, 20 and 50 ng mL⁻¹ spiked samples, respectively, as presented in Table 2.

Fig. 2:

An overlay chromatogram of 50 ng mL⁻¹ aqueous standard and tap water samples spiked at 50 ng mL⁻¹ for 4-n-octylphenol (m/z 107).

Fresh water samples

The type of aquifer of a water body determines the mineral content of the water, where those with high calcium and magnesium content are generally termed hard water which has been reported to hinder the extraction efficiency of analytes [24], thus, low analyte recovery affects accurate analyte quantification. The lake and well water samples were spiked after blank determinations, and their respective percent recoveries were determined with deionized water standards. The percent recoveries calculated for the analytes in the two well water samples ranged between 75 and 116%, but using one well sample as standard relative to the other gave results between 90 and 103% as shown in Table 3. Two of the lake water samples are fresh water but the third lake water sample is a brackish water, which has salinity levels between sea water and fresh water. Its high salt content makes it inconvenient for drinking or irrigation but supports only limited aquatic life such as darekh (herring species). Presented in Fig. 3 is an overlay chromatogram of 4-n-nonylphenol for the three lake water samples spiked at 50 ng mL⁻¹. The high salinity content (29.2 mS/cm) of the third lake sample can be seen to have lowered the extraction efficiency of 4-n-nonylphenol and similarly the other analytes. The salinity level determined for the two fresh lake water samples based of based on electrical conductivity was 6.29 and 3.7 mS/cm. The closeness in results of the fresh lake samples makes them suitable for matrix matching as presented in Table 3.

Fig. 3:

An overlay chromatogram of 50 ng mL⁻¹ aqueous standard lake water samples spiked at 50 ng mL⁻¹ for 4-n-nonylphenol (m/z 107).

Sea water

Salinity and density of matrix are two important properties affecting the mass transfer of analytes from aqueous solution into extraction solvents. Analyte mass transfer in hindered by high density, resulting in low extraction efficiencies. Salinity on the other hand when in the right amount could lower analyte solubility and enhance its migration into extraction solvent, and also facilitate phase separation [25]. Two sea samples were sampled from different regions in İstanbul. The sampling depth was about one meter for both samples. The pH and density determined for sea sample A were 7.39 (at 24°C) and 1.026 g mL⁻¹, respectively, and 7.45 (at 24°C) and 1.019 g mL⁻¹, respectively for sea sample B.

The sea samples were spiked to 10, 20 and 50 ng mL⁻¹ after analytes were not detected in blank determinations. Due to the wide applicability of alkylphenols (nonylphenol and octylphenol) and BPA, they are commonly detected in industrial waste sites and sea water [26]. Special attention was given to the blanks of these analytes due to their possible presence in the plastic tubes used. Potassium iodide was not added to the sea samples due to their natural salinity content, but the aqueous standard solutions were extracted with the addition of salt according to the optimized method. Integrated peak area values of both sea water samples were lower than their aqueous standard analogues of similar concentrations. The effect of potassium iodide was therefore investigated by adding 0.50, 1.0 and 2.0 g to both samples spiked at 50 ng mL⁻¹. The extraction efficiencies decreased with increasing salt amount and average percentage differences calculated between saltless extraction, and 0.50, 1.0 and 2.0 g were 14, 25 and 39%, respectively. The decrease in extraction output for the salt added samples might be due to saturated sea samples hindering analytes movement into the extraction solvent.

The percent recoveries recorded for the analytes were similar at all concentrations, having less than 7% percent difference. The lowest percent recovery (67%) was obtained for Endosulfan α as presented in Table 4. Sea sample B, with lower salinity content and density had a relatively higher percent recovery and was thus used as a matching standard to calculate percent recoveries of the sea sample A. The matrix matched percent recoveries obtained were between 90 and 110% as presented in Table 4. The increased percent recovery signifies an increase in accuracy for quantitative determinations.

Soil

Soil is a typical environmental sample having complex matrix. This complexity is a result of its composition which includes minerals, gases, liquids, organic matter, and several living organisms that help maintain the balance of the ecosystem [27]. Components of soil can either suppress the determination of an analyte or boost its output, both a misinterpretation of the true amount. Some pesticides and other chemicals having very low solubility in water do not easily leach down the layers of soil into the water table. Nonporous soils also tend to retain chemicals for longer periods. Acetonitrile is an appropriate solvent for most pesticides and was therefore used as extraction solvent to extract 1.0 g soil samples spiked at 10, 20 and 50 ng mL⁻¹ as final concentration in the extraction solvent.

Due to the miscibility of chloroform and acetonitrile, the extraction method could not be directly applied to the soil extract. Appropriate aliquots from extracted soil sample were diluted 4.0, 8.0 and 16 folds in deionized water. The dilution facilitated application of dispersive liquid–liquid microextraction, and at the same time diluted the soil extract for less matrix effects. High dilution factors decrease or eliminate matrix effects but it also pose the risk of diluting analytes below their detection limits. The 8.0 fold dilution presented the best extraction output for all 12 analytes. The 4.0 fold dilution was expected to give the highest extraction output due to its lower dilution factor, but the amount of acetonitrile in the diluted sample turned out to be too high. This resulted in nearly complete loss of chloroform after extraction and the settled phase could not be taken for analysis. Acetonitrile on its own is a good dispersive solvent and its high amount resulted in over dispersion of chloroform. Using the 8.0 folds dilution, spiked samples were compared to their calibration analogues prepared in deionized water. The spiked soil samples recorded very high signals and their percentage recoveries were accordingly higher than 120%. However, blank determination of the soil sample did not record a signal for all analytes. The amount of acetonitrile in the diluted soil sample clearly increased the extraction efficiency of analytes with respect to aqueous standards. To overcome this effect, 1.0 g of unspiked soil sample was taken through the same extraction procedure performed for spiked samples. Equivalent amounts from the blank soil extracts was then used to prepare calibration standards. This resulted in the matching of both acetonitrile amount (12.5%) and soil matrix. The percent recoveries recorded with this method therefore was close to 100% for all analytes as shown in Table 5.