Dispersive liquid-liquid microextraction for the isolation and HPLC-DAD determination of three major capsaicinoids in Capsicum annuum L.

Dispersive liquid-liquid microextraction (DLLME) was combined with high-performance liquid chromatography-diode-array detector (HPLC-DAD) for the extraction and quantitation of three major capsaicinoids (i.e. capsaicin, dihydrocapsaicin and nordihydrocapsaicin) from pepper (Capsicum annuum L.). Chloroform (extraction solvent, 100 μL), acetonitrile (disperser solvent, 1250 μL) and 30 s extraction time were found optimum. The analytes were back-extracted into 300 μL of 50 mM sodium hydroxide/ methanol, 45/55% (v/v), within 15 s before being injected into the instrument. Enrichment factors ranged from 3.3 to 14.7 and limits of detection from 5.0 to 15.0 µg g-1. Coefficients of determination (R2) and %RSD were higher than 0.9962 and lower than 7.5%, respectively. The proposed method was efficiently applied for the extraction and quantitation of the three capsaicinoids in six cultivars of Capsicum annuum L. with percentage relative recoveries in the range of 92.0%–108.0%. DLLME was also scaled up for the isolation of the three major capsaicinoids providing purity greater than 98.0% as confirmed by liquid chromatography-mass spectrometry (LC-MS) and nuclear magnetic resonance (NMR) analysis, which significantly reduced the extraction time and organic solvent consumption.

achieve rapid, highly efficient and highly selective separation of food matrices without an efficient sample preparation step [17] owing to the high complexity of such samples. The drawbacks of conventional sample preparation techniques such as liquid-liquid extraction (LLE) and solid-phase extraction (SPE) are well known and documented in the literature, some worth mentioning are the tedious procedures and large consumption of toxic organic solvents involved in LLE, which are harmful to the researcher, living organisms and the environment. SPE uses smaller volumes of organic solvents, but it is still considered significant. In addition, SPE cartridges are cost-prohibitive and generate chemical waste [18].
The emphasis has recently been diverted towards the design of reliable miniaturized sample preparation techniques. Dispersive liquid-liquid microextraction (DLLME) [19] has achieved broad acceptance and recognition in food analysis [20][21][22] due to its rapidness, environmental friendliness, high extraction efficiency and affordability. In DLLME, the maximum extraction efficiency is maintained almost instantly thanks to the extremely large surface area of the extraction solvent available for the analyte. Although gas chromatography (GC) was the first instrument to be used with DLLME [19], in which the extract can be directly injected due to the compatibility of the final extract with the instrument, HPLC is so far, the most frequently used one after DLLME [23]. Evaporating the solvent to dryness and reconstituting in the mobile phase or back-extraction into a compatible solvent are common sample introduction methods of the final extract into reversed-phase HPLC after DLLME [24]. Back-extraction of ionizable compounds is a faster choice that can also provide additional selectivity, improved sample clean up and better regulation of the ionic strength in the final extract.
The objective of this research was to design a fast, robust and efficient extraction method using DLLME followed by back-extraction for preconcentration and determination of the three major capsaicinoids (i.e. NDHC, CAP and DHC) from different cultivars of Capsicum annuum L. using HPLC-diode-array detector (DAD). To date, this is the first study on DLLME of capsaicinoids from pepper samples. Influential experimental parameters on the extraction and backextraction efficiencies were optimized. This was also the first attempt to scale up DLLME for the isolation of the three major capsaicinoids from Capsicum annuum L.

Chemicals and reagents
Analytical grade standards of CAP (logP = 3.75), DHC (logP = 4.11) and NDHC (logP = 3.67) were isolated on a preparative scale with purity above 98% from Capsicum annuum L. in our pharmacognosy laboratory, as confirmed by LC-MS and NMR analysis (supplementary material). The chemical structures of the three analytes are shown in Figure 1. HPLC-grade methanol, acetonitrile, tetrahydrofuran, chloroform, dichloromethane, sodium chloride, sodium hydroxide (NaOH) and ethyl acetate were acquired from Sigma-Aldrich (Darmstadt, Germany). Acetic acid, n-hexane and toluene were obtained from Honeywell Riedel-de Haën (Seelze, Germany). Silica gel was obtained from Merck (Søborg, Denmark). All reagents used were of analyticalgrade. Deionized (DI) water (18.2 MΩ.cm), treated with Purelab Ultra Analytic (ELGA LabWater, Leicestershire, UK), was used for the preparation of aqueous solutions. MarvinSketch (Version 20.11.0, ChemAxon, Boston, MA, USA) was used for logP and pK a calculations.

Standard solutions of capsaicinoids
Individual standards of the three capsaicinoids were prepared at 1000 µg mL -1 in methanol and stored at -4 °C. Mixed standard solutions were freshly prepared by diluting the stock solution with 50 mM sodium hydroxide/ methanol (45/55%, v/v), hereafter, referred to as the back-extraction solution (BES). Mobile phases and aqueous solutions were filtered using a vacuum solvent filtration system through 0.20 µm regenerated cellulose-membrane filters (Isolab, Eschau, Germany), whereas, samples were filtered through 0.22 µm nylon syringe filters (Microlab Scientific, Shanghai, China).

Sample preparation
Fresh cultivars of Capsicum annuum L. and green pepper pickles were collected from local markets in Nicosia, TRNC. The samples were washed with plenty of DI water, left to dry at room temperature, and cut into small pieces with a stainless steel knife. Then, they were dried overnight in the oven at 60 °C. Dry samples were powdered using a blender with stainless steel blades and stored in well-sealed glass bottles in a dark place till analysis. An initial solid-liquid extraction (SLE) was carried out by weighing 1.00 ± 0.01 g of the sample and extracting it with 50.0 mL of 50/50 (%, v/v) acetonitrile/ water through ultrasonication for 30 min at 25 °C. The mixture was filtered through a cotton wool and a 0.45 µL syringe filter and the filtrate was transferred into a 50 mL volumetric flask and made up to the mark with 50/50 (%, v/v) acetonitrile/water (referred to hereafter as the sample solution). A portion of 5.0 mL of the sample solution was mixed with 2.0 mL of saturated sodium chloride solution. The mixture was vortexed for 30 s and centrifuged for 3 min at 6000 rpm to induce salting-out extraction (SOE). Out of the 1.2 mL of acetonitrile that salted out, 1.0 mL was used in DLLME.

Dispersive liquid-liquid microextraction
One milliliter of acetonitrile, resulting from SOE, was completed to 10.0 mL with DI water into a 15 mL centrifuge tube after adding an extra volume of 250 μL of acetonitrile, which acted as the disperser solvent in DLLME. Upon the addition of chloroform and acetic acid (100 µL, each), the mixture was vortex mixed for 30 s and centrifuged (3 min at 6000 rpm). Chloroform, which settled to the bottom, was transferred completely into a snaplock microtube and the analytes were back-extracted into 300 µL of BES through vortexing for 15 s and centrifugation (3 min, 6000 rpm). A portion of the aqueous extract (i.e. 5 µL) from the upper layer was injected into HPLC-DAD for analysis.

Optimization of extraction conditions 3.1.1. Type and volume of the extraction solvent
Selecting a suitable extraction solvent for DLLME is based on its ability to extract the analyte(s) from the sample solution and its immiscibility with the aqueous phase [25]. In conventional DLLME, extraction solvents denser than water, such as chloroform, dichloromethane and tetrachloromethane, are commonly used [19]. However, the use of less dense solvents has also been proposed, which include toluene, n-hexane, 1-undecanol and 2-dodecanol [26]. Chloroform, dichloromethane, toluene and n-hexane were tested as the extraction solvents considering extractability and solvent recovery. Although dichloromethane gave the highest extraction efficiency (Figure 2a), the recovered volume of chloroform (Figure 2b) was much higher (i.e. 170 ± 5 µL), making it an optimum extraction solvent to proceed with. Volume of the selected solvent can significantly influence the extraction efficiency [21]. The effect of chloroform volume was monitored starting from 100 to 300 μL within 50 μL intervals. Peak areas rapidly decreased as the volume of chloroform was increased due to dilution (Figure 2c). An optimum volume would ensure the highest enrichment factor (EF) and high solvent recovery after extraction [27]. When lower volumes of chloroform below 100 μL were tested, phase separation was not observed. Hence, 100 μL of chloroform was taken as optimum for subsequent experiments.

Type and volume of the disperser solvent
Due to their high miscibility with the extraction solvent, acetonitrile, methanol, acetone and ethanol are common disperser solvents in DLLME [23]. Since SOE was used prior to DLLME, it was decided to use acetonitrile as the disperser solvent being the most suitable solvent for SOE [28]. Volume of the disperser solvent has a significant impact on the formation of a cloudy solution, which in turn determines the extraction efficiency [23]. The effect of this parameter was investigated by using 1000 μL of the salted-out acetonitrile and adding increasing volumes of pure acetonitrile. The optimum volume of acetonitrile was found as 1250 μL (Figure 2d), beyond which peak areas decreased due to increased miscibility and solubility of chloroform and the analytes in the aqueous phase, respectively.

Ionic strength and extraction time
In LLE, the addition of inorganic salts that increase the ionic strength has been reported to reduce the solubility of CAP in aqueous solutions [29]. However, in DLLME, such salts can also act as demulsifiers, which can eventually lead to faster phase separation, yet with reduced extraction efficiency [30]. The effect of ionic strength was evaluated by adding increasing concentrations of sodium chloride to the sample solution. It was observed that its addition caused a negative effect on the extraction efficiency (Figure 2e). Therefore, no salt was added to the sample solution in subsequent experiments. Extraction, or vortex, time corresponded to the period between the injection of the mixture of disperser/extraction solvent into the sample solution to the moment the mixture was centrifuged [19]. The effect of vortex time was monitored from 0 to 120 s. It was noticed that equilibrium was achieved within 30 s such that this value was considered optimum (Figure 2f).  c. Coefficient of determination. d. Percentage relative standard deviation, n = 3. e. Limit of detection (µg mL -1 ). f. Limit of quantitation (µg mL -1 ). g. Linear dynamic range (µg mL -1 ). h. Enrichment factor: ratio of calibration slope with DLLME-HPLC-DAD to that with external aqueous calibration (conventional HPLC-DAD).

Optimization of back-extraction conditions
Chloroform, obtained from DLLME was incompatible with reversed-phase HPLC. Hence, back-extraction of the analytes into an aqueous solution facilitated injection of the BES into the instrument. Impact of the BES volume on extraction efficiency was assessed starting from 100 to 400 μL. Extraction performance improved with an increase of up to 300 μL BES volume (Figure 2g), beyond which it started to decrease due to dilution of the analytes in excess volume of the BES. Thus, 300 μL was considered as an optimum BES volume. The effect of back-extraction time was evaluated within the range of 0 s to 60 s. Equilibrium was achieved within 15 s only due to microemulsion formation ( Figure 2h). Hence, the optimum back-extraction time was taken as 15 s.

Analytical performance
Both external aqueous and standard-addition calibration graphs were plotted as a function of peak area versus concentration of the three capsaicinoids under optimized conditions to test the analytical performance. External aqueous calibrations were constructed by dissolving appropriate amounts of the analytes in the BES at the concentration range of 25-150 µg mL -1 , whereas, standard-addition calibrations were obtained by spiking the pepper samples, prior to SLE, with the standards at concentrations within the range of 2.5-15 µg mL -1 and applying DLLME followed by the back-extraction step.
A good linearity was achieved as reflected by coefficients of determination (R 2 ) in the range of 0.9962 to 0.9999 (Table 1). Intra-and interday precision, expressed as percentage relative standard deviation (%RSD), was found to be less than 6.1 and 7.5%, respectively, indicating good repeatability. Limits of detection (LOD), ranged between 0.1 and 0.3 µg mL -1 (or 5.0 and 15.0 µg g -1 ). Limits of quantitation (LOQ) ranged from 0.3 to 1.0 µg mL -1 (or 15.0 and 50.0 µg g -1 ). Linear dynamic range (LDR) extended from the LOQ to 15.0 µg mL -1 (0.75 mg g -1 ). The method's EF is a measure of how sensitive this method is. The higher the EF, the more sensitive it is. EFs ranged from 3.3 to 14.7, which are significant considering the high complexity of the food samples analyzed. EFs ranged between 3.3 and 14.7 times. The proposed method's figures merit of are summarized in Table 1.

Recovery studies and matrix effect
Applicability, recovery and matrix effect were evaluated by analyzing six pepper samples for their capsaicinoid content. Typical chromatograms, shown in Figure 3, show the absence of extraneous peaks from the samples at the retention times of the three capsaicinoids, confirming high selectivity of the method. Percentage relative recoveries (%RR) obtained fro   spiked samples at two concentration levels (i.e. 2.5 and 5.0 µg mL -1 ) were calculated to be in the range of 92.0-108.0 (Table  2). Possible matrix effect was assessed from the slopes of the calibration graphs for the three analytes and the samples (Table 1) and the p-value calculated using a single-factor analysis of variance (ANOVA). The difference was statistically significant (P < 0.05), suggesting the presence of matrix effect due to the wide variety of the studied samples. The standardaddition method was therefore appropriate to suppress this effect. 3.5. Comparison with other methods DLLME-HPLC-DAD was contrasted with other extraction methods used to analyze capsicum samples in terms of parameters like extraction time, consumption of organic solvent used per sample, LOD, LOQ, R 2 and %RSD. The shortest extraction time (i.e. 0.75 min) and the least volume of organic solvent (i.e. 1.9 mL) are the key benefits of this approach over the others (Table 3), with the exception of the pressurized hot-water extraction method (PHWE) suggested by Bajer et al. [31], where water was used as the extraction solvent. Nevertheless, PHWE required special instrumentation, incurring extra capital cost. Supercritical fluid extraction (SFE) proposed by Santos et al. [32] used carbon dioxide as a 'green' extraction solvent, however, the final extract had to be conditioned with 10 mL of methanol before injection into HPLC. The lower LODs and LOQs reported by Bajer et al. [31] and Barbero et al. [33] compared to this study were largely due to the higher sensitivity of the MS detector used in their method. MS detectors are more sensitive than UV, but they are more complicated and expensive.

Conclusion
DLLME, combined with a back-extraction step, was proposed prior to HPLC-DAD for the extraction of the three major capsaicinoids from different cultivars of pepper (Capsicum annuum L.) samples. Several advantages of this method as compared to other existing extraction methods include cost-effectiveness, rapidness, high extraction efficiency, good selectivity and sample clean up as well as minimum consumption of organic solvents. Despite the complexity of the studied matrix, high recoveries, good linearity, reproducibility and interference-free chromatograms were attained. These advantages of DLLME-HPLC-DAD encourages its use for the quantitation of capsaicinoids in pepper samples in routine food analysis and quality control laboratories. The ability to scale-up DLLME for the isolation of capsaicinoids with high purity suggests that this method be applied for the isolation of various intermediate-to low-polarity compounds from natural products, which reduces the extraction time and organic solvent consumption, satisfying the criteria for green chemistry.