Solid-Phase Extraction and Large-Volume Sample Stacking-Capillary Electrophoresis for Determination of Tetracycline Residues in Milk

Solid-phase extraction in combination with large-volume sample stacking-capillary electrophoresis (SPE-LVSS-CE) was applied to measure chlortetracycline, doxycycline, oxytetracycline, and tetracycline in milk samples. Under optimal conditions, the proposed method had a linear range of 29 to 200 µg·L−1, with limits of detection ranging from 18.6 to 23.8 µg·L−1 with inter- and intraday repeatabilities < 10% (as a relative standard deviation) in all cases. The enrichment factors obtained were from 50.33 to 70.85 for all the TCs compared with a conventional capillary zone electrophoresis (CZE). This method is adequate to analyze tetracyclines below the most restrictive established maximum residue limits. The proposed method was employed in the analysis of 15 milk samples from different brands. Two of the tested samples were positive for the presence of oxytetracycline with concentrations of 95 and 126 µg·L−1. SPE-LVSS-CE is a robust, easy, and efficient strategy for online preconcentration of tetracycline residues in complex matrices.

Recently, capillary electrophoresis (CE) has received considerable attention in the development of online preconcentration systems such as transient isotachophoresis (tITP) [17], dynamic pH junction [18], sweeping [19,20], and eld-ampli ed stacking. e main advantages of these methods compared to o -line techniques include higher e ciency, shorter analysis time, and lower reagent and sample consumption [21][22][23]. Online preconcentration in CE is based on injection of a larger-than-normal sample volume into the capillary via hydrodynamic or electrokinetic methods [24].
Field-ampli ed stacking was developed for preconcentration of several analytes based on the charges of the analytes. Figure 1 shows a large-volume sample stacking (LVSS) system, which involves a series of polarity switches in CE. e rst step (Figure 1(a)) is hydrodynamic injection of a large sample volume into the capillary. Subsequently (Figure 2(b)), a voltage is applied (reverse polarity) promoting concentration of the analytes and removal of the cationic and nonionic compounds contained in the sample matrix. Finally (Figures 1(c) and 1(d)), analytes are separated in normal polarity in the background electrolyte (BGE) [25,26].
Tetracyclines (TCs) are broad-spectrum antibiotics frequently employed in veterinary medicine for therapeutic purposes [5,27] or incorporated into livestock feed at subtherapeutic doses as growth promoters. However, their indiscriminate use can produce enhanced bacterial resistance, allergic reactions, liver damage, and gastrointestinal issues [28,29].
In order to protect human health from exposure of TC residues in milk, the European Union has established a maximum residue limit (MRL) of 100 µg·kg −1 for chlortetracycline (CT), oxytetracycline (OT), and tetracycline (TC) [30]; the Food and Drug Administration (FDA) has established a MRL of 300 µg·kg −1 for the combined residues CT, OT, and TC [31]; the Codex Alimentarius recommends a limit of 200 µg·kg −1 in milk for the combined residues CT, OT, and TC [32].
In recent years, due to the concerns caused by veterinary drugs contained in food samples, there were developed a large variety of analytical methodologies for the determination of TC residues at μg·kg −1 or μg·L −1 levels in di erent matrices. ese methods included chemiluminescence [33], microbiological assays [34], high-performance liquid chromatography (HPLC) [35,36], or capillary electrophoresis (CE) [37].
Taking into account the MRLs and the complexity of milk, this work develops a CE method using SPE and LVSS-CE for determination of TCs in milk that was demonstrated to be rapid, simple, and e cient. Additionally, the developed method showed higher sensitivity and accuracy than those reported by conventional methods using CZE aimed at the detection and quanti cation of TC residues in milk.

Experimental
2.1. Reagents and Chemicals. All solutions were prepared by dissolving the respective analytical grade reagent in deionized water with a resistivity not less than 18.0 MΩ·cm, which was provided by a Milli-Q system (Millipore, Bedford, MA, USA). Sodium phosphate was obtained from Sigma (St. Louis, MO, USA). EDTA sodium salt, sodium hydroxide, and hydrochloric acid were obtained from J.T. Baker (Phillipsburg, NJ, USA). Methanol was obtained from Mallinckrodt Baker (Xalostoc, Mexico), and 2-propanol was obtained from Fluka (St. Gallen, Switzerland).
Single stock standards of 100 mg·L −1 were prepared in methanol. e stock solutions were stored at −4°C. Mixed standard working solutions were prepared by diluting the standard stock solution immediately before use. e BGE solution consisted of 30 mM sodium phosphate, 2 mM EDTA disodium salt, and 2% 2-propanol. e solution pH was adjusted to 12.0 with 0.1 M·NaOH.

Apparatus.
Electrophoresis was performed using a Beckman Coulter P/ACE 5500 (Fullerton, CA, USA) with a photodiode array detector. Data were collected and analyzed with a Beckman P/ACE system with MDQ version 2.3 software. TC separations were performed in a fused silica capillary (41.7 cm × 75 µm ID). A pH/ion analyzer (model 450; Corning Science Products, NY, USA) was used to accurately adjust the pH of the electrolyte solution to within 0.01 pH units.
At the beginning of each working day, the capillary was activated with 1.0 M NaOH at 35°C for 15 min, followed by 0.1 M NaOH for 10 min, deionized water at 25°C for 10 min, and then electrolyte solution at 25°C for 10 min. e capillary was washed out between successive analyses using 1.0 M NaOH for 4 min, 0.1 M NaOH for 2 min, deionized water for 2 min, and electrolyte solution for 4 min. e detector wavelength ( y ) was set at 360 nm, and the capillary was kept at 25°C. Peaks were identi ed by migration times and coinjection of standard solutions [5].

Sample Treatment and Analysis
. A 1.0 mL milk sample was forti ed with an internal standard (50 µg·L −1 ) in polypropylene tubes. Proteins were precipitated by adding 0.2 mL of 2% acetic acid (v/v), followed by heating for 5 min (65°C) in a water bath and centrifuging at 3200 rpm for 15 min. Once completed, the protein-free liquid phase was diluted to 10 mL with deionized water. e solution was then passed through a cartridge (Sep-Pak Vac C 18 cartridges, 1 g, 6 cc, Waters) previously activated with 5 mL of methanol, followed by 5 mL of methanol and 5 mL of deionized water at a maximum ow rate of 1 mL·min −1 . Analytes retained on the SPE cartridge were washed with 2.0 mL of 5.0% methanol. Retained TCs were eluted with 3.0 mL of methanol. e eluted solution was evaporated to dryness, and the residue was dissolved in 1 mL of 0.01 M NaOH containing 50 µg·L −1 picric acid as an internal standard.
Samples treated by SPE were introduced by hydrodynamic injection at 5 psi for 180 s (around 98% of capillary capacity). e capillary was then set in BGE vials, and a potential of 12 kV was applied for 120 s (reverse polarity) to preconcentrate TCs at the inlet, while water and other ions were removed from the capillary. Finally, polarity was returned to normal (14 kV), and CE separation was carried out.

LVSS Optimization.
Development of an LVSS preconcentration technique for capillary electrophoresis requires optimization of control variables. Optimization involves selection of factors that in uence the analytical signal and enrichment factor. Box-Behnken design (BBD) was selected for optimization because it allows evaluation of control factors using an adjusted surface response. e experimental design matrix describes the combination of factors in each experiment and allows simultaneous evaluation of several variables. Optimization of the system with BBD involves four steps: (i) identifying the output variable to optimize, (ii) identifying and selecting factors and levels that a ect the LVSS system, (iii) data analysis and tting of the surface response model, and (iv) con rmation under the optimal conditions obtained.
For LVSS, the output variable selected is the sum of the enrichment factors of the four TCs. e variables optimized in the procedure were the injection time (min) in the hydrodynamic mode using a pressure of 5 psi, applied potential (kV), and time (min) of reverse polarity. Injection time was varied between 2.0 and 3.0 min to evaluate the time required to ll the capillary. e reverse potential was evaluated between 8.0 and 12 kV. ese values were selected to ensure  su cient stacking time to remove the sample matrix from the capillary without losing analytes. Additionally, time during preconcentration (2.0-3.0 min) must be su cient to increase analyte enrichment. Table 1 shows the design matrix produced and the output variable in function of the sum of each enrichment factor obtained in each condition. All experiments were performed using 1 mL of a standard solution of TCs at a concentration of 1.0 mg·L −1 . Enrichment factors were estimated as the area ratio of the signals obtained with and without application of online LVSS.
Results were analyzed using MINITAB ® version 17 software. Data were adjusted to the quadratic model according to the analysis of variance (ANOVA). e coe cient of determination (r 2 ) for the adjustment was 0.785, and the equation for the surface response was where Y1 is the sum of the enrichment factor, X1 is the injection time (min), X2 is the inversion electric current (kV), and X3 is the applied time in the inversion electric current (min). e critical variables during LVSS are the reverse potential and applied time (p > 0.05). e lack-of-t test is designed to determine if the proposed model is adequate for the observed data. e test is performed by comparing the variability of residuals from observations at replicate settings of the factors. Since the p value for lack of t in the ANOVA table (0.744) is greater than 0.05, the model is adequate for the observed data at the 95.0% con dence level.
Based on the response surfaces (Figure 2), a clear interaction between the variables is observed, which is commonly observed for preconcentration systems employing LVSS-CE. Optimal conditions determined by BBD were X1: injection time (3.0 min), X2: reverse potential (12 kV), and X3: preconcentration time (2.0 min). e proposed methodology (LVSS-CE) was applied for the determination of TCs in commercial milk samples using a modi cation of the method proposed by Islas et al. [4]. However, di erent electrophoretic mobilities were obtained for the internal standard, which can be attributed to the ionic strength of the sample. Ionic strength signicantly increases the electrophoretic mobility of analytes, thereby a ecting LVSS preconcentration and causing loss of analyte if care is not taken when applying the negative polarity [38].
For these reasons and given the complexity of the sample, one of the most important steps in LVSS-CE analysis is sample cleanup. However, this may be di cult for analysis of antibiotics. For these reasons, an extraction and cleanup step was used previous to preconcentration and analysis by LVSS-CE. SPE was used for extraction and cleanup of TCs in milk samples. is technique decreases ionic strength e ects, making samples suitable for analysis by LVSS-CE. For sample pretreatment, following protein removal from the milk sample, the liquid phase is diluted to 10.0 mL with deionized water and then passed through an activated C 18 SPE cartridge. Analytes retained on the SPE cartridge were washed with 2.0 mL of 5.0% methanol. Retained TCs were eluted with 3.0 mL of methanol. e eluted solution was evaporated to dryness and redissolved in 1.0 mL of 0.01 M NaOH containing 50 µg·L −1 picric acid [39].

Analytical Parameters.
Under optimal conditions, analytical parameters of the LVSS-CE method were evaluated at concentrations of 0-200 µg·L −1 for each TC. Each standard was prepared and analyzed in triplicate using the proposed methodology. Peak areas were measured, and calibration curves were constructed from the peak area ratios (analyte : internal standard). Calibration curves showed  a linear dependence on TC concentration. Calibration regression parameters are shown in Table 2. LODs were calculated for a signal-to-noise ratio of 3.29 according to IUPAC recommendations [40]. e accuracy and precision of the method proposed was measured in terms of intra-and interday repeatabilities for migration times and peak areas. Results were determined as the relative standard deviation (%RSD) obtained in the analysis of TCs at two concentrations (75 and 150 µg·L −1 ). Based on these results and using the most restrictive MRLs established by EU regulations, the LVSS is adequate for analysis of TCs in milk samples.

Application.
e proposed SPE-LVSS-CE method was applied for the determination of TCs in 15 commercial milk samples from di erent brands. ree replicate determinations of each analyte in the selected samples were performed. Two samples were determined to be positive for the presence of OT with concentrations of 95 and 126 µg·L −1 , respectively, which was identi ed by their migration times. In order to con rm the presence of the analyte, a standard addition was made to the sample extract. An increase in the peak area conrmed the presence of the antibiotic residue. Samples with TC concentrations outside the linear response range were diluted tenfold with deionized water. Con rmation using mass spectrometry is also required. e electropherograms obtained are shown in Figure 3.

Conclusions
e proposed SPE-LVSS-CE technique provided sensitive, rapid, simple, and e cient online preconcentration of TC residues in complex matrices such as milk. is methodology only required 1.0 mL of milk, whereas traditional methods require about 100.0 mL to reach the MRLs established by international regulations.
Additionally, this technique provides good sensitivity and accuracy compared to CZE and has a much higher

Conflicts of Interest
e authors declare that they have no con icts of interest.