Article pubs.acs.org/JAFC
Development and Evaluation of a Molecularly Imprinted Polymer for the Detection and Cleanup of Benzylpenicillin in Milk Geert Van Royen,*,† Peter Dubruel,‡ and Els Daeseleire† †
Technology and Food Science Unit, Institute for Agricultural and Fisheries Research (ILVO), Government of Flanders, Brusselsesteenweg 370, B-9090 Melle, Belgium ‡ Department of Organic Chemistry, Department of Polymer Chemistry, and Biomaterials Research Group (PBM), Faculty of Sciences, Ghent University, Krijgslaan 281 S4, B-9000 Gent, Belgium ABSTRACT: A molecularly imprinted polymer (MIP) was designed for benzylpenicillin via suspension polymerization. The specific absorption of benzylpenicillin to the MIP, applied in a molecularly imprinted solid-phase extraction (MISPE), was compared to the nonspecific binding using a NIP (nonimprinted polymer without a target molecule) in a non-molecularly imprinted solid-phase extraction. This validation was performed successfully in acetonitrile solutions and milk extracts spiked with benzylpenicillin. Significant differences in absorption were observed. In acetonitrile, the recoveries using MISPE (90−95%) were a fraction higher than those in milk extracts (70−80%). The validation revealed the limit of detection and the limit of quantitation for the MISPE application in milk samples to be 0.51 and 1.02 μg/kg, respectively. In addition, comparing the results of the analysis of positive milk samples using MISPE with those using a classic sample preparation step showed a Pearson correlation of 0.989. Finally, cross reactivity tests using other antibiotics showed a certain cross reactivity, but non-β-lactams were barely bound. KEYWORDS: molecularly imprinted polymer, benzylpenicillin, milk, MISPE, validation
■
INTRODUCTION β-Lactam antibiotics are frequently used to treat various bacterial infections in humans and cattle. The continuous application of usually small doses in cattle stimulates the development of resistance. This has led to the development of new resistant strains that can be passed from animals to humans and therefore pose a threat to human health.1,2 Because of the essential role that β-lactam antibiotics play in the dairy industry, their use is expected to continue on a massive scale. The risk of increased resistance is not the only problem, however: residues in dairy products can cause serious allergic reactions in the consumer and can also have inhibitory effects on dairy cultures. A Flemish study covering 181 antibiotic-positive milk samples showed that almost 90% of the antibiotics detected were of the β-lactam group.3 An effective tool for detecting β-lactam antibiotics would therefore be highly useful. The purpose of this study was to develop such a tool based on a molecularly imprinted polymer (MIP) designed for benzylpenicillin. At present, immunoaffinity columns (IACs) are regarded as the most specific cleanup method in the field of residue analysis. The most important disadvantage of IACs is the (un)availability of antibodies. The production of antibodies is a long and expensive process with only uncertain chances of success. Commercially available IACs are furthermore often limited to a single use, which implies that the cost of an IAC analysis is (too) high.4 In contrast, MIPs are easier, cheaper, and faster to produce. They are more stable under extreme conditions (e.g., organic solvents, acid/base solutions, high temperatures, etc.) and therefore have a longer shelf life. Finally, it is possible to use them several times.5 These properties make MIPs very suitable © 2014 American Chemical Society
for use in sensor technology, cleanup of samples, separation applications, etc.6 MIPs are composed of several components, including a template molecule, a monomer, a cross-linker, a solvent, and an initiator. They are prepared following several steps.7 First, the selected template molecule is mixed with the monomers, and a complex is formed between template molecules and monomers. This complex is polymerized through cross-linkers in an inert solvent to form a rigid polymer. Lastly, the template molecule is eliminated by rinsing. This results in the formation of a polymer with cavities that are the same shape and size as those of the template molecule. MIPs are available in several forms, e.g., as bulk polymers, monolithic polymers, surface-embedded polymers, and nanoparticle polymers.8 Bulk polymers are prepared through mass polymerization, while the nanoparticle MIPs are produced through suspension polymerization. The advantages of the latter procedure include higher production yields, a shorter production time (no grinding and sieving required), and the highly reproducible production of very regular shapes.9 The production of MIPs for several antibiotics has been described previously. For benzylpenicillin, MIP development was performed through mass polymerization using methacrylic acid as a monomer and trimethylol propane trimethacrylate as a cross-linker.10 For sulfamethazine, a molecule from the sulfonamide group, solid-phase extraction (SPE) MIPs were generated for detection in milk by means of voltammetry.11,12 Received: Revised: Accepted: Published: 8814
February 24, 2014 July 15, 2014 July 17, 2014 July 17, 2014 dx.doi.org/10.1021/jf502331h | J. Agric. Food Chem. 2014, 62, 8814−8821
Journal of Agricultural and Food Chemistry
Article
spectrometer (Micromass, Altrincham, U.K.) equipped with a zspray electrospray ion interface or an Acquity Ultra Performance liquid chromatograph (Waters) coupled to a Xevo TQ mass spectrometer (Waters) also equipped with a z-spray system. The mass spectrometers were fully controlled by MASSLYNX version 4.0 or 4.1 (Waters). The separations on the LC system were conducted on an Xterra MS C18 column [Waters, 5 μm, 150 mm × 2.1 mm (inside diameter)] protected with an Alltima C18 guard column [Alltech, Deerfield, IL, 5 μm, 7.5 mm × 2.1 mm inside diameter)]. The ultra performance liquid chromatography (UPLC) separations were performed on a Waters UPLC BEH C18 column [Waters, 1.7 μm, 100 mm × 2.1 mm (inside diameter)]. Synthesis of the MIP for Benzylpenicillin. Benzylpenicillin (2 mmol or 0.7 g), methacrylic acid (16 mmol or 1.4 g), trimethylolpropane methacrylate (7 mmol or 2.4 g) or ethylene glycol dimethacrylate (7 mmol or 1.4 g), and azobis(isobutyronitrile) (0.4 mmol or 0.07 g) were dissolved in 6 mL of AcN (bulk polymerization) or in 70 mL of AcN (suspension polymerization). These solutions were polymerized for 20 h in a water bath at 75 °C under nitrogen saturation. After the MIP had been dried in an oven at 60 °C for 48 h, the template molecule was washed out with a MeOH/H2O mixture [95/5 (v/v)] according to the Soxhlet principle. Equilibrium Binding Experiments in Acetonitrile or Milk Extracts. Ten milligrams of MIP or NIP was incubated on a shaker for 4 h in 1 mL of AcN or milk extract containing different amounts of benzylpenicillin. For the binding experiments in AcN, the following protocol was used. First, the solution was centrifuged for 15 min at 4302g, after which 500 μL of the supernatant was dried at 60 °C under a flow of nitrogen. Then, the residue was redissolved in 1 mL of a 1/1 (v/v) AcN/H2O mixture and 1 mL of internal standard (containing 1 μg/mL nafcillin or piperacillin solution in AcN). This solution was filtered through a syringe-driven filter unit with a pore size of 0.22 μm (Millipore, Bedford, MA) and analyzed via LC−MS/MS. The protocol for the binding experiments in milk extracts was similar. These binding experiments were used for the evaluation of the binding capacities, the determination of the cross reactivity, and the Scatchard analysis of the developed MIP. Extraction of Bound Benzylpenicillin from a Previously Used MIP or NIP. The suspension of 10 mg of MIP or NIP in 1 mL of AcN containing 5 μg of benzylpenicillin used for the equilibrium binding experiments was passed through a syringe-driven filter unit with a pore size of 0.22 μm (Millipore). A light vacuum (maximum of 0.17 bar) was produced, and the solvent that came through the filter was analyzed using LC−MS/MS as fraction 1 using the same preparative steps described above (see Equilibrium Binding Experiments in Acetonitrile or Milk Extracts). The vial used during the equilibrium binding experiment was rinsed with AcN to pass the total amount of MIP or NIP used during the binding experiment through the filter unit. Again a light vacuum was developed, and the collected solvent was analyzed as fraction 2. The total amount of MIP or NIP on the filter was then washed with MeOH to recover the bound benzylpenicillin. This was repeated in fractions of 1 mL. These fractions (3−6) were also analyzed using LC−MS/MS. These tests were performed to prove the reversibility of the binding of benzylpenicillin to the MIP or NIP. Extraction of the Milk Samples. The blank milk samples were skimmed (centrifugation at 3399g for 10 min). An equal amount of AcN was added to the skimmed milk sample together with NaCl (0.2 g/mL). After the solution had been mixed with a vortex mixer and centrifuged for 10 min at 3399g, a portion of the supernatant was evaporated and spiked with different concentrations of benzylpenicillin, dissolved in AcN, before application to the MIP or NIP. MISPE and Non-molecularly Imprinted Solid-Phase Extraction (NISPE) for the Analysis of AcN Solutions or Milk Samples. An empty SPE cartridge was packed with 50 mg of the (non)imprinted polymer. The polymers were packed between two glass wool frits and washed with 50 mL of MeOH and 15 mL of a MeOH/H2O mixture [95/5 (v/v)] to remove the remaining benzylpenicillin after Soxhlet treatment. The last 5 mL was collected and analyzed by HPLC−MS/ MS to rule out template bleeding.
Fernandez-Gonzales et al. evaluated molecularly imprinted sol− gel materials for the detection of nafcillin.13 For the β-lactam antibiotic cephalexin, MIPs were synthesized and evaluated by Guo and He.14 Skudar et al. produced MIPs for penicillin V and oxacillin. However, the developed MIPs also possessed an affinity for other antibiotics of the β-lactam group.15 Cederfur et al. synthesized MIPs for benzylpenicillin using a molecularly imprinted polymer library.10 Caro et al. developed two MIPs using tetracycline and oxytetracycline as template molecules.16 Zhang et al. investigated the use of MIPs for the determination of benzylpenicillin in milk, but their detection limit was only 1 mg/kg.17 Kempe et al. tested the influence of salt ions on the binding properties of MIPs and applied this to benzylpenicillin in milk.18 The recoveries obtained amounted to 87%, but the control polymer showed also a high recovery of 73%, indicating a high level of nonspecific binding. Finally, Guardia et al. developed molecularly imprinted sol−gel materials for the determination of nafcillin in milk-based products, and a detection limit of 13.2 mg/kg was obtained.19 Several MIPs for the detection of other antibiotics in different fields have been developed: for example, trace analysis of quinolones in eggs,20 determination of nitroimidazoles and fluoroquinolones in egg-based products,21,22 extraction of sulfonamides from honey,23 and determination of sulfonamides in pork and chicken.24 Even MIPs for extraction of several antibiotics from complex matrices are already commercially available. In this study, MIPs were developed through suspension polymerization for application in an SPE procedure for the detection of benzylpenicillin in milk, which has a maximal residue level (MRL) of 4 μg/kg. Binding properties (cross-reactivity, recoveries, number of binding sites, etc.) were evaluated via binding experiments. The MIPs were validated in acetonitrile and then in a milk matrix, where the limit of detection (LOD) and limit of quantitation (LOQ) were determined. Finally, the molecularly imprinted solid-phase extraction (MISPE) procedure was compared (Pearson correlation) with a traditional sample preparation method performed in an accredited laboratory.
■
MATERIALS AND METHODS
Solvents and Solid Reagents. All products [benzylpenicillin, nafcillin, piperacillin, ampicillin, amoxicillin, oxacillin, cloxacillin, dicloxacillin, cephalexin, cephapirin, sulfadiazin, oxytetracycline, methacrylic acid, trimethylolpropane methacrylate, ethylene glycol dimethacrylate, and azobis(isobutyronitrile)] were purchased from Sigma-Aldrich (St. Louis, MO), except for acetonitrile (AcN), formic acid, and methanol (MeOH), which were purchased from Biosolve (Valkenswaard, The Netherlands). Water (H2O) was purified by a Millipore (Billerica, MA) Milli-Q gradient system to high-performance liquid chromatography (HPLC) grade. Standard Solutions of the Antibiotics. Analytical standards were dissolved at a concentration of 1 mg/mL (stock solutions) in a 1/1 (v/v) AcN/H2O mixture. The working standard solutions were made by dilution with AcN. The stock solutions were kept at −20 °C, and the working solutions were freshly prepared before use. Apparatus. A shaker (Bühler, Hechingen, Germany), a centrifuge (model 4K15, Sigma, Osterode am Harz, Germany), a vortex mixer (Scientific Industries, Bohemia, NY), and a vacuum manifold (Varian, Palo Alto, CA) were used during the experiments and sample preparation. A high-performance liquid chromatograph combined with a mass spectrometer was used for the detection and quantification of benzylpenicillin. The high-performance liquid chromatographic− tandem mass spectrometry (LC−MS/MS) system consisted of a LC system (Alliance, Waters, Milford, MA) with a model 2695 separation module coupled to a quattro LCZ tandem quadrupole mass 8815
dx.doi.org/10.1021/jf502331h | J. Agric. Food Chem. 2014, 62, 8814−8821
Journal of Agricultural and Food Chemistry
Article
Before the sample was loaded, the polymers were conditioned with 5 mL of a MeOH/AcN mixture [50/50 (v/v)] to remove the remaining impurities. Next, the AcN solution or the extract of the milk sample was loaded onto the MIP dissolved in 5 mL of AcN. The loading was followed by a washing step with 2 mL of AcN, and finally, the MIP was treated with 6 mL of a MeOH/H2O mixture [50/50 (v/ v)] to elute the benzylpenicillin. This eluent was collected, dried at 60 °C under a flow of nitrogen, and redissolved in 1 mL of a 1/1 (v/v) water/acetonitrile mixture and 1 mL of internal standard (containing 1 μg/mL piperacillin solution in acetonitrile). This solution was filtered through a syringe-driven filter unit with a pore size of 0.22 μm (Millipore) and analyzed using LC−MS/MS. Validation Study in Acetonitrile and Milk. The validation of the performance of the MIP is based on the criteria mentioned in EU Commission Decision 2002/657/EG. Solutions of AcN and spiked milk sample extracts were percolated through the MIP and NIP. The experiments were performed in triplicate, and the recovery was calculated. The recovery is expressed in percent, i.e., the amount of benzylpenicillin obtained after the cleanup step (over MIP or NIP) in relation to the amount of benzylpenicillin added to the AcN solution or milk extract. The goal of the study is to validate the MIP cleanup procedure and not the extraction protocol. That is why we chose to spike the milk samples after extraction to prevent variations originating from the extraction step. Validation of the developed method (including extraction and cleanup step) was conducted during the analysis of contaminated milk samples (see below). The precision was evaluated by calculating the standard deviation for the three experiments executed on different days. The LOD and LOQ were calculated using a calibration curve and the equations
LOD = 3 ×
SEB M
LOQ = 6 ×
SEB M
column was conditioned with 100% A until the pressure was stable for at least 15 min. Mass Spectrometry. The method and optimized ESI+ mass spectrometric conditions described by Daeseleire et al.25 for detecting penicillins in milk were used. The data obtained were analyzed using MassLynx version 4.0 or 4.1 (Waters). Statistical Analysis. To test the influence of the applied crosslinker on the polymerization yield, one-way analysis of variance (ANOVA) was conducted. A Sheffé post hoc test was performed. In addition, one-way ANOVA was used to determine the effect of the application of NIP and MIP on the binding capacity toward benzylpenicillin. SAS version 9.4 (SAS Institute Inc., Cary, NC) was used for all analyses. Significant differences were assumed when p < 0.05.
■
RESULTS AND DISCUSSION This study builds upon the literature.14,15 Our experiments for the production of MIPs used benzylpenicillin as the template molecule, methacrylic acid (MAA) as the monomer, trimethylolpropane methacrylate (TRIM) or ethylene glycol dimethacrylate (EDMA) as the cross-linker, and AcN as the solvent. The MIP was produced through polymerization using azobis(isobutyronitrile) (AIBN) as the heat-induced initiator. The first step was to select the most suitable polymerization technique for the production of MIPs for benzylpenicillin (bulk polymerization vs suspension polymerization). The initial screening tests we performed indicated that the latter technique generated the best results in terms of higher production yields (data not shown) and the more refined way of working that could be applied, i.e., no grinding and sieving required. Second, using the selected suspension polymerization technique, the possible effect of the selected cross-linker on the polymerization yield was evaluated. EDMA and TRIM were selected as bi- and trifunctional cross-linkers, respectively (Table 1).
where SEB is the standard error value on the intercept of the calibration curve and M the slope of the calibration curve. Analysis of Contaminated Milk Samples. To evaluate the implementation of the developed MIP as an SPE sorbent, the concentration of benzylpenicillin present in contaminated milk samples was determined. The results were compared with those obtained after the execution of the traditional protocol (extraction with AcN) for the determination of benzylpenicillin in milk samples in a Beltest-accredited laboratory. Liquid Chromatography. The HP liquid chromatographic separation of benzylpenicillin was performed at room temperature on a C18 column. Mobile phase A consisted of H2O and phase B AcN. In both phases, 0.1% formic acid was added to increase the ionization efficiency. To increase the separation capacity and to decrease the retention times, gradient elution was applied. The gradient program was as follows: 100% A (from 0 to 0.5 min), 100 to 55% A and 45% B (from 0.5 to 0.6 min), 55% A and 45% B to 35% A and 65% B (from 0.6 to 8.5 min), 35% A and 65 to 100% B (from 8.5 to 8.6 min), 100% B (from 8.6 to 18.6 min), 100% B to 100% A (from 18.6 to 20.0 min), and 100% A (from 20.0 to 30.0 min) to re-equilibrate. Before the start of the analyses, the column was conditioned with 100% A until the pressure was stable for at least 15 min. The flow rate was 0.25 mL/ min, and a volume of 40 μL of sample extract was injected into the LC−MS/MS apparatus. The UPLC separation was conducted on a Waters HPLC BEH C18 column held at 30 °C. The injection volume was set at 10 μL and the eluent flow at 0.40 mL/min. Mobile phase A consisted of a H2O/AcN mixture [95/5 (v/v)] with 0.3% acetic acid and phase B an AcN/H2O mixture [95/5 (v/v)] with 0.3% acetic acid. Gradient elution was applied here, as well. The gradient program was as follows: 100% A (from 0 to 2 min), 100 to 70% A and 30% B (from 2 to 8 min), 70% A and 30 to 100% B (from 8 to 12 min), 100% B (from 12 to 13 min), 100% B to 100% A (from 13 to 13.01 min), and 100% A (from 13.01 to 14.6 min) to re-equilibrate. Before the start of the analysis, the
Table 1. Effects of Cross-Linker Type on Polymerization Yield (expressed in percent) applied crosslinker TRIM TRIM blank EDMA EDMA blank
batch 1 batch 2 batch 3 81.0 94.0 78.0 80.8
90.1 89.3 81.3 78.8
89.0 90.1 80.9 79.3
mean
standard deviation
86.7 91.1 80.1 79.6
5.0 2.5 1.8 1.0
The yields of the MIPs produced using TRIM as a crosslinker were higher, although not significantly so (p > 0.05) (Table 1). The higher yields can most likely be explained by the differences in the number of reactive functional groups between EDMA (bifunctional) and TRIM (trifunctional). On the basis of these data, we decided to continue all further experimental work using TRIM as the cross-linker. After optimization of the template/functional monomer/ cross-linker molar ratio, the composition applied for the development of a benzylpenicillin selective MIP was as follows: benzylpenicillin (2 mmol), MAA (16 mmol), TRIM (7 mmol), and AIBN (0.4 mmol) dissolved in 70 mL of AcN. Using this formulation, the MIP produced was visualized by scanning electron microscopy (SEM). Uniformly sized microspheres were obtained by applying the suspension polymerization method (Figure 1). The microspheres had an average size of 1.2 μm. To determine the binding specificity of the developed MIP for benzylpenicillin in AcN, we first performed an equilibrium 8816
dx.doi.org/10.1021/jf502331h | J. Agric. Food Chem. 2014, 62, 8814−8821
Journal of Agricultural and Food Chemistry
Article
NIP with respect to structural analogues of benzylpenicillin. This was done by performing the equilibrium binding experiments in AcN with solutions containing seven molecules belonging to the group of the β-lactam antibiotics (benzylpenicillin is one of them). The selected molecules included ampicillin, amoxicillin, oxacillin, cloxacillin, dicloxacillin, cephalexin, and cephapirin. In addition, one molecule of the group of sulfonamides (sulfadiazin) and one of the group of the tetracyclines (oxytetracycline) were also included. All equilibrium binding tests were repeated three times (Table 3). Table 3. Results of the Cross-Reactivity Tests for the Developed NIP
Figure 1. Size of benzylpenicillin selective MIP as visualized by SEM.
experiment by comparing the produced MIP and NIP. The obtained results of three independently performed equilibrium binding studies show a statistically significant (p < 0.01) difference in binding capacity between the MIP or NIP and benzylpenicillin (68.3 ± 3.1 or 13.2 ± 8.2%, respectively). This represented the first convincing proof that the developed MIP possessed a high affinity for benzylpenicillin. The reversibility of the binding of benzylpenicillin to the MIP or NIP was proven by washing out the MIP or NIP used during an equilibrium binding experiment. The amounts of benzylpenicillin detected in the different fractions are expressed in relation to that of an internal standard (nafcillin). The experiments were executed in duplicate, and the results (Table 2) clearly show that the total amount of benzylpenicillin (5 μg) Table 2. Results of the Recovery of Benzylpenicillin out of MIP or NIP of an Equilibrium Binding Experiment a
benzylpenicillin/ nafcillin ratio
MIP (first experiment)
MIP (second experiment)
NIP (first experiment)
NIP (second experiment)
fraction 1a fraction 2a fraction 3b fraction 4b fraction 5b fraction 6b sum fractions standard 5 μg of benzylpenicillin
0.530 0.506 0.758 0.004 0.000 0.001 1.799 1.805
0.541 0.514 0.774 0.003 0.000 0.002 1.834 1.805
1.177 0.735 0.044 0.003 0.001 0.001 1.961 1.805
1.168 0.735 0.067 0.007 0.001 0.003 1.981 1.805
% binding components
first test
second test
third test
mean
standard deviation
benzylpenicillin ampicillin amoxicillin oxacillin cloxacillin dicloxacillin cephalexin cephapirin sulfadiazine oxytetracycline benzylpenicillin ampicillin amoxicillin oxacillin cloxacillin dicloxacillin cephalexin cephapirin sulfadiazin oxytetracycline
3.8 92.0 94.5 11.0 −a 11.0 72.1 90.2 6.0 73.1 66.9 99.1 99.6 63.5 50.3 33.8 97.1 97.4 15.6 87.8
16.9 91.0 94.2 14.6 32.2 3.9 68.0 94.9 5.0 86.3 71.8 98.6 99.6 61.2 51.6 45.9 97.2 99.4 −a 94.2
18.8 94.0 91.5 11.0 12.3 12.9 71.0 93.1 10.7 87.3 66.1 99.2 98.8 58.5 43.6 46.2 93.5 99.6 21.7 91.5
13.17 92.33 93.40 12.20 22.25 9.27 70.37 92.73 7.23 82.23 68.27 98.97 99.33 61.07 48.50 41.97 95.93 98.80 18.65 91.17
8.17 1.53 1.65 2.08 14.07 4.74 2.12 2.37 3.04 7.93 3.09 0.32 0.46 2.50 4.29 7.07 2.11 1.22 4.31 3.21
Results not included because of problems with the internal standard.
The results of the tests for both the NIP and the MIP clearly show that the compounds included in this study can be classified into three separate groups. A first group showed binding efficiencies of >90% for the MIP and >70% for the NIP (ampicillin, amoxicillin, cephalexin, cephaphirin, and oxytetracycline). A second group showed binding percentages between 42 and 68% for the MIP and between 10 and 22% for the NIP (oxacillin, cloxacillin, and dicloxacillin). The last group showed binding degrees of
