Preparation and Evaluation of Core–Shell Magnetic Molecularly Imprinted Polymers for Solid-Phase Extraction and Determination of Sterigmatocystin in Food

Magnetic molecularly imprinted polymers (MMIPs), combination of outstanding magnetism with specific selective binding capability for target molecules, have proven to be attractive in separation science and bio-applications. Herein, we proposed the core–shell magnetic molecularly imprinted polymers for food analysis, employing the Fe3O4 particles prepared by co-precipitation protocol as the magnetic core and MMIP film onto the silica layer as the recognition and adsorption of target analytes. The obtained MMIPs materials have been fully characterized by scanning electron microscope (SEM), Fourier transform infrared spectrometer (FT-IR), vibrating sample magnetometer (VSM), and re-binding experiments. Under the optimal conditions, the fabricated Fe3O4@MIPs demonstrated fast adsorption equilibrium, a highly improved imprinting capacity, and excellent specificity to target sterigmatocystin (ST), which have been successfully applied as highly efficient solid-phase extraction materials followed by high-performance liquid chromatography (HPLC) analysis. The MMIP-based solid phase extraction (SPE) method gave linear response in the range of 0.05–5.0 mg·L−1 with a detection limit of 9.1 µg·L−1. Finally, the proposed method was used for the selective isolation and enrichment of ST in food samples with recoveries in the range 80.6–88.7% and the relative standard deviation (RSD) <5.6%.


Introduction
Sterigmatocystin (ST), a carcinogenic mycotoxin, is a secondary metabolite produced mainly by Aspergillus versicolor, Aspergillus flavus, Aspergillus nidulans, fine wrinkles and other fungi Aspergillus [1][2][3]. With a basic structure composed of two furan rings with oxygen hetero anthraquinone, ST possess a similar structure to Aflatoxin B1 and its toxicity is second only to Aflatoxin that can induce liver cancer, lung cancer and other cancers [4,5]. Moreover, ST spreads widely in nature, and could contaminate most food and forage, especially wheat, corn, peanuts and forage, raising great concerns of public society [6,7]. For the quantification of ST in food samples, usually with complex matrix, liquid-liquid extraction or solid phase extraction coupled with high-performance were obtained from Sigma-Aldrich (St. Louis, MO, USA). 2,2-azobisisobutyronitrile (AIBN; 99%) was purchased from Tianjin Kermel Chemical Reagents Co. Ltd. (Tianjin, China). Acetonitrile, methanol, ethanol, n-hexane, acetone, and glacial acetic acid were obtained from Sinopharm Group Co. Ltd. (Tianjin, China). Iron (II) dichloride and iron (III) chloride were obtained from Chemicals Co. Ltd. (Tianjin, China). Highly purified water was obtained from a Pro Water System (Millipore Co. Ltd., Billerica, MA, USA). The food samples were randomly obtained from some local supermarkets (Tianjin, China). The samples of wheat and rice millet were purchased from a local supermarket (Tianjin, China). The ST-free samples were detected by HPLC. Caution is necessary when operating with toxins, wear the necessary personal protective equipment (PPE) including gloves and protective facial mask.

Instrumentation
Ultraviolet-visible (UV-vis) spectra over 200-800 nm was recorded on an Evolution 300 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Scanning electron microscopy (SU 1510, Hitachi, Tokyo, Japan) was used to observe the shape, size and surface morphologies of Fe 3 O 4 , Fe 3 O 4 @SiO 2 , Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs (non-imprinted polymers). Transmission electron microscope (TEM) images were obtained on a 2010 FEF microscope (JEOL, Tokyo, Japan). The infrared spectra were observed using a Tensor 27 FT-IR spectrophotometer (Bruker Company, Berlin, Germany). The magnetic intensity was evaluated using a 7410 VSM (Lake Shore Company, MA, USA). The analysis of HPLC was performed using HPLC (Shimadzu, Tokyo, Japan) with a variable wavelength UV-visible detector. A ZORBAX Eclipse XDB-C 18 (5 µm, 250 mm × 4.6 mm, Shimadzu, Tokyo, Japan) analytical column was used for the separation of analytes. The mobile phase was methanol/H 2 O (4/1, v/v), and the flow rate was 0.7 mL·min −1 at 35 • C. The injection volume was 20 µL, and the wavelength of the UV detector was proceeding at 246 nm.

Synthesis of Silica Coated Fe 3 O 4 Particles
Referring to the reported solvothermal method [31], the FeCl 3 ·6H 2 O (1.35 g, 5 mmol) was dissolved in ethylene glycol (EG, 40 mL), with continued stirring to form a clear solution, followed by the addition of NaAc (3.6 g) and polyethylene glycol (1.0 g). The mixture solution was stirred vigorously for another 30 min, and sealed in a teflon lined stainless-steel autoclave. The autoclave was heated to 200 • C and maintained for 8 h, then cooled to room temperature after complete reaction. The black products were washed three times using ethanol with the assistance of external magnet isolation. The obtained magnetic Fe 3 O 4 particles product was dried at 60 • C for 6 h (0.26 g with a yield of 68.4%).
Thirty mg of Fe 3 O 4 particles were dispersed in ethanol (50 mL), and 0.4 mL of TEOS was added, followed by cautious addition of 2 mL of ammonia water (20%) dropwise. The reaction was allowed to proceed under continuous stirring for 8 h at room temperature, and the resultant product was repeatedly washed four times using ethanol by external magnets, dried in a vacuum oven at 60 • C for 10 h (38.4 mg with a yield of 86.7%).
One hundred mg of thin silicon coated Fe 3 O 4 particles were dispersed in isopropanol (18 mL), followed by addition of 0.5 mL of aqueous ammonia solution (30%) under continued stirring. Then, excessive γ-methacryloxypropyl triethoxysilane (γ-MAPs) was added drop by drop, and the mixture was allowed to react for 24 h at room temperature under continuous stirring. The resultant product was collected by an external magnetic field and rinsed with ethanol four times thoroughly, dried in the vacuum (98.2 mg with a yield of 98.2%).

Preparation of Magnetic Core-Shell Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs Particles
The MIPs were prepared by a precipitation polymerization approach. First, DT (120 mg) and acrylamide (AM) (142.2 mg) was mixed with 30 mL of acetonitrile, with continuous stirring until completely dissolved, followed by addition of 100 mg of Fe 3 O 4 @SiO 2 -MAPs. The mixed solution was kept under ultrasonic treatment (300 W, SB-5200D, SCIENTZ, Ningbo, China) for 15 min followed with forming template-monomer complex at 25 • C for 3 h. EGDMA (1.981 g) and AIBN (60 mg) were added and then mixed the solution adequately with a 15-min ultrasonic treatment. The mixed solution was transferred to a 250 mL three-necked flask followed by the addition of 70 mL of acetonitrile. Finally, nitrogen was purged in to remove oxygen and reaction proceeded by thermal polymerization at 60 • C for 24 h. The products were thoroughly grinded, filtrated through a 200-mesh sieve, and treated with 200 mL of acetone and acetonitrile (9:1, v/v) to remove the template. The washing efficiency was evaluated via UV-vis measurement (429 nm) of DT amount in supernatant solutions. Procedure of preparation of non-imprinted Fe 3 O 4 @NIPs was the same to that of Fe 3 O 4 @MIPs in the absence of the template DT. The obtained polymer products (MIP~680 mg, NIP~710 mg) was thoroughly washed by ethanol three times and dried under vacuum.

Determination of Sterigmatocystin via Magnetic Fe 3 O 4 @MIPs
To investigate the recognition properties of Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs, the adsorption experiments were performed. Every test was performed thrice in parallel. Blank tests were performed by incubating MIP or NIP with water and measuring by HPLC to ensure there was no interferent existing in the polymers.
In the adsorption kinetics experiment, 5 mg of Fe 3 O 4 @MIPs or Fe 3 O 4 @NIPs were added to 3 mL of ST solution (20 mg·L −1 ), and incubated at regular time intervals from 10 min to 300 min at room temperature. After separating the supernatants and polymers using an external magnetic field, the concentration of ST in the supernatants was measured by a UV-visible spectrophotometer (327 nm). The amount of ST bound to the Fe 3 O 4 @MIPs or Fe 3 O 4 @NIPs was figured out by the formula.
In the isothermal binding experiment, 5 mg of Fe 3 O 4 @MIPs or Fe 3 O 4 @NIPs were added to 3 mL methanol solution of ST of various concentrations from 2 to 50 mg·L −1 and the mixture was incubated for 10 min at room temperature respectively. After incubation, the supernatants and polymers were separated by an external magnet, and the remaining ST in the supernatants was determined by UV-visible spectroscopy at 327 nm.
In the selectivity experiments, AFB1 was selected as the structural analog. The Fe 3 O 4 @MIPs or Fe 3 O 4 @NIPs (25 mg) was placed in methanol solution of ST or AFB1 (20 mg·L −1 , 10 mL). After incubating for 4 h at room temperature, the supernatant and polymers were separated using an external magnetic field and the concentration of ST and its analog in the supernatant was determined by HPLC-UV. Moreover, to further verify the competitive recognition ability, Fe 3 O 4 @MIPs (20 mg) was mixed with 10 mL methanol solution of ST and AFB1 (20 mg·L −1 each). The extraction and determination procedures were then performed as described earlier for the static adsorption experiments. The same procedure was performed for the Fe 3 O 4 @NIPs.

Adsorption Isotherms
The Langmuir and Freundlich isotherm models were employed to evaluate the adsorption process. The widely used Langmuir equation, which was valid for monolayer sorption on a surface with a finite number of identical sites, is given by where Q m (mg·g −1 ) is the maximum adsorption capacity of adsorbent at monolayer, and K l is the Langmuir constants. The essential characteristics of the Langmuir isotherm can also be expressed in terms of where R l is a dimensionless constant of separation factor or equilibrium parameter, which indicates the shape of adsorption isotherm. The widely used empirical Freundlich equation, based on sorption on a heterogeneous surface, is given as: where K F and n are Freundlich constants indicating adsorption capacity and intensity, respectively, which can be calculated from linear plot of ln q e against ln C e .

Adsorption Kinetics Investigations
To investigate the mechanism of sorption and potential rate controlling steps, the pseudo-firstorder, pseudo-second-order, intra-particle diffusion and Elovich model have been used to test the experimental data. The rate constants for four models have been determined and the correlation coefficients have been calculated in order to assess which model provides the best fit of the predicted data with the experimental results.
The pseudo-first-order kinetic model known as the Lagergern equation: where q t and q e are the amounts of ion adsorbed at time t and at equilibrium (mg·g −1 ), respectively, and K 1 is the rate constant of pseudo-first-order adsorption process (min −1 ). After integration and applying boundary conditions, for t = 0, q = 0, the integrated form of equation becomes: The pseudo-second-order equation based on adsorption equilibrium capacity can be expressed as: where K 2 is the rate constant of pseudo-second-order sorption (g·mg −1 ·min −1 ). For t = 0, q = 0, it was given as: The Elovich equation was also be applied to analyze the adsorption data. Its linear form was given as: where α is the initial sorption rate constant (mg·g −1 ·min −1 ), and the parameter β is related to the extent of surface coverage and activation energy for chemisorption (g·mg −1 ). α, β can be obtained from the slope and intercept of the plot of q t versus ln t.
The intra-particle diffusion model was also tested. The initial rate of intra-particle diffusion is as follows: where k int is the intra-particle diffusion rate constant (mg·g −1 ·min −1/2 ), and C is the intercept.

Real Sample Analysis
All food samples were free of ST, and the spiking concentrations were 50 µg·kg −1 , 100 µg·kg −1 , and 200 µg·kg −1 . The food sample (1.0 g) was accurately weighed in a 25-mL conical flask with a stopper, and then different amounts of ST standard solutions were added (5 mg·L −1 , 50 µL, 100 µL, and 200 µL, dissolved in methanol). After thoroughly incorporated, the mixture stands overnight. Ten milliliters of n-hexane were added, and the mixture was treated by ultrasonic for 10 min. Exactly 3.0 mL of the supernatant solution was added to another 25-mL conical flask. Typically, 15.0 mg of Fe 3 O 4 @MIPs were added and shaken at 120 rpm for 50 min at room temperature on a shaker (MS 3 digital, IKA, Berlin, Germany). The supernatant and polymers were separated using an external magnetic field. After removing the supernatant solution, the Fe 3 O 4 @MIPs was washed with 1 mL of n-hexane/ether (4:1, v/v) to eliminate the co-extracted impurities. Then, the ST was eluted from the Fe 3 O 4 @MIPs with 3 × 1.0 mL of chloroform, and the elutes were combined together and evaporated to almost dryness under a stream of nitrogen. The residue was re-dissolved by 1.0 mL of chromatographic pure methanol, and filtered through a 0.45 µm organic membrane, and finally detected by HPLC.

Preparation and Characterization of Imprinted Magnetic Nanoparticles
When constructing the MIP materials, a dummy template molecule is often employed as an alternative template for imprinting instead of the target analyte, in the case that the target analyte is highly toxic, expensive or unavailable [15,32]. The dummy template should possess the similar structure and physicochemical property to the target molecules and scarcely present in the related sample matrix to avoid possible interference. In the previously-reported work, DT was a widely-used dummy template molecule for imprinting of ST, due to the high price and toxicity of ST [22,24]. Therefore, the magnetic molecularly imprinted polymers for recognizing ST specifically were synthesized using DT as the dummy template molecule, which has similar structure with ST and hardly presents in the grain samples. The preparation procedure of Fe 3 O 4 @MIPs by the precipitation polymerization is shown in Figure 1. Super paramagnetic Fe 3 O 4 nanoparticles were synthesized by the solvothermal method to provide a good magnetic core, supporting the MIPs material magnetic response. The surface of the Fe 3 O 4 core was encapsulated with SiO 2 shell by TEOS to avoid the oxidation and provide silanol groups at the surface, which make them biocompatible and easily modified with various functional groups. Furthermore, silica shell could not only shield the magnetic dipolar attraction between magnetic particles, in favor of the dispersion of magnetic particles in solvent, but also protect Fe 3 O 4 from dissolving in an acidic environment.
Polymers 2017, 9, 546 6 of 13 from the Fe3O4@MIPs with 3 × 1.0 mL of chloroform, and the elutes were combined together and evaporated to almost dryness under a stream of nitrogen. The residue was re-dissolved by 1.0 mL of chromatographic pure methanol, and filtered through a 0.45 µm organic membrane, and finally detected by HPLC.

Preparation and Characterization of Imprinted Magnetic Nanoparticles
When constructing the MIP materials, a dummy template molecule is often employed as an alternative template for imprinting instead of the target analyte, in the case that the target analyte is highly toxic, expensive or unavailable [15,32]. The dummy template should possess the similar structure and physicochemical property to the target molecules and scarcely present in the related sample matrix to avoid possible interference. In the previously-reported work, DT was a widely-used dummy template molecule for imprinting of ST, due to the high price and toxicity of ST [22,24]. Therefore, the magnetic molecularly imprinted polymers for recognizing ST specifically were synthesized using DT as the dummy template molecule, which has similar structure with ST and hardly presents in the grain samples. The preparation procedure of Fe3O4@MIPs by the precipitation polymerization is shown in Figure 1. Super paramagnetic Fe3O4 nanoparticles were synthesized by the solvothermal method to provide a good magnetic core, supporting the MIPs material magnetic response. The surface of the Fe3O4 core was encapsulated with SiO2 shell by TEOS to avoid the oxidation and provide silanol groups at the surface, which make them biocompatible and easily modified with various functional groups. Furthermore, silica shell could not only shield the magnetic dipolar attraction between magnetic particles, in favor of the dispersion of magnetic particles in solvent, but also protect Fe3O4 from dissolving in an acidic environment. To introduce vinyl groups, the hydroxyl groups on surface of Fe3O4@SiO2 were further reacted with MPS, which subsequently reacted in the synthesis of MIPs or NIPs. In addition, the Fe3O4@SiO2 particles had a small diameter with an extremely high surface area-to-volume ratio, so that the MIPs formed easily at the surface of the magnetic particles. The MIPs shells were coated on the surface of Fe3O4@SiO2 by the copolymerization of functional monomer (AM), cross-linking agent (EGDMA), initiator (AIBN) and template molecule (DT). After removal of the templates, the Fe3O4@MIPs particles were obtained. Overall, the Fe3O4@MIPs particles could recognize and adsorb the targets effectively and were also easily collected using an external magnetic field. Meanwhile, Fe3O4@NIPs were also prepared with the same procedure, but without the addition of template DT.
The FT-IR spectroscopy of Fe3O4, Fe3O4@SiO2, Fe3O4@MIPs and Fe3O4@NIPs is shown in Figure  2A.  To introduce vinyl groups, the hydroxyl groups on surface of Fe 3 O 4 @SiO 2 were further reacted with MPS, which subsequently reacted in the synthesis of MIPs or NIPs. In addition, the Fe 3 O 4 @SiO 2 particles had a small diameter with an extremely high surface area-to-volume ratio, so that the MIPs formed easily at the surface of the magnetic particles. The MIPs shells were coated on the surface of Fe 3 O 4 @SiO 2 by the copolymerization of functional monomer (AM), cross-linking agent (EGDMA), initiator (AIBN) and template molecule (DT). After removal of the templates, the Fe 3 O 4 @MIPs particles were obtained. Overall, the Fe 3 O 4 @MIPs particles could recognize and adsorb the targets effectively and were also easily collected using an external magnetic field. Meanwhile, Fe 3 O 4 @NIPs were also prepared with the same procedure, but without the addition of template DT.  Magnetic property is crucial to the magnetic particles for their application in fast separation. VSM was employed to study the magnetic properties of the synthesized magnetic particles, and the magnetic hysteresis loop of the dried samples at room temperature is illustrated in Figure 2B. It is obvious that there is no hysteresis, both remanence and coercivity are zero, suggesting that the samples are superparamagnetism. The saturation magnetization values obtained at room temperature were 88.73 emu·g −1 , 59.45 emu·g −1 , 31.36 emu·g −1 , and 15.17 emu·g −1 for Fe3O4, Fe3O4@SiO2, vinyl-modified Fe3O4@SiO2 and Fe3O4@MIPs, respectively. The theoretical value of saturation magnetization for bulk magnetite is reported to be 92 emu·g −1 . The decrease in magnetization value can be attributed to the small particle surface effect such as magnetically inactive layer containing spins that are not collinear with the magnetic field. The saturation magnetization of Fe3O4@MIPs was reduced to 15.17 emu·g −1 in comparison with the pure Fe3O4, but remained strongly magnetic at room temperature and qualified as effective magnetic separation carriers.
TEM, SEM, and dynamic light scattering (DLS) characterizations revealed that the as-synthesized Fe3O4 particles possessed an average diameter of ~370 nm. After the modification with TEOS, a thin silicon layer coated on the surface of Fe3O4 could be distinguished, with the size increased to ~480 nm, which also confirmed the successful preparation of the core-shell structure Fe3O4@SiO2. In comparison, the Fe3O4@MIPs, with an average diameter of ~830 nm, structures seemed more rough and dense than the Fe3O4@NIPs (average diameter of ~1.23 µm), indicating the template molecule showed obvious influence over the surface topography. Consequently, this uniform structure of MIP materials would facilitate the mass transport between solution and the Magnetic property is crucial to the magnetic particles for their application in fast separation. VSM was employed to study the magnetic properties of the synthesized magnetic particles, and the magnetic hysteresis loop of the dried samples at room temperature is illustrated in Figure 2B. It is obvious that there is no hysteresis, both remanence and coercivity are zero, suggesting that the samples are superparamagnetism. The saturation magnetization values obtained at room temperature were 88.73 emu·g −1 , 59.45 emu·g −1 , 31.36 emu·g −1 , and 15.17 emu·g −1 for Fe 3 O 4 , Fe 3 O 4 @SiO 2 , vinyl-modified Fe 3 O 4 @SiO 2 and Fe 3 O 4 @MIPs, respectively. The theoretical value of saturation magnetization for bulk magnetite is reported to be 92 emu·g −1 . The decrease in magnetization value can be attributed to the small particle surface effect such as magnetically inactive layer containing spins that are not collinear with the magnetic field. The saturation magnetization of Fe 3 O 4 @MIPs was reduced to 15.17 emu·g −1 in comparison with the pure Fe 3 O 4 , but remained strongly magnetic at room temperature and qualified as effective magnetic separation carriers.
TEM, SEM, and dynamic light scattering (DLS) characterizations revealed that the as-synthesized Fe 3 O 4 particles possessed an average diameter of~370 nm. After the modification with TEOS, a thin silicon layer coated on the surface of Fe 3 O 4 could be distinguished, with the size increased to~480 nm, which also confirmed the successful preparation of the core-shell structure Fe 3 O 4 @SiO 2 . In comparison, the Fe 3 O 4 @MIPs, with an average diameter of~830 nm, structures seemed more rough and dense than the Fe 3 O 4 @NIPs (average diameter of~1.23 µm), indicating the template molecule showed obvious influence over the surface topography. Consequently, this uniform structure of MIP materials would facilitate the mass transport between solution and the shell surface of Fe 3 O 4 @MIPs as SPE adsorbents (Figures 2C and 3).
Experimental conditions, including MMIPs amount and organic media used for SPE assay, have been fully optimized. Results in Figure 4 indicated the MMIPs amount showed insignificant influence over the recovery, thus chose the relative small amount of 5 mg for all assays. When using n-hexane as the sampling media and chloroform as the eluting media, best SPE performance was achieved. It was also found that pH showed little influence on the adsorption, probably because there was limited electrostatic interaction in the adsorption process that carried out all in non-aqueous solvents. n-hexane as the sampling media and chloroform as the eluting media, best SPE performance was achieved. It was also found that pH showed little influence on the adsorption, probably because there was limited electrostatic interaction in the adsorption process that carried out all in non-aqueous solvents.

MIPs Recognition of Sterigmatocystin
The binding isotherms of ST onto Fe3O4@MIPs and Fe3O4@NIPs were determined in the concentration range of 0-50 mg·L −1 , and the results are shown in Figure 4A. Molecular recognition of Fe3O4@MIPs and Fe3O4@NIPs particles increased rapidly with increasing initial concentration, and became relatively flat and reached its saturation at high ST concentration. The amount of ST bound to the Fe3O4@MIPs was significantly higher than that of the Fe3O4@NIPs at the same initial concentration. The recognition ability of Fe3O4@MIPs particles towards ST was investigated by adsorption kinetics. The adsorption kinetics of 20 mg·L −1 ST solution to Fe3O4@MIPs and Fe3O4@NIPs are shown in Figure 4B. The adsorption capacity increased with time, and the Fe3O4@MIPs showed a fast adsorption rate. The adsorption capacity increased rapidly in the first 30 min and almost reached equilibrium after 2 h. Most of the recognition sites of the imprinted polymers are on the surface of the imprinted magnetic particles, facilitating high adsorption efficiency.
The adsorption process was interpreted both with Langmuir and Freundlich isotherm models. Comparison of the calculated data from Langmuir and Freundlich isotherm models indicated the obtained data is better fitted with Langmuir model than with Freundlich model, revealing the adsorption was more similar to a monolayer adsorption process rather than a multiple process. The Langmuir constant Rl was in the range of 0-1, which indicated favorable adsorption of MMIP to the analytes ( Table 1). The models of the pseudo-first-order, pseudo-second-order, intra-particle diffusion and Elovich model were employed to evaluate the kinetic mechanism. It was noticed that pseudo-second-order equation provided the better R 2 and agreement between calculated qe values and the experimental qe (exp) value (0.95), whereas the pseudo-first-order, Elvoich and intra-particle

MIPs Recognition of Sterigmatocystin
The binding isotherms of ST onto Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs were determined in the concentration range of 0-50 mg·L −1 , and the results are shown in Figure 4A. Molecular recognition of Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs particles increased rapidly with increasing initial concentration, and became relatively flat and reached its saturation at high ST concentration. The amount of ST bound to the Fe 3 O 4 @MIPs was significantly higher than that of the Fe 3 O 4 @NIPs at the same initial concentration. The recognition ability of Fe 3 O 4 @MIPs particles towards ST was investigated by adsorption kinetics. The adsorption kinetics of 20 mg·L −1 ST solution to Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs are shown in Figure 4B. The adsorption capacity increased with time, and the Fe 3 O 4 @MIPs showed a fast adsorption rate. The adsorption capacity increased rapidly in the first 30 min and almost reached equilibrium after 2 h. Most of the recognition sites of the imprinted polymers are on the surface of the imprinted magnetic particles, facilitating high adsorption efficiency.
The adsorption process was interpreted both with Langmuir and Freundlich isotherm models. Comparison of the calculated data from Langmuir and Freundlich isotherm models indicated the obtained data is better fitted with Langmuir model than with Freundlich model, revealing the adsorption was more similar to a monolayer adsorption process rather than a multiple process. The Langmuir constant R l was in the range of 0-1, which indicated favorable adsorption of MMIP to the analytes ( Table 1). The models of the pseudo-first-order, pseudo-second-order, intra-particle diffusion and Elovich model were employed to evaluate the kinetic mechanism. It was noticed that pseudo-second-order equation provided the better R 2 and agreement between calculated q e values and the experimental q e (exp) value (0.95), whereas the pseudo-first-order, Elvoich and intra-particle diffusion equations did not give a good fit to the experimental data for the adsorption of ST (Table 1).
The regeneration ability and the stability of Fe 3 O 4 @MIPs materials were evaluated by spacing and non-spacing adsorption and elution cycles, and the SPE efficiency was assessed by observing the changes of the recovery. The results indicated that the MIPs were stable in the cycle tests, specifically in more than 20 adsorption-elution cycles with stable SPE recoveries of the target analyte ( Figure 4C).   To evaluate the specificity of the developed Fe3O4@MIPs materials, the imprinting factors are introduced for comparison. The adsorption of the Fe3O4@MIPs and Fe3O4@NIPs to both ST and the structural analog, AFB1 ( Figure 4D) were carried out at the same experimental conditions. The imprinting factor was defined as the ratio of QMIPs to QNIPs, which represented adsorption capacity of MIP and NIP, respectively. As shown in Figure 5, the Fe3O4@MIPs showed a significantly higher adsorption capacity of ST than AFB1, while the Fe3O4@NIPs did not show such a difference, indicating that the template molecule had a relatively higher affinity for the imprinted polymer than its analog. Moreover, the imprinting factor of ST (2.8) was also much higher than its analog   To evaluate the specificity of the developed Fe 3 O 4 @MIPs materials, the imprinting factors are introduced for comparison. The adsorption of the Fe 3 O 4 @MIPs and Fe 3 O 4 @NIPs to both ST and the structural analog, AFB1 ( Figure 4D) were carried out at the same experimental conditions. The imprinting factor was defined as the ratio of Q MIPs to Q NIPs , which represented adsorption capacity of MIP and NIP, respectively. As shown in Figure 5, the Fe 3 O 4 @MIPs showed a significantly higher adsorption capacity of ST than AFB1, while the Fe 3 O 4 @NIPs did not show such a difference, indicating that the template molecule had a relatively higher affinity for the imprinted polymer than its analog. Moreover, the imprinting factor of ST (2.8) was also much higher than its analog (1.4), further confirming the excellent recognition performance of Fe 3 O 4 @MIPs to ST. Furthermore, co-existing experiments demonstrated similar results as above. Although ST and AFB1 have similar scaffold (xanthene), the differences in their spatial structures and functional groups caused a mismatch in the holes and binding sites, leading to the good selectivity to ST against AFB1. Due to utilizing DT as template for imprinting, the selectivity of DT is comparable to ST. However, in most cases, DT scarcely exist together with ST in grain samples; even during coexistence, the interference could be effectively reduced or totally eliminated via the extraction procedures [22]. Besides, owing to the well-constructed imprinting polymers, the Fe 3 O 4 @MIPs showed insignificant response to the interferents, such as OTA, AFB2, AFM1, MC-LR, DON, and ZEN, compared with that of target ST.

Intra-Particle Diffusion Elovich
The proposed fluorescent MIPs method gave a linear range of 0.05-5.0 mg·L −1 with a detection limit (3 s) of 9.1 µg·L −1 for the detection of ST. The precision (relative standard deviation (RSD)) for eleven replicate detections of 0.5 mg·L −1 ST was 2.1%. These results indicate that the Fe 3 O 4 @MIPs can be used for the sensitive and selective SPE and determination of ST in complex samples. Compared with the previously-reported methods for ST determination in terms of sensitivity and linear range, the developed MIP method showed comparable performance ( Table 2).

Real Sample Analysis
To demonstrate the applicability of the developed magnetic MIP-HPLC method for real sample analysis, it was applied for the selective isolation and enrichment of ST in food samples via spiked recovery experiments. As can be seen in the chromatograms of the cereal samples before and after being spiked with ST at 50 µg·kg −1 , ST appeared at 7.86 min after being spiked, and other irrelevant compounds in the sample showed insignificant interference to the measurement ( Figure 6). As shown in Table 3, the obtained recoveries of the spiked samples were in the range of 85.2-88.1%  GC-MS: gas chromatography-mass spectrometry; LC-MS: liquid chromatography-mass spectrometry; HPLC: high-performance liquid chromatography; NA: Not applicable.

Real Sample Analysis
To demonstrate the applicability of the developed magnetic MIP-HPLC method for real sample analysis, it was applied for the selective isolation and enrichment of ST in food samples via spiked recovery experiments. As can be seen in the chromatograms of the cereal samples before and after being spiked with ST at 50 µg·kg −1 , ST appeared at 7.86 min after being spiked, and other irrelevant compounds in the sample showed insignificant interference to the measurement ( Figure 6). As shown in Table 3, the obtained recoveries of the spiked samples were in the range of 85.2-88.1% for wheat, 80.6-88.7% for rice, and 82.9-88.6% for millet, with relative standard deviation (RSD) less than 5.6%.

Conclusions
In this study, MIPs were synthesized onto Fe3O4@SiO2 magnetic particles with a uniform coreshell structure by surface imprinting and nanotechnology. The Fe3O4@MIPs showed remarkable specificity target ST along with sensitive response. The successful selective separation and enrichment of ST in food samples indicated that the Fe3O4@MIPs was ideal solid-phase extraction material and had the potential of applying to detect the illegal addition of ST in food.

Conclusions
In this study, MIPs were synthesized onto Fe 3 O 4 @SiO 2 magnetic particles with a uniform core-shell structure by surface imprinting and nanotechnology. The Fe 3 O 4 @MIPs showed remarkable specificity target ST along with sensitive response. The successful selective separation and enrichment of ST in food samples indicated that the Fe 3 O 4 @MIPs was ideal solid-phase extraction material and had the potential of applying to detect the illegal addition of ST in food.