Development of ELISA-Like Fluorescence Assay for Melamine Detection Based on Magnetic Dummy Molecularly Imprinted Polymers

We present a directly competitive fluorescence assay for highly sensitive detection of melamine in milk using magnetic dummy molecularly imprinted polymers (MDMIPs). The detection principle is based on competitive binding between the fluorescent label and melamine on the MDMIPs. The fluorescent label was obtained by combining fluorescein isothiocyanate (FITC) with melamine in ethanol and water. MDMIPs were prepared on the surface of Fe3O4@SiO2 nanoparticles using 2,4-diamino-6-methyl-1,3,5-triazine as dummy template. The MDMIPs were characterized and their adsorption capacity was evaluated based on their static adsorption and Scatchard analysis. Results suggest that MDMIPs were successfully coated on the Fe3O4@SiO2 surface and had a core–shell structure. Adsorption experiments suggested that the MDMIPs had higher specific recognition capacities for melamine and FITC–melamine (FITC-Mel) than did magnetic dummy molecularly non-imprinted polymers. Competitive binding between FITC-Mel and melamine was performed under the optimum conditions to determine melamine quantitatively. The linear range of this fluorescence assay was 0.1–20 mg/L for melamine detection. The detection limit was 0.05 mg/L in negative milk samples. The assay was also successfully employed to detect melamine in spiked milk samples, with satisfactory recoveries, i.e., between 70.2% and 92.7%.

Surface molecular imprinting on various surfaces, including SiO 2 nanoparticles [16], carbon materials [17], polymeric supports [18], metal oxide particles [19], and magnetic materials [20], has been used to overcome these problems. Magnetic nanoparticles have been widely used in the preparation of magnetic molecularly imprinted polymers (MMIPs) because of their properties such as small size, high magnetic susceptibility, and high surface-to-volume ratio. MMIPs can be easily separated from a complex matrix using an external magnet, enabling convenient and efficient preconcentration to avoid using time-consuming solid-phase extraction and column packing [21,22]. Potential interference from leakage of residual template molecules embedded in MIPs can be prevented by using a dummy molecular imprinting technique, in which a structural analog of the target compound is used as a template for MIP preparation [23].
Melamine, which is a triamine and contains a nitrogen heterocyclic ring, is an important organic chemical; it is used in the production of melamine resins, which are widely used in wood processing, plastics, paints, papermaking, textiles, leather, electrical goods, and medicines [24]. Melamine detection became a hot topic after cases of infant kidney stones caused by melamine-poisoned infant formula powders in China, and pet deaths from melamine-poisoned feeds in USA [25,26]. Melamine had been illegally added to infant formula powders, dairy products and pet feeds to enhance the protein content measured by the Kjeldahl method. Since 2008, the Chinese government has forbidden the artificial addition of melamine to any food, and the temporary limitation of the volume of melamine in liquid milk is 0.05 mg/Kg. Analytical methods such as high-performance liquid chromatography (HPLC) [27], gas chromatography-mass spectrometry (GC-MS) [28], HPLC-MS/MS [29], surface-enhanced Raman scattering [30], electrochemical immunosensing [31], and enzyme-linked immunoassays (ELISA) [32] have been used for melamine determination. Although these analytical methods are highly sensitive, selective, and efficient, they are time consuming, expensive, and need sophisticated instruments and highly skilled personnel.
Fluorescent sensors provide a promising method for the detection of trace pollutants because such methods are simple, rapid, and have high sensitivity and selectivity [33,34]. The use of combinations of MIPs and fluorescence assays to detect melamine in milk and infant formula milk powder has been reported [35][36][37]. Fluorescent MIP sensors can be divided into two categories based on their preparation method. In the first method, fluorescent materials such as a fluorescent functional monomers or quantum dots are used for molecular imprinting. The interactions between template molecules and the fluorescent functional monomers or quantum dots in the MIPs cause changes in the fluorescence signals, which can be monitored to detect target molecules quantitatively [38,39]. However, the synthesis and modification of fluorescent functional monomers are complicated and quantum dots cannot be uniformly embedded into MIPs. These problems prevent the widespread use of this type of fluorescent MIP sensor. The other method combines the binding properties of MIPs (plastic antibodies) with a competitive fluorescence assay, similar to fluorescence ELISA [2,[40][41][42]. For example, our research group designed a novel fluorescent MIP sensor involving direct competitive binding between triazine herbicide and 5-(4,6-dichlorotriazinyl)aminofluorescein on MMIPs for atrazine detection in tap water [2].
Based on the above, we designed a novel magnetic dummy molecularly imprinted polymer (MDMIP) for fluorescence detection of melamine in milk. The principle is based on competitive absorption between a melamine molecule and a fluorescein isothiocyanate (FITC)-melamine (FITC-Mel) conjugate on the surface active sites of the MDMIPs. The MDMIPs were prepared on the surfaces of Fe 3 O 4 @SiO 2 to absorb specifically melamine molecules. The MDMIPs were characterized, and their adsorption capacity was evaluated based on their static adsorption and Scatchard analysis. Competitive binding between FITC-Mel and melamine was used for quantitative detection of melamine in spiked milk samples. Compared to the above-mentioned melamine detection methods, the detection method established in this paper needs no sophisticated experimental procedure and instruments.
In addition, the present competitive fluorescence assay based on MDMIP is simpler than current Quantum Dots-MIP ( QDs-MIP) fluorescent sensors with embedding of fluorescent QDs into the MIPs.

MDMIP Synthesis
Fe 3 O 4 nanoparticles were prepared using a modified version of a previously reported hydrothermal method [43]. FeCl 3 ·6H 2 O (0.665 g), anhydrous sodium acetate (1.80 g), PEG 1000 (0.5 g), and ethylene glycol (20 mL) were placed in a Teflon-lined stainless-steel autoclave. The autoclave was heated at 180 • C for 8 h. The Fe 3 O 4 nanoparticles were separated magnetically and washed several times with ethanol and water. The precipitate was dried at 60 • C in a vacuum oven.
The Fe 3 O 4 particles were modified with SiO 2 using the method reported by Ni et al. [44]. Fe 3 O 4 (200 mg) was dispersed in 0.1 M hydrochloric acid solution (50 mL) by sonication for 10 min. The Fe 3 O 4 was separated using an external magnet and mixed with ethanol (160 mL) and ultrapure water (40 mL) by sonication for 15 min. Ammonium hydroxide (5.0 mL, 25 wt %) and TEOS (0.7 mL) were then added sequentially. The mixture was stirred at 40 • C for 12 h. The Fe 3 O 4 @SiO 2 nanoparticles were collected magnetically, thoroughly washed with ethanol and water, and dried in a vacuum for 24 h.
The Fe 3 O 4 @SiO 2 nanoparticles were modified with vinyl groups. The nanoparticles (200 mg) were dissolved in methanol (50 mL) by sonication for 15 min, followed by addition of MPS (3 mL). The mixture was incubated at room temperature for 24 h. The Fe 3 O 4 @SiO 2 @MPS nanoparticles were separated using an external magnet, washed with water, and dried in a vacuum for 24 h.
EGDMA (5 mm) and AIBN (25 mg) were then added sequentially to the mixture. The solution was deoxygenated with N 2 for 10 min and the reaction was performed at 60 • C for 24 h, with mechanical stirring. The MDMIPs were separated magnetically and washed with methanol/acetic acid (9:1, v/v) until no template was found by HPLC. Magnetic dummy molecularly non-imprinted polymers (MDNIPs) were prepared using the same procedures but without the dummy template. The procedure for MDMIP preparation is shown in Figure 1A.

Synthesis of FITC-Mel Fluorescent Label
FITC-Mel was synthesized using a modified version of a previously reported method [46]. Melamine (20 mg) was dissolved in ethanol (15 mL) by sonication for 15 min, followed by addition of FITC (20 mg). The mixture was stirred for 24 h at room temperature in the dark. The product was purified by cation-exchange solid-phase extraction and the purified FITC-Mel was identified using electrospray ionization MS (ESI-MS).

Characterization of MDMIPs
The MDMIP morphology and size were examined using TEM. FT-IR spectra of solid samples dispersed in KBr powder were recorded in the range 400-4000 cm −1 . The magnetization was measured at room temperature using vibrating sample magnetometer (VSM).

Adsorption Experiments
The adsorption capacity and selectivity of the MDMIPs were investigated by performing static adsorption experiments. MDMIPs or MDNIPs (2 mg) were added to a 2 mL polyethylene tube containing melamine solutions (1 mL) of various concentrations (0.1-20 mg/L). After incubation for 2 h at room temperature, the MDMIPs or MDNIPs were collected using an external magnet and the supernatant solutions were examined using HPLC. The static binding capacities of the MDMIPs and MDNIPs in melamine adsorption were calculated using the equation Q = (C 0 − C e ) × V/M, where Q is the static binding capacity, C 0 is the initial melamine concentration, C e is the equilibrium concentration of melamine, V is the total volume of the initial melamine solution, and M is the mass of MDMIPs or MDNIPs.
The binding capacity and binding site properties were studied by Scatchard analysis using the equation Q/C e = (Q max − Q)/K D , where C e is the equilibrium concentration of the analyte, K D is the dissociation constant, and Q max is the maximum amount of apparent binding. Plots with Q/C e as the ordinate and Q as the abscissa enabled identification of the types of binding sites, and K D and Q max were calculated based on the linear regression curves and corresponding equations.

Fluorescence Measurements
Fluorescence detection was performed at an excitation wavelength of 478 nm and the emission was recorded from 490 to 800 nm. After template removal, MDMIPs (2 mg) were mixed with a methanol solution (1 mL) containing FITC-Mel (2.08 mg/L) and various concentrations of melamine. The mixture was shaken at room temperature in the dark for 2 h. The supernatant was separated using an external magnetic field. The fluorescence intensity of the supernatant solution was monitored using a microplate reader. The mechanism of competitive fluorescence assay of melamine is shown in Figure 1B.

Preparation of Milk Samples
Blank milk samples (1 g) were spiked with melamine at levels of 0.25, 0.5, 1.0 and 25 mg/kg. The spiked milk was mixed with acetonitrile (9 mL) and the solution was vortexed, then sonicated for 15 min, and centrifuged at 10,000 rpm for 5 min. The supernatant was purified using a PCX-SPE cartridge, as previously reported [47]. The cartridge was first conditioned with methanol (3 mL) and deionized water (5 mL). The milk sample (5 mL) was diluted with water (5 mL). The diluted sample was passed through the cartridge and then the cartridge was washed with methanol (3 mL) and deionized water (3 mL) to remove matrix interference. The analyte was recovered by eluting the cartridge with 5% ammonium hydroxide in methanol (6 mL). The eluate was evaporated to dryness and the residues were redissolved in methanol (1 mL). The analyte solution was filtered through a 0.22 µm poly(vinylidene fluoride) filter membrane and subjected to competitive fluorescence analysis.
The potential use of this competitive fluorescence assay in practical applications was evaluated based on the recoveries from the spiked milk samples. As a reference, HPLC-MS/MS was used to determine the recovery values and relative standard deviations (RSDs) of the spiked milk samples.

Mass Spectrum of FITC-Mel
FITC-Mel was examined using ESI-MS in positive-and negative-ion modes. The molecular weights of FITC and melamine are 389.38 and 126.12, respectively. Figure 2 shows that FITC-Mel has three hydroxyl and two amino groups. In positive-ion mode, the molecular ion peak should appear at 516.5 (M + H) + or 517.5 (M + 2H) 2+ , but we did not observe this molecular ion peak. In negative-ion mode, the molecular ion peak should appear at 514.5 (M − H) − , 513.5 (M − 2H) 2− , or 512.5 (M − 3H) 3− . Figure 2 shows a molecular ion peak at 512.8, corresponding to FITC-Mel and indicating successful combination of FITC and melamine. In the previous literature [48], a commercial fluorescent, 5-(4,6-dichlorotriazinyl) amino fluorescein (DTAF) was used as fluorescent label and structural analog to monitor melamine by performing directly competitive adsorption. However, there is a relatively large difference in structure between DTAF and melamine, which restricts the sensitivity and selectivity of the fluorescent assay. In this work, we prepared FITC-Mel as a novel fluorescent label by combining FITC and melamine using chemical modification. The structure of FITC-Mel has a melamine molecule which will endows it high similarity and provides the fluorescent MIP sensor developed in this paper with good recognition and sensitivity.   Figure 3b shows that the diameter of Fe 3 O 4 @SiO 2 @MPS was approximately 300 nm; this confirms that SiO 2 and MPS layers were uniformly coated on Fe 3 O 4 . After imprinting, the MDMIP diameter was about 400 nm, which suggests successful synthesis of core-shell MDMIPs.   The magnetic properties of Fe 3 O 4 , Fe 3 O 4 @SiO 2 @MPS, and the MDMIPs were studied using VSM. Figure 4B shows that the saturation magnetization values of Fe 3 O 4 , Fe 3 O 4 @SiO 2 @MPS, and the MDMIPs were 55.078, 35.247, and 22.149 emu/g, respectively. The magnetic hysteresis loops show that the remanence and coercivity values were zero, indicating that Fe 3 O 4 , Fe 3 O 4 @SiO 2 @MPS, and the MDMIPs were all superparamagnetic.

Binding Properties
The isothermal adsorption capacities of the MDMIPs and MDNIPs for melamine are shown in Figure 4C. The amounts of melamine adsorbed on the MDMIPs and MDNIPs increased with increasing initial melamine concentration. The binding capacity of the MDMIPs was higher than that of the MDNIPs. The reason might be that the MDMIPs have specific binding sites that match the size and structure of melamine, whereas the MDNIPs have only non-specific binding sites. Scatchard analysis was used to investigate the adsorption site heterogeneities of the MDMIPs and MDNIPs. Figure 4D shows that the Scatchard plot for melamine adsorption on the MDMIPs was not a single linear plot but two linear plots with different slopes. The slopes and intercepts of the two linear regression curves can be used to calculate K D and Q max . The linear regression equation for low initial concentrations was y = −2.8043x + 2400.1; the K D and Q max values were 0.36 µg/mL and 855.8 µg/g, respectively.
The linear regression equation for high initial concentrations was y = −0.5702x + 833.8; the K D and Q max values were 1.75 µg/mL and 1.46 mg/g, respectively. As shown in Figure 4D, two parts of the Scatchard plot curve indicated there were two types of binding sites, nonspecific and specific binding sites. At the low initial concentration of melamine, MIP showed nonspecific adsorption. However, specific recognition binding sites will be activated at the relative high melamine concentrations.

Determination of Melamine in Milk Using Fluorescence Assays
Based on its good selectivity and purification ability, an MIP-SPE cartridge was used for solid-phase extraction before the competitive fluorescence assays. Figure 5A shows the competitive binding reaction between FITC-Mel and melamine at various melamine concentrations. Lower melamine adsorption on the MDMIPs resulted in greater FITC-Mel binding to the MDMIPs and a decrease in the fluorescence intensity of the supernatant. Figure 5B shows that the fluorescent intensity of the supernatant increased linearly with the logarithm of the melamine concentration in the range 0.1-20 mg/L. The linear regression equation for the curve was y = 0.6386x + 0.7098 and the correlation coefficient (R 2 ) was 0.9345. The limit of detection was 0.05 mg/L in negative milk samples (signal/noise = 3). To investigate method selectivity, some common potential interfering substances such as dicyandiamide, trazine, cyanuric acid and chloramphenicol were tested by conducting experiments at the same concentration (1.0 mg/L). As shown in Figure 6, only melamine could induce a considerable change of the fluorescence signal at 526 nm. No dramatic changes at 526 nm were observed in the presence of dicyandiamide, trazine, cyanuric acid and chloramphenicol.
Milk samples spiked with different concentrations of melamine (0.25, 0.5, 1.0, and 2.5 mg/L) were pretreated using liquid-liquid extraction and MIP-SPE. The melamine in the spiked milk samples was determined using the developed competitive fluorescence assay. HPLC/MS/MS was also used to determine the melamine recoveries and RSDs for the spiked milk samples. The results are shown in Table 1; the recovery means for melamine in the spiked milk samples were between 70.2% and 92.7%, and the RSD values were in the range 5.4-12.4% (n = 3). The successful use of the competitive fluorescence assay to determine melamine in spiked milk samples suggests that the developed method would be useful for monitoring melamine in milk and formula powdered milk.

Comparison
The previous literature reported that a fluorescent competitive detection method based on dummy molecularly imprinted polymers (DMIPs) for melamine was developed [48]. DMIPs were prepared using a traditional molecular imprinting technique, which has some limitations, such as poor binding capacity, incomplete template removal, slow mass transfer and poor selectivity. Compared to conventional MIPs, we synthesized a novel magnetic surface MIP by using Fe 3 O 4 @SiO 2 nanoparticles as a support and magnetic core. The novel MDMIPs have a core-shell sphere structure which enhances the absorption capacity and selectivity for melamine and accelerates the mass transfer. More importantly, most adsorption binding sites were distributed on the surface of MIPs, which indicate a good template removal performance. In addition, MDMIPs possess a high magnetic property that allows for the rapid separation of melamine from the complex sample matrix without any additional operations such as centrifuge.

Conclusions
An ELISA-like fluorescence assay for the simple and sensitive detection of melamine in milk was developed. The detection principle of the fluorescence assay was based on competitive binding between melamine molecules and FITC-Mel conjugates on the surface active sites of MDMIPs, similar to fluorescence ELISA. The increase in the fluorescence intensity of the supernatant was linearly related to the logarithm of the melamine concentration in the range 0.1-20 mg/L. The detection limit for melamine in milk was 0.05 mg/L. The MDMIPs in this ELSA-like fluorescence assay are more easily prepared than current fluorescent MIP sensors such as fluorescent functional monomers or quantum dots for molecular imprinting and do not require complicated synthesis and modification. More importantly, the fluorescence assay is easily performed, without the need for professionals, and sophisticated experimental techniques and instrumentation. The results show that this method could be used for the determination of melamine in milk.