Development of hydrophilic magnetic molecularly imprinted polymers by directly coating onto Fe₃O₄ with a water-miscible functional monomer and application in a solid-phase extraction procedure for iridoid glycosides

Abstract

Hydrophilic magnetic molecularly imprinted polymers (HMMIPs) have been synthesized by directly coating onto Fe₃O₄ alkenyl glycosides glucose as a novel water-miscible functional monomer, which introduces an abundance of hydrophilic groups into polymers. Photographs of the dispersion properties and water contact angles demonstrated that these HMMIPs have excellent hydrophilicity compared with those prepared using the traditional hydrophilic functional monomer, methacrylic acid. HMMIPs were characterized by Fourier transform infrared spectroscopy, vibrating sample magnetometry, transmission electron microscopy, scanning electron microscopy, and X-ray diffraction. Binding experiments indicated that HMMIPs had an excellent imprinting effect and high selectivity. Under optimum magnetic molecular imprinted solid phase extraction (MMISPE) conditions, a wide linear range (0.01–50.0 μg mL⁻¹), with low limits of detection and quantification (0.005–0.01 μg mL⁻¹ and 0.019–0.033 μg mL⁻¹, respectively), was achieved for eight iridoid glycosides (IGs). Typical chromatograms obtained using MMISPE showed that major interferences around the IGs were eliminated efficiently and matrix interference was minimized. The results suggest that our newly developed method, combining MMISPE with HPLC, could be used for the selective enrichment and determination of IGs.

---

Introduction

Traditional Chinese medicines (TCMs) have been widely used in China and other Far Eastern countries for thousands of years. TCMs contain several herbal ingredients in a specified ratio and are usually administered as an aqueous solution. The formulated preparations are thus extremely complex matrices containing diverse hydrophilic substances. Enrichment of target hydrophilic compounds from highly complex matrices is still a challenge¹ and this causes problems in the analysis of TCMs. Magnetic molecular imprinted solid phase extraction (MMISPE) has attracted increasing research interest worldwide since it allows rapid and selective enrichment and isolation of the target molecule and analogues.^2–5 Magnetic molecularly imprinted polymers (MMIPs) are, however, normally prepared in hydrophobic organic solvents and typically show poor molecular recognition in aqueous media,⁶ which limits their application in the quality control of TCMs. Hydrophilic magnetic molecularly imprinted polymers (HMMIPs) have been developed using immobilization of hydrophilic Fe₃O₄ onto molecularly imprinted polymers (MIPs),^6,7 modification of the surface of Fe₃O₄ (ref. 8) and Fe₃O₄@SiO₂,^9–13 and modification of graphene oxide.¹⁴ However, the time-resumed polymerization process and relatively rigorous experimental conditions limit their practical application. There are few reports about directly coating onto Fe₃O₄ with a water-miscible functional monomer to prepare HMMIPs. Selection of appropriate hydrophilic functional monomers is essential to guarantee the hydrophilicity.

Iridoid glycosides (IGs), are a group of natural monoterpenoids with an attached glucose moiety that can be isolated from many different plants.^15,16 IGs have been shown to possess a number of pharmacological properties, including cancer chemopreventive, antiviral, immunomodulatory, and cardiovascular activities^17–19 and some IGs have been chosen as marker compounds for the quality control of TCMs. For example, in the Chinese Pharmacopoeia (2015 edition), gardenoside (GDS) is specified for the quality control of Gardenia Fructus²⁰ and Zhizi Jinhua pills (ZJP).²¹ ZJP are a traditional Chinese herbal product that have long been widely used to treat erythema, canker sores, strep throat, constipation, vertigo, and swollen gums. ZJP are prepared from eight different herbs: Gardenia Fructus, Coptis Radix, Scutellaria Radix, Phellodendron stem, Rheum officinale Radix, Lonicera flower, Anemarrhenae Rhizoma, and Snakegourd Fructus. There have been many reports describing the simultaneous analysis of different IGs in a single herb using HPLC-UV.^22–28 However, for the analysis of complex mixtures such as ZJP, which have varying concentrations of different components, the sample often needs to be cleaned up before instrumental analysis. Moreover, multi-component quantification is also more rigorous than analysis of a single active ingredient in the quality control of TCMs.

In previous work, we successfully prepared superhydrophilic MIPs for gastrodin recognition, using alkenyl glycosides glucose (AGG) as the novel water-miscible functional monomer.²⁹ The aim of the present work was to develop HMMIPs to capture a set of IGs in herbal products. MMIPs have normally been prepared using the target compound as a template but natural products may be rare or difficult to produce.^30,31 Because of the special structural skeleton of IGs and their high cost, geniposide (GPS) was chosen as a dummy template in this study. Importantly, the prepared Fe₃O₄ microspheres can be directly coated by the polymer shell with AGG and different cross linkers without any surface modification through precipitation polymerization. Because of its excellent water-compatibility, MMISPE has been successfully used for the targeted extraction of IGs from complex sample matrices.

Experimental procedures

Chemicals and materials

Gardenoside (GDS), geniposide (GPS), geniposidic acid (GPA), genipin-1-O-gentiobioside (GGB), and shanzhiside methyl ester (SAM) were purchased from Yiyan Bio-Tech Co., Ltd. (Shanghai, China). Loganin (LOG), morroniside (MOS), and secoxyloganin (SOL) were obtained from Shilan Technology Co., Ltd. (Tianjin, China). Allyl alcohol and trifluoroacetic acid (TFA) were provided by Xiya Chemistry Co., Ltd (Chengdu, China). D-(+)-Glucose, acetyl chloride, FeCl₃, sodium acetate (NaOAc), and ethylene glycol were purchased from J&K Scientific Ltd. (Beijing, China). Arbutin (ARB), genipin (GEP), N,N′-methylene diacrylamide (MBA, cross linker), divinylbenzene (DVB, cross linker), ethylene glycol dimethacrylate (EGDMA, cross linker), methacrylic acid (MAA, functional monomer), N,O-bismethacryloyl ethanolamine (NOBE, cross linker), and 2,2-azoisobutyronitrile (AIBN, initiator) were purchased from Alfa Aesar (Shanghai, China). HPLC grade reagents were obtained from Fisher Scientific (Fair Lawn, NJ, USA). Other analytical grade solvents were obtained from Aladdin Chemistry Co., Ltd (Shanghai, China). Chemical structures of investigated compounds are shown in Fig. 1.

Fig. 1 Chemical structures of investigated compounds.

Synthesis of MMIPs

AGG was synthesized as previously described²⁹ (see ESI†).

The Fe₃O₄ magnetic microspheres were synthesized by the solvothermal reaction as previously described.^32,33 Briefly, FeCl₃ (5.40 g) was dissolved in ethylene glycol (200 mL) with magnetic stirring. After 40 min, a clear yellow solution was obtained and NaOAc (14.80 g) was then added. After stirring for another 40 min, the resultant solution was transferred into a Teflon-lined stainless-steel autoclave with a capacity of 400 mL. The autoclave was sealed and heated at 200 °C for 8 h, and then cooled to room temperature. The black magnetic microspheres were collected using a magnet, washed three times with water and ethanol, and then dried under vacuum at 60 °C for 12 h.

Synthesis of MMIPs: GPS (model template, 3.0 mmol), AGG (functional monomer, 9.0 mmol), cross-linker (36.0 mmol), and AIBN (initiator, 60 mg) were dissolved in CH₃CN (50 mL). Fe₃O₄ microspheres (350 mg) were then added to the mixture. Pre-polymerization was carried at 60 °C for 3 h, and the final polymerization was completed by mechanical mixing under a nitrogen atmosphere at 80 °C for 6 h. The polymers were collected using an external magnetic field. The template molecules were removed by rinsing with ethanol–Et₃N (90 : 10, v/v) and ethanol until no GPS absorption was detected by HPLC. Finally, MMIPs were dried under vacuum at 60 °C.

Magnetic non-imprinted polymers (MNIPs) were prepared without the introduction of templates.

Instruments and operation parameters

The magnetism of polymers was determined using a SQUID VSM magnetometer (QUANTOM, USA). Fourier transform infrared spectroscopy (FT-IR) was carried out using a Nicolet 710 spectrometer (Foster City, CA, USA). Transmission electron microscopy (TEM) images were obtained using a JEM-2010F transmission electron microscope (JEOL, Tokyo, Japan) and scanning electron microscopy images (SEM) were obtained using a SWPRATM55 scanning electron microscope (Carl Zeiss, AG, Aalen, Germany). The hydrophilic properties of materials were determined using OCA40 video optical contact angle (CA) measuring equipment (Dataphysics, Germany). X-ray diffraction (XRD) was carried out using a D/max-rB diffractometer (Rigaku, Tokyo, Japan). The thermogravimetric analysis was performed on the STA 449F3-QMS403C system (Netzsch, Germany) from the room temperature to 500 °C.

HPLC analysis

An Agilent Series 1120 HPLC system (Agilent Technologies, Santa Clara, CA, USA) was used to analyze the samples. Chromatographic separations were carried out on a YMC-Pack ODS-A analytical column (5 μm, 4.6 × 250 mm). The mobile phase was methanol : H₂O (15 : 85, v/v) at 30 °C. The flow rate was 1.0 mL min⁻¹. Spectra were monitored at 238 nm. The injection volume was 10 μL.

Binding experiments with polymers

The static adsorption properties and dissociation constants of sorbents were investigated by binding experiments using GDS as a model compound. GDS standard solutions (0.01–1.0 mmol L⁻¹) were prepared in H₂O. Polymers (10 mg) were added to each solution (6.0 mL) and the mixtures were shaken for 30 min on a horizontal shaker. The sorbents were then isolated magnetically and the supernatant solutions were filtered through 0.45 μm PTFE membranes. The concentration of GDS in the filtrates was analyzed by HPLC. The adsorption capacity (Q_e, mmol g⁻¹) and imprinting factor (α) were calculated according to the following equation:³⁴

Q_e = (C_i − C_e) × v/m

α = Q_{e MMIPs}/Q_{e MNIPs}

where C_e (mmol L⁻¹) and C_i (mmol L⁻¹) are the equilibrium and initial concentrations of GDS, v (mL) is the volume of solution and m (mg) is the mass of sorbent.

The data were further processed using the Scatchard equation:³⁵

Scatchard equation: Q_e/C_e = (Q_max − Q_e)/K_d

where Q_max (mmol g⁻¹) is the amount of GDS adsorbed at saturation and K_d (mmol L⁻¹) is the equilibrium dissociation constant.

Selectivity was investigated in experiments with ten standard solutions (0.3 mmol L⁻¹) of eight IGs and two analogues (ARB and GEP). 6.0 mL of each solution was shaken with 10 mg of sorbents for 30 min. The sorbents were then isolated magnetically and the supernatant solutions were filtered through 0.45 μm PTFE membranes. The concentration of each analyte in the filtrates was analyzed by HPLC.

Sample preparation

Powdered ZJP (0.1 g) was extracted with water (100 mL) using ultrasonication at 25 °C for 30 min. The supernatant was filtered through a 0.45 μm PTFE membrane and the mixture was accurately diluted with water to 100 mL and thoroughly mixed.

Application of sorbents to real samples

Sorbents (100 mg) were dispersed in ZJP solution (100.0 mL) and the mixtures were shaken for 10 min. The sorbents were then collected rapidly from the mixtures using an external magnetic field. The IGs were eluted with 10 mL of ethanol–Et₃N (90 : 10, v/v) at 65 °C. The eluate was analyzed by HPLC.

Results and discussion

Optimization of the synthesis of MMIPs

According to the literature, the synergetic hydrogen bond interaction between different active groups and the hydroxyl groups of Fe₃O₄ microspheres was strong enough for Fe₃O₄ to capture the oligomer during the polymerization process.^36,37 In this study, because polymerization systems was abundant with hydroxyl groups, carbonyl groups, and amide groups, the precipitation polymerization was performed to afford the core–shell microspheres in the presence of Fe₃O₄ as seeds without any surfactant and stabilizer.

Molecular recognition is manly based on intermolecular interactions between functional monomer and template, such as electrostatic, π–π, hydrogen bond, coordination bond or van der Waals interactions. Hydrogen bonds may be formed between GPS and AGG as the interaction for binding sites, whereby abundant hydrogen bond acceptors and donors existed on GPS and AGG. The mole ratio of template : functional monomer : cross-linker is a key factor that influences the specific adsorption properties of MMIPs. Therefore, several different MMIPs, containing five different cross-linkers, were designed and prepared. The specific adsorption properties, Q_e and α, of MMIPs were investigated. MMIPs prepared at a ratio of 1 : 3 : 12, which used AGG as the functional monomer and MBA, EGDMA, NOBE, and DVB as cross-linkers, had higher α values (Table S1† entries 3, 9, 15, and 21). Compared with the traditional water-miscible functional monomer (MAA), the values of Q_e and α for MMIPs/AGG were also higher (Table S1† entry 27). Taking into account Q_e and α values, a mole ratio of 1 : 3 : 12 for template : functional monomer : cross-linker was selected to prepare MMIPs. Moreover, to evaluate the effect of shell thickness on the adsorption property, different MMIPs were fabricated by altering the polymerization time. As shown in Fig. S1,† the imprinting shell obviously thickened from 13 to 45 nm with increasing the polymerization time from 3 to 9 h. The values of Q_e for MMIPs also increased from 0.034 to 0.11 mmol g⁻¹. The shell thickness was no change with the continuing prolongation of reaction time. Therefore, 9 h of polymerization time was used to prepare MMIPs.

Characterization of MMIPs

FT-IR spectroscopy was used to monitor the successful preparation of the MMIPs. As shown in Fig. 2A, the spectrums were relative to MMIPs made with MBA, EGDMA, NOBE, and MAA as cross-linker, respectively. Obviously, strong absorption peaks for Fe–O vibration seen at ∼586 cm⁻¹ represented the existence of Fe₃O₄. The typical C=O stretching band was also present in spectrum Fig. 2A-a (1653 cm⁻¹ for MBA), 2A-b (1731 cm⁻¹ for EGDMA), 2A-c (1722 cm⁻¹ for NOBE), and 2A-e (1655 cm⁻¹ for MAA). The absorption peak at 1603 cm⁻¹ arising from the aromatic skeletal vibration of benzene ring was characteristic for DVB (Fig. 2A-d). These results suggested that the Fe₃O₄ molecules were successfully encased during the coating polymerization process.

Fig. 2 (A) FT-IR spectra of MMIPs. (B) The profiles of a water drop on the films of MMIPs. (C) The detailed photographs for the dispersion stability of polymers in water (1.0 mg mL⁻¹) at 25 °C after their ultrasonically dispersed solutions being settled down for 2 h. (D) Magnetization curves of Fe₃O₄ and MMIPs at 300 K. (E) Typical X-ray powder diffraction pattern of MMIPs. Polymers information: (a–d) MMIPs prepared with AGG as the functional monomer and MBA (a), EGDMA (b), NOBE (c), and DVB (d) as cross linker, respectively. (e) MMIPs prepared with MAA as the functional monomer and MBA as cross linker.

The hydrophilicity of MMIPs is a critical factor for enrichment and separation of analytes from aqueous media.³⁸ The static water contact angle (CA) has previously been used to evaluate the degree of hydrophilicity.³⁹ The static water CAs of MMIPs prepared with AGG as the functional monomer (except those prepared using DVB as cross-linker) were all <30° (Fig. 2B and Table S1† entries 1–18). The high hydrophobicity of DVB greatly restrains the water-compatibility of MMIPs. Reference magnetic molecularly imprinted polymers (R-MMIPs) exhibited water CAs >120° (Table S1† entries 25–30) and had poor dispersion stability in pure water (Fig. 2C; see also Fig. S2†). In our previous work, the static water CA of MIPs, prepared by using the bulk polymerization with AGG as the functional monomer, was close to 0°. In sharp contrast, the CA of MIPs prepared by using MAA was 140 ± 2°.²⁹ Residues of the relatively polar compound MAA in MMIPs are not sufficiently hydrophilic to overcome the hydrophobicity of other aliphatic chains. Similar results were obtained with blank magnetic molecularly imprinted polymers (B-MMIPs) (Table S1† entries 31–33), suggesting that AGG residues could improve the hydrophilicity of MMIPs than using hydrophilic cross-linkers or hydrophilic functional monomers.

Magnetic characteristics of Fe₃O₄ microspheres and typical polymers (Table S1† entries 3, 9, and 15) were measured by vibrating sample magnetometry (VSM) at 300 K. Saturation magnetization values were 79.8, 39.9, 33.9, and 9.8 emu g⁻¹, respectively (Fig. 2D). The decrease in saturation magnetization values also indicated that Fe₃O₄ microspheres were encapsulated during polymerization. Efficient separation from aqueous solution was easily achieved by an external magnetic field in <10 s.

A typical XRD pattern of Ms-2 is shown in Fig. 2E. Six relatively discernible diffraction peaks, corresponding to Fe₃O₄ (2θ = 30.1°, 35.5°, 43.1°, 53.4°, 57.3°, and 62.7°), were observed in the curves of MMIPs-2. These six peaks correspond well with the 2θ values of magnetite, which are indexed as (220), (311), (400), (422), (511) and (440) in the database of the Joint Committee on Powder Diffraction Standards (JCPDS card: 19-629) file. This indicates that the crystal structure of Fe₃O₄ was not changed during the polymerization process.

To further prove Fe₃O₄ directly coated by MIPs, information about the morphology of MMIPs-2, MMIPs-5 and MMIPs-8 was assessed by SEM and TEM (Fig. 3). TEM images of MMIPs showed the Fe₃O₄ microspheres as a core (dark), surrounded by a layer (gray) of MIPs, indicating that MIPs were coated onto the surface of the Fe₃O₄ microspheres. Moreover, the average diameters of Fe₃O₄ microspheres and MMIPs-2 were 377 and 411 nm with the standard deviation (SD) of 68 and 43, respectively, which were calculated from the SEM data measuring the size of, at least, 100 particles (Fig. S3†). Larger size also implied that Fe₃O₄ microspheres were coated during polymerization.

Fig. 3 SEM images of MMIPs-2 (A), MMIPs-5 (B), and MMIPs-8 (C). TEM images of MMIPs-2 (D), MMIPs-5 (E), and MMIPs-8 (F).

Because of the superior hydrophilicity and saturation magnetization of MMIPs/MBA, MMIPs-2 was chosen as the sorbent in the following studies.

Binding properties of MMIPs and MNIPs

The adsorption capacities of MMIPs and MNIPs increased rapidly with increasing initial GDS concentration (Fig. 4A). MMIPs exhibited a significantly higher adsorption capacity than MNIPs over the range of concentrations tested (0.01–1.0 mmol L⁻¹). The saturation binding was observed with the initial concentration of 0.8 mmol L⁻¹ and 0.6 mmol L⁻¹ for MMIPs and MNIPs, respectively. These results indicated that recognition sites had been created during the molecular imprinting process.

Fig. 4 (A) The adsorption isotherms of GDS on MMIPs-2 and MNIPs-2. Error bars represent one standard deviation for three measurements. (B) Scatchard plots of the MMIPs-2 and MNIPs-2 isotherms.

The binding data for MMIPs and MNIPs were further processed using Scatchard analysis (Fig. 4B). The binding data of MMIPs produced two different straight lines, suggesting that two types of binding site were present in the MMIPs: lower affinity binding sites (MMIPs low) and higher affinity binding sites (MMIPs high).⁴⁰ The linear regression equations for the two different straight lines were Q_e/C_e = −6.17Q_e + 1.79 (R² = 0.9798) and Q_e/C_e = −122.86Q_e + 6.80 (R² = 0.9749). The Q_max and K_d values were calculated as 0.29 mmol g⁻¹ and 0.16 mmol L⁻¹, respectively, for the lower affinity binding sites, and as 0.055 mmol g⁻¹ and 0.0081 mmol L⁻¹, respectively, for the higher affinity binding sites. The linear regression equation for adsorption by MNIPs was Q_e/C_e = −1.10Q_e + 0.15, indicating homogeneous binding sites, with Q_max and K_d values of 0.14 mmol g⁻¹ and 0.91 mmol L⁻¹, respectively. The higher affinity binding sites have a lower dissociation rate and displayed a smaller adsorption capacity than the lower affinity binding sites. The overall adsorption capacity of MMIPs for GDS is higher than that of MNIPs, as indicated by the K_d values. Higher Q_max and lower K_d values suggest that MMIPs possess higher affinity and binding capacity than MNIPs.

Evaluation of selectivity of MMIPs

ZJP contain eight typical IGs, GDS, GPS, GPA, GGB, SZM, LOG, MOS, and SOL. Ten standard solutions (0.3 mmol L⁻¹) of eight IGs and two analogues (ARB and GEP) were used to assess the selectivity of the MMIPs. MMIPs-2 exhibited a higher IF value for the template molecule GPS than for the other seven IGs (Fig. 5). α values were 4.00 for GPS, 3.55 for GDS, 3.12 for GPA, 3.45 for SAM, 2.62 for GGB, 3.03 for LOG, 2.84 for MOS, and 2.50 for SOL. Although imprinting cavities exhibit enough mobility and flexibility to extend recognition towards analogues of the template, even owning larger molecular size than that of template,^41–44 the lower α values for GGB and SOL may reflect the slight differences in the molecular skeletons of GGB and SOL compared with that of GPS. α values <2.0 were obtained for ARB and GEP, indicating the presence of non-specific binding sites in the MMIPs for analogues with completely different molecular structures to that of GPS. The significantly high values for the selectivity coefficient (k^sel) and relative selectivity coefficient (k^rel) of MMIPs-2 demonstrated the higher selectivity of MMIP2 (Table S2†). MMIPs-14, prepared using MAA, exhibited IF values <2.0 for all eight IGs (Fig. 5), together with low k^sel and k^rel values (Table S3†). This indicated that MMIPs-14, which was prepared using a traditional hydrophilic functional monomer, absorbed IGs non-specifically from aqueous media.

Fig. 5 The specific binding behaviors of MMIPs and MNIPs for eight IGs and two analogues. Error bars represent one standard deviation for three measurements.

Chemical/thermal stability and reusability of MMIPs

The chemical/thermal stability of sorbents is one of the most important factors for practical applications. The specific recognition ability of MMIPs did not noticeably deteriorate after treatment with an acidic solution (30% hydrochloric acid for 12 h) or a basic solution (20% NaOH solution for 12 h), or when the MMIPs were heated to a high temperature (100 °C for 12 h) (Table S4,† IF > 2.8). Moreover, MMIPs before and after damaged at 100 °C for 12 h did not showed appreciable differences in morphology (Fig. S4†). The result implied that MMIPs did not shrink or crack when completely dried. Additionally, the thermogravimetric analysis was investigated to ensure that the decomposition temperature of MIPs layer was approximately 330 °C (Fig. S5†). The excellent chemical/thermal stability may be due to the strong chemical bonds of the polymers and suggests that MMIPs could be used in extreme conditions. What is move, the adsorption capacity of MMIPs was also essentially constant above 0.09 mmol g⁻¹ over ten consecutive adsorption–desorption cycles (Table S4†), indicating that the adsorption process is reversible during the purification procedure. These results further demonstrate that these sorbents are promising candidates for economic large scale application.

Optimization of MMISPE parameters

Molecularly imprinted polymers are artificial materials with selective binding cavities that specifically recognize template molecules, or structurally related compounds, in appropriate solvents.^45,46 We have optimized the group extraction efficiency of MMIPs for eight structurally similar IGs (GDS, GPS, GPA, GGB, SZM, LOG, MOS, and SOL) by investigating the effect of sample pH, extraction time, amount of sorbent, and elution solvent.

The effect of pH on the stability of the eight IGs was firstly evaluated over the pH range 2.0–12.0 in the absence of MMIPs-2. At pH 2.0–6.0, after shaking for 30 min, concentrations of the model molecule GDS detected by HPLC were below 0.25 mmol L⁻¹, whereas, at pH 7.0–12.0, the concentration remained almost constant at the initial value of 0.3 mmol L⁻¹ (Fig. S6†). The unsatisfactory stability of IGs under acidic conditions is mainly due to hydrolysis of the glycosidic bonds. Therefore, the pH of samples was investigated in a range from 7.0 to 11.0. Recoveries were stable over the pH range 7.0–8.0 and decreased significantly at pH 11.0 for all eight IGs (Fig. 6A). As the pH increased further, the recoveries of all eight IGs decreased steadily, especially for GPA and SOL. This might be mainly related to the dissociation constants (pK_a). The pK_a values, calculated using the SciFinder Scholar database,⁴⁷ of the eight IGs were 12.80 ± 0.70 (GPS), 12.28 ± 0.60 (GDS), 4.49 ± 0.60 (GPA), 12.80 ± 0.70 (GGB), 12.78 ± 0.70 (SAM), 12.80 ± 0.70 (LOG), 12.54 ± 0.70 (MOS), and 4.40 ± 0.70 (SOL). The low recoveries at pH 11.0 are likely due to formation of organic salts by ionization, especially for GPA and SOL. The imprinting interactions between the analytes and sorbents were further decreased by changes in the 3D structures of the IGs. Over the pH range 7.0–11.0, MNIPs-2 showed lower recoveries than MMIPs-2 for all eight IGs (Fig. 6B). Group-selective enrichment for the eight IGs was thus obtained through MMISPE over the pH range 7.0–8.0. Most aqueous solutions of TCMs used for quality control have a pH in the range 7.0–8.0 and could, therefore, be directly extracted using the MMISPE method developed in this work.

Fig. 6 Extraction and desorption parameters optimization: (A) pH effect on MMIPs-2; (B) pH effect on MNIPs-2. Error bars represent one standard deviation for three measurements.

To optimize extraction efficiency, the ability of different amounts (4, 6, 8, 10, 20, 30 mg) of MMIPs-2 to extract IGs (GDS, GPA, LOG and SOL) from spiked water samples (10 mL, 1.0 μg mL⁻¹ of each IG) was investigated. When the amount of MMIPs-2 was increased to 10 mg, the recoveries of GPS, GDS, GPA, SAM, GGB, LOG, MOS, and SOL improved to 95.7%, 94.5%, 88.2%, 92.1%, 91.2%, 93.3%, 94.1%, and 87.1%, respectively (Fig. S7†). There was no further increase in recovery with increased amounts of sorbent. Therefore, 10 mg was chosen as the amount of sorbent for the following experiments.

Extraction time is one of the key factors during the adsorption process. The amount of the model molecule GDS that was extracted from spiked water samples (10 mL, 1.0 μg mL⁻¹) with different extraction times (2–40 min) was investigated. MMIPs-2 reached adsorption equilibrium after 10 min (Fig. S8†). In sharp contrast, satisfactory extraction efficiency was not achieved with either MNIPs-2 or MMIPs-14 after 10 min and recoveries were low. The excellent hydrophilicity and strong recognition sites of MMIPs-2 may account for the higher extraction efficiency of this sorbent. Based on these results, an extraction time of 10 min was used in the subsequent experiments.

It is essential to ensure that the analytes are completely eluted from the sorbent. First, ethanol was used as the eluting solvent. Unfortunately, the recoveries of IGs were below 50% even at 80 °C. Because of the instability of IGs in acidic media and their insolubility in organic solvents such as acetonitrile, the ability of ethanol–Et₃N (90 : 10 (v/v), 10 mL) to elute eight IGs from MMIPs-2 was examined at different temperatures. Higher temperature can cause the molecules to move faster and improve the mass transfer rate for leasing the analytes. As the temperature of the eluent was increased, the recoveries of eight IGs showed an upward trend and reached a peak at 65 °C (Fig. S9†). Increasing the temperature further did not improve recoveries. Therefore, ethanol–Et₃N (90 : 10 (v/v) 10 mL) at 65 °C was selected as the eluting solvent.

Method validation

Method validation was carried out to evaluate the feasibility of using the newly developed MMISPE-HPLC method to analyze IGs in real aqueous samples of TCMs. Linearity, accuracy, precision, limit of detection (LOD) and limit of quantitation (LOQ) were evaluated. The accuracy (recovery) and precision (RSD) of the method were determined using blank samples spiked at three levels of IGs (0.500, 4.00 and 10.0 μg mL⁻¹). The blank samples, which were all confirmed to lack detectable IGs, were solutions obtained by extraction of Scutellaria Radix, Coptis Radix, and Phellodendron stem. The recoveries of IGs in the spiked samples were in the range 87.6–98.8%, with relative standard deviation (RSD) values <6.1% (Table 1). The linearity range, LOD, and LOQ values for the eight IGs under optimum extraction conditions are listed in Table 2. Excellent linearity was obtained over the range 0.01–50 μg mL⁻¹, with a correlation coefficient (R²) ≥ 0.990. The LOD was 0.005–0.01 μg mL⁻¹ and the LOQ was 0.019–0.033 μg mL⁻¹, values that are lower than those reported earlier by HPLC-UV analysis. (Table S5†). The LOD were higher than those found for HPLC/MS analysis method. However, the expensive instruments have limited its wide application. These results demonstrate that an applicable, reliable, and sensitive MMISPE-HPLC method has been established.

Table 1 Analysis of eight IGs in different spiking levels using MMIPs-2-SPE-HPLC

IGs	Spiking level (μg mL^−1)	Scutellaria Radix		Coptis Radix		Phellodendron stem
		Recovery (%)	RSD (%, n = 3)	Recovery (%)	RSD (%, n = 3)	Recovery (%)	RSD (%, n = 3)
GPS	0.500	95.2	4.61	94.1	5.12	94.4	5.33
	4.00	95.9	4.73	93.4	4.24	97.1	4.64
	10.0	93.2	4.95	96.7	4.86	98.2	3.85
GDS	0.500	97.1	3.37	97.3	4.58	95.3	3.77
	4.00	96.7	3.92	95.6	4.31	95.4	3.34
	10.0	98.8	4.14	97.2	5.73	97.2	4.92
GPA	0.500	88.5	5.16	88.9	4.15	87.9	4.71
	4.00	91.2	4.48	92.2	5.37	90.5	3.72
	10.0	95.2	5.21	94.4	4.48	94.7	3.54
SAM	0.500	90.2	4.33	92.7	4.46	93.1	4.65
	4.00	94.7	5.15	94.3	5.65	92.1	5.86
	10.0	95.1	3.77	96.3	3.84	94.4	3.53
GGB	0.500	87.6	5.19	88.1	4.34	87.9	4.12
	4.00	88.2	4.32	89.3	5.23	88.9	4.71
	10.0	89.2	6.14	89.5	5.42	89.6	4.32
LOG	0.500	97.3	4.76	98.1	3.86	97.5	4.53
	4.00	97.1	5.18	97.5	4.67	96.5	4.65
	10.0	96.4	3.41	96.7	4.88	95.4	4.36
MOS	0.500	95.3	4.73	96.2	3.99	96.7	4.13
	4.00	93.2	5.15	95.4	5.24	95.6	4.51
	10.0	98.1	4.27	97.7	5.13	97.3	5.22
SOL	0.500	87.8	4.29	88.1	3.92	89.4	4.63
	4.00	88.2	4.92	88.9	4.31	90.2	5.91
	10.0	88.3	5.34	87.9	5.05	89.5	5.16

---

Table 2 The validation data of MMIPs-2-SPE-HPLC

IGs	Linear range (μg mL^−1)	R^2	LOD (μg mL^−1)	LOQ (μg mL^−1)
GPS	0.01–50	0.993	0.01	0.032
GDS	0.01–50	0.994	0.007	0.024
GPA	0.01–50	0.992	0.01	0.033
SAM	0.01–50	0.997	0.005	0.019
GGB	0.01–50	0.995	0.008	0.028
LOG	0.01–50	0.999	0.007	0.023
MOS	0.01–50	0.997	0.01	0.031
SOL	0.01–50	0.992	0.01	0.032

---

Application in real sample analysis

The molecular recognition ability of MMIPs and the practicability of the MMISPE-HPLC method were further evaluated using real ZJP samples. The samples were pretreated using the method described in section “Sample preparation”. In order to make a direct comparison, the eight IGs were determined in 100 mL samples treated as following three methods: (1) no pre-treatment and direct injection, (2) concentrated to 10 mL, and (3) MMSIPE. Typical chromatograms are shown in Fig. 7. Without pretreatment, chromatographic peaks for the IGs were barely visible since the concentrations were very low (Fig. 7a). Although peaks for the IGs were seen after the samples were concentrated to 10 mL, the impurity peaks were also very obvious and matrix interference was maximized (Fig. 7b). Both direct injection and sample concentration are thus inadequate for ZJP quality control. After selective enrichment of the IGs using MMISPE, the large interferences around the IGs were eliminated efficiently and matrix interference was minimized (Fig. 7e). Compared with the cleaner chromatograms obtained using MMISPE, impurity peaks around GDS and GPA were still obvious after using commercial C18-SPE materials (Fig. 7c). Moreover, the peaks obtained in C18-SPE materials and MNISPE (Fig. 7d) were obviously lower than the peaks obtained in MMISPE, demonstrating the lower recovery of IGs.

Fig. 7 Chromatograms of the ZJP samples. ZJP extracts without any pre-treatment (direct injection) (a); ZJP extracts with concentration (100.0 mL extracts were concentrated to 10.0 mL) (b); eluting solutions from C18-SPE column (c); desorption solutions from MNISPE (d); desorption solutions from MMISPE (e); standards of eight IGs (f). Peak identifications: 1, GPA; 2, GDS; 3, SAM; 4, MOS; 5, GGB; 6, GPS; 7, LOG; 8, SOL.

The newly established quantitative method was then used to determine the concentrations of eight IGs in five commercially available ZJP products. It can be seen from Fig. 7B that only four chromatographic peaks for IGs, including GGB, GPS, LOG, and SOL, had no matrix interference. As summarized in Table S6,† the HPLC detection concentrations of these four IGs in ZJP (TRT-4082322) were 4.39, 7.90, 3.36, and 3.52 mg g⁻¹, respectively. After treated with MMISPE, the concentrations were 3.87, 7.51, 3.35, and 3.13 mg g⁻¹, respectively. The recoveries of four IGs were all above 87%, in sharp contrast to SampliQ C18 as the conventional SPE column (below 40%). These results indicate that the MMIPs prepared in this work have excellent molecular recognition ability and that the proposed MMISPE-HPLC method can be applied to enrich and detect IGs in aqueous matrices.

Conclusions

Hydrophilic magnetic molecularly imprinted polymers have been synthesized by directly coating onto Fe₃O₄ with AGG as a novel water-miscible functional monomer. This introduced abundant hydrophilic groups into the MMIPs and made the magnetic materials very compatible with water. In the present study, eight IGs were chosen as marker compounds for the quality control of ZJP, whereas only GDS is specified in the Chinese Pharmacopoeia. MMIPs prepared using GPS as the dummy template displayed excellent molecular recognition towards IGs in aqueous media. The MMISPE-HPLC method presented here can be used for the selective enrichment and determination of IGs from aqueous solutions of TCMs. Because of its excellent hydrophilicity, high selectivity, and high sensitivity, MMISPE-HPLC has the potential to be an outstanding tool for quality assessments of TCMs.

Acknowledgements

This study was financially supported by the Natural Science Foundation of Shandong (ZR2015BQ005) and a research grant from Shandong Analysis and Test Center.

References

1. X. Wang, A. Zhang, G. Yan, Y. Han and H. Sun, TrAC, Trends Anal. Chem., 2014, 63, 180–187.
2. T. Kubo and K. Otsuka, TrAC, Trends Anal. Chem., 2015 DOI:10.1016/j.trac.2015.08.008.
3. J. Javidi, M. Esmaeilpour and M. R. Khansaria, RSC Adv., 2015, 5, 73268–73278.
4. L. Ye, Anal. Bioanal. Chem., 2016, 408, 1727–1733.
5. Z. Yang, Q. Cai, N. Chen, X. Zhou and J. Hong, RSC Adv., 2016, 6, 12193–12204.
6. M. Zhao, C. Zhang, Y. Zhang, X. Guo, H. Yan and H. Zhang, Chem. Commun., 2014, 50, 2208–2210.
7. J. Pan, W. Zhu, X. Dai, X. Yan, M. Gan, L. Li, H. Hang and Y. Yan, RSC Adv., 2014, 4, 4435–4443.
8. Y. Hao, R. Gao, L. Shi, D. Liu, Y. Tang and Z. Guo, J. Chromatogr. A, 2015, 1396, 7–16.
9. F. Lu, H. Li, M. Sun, L. Fan, H. Qiu, X. Li and C. Luo, Anal. Chim. Acta, 2012, 718, 84–91.
10. J. Zhou, N. Gan, T. Li, F. Hu, X. Li, L. Wang and L. Zheng, Biosens. Bioelectron., 2014, 54, 199–206.
11. A. Mehdinia, T. B. Kayyal, A. Jabbari, M. O. Aziz-Zanjani and E. Ziaei, J. Chromatogr. A, 2013, 1283, 82–88.
12. F. Chen, X. Xie and Y. Shi, J. Chromatogr. A, 2013, 1300, 112–118.
13. L. Xu, J. Pan, J. Dai, X. Li, H. Hang, Z. Cao and Y. Yan, J. Hazard. Mater., 2012, 233–234, 48–56.
14. S. Pan, X. Chen, X. Li, M. Cai, H. Shen, Y. Zhao and M. Jin, J. Chromatogr. A, 2015, 1422, 1–12.
15. C. A. Boros and F. R. Stemrmitz, J. Nat. Prod., 1990, 53, 1055–1147.
16. C. Seger, M. Godejohann, L. H. Tseng, M. Spraul, A. Girtler, S. Sturm and H. Stuppner, Anal. Chem., 2005, 77, 878–885.
17. K. Takao, T. Midori, T. Harukuni and N. Hoyoku, Cancer Lett., 2000, 157, 87–92.
18. H. J. Koo, K. H. Lim, H. J. Jung and E. H. Park, J. Ethnopharmacol., 2006, 103, 496–500.
19. Y. Y. Sung, A. Y. Lee and H. K. Kim, J. Ethnopharmacol., 2014, 156, 33–40.
20. Chinese Pharmacopoeia Commission, Pharmacopoeia of the People's Republic of China, People's Medical Publishing House, Beijing, 2015, vol. 1, pp. 248–249.
21. Chinese Pharmacopoeia Commission, Pharmacopoeia of the People's Republic of China, People's Medical Publishing House, Beijing, 2015, vol. 1, pp. 1165–1166.
22. Y. Song, S. L. Li, M. H. Wu, H. J. Li and P. Li, Anal. Chim. Acta, 2006, 564, 211–218.
23. X. Y. Cao and Z. Z. Wang, Phytochem. Anal., 2010, 21, 348–354.
24. L. Yang, Y. Wang, L. Wang, H. Xiao, Z. Wang and Z. Hu, J. Chromatogr. A, 2009, 1216, 2098–2103.
25. M. L. He, X. W. Cheng, J. K. Chen and T. S. Zhou, Chromatographia, 2006, 64, 713–717.
26. M. C. Bergonzi, C. Righeschi, B. Isacchi and A. R. Bilia, Food Chem., 2012, 134, 1199–1204.
27. S. A. Coran, S. Mulas and A. Vasconi, J. Chromatogr. A, 2014, 1325, 221–226.
28. Y. Han, J. Wen, T. Zhou and G. Fan, Food Chem., 2015, 188, 648–657.
29. W. Ji, M. Zhang, D. Wang, X. Wang, J. Liu and L. Huang, J. Chromatogr. A, 2015, 1425, 88–96.
30. J. Yang, Y. Li, J. Wang, X. Sun, R. Cao, H. Sun, C. Huang and J. Chen, Anal. Chim. Acta, 2015, 872, 35–45.
31. W. Rao, R. Cai, Y. Yin, F. Long and Z. Zhang, Talanta, 2014, 128, 170–176.
32. X. Xu, C. Deng, M. Gao, W. Yu, P. Yang and X. Zhang, Adv. Mater., 2006, 18, 3289–3293.
33. F. Liu, H. Zhang, C. Lin and X. Zhang, J. Am. Soc. Mass Spectrom., 2008, 19, 865–873.
34. W. Ji, L. Chen, X. Ma, X. Wang, Q. Gao, Y. Geng and L. Huang, J. Chromatogr. A, 2014, 1342, 1–7.
35. S. Zeng, Y. She, B. Jiao, G. Liu, J. Wang, X. Su, X. Ma, M. Jin, F. Jin and S. Wang, RSC Adv., 2015, 5, 94997–95006.
36. G. Liu, L. Li and X. Yang, Polymer, 2008, 49, 4776–4783.
37. B. Liu, W. Zhang, F. Yang, H. Feng and X. Yang, J. Phys. Chem. C, 2008, 112, 6245–6252.
38. Y. Ma, G. Pan, Y. Zhang, X. Guo and H. Zhang, Angew. Chem., Int. Ed., 2013, 52, 1511–1514.
39. L. Zhang, J. Wu, Y. Wang, Y. Long, N. Zhao and J. Xu, J. Am. Chem. Soc., 2012, 134, 9879–9881.
40. P. Lenain, J. D. D. Mavungu, P. Dubruel, J. Robbens and S. D. Saeger, Anal. Chem., 2012, 84, 10411–10418.
41. M. Contin, S. Flor, M. Martinefski, S. Lucangioli and V. Tripodi, Anal. Chim. Acta, 2014, 807, 67–74.
42. M. Szumski, D. Grzywiński, W. Prus and B. Buszewski, J. Chromatogr. A, 2014, 1364, 163–170.
43. W. Du, Q. Fu, G. Zhao, P. Huang, Y. Jiao, H. Wu, Z. Luo and C. Chang, Food Chem., 2013, 139, 24–30.
44. J. Luo, L. Zhang, D. Chen, P. Wang, J. Zhao, Y. Peng, S. Du and Z. Zhang, Analyst, 2012, 137, 2891–2902.
45. D. Xiao, P. Dramou, N. Xiong, H. He, D. Yuan, H. Dai, H. Li, X. He, J. Peng and N. Li, Analyst, 2013, 138, 3287–3296.
46. R. Zhang, X. Guo, X. Shi, A. Sun, L. Wang, T. Xiao, Z. Tang, D. Pan, D. Li and J. Chen, Anal. Chem., 2014, 86, 11705–11713.
47. Software calculated value from SciFinder Scholar database 2008, http://www.cas.org/products/sfacad/.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra06935e

‡ These authors have equal contribution to this work. Both persons are the first authors.

---