Magnetic solid-phase extraction based on Fe₃O₄ nanoparticle retrieval of chitosan for the determination of flavonoids in biological samples coupled with high performance liquid chromatography

Abstract

A novel and facile magnetic solid-phase extraction method (MSPE) based on the two-step magnetic retrieval of chitosan was developed and applied for the first time in bio-matrix samples for the simultaneous extraction and determination of trace flavonoids. A systematic study of the different types of chitosan, Fe₃O₄ nanoparticles (NPs), the analytes and the matrices is presented. Owing to a higher extraction efficiency and capacity for analytes, chitosan with a 95% degree of deacetylation and average molecular weight of 1.0 × 10⁶ and Fe₃O₄ NPs synthesized by the solvothermal method were selected as MSPE materials. Three analytes of luteolin, quercetin and kaempferol can be quantitatively extracted and simultaneously determined coupled with high performance liquid chromatography (HPLC) in urine and serum samples. No interferences were caused by proteins or endogenous compounds. Good linearities (r² > 0.9990) for all calibration curves were obtained, and the limits of detection (LODs) for quercetin, luteolin, and kaempferol were 1.0, 0.5 and 0.7 ng mL⁻¹ in urine samples and 10, 2 and 5 ng mL⁻¹ in serum samples, respectively. Satisfactory recoveries (90.1–106.5%, 91.1–105.5% and 93.5–108.8% for quercetin, luteolin and kaempferol) in biological samples were achieved.

---

1. Introduction

Flavonoids, which are one of the largest groups of natural phenols present in medicinal plants, have drawn considerable public attention over several previous decades due to antioxidant activity.^1,2 Numerous studies have revealed that flavonoids exert a positive influence on health owing to alleviating and preventing many serious diseases, such as inflammation, cancer, cardiovascular problems, arteriosclerosis, bleeding, allergies and so on.^3,4 Therefore, flavonoids are always widely used as remedies because of their biological and physiological importance. In order to investigate different flavonoid consumptions and metabolisms more efficiently, there is a necessity to develop analytical methods to achieve rapid and sensitive simultaneous measurement of trace-level flavonoids in human fluids. To date, several approaches have been described for the determination of flavonoids in biological samples, including high-performance liquid chromatography (HPLC) coupled with different detectors, such as ultraviolet detection (UV),^5,6 fluorometric detection (FD),⁷ electrochemical detection (ED),⁸ liquid chromatography-mass spectrometry (LC-MS),⁹ capillary electrophoresis (CE)¹ and gas chromatography-mass spectrometry (GC-MS).^4,10 Even though some of these methods are sensitive and capable of low detection limits, they are unsatisfactory for the quantitative determination of flavonoids because of the interference of the complex matrix and their extremely low concentration in body fluids. Therefore, it is crucial to separate and enrich flavonoids prior to determination. However, the traditional pretreatment procedures were always time-consuming and solvent dependent. In order to solve the aforementioned problems, it is necessary to develop a practicable enrichment material with high extraction efficiency and time-saving properties for the separation and determination of flavonoids.

Nowadays, bioadsorbents have attracted extensive attention owing to some outstanding advantages, such as their nontoxic, biodegradable, and biocompatible properties and so on. In particular, chitosan is known as a superior bioadsorbent and widely used in many areas.¹¹ Chitosan is one of the high-performance natural polysaccharide materials and derives from the deacetylation of chitin. Owing to the presence of large amounts of amino (–NH₂) and hydroxyl (–OH) groups, chitosan exhibits a high adsorption capacity and fast adsorption rate on a substantial number of materials, such as dyes, heavy metal ions, phenolic compounds, and so on.^12–14 However, its practical application is hampered by two predominant defects existing in the chitosan. On the one hand, the fact that chitosan dissolves in an acid environment can severely impose a great limitation on the chitosan as an effective adsorbent.^15,16 On the other hand, the powdery chitosan is difficult to separate and recover except by high-speed centrifugation and filtration, which is too time-consuming and tedious to meet the high throughput and fast enrichment and purification needed in biological samples.¹⁷

In order to overcome the above barriers, magnetic carrier technology (MCT) has gradually attracted the attention of many scientists and technicians as a rapid and effective technology for magnetic separation.¹⁸ A distinct advantage of MCT is the utilization of magnetic carrier materials. Among the various magnetic carriers, magnetic nanoparticles (MNPs), such as Fe₃O₄ NPs, are promising candidates for carrier technology.^19–21 Hence, special attention has been directed to combine magnetic nanoparticles (MNPs) with chitosan to make most use of their relative virtues and have the potential to solve the existing problems.^22,23 With the advent of Fe₃O₄ nanoparticles, the chitosan is endowed with a magnetic property, which makes sampling and collection easier and faster. However, much time-consuming and arduous work is still dedicated to synthesize Fe₃O₄@chitosan composites compared to their pretreatment processes. Moreover, the adsorption efficiency and capacities of Fe₃O₄@chitosan composites become lower because the synthesis procedure is involved in the reaction of amino and hydroxyl groups, which are expected to play a great part in the adsorption process.²⁴ In this sense, enforcements are still necessary for developing a new extraction technique offering a simple but effective approach to achieve the extraction. In the present work, a new two-step extraction technique based on the magnetic retrieval of chitosan was developed as a rapid and efficient sample preparation method, which could not only avoid the tedious procedures involved in the complex chemical modification, but also realize the retrieval and separation of chitosan from the dispersion rapidly and effectively. The chitosan substituted for the traditional SPE adsorbent was utilized to extract the analytes in the first process. Then the MNPs were used to retrieve the chitosan enriched with the analytes in the second step.²⁵

To the best of our knowledge, some researchers have studied the two-step extraction mode in many matrices.^26–29 Zhang et al. have applied the magnetic retrieval of chitosan solid extraction in green tea beverage samples.²⁵ Li et al. used the magnetic retrieval of ionic liquid in environmental water samples.²⁸ This is the first report to study the mode based on the combination of Fe₃O₄ nanoparticles synthesized by a solvothermal method, and chitosan in complex biological sample preconcentrations of organic compounds. Besides, it is worth noting that chitosan with different degrees of deacetylation and average molecular weight and Fe₃O₄ nanoparticles synthesized by the different methods were investigated systematically. Especially, a comparative study among different bio-matrix samples and analytes was also examined in detail.

2. Experimental

2.1. Chemicals and materials

All reagents were of analytical reagent grade and used as supplied. Luteolin, quercetin and kaempferol standards were purchased from Sinopharm Chemical Reagent Co., Ltd., China. Their structures are shown in Fig. 1. Chitosan with 100 mesh, 95% degree of deacetylation and average-molecular weight of 1.0 × 10⁶ was supplied by Qingdao Baicheng Biochemical Corp. Other reagents include ferric chloride hexahydrate (FeCl₃·4H₂O), ferrous sulfate heptahydrate (FeSO₄·7H₂O), ethylene glycol (EG), diethylene glycol (DEG), sodium acrylate (CH₂=CHCOONa, Na acrylate), and sodium acetate (CH₃COONa, NaAc). The water used in all experiments was prepared using a compact ultrapure water system from Ulupure Corporation (Chengdu, China).

Fig. 1 Structures of quercetin, luteolin and kaempferol.

2.2. Instruments

The size and distribution of the as-synthesized nanoparticles were studied using a FEI TecnaiG2 F20 transmission electron microscope (TEM). The surface groups on the nanoparticles were measured with a 8400s FT-IR spectrometer (Shimadzu Corporation, Japan). Phase identification was done by X-ray powder diffraction pattern (XRD), using an X' TRA X-ray diffractometer with Cu Ka irradiation at c = 0.1541 nm. The magnetic properties were studied using a LDJ 19600-1 vibrating sample magnetometer (VSM) operating at room temperature with applied fields up to 10 kOe. The specific surface area of the as-synthesized nanoparticles in the dry state was determined by a multipoint Brunauer–Emmett–Teller (BET) apparatus (3H-2000PS2, Beishide Instrument, China). Zeta-potential measurements of two kinds of Fe₃O₄ NPs and the chitosan were performed with a Zeta Plus Zeta Potential Analyzer (Brookhaven, USA).

2.3. Preparation of magnetic Fe₃O₄ NPs

Fig. 2 illustrates the complete procedure of the two-step magnetic retrieval of chitosan and its application as an MSPE adsorbent for the simultaneous extraction and preconcentration of targeted analytes in urine and serum samples. The Fe₃O₄ NPs prepared by the solvothermal and chemical co-precipitation methods were used and compared in the MSPE process. Firstly, the following is the solvothermal method for preparing Fe₃O₄ NPs.³⁰ Briefly, 2.4 g FeCl₃·6H₂O, 3.4 g CH₂=CHCOONa and 3.4 g NaAc were added into a mixture of ethylene glycol (EG, 22.5 mL) and diethylene glycol (DEG, 22.5 mL) under ultrasonication for about 1 h. The resulting homogeneous black solution was then transferred and sealed into a Teflon-lined stainless-steel autoclave. The autoclave was heated at 200 °C for 10 h, and then cooled to room temperature. After the reaction, the obtained Fe₃O₄ NPs were washed with ethanol and water several times, and then dried in vacuum at 65 °C for 10 h. The chemical co-precipitation method is as follows.³¹ Briefly, 100 mL of 0.02 mol Fe²⁺ and 0.04 mol Fe³⁺ solutions were prepared with deionized water in two beakers, and then transferred together to a 500 mL three-necked flask. When the solution was heated to 80 °C, ammonia solution (100 mL) was added dropwise under nitrogen gas protection with vigorous mechanical stirring until the pH was between 10 and 11. After the addition of ammonia, the solution immediately turned black indicating the formation of iron oxide (Fe₃O₄) in the system. The solution continued to be heated at 80 °C for 2 h, and then the precipitated powders were collected by magnetic separation. The obtained magnetic nanoparticles were washed immediately with deionized water several times. The final product was dried to a powder at 40 °C in a vacuum oven.

Fig. 2 Scheme of the two-step magnetic retrieval of chitosan and its application as MSPE adsorbents for the simultaneous extraction and preconcentration of targeted analytes in urine and serum samples.

2.4. Urine and serum sampling

Standard stock solutions of luteolin (1 mg mL⁻¹), quercetin (1 mg mL⁻¹) and kaempferol (1 mg mL⁻¹) were prepared in methanol and then diluted to the desired concentration. Blank urine and serum samples were collected from volunteers in the China Pharmaceutical University (Nanjing, China). Appropriate stock solutions of luteolin, quercetin and kaempferol were spiked to the blank urine and serum solutions. The concentrations of urine samples were prepared with 3.0–2000 ng mL⁻¹ of quercetin, 2.0–2000 ng mL⁻¹ of luteolin and 2.5–2000 ng mL⁻¹ of kaempferol, respectively, while the serum solutions were prepared with 20.0–2000 ng mL⁻¹ of quercetin, 10.0–2000 ng mL⁻¹ of luteolin and 15.0–2000 ng mL⁻¹ of kaempferol, respectively. All the solutions were stored at 4 °C.

2.5. SPE based on magnetic retrieval of chitosan

Firstly, 2 mL urine samples were added into a 5 mL vessel, the pH of which was adjusted to 5 using 0.01 mol L⁻¹ phosphoric acid. Secondly, 10 mg chitosan were added into the above solution, and then shaken gently for 30 s. For serum samples, the pH was adjusted to 7 and then 2 mL serum samples were added into another vessel. Secondly, 15 mg chitosan were added into the above solution, and then shaken gently for 30 s, respectively. The obtained mixtures were kept still for 10 min to completely trap the target analytes. Secondly, 2 mg and 3 mg Fe₃O₄ NPs prepared by the solvothermal method were put into the urine and serum vessels, respectively, and then under ultrasonication for 1 min, and therefore Fe₃O₄ NP retrieval chitosan adsorbents were isolated from the solution by placing a strong magnet at the side of the vessel. The suspension became limpid after 1 min, the supernatant was decanted, and the collected Fe₃O₄ retrieval chitosan adsorbents were eluted with 1 mL acetonitrile containing 5% HAc under ultrasonication for 2 min (0.5 mL each time and eluted twice). Finally, the eluted solution was collected and then dried under a stream of nitrogen at 60 °C and redissolved in 500 μL methanol. After filtration through a 0.45 μL membrane, 10 μL of the solution was injected into the HPLC system for analysis.

2.6. HPLC analysis of urine and serum sampling

The three flavonoids (luteolin, quercetin and kaempferol) were separated and quantified by using high performance liquid chromatography with an automatic sampler (Agilent). The analytical column was a ZORBAX Eclipse XDB-C18 column (4.6 × 150 mm, 5 μm) supplied by Agilent. The mobile phase consisted of methanol–0.2% aqueous phosphoric acid solution (48 : 52, v/v) and the flow-rate was set at 1 mL min⁻¹. The column temperature was 30 °C. The injection volume was 10 μL, and the effluent was analyzed by HPLC equipped with a UV detector at 360 nm (ref. 32).

3. Results and discussion

3.1. Characterization

TEM images of the Fe₃O₄ NPs prepared by the solvothermal and chemical co-precipitation methods are shown in Fig. 3A and B, respectively. The result shows that the mean diameters of nanoparticles prepared by the solvothermal method were mainly distributed in the range 100–200 nm, which is smaller than that of the Fe₃O₄ nanoparticles (200–300 nm) prepared by the chemical co-precipitation method. The specific surface area of the Fe₃O₄ NPs prepared by the solvothermal method was found to be 11 m² g⁻¹, which was apparently higher than that of the Fe₃O₄ NPs prepared by the chemical co-precipitation method (9 m² g⁻¹).

Fig. 3 TEM images of Fe₃O₄ NPs prepared by the solvothermal method (A) and Fe₃O₄ NPs prepared by the chemical co-precipitation method (B).

The XRD patterns of pure Fe₃O₄ NPs prepared by the solvothermal method and the Fe₃O₄ retrieval chitosan composites are presented in Fig. 4A. In the 2θ region of 20–70°, six characteristic peaks marked by their indices (220), (311), (400), (422), (511), and (440) (JCPDS card 19-0629 for Fe₃O₄) were observed for pure Fe₃O₄ NPs and the Fe₃O₄ retrieval chitosan composites. Because of the encapsulation by chitosan on the magnetic nanoparticles, the intensity of the peaks decreased slightly from Fe₃O₄ NPs to the Fe₃O₄ retrieval chitosan composites.

Fig. 4 (A) XRD patterns for Fe₃O₄ NPs prepared by the solvothermal method (a) and Fe₃O₄ NP retrieval chitosan composites (b). (B) FT-IR spectra of Fe₃O₄ NPs prepared by the solvothermal method (a), chitosan (b) and Fe₃O₄ NP retrieval chitosan composite (c). (C) VSM magnetization curves of Fe₃O₄ NPs prepared by the solvothermal method and chemical co-precipitation method. (D) zeta-potential at different pHs of chitosan, Fe₃O₄ NPs prepared by the solvothermal method and Fe₃O₄ NPs prepared by the chemical co-precipitation method.

Fe₃O₄ NPs prepared by the solvothermal method, chitosan and Fe₃O₄ retrieval chitosan composites were characterized by FT-IR.^33,34 As observed in Fig. 4B, all displayed many common characteristics in their spectra. As shown in Fig. 4B-a, the Fe–O characteristic band at 583 cm⁻¹ is indicative of Fe₃O₄ NPs. At the same time, the band appearing at 2919 cm⁻¹ is the stretching of C–H from the methyl group (–CH₂, –CH₃), and the peaks at 1076 cm⁻¹ (C–O) and 1645 cm⁻¹ (N–H) are also characteristic of chitosan, and which are demonstrated in Fig. 4B-b. Moreover, as displayed in Fig. 4B-c, the stretching vibration of the C–H band at 2918, 2852 cm⁻¹, the band at 3430 cm⁻¹ corresponding to the hydroxyl (–OH) groups and the vibration of N–H in imidazole ring at 1470 cm⁻¹ appeared in the spectrum of the Fe₃O₄ NP retrieval chitosan sorbents, which indicated that chitosan was successfully adsorbed on the Fe₃O₄ NPs.

To enable the practical application of MCT, it is critical that the adsorbents should possess sufficient magnetic properties in MSPE application. Fig. 4C shows VSM magnetization curves of Fe₃O₄ NPs prepared by the solvothermal and chemical co-precipitation methods at room temperature. It was found that the maximum saturation magnetization was 64.60 emu g⁻¹ for Fe₃O₄ NPs prepared by the solvothermal method and 51.34 emu g⁻¹ for Fe₃O₄ NPs prepared by the chemical co-precipitation method. Although the two prepared kinds of Fe₃O₄ NPs could be separated from their dispersion quickly once an external magnetic field was applied, more time was needed for the Fe₃O₄ NPs prepared by the chemical co-precipitation method.³⁵

The isoelectric point (IEP) is known to be an important characteristic. The charge density is a predominant factor influencing the interaction between Fe₃O₄ NPs and chitosan. The isoelectric points (IEPs) of chitosan, Fe₃O₄ NPs prepared by the solvothermal method and Fe₃O₄ NPs prepared by the chemical co-precipitation method were measured under different pH (shown in Fig. 4D). The IEPs of Fe₃O₄ NPs prepared by the solvothermal and chemical co-precipitation methods were found to be at pH 3.47 and 3.56, respectively. For chitosan, the IEP was found to be at pH 8.16, which approached data previously reported for them.^36,37

3.2. Optimization of the analysis conditions

In order to obtain a high recovery for three flavonoids while eliminating most of the interference originating from the urine and serum samples, the significant factors affecting the extraction recoveries of the MSPE-HPLC-UV method, including the amount of chitosan, the amount of Fe₃O₄ NPs, pH, and the adsorption time, ionic strength and desorption conditions, were studied. It is worth noting that this work represents the first attempt to systematically investigate the types of chitosan and Fe₃O₄ NPs in this two-step MSPE procedure.

3.2.1. Effect of the chitosan type. As a bioadsorbent, chitosan exhibits many bio-properties, among which the degree of deacetylation and average molecular weight are two predominant characteristics that play a great part in the adsorption efficiency.³⁸ In order to investigate different kinds of chitosan, three types of chitosan possessing different degrees of deacetylation and average molecular weight were evaluated in the urine and serum samples. The results are shown in Fig. 5A and B, which imply that with the same average molecular weight, the chitosan with 95% degree of deacetylation showed a stronger adsorption efficiency than that of chitosan with 85% degree of deacetylation. This may be attributed to the fact that with an increasing degree of deacetylation, more free amino groups exist in the chitosan, which could react with more analytes through hydrogen bonding. On the other hand, based on the same degree of deacetylation, the adsorption efficiency of the chitosan with an average molecular weight of 1.0 × 10⁶ is higher than that of chitosan with a 4.0 × 10⁶ average molecular weight, which can be explained by the fact that the chitosan with a higher average molecular weight dissolves less in the solution due to the higher degree of twisting among the chitosan molecules. Based on the above results which imply that the choice of chitosan was very important, chitosan with a 95% degree of deacetylation and average molecular weight of 1.0 × 10⁶ was selected in the following experiments.

Fig. 5 Effect of the types of chitosan on the adsorption of luteolin, quercetin and kaempferol performed in urine samples (A) and serum samples (B). Operation in the batch mode. Concentration of each analyte: 60.0 ng mL⁻¹. In urine samples: amount of chitosan: 10 mg; amount of Fe₃O₄ NPs prepared by the solvothermal method: 2 mg; pH: 5.0. In serum samples: amount of chitosan: 15 mg; amount of Fe₃O₄ NPs prepared by the solvothermal method: 3 mg; pH: 7.0.

3.2.2. Effect of the type and amount of Fe₃O₄ NPs. Magnetic nanoparticles have been used as better adsorbents for their high surface areas and strong magnetism. In the novel SPE mode, Fe₃O₄ NPs act as a carrier to separate the chitosan loading three flavonoids based on the electrostatic attraction and complexation. In the experiment, two kinds of Fe₃O₄ NPs synthesized by the solvothermal and chemical co-precipitation methods were investigated and compared in urine and serum samples. To find the optimized amount of Fe₃O₄ NPs for the extraction and make a comparative study between the two kinds of Fe₃O₄ NPs, amounts of two kinds of Fe₃O₄ NPs that ranged from 0 to 12 mg were tested in the urine and serum samples. As shown in Fig. 6, the extraction recoveries increased with increasing Fe₃O₄ NPs; further increasing the amount of the Fe₃O₄ NPs showed no significant improvement for the recoveries of flavonoids. From Fig. 6A and C in the urine sample and Fig. 6B and D in the serum sample, it was inferred that a higher extraction efficiency and greater adsorption capacity could be obtained both in the urine and serum samples by using the Fe₃O₄ NPs prepared by the solvothermal method, which can be attributed to the fact that the Fe₃O₄ NPs prepared by the solvothermal method possess a higher saturation magnetization and larger specific surface area compared to the co-precipitation Fe₃O₄ NPs in the urine and serum samples. According to these results, both two kinds of Fe₃O₄ NPs could be successfully used to retrieve the chitosan and exploited for the SPE mode. The Fe₃O₄ NPs of the solvothermal method were selected because fewer amounts of nanoparticle adsorbents could achieve a higher extraction efficiency, and therefore, 2 mg and 3 mg of Fe₃O₄ NPs prepared by the solvothermal method were selected and adequate for extracting the three analytes from the urine and serum samples, respectively, in the following studies.

Fig. 6 Effect of the type and amount of Fe₃O₄ NPs on the adsorption of luteolin, quercetin and kaempferol in urine samples (A) and (C) and serum samples (B) and (D). Operation in the batch mode. Concentration of each analyte: 60.0 ng mL⁻¹. In urine samples: amount of chitosan: 10 mg; pH: 5.0. In serum samples: amount of chitosan 15 mg; pH: 7.0.

3.2.3. Effect of the amount of chitosan. The percentage of the retained three flavonoids is a key parameter depending on the amount of chitosan added. To achieve good extraction recoveries towards the three flavonoids, the amount of chitosan was investigated from 0 to 25 mg in the urine and serum samples, as shown in Fig. 7A and B. From the results, it can be concluded that the flavonoids were hardly adsorbed onto the surface of Fe₃O₄ NPs in the absence of chitosan, indicating that the Fe₃O₄ NPs almost have no enrichment ability towards the three flavonoids. The adsorption ratio of luteolin, quercetin and kaempferol increased remarkably with the increasing amount of chitosan added into the solution, then kept relatively invariant. Comparing Fig. 7A and B, more chitosan is required for serum samples than for urine samples because more protein or endogenous compound is presented in the former. Given the above findings, 10 mg and 15 mg chitosan were employed as the final addition amount of chitosan in urine and serum samples, respectively, in the following studies.

Fig. 7 (A) and (B) Effect of the amount of chitosan on the adsorption of luteolin, quercetin and kaempferol performed in urine samples (A) and serum samples (B). Operation in the batch mode. Concentration of each analyte: 60.0 ng mL⁻¹. In urine samples: amount of Fe₃O₄ NPs prepared by the solvothermal method: 2 mg; pH: 5.0. In serum samples: amount of Fe₃O₄ NPs prepared by the solvothermal method: 3 mg; pH: 7.0. (C) and (D) Effect of pH on the adsorption of luteolin, quercetin and kaempferol in urine samples (C) and serum samples (D). Operation in the batch mode. Concentration of each analyte: 60.0 ng mL⁻¹. In urine samples; amount of chitosan: 10 mg; amount of Fe₃O₄ NPs prepared by the solvothermal method: 2 mg. In serum samples: amount of chitosan: 15 mg; amount of Fe₃O₄ NPs prepared by the solvothermal method: 3 mg.

3.2.4. Effect of solution pH. The pH is one of the prime factors to influence the extraction efficiency by affecting both the existing forms of analytes and the charge species and density on the adsorbent surface. In the present study, the isoelectric points (IEPs) of Fe₃O₄ NPs and chitosan were around 3.0 and 8.16, respectively, and the effect of pH was investigated by varying the pH values from 3.0 to 10.0 in the urine and serum samples. Fig. 7C and D imply that maximum adsorption performance occurred at pH 5.0 in the urine sample and 7.0 in the serum sample, which are between the isoelectric points (IEPs) of Fe₃O₄ NPs and chitosan. When the pH value was around 3.0, the charge density of Fe₃O₄ NP surface was very low and most of these functional groups present in chitosan are protonated and presented in positively charged form. A small amount of chitosan was absorbed on the Fe₃O₄ NPs and therefore could not be retrieved based on the electrostatic attraction.²⁵ In addition, the presence of a large number of H⁺ and H₃O⁺ ions in the aqueous solution may compete with the three flavonoids for the adsorption sites available on chitosan. Thus low extraction recoveries were observed at low pH. The extraction recoveries increased with the increase of pH from 4 to 8, which could be explained by the fact that with the increase of pH, there are more free –NH₂ and –OH functional groups in chitosan to react with the three flavonoids through hydrogen bonds. Moreover, the charges on chitosan and the Fe₃O₄ NPs changed. Therefore, chitosan could be retrieved enough by the Fe₃O₄ NPs based on the strong electrostatic attraction.

However, when the pH was above 8, the extraction recoveries of the three flavonoids decreased with increasing of pH, which is due to the fact that chitosan and Fe₃O₄ NPs both became negatively charged, unfavorable to the retrieval process because of the electrostatic repulsion. In addition, the pK_a values of luteolin, quercetin and kaempferol are 7.04, 7.36 and 8.09, respectively.³⁹ Under these conditions, the three flavonoids were ionized and therefore electrostatic repulsion will occur between the three flavonoids and the negatively charged chitosan surface. For further studies, pH 5.0 and 7.0 were selected as the optimal pH in the urine and serum samples, respectively, for the rest of experiments, since sufficient extraction recoveries were achieved of all the three flavonoids at those pH values.

3.2.5. Effect of extraction time. Generally, sufficient time is required to achieve the adsorption equilibrium for the analytes on the adsorbent. In this study, the effect of the extraction time on the extraction efficiency of the three flavonoids was investigated by changing the time from 2 to 30 min under the above optimal condition in the urine and serum samples. The results are shown in Fig. 8A and B. The extraction recoveries for all three flavonoids reached their maxima when the extraction time was increased to 5 min, and prolonged extraction times did not increase the extraction recoveries of the analytes any further, indicating that the extraction equilibrium could be achieved in a very short time. However, a slightly longer time was beneficial to get good reproducibility. Therefore, an extraction time of 10 min was selected, which not only enabled these three flavonoids to be completely absorbed on the chitosan, but also ensured good reproducibility.

Fig. 8 (A) and (B) Effect of extraction time on the adsorption of luteolin, quercetin and kaempferol in urine samples (A) and serum samples (B). (C) and (D) Effect of ionic strength on the adsorption of luteolin, quercetin and kaempferol in urine samples (C) and serum samples (D).

3.2.6. Effect of ionic strength. Ionic strength, which is examined through the addition of salt, may enhance the partition of analytes, thereby influencing the enrichment performance.¹⁶ To investigate the effect of salt on the proposed method, various concentrations of sodium chloride from 0 to 12% were examined in the urine and serum samples. As shown in Fig. 8C and D, the result indicated that no significant effect on the extraction recovery was observed for luteolin and kaempferol in the NaCl concentration range investigated, while, for quercetin, a slightly decreased extraction recovery was observed with increased concentrations of NaCl from start to finish. Therefore, no addition of NaCl into the sample solution was adopted for all the subsequent experiments in urine and serum samples.

3.2.7. Desorption conditions. A suitable desorption solvent plays an important role in MSPE. In the study, acetonitrile and methanol were studied for the desorption of analytes from the chitosan adsorbents. In the first place, it was found that the analytes could not be desorbed from the adsorbents completely, even when desorption time was prolonged to 4 h, but the desorption ability of acetonitrile was superior to that of methanol. This phenomenon may be attributed to the fact that the multi-interactions, such as hydrogen bonds between the adsorbent and analytes, could not be disrupted completely by the above solvent. Therefore, acetonitrile with different contents of acetic acid (1–10%, v/v) was used as the desorption solvent to desorb the three flavonoids. The results indicated that with an increasing amount of acetic acid, the recoveries of luteolin, quercetin and kaempferol increased remarkably, but then the adsorption of analytes decreased gradually. Because the addition of acetic acid not only caused the analytes to exist in the molecular form, which was prone to be soluble in organic solvent, but also made the chitosan dissolve faster, while excess acetic acid could disrupt the desorption ability of acetonitrile.⁴⁰ At last, acetonitrile containing 5% HAc was selected as the final desorption solvent and quantitative recoveries of the flavonoids were achieved with 1 mL acetonitrile containing 5% HAc (0.5 mL every time and washed two times).

3.3. Method validation

Based on the above results, the conditions of the MSPE were as follows: 10 mg chitosan was added into urine samples, and next 2.0 mg Fe₃O₄ NPs were added to the solution to retrieve the chitosan, 15 mg chitosan was added into serum samples, and 3.0 mg Fe₃O₄ NPs were added to the solution to retrieve the chitosan. The extraction time was 10 min, and acetonitrile containing 5% HAc was used for the desorption. To evaluate the accuracy and feasibility of the method developed, Fig. 9 show the chromatograms of blank and spiked urine and serum samples, respectively. The samples were treated with the same proposed SPE before injection to HPLC. The retention times of quercetin, luteolin and kaempferol were 9.57 min, 11.94 min and 17.69 min, respectively. No matrix effects, such as the proteins and endogenous components in urine and serum samples, were observed, which implies excellent specificity for the determination of quercetin, luteolin and kaempferol with this novel two-step MSPE technique.

Fig. 9 HPLC-UV chromatograms of samples with the two-step MSPE procedure: (A) blank urine sample; (B) urine sample spiked with 5 ng mL⁻¹ standard solution of three flavonoids; (C) blank serum sample; (D) serum sample spiked with 20 ng mL⁻¹ standard solution of three flavonoids (1. quercetin; 2. luteolin; 3. kaempferol).

3.3.1. Linearity, LOD and LOQ. Under the optimized conditions, the quantitative parameters of the proposed method including linear range, correlation coefficients, precision, limits of detection (LOD), limits of quantification (LOQ) and recovery were evaluated in urine and serum samples. As shown in Table 1, each analyte exhibited good linearity with correlation coefficient R² > 0.9990 in the studied range. The LODs of the investigated compounds in urine samples should be the minimum concentrations determined based on a signal-to-noise ratio of 3 (S/N = 3). They were estimated to be 0.50 ng mL⁻¹, 0.10 ng mL⁻¹ and 0.20 ng mL⁻¹ for quercetin, luteolin and kaempferol, respectively, in urine samples and 10 ng mL⁻¹, 2.0 ng mL⁻¹ and 5.0 ng mL⁻¹ for quercetin, luteolin and kaempferol, respectively, in serum samples. The LOQ values based on a signal-to-noise ratio of 10 (S/N = 10) were 0.80 ng mL⁻¹, 0.25 ng mL⁻¹ and 0.50 ng mL⁻¹ for quercetin, luteolin and kaempferol, respectively, in urine samples and 15 ng mL⁻¹, 5.0 ng mL⁻¹ and 8.0 ng mL⁻¹ for quercetin, luteolin and kaempferol, respectively, in serum samples.

Table 1 Analytical parameters of the proposed method

Biological samples	Compound (n = 6)	Linear range (ng mL^−1)	Linearity (R^2)	LOD (ng mL^−1)	LOQ (ng mL^−1)
Urine	Quercetin	3.0–2000	0.9995	1.0	2.0
	Luteolin	2.0–2000	0.9994	0.50	1.0
	Kaempferol	2.5–2000	0.9997	0.70	1.3
Serum	Quercetin	20–2000	0.9994	10	15
	Luteolin	10–2000	0.9991	2.0	5.0
	Kaempferol	15–2000	0.9992	5.0	8.0

---

3.3.2. Precision and recovery. The overall performance of the method in terms of the intra-day and inter-day precision and accuracy was evaluated by applying the proposed MSPE to six replicate spiked urine and serum samples at three different concentration levels (low, middle and high quantification concentrations) of the targets in the same day and over three consecutive days. The relative standard deviations (RSDs) of the intra-day and inter-day precision and accuracy values for spiked urine and serum samples are summarized in Table 2. The RSDs of the intra-daily tests are less than 5.0%, and the RSDs of inter-daily tests are less than 5.4%. These results indicate that the two-step MSPE method has good accuracy and precision.

Table 2 Precision and accuracy for the detection of three flavonoids in urine and serum samples

Biological samples	Compound (n = 6)	Conc. (ng mL^−1)	Inter-day (n = 6)		Intra-day (n = 6)
			Mean accuracy (%)	RSD (%)	Mean accuracy (%)	RSD (%)
Urine	Quercetin	10	94.3	4.6	93.5	3.3
		100	93.6	3.2	93.2	4.2
		1500	93.7	3.5	91.5	3.5
	Luteolin	5	97.1	4.5	95.6	3.4
		100	91.1	2.8	97.8	4.8
		1500	95.2	3.1	98.5	4.1
	Kaempferol	5	98.4	3.6	96.3	4.9
		100	93.5	4.2	97.5	3.5
		1500	98.1	4.5	95.4	4.6
Serum	Quercetin	20	94.5	5.0	93.6	5.4
		200	93.4	3.6	94.5	3.6
		1500	90.1	2.9	97.6	3.0
	Luteolin	10	95.6	4.5	90.1	4.5
		200	96.4	3.8	91.4	3.5
		1500	95.8	2.4	96.5	3.1
	Kaempferol	10	97.4	4.1	90.9	3.8
		200	96.4	4.0	96.8	4.0
		1500	93.4	3.1	97.4	3.1

---

The recovery of the method was assessed using the standard addition method. Table 3 shows the recoveries calculated after spiking three different concentration levels (low, middle and high quantifications; mean value of three independent determinations). It can be seen that a satisfactory recovery range from 93.5% to 108.8% and the RSDs of recoveries range from 3.0% to 5.0% were obtained using the proposed method. Thus the method is suitable for analyzing the three flavonoids in biological samples.

Table 3 Recovery for three flavonoids from blank urine and serum samples

Biological samples	Compound (n = 6)	Conc. (ng mL^−1)	Mean recovery (%)	RSD (%)
Urine	Quercetin	10	99.0	4.8
		100	93.5	3.0
		1500	106.5	2.5
	Luteolin	5	99.8	4.9
		100	96.6	2.8
		1500	105.5	3.4
	Kaempferol	5	108.8	5.0
		100	98.1	4.1
		1500	100.1	3.6
Serum	Quercetin	20	98.7	4.5
		200	105.2	3.6
		1500	100.4	3.7
	Luteolin	10	104.2	5.1
		200	99.7	3.4
		1500	98.4	4.0
	Kaempferol	10	95.8	4.0
		200	97.4	3.0
		1500	102.1	4.5

---

3.3.3. Comparison of the present two-step MSPE with other analytical methodologies. The present work established an easy-to-handle, cheap and accurate two-step extraction mode for the determination of quercetin, luteolin and kaempferol in human urine and serum samples. Table 4 compares some analytical performances of previous reported methods for the determination of the flavonoids. As can be seen, the present method exhibited some remarkable strong points from the viewpoint of LOD, linearity, extraction time, recovery, accuracy, and especially the adsorbent used.

Table 4 Comparison of different methods applied to extract the flavonoids

Method	Sample matrix	Recovery (%)	Linear range (ng mL^−1)	LOD (ng mL^−1)	LOQ (ng mL^−1)	RSD (%)	Extraction time (min)	Ref.
a HF-LPME-HPLC-UV: hollow fiber liquid phase microextraction-high-performance liquid chromatography-ultraviolet detector.b UAE-HPLC-UV: ultrasonic-assisted extraction-high-performance liquid chromatography-ultraviolet detector.c SPE-HPLC-UV: solid phase extraction-high-performance liquid chromatography-ultraviolet detector.d LLE-HPLC-QQQ MS: liquid-liquid extraction-high-performance liquid chromatography-triple quadrupole mass spectrometry.e SPE-CZE: solid phase extraction-capillary zone electrophoresis.f SPE-UHPLC-UV: solid phase extraction-ultra high-performance liquid chromatography-ultraviolet detector.g MMHSPE-HPLC-UV: magnetic mixed hemimicelles solid-phase extraction method-high-performance liquid chromatography-ultraviolet detector.
HF-LPME-HPLC-UV^a	Echinophora platyloba DC	92.0–99.0	3.0–500	0.5–7.0	—	3.18–11.00	—	43
UAE-HPLC-UV^b	Dried celery	72.7–89.5	1 × 10^2–2.5 × 10^5	70	—	2.5–4.5	90	44
SPE-HPLC-UV^c	Plasma and urine	—	4.0–1 × 10^3	0.35–7	35	1.5–9.4	∼45	5
LLE-HPLC-QQ-MS^d	Plant	93.03–98.06	5.0–2 × 10^3	1.0	5	1.3–3.0	>60	9
SPE-CZE^e	Flos Lonicer	93–104	8 × 10^3–1.59 × 10^5	60	1800	2.57–4.36	>60	42
SPE-UHPLC-UV^f	Urine	70.35–96.58	0.05–5.0	15.4	46.2	3.9–5.0	>60	41
MMHSPE-HPLC-UV^g	Urine	90.1–97.6	0.5–1500	0.1–0.5	0.25–0.8	3.3–5.0	90	39
MSPE-HPLC-UV	Urine and serum	91.2–99.7	2.5–2000	0.5–10	1–15	2.5–5.0	10	This method

---

Firstly, compared to traditional SPE methods,^41,42 our present method not only has lower LODs and wider linear ranges but also possesses some attractive merits. An important aspect should be pointed out that the chitosan is low-cost, accessible and environmentally friendly, which turns the method into a potential technique for the routine analysis of actual samples. The Fe₃O₄ NPs used in the study possess strong magnetization characteristics and a large specific surface area, which enable the Fe₃O₄ NPs to be favorable enough for MCT to retrieve the chitosan containing the targeted analytes. With the aid of Fe₃O₄ NPs, the present method greatly simplifies sample pretreatment and eliminates the time-consuming column passing and filtration operation and shows great analytical potential in pretreatment.

Secondly, it is known to us that the magnetic mixed hemimicelle solid-phase extraction method is widely used in sample pretreatment.^36,39,40 However, the present method still has some remarkable superiority over it. In the first place, as shown in Table 4, the extraction time of the present method is shorter than that of the magnetic mixed hemimicelle solid-phase extraction method, because the adsorbent chitosan exhibits several desirable properties such as dispersing ability, high adsorption capacity and fast adsorption rate, which render the method capable of a high extraction efficiency and short equilibrium time. Besides, the strong hydrogen bonds between the chitosan and the targeted analyte have afforded a high affinity and selectivity toward the three flavonoids, superior to other methods. Accordingly, the successful application in the enrichment and determination of the three trace flavonoids from serum and urine samples suggests that the two-step MSPE method based on magnetic carrier technology (MCT) and chitosan could be a promising alternative for the fast and selective extraction of trace amounts of natural therapeutic substances from biological fluids.

4. Conclusions

In conclusion, a rapid, sensitive and simple two-step SPE method based on the magnetic retrieval of chitosan was, for the first time, applied in biological samples for the extraction and preconcentration of the active compounds. Three types of chitosan with different degrees of deacetylation and average-molecular weight were investigated. The experimental results demonstrated that chitosan with a 95% degree of deacetylation and average-molecular weight of 1 × 10⁶ was more appropriate for the SPE of luteolin, quercetin and kaempferol. Two kinds of magnetic nanoparticles synthesized by solvothermal and chemical co-precipitation methods were compared and the solvothermal method was selected because a higher extraction efficiency and greater adsorption capacity were achieved, which were attributed to the higher saturation magnetization and larger specific surface area of Fe₃O₄ NPs prepared by this method.

It is widely known to us that the efficient preconcentration of trace compounds in biological samples remains a challenge. Although some researchers have conducted a preliminary exploration of the extraction mode based on the magnetic retrieval of chitosan, this is the first attempt to apply it in complex biological samples and study it more comprehensively and systematically. Moreover, good linearities and recoveries with serum and urine samples were obtained, which indicates that this proposed method can be successfully applied in the sample preparation of biological samples and will hopefully have a high analytical potential for the preconcentration of trace analytes from complex samples.

Acknowledgements

This work was supported by Graduate Students Innovative Projects of Jiangsu Province, National Training Programs of Innovation and Entrepreneurship for Undergraduates (J1030830) and Excellent Youth Foundation of Yancheng Health Vocational and Technical College (no. 20114105). We are delighted to acknowledge discussions with colleagues in our research group.

References

1. S. P. Wang and K. J. Huang, J. Chromatogr. A, 2004, 1032, 273–279.
2. Y.-q. Xia, T.-y. Guo, H.-l. Zhao, M.-d. Song, B.-h. Zhang and B.-l. Zhang, J. Sep. Sci., 2007, 30, 1300–1306.
3. Y. Nolvachai and P. J. Marriott, J. Sep. Sci., 2013, 36, 20–36.
4. S. Magiera, C. Uhlschmied, M. Rainer, C. W. Huck, I. Baranowska and G. K. Bonn, J. Pharm. Biomed. Anal., 2011, 56, 93–102.
5. K. Ishii, T. Furuta and Y. Kasuya, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2003, 794, 49–56.
6. M. N. Irakli, V. F. Samanidou, C. G. Biliaderis and I. N. Papadoyannis, J. Sep. Sci., 2012, 35, 1603–1611.
7. P. C. H. Hollman, J. M. P. van Trijp, M. N. C. P. Buysman, M. S. v. d. Gaag, M. J. B. Mengelers, J. H. M. de Vries and M. B. Katan, FEBS Lett., 1997, 418, 152–156.
8. A. Bolarinwa and J. Linseisen, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2005, 823, 143–151.
9. K. Zhang, J. Zhang, S. Wei, W. Jing, Y. Wang and A. Liu, J. Sep. Sci., 2013, 36, 2407–2413.
10. H. Alvarez-Ospina, I. Rivero Cruz, G. Duarte, R. Bye and R. Mata, Phytochem. Anal., 2013, 24, 248–254.
11. Z. Yu, X. Zhang and Y. Huang, Ind. Eng. Chem. Res., 2013, 52, 11956–11966.
12. D. Hritcu, M. I. Popa, N. Popa, V. Badescu and V. Balan, Turk. J. Chem., 2009, 33, 785–796.
13. Y. H. Yu, B. Xue, Y. Sun and B. L. He, Acta Polym. Sin., 2000, 340–344.
14. H. Yan, L. Yang, Z. Yang, H. Yang, A. Li and R. Cheng, J. Hazard. Mater., 2012, 229–230, 371–380.
15. H.-J. Chen, Z.-H. Zhang, L.-J. Luo and S.-Z. Yao, Sens. Actuators, B, 2012, 163, 76–83.
16. T. S. Anirudhan, D. Dilu and S. Sandeep, J. Magn. Magn. Mater., 2013, 343, 149–156.
17. Y. Liu, S. Jia, Q. Wu, J. Ran, W. Zhang and S. Wu, Catal. Commun., 2011, 12, 717–720.
18. G. Giakisikli and A. N. Anthemidis, Anal. Chim. Acta, 2013, 789, 1–16.
19. L. Ye, Q. Wang, J. P. Xu, Z. G. Shi and L. Xu, J. Chromatogr. A, 2012, 1244, 46–54.
20. S. R. Yazdinezhad, A. Ballesteros-Gomez, L. Lunar and S. Rubio, Anal. Chim. Acta, 2013, 778, 31–37.
21. A. A. Rajabi, Y. Yamini, M. Faraji and S. Seidi, Med. Chem. Res., 2013, 22, 1570–1577.
22. L.-M. Zhou, C. Shang and Z.-R. Liu, Acta Phys.-Chim. Sin., 2011, 27, 677–682.
23. Y. C. Chang and D. G. H. Chen, Macromol. Biosci., 2005, 5, 254–261.
24. H. Y. Zhu, R. Jiang, L. Xiao and G. M. Zeng, Bioresour. Technol., 2010, 101, 5063–5069.
25. H. F. Zhang and Y. P. Shi, Analyst, 2012, 137, 910–916.
26. N. Y. Wang, R. Shen, Z. H. Yan, H. Feng, Q. Y. Cai and S. Z. Yao, Anal. Methods, 2013, 5, 3999–4004.
27. Y. B. Luo, Z. G. Shi, Q. A. Gao and Y. Q. Feng, J. Chromatogr. A, 2011, 1218, 1353–1358.
28. J. H. Zhang, M. Li, M. Y. Yang, B. Peng, Y. B. Li, W. F. Zhou, H. X. Gao and R. H. Lu, J. Chromatogr. A, 2012, 1254, 23–29.
29. M. Li, J. H. Zhang, Y. B. Li, B. Peng, W. F. Zhou and H. X. Gao, Talanta, 2013, 107, 81–87.
30. D. Xiao, P. Dramou, N. Xiong, H. He, H. Li, D. Yuan and H. Dai, J. Chromatogr. A, 2013, 1274, 44–53.
31. P. Dramou, P. Zuo, H. He, L. A. Pham-Huy, W. Zou, D. Xiao and C. Pham-Huy, J. Chromatogr. A, 2013, 1317, 110–120.
32. D. Xiao, D. Yuan, H. He, C. Pham-Huy, H. Dai, C. Wang and C. Zhang, Carbon, 2014, 72, 274–286.
33. Y. Wu, Y. Wang, G. Luo and Y. Dai, Bioresour. Technol., 2009, 100, 3459–3464.
34. Z. Wang, T. Yue, Y. Yuan, R. Cai, C. Niu and C. Guo, Int. J. Biol. Macromol., 2013, 58, 57–65.
35. H. Zhu, M. Zhang, Y. Liu, L. Zhang and R. Han, Desalin. Water Treat., 2012, 37, 46–54.
36. L. Zhu, D. Pan, L. Ding, F. Tang, Q. Zhang, Q. Liu and S. Yao, Talanta, 2010, 80, 1873–1880.
37. C.-H. Zhu, H.-B. Shen, R.-Y. Xu, H.-Y. Wang and J.-M. Han, Acta Phys.-Chim. Sin., 2007, 23, 1583–1588.
38. K. Donadel, M. D. V. Felisberto, V. T. Favere, M. Rigoni, N. J. Batistela and M. C. M. Laranjeira, Mater. Sci. Eng., C, 2008, 28, 509–514.
39. H. He, D. Yuan, Z. Gao, D. Xiao, H. He, H. Dai, J. Peng and N. Li, J. Chromatogr. A, 2014, 1324, 78–85.
40. Q. Cheng, F. Qu, N. B. Li and H. Q. Luo, Anal. Chim. Acta, 2012, 715, 113–119.
41. I. Baranowska and S. Magiera, Anal. Bioanal. Chem., 2011, 399, 3211–3219.
42. Z. Ning, H. Zhong, C. Kong and Y. Lu, Asian J. Chem., 2012, 24, 2965–2968.
43. M. Hadjmohammadi, H. Karimiyan and V. Sharifi, Food Chem, 2013, 141, 731–735.
44. D. Han and K. H. Row, J. Sci. Food Agric., 2011, 91, 2888–2892.

---

Footnote

† These authors contributed equally to this work and should be considered joint first authors.

---