Selectively Adsorptive Extraction of Phenylarsonic Acids in Chicken Tissue by Carboxymethyl α-Cyclodextrin Immobilized Fe3O4 Magnetic Nanoparticles Followed Ultra Performance Liquid Chromatography Coupled Tandem Mass Spectrometry Detection

Carboxymethyl α-cyclodextrin immobilized Fe3O4 magnetic nanoparticles (CM-α-CD-Fe3O4) were synthesized for the selectively adsorptive extraction of five phenylarsonic acids including p-amino phenylarsonic acid, p-nitro phenylarsonic acid, p-hydroxy phenylarsonic acid, p-acylamino phenylarsonic acid and p-hydroxy-3-nitro phenylarsonic acid in chicken tissue. Using ultra performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS), a highly sensitive analytical method was proposed for the determination of five phenylarsonic acids. It was shown that CM-α-CD-Fe3O4 could extract the five phenylarsonic acids in complex chicken tissue samples with high extraction efficiency. Under the optimal conditions, a high enrichment factor, ranging from 349 to 606 fold, was obtained. The limits of detection (LODs) (at a signal-to-noise ratio of 3) were in the range of 0.05–0.11 µg/kg for the five phenylarsonic acids. The proposed method was applied for the determination of five target phenylarsonic acids in chicken muscle and liver samples. Recoveries for the spiked samples with 0.2 µg/kg, 2.0 µg/kg and 20 µg/kg of each phenylarsonic acids were in the range of 77.2%–110.2%, with a relative standard deviation (RSD) of less than 12.5%.

Introduction 4-Hydroxy-3-nitrobenzenearsonic acid, also known as roxarsone (ROX), has been used since 1944 as a feed additive in the poultry industry to promote growth and to control coccidiosis, a parasitic disease that infects the intestinal tract of poultry [1]. Besides ROX, some other organic arsenic compounds, including p-amino phenylarsonic acid (p-APAA), p-nitro phenylarsonic acid (p-NPAA), p-hydroxy phenylarsonic acid (p-HPAA) and pacylamino phenylarsonic acid (AAPAA) (Fig. 1) have been successively employed for the same purposes. Their slight structural difference, i. e. different substituent groups on the aromatic ring, results in different growth-promoting and diseasecontrolling effects [2]. Phenylarsonic acids have been approved as feed additives by many countries at levels of 25-50 mg/kg [3]. Recent studies showed that phenylarsonic acids in the environment might be converted into elemental arsenic and other inorganic arsenic compounds, which are known to be strongly carcinogenic [4]. Some countries in EU strictly control the use of phenylarsonic acid additives, while in the U.S., Tyson Foods, the country's largest poultry producer, stopped the use of arsenic compounds in 2004. After the release of 2011 FDA report of elevated inorganic arsenic in the livers of chickens treated with ROX, Pfizer Animal Health, the US manufacturer of ROX, quickly suspended ROX sales [5]. In order to monitor the residues of phenylarsonic acid in animal products, it is of great significance to establish a convenient, sensitive and reliable method to analyze the organic arsenic in samples [6].
Several analytical methods for the determination of phenylarsonic acids in the environment have been reported, including liquid chromatography (LC) coupled to atomic absorption spectroscopy (AAS) [7], atomic emission spectroscopy (AES) [8] or atomic fluorescence spectrometry (AFS) [9] as well as gas chromatography-mass spectrometry (GC-MS) [10], capillary electrophoresis (CE) coupled to ultraviolet and visible light detector [11] or inductively coupled plasma-mass spectrometry (ICP-MS) [12]. LC has been demonstrated to be the most effective method in arsenic separation, and ICP-MS can provide low detection limit. Therefore, LC-ICP-MS is one of the most powerful research means for analysis of organic arsenic in complex samples [2,13]. The LC-MS/MS is a powerful separation and detection platform in multi residues analysis. Pergantis et al [14] have developed a stable method for determination of 5 phenylarsonic acids including ROX, p-APAA, p-NPAA, AAPAA, p-HPAA and other organic arenics using LC-MS/MS in positive ionization mode. The LODs of the developed method achieved sub ng/g level, whereas the analytical time was too long. Furthermore, compared with other arsenic speciation methods, the LC-MS/MS could provide more structural information of the phenylarsonic acids.
After feeding, nearly all phenylarsonic acids are excreted unchanged to the environment through the disposal of poultry litter, so the residues in animal tissues are very low [15,16]. Some studies indicated that the approved conditions of use mandate a 5day withdrawal period from the medicated feed before animals are slaughtered, and limits are in place for total residues of combined arsenic (As) in meat from ROX treated animals [0.5 mg kg 21 As in muscle tissue and eggs, and 2 mg kg 21 As in liver and kidney] [17]. To further lower the detection limits of phenylarsonic acids in complex biological samples, nano-materials have been used to selectively extract and concentrate arsenic compounds. The most commonly used nano-materials for the adsorption of arsenic compounds include goethite [18], titania [19], iron oxide [20,21], Fe 3 O 4 nanoparticles [22,23] or modified Fe 3 O 4 nanoparticles [24]. Among these nano-materials, Fe 3 O 4 nanoparticles are well suited for arsenic analysis due to the following advantages. First, they can be easily isolated from solutions by applying an external magnetic field [25], which ensures simplified sample adsorption and elution processes. Second, Fe 3 O 4 nanoparticles have been demonstrated to have higher affinity toward arsenic element than other nano-materials, resulting in higher arsenic extraction efficiency. Fe 3 O 4 magnetic nanoparticles have been used for removing inorganic arsenic ions in water sample. For example, Fe 3 O 4 nanoparticles dispersed in chelating resins or coated with adequate chelating agents have been used for the removal of a wide range of metal ions from wastewater, overall displaying higher adsorption capacity than traditional materials [26][27][28][29]. More recently, b-CD coated Fe 3 O 4 nanoparticles have been successfully applied for the removal of methylene blue and copper ions [30,31]. It is found that the cavity of cyclodextrin and its surface hydroxyl group can impart better binding capability and chemical stability to the magnetic particle [32].
In this work, a highly sensitive determination method was established to monitor five phenylarsonic acids in chicken tissues. In our first attempt, it was found that the adsorption of phenylarsonic acids by Fe 3 O 4 magnetic nanoparticles was not as good as inorganic arsenic compounds. It was then decided to use CM-a-CD to couple on the surface of Fe 3 O 4 providing CD cavities to fit the benzene rings in the structures of phenylarsonic acid compounds. In addition, hydrogen bonding and electrostatic interactions between hydroxyl/carboxyl groups of modified CMa-CD and amino/nitro of phenyl arsenic acids were also increased. The adsorption properties of modified Fe 3 O 4 magnetic nanoparticles to phenylarsonic acids were studied, and the interactions between nanoparticles and phenylarsonic acids were examined. The synthesized materials were successfully applied in the sample clean-up and pre-concentration of phenylarsonic acids in chicken muscle and liver samples, which were subsequently separated and detected by UPLC-MS/MS.

Materials and Methods
Apparatus ICP-MS (Agilent7500Ce, USA) was used to study the adsorption properties of synthesized material for phenylarsonic acids. The optimum operation parameters of ICP-MS were selected by tuning. The power was 1550W, the flow rates of cooling air, auxiliary air and carrier were 15.0 L/min, 1.0 L/min and 1.12 L/ min, respectively. The sample rate of ascension by using peristaltic pump was set as 1.0 mL/min. The integration time for arsenic was 0.3 s/isotope. The operating parameters of UPLC-MS/MS (Waters Xevo TQ, USA) were as follows: capillary voltage = 2.8 kv, desolvation temperature = 450uC, desolvation gas flow rate = 600 L/Hr. The mobile phase was a mixture of acetonitrile (solvent A) and water containing 0.1% formic acid (solvent B) at a flow rate of 0.3 mL/min. All chromatographic separations were carried out in linear gradient mode as follows. In the first minute, solvent A was maintained at 98% (v/v). Solvent B quickly dropped to 30% from 1 to 3 min, followed by dramatic increase back to 98% from 3 to 5 min. MS parameters of UPLC-MS/MS are showed in Table 1.
The closed microwave digestion system (CEM MARS, American) was used to digest samples for the determination of the total arsenic in the crude samples. The homogenizer (IKA, Germany) was applied to sample pretreatment and rocking hammock bed was from Zhicheng, Shanghai, China.

Standard solutions and reagents
Five phenylarsonic acids (98%) were obtained from the Chinese Academy of Agricultural Sciences. The standard stock solutions were prepared by dissolving each arsenic species in pure water at an arsenic concentration equivalent to that of 1 mg/mL phenylarsonic acids and stored at 4uC in the dark.
Reagents for preparing magnetic nanoparticles: FeCl 2 ?4H 2 O and FeCl 3 ?6H 2 O (analytical reagent grade) were purchased from Tianjin Guangfu Fine Chemical Research Institute. Sodium chloroacetate (98%) was bought from Alfa Aesar. a-cyclodextrin (98%) was purchased from Beijing Dilang Biochemical Technology Co., Ltd., China. Other reagents including methanol, ethanol, acetone, toluene, formic acid and sodium hydroxide were of analytical reagent grade and all bought from Beijing Chemical Plant. The water used throughout the experiment was purified using a Milli-Q water purification system (Millipore, Germany).

Preparation of CM-a-CD stabilized magnetic nanoparticles
Synthesis of CM-a-CD. CM-a-CD was prepared according to the following procedures. a-CD (3.55 mmoL) and NaOH (90.2 mmoL) were first dissolved into 20 ml water. The solution was then heated at 90uC for 5 min, followed by the addition of 74.6 mmoL sodium chloroacetate. The solution was heated for 3 h at 90uC under stirring. Once cooled to room temperature, the pH of the solution was adjusted to 6-7 by hydrochloric acid. The nearly neutral solution obtained was then poured into about 500 mL methanol. CM-a-CD was precipitated out as white solids, which was filtered and washed with methanol for a few times and then dried under vacuum for 3 d at 50uC and 0.085 MPa. The melting point of the CM-a-CD product was about 245.5uC as determined by micro melting point apparatus.
Synthesis of Fe 3 O 4 nanoparticles. Fe 3 O 4 magnetic nanoparticles were prepared according to the conventional coprecipitation method [33]. A mixture of 2.0 g FeCl 2 ?4H 2 O and 5.2 g FeCl 3 ?6H 2 O was added into a 500 mL conical flask containing 200 ml 0.05 M HCl. After dissolution of the solids, 250 ml 0.75 M NaOH solution was poured into the flask under a blanket of N 2 . The mixture was stirred for another 2 h at 80uC. The Fe 3 O 4 nanoparticles were then obtained in the form of black precipitates, which were separated with a magnet and washed subsequently by water (3 times) and ethanol (twice). It should be noted that both the HCl and NaOH solutions were degassed by a sonicator for 20 min before use.

CM-a-CD modified Fe 3 O 4 nanoparticles (CM-a-CD-
The Fe 3 O 4 nanoparticles prepared in the previous step were added into 60 ml PBS buffer solution (pH = 6.6) containing 1.6 g of CM-a-CD. The suspension was sonicated for 3 min and then stirred for 3 h at 80uC. After cooling to room temperature, the nanoparticles were washed several times by PBS buffer solution to remove excess CM-a-CD. The CM-a-CD modified Fe 3 O 4 nanoparticles were then dried at 80uC in a vacuum oven.

Adsorption procedure
Static adsorption. In a 10 mL centrifuge tube, 5 mg CM-a-CD-Fe 3 O 4 nanoparticles were mixed with 8 mL standard solution of phenylarsenic acid with a given concentration. The centrifuge tube was placed on a rocking hammock bed at a rate of 270 rpm. After equilibrating for 30 min, the magnetic nanoparticles were separated from the solution with external magnetic field. The nanoparticles were rinsed twice with ethanol and dried in N 2 . To desorb target compounds, 1.0 mL pure water was added to the nanoparticles followed by equilibration for 10 min. The aqueous solution containing target compounds was filtered through a 0.22 mm Poly (ether sulfones) (PES) syringe filter and analyzed by ICP-MS. Standard phenylarsenic acid solutions of other concentrations were analyzed in the same way.
Dynamic adsorption. In a 10 mL centrifuge tube, 5 mg CM-a-CD-Fe 3 O 4 nanoparticles were mixed with 8 mL standard solution of phenylarsenic acid with a given concentration. A number of centrifuge tubes were prepared in this manner for a given concentration of standard. The centrifuge tubes were placed on an orbital shaker at a rate of 270 r/min. At different time points, one tube was removed and the magnetic nanoparticles contained in the tube were separated from the solution with external magnetic field. The following washing, desorption and ICP-MS analysis procedures were the same as those in the static adsorption step. Standard phenylarsenic acid solutions of other concentrations were analyzed in the same way.

Sample analysis
Chicken tissues including meat and liver were bought from a supermarket in Beijing. Chicken meat and chicken liver samples were pulverized and freeze-dried for 24 h. The freeze-dried sample was homogenized by grinding and frozen until analysis. In a 50 mL centrifuge tube, 5.0 g freeze-dried chicken tissue sample and 10 mL ethanol were added. The extraction was repeated twice, each lasting 30 min. The combined extracts were equilibrated with 5 mg CM-a-CD-Fe 3 O 4 nanoparticles for 10 min. Then the magnetic nanoparticles which adsorbed target analytes were separated under external magnetic field, the analytes adsorbed on the nanoparticles were then desorbed with 2 mL deionized water. The aqueous solution containing target compounds was filtered through a 0.22 mm PES syringe filter and analyzed by UPLC-MS-MS.

Determination of total arsenic
The total amount of arsenic in chicken tissue samples was determined by microwave digestion ICP-MS according to reference [34]. Each digestion can containing 0.5 g chicken tissues samples was added 5 mL 65% nitric acid. Stages digestion method by controlling temperature was used. The obtained digestion solution was diluted until the concentration of nitric acid fell below 5% and then subjected to ICP-MS analysis.

Optimization of adsorption efficiency
For optimal adsorption efficiency, the amounts of CM-a-CD for the modification of

Characterization of magnetic CM-a-CD-Fe 3 O 4
The transmission electron microscope (TEM), fourier transform infrared spectrometry (FTIR) and thermo gravimetric analyzer (TGA) were used to characterized the magnetic nanoparticles. The TEM images of Fe 3 O 4 and CM-a-CD-Fe 3 O 4 nanoparticles are shown in Fig. 4 (a) and Fig. 4 (b), respectively. Unmodified  Fig. 4 (b) that the composite of modified nanoparticles was more compacted and displayed roughly spherical shapes.
The FTIR spectrums of Fe 3 O 4 , CM-a-CD and CM-a-CD-Fe 3 O 4 nanoparticles respectively were scanned (See Fig. S1). In all three samples, a strong characteristic O-H absorption band at

Selection of adsorption solvent
The solvent for phenylarsonic acid solutions is an important factor affecting the adsorption efficiency. Five candidate solvents, i.e., water, methanol, ethanol, acetone and toluene, were compared (Fig. 5). The adsorption efficiency of all target arsenic species could reach 100% in ethanol, acetone and toluene. Ethanol was selected as the solvent in the following experiments  due to the relatively low toxicity. In water and methanol, the adsorption efficiencies of the five target compounds are very poor, which might be because of the water and methanol possess higher polarity than ethanol, acetone and toluene.

Static adsorption
Saturation of adsorption. The adsorption saturation curve is shown in Fig. 6. The point of saturation was reached at 40 mg for p-HPAA, p-APAA, and AAPAA, whereas it was 60 mg for p-NPAA and ROX, corresponding to about 0.2 mmol of each phenylarsonic acid.
The amount of CM-a-CD coated on the surface of Fe 3 O 4 could be estimated from the saturation of adsorption. Since one phenylarsonic acid molecule was supposed to fit one CD cavity, there should be approximately equal amount of CD and phenylarsonic acid. Thus, the amount of CM-a-CD might be calculated through its molecular mass of 1029 according to equation (2). The estimated amount of grafted CM-a-CD was about 41.2 mg/g, which was similar to 43 mg/g calculated by TGA in 3.2.3.  Table 2.

Dynamic adsorption
Optimization of adsorption time and desorption time. The water was selected as desorption solvent. The adsorption efficiency was optimized by varying the adsorption time in the range of 1-40 min with all other parameters held constant. The adsorption efficiency increased with the adsorption time from 1 to 30 min followed by a plateau. The rate of adsorption was so high that 5 min was enough to adsorb the target compounds. Similarly, desorption efficiency was optimized by varying desorption time in the range of 1-10 min. The desorption efficiency increased with desorption time till 10 min, at which point it plateaued. Therefore, 10 min was selected as desorption time and desorption efficiency was achieved above 75%.
The tolerance of coexisting inorganic ions and organic analogues. By fixing the concentration of each phenylarsonic acids at 50 mg/L and 12 coexisting inorganic ions including Mn 2+ , Cu 2+ , K + , Zn 2+ , Ba 2+ , Fe 3+ , Mg 2+ , Ca 2+ , Pb 2+ , Cr 3+ , Cd 2+ , and  Table 3. Concentration (mg/L) of coexisting analogues when tolerance factor less than 5%.    analogues respectively on the adsorption of target arsenic species was investigated. As shown in Table 3, the allowed concentrations of coexisting analogues were about 500 mg/L or 250 mg/L with a tolerance factor of less than 5%. However, p-NPAA was an exception. The tolerance concentration of p-hydroxybenzoic acid was only 50 mg/L. Possible explanation is shown in Fig. 7.
Despite other interactions, the hydrogen bonding between hydroxyl on the surface of CM-a-CD-Fe 3 O 4 and amino groups of p-APAA and AAPAA was very strong but there was hardly any hydrogen bonding between the hydroxyl and nitro of p-NPAA. In any case, phenylarsonic acid could form inclusion complex with a-CD and therefore the arsenate, hydroxyl or amino group were forced closer to the magnetic particles, leading to stronger interactions between the modified material and phenylarsonic acids. Both selectivity and adsorption efficiency were improved compared to that of Fe 3 O 4 nanoparticles without modification.

Analytical performance
Optimization of UPLC-MS-MS chromatograph separation condition. A mixture of five arsenic compound standards, each at a concentration of 2.0 mg/L, was successfully separated and analyzed by UPLC-MS/MS in less than 3 min on a C18 column. Under the optimal separation conditions, baseline separation was achieved for every arsenic compound. The multireaction monitoring (MRM) chromatograms of arsenic compounds are shown in Fig. 8. Optimization of sample pretreatment condition. To extract phenylarsonic acids from chicken tissue samples, different extraction methods (including ultrasonic extraction, microwave extraction), extraction time (10,20,30,40, 50 and 60 min) and extraction solvents (methanol, ethanol and toluene) were studied. The optimal recovery was above 75% with the ultrasonic extraction method and toluene as the extraction solvent. The proper extraction time was found to be 30 min and extraction should be conducted twice.

Method evaluation
The developed method was validated by determining the linearity and LOD of arsenic species listed in Table 4. A linear response can be seen in the concentration range of 0.20-10 mg/L (enrichment factor 2.5), with the R 2 ranging from 0.9951 to 1.0000. The repeatability study was performed for each of the phenylarsonic acids under the optimal conditions. The LOD of each phenylarsonic acid was estimated by analyzing blank samples spiked at 0.2 mg/kg of each target analytes and they were determined as the lowest concentrations of the analyte for which signal-to-noise ratios were 3 respectively. The resultant repeatabilities expressed as RSD varied from 0.85% to 4.49%. These results show that the method has a high sensitivity and good repeatability.

Sample analysis
The phenylarsonic acid in tissue was stored with the prototype compound. So the sample preparation of real tissue samples is same as the spiked samples [17]. In order to validate the suitability of the developed method, the method was applied to analyze spiked chicken tissue samples. For comparison, the total arsenic in these samples was also determined.
In the samples of chicken meat and liver, 0.2, 2 and 20 mg/kg of each phenylarsonic acids were spiked, respectively. The recoveries of five phenylarsonic acids in chicken tissue samples, as shown in Table 5, fell in the ranges of 77.2%-110.2%, with a RSD less than 12.5%. The chromatograms of spiked chicken meat sample (2.0 mg/kg) and blank chicken meat sample are shown in Fig. 9.

Conclusions
In this work, CM-a-CD-Fe 3 O 4 nanoparticles were synthesized to selectively extract and enrich phenylarsonic acids. CM-a-CD-Fe 3 O 4 nanoparticles exhibited excellent selectivity and adsorption efficiency for five phenylarsonic acids because of the size selectivity of a-CD and the affinity of Fe 3 O 4 to arsenic. In the sub ppb level of phenylarsonic acids, the enrichment factor was higher than 400 and the extraction efficiency higher than 70%. Coupled with UPLC-MS/MS, a fast, selective and convenient analytical method for the determination of phenylarsonic acid was developed. Comparing with published method [14], the developed method showed satisfactory sensitivity due to the selectively adsorption of CM-a-CD-Fe 3 O 4 nanoparticles for phenylarsonic acids in complex sample matrix.