Hollow fibre cell fishing with high performance liquid chromatography for screening bioactive anthraquinones from traditional Chinese medicines

Journal of Chromatography A, 1322 (2013) 8–17

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Hollow fibre cell fishing with high performance liquid chromatography for screening bioactive anthraquinones from traditional Chinese medicines Yunyan Yan, Yaomei Hao, Shuang Hu, Xuan Chen, Xiaohong Bai ∗ School of Pharmacy, Shanxi Medical University, Taiyuan 030001, PR China

a r t i c l e

i n f o

Article history: Received 2 April 2013 Received in revised form 26 October 2013 Accepted 28 October 2013 Available online 4 November 2013 Keywords: Hollow fibre cell fishing High performance liquid chromatography Activity screening Cell fishing factor Anthraquinones Traditional Chinese medicines

a b s t r a c t Hollow fibre cell fishing with high performance liquid chromatography (HFCF-HPLC) is a newly developed method used to screen and fish bioactive compounds in traditional Chinese medicines (TCMs). In the study, colorectal cancer cell HCT116 was first seeded in a hollow fibre and used for screening and fishing active compounds from TCMs. The surface properties of the hollow fibre seeded with HCT116 cells, the non-specific binding between active centres in the fibre and the target compounds, the cell survival rate under different conditions before and after screening, the repeatability and recovery of HFCF-HPLC were investigated in detail. The cell fishing factor of active compound was defined in HFCFHPLC. We employed HFCF-HPLC to screen and fish anthraquinones active compounds group from extracts of Polygonum cuspidatum, Cecropia obtusifolia L. and Polygoni multiflori radix praeparata. Some of the anthraquinones structures screened from TCMs were identified by comparing to the retention time of the reference substances confirmed by mass spectrometry. The ability of permeable membrane of anthraquinones screened by HFCF-HPLC was further described. Indomethacin was used as the positive control substance. Results demonstrated that HFCF-HPLC is an effective, stable and reliable method to screen and analyse bioactive compounds. Other bioactive compounds from TCMs could also be screened. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Traditional Chinese medicines (TCMs) have been used clinically and exhibited reliable therapeutic efficacy. TCMs have attracted increasing global attention as sources of bioactive compounds used to discover new drugs. However, the complex chemical composition of different structural types and low concentrations of active compounds in TCMs complicate the screening and analysis of TCMs. In the body, the effective part of TCM is attributed to a group of active compounds but not to a certain active compound; meanwhile the therapeutic effect may be caused by the active compounds themselves and/or their metabolites. Such conditions complicate the study of pharmacological activity and/or toxicity of drugs. Therefore, bioactive compound groups should be simultaneously screened and analysed to seek the really effective active compound group in TCMs and elucidate the therapeutic principle of TCMs.

∗ Corresponding author. Tel.: +86 13935105965; fax: +86 351 4690114. E-mail addresses: [email protected], [email protected] (X.H. Bai). 0021-9673/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.chroma.2013.10.084

Cell membrane chromatography (CMC) [1–3] is a biological affinity chromatographic method. It has been proved to be a useful tool to investigate the binding interactions of drug and receptors and to screen active compounds in natural products. Cardiac muscle membrane chromatography has been performed to investigate the interactions between cardiac muscle membrane receptors and four compounds in Ligusticum chuanxiong rhizome [4] and L. chuanxiong Hort extract [5]. The chromatography affinity of nine ligands to receptors of EVC304 receptors has also been investigated [6]. CMC with liquid chromatography has been conducted to detect, separate and identify the active compounds of sinapine [7], vauquline and strychnine [8], berberine [9], corynoxeine, isorhynchophylline, isocorynoxeine and rhynchophylline [10] obtained from TCMs. However, this process has several drawbacks: (1) the biomembrane separated from tissues or cells can affect cellular characteristics; (2) professional skill is required to prepare the cell membrane column and this can cause non-specific binding of Si OH to the drugs. Dialysis separation method can also be used to screen and analyse potential bioactive compounds when cells are incubated with TCM extracts in a suitable culture medium. This method has been used to predict bioactive compounds in the extract of Radix scrophulariae [11], Rhizoma salviae miltiorrhizae [12] and Danggui Buxue [13]. However, (1) the impurities and contaminants in TCM extracts can

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

9

Table 1 Formulas and stage mass fragment ions of five anthraquinones in reference solution. Number

Compound

Formula

1

Aloe-emodin

2

Structure

Product ion (m/z)

Precursor ion [39] (m/z)

C15 H10 O5

268.7, 239.8, 210.9, 194.8, 182.8, 167.0

268.7/239.8:

(CHO)

Rhein

C15 H8 O6

282.9, 256.9, 238.8, 210.9

282.9/238.8:

(CO2 )

3

Emodin

C15 H15 O5

268.6, 240.8, 224.9, 209.7

268.6/240.8:

(CO)

4

Chrysophanol

C15 H10 O4

252.8, 234.8, 224.5

252.8/224.5:

(CO)

5

Physcion

C16 H12 O5

282.8, 267.9, 239.8, 211.8

282.8/239.8:

(CH3 + CO)

affect the screened bioactive compounds and (2) the dialysis membrane pores can be blocked; cell mortality rate increases in this time-consuming dialysis. At the end of the 20th century, hollow fibre liquid phase microextraction (HF-LPME) [14] was introduced as a simple, rapid, cheap, efficient and sensitive sample pre-treatment technology. For several years, HF-LPME coupled with HPLC has been used successfully to determine the active compounds present in essential oils [15], alkaloids [16], orange peel glucosides [17], valerenic acid [18], and flavonoids [19,20] in TCMs. Related studies have been conducted in our laboratory to quantify the active compounds of benzene acrylics [21], hydroxybenzoic acids [22], anthraquinones [23], oxymatrine and matrine [24], ephedrine and pseudoephedrine [25], jatrorrhizine, coptisine, palmyatine and berberine [26] in TCMs. The binding parameters of the interactions between active compounds and proteins have also been investigated [27–29]. Our research group used a hollow fibre filled with liposome in the wall pores and the lumen to insert the TCM extract solution; thus, preliminarily biomembrane-permeable compound group in TCMs can be screened and analysed. This process is referred to as hollow fibre liposome microscreening (HFLMS). Employing HFLMS-HPLC, we screened and analysed biomembrane-permeable compound groups of flavonoids and anthraquinones [30] as well as coumarins and lignans [31] from TCMs. We proposed a novel method, called hollow fibre cell fishing with HPLC (HFCF-HPLC) [32] to simulate the actual screening conditions and investigate the interactions between drugs and cells. HFCF-HPLC can also be performed to screen and fish coumarins bioactive compounds and the lignans bioactive compounds under the simulated human drug absorption conditions. In this process, a certain amount of living cells suspension was injected into the fibre lumen by using a syringe, then the hollow fibre was bent into a U-shape and inserted into the TCM extract to screen and fish active compounds. The method was simple, fast, effective, and reliable. However, it could damage the quality and reduce the quantity of cells, and cells in Ushape fibre would be layered or appeared concentration gradient in HFCF screening process because of cell gravity. The active anthraquinone compounds extracted from TCMs, mainly including aloe-emodin, rhein, emodin, chrysophanol and physcion (their chemical structures are shown in Table 1), have specific pharmacological activities such as antibacterial [33],

anticancer [34], antioxidation activities [35], hepatoprotection, blood activation and stasis dissolution [36]. In the present study, colorectal cancer cell HCT116 was first seeded in the inner wall of the hollow fibre which was used as an activity screening terrace to screen and fish active anthraquinone compounds. The seeded process could improve the quality and increase the quantity of living cells. Except that, cells were more adhered to the inner wall, distributed uniformly in fibre and not layered or appeared concentration gradient in HFCF process. Finally, according to the cell type and experiment result, we chose 24 h as the seeding time. So the method was more stable and reliable for screening result than the previous method. In the paper, the surface properties of hollow fibre seeded HCT116 cell, non-specific binding between the active centre in the fibre and the target compounds, cell survival rate under different conditions, the repeatability and recovery of hollow fibre HCT116 cell fishing with HPLC were analysed and researched in detail. The cell fishing factor of target compound was defined and used as an index of binding ability of target compound and cell in HFCF-HPLC. We employed HFCF-HPLC to screen and fish anthraquinones active compounds group from Polygonum cuspidatum, Cecropia obtusifolia L. and Polygoni multiflori radix praeparata extracts. Some of the structures of anthraquinones active compounds screened from TCMs were identified by comparing the retention times with those of the reference substances confirmed by mass spectrometry. The ability of permeable membrane of anthraquinones screened by HFCF-HPLC was further described. Meanwhile, indomethacin with an effective mechanism on HCT116 cells was used as the positive control substance. The reliability of the screening results determined by hollow fibre HCT116 cell fishing was further verified. The test results demonstrate that HFCF-HPLC is an effective, stable and reliable method to screen and analyse bioactive compounds from TCMs. 2. Experimental 2.1. Instruments and apparatus The following instruments were procured and utilised: a 1200 Series liquid chromatograph (Agilent Technologies, Palo Alto, CA, USA) equipped with two G1311A pumps, a G1316A thermostat, and a VWD UV-detector. Mass spectrum (MS) analyses of

10

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

five anthraquinones were performed on a 3200 QTRAP mass spectrometer (Applied Biosystems Company, Foster, CA, USA) equipped with an electrospray ionisation source. FACSCalibur flow cytometry (Becton, Dickinson and Company, Franklin, NJ, USA); CO2 incubator (Thermo Fisher Scientific, Waltham, MA, USA), low-speed centrifuge (LD-5-2A, Beijing, China), pressure steam steriliser (YXQ-LS-30S11, Shanghai, China), electro-heating standing-temperature cultivator (78-1, Hubei, China): Ronghua Group 85-2 magnetic heater with stirrer (Jiangsu, China) and 8 mm × 4 mm stirring bars, XDS-1B inverted biological microscope and a clean bench were also used. The morphology of the hollow fibre was examined under a scanning electron microscope (SEM; JEOL JSM-35C, Tachikawa, Tokyo, Japan). 2.2. Chemicals The analytical reference substances for aloe-emodin (Batch number: A0577, purity ≥98%), rhein (Batch number: A0043, purity ≥98%), emodin (Batch number: A0044, purity ≥98%), chrysophanol (Batch number: A0046, purity ≥98%), and physcion (Batch number: A0045, purity ≥98%) were obtained from the National Institute for the Control of Pharmaceutical and Biological Products, and Chengdu Must Bio-Technology Co., Ltd. Indomethacin tables, P. cuspidatum, C. obtusifolia L. and P. multiflori radix praeparata were purchased from a local store (Taiyuan, Shanxi, China). The plants were identified by Professors Jian-ping Gao and Guan-e Yang (School of Pharmacy, Shanxi Medical University). 2.3. Materials and reagents The polypropylene fibre (PP, with an inner diameter of 0.55 mm and a pore size of 0.18 ␮m) was purchased from the Sea-Water Desalination and Membrane Technology Centre of Tianjin University. Two types of polyvinylidene fluoride fibre (MIF-1a, with an inner diameter of 0.5 mm and a pore size of 0.2 ␮m; MOF-1b, with an inner diameter of 0.5 mm and a pore size of 0.1 ␮m) were purchased (Tianjin Motianmo Engineering Co., Ltd., Tianjin, China). HPLC-grade methanol was obtained from Tianjin Kemio Chemical (Tianjin, China). All of the reagents, including sodium dihydrogen phosphate, disodium hydrogen phosphate, sodium chloride, potassium dihydrogen phosphate (KH2 PO4 ), hydrochloric acid (HCl), phosphoric acid (H3 PO4 ), glacial acetic acid (CH3 COOH), sulphuric acid (2.5 mol L−1 ) and sodium hydroxide (NaOH, 0.1 mol L−1 ) used in this study were of analytical grade. Double-distilled water was used throughout this study. Stable cell lines of human colorectal cancer cell HCT116 were procured from the Pharmacological Experiment Centre of Shanxi Medical University. Improved IMDM medium (Wuhan Boster Bio-Engineering Co., Ltd., Wuhan, China), foetal bovine serum (100 mL, Zhejiang, China) and trypsin solution were purchased (0.25%, Cat. no. T1300-100, Beijing Solarbio Science & Technology Co., Ltd., Beijing, China). Penicillin–streptomycin was used in cell culture (Beijing Solarbio Science & Technology Co., Ltd., Beijing, China). l-␣-Phosphatidylcholine (Cat. no. L8050, Beijing Solarbio Science & Technology Co., Ltd., Beijing, China) and cholesterol (Batch number: 20030812, Beijing Aoboxing Bio-Technology Co., Ltd., Beijing, China) were prepared for liposome. 2.4. Preparation of solutions 2.4.1. Preparation of reference and sample solutions The reference solutions of aloe-emodin (120 ␮g mL−1 ), rhein (78 ␮g mL−1 ), emodin (74 ␮g mL−1 ), chrysophanol (54 ␮g mL−1 ) and physcion (40 ␮g mL−1 ) were prepared in methanol. All of the solutions were stored at approximately 4 ◦ C before use. The mixed reference solutions of aloe-emodin, rhein, emodin, chrysophanol and physcion were 9.6, 12.0, 9.6, 11.2 and 9.6 ␮g mL−1 , respectively.

Powdered Rhizoma P. cuspidatum (50 mesh; 0.1 g) was extracted using 25 mL of chloroform and 20 mL of 2.5 mol L−1 H2 SO4 solution by refluxing for 2 h in a water bath (at 80 ◦ C). After the solution in the extraction procedure was cooled to room temperature, the weight loss of this solution in the extraction procedure was compensated with chloroform. The chloroform layer solution (20 mL) was collected and evaporated to dryness. The residue was dissolved with methanol, transferred to a 25 mL volumetric flask and diluted to scale. The solution was stored at 4 ◦ C before use [36]. A sample (0.1 g) of powdered C. obtusifolia L. (50 mesh) was dissolved in 25 mL of methanol and refluxed for 30 min in a water bath. After this solution was cooled, the weight loss was compensated with methanol. The extract was filtered and 5 mL of the filtrate was added to 20 mL of 10% HCl solution, refluxed for 1 h and cooled immediately. This two-phase mixture was transferred to a separatory funnel and extracted four times by using 20 mL of chloroform. The bound extract was gathered and dried. The residue was dissolved with methanol, transferred to a 25 mL volumetric flask and diluted to scale. The solution was stored at 4 ◦ C before use [36]. A sample (0.1 g) of powdered P. multiflori radix praeparata (50 mesh) was dissolved in 25 mL methanol and refluxed for 1 h in a water bath. After this solution was cooled, the weight loss was compensated with methanol. Afterwards, 5 mL of the filtered extract was dried and 10 mL of 8% HCl solution was added to the mixture by ultrasound for 2 min. Chloroform (10 mL) was subsequently added; the resulting mixture was refluxed for 1 h and then cooled. This two-phase mixture was transferred to a separatory funnel and extracted thrice by using 10 mL of chloroform. The bound extract was gathered and dried. The residue was dissolved with methanol, transferred to a 10 mL volumetric flask and diluted to scale. The solution was stored at 4 ◦ C before use [37]. Ten indomethacin tablets were ground. The powdered indomethacin (50 mg) was precisely weighted and dissolved in a 100 mL volumetric flask. Methanol (30 mL; 50%) was added and the mixture was dissolve by ultrasound. After this solution was cooled, the weight loss was compensated with 50% methanol. The extract was filtered and 2 mL of the filtrate was added to 10 mL of 50% methanol [38]. 2.4.2. Preparation of l-˛-phosphatidylcholine and liposome solutions l-␣-Phosphatidylcholine (0.8 g) was weighted and dissolved in 20 mL of phosphate buffer (1.7 g KH2 PO4 , 98.75 mL of 0.1 mol L−1 NaOH; pH 7.4) by ultrasonic. The solution was stored at 4 ◦ C before use. l-␣-Phosphatidylcholine (0.8 g) and cholesterol (0.1 g) were weighted and dissolved in 15 mL of chloroform. After the solvent removed by reduced pressure distillation, the residue was dissolved in 20 mL of phosphate buffer (1.7 g KH2 PO4 , 98.75 mL of 0.1 mol L−1 NaOH; pH 7.4) by ultrasonic. The solution was stored at 4 ◦ C before use. 2.5. Chromatographic conditions Chromatographic separations were performed using a Zorbax Eclipse XDB-C18 column (5 ␮m, 4.6 mm × 150 mm, Agilent, USA). The injection volume of the samples was 20 ␮L. For the anthraquinones, optimum chromatographic separation was achieved in a short period with a binary mobile phase under gradient conditions. The mobile phases A and B were 0.6% H3 PO4 and methanol, respectively. The gradient elution programmes were listed as follows: 0–6 min, 70% B (0.8 mL min−1 ); 6–12 min, 78% B (1.0 mL min−1 ); and 12–13 min, 85% B (1.0 mL min−1 ). The detection wavelength was set at 434 nm. The column was maintained at 37 ◦ C during the experiment. For indomethacin, the mobile phases A and B were 0.1 mol L−1 CH3 COOH and acetonitrile (20:80),

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

11

Fig. 1. Procedure of the hollow fibre seeded with HCT116 cells. (a) Cells were drawn into a syringe without a needle; (b) the needle syringe was installed; (c) the hollow fibres were placed in the culture flask; (d) the cells liquid was seeded into the fibres until another port released cell liquid; (e) 5 mL of the nutrient solution was added to the culture flask and cultivated for 24 h.

Fig. 2. Procedure of the hollow fibre HCT116 cells fishing bioactive compounds from the TCMs extracts. (a) Both the ends of the hollow fibre with the cells were sealed with cotton thread; (b) the hollow fibre was inserted into the extract; (c) the bioactive compounds were screened by stirring at 37 ◦ C; (d) the fibre was removed and the seals at the two ends of the fibre were cut; (e) the cell acceptor phase in the fibre lumen was transferred to a Eppendorf tube and the fibre was washed with methanol.

respectively, at a flow rate of 1.0 mL min−1 . The detection wavelength was set at 228 nm. The column was maintained at 30 ◦ C during the experiment [38].

2.8. Cell fishing factor of active compound

2.6. Cell culture and seeding Human colorectal cancer cell HCT116 was selected and placed in a nutrient medium containing 10% foetal bovine serum and 1% penicillin–streptomycin containing 10,000 U mL−1 penicillin and 10,000 ␮g mL−1 streptomycin (v/v) at 37 ◦ C in 5% humidified CO2 and 95% air. The medium was replaced every other day. After 3–4 d, confluent cell monolayers in the nutrient medium were obtained. Under an inverted microscope (200× magnification), the neatly arranged cells with a complete cytoplasm exhibited acceptable growth conditions and uniformly covered the bottom of the bottles; these cells were collected for subsequent experiments. The cells were then digested with trypsin at 37 ◦ C for 5 min. The trypsinized cells (approximately 107 mL−1 ) were suspended in a nutrient solution. The trypsinized cells were transferred to a centrifugal tube and then centrifuged at 1000 rpm for 3 min. The supernatant fluid was removed and a certain amount of nutrient solution was added. The cells were drawn into a syringe without needle (Fig. 1a) and a needle was then installed (Fig. 1b). After the hollow fibres were placed in the culture flask (Fig. 1c), the cells in the syringe were injected and seeded into the fibres until another port began releasing cells liquid (Fig. 1d). The nutrient solution (5 mL) was added to the culture flask and cultivated for 24 h (Fig. 1e). 2.7. HFCF-HPLC procedure The HFCF-HPLC procedure was conducted according to the following procedures. TCMs extract solution (800 ␮L) and phosphate buffer solution (PBS, 7.2 mL; pH 7.4; at 37 ◦ C) were placed in a 10 mL flat-bottom vial with a screw cap and a magnetic stirring bar (8 mm × 4 mm). The vial was then fastened onto a magnetic heater. A 7 cm hollow fibre seeded with cells was sealed with a cotton thread tied on both ends of the fibre (Fig. 2a) and then inserted into the vial filled with the TCMs extract (Fig. 2b). The TCMs extract was stirred at 600 rpm at a constant 37 ◦ C for 3 h (Fig. 2c). Afterwards, the hollow fibre was removed from the vial and the seals at both ends were cut (Fig. 2d). The cell acceptor phase in the fibre was washed with 40 ␮L of methanol and then transferred to a clean, dry Eppendorf (EP) tube (Fig. 2e). The EP tube containing the cell acceptor phase was centrifuged at 10,000 rpm for 20 min; 20 ␮L of the supernatant fluid was subjected to HPLC analysis. This process was performed thrice for screening. A hollow fibre seeded with nutrient solution was simultaneously used for screening.

The nutrient hollow fibre (i.e., without cells) did not detect anything or detected less active compounds, but the hollow fibre with cells could detect many active compounds from TCMs. This result indicated that the active compounds detected by HFCF-HPLC were produced because of combination with compounds to the cells rather than the physical and chemical reactions of compounds with other reagents (except cells) or the hollow fibre. Therefore, the cell fishing factor (CFF) of the active compound is defined as the ratio of the active compound concentration in the hollow fibre with cells (Ccell ) to the active compound concentration in the nutrient hollow fibre (Cnutr ) after screening balance. Under certain conditions, the active compound concentration is proportional to the peak area (A): CFF =

Ccell A = cell Cnutr Anutr

(1)

where Acell and Anutr are the chromatographic peak areas of the active compounds in the hollow fibre with cells and hollow fibre with nutrient after screening balance, respectively. Thus, CFF is the index of cell–drug binding ability that we investigated by HFCFHPLC under certain conditions. The higher CFF corresponds to the stronger binding ability and the lower CFF corresponds to the weaker binding ability. In order to further verify the interaction mechanism of drug–cell, we also defined the phosphatidylcholine fishing factor (PFF) and liposome fishing factor (LFF) which were the indexes to investigate binding ability of phosphatidylcholine–drug and liposome–drug under certain conditions. 3. Results and discussion 3.1. Characterisation of MOF-1b hollow fibre seeded with cells In order to characterise the binding of cells to the inner wall of the hollow fibre, we investigated and compared the SEM images of the internal surface of the MOF-1b hollow fibres containing nutrients and overgrown cells. These fibres were dehydrated using an ethanol gradient (25%, 50%, 75% and 100%) for 5 min each; ethanol was then evaporated. Afterwards, the hollow fibres were then cut longitudinally to reveal their internal surface. The internal surfaces of the hollow fibres were scanned under an SEM at 25 kV and at low 1000 and 2000 amplification factor. The inner surface of the hollow fibre filled with nutrients (Fig. 3a and c) shows that inner wall of the fibre contained a meshed structure revealing numerous smooth and uniformly distributed fibre pores and small pits. This

12

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

Fig. 3. SEM micrographs of the internal surface of the MOF-1b hollow fibre containing the nutrient and the overgrowed cells. (a) Fibre with nutrient solution at 25 kV × 1000; (b) fibre with cells at 25 kV × 1000; (c) fibre with the nutrient at 25 kV × 2000; and (d) fibre with the cells at 25 kV × 2000.

observation indicated that the inner wall and wall pores of the fibre did not contain any additives or cells. The SEM micrograph of the internal surface of the hollow fibre seeded with cells (Fig. 3b and d) shows that uniformly distributed and tightly arranged cells covered the fibre wall pores. This result also indicated that the closely arranged cells adhered to the internal surface and covered the holes on the surface of the hollow fibres. The cells can healthily grow in the hollow fibre and remain alive in the screening process because (1) the hollow fibre exhibits a porous polymer three-dimensional (3D) structure (with a hole rate >80%) and a large internal surface area that allows cell seeding, causes cells to adhere to the walls and induces the cells to permeate the nutrient solution; (2) the hollow fibre shows good histocompatibility that provides cells with an interface and a microenvironment conducive to growth and proliferation; and (3) the hollow fibre can be bent to form different shapes when TCMs are screened and filtered because the fibre exhibits mechanical strength and toughness. 3.2. Cell survival rate under different conditions before and after screening The cell death rates before and after screening under different screening conditions were determined by flow cytometry (Fig. 4). After the cells were treated, the fluorescence signals were detected using the Cell Quest software and the resulting data were analysed. The abscissa denotes the fluorescence intensity of PI and the ordinate denotes the relative cell count. We then determined the proportion of the living cells based on the negative M1 value. The death rate of the cells suspension cultivated in the culture flask was 2.4% (Fig. 4a). For the cells seeded in a fibre lumen by enzyme digestion before screening, the death rate was 22.8% (Fig. 4b) separately. Therefore, the fibre could be used for cell seeding as well as the subsequent screening and fishing processes, although the survival rate of the cells seeded in the hollow fibre was approximately 20% less than those cultivated directly in the culture flask. Fig. 4c and d shows the separate screening of the PBS blank solution and the sample solution in a static state (stirring at a speed of 0 rpm) for 3 h, in which the fibre was exposed. Fig. 4c and d also shows that the cell death rate in the screened sample solution was approximately 30% higher than that in the blank solution. This result indicated that the active compounds in the TCMs extract could damage colorectal cancer HCT116 cells. Fig. 4d–f shows the

Fig. 4. Cell survival rate determined by flow cytometry analysis under different experimental conditions before and after screening. (a) Suspension cells in the culture flask by enzyme digestion; (b) suspension cells growing in the fibre by enzyme digestion before screening; (c) suspension cells growing in the fibre by enzyme digestion after screening with PBS blank solution at a static state; (d) suspension cells growing in fibre by enzyme digestion after screening the sample solution in idle state; (e) suspension cells growing in fibre by enzyme digestion after screening sample solution at 300 rpm; and (f) suspension cells growing in fibre by enzyme digestion after screening sample the solution at 600 rpm.

screening of the active compound in the sample solution at 0, 300 and 600 rpm for 3 h. Under the same conditions, the cell death rate increased from 65.44% to 82.17% as the stirring speed increased. This result probably occurred because the increase in the stirring speed accelerated the transfer and diffusion speed of the active compounds; this increase in the stirring speed also increased the concentration of the active compounds reaching the cells surface and the ability to eradicate HCT16 cells. These results show that the cell survival rates ranged from 77% (Fig. 4b) to 18% (Fig. 4f) in the whole screening process at 600 rpm, indicating the interaction between active compounds and living cells. Therefore, the screened active compounds were reliable. 3.3. Selection of experimental conditions 3.3.1. Selection and preparation of hollow fibres In HFCF, the porous hollow fibre as a carrier of the living cells not only protects cells but also allows active compounds to diffuse into and out of the hollow fibre lumen, blocks the entry of contamination particles into and prevents the movement of biological macromolecules out of the fibre lumen. The frequently used hollow fibre

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

13

includes polypropylene fibre (PP) and polyvinylidene fluoride fibre (MIF-1a or MOF-1b) in HFCF-HPLC. These fibres are high polymer materials with good physical properties (non-toxic, tasteless and oyster white appearance); these substances also exhibit extremely stable chemical properties, including resistance to strong acid and strong alkali. The 3D structure of the porous wall and its good biological compatibility provide the cells with a microenvironment favourable for healthy growth and proliferation. In this study, PP, MIF-1a and MOF-1b used to seed the cells were investigated. In the process, a syringe was needed to transfer the cells into or out of the fibre lumen. Hence, the fibre should have sufficient inner diameters. In addition, the repeatability of HFCF-HPLC, which is determined mainly by the non-specific binding between activity centres of the hollow fibre and the active compounds, is significant to screen and isolate bioactive compounds from the TCM extract solution accurately and efficiently. To minimise nonspecific binding between the activity centres of the hollow fibre and the bioactive TCM compounds, we should select the hollow fibre with the largest peak area and the most peak number of the bioactive compounds for the cell hollow fibre and the smallest peak area and the least peak number of bioactive compounds for the nutrient hollow fibre. The experimental results indicated that the MOF-1b hollow fibre could meet the aforementioned conditions. Therefore, MOF-1b hollow fibre was used in this study. The MOF-1b hollow fibre was washed by ultrasonication using acetone, methanol, acid, alkali and distilled water for 5–10 min and then dried directly in air. The hollow fibre was cut to a size of 7 cm. High pressure sterilisation was performed and the resulting product was collected to seed the cells. The lumen and the pores of the hollow fibre were overgrown cells, which covered the active centres of the hollow fibre, thereby suppressing fibre activity to a certain degree. 3.3.2. Selection of string time The sample solution phase was stirred using an agitator in HFCF process to obtain the screening equilibrium between the sample solution phase and the cell receptor phase of the active anthraquinones. A suitable stirring speed is important because an increase in the stirring speed can accelerate the transfer and diffusion speed of active compounds, increase the concentration of active compounds reaching the cell surface and enhance the ability to kill HCT16 cells (Fig. 4d–f). Thus, we selected 600 rpm as the optimal speed. At this speed, the effects of stirring time (1–4 h) on the screening efficiencies of the five target analytes were investigated. We found that almost no interaction occurred between the target analytes and the cells when screening was performed for 1 h. By contrast, the interaction increased as screening was performed at an extended time. However, the difference of the results obtained at 3 h and 4 h was not significant. Therefore, the optimal screening time in this study was 3 h. 3.4. Identification of anthraquinone model compounds and chromatographic analysis of anthraquinones in TCM extract solution At an ESI negative ion mode, five anthraquinone reference substances, namely, aloe-emodin, rhein, emodin, chrysophanol and physcion were used as model compounds. Each compound produced a good signal and generated a strong [M−H] excimer ion. Table 1 provides an overview of the MS parameters, including a list of the product ions from the MS/MS spectra. After HFCF-HPLC, some of the structures of the active anthraquinones screened from TCMs were identified by comparing with the retention time of the reference substances confirmed by MS. Using the conditions described in Section 3.2, we investigated the chromatographic behaviours of the five anthraquinones obtained from TCMs. The chromatograms of the reference

Fig. 5. Chromatograms of anthraquinone compounds and reference substances. 1, Aloe-emodin; 2, rhein; 3, emodin; 4, chrysophanol; 5, physcion. (a) reference substances; (b1 –b3 ) TCM extract. (A) P. cuspidatum, (B) C. obtusifolia L., (C) P. multiflori radix praeparata.

substances and the target analytes in P. cuspidatum, C. obtusifolia L. and P. multiflori radix praeparata were shown in Fig. 5A–C. Under the optimal chromatographic condition, most of the analytes could exhibit ideal peak shape and good resolution.

14

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

3.5. Binding between activity centres of hollow fibre and bioactive compounds The chromatographic behaviours of the active compounds were investigated after screened and isolated to verify whether the main interaction mechanism is the binding between cells and bioactive compounds and not the binding between activity centres of MOF1b hollow fibre and bioactive compounds during HFCF-HPLC. In Line a (Fig. 6A–C), the chromatographic peak areas and the number of the chromatographic peaks of the compounds screened using the MOF-1b hollow fibres with the nutrient solution were smaller and less than those screened using the MOF-1b hollow fibres with HCT116 cells. These results indicated that binding slightly occurred between the MOF-1b hollow fibre and the compounds. The main mechanism of the bioactive compounds screened and isolated by HFCF-HPLC was the interaction between living cells and bioactive compounds. 3.6. Repeatability and recovery of screening result The repeatability of the screening of bioactive compounds from TCM extract solutions can greatly affect the reliability of experimental result. Therefore, we investigated the repeatabilities of the retention times (tR s) and the relative peak areas (RPAs, the ratio of the peak area of each active compound to A1, that is, aloe-emodin, an active compound) of the anthraquinones screened by HFCFHPLC. The repeatability results of tR s and RPAs of anthraquinones screened by HFCF-HPLC using HCT116 cells were shown in Table 2. The relative standard deviation (RSDs) of the tR s (n = 3) and the RPAs (n = 3) of the screened analytes were <0.92% and <9.67%, respectively. This result indicated that the HFCF-HPLC method used to screen anthraquinones from the TCM extract solution has good repeatability. The recovery is another index parameter affecting the reliability of experimental result. Therefore, we tested the recoveries of the known anthraquinones screened by HFCF-HPLC. The recoveries of the spiked samples of P. cuspidatum, C. obtusifolia L., and P. multiflori Radix praeparata ranged from 33.3% to 57.1%, 25.0% to 50.0% and 44.4% to 54.3% (as Table 3), respectively. These results show that recoveries of compounds screened in TCM samples by HFCFHPLC are low. However, as a qualitative analysis method, they are acceptable. 3.7. Application of HFCF-HPLC In the work, five anthraquinones, namely aloe-emodin, rhein, emodin, chrysophanol and physcion were chosen as model bioactive compounds and the bioactive compounds in P. cuspidatum, C. obtusifolia L., P. multiflori radix praeparata were screened by HFCFHPLC. Fig. 6 shows that the peak area and the number of peaks of the active compounds screened and isolated using the hollow fibres seeded with HCT116 cell were higher and more than those of hollow fibres with nutrient solution. Different retention times of the bioactive compounds revealed the following results: P. cuspidatum (Fig. 6A), peak 1 aloe-emodin, peak 3 emodin and peak 5 physcion; P. multiflori radix praeparata (Fig. 6C), peak 1 aloe-emodin and peak 5 physcion; and C. obtusifolia L. (Fig. 6B), peak 2 rhein, peak 4 chrysophanol and seven unidentified active compounds (peaks X1 –X7 ) except the three active compounds screened from P. cuspidatuma and P. multiflori radix praeparata. In addition, only strong interactions between the compounds and the cells were detected. Weak interactions among the constituents were not observed because of low sensitivity. The interactions between the nutrient solution and the TCM extracts were weak or non-existent. The screening results also indicated that some compounds of P. cuspidatum, C. obtusifolia L. and P. multiflori radix praeparata extracts can react

Fig. 6. Chromatograms of the hollow fibre HCT116 fishing active anthraquinones in TCMs. 1, Aloe-emodin; 2, rhein; 3, emodin; 4, chrysophanol; 5, physcion; X1 –X7 , unidentified active compounds. (a) Screening with nutrient; (b) screening with HCT116 cells. (A) P. cuspidatum, (B) C. obtusifolia L., (C) P. multiflori radix praeparata.

with HCT116 cells. A group of effective compounds rather than single compound were screened out from these TCM extracts. The information further showed that the active compounds in TCMs are multi-component, synergistic and holistic.

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

15

Table 2 The repeatability of screening results for anthraquinone active compounds in TCM extract solutions using hollow fibre HCT116 cell fishing with HPLC. Average tR (min) and RPA

RSD (%), n = 3

5.362 1.00 14.016 170.74 18.813 1.17

5.400 1.00 14.066 173.36 18.834 1.08

0.69 – 0.53 6.94 0.23 5.90

4.521 10.79 4.818 1.23 5.425 1.00 6.611 0.59 7.644 0.66 8.986 1.56 11.585 2.39 12.233 0.76 14.035 1.92 13.919 0.60 16.557 1.12 18.980 0.49

4.514 11.39 4.810 1.30 5.538 1.00 6.483 0.57 7.578 0.67 8.898 1.51 11.600 2.42 12.325 0.67 13.840 2.35 14.035 0.65 16.897 1.04 18.888 0.46

4.502 11.39 4.802 1.24 5.474 1.00 6.547 0.60 7.657 0.65 8.950 1.56 11.653 2.35 12.218 0.73 13.931 2.17 14.046 0.64 16.691 1.16 18.983 0.45

0.49 4.30 0.40 3.66 0.86 – 0. 80 4.91 0. 92 4.06 0. 42 2.62 0. 73 3.35 0.76 5.81 0.57 8.45 0.77 4.59 0.59 9.67 0.41 8.31

5.271 1.00 19.056 1.44

5.365 1.00 18.908 1.52

5.323 1.00 18.923 1.48

0.73 – 0.54 2.17

TCMs

Active compound group

Parameter

I

II

III

P. cuspidatum

A1

tR RPA tR RPA tR RPA

5.388 1.00 14.172 189.23 18.797 1.03

5.451 1.00 14.009 160.10 18.894 1.04

tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA tR RPA

4.471 11.99 4.777 1.19 5.461 1.00 6.547 0.64 7.749 0.61 8.965 1.61 11.773 2.24 12.098 0.76 13.919 2.25 14.184 0.67 16.718 1.31 19.081 0.40

tR RPA tR RPA

5.334 1.00 18.805 1.48

A3 A5 C. obtusifolia L.

X1 X2 A1 X3 X4 X5 X6 A2 X7 A3 A4 A5

P. multiflori radix praeparata

A1 A5

tR (min) and RPA

In order to further verify the effect target screened by the cell in HFCF-HPLC and describe their ability of permeable membrane, we used l-␣-phosphatidylcholine and liposome in the fibre instead of the cell to screen target compounds from the TCM extracts under the same condition. Meanwhile, a hollow fibre with phosphate buffer was simultaneously used for screening. The combining

index CFFs, LFFs and PFFs of the screened active compounds are also shown in Table 3. These results show that (1) the CFFs of the anthraquinone compounds were <2.2, indicating that the combined abilities of HCT116 cells and anthraquinones were week. (2) P. cuspidatum and C. obtusifolia L. contained emodin with similar CFFs; aloe-emodin and physcion in the three TCMs were also similar,

Table 3 Screening results of the hollow fibre HCT116 cell, liposome and phosphatidylcholine fishing with HPLC for anthraquinone compounds in TCMs (n = 3). TCMs

P. cuspidatum

C. obtusifolia L.

Number of being cell screened compounds 3

Retention time (min)

Compared with reference substances

Recovery (%)

Cell fishing factor, CFF

5.400 14.066 18.834

Aloe-emodin Emodin Physcion

33.3 50.6 57.1

1.5 1.9 1.9

– 1.3 1.4

– – 1.0

A3 A4 A5

4.502 4.802 5.474 6.547 7.657 8.950 11.653 12.218 13.931 14.046 16.691 18.983

Unidentified Unidentified Aloe-emodin Unidentified Unidentified Unidentified Unidentified Rhein Unidentified Emodin Chrysophanol Physcion

– – 25.0 – – – – 33.3 – 25.0 48.5 50.0

3.8 1.4 1.8 1.1 1.4 4.0 1.5 1.2 3.3 1.2 2.2 1.3

– – – 1.0 – 2.9 1.1 – – 1.0 1.1 1.4

– – 1.0 – 1.2 – 1.3 1.0 1.8 1.5 1.5 1.0

A1

5.323

Aloe-emodin

44.4

1.2

–

–

A5

18.923

Physcion

53.3

1.3

1.5

1.2

Known compounds peak no.

Unknown compounds peak no.

A1 A3 A5

12

X1 X2 A1 X3 X4 X5 X6 A2 X7

P. multiflori radix praeparata

2

Liposome fishing factor, LFF

Phosphatidylcholine fishing factor, PFF

16

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17

4. Conclusions

Fig. 7. Chromatograms of the hollow fibre HCT116 fishing indomethacin. (a) Screening with nutrient; (b) screening with HCT116 cells.

indicating that the combined abilities of the same active compound in different TCMs were similar. (3) in C. obtusifolia L., the CFFs of the active compounds of the fished and unidentified X1 (CFF = 3.8), X5 (CFF = 4.0), X7 (CFF = 3.3) were higher, indicating that the X1 , X5 and X7 could be the potential biological active compounds; (4) except that, we found most of the LFFs and PFFs of the target analytes were smaller than their CFFs. The closer PFFs or LFFs are to CFFs (such as X3 , X4 , X6 , A2 , A3 and A5 ), the more possibility they are nonspecific effect of the target analyte and cell membrane molecules. The compounds screened with smaller LFFs and PFFs (such as A1 , X5 , X7 and A4 ) maybe have non-specific effect with phosphatidylcholine and liposome and specific binding with specific acceptors. However, compounds, such as X1 and X2 screened by liposome or phosphatidylcholine (without LFFs and PFFs), are possible to enter cell through cell channel or combine with acceptors. The effect target compounds screened and fished by the HCT116 cell still need to be further studied. Using the HFCF-HPLC method, we screened and determined numerous anthraquinones including aloe-emodin, rhein, emodin, chrysophanol and physcion, which exhibit different levels of binding ability with the colorectal cancer cell HCT116. The active compound with higher CFF has important significance to find possible new anti-colorectal cancer anthraquinone compounds in TCMs. In general, the active compounds, absorbed by the body after TCM is taken, can elicit a therapeutic action. However, numerous researches have shown that most of the metabolites of the active compounds also exhibit biological activities. Hence, the therapeutic effect of active compounds maybe caused by the drug itself and/or their metabolites when drug enters the body. Tian et al. [40] analysed and researched emodin and its metabolites in the body and indicated that these metabolites mainly included aloe-emodin, rhein, chrysophanol and physcion. In this study, these metabolites were the active compounds screened by HFCF-HPLC. This result indirectly indicated that emodin and its metabolites could be the key active compounds in the body. Therefore, HFCF-HPLC can be performed to find new active metabolites in the body. Under the conditions described in Section 3.2, the chromatographic behaviour of indomethacin [41–43] (positive control substance), which can interact with HCT116 cells, is showed in Fig. 7. The peak value of the control group was higher than that of the blank group. CFF is 5.8, further indicating that screening result by HFCF-HPLC is reliable. Similarly, in order to verify the effect target of positive control substance, we tested its ability of permeable membrane. The LFF and PFF for indomethacin were 3.5 and 5.6, which indicated the major interaction can be indomethacin and phosphatidylcholine. Therefore, the proposed method could be used to screen and detect active compounds in TCMs.

The proposed HFCF-HPLC method with colorectal cancer cell HCT116 seeded in a hollow fibre has been successfully used to screen and fish anthraquinones bioactive compounds from TCMs. In contrast to the former HFCF-HPLC [32], the proposed method has the following advantages. (1) Hollow fibre seeded cells is first employed as an activity screening terrace. In the screening process cells were more adhered to the inner wall, uniformly distributed in fibre and not layered or appeared concentration gradient in HFCF process. So the procedure is more reliable and stable. (2) The CFF of active compound is defined and used as the index of the binding ability of active compounds and cells in HFCF-HPLC, which has important significance to find new active compounds in TCMs. The method, however, has the following disadvantages. (1) The target receptors, action mechanism and pharmacologic effects of the bioactive compounds screened and fished by HFCF require further research by using more extensive bioassays. (2) Sufficient quantity of active compounds should be obtained to verify pharmacological activity and identify the structures of compounds, but this procedure is difficult to perform with HFCF-HPLC. (3) The active compounds without reference substances need subsequent LC–MS analysis, spectrum control and literature control to identify their possible structure. Hence, some aspects of the method should be improved. In all, the successful application of HFCF-HPLC method can provide a stable, reliable, efficient and universal way to screen the activities of TCMs. This method can also be applied to study multi-component, multi-target, entirety and coordination research of TCMs. Acknowledgements This study was supported by the National Nature Science Foundation of China (No. 81041084), the Nature Science Foundation of Shanxi Province, China (No. 2011011035-2), and the Programme for the Top Science and Technology Innovation Teams of Higher Learning Institutions of Shanxi Province, China (2011). References [1] L. Feng, Y.H. Xua, S.S. Wang, W.A. yeung, Z.G. Zheng, Q. Zhu, P. Xiang, J. Chromatogr. B 881–882 (2012) 55. [2] L.C. He, G.D. Yang, X.D. Geng, Chin. Sci. Bull. 44 (1999) 826. [3] L.C. He, S.C. Wang, X.D. Geng, Chromatographia 54 (2001) 71. [4] Y.N. Zhang, X.F. Yue, Z.Q. Zhang, China J. Chin. Mater. Med. 29 (2004) 660. [5] X.F. Yue, Y.N. Zhang, Z.Q. Zhang, Z.J. Tian, J.X. Yang, F.R. Li, China J. Chin. Mater. Med. 30 (2005) 129. [6] D. Zhang, B.X. Yuan, X.L. Deng, G.D. Yang, L.C. He, J. Xian Jiao Tong Univ. 30 (2009) 315. [7] T. Zhang, S.L. Han, J. Huang, S.C. Wang, J. Chromatogr. B 912 (2013) 85. [8] M. Sun, Y. Guo, B. Dai, C. Wang, L. He, Rapid Commun. Mass Spectrom. 26 (2012) 2027. [9] J. Liu, J. Yang, S. Wang, J.Y. Sun, J.F. Shi, G.Z. Rao, A. Li, J.Z. Gou, J. Chromatogr. B 904 (2012) 115. [10] J. He, S. Han, F. Yang, N. Zhou, S. Wang, J. Chromatogr. Sci. (2012), http://dx.doi. org/10.1093/chromsci/bms188. [11] Y.F. Zhu, Z.M. Bi, C.W. Liu, M.T. Ren, F.H. Wu, P. Li, J. China Pharm. Univ. 39 (2008) 228. [12] Q.Q. Dong, P. Li, Y. Song, Z.M. Bi, Chin. J. Anal. Chem. 35 (2007) 648. [13] X. Zhang, L.W. Qi, L. Yi, et al., Biomed. Chromatogr. 22 (2008) 157. [14] S. Pedersen-Bjergaard, K.E. Rasmussen, Anal. Chem. 71 (1999) 2650. [15] C.H. Dong, N. Yao, A.Q. Wang, X.M. Zhang, Anal. Chim. Acta 536 (2005) 245. [16] H.W. Lu, H.Q. Sun, X. Liu, S.X. Jiang, Chem. Papers 63 (2009) 351. [17] J. An, F. Zhang, Y. Lu, Y. Jiang, J. Chin. Med. Mater. 36 (2011) 455. [18] M. Mirzaei, H. Dinpanah, J. Chromatogr. B 879 (2011) 1870. [19] Y. Xian, B.C. Wang, X. Zhang, S.P. Yang, W. Li, Q. Tang, K. Gurinder Singh, J. Pharm. Biomed. Anal. 54 (2011) 311. [20] M. Hadjmohammadi, H. Karimiyan, V. Sharifi, Food Chem. (2013), http://dx.doi. org/10.1016/j.foodchem.2013.02.083. [21] X.Y. Wang, X. Chen, X.H. Bai, Chin. J. Anal. Chem. 37 (2009) 35. [22] Y. Zhu, X. Chen, F.L. Zheng, X.H. Bai, Chin. J. Chromatogr. 27 (2009) 769. [23] Y. Zhu, X. Chen, X.H. Bai, J.C. Guo, Chin. Pharm. J. 44 (2009) 1334. [24] X.H. Bai, X. Yang, X. Chen, L.H. Wang, Chin. J. Anal. Chem. 36 (2008) 182.

Y. Yan et al. / J. Chromatogr. A 1322 (2013) 8–17 [25] X. Chen, X.H. Bai, X. Wang, J. Wang, W. Bu, Chin. J. Chromatogr. 28 (2010) 1144. [26] J. Zhang, X. Chen, X.H. Bai, W.Y. Ma, L.S. Zhang, Chin. Remed. Clin. 3 (2012) 277. [27] G.C. Xi, S. Hu, X.H. Bai, Chem. Res. Chin. Univ. 27 (2011) 764. [28] S. Hu, Y.J. Zhang, X.H. Bai, Chromatographia 74 (2011) 503. [29] X. Zhang, X. Chen, X.H. Bai, Chin. J. Anal. Chem. 39 (2011) 857. [30] X. Chen, X. Zhang, J. Tian, X.H. Bai, Chromatographia 75 (2012) 1395. [31] L.H. Li, X. Xue, C. Peng, X.H. Bai, Chin. J. Anal. Chem. 41 (2013) 337. [32] X. Xue, L.H. Li, X. Chen, S. Hu, X.H. Bai, J. Chromatogr. A 1280 (2013) 75. [33] L.R. Zhou, Q. Peng, Mod. Hosp. 11 (2011) 34. [34] Q.X. Zhang, B. Cao, J. North Pharm. 9 (2012) 40.

17

[35] J.M. Cherng, W. Chiang, J.H. Wang, C.M. Lin, C.Y. Lee, C.M. Shih, L.C. Chiang, Food Chem. 107 (2008) 1576. [36] J. Tian, X. Chen, X.H. Bai, Chin. J. Chromatogr. 38 (2010) 1593. [37] Committee of National Pharmacopoeia, The Pharmacopoeia of the People’s Republic of China, vol. 1, China Medicine Science Press, Beijing, 2010. [38] Y.H. Sun, Chin. Pharm. Affairs 26 (2012) 172. [39] J. Dong, H. Wang, L.R. Wan, Y.S. Duan, S.Z. Chen, Chin. J. Chromatogr. 27 (2009) 425. [40] J. Tian, X. Chen, X.H. Bai, J. Sep. Sci. 35 (2012) 145. [41] M.H. Xu, G.Y. Zhang, China J. Mod. Med. 5 (2005) 667. [42] H.M. Wang, G.Y. Zhang, J. First Mil. Med. Univ. 24 (2004) 184. [43] G.Q. Wang, Y.J. Fang, Z.H. Lu, Z.Z. Pan, D.S. Wan, J. Pract. Med. 28 (2012) 1756.