Kinetic Cellular Phenotypic Profiling: Prediction, Identification, and

ARTICLE pubs.acs.org/ac

Kinetic Cellular Phenotypic Profiling: Prediction, Identification, and Analysis of Bioactive Natural Products Huiying Fu,† Wenqing Fu,‡ Mingjiao Sun,† Qiyang Shou,§ Yunyan Zhai,† Hongqiang Cheng,† Li Teng,† Xiaozhou Mou,† Yanwei Li,† Shuying Wan,† Shanshan Zhang,† Qinqin Xu,† Xue Zhang,† Jiucun Wang,‡ Jenny Zhu,|| Xiaobo Wang,|| Xiao Xu,|| Guiyuan Lv,§ Li Jin,‡ Wensheng Guo,‡,^ and Yuehai Ke*,† †

)

Program in Molecular Cell Biology, Department of Basic Medical Sciences, Zhejiang University School of Medicine, Hangzhou 310058, China ‡ State Key Laboratory of Genetic Engineering, School of Life Sciences, Fudan University, Shanghai 200433, China § Zhejiang University of Traditional Chinese Medicine, Hangzhou 310053, China ACEA Biosciences Incorporated, San Diego, California 92126, United States ^ Center for Clinical Epidemiology and Biostatistics, School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania 19104, United States

bS Supporting Information ABSTRACT: Natural products have always been a major source of therapeutic agents; however, the development of traditional herbal products has been currently hampered by the lack of analytic methods suitable for both high-throughput screening and evaluating the mechanism of action. Cellular processes such as proliferation, apoptosis, and toxicity are wellorchestrated in real time. Monitoring these events and their perturbation by natural products can provide high-rich information about cell physiological relevancies being involved. Here, we report a novel cell-based phenotypic profiling strategy that uses electronic impedance readouts for real-time monitoring of cellular responses to traditional Chinese medicines (TCMs). The utility of this approach was used to screen natural herbs that have been historically documented to cure human diseases and that have been classified into seven clusters based on their mechanisms of action. The results suggest that herbal medicines with similar cellular mechanisms produce similar time/dosedependent cell response profiles (TCRPs). By comparing the TCRPs produced by the Chinese medicinal Cordyceps sinensis with similar TCRPs of chemical compounds, we explored the potential use of herbal TCRPs for predicting cellular mechanisms of action, herbal authentications, and bioactive identification. Additionally, we further compared this novel TCRP technology with highperformance liquid chromatography (HPLC)-based methods for herbal origin-tracing authentication and identification of bioactive ingredients. Together, our findings suggest that using TCRP as an alternative to existing spectroscopic techniques can allow us to analyze natural products in a more convenient and physiologically relevant manner.

H

istorically, most of the therapeutic agents used to combat human diseases have been derived from natural products. Indeed, natural products or their derivatives have always attracted considerable attention from pharmaceutical industrials because of their biological integrity, structural diversity, and low production costs. However, rapid developments in combinatorial chemistry have led to a decline in the use of traditional natural medicines as a major source of drug discovery in the past few years.1
3 Factors that have contributed to this decline are (i) difficulty in assessing the biological properties of natural products by high-throughput screening, (ii) the high costs and tedious nature of conventional methods for identifying the active r 2011 American Chemical Society

ingredients in natural products, and (iii) the lack of standardization and quality control for mass-manufactured natural products.1
3 However, the current industry model of hit-to-lead discovery has not been as successful as promised. The use of natural products as biologically validated starting points may be set for a renaissance. Therefore, new approaches for informationrich biological screening and characterization of natural products remain an unmet need to further our understanding of traditional Received: March 9, 2011 Accepted: July 27, 2011 Published: July 27, 2011 6518

dx.doi.org/10.1021/ac201670e | Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry herbal medicine and to improve and develop novel therapeutic treatments based on natural products.3
6 Traditional Chinese medicines (TCMs) have been invaluable as a source of therapeutic agents in China and other East Asian countries for centuries. Numerous widely used drugs derived from TCMs have been developed as anti-infection, anti-inflammation, and anticancer agents, such as astragalus, vinblastine, vincristine, and colchicine.7
9 Importantly, TCMs are largely based on clinical practices with unique and complicated theoretical systems, opening a different venue to the modern hit-to-lead drug discovery. However, the development of TCMs for modern medicines has been hampered to a large extent by the lack of compatible analytical tools. Most herbal medicines can be regarded as efficient and safe because of their long history in traditional therapy, but concerns regarding the poor quality control of herbal products are apparent. In particular, the lack of consistency in the levels of active ingredients contributes to the variation in efficacy.1,2,5,9 Although spectroscopic methods for quality control of herbal medicines are useful, they can only address the integrated sameness and stability of the sample.2,4,10 The complex relationships between the chemical fingerprint and the efficacy of herbal medicines have not been taken into account. In this work, we introduce a novel cell-based screening assay for integrating biological mechanisms of action with herbal authentication and quality control. We previously described this cell-based screening assay for profiling and identifying pattern-based mechanisms. This technique is based on continuous monitoring of cellular impedance in real time, which produces specific time/dosedependent cell response profiles (TCRPs) upon treatment with biologically active compounds.11,12 We propose that TCRPs can be predictive of an agent’s mechanism of action because compounds with similar biological activities often generate similar TCRPs. Herein, we have extended these studies and have focused on signature TCRPs induced by TCMs. Furthermore, we used a variety of well-studied molecular and cell biological tools to characterize these featured patterns, thus demonstrating the utility of this profiling approach for the classification of natural products, maintaining quality control, and facilitating active ingredients identification. In comparison with high-performance liquid chromatography (HPLC)-based methods, our work suggests that this label-free, noninvasive TCRP methodology could address most of the above-described challenges of studying TCMs, including global biological screening, predicting mechanisms of action, and the authentication and identification of bioactive ingredients in a more convenient and physiologically relevant manner.

’ EXPERIMENTAL SECTION Cell Culture. All the adherent cell lines [human lung epithelial cells (A549); rat basophilic leukemia cells (RBL-2H3); liver hepatocellular cells (HepG2); human cervical epithelial cell (Hela); kidney fibroblasts (COS1); human breast epithelial cells (MCF7-adr); embryonic fibroblast cell (NIH-3T3)] were originally purchased from ATCC (American Type Culture Collection). Cells were cultured in a humidified incubator at 37 C with 5% CO2 in accordance to optimal media and growth conditions specified by ATCC. Natural Herbal Medicines and Compounds. All natural herbal materials used in this study were purchased from four traditional medicinal plant dealers: Bozhou-Anhui, Chengdu-

ARTICLE

Sichuan, Pan’an-Zhejiang, and Yuzhou-Henan. All herbs were organoleptically, macroscopically inspected by three TCM experts, J. W. Lian (Zhejiang University of Traditional Chinese Medicine, Hangzhou, China), J. P. Yu (Zhejiang Institute for Food and Drug Control), and C. K. Ke (Hangzhou WenSongTang Herbal Pharmaceuticals Inc., Hangzhou, China), who analyzed the appearance, shape, color, texture, odor, and other typical characteristics of these herbs according to the Pharmacopoeia of the People’s Republic of China (PPRC).13 All diverse sources of wild Cordyceps sinensis, Hirsutella sinensis, and other five species of Cordyceps (Cordyceps militaris, Cordyceps liangshanensis, Cordyceps hawkesii gray, Mortierella hepiali, Paecilomyces hepiali, referred as fake C. sinensis as negative controls) were kindly provided and identified by N. Y. Shen (Qinghai Academy of Animal and Veterinary, Qinghai, China). All Cordyceps and Radix Stephaniae tetrandra, Herba Lysimachiae, Rhizoma Cynanchi stauntonii, which are methodologically discussed in this work, were further analyzed and traced by HPLC. To isolate various ingredients of C. sinensis, dried sample (1 g) was pulverized and extracted three times with petroleum ether (80:1, v/w). Petroleum ether extract was evaporated to dryness under vacuum to obtain the C4 component. The residue was sequentially extracted three times with acetone, ethyl acetate, n-butyl alcohol, and ethanol (80:1, v/w), and the extracts were evaporated to dryness to obtain components C2, C6, C1, and C5, respectively. Distilled water was added to the residue before sonication for 30 min at 30 C, and the water solution was evaporated to obtain the C3 component. The small molecular compounds (docetaxel, colchicine, chelidonine, paclitaxol, vincristine, and vinblastine) with antimitotic action as previously described12 were originally purchased from MicroSource Discovery (New Milford, CT). Time/Dose-Dependent Cell Response Profiles (TCRPs) Detected by the xCELLigence System. The detailed procedures for the xCELLigence system (Roche Applied Science, Basel, Switzerland), also known as the real-time cell electronic sensing system (RT-CES, ACEA Biosciences, San Diego), have been previously described.11,12,14 Briefly, to measure TCRPs, 50 μL of medium was added to 96-well E-plates to obtain background readings followed by the addition of 100 μL of cell suspension. The E-plates containing the indicated initial number of cells were allowed to incubate at room temperature for 30 min before being placed onto the reader in the incubator for continuous recording of impedance as reflected by cell index (CI). The cells were allowed to attach and grow for 18
24 h to reach a stable baseline before the addition of the indicated extracts of herbal medicines or compounds. To prepare for observation of the TCRPs generated by natural herbal medicines, 10 μL of sample was added to the E-plate containing the cells. The cells were monitored every 2 min for 1
2 h after treatment to capture the short-term cell response and every 15 min for 48
96 h to capture the long-term TCRPs. Cell Proliferation, Apoptosis, and Mitotic Assays. Cell proliferation was assayed using the bromodeoxyuridine (BrdU) detection kit (Roche Diagnostics, Indianapolis, IN) following the manufacturer’s suggested protocol. Apoptosis was evaluated using a DNA fragmentation assay kit (BD Biosciences, Franklin Lakes, NJ) according to the user manual. Cell cycle analysis was achieved using a flow cytometer (Beckman-Coulter BFC500 MCL, Brea, CA). Cells were thoroughly suspended (106
107 cells) in 0.5 mL of PBS, fixed with 70% ethanol for 2 h, and centrifuged. The fixed cells were then suspended in propidium 6519

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry iodide (PI)/TritonX-100 staining solution with RNase A. Cell fluorescence was analyzed with a flow cytometer. The pulsewidth
pulse-area signal was evaluated to discriminate between G2/M cells and cell doublets, with the latter gated out. The data were analyzed with deconvolution of DNA content histograms. High-Performance Liquid Chromatography. HPLC analyses of R. S. tetrandra, H. Lysimachiae, R. C. stauntonii, C. sinensis, and other Cordyceps were performed with the Agilent XDB-ODS column (250 mm  4.6 mm, 5 μm). The mobile phase was acetonitrile with 0.1% phosphoric acid, and the gradient elution mode was applied. The flow rate was set at 1 mL/min, and the column temperature was kept at 25 C. Statistical Analysis. A B-spine-based semiparametric nonlinear mixed effects model was proposed for TCRPs, which allowed us to align the raw data set along the axis of time. The maximum likelihood method was used for parameter estimation. As for TCRPs of wild C. sinensis samples, 95% confidence intervals were computed using the bootstrap based on the mixed effects model. In particular, this model could be extended to a multidimensional data set, which could allow the analysis of time/dosage-dependent cell response profiles. A statistical method for the analysis of TCRP-based authentication and classification has been developed and described previously in detail.15 The chromatogram of C. sinensis was handled with the method of dynamic time warping (DTW). DTW distance was calculated to measure the similarity between two sequences which vary in time. The neighbor tree was drawn based upon the pairwise DTW distance matrix by Phylip 3.67.

’ RESULTS Global Monitoring and Functional Analysis of Cell Kinetic Responses to Natural Herbs. Real-time impedance-based

monitoring of cellular responses to biologically active compounds produces TCRPs, which can be predictive of a compound’s biological mechanism of action in cells.11,12 We first implemented this noninvasive, cell-based methodology for global monitoring of cell kinetic responses to a number of traditional Chinese natural products, which are referred as the origin-tracing (“dao-di”) herbal medicines and are historically well-documented in ancient Chinese pharmacological literature. A representative panel of typical cell kinetic curves produced in lung epithelial cells (A549) is shown in Figure 1A. Featured TCRPs mirror the alterations of cell conditions, including cellular morphology, adhesion, and growth, which are modulated by cellular interactions with a variety of treatments.11,12,16,17 The tested herbal products were hierarchically clustered on the basis of the similarity of the TCRPs using statistical analysis. By comparison with previously well-studied cellular TCRPs with known mechanistic chemical compounds,11,12,16
20 we identified seven functional subclusters of actions for 146 herbal species, including the impact of protein synthesis, DNA-damaging agents, antimitotic agents, cytotoxicity agents, promotion of cell growth, short-term cell stress, and calcium-release events (Figure 1B and Supporting Information Figure 1). Next, representative TCMinduced TCRPs were chosen for further validation of the predicted TCRP-based mechanisms of action by the application of independent assays (Figure 1C
E). Similar to previous results,12,17 we observed that different parts of the TCRP curve produced by the administration of R. S. tetrandra (Supporting Information Figure 2A) to A549 cells were consistent with the cellular morphological alterations (Figure 1C). The TCRP

ARTICLE

produced by adding H. Lysimachiae (Supporting Information Figure 2B) to cells suggest that H. Lysimachiae functions in promoting cell proliferation, which was further confirmed by BrdU incorporation assay as shown in Figure 1D. In another example, the TCRP suggested that R. C. stauntonii (Supporting Information Figure 2C) induced cell apoptosis, which was confirmed in an accompanying DNA fragmentation assay (Figure 1E). On the basis of these findings, we proposed that natural herbal medicines may have unique TCRP-based profiles, when systematically produced in diverse cell lines or each herbal medicine at a wide range of concentrations, providing a potential rich information tool for herbal authentication, prediction of mechanism of action, and identification of bioactive ingredients in a more convenient and physiologically relevant manner. Feature TCRP Profiles for Herbal Authentication. The lack of quality control and manufacturing standards for natural products is a constant concern that has impeded the use of herbal medicines in the modern pharmaceutical industry.1,2,10 As such, we first explored the application of TCRP-based assays for herbal authentication. Herein, a well-known and highly valued TCM, C. sinensis, is discussed as a methodological example. In recent years, the high demand and high price of wild C. sinensis have resulted in the excessive exploitation of C. sinensis, which has resulted in severe damage to its native environment. As an alternative, mycelium production of C. sinensis by submerged fermentation could satisfy the needs of consumers for wild C. sinensis. However, there is no reliable standard for C. sinensis mycelium.21
24 In this study, 17 wild C. sinensis samples collected from different regions in the Qinghai
Tibetan plateau produced the highly similar TCRPs (Figure 2A). Compared with wild C. sinensis, we found that 29 lab-cultivated C. sinensis mycelia exhibit comparable TCRPs with significant variability (data not shown). We next asked whether this featured TCRP for the wild representative specimen could be used as a population average standard for C. sinensis authentication. Given the random variations that occur in cell-based experiments, a statistical model was developed so that the TCRP-based herbal authentication could be performed in a reliable and practical manner (Figure 2B). A B-spline-based semiparametric nonlinear mixed effects model was used to fit the TCRPs under the assumption that the underlying shape functions of the TCRPs are relatively smooth. According to the maximum log-likelihood method, the average (solid line in black) and 95% confidence interval (dotted lines in black) of the predicted pattern could be obtained on the basis of the 17 wild C. sinensis-derived TCRPs (Figure 2B). Another five species of Cordyceps (fake C. sinensis), which have almost the same appearance as C. sinensis, showed significant differences from the standard TCRPs. Those evidences demonstrated that, by comparing a TCRP with the population average, a variety of samples, including wild C. sinensis and other five fake C. sinensis, can be authenticated semiquantitatively using TCRP technology integrated with statistical analysis (Figure 2C). Next, we compared the TCRP-based assay with an existing spectroscopic method (HPLC) for herbal authentication. C. sinensis is currently authenticated by appearance and HPLC using adenosine (or 30 -deoxyadenosine) as a standard, according to the Pharmacopoeia of the People’s Republic of China (2005 version).13 The same tested samples in Figure 2C were processed using HPLC. As described in Figure 2D, it has been obviously shown that all samples (four wild and five fake C. sinensis) have the adenosine content, suggesting determination of adenosine in C. sinensis using HPLC is insufficient to authenticate C. sinensis. 6520

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry

ARTICLE

Figure 1. Global analysis of kinetic cellular responses to natural products. (A) Representative examples of kinetic cellular responses to herbal medicines. The human epithelial alveolar cell line A549 was seeded at a density of 4000 cells/well containing 10% fetal bovine serum (FBS) in medium. At the indicated time points (arrow), the medium was replaced with a serum-free medium that contained the indicated aqueous extracts, and the cellular responses were monitored using the xCELLigence system. An equal amount of distilled H2O was added to the FBS-containing or serum-free medium as a control. (B) Agglomerative hierarchical clustering analysis of kinetic cellular dynamic curves in the long-term (left panel) and short-term (right panel) responses. The cell index values were colorized according to the indicated scale and plotted as a function of time, with time zero at the left. As indicated, the herbs that elicited similar responses were subgrouped together. (C) A typical dynamic cell response curve is indicative and predictive for cellular morphological changes following the addition of herbs. A549 cells were seeded onto the electronic sensor device (E-plate, xCELLigence) and regular culture dishes in a simultaneous test. Cells were treated with R. S. tetrandra at the indicated time points (arrow), and the cellular morphology was visualized and photographed using an electronic microscope before treatment (i) and at 2 (ii), 10 (iii), and 24 h (iv) post-treatment, scale bar = 20 μm. (D) A typical dynamic cell response curve is indicative and predictive for the promotion of cell growth following the addition of herbs. As described in panel C, A549 cells in serum-free medium that were exposed to H. Lysimachiae were analyzed at the indicated time point (20 h post-treatment) with a BrdU assay, and BrdU+ cells were visualized by fluorescence microscopy, scale bar = 300 μm. (E) A typical dynamic cell response curve is indicative and predictive for the induction of cell apoptosis following the addition of herbs. As described in panel C, A549 cells in serum-free medium that were exposed to R. C. stauntonii were analyzed at the indicated time point (24 h post-treatment) with a TUNEL DNA fragmentation assay. Apoptotic cells (green) were visualized by fluorescence microscopy, scale bar = 300 μm.

Although current spectroscopic fingerprinting offers the possibility to determine all chemical compositions in C. sinensis, our result suggests that TCRP technology, as an alternative to chemical methods, provides a more convenient way for herbal authentications. “Dao-di” in the concept of TCM refers to herbal medicines of particular species from the best regional origin with the best efficacy, which is broadly discussed for authentication of herbal medicines. Obviously, this ancient notion of “dao-di”, to some extent, has centered on the effect of cultivating environmental factors on herbal products. Nevertheless, until now, limited methods have been available to evaluate “dao-di”. In this context, we expand the application of the TCRP-based method for herbal classification. Seventeen wild C. sinensis-derived TCRPs and HPLC profiles were produced (Supporting Information Figure 3, parts A and B), and all 17 tested samples exhibit similar profiles, suggesting the high degree of comparability and

consistency between the TCRP and HPLC-based methods. The two profiles were further analyzed with a normalized phylogenetic trees using DTW distance estimation (Figure 2, parts E and F). Overall, our findings suggest that samples of C. sinensis from Yushu
Nakqu (historically documented as the best quality zone) exhibit unique clustering patterns (high identical or close relative) in both methods. Interestingly, it is noted that C. sinensis from Dege, which is not geographically located in the traditional best quality zone, was clustered into the Yushu
Nakqu subgroup using HPLC-based phylogenetic analysis, compared with TCRP-based data. In this case, we suggest that the TCRP-based assay exhibits favorable value for source-tracing herbal authentication and, more important, their potential for predicting the mechanism of action of herbal medicines. TCRP-Based Identification of the Mechanism of Action in Herbal Medicines. We next examined the ability of the 6521

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry

ARTICLE

Figure 2. TCRP-based authentication of herbal medicines. (A) Cell response phenotypic profiles of wild C. sinensis collected from 17 regions in the neighboring Qinghai
Tibetan plateaus. Mast cells (RBL-2H3) were seeded onto the E-plate and treated with different C. sinensis samples in serum-free medium at the indicated time points and continuously monitored for 50 h. (B) The TCRP-based population average model for herbal C. sinensis. A semiparametric nonlinear mixed effects model was used for quantitative authentication. The average predicted CI (solid line) and its 95% confidence interval (dotted) are shown in black. A nonlinear mixed effect model was applied to the original CI. After smoothing and alignment at the axis of time, the predicted CI with time (h) was obtained for each sample. The average predicted CI can also be obtained by including the fixed effect and excluding the random effect of the mixed effect model. The 95% confidence interval was obtained by a bootstrap method. Predicted changes of CI with time (h) for the 17 wild C. sinensis samples are shown in the corresponding color. (C) Examples of the use of the TCRP-based population average model for C. sinensis herbal authentication. The maximum log-likelihood fit of the TCRP to the mixed effects model derived from wild C. sinensis-derived TCRPs (A) was used as a score (P-value) for testing samples, as indicated. The P-value shown corresponds to a null distribution generated by 10 000 simulation trials of the TCRP-based population average model (B). The C. sinensis-derived TCRPs conformed to this inferred mixed model (P > 0.05). In contrast, other Cordyceps could be easily distinguished using this approach (P < 10
4). This quantitative model can identify lab-derived cultivated C. sinensis mycelia (labeled as H. sinensis 1 and 2), whose TCRPs conformed to this mixed effects model (P > 0.05). The details for this statistical approach have been previously developed and described (ref 15). (D) HPLC profiles for two wild C. sinensis, two C. sinensis mycelia, and five other fake C. sinensis (S1, M. hepiali; S2, Nakqu; S3, Yushu; S4, H. sinensis 1; S5, H. sinensis 2; S6, C. militaris; S7, P. hepiali; S8, C. hawkesii; S9, C. liangshanensis). The star shows the adenosine peak. (E and F) Phylogenetic tree analysis of the DTW distance estimation for the 17 wild C. sinensis using TCRPs (E) and HPLC profiles (F). Distances were estimated using the curve-clustering algorithm. The DTW distance in an n-space was used as the distance between signal curves having n data points. Hierarchical cluster analysis was performed, and the dendrogram was plotted in accordance with distance. The scale bar indicates the DTW distance. The best quality of C. sinensis in Yushu
Nakqu was clustered as labeled by the broken line, except the sample from Dege indicated by an asterisk.

TCRP-based methodology to identify the mechanism of action in herbal medicines. C. sinensis has been widely believed to have many beneficial properties for centuries; however, the cellular mechanism responsible for its efficacy remains largely elusive.21,25 As shown in Supporting Information Figure 1, we have classified C. sinensis into the antimitosis cluster on the basis of its similar kinetic cellular response with small molecules with known antimitotic actions. To confirm this prediction, colchicine, a well-known antimitotic drug, was tested using the TCRP

assay and produced a similar pattern as C. sinensis (Figure 3, parts A and B). To further probe this biological event, cells treated with C. sinensis were analyzed with a cell cycle assay. This assay showed that C. sinensis induces growth inhibition and G2/M-phase arrest of the cells compared with the control treatment, which is consistent with our antimitotic prediction (Figure 3C). To check whether the antimitotic activity of C. sinensis is unique to our tested cells (A549), we treated several other cell lines with C. sinensis (Supporting Information Figure 4). Overall, the 6522

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry

ARTICLE

Figure 3. TCRP-based predictions of the mechanism of action of C. sinensis. (A and B) Comparison of TCRPs of the crude aqueous extract of C. sinensis (A) and the compound colchicine (B). (C) Cell cycle arrest at mitosis in response to C. sinensis (6 mg/mL). Treated or untreated (control) A549 cells were suspended and analyzed with flow cytometry (Beckman-Coulter). Cell cycle stages were defined by measuring the nuclear DNA content using the total intensity of a DNA binding dye (PI). Treatment with C. sinensis resulted in a significant increase in the number of cells at G2/M, compared with the control (// P < 0.01, n = 3).

TCRPs of all cell lines treated with C. sinensis, not including the initial 2 h response, exhibit comparable patterns: a steady decline in the CI, followed by an increase approximately 10
20 h after treatment, which is similar to typical cell cycle arrest patterns produced by a variety of well-known antimitotic agents (Supporting Information Figure 5). Together, these findings suggest that the TCRP could provide important information for identifying herbal bioactivities in herbal mixtures. TCRP-Based Identification of Herbal Active Ingredients. Chromatographic methods are commonly used for bioactive analysis in herbal mixtures. We next asked whether this TCRP is beneficial in isolating herbal active ingredients using chemical separation techniques. In recent studies, FTY720 (2-amino2-[2-(4-octylphenyl)ethyl]-1,3-propanediol hydrochloride), derived from Cordyceps sinclairii, has been widely reported as a newly synthesized immunosuppressive agent.26
28 To test the feasibility of TCRP-based separation, we first produced the TCRPs of FTY720 in immune cells (RBL-2H3), which appear similar to the ones of C. sinensis (Figure 4A). This interesting observation prompted us to isolate active ingredients in C. sinensis on the basis of this featured TCRP. The wild C. sinensis were serially extracted with organic solvents and separated into six fractions, among which the fraction 3 (C3) produces the similar TCRP with wild C. sinensis in RBL-2H3 cells (Figure 4B).

Moreover, the C3 was purified using membrane size exclusion filtration into three subfractions, designated as S1 (molecular weight over 2000 Da), S2 (molecular weight ranges between 500 and 2000 Da), and S3 (molecular weight less than 500 Da). As shown in Figure 4C, TCRP of subfraction S3 appears to be similar to fraction C3, suggesting the major active ingredient in the subfraction with molecular mass under 500 Da (Figure 4C). This finding is consistent with HPLC analysis, which detected the occurrence of a single overlapping peak in the purified subfraction S3 and FTY720 upon the retention time (Figure 4D). Despite the fact that further investigations are needed to determine the chemical composition of active ingredients in C. sinensis, the case presented here suggests that TCRP provides a unique tool for identifying bioactive components in natural products

’ DISCUSSION In this report, we characterized impedance time/dosagedependent cell response profiling technology for biological analysis of natural products. We found that the cell kinetic responses upon treatment with various TCMs can generate biologically specific TCRPs that can help predict the mechanism of action in these herbal medicines. Using a valued TCM, C. sinensis, as an example, we further discussed the use of TCRPs 6523

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry

ARTICLE

Figure 4. TCRP-based identification of bioactive ingredients of C. sinensis. (A) Immune cells (RBL-2H3) were treated with FTY720 and C. sinensis at the indicated concentration for continuous monitoring of long-term cellular responses. The inserted image shows the chemical structure of FTY720, a synthetic myriocin analogue that was originally isolated from a fungal metabolite of Cordyceps. (B) TCRPs of six fractions (C1
C6) of C. sinensis using organic solvent extractions (left, combined TCRPs; right, individual TCRP of each fraction with the indication). The fraction C3 produces the similar TCRP as wild C. sinensis, indicated by asterisks in the right panel. (C) TCRPs of three subfractions (S1
S3) using size exclusion filtration (left, combined TCRPs and fraction C3 as control; right, individual TCRP of each subfraction with the indication). The subfraction S3 (molecular mass less than 500 Da) has the same cellular response as the fraction C3, indicated by asterisks in the right panel. (D) HPLC profiles for wild C. sinensis, purified subfraction S3, and FTY720; the results show overlapping chromatographic separations upon the retention time, indicated by an asterisk in the panel.

for herbal authentication, identification, and isolation of bioactivities. Herein, our observation demonstrates that TCRP technology is a unique biological analytic tool for natural products that compliments existing chemical tools that lack biological integrations for the natural products. In recent years, a considerable amount of attention has been given to various spectroscopic-based chemical profiles, which have been used as fingerprinting tools for authenticating and assessing the quality of natural products. The standardized HPLC fingerprints show high stability and reproducibility and thus can be used effectively for the screening analysis or quality control of natural products. However, chemical measurements are often unable to detect the biological integrity of herbal products and underestimate the complexities of the interactions of purified components.6,10,29,30 A TCRP reflects the dynamic changes in three cellular parameters: cell number, morphology, and adhesive quality. These parameters are intricately associated

with the regulation of cell physiology and are therefore amenable to modulation by biologically active ingredients in natural products in a prompt and unbiased manner.16 In this work, we compared TCRP technology with HPLC-based method for herbal authentication. Our results suggest considerable comparability and consistency for both methods. In some cases, the featured TCRP appears to have a unique advantage in the authentication or source-tracing the geographical origins of herbal medicines. More important, this novel TCRP-based signature integrates bioactive properties with herbal authentication, which is limited in spectroscopic-based chemical profiles. The bioactive correlation seems to be the most important aspect for the quality control of herbal medicines. As shown in this work, we further discussed a novel TCRP-based signature method for the quality control and meaningful bioactive identification of herbal products. Our results suggest that the TCRP-based method, as an alternative to chemical fingerprinting, could provide increased efficiency in quality control and authentication of herbal medicines. 6524

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry In this work, we only tested cellular responses to very limited cell types (mostly in lung epithelial cell line) for the methodological purpose. In fact, TCRPs generated by natural products tested across multiple functional/tissue-specific cell panels would allow a broader coverage of biological screening information with more specialized predictive values. Moreover, TCRP-based assays monitor the effect that a compound has on cells and are not limited to a single cell line with simple phenotypic end points. More important, genetically modified cells, stress-induced cells, or even disease-associated induced pluripotent stem cell (iPS)derived lines can also be used to meet specific requirements of drug discovery, which makes this assay more physiologically relevant. In recent years, there has been considerable interest in the use of this TCRP-based technique for the construction of cell-based screening models in drug discovery, including models of RTKs,16,18 GPCRs17,20 and NK cell-mediated cytotoxicity.31 It is likely that the application of this technique will provide better analytical tools for interpreting cellular information obtained from the TCRP-based model. Cell-based TCRP technology in which multiple physiological changes are evaluated would provide richer information in biological properties for natural products than some spectroscopic methods; however, the inherent variability and complexity of the cells should be taken into full consideration with regard to experimental design, implementation, and analysis.32 To address this technique challenging issue in TCRP-based assay, herein we introduced statistical model of TCRPs using a nonlinear mixed effect model, which allows for derivation of cell responses in terms of CI values to herbal treatments. In fact, TCRP measurement due to the additional dimension of time/dosage during monitoring cell responses, offers the possibility to construct multiple variables in mathematical simulation for kinetic measurement of cell phenotypic changes. In the near future, we reason that more statistical analyses would shed light on herbal identification and biological prediction in the practical manner. In summary, in this study, the TCRP-based technique is used for the first time to perform biological assays on natural herbal products based on its utility in various applications, including bioactive screening, quality control and authentication of natural products, and identification of active ingredients. The technique provides a unique window for dissecting and understanding the mechanisms of action of natural products. It is our belief that the TCRP technique, together with other emerging advances in chemical and biological technologies, opens a promising new avenue for the renaissance of natural herbal medicines.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. Phone: (86.571) 8820.8713. Fax: (86.571) 8820.8583.

’ ACKNOWLEDGMENT We thank Drs. Q. M. Xie, X. M. Wu, and H. Hu (Zhejiang University School of Medicine, Hangzhou, China) for critical

ARTICLE

review of the manuscript. This work was supported by key program of Zhejiang Provincial TCM Administration N20090004 to Y.H.K., National Natural Science Foundation of China 30900849 to Y.H.K., and 81001617 to H.Y.F.

’ REFERENCES (1) Li, J. W.; Vederas, J. C. Science 2009, 325, 161–165. (2) Harvey, A. L. Curr. Opin. Chem. Biol. 2007, 11, 480–484. (3) Koehn, F. E.; Carter, G. T. Nat. Rev. Drug Discovery 2005, 4, 206–220. (4) Haustedt, L. O.; Mang, C.; Siems, K.; Schiewe, H. Curr. Opin. Drug Discovery Dev. 2006, 9, 445–462. (5) Clardy, J.; Walsh, C. Nature 2004, 432, 829–837. (6) Eldridge, G. R.; Vervoort, H. C.; Lee, C. M.; Cremin, P. A.; Williams, C. T.; Hart, S. M.; Goering, M. G.; O’Neil-Johnson, M.; Zeng, L. Anal. Chem. 2002, 74, 3963–3971. (7) Zhang, X. W.; Yan, X. J.; Zhou, Z. R.; Yang, F. F.; Wu, Z. Y.; Sun, H. B.; Liang, W. X.; Song, A. X.; Lallemand-Breitenbach, V.; Jeanne, M.; Zhang, Q. Y.; Yang, H. Y.; Huang, Q. H.; Zhou, G. B.; Tong, J. H.; Zhang, Y.; Wu, J. H.; Hu, H. Y.; de The, H.; Chen, S. J.; Chen, Z. Science 2010, 328, 240–243. (8) Wang, L.; Zhou, G. B.; Liu, P.; Song, J. H.; Liang, Y.; Yan, X. J.; Xu, F.; Wang, B. S.; Mao, J. H.; Shen, Z. X.; Chen, S. J.; Chen, Z. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 4826–4831. (9) Wang, M. W.; Hao, X.; Chen, K. Philos. Trans. R. Soc. London, Ser. B 2007, 362, 1093–1105. (10) Koehn, F. E. Prog. Drug Res. 2008, 65 (175), 177–210. (11) Ke, N.; Xi, B.; Ye, P.; Xu, W.; Zheng, M.; Mao, L.; Wu, M. J.; Zhu, J.; Wu, J.; Zhang, W.; Zhang, J.; Irelan, J.; Wang, X.; Xu, X.; Abassi, Y. A. Anal. Chem. 2010, 82, 6495–6503. (12) Abassi, Y. A.; Xi, B.; Zhang, W.; Ye, P.; Kirstein, S. L.; Gaylord, M. R.; Feinstein, S. C.; Wang, X.; Xu, X. Chem. Biol. 2009, 16, 712–723. (13) Kuang, L.; Zhang, K. Pharmacopoeia of The People’s Republic Of China; People’s Medical Publishing: Bejing, China, 2005. (14) Ke, Y.; Zhang, E. E.; Hagihara, K.; Wu, D.; Pang, Y.; Klein, R.; Curran, T.; Ranscht, B.; Feng, G. S. Mol. Cell. Biol. 2007, 27, 6706–6717. (15) Fu, W.; Fu, H.; Sun, M.; Mou, X.; Jin, L.; Ke, Y.; Guo, W. Unpublished work, 2010. (16) Xi, B.; Yu, N.; Wang, X.; Xu, X.; Abassi, Y. A. Biotechnol. J. 2008, 3, 484–495. (17) Yu, N.; Atienza, J. M.; Bernard, J.; Blanc, S.; Zhu, J.; Wang, X.; Xu, X.; Abassi, Y. A. Anal. Chem. 2006, 78, 35–43. (18) Atienza, J. M.; Yu, N.; Wang, X.; Xu, X.; Abassi, Y. J. Biomol. Screening 2006, 11, 634–643. (19) Atienza, J. M.; Zhu, J.; Wang, X.; Xu, X.; Abassi, Y. J. Biomol. Screening 2005, 10, 795–805. (20) Abassi, Y. A.; Jackson, J. A.; Zhu, J.; O’Connell, J.; Wang, X.; Xu, X. J. Immunol. Methods 2004, 292, 195–205. (21) Zhou, X.; Gong, Z.; Su, Y.; Lin, J.; Tang, K. J. Pharm. Pharmacol. 2009, 61, 279–291. (22) Wang, Y.; Wang, M.; Ling, Y.; Fan, W.; Wang, Y.; Yin, H. Am. J. Chin. Med. 2009, 37, 977–989. (23) Xie, C. Y.; Gu, Z. X.; Fan, G. J.; Gu, F. R.; Han, Y. B.; Chen, Z. G. Appl. Biochem. Biotechnol. 2009, 158, 483–492. (24) Li, S. P.; Yang, F. Q.; Tsim, K. W. J. Pharm. Biomed. Anal. 2006, 41, 1571–1584. (25) Zhou, X.; Luo, L.; Dressel, W.; Shadier, G.; Krumbiegel, D.; Schmidtke, P.; Zepp, F.; Meyer, C. U. Am. J. Chin. Med. 2008, 36, 967–980. (26) Liu, Q.; Alinari, L.; Chen, C. S.; Yan, F.; Dalton, J. T.; Lapalombella, R.; Zhang, X.; Mani, R.; Lin, T.; Byrd, J. C.; Baiocchi, R. A.; Muthusamy, N. Clin. Cancer Res. 2010, 16, 3182–3192. (27) Wolf, A. M.; Eller, K.; Zeiser, R.; Durr, C.; Gerlach, U. V.; Sixt, M.; Markut, L.; Gastl, G.; Rosenkranz, A. R.; Wolf, D. J. Immunol. 2009, 183, 3751–3760. (28) Aktas, O.; Kury, P.; Kieseier, B.; Hartung, H. P. Nat. Rev. Neurol. 2010, 6, 373–382. 6525

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526

Analytical Chemistry

ARTICLE

(29) Mishra, K. P.; Ganju, L.; Sairam, M.; Banerjee, P. K.; Sawhney, R. C. Biomed. Pharmacother. 2008, 62, 94–98. (30) Harrigan, G. G.; Goetz, G. H. Comb. Chem. High Throughput Screening 2005, 8, 529–534. (31) Zhu, J.; Wang, X.; Xu, X.; Abassi, Y. A. J. Immunol. Methods 2006, 309, 25–33. (32) Inglese, J.; Auld, D. S.; Jadhav, A.; Johnson, R. L.; Simeonov, A.; Yasgar, A.; Zheng, W.; Austin, C. P. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 11473–11478.

6526

dx.doi.org/10.1021/ac201670e |Anal. Chem. 2011, 83, 6518–6526