Rim

YYa

b

h

•

•

•

•

ARRAA

KNPRTT

(

0h

Analytica Chimica Acta 777 (2013) 49– 56

Contents lists available at SciVerse ScienceDirect

Analytica Chimica Acta

j ourna l ho mepage: www.elsev ier .com/ locate /aca

apid screening of bioactive compounds from natural products byntegrating 5-channel parallel chromatography coupled with on-line

ass spectrometry and microplate based assays

ufeng Zhanga, Shun Xiaoa, Lijuan Suna, Zhiwei Gea, Fengkai Fangb, Wen Zhangb,∗,i Wanga,∗∗, Yiyu Chenga

College of Pharmaceutical Sciences, Zhejiang University, Hangzhou, Zhejiang 310058, ChinaResearch Center for Marine Drugs, School of Pharmacy, Second Military Medical University, 325 Guo-He Road, Shanghai 200433, China

i g h l i g h t s

A 5-channel parallel chromatogra-phy coupled with mass spectrometrymethod was established to screenactive compounds.The proposed approach has satisfiedsensitivity and stability.Three case studies of this approachidentified several active compounds.Agelasine B and (−)-agelasine D frommarine sponge were found to haveanti-tumor effect.

g r a p h i c a l a b s t r a c t

a r t i c l e i n f o

rticle history:eceived 28 December 2012eceived in revised form 8 March 2013ccepted 12 March 2013vailable online 21 March 2013

eywords:atural productsarallel chromatography

a b s t r a c t

A high throughput method was developed for rapid screening and identification of bioactive compoundsfrom traditional Chinese medicine, marine products and other natural products. The system, integratedwith five-channel chromatographic separation and dual UV–MS detection, is compatible with in vitro 96-well microplate based bioassays. The stability and applicability of the proposed method was validated bytesting radical scavenging capability of a mixture of seven known compounds (rutin, dihydroquercetin,salvianolic acid A, salvianolic acid B, glycyrrhizic acid, rubescensin A and tangeretin). Moreover, theproposed method was successfully applied to the crude extracts of traditional Chinese medicine and amarine sponge from which 12 bioactive compounds were screened and characterized based on their anti-

apid identificationraditional Chinese medicineargeted screening

oxidative or anti-tumor activities. In particular, two diterpenoid derivatives, agelasine B and (−)-agelasineD, were identified for the first time as anti-tumor compounds from the sponge Agelas mauritiana, showinga considerable activity toward MCF-7 cells (IC50 values of 7.84 ± 0.65 and 10.48 ± 0.84 �M, respectively).Our findings suggested that the integrated system of 5-channel parallel chromatography coupled withon-line mass spectrometry and microplate based assays can be a versatile and high efficient approachfor the discovery of active compounds from natural products.

© 2013 Elsevier B.V. All rights reserved.

∗ Corresponding author.∗∗ Corresponding author. Tel.: +86 571 88208426; fax: +86 571 88208426.

E-mail addresses: wenzhang1968@163.com (W. Zhang), mysky@zju.edu.cnY. Wang).

003-2670/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.aca.2013.03.028

1. Introduction

Natural products have become one of the most important

resources for discovering bioactive compounds. The successfulstories of artemisinin, taxol, and digitalis encourage a contin-uous investigation on screening drug candidates from naturalsources [1]. The classical and commonly used procedures for

5 Chimic

ltoiwtcfrnstbradoecdhawesawbitfttvb

a[hunepfobdttwanrin(

2

2

diS

0 Y. Zhang et al. / Analytica

ead-compound discovery start with the isolation and struc-ural elucidation of pure compounds from the crude extractsf natural products, followed by subsequent bioassays of thesolates. Alternatively, a bioassay-guided isolation approach, in

hich a bio-screening was conducted prior to repeated isola-ion, has gained valuable findings in the discovery of bioactiveompounds since the past three decades. However, the processor isolating and identifying bioactive components from natu-al products is still time-consuming and labor-intensive. Hence,ew techniques and efficient approaches coping with the rapidcreening of active compounds from natural products are con-inuously in great demand. Several on-line separation followedy (bio)chemical detection methods based on liquid chromatog-aphy (LC) have been developed [2,3]. The on-line screening ofntioxidant by reacting post-column eluent with appropriate oxi-izing reagents is one of the most successful cases [4]. Otherwise,n-line enzyme assays were developed by mixing post-column elu-nts with buffer solutions of enzymes in micro-reactors where theorresponding bioactive compounds would be consumed or pro-uce ligand–enzyme complexes for detection. Such an approachas been proven to be successful in recognizing several types ofctive compounds from natural products, whilst avoiding tediousork of purification [5]. However, the application of on-line

nzyme assays are very limited because it relies on simple andtraightforward reactions between substrates with enzymes suchs acetylcholinesterase (AChE) and phosphodiesterase (PDE) [6,7],hile such in vitro assays rarely simulate real circumstances in

ody. Furthermore, the on-line separation cannot be coupled withn vitro cellular assays since two prerequisites have to be obligatedo fulfill the experiments: (1) reactions in micro-reactors should beast enough to allow the continuous post-column supply, and (2)he mobile-phase composition should be compatible with the reac-ion materials (e.g. organic solvent is harmful to enzyme and cellitalities) as well as the instrument (e.g. non-volatile salts shoulde avoided for mass spectrometry).

Recently, we made an endeavor in developing a cell-image basedpproach for screening active compounds from medicinal plants8]. The aim of the present study is to establish a new method forigh-efficient in vitro bio-screening of active compounds from nat-ral products, including traditional Chinese medicine and marineatural products. A five-column parallel separation platform wasstablished to isolate complex mixtures of natural products by higherformance liquid chromatography. One channel was employedor monitoring and obtaining chemical information of samples byn-line UV and mass spectrometry (MS). In order to obtain enoughiological replicates for bioassays of each fraction, four indepen-ent samples were collected by automated micro-plate collectorso generate reliable bioactivity results (one blank control andhree replicates or four replicates). The fractions in micro-platesere evaporated by centrifugal concentrating, and subsequently

nalyzed by micro-plate based enzyme and cellular assays. Thisewly developed method was validated in terms of precision andeproducibility using seven marker compounds, and applied todentify bioactive compounds from the extracts of traditional Chi-ese medicine (Donglingcao and Sangye) and one marine organismAgelas mauritiana).

. Materials and methods

.1. Material, chemicals and reagents

Donglingcao and Sangye were obtained from Hangzhou Tra-itional Chinese Herbal Medicine Factory (Hangzhou, China) and

dentified by Assoc. Prof. Chen Liurong (College of Pharmaceuticalciences, Zhejiang University) as Rabdosia rubescens (Hemsl.) Hara

a Acta 777 (2013) 49– 56

and Morus alba L., respectively. Voucher specimens were depositedin College of Pharmaceutical Sciences, Zhejiang University. Thesponge A. mauritiana was collected off the Coast of Xisha Islandin December 2011, at a depth of 20 m, and identified by comparingwith the animals previously collected at the same place and authen-ticated by Prof. Jin-He Li (Institute of Oceanology, Chinese Academyof Sciences). A voucher specimen (LG-1) was deposited in theSecond Military Medical University. Reference compounds, rutin(RT), dihydroquercetin (DQ), salvianolic acid A (SA), salvianolic acidB (SB), rosmarinic acid (RA), glycyrrhizic acid (GA), rubescensinA (RuA) and tangeretin (TG) were provided by Shanghai Win-herb Medical S & T Development Co., Ltd. (Shanghai, PR China).Chlorogenic acid (3-CQA) was purchased from the National Insti-tute for the Control of Pharmaceutical and Biological Products(Beijing, China). Thiazolyl blue tetrazolium bromide (MTT), 2,2-diphenyl-1-picrylhydrazyl (DPPH), fluorescein diacetate (FDA) anddimethyl sulfoxide (DMSO) were purchased from Sigma–Aldrich(USA). Vitamin C (Aladdin-Reagent Co., Ltd., Shanghai, China) anddoxorubicin hydrochloride (National Institute for Food and DrugControl, Beijing, China) were selected as positive controls. HPLCgrade acetonitrile (Fisher, NJ, USA), formic acid (MERDA, USA) andultrapure water (Milli-Q Plus, Millipore Co., Ltd., USA) were used inall experiments. All other chemicals and solvents are of analyticalgrade.

2.2. Preparation of sample extracts

Air-dried and finely powdered herbs of Donglingcao (250 g)were extracted twice with 1000 mL of 70% ethanol for 2 hunder reflux. The extracts were evaporated in vacuo to obtaina Donglingcao crude extract, which was subjected to ODS col-umn chromatography and subsequently eluted with 5% and 60%methanol in water. The 60% methanol eluents were collectedand concentrated to dryness to obtain a refined Donglingcaofraction (DLF). The Sangye fraction (SYF) was isolated as pre-viously described [9]. Briefly, after being extracted using 70%ethanol, the crude extract of Sangye was further purified on aD101 macroporous resin and a polyamide resin column chro-matography, respectively. The eluents of 50% ethanol from thepolyamide resin column was concentrated in a rotatory evapora-tor in vacuo to give SYF. Samples for parallel preparative separationand LC–MS analyses were prepared by dissolving DLF and SYF in50% methanol–water, respectively, to obtain a final concentrationof 25 mg mL−1.

The frozen sponge of A. mauritiana (490 g, wet weight) was cutinto small pieces and extracted ultrasonically with acetone (3×0.5 L) at room temperature for 30 min. The combined extracts weresuspended in H2O and extracted subsequently with Et2O and n-butanol. The butanol extract was concentrated in vacuo to give adark residue (LG-1-nB, 5.4 g). This residue was dissolved in MeOHto give a final concentration of 5 mg mL−1.

2.3. Hardware configuration

A photogram and a schematic representation of the five-channelparallel LC/UV/MS system used in this work are shown in Fig. 1.The system consists of the following: a WellChrom HPLC pump (K-1001, Knauer, Berlin, DE), a K-1500 solvent organizer (Knauer), afour-channel degasser (Knauer), a K-2600 UV detector (Knauer), anAS-3800 autosampler (Knauer), four D-14163 injectors (Knauer),four Foxy R1 fraction collectors (Teledyne Isco, Lincoln, NE) and aLCQ Deca XPplus ion trap mass spectrometer (ITMS) (Thermo Finni-

gan, San Jose, CA, USA). The WellChrom pump was run at a total flowrate of 2 mL min−1 (for TCM, or 0.5 mL min−1 for marine spongesamples), followed by splitting equally into five 0.4 mL min−1 (or0.1 mL min−1 for marine sponge samples) flow streams by a 6-port

Y. Zhang et al. / Analytica Chimica Acta 777 (2013) 49– 56 51

llel HP

maE(etbMfgfbsf

2

at3sUlrmmwf

2

nhm31

Fig. 1. Schematic diagram of the para

anifold tee. One stream passed through the AS-3800 autosamplernd the others passed through one of the four D-14163 injectors.ach stream was then connected to one Zorbax SB-C18 column4.6 mm × 250 mm, 5 �m, Agilent Technologies), respectively. Theluent from the first column passes through the K-2600 UV detec-or and ITMS in series for on-line monitoring and identification. Theack-pressure from UV detector is negligible (<1 bar) but that fromS is as high as about 15 bar at a flow rate of 0.4 mL min−1. Thus,

used silica capillaries (I.D. 0.1 mm, O.D. 0.19 mm, Thermo Finni-an) identical to those on MS are assembled between the otherour columns and the four fraction collectors to guarantee that theack-pressure values of the five columns are all the same. Usinguch an assembly, four 96-well plates (GIBCO®, USA) with identicalractions of herbs could be obtained for each run.

.4. Experimental conditions for HPLC and fraction collectors

In all the experiments, the mobile phase was 0.05% formiccid–water (A) and acetonitrile (B). Linear gradient programs werehe following: 0/20 and 30/60 (min/B%) for DLF; 0/20, 20/40, and0/75 (min/B%) for SYF; 0/5, 30/95, and 40/95 (min/B%) for marineponge. For DLF and SYF, flow rate was set at 2.0 mL min−1. TheV detector was set at 254, 300 and 400 nm. Fractions were col-

ected into 84 of the 96 wells from 5 to 30 min (18 s/well). Theemaining 12 wells were used for negative control samples. Forarine sponge, flow rate was 0.5 mL min−1 and UV detector wasonitored at 254 nm. Fractions were collected into 84 of the 96ells from 5–40 min (25 s/well). The remaining 12 wells were used

or negative control samples.

.5. Experimental conditions for mass spectrometry

Herbal and marine sample extracts were characterized usingormal mass scanning and product ion scanning. For the analysis

erbal extracts, negative ionization mode was used. Typical instru-ent settings were: source voltage −3.0 kV, capillary temperature

50 ◦C, sheath gas (N2) 60 arbitrary units (arb.), auxiliary gas (N2)0 arb. and tube lens offset −50 V. For the analysis of marine sponge,

LC/UV/MS based screening platform.

positive ionization mode was employed. Typical instrumentsettings were: source voltage 4.5 kV, capillary temperature 350 ◦C,sheath gas (N2) 60 arbitrary units (arb.), auxiliary gas (N2) 20 arb.and tube lens offset 50 V.

To further evaluate/confirm the potential marker compounds(for herbal samples), accurate mass was acquired using a LTQ Orbit-rap XL hybrid mass spectrometer (Thermo Fisher Scientific, SanJose, CA, USA) equipped with a MAX source. Operation parameterswere set as follows: negative ion mode, m/z range of 100–1000 withresolution set at 30,000. The data-dependent MS/MS events wereperformed on the most intense ions detected from full scanning.ESI parameters were as follows: source voltage −3.0 kV; sheathgas (N2) 50 arb; auxiliary gas (N2) 10 arb; capillary voltage −35.0 V;capillary temperature 300 ◦C; tube lens offset −110.0 V.

2.6. Sample pretreatment in 96-well plates

Samples (fractions) collected in the 96-well plates were evap-orated to dryness on a SpeedVac evaporator (SPD121P, ThermoElectron, USA). The evaporator was operated in “auto time” modeand a total of four plates could be dried at the same time. The plateswere used for subsequent bioassays.

2.7. Radical scavenging activity (DPPH assay)

One of the four 96-well plates from a parallel-separation runwas used as blank (background) samples (fractions) to precludethe interference from those compounds that have absorption peaksaround 517 nm. The other three plates were selected to evaluatethe free radical scavenging activities of herbal fractions. To eachwell of three sample plates, 100 �L each of H2O and alcoholic solu-tion of DPPH (500 �mol L−1) was added, and the microplates wereshaken in a Thermomixer (Eppendorf AG, Hamburg, Germany) for1 h (700 rpm, 37 ◦C). The remaining 12 wells of each plate were

used as negative controls. While the plate of blank samples (con-trols) were prepared using the same procedure above except fornot adding DPPH but 100 �L of ethanol. Vitamin C was served aspositive controls (2–800 �g mL−1). Changes in the absorbance of

5 Chimica Acta 777 (2013) 49– 56

sMoc

S

w(O

2

1(t1MfteSlc

S

wOr

sgCTirue

2m

mBsmF1ownCp

(a5Cur

a

Fig. 2. Results on method validation using a standard mixture (2 �M each) of antiox-

2 Y. Zhang et al. / Analytica

amples were measured spectrophotometrically at 517 nm (ELx800icroplate Reader, BIO-TEK®, USA). Radical scavenging activities

f fractions were expressed as the inhibition percentages and werealculated using the following formula:

cavenging rate (%) =[

1 − (ODsample − ODblank)

ODcontrol

]× 100

here ODcontrol is the absorbance of negative control samplessolvents + DPPH); ODsample is the absorbance of fraction samples;Dblank is the absorbance of blank (samples but without DPPH).

.8. Anti-tumor activity on tumor cell line

Dried fractions in 96-well plates were reconstituted using00 �L culture medium and logarithmic phase growth HL-60 cells1 × 104/mL) were dispensed into each well. The culture plates werehen incubated at 37 ◦C, 5% CO2 for 48 h. After centrifugation at000 rpm for 10 min, the supernatant was discarded and 100 �LTT solution (500 �g mL−1) was added to each well, followed by

urther incubation for 4 h. The supernatant was removed after cen-rifugation (1000 rpm, 10 min), and 150 �L DMSO was added intoach well. The absorbance was measured by an ELx800 Microplatepectrophotometer (BIO-TEK®) at 550 nm after the plate was oscil-ated for 10 min under 350 rpm, 37 ◦C. The cell survival rate isalculated as:

urvival rate (%) = ODsample

ODnormal× 100

here ODnormal is the absorbance of wells without samples;Dsample is the absorbance of sample (collected fraction). Doxo-

ubicin (2 �mol L−1) was served as positive control group.Dried fractions of marine sponge in 96-well plates were recon-

tituted using 100 �L culture medium and logarithmic phaserowth MCF-7 cells (5000/mL) were dispensed into each well.ulture plates were then incubated under 37 ◦C, 5% CO2 for 48 h.he anti-tumor activity of samples was evaluated by fluorescence-mage based automatic microscopy screening (FAMS) to obtainepresentative images of tumor cells on microplates [10]. To eval-ate anti-tumor effects of purified compounds, three independentxperiments were performed.

.9. Isolation and identification of anti-tumor compounds fromarine sponge

For the characterization of the two unknown compounds, theelting points of purified compounds were read directly from a

UCHI B-540 melting point apparatus. Optical rotations were mea-ured with a JASCO P1010 polarimeter. The mass spectra (normalass scanning and product ions scanning) were performed on a

innigan LCQ Deca XPplus mass spectrometer coupled to an Agilent100 HPLC device. 1H and 13C NMR spectra were recorded at 323 Kn a Bruker Avance III 500 M NMR spectrometer. Chemical shiftsere reported in parts per million (ı), using the residual CHCl3 sig-al (ıH 7.26 ppm) as an internal standard in 1H spectroscopy andDCl3 (ıC 77.0 ppm) in 13C NMR; and the coupling constant (J) wasresented in Hz.

To isolate anti-tumor compounds from marine sponge extractas prepared in Section 2.2), preparative HPLC was carried out on

Agilent 1200 system using XDB-C18 column, 30 mm × 250 mm, �m (Agilent). Pre-coated silica gel plates (HSGF-254, Yantai,hina) were used for analytical TLC. Spots were detected on TLC

nder UV light or by heating after spraying with ethanol H2SO4eagent.

The HPLC fractions of sponge A. mauritiana were screeneds described in Sections 2.4–2.8. Fraction LG-1-nB displayed a

idants (RT, DQ, SB and SA) and non-antioxidants (GA, RuA and TG). (Upper) LCchromatogram at 254 nm; (lower) radical scavenging activity of each collected frac-tion.

considerable activity with IC50 values of 1.91 �g mL−1 and it wasfurther purified by the preparative RP-HPLC (XDB-C18, 35% CH3CN,18 mL min−1) to obtain pure compounds 11 (18.5 mg, 19.3 min)and 12 (30.5 mg, 18 min).

On the basis of detailed spectroscopic analysis and comparisonwith reported data, the two compounds were identified:

(−)-Agelasine D (11). White amorphous powder, m.p.188.3–190.2 ◦C, [�]20

D −12.3◦ (c 1.0, in MeOH) [lit. −19.8◦ (c0.5, in MeOH)]. 1H and 13C NMR data, and MS agreed with theliterature values [11].Agelasine B (12). White amorphous powder, m.p. 183.8–184.7 ◦C,[�]20

D −17.1◦ (c 1.0, in MeOH) [lit. 167–170 ◦C and −21.5◦ (c 1.0, inMeOH)]. 1H and 13C NMR data, and MS agreed with the literaturevalues [12].

3. Results and discussion

The developed 5-channel parallel chromatographic mass spec-trometry and microplate based assay (Fig. 1) allowed the collectionof four identical sample fractions into 96-well plates and simulta-neous structural information of each group of fractions by UV–MSdetection. The fractions collected in the microplates could then beutilized for a broad range of in vitro screening such as enzyme- orcell-based assays.

3.1. Performance of the developed method

The suitability of the developed method for bioactivity screeningwas verified using 4 antioxidants (RT, DQ, SA and SB) and 3 non-antioxidants (GA, RuA and TG) at 2 �M. Firstly, the stability of theparallel separation system was evaluated by studying the column-to-column variation and day-to-day reproducibility. The retentiontime of each analyte eluted from the five chromatographic columnswere compared by placing a UV detector post-column (the backpressure of the UV detector is negligible (<1 bar) and hence wouldnot affect the retention time of the analytes). The chromatogram ofthe 7 standard mixtures is represented in Fig. 2 (upper). As shownin Table 1, the variations in retention time and peak area of RT, DQ,SB, SA, GA and TG eluted from one LC channel were determined to

be less than 1.1 and 5.1% RSD, and within 1.2 and 6.0% RSD amongthe five LC channels, respectively (RuA was not evaluated due toits low UV absorption). The result suggests that good repeatabilityand reproducibility of the 5-channel system can be achieved.

Y. Zhang et al. / Analytica Chimica Acta 777 (2013) 49– 56 53

Table 1Repeatability and reproducibility of one column, and reproducibility between five columns for the parallel separation system.

Compound Repeatability, %RSD Reproducibility, %RSD (3 days) Among five columns, %RSD

tR Area tR Area tR Area

RT 1.1 1.3 1.1 3.2 1.2 3.7DQ 0.4 0.9 0.9 3.1 0.6 6.0SB 0.4 2.4 0.8 3.2 0.7 4.8SA 0.1 3.9 0.5 5.1 0.7 4.5GA 0.1 2.6 0.2 3.9 0.3 3.5TG 0.2 2.8 0.2 1.3 0.2 3.6

0

A

sf1ogt

Rifaomim(ttc

3

aeMii

Fei

Average 0.2 2.5

ll experiments were conducted in triplicate analysis.

Secondly, the feasibility of the microplate bioassays with thosetandards collected from 96-well plates was evaluated. The eluentsrom the four columns were collected into four 96-well plates from0 to 35 min simultaneously at 18 s per well. After fractionation,ne microplate was used for blank correction (so as to exclude back-round interference), and other microplates were used to evaluatehe radical scavenging activity.

As shown in Fig. 2 (lower), fractions containing antioxidantsT, DQ, SB and SA (at 2 �M level) exhibited 18–58% scaveng-

ng rate, whilst low/negligible antioxidant activity was observedor GA, RuA and TG. The variation (presented as standard devi-tion) in scavenging rates is big (average of 4%, and maximumf 18%), suggesting that the main variations of the developedethod were attributed to the in vitro bioactivity experiments

nstead of the chromatographic system. The variations betweenicroplates were also tested by adding a series of concentrations

2–800 �g mL−1) of reference compounds, which further confirmedhe above assumption (data not shown). The results demonstratedhe suitability and feasibility of the developed 5-channel parallelhromatography–microplate based assay.

.2. Screening of active components from herbal extracts

The developed 5-channel parallel LC–MS-microplate-basedssay was applied to screen active compounds from two herbal

xtracts, i.e., Rabdosia rubescens (Donglingcao extract, DLF) andorus alba (Sangye extract, SYF). Aliquots of 20 �L extracts were

njected to each of five columns and the eluents were collectednto four 96-well plates and one into MS for mass detection. The

ig. 3. (Upper) LC chromatograms of SYF (A) and DLF (B) at 254 nm. (Lower) Bioactivity

xhibiting a scavenging rate higher than 40% and DLF fractions having a survival rate less tn this figure legend, the reader is referred to the web version of the article.)

.5 3.3 0.5 4.5

collected fractions of SYF were evaluated for antioxidant activityusing DPPH assay, and those for DLF were tested for cytotoxicity byMTT assay. The LC chromatograms and their corresponding activityresults were given in Fig. 3. For SYF fractions exhibiting a scaveng-ing rate higher than 40% and DLF fractions having a survival rateless than 22.4% (equivalent to the MTT results using 2 �M doxo-rubicin) were identified having the related bio-activity. The massidentification of the major peak or potential active componentsin these fractions (as indicated by their corresponding retentiontimes) could then be revealed by MS and MS/MS. Totally 6 SYF and4 DLF compounds having pronounced bioactivity were identified.

Once the active compounds had been identified, further confir-mation was made by determining their accurate mass using highresolution mass spectrometry. Using the molecular ion and MS/MSspectrum of the selected compound from the parallel LC/UV/MSn

system as a tag, we could only focus on the identification ofbioactive compounds, which greatly reduced both the time andworkload. The ten screened active compounds in Fig. 3 were furtheranalyzed by a LTQ Orbitrap mass spectrometer and the proposedelementary compositions with an error of less than 4 ppm wereshown in Table 2. Furthermore, MS3 of the ten components wereconducted using the LCQ ion trap to obtain more structural infor-mation.

Compound 1 is identified as 3-caffeoylquinic acid (3-CQA) bycomparing with a reference compound. Compounds 2 and 3 possess

the same molecular formula C16H18O9 with that of compound 1suggesting that they would be positional isomers. MS2 and MS3

(Table 2) of the three compounds are found to be consistent withthose of 3-, 5- and 4-CQA as reported by Clifford et al. [13,14].

results of DPPH (A) and MTT (B) assays for SYF and DLF fractions. For SYF fractionshan 22.4% were indicated by red color. (For interpretation of the references to color

54 Y. Zhang et al. / Analytica Chimica Acta 777 (2013) 49– 56

Table 2Characterization of the active compound from SYF and DLF using HRMS (a) and LC–MSn (b).

Extract No. tR (min) (a) [M−H]− Proposed formula Error (ppm) (b) ESI/MSn m/z (%) Identification

SYF 1 6.3 353.08725 C16H18O9 1.5 MS2[353]: 191(100), 179(69), 135(15)MS3[353→179]: 135(100)MS3[353→191]: 191(12), 173(30), 171(22), 155(9),127(76), 111(27), 93(16), 85(100)

3-CQAa

2 8.2 353.08698 C16H18O9 0.8 MS2[353]: 191(100)MS3[353→179]: 179(10), 135(100)MS3[353→191]: 191(55), 173(100), 171(17), 153(6),137(8), 127(79), 111(45), 109(34), 98(8), 93(33),87(34), 85(100), 81(11)

5-CQA

3 9.3 353.08743 C16H18O9 2.0 MS2[353]: 191(100), 179(98), 173(95), 135(14)MS3[353→179]: 135(100)

4-CQA

4 13.7 609.14441 C27H30O16 −1.0 MS2[609]: 343(8), 301(100), 300(75), 271(10), 255(7)MS3[609→300]: 300(41), 299(7), 271(100), 257(9),255(47), 179(26), 151(31)

Rutina

5 15.4 463.08704 C21H20O12 0.1 MS2[463]: 301(100), 300(52)MS3[463→300]: 301(10), 300(13), 271(100), 255(45),254(7), 179(28), 151(14)

Hirsutrin

6 15.8 593.15057 C27H30O15 0.8 MS2[593]: 327(4), 285(100), 284(5), 257(4)MS3[593→285]: 285 (100), 284 (60), 255 (40), 227 (2)

Nicotiflorin

DLF 7 8.3 593.15055 C27H30O15 −1.1 MS2[593]: 593(67), 575(10), 503(30), 473(100), 383(8),353(13)MS3[593→473]: 455(3), 383(17), 353(100)

Vicenin-2

8 15.6 717.14502 C36H30O16 −1.5 MS2[717]: 537(3), 519(100), 475(25), 339(3)MS3[717→519]: 475(100), 365(6), 339(8)

Rabdosiin

9 17.6 701.15034 C36H30O15 −1.2 MS2[701]: 537(3), 519(100), 503(15), 475(25), 459(4),339(5)MS3[701→519]: 475(100), 365(8), 339(9)

DehydroxylRabdosiin

10 18.6 359.07596 C18H16O8 −3.6 MS2[359]: 223(9), 197(20), 179(20), 161(100), 133(4)3

Rosmarinica

CC3cpi3tb3g([wmTntr

imfr5ivuti9soeb

a Verified with standard compounds.

ompounds 2 and 3 were therefore identified as 5-CQA and 4-QA, respectively. Compound 4 was identified as rutin (quercetin-rutinoside) by comparing with reported data and a referenceompound. Similar to compound 4, MS2 of compound 5 (m/z 463)roduced a fragment ion at m/z 301 (most abundant) and a rad-

cal ion at m/z 300, which are characteristic fragment ions of-substituted flavonoid glycosides [15]. Further fragmentation ofhe radical ion at m/z 300 led to the same mass spectrum foroth compounds (Table 2), suggesting that compound 5 is also a-substituted quercetin. Considering the neutral loss of 162 Da (alucose unit) in its MS2, compound 5 was deduced to be hirsutrinquercetin 3-glucoside). The major fragment ion (m/z 285) fromM−H]− of compound 6 was the loss of a rutinose unit (308 Da),hich is rather stable and hard to be cleaved. The product ionass spectrum of m/z 285 was similar to that of kaempferol [16].

he formation of a radical ion at m/z 284 suggested that the ruti-ose is connected to the 3-position of the aglycone. All these led tohe identification of compound 6 as nicotiflorin (kaempferol-3-O-utinoside).

Compound 10 was identified as rosmarinic acid by comparingts MSn and retention time with a reference standard. The exact

ass of compound 7 was m/z 593.15055, indicating a molecularormula C27H30O15. Further fragmentation of m/z 593 ([M−H]−)esulted in fragment ions at m/z 473 (loss of 120 u) and m/z03 (loss of 90 u), which are consistent with the characteristic

ons of flavonoid C-glycosides [17]. It was tentatively identified asicenin-2 (apigenin 6,8-di-C-glucoside). Compound 8, with molec-lar formula of C36H30O16, was found to be the same compoundhat we have recently identified from Isodon amethystoides, belong-ng to the same Rabdosia genus [18]. The [M−H]− of compound

was 16 Da (O) less than that of compound 8. Both compounds

hared similar dissociation pathways when losing a neutral unitf 198 (danshensu) or 182 Da, suggesting they were only differ-nt in one side chain. The neutral loss of 182 Da was deduced toe 3-(4-hydroxyphenyl) lactic acid, a natural phenolic compound

MS [359→161]: 161(67), 133(100) acid

commonly found in herbs. Thus, compound 9 was plausibly iden-tified as dehydroxyl rabdosiin.

Compounds 1 to 6 identified from SYF extract are caf-feoylquinic acids (3-CQA, 5-CQA and 4-CQA) and flavonoids (rutin,hirsutrin and nicotiflorin), which are well-known antioxidants. C-Glycopyranoside (compound 7) and three caffeic acid metabolites(rabdosiin, dehydroxyl rabdosiin and rosmarinic acid, compounds8 to 10) found in DLF extract that exhibit cytotoxicity are alsoin accord with the reported data [19–21]. Our results indicatedthat the developed method is applicable to identify potentialactive compounds from herbal mixtures and compatible withbiochemical- or cell-based in vitro screening models. Since fouridentical microplates are prepared in each run, in vitro experimentscould be performed in triplicates with another one as blank control.Moreover, in pilot studies, the four microplates can also be used toscreen different types of active compounds, reducing the overalltime required.

3.3. Screening of anti-tumor compounds from marine sponge

The developed method was also applied to screen anti-tumorcompounds from marine sources. Aliquots of 20 �L marine sponge(A. maruitiana) extracts were injected to each of the five columnsand eluents were monitored by MS or collected into 96-wellmicroplates. Fractions of marine product from one microplate wereevaluated for anti-tumor activity. The total ion chromatogram(positive ionization mode) and the anti-tumor activity of the frac-tions are presented in Fig. 4. The hit compounds (fractions elutedat 25–28 min) were then purified and identified on the basis ofdetailed spectroscopic analysis (1H and 13C NMR, ESI-MS) andcomparison with reported data. Agelasine B (12) was an optically

active, white amorphous solid (refer to Section 2.9). The molec-ular formula of C26H40N5Cl was deduced from the MS, 13C NMR,and DEPT spectra. The presence of a quaternary 9-methylladeninemoiety was indicated by the observation of the characteristic 1H

Y. Zhang et al. / Analytica Chimica Acta 777 (2013) 49– 56 55

-tumo

a(a1ssotNcddtnDgrtMatc

F

Fig. 4. (Upper) Total ion chromatogram of marine sponge extract. (Lower) Anti

nd 13C NMR signals assigned to NH2 (ı 6.84, br s, 2H), N CH3ıH 4.10, s; ıC 32.4, CH3), and five aromatic groups (ıH 10.89, snd 8.50, s; ıC 142.5 and 156.2, each CH, and 152.5, 149.8, and10.0, each C). The other NMR signals for the molecule were con-istent with those of a clerodane diterpene. The 1H and 13C NMRignals of compound 12 were identical to those of agelasine B, onef the 9-methyladenninium chromophore derivates isolated fromhe Okinawan sponge A. nakamurai. The similarity of the 1H and 13CMR signals of compounds 12 and 11 suggested that 11 is a chemi-al analog, though a difference was recognized in the signals for theiterpene moiety. Compound 11 (refer to Section 2.9) was readilyetermined as (−)-agelasine D by comparing its NMR data withhose reported ones. This compound was found in the sponge A.akamurai in the Indonesian sea. Both agelasine B and (−)-agelasine

were reported to have inhibitory effect on the growth of microor-anisms, contractive response of smooth muscles and enzymaticeaction of Na+,K+-ATPase. These purified compounds were thenested in vitro for the growth inhibitory activity toward tumor cell

CF-7 (Fig. 5). The IC50 values were found to be 7.84 ± 0.65 �M

nd 10.48 ± 0.84 �M for agelasine B and (−)-agelasine D, respec-ively. This is the first report of the anti-tumor activity of theseompounds.

ig. 5. The survival rate of MCF-F cells treated with agelasine B and (−)-agelasine D.

r activity (presented as % survival rate) of fractions of marine sponge extracts.

4. Conclusion

In this paper, a system composed of five-column parallel sepa-rations, on-line UV/MSn and 96-well plate assays, was developedto provide a fast and versatile tool for screening bioactive com-pounds from natural sources. This method retains the high efficientchromatographic separation whilst avoiding harsh prerequisitesof on-line post-column bioassays. Furthermore, sample identifi-cation and four replicates of fractions can be conducted at thesame time. Since the bioassays are performed virtually off-line in96-well plates, free choices in incubation time, necessary biochem-ical additives, and readout formats can be made. The proposedmethod may significantly reduce the time as requested in a com-mon bioactivity-guided isolation approach. The procedure wasfound to be applicable for primary screening of a broad range ofactive compounds in complex natural products.

Acknowledgements

This study was supported by the National Science and Tech-nology Major Project of China (Nos. 2009ZX09502-012 and2012ZX09304-007). The authors thank Dr. Siu-Kwan Wo (CUHK)for English polish.

References

[1] F. Cheung, Nature 480 (2011) S82–S83.[2] S.-Y. Shi, H.-H. Zhou, Y.-P. Zhang, X.-Y. Jiang, X.-Q. Chen, K.-L. Huang, Trends

Anal. Chem. 28 (2009) 865–877.[3] A.R.d. Boer, H. Lingeman, W.M.A. Niessen, H. Irth, Trends Anal. Chem. 26 (2007)

867–883.[4] H.A. Niederlander, T.A. van Beek, A. Bartasiute, I.I. Koleva, J. Chromatogr. A 1210

(2008) 121–134.[5] K. Ingkaninan, C.M. de Best, R. van der Heijden, A.J. Hofte, B. Karabatak, H. Irth,

U.R. Tjaden, J. van der Greef, R. Verpoorte, J. Chromatogr. A 872 (2000) 61–73.[6] C.F. de Jong, R.J. Derks, B. Bruyneel, W. Niessen, H. Irth, J. Chromatogr. A 1112

(2006) 303–310.[7] T. Schenk, G.J. Breel, P. Koevoets, S. van den Berg, A.C. Hogenboom, H. Irth, U.R.

Tjaden, J. van der Greef, J. Biomol. Screen. 8 (2003) 421–429.[8] Y. Wang, X. Zhao, X. Gao, X. Nie, Y. Yang, X. Fan, Anal. Chim. Acta 702 (2011)

87–94.[9] W. Tao, Z. Deqin, L. Yuhong, L. Hong, L. Zhanbiao, Z. Chunfeng, H. Limin, G.

Xiumei, J. Ethnopharmacol. 128 (2010) 575–582.

5 Chimic

[

[

[

[

[[

[[

[

[

6 Y. Zhang et al. / Analytica

10] Y. Jin, X. Zhao, Y. Zhang, X. Li, X. Nie, Y. Wang, X. Fan, Int. J. Toxicol. 30 (2011)287–299.

11] T. Hertiani, R. Edrada-Ebel, S. Ortlepp, R.W.M. van Soest, N.J. de Voogd, V. Wray,U. Hentschel, S. Kozytska, W.E.G. Müller, P. Proksch, Bioorg. Med. Chem. 18(2010) 1297–1311.

12] H. Wu, H. Nakamura, J. Kobayashi, M. Kobayashi, Y. Ohizumi, Y. Hirata, Bull.

Chem. Soc. Jpn. 59 (1986) 2495–2504.

13] M.N. Clifford, K.L. Johnston, S. Knight, N. Kuhnert, J. Agric. Food Chem. 51 (2003)2900–2911.

14] M.N. Clifford, S. Knight, N. Kuhnert, J. Agric. Food Chem. 53 (2005) 3821–3832.15] E. Hvattum, D. Ekeberg, J. Mass Spectrom. 38 (2003) 43–49.

[

[

a Acta 777 (2013) 49– 56

16] M. Ye, Y. Yan, D.A. Guo, Rapid Commun. Mass Spectrom. 19 (2005) 1469–1484.17] J. Han, M. Ye, M. Xu, J. Sun, B. Wang, D. Guo, J. Chromatogr. B: Anal. Technol.

Biomed. Life Sci. 848 (2007) 355–362.18] Y. Zhang, J. Pan, J. Zhong, Y. Wang, X. Fan, Y. Cheng, J. Chromatogr. A 1227 (2012)

181–193.19] Y. Kashiwada, K.F. Bastow, K.H. Lee, Bioorg. Med. Chem. Lett. 5 (1995)

905–908.20] J. Lee, E. Jung, Y. Kim, J. Park, S. Hong, C.G. Hyun, D. Park, Y.S. Kim, Br. J. Pharmacol.

148 (2006) 366–375.21] M.M.D. Mohammed, M. Chen, L. Zhai, N.A. Ibrahim, Eur. J. Chem. 1 (2010)

110–114.