The barks of R. davurica Pall. (growing in China’s northeast) were purchased from Jikang pharmaceutical Co., Ltd. (Hebei, China). The oven-dried samples were ground using a domestic blender, and packed in sealed polyethylene bags, then stored in a refrigerator at 4°C until use. The authentication and identification of the specimen was kindly assisted by our taxonomist (Professor Guangwan Hu) at the Key Laboratory of Plant Germplasm Enhancement and Specialty Agriculture, Wuhan Botanical Garden, Chinese Academy of Sciences. A voucher specimen (No. 0031) was deposited in herbarium of the Key Laboratory.

The reference standards of apigenin, quercetin and camptothecin were purchased from Shanghai Tauto Biotech (Shanghai, China) and Aladdin Industrial Corporation (Shanghai, China). Formic acid, methanol and acetonitrile (ACN) of HPLC grade were purchased from TEDIA Company Inc. (Fairfield, OH, United States). Polyamide was obtained from an industrial chemical company affiliated to Nan Kai University (Tianjin, China). DNA Topo I was purchased from NEW ENGLAND Biolabs (Beijing) Inc. (Beijing, China), and centrifugal ultrafiltration filters (YM-30, 30 kDa) were provided by Millipore Co. Ltd. (Bedford, MA, United States). Water for HPLC and LC-MS/MS was prepared with EPED (Nanjing Yeap Esselte Technology Development Co., Nanjing, China). All other chemicals and solvents were of analytical grade.

Firstly, 100 g raw powder sample of R. davurica was accurately weighted and then extracted in an ultrasonic bath with 60% ethanol for 30 min at room temperature, and the residue was re-extracted twice as described above. After being filtered by a quantitative filter paper under reduced pressure, the combined filtrates were concentrated in a rotary evaporator under reduced pressure at 40°C to afford the crude syrup extract. Later, the crude extract was dispersed in water (100 mL) and subjected to liquid-liquid fractionation with petroleum ether (PE, b.p. 60–90°C, to remove chlorophyll), and ethyl acetate (EA), successively. Afterward, a part of EA fraction was eluted on a polyamide column (45 cm × 5.3 cm) with distilled water to nearly colorless, and subsequently with 30 and 60% ethanol to give the two main fractions (R.d-30 and R.d-60), respectively. Finally, the two fractions were lyophilized in a freeze dryer to dryness for the subsequent analysis.

The antiproliferation assays using two human cell lines: SGC-7901 (gastric carcinoma) and HT-29 (colon carcinoma) were performed according to our previous study with slight modification (Tian et al., 2015). Briefly, both cell lines were seeded into a 96-well plate at a density of 5000 cells in DMEM (Dulbecco’s Modified Eagle Medium), supplemented with 10% fetal bovine serum (FBS), 1% penicillin- streptomycin and incubated in a humidified atmosphere containing 5% CO₂ at 37°C for 24 h, respectively. The fractions of R.d-30 and R.d-60 were dissolved in dimethyl sulfoxide (DMSO), then diluted with the medium, and further dispensed into each well to a final concentration of 150 μg/mL. The solvent, DMSO, was used as the blank control. After incubating mixtures of cells and tested fractions for 72 h, 20 μL of MTT [3-(4,5-dimethyl-2-thiazolyl)- 2,5-diphenyl-2-H-tetrazolium bromide] solution (5.0 mg/mL) was added into each well, and incubated for a further 2 h. The optical density (OD) values of the plate were detected using a Tecan plate reader at the excitation wavelength of 492 nm, and emission wavelength of 595 nm. Mean absorbance values of each fraction were calculated and the control values were subtracted from the mean values of the corresponding samples. The inhibition rate (%) was calculated using the equation: Inhibition rate (%) = (ODC – ODT)/ODC × 100%, where ODC and ODT were the OD values of control and fractions of R. davurica, respectively. Besides, the activity of camptothecin was also tested as a positive control. Each sample was tested in triplicate, and the results were expressed as means ± SD (standard deviation).

The present procedure was modified according to our most recent work (Chen et al., 2016b). In Brief, 100 μL of tested sample solution (R.d-60, 1.0 mg/mL) was mixed with 10 μL Topo I (0.5 U/μL) and incubated for 30 min at 37°C in a 0.2 mL Eppendorf (EP) tube. After that, the mixture was filtered through a 30 kDa molecular weight cut-off ultrafiltration membrane (YM-30), and then centrifuged at 10,000 rpm for 10 min at room temperature. The filtrate was washed three times by centrifugation with 200 μL of NEBuffer (pH 7.9) to remove the unbound compounds. Then, the bound compounds were dissociated by adding 200 μL of ACN/water (90:10, v/v) and then centrifuged at 10,000 rpm for 10 min with three times. The filtrates were combined and dried by lyophilization using a centrifugal evaporator, and later the residue was dissolved in 50 μL 90% aqueous ACN (containing 2.0 μg/mL isoschaftoside as the internal standard) and directly analyzed using HPLC-ESI-MS/MS. Following the similar procedure, inactivated Topo I solution (boiled for 10 min in water bath) was used as the negative control. The relative contents of compounds of interest were calculated based on the peak areas from the HPLC chromatography against that of the added standard (isoschaftoside).

The phytochemicals in the R.d-60 fraction before and after ultrafiltration were directly analyzed by HPLC-ESI-MS/MS using a TSQ Quantum Access MAX mass spectrometer (Thermo Fisher Scientific, San Jose, CA, United States) equipped with a Thermo Accela 600 HPLC system. The HPLC-UV system consisted of a quaternary pump (1250), on-line vacuum degasser, autosampler and thermostatic column compartment, connected in line to a UV detector prior to the mass spectrometer. A reverse C18 column (4.6 mm × 250 mm, 5.0 μm; Waters Symmetry, United States) was used at a flow rate of 0.6 mL/min. The powder of R.d-60 was dissolved in methanol (1.0 mg/mL), and used directly for HPLC-ESI-MS/MS analysis. A 10 μL aliquot of sample solution was injected into the HPLC system, and a binary gradient HPLC conditions were: solvent A (0.1% formic acid-H₂O), and solvent B (acetonitrile, ACN). The elution procedures were conducted at 30°C and used as follows: 20% B in 0–5 min, 20 to 70% B in 5–30 min, and 70% B in 30–35 min. The UV detection wavelength was set at 360 nm. The ESI-MS/MS experiments were carried out using a TSQ Quantum Access MAX mass spectrometer with an ESI source operating in Auto-MSⁿ mode. The MSⁿ conditions were set as follows: the electrospray capillary voltage, 3,000 V; nebulizing gas flow rate, 6.0 L/min; sheath gas (N₂) pressure, 40 arbitrary units; Aux gas (N₂) pressure, 10 arbitrary units; vaporizer temperature, 300°C; capillary temperature, 250°C; collision energy (CE), 10 V; collision energy grad (CE grad), 0.035 V/m. Mass spectrometry data were acquired in the Data-Dependent Scan mode (mass range m/z 150–1500). All of MSⁿ analyses were operated in the negative ion (NI) mode, and data acquisition and analysis was performed in the Thermo Xcalibur ChemStation.

In our previous study, the EA extracts of R. davurica were found to exhibit potential anti-proliferative effects on SGC-7901 and HT-29 cells with the inhibition rates at 43.64 and 45.53%, respectively. In this study, we focused on the anti-proliferative fraction of R. davurica in order to explore its active components. To this end, the comparisons of the anti-proliferative activity of R.d-30/60 on SGC-7901 and HT-29 in vitro were conducted. As shown in Figure 1, it is found that the R.d-60 showed significantly higher inhibitory rates against SGC-7901 and HT-29 at 61.72 and 72.29% (p < 0.001) at the same dose, compared with those of the R.d-30 at 6.37 and 8.92%, respectively, which indicated the R.d-60 was the active fraction and conducive to the further identification of its active compounds.

Figure 2 shows the ultrafiltration-HPLC analysis of the constituents from R.d-60 fraction, and some components in this fraction exhibited specific binding to Topo I. Obviously, the chromatogram of R.d-60 is very different before and after incubation with Topo I. In total, eight peaks of activated group gave bigger peak areas than the inactivated control group after incubation, which implied these eight compounds showed specific binding toward Topo I and were considered as potential ligands of topoisomerase with different anti-proliferative activities. Meanwhile, the relative contents of the eight peaks from activated and inactivated group were calculated based on the peak areas of the corresponding compounds against that of IS as shown in Table 1. It’s interesting to find that the eight compounds from activated group exhibited relatively higher contents than those of the control group incubated with inactivated enzyme. At the same time, other compounds in the R.d-60 fraction were not considered as Topo I ligands, which were supposed not to be distinguished by the ultrafiltration screening assay.

Based on the variations of the peak areas before and after incubation with Topo I, the EFs can be used to determine the degree of affinity between the ligand and the enzyme. The EF was calculated as follows:

where At, Ac, and A0 represent the peak areas obtained from incubation of the R.d-60 fraction with activated, inactivated and without Topo I in Figure 2 (Zhu et al., 2013), respectively. Among those chemical constituents from R.d-60, the unique EF could assess binding affinity between each compound to Topo I, and also imply the characteristic anti-proliferative activity. The results showed that compound 5 possessed the greatest degree of affinity (5.6%), followed by compounds 3 (5.2%), 8(5.1%), 7 (3.3%), 4 (3.0%), and 6 (1.7%) in Table 1. As expected, the EFs for each compound were different from each other, and even the lower abundant peak corresponding to the compound 5, for example, exerted a relative higher bio-affinity to Topo I. Therefore, the discrepant EFs may be due to their competitively distinguished interactions with Topo I.

After incubation with Topo I and ultrafiltration affinity screening, the eight compounds from R.d-60 with different affinities to Topo I were identified or characterized by both their HPLC retention time and MS/MS spectra in the negative ion mode. Also, the retention times (t_R), molecular masses, and contents and fragment ions of these eight components are listed in Table 1, and their corresponding structures are shown in Figure 3. Furthermore, the representative fragmentation pathways of compounds 6 and 8 were also proposed according to their corresponding MS/MS spectra and shown in Figure 4.

By comparing the MS/MS data with those of the corresponding standards (Akiyo et al., 1983; Malla et al., 2012; Chirumbolo, 2014), compounds 1, 2, and 6 (ESI-MS: m/z 287 [M-H]^-, 271 [M-H]^-, 285 [M-H]^-, respectively) were easily characterized as aromadendrin, naringenin, and sakuranetin, respectively; they commonly shared a same basic parent skeleton of dihydroflavone as shown in Figure 3. Interestingly, compounds 3, 4, and 5 (ESI-MS: [M-H]^- at m/z 269, m/z 301, m/z 299, respectively) had the same basic parent skeleton of flavone (Figure 3) and were identified as apigenin, quercetin and rhamnocitrin according to the corresponding standards and previous study (Mohamed et al., 1999; Sandra and Paula, 2010). Most of the fragment ions for flavone skeleton were also observed by ESI-MS/MS, and further fragment ions could be more informative for structural elucidation. For example, the deprotonated precursor ion of compound 5 at m/z 299 generated some fragment ions, such as [M-H-CH₃]^- at m/z 284, [M-H-CH₃-CO₂-COH]^- at m/z 211, and m/z 150 produced by the RDA-cleavage of C ring. With the same [M-H]^- ion at m/z 283 in the ESI-MS in the negative ion mode (Table 1), compounds 7 and 8 were identified as questin and physcion in accordance with a C₁₆H₁₂O₅ formula, considering the retention times, elution profile in C18 columns and the similar collision-induced dissociation (CID) pathways reported for the two isomers (Gauthier et al., 2012; Yan et al., 2014). Additionally, the same ion at m/z 268 corresponds to the loss of a methyl moiety (15 Da).

Flavonoids and anthraquinones are widely consumed as food supplements or functional foods from natural products, and have shown remarkable biological or pharmacological activities (Chowdhury et al., 2002). In Figure 3, six flavonoids could be classified into two subtypes, dihydroflavones and flavones, and two anthraquinones were detected in the R.d-60 according to their MS/MS spectra after UF-HPLC-MS assay. In order to further elucidate the structures of flavonoids and anthraquinones, the MS/MS spectra of representative compounds 6 and 8 were applied to elaborate the proposed fragmentation pathways as shown in Figure 4. Compound 6 was identified as sakuranetin (4′, 5-dihydroxy-7- methoxyflavanone, calculated as C₁₆H₁₄O₅, with Mw. 286, Figure 4A), which was indicated by a deprotonated molecular ion [M-H]^- at m/z 285 in the negative ion mode, [M-H-CH₃]^- at m/z 270, [M-H-CO-C]^- at m/z 243, and [M-H-C₈H₈O]^- at m/z 165. Additionally, the ion at m/z 119 corresponds to the characteristic RDA-cleavage of C ring, and further to the dissociation of acetenyl moiety (24 Da). By comparison of the proposed fragmentation pathways and the MS/MS spectra with those of previous study, compound 6 was suggested as sakuranetin (Chirumbolo, 2014). For compound 8 ([M-H]^- at m/z 283), several fragment ions at m/z 268, 239, 211, and 195 were obtained. Fragmented ions of [M-H-15]^- at m/z 268, [M-H-15-29]^- at m/z 239 and [M-H-15-29-44]^- at m/z 195 were produced by the successive loss of methyl moiety (15 Da), neutral CO+H moiety (29 Da) and CO₂ moiety (44 Da). Besides, the fragment ion at m/z 211 implied a neutral loss of CO moiety (28 Da) from [M-H-CH₃-COH]^- at m/z 239. As a result, compound 8 was deduced to be physcion (1,8-dihydroxy-3-methoxy-6-methyl-9,10- anthracenedione, calculated as C₁₆H₁₂O₅, Mw. 284, Figure 4B), whose fragmented ions were also consistent with those observed using ESI-MS/MS (Yan et al., 2014).

In this study, two representative bioactive compounds (3 and 4) with higher EFs were investigated against SGC-7901, HT-29, and Hep G2 cells in a concentration range of 0.37–30.0 μg/mL to further validate the screening method and the potential anti-proliferative effects. Camptothecin was served as the positive control, which was the first small molecule identified as a Topo I inhibitor (Majumdar et al., 2015). The assay results were listed in Table 2. It was observed that compounds 3 and 4 showed notably inhibition to proliferation of all of the three cancer cells tested.

At the same time, cell populations and morphological changes were observed with a phase-contrast microscopy after treated with apigenin and quercetin for 72 h. The concentrations here were approximately equal to the IC₅₀ values of corresponding cancer cell lines acquired in the anti-proliferative assay in vitro, respectively. Two compounds mentioned above effectively inhibited cell proliferation in a concentration-dependent manner, which were also in accordance with the previous MTT results. Meanwhile, significant reduction of viable cells caused by the drug treatment was observed using microscopic observations as shown in Figure 5. Furthermore, cells treated with apigenin and quercetin showed distinct morphological changes, including cell shrinkage, decreased intercellular adhesion, scattering and expanded intercellular spaces compared with those untreated cells. In the early stages of apoptosis, many morphological changes of apoptotic features, such as cell shrinkage, membrane blebbing and so on, occurred commonly in HCT-116, Hep G2, and ovarian cells (Lee et al., 2014; Yi et al., 2014; Zhao et al., 2014). Intervention with apigenin and quercetin, caused significant pharmacodynamic effects on the cellular morphology of those cancer cells, which were very similar to the results of the previous studies (Wei et al., 1990; Casagrande and Darbon, 2001; Zhao et al., 2014).

The traditional method to screen inhibitors of an enzyme usually involves both phytochemical isolation of pure compounds, and subsequent bioactivity assays using the corresponding enzyme, which is both labor-intensive and time/material -consuming. Later, bio-affinity methods based on the interaction between the ligands and the active sites of enzymes were developed (Zhang et al., 2004), and a further combination of bio-affinity methods (such as UF) with HPLC-MS techniques could further offer vital insight into biomolecule structures and ligand-receptor binding properties (Zhou et al., 2011; Qin et al., 2015). If one compound in the constituents of a mixture is able to interact with a specific enzyme, such as Topo I, the HPLC peak area of the binding compound will significantly increase in the HPLC chromatogram after its desorption from the binding enzyme. In this way, UF-HPLC-MS assay could be applied to screen and identify the ligand-receptor complexes from unbound compounds by directly comparing the chromatogram peak areas between activated and inactivated Topo I after ultrafiltration.

Rhamnus davurica has long been used as a kind of folk medicines in many Asian countries. Phytochemical investigations have found that the major constituents of the Rhamnus species are flavonoids, anthraquinones, naphthols, anthrone, triterpenes and their glycosides (Berhanu and Martin, 1995; Mohamed et al., 1999; Mai et al., 2001; Sharp et al., 2001; Chen et al., 2016a). The anti-proliferative assays showed that the R.d-60 fraction showed significantly higher inhibitory rates against SGC-7901 and HT-29, compared with the R.d-30 fraction. The efficacy of TCM usually depended on the characteristic of its chemical components. In order to further clarify and analyze the active compounds of R.d-60, ultrafiltration with Topo I coupled to HPLC-ESI-MS/MS analysis was developed and detailed afterward.

The ultrafiltration-HPLC analysis revealed that eight compounds from R.d-60 fraction showed specific binding toward Topo I. With the highest bio-affinity of the compound 5, the eight compounds with discrepant EFs disclosed competitively distinguished interactions with Topo I. To this end, these eight compounds were further identified or characterized by both their HPLC retention time and MS/MS spectra, and six of which, namely apigenin, quercetin, rhamnocitrin, sakuranetin, questin, and physcion, showed more potential antitumor activity with higher EFs.

Flavonoids are ubiquitously occurring and well-documented for their anticancer activities. To date, more than 8,000 natural flavonoids have been identified with diversified activities, and their discrepant biological activities are mainly due to their structural diversity and various chemical modifications (Malla et al., 2012). In this study, the two representative compounds, apigenin and quercetin, exerted notably inhibition to proliferation of all of the three cancer cells tested in Table 2.

Since Topo I relaxes supercoils by reversibly nicking duplex DNA to control DNA replication (Kim et al., 2011), hypothesis is that the two compounds could reversibly block Topo I -mediated cleavage of DNA complex, ultimately leading to DNA strand breaks and activation of apoptosis (Majumdar et al., 2015). Meanwhile, as a potent inhibitor of epidermal ornithine decarboxylase for the topical application, apigenin and quercetin significantly inhibited the incidence of skin tumorigenesis initiated by 7, 12-dimethylbenz (a) anthracene (DMBA) and promoted by 12-O- tetradecanoylphorbol-13-acetate (TPA) in SENCAR mice (Wei et al., 1990). Depending on the different chemical structures, flavones which contain the C2-C3 double bonds have been reported to arrest human gastric cancer cells in G1 phase, while the isoflavones which lack the C2-C3 double bonds have shown a distinct block at the G2/M transition (Matsukawa et al., 1993). Casagrande and Darbon (2001) also discovered that quercetin and apigenin showed remarkable inhibitory effect on human melanoma cells (OCM-1), the mechanisms were due to the presence of a hydroxyl group at the 3′-position of the B ring of quercetin, which could arrest cell cycle in the G1 phase, while its absence in apigenin correlated to a G2 block.

It is well known that flavonoids are the secondary metabolites of plants and have exhibited well-documented pharmacological effects, such as anti-inflammatory, anti-mutagenic and anticancer activities (Wu et al., 2013). Generally, the structural diversity of flavonoids based on substituent groups could largely contribute to their diversified biological activities. Previous studies further revealed that the structure-activity relationships involved the following respects: (1) the oxo functional group at position 4 and a C2-C3 double bond of C ring were both required for maximal anti-proliferative activity; (2) the presence of hydroxyl groups existing at the 3′- and 4′- positions of B ring conferred the maximal inhibitory effect; (3) the 3′ -OH of the B ring might be important for the arrested cell cycle level (Casagrande and Darbon, 2001; Xiao J.B. et al., 2014; 2015). For example, apigenin and myricetin exhibited 3.8 times higher bio-affinities against γ-globulin than the naringenin and dihydromyricetin (Xiao et al., 2011a). Meanwhile, flavone, 6-hydroxyflavone, 6-methoxyflavone, and myricetin showed lower binding affinities for common human plasma proteins (CHPP) about 10.02 to 17.82 times after hydrogenation of C2-C3 double bond of C ring (Xiao et al., 2011b). Interestingly, this coincided to our data summarized in Table 1 and Figure 3, which revealed the reason that those flavones of compounds 3, 4, and 5 exhibited higher EFs than the dihydroflavones of compounds 1, 2, and 6, and confirmed the requirement of a C2-C3 double bond of C ring for the bio-affinities between the active components and Topo I. In addition, compound 4 showed markedly inhibition to proliferation on SGC-7901 (IC₅₀: 12.70 ± 0.71 μg/mL) and Hep G2 (IC₅₀: 4.78 ± 0.28 μg/mL), which implied that compound 4 could be a better Topo I inhibitor. A similar study also reported that quercetin significantly inhibited the cell proliferation of the gastric and colon cancer cells, as well as in leukemic cells during the cell cycle at the G1/S transition (Yoshida et al., 1992). In consideration of the above results, the successful screening and validation of these bioactive components from R.d-60 demonstrated the UF-HPLC-MS could be an effective method to rapidly screen and identify bioactive compounds from the complex natural products in order to discovery better natural inhibitors against Topo I and more potential natural anti-cancer candidates.