Zebrafish Bioassay-Guided Natural Product Discovery: Isolation of Angiogenesis Inhibitors from East African Medicinal Plants

Natural products represent a significant reservoir of unexplored chemical diversity for early-stage drug discovery. The identification of lead compounds of natural origin would benefit from therapeutically relevant bioassays capable of facilitating the isolation of bioactive molecules from multi-constituent extracts. Towards this end, we developed an in vivo bioassay-guided isolation approach for natural product discovery that combines bioactivity screening in zebrafish embryos with rapid fractionation by analytical thin-layer chromatography (TLC) and initial structural elucidation by high-resolution electrospray mass spectrometry (HRESIMS). Bioactivity screening of East African medicinal plant extracts using fli-1:EGFP transgenic zebrafish embryos identified Oxygonum sinuatum and Plectranthus barbatus as inhibiting vascular development. Zebrafish bioassay-guided fractionation identified the active components of these plants as emodin, an inhibitor of the protein kinase CK2, and coleon A lactone, a rare abietane diterpenoid with no previously described bioactivity. Both emodin and coleon A lactone inhibited mammalian endothelial cell proliferation, migration, and tube formation in vitro, as well as angiogenesis in the chick chorioallantoic membrane (CAM) assay. These results suggest that the combination of zebrafish bioassays with analytical chromatography methods is an effective strategy for the rapid identification of bioactive natural products.


Introduction
Small molecules from natural sources are recognized as evolved, privileged structures with greater likelihood than many synthetic compounds to exhibit specific bioactivities [1]. For example, 73% of cancer therapeutics approved to date are either natural products or derivatives thereof [2]. Nevertheless, the use of natural products in drug discovery has significantly declined in the past two decades, due in part to persisting difficulties in the systematic isolation and synthesis of such molecules [3]. One promising strategy to better exploit the therapeutic potential of natural products could be the use of more biomedically relevant assaysideally in vivo models -for the screening and bioactivity-guided fractionation of plant, fungal and microbial extracts.
Many currently known bioactive natural products were originally identified using in vitro assays (e.g. cytotoxicity in tumor cells) for their activity-guided isolation from extracts. The biological activity of many other natural products was determined only after their initial isolation on the basis of physical characteristics (e.g., preparative chromatography followed by mass spectrometry and NMR spectroscopy analysis). Because of the low throughput of conventional in vivo models such as mice and rats, in addition to the relatively large amounts of compound required for testing in these systems, in vivo assay-guided fractionation is currently not a widely-used approach for the discovery of drug-like natural products.
Novel opportunities for in vivo natural product discovery have arisen through the recent emergence of zebrafish as an effective model system for the identification of disease-relevant genes and bioactive small molecules [4]. Large-scale genetic screens in zebrafish carried out since the early 1990s have led to the identification of therapeutically relevant genes in several indication areas, including cardiovascular, neurological, gastrointestinal, musculoskeletal, and metabolic disorders [5,6,7,8]. In addition, small-molecule screens carried out in zebrafish within the past decade have confirmed the ability of this model system to identify bioactive compounds in a target-independent manner, thereby enabling the discovery of novel mechanisms of action [9,10,11,12,13].
The primary advantages of zebrafish for drug discovery include their high genetic, physiologic, and pharmacologic similarity with humans, as well as the small size, optical transparency, rapid development, and large numbers of their embryos and larvae, which are the primary system for experimental analysis. Because of their small size (1 to 5 mm), zebrafish embryos and larvae are compatible with microtiter plates for screening (primarily 24-and 96-well plates, but even 384-well plates are possible), thereby requiring only microgram amounts of each extract, fraction, or compound to be tested. Because of the high fecundity of zebrafish, large numbers of embryos and larvae can be produced and analyzed in a more cost-effective manner than, for example, mice and rats. Combined, these features define zebrafish as an ideal in vivo model for the systematic identification of bioactive natural products with therapeutic potential [4].
For an initial evaluation of zebrafish as a platform for natural product discovery, we opted to identify novel inhibitors of angiogenesis. Despite the recent regulatory approval of recombinant antibodies and small molecules targeting the vascular endothelial growth factor (VEGF) pathway, the clinical efficacy of these therapies for various cancers is limited [14]. Also, despite the large number of compounds targeting this pathway [15], many of these have shown limited or insufficient efficacy in clinical trials, or are associated with toxicities such as arterial thromboembolic events [16]. For these reasons, there is still a need for novel antiangiogenic compounds with different mechanisms of action, some of which might be suitable for use in combinatorial therapy strategies [17].
Numerous in vivo and in vitro assays have been developed since the 1970s for the evaluation of anti-angiogenic molecules [18], yet because of various disadvantages (low throughput, high cost, and the requirement for larger compound amounts in the case of in vivo assays, and limited predictive value in the case of in vitro assays), these are not ideal as front-line assays for natural product discovery (i.e. for high-throughput screening and rapid bioassay-guided fractionation). Zebrafish, however, offer an interesting combination of (1) being an in vivo model and (2) enabling high-throughput, lowvolume screening. Within the past decade, zebrafish embryos have become well-established as an in vivo model for the analysis of angiogenesis and vascular development [9,19,20,21,22,23]. To test the suitability of zebrafish as an in vivo frontline assay for the bioassay-guided fractionation of complex natural extracts, we therefore combined an embryonic zebrafish angiogenesis assay with analytical chromatography methods, with the goal of rapidly isolating phytochemicals from medicinal plant extracts capable of inhibiting vascular outgrowth in this assay.

Results and Discussion
We chose an angiogenesis assay based on the evaluation of intersegmental vessel (ISV) outgrowth in fli-1:EGFP transgenic embryos [19], which exhibit vasculature-specific expression of enhanced green fluorescent protein (EGFP) in the trunk and tail during embryonic and larval development (Figs. 1a, b). With respect to natural product research, fli-1:EGFP zebrafish have been used to characterize the angiogenic activity of Angelica sinensis (dong quai) [24], as well as the anti-angiogenic activity of solenopsin, an alkaloid isolated from Solenopsis invicta (fire ants) [25]. Similar transgenic lines, with fluorescent reporter proteins expressed under the control of the endothelial cell-specific flk-1/ VEGFR2 promoter, have recently enabled an ENU mutagenesis screen to identify genetic determinants of vascular development [26] and a small-molecule screen to identify novel angiogenesis inhibitors [27]. To test the utility of this zebrafish assay for natural product discovery, we screened crude methanolic extracts from over 80 East African medicinal plants. Two extracts, from Oxygonum sinuatum (Meisn.) Dammer (Polygonaceae) and Plectranthus barbatus Andrews (Lamiaceae), inhibited ISV outgrowth in fli-1:EGFP embryos in a dose-dependent manner (Figs. 1c, d and 2a, b).
In terms of known bioactivities for these plants, O. sinuatum has been documented as an ethnobotanical treatment in Kenya for several unrelated disorders [28]. No phytochemical analysis of this plant has been reported to date. P. barbatus (previously also known as Coleus forskohlii Briq.) is widely used in traditional medicine in Africa and Latin American to treat a range of human ailments [29]. This species is also well-known as the primary source of forskolin, a labdane diterpenoid and activator of cAMP signaling [30,31]. Intruigingly, although forskolin has been shown to inhibit angiogenesis in the chick chorioallantoic membrane assay [32] and in vitro [33], it is also known to upregulate VEGF expression [34], making its overall effect on angiogenesis in vivo difficult to predict. We determined that forskolin does not inhibit angiogenesis in zebrafish (data not shown) and since it is isolated from P. barbatus roots (an extract from leaves being the subject of this analysis), we concluded that the anti-angiogenic activity seen in zebrafish embryos for the P. barbatus extract is likely due to the bioactivity of another compound.
We next sought to isolate from O. sinuatum and P. barbatus extracts the principle components responsible for their anti-angiogenic effects. Both crude methanolic extracts were fractionated via thinlayer chromatography (TLC), using toluene/ethyl formate/formic acid (5:4:1) as the solvent. A single analytical-scale TLC plate (20620 cm) was used to separate 10 mg of each extract, and was subsequently divided into 10-15 horizontal strips based on the presence of UV 254 -absorbing and UV 365 -emitting components (Figs. 3 a, b). The silica was removed from these strips and extracted with methanol, after which the eluted constituents were subjected to bioactivity analysis in zebrafish, followed by high-resolution electrospray ionization mass spectroscopy (HRESIMS) for those exhibiting anti-angiogenic activity. For both O. sinuatum and P. barbatus, single TLC fractions were identified in this manner which phenocopied the anti-angiogenic activity shown by the crude extracts.
O. sinuatum yielded a single bioactive compound whose molecular formula was determined to be C 15 H 10 O 5 based on the deprotonated molecular ion at m/z 269.0457 identified by HRESIMS analysis (negative mode), suggesting 6-methyl-1,3,8trihydroxyanthraquinone (emodin) -a known product of several other Polygonaceae species. Emodin has recently been reported as an inhibitor of angiogenesis in vitro [35] and in vivo [36], and has been shown to be an inhibitor of the protein kinases Lck [37], HER-2 [38], and CK2 [39]. More recently, emodin was shown to inhibit angiogenesis in vitro at least in part by restricting the phosphorylation of VEGFR2 [40]. In addition; CK2 has been found to directly phosphorylate Akt [41], one of several downstream elements of VEGF signaling, and this modification has been shown to be essential for nuclear retention of FOXO1, an important cytoplasmic inhibitor of angiogenesis [42]. Semi-synthetic emodin revealed an MS spectrum identical to the bioactive compound isolated from O. sinuatum (data not shown) and, importantly, phenocopied both this compound and the crude extract (Fig. 1e, 2c), thereby confirming emodin as the primary constituent responsible for this plant's bioactivity. Furthermore, emodin and other anthraquinones synthesized by Rheum species (rhubarb root, or Dahuang) have recently also been shown to inhibit vascular outgrowth in zebrafish using a histochemical assay to visualize blood vessels [43].
P. barbatus yielded a bioactive molecule with an apparent M w of 355.1190 based on HRESIMS analysis and the predicted molecular formula C 20 H 19 O 6 , suggesting coleon A lactone, a known product [44] of another Lamiaceae species but with no previously described bioactivity. Following isolation by preparative-scale chromatography, the identity of coleon A lactone (hereinafter referred to as coleon AL) was confirmed by NMR (Table 1).
Zebrafish have only recently been utilized for the systematic identification of bioactive small molecules [12], so the predictive power of zebrafish assays for drug discovery will only become clear as molecules found using this platform are advanced into the clinic. In any case, to further evaluate the therapeutic potential of natural products identified using embryonic or larval zebrafish models, bioactive compounds should subsequently be validated using a series of additional in vitro and in vivo assays. To this end, in vitro anti-angiogenesis assays were carried out to further characterize the anti-angiogenic activity of the bioactive natural products isolated in this study, revealing both emodin and coleon AL to inhibit the proliferation, migration and tube formation of mammalian endothelial cells ( Table 2, Fig. 4, 5). In addition, both compounds inhibited blood vessel formation in the chick chorioallantoic membrane (CAM) assay (Fig. 6). Emodin and coleon AL inhibited the proliferation of mouse aortic endothelial cells (MAECs) with an IC 50 similar to that of the vascular endothelial growth factor (VEGF) receptor inhibitor SU5416, a synthetic indoline derivative [45], and inhibited the proliferation of bovine aortic endothelial cells (BAECs) with an IC 50 similar to that of the PI3K inhibitors wortmannin, a fungal furanosteroid [46], and LY294002, a synthetic chromone derivative [47]. In the CAM assay, we determined the anti-angiogenic activity of emodin and coleon AL to be similar to that of SU5416.
As a next step, in vivo angiogenesis assays should be carried out for these compounds in mammalian models, including mouse tumor assays [18]. Additional in vitro assays should also be performed to determine the activity of anti-angiogenic compounds under hypoxic conditions, an important condition for possible application in antitumor therapy [48].
Zebrafish were first proposed as a model for in vivo drug discovery in 1957 by Jones and Huffmann of the Oklahoma Research Foundation [49], and soon thereafter were used for the first time to analyze the bioactivities of natural products [50]. Only more recently, however, have zebrafish been widely used for the function-based identification of disease-relevant genes and bioactive compounds [12]. An important advantage of using zebrafish embryos and larvae for the identification of bioactive molecules is that they are living organisms, thereby enabling the rapid in vivo evaluation of compounds not only in terms of their pharmacological activity but also of their possible toxicity (particularly cardio-, neuro-, and hepatotoxicity) early in the drug discovery process [51,52]. Within the past decade, zebrafish have furthermore emerged as a powerful model for chemical genetics, both with respect to the in vivo dissection of signaling pathways [20,53] and to the elucidation of the mechanism of action of bioactive natural    products [54,55]. In the area of natural product discovery, one recent report describes the application of a histochemical assay in zebrafish to guide the isolation of anti-angiogenic terpenoids from Tripterygium wilfordii, an anti-inflammatory Chinese medicinal plant, using preparative chromatography methods [56].
Here, we demonstrate the utility of zebrafish bioassay-guided fractionation by analytical chromatography techniques, and further establish zebrafish as an in vivo platform for the discovery of bioactive natural products. Based on these initial results, it appears possible that zebrafish can help address a critical bottleneck in natural product discovery by enabling the rapid, in vivo and microgram-scale bioactivity analysis, bioassay-guided fractionation, and dereplication of complex natural extracts. While data described here were obtained using zebrafish bioassay-guided TLC fractionation, additional advantages for accelerating natural product discovery will be realized through the combination of zebrafish bioassays with more advanced HPLC techniques -in particular, those enabling microfractionation [57,58]. With a wide array of biomedically relevant assays now established in zebrafish [4], the advantages of this in vivo system for natural product discovery should facilitate the systematic identification of a new generation of bioactive natural products with potential utility in drug discovery.

Zebrafish
The fli-1:EGFP transgenic line [19] was obtained from the Zebrafish International Resource Center at the University of Oregon (Eugene, Oregon, USA). Zebrafish husbandry, embryo collection, and embryo and larvae maintenance were performed as described earlier [59,60]. Zebrafish assays were standardly performed in 24-well microtiter plates using 10 embryos per well in 1 ml of 0.36 Danieau's medium (17 mM NaCl, 2 mM KCl, 0.12 mM MgSO 4 , 1.8 mM Ca(NO 3 ) 2 and 1.5 mM HEPES, pH 7.6). Embryos were exposed to extracts and compounds at 16 hours post-fertilization (hpf) -approximately 8 hours prior to the initiation of intersegmental vessel (ISV) outgrowth -and scored for relative vascular outgrowth at 40 hpf. Extracts and compounds were solubilized in dimethyl sulfoxide (DMSO, Agros Organics, Geel, Belgium), and were added to the medium up to a maximum DMSO concentration of 1%. The extent of outgrowth of intersegmental vessels (ISVs) was determined using a scoring method that takes into account both the approximate number of outgrowing vessels (100, 75, 50, 25, or 0%) and the average degree to which these vessels have extended into the trunk from the dorsal aorta/posterior cardinal vein (DA/PCV) (100, 75, 50, 25, or 0%). These two values are multiplied to give the relative vascular outgrowth (RVO) score.  Table 2. Emodin and coleon AL inhibit endothelial cell proliferation in vitro.

Compounds
Semi-synthetic emodin was obtained from Janssen Chimica (Geel, Belgium). SU5416 and wortmannin were obtained from Sigma-Alrich (Bornem, Belgium) and LY294002 was obtained from Cayman Chemical (Talinn, Estonia). Coleon A lactone (coleon AL) was isolated from leaves of Plectranthus barbatus collected in Handeni, Tanga Region, Tanzania. Leaves were dried at ambient temperature under sunlight, homogenized, and extracted 3 times with chloroform (1 l chloroform per 100 g leaves). This chloroform extract was concentrated 100:1 on a rotary evaporator and subjected to chromatographic separation on a LaFlash chromatography apparatus from VWR (West Chester, Pennsylvania, USA) using VWR SuperVarioFlash silica cartridges (30 g, Si60, pore size 15-40 mM), with chloroform:acetic acid 200:1 as the solvent. Coleon AL was isolated as the second major peak showing absorbance at 254 nm.
Thin-layer chromatography TLC plates were obtained from Macherey-Nagel (Düren, Germany). For the experiments described here, 20620 cm aluminum plates coated with TLC silica gel 60 (layer thickness, 0.2 mm) containing a UV 254 fluorescence indicator were used (ALUGRAM SIL G/UV 254 , Machery-Nagel product number 818133). Plates were loaded manually, using a finely tapered micropipette tip, with 10 mg of crude extract (50 mg/ml in methanol), dried for 15 seconds with a hair dryer at low heat, and placed in an enclosed, upright 25625610 cm glass chamber containing 100 ml toluene/ethyl formate/formic acid 5:4:1 (preequilibrated for at least 15 minutes, and with 3 vertical faces covered with solvent-saturated filter paper).

High-resolution electrospray ionization mass spectrometry
Electrospray ionization (ESI) mass spectra were recorded in positive and negative mode on an orthogonal acceleration quadrupole time-offlight mass spectrometer (Q-Tof 2, Micromass, Manchester, UK). The electrospray needle voltage was set to 3000 V or 22850 V for the positive and negative mode respectively. Fragment ion spectra were  obtained by selecting the precursor ion in the quadrupole and collisional activation with argon gas in the collision cell. Accurate mass measurements were performed at a resolution of 9000 using the protonated leucine-enkephaline (YGGFL) ion as lock mass.
NMR spectroscopy 1 H and 13 C NMR spectra were recorded on a Bruker (Fällanden, Switzerland) Avance II 500 spectrometer operating at 500.130 MHz for 1 H and at 125.758 MHz for 13 C, and using a gradient-equipped inverse 5 mm triple probe with p/2 pulses of 6.5, and 14.5 ms respectively. The standard Bruker Topspin 2.1 software under Windows XP was used throughout. All experiments were performed at 22 uC in deuterochloroform solution with the solvent peak as internal standard set at 7.27 ppm ( 1 H) or 77.0 ( 13 C) vs. TMS respectively. First-order analysis was applied throughout, and firstorder multiplets or apparent first-order multiplets were denoted as follows: s = singlet, d = doublet, dd = double doublet, t = triplet. J-values were extracted directly from the splittings in the spectrum, and are not optimised. Spectral assignments were based not only on the usual chemical shift rules and coupling patterns, but especially on routine 2D-correlations such as COSY45-(homonuclear H,H Jcorrelations), GHSQC-(single bond C,H 1 J-correlations) and GHMBC-experiments (multiple bond C,H 3 J-correlations). The data for coleon AL are summarized in Fig. 4 and compared with previously reported values [44].

Imaging
Zebrafish were screened for GFP fluorescence using an Axiovert 40 CFL microscope from Zeiss (Oberkochen, Germany) equipped with an MBQ 52 AC fluorescence lamp from LEJ (Jena, Germany). Micrographs of zebrafish embryos were taken on a Stemi 2000 stereo microscope from Zeiss equipped with a DP200 CMOS digital camera and using DpxView Pro EE EF software, both from Deltapix (Maalov, Denmark). Confocal fluorescence micrographs of zebrafish embryos were acquired using a Nikon A1R confocal unit mounted on a Ti2000 inverted microscope (Nikon, Tokyo, Japan). The microscope was equipped with 46 (0.2 N.A.) and 106 (0.45 N.A.) objective lenses, and fluorescence was revealed using a 488 nm laser line (CVI Melles Griot, Albuquerque, NM, USA). For imaging, zebrafish embryos were anesthetized using 0.1 mg/ml ethyl 3-aminobenzoate methanesulfonate (Sigma) in 0.36 Danieau's solution.

Cell proliferation assays
Cells (MAEC or BAEC) were seeded in 48-well plates at 10,000 cells per cm 2 . After 16 h, the cells were incubated in fresh medium in the presence of different concentrations of the test compounds (i.e. 100 -20 -4 -1 -0.2 mM). On day 5, cells were trypsinized and counted by a Coulter counter (Analis, Belgium). The compound concentration that inhibits cell growth by 50 % (i.e. IC 50 ) was calculated based on cell counts in control cultures.

Cell migration assay
Wounds were created in confluent MAE cell monolayers with a 1.0-mm wide micropipette tip. Then, cells were incubated in fresh medium containing 10% FCS in the presence of the test compounds. After 8 h, the wounds were photographed, and endothelial cells invading the wound were quantified by computerized analysis of the digitalized images.

Tube formation assay
Wells of a 96-well plate were coated with 60 ml matrigel (10 mg/ml, BD Biosciences, Heidelberg, Germany) at 4 uC. After gelatinization at 37 uC during 30 min, BAEC (60,000 cells) were seeded on top of the matrigel in 200 ml DMEM containing 1% FCS and the test compounds. After 6 hours of incubation, the cell structures were photographed at 1006magnification. Tube formation was quantified by counting the number of branching points.

Chorioallantoic membrane (CAM) assay
The in vivo CAM angiogenesis model was performed as described with slight modifications [61]. Fertilized chicken eggs were incubated for 3 days at 37 uC when 3 ml of albumen was removed (to detach the shell from the developing CAM) and a window was opened on the eggshell exposing the CAM. The window was covered with cellophane tape and the eggs were returned to the incubator until day 9 when the compounds were applied. The compounds were placed on sterile plastic discs (Ø 8 mm), which were allowed to dry under sterile conditions. A solution of cortisone acetate (100 mg/disc, Sigma-Aldrich, St. Louis, MO, USA) was added to all discs in order to prevent an inflammatory response. A loaded and dried control disc was placed on the CAM approximately 1 cm away from the disc containing the test compound(s). Next, the windows were covered and the eggs further incubated until day 11 when the area around the discs was cut-off and photographed. Then, 2 concentric circles were positioned on the digitalized pictures and all vessels intersecting these circles were counted. A two-tailed paired Student's t-test was performed to assess the significance of the obtained results.