Antioxidant and anti-infectious properties of C. alata extracts are well documented in the literature [1,2,10,11]. In order to increase the relative amount of probable bioactive compounds of the aqueous extract of C. alata leaves, we decided to refine the aqueous extract using Reverse Phase (C18)-Solid Phase Extraction obtaining a 17-fold concentrated extract named CaRP that showed much more intensive bands in TLC than the original aqueous extract (data not shown). CaRP also presented antimicrobial activity against Staphylococcus aureus, S. epidermidis, Bacillus subtilis and Pseudomonas aeruginosa (data not shown). We used the DPPH assay to also search for antioxidant properties of CaRP.

In order to calculate the IC₅₀, calibration curves of each standard (ascorbic acid and Trolox) and of CaRP were determined by 2,2-diphenyl-1-picrylhydrazyl (DPPH) assay as described in Material and Methods. The curves proved to be linear in the concentration range shown in Table 1 using linear regression and resulted in a correlation coefficient (r) >0.99 to standards and CaRP.

At the conditions used, antioxidant activity of ascorbic acid was higher than that of Trolox. Interestingly, CaRP had higher antioxidant activity than ascorbic acid and Trolox, i.e., 0.5639 g of antioxidant activity of CaRP is equivalent to 1 g of ascorbic acid, or 0.5000 g of CaRP is equivalent to 1 g of Trolox. Compared to the literature, the IC₅₀ of CaRP was about 50× less (IC₅₀ of CaRP: 2.25 μg/mL; IC₅₀ of ethanolic C. alata extract: 112.46 μg/mL) than ethanol-extracted without any refined method [11].

Since CaRP may be a valuable source of compounds, especially of AST, with its different biological and pharmacological effects, we decided to evaluate the effect of CaRP using Saccharomyces cerevisiae mutants lacking synthesis of endogenous antioxidants, DNA repair enzymes or membrane constituents as model cells to find possible targets that could hint at mechanisms of action.

S. cerevisiae null mutants that lack only one gene in comparison to the isogenic wild type offer the possibility to study the relevance of a missing specific metabolic pathway on the response to the defined treatment with the chemical studied. The absence of the protein codified by that gene leads to specific responses described as phenotypes [12] that characterize the metabolic pathway in which it is involved, i.e., mutants in proteins involved in DNA repair are more sensitive to drugs that cause DNA damage that is repaired by that biological pathway. Therefore, this is a very powerful system to rapidly identify a putative mechanism of drug action in a complex cell [13]. Here we used a set of 10 yeast null mutants involved in DNA repair, representing the three so-called epistasis groups, which are thought to reflect three major DNA repair pathways: RAD3 (involved in nucleotide excision repair [NER], Figure 2A), RAD52 (recombinational repair, Figure 2B) and RAD6 (error-prone repair, Figure 2C) epistasis group genes [14–16]; 3 yeast mutants defective in membrane composition or transport across the membrane (Figure 2D); and 3 yeast mutants defective in oxidative stress protection mechanisms (Figure 2E).

Although CaRP (12 mg/plate) did not show any cytotoxic effect in wild type strains (BY4741, SOD2), isogenic DNA repair mutants showed higher sensitivity when treated with CaRP (Figure 2A–C). Interestingly, all repair mutants of the three epistasis groups of DNA repair (RAD3, RAD52 and RAD6 epistasis group) showed some degree of sensitivity. These results suggest that CaRP contains one or more substances that interact with DNA individually or simultaneously in distinct ways, considering that the repair pathway seemed not specific. Additionally, all mutant strains in the RAD3 (nucleotide excision repair—NER, Figure 2A), RAD52 (homologous recombination, Figure 2B) and RAD6 (translesion repair, Figure 2C) epistasis group had similar sensitivity. The higher than wild-type sensitivity to pso2Δ (snm1Δ) (Figure 2B) suggested that some substance of CaRP might intercalate with DNA and induce lesions that link the two DNA strands. Evidence of this comes from some studies that showed pso2Δ mutant specifically blocked in repair of interstrand cross-links [13]. These results could be promising because many antitumoral drugs act by intercalating into DNA and forming interstrand cross-links, e.g., photoactivated psoralens, mitomycin C and cis-platinum [17].

Mutants defective in membrane transport such as aus1Δ (lacking the function of uptake of exogenous sterols and its subsequent incorporation into the plasma membrane) and yor1Δ (lacking multidrug transporter which mediates export of many different organic anions) had the same sensitivity as the DNA repair-mutant strains except for erg3Δ (lacking C-5 sterol desaturase) that was almost unaffected, i.e., showed a wild type response (Figure 2D). Erg3p is known to protect cells specifically from drugs inducing lipid peroxidation [18]. Lack of sensitivity in erg3Δ mutant, therefore, indicates that CaRP did not induce lipid peroxidation.

Conversely, mutants lacking endogenous antioxidants such as superoxide dismutase (sod2Δ) or catalase (ctt1Δ and cta1Δ) had identical response to CaRP as the isogenic wild type that is proficient in all these enzymes (Figure 2E). This result corroborates with the DPPH assay (Table 1) and suggests that CaRP in the applied concentrations is not pro-oxidant in these eukaryotic cells.

Since CaRP showed cytotoxicity in DNA repair-deficient yeast mutants and in strains with defects in membrane transport, we decided to test if CaRP compounds could bind to DNA in vitro. For this, centrifugal ultrafiltration sampling followed by HPLC analysis was used, a fast screening for bioactive compounds binding to DNA [19].

The Amicom filtrates from CaRP extract and from the mixed solution of CaRP + DNA were collected and analyzed by HPLC (Figure 3). Because the free concentration of ligands decreases in the solution after the interaction (DNA remains in the Amicon membrane and only unbound metabolites with medium polarity can pass through) the peak areas of these compounds decrease in the chromatogram; for those not DNA binding, there should be no or little change in their peak areas before and after the interaction. Therefore, comparing the chromatograms of the Amicon filtrates after the reaction of the control (without DNA) and with DNA, the DNA-binding compounds in the CaRP can be easily distinguished from those that do not have affinity to DNA.

The binding degree of any component to DNA was calculated according to Zhou et al. [19]:

where the peak areas of a compound before (A_b) and after (A_c) the interaction with DNA in the HPLC chromatograms, were measured.

Figure 3 shows the bio fingerprinting chromatograms at 254 nm. The integration of the peak areas showed that 3 peaks in CaRP decreased significantly after interaction with DNA: the respective decreases were 64% for peak 10, 72% for peak 11, and 78% for peak 20.

We chose to study by FTIR spectroscopy the CaRP fraction equivalent to peak 11 (Figure 3) in its interaction with calf thymus DNA since it is the major fraction (yield of 52%) and also showed cytotoxicity against prokaryotic cells, such as S. epidermidis and s. aureus [20].

The equivalent fraction of HPLC peak 11 was obtained by LC-microfractionation (Figure 4). Identification was carried out using FTIR, high-resolution mass spectrometry (ESI-qTOF-MS/MS) and H¹ and C¹³ NMR.

Fraction 11 (F11): yield of 0.523 mg/mg. Green powder, IR (KBr) cm⁻¹: 3364 (OH), 1782 (C=O), 1656, 1607, 1505, 1452 (C=C, Ar), 1286 (=C–O–C), 1118 (C–OH); Negative high resolution ESI-qTOF-MS m/z 447.1187 [M-H]⁻ (32%), MS-MS fragmentation of m/z 447.1187: 284.0699 [M-H-Glu]⁻, 255.0648, 227.0681; H¹ NMR (MeOD) and C¹³ NMR (MeOD) were carried out and after comparison to data in the literature [3,4], F11 was identified as kaempferol-3-O-β-d-glucoside (astragalin) (Figure 5).

Only a little quantity of AST was isolated from various plants such as persimmon leaves, Eucommia ulmoides leaves, Nelumbo nucifera (Lotus leaves), Solenostemma argel, garlic leaf and shoot, Flaveria bidentis, Cuscuta chinensis, and the seeds of Centaurea schischkinii [3–7].

Although this compound has been isolated from C. alata leaves before [1,2,21], our results showed that solid phase extraction is an improved technique to obtain the enriched extract CaRP, having also the advantage of an easy and rapid purification method.

The infrared spectrum of AST-DNA mixture offered evidence for the direct binding of this compound to DNA (Figure 6 and Table 2). The vibrational frequencies of NH and OH groups of free DNA appeared around 3469–3430 cm⁻¹, which is in accordance to Usha and co-workers [22], and in the AST-DNA complex had a little shift to 3469–3435 cm⁻¹. Therefore, the vibrational frequencies of NH band in DNA and OH in the fraction has changed in the complex, evidencing the effective interaction of AST OH with NH of DNA bases, possibly through H-bond interaction.

The vibrational frequency of C=O of the free DNA at 1707 cm⁻¹ shifted to 1714 cm⁻¹ in the complex. According to Alex and Dupuis [23] further indication for H-bonding interaction of AST with DNA base pairs such G-C and A-T and with phosphate groups comes from major spectral changes of DNA in-plane-vibrations in the region of 1707–1080 cm⁻¹ (Figure 6, Table 2). The band at 1707 cm⁻¹ (Gua, Thy) related to mainly Gua, 1602 cm⁻¹ (Ade, Cyt) related to mainly Ade, and 1484 cm⁻¹ (Cyt, Gua) related to mainly Cyt, shifted towards higher frequency with the change in intensity in complex at 1714, 1604 and 1488 cm⁻¹, respectively. The band at 1649 cm⁻¹ (Thy, Gua, Cyt), mainly for Thy, shifted to lower frequency at 1647 cm⁻¹. The observed changes in frequency can be related to AST interaction with Gua, Cyt, Ade and Thy bases. But there is some evidence that AST binds preferentially to Gua beyond the major shift at 1707 cm⁻¹ (Gua C6=O stretching) to 1714 cm⁻¹, where we can see also the shift of the band at 1572 cm⁻¹ (purine N7 stretching) to 1574 cm⁻¹ and shift of the band at 1532 cm⁻¹ (in-plane vibration of C≡G) to 1530 cm⁻¹.

Additional evidence for DNA-AST interaction is obtained from the spectral shifting of AST vibrational frequencies upon DNA binding. Besides the functional OH group of AST, the symmetrical and asymmetrical stretching is also essential in the binding of the aromatic ring to DNA. The frequency vibration of C=C stretching of free AST at 1653 (asymmetrical C=C stretching) and 1613 cm⁻¹ (symmetrical C=C stretching) shifted to 1686 and 1628 cm⁻¹, respectively. The out-of-plane C–H bending at 806 cm⁻¹ in AST shifted to 795 cm⁻¹ in the complex. Other minor changes (shifting/intensity variation) were also observed in AST at 1362 cm⁻¹ (in plane OH bend), 1208 cm⁻¹ (C–O stretching), 1508 and 1448 cm⁻¹ (C–C stretching (in ring)) [24].

Minor variation in PO₂ ⁻ asymmetric and symmetric vibration at 1237 to 1236 cm⁻¹ and 1098 to 1094 cm⁻¹, infrared B-marker bands at 882 cm⁻¹ (sugar-phosphate stretch) and 817 cm⁻¹ (phospho-diester mode) present in free DNA did not appreciably shift in the DNA-AST complex. This result shows that the DNA remains in the B-DNA conformation.

Deng and co-workers [9], using fluorescence and UV-vis absorption spectroscopy analyzed the interaction between Lotus leaf derived AST and DNA and computed the binding constant to K = 3.41 × 10⁴ M⁻¹ and concluded that AST could interact with DNA by intercalative binding. Their data corroborate with ours. By applying FTIR spectroscopy we could obtain the additional information that AST binding to DNA occurs preferentially with Gua-Cyt rather than Ade-Thy base pairs.

It is generally accepted that small molecules might bind to a DNA double helix by three modes: electrostatic binding, groove binding and intercalative binding [25]. Normally, small molecules have to some extent selectivity except for their binding to DNA by electrostatic interaction (along the external DNA double helix) [19]. Intercalative binding is usually affected by planarity of the bound candidates and groove binding is related to the two types of grooves in DNA, i.e., major and minor groove. In fact, each small molecule has a particular structure, which possibly has different binding modes with DNA and each mode resulting in a characteristic distortion of DNA, and this in theory giving rise to a specific pharmacological effect [26]. Indeed, in vitro studies of compound-DNA interaction allow valid predictions to the possible in vivo interactions with DNA and their cytotoxic/genotoxic consequences (Figure 2).

However, it should be stressed that we have no direct prove that AST reaches nuclear DNA. TOXNET database [27] has so far no entry related either to mutagenicity, genotoxicity nor to carcinogenicity of this compound. Our S. cerevisiae data (Figure 2) give some hint that nuclear uptake of CaRP compounds may occur. Whether it is AST or one of the metabolites cannot be answered by our experiments, although the significantly higher sensitivity of yeast DNA repair mutants and the abundance of AST (> 50%) in CaRP suggests this.

To determine its preferred binding sites on sequence of DNA [CGCGAATTCGCG (PDB ID:1D30)], AST was docked to DNA using HexServer [28]. The docking result is shown in Figure 7. The model shows that AST is surrounded by C9.A, G10.A, C11.A, C15.B, G16.B and A17.B of the oligonucleotide 1D30 (Figure 7). For the AST-1D30 complex, docking results indicated the presence of a hydrogen bond between H hydroxyl (C7) of AST and O2 of C9.A and O (C ring), which can form a hydrogen bond to both N2 of G10.A and N2 of G16.B (Figure 7). The binding energy of AST (E_total −289.30) was found to be higher than that of kaempferol (E_total −238.29) when docking with the same method/form and this result could be due to the fact that kaempferol only can form a hydrogen bond between O (C ring) and N2 of G10.A and N2 of G16.B (figure not shown).

These docking data are compatible with our FTIR results that showed intercalation of AST into DNA (mainly with GC base pairs) stabilized by hydrogen bonds.

Fresh leaves (950 g) were freeze-dried and ground to obtain 320 g of dry material, which was extracted by decoction (1:20 p/v, 80 °C per 20 min) followed by filtration, rotary evaporation and freeze-drying to obtain 80 g of crude extract. Reverse phase-solid phase extraction was performed using the column Strata C18E 5 g/20 mL Giga Tubes (Phenomenex, USA). For each batch, 1 g of crude extract was diluted in 100 mL of distilled water. The cartridge was conditioned with acetone (25 mL) and washed with 5 mL of water before loading the sample (100 mL). It was then washed with 20 mL of water and eluted with 25 mL of ethyl acetate. After evaporation this eluent yielded the CaRP refined extract. The 80 g of crude extract yielded 4 g of CaRP that was stored in desiccators at room temperature in the dark until further use.

The free radical scavenging capacity of CaRP and of reference substances was determined using 2,2-diphenyl-1-picrylhydrazyl (DPPH) method according to Brand-Willians and co-workers [29]. Five concentrations of CaRP, ascorbic acid (AA), and Trolox in methanol were used to obtain the curves (Table 1). CaRP and standards (200 μL in the range of 2.66 to 7.09 μg mL⁻¹) in methanol were added to 100 μL of DPPH (10 mg/100 mL MeOH) and allowed to stand for 30 min before absorbance was measured at 515 nm using a Molecular Devices (Sunnyvale, USA) spectrophotometer model Versamax Tunable Plate Reader. The DPPH solution (100 μL) and methanol (200 μL) were used as a negative control. This experiment was conducted in triplicate. Antioxidant activity was expressed as IC₅₀ (inhibitory concentration in μg/mL necessary to reduce the absorbance of DPPH by 50% compared to the negative control). The lower the IC₅₀, the higher was the antioxidant activity. Results were also expressed as AEAC (AA equivalent antioxidant capacity) or TEAC (Trolox equivalent antioxidant capacity) in grams and calculated as follows:

The relevant genotypes of the yeast strain used in this work are given in Table 3. Media and solutions were prepared according to [30]. Complete medium (YPD) was used for routine growth of yeast cells and minimal medium (MM) was supplemented with the appropriate amino acids and uracil to yield SC (synthetic complete medium). Stationary (STAT) phase cultures with 2 × 10⁸ cells/mL were obtained after inoculation of an isolated colony into liquid YPD and after 72 h incubation at 30 °C with aeration by shaking. Cells were washed and diluted in saline (0.9% NaCl, pH 5.0) before all treatments.

The sensitivity of different yeast strains (STAT cells of wild type [WT] and of isogenic mutants) was evaluated on freshly prepared solid SC medium containing either no drug or 12 mg of CaRP. Cells were diluted in 1:10 steps in sterile saline and 5 μL of each dilution (suspensions containing from 10⁷ to 10³ cell/mL) was placed on agar surface of SC solid medium containing the appropriate compound. Cellular growth on SC was determined after 5 days incubation at 30 °C. Photos were taken using a Canon PowerShot G10 camera. Photos represent one of at least three independent experiments with similar results.

Infrared spectra of dried fraction 11, obtained from the LC-microfractionation, were recorded on a Perkin Elmer Spectrum 400 FT-IR/FT-NIR spectrometer (Perkin Elmer, UK). Absorbance spectra were acquired at 4 cm⁻¹ resolution and signal-averaged over 8 scans. For DNA interaction analysis, the fraction 11 (F11) was mixed with DNA separately in 1:2 ratio in BPES buffer pH 7.0 in sterile Eppendorf tubes; after a thorough mixing and overnight incubation at 37 °C, the samples were concentrated by speed-vac centrifugation (Eppendorf concentrator mod. 5301). DNA-F11 complexes were prepared and studied in solid-state using KBr pellet according to Usha and co-workers [22]. Spectra were processed using the SPECTRUM software from Perkin-Elmer. The de-convoluted parameters were set with a gamma value of 8.0 and a smoothing length of 90%.

The crystal structure of DNA-ligand complex was selected from Protein Data Bank PDB ID: 1D30 [32]. We docked AST onto the oligonucleotide extracted from the crystal structure. To determine the preferred binding sites on DNA and the binding energy of AST, docking studies were performed using HexServer [28]. HexServer is a fast Fourier transform (FFT)-based protein docking server to be powered by graphics processors. FFT-based approaches assume that the proteins (receptor) to be docked are rigid, but they sample densely all possible rigid-body orientations in the 6D search space [26]. The ligand structure was extracted from Pubchem (SDF: CID 5282102) file. To display docking results, we used UCSF Chimera 1.5.3 [33].