Comparative studies of two vegetal extracts from Stokesia laevis and Geranium pratense: polyphenol profile, cytotoxic effect and antiproliferative activity

Abstract In this study, two ethanolic extracts, from Stokesia aster (Slae26) and Geranium pratense (Gpre36) respectively, were evaluated in order to assess the cytotoxic activity and potential antiproliferative activity upon the nontumorigenic human epithelial cell line derived from the mammary gland (MCF-12A) and the human breast tumor cell line (BT-20). The selection of the plant species was done on the basis of their chemical composition, specifically combinations of luteolin derivatives with caffeic and gallic acid derivatives. Therefore, the S. laevis ethanolic extract proved its capacity to inhibit the viability of both normal and tumor breast cell lines (i.e., up to 90% cell viability inhibition, IC50 = 42 µg/mL). On the contrary, the G. pratense ethanolic extract proved weak stimulatory effects on the viability of the two human breast cell lines studied. The obtained results were discussed in the contexts of computational studies and drug-likeness bioactivity of seven common luteolin derivatives: luteolin, luteolin-7-O-glucoside/cynaroside, luteolin-5-O-glucoside/galuteolin, luteolin-6-C-glucoside/isoorientin, luteolin-8-C-glucoside/orientin, luteolin-3′,4′-di-O-glucoside and luteolin-7,3′-di-O-glucoside. Computational studies have revealed that the hydrophilic behavior of luteolin derivatives (log P values) does not follow other tested parameters (e.g., polar surface area values), possibly explaining different efficacy concerning the biological properties in vitro. These predictions could be a starting point for studies on the biochemical mechanism by which luteolin derivatives induce biological effects. Graphical Abstract The localization of frontier molecular orbitals (FMOs) for LUTEOLIN: HOMO (the highest occupied molecular orbital) and LUMO (the lowest unoccupied molecular orbital).


Introduction
Scientific data of recent years indicate a greater attention to luteolin derivatives, especially as a result of in vitro evidence attesting their inhibitory potential upon breast (cancer) cell line development.
In this context, two ethanolic extracts found to contain luteolin derivatives aside caffeic and gallic/ellagic acid derivatives, from Stokesia laevis and Geranium pratense plant species, were studied concerning their in vitro cytotoxic activity and potential antiproliferative activity of two cell lines: a nontumorigenic epithelial cell line derived from mammary gland (MCF-12A) and a human breast tumor cell line . The results were discussed in the context of computational studies and drug-likeness bioactivity of some of the most common luteolin derivatives usually found in plant species.
Regarding the two plant species selected, S. laevis (J. Hill; family Asteraceae), commonly known as "Stokes' aster", is a perennial species native to the United States (southeastern region) and mainly used as a decorative plant [1]. In the 1980s, there was an increased economical interest on this plant species as a result of Stokes' aster achenes content in vernolic acid, which is used in the chemical industry [2,3]; vernolic acid is an epoxide-type structured compound acting as a nonvolatile solvent useful for manufacturing oil-based paints, varnishes, adhesives and other industrial coatings. Data on the phytochemical composition and potential biological effects of S. laevis are scarce; the only data found argue the way in which the copigment luteolin and the pigments petunidin, pelargonidin and cyanidin result in the final coloration of Stokesia aster flower heads [1]. Aster species are reported to be nontoxic to dogs, except for the woody aster, Xylorhiza orcuttii [4].
Geranium species (over 280 species are found worldwide) are particularly known for their geraniol (a volatile oil) and geraniin (a gallic/ellagic acid derivative) contents. The former has been shown to have antimicrobial [5][6][7][8] and antitumor potencies [9][10][11], while the latter shows augmented antioxidant and anti-inflammatory activities [12,13] and antitumor potency [14]. Geraniin compound also demonstrated antiviral properties on herpes simplex virus types 1 and 2 [15] and human enterovirus 71 [16] in vitro and the capacity to reduce the development of Alzheimer's disease by inhibiting β-secretase activity [17]. Geraniol has been proved to be an effective insecticidal and antimosquito compound [18]. Owing to their specific phytochemical content, the extracts from Geranium species proved to have numerous biological effects and health benefits on human. For example, G. schiedeanum extracts demonstrated hepatoprotective properties against ethanol-induced hepatic injuries [19]; G. robertianum and G. molle extracts showed toxicity against breast, lung, cervical and hepatocellular carcinomas [20,21]; G. robertianum demonstrated strong antioxidant and anti-inflammatory properties [22]; and G. ayavacense indicated antidiabetic activity [23]; while G. carolinianum demonstrated antiviral properties on hepatitis B virus [24].
G. pratense L. plant species (commonly known as "meadow cranesbill") are proved to be effective against free radical-induced impairment of endothelium-dependent relaxation in isolated rat aorta [25]. The methanol extract from the roots of G. pratense decreased the area with scab lesions on potato tubers; its impact against the pathogen Streptomyces scabies has been attributed to the antimicrobial activity of the geraniin compound [26]. The aqueous extract from the aerial part of G. pratense subsp. finitimum (Woronow) Knuth significantly inhibited the formation of carrageenan-induced hind paw edema (dose of 100 mg/kg) and also proved to have antinociceptive activity [27]. Finally, the Encyclopedia of Traditional Chinese Medicine recommends the utilization of the aerial parts of G. pratense in order to dispel wind damp, free channels and network vessels and treat dysentery and diarrhea [28] The two plant materials (the aerial part, herba et flores) were shade dried, ground to medium-size powder and then used in technological studies.

Vegetal extract preparation
S. laevis plant powder of 50 g was extracted with 1,000 mL of 70% (v/v) ethanol, after 1 h at the reflux temperature. After paper filtering, 750 mL of the whole ethanolic extract with a content of 0.70 mg gallic acid equivalent (GAE) per 1 mL extract was obtained. Furthermore, 300 mL of the whole ethanolic extract was concentrated at the residue, which was dissolved into 42 mL of 40% (v/v) ethanol, and then filtered on a glass fiber filter. The resultant extract, codified as Slae26, presents as a homogeneous brown liquid having a content of 5 mg GAE per 1 mL sample.
Similarly, 50 g of G. pratense plant powder was extracted with 500 mL of 70% (v/v) ethanol solvent after 1 h at reflux temperature. The whole ethanolic extract of 330 mL with a content of 3.50 mg GAE per 1 mL extract was obtained. The whole ethanolic extract of 100 mL was further concentrated as the residue, which was dissolved into 70 mL of 40% (v/v) ethanol and then filtered on the glass fiber filter. The resultant extract, codified as Gpre36, presents as a homogeneous light brown liquid with the exact content of 5 mg GAE per 1 mL sample.
The standardized extracts, Slae26 and Gpre36, were used in pharmacological studies. Figure 1 shows the general method for obtaining the two standardized extracts.
(first author of this study) [32].  [34]. The MTS test allows evaluation of either the cytotoxic effect or the antiproliferative effect. For the (anti) proliferation assay, the application of the modulating factor (in this case the two vegetal extracts, Slae26 and Gpre36) is done on "sub-confluent" cell cultures (about 30%), and the determination of the cell population is performed after at least one division cycle (population doubling interval); for the cytotoxicity assay, the cells are exposed to "semi-confluent" cultures (about 70%) and the activity is measured for a shorter time period, rather than a doubling time (usually 6-12 h). Briefly, each test vegetal extract and dilution point series were prepared as triplicates and compared to a control sample with identical concentration of test vegetal solvent sample, at 40% (v/v) ethanol solution. After 20 h (cytotoxicity test) and 20 and 44 h (antiproliferative test) of exposure, the culture medium was removed. After another 2 h of incubation with the MTS solution, the viability of the adherent cells was determined via CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega, USA). The absorbance of the treated and control samples was measured at 490 nm with a Microplate Reader (Chameleon V Plate Reader; LKB Instruments, Australia) and the recorded values were used for cell viability estimation (see formula below). The results are calculated as mean ± SD, n = 3:

Instruments used
The instruments used for chemical analyses included the UV/Vis spectrophotometer (Hélios γ; Thermo Electron Corporation, UK) and Linomat5 TLC visualizer (CAMAG, Muttenz, Switzerland). Pharmacological in vitro cell proliferation tests were done by xCELLigence ® DP Real-Time Cell Analysis (ACEA Biosciences, USA). In Figure 3, chromatograms A, B and C present the polyphenol profile of the G. pratense ethanolic extract i.e., Gpre36 test sample, compared to several reference compounds (tracks T1, T3, T4 and T5). System A setting study of the Gpre36 extract (chromatogram A, tracks T2) reveals an important number of polyphenol compounds, namely, luteolin (yellow fl., spots s5, s8), ellagic (light blue fl., spots s1, s2 and s4) and caffeic acid (blue fl., spots s7 and s10) derivatives and at least one kaempferol derivative (blue-green fl., spot s11).
System B setting study of the Gpre36 test sample, before and after hydrolysis in 4 N HCl medium (chromatogram B, tracks T2 and chromatogram C, tracks T2H) confirmed the presence of gallic acid aglycone (intense blue-indigo fl. spot s7, Rf ∼ 0.55) and condensed derivatives, ellagic acids (intense blue-indigo fl. spot s1, s2 and s4); chromatogram C also revealed augmented quantities of luteolin (yellow fl. spot s8, Rf ∼ 0.75) and kaempferol (green-blue fl. spot s9, Rf ∼ 86) aglycones as well as low quantities of caffeic acid aglycone (blue fl. spot s8′ covered by luteolin s8). Other high-performance liquid chromatography studies [25]   By comparison, qualitative (HP)TLC studies revealed that the ethanolic extract from S. aster plant species, caffeic acid/s7, chlorogenic acid/s3 and one luteolin monoglycoside (likely luteolin-7-O-glucoside/s4) were found, and the ethanolic extract from G. pratense plant species confirmed numerous gallic/ellagic acid derivatives/s1, s2, s4 and one augmented luteolin polyglycoside/s5. Concerning the quantitative aspects, the two ethanolic extracts, Slae26 and Gpra36, were made in a manner to assure the exact content of 5 mg total phenol content, GAEs per 1 mL sample.

Cytotoxic and antiproliferative potential assays in vitro
In vitro cytotoxicity and antiproliferative assessments were done on Slae26 and Gpre36 test vegetal extracts (characterized by 5 mg GAE/mL sample) by using six dilution series, that is, 1, 5, 10, 25, 50 and 100 µg GAE/ mL samples. The two test vegetal dilution series and the corresponding control sample dilution series (40% ethanol solvent series) were applied on "semi-confluent" and "sub-confluent" mammary gland cell culture MCF-12A and human breast tumor cell culture BT-20, as described in Section 2. Due to the combination of luteolin derivatives with gallic or ellagic acid derivatives, which were shown to have strong antitumor, antimetastatic and antiangiogenesis activities [43,44], it is expected that the G. pratense ethanolic extract (Gpre36) has a strong antiproliferative potential. However, when applied on the normal mammary gland cell line MCF-12A (Figure 4a), it indicated a weak stimulatory activity in the cytotoxicity test (up to 15% cell viability stimulation); during the antiproliferative test, at first stimulatory effects were noticed (up to 50% cell viability stimulation at 24 h), followed by inhibitory effects (up to 8% cell viability inhibition at 48 h). Gpre36 test sample tested on human breast tumor cell line BT-20 (Figure 4b) indicated weak cytotoxic effects (up to 16% cell viability inhibition at 24 h), and no antiproliferative activity; in fact, 7% and 8% cell viability stimulatory effects were measured at 24 and 48 h, respectively. The results are available in Supplementary Tables 1s, 2s, 3s and 4s.

Computational results
Computational study of the seven luteolin derivatives, molecular descriptors and physicochemical properties was planned in order to find a possible explanation for the differences in pharmacological activity of the twotest vegetal extracts. Although the G. pratense ethanolic  extract has a theoretically more efficient chemical composition, in vitro assessments indicated the lack of antiproliferative activity of Gpre36 versus certain antiproliferative activity of S. aster ethanolic extract, Slae26. Therefore, it appears that the specific chemical structure of luteolin derivatives (the position and the number of substituents) and less combination with caffeic or gallic/ ellagic acid derivatives lead to the final antiproliferative potential of the two vegetal extracts. Table 1 summarizes the quantum chemical parameters for luteolin and the six most common luteolin derivatives: luteolin-7-O-glucoside/cynaroside, luteolin-5-O-glucoside/galuteolin, luteolin-6-C-glucoside/isoorientin, luteolin-8-C-glucoside/orientin, luteolin-3′,4′-di-O-glucoside and luteolin-7,3′-di-O-glucoside.
First of all, it is obvious that the energy of the highest occupied molecular orbital (HOMO) orbital (E HOMO ) of all studied compounds is significantly smaller than that of the energy of the lowest unoccupied molecular orbital (LUMO) orbital (E LUMO ). Energetically, from the HOMO orbital, it is easiest to donate electrons to form new bonds to be involved in oxidation. The LUMO is energetically more able to receive electrons, possibly involved in reduction reactions.
Second, the ΔE gap parameter (the energy difference between HOMO and LUMO) providing information about chemical reactivity and kinetic stability of the molecules indicated close values, ranging between 4.14 and 4.26 eV. Table 2 presents the values obtained for physicochemical properties and molecular descriptors involved when discussing the drug-likeness and matching with the statements of Lipinski's rule of five concerning oral bioavailability. For each structure, the first line of values corresponds to Spartan computations, and on the second line are the values obtained from the Molinspiration platform. A small difference is due to the methods used for computation. Molinspiration employed the semiempirical AM1 method for the optimized 3D molecular geometries, wherein volumes and polar surface area (PSA) are approximated based on the fragment contributions. Table 2 clearly demonstrates luteolin aglycone as the most feasible drug candidate due to zero violation of Lipinski's rule [42]. Although not sufficient and mandatory to send a potential drug compound forward for pharmaceutical development, these limits of physicochemical parameters are used, rather than excluding from screening the compound with poor success rate. As observed from the results of computations, the correlation between the molecular weight and the PSA indicated that, generally, by increasing the molecular weight, from luteolin to luteolin diglucoside, the number of added hydroxyl groups and PSA increases. The biggest structures with the highest hydrophilicity, luteolin-3′,4′-diglucoside and luteolin-7,3′-di-glucoside, showed the most deviations from drug-likeness criteria (Lipinski's rule), also presenting the highest number of flexible bonds (Nrotb).
Yet, even if luteolin derivatives present a more augmented lipophilic character due to the lack of one hydroxyl group in position 3, the comparison of log P values (the calculated values of octanol-water partition coefficient log P by Spartan computations) with those   computed for quercitrin derivatives [34] indicates that (1) luteolin derivatives are more hydrophilic compounds capable of exhibiting greater oral bioavailability and (2) their log P values do not follow PSA values, possibly explaining their different efficacies concerning the biological properties and function of test conditions. The correlation between PSA, molecular volume (vol) and water-octanol partition coefficient (log P) for the seven luteolin derivatives studied ( Figure 6) indicated that luteolin, luteolin-5-O-glucoside, luteolin-6-Cglucoside and luteolin-3′,4′-O-diglucoside have significantly different log P values in comparison with luteolin-7-O-glucoside, luteolin-8-C-glucoside and luteolin-7,3′-O-diglucoside, which could support the differences between the pharmacological activity of the vegetal extracts.
In Table 3 are listed the bioactivity scores predicted using the Molinspiration software toward GPCR ligands, ion channel modulators, kinase inhibitors, nuclear receptor ligands, protease inhibitors and enzyme targets. It is known that if the bioactivity score is larger than 0.0, the drug candidate is classified as active; if score range is between −5 and 0, then the structure is moderately active and if the score is less than −5, then inactive. Therefore, luteolin exhibits larger score values for kinase inhibitor (0.26the largest of all the tested structures), nuclear receptor ligand (0.39, also the highest score, compared with its derivatives) and enzyme inhibitor  (0.28), thus being considered as the most promising structure, the lead compound. This conclusion is also supported by the results presented in Table 2, showing no deviation from the limiting criteria postulated by Lipinski's rule of five. On the contrary, luteolin-7,3′-di-O-glucoside, the most flexible molecule presenting seven rotatable bonds and three violations from Lipinski's rule (molecular weight larger than 500 Da; more than five hydrogen bond donors, more than ten hydrogen bond acceptors), shows the lowest scores for all targets. Thus, it is classified as a moderately active candidate. None of the luteolin derivatives can be considered inactive, as they all show scores larger than −5.0. Table 3 also indicates that all compounds show greater scores for enzyme inhibition, all values being positives, acting as active drug candidates. Based on this, we can declare that the most active ligand is luteolin 6-C-glucoside with 0.46 score for enzyme inhibition. Concerning protease inhibition, the most active ones are luteolin 6-C-glucoside and luteolin 8-Cglucoside (score = 0.01). Luteolin aglycone is the most active when compared with its derivatives, because it is the ligand that is capable of binding to nuclear receptor (score = 0.39) and kinase inhibitor (score = 0.26). Luteolin 5-O-glucoside and luteolin 8-C-glucoside were found to be highly bioactive toward GPCR ligand (score = 0.12). Regarding ion channel modulator bioactivity, luteolin 6-C-glucoside shows a promising score (0.01).
Considering all the above observations of the druglikeness parameters ( Table 2) and bioactivities ( Table 3), it can be concluded that all compounds show great potential in terms of bioactivity as ligands for the tested targets, starting with their lead compound, luteolin. Therefore, the interest in luteolin and luteolin derivatives in the recent years is argued. The advantages offered by luteolin derivatives as opposed to other flavonoid subclasses are drawn from their chemical structures: they are more resistant to the (auto)oxidation process explained by the lack of a hydroxyl group in position 3 of the flavan ring. They are resistant to acidic hydrolysis [32], at the same time more stable and safe and able to pass through the cell membranes more easily, while also showing great bioactivity potential.
Proving these, scientific data indicate that luteolin stimulates breast cancer cell apoptosis [45] and inhibits fatty acid synthase (FASN) activity [46][47][48][49], essential for normal and tumor breast cell growth and function. FASN is the only enzyme capable of inhibiting the novo synthesis of fatty acids with long chains (starting from acetyl-CoA, malonyl-CoA and nicotinamide adenine Moreover, due to the complex actions that compete for the final effect, luteolin aglycone was proved to be an effective inhibitor of mammary and prostate cancer metastases [50]. For instance, luteolin inhibits the angiogenesis process induced by the vascular endothelial growth factor, which reduces the chance of colonization of metastatic targets [51]; luteolin inhibits the expression of specific matrix metalloproteinases involved in basement membrane degradation, a key step in the neovascularization and tumor invasion process [52]; luteolin inhibits the secretion of interleukin (IL)-8 [53] and IL-6, acting as an amplifying signal of cancer cell growth and invasion [54]; luteolin inhibits epithelial-mesenchymal transition considered as a pinnacle step in the process of cancer cell invasion and metastasis [55]; luteolin acts as a p90 ribosomal S6 kinase inhibitor that suppresses Notch4 signaling [56]; luteolin promotes cell cycle arrest in breast cancer and inhibits the expression of human epidermal growth factor receptors HER2 [57] and EGFR/aka HER1 [58], related to the new breast cancer cases; luteolin acts as an insulin-like growth factor-mediated proliferation antagonist [59] and stimulates breast cancer cell apoptosis through extrinsic (via caspase 9) [60] and intrinsic (via caspases 8 and 10) [61] pathways, thus being the subject of numerous review articles [62][63][64].
Regarding aspects of bioavailability, luteolin derivatives appear to be generally degraded to smaller metabolites by the intestinal bacterium Eubacterium cellulosolvens within 24 h; in contrast, luteolin C-polyglycosides appear unchanged when they are absorbed in the intestine, before being distributed to the body tissues [65].
Another aspect to be mentioned is the capacity of polyphenols, especially of flavonoid subclasses, to bind iron from the medium; flavonoid C-glycosides fulfil the role of siderophores in plant tissues [66,67]. The sequestration of iron from the intestine by polyphenolbased supplements leads to several shortcomings in humans: the inhibition of heme iron absorption possibly promotes anemia [68], the intestinal inflammation and colon tumor development. Data suggest that good intestinal bacteria are significantly affected by iron depletion. This is because their metabolism is dependent on numerous iron enzymes, while pathogenic bacteria necessitate low or no levels of iron to survive and proliferate [69]. A possible way to oppose iron sequestration by polyphenol-based supplements is by the addition of Fe and EDTA in tea in a molar ratio of 1:2, to avoid iron sequestration by green tea catechins [70].

Conclusion
The present study aims to evaluate the antiproliferative potencies of two vegetal extracts combining luteolin derivatives with caffeic and, respectively, gallic/ellagic acid derivatives isolated from S. laevis and G. pratense plant species.
In vitro pharmacological studies indicated that the ethanolic extracts of S. laevis have the ability to inhibit the viability of both sub-confluent and semi-confluent nontumorigenic epithelial mammary gland cell line MCF-12A and sub-confluent and semi-confluent human breast tumor cell line BT-20. This suggests both potential toxic effects and potential antiproliferative activity of human breast (cancer) cell development. Also, the fact that Stokes' aster ethanolic extract inhibited either normal or tumor breast cell line sustains the involvement of luteolin derivatives, which proved to inhibit the activity of FASN involved in both normal and tumor mammary cell metabolism.
G. pratense ethanolic extract yielded in vitro results that rather seem to suggest stimulatory effects on normal and tumor human breast cell development.
Computational studies made on luteolin and several common luteolin derivatives conducted to a reactivity analysis and to drug-likeness and bioactivity prediction approaches. For example, it was shown that luteolin, the structure that respects all limitations imposed by Lipinski's drug-likeness criteria, is a promising drug molecule, with a potential for higher activity against several biological targets. The predictions also indicate that all of its studied derivatives could exhibit good or moderate activity toward used receptors. The fact that log P values computed for luteolin derivatives do not follow the PSA values also explains to some extent the results of in vitro pharmacological studies, and antiproliferative effectiveness of Slae26 and Gpre36.
In summary, prediction studies are key factors for developing new drugs, giving a preview of the hydrophilic-lipophilic character, absorption, transport and distribution through the physiological media of the modeled compounds; it must be noticed that the computational scores are based on experimental data collected from the medicinal chemistry literature (about 10,000 data points for every target class) [71].
Overall, for the first time in the literature, the potential antiproliferative activity of the S. aster ethanolic extract was proved on tumor human breast cell line MT-20, but further studies are required to assess the exact chemical composition and to separate more active fractions to decrease the IC 50 value. This way, the S. laevis plant species turns out to be a good source of luteolin derivatives for pharmaceutical purposes, but not for alimentary or phytomedicine purposes, since in vitro studies on the normal human epithelial mammary cell line MCT-12A indicate potential toxic effects (up to 88% cell viability inhibition) of the polar extracts from the aerial part of Stokes' aster.