Verbascum nigrum: Cytotoxicity Evaluation in A431 Epidermoid Carcinoma Cells and Untargeted LC‐HR‐MS/MS Metabolite Profiling

The crude methanolic extract obtained from Verbascum nigrum aerial parts (VNE) and its six fractions (VNF1–VNF6) were initially screened regarding their effects on the viability of immortalized HaCaT keratinocytes and A431 epidermoid carcinoma cells (MTT assay, 24 h). None of the tested samples affected the viability of HaCaT cells in a concentration range of 25–150 μg/mL. VNE and VNF4 exhibited significant cytotoxic effects in A431 cells, with IC50 values of 81.92 and 12.27 μg/mL, respectively; the selectivity index was higher than 10 for VNF4. The untargeted LC/HR‐MS/MS metabolite profiling led to the tentative annotation of a total number of 23 compounds. Of these, VNE comprised mainly iridoid glycosides (harpagoside, laterioside, acylated aucubin derivatives), whereas VNF4 showed a high abundance of triterpene saponin glycosides (ilwensisaponins A and C, songarosaponins A and B), constituents known for their selective cytotoxic potential.


Introduction
Native to Europe and Asia, Verbascum genus (Scrophulariaceae) comprises more than 360 species distributed not only in the two continents of origin, but also in North America and Northeast Africa. [1] The roots, leaves and flowers collected from different species have been empirically used since Ancient times in the treatment of a wide range of human ailments. [2,3] For instance, the aerial parts of Verbascum nigrum L. (dark mullein) are traditionally or homeopathically recommended in epilepsy, asthma, acute respiratory infections, gastralgia, diarrhea, dysentery or renal dysfunctions. [4,5] Regarding previous chemico-biological investigations on this species, the literature is considerably scarce. Several extracts of black mullein were documented to possess anti-inflammatory, [6] anticoagulant, antihypotensive and diuretic effects. [4] These bioactivities might be correlated with the presence of iridoid glycosides (harpagoside, laterioside, aucubin, sinuatol, nigrosides I -V), phenylethanoid glycosides (verbascoside), flavonoids and saponin glycosides (ilwensisaponins A and C) as the main phytochemicals reported in the flowers, leaves or roots of V. nigrum. [5,7,8] Squamous cell carcinoma, the second most common skin cancer (20 % of keratinocyte carcinomas), is associated with a poor prognosis as it easily metastasizes. Both systemic and topical chemopreventive drugs (retinoids, nicotinamide, non-steroidal anti-inflammatory drugs and 5-fluorouracil alone or in combination with calcipotriol, respectively) have limitations with respect to tolerability. [9,10] Therefore, identification of novel prophylactic and therapeutic strategies is of great significance. As part of our continuous research interest in finding new phytochemicals that could be used as potential leads for cancer chemoprevention and therapy, [2,11,12] some preliminary investigations on V. nigrum, a less explored species of mullein, were performed. [13] Consequently, the current work investigated the in vitro cytotoxic activity of a crude methanolic extract obtained from V. nigrum and its six fractions in human skin squamous cell carcinoma A431 cells. Furthermore, to potentially correlate the observed biological outcomes with the phytochemical composition, the systematic metabolite profiling of the active samples by liquid chromatography hyphenated with tandem high-resolution mass spectrometry (LC/HR-MS/MS) was performed. The indepth LC/HR-MS/MS dereplication strategies for the fast metabolite annotation was already established by our group for other Verbascum species, such as V. blattaria L. and V. ovalifolium Donn ex Sims, [1,2] but not for V. nigrum.

Cytotoxic Activity
The crude extract of V. nigrum aerial parts (VNE) and its six fractions (VNF1 -VNF6) obtained by vacuum liquid chromatography (Scheme 1) were initially investigated for their cytotoxic effects by MTT assay (24 h incubation) in spontaneously transformed immortalized HaCaT keratinocytes and A431 epidermoid carcinoma cells. As depicted in Figure 1a, none of the tested samples induced a significant reduction of HaCaT cell viability as compared to vehicle control. A similar trend was observed in A431 tumor cells for VNF1-3 and VNF5-6. However, VNE exhibited impor-tant cytotoxic effects of 47 % and 58 %, especially at higher concentrations (Figure 1b), with IC 50 = 81.92 μg/ mL. In contrast to this, VNF4 reduced the viability of A431 cells by more than 95 % over the concentration range of 25-150 μg/mL. As compared to this, the positive control (resveratrol 65 μM) produced cytotoxic effects of only 90 %. Successive dilutions were made in order to obtain the IC 50 value of 12.27 μg/mL for VNF4. Taking into consideration that the IC 50 value for VNF4 in the non-tumor HaCaT cells would be > 150 μg/mL, it can be assumed that this fraction exhibited highly selective cytotoxic effects [the selectivity index (SI), calculated by dividing the IC 50 value in the non-tumor cells to that in the tumor cells, would be > 10; samples with SI > 3 are considered highly selective]. [10] Previ-Scheme 1. Verbascum nigrum aerial parts: extraction and fractionation. Chem. Biodiversity 2020, 17, e2000644 ously we have shown that a crude methanolic extract obtained from the aerial parts of V. ovalifolium and its liquid-liquid partitioning fractions exhibited a concentration-dependent reduction of malignant melanoma SK-MEL-2 cell viability over the concentration range of 25 -200 μg/mL, with butanol (IC 50 = 77.98 μg/mL), hexane (IC 50 = 144.02 μg/mL) and ethyl acetate (IC 50 = 189.31 μg/mL) fractions as the most active samples. [2] In another study, [14] the ethanol, chloroform, hexane and methanolic flower extracts of V. sinaiticum Benth. showed IC 50 values in breast adenocarcinoma MCF-7 cells of 87.30, 95.43 and 119.04 μg/mL, respectively. Moreover, the hydroethanolic leaf extract of the same species exerted a selective cytotoxicity in hepatocellular carcinoma HepG2 cells (IC 50 = 80.6 μg/mL) vs. pulmonary MRC-5 fibroblast cells (IC 50 > 500 μg/mL). [15] Various plant extracts have been previously investigated for their cytotoxic potential against A431 human squamous carcinoma cells. Vitis vinifera seed and peel aqueous extracts inhibited the growth of A431 cells with IC 50 values of 111.11 and 319.14 μg/ mL, respectively, showing no cytotoxicity in HaCaT cells. [16] Another study reported a lower cytotoxicity of grape seed extract against A431 cells (IC 50 = 480 μg/ mL). [17] Methanolic extracts of Catharanthus roseus and Calystegia sepium leaves, methanolic and ethanolic extracts of Narcissus tazetta bulbs reduced the viability of A431 cells with IC 50 values higher than 50 but not exceeding 150 μg/mL. Only the methanolic extract of Calystegia sepium leaves showed less cytotoxicity in normal cells, namely human gingival fibroblast (HGF-1) cells (IC 50 > 450 μg/mL). [18] Ethanol extracts of Helichrysum odoratissimum (mixture of leaves and stems) and Vanilla planifolia (leaves) were found to be cytotoxic against A431 cells showing IC 50 values of 15.5 and 31.2 μg/mL, respectively. [19,20] Moreover, H. odoratissimum extract showed a SI of 2.39, expressing its selectivity against A431 cells vs. the non-tumor human embryonic kidney (HEK-293) cells. [19] Extracts of different polarities obtained from Cissus quadrangularis stems were also cytotoxic with the acetone extract and its purified fraction being the most active (GI 50 = 8 and 4.8 μg/mL, respectively, with GI 50 standing for growth inhibition 50 % value). [21] As compared to these plant extracts, VNF4, showing an IC 50 value of 12.27 μg/mL in A431 cells and higher than 150 μg/mL in non-tumor HaCaT cells with SI > 10, is undoubtedly a promising candidate for the development of novel therapies against cutaneous squamous cell carcinoma.
Next, the cytotoxicity of VNE and its fractions was evaluated by quantifying intracellular LDH released to the cell culture supernatant (marker of cellular dam-age) in A431 tumor cells. As compared to the negative control (untreated cells), only VNF4 produced significant cytotoxic effects (IC 50 = 28.83 μg/mL), whereas VNE and all the other fractions were considered inactive (IC 50 > 150 μg/mL). At concentrations > 50 μg/ mL, the LDH release induced by VNF4 was comparable to the positive control, in which the cell lysis solution was added (Figure 1c).

LC/HR-MS/MS Metabolite Profiling
To find the possible correlations between the observed cytotoxic effects exhibited by VNE and VNF4 in A431 cells and their phytochemical composition, untargeted LC/HR-MS/MS metabolite profiling was next performed. It was noticed that VNE comprised mainly iridoid glycosides, whereas VNF4 showed a high abundance of saponin glycosides ( Table 1) Table 1).  Two phenolic acid derivatives, namely caffeic acid hexoside (1) and chlorogenic acid (2) were tentatively annotated. [1,2] Next, a total number of 12 iridoid glycosides were identified in the analyzed samples, with ten acylated aucubin-type iridoid diglycosides. The molecular formulas of these distinguishable derivatives were suggested from the HRMS data (with mass errors < 5.0 ppm), whereas several diagnostic ion fragments were observed in their MS/MS spectra. For instance, the sequential neutral elimination of glucosyl/glucose (162/180 Da) and aucubigenyl/dehydroaucubigenyl (166/148 Da) afforded ion fragments composed of the acyl unit(s) attached to the rhamnose group. Further intramolecular cleavage of the latter group would eventually give rise to the deprotonated cinnamic acid-derived ions [caffeate, p-coumarate, (iso) ferulate, cinnamate]. These latter ions can suffer dehydration or decarboxylation, generating specific smaller fragment ions. Furthermore, the ion fragment with m/z 165.0485 was essential for the diagnostic of the aucubigenin nature of the aglycon part of these derivatives. Consequently, four monoacyl aucubintype iridoid diglycosides (3, 6, 7 and 12) and six diacyl aucubin-type iridoid diglycosides (9, 10, 13, 16, 18 and 19) were tentatively labeled in the two extracts ( Table 1). Furthermore, the MS/MS fragmentation pathways for compound 3 were proposed in Figure 4, together with general structure of the ten aucubin derivatives. Aucubin-type iridoid derivatives have been previously reported in V. nigrum, with nigrosides I  [7,8] Also, harpagoside (14) and laterioside (15) were observed in VNE. Harpagoside was identified by comparing its chromatographic and spectral data with those of a reference compound, whereas laterioside was tentatively labeled based on a previous report. [2] Both harpagoside and laterioside are acylated iridoid monoglycosides, with cinnamoyl as their acyl group, being already reported in V. nigrum. [7,22] It is worth mentioning that, from the 12 iridoid glycosides, only a limited number (3, 7, 12 and 18) was spotted in VNF4; this could be correlated with the relatively polar nature of these derivatives which gives them a lower affinity for high concentrations of organic solvents (fraction VNF4 was eluted with 100 % methanol, Scheme 1).
Four phenylethanoid glycosides were identified in the analyzed samples. Verbascoside was unequivocally assigned as peak 5 by comparing its retention time, HR-MS and MS/MS spectra with those of a commercially available standard. Having the same suggested molecular formula and fragment ions, peak 8 was ascribed as isoverbascoside. Lastly, compounds 4 ([MÀ H] À at m/z 755.2281) and 11 ([MÀ H] À at m/z 651.2294) were putatively considered as forsythoside B and martynoside, respectively. [1,2] Verbascoside and forsythoside B were previously described in V. nigrum, [22] whereas isoverbascoside and martynoside are very well-known constituents of Verbascum genus. [23] A number of five oleanane-type triterpene tetradesmosidic saponin glycosides, namely buddlejasaponin I, songarosaponins A and B, ilwensisasaponins A and C were tentatively identified in VNF4, with only one of these present in VNE ( Table 1). Their MS dereplication was performed as previously described by our group. [ These spectral data were in agreement with those reported by Hartleb and Seifert [24] for the structure of buddlejasaponin I. The same sequential neutral losses of rhamnosyl, two glucosyl and fucosyl groups were observed in the case of peaks 20 (m/z 1089.5803), 21 (m/z 1103.5944), 22 (m/z 1071.5714) and 23 (m/z 1071.5733) leading to proposing their tentative structures as songarosaponin B, ilwensisaponin C, songarosaponin A and ilwensisaponin A, respectively. [25] Ilwensisaponins A and C were previously reported in V. nigrum, [5] whereas the other saponins might be new for V. nigrum, being formerly shown in V. songaricum Schrenk. [24,26] The fact that VNF4 showed higher cytotoxic effects in A431 tumor cells could be attributed to the predominance of saponins, as compared to VNE in which iridoids and phenylethanoids were found its main constituents. For instance, ilwensisaponins A and C were shown to reduce the viability of human myelomonocytic HL-60 cells, with IC 50 values of 4 and 33 μg/mL, respectively, after an incubation time of 18 h. On the other hand, verbascoside was also tested under the same conditions, but its IC 50 value was > 62.5 μg/mL. [27] Furthermore, in another study, buddlejasaponin IV and songarosaponin D decreased the viability of human renal cell adenocarcinoma ACHN (IC 50 = 7.9 and 32.7 μM, respectively), malignant melanoma A375 (IC 50 = 14.6 and 12.4 μM, respectively), amelanotic melanoma C32 (IC 50 = 26.9 and 14.0 μM, respectively), lung carcinoma A549 (IC 50 = 38.3 and 41.1 μM, respectively), lung large cell carcinoma COR-L23 (IC 50 = 0.4 μM for both saponins) and colon adenocarcinoma Caco-2 (IC 50 = 40.7 and 16.7 μM, respectively) cells, after 48 h exposure. Nevertheless, the viability of human skin fibroblast 142BR cells was not affected (IC 50 values > 50 μM); this indicates the selective cytotoxic profile of Verbascum saponins, [28] which is in agreement with the highly selective cytotoxicity exhibited by VNF4, a saponin-enriched fraction of V. nigrum.

Conclusions
In this study, a crude methanolic extract obtained from the aerial parts of V. nigrum and its six fractions were evaluated for their effects in immortalized HaCaT cells and tumor A431 cells by MTT and LDH assay. Since the crude extract (VNE) and VNF4 showed the highest cytotoxic effects, their metabolite profiling by LC/HR-MS/MS was subsequently performed. The aim was to putatively identify the constituents that could be responsible for the observed bioactivities. The analysis revealed that VNE comprised mainly iridoid glycosides, whereas VNF4 showed a high abundance of saponin glycosides, suggesting that the latter category of specialized metabolites might be endowed with an important selective cytotoxicity as compared to other types of constituents present in V. nigrum. To conclude, fraction VNF4 is a good candidate for isolation of phytochemicals with selective cytotoxicity against human skin squamous cell carcinoma A431 cells.

Extraction and Fractionation
For extraction and fractionation, we used a procedure described in a previous work [13] with some modifications. The air-dried aerial parts of Verbascum nigrum were powdered using a mill and then, 100 g were extracted with methanol at room temperature (1000 mL, 48 h). The liquid extract was then filtered and evaporated to dryness in a rotary vacuum evaporator at temperatures below 40°C, affording the crude methanolic extract (VNCE, 17.7 g). Next, 15.0 g of VNCE were suspended in water (300 mL) and partitioned between water and diethyl ether (5 × 300 mL). After concentration and lyophilization, the aqueous phase yielded a defatted extract (VNE, 10.0 g). 5 g of this extract were subjected to vacuum liquid chromatography on C 18 Polygoprep column (50-60 μm, 2 × 40 cm, Macherey-Nagel) and eluted at a flow-rate of 10 mL/min, with 1500 mL water, 3650 mL water/methanol 1/1 (v/v) and 3650 mL methanol. Fractions of 50 mL were collected and they were pooled together based on their TLC profile: Silica gel 60 F 254 (Merck) as stationary phase; ethyl acetate/ methyl ethyl ketone/water/formic acid 50 : 30 : 10 : 10 as mobile phase; derivatization with anisaldehyde-sulfuric acid reagent (105°C, 5 min) and visualization under normal light. Six fractions (Scheme 1) were eventually pooled, as follows: VNF1 (Frs.

Cell Lines
Human immortalized keratinocyte HaCaT cells (obtained from Cell Line Service, CLS, Heidelberg, Germany) and human epidermoid carcinoma A431 cells (ATCC CRL-1555) were grown in DMEM/F12 medium, supplemented with 1 % GlutaMAX™ and 10 % FBS. Cells were cultured at a density of 10 5 cells/mL, 37°C, in a 5 % CO 2 atmosphere, with 95 % humidity. When cells reached 70-80 % confluence, they were harvested using trypsin-EDTA and DPBS. Cell counting was performed by staining with trypan blue solution and visualization on a hemocytometer under a microscope.

MTT Assay
Cell viability was measured using a tetrazolium-based colorimetric assay (MTT assay). [13,29] Briefly, cells were seeded in 96-well plates (10 4 cells/well), allowed to grow for 24 h and then treated with different concentrations of VNE and its fractions (VNF1 -VNF6) for 24 h. At the end of the incubation time, MTT (10 μL/100 μL) was added to each well and cells were further incubated for 4 h; the medium was then removed and DMSO (150 μL/well) was added for 10 min to solubilize the formazan crystals. The absorbance was recorded at 570 nm, using with a BioTek™ microplate reader. Results were expressed as: cell viability % ð Þ ¼

Absorbance of cells treated with extracts
Absorbance of untreated cells � 100 and as half maximal inhibitory concentration (IC 50 ) determined by a sigmoidal fit with Boltzmann function in OriginPro2020 software.

LDH Assay
The levels of lactate dehydrogenase (LDH) were evaluated in the culture medium, using CytoTox 96® assay kit (Promega AG, Dübendorf, Switzerland). The assay was performed according to the manufacturer's instructions, with cells being incubated with VNE and its VNF4 fraction for 24 h. Results were expressed as:  . The flow-rate was 0.2 mL/min and the injection volume was 10 μL. The operating ESI-Q-TOF parameters in negative ionization mode were: mass range m/z 100-1700, gas temperature 350°C, gas flow 12 L/min, nebulizer 40 psi, capillary voltage À 4000 V, fragmentor 120 V and fixed collision energy 40 V. Data were analyzed with a Mass Hunter Qualitative Navigator B.08.00 software.

Statistical Analysis
Values are presented as means � standard error of mean (SEM) of three individual experiments, each with eight replicates. Statistical analysis was performed by comparing treated samples with untreated controls. The statistical significance was evaluated using Student's t-test for unpaired samples, with p < 0.05 considered statistically significant. All statistical tests were conducted using OriginPro2020.