Discrimination of the Essential Oils Obtained from Four Apiaceae Species Using Multivariate Analysis Based on the Chemical Compositions and Their Biological Activity

The chemical composition of the essential oils obtained from the aerial parts of four Apiaceae species, namely Elaeosticta allioides (EA), E. polycarpa (EP), Ferula clematidifolia (FC), and Hyalolaena intermedia (HI), were determined using gas chromatography. Altogether, 100 volatile metabolites representing 78.97, 81.03, 85.78, and 84.49% of the total components present in EA, EP, FC, and HI oils, respectively, were reported. allo-Ocimene (14.55%) was the major component in FC, followed by D-limonene (9.42%). However, in EA, germacrene D (16.09%) was present in a high amount, while heptanal (36.89%) was the predominant compound in HI. The gas chromatographic data were subjected to principal component analysis (PCA) to explore the correlations between these species. Fortunately, the PCA score plot could differentiate between the species and correlate Ferula to Elaeosticta species. Additionally, the antioxidant activity was evaluated in vitro using the 2,2-diphenyl-1-picryl-hydrazyl-hydrate (DPPH), 2,2-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) (ABTS), and the ferric reducing power (FRAP) assays. In addition, the antimicrobial activity using the agar diffusion method was assessed, and the minimum inhibitory concentrations (MICs) were determined. Furthermore, the cell viability MTT assay was performed to evaluate the cytotoxicity of the essential oils against hepatic (HepG-2) and cervical (HeLa) cancer cell lines. In the DPPH assay, FC exhibited the maximum activity against all the antioxidant assays with IC50 values of 19.8 and 23.0 μg/mL for the DPPH and ABTS assays, respectively. Ferula showed superior antimicrobial and cytotoxic activities as well. Finally, a partial least square regression model was constructed to predict the antioxidant capacity by utilizing the metabolite profiling data. The model showed excellent predictive ability by applying the ABTS assay.


Introduction
Central Asia, particularly its hilly regions, is one of the richest and significantly diverse areas for the growth of family Apiaceae in the world. Kazakhstan, Kyrgyzstan, activities. Finally, a partial least square (PLS) regression model was constructed to investigate the correlation between the chromatographic fingerprint and the antioxidant activity.
Monoterpenes and oxygenated monoterpenes are the predominant classes of metabolites detected in the F. clemaditifolia essential oil, where they represent 42.28 and 31.05% of the total components, respectively. Similar results could be noticed with these classes in many Ferula species. For instance, terpinolene (77.62%) was the major component in F. macrecolea [15], while β-pinene (60.84%) was the predominant metabolite in F. gummosa [16]. In F. tunetana, α-pinene (39.8%) was the main component [17] and α-thujene (13.5%) is present in the highest concentration in F. tingitana [18]. On the same context, these results were following our previous studies on other Ferula species where D-limonene was a major identified component in 6 different Ferula species [19]. These diverse metabolic profiles of Ferula, can be attributed to the wide spread of the genus in many continents and the subsequent great variability in the climatic and the soil characteristics, which have a great impact on the essential oils composition in the plants [20].
Oxygenated monoterpenes are also the main class in E. polycarpa, representing 19.7% of the detected components in its essential oil. However, oxygenated sesquiterpenes are the dominant class of metabolites in E. allioides (34.45%). Rare studies were carried out on this small genus. Hamedi et al. studied the essential oil composition of E. glaucescens [21]. The aliphatic ester constituted 52.0%, followed by the sesquiterpene hydrocarbon (6.1%), which are the major reported classes of the secondary metabolites. Conversely, aldehyde and ketones are highly predominant in H. intermedia, accounting for 53.79% of the identified components. It was the first report on the chemical composition of the volatile components in the genus.

Multivariate Data Analysis Explorative Data Analysis
A useful approach for differentiation between closely related species is applying multivariate analysis based on various chromatographic techniques [22,23]. Multivariate analysis was used based on the GC data to determine the relationships between these Apiaceae species and explore similarities and dissimilarities between them. Principal Component Analysis (PCA) was primarily used to categorize data and to find correlations between samples and variables. After that, Hierarchical Cluster Analysis (HCA) was used for sorting samples into clusters based on their vicinity to each other. The relative peak areas of all compounds in the chromatograms from the studied species were subjected to both PCA and HCA analyses.
The score and loading plots for the data set are shown in Figure 1. The PCA scores plot for principal components (PCs), namely PC1 versus PC2, is shown in Figure 1a. The total variance of 82% was explained with two PCs. Different Apiaceae species could be discriminated from each other into three main groups, where they are positioned at three different quadrants. Both F. clemaditifolia and H. intermedia were positioned in separate quadrants highlighting their obvious dissimilarity from the remaining species. However, the PCA score plot showed F. clemaditifolia closeness to genus Elaeosticta, as both were positively correlated to PC1. This correlation may be attributed to the chemotaxonomic similarity due to common components such as D-limonene, dihydro-linalool acetate, hexadecanoic acid, phytol, methyl stearate, and ethyl octadecanoate. This correlation is quite interesting as it is the first chemotaxonomic classification carried out on these two species. Most morphological and anatomical classifications typically correlate both Hyalolaena and Elaeosticta genera together, and Ferula is usually separated [24]. These results encourage researchers to perform more studies on the chemotaxonomy of Apiaceae. Moreover, E. allioides and E. polycarpa were falling together in the same upper quadrant distanced from each other (as both are from the same genus). By clear inspection of the loading plot Figure 1b, it revealed the effective variables for each PCs where allo-ocimene, heptanal, and germacrene D were the key makers accountable for the segregation of F. clemaditifolia, H. intermedia, and E. allioides, E. polycarpa, respectively. HCA clustering was performed to have a better insight into species classification. Figure 2 shows the HCA dendrogram for GC fingerprints. HCA dendrogram clustered Apiaceae species into three main clusters. Cluster I, II, and III displayed F. clemaditifolia, H. intermedia and E. allioides, E. polycarpa, respectively. The dendrogram confirmed the results obtained from PCA, revealing the similarity of E. allioides and E. polycarpa as both grouped in one cluster.

Antibacterial Activity
The disc diffusion and micro-dilution techniques were applied to investigate the antimicrobial activity of the essential oils obtained from the aerial parts of F. clemaditifolia, E. allioides, E. polycarpa, and H. intermedia. Staphylococcus aureus (ATCC 25923), and Pseudomonas aeruginosa (ATCC 85327) were used as representatives of Gram-positive and Gram-negative bacteria, respectively. The mean diameter of inhibition zones (DIZ) and the minimum inhibition concentrations (MIC) values were determined, and the results HCA clustering was performed to have a better insight into species classification. Figure 2 shows the HCA dendrogram for GC fingerprints. HCA dendrogram clustered Apiaceae species into three main clusters. Cluster I, II, and III displayed F. clemaditifolia, H. intermedia and E. allioides, E. polycarpa, respectively. The dendrogram confirmed the results obtained from PCA, revealing the similarity of E. allioides and E. polycarpa as both grouped in one cluster.
Plants 2021, 10, x FOR PEER REVIEW HCA clustering was performed to have a better insight into species classif Figure 2 shows the HCA dendrogram for GC fingerprints. HCA dendrogram clu Apiaceae species into three main clusters. Cluster I, II, and III displayed F. clemad H. intermedia and E. allioides, E. polycarpa, respectively. The dendrogram confirm results obtained from PCA, revealing the similarity of E. allioides and E. polycarpa grouped in one cluster.

Antibacterial Activity
The disc diffusion and micro-dilution techniques were applied to investig antimicrobial activity of the essential oils obtained from the aerial parts of F. clemad E. allioides, E. polycarpa, and H. intermedia. Staphylococcus aureus (ATCC 25923 Pseudomonas aeruginosa (ATCC 85327) were used as representatives of Gram-posit Gram-negative bacteria, respectively. The mean diameter of inhibition zones (DI

Antibacterial Activity
The disc diffusion and micro-dilution techniques were applied to investigate the antimicrobial activity of the essential oils obtained from the aerial parts of F. clemaditifolia, E. allioides, E. polycarpa, and H. intermedia. Staphylococcus aureus (ATCC 25923), and Pseudomonas aeruginosa (ATCC 85327) were used as representatives of Gram-positive and Gram-negative bacteria, respectively. The mean diameter of inhibition zones (DIZ) and the minimum inhibition concentrations (MIC) values were determined, and the results were presented in Table 2. Three of the studied essential oils showed significantly low antimicrobial activity against the examined bacterial strains with MIC values >500 µg/mL. However, F. clemaditifolia oil exerted promising activity comparable to other tested oils with MICs values of 125 µg/mL against the Gram-positive S. aureus strains. These results were following previously reported data for the antiinfective activity of the essential oils obtained from six other Ferula species that exhibited moderate antimicrobial activity [18,[25][26][27]. At the same time, nothing could be found in the literature related to the antibacterial activity of the other two genera Elaeosticta or Hyalolaena, and here we report for the first time about their biological activities. Although many studies deal with the exact mechanism of antibacterial activities of many essential oils and their components, it is not easy to correlate the activity to a single component. These oils are a complex mixture composed of many active ingredients that work synergistically at many targets in the microbes. Most of the volatile secondary metabolites can diffuse through the bacteria's cell membranes, leading to destructing many organelles, morphological changes, and interference of the respiratory chain, causing bacterial death [28]. This promising activity of Ferula compared to other oils may be attributed to the high content of the oxygenated monoterpenes, especially linalool, citronellol, allo-ocimene, and 4-terpineol. These compounds were reported to have a relevant bactericidal activity against a wide range of Gram-positive and Gram-negative bacteria [29][30][31][32][33]. Data are presented as means ± S.D. n = 3; * The positive control was taken as ampicillin for Gram-positive bacteria and gentamycin for Gram-negative bacteria; In all assays, 30 mg essential oil, 10 µg of the standard antibiotic in 1 mL DMSO, and 100 µL were applied. DIZ, the diameter of inhibition zone is measured in (mm) by the agar diffusion method.

Antioxidant Activity
The DPPH, ABTS, and FRAP assays were used to determine the in vitro antioxidant activity of the essential oils. The IC50 values of the antioxidant activity of the essential oils obtained from the aerial parts of F. clemaditifolia, E. allioides, E. polycarpa, and H. intermedia and ascorbic acid as a positive control were displayed in Table 3. In all the experiments used, F. clematidifolia showed superior antioxidant activity with IC50 values of 19.8 and 23.0 µg/mL for the DPPH and ABTS assays, respectively. In the FRAP assay, it exhibited a substantial reducing power of 1308.1 ± 8.9 mM Fe(II)/g. In general, measuring the antioxidant activity by in vitro experiments depends on the power of the sample to scavenge the generated free radicals produced by the system. Phenolic compounds such as acids and flavonoids are the main target secondary metabolites that can strongly suppress these active species. However, oxygenated hydrocarbons can also work in the same context, but to a lesser extent [34]. Regarding the studied samples, the absence of high content of phenolic compounds is the main cause of the moderate antioxidant activity. However, the antioxidant effect may be attributed to the whole oxygenated components of the oils. Similar results for Ferula species were explained in previous work based on the mechanism of preventing the free-radical initiation and decomposition of the peroxide radicals [35,36]. Our results are following the previously published data about the antioxidant activity of Ferula assafoetida oleo-gum-resin. Both Ferula essential oils showed promising antioxidant activity with IC50 values ranging from 10-50 µg/mL [37]. Data are presented as means ± S.D. n = 3.

Cytotoxic Activity
The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was used to assess the cytotoxic activity of the essential oils obtained from E. allioides, E. polycarpa, F. clemaditifolia, H. intermedia, and the standard drug doxorubicin against two different cancer cells (HeLa and HepG-2). Most of the evaluated oils showed moderate activity against HeLa cells with IC50 values 138.5-171.2 µg/mL. F. clemaditifolia was the most potent cytotoxic agent among these oils against both HeLa and HepG-2 with IC50 values of 138.5 and 252.2 µg/mL, respectively. However, E. allioides showed the weakest activity with IC50 values of 233.1 and 387.4 µg/mL, respectively. Nevertheless, it is important to highlight that all the tested oils showed very weak anti-proliferative effect against HepG-2 cells, with IC50 values 252.2-387.4 µg/mL compared to HeLa cells as reported in Table 4. The cytotoxic properties of F. clemaditifolia may result from the synergistic action of the different constituents, including limonene, eucalyptol, linalool, 4-terpineol, and citronellol [38][39][40][41][42][43]. These components have been reported to specifically target many vital pathways and processes inside cancer cells, including apoptosis, cell cycle arrest, membrane damage, DNA synthesis inhibition, and ABC transporter modulation [44]. This pattern of antiproliferative activity is typically noticed with Zanthoxylum essential oil. This oil has a similar metabolic profile, and the activity was referred to the ability to induce S phase arrest apoptosis and increase the expression of cleaved caspases in HaCaT cells [45].

Partial Least Square Regression Analysis
Partial least square regression (PLS-R) analysis was conducted to establish a link between the gas chromatography fingerprints and their antioxidant activities of various studied species. PLS-R model was constricted by the data matrix X containing the peak area of the GC fingerprints and the response vector y containing the antioxidant activity data. The ideal number of latent variables (LVs) in the PLS model was calculated using minimum root mean square error (RMSE) values obtained by cross-validation (CV). The performance of the model was assessed by the parameters of root mean square error of calibration (RMSEC), root mean square error of validation (RMSEV), and correlation (R 2 ). The PLS-R model parameters, including slope, offset, RMSEC, RMSEV, and R 2 , have been shown in Table 5, indicating a superior prediction ability of the PLS regression model. PLS-R models showed excellent linearity and accuracy with R 2 > 0.99, slope close to 1, intercept close to 0, low RMSEC and RMSEV (close to 0), and low differences between RMSEC and root mean square error of prediction (RMSEP) reveal the robustness of the model. In general, while employing ABTS data, the RMSEV value was less, implying that ABTS findings are more representative than other techniques used to assess antioxidant activity. The prediction performance for the developed models was shown in Table 6. The results showed that the antioxidant activity is correctly predicted with ±5% accuracy.

Essential Oil Isolation
At room temperature, the aerial parts of E. allioides, E. polycarpa, H. intermedia, and F. clematidifolia were air-dried in the shade for one week (moisture content not exceeding 10-12%). The dried plant samples (300 g each) were hydro-distilled for 2 h using a Clevenger-type apparatus, yielding 0.14, 0.12, 0.19 and 0.38% v/w, respectively, after being trapped in dichloromethane. The oils were dried by passing over anhydrous sodium sulfate and kept at −4 • C until being used for further experiments.

Gas Chromatography Analysis
Gas chromatography (GC) was performed on an Agilent 7890B gas chromatograph equipped with a DB-5Ms fused silica column (30 m × 0.25 mm, film thickness 0.25 µm, Agilent Technologies, Middelburg, The Netherlands), which interfaced with an Agilent mass selective detector 5977A (Agilent Technologies, Middelburg, The Netherlands). The interface temperature was 280 • C; the source temperature was 230 • C; ionization energy: 70 eV; and the scan range: 45-950 atomic mass units. A GC autosampler was used to inject 0.5 µL of the sample. The temperature programming was set to supply an oven temperature at 50 • C for 5 min, rising from 50 • C to 280 • C at 5 • C/min, and then finally held isothermally at 280 • C for 15 min. The injector temperature was 250 • C; detector temperature 270 • C; carrier gas helium (0.9 mL/min); and split mode (split ratio, 1:20). The retention indices were calculated based on the retention times of the standard alkane series (C7-C40) purchased from Sigma-Aldrich (Sigma Aldrich GmbH, Sternheim, Germany). Enhanced ChemStation software, version MSD F.01.01.2317 (Agilent Technologies, Middelburg, The Netherlands), was used for recording and integrating the chromatograms. Quantitation was carried out based on the normalization method using the reading of three chromatographic runs. The compounds were identified by comparison of their mass spectral data and their retention indices (RIs) with those reported in the Wiley Registry of Mass Spectral Data (9th Ed.), NIST Mass Spectral Library (2011), and references [46,47].

Antibacterial Assay
The antimicrobial activity was determined using two commonly known standard bacterial strains: Staphylococcus aureus (ATCC 25923) as an example of Gram-positive bacteria and Pseudomonas aeruginosa (ATCC 85327) as representative of Gram-negative microbes. The antibacterial activity of the essential oils was evaluated using modified agardisc diffusion and broth micro-dilution techniques (MIC) that were reported previously in details. Briefly, the suspensions of the utilized microorganisms were prepared to a final concentration of approximately 1 × 10 6 CFU/mL, that followed by inoculating the Mueller Hinton agar (Biomerieux, Marcy l'Étoile, France) with the pathogens. Wells with a diameter of 6 mm were obtained and loaded with 100 µL of 30 mg/mL of essential oil. DMSO, ampicillin (10 µg/mL), and gentamycin (10 µg/mL) were used as controls. The diameters of the growth inhibition zones were determined in triplicate after incubation at 37 • C for 24 h. The MICs of the essential oils were determined by micro-dilution method where the oils were dissolved in 5% DMSO and then were diluted with the broth in 96-well plates to obtain (4-0.007 mg/mL) concentrations. The adjusted bacterial suspension concentrations were then added and the plates were incubated at 37 • C for 24 h (bacteria). The concentration in the first well showed no visible turbidity matching, with a negative control defined as the MIC. Each test was performed in triplicate [48].

Antioxidant Assay
The 2,2-diphenyl-1-picryl-hydrazyl-hydrate (DPPH), 2,2-azino-bis(3-ethylbenzothiazo line-6-sulphonic acid) (ABTS) radical scavenging, and ferric reducing power (FRAP) assays were adopted to determine the in vitro antioxidant activity of the essential oils. The methods were described with all details previously by the same team [49]. In the DPPH assay, the methanol solutions (1 mL) containing 1-500 µg/mL of the essential oils was mixed with 4 mL of a 0.004% methanol solution of DPPH, and the absorbance was measured at 517 nm after incubation in the dark for 30 min at room temperature, while in ABTS assay the diluted methanol essential oils solutions (1 mL) were mixed with freshly prepared ABTS solution (2 mL). The sample absorbance was read at 734 nm after a 30 min incubation at room temperature. However, in FRAP assay, sample solution (0.1 mL) was added to premixed FRAP reagent (2 mL) containing acetate buffer, 2,4,6-tris(2-pyridyl)-s-triazine in HCl and ferricchloride in a ratio of 10:1:1 (v/v/v). Then, the absorbance was detected at 593 nm after a 30 min incubation at room temperature. Citric acid was used as a positive control [50].

Cytotoxicity Assay
The 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was used to assess the cytotoxic activity of the essential oils obtained from E. allioides, E. polycarpa, F. clemaditifolia, H. intermedia, and the standard drug doxorubicin against two different cancer cells; HeLa and HepG-2. All the details are mentioned in previous work by the team. The cells (2 × 10 4 cells per well) were seeded in 96-well plates, and incubated for 24 h. The stock solution (100 mg/mL) of essential oils in DMSO was diluted with the media, and each sample was incubated with cells for 24 h. The MTT solution (0.5 mg/mL) was added and incubated for 4 h. The DMSO (100 µL) was used to dissolve the formazan crystals. The absorbance was measured at 570 nm. Experiments were performed in triplicate [40,[51][52][53].

Statistical Analysis
Unless otherwise specified in the technique, all tests were repeated three times. Continuous variables were expressed as mean ± SD. The IC50 value was defined as the concentration of the substance that resulted in a 50% decrease or inhibition of biological activity. A one-way analysis of variance (ANOVA) was used to determine statistical significance, followed by a Tukey's post hoc test, with a significance level of p < 0.05.

Multivariate Analysis
GC-MS data were subjected to multivariate analysis. Principal component analysis (PCA) is the first step in studying the data to offer an overview of all observations and samples to detect and analyse groups, trends, and significant outliers. Hierarchical cluster analysis (HCA) was then employed to enable sample clustering. The clustering patterns were built adopting the entire linkage approach used for group formation; this presentation is more efficient when the Euclidean approach calculates the distance between samples (points). A quantitative calibration model, partial least squares (PLS), was designed to develop a linear relationship between the GC-MS peak area (X) matrix and antioxidant activity (Y) matrix. In this situation, there was no partition of data into model and test set as just 12 samples were assessed (small data set). Therefore, predicting antioxidant activity for fresh samples was not our main concern. The root mean square error (RMSE) and correlation coefficient assessed the PLS-R model capability (R2). PCA, HCA, and PLS were accomplished using CAMO's Unscrambler ® X 10.4 software (Computer-Aided Modeling, As, Norway).

Conclusions
In this study, the chemical profiles, antimicrobial, cytotoxic, and antioxidant activities of four essential oils obtained from the aerial parts of Elaeosticta allioides, Elaeosticta polycarpa, Hyalolaena intermedia, and Ferula clematidifolia were reported for the first time. Altogether, 100 components were identified in their volatile oils, in which the oxygenated monoterpenes and sesquiterpenes represent the major constituents. Multivariate analysis was adopted to correlate the studied species based on the chemical and biological data, revealing a close relation between Ferula and Elaeosticta. The partial least square regression model showed excellent prediction results for the antioxidant activity based on GC-MS metabolic profiling.
In addition, Ferula showed superior biological activity in all experiments that might present a good source for many drug leads, especially for those diseases associated with the high level of oxidative stress. These products are widely accepted among the populations since most of these plants have been used for ages in culinary purposes as food. Therefore, it will be of great value if these plants are applied for therapeutic uses. However, more studies are required to ensure the efficacy and the safety of the drug before its approval.