GC-MS and LC-DAD-MS Phytochemical Profiling for Characterization of Three Native Salvia Taxa from Eastern Mediterranean with Antiglycation Properties

Salvia fruticosa and S. pomifera subsp. calycina are native to Eastern Mediterranean and S. pomifera subsp. pomifera is endemic to Greece. The primary aim of this study was to develop an analytical methodology for metabolomic profiling and to study their efficacy in combating glycation, the major biochemical complication of diabetes. After sequential ultrasound-assisted extraction of 2 g of leaves with petroleum ether and 70% methanol, the volatile metabolites in the petroleum ether extracts were studied with GC-MS (Gas Chromatography-Mass Spectrometry), whereas the polar metabolites in the hydroalcoholic extracts were determined and quantified by UHPLC-DAD–ESI-MS (Ultra-High Performance Liquid Chromatography–Diode Array Detector–Mass Spectrometry). This methodology was applied to five populations belonging to the three native taxa. 1,8-Cineole was the predominant volatile (34.8–39.0%) in S. fruticosa, while S. pomifera had a greater content of α-thujone (19.7–41.0%) and β-thujone (6.0–39.1%). Principal Component Analysis (PCA) analysis of the volatiles could discriminate the different taxa. UHPLC-DAD-ESI-MS demonstrated the presence of 50 compounds, twenty of which were quantified. PCA revealed that not only the taxa but also the populations of S. pomifera subsp. pomifera could be differentiated. All Salvia samples inhibited advanced glycation end-product formation in a bovine serum albumin/2-deoxyribose assay; rosmarinic and carnosic acid shared this activity. This study demonstrates the antiglycation activity of S. fruticosa and S. pomifera extracts for the first time and presents a miniaturized methodology for their metabolomic profiling, which could aid chemotaxonomic studies and serve as a tool for their authentication and quality control.


Introduction
Salvia L. (Lamiaceae) includes approximately 980 species that are distributed almost worldwide [1]. A few of these are aromatic species that are used as flavoring agents and spices, but also as medicinal herbal products with commercial value. More than 100 volatiles have been found in the essential oil of the studied Salvia species, belonging to the classes of monoterpenes, sesquiterpenes, diterpenes, and non-isoprenoid compounds, usually with thujone, camphor, and 1,8-cineole as the most dominant ones [2]. Regarding non-volatiles, about 160 polyphenolic compounds have been identified from sage plants: flavonoids and their glycosides, anthocyanins, and phenolic acids with characteristic caffeic acid derivatives, such as rosmarinic acid, and phenolic diterpenes such as carnosic acid [3,4]. Most studies on the phytochemistry of Salvia taxa have focused on essential oils and have demonstrated that the chemical composition of essential oils varies greatly not only among different taxa but also within the same taxon [5][6][7][8].

Results and Discussion
Leaves from five populations from three Salvia taxa were collected and analyzed (two biological samples/population). In detail, two of Salvia fruticosa Mill. (from Fournoi island (North Aegean) and Rodini (Achaia, Peloponnese), mentioned as SF-S and SF-A, respectively), two of Salvia pomifera L. subsp. pomifera (from Elafonisi area and Sfakiano gorge (Crete), mentioned as SPP-E and SPP-FS, respectively) and one population of Salvia pomifera subsp. calycina (Sm.) Hayek from Parnitha Mountain in Attica (SPC-A).

GC-MS Analysis of Petroleum Ether Extracts for Volatile Profiling
The average petroleum ether (PE) extract yield was 8.96%, 4.15%, 12.76%, 7.40%, and 5.17% (v/w) for SPC-P, SPP-E, SPP-FS, SF-S, and SF-A, respectively. High yield values are explained by the fact that such organic solvents have the capacity to extract lipophilic compounds such as fatty acids and aliphatic hydrocarbons along with essential oil ingredients. Literature on PE extraction from Salvia species is limited; however, our results agree with those of Velickovic et al. [25] for other sage taxa (4.9 and 2.7%).
In total, forty-one (41) volatiles were identified in the PE extracts of the five examined Salvia populations, and fifteen were common in all samples (Table 1). Representative chromatograms are provided in Supplementary Information (Figures S1-S3). The total identified compounds ranged from 87 to 93% of the total peak area and were classified as monoterpenes and sesquiterpenes. Oxygenated monoterpenes and sesquiterpenes dominated every Salvia PE extract (Table 1). Table 1. The major volatile compounds in petroleum ether extracts of the leaves of the five Salvia populations using GC/MS. The results are expressed as a percentage of the total peak area (means ± standard deviation), while the limit for identification and quantification was set at 0.10%. The composition of volatiles in the S. pomifera subsp. pomifera populations were quite distinct from that in the S. pomifera subsp. calycina population, but thujones were the dominant ingredients in both taxa. Specifically, the main compounds in S. pomifera subsp. calycina (SPC-P) extracts were α-thujone and cubebol (approximately 20% and 10%, respectively). Additionally, β-thujone, (E)-caryophyllene, and myrcene were found in percentages of over 4.5%. The high values of cubebol and myrcene are in accordance with previous studies on the taxon [7,26]. On the other side, the main ingredients of the S. pomifera subsp. pomifera extracts were α-thujone (>25%) and β-thujone (>20%). In both S. pomifera subsp. pomifera extracts, (E)-caryophyllene was present at relatively high concentrations (>7.5%), whereas camphor was present at low concentrations (<1%) in agreement with the results of Karousou et al. [27]. Furthermore, the predominant compound in SPP-E was α-thujone whereas SPP-FS was dominated by β-thujone.

LC Profiling and Determination of Polar Phenolic and Diterpene Metabolites in Hydroalcoholic Extracts
The yield of hydroalcoholic extracts of the Salvia populations ranged from 8.82 to 10.96%. The metabolites were analyzed and identified by UHPLC-DAD-ESI-MS analysis using both positive and negative ionization modes. Representative chromatograms are presented in Supplementary Information (Figures S4-S6). Table 2 illustrates the characterization of the 50 compounds that were detected in the hydroalcoholic Salvia extracts, and Table 3 presents the quantification results. The linearity calibration curves are shown in Figure S7. Table 2. Polar metabolites identified in hydroalcoholic extracts of the five Salvia samples using UHPLC-DAD-ESI-MS on a C18 column. The retention times, the molecular weight, the ions observed after positive or negative ionization with a description of the ion origin with their relative abundance in parentheses, and the UV-vis absorption maxima are presented herein. The Salvia sample in which each compound was determined is presented in the eighth column. The previous studies that helped the characterization are provided in the last column.

Peak
No.     Table 3. Concentration of the major polar metabolites in mg g −1 dry extract weight in each population. The first column shows the peak numbers as presented in Table 2 and the second column, the wavelength at which the quantification was performed.

SPC-P SPP-E SPP-FS SF-S SF-
Rosmarinic acid, the main metabolite in all populations, ranged from 3.80 to 12.55 mg/g DW (Table 3); the SPC-P population showed the highest value of rosmarinic acid (12.55 mg/g DW), while the two S. pomifera subsp. pomifera populations, the lowest values (<4 mg/g DW). S. fruticosa populations SF-S and SF-A contained 7.22 and 11.02 mg/g DW rosmarinic acid. Our results are in accordance with previous studies in S. pomifera and S. fruticosa samples [12]; in that study, rosmarinic acid in 15 S. fruticosa populations ranged from 1.00 to 10.72 mg/g DW and in two S. pomifera populations 2.46-6.74 mg/g DW. Vergine et al. [15] quantified 6.55 mg/g DW of rosmarinic acid in S. fruticosa. Sarrou et al. [11] reported a decrease in rosmarinic acid in S. fruticosa from 35.354 mg/g DW to 26.355 mg/g DW from August to September.
Salvianolic acid K (compound 25) was also detected in all Salvia populations but was quantifiable only in SF ones (2.29-2.44 mg/g DW). Cvetkovikj et al. [12] reported the presence of <1.00-1.31 mg/g DW salvianolic acid K in 15 S. fruticosa populations.
Additionally, two unidentified compounds with MW of 598 (compound 9) and 778 (compound 19) were found in the S. pomifera populations, but to our knowledge, no study has reported any information about them. Both 9 and 19 were quantified only in the SPC-P and SPP-E populations in a range of concentrations of 2.36-0.82 mg/g DW and 1.32-1.74 mg/g DW, respectively.
Seven abietane diterpenes were present in the two S. fruticosa populations, specifically two rosmanol isomers (33 and 35), methoxy rosmanol or methoxycarnosol (44), two carnosol isomers (45) [38]. Decarboxylation-derived fragments were also present in mass spectra of carnosol (m/z 285) and carnosic acid (m/z 287). Additionally, rosmanol displayed the fragment m/z 283, which corresponds to a further loss of water [38]. Unknown compound 39, with an MW of 362, is possibly an abietane diterpene as well, due to the presence of the fragment m/z 287. Carnosic acid was present only in S. fruticosa populations at 6.35 and 7.22 mg/g DW. Carnosol (45) ranged between 7.51 and 9.11 mg/g DW in S. fruticosa populations. Those findings agree with previous literature where the same compound displayed 6.18 mg/g DW in an S. fruticosa population [12]. Moreover, an unidentified diterpene 40 was found in the S. pomifera populations as described in the study by Koutsoulas et al. [14] along with compound 43. Both 40 and 43 were detected in quantifiable amounts in all S. pomifera populations and ranged between 3.24-4.83 mg/g DW and 5.06-16.38 mg/g DW, respectively. Compound 49 was present in the S. pomifera populations as previously mentioned [14]; SPP-E and SPP-FS contained approximately 5 mg/g DW of compound 49.

Multivariate Analysis
In order to test the power of this analytical toolkit to characterize the Salvia taxa and populations, biplots were produced to determine the association of the volatile and non-volatile chemical compounds with the samples (Figures 1 and 2, respectively).
The concentrations of the 20 most abundant ingredients of the polar extract were also subjected to PCA, and 82.17% of the total variability of the data was explained by the first three components ( Figure 2); PC 1 and PC 2 explained 46.10% and 36. 07%, respectively. Based on the biplot analysis, all S. fruticosa samples were associated with salvionic acid K (25), carnosol (45), carnosic acid (48), and compound 50 (abietane diterpene); S. pomifera ssp. pomifera samples were not grouped together, but were associated with salvigenin (42), salviol (47), and compound 41 (7-methoxy rosmanol or methoxycarnosol). In addition, SPP-E and SPC-P samples were associated with compounds 9 and 19 and luteolin-7-Oglucoside (12); however, S. pomifera ssp. calycina samples were not closely associated with any of the quantified non-volatile ingredients. It is worth mentioning that the two SF samples in our study were collected at different times (SF-S in July and SF-A in September), but they were clustered together in both Figures 1 and 2, showing that time-dependent changes in the concentration of the metabolites that may exist do not affect the discrimination from the other taxa.
in mass spectra of carnosol (m/z 285) and carnosic acid (m/z 287). Additionally, rosmanol displayed the fragment m/z 283, which corresponds to a further loss of water [38]. Unknown compound 39, with an MW of 362, is possibly an abietane diterpene as well, due to the presence of the fragment m/z 287. Carnosic acid was present only in S. fruticosa populations at 6.35 and 7.22 mg/g DW. Carnosol (45) ranged between 7.51 and 9.11 mg/g DW in S. fruticosa populations. Those findings agree with previous literature where the same compound displayed 6.18 mg/g DW in an S. fruticosa population [12]. Moreover, an unidentified diterpene 40 was found in the S. pomifera populations as described in the study by Koutsoulas et al. [14] along with compound 43. Both 40 and 43 were detected in quantifiable amounts in all S. pomifera populations and ranged between 3.24-4.83 mg/g DW and 5.06-16.38 mg/g DW, respectively. Compound 49 was present in the S. pomifera populations as previously mentioned [14]; SPP-E and SPP-FS contained approximately 5 mg/g DW of compound 49.

Multivariate Analysis
In order to test the power of this analytical toolkit to characterize the Salvia taxa and populations, biplots were produced to determine the association of the volatile and nonvolatile chemical compounds with the samples (Figures 1 and 2, respectively).    Table 2.
The concentrations of the 20 most abundant ingredients of the polar extract were also subjected to PCA, and 82.17% of the total variability of the data was explained by the first three components (Figure 2); PC 1 and PC 2 explained 46.10% and 36. 07%, respectively. Based on the biplot analysis, all S. fruticosa samples were associated with salvionic acid K (25), carnosol (45), carnosic acid (48), and compound 50 (abietane diterpene); S. pomifera  Table 2.

Antiglycation Studies
First and foremost, the optimization of the method was performed. According to Maietta et al. [39], fluorescence monitoring with the 335/420 pair gives information on pentosidine-like AGEs, while the 370/440 pair, on vesperlysine-like AGEs. 2-Deoxy-Dribose was chosen due to its higher glycation capacity as described by Monnier [40]. Trichloroacetic acid (TCA) was used to precipitate proteins and remove contaminants from the solution. Matsuura et al. [41] revealed that the use of 10% TCA did not affect AGEs structure. Aminoguanidine at the concentration of 7.5 mM showed 88.3 ± 2.2% and 86.4 ± 2.9% inhibition of pentosidine-like and vesperlysine-like AGE formation, respectively; those values agree with the literature since its reported IC50 is 5 ± 3 mM [42].
The dose-response curves showing the % inhibition of vesperlysine-like and pentosidinelike AGEs formation by extracts in the concentration range of 15 to 100 µg/mL are presented in Figure 3. At the concentration of 100 µg/mL, all Salvia extracts showed strong inhibitory activity against AGEs formation (>77.9%). At lower concentrations, SF extracts scored higher inhibitory action than the S. pomifera ones. This finding might be due to the fact that SF populations contained carnosic acid (not present in S. pomifera ones) and high values of rosmarinic acid. respectively; those values agree with the literature since its reported IC50 is 5 ± 3 m The dose-response curves showing the % inhibition of vesperlysine-lik pentosidine-like AGEs formation by extracts in the concentration range of 15 to 100 are presented in Figure 3. At the concentration of 100 μg/mL, all Salvia extracts s strong inhibitory activity against AGEs formation (>77.9%). At lower concentratio extracts scored higher inhibitory action than the S. pomifera ones. This finding m due to the fact that SF populations contained carnosic acid (not present in S. pomifer and high values of rosmarinic acid.  Previous studies have revealed the antiglycation activity of those compounds, and we confirmed this in the current experimental set-up ( Figure 4) [43,44]. Previously, rosmarinic acid and carnosic acid inhibited the formation of AGEs, carboxymethyl lysine and carboxyethyl lysine production in the BSA-glucose, BSA/glyoxal, and BSA/methylglyoxal assay systems, and significantly decreased the concentration of methylglyoxal and protein carbonylation [44].
In our BSA/2-deoxyribose assay system, both compounds showed nearly 100% and 90% inhibition of pentosidine-like and vesperlysine-like AGEs formation at the concentration of 1 mM, respectively. At a higher concentration, they seem to be less effective, but this apparent increase in AGEs production might be due to the quenching effect of the fluorescence of the tested compounds at higher concentrations, which has been repeatedly reported for natural products [42]. Notwithstanding, the beneficial inhibitory activity of AGEs formation has been reported for other sage ingredients, e.g., for caffeic acid derivatives and luteolin glycosides [24,43], as well; thus, the other ingredients might also contribute to the activity of the extracts. rosmarinic acid and carnosic acid inhibited the formation of AGEs, carboxymethyl lysine and carboxyethyl lysine production in the BSA-glucose, BSA/glyoxal, and BSA/methylglyoxal assay systems, and significantly decreased the concentration of methylglyoxal and protein carbonylation [44]. In our BSA/2-deoxyribose assay system, both compounds showed nearly 100% and 90% inhibition of pentosidine-like and vesperlysine-like AGEs formation at the concentration of 1 mM, respectively. At a higher concentration, they seem to be less effective, but this apparent increase in AGEs production might be due to the quenching effect of the fluorescence of the tested compounds at higher concentrations, which has been repeatedly reported for natural products [42]. Notwithstanding, the beneficial inhibitory activity of AGEs formation has been reported for other sage ingredients, e.g., for caffeic acid derivatives and luteolin glycosides [24,43], as well; thus, the other ingredients might also contribute to the activity of the extracts.

Plant Material
The aerial parts of five native Greek sage populations were collected on July of 2018 except SF-A that was collected in September of the same year. Plant material of SPC-P, SPP-E, and SPP-FS was collected and identified by Prof. Dr. P. Trigas, while their vouchers (ACA 6459, ACA 6433, ACA 6457) were deposited at the Agricultural University of Athens Herbarium (ACA). SF-S and SF-A were collected by K. Zeliou and M. D. Gkioni, respectively, and identified by the UPA Herbarium staff, where vouchers were deposited (UPA 22928, UPA 22929). Plant material was dried at room temperature in the dark and stored until extraction.

Extraction
For the extraction, previously described protocols were modified [12,25]. Two grams of dried leaves were grounded in the presence of liquid nitrogen and subsequently extracted with 20 mL petroleum ether (three times) and 20 mL 70% (aq.) methanol (three times) for 20 min each time in an ultrasonic bath (120 W, 40 kHz) at a temperature under 40 • C. Two biological replicates per population were extracted. The petroleum ether (PE) extracts obtained were condensed, dried over anhydrous sodium sulfate, and stored at −20 • C. The hydroalcoholic extracts obtained were freeze-dried (Freezone 6, Labconco, MO, USA) and stored at −20 • C.

Gas Chromatography-Mass Spectrometry (GC-MS)
The PE extracts were analyzed by GC-MS on a non-polar HP-5MS (30 m × 0.25 mm × 0.25 µm film thickness) column, using a 6890N GC interfaced with an 5975B mass spectrometer (Agilent Technologies Inc., Santa Clara, CA, USA) using electron impact (70 eV) ionization mode. Helium was used as the carrier gas; the flow rate was 1 mL/min and the injected volume was 1 µL in the splitless mode. Injection temperature was set at 280 • C, and the ion source was heated to 230 • C. The oven temperature program was 59

Anti-Glycation Activity Determination Assay
The method was based on previous studies with slight modifications [39,41]. Bovine Serum Albumin (BSA) (0.625 mg/mL) was mixed with 2-deoxy-D-ribose (25 mM) and sodium azide (0.75 mM) in the presence or absence of inhibitors and samples. All solutions were prepared in phosphate buffer (pH 7.4; 0.1 M). Sage hydroalcoholic extracts were prepared at various concentrations (15-100 µg/mL), while aminoguanidine at 7.5 mM was used as a standard inhibitor. Rosmarinic and carnosic acid were examined in a concentration range of 0.14-3.00 mM. Samples were incubated at 37 • C in the dark for 6 days. At the end of the incubation, 10% w/v TCA was added, and the samples were centrifuged. Sediments were redissolved in the phosphate buffer, and fluorescence intensity was measured. The excitation/emission wavelength pairs were 335/420 nm and 370/440 nm.

Multivariate Analysis
The quantified volatile components, as well as the main polar metabolites of the hydroalcoholic extracts, were analyzed using principal component analysis (PCA) (followed by Varimax rotation with Kaiser normalization) with SPSS version 25.0 (IBM Corp., Armonk, NY, USA). The variables were standardized for a normalized PCA; the value of 0.001 was used for compounds that were not detected. For the volatile compounds, a matrix was created with 41 variables × 10 samples = 410 data points, while for methanolic extracts, the matrix was generated from 20 variables × 10 samples = 200 data points. Each component value of the Loading Plot graph was calculated in relation to the other, adjusting each value for the mean of each extraction. Biplots were produced with CATPCA (IBM Corp., Armonk, NY, USA).

Conclusions
With a miniaturized extraction approach of the plant material, we could monitor simultaneously the volatile and polar metabolites of Salvia leaves with only 2 g of starting material. Thus, thorough monitoring of the chemovariability of five populations of the three taxa was possible for the first time. A virtue of this approach is that it may facilitate studies on rare endemics and range-restricted sage species. The non-polar leaf extracts of S. pomifera populations had a high content of α-thujone (19.7-41.0%) and β-thujone (6.0-39.1%), while S. fruticosa extracts exhibited a greater content of 1,8-cineole (34.8-39.0%). UPLC-DAD-ESI-MS demonstrated the presence of a total of 50 compounds; 14 were present in all studied populations. Carnosic acid was determined only in S. fruticosa (6.35-7.22 mg/g DW), while rosmarinic acid concentration ranged from 3.80 to 12.55 mg/g DW; the lowest values were noted in SPP ones. Multivariate analysis showed that this analytical methodology can discriminate the three Salvia taxa and can be used in chemotaxonomic and authentication efforts with larger numbers of populations. For the first time, we show the dose-dependent high antiglycation activity of the polar extracts of those three taxa and confirm in our experimental set-up (BSA/2-deoxy ribose) previous studies showing the antiglycation activity of rosmarinic and carnosic acid, showing that those two compounds contribute greatly to the antiglycation activity of the extracts. In summary, the secondary metabolome of Greek and Cretan sage populations was recorded with the usage of a strategy requiring minute amounts of plant material. The promising anti-glycation activities of the methanolic extracts of the leaves are presented.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/molecules28010093/s1. Figure S1: Representative total ion GC chromatogram of the petroleum ether extract of the leaves of the population SPC-A, Figure S2. Representative total ion GC chromatogram of the petroleum ether extract of the leaves of the population SPP-E, Figure S3. Representative total ion GC chromatogram of the petroleum ether extract of the leaves of the population SF-S, Figure S4. Representative chromatogram from the UHPLC-DAD-MS analysis of the hydroalcoholic extract of the leaves of the population SPC-A at 280 nm (1, black), at 330 nm (2, green), the total ion current after positive ionization (3, red) and negative ionization (4, blue), Figure S5. Representative chromatogram from the UHPLC-DAD-MS analysis of the hydroalcoholic extract of the leaves of the population SPP-FS at 280 nm (1, black), at 330 nm (2, green), the total ion current after positive ionization (3, red) and negative ionization (4, blue), Figure S6. Representative chromatogram from the UHPLC-DAD-MS analysis of the hydroalcoholic extract of the leaves of the population SF-S at 280 nm (1, black), at 330 nm (2, green), the total ion current after positive ionization (3, red) and negative ionization (4, blue), Figure S7. Linearity calibration curves of luteolin-7-O-glucoside, rosmarinic acid, and carnosic acid that were used as external standards in the UHPLC-DAD-MS analysis of the hydroalcoholic extracts of the leaves of the various Salvia taxa. Each one was used for a different category of compounds. Funding: This work was partly supported by the project "PlantUP" (MIS 5002803), which is implemented under the Action "Reinforcement of the Research and Innovation Infrastructure", funded by the Operational Programme "Competitiveness, Entrepreneurship, and Innovation" (NSRF, 2014-2020) and co-financed by Greece and the European Union (European Regional Development Fund).