Natural Deep Eutectic Solvents for the Extraction of Triterpene Saponins from Aralia elata var. mandshurica (Rupr. & Maxim.) J. Wen

The roots of the medicinal plant Aralia elata are rich in biologically active natural products, with triterpene saponins constituting one of their major groups. These metabolites can be efficiently extracted by methanol and ethanol. Due to their low toxicity, natural deep eutectic solvents (NADES) were recently proposed as promising alternative extractants for the isolation of natural products from medicinal plants. However, although NADES-based extraction protocols are becoming common in routine phytochemical work, their application in the isolation of triterpene saponins has not yet been addressed. Therefore, here, we address the potential of NADES in the extraction of triterpene saponins from the roots of A. elata. For this purpose, the previously reported recoveries of Araliacea triterpene saponins in extraction experiments with seven different acid-based NADES were addressed by a targeted LC-MS-based quantitative approach for, to the best of our knowledge, the first time. Thereby, 20 triterpene saponins were annotated by their exact mass and characteristic fragmentation patterns in the total root material, root bark and root core of A. elata by RP-UHPLC-ESI-QqTOF-MS, with 9 of them being identified in the roots of this plant for the first time. Triterpene saponins were successfully extracted from all tested NADES, with the highest efficiency (both in terms of the numbers and recoveries of individual analytes) achieved using a 1:1 mixture of choline chloride and malic acid, as well as a 1:3 mixture of choline chloride and lactic acid. Thereby, for 13 metabolites, NADES were more efficient extractants in comparison with water and ethanol. Our results indicate that new, efficient NADES-based extraction protocols, giving access to high recoveries of triterpene saponins, might be efficiently employed in laboratory practice. Thus, our data open the prospect of replacing alcohols with NADES in the extraction of A. elata roots.


Introduction
Aralia elata var. mandshurica (Rupr. & Maxim.) J. Wen (1994) (syn. A. elata) [1] is a small, fast-growing tree or shrub that has shoots covered with thorns. It is also known as the "thorn tree" or "devil's tree". It is widely spread in Japan, eastern Siberia, South Korea and northeastern China [1,2]. The root and stem of A. elata, and especially their bark, are often used in traditional Chinese, Japanese and Korean medicine to treat rheumatic arthritis, soreness of the waist and knees, traumatic injury, lumps, abscesses, diabetes mellitus,

Identification of Triterpene Saponins in A. elata Roots
At the first step, preliminary annotation relied on the data, acquired by reversed phaseultra-high-performance liquid chromatography-electrospray ionization-quadrupole-timeof-flight-mass spectrometry (RP-UHPLC-ESI-QqTOF-MS) using a simulated sequential window acquisition of all theoretical mass spectra (SWATH) approach. For this, a comprehensive literature mining was accomplished and a representative list of triterpene saponins earlier annotated in A. elata roots was composed (Supplementary Information S1, Table S1). The core of this library relied on the work of Yu Xi et al., 2022 [2], which collected all known triterpene saponins, organized by different groups. Based on this list, a total of 104 metabolites could be annotated at the MS1 level in the samples of A. elata whole roots, bark and core. The exploration of the SWATH data revealed 25 hits, all of which could be assigned as prospective derivatives of caulophyllogenin, hederagenin and oleanolic and echinocystic acids (Table S2). However, the analysis of the MS1 data showed that annotated structures could potentially be not only the target saponins, but also the products of in-source fragmentation. This phenomenon is known in ESI-MS [42], and we addressed the probability of this scenario in detail.
One of the reasons for enhanced in-source fragmentation could be the design of the interface of the Shimadzu QqTOF mass spectrometer. Indeed, this phenomenon might be associated with not only heat-assisted desolvation (which is commonly employed to facilitate the ESI process), but also further droplet size degradation in a heated transfer capillary. Moreover, the instrument relied on Ar as a collision gas. Therefore, to reduce these effects and to verify our annotations, we employed another instrument-namely, Sciex TripleTOF 6600-for the acquisition of the MS/MS information. The interface of this instrument relies on curtain gas, which flows between the atmosphere pressure ionization (API) ion source and skimmer and further ion optics behind it [43]. Both of the curtain and collision gases were nitrogen, which resulted in lower energies of collision. With this mass spectrometer, we first performed two alternative experiments with a SWATH and DDA setup, in combination with a longer separation gradient, with the intention of obtaining fragmentation for all m/z, which could be considered as both prospective precursors and prospective in-source fragments in the data obtained using the Shimadzu instrument (listed in Table S3). This setup could present us an opportunity to address all unresolved issues by further targeted product ion scanning MS/MS experiments, if necessary.
All annotated triterpene saponins showed well-interpretable MS/MS spectra in the SWATH and DDA experiments, i.e., no targeted RP-UHPLC-QqTOF-MS/MS experiments were required. As was originally proposed, this analysis allowed the exclusion of most of the tentatively annotated species from the final list of the triterpene saponins discovered in the A. elata roots and the appearance of several new hits. The interpretation of the data was straightforward, although another QqTOF-MS instrument was used for the fragmentation experiments. Thus, only 20 metabolites of this class could be confirmed in the roots of A. elata based on their characteristic fragmentation patterns and the exclusion of false positive hits related to in-source fragmentation (Table S4). All these hits were in agreement with the spectral data acquired using the Shimadzu instrument, i.e., these analytes could be successfully re-annotated in the first dataset with both MS1 and MS2 information ( Table 1). All finally annotated compounds were the derivatives of oleanolic acid (Figures S1-S14) and their patterns were in good agreement with the classical work of Yu Xi et al., 2022 [2]. Among the annotated compounds, a group of oleanolic acid O-glycosides of hexose and pentose was noted (compounds 1, 2, 8, 9, 11, 15, 19 and 20), along with several uronic acid derivatives (compounds 3, 4, 6, 7, 10, 12, 13, 14-16 and 18). In addition, the presence of isomeric structures characterized by identical m/z but different t R s (e.g., compounds 1 vs. 15 vs. 20, 3 vs. 6, 7 vs. 13, 14 vs. 16 and 12 vs. 18) should be noted among the annotated compounds. Although most of the annotated compounds have previously been reported in A. elata roots [2], the metabolites 1, 3, 4-7, 13, 15 and 20 were identified in this organ for the first time. Thus, the earlier characterized metabolites 1, 4, 5, 6, 15 and 20 were found in leaves, compound 3 was found in steams and 7 and 13 were found in leaves and buds [2,20].

Extraction of Triterpene Saponins with NADES
The freshly prepared NADES, i.e., the mixtures of choline chloride, organic acids (malic acid, lactic acid and oxalic acid) and sugar alcohol (sorbitol) were employed for the extraction of A. elata roots. Thereby, choline chloride acted as hydrogen bond acceptors, whereas malic acid, lactic acid, oxalic acid and sorbitol were used as hydrogen bond donors [23,24]. While studying the mixture of sorbitol and malic acid, Dai and coworkers noted that sugars and organic acids can behave similarly to donors and acceptors of hydrogen bonds [23].
All the tested NADES proved to be applicable for the extraction of triterpene saponins from the roots of A. elata, although their efficiency as extractants differed. The maximal number of extracted metabolites is annotated by the ND1 (choline chloride-malic acid: 1:1) and ND3 (choline chloride-lactic acid: 1:3) extracts. However, although all twenty saponins could be identified both in the ND1 and ND3, the relative contents of the individual metabolites in the extract ND1 were higher than in the ND3. The typical chromatograms of the ND1 and ND3 extracts are presented in Figure 1 (for the whole t R range, please see Figure S15).

Extraction of Triterpene Saponins with NADES
The freshly prepared NADES, i.e., the mixtures of choline chloride, organi (malic acid, lactic acid and oxalic acid) and sugar alcohol (sorbitol) were employed extraction of A. elata roots. Thereby, choline chloride acted as hydrogen bond acc whereas malic acid, lactic acid, oxalic acid and sorbitol were used as hydrogen donors [23,24]. While studying the mixture of sorbitol and malic acid, Dai and cow noted that sugars and organic acids can behave similarly to donors and acceptors drogen bonds [23].
All the tested NADES proved to be applicable for the extraction of triterpen nins from the roots of A. elata, although their efficiency as extractants differe maximal number of extracted metabolites is annotated by the ND1 (choline ride-malic acid: 1:1) and ND3 (choline chloride-lactic acid: 1:3) extracts. Howev hough all twenty saponins could be identified both in the ND1 and ND3, the r contents of the individual metabolites in the extract ND1 were higher than in th The typical chromatograms of the ND1 and ND3 extracts are presented in Figure the whole tR range, please see Figure S15). To address the relative recovery efficiencies of the individual triterpene sa achieved with different NADES, we normalized them to the specific recoveries ob in the aqueous and ethanolic extracts. Thus, the triterpene recoveries obtained w NADES were expressed as fold changes relative to those observed in the aqueo ethanolic extracts of the A. elata roots and root parts ( Figure 2 and Table S4). Th fold change exceeding one indicated better recoveries obtained with NADES, i.e. extraction efficiency was demonstrated by these extractants. As can be seen in Fi some NADES were inefficient for some (ND2, ND3 and ND4) or all (ND5) meta addressed, as can be judged from the comparison of their recoveries with aque ethanolic extracts.  Table 1.
To address the relative recovery efficiencies of the individual triterpene saponins achieved with different NADES, we normalized them to the specific recoveries observed in the aqueous and ethanolic extracts. Thus, the triterpene recoveries obtained with all NADES were expressed as fold changes relative to those observed in the aqueous and ethanolic extracts of the A. elata roots and root parts (Figure 2 and Table S4). Thus, the fold change exceeding one indicated better recoveries obtained with NADES, i.e., better extraction efficiency was demonstrated by these extractants. As can be seen in Figure 2, some NADES were inefficient for some (ND2, ND3 and ND4) or all (ND5) metabolites addressed, as can be judged from the comparison of their recoveries with aqueous or ethanolic extracts. (oleanolic acid-3-O-(methyldioxy-trihexopyranosyl-1-3-pentopyranosyl)-28-1-hexopyranosyl ester, (c)), 10 (araloside (b,d)), 12 (oleanolic acid-3-O-(hexosyl)-28-1-hexouronide ester isomer 1, (e)) and 13 (calendulaglycoside C isomer 2, (f)) expressed as the ratio (fold) in comparison to those observed in aqueous (blue) and ethanolic (orange) extracts. The compounds are numbered as in Table 1.

Discussion
Interestingly, NADES proved to be more efficient solvents for the extraction of thirteen triterpene saponin metabolites than water or ethanol. The recoveries of 5-10, 12-16, 18 and 19 (i.e., the relative contents of these compounds in corresponding extracts) were higher in NADES than in aqueous or ethanolic isolates. For example, NADES appeared to be more efficient than ethanol for the extraction of compound 10. The saponin 5 showed better recoveries using NADES than using water, while ethanol was a weaker extractant. The highest recoveries of compounds 6, 7, 13-15 and 18-19 were found when the 1:1 mixture of choline chloride and malic acid (ND1) was applied. On the other hand, compound 5 was better recovered using the 30% v/v aqueous 1:3 mixture of choline chloride and lactic acid (ND4), whereas compounds 8-10, 12 and 16 were better recovered using the 20% v/v aqueous 1:2 mixture of sorbitol and malic acid (ND7). Specifically, the relative recoveries of saponin 6 using the NADES extracts were higher than when using the ethanolic and aqueous extracts. Moreover, the recovery of compound 6 using ND1 was 13-fold higher than in the ethanolic extract and 9-fold higher than in the aqueous extract (Figure 2a). In contrast, the relative recoveries of compound 7 in NADES extracts appeared to be higher than in aqueous extracts, but not higher than ethanolic extracts, with the best effect observed using ND1 and ND7 (Figure 2b). The recoveries of saponin 9 using ND7 were 70-fold higher than the use of aqueous and 60-fold higher than the use of ethanolic extracts of A. elata (Figure 2c). The relative recoveries of compound 10 using NADES exceeded those when using ethanolic extracts, but not those when using the aqueous extracts (Figure 2d). This tendency was the clearest for ND7. The relative recoveries of compound 12 using NADES exceeded the contents of this compound recovered using ethanolic and aqueous extracts. This could be most clearly seen for ND7 (Figure 2e). On the other hand, the recoveries of triterpene saponin 13 were the highest using ND1 (Figure 2f). The pattern for compound 14 (araloside A isomer 1- Figure S16a) was similar to compound 9 (oleanolic acid-3-O-(methyldioxy-trihexopyranosyl-1-3-pentopyranosyl)-28-1-hexopyranosyl ester, Figure 2c); furthermore, the pattern for 18 (oleanolic acid-3-O-(hexosyl)-28-1-hexouronide ester isomer 2- Figure S16b) was similar to the pattern for the closely related structure of compound 12 (oleanolic acid-3-O-(hexosyl)-28-1-hexouronide ester isomer 1- Figure 2e). These similarities are explained by the similar structures of these compounds.
It is important to note that, despite the sufficient length of the LC gradient, some triterpene saponins co-eluted. Specifically, this was observed for compounds 1 and 2, 3 and 4, 17 and 18 and 6, 7, 8 and 9. In all cases, this chromatographic behavior can be explained by the similarities in the structures of their molecules ( Figure 2).
Thus, the patterns of the NADES extraction activities, in respect to individual triterpene saponins, differed essentially (Figure 2), which can mostly be explained by the differences in analyte structures. Moreover, isomerism also impacts this phenomenon. For example, the recovery patterns of two isomers of calendulaglycosides C (Figure 2b,f) clearly differed from each other. It is also apparent that different types of NADES have different affinities for triterpene saponins of A. elata. Obviously, the structures of the individual constituents of NADES extracts, their ratio and their solvation degree directly affect the interactions between the molecules of solvents and target analytes [50,51].
Due to the fact that all of these mentioned above-ND3, 4, 6 and 7-showed much less viscosity in comparison to ND1 and ND5; the mixtures ND1, ND3 and ND7 were characterized with better mass transfer for triterpene natural products, which was most clearly seen for the mixtures based on malic acid or lactic acid. On the other hand, the solvent ND5 (based on the mixture of choline chloride and oxalic acid) showed weak extraction efficiency (Figure 2). Our results were in agreement with the studies of Lanjekar and coworkers, who demonstrated that choline chloride-lactic acid mixture (1:1 v/v) was efficient for the extraction of triterpene saponin glycyrrhizic acid from Glycyrrhiza glabra [41,52]. Furthermore, Suresh et al., 2022 showed that this NADES was also efficient in the extraction of steroidal saponins from Trillium govanianum [53]. Choline chloridelactic acid mixture (1:2 v/v) proved to be efficient in the extraction of saponins from Acanthopanax senticosus [54]. Similar data for lactic acid-based NADES were obtained by Liu et al., 2023 for steroidal saponins from Polygonatum cyrtonema [55]. Increasing the relative contents of the aqueous component in NADES leads to better solvation of the polar analytes and, therefore, improved extractive capacity. The addition of water reduced viscosity. It promoted diffusion, while hydrogen bonds between the components weakened it [56]. It is noteworthy that for the extraction of kalopanax-Saponin F isomer 2 (6), calendulaglycoside C isomer 1 (7), calendulaglycoside C isomer 2 (13), araloside A isomer 1 (14), guaiacin B isomer 2 (15), oleanolic acid-3-O-(hexosyl)-28-1-hexouronide ester isomer 2 (18) and oleanolic acid-3-O-hexuronide-(1-3-pentafuranoside) (19), the most appropriate solvent was ND1 (choline chloride and malic acid in the ratio of 1:1). This solvent was also reported to be efficient for the extraction of ginsenoside Rb1 from Panax ginseng stems [57] and steroidal saponins from Dioscoreae nipponicae [40]. Other researchers remarked upon the efficiency of the acetylcholine chloride-malic acid-water mixture (ratio 1:2:2 v/v/v) for the extraction of triterpene saponins-madecassoside and asiaticoside from Centella asiatica [58].
It is important to note that the application of NADES not only gives access to a better efficiency of extraction, but also provides improved stabilities of the resulting extracts. On the other hand, the aqueous extracts are prone to microbial contamination and therefore have limited shelf life [24,34]. This might be underlined by the presence of additional hydroxyl and carboxyl groups in their structure of NADES, which are readily involved in the formation of hydrogen bonds with triterpene saponins [23,59].

Plant Material and Extraction Procedures
The roots of A. elata were collected in the Far East (Khabarovsk region) of Russia and the species identity was confirmed in the Department of Pharmacognosy of the Saint Petersburg State Chemical Pharmaceutical University (voucher of specimens MW0107420). All collected samples were dried without direct irradiation at room temperature. Whole roots (Y), isolated root bark (X) and root core (Z) were extracted using water, ethanol and different NADES. About 1.0 g of dried raw material (X, Y and Z) was extracted with 100 mL of hot boiling water with reflux for 2 h. Ethanol extracts were obtained by the extraction of 1.0 g of each part of raw material (X, Y and Z) with 100 mL of 96% (v/v) EtOH in Soxhlet. The resulting extracts were filtered, concentrated in a vacuum and freeze dried. Further, 10.0 mg of each lyophlisate was dissolved in 1 mL of purified water and centrifuged for 10 min at 10,000 rpm (4 • C), and the supernatant (180 µL) was analyzed by RP-UHPLC-QqTOF-MS.
All NADES were prepared by the heating and stirring method [23]. In detail, the mixtures of hydrogen bond donor and hydrogen bond acceptor with a certain amount of water (Table 2) were slightly heated (at the temperatures not exceeding 80 • C) with continuous gentle agitation before forming a clear transparent liquid (about 120 min) [23]. The composition of NADES (i.e., the composing solvents and their ratios) relied on the literature data [23,24] and our previous experiment. Extraction using NADES was performed by maceration with continuous stirring. A total of 1.0 g of dried raw material and 40.0 g of NADES were transferred in flask. The mixture was stirred at 300 rpm for 60 min at 35 • C. The residue was filtered. Prior to metabolic profiling, NADES extracts were dissolved in water (in a ratio of 2:3) and placed in an ultrasonic bath for 15 min (35 kHz): 200 µL was centrifuged for 10 min at 10,000 rpm (4 • C). The supernatant (180 µL) was further analyzed using RP-UHPLC-QqTOF-MS.

Metabolite Analysis
The samples were analyzed using the randomization/standardization strategy by reversed phase-ultra-high-performance liquid chromatography-electrospray ionizationquadrupole-time-of-flight mass spectrometry (RP-UHPLC-ESI-QqTOF-MS) using a liquid chromatograph-mass spectrometer SHIMADZU LCMS-9030 System (Shimadzu Corporation, Kyoto, Japan) operated in negative ion mode. The data acquisition relied on the sequential window acquisition of all theoretical mass spectra (SWATH) mode. The chromatographic and mass spectrometric settings are specified in Supplementary Information S1 (Table S5). For the interpretation of the LC-MS data, Shimadzu LabSolution (Shimadzu Corporation, Kyoto, Japan), MSDial (version 3.12, free available via http://prime.psc.riken.jp/Metabolomics_Software/MS-DIAL/index2.html (accessed on 3 March 2023)), and manual mass spectra interpretation were used. The quantitation relied on the integration of the corresponding extracted ion chromatograms (XICs, m/z ± 0.025) at specific retention times (t R ). The intensity matrix was generated after data processing in MSDial. Based on the intensities, the fold changes were calculated by the comparison of the amounts in the aqueous and ethanolic extracts. Data visualization was performed using Microsoft Excel 2016 ( Supplementary Information S2 (Figure S17)).
For all features, which were annotated based on their [M-H] − ions observed in the MS spectra (MS1 scans) of triterpene glycoside structures with a mass accuracy of better than 10 ppm but did not yield unambiguously interpretable fragmentation patterns in the first SWATH mode experiments (typically due to the uncertainty of the fragmentation patterns and simultaneous fragmentation of two or more intense m/z), additional SWATH and data-dependent acquisition (DDA) RP-UHPLC-MS/MS experiments were conducted using a TripleTOF 6600 mass spectrometer (AB Sciex, Darmstadt, Germany), using the LC conditions summarized in Supplementary Information S1 (Table S6). The nebulizer (GS1), drying (GS2) and curtain (CUR) gases were set to 60, 70 and 55 psig, while the ion spray voltage was set to −4500 V. The MS experiments were accomplished in the TOF-scan mode (the accumulation time was 50 or 75 ms in SWATH and DDA experiments, respectively) in the m/z range of 65-1250. The MS/MS conditions in the SWATH method were set as follows: each analysis was performed with 60 ms of accumulation time at a collision energy of −45 V with a collision energy spread (CES) of 35 V and a declustering potential (DP) of −35 V. The MS/MS conditions in the IDA method were set as follows: each analysis was performed with an accumulation of time 175 ms and a range of collision energy of −45 V with a collision energy spread (CES) of 35 V and a declustering potential (DP) of −35 V. Nitrogen was used as the CAD gas.

Conclusions
In this study, a total of twenty triterpene saponins (derivatives of oleanolic acid) were identified in the roots, root bark and root core of A. elata by RP-UHPLC-ESI-QqTOF-MS. Thereby, compounds 1, 3, 4-7, 13, 15 and 20 were identified in the roots of A. elata for, to the best of our knowledge, the first time. Furthermore, seven acid-based NADES were successfully applied for the extraction of triterpene saponins from the roots of A. elated for the first time. All of the tested NADES were able to extract triterpene saponins. The maximal number of triterpene saponins was identified using the ND1 (choline chloride with malic acid: 1:1) and ND3 (choline chloride with lactic acid: 1:3) extracts. Remarkably, NADES proved to be more efficient solvents than water or ethanol for the extraction of thirteen triterpene saponins. The relative recoveries of the compounds 5-10, 12-16, 18 and 19 were higher using NADES than when using water or ethanol. For the extraction of compounds 6-10, 12, 15, 16, 18 and 19, the addressed NADES appeared to be more efficient than conventional solvents. Our data open the prospect of replacing aqueous and alcohols with non-toxic, "green" NADES for the extraction of A. elata.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/molecules28083614/s1. Supplementary Information S1- Table  S1: the list of triterpene saponins predicted in different parts of A. elata based on literature mining. Table S2: triterpene saponins annotated in the extracts of A. elata roots by RP-HPLC-QqTOF-MS in untargeted SWATH experiments. Table S3: list of compounds confirmed by MS2 analysis in the roots of Aralia elata. Table S4: triterpene saponins annotated in the extracts of A. elata roots by RP-UHPLC-QqTOF-MS and MS/MS in SWATH and DDA experiments. Table S5: the conditions of RP-HPLC separation and the settings ESI-QqTOF-MS applied for the profiling of A. elata root semi-polar secondary metabolites. Table S6: the conditions of RP-UHPLC separation and the settings for ESI-QqTOF-MS applied for the SWATH and DDA MS/MS analysis of A. elata root semi-polar secondary metabolites. Figures-Figures S1-S14: spectral data of triterpene saponins annotated in the extracts of A. elata roots. Figure S15: full t R range in the chromatograms of ND1 and ND3 extracts of the whole roots of A. elata. Figure S16: structures and relative recoveries of 14 (araloside isomer 1, A) and 18 (oleanolic acid-3-O-(hexosyl)-28-1-hexouronide ester isomer 1, B), expressed as the difference (fold) in comparison to those observed in aqueous and ethanolic extracts. Supplementary Information S2- Table S7: results of untargeted metabolomic analysis of A. elata with MSDial software. Table  S8: relative recoveries of triterpene saponins annotated in A. elata roots, extracted by conventional solvents and different types of NADES. Table S9: relative recoveries of individual triterpene saponins in aqueous, ethanolic and NADES extracts of A. elata whole roots, root core and bark. Figure S17: relative recovery diagrams of the triterpene saponins annotated in A. elata roots extracted with conventional solvents and different types of NADES. References [2,3,19,20,[44][45][46][47][48][49] are cited in the supplementary materials.

Data Availability Statement:
The data presented in this study are available on request from the corresponding authors.