Lanostane-Type Saponins from Vitaliana primuliflora

The phytochemistry of the genera Androsace, Cortusa, Soldanella, and Vitaliana, belonging to the Primulaceae family is not well studied so far. Hence, in this paper, we present the results of UHPLC-MS/MS analysis of several primrose family members as well as isolation and structure determination of two new saponins from Vitaliana primuliflora subsp. praetutiana. These two nor-triterpenoid saponins were characterized as (23S)-17α,23-epoxy-29-hydroxy-3β-[(O-β-d-glucopyranosyl-(1→2)-O-α-l-rhamnopyranosyl-(1→2)-O-β-d-glucopyranosyl-(1→2)-O-α-l-arabinopyranosyl-(1→6)-β-d-glucopyranosyl)oxy]-27-nor-lanost-8-en-25-one and (23S)-17α,23-epoxy-29-hydroxy-3β-[(O-α-l-rhamnopyranosyl-(1→2)-O-β-d-glucopyranosyl-(1→2)-O-α-l-arabinopyranosyl-(1→6)-β-d-glucopyranosyl)oxy]-27-nor-lanost-8-en-25-one, respectively. Their structures were determined by high resolution mass spectrometry (HRMS), tandem mass spectrometry (MS/MS), one- and two-dimensional nuclear magnetic resonance spectroscopy (1D-, and 2D-NMR) analyses. So far, the 27-nor-lanostane monodesmosides were rarely found in dicotyledon plants. Therefore their presence in Vitaliana and also in Androsace species belonging to the Aretia section is unique and reported here for the first time. Additionally, eleven other saponins were determined by HRMS and MS/MS spectra. The isolated lanostane saponins can be considered as chemotaxonomic markers of the family Primulaceae.


Introduction
Some plants are rich in secondary metabolites of a specific class, e.g., saponins. They can reach up to 10% of dry mass and, thus, are attractive for industrial usage. Among them, steroidal saponins are particularly abundant in monocotyledons, while triterpenoid saponins are abundant in eudicotyledons, with several exceptions. Among a few cases of medicinally valuable steroidal glycosides present in angiosperms, cardiac glycosides and their open-lactone analogs should be mentioned [1]. Other economically important compounds are appetite suppressants from the South African Hoodia sp., Euphorbiaceae [2] or male hormone imitators from Tribulus sp., Zygophyllaceae [3]. On the other hand, typical triterpenoids (C 30 ) are rare in monocotyledons [4].
Taking under consideration the nomenclatural ambiguity of saponins, some researchers classify tetracyclic triterpenoids, for example, ginseng dammaranes, as steroids, and this term regularly appears in some papers [5]. With lanostanes: Some classify them as steroids because of the biosynthesis step of squalene cyclization and mutual conformation of the resulting rings [6][7][8], while others catalog them as triterpenoids by the presence of two methyl substituents in position 4 and count the total carbon number of this aglycone as 30 [7,9].
Lanostane homologs are uncommon in dicotyledonous plants [7]. They can be found in many Asparagaceae members like in ornamental muscari or squills [10] and conifers [11]. A variety of sea cucumbers should be mentioned as a non-vegetable lanostane source [12,13]. A considerable number of bioactive, non-glycosidic lanostanes was reported in some fungi, including the famous Ganoderma sp. [14,15].
Up to now, the Primulaceae family was known to be a source of triterpenoid saponins of the oleanane type [16] and several cucurbitacins [17], beside several unique flavonoids [18,19].
Performing the phytochemical screening and characterization of this family, we have developed rapid, useful, and universal UHPLC-MS and -MS/MS methods for saponin determination. Moreover, we described the isolation and identification of two dominant saponins from Vitaliana primuliflora Bertol., that were previously observed on thin layer chromatography (TLC) only [20]. Finally, we proved the occurrence of these two 27-nor-lanostane saponins in some species belonging to Androsace (in Aretia and Douglasia sections only). It is also the first communication describing plant nor-lanostanes outside the Asparagaceae family.

Results
Hydroalcoholic plant extracts were prepared routinely and analyzed by UHPLC-MS and UHPLC-MS/MS in the negative mode as a part of a more comprehensive screening. The first examination of MS chromatograms revealed two main unidentifiable ions of high intensity, especially in Vitaliana species extracts. Later on, about 13 g of underground parts of V. primuliflora subsp. praetutiana were taken for extraction. Subsequently, with the help of semi-preparative flash chromatography on silica and HPLC on reversed phase, we have obtained approximately 40 mg of 12 and 36 mg of 13, as amorphic white solids (almost 0.3% of starting dry mass each; Figure 1). Both compounds could form a stable foam at a concentration of 0.1 mg/mL. The isolation was performed in mild conditions to avoid any artifact formation. All fractionation steps were monitored by TLC/HPLC.

Structure Elucidation of New Saponins
The HRMS spectrum revealed the molecular formula of 12, the most abundant peak (R t = 13.98 min, UV max = 199 nm) to be C 58  . Sugar identity (glucose, arabinose, rhamnose) was confirmed after acidic hydrolysis on TLC as described previously [21].
1 H-NMR and heteronuclear single quantum coherence (HSQC) spectrum of 12 showed five anomeric signals. Two of them, observed as narrow doublets, were initially assigned to α-L sugars while three wide doublets were key to β-D sugars [22]. Overlapping 1 H signals were resolved by the examination of HSQC and heteronuclear multiple bond correlation (HMBC) together with total correlated spectroscopy (TOCSY) and finally defined by heteronuclear two bond correlation (H2BC).
Sugar chain linearity and positions of substitution were confirmed by 2D-NMR as shown in Figure 2a,b. Briefly, anomeric hydrogen of first glucose moiety was HMBC correlated with C3 of the aglycone and HMBC correlation from H3 to C1 of the first glucose was observed. This pointed out the place of substitution of the aglycone with the sugar chain. The C6 signal of the first Glcp was shifted downfield by 6 ppm relative to the non-substituted Glcp. This chemical shift difference suggested Arap was linked to the C6 of the first Glcp. Anomeric hydrogen of the second glucose unit was correlated not only with C2 of arabinose but also with C1 of arabinose. That was the reason the second glucose was linked (1→2) to arabinose. C2 signal of the second glucose was shifted downfield relative to the non-substituted Glcp. That remarked that this sugar was substituted with the next one (rhamnose). HMBC correlations verified that supposition. The observation of highly shifted C2 of Rhap (up to 83 ppm) together with long-range HMBC of an anomeric proton from the third Glcp determined the position of substitution at the end of the sugar chain. The 13 C chemical shifts of the glycone part of 12 were strictly similar to those reported in the literature [23][24][25].
The general pattern of the 13 C-NMR spectrum showed similarity of the aglycone of 12 to the well-known eucosterol [25]. Seven methyl groups (two of them split by nodal hydrogen), together with 14 methylene groups and eight quaternary carbons were observed. One of the methyl doublets was ascribed to the sugar unit (Rhap) while the second was ascribed to C21 of eucosterol. The examination of long-range HMBC correlations of the rest of the methyls and of well-separated H3 and H5, let us build a skeleton of the lanosterol-like molecule. Among quaternary carbon signals, the highest shift at 208.6 ppm was assigned to the carbonyl. The two double-bond forming carbons were observed at about 135 ppm (close to pyridine-d5 signals). Spiro carbon was visible at 96.2 ppm. The remaining four quaternary carbons were fixed to aglycone nodes, each substituted with the methyl group. Four methylene groups were assigned to terminal carbons of sugars (3× glucose, 1× arabinose). Another one was assigned to oxygenated methyl (C29, -CH 2 OH) in the close neighborhood of C5 and methyl C28 (based on HMBC and nuclear Overhauser effect spectroscopy (NOESY)).
Compared to eucosterol, C15 of 12 remained unsubstituted (based on TOCSY and HMBC correlations), while the side chain (C24-C26) seemed to be modified. It was observed, that the side chain showed a significant downfield shift of the terminal CH 3 group from 7 to 30 ppm compared to reference [25]. Quaternary carbonyl was the nearest neighbor to this CH 3 group and two CH 2 hydrogens were HMBC correlated both with carbonyl and epoxy ring hydrogens and carbons. Thus the order of =CH 2 and =CO in the C24-C26 chain had to be reversed compared to eucosterol. Detailed HMBC, TOCSY, and NOESY examination revealed the rest of well-separated signals typical for eucosterol (Tables 1 and 2). All key correlations are visualized in Figure 2a,b.
We based the stereochemistry of the epoxy ring of 12 on an earlier report [4]. The 13 C shifts of C16 and C17 together with strong NOE correlation between CH 3 18 and CH 3 21 methyl groups indicated the 17S and 20R conformation [26]. To prove it, we found that the H11 ax at 1.95 ppm was NOE correlated with both methyl groups CH 3 18 at 0.90 and CH 3 19 at 0.94 ppm. Besides, CH 3 18 was NOE correlated with both H20 at about 2.02 ppm and CH 3 21, that suggested the β position of C20 in relation to the skeleton (resulting in 17S conformation). Secondly, a well-separated signal of H23 at 4.61 ppm was simultaneously NOE correlated with the methyl group CH 3 21 at 1.02 ppm and H16 eq at 1.74 ppm. Moreover, none of the two H24 hydrogens was NOE correlated with the main skeleton hydrogens. Thus both H23 and CH 3 21 were assigned as α in relation to the skeleton (resulting in 20R and 23S conformation).
The neutral molecular formula of 13, noticed as the second most prominent peak in MS chromatogram of V. primuliflora hydroalcoholic extract (R t = 14.69 min, UV max = 199 nm) was revealed to be C 52 H 84 Tables 1 and 2. More HRMS fragmentation data together with retention times for compounds 12 and 13 with other observed and tentatively described triterpenoid glycosides 1-11 are arranged in Table A2.

Discussion
Taking under consideration all the Primulaceae species listed in Table A1, the newly described triterpenoid saponins 12 and 13 were detected by UHPLC-MS and UHPLC-MS/MS in the underground parts of the genus Androsace, in sections Aretia (five of six samples) and Douglasia (one of one samples) as well as in the cognate genus Vitaliana (four of four samples). The 27-nor-lanostane glycosides 12 and 13 were dominant among saponins detected by ESI-MS in the negative mode in all Vitaliana plants. The richest sample was V. primuliflora subsp. praetutiana (VPPR_B_2015), from which these two compounds were isolated (VPPR_B_2017). In samples derived from Androsace cylindrica and A. obtusifolia, the intensity of 12 was on a similar level as in V. primuliflora (VPRI_B_2015). In the remaining samples containing 12, the concentration of this compound was lower than that in V. primuliflora subsp. assoana (VPAS_TK_2015). MS signal intensity of 12 was significantly higher than that of 13 in all Vitaliana plants and in Androsace calderiana, A. lactiflora, A. obtusifolia, A. mathildae, and A. montana (= Douglasia montana). In most samples containing 12, it was also the best detectable saponin on an MS chromatogram (except A. montana, A. mathildae, and A. obtusifolia). Androsace lehmannii, the only examined Aretia member originating from Asia was transferred to section Aretia based on its morphological features [27]. We did not prove any of the newly discovered saponins in the analyzed sample of A. lehmannii. It may indicate that this species is chemically different from the rest of the Aretia members, and thus should be more precisely analyzed, also by botanists.
This is the first report on the occurrence of plant nor-lanostane saponins outside the Asparagaceae family. The new compounds were detected in samples originating from different locations, that makes our observations more reliable. All isolation procedures were conducted in mild and acid-free conditions. Nonetheless, we cannot exclude the possibility of the endophytic origin of isolated 27-nor-lanostanoids, as such observations were already published [28].
The assumption of extensive studies presented in this paper in part only was to develop an integrated model to screen many samples for the detection of a specific group of secondary metabolites, namely saponins. In our opinion, a detailed analysis of a broad range of related organisms may lead to observation and notification of some statistically significant rules. Usually, the problem is that only some of the plants in a selected botanical group are treated as important for public opinion. The reasons are enormous biomass gain or well-established pharmaceutical or medical properties. Our study shows that less popular plants could also be an interesting source of natural compounds.
Additionally, we would like to point out that amateur sourcing, breeding, or widespread cultivation of ornamental and exceptional plants (e.g., so-called 'alpines') may be a good measure in order to collect phytochemically interesting plants.

Plant Material
The plants were obtained from commercial suppliers, mainly from reputable nurseries. A precise list of seedlings suppliers and sample acronyms is available in Appendix A, Table A1.
For screening purposes, three seedlings each of: V. primuliflora Bertol., V. primuliflora subsp. assoana M. Laínz, V. primuliflora subsp. praetutiana (Buser ex Sünd.) I. K. Ferguson were used. Plants were documented by the author (M.W.). Vouchers (VPRI_B_2015, VPAS_TK_2015, VPPR_B_2015) were stored in a herbarium of the Department of Pharmacognosy and Herbal Medicines, Wroclaw Medical University. The plants were separated from the soil, carefully cleaned with tap water, separated into the underground and aboveground parts, allowed to dry in the shade for two weeks, and then stored in paper bags. Parts of plants were separately powdered right before further processing (Basic A11; IKA, Staufen, Germany), sieved (0.355 mm), and stored in darkness in airtight containers. Androsace, Soldanella, and other used plant species were processed in the same way.

Chemicals
LC-MS grade water and formic acid were purchased from Merck (Darmstadt, Germany) while acetonitrile was obtained from Honeywell (Morris Plains, NJ, USA). Analytical grade chloroform was sourced from Chempur (PiekaryŚląskie, Poland) and methanol from POCh (Lublin, Poland).

Analytical Samples Preparation
Samples for UHPLC analyses were extracted (100 mg sample per 2 mL of 70% methanol) in an ultrasonic bath (15 min, 25 • C, 50% of power; Bandelin, Berlin, Germany). Then, 1 mL of each extract was filtered through a 0.22 µm PTFE syringe filter (Merck-Millipore, Darmstadt, Germany), diluted 100 times with 50% acetonitrile in water (LC-MS class) and stored at 4 • C before the analysis.

UHPLC-MS and UHPLC-MS/MS Analysis
The UHPLC-MS instrument was operated in negative mode. The HRMS detector was calibrated in the dead time of every single run with the Tunemix TM mixture (Bruker Daltonics) with m/z standard deviation below 0.5 ppm. The analysis of the obtained mass spectra was carried out using Data Analysis 4.2 software (Bruker Daltonics). The key instrument parameters were: Scan range 50-2200 m/z, low mass set at 200 m/z, nebulizer pressure 1.5 bar, dry gas N 2 with flow 7.0 L/min, temperature 200 • C, capillary voltage 2.2 kV, ion energy 5 eV, collision energy 10 eV and 30 eV (in separate runs). MS 2 analyses were performed for ions in the range 400-1900 m/z, with collision energy gradient 40→170 eV. The gradient elution system consisted of 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B). At the flow rate of 0.3 mL/min, the following elution program was used: 0→1 min (2→30% B), 1→21 min (30→50% B), 21→21.5 min (50→100% B), 21.5→25.5 min (100% B). The column was equilibrated for 5 min before the next analysis. Blanks were run after each sample to avoid any cross-contamination. Other parameters were: Column oven temperature 30 • C, injection volume 5 µL. The MS/MS fragmentations of 12 and 13 are presented in Supplementary Materials ( Figure S1). LC-MS results are presented in Appendix A, in Table A2.

Isolation of Compounds 12 and 13
Based on UHPLC-MS results, VPPR_B_2017 underground parts were selected for semi-preparative purposes. The total amount of 12.875 g of underground parts was macerated three times with 70% methanol in water in the ratio 1:10. Combined extracts were diluted with water and applied to SPE. The fraction eluted with 70% to 85% methanol was concentrated to dryness in vacuo in 40 • C (Rotavapor V-100; Büchi, Flavil, Swiss), giving 346 mg (2.7% of initial dry mass). The first separation was performed by FC on 85 g of Si 60 (Merck, Darmstadt, Germany), with chloroform→methanol gradient at flow rate 5 mL/min. Further, selected fractions were purified by semi-preparative HPLC on RP-18 in 75% methanol (isocratically) with a flow of 3 mL/min. As a result, 36.1 mg of 13 and 40.8 mg of 12 were obtained, corresponding to 0.28% and 0.32% of starting dry mass. To assure, that no artifacts were collected, 12 and 13 were confronted with primary extract by UHPLC-MS (conditions as described in Section 4.5).

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.