Introduction

The pineapple sage (Salvia elegans Vahl) is a hardy shrub that originates from Mexico and other Central American countries like Guatemala and Honduras. Locally its leaves and flowers are used to make drinks and salads, and it is used in traditional remedies to relieve anxiety and insomnia1. In a recent study, the triterpene ursolic acid and flavonoid 5-O-(6-rhamnosylglucoside)−7-hydroxy-4’-methoxyflavanone were the main active principles linked to the species anxiolytic property2. Overall, species of Salvia are primarily known to produce caffeic acid, polyphenolic compounds such as rosmarinic and salvianolic acids, and diterpenes3.

Salvia diterpenes are represented by bi- and tricyclic classes including abietanes, (iso)pimaranes, clerodanes and labdanes3,4. The structures of labdanes usually feature a fourth ring (usually derived from furane) from the oxidation of the compound side chain, whereas the C-ring of abietanes are commonly overoxidized to royleanones, a abietane-type characteristic of species of Coleus4. Some abietanes from species of Salvia have their ring A or B of the perhydrophenanthrene system sectioned leading to secoabietanes3,4. The less common group of diterpenes in species of Salvia is (iso)pimaranes of which only one secoisopimarane have been reported so far and that was from the leaves of S. cinnabarina M.Martens & Galeotti and featuring a 3,4-section of the diterpene ring4,5.

Morphologically, S. elegans is similar to S. cinnabarina as both plants exhibit similar foliage and flowers of the same shape and color. However, S. cinnabarina grows taller and produces a greater number of flowers than S. elegans. A chloroplast DNA study of species of Salvia within the Calosphace subgenus, where S. cinnabarina and S. elegans belong, shows a strong molecular relationship between the two species6. Therefore, it could be hypothesized that S. elegans could contain similar diterpenoid groups to S. cinnabarina such as the secoisopimarane.

This study was undertaken to see if this was in fact the case and to evaluate whether secoisopimaranes like some of the other diterpenes have activity against Gram-positive bacteria like Staphylococcus aureus. It is suggested that diterpenes in general disrupt microbial cell walls of Gram-positive bacteria while remaining ineffective on the membranes of Gram-negative bacteria. However, the specific chemical characteristics shared among diterpenes that account for this activity remain unclear. The study involved plants from the living and herbarium collections at the Royal Botanic Gardens, Kew.

Results and discussion

Salvia elegans leaves were serially extracted in n-hexane, ethyl acetate (EtOAc) and 60% methanol. Extracts were tested for antimicrobial activity (both minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC)) against Escherichia coli, Staphylococcus aureus and Aspergillus brasiliensis. None of the extracts tested against E. coli and A. brasiliensis exhibited activity, whereas the hexane extract inhibited S. aureus with a MIC of 62.5 µg/mL (Table 1). Its fractionation with a flash chromatography system resulted in six fractions H1-H6, of which H2 and H3 exhibited moderate to strong inhibitory potential against S. aureus with MICs of 250 and 15.6 µg/mL, respectively. The active fractions were examined following an NMR-based metabolomic approach described previously7. Fraction H3 (Fig. S2) was mainly composed of 3,4-secoisopimara-7,15-dien-3-oic acid (1) (Fig. 1 and S3) while the 1H NMR spectrum (Fig. S4) of H2 evidenced signals of a fatty acid. Therefore, compound 1 was purified during an untargeted analysis of both active fractions H2 and H3 on a HPLC system alongside two additional minor components, the secoisopimaranes (2–3) and labdane derivatives (4–5) (Fig. 1). Their structures were established by means of extensive NMR and MS analysis.

Table 1 Results of the antibacterial activity (µg/mL) of extracts from S. elegans, fractions (H1-H6) and isolated compounds 1–3 against Staphylococcus aureus.
Full size table
Fig. 1
figure 1

Structures of isolated compounds 1–5.

Full size image

Structure determination of new compounds

Compound 2 was a colorless oil. Its negative-mode HRESIMS exhibited a deprotonated molecular ion peak [M-H] − at m/z 319.2275 for the molecular formula C20H31O3− (calcd for C20H31O3−, m/z 319.2279). Its NMR spectra (Fig. S7-S12) provided evidence of a vinyl group at δH 5.82 (dd, J = 10.6, 17.7 Hz, H-15)/δC 150.1 (C-15), 4.90 (dd, J = 1.3, 10.6 Hz, H-16a) and 4.95 (dd, J = 1.3, 17.7 Hz, H-16b)/δC 109.5 (C-16) along with a broad doublet of olefin at δH 5.34 (J = 3.5 Hz, H-7)/δC 121.1 (C-7) and four angular methyls at δH 0.89 (s, H-17)/δC 21.4 (C-17), 1.03 (s, H-20)/δC 17.9 (C-20), 1.29 (s, H-19)/δC 34.4 (C-19) and 1.40 (s, H-18)/δC 26.5 (C-18). As it stands, one might think that these characteristics are those of the diterpene isopimaric acid, which was isolated from Aeollanthus buchnerianus7. However, the methyls H-18 and H-19 were constitutive of a 2-hydroxyisopropyl unit, supported by the HMBC cross peaks (Fig. 2) from one another as well as from the correlations either methyls exhibited to a methine at δC 49.3 (C-5) and an oxygenated carbon at δC 76.3 (C-4). This suggested an opened A-ring in compound 2 unlike isopimaric acid. Indeed, the methyl H-20 also showed correlations to the methine C-5, as expected in the A-ring of terpene skeletons, a second methine at δC 44.2 (C-9) and a methylene at δC 32.3 (C-1). The latter, with protons resonating at δH 1.84/2.28, was further engaged in a spin coupling with a second methylene at δH 2.28/2.72 as evidenced by the1H-1H COSY spectrum (Fig. S9) of compound 2. The HMBC spectrum also exhibited correlations from H-1 to C-2 (δC 29.7) and a carbonyl at δC 177.4 (C-3). Further evidence in the1H1, H COSY spectrum of compound 2 supported a spin system between the methine H-9 and the methylenes H-11 (δH 1.39/1.56) and H-12 (δH 1.39/1.52). Which, when combined to the fact that the HMBC spectrum (Fig. S11) exhibited correlations from the methyl H-17 to C-12 (δC 35.9), C-15, C-13 (δC 36.8) and C-14 (δC 45.9), confirmed the (iso)pimarane nature of the structure of 2. The relative stereochemistry of compound 2 was similar to that of (iso)pimaranes where the methyl H-20 and the methine H-5 are antiperiplanar while H-5 and H-9 are cis as judged by the correlations evidenced in its NOESY spectrum between H-20 and H-18; H-5 and H-19 or H-5 and H-9. In addition, the remaining stereochemistry at C-13 was established as S as the methyl C-17 resonated at 21.4 ppm in accordance with previously detailed method7,8. Thus, compound 2 is a new derivative of secoisopimaranes and characterized rel−4-hydroxy-3,4-secoisopimara-7,15-dien-3-oic acid.

Fig. 2
figure 2

Selected COSY (bold lines), HMBC (blue arrows) and NOESY (red double arrows) correlations in compounds 2 and 3.

Full size image

Likewise, the positive-mode HRESIMS of compound 3 exhibited a protonated molecular ion peak [M + H]+ at m/z 303.2318 for the molecular formula C20H31O2+ (calcd for C20H31O2+, m/z 303.2319). Its NMR spectra (Fig. S15-S20) exhibit some of the resonances captured above for compound 2. Mainly, the vinyl group ABX system at δH 5.82 (dd, J = 10.8, 17.6 Hz, H-15)/δC 149.8 (C-15), 4.91 (dd, J = 1.2, 10.8 Hz, H-16) and 4.96 (dd, J = 1.2, 17.6 Hz, H-16)/δC 109.6 (C-16), the olefin at δH 5.41 (m, H-7)/δC 121.0 (C-7) and the angular methyls at δH 0.90 (s, H-17)/δC 21.5 (C-17), 1.06 (s, H-20)/δC 15.3 (C-20), 1.45 (s, H-19)/δC 34.0 (C-19) and 1.58 (s, H-18)/δC 24.2 (C-18) (Table 2). Similarly, the HMBC correlations (Fig. 2) of all methyl groups support a isopimarane skeleton for compound 3 including a hydroxyisopropyl fragment at C-5 as for compound 2, but with slight differences. Indeed, the carbonyl C-3 (δC 175.8) was shielded by 1.8 ppm while C-4 (δC 86.3) was deshielded by 10 ppm compared to the same carbon atoms in compound 2. Surprisingly, compound 3 was not ionizable in negative mode but in positive mode which indicates that its structure lacks the acid or hydroxyl functions that have made compound 2 active in both modes. Thus, the structure of 3 was proposed with a lactone between C-3 and C-4. This ring closure can also justify the shielding of H-20 and C-18 or the deshielding of C-5, H-18 and H-19. The relative stereochemistry of compound 3 is similar to that of compound 2 where the methyl H-20 and the methine H-5 are periplanar while H-5 and H-9 are cis and the configuration around C-13 is S as its NOESY spectrum provided the same evidence as with compound 2 including correlations between H-20 and H-18 as well as between H-5 and H-19. Compound 3 is a new derivative of compound 2 and characterized as rel−3,4-secoisopimara-7,15-dien-3,4-olide.

Table 2 [1]H (400 MHz) and 13C (100 MHz) NMR spectroscopic data for compounds 2 and 3 (δ in ppm).
Full size table

Considering the significant yield of the plant material in the secoisopimarane 1 as compared to either compounds 2 or 3, one can consider the leaves of S. elegans as the biofactory of compound 1. Its biosynthesis is proposed to have begun with the oxidation of isopimara-7,15-diene (6) to the known isopimara-7,15-dien-3-one (7) which then undergoes a Baeyer-Villiger oxidation to compound 3. The oxidative hydrolysis of 3 followed and leads to compound 2 which continues following an elimination (dehydration) reaction to compound 1 (Fig. 3). This probable protocol can be achieved by either enzymatic or chemical routes.

Fig. 3
figure 3

Proposed biosynthesis route to compounds 1–3.

Full size image

Antibacterial assay results

Compound 1 was the most abundant compound of either fraction H3 or extract accounting for 0.11 mg/g of the dried materials. It was also the only compound tested for potency against S. aureus as the other compounds were isolated in very low yields. Compound 1 replicates the activity of the active fraction H3 with a MIC of 15.6 µg/mL (51.6 µM) similar to the activity of the positive reference, piroctone olamine, (MIC 15.6 µg/mL or 52.3 µM). In addition, compound 1 was bactericidal while the hexane extract was bacteriostatic. This is the first report on the antibacterial activity of the hexane extract of S. elegans and the secoisopimarane 1. None of the other more polar extracts showed activity against S. aureus and none of the samples were active against the other microbes tested9.

Surprisingly, this is only the second report of secoisopimaranes in Salvia, following the work of Romussi et al.5 which first highlighted the occurrence of compound 1 in S. cinnabarina. Secoisopimaranes have been mainly reported in species of Isodon and Orthosiphon, two closely related genera to Salvia, where the preference for secoisopimaranes have always been for the 3,4-seco type10,11,12,13,14 as opposed to the 9,10-seco type found exclusively in Kaempferia sp15,16,17.

Distribution of the isopimaranes across salvia species

If secoisopimarane 1 is to be utilized as an antibacterial compound then it is important to know what other species of Salvia produces it and at what yields. The genus Salvia contains around 1000 species grouped in ten subgenus made of various clades18. For instance, S. elegans is consistently placed in the Calosphace subgenus but in different clades6,18. In the general classification of the genus, S. elegans forms a clade (Fig. S21A) with S. muelleri Epling, S. lycioides A. Gray, S. coahuilensis Fernald, S. chamaedryoides Cav. S. microphylla Kunth, S. darcyi J.Compton, S. karwinskii Benth., S. curviflora Benth. and S. wagneriana Pol18. In contrast, a clade within the classification of the Calosphace subgenus relates S. elegans to S. cinnabarina, S. clinopodioides Kunth, S. ramosa Brandegee and S. regla Cav. (Fig. S21B)6. Based on the position of S. elegans in these papers, leaves from 41 specimens (Table 3) were sampled from the Herbarium at RBG Kew. The sampling included 2–5 wild specimens per species and 1–2 specimens of the same species cultivated in Europe. Collected materials were ground separately into powder and extracted in n-hexane. After removal of the solvent on a centrifugal evaporator, the extracts were dissolved in a known amount of acetonitrile (ACN) and injected into the Orbitrap Exploris 120 LC-MS along with the same solution from dried plant materials of S. elegans collected from the Living Collection at RBG Kew. Generated LC-MS data were processed on MS Dial and aligned referenced to the pool sample data. The peak areas of detected ions in each sample were extracted and only the corresponding features of compounds 1–3 were retained for comparison.

Table 3 Specimen details collected from the herbarium at RBG kew.
Full size table

The results (Fig. S22) show that the three secoisopimaranes are only detected in various specimens of S. cinnabarina and S. elegans. Both wild and cultivated materials showed relative levels of the target compounds. The three compounds were more abundant in specimens of S. cinnabarina than S. elegans. Therefore, the former could be a better source of the three compounds. The absence of compounds 1–3 in the other species does not enforce any particular phylogeny, although it does support the molecular phylogeny of Lara-Cabrera et al.6 that places both these species together. The lack of the secoisopimaranes 1–3 in other species could be an indication to consider them for authentication of both species. This could be assessed by extending the qualitative analysis to other species and subgenus of Salvia. Such an assessment would provide additional insights to solve the still opened question of the distribution of (iso)pimaranes in Salvia.

Methods

General experimental procedure

LC–MS grade solvents (acetonitrile, methanol) and formic acid were obtained from Fisher Scientific (Loughborough, UK) and milliQ water was used for HPLC and LC-MS analysis. NMR spectra were acquired on a Bruker Avance-III 1H NMR: 400 MHz and 13C NMR: 100.1 MHz) spectrometer equipped with a 5 mm cryoprobe. Chemical shifts were referenced to residual solvent signals and reported in parts per million (ppm). Spectra were processed using Bruker NMR academic Topspin software. Mass spectra were collected on a Orbitrap Exploris mass spectrometer, equipped with an Orbitrap Exploris 120 with a heated ESI source (Thermo Scientific, Germany), acquired in both negative and positive modes with a resolution of 60,000 over m/z 125–1800 under various acquisition parameters Like source voltages, sheath gas, auxiliary gas, sweep gas and capillary temperature set to 2.5 kV (negative mode) and 3.5 kV (positive mode), 50 (arbitrary units), 10 (arbitrary units), 1 (arbitrary units) and 350 °C, respectively. Automatic MS–MS fragmentation was performed on top four ions of the TIC using an isolation width of m/z 2. High-energy C-trap Dissociation with a normalized collision energy of 40 and an activation time of 0.1 ms was served to fragment ions. The MS unit is interfaced with a Vanquish Core UHPLC system, which includes a Vanquish diode array detector (VH-D10) operating at four wavelengths: 210 nm, 254 nm, 300 nm, and 366 nm. Samples are injected using an autosampler maintained at 30 °C, while the analytical column is kept at a constant temperature of 35 °C. The injection volume for each sample is 1 µL with a gradient of acetonitrile (B) in water (A) (0–5 min, 10% B; 5–20 min, 10 to 100% B, 20 to 30 min, 100% B and 30 to 35 min, 10% B). Collected data were inspected using Xcalibur v. 4.2.47 (Thermo Fisher Scientific). Chemical profiling of extracts was conducted on a Biotage® Isolera One system for splitting extracts into small fractions and a Waters Alliance 2695 HPLC system for isolation of compounds. A reversed-phase Discovery HS C-18 column (5 µm, 10 mm × 250 mm i.d., Supelco, UK) maintained at 35 °C served in compounds isolation and purification over gradient of acetonitrile + 0.1% formic acid (A) and water (B).

Plant material

The aerial parts of Salvia elegans were collected from the Living collection at RBG Kew. The specimen was fully verified by Dr Sven Landrein at RBG Kew where it is cultivated under the accession No 1994–2122. The material was freeze-dried, milled to a fine powder, and kept in the dark before being used.

Extraction and purification of compounds

The plant powder (500 g) was serial extracted in n-hexane, EtOAc and 60% MeOH affording dried extracts of 1.14 g, 1.86 g and 2.56 g, respectively. The hexane extract was split into six fractions (H1-H6) following flash chromatography on a semi-prep Biotage Isolera system. The gradient was a stepwise increase of isopropanol (B) in MeOH (A) (3CV, 5% B; 3CV, 10% B, 3CV, 20% B, 3CV, 30% B; 3CV, 40% B, 3CV, 50% B and 3CV, 100% B), flowrate 30 mL/min, on a SNAP Ultra C18 60 g cartridge from Biotage. The fractions were dissolved in CDCl3 and submitted to1H NMR then to 2D NMR for chemical profiling. As a result of bio-guided protocol, only fractions H2 (30.9 mg) and H3 (224.3 mg) were further purified. Fraction H3 was dissolved in 4 mL of ACN + 10% DMSO and injected into the Waters system, eluting with a constant flow rate of 2 mL/min of a linear gradient of acetonitrile (D) in water (C) (0–5 min, 70% D; 5–60 min, 70 to 90% D, 60 to 70 min, 100% D and 70 to 75 min, 70% D). Compounds were detected at 210, 254, 300 and 354 nm and collected by time into glass tubes. Cumulative fractions from 32 injections of 100 µL each were collected and dried using a GeneVac concentrator (Genevac, Suffolk, UK). Fraction H2 was not further purified. Collected fractions were dried and analyzed by NMR leading to compounds 1 (56.4 mg), 2 (3.1 mg), 3 (2.5 mg), 4 (< 1 mg), 5 (< 1 mg), oleanolic acid (4.2 mg) and ursolic acid (3.6 mg).

rel-4-hydroxy-3,4-secoisopimara-7,15-dien-3-oic acid (2). Colorless oil; UV (MeOH) λmax 218, 266 nm; (-)-HRESIMS, [M-H] − of m/z 319.2275 (calcd for C20H31O3−, 319.2279)1H and 13C NMR spectroscopic data are summarized in Table 2.

rel-3,4-secoisopimara-7,15-dien-3,4-olide (3). Colorless oil; UV (MeOH) λmax 218, 266 nm; (+)-HRESIMS, [M + H]+ of m/z 303.2318 (calcd for C20H31O2+, 303.2319);1H and 13C NMR spectroscopic data are summarized in Table 2.

Qualitative and comparison analysis

A total of 41 specimens of species (aerial parts) closely related to Salvia elegans were collected from the Herbarium at RBG Kew. For each specimen, the milled plant material (15 mg) was suspended in 1.2 mL of n-hexane, vortexed vigorously for 10 s, heated in a water bath kept at 50 °C for 10 min and centrifuged at 15,600 rpm for 10 min. Part of the supernatant (1 mL) was transferred to a clean flask and dried using a centrifugal evaporator (Genevac, Suffolk, UK). These extracts were then dissolved in 1 mL ACN, sonicated and spined at 15,600 rpm for 10 min, 300 µL of each extract was transferred into a glass autosampler vial for LC-MS analysis. The samples were all injected into the Orbitrap Exploris 120 using a gradient of ACN in H2O, the other parameters listed above remain unchanged. The LC-MS data generated were deconvoluted, aligned using a pool of all samples and extracted as peak areas using MS-Dial (https://systemsomicslab.github.io/compms/msdial/main.html). The processing of data was done using the mass tolerance, rt tolerance and minimum peak height of 0.01 Da, 0.04 min and 104, respectively. The other parameters remained the same as default settings and no database was added for peak annotation. Both negative and positive modes were retrieved separately.

Antimicrobial assays

Minimum Inhibitory Concentration (MIC) and Minimum Biocidal Concentration (MBC) assays were used to determine the minimum concentration of an active required to inhibit half of the growth of microorganisms. Samples were evaluated against a range of organisms in triplicate including the Gram-positive bacterium, Staphylococcus aureus ATCC6538, the Gram-negative strain, Eschierichia coli ATCC8739 and the mould Aspergillus brasiliensis ATCC16404. Microbial solutions were prepared in saline (bacteria) and saline with tween (mould) and adjusted turbidometrically to a target concentration of 107 −108 CFU/mL. This inoculum solution was further diluted in Tryptic Soy broth (S. aureus and E. coli) or Soya Dextrose broth (A. brasiliensis) to achieve a final inoculum level of approximately 105 CFU/mL assay use. A stock 96-well plate of the extracts were prepared in dimethyl sulfoxide (DMSO) at a concentration of 20 mg/mL and serially diluted in DMSO. For each test plate, 5 µL of each dilution (each well from the stock plate) was transferred to a new test plate and 195 µL of inoculum in broth was added to each well. The S. aureus and E. coli plates were incubated on an orbital sharker for 18 ± 2 h at 32.5 °C. The A. brasiliensis plates were inoculated at 20–25 °C for 5 days. After incubation, the plates were visually assessed and the MICs were determined as the most dilute well with reduced growth (~ 50%) compared to a growth control. To determine the Minimum Biocidal Concentration (MBC), 10 µL from each well was pipetted onto neutralizing agar (Modified Letheen Agar with Tween) and incubated for 24 h at 32.5 °C (S. aureus and E. coli) or 5 days at 20–25 °C for 5 days (A. brasiliensis). The MBC was determined as the most dilute concentration with no visible growth.