4.1 Sampling of plant material

The sampling of plant material occurred during the blooming season (February and March) of 2021 as described in De Agostini et al. [10]. In fact, while reports indicate anthesis occurring as early as in November in coastal regions of the southern distribution, the peak bloom typically occurs during the winter months from January to April.

Leaves of H. robertianum were gathered from six distinct localities across Sardinia Island (Italy) (Fig. 8), characterized by diverse ecological features. Sampling sites were chosen taking into account the results of our previous study, where the variability of volatile organic compounds (VOCs) produced by H. robertianum inflorescences was explored [10].

The sampling sites will be referred in the text as BAO, CAM, DOM, JER, SIS and SUP. The labels refer to the toponymy of the sampling sites: BAO = Bao Onu (municipality of Laconi); CAM = Capo Mannu (municipality of Cabras); DOM = Domusnovas; JER = Jerzu; SIS = Sant’Isidoro (municipality of Quartucciu); SUP = Su Planu (municipality of Selargius). Coordinates, altitude, climatic conditions, lithology, and vegetation types are summarized in Fig. 8 (map obtained by Qgis Software and further modified on Microsoft PowerPoint).

Climatic data are from the climatic monitoring authority of Sardinia (data available at https://www.sar.sardegna.it/pubblicazioni/riepiloghimensili/mensili.asp) and consider the mean values of January and February 2021 as they reflected the climate conditions during plant development. Lithological information derived from Aru et al. [43, 44]. Vegetation data were based on authors’ observations. H. robertianum leaves were sampled from three individuals per location. The leaves were carefully stored in zip-locked polyethylene bags containing silica gel for transport, and freeze-dried upon arrival at the laboratory. They were then powdered using an electrical grinder (1 g of fresh material yielded 0.1 g of freeze-dried powder). Specimens (one for each population) were deposited in the Herbarium CAG of the Department of Life and Environmental Sciences of the University of Cagliari, with the specimens’ vouchers CAG1305/V1a-f, as already reported in De Agostini et al. [10].

4.2 Chemicals

Deuterium oxide (D2O, 99.90% D) and deuterated methanol (CD3OD, 99.80% D) were purchased from Eurisotop (Cambridge Isotope Laboratories, Inc, France). Standard 3-(trimethylsilyl)-propionic-2,2,3,3-d4 acid sodium salt (TMSP), sodium phosphate dibasic anhydrous and sodium phosphate monobasic anhydrous and all the other chemicals and solvents were purchased from Sigma-Aldrich Co. (St. Louis, MO, USA).

4.3 Extracts preparation for 1H-NMR profiling

Thirty mg of powdered freeze-dried leaves per individual were extracted with 1 mL of mixture (1:1) of phosphate buffer (90 mM; pH 6.0) in D2O (containing 0.01% TMSP) and CD3OD by ultrasonication (TransSonic TP 690, Elma, Germany) for 20 min. After this procedure, samples were centrifuged for 5 min (17000 × g), and then 700 μL of supernatant were transferred into NMR tubes.

4.4 NMR analysis and data treatment

1H NMR, homonuclear (COSY and J resolved) and heteronuclear 2D correlation experiments (HMBC, HSQC) were recorded at 25 °C on a Bruker Avance Ascend 600 instrument equipped with autosamplers and a cryoprobe Prodigy. For 1H NMR profiling, the instrument operated at 1H NMR frequency of 600.13 MHz, and CD3OD was used as internal lock. Each 1H NMR spectrum consisted of 46 scans with a relaxation delay (RD) of 2 s and spectral width of 9595.8 Hz (corresponding to δ 16.0), the measurement lasted 4 min. A presaturation sequence (PRESAT) was used to suppress the residual water signal at δ 4.83.

The spectra were manually phased, baseline corrected, and calibrated to the internal standard trimethyl silyl propionic acid sodium salt (TMSP) at δ 0.0, this was also used as a standard for semiquantitative analysis. Spectral intensities (in the region from δ 0.0 to 10.0), were reduced to the integrated regions of equal width (δ 0.04) and normalized by total area using the NMR MestReNova 12 software (Mestrelab Research, Spain).

4.5 Multivariate data analysis and statistics

The obtained data matrix was subjected to multivariate data analysis using the software SIMCA P + (v. 18, Sartorius), the models were built using Pareto scaling. The spectral regions between δ 5–4.5 and δ 3.34–3.30 were excluded from the analysis because of the residual solvent signals. Metabolites were identified on the basis of literature data [45, 46], in-house database, and further phytochemical analysis. All the spectra are available on Zenodo repository (https://doi.org/10.5281/zenodo.14960480).

Values of metabolite concentration (calculated by semi-quantitative analysis) were expressed as μg/mg (leaf DW) as mean and standard deviation (SD) of three individuals. Statistical analyses were performed using Graph Pad Prism 4 software (La Jolla, CA, USA) by one-way analysis of variance (ANOVA), followed by Tukey’s honestly significant difference (HSD) post-hoc test, considering significant differences at P values < 0.05.

4.6 Pre-purification procedures

Pre-purification procedures were carried out to characterize the most abundant secondary metabolites, whose structures could not be elucidated just from the NMR profile. In order to extract preferentially the specialized metabolites, 6.7 g of freeze-dried leaves (obtained by pooling material from all samples) were extracted with 90 mL of MeOH/H2O (70:30), filtered on Büchner funnel, and dried in rotary evaporator. The procedure was repeated six times obtaining 3.44 g of dried extract (51.3% w/w). This latter was dissolved in 60 mL of H2O and partitioned with ethyl acetate (EtOAc) for three times. The two fractions obtained through this partitioning were dried in rotary evaporator yielding H2O fraction (FrW = 3 g) and EtOAc fraction (FrEt = 440 mg). Subsequently, FrEt and FrW were further fractionated by Medium Pressure Liquid Chromatography (MPLC) (Reveleris®, Bȕchi, Switzerland). In particular, FrEt was dissolved in 1.5 mL of MeOH, injected in C18 column (Select C18 30 μm spherical 4 g, Buchi, Switzerland), and eluted with a gradient of H2O (solvent A) and MeOH (solvent B) starting from 5% MeOH up to 100% MeOH in 50 min. The flow rate was 5 mL/min. The three detection wavelengths used were λ 220 nm, 256 nm, and 278 nm. The fractions were collected by volume (5 mL each tube), obtaining 50 tubes. Each fraction was dried in rotary evaporator and analyzed by 1H NMR, which guided the selection of the fractions containing the metabolite of interest. Subfractions 10 (9.8 mg) e 22 (5.3 mg) showed respectively the presence of 4-hydroxybenzyl alcohol (gastrodigenin) and bis(4-hydroxybenzyl)ether identified by means of both 2D NMR and MS experiments.

FrW (1.5 g) was dissolved in 2 mL of H2O, injected in C18 column (Select C18 50 μm spherical 80 g, Bȕchi, Switzerland), and eluted with a gradient of H2O (solvent A) and MeOH (solvent B). The gradient was composed of an isocratic phase of 4.6 min (95% A and 5% B), a gradient to 10% B in 4.6 min, an isocratic phase of 4.6 min (10% B), a gradient 20% B in 4.6 min, an isocratic phase of 4.6 min (20% B), a gradient 30% B in 4.6 min, an isocratic phase of 4.6 min (30% B), a gradient from 100% B in 18.4 min and an isocratic phase of 9.2 min (100% B). The flow rate was 30 mL/min and the run length was 60 min. The three detection wavelengths used were λ 220 nm, 256 nm, and 278 nm. The fractions were collected by volume (25 mL each tube) obtaining 72 tubes. Analogously to FrEt, NMR was employed to guide the selection of the subfraction of interest, namely subfraction 9 (10.3 mg), yielding parishin E, and subfraction 38 (3 mg), yielding parishin A, both elucidated by means of NMR and MS experiments.

4.7 Mass spectrometry

UHPLC-UV–MS analysis was run on a Waters ACQUITY ARC UHPLC/MS system consisting of a QDa mass spectrometer equipped with an electrospray ionization interface and a 2489 UV/Vis detector. The detected wavelengths (λ) were 268 nm and 370 nm. The analyses were performed on a XBridge BEH C18 column (100 × 2.1 mm i.d.; particle size 2.5 μm) with a XBridge BEH C18 VanGuard Cartridge precolumn (5 mm × 2.1 mm i.d.; particle size 1.8 µm). The mobile phases were H2O (0.1% formic acid) (A) and MeCN (0.1% formic acid) (B). Gradient: 0–0.78 min, 5% B; 0.78–10.00 min, 5 − 50% B; 10.00–11.00 min, 50–95% B; 11.00–12.00 min, 95% B; 12.00–13.00 min, 95–5% B; 13.00–14.50 min, 5% B. Flow rate: 0.8 mL/min. Injection volume: 4 μL. Electrospray ionization (ESI) in positive and negative mode was applied in the mass scan range of 50–1200 Da. Raw extracts were injected at a concentration of 10 mg/mL, while pre-purified fractions were injected a concentration of 1 mg/mL.

To obtain the exact molecular weight, FrEt subfractions 10 and 22, and FrW subtraction 38 were diluted to 1 µg/mL and analysed in a Xevo G2-XS QTof system through direct infusion. ESI in positive and negative modes was applied in the mass scan range of 50–1200 m/z. ESI source conditions were as follows: capillary = 3 kV, cone = 30 V, source temperature = 120 °C, desolvation temperature = 600 °C, cone gas flow = 50 L/h, and desolvation gas flow = 1 L/h.

4.8 Structure elucidation4.8.1 4-hydroxybenzyl alcohol (gastrodigenin)

1H NMR (600 MHz, D2O): δ 7.22 (d, 2H, J = 8.46 Hz, H2, H6), 6.83 (d, 2H, J = 8.46 Hz, H3, H5), 4.46 (s, 2H, H7a, H7b).

13C NMR: δ 155.23 (C4), 132.29 (C1), 129.52 (C2, C6), 115.23 (C3, C5), 63.48 (C7).

QToF-MS: m/z 107.0486 [M-H2O + H]+

4.8.2 bis(4-hydroxybenzyl)ether

1H NMR (600 MHz, CD3OD): δ 7.15 (d, 2H, J = 8.50 Hz, H2, H6), 6.76 (d, 2H, J = 8.50 Hz, H3, H5), 4.39 (s, 2H, H7a, H7b).

13C NMR: δ 156.87 (C4), 128.74 (C1), 129.66 (C2, C6), 114,75 (C3, C5), 71.20 (C7).

QToF-MS: m/z 229.0870 [M-H]−, m/z 123.0450 [M/2-H]−.

4.8.3 Parishin A

1H-NMR (600 MHz, CD3OD:D2O 50:50): δ 7.23 (4H, d, J = 8.98 Hz, H-2, 2", 6, 6"); 7.06 (2H, d, J = 8.98 Hz, H-2', 6'); 7.04 (4H, d, J = 8.98 Hz, H-3, 3", 5, 5"); 6.99 (2H, d, J = 8.98 Hz, H-3', 5'); 4.92 (4H, d, H-7, 7"); 4.89 (2H, d, H-8, 8"); 4.77 (1H, d, H-8'); 4.76 (2H, d, H-7'); 3.80 (6H, dd, J1 = 1.96 Hz, J2 = 12.50 Hz, H-13a, 13'a, 13"a); 3.68 (6H, dd, J1 = 5.60 Hz, J2 = 12.50 Hz, H-13b, 13'b, 13"b); 3.48 (3H, overlapping, H-10, 10', 10"); 3.46 (3H, overlapping, H-9, 9', 9"); 3.41 (3H, overlapping, H-11, 11', 11"); 3.32 (3H, overlapping, H-12, 12', 12"); 2.93 (2H, d, J = 15.3 Hz, H-15a, 15a"); 2.74 (2H, d, J = 15.4 Hz, H-15b, 15b").

13C-NMR: δ 173.3 (C-14’); 170.3 (C-14, 14"); 157.3 (C-4, 4', 4"); δ 130.1 (C-2, 2’, 2’’, 6, 6’, 6’’); 130.03 (C-1, 1’, 1’’); 116.4 (C-3, 3’’, 5, 5"); 116.3 (C-3', 5'); 102.6 (C-8'); 100.3 (C-8, 8"); 76.30(C-12, 12', 12"); 75.9 (C-10, 10', 10"); 73.5 (C-15'); 73.12 (C-9, 9', 9"); 69.2 (C-11, 11', 11"); 66.9 (C-7'); 66.4 (C-7,7"); 60.6 (C-13, 13', 13"); 43.3 (C-15, 15").

QToF-MS: m/z 1041.3093 [M + HCOO]−, m/z 995.30483 [M-H]−.

4.8.4 Parishin E

1H-NMR (600 MHz, D2O): δ 7.42 (d, 2H, J = 8.70 Hz, H2, H6), 7.14 (d, 2H, J = 8.70 Hz, H3, H5), 5.15 (d, 1H, J = 7.57 Hz, H8), 5.12 (d, 2H, J = 2.20 Hz, H7a, H7b), 4.20 (H9), 3.91 (H13a), 3.74 (H13b), 3.7 (H10), 3.6 (H11), 3.6 (H12), 2.94 (d, 1H, J = 15.20 Hz, H15a), 2.80 (d, 1H, J = 15.20 Hz, H15b), 2.69 (d, 1H, J = 15.20 Hz, H17a), 2.58 (d, 1H, J = 15.20 Hz, H17b).

13C NMR: δ 179.47 (C19), 177.49 (C18), 171.1 (C14), 156.33 (C4), 134.6 (C1), 130.04 (C2, C6), 116.62 (C3, C5), 100.31 (C8), 81.5 (C12), 76.8 (C10), 74.24 (C16), 73.4 (C9), 71.5 (C11), 66.27 (C7), 60.49 (C13), 44.9 (C17), 43.1 (C15).

UHPLC-MS: m/z 478 [M + NH4]+, m/z 459 [M-H]−