NMR-based metabolomics for frauds detection and quality control of oregano samples

Abstract


Introduction
The global herbs and spices market is increasingly threatened by accidental or intentional adulteration, also facilitated by the complexity of the supply chains (Galvin-King et al., 2018).Herbs are generally sold chopped or powdered, making it relatively easy to counterfeit by means of adding cheaper bulking agents (i.e.chemicals, extraneous material from other plants or foreign species) (Van Ruth et al., 2018).
In this scenario, the development of rapid, robust, cost-effective and environmentally friendly methods for authentication and quality control of spices and herbs takes on great importance (Wadood et al. 2020).
Oregano is one of the most adulterated herbs (Black et al., 2016;L. Drabova et al., 2019).Despite the high heterogeneity of Origanum genus, only a few species are accepted on the market.According to ISO 7925:1999, all Origanum species and sub-species, except Origanum majorana L. are considered marketable, while according to the European Spice Association (ESA) only Origanum vulgare L. and Origanum onites L. (or products made of a mixture of the two) are considered 'true' oregano (List of culinary herbs and spices, ESA, 2018), with the limit of permitted impurities set at 2% (w/w) (Quality minima document, ESA, 2015).With respect to oregano authentication, the European Pharmacopoeia makes a further restriction: in addition to O. onites, it only considers a specific subspecies (subsp.) of O. vulgare as 'true' oregano, namely O. vulgare subsp.hirtum (Link) Ietsw (9 th European Pharmacopoeia, 2017).
Besides the frauds ascribable to the sale of non-marketable species, oregano is often contaminated with a significant amount of other plants such as olive tree leaves, myrtle, sumac, cistus, savory and others (Black et al. 2016, Marieschi et al., 2009;Marieschi et al., 2010;Marieschi et al., 2011a;Marieschi et al., 2011b).
Generally, companies dealing with large-scale sales buy oregano from third parties.The contractor sends a representative sample and the company analyzes it in order to establish, on the basis of the sample quality, whether to buy it or not.In this stage, in order to determine the authenticity and purity of samples, companies generally rely on morphological analysis performed by microscopy (MA).
Nevertheless, this technique is time-consuming, and thus not suitable for quickly analyzing a high number of samples; furthermore, its precision strongly depends on the operator performing the analysis.This makes it of great importance to develop new efficient, quick and cost-affordable methods for oregano quality control, and to contribute to building databases, apt to facilitate fraud detection worldwide.Marieschi et al. (2009Marieschi et al. ( , 2010Marieschi et al. ( , 2011aMarieschi et al. ( , 2011b) ) published several works focused on the assessment of oregano authenticity by sequence-characterized amplified region makers (SCARs), which also allowed them to recognize a number of specific adulterants.The technique proposed is extremely precise and accurate, however, it does not provide information about the phytochemical features of the samples, which is a further element of interest to establish the quality of an herbal product.In fact, the presence of specific metabolites confers the characteristic taste and flavor to the herb.Moreover, several plant secondary metabolites are organ-specific, this implies, taking for instance the case of oregano, that a high-quality sample should be made of leaves and bracts predominantly, with the lowest possible amount of stem.Hence, the analysis of the phytochemical profile is an important additional step for the quality assessment of herbs, also considering that the metabolites production in plants is significantly affected by a number of factors, including seasonal variation, altitude, biotic and abiotic factors and so on (Mandrone et al 2021, Anđelković et al., 2017;Scognamiglio et al., 2015;Salomè et al., 2020).
In the field of phytochemistry applied to food or herb quality control, in addition to targeted analysis, new untargeted strategies, relying on an inductive approach, have recently been emerging.In this case, the analytical protocols used are set to detect the widest possible range of metabolites (metabolome) (Riedl et al., 2015).According to this approach, the insights are gained from comparison, on a pattern level, of a high number of samples by multivariate data analysis (MVDA).This workflow was successfully applied by Black et al. (2016) to determine oregano quality by both FTIR and LC-HRMS untargeted analysis.In this context, a further analytical option is offered by NMR-based metabolomics, increasingly employed for untargeted food and herbs quality control (Sobolev et al., 2019).NMR is non-destructive, quickly performed and environmental-friendly, and in addiction it provides numerous detailed chemical information in a single spectrum (Pontes et al., 2017).
This study explores the potentiality of 1 H NMR metabolomics to be employed for oregano quality control.In particular, spectral data and results obtained by MA (performed by stereomicroscope and optical microscope) were combined by MVDA.Predictive multivariate data models, able to quickly provide a wide spectrum of information on oregano samples, were obtained, and oregano biomarkers of quality easily detectable by 1 H NMR profiling were identified.

Oregano samples and Cistus creticus
Different Italian companies provided a total of twenty-seven samples of (supposed) oregano to be analyzed in order to assess purity and authenticity.According to what declared by the companies: 14 samples were cultivated in Turkey, 3 in Sicily, 3 in Albania, 2 in Peru, while for 5 samples this information was not available.Vouchers of crude drugs were deposited in Department of Pharmacy and Biotechnology, University of Bologna (via Irnerio 42, Bologna, Italy) and reported in Table 1, together with the results obtained by the MA performed in this work.

Morphological analysis
The identification of oregano species was performed according to the oregano taxonomic key presents in Ietswaart (1980).Morphological analysis (MA) was performed following the method reported by Marieschi et al. 2009.In particular, twenty-five milligrams of each dried sample were analyzed by a stereoscopic microscope (Nikon SMZ-1000) and a microscope (Nikon Eclipse E600).The pictures were obtained by Scanning Electron Microscopy.Specifically, dried calices of Origanum onites and Origanum vulgare were mounted on aluminum stubs with double-stick carbon tape, sputter coated with gold nanoparticles (5 nm) using a Emitech K500 coater and observed with a Philips SEM 515 at 20.0 and 19.9 kV respectively.

Sample preparation for metabolomics
Samples were constituted by dried and shredded plant material.They were quartered and powdered using an electrical grinder (IKA, A11 basic, Merck, Italy) Then 50 mg were extracted with 1 mL of mixture (1:1) of phosphate buffer (90 mM; pH 6.0) in H 2 O-d 2 (containing 0.1% TMSP) and MeOH-d 4 by ultrasonication (TransSonic TP 690,Elma,Germany) for 20 minutes at 45°C.After this procedure, samples were centrifuged for 10 min (17000 x g), then 700 μL of supernatant were transferred into NMR tubes.For each sample, two different extracts were prepared to test reproducibility.

NMR and ESI-MS measurement
1 H NMR spectra, J-resolved (J-res), 1 H-1 H homonuclear (COSY) and inverse detected 1 H-13 C correlation experiments (HMBC, HSQC) were recorded at 25°C on a Varian Inova instrument (equipped with a reverse triple resonance probe).For 1 H NMR profiling the instrument was operating at 1 H NMR frequency of 600.13 MHz, and H 2 O-d 2 was used as internal lock.Each 1 H-NMR spectrum consisted of 256 scans (corresponding to 16 min) with the relaxation delay (RD) of 2 s, acquisition time 0.707 s, and spectral width of 9595.8Hz (corresponding to δ 16.0).A presaturation sequence (PRESAT) was used to suppress the residual water signal at δ 4.83 (power = -6dB, presaturation delay 2 s).For 1D and 2D NMR analysis of pure compounds MeOH-d 4 was used as lock.J-res spectra has been measured in order to clarify or confirm splitting patterns and coupling constants values.This approach is especially useful in metabolomics where a mixture of compounds is visible in a single 1 H NMR spectrum making more difficult to establish these parameters (Emwas et al, 2019).
For ESI-MS analyses, dried pure compounds were dissolved in MeOH, and analyzed by WATERS ZQ 4000 (Milford, MA USA) mass spectrometer.in negative and/or positive ion modes according to the more ionizable chemical groups of samples.A direct infusion of 20 μL/min, source temperature of 80 °C and desolvation (nitrogen) gas (flow rate of 200 L/h) were common parameters used in both positive and negative ion modes.Capillary potential and source cone were 3.54 Kv and 20 V in positive ion mode, and 2.53 Kv and 30 V in negative ion mode.The mass range was from 0 to 1000 m/z.

Multivariate data analysis
The spectra were manually phased and baseline corrected, and calibrated to the internal standard trimethyl silyl propionic acid sodium salt (TMSP) at δ 0.0 using Mestrenova software (Mestrelab Research, Spain).The regions of δ 5−4.5 and δ 3.34−3.30were excluded from the analysis because of the residual solvent signals.Then spectral intensities were reduced to integrated regions of equal width (δ 0.04) corresponding to the region from δ 0.0 to 10.0, and normalized by total area.The analysis of 1 H NMR profiles of extracts was performed based on an in-house library and comparison with literature (Mandrone et al., 2017;Mandrone et al., 2019b).
SIMCA P+ software (v.15.0, Umetrics, Sweden) was used in order to perform PCA, OPLS, OPLS-DA, PLS-DA models.For multivariate analysis, data were subjected to Pareto scaling.Supervised models were evaluated by the goodness of fit (R 2 x (cum) and R 2 y(cum)) and goodness of prediction (Q 2 (cum)), together with the parameters given by cross validation tests: permutation test (performed using 200 permutations) and CV-ANOVA (Malzert-Fréon et al., 2010).

Purification and structure elucidation of biomarkers
In order to elucidate the structure of the metabolites of interest, different NMR-guided isolation procedures ware carried out for O. onites, O. vulgare and Cistus creticus.The NMR spectra of the isolated compounds and/or fractions are given in supporting material (Fig. S1 to S8).

Isolation of biomarkers from O. onites and O. vulgare
Forty g of O. onites (degree of purity of 99.88%) or O. vulgare (degree of purity of 98.74%) were extracted using 300 mL of CH 3 OH:H 2 O (1:1), sonicating for 30 min and centrifuging for 10 min (2469 x g).The obtained supernatant was dried in rotary evaporator at 40°C yielding 18% w/w for O. onites and 15% for O. vulgare.In both cases, the dry extract was suspended into 300 mL of water to undergo liquid-liquid partition with CHCl 3 , EtOAc and n-BuOH used in sequence, repeating the procedure twice for each solvent.Fractions were dried in rotary evaporator, and for O. onites 5 g were obtained from the water fraction, 1.5 g from the n-BuOH fraction, 0.36 g from the EtOAc fraction, and 0.2 g from the CHCl 3 fraction.While for O. vulgare 4.50 g were obtained from the water fraction, 0.96 g from the n-BuOH fraction, 0.27 g from the EtOAc fraction and 0.22 from the CHCl 3 fraction.
The biomarkers of O. onites were found (by 1 H-NMR analysis) in the CHCl 3 and the n-BuOH fractions, while the water fraction contained the biomarker of O. vulgare.Fractions in EtOAc obtained from both species contained rosmarinic acid, whose structure was elucidated by NMR and ESI-MS experiments (Exarchou, et al. 2003).The CHCl 3 fraction obtained from O. onites was analyzed by 1 H NMR and ESI-MS (by direct infusion) allowing to confirm the structures of thymol and p-cymene (Exarchou, et al. 2003).For purification and structure elucidation, 200 mg of n-BuOH and water fractions from O. onites and O. vulgare, respectively, were further fractionated by column chromatography, using in both cases a chromatography column (1800 mm x 25 mm) filled with 220 g of Sephadex (LH-20) and, as eluent, methanol.The fractions were suspended in the minimum amount of CH 3 OH:H 2 O (1:1) to be chromatographed.The flow rate was of 0.4 mL/min.Each fraction was concentrated and analyzed by 1 H-NMR.Fraction 114 contained the biomarker of interest of O. onites, which has been identified as apigenin (0.1 mg) (Exarchou, et al. 2003).In case of O. vulgare fractions from 25 to 27 contained the biomarker of interest identified as salvianolic acid B (10.9 mg) (Sun et al., 2009).Table S1 summarizes the results of the NMR experiments performed in order to elucidate the structure of salvianolic acid B ( 1 H NMR, HMBC, HSQC, COSY).

Isolation of pinitol as biomarker of C. creticus
In order to characterize the biomarker of Cistus creticus L., 9.8 g of dried and grounded plant material were extracted using 300 mL of CH 3 OH:H 2 O (1:1); sonicated for 40 min; centrifuged for 10 min (2469 x g).This extraction procedure was repeated twice on the same plant material.The extract was dried in rotary evaporator (yield = 30.6%w/w) and suspended in 300 mL of water to undergo liquid-liquid partition with CHCl 3 and EtOAc (for two times each), obtaining 2.9 g of material from the water fraction and 0.1 g from the EtOAc fraction.According to 1 H-NMR analysis, water fraction contained the metabolite of interest, thus 300 mg of this fraction were suspended in 500 µ L of water and injected in MPLC instrument (Reveleris®, Büchi, Switzerland) connected to C 18 column (4 g).A gradient of water (solvent A) and methanol (solvent B) was used as eluent.The steps gradient was composed by an isocratic phase of 10 min (95% A and 5% B), a gradient from 95% A to 90% A in 10 min, an isocratic phase of 10 min (90% A and 10% B), a gradient from 90% A to 80% A in 10 min, an isocratic phase of 10 min (80% A and 20% B), a gradient from 80% A to 0% A in 20 min.The flow rate was 3 mL/min, and the run length was 70 minutes.A total of 43 fractions were collected, and the metabolite of interest was found in was in fraction 3 and fraction 4.

Discrimination between Origanum species
The samples analyzed in this work were provided by different companies, who, in turn, obtained them from third parties.The samples were under evaluation by the companies in order to decide whether to buy and put them on the market or not.These samples were subjected both to MA and 1 H NMR profiling.Species, percentage of stem, degree of total impurity, and percentage of specific contaminants were firstly assessed by MA, as usually done by the companies in order to verify the quality and the authenticity of the samples (Table 1).
The most evident morphological difference between O. vulgare and O. onites is the shape of their calices, O. vulgare is characterized by actinomorphic calix (radial symmetry), with five equal lobes (teeth), while O. onites presents a cone shaped zygomorphic calyx (with only one plane of symmetry, bilaterally symmetrical) open on one side (Fig. 1A) (Ietswaart, 1980).According to the MA analysis, out of twenty-seven samples, twelve were O. vulgare, eleven O. onites, two samples resulted composed by a mixture of the two species, and two were recognized as non-marketable species.
After recording the 1 H NMR spectra, a first overview of the results obtained was acquired by performing Principal Components Analysis (PCA), using bucketed 1 H NMR spectra as x variables.The results obtained by PCA analysis agreed with what observed by MA, except for the samples coming from Sicily, thus they were excluded from the first PCA model developed (this result will be discussed later in the text).Excluding all unknown species, sixteen components (PCs) maximized the explained 98.6% of the variance in the data set (given by R 2 x(cum)), while the obtained Q 2 (cum) was 89.6%, indicating very good predictability (Q 2 must be equal or higher than 50%).This PCA model facilitated the detection of similarities/differences among the metabolomic profiles of the samples.As highlighted by the score scatter plot (Fig. 1A), the model was able to make a distinction between the two marketable oregano species along the component t[1], and, consistently, the samples made up of the two species blended were placed in the middle of the plot (violet dots).On the other hand, the shift along the component t[2] was due to sample degree of purity, which decreases along positive t [2].
In order to facilitate the identification of the main biomarkers responsible for the differentiation between the two marketable species, an OPLS-DA (Orthogonal Partial Least Squares-Discriminant Analysis) model was build (Fig. 1B), using as discriminant classes the two species of oregano (O.onites and O. vulgaris) previously identified by MA.Considering the objective of this analysis, the model was performed excluding from the data set the samples constituted by the mix of the two species, and the samples heavily adulterated with other herbs (percentage of contamination higher than 15%).
The OPLS-DA analysis find a perfect fit to the response using two components, with goodness of fit (R 2 y(cum)) of 100% and goodness of prediction (Q 2 (cum)) of 98.5%.Considering that these two parameters have to be as close as possible to 100% (Malzert-Fréon et al., 2010), the obtained model is interpretable and strongly predictable.This was further confirmed by permutation test ( 200permutations) (Fig. S9), giving Q 2 (cum) of 99% and intercept on y axis of Q-line of -0.55; while R 2 (cum) was 99% and intercept on y axis of R-line was 0.19, and significance testing of the model based on ANOVA of the cross-validated residuals (CV-ANOVA) giving p = 4.64 x 10 -17 and F = 313.
Loading plot and S-plot of OPLS-DA model show the relationships between x variables and sample classes, in this case, x were 1 H NMR spectral signals and classes were oregano species (Fig. S9).In order to elucidate the molecular structure of the metabolites responsible for these 1 H NMR signals, they were isolated by NMR-guided fractionation and subjected to 1D, 2D-NMR and ESI-MS analysis.After this procedure, salvianolic acid B was found to be a biomarker of O. vulgare, while the flavonoid apigenin and the terpene p-cymene resulted the biomarkers of O. onites (Fig. 2 and Fig. S10).To the best of our knowledge, this is the first detailed report of the main biomarkers useful to distinguish between these two oregano species.Although also Blake et al. 2016 developed a method to detect adulteration in oregano samples through both FTIR and LC-HRMS untargeted analysis, their results are only at a pattern level, while no biomarkers identification was reported in their work.
Other aromatic signals (number 15 in Fig. 2) emerged from the OPLS-DA as characteristic of O. onites, however, the metabolite(s) responsible for these signals was not fully characterized in this work.
Besides the possibility to discriminate between O. vulgare form O. onites, metabolomics also individuated non-marketable samples.This was highlighted when the two samples, classified by MA as unknown plant material (thus non-marketable), were added to the data matrix and processed by PCA.
These samples resulted, in fact, outliers (Fig. 1C), since their phytochemical profile was completely different from the marketable oregano species.In this specific case, the non-marketable samples were lacking the main components of the essential oil (among which thymol), and they showed a different aromatic pattern, with high content of amino acid tyrosine, which was not detected in marketable oregano species (Fig. S11).Conversely, rosmarinic acid was detected both in 'true oregano' and in these non-marketable samples.
As above described, the results obtained by PCA and successively by OPLS-DA were consistent with those obtained by MA, making reliable the discrimination of oregano species based on their 1 H NMR profiles.However, in this study, the results of MA and NMR-metabolomics disagreed on three samples (all three coming from Sicily), which were classified by MA as O. vulgare (based on morphological traits), these samples resulted extremely different from both O. vulgare and O. onites when analyzed by NMR-based PCA (Fig. 3A).Thus, according to metabolomics approach, these samples should be considered non-marketable.In particular, instead of thymol (which was not detected) they were enriched in carvacrol, which was not found in the NMR profiles of marketable oregano species (Fig. 3B-C).These two positional isomers were easily distinguished by 1 H NMR spectroscopy, in fact, the signal of the methyl group (in para position to the isopropyl group on the aromatic ring) of carvacrol is more downfield shifted (δ 2.22) compared to the spectrum of its positional isomer thymol (δ 2.14) (Fig. 3C).This spectral feature is due to the β-effect of the hydroxyl group.Conversely, the spectral signal of the same substituent in thymol resulted de-shielded, since the hydroxyl group decreases electron density due to smaller inductive and resonance effect.
These samples are likely a Sicilian subspecies of O. vulgare (or a chemotype), hardly recognizable by dichotomous keys.This result highlights the potentialities of NMR-based metabolomics approach for oregano quality control.

Assessment of samples degree of purity
Two independent OPLS models were built (one for each species of marketable oregano), using as y variable the percentage of total impurity found in the samples by MA (Fig. 4A-B).
For O. vulgare, the model was fitted by six components (R 2 x(cum) = 90.4%,R 2 y(cum) = 96.9%,and Q 2 (cum) = 91.1%).Its predictability was confirmed by R 2 (cum) and Q 2 (cum) obtained by permutation test (200 permutations), which were 96.9% and 91.1%, respectively, while intercept on y axis of Q-line was -1.38, of R-line intercept was 0.696.CV-ANOVA F and p were 9.11 and 0.002, respectively.The value of R 2 obtained in the observed vs predicted plot was 0.9695, indicating a consistency between prediction based on NMR profiling and observations made by MA (Fig. 4A).However, one sample (BO19UOR in Table 1) was considered less pure than what was established by MA, this result might be explained by the content of stem (found in this sample), which likely lower the amount of the biomarkers, leading the model to classify it as less pure.
For O. onites, the model was fitted by four components (R 2 x(cum) = 79.2%,R 2 y(cum) = 98.8%, and Q 2 (cum) = 96.1%).Its predictability was confirmed by R 2 (cum) and Q 2 (cum) obtained by permutation test (200 permutations), which were 98.8% and 96%, respectively, while intercept on y axis of Q-line was -1.07, of R-line intercept was 0.478, and F = 39.85 and p = 6.5 x 10 -8 by CV-ANOVA.The value of R 2 obtained for observed vs predicted plot was 0.9883, indicating, also in this case, a consistency between NMR profiling and MA results, with some discrepancies especially at very low level of impurity (Fig. 4B, S12).
The OPLS models, built for the two Origanum species, individuated thymol and rosmarinic acid as important biomarkers of purity.Thymol had a VIP coefficient (calculated for the signal at δ 2.14) of 2.19 for O. onites and 2.41 for O. vulgare.Rosmarinic acid VIP coefficient (δ 7.11) was 1.69 and 1.60 for O. onites and O. vulgare, respectively.These two compounds were easily recognizable by 1 H NMR profiling, however their molecular structures were confirmed by further NMR and ESI-MS analysis performed on pre-purified fractions obtained by liquid-liquid partition.
Thymol is one of the main components of oregano essential oil (Khan et al., 2019), and rosmarinic acid is reported to be the main phenolic acid found in oregano extracts (Ozkan et al., 2010).In addition to be important for oregano flavor, thymol is also relevant for its numerous biological activities (Dheer et al., 2019).Rosmarinic acid is another important bioactive metabolite of oregano, especially for its antioxidant potential (Guitard et al., 2016;Villalva et al., 2018).

Detection of Cistus creticus impurity
Cistus was the most frequent adulterant identified by MA from its characteristic trichomes (Fig. 5A).
The adulteration of oregano with cistus, specifically Cistus creticus (ex.incanus) L., has been also reported by Marieschi et al., 2010.In order to get the adequate model for the prediction of cistus contamination, samples with specific percentage of cistus were prepared in laboratory, spiking O. onites and O. vulgare samples with diverse cistus percentage (from 2 to 60 %) (Fig. 5A).The NMRmetabolomic profile of these samples was added to the dataset, previously constituted only by the samples provided by the companies.An OPLS model was built using as y variable the percentage of cistus (found in the samples by MA, and known in the samples prepared in laboratory).The model was fitted by four components, with R 2 x(cum) = 76.9 %, R 2 y(cum) = 93.2%,Q 2 (cum) = 88.9%, and it was further validated by R 2 (cum) and Q 2 (cum) obtained by permutation test (200 permutations) (Fig. S13), which were 93.2% and 88.9%, respectively, Q-line intercept on y axis was -0.518 and R-line intercept was 0.198, while F = 47.2 and p = 6.77x 10 -20 given by CV-ANOVA.This OPLS model showed a correlation between samples metabolomic profiles and the used y variable.The analysis of the S-plot (Fig. 5B) highlighted a correlation between percentage of cistus contamination and the intensity of a specific 1 H NMR bucket (from δ 3.45 to 3.49; VIP = 5.16).In fact, the spectral signal resonating at δ 3.47 was present only in 1 H NMR spectra of contaminated samples.J-res experiment confirmed that this signal was a singlet (Fig. S14).Hence, a sample of Cistus creticus L. was harvested and analyzed by 1 H NMR profiling, exhibiting a prominent signal at δ 3.47, which confirmed the result given by the OPLS model (Fig. 5C).Through a pre-purification procedure, performed on cistus extract, it was

Conclusions
Combining results obtained from morphological analysis and 1 H NMR metabolomics, it was possible to rapidly identify non-marketable species of oregano, as well as to discriminate between the two marketable ones (Origanum vulgare and O. onites).Apigenin and p-cymene were the main biomarkers of O. onites, while salvianolic acid B of O. vulgare.The developed PCA model was also able to individuate samples made up of a blend of these two species.
Moreover, sample degree of purity could be predicted by OPLS.Thymol and rosmarinic acid were very important indicators to predict general oregano purity, together with the specie-specific apigenin, pcymene.A further model was built to detect the degree of cistus contamination.
Metabolomics proved to be a valuable approach for oregano quality control, giving information about oregano species and phytochemical composition of samples, which is an important data, also correlated to the organoleptic properties.The obtained results encourage the use of metabolomics to make predictions regarding oregano quality on the basis of samples NMR profile, supporting also the possibility of building NMR motabolomics-based databases, to be used worldwide for oregano quality control, giving a significant number of information in a considerably short period of time.Samples constituted by unknown species or a mixture of the two marketable species were excluded from this analysis.Table 1.Vouchers, origin, and results of the morphological analysis (MA) performed on the samples analyzed in this work.Samples were provided by different companies, who received, in turn, from third parties, who give the information related to sample origin.Samples defined as 'unknown' species have to be considered non-marketable as oregano.Samples coming from Sicily were identified as O. vulgare by MA while as 'unknown' species by NMR-metabolomics (this information has been reported in the Table ).
obtained a fraction containing a compound, whose positive ESI-MS yield m/z: 195 [M + H] + , calculated as 194.18 for C 7 H 14 O 6 and negative ESI-MS was m/z: 193 [M -H] − , corresponding to pinitol molecular weight.This cyclic alcohol, in fact, presents a singlet in NMR spectrum at δ 3.47, due to the methoxy group in position 3 of the ring.

Fig. 1 A
Fig. 1 A) Morphological differences between O. vulgare e O. onites calix, and NMR-based PCA score

Fig. 3 A
Fig. 3 A) PCA score scatter plot where unknown samples (yellow dots) are discriminated from the two

Fig. 4
Fig. 4 OPLS score scatter plot (on top) and observed vs predicted plot (bottom).Percentage of total

Fig. 5 A
Fig. 5 A) Different trichomes of cistus (on the right) and oregano (on the left) and OPLS score scatter