Source of retail samples and preparation of extracts

The seven boldo samples were sourced from various natural stores within the "Central de Abastos" of Mexico City, as well as online providers. Prior to analysis, rigorous cleaning was carried out, with only leaves of uniform appearance and optimal preservation being chosen. Any residual flowers, twigs, or leaves displaying dissimilar characteristics were excluded and analyzed separately. The selected leaves were then finely ground to achieve a consistent particle size (1-4 mm). Subsequently, 4 g of each sample was treated with 25 mL of hot water (80 °C) and allowed to steep for 5 minutes. Following this, the resulting suspension underwent a 15-minute sonication at 60 °C, followed by decantation and filtration through a cotton filter, and centrifugation at 10,500 rpm for 20 minutes. Finally, the supernatant was filtered through a 0.22 µm hydrophilic filter. For each sample, this process was repeated ten times.

Sample preparation for NMR analysis

To prepare the samples for NMR analysis, a 400 µL aliquot of infusion was poured into an NMR tube, followed by 140 µL of phosphate buffer at pH 5 enriched with 1 % sodium azide, and 60 µL of a 10 mM TSP standard solution in D2O. NMR analysis of each brand of boldo tea was performed by ten replicates. A concentration of 1 mM EDTA was added to the samples to observe with better resolution the signals corresponding to citric, formic, malic and succinic acid compared to aqueous extracts without EDTA.

¹H NMR analysis

The ¹ H NMR spectra used for statistical analysis were obtained at 298 K in a Bruker Avance III 750 MHz (17.6 T) spectrometer (Bruker Biospin, Rheinstetten, Germany) equipped with a 5 mm TXI probehead (PA TXI 750S4 H-C/N-D-05 Z) able to perform z-gradient. The following acquisition parameters were used: 4 dummy scans, 256 scans, FID size 64 K, pulse width 8 ms, spectral width 15 ppm, receiver gain 37.9, acquisition time 2.18 s, relaxation delay 5 s and digital resolution 0.46 Hz. For aqueous extracts, ¹H NMR spectra were acquired using the NOESYPR1D pulse sequence to suppress the residual signal of water. Spectra for quantification were acquired at 298 K in a Bruker Avance III 600 MHz (14.1 T) equipped with a 5 mm BBO probehead (PA BBO 600S3 BBF-H-D-05 Z SP) able to perform z-gradient: 4 dummy scans prior to 256 scans, FID size 64 K, spectral width 13,000 Hz, receiver gain 189.9, acquisition time 1.70 s, relaxation delay 5 s, and digital resolution 0.58 Hz.

Analysis by HR-MS-ESI

High-resolution mass spectra were acquired on a Bruker solariX Ion Cyclotron Resonance Fourier Transform Mass Spectrometer (Bruker Daltonics GmbH, Bremen, Germany) equipped with a 7 Tesla superconducting magnet (Magnex Scientific Inc., Yarton, GB) and an APOLO II ESI source (Bruker Daltonics GmbH, Bremen, Germany) operated in negative/positive ionization mode. The FTICR spectra were exported to peak lists with a cut-off of signal-to -noise ratio (S/N) of 6 using the DataAnalysis 5.2 software. Ions were measured in the ICR cell from m/z 43 to 3000.00 using the broadband detection mode with 2 M data points. Precursor ions were excluded after one MS/MS spectrum and released after 0.5 min. The equipment was calibrated according to the mass of boldine, favoring the formation of its molecular ion in the rest of the samples.

Identification and quantification of metabolites

The chemical profiling of boldo was determined by consulting databases such as Chenomx (Chenomx NMR Suite 8.0, Chenomx Inc., Edmonton, Alberta, Canada) (Fig. S1), HMDB (The Human Metabolome Database), PubChem and BMRB (Biological Magnetic Resonance Bank). Identification of metabolites was also supported by performing 2D NMR experiments such as ¹H-¹H COSY, HSQC, J-res and TOCSY.

Metabolite quantification from the metabolome was performed using Chenomx software, referencing the concentration of TSP at 1 mM. Metabolites were quantified at pH 5±0.5, only those that did not show overlapping signals with those of other metabolites. Quantification of metabolites was subjected to analysis of variance and Tukey's test using InfoStat software (v2008).

Pre-processing of the ¹H NMR spectra

The ¹H NMR spectra were baseline corrected, phased, referenced and normalized to the TSP signal (δ 0.0). A 0.30 Hz exponential apodization and Whittaker smoothing were performed with MestReNova software (v14.2.0 -26256; MestReC, Santiago de Compostela, Spain). Integral of the full spectrum was normalized to the TSP signal. All ¹H NMR spectra were stacked, and the binning size was set to 0.04 ppm.

Multivariate statistical analysis

Multivariate data analysis was performed with SIMCA software (v14.1 Umetrics, Sweden). Principal Component Analysis (PCA), Least Squares Regression (PLS) and Orthogonal Partial Least Squares Discriminant (OPLS-DA) were performed using the binning results from MestReNova. This facilitated the discernment of variations between the different samples. A second statistical analysis, scaled with Pareto, was performed using the concentration data obtained with the Chenomx software, which allowed to observe the relationship between metabolites and their presence or absence in each boldo leaf sample. The PLS and OPLS-DA models were validated by permutation analysis, while the discriminating metabolites were identified by their VIP value (-1≤ VIP ≥1). S-Plot and B-Plot Figures were constructed by considering their p(corr) values (-0.5≤ p(corr) ≥0.05). A third statistical analysis was performed comparing the two most different samples with each other. The statistical analysis performed at 600 MHz on the entire metabolome revealed a consistent trend compared to the analysis conducted on the 750 MHz equipment. This consistency indicates the reproducibility of the findings, irrespective of the magnetic field used (Fig. S2).

Statistical multivariate analysis

Principal Component Analysis (PCA) accounted for 89.3 % of the total variance (R²X = 0.938 and Q² 0.928). In particular, it showed a clear tendency to separate two main groups: one comprising samples 2 and 4, and the other including samples 1, 3, 5-7. Notably, sample 1 stood out from the second group (Fig. 4(a)). This pattern became even more pronounced when supervised Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was applied to the NMR data (Fig. 4(b)). In this case, components 1 and 2 accounted for 87.1 % of the total variance (R²X = 0.981, R²Y = 0.829, and Q² = 0.814).

Fig. 4 Multivariate analysis of the infusions of the seven retail samples of Peumus boldus leaves. (a) PCA and (b) OPLS-DA scores plot. (c) Loading plot showing metabolites with the most significant influence. (d) Hierarchical clustering revealing the grouping between the different boldo tea samples. 

The loading plot (Fig. 4(c) ) and hierarchical clustering (Fig. 4(d)) were generated based on the outcomes of the OPLS-DA analysis. The loading plots visually depict the metabolites that exert the greatest discriminatory impact, specifically those that significantly determine the variance of the data. Notably, amino acids such as asparagine, glutamine, proline, alanine, valine, tyrosine, phenylalanine, as well as malic, succinic, and acetic acids where shown to be the metabolites with the highest discriminatory power.

To quantify their impact, we rely on the Variable Importance in Projection (VIP) scores. Among these, asparagine stands out with a VIP score of 3.10, making the most pronounced contribution to intergroup variance. In contrast, phenylalanine (VIP score: 1.26), exerts a comparatively lower discriminatory effect. The dendrogram depicted in Fig. 4d visually illustrates the hierarchical relationships among sample groups, organized according to their similarities or differences in multivariate data profiles. Within this graphical representation, the initial cluster comprises sample 2 and sample 4, while the second cluster encompasses all remaining samples. As was observed in the scores plot generated from PCA and OPLS-DA (Fig. 4(a- b)), within the hierarchical cluster, sample 1 exhibits the highest variance compared to the other samples in its group. Conversely, samples 3, 5 and 7 exhibit the closest similarity to each other within this framework. The validation of the OPLS-DA models was performed by permutation analysis (Fig. S3).

The inner relationship analysis revealed that samples 2 and 6 exhibit the greatest dissimilarity among all samples (Fig. S4). An s-plot derived from the OPLS-DA (R²X = 0.998, R²Y = 0.998 and Q² = 0.996) analysis of samples 2 and 6 indicated negative correlation in the levels of asparagine (-0.993), alanine (-0.993), glutamine (-0.994), proline (-0.994), and tyrosine (-0.981). Therefore, the concentration of these metabolites decreases between the samples. Conversely, there was a positive correlation with succinic acid (0.730) and formic acid (0.112), leading to an increase in their concentration between the samples.

Quantification of the identified metabolites

Relative quantification was performed using the Chenomx software. Free amino acids play a pivotal role in defining the distinctive flavor and aroma profiles. Among the twenty-six varieties of amino acids identified in the infusion, accounting for the 1 % to 4 % of the plant’s dry weight ²², proline (1.42 - 5.68 µM) is the most abundant amino acid in boldo infusion. Conversely, alanine (0.17 - 0.34 µM) and phenylalanine (0.13 - 0.18 µM) exhibit the lowest concentration. Tyrosine (0.29 - 0.60 µM) and valine (0.27 - 0.48 µM) are present in minimal concentrations, while glutamine (1.61 - 3.34 µM) and asparagine (0.85 - 1.86 µM) fall in between as aminoacids of intermediate concentration levels (Fig. 5).

Fig. 5 Relative concentration (μM) of metabolites identified in boldo infusion. Lowercase letters represent the significant differences (ANOVA, p-value ≤ 0.05, ±SD) between different tea brands. 

Alanine, arginine, asparagine, aspartic acid, glutamic acid, isoleucine, histidine, leucine, phenylalanine, serine, threonine and tyrosine are the free amino acids found in each infusion ²³. Therefore, the discriminating amino acids defining boldo infusion encompass alanine, asparagine, phenylalanine and tyrosine. These amino acids, crucial for both the flavor profile and health benefits of the infusion, significantly impact its quality ²⁴. Moreover, they serve as vital precursors, facilitating the generation of aromatic compounds via Maillard reactions, oxidations, and degradation processes ²⁵.

Research has shown that leucine can serve as a precursor in the biosynthesis of other alkaloids. It can transform into the respective aldehyde and then combine with dopamine, thereby leading to the formation of isoquinoline-derived alkaloids ²⁶. Some amino acids are important precursors of essential metabolites in plants. An example of this is glycosylates, sulfur-rich natural products that, upon conversion by thioglucosidases through hydrolysis, yield compounds as isothiocyanates, thiocyanates, and nitriles. These compounds exhibit anticancer properties or act as pesticides ²⁵ and can be synthesized from branched chain aromatic amino acids, methionine or alanine, which are notably found in boldo infusion ^27,10. Tyrosine arises through the shikimic acid pathway, giving rise to a diverse array of molecular structures such as phenylpropanoids, betalains, catecholamines, tocopherols and others derivates. These compounds hold significant nutritional and pharmacological value ²⁸. Aspartic acid plays a pivotal role in plants as a precursor of various metabolites, including threonine, leucine and glycine, which are crucial for photorespiratory functions ²⁹.

The most abundant short-chain organic acids are citric (5.04 - 7.91 µM), glyceric (4.59 - 7.80 µM) and malic (2.96 - 5.36 µM) acids, while succinic acid (0.05 - 0.17 µM) is present in lower concentration. Those found in intermediate concentration are malonic (0.81 - 1.16 µM), acetic (0.54 - 1.10 µM) and formic (0.28 - 0.83 µM) acids (Fig. 5). Studies indicate that asparagine can be catabolized to produce malic acid. This process involves converting asparagine to oxosuccinamic acid, which is further reduced to hydroxysuccinamic acid, ultimately leading to the production of malic acid ²⁷. While malic acid could not be quantified due to signal resolution, it was identified through other NMR experiments.

Glucose (16.09 - 29.76 µM) is the predominant sugar in the metabolome, while sucrose (1.11 - 5.17 µM) stands out as the sugar with the lowest concentration within boldo infusion. Notably, fructose (12.18 - 16.47 µM) exhibits the least variation among samples (Fig. 5). Both glucose and fructose serve as metabolites primarily responsible for fueling diverse energy-dependent processes ²¹. In general, the samples to prepare the boldo infusion were those selected based on morphology related to the Peumus boldus Molina variety. These presented the same metabolomic profile (Fig. 1), whose main difference was found in the relative concentration of metabolites, which may be related to the storage time of the P. boldus samples ³⁰.

B-plot (Fig. 6) resulted from an OPLS-DA analysis (R²X = 0.912, R²Y = 0.672, and Q² = 0.403) as outlined in the supplementary materials (Fig. S5), derived from the individual quantification of ¹H NMR spectra using the Chenomx software. Samples 1 and 3 exhibit negative correlations with sucrose, glyceric acid, and malic acid, whereas sample 5 shows positive correlations with these metabolites but negative ones with citric acid. On the other hand, samples 6 and 7 demonstrate negative correlations with valine, asparagine, glutamine, proline, phenylalanine, alanine, succinic, malonic, acetic acid, formic acid, glucose, and fructose. In contrast, sample 2 displays positive correlations with the same set of metabolites.

Fig. 6 B-Plot of the quantified metabolites in the boldo infusion metabolome. 

The infusion comprises numerous chemical components, such as amino acids, polyphenols, polysaccharides, volatile oils, vitamins, minerals and alkaloids ²⁶. Upon examining the correlations among the identified metabolites (Fig. 6), it is observed that the concentration of citric acid rises as the level of sucrose, while malic and glyceric acids decrease. Moreover, tyrosine, glucose, phenylalanine, alanine and glutamine show positive correlation with each other, suggesting a potential connection along the same metabolic pathway or related pathways.

Mass spectrometry analysis

The rising demand for medicinal plants has encouraged their commercialization. Consequently, ensuring the accurate identification of these plants is crucial. Mistaken identity can have severe repercussions, ranging from a lack of the intended therapeutic benefits to potential intoxication. The species Peumus boldus Molina are known as true boldo, while Plectranthus ornatus³¹, Plectranthus neochilus³², Plectranthus amboinicus³³, Plectranthus barbatus³⁴, Plectranthus grandis³⁵, Plectranthus ornatus³⁶, Vernonia condensate³⁷, and Coleus barbatus³⁸ are identified as false boldo. The morphological similarity between these species and the fragmentation of the dried leaves favors the false commercialization of Peumus boldus Molina ³⁹. The marketing of adulterated samples may contain one or more mixed species or even complete substitution of the correct one, thus having a lower antioxidant potential or even health disorders. For instance, Coleus barbatus has been used in folk medicine to interrupt pregnancy, delay fetal development and inhibit implantation; indeed, these effects justify its popular use ³⁸. It is evident the importance to identify biomarkers to detect intentional or mistaken adulterations in medicinal plants available for human consumption.

Boldine is an aporphine alkaloid present in authentic samples of Peumus boldus Molina ⁴⁰. It is a strong antioxidant with several pharmacological activities, such as anti-inflammatory, antipyretic, antiatherogenic, antiplatelet, antitumor, cytoprotective and tyrosine inhibitor ⁴¹. Although boldine is neither the only nor the most abundant alkaloid in boldo leaves, it is an important biomarker to authenticate Peumus boldus Molina ³⁰.

In this sense, a complementary mass spectrometry analysis was carried out with a SolariX Bruker equipment in order to identify boldine or its isomers such as isoboldine, laurotetanine or norisocorydine. Through positive ionization, it was possible to observe the molecular ion corresponding to boldine or boldine isomer in all samples (m/z = 328.15) and N-methyllaurotetanine (m/z = 342.17) or its isomer (Norglaucine or Isocorydine) (Fig. S6), it is the main alkaloid present in boldo leaves ³⁰.

Positive ionization results were analyzed using MetaboScape software (2022b), through which another 10 characteristic metabolites of true boldo were identified according to the literature ^30,42,43: chlorogenic acid (m/z = 355.08306), coclaurine (m/z = 286.14311), epicatechin (m/z = 291.08672), epicatechin gallate (m/z = 443.17429), laurolitsine (m/z = 314.14603), N-methylcoclaurine (m/z = 300.14712), pronuciferine (m/z = 312.16252), quercetin (m/z = 303.10240) and reticuline ( m/z = 330.16861). In negative ionization mode, 11 metabolites characteristic of the authentic boldo plant were identified ^43,44,45: 6α,7-dehydroboldine (m/z = 324.13040), bornyl acetate (m/z = 195.04947), caffeic acid (m/z = 179.05491), camphene (m/z = 135.02995), catechin (m/z = 289.07215), chlorogenic acid (m/z = 353.14590), epicatechin (m/z = 289.06862), kaempferol (m/z = 285.07597), p-cymene (m/z = 133.01351), palmitic acid (m/z = 255.22888) and thymol (m/z = 149.04958).