Skip to main content

Biodynamic, organic and integrated agriculture effects on cv. Italia table grapes juice, over a 3-year period experiment: an 1H NMR spectroscopy-based metabolomics study



The new trend demanding for “natural” agri-food products has encouraged the application of more sustainable and eco-friendly farming methods, which limit or avoid the use of synthetic chemicals. This approach is increasing in viticulture, one of the sectors with the highest commercial value since grapes and derived products are largely consumed foodstuffs, with appreciated nutritional and sensory features. In this work, 1H Nuclear Magnetic Resonance spectroscopy (1H NMR) was applied for the metabolic profiling of cv. Italia table grapes samples, from the same origin area, cultivated with different treatments (biodynamic, organic and integrated) and collected in three subsequent vintages. Multivariate statistical analysis was performed on NMR-data with the aim of comprehensively researching the possible influences on metabolites due to the use of diverse agricultural practices.


Both inter-annual variability (2020, 2021 and 2022 vintages) and different vineyard treatments (biodynamic, organic and integrated) resulted as significant drivers for samples differentiation in the preliminary unsupervised analysis of the (1H NMR spectra derived) metabolic profile data. Nevertheless, supervised data analyses showed that inter-vineyards variability, due to application of diverse farming methods, had a comparable discriminating effect with respect to harvesting years. Ethanol, sugars (as α-/β-glucose), organic acids (as malate) and amino acids (as arginine, leucine, glutamine) resulted the most viticultural practices-dependent metabolites. Interestingly, results from pairwise comparisons between treatments indicated the biodynamic samples with respect to the organic ones as the best-observed differentiation. This was followed by the biodynamic vs integrated and organic vs integrated samples comparisons, in decreasing discrimination order, as confirmed by the descriptiveness and predictive ability parameters of the corresponding pairwise OPLS-DA models.


Results highlighted that metabolites’ composition in cv. Italia table grapes juice is significantly affected by the use of different kinds of vineyard managements (biodynamic, organic and integrated, here investigated). Metabolomics study, here employing 1H NMR spectroscopy combined with multivariate statistical analysis, offers powerful tools to elucidate the metabolic differences among classes of samples.

Graphical Abstract


Grapes are considerably appreciated from their consumption as table fresh fruit to the production of wines, juices and jams. Thanks to its interesting versatility, this agri-food product is one of the most widely cultivated crops in the world. Its success is also due to vine’s ability to adapt to very different pedoclimatic conditions, allowing it to conquer large areas of cultivation [1]. In 2022, the Italian land intended just for the production of table grapes amounted to 47 583 hectares and the total harvest was of 9 662 593 quintals [2]. Grapes have a great nutritional value thanks to their content in soluble sugars, minerals, nutraceutical and bioactive compounds [1, 3]. Moreover, the specific chemical composition can be influenced by various factors, including the cultivar, the farming method, the local climate [4], the microclimate in the vineyard [5, 6] and the soil features [7,8,9,10].

Regarding to farming method, a renewed awareness in consuming quality foods, in preserving both agro-ecosystems health and a good harvest yield, as well as in promoting a sustainable use of soils and resources, encourages study and adoption of environmentally friendly agronomic strategies [11, 12]. This is also a key issue to support the needs of a fast-growing human population [13]. Some of the contemporary regenerative farming methods are biodynamic and organic, which offer more sustainable food production systems in comparison to the industrial model. Industrialized production may obtain high yields over the short term through, for example, soil maximization and the help of chemical compounds. Nevertheless, over the long term, this may occur with high costs associated with losses of soil fertility, biodiversity and crop nutritional quality [14, 15]. The diffusion of more sustainable agricultural practices is a remarkable trend in viticulture [16,17,18]. These are usually accompanied by other management techniques in the vineyard aimed at improving production (e.g., trunk girdling or thinning of leaves, bunches and berries). Apart from the considered eco-friendlier agricultural methods, including biodynamic and organic, where the use of chemical compounds is strictly limited or banned, other methods seek to use strategies that are as sustainable as possible. One of them is the integrated treatment, which employs a low-pesticide-input pest management [19]. Large differences among vine cultivated with diverse treatments, resulting in a most vigorous growth for integrated protocol, were found by Meissner et al. [20]. Similarly, Döring et al. [21] assessed an increase of growth and yield for the integrated viticulture in comparison to the organic and biodynamic grapevines. These latter appeared characterized by higher disease incidence and severity of downy mildew, which partially accounted for the observed results [21]. By contrast, grapes from biodynamic and organic vineyards seemed to be more resistant to deterioration when compared with those ones from integrated treatment [22]. Furthermore, biodiversity at different trophic levels was improved in organic and biodynamic viticulture compared to integrated cultivation, including an increase in earthworm abundance [20, 23]. However, stimulation of soil nutrients cycling requires years to produce an important effect on N levels and microbial activity in the soil [23]. This is usually accomplished by application of compost, use of varied cover crop mixtures and denial of chemical compounds, as promoted in organic and biodynamic viticulture.

In particular, if biodynamically and organically managed vineyards are compared, Reeve et al. [24] observed no differences in yield, in soil quality and in microbial efficiency. On the other hand, in the biodynamic grapes, a great quality was also specifically found, in terms of significantly higher levels of soluble solids (i.e., °Brix), total phenols and anthocyanins [24]. Regarding to the physiological responses of grapevines, lower stomatal conductance and lower leaf water potential were found in biodynamic with respect to organic Vitis vinifera L. cv. Sangiovese. Moreover, an increase in leaf enzymatic activities of endochitinase, exochitinase (β-N-acetylhexosaminidase and chitin 1,4-β-chitobiosidase) and β-1,3-glucanase, which are typically associated with plant biotic and abiotic stresses and with induced plant resistance, was also observed in biodynamic plants in comparison to those ones organically managed [25].

Influence of different agricultural management systems on the metabolites fingerprint of grapes was confirmed by some studies [24, 26, 27], although other authors did not observe significant variations in the metabolome of samples coming from different viticultural methods [28, 29]. NMR-based metabolomics studies on grapes from diverse agricultural practices are few in literature, mainly concerning winegrapes and wines [30,31,32]. By 1H NMR, Picone et al. [32] evidenced various winegrape berries (cv. Sangiovese) metabolic profiles that could be correlated with the different physiological response of plants to the biodynamic and organic treatments. About table grapes, Gallo et al. [33], using 1H NMR, evaluated the effects of conventional local practices (CONV) and organic management on the chemical composition for cv. Superior Seedless, as well as the inter-vineyards variability for cv. Red Globe grown with different farming techniques established by each producer. They observed that sugar concentrations were appreciably affected by the application of different cultivation practices, which, instead, resulted to have a less pronounced effect on organic acids content.

High-resolution Nuclear Magnetic Resonance (NMR) spectroscopy is normally combined with multivariate statistical analysis (MVA) in metabolomics studies. NMR spectroscopy is a very powerful tool, generally characterized by rapid sample preparation procedures and extremely reproducible results [34]. Nowadays, high field NMR instruments, gradient assisted and equipped with cryo-probes, also allowed to overcome the original low sensitivity problems with respect to other analytical techniques. Well-suited for metabolomics studies, capturing a “snapshot” of the metabolic profile of the analyzed sample at a given time, NMR spectroscopy provides detailed information on metabolites in complex mixtures. In the food science, NMR offers a wide range of applications: from studies of traceability, authenticity and safety of food [35, 36] to the monitoring of wine alcohol fermentation and aging [34], including also the analysis of a large variety of agri-food matrices [37,38,39,40,41,42,43,44]. The present work proposes a comparative study of metabolic profiles for cv. Italia table grapes juice from biodynamic, organic and integrated products, supplied by certified producers. Thus, this study may represent a useful completion of the investigations already carried out, based on the NMR-based profiling of grapes juice [32, 33], including for the first time a direct comparison of biodynamic (BD), organic (ORG) and integrated (INT) agriculture. The general biodynamic definition reported in the present work, specifically refers, for the investigated case, to the organic vineyards certified Reg. CE 848/18 [45] DEMETER production rules [46].


Origin of grape berries samples and sampling methods

A total of 210 grapes samples was obtained from bunches of Vitis vinifera cv. Italia table grapes, harvested at maturation according to the usual commercial practices. Grapes were collected from 2020 to 2022. They came from different specialized Italian farms placed in the same limited geographical area “Masseria Gaudella” included in the countryside of Castellaneta (Taranto province), in the south of Apulia Region (Italy) (Fig. 1a). Their geographic coordinates are: 40°34ʹ46.13″N 16°52ʹ6.20″E (BD farm), 40°34ʹ26.43″N 16°53ʹ0.61″E (ORG farm) and 40°34ʹ52.23″N 16°53ʹ10.57″E (INT farm). The vineyards selected for the study are located within a radius of approximately 1km (Fig. 1b), to homogenize as much as possible the pedoclimatic characteristics of the grapes growing area, as well as to minimize the possible variability effects of both occurring climatic conditions and soil properties on grape metabolism. According to the regions and provinces of Italian soil map [47], the selected vineyards are characterized by similar features with specific homologous pedological classification, as indicated in a detailed portion of the local map [48], more recently confirmed [49] (Additional file 1: Figure S1) and available from the regional database [50]. On the other hand, the soil characteristics could be influenced by the different land use in relation to the agronomic practices adopted, in particular for long-lasting management protocols, as recently demonstrated for the specific area of investigation [51]. The vine training system was a covered tendone with a layout measuring 3x3 m. The average of cv. Italia yield was, respectively, ca. 20–25 tons ha-1 year-1 for BD vineyard, ca. 20–30 tons ha-1 year-1 for ORG vineyard and ca. 35–50 tons ha-1 year-1 for INT vineyard. The field sampling was carried out during the ripening period of the cv. Italia grapes corresponding to the first half of September, collecting two bunches per plant. Selection of the plants to be sampled was randomized (as represented in the scheme, Fig. 1c) to guarantee for each vineyard a representative set of samples in the identified plots. The vineyards were managed according to the biodynamic, organic and integrated farming methods. Supplier farms are certified producers, in particular the biodynamic (BD) farm has organic vineyards certified Reg. CE 848/18 [45] DEMETER production rules [46], organic (ORG) farm has organic vineyards certified Reg. CE 848/18 production rules and integrated (INT) farm has integrated pest management (IPM) vineyards certified GDO production rules [52]. Questionnaires, based on bibliographic research [20,21,22, 53] and consisting in open/close-ended questions (Additional file 1: Figure S2), were supplied to farms to know detailed information about their viticultural practices (reported in Table 1). Immediately after collecting, the bunches of grapes were transported to the laboratory in three separated lots, respectively, of BD, ORG and INT grapes. In the same day of field sampling, fresh berries were selected for the preparation of samples (consisting in 10–15 berries), which were rapidly stored in freezer at − 20 ℃. For each farming method, berries were randomly selected from the bunches, to be representative of the total studied population and to minimize bias. During the experiment, every harvesting year, homogeneous sample replicates were derived from each provided lot (10 samples for 2020 which were increased to 30 samples for 2021 and 30 samples for 2022), reaching a total amount of 30 samples for 2020, 90 samples for 2021 and 90 samples for 2022, as reported in Table 2.

Fig. 1
figure 1

Geographical area of samples origin, vineyard sites and sampling methods. a Map of Italy with place of origin of analyzed table grapes samples; b vineyards location area; c scheme of field sampling, green = rows of vine plants, yellow squares = sampled plants, brown = inter-row space

Table 1 Summary of treatments indicated by supplier farms on questionnaires
Table 2 Summary of analyzed samples divided per year and farming methods

Samples preparation for NMR analysis

From each sample, few grape berries were defrosted and squeezed to obtain the grapes juice. Then, 100 μL of phosphate buffer (1M KH2PO4, 0.05 % TSP, trimethylsilylpropanoic acid, as internal standard) and 2mM NaN3 to prevent microbial contamination, at pH 3.3, according to Girelli et al. [54], were added to 900 μL of juice and subsequent centrifuged (at 10.000 g at room temperature for 5 min) to separate the solid phase from the liquid one. For 1H-NMR direct analysis on juice, 700 μL of the supernatant were used to fill a 5 mm NMR tube.

NMR experiments and data processing

NMR spectra were acquired using a Bruker Avance III 600 Ascend NMR spectrometer (Bruker, Ettlingen, Germany), operating at 600.13 MHz for 1H observation, using a TCI cryo-probe, equipped with a z axis gradient coil and automatic tuning-matching (ATM). After loading individual samples on a Bruker Automatic Sample Changer interfaced with the software IconNMR (Bruker), all the experiments were carried out at a temperature of 300 K in automation mode. An automated procedure consisting of locking, tuning, matching and shimming with a zgcppr Bruker standard pulse sequence was used for the spectra acquisition, alongside NOESY experiment to suppress the residual water signal. A total of 64 transients (with 16 dummy scans) were collected into 64 k data points with relaxation delay set to 5.0 s. A spectral width of 20.0276 ppm (12019.230 Hz) and an acquisition time of 2.7262144 s were used. Free induction decays (FIDs) were Fourier transformed, after being multiplied by an exponential weighting function corresponding to a line broadening of 0.3 Hz. For every spectrum, the phase was manually adjusted, the baseline was corrected and the calibration was done by aligning TSP singlet to 0 ppm. All acquired 1H NMR spectra were processed by using Topspin 3.5 (Bruker), which also allows for simultaneous visual inspection. The metabolites were assigned to the spectral peaks on the basis of 2D NMR spectra analysis (2D 1H Jres, 1H COSY, 1H–13C HSQC and HMBC) (Additional file 1: Figures S3, S4, S5) and by comparison using Chenomx software (Chenomx Inc., Edmonton, Canada) and literature data [8, 32, 33, 54,55,56,57,58]. The acquisition and processing of 2D spectra were performed as follows: a cosygpprqf Bruker standard pulse sequence with presaturation relaxation delay using gradient pulses for coherence selection allowed to acquire a bi-dimensional 1H COSY spectrum. A spectral width in both dimensions 12,019 Hz (20.028 ppm), 2048 data points in f2, 1024 increments in f1, processed with an unshifted sine-bell squared (QSINE) window function in both dimensions before Fourier transform. An 1H homonuclear J-resolved spectrum was obtained with a spectral width of 12 kHz for f2 dimension and 80 Hz for f1, 8096 data points in f2, 256 increments in f1, processed with zero filling in f1 to 4096 real data points, unshifted sine-bell squared window functions in both dimensions before Fourier transform. The 1H–13C gradient-selected HSQC spectrum was acquired with 1H–13C decoupling, 9 and 37 kHz spectral widths in the 1H and 13C dimensions, respectively, 2048 data points in f2, 256 increments in f1, forward linear prediction with 32 coefficients, zero filling to 4096 data points for the f1 dimension. Unshifted sine-bell squared (SINE) window functions were also applied in both dimensions before Fourier transform. The 1H–13C HMBC spectrum was obtained with 9 and 45 kHz spectral widths in the 1H and 13C dimensions, respectively, 2048 data points in f2, 512 increments in f1, forward linear prediction with 32 coefficients, zero filling to 4096 data points in f1. Unshifted sine-bell squared window functions were also applied in both dimensions before Fourier transform.

Multivariate statistical analysis on NMR spectroscopy data

The 1H NMR spectra were divided into segments, called buckets or bins, of fixed 0.04 ppm width. The bucketing was performed by Amix 3.9.13 (Analysis of Mixture, Bruker, Biospin, Italy), as the subsequent step in NMR data processing. The spectral region between 4.78 and 5.00 ppm was excluded due to the presence of residual water signal and the remaining 233 buckets in the range 10.00–0.48 ppm were normalized to total area, mean-centered and integrated by Amix software. The total sum normalization was applied to minimize small differences due to sample concentration and/or little changes in experimental conditions among samples. After the alignment of buckets rows reduced spectra, the resulting data table was used in the MVA, employing the tools offered by Simca-P version 14 (Sartorius Stedim Biotech, Umetrics, Umeå, Sweden). In this work, a particularly careful alignment process was required to overcome shifts of some organic acid signals, despite the use of the buffer. Nevertheless, essentially for check purposes, a further data table was produced after manual adjustment of some buckets by merging, into wider buckets, bins containing signals of tartaric (4.5; 4.46 ppm), malic (2.62; 2.66; 2.70 ppm) and citric (2.74; 2.78 ppm) acids, without any overlapping neighboring resonances [59]. The results from analogous MVA performed on the latter data table did not show significant variations with respect to the conclusions from the statistical analyses on NMR data with regular bucketing (fixed buckets of 0.04 ppm width), here reported.

A data table consists of rows (the observations or samples) and columns (the variables or buckets). Buckets represent the variables used as descriptors for each sample in chemometric analysis. The Pareto scaling method was applied to the variables, which is performed by dividing the mean-centered data by the square root of the standard deviation [60]. Then, the data matrix was analyzed according to unsupervised and supervised pattern recognition methods useful to have an overview of the intrinsic data variation. The first multivariate analysis is the Principal Component Analysis (PCA), an unsupervised method, which provides a summary of all observations, founding groupings, trends, outliers [61]. Starting from the data table, PCA can extract and show the systematic variation in a data matrix X, without using information on group identity to construct the model. Supervised methods, such as the Partial Least Squares Discriminant Analysis (PLS-DA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) are, then, used to discriminate samples grouped in classes with different characteristics. The PLS-DA allows to stress the maximum separation between groups of observations and to obtain information about the variables responsible for the observed separation. It is performed by rotating the main components, that are the axes expressing the variance of the data [62]. A modified PLS-DA method, that is the OPLS-DA, is able to powerfully reveal the most discriminating variables, filtering out variation that is not directly related to the focused discriminating response. For this purpose, the portion of the variance useful for predictive comparisons is separated from the not-predictive, which is, therefore, considered as orthogonal. As a result, the model interpretability is considerably improved [63]. The statistical models were validated using the internal cross-validation default method (7-fold) and the permutation test (100 permutations), available on the SIMCA-P software [64]. In particular, the permutation tests were used to validate the PLS-DA and OPLS-DA models, checking the degree of overfit and the possibilities of false-positive findings. Models were considered valid if: the Q2 value on the actual data set is higher than all Q2 values from the permuted data set, as well as the Q2 regression line has a negative value of intercept on the y-axis [65]. There are some parameters useful to evaluate the quality of the MVA models: R2, Q2 and p [CV-ANOVA]. The first (R2) represents the part of the original data that is explained by the model. Therefore, it is a measure of the overall fit of the model. The second (Q2) is the part of the original data which the cross-validated model explains. It is a measure of the predictive power of the model, describing how well it can predict correct classification for a test sample. In supervised methods, the cross-validated analysis of variance (CV-ANOVA) provides a p value, a tool to understand the level of significance of the observed group separation analysis. An overview on these parameters allows to evaluate the model robustness. An ideal model should have closest to unity and similar R2 and Q2 values. In a real MVA, the model shows Q2 value lower than R2, but if it is particularly lower the robustness is compromised [66]. Score scatter plot, loading scatter plot and S-line plot are tools available on SIMCA-P software to display the results and they were performed for the developed models. The score scatter plot displays the scores, which summarize the relationship among the observations of a model, in a 2D (or 3D) scatter plot revealing the possible presence of outliers, groups, and other patterns in the data. The loading scatter plot shows in a 2D (or 3D) scatter plot the contribution of the original variables, describing each sample (in our case the buckets obtained from its NMR spectrum), in determining the sample position in the score scatter plot. The S-line plot for an OPLS-DA model with two classes visualizes the centred loading vector p(ctr) colored according to the absolute value of the correlation loading, p(corr). Changes in the metabolites content between two groups of observations can be evaluated as relative quantification. For this purpose, the Fold Change (FC) ratios of the normalized median intensity of selected bucket-reduced NMR unbiased signals of discriminant metabolites in the pairwise comparisons of treatment were calculated and reported as Log2FC [54]. The statistical significance of the differences between the means, for each variable of the two groups, was assessed using the Student t test. A p value <0.05 (confidence level 95%) was considered statistically significant [54]. In this work, only metabolites with variable importance in projection (VIP) computed from all extracted components and correlation coefficient p(corr) absolute values higher than 1.0 and 0.5, respectively, were considered as potential significant discriminant compounds [67, 68] and taken into account for relative quantification.


Cv. Italia table grapes composition

A representative spectroscopic fingerprinting of cv. Italia table grapes juice extracts can be observed in Fig. 2, where one of the acquired 1H NMR spectra is reported. A summary of the assigned peaks is reported on Table 3. To better visualize the peaks and the metabolites assigned to their relative chemical shifts, three main regions can be identified in the 1H NMR spectra of table grapes juice: the aliphatic, the carbohydrate and the aromatic region [32]. The aliphatic region (range 0.9–3.2 ppm) is characterized by resonances mainly ascribable to amino acids and organic acids, which remarkably influence the organoleptic properties of grapes (Fig. 2a). Here, the most intense signals belong to malate, but a great variety of peaks of lower intensity corresponding to valine, leucine, isoleucine, lactate, alanine, arginine, proline, GABA, glutamine and citrate are also present. Moreover, the methyl signal of ethanol is detected, which gives important information about the active metabolic changes in grapes, such as some fermentation pathways of sugars occurring during fruit ripening and responsible for the production of organic acids [3]. The most intense peaks are recorded in the middle field region of the spectrum (range 3.2–5.6 ppm) (Fig. 2b), that is, the carbohydrate region. The dominant resonances of the whole spectrum, indeed, belong to sugar compounds that are present in high concentration in the grapes juice extract. Then, the aromatic region, ranging from 5.6 to 10 ppm (Fig. 2c), shows signals assigned mainly to aromatic compounds: syringate, pyridine alkaloids as trigonelline, amino acids as phenylalanine and tyrosine. Very important are the peaks associated with polyphenols as the hydroxycinnamic acids that have well-known antioxidant properties [69]. Moreover, the signals ascribable-to-not-aromatic compounds, such as formate and fumarate, are also present.

Fig. 2
figure 2

Typical 1H NMR spectrum of cv. Italia table grapes juice sample. Main resonances are labelled with the assigned metabolite. a Aliphatic region, mainly characterized by signals of amino acids and organic acids; b carbohydrate region, collecting mainly signals belonging to sugars and c Aromatic region (approximate zoom factor × 10), where signals principally of phenolic and other aromatic compounds are present

Table 3 1H-NMR chemical shift (δ) of metabolite resonances of cv. Italia table grapes juice

Statistical analysis

To reveal the possible data grouping of the grapes samples, an unsupervised PCA was applied on the whole cv. Italia NMR data set, where all the vintages (2020, 2021 and 2022) were considered. In the present PCA analysis [R2X (cum) = 0.829, Q2(cum) = 0.728], five components explained 82.9% of the total variance (38.9%, 23.7%, 9.9%, 5.6% and 4.7% for t[1], t[2], t[3], t[4] and t[5], respectively), describing the samples distribution in the space. The same t[1]/t[2] score scatter plot of the model was labeled according to both vintage (Fig. 3a) and farming method (Fig. 3b). Two main clusters are visible in Fig. 3a: the group of the vintages 2021–2022 shifted at more positive values of t[1] with respect to the group of the vintage 2020 mainly at negative values. Moreover, a further separation of samples can be observed along t[2]: the group of the vintages 2020–2021 at positive values of t[2] and the group of the vintage 2022 at negative values. A less evident clustering according to the different agricultural practices is observed in Fig. 3b. Nevertheless, an indication of differentiation along t[2], between the ORG and INT samples, could be observed in the t[1]/t[2] PCA score scatter plot (Fig. 3b). To visualize the separation of the BD samples with respect to the others, the analysis of the 3D PCA score scatter plot also including t[3] was required (Fig. 3c). Interestingly, at first glance, the observation of the 3D PCA score scatter plot (Fig. 3c) suggests the possible existence of specific characteristics in the metabolic profiles, differentiating essentially BD from ORG samples. Loading scatter plot for the PCA model was also performed, aiming to reveal the metabolites responsible for the observed separation of the samples groups (Additional file 1: Figure S6).

Fig. 3
figure 3

PCA score scatter plot for all grapes juice samples. Model parameters: five components give R2X (cum) = 0.829, Q2(cum) = 0.728; PCA score scatter plot a colored by vintage and b colored by farming method; c 3D PCA score scatter plot; BD = biodynamic, INT = integrated, ORG = organic agriculture

The observed trends (Fig. 3) in the PCA analysis indicated that the inter-annual variability due to the occurring climatic conditions might influence the metabolism of grapes, besides those produced by other factors, such as the application of different cultivation methods. However, to clearly identify and possibly quantify all the differences resulting from the 1H NMR metabolic profiling of grapes juice samples, supervised MVA methods were, therefore, used.

The separation among samples classes could be improved by PLS-DA analysis. Two supervised (PLS-DA) analyses were performed and compared, based on the PCA clustering indications (Fig. 4). One PLS-DA analysis was set on the class “harvesting year” (score scatter plot in Fig. 4a) and the other one was set on the class “viticultural treatment” (score scatter plot in Fig. 4c). The two models, built using five components, were both reasonably good: the first one with R2X (cum) = 0.789, R2Y(cum) = 0.742, Q2(cum) = 0.689; the second one with R2X (cum) = 0.802, R2Y(cum) = 0.723, Q2(cum) = 0.705. Interestingly, the model parameters (descriptiveness/predictive-ability) for the differentiation according to “viticultural treatment” were comparable to those characterizing the differentiation according to “harvesting year”. Therefore, this result fully justified the further investigations on farming methods effects. Differences in the samples number per class, in the two PLS-DA models, should be also taken into account. Nevertheless, the results clearly indicate that both occurring climatic conditions specific for each vintage and viticultural treatments influence the characteristics of the analyzed grapes juice samples, and the observed effects resulted at least comparable in importance. Permutation tests (n=100) were used to assess the validity and the degree of overfit of the two PLS-DA models (Additional file 1: Figure S7a).

Fig. 4
figure 4

PLS-DA analyses for all grapes juice samples. a PLS-DA score scatter plot from model set on class “harvesting year” (five components give R2X (cum) = 0.789, R2Y(cum) = 0.742, Q2(cum) = 0.689), b PLS-DA loading scatter plot from model set on class “harvesting year”, c PLS-DA score scatter plot from model set on class “viticultural treatment” (five components give R2X (cum) = 0.802, R2Y(cum) = 0.723, Q2(cum) = 0.705) and d PLS-DA loading scatter plot from model set on class “viticultural treatment”. Loading scatter plots for the models are coloured according to the absolute value of the correlation loading, p(corr). The colour bar associated with the plot indicates the correlation of the metabolites that discriminate the classes. w*c[1] and w*c[2] axes represented the weighted correlation vectors

Both PLS-DA analyses score scatter plots (Fig. 4a, c) showed a great separation among samples classes. In Fig. 4a, the t[1]/t[2] plot displayed the grapes samples belonging to the vintages 2020–2021 both at positive values of t[1], while the 2022 samples placed at negative values. This means that 2020–2021 grapes samples were more similar to each other with respect to those from 2022. The corresponding loading scatter plot (Fig. 4b) clearly indicates the metabolites responsible for vintages differentiation. In particular, table grapes samples from the 2020 to 2021 vintages appeared richer in α-/β-glucose, malate and citrate than those from 2022, while instead the 2022 samples contained more fructose and alanine.

In Fig. 4c, an important degree of separation among cultivation methods was observed in the t[1]/t[2] score scatter plot, with BD and INT table grapes samples occurring mainly at positive values of t[1] and well-differentiated from ORG samples placed, instead, at negative values. This result confirmed the PCA grouping indication (Fig. 3c). However, there are clearly other intrinsic aspects which determine the mutual differentiation between BD and INT-ORG samples along the t[2].

The PLS-DA loading scatter plot for the model set according to “viticultural treatment” classification shows the discriminating metabolites responsible of the observed separation within the farming methods clusters (Fig. 4d). In particular, a higher relative content of malate and ethanol characterized BD and INT grapes; while citrate, monosaccharides as α-/β-glucose and amino acids as arginine were more abundant in ORG samples. Noteworthy, from the comparative analysis of Fig. 4 loading scatter plots clearly results that the metabolites responsible for “viticultural treatment” discrimination do not correspond to those ascribable to “harvesting year” differentiation. These results further buttress the need to thoroughly study the specific effects of different agricultural managements.

To better investigate differences of metabolites, which discriminate BD, ORG and INT products, separate pairwise OPLS-DA analyses were performed, setting “viticultural treatment” as classification category (Fig. 5). The score scatter plots of the three obtained models (Fig. 5a, c, e) displayed a good separation degree in each pairwise comparison (BD vs ORG, BD vs INT and ORG vs INT), confirmed by reasonably sound model parameters (Table 4). Remarkably, the BD vs ORG and BD vs INT OPLS-DA pairwise models showed higher descriptiveness and predictive ability parameters with respect to the ORG vs INT model. The present result evidently testifies a better differentiation of BD from both ORG and INT samples, when compared with the ORG vs INT discrimination. Indeed, a greater discrimination of samples usually relates to a higher predictive ability of the corresponding models in the pairwise OPLS-DA evaluation [70]. Permutation tests (n=100) were used to check the validity and the degree of overfit of the three OPLS-DA models (Additional file 1: Figure S7b).

Fig. 5
figure 5

OPLS-DA analyses, pairwise comparisons. a pairwise OPLS-DA score scatter plot (one predictive and four orthogonal components, R2X(cum) = 0.83, R2Y(cum) = 0.912, Q2(cum) = 0.889) for biodynamic (BD) and organic (ORG) treatments, b S-line for the model; c pairwise OPLS-DA score scatter plot (one predictive and four orthogonal components, R2X(cum) = 0.84, R2Y(cum) = 0.874, Q2(cum) = 0.835) for biodynamic (BD) and integrated (INT) treatments, d S-line for the model; e pairwise OPLS-DA score scatter plot (one predictive and four orthogonal components, R2X(cum) = 0.794, R2Y(cum) = 0.824, Q2(cum) = 0.779) for organic (ORG) and integrated (INT) treatments, f S-line for the model. In the score scatter plots, the colour intensity of the dots increases progressively according to the vintage (light = 2020, medium = 2021, dark = 2022). The S-line plots for the OPLS-DA models visualize the centred loading vector (p(ctr)), colored according to the absolute value of the correlation loading, p(corr).The colour bar associated with the plot indicates the correlation of the metabolites that discriminate the classes

Table 4 Descriptiveness and predictive ability parameters of pairwise OPLS-DA analyses models

Interestingly, in all the comparisons (Fig.5), a very low differentiation according to the harvesting years was observed along the orthogonal component of the OPLS-DA score scatter plots. In addition, this result reinforces the already discussed comparable importance of vintages effects and viticultural treatments influences. For the pairwise OPLS-DA models, the corresponding S line plots (Fig. 5b, d, f) clearly indicate the metabolites responsible for BD vs ORG, BD vs INT and ORG vs INT differentiations. In the BD vs ORG comparison, the S-line plot for the OPLS-DA model (Fig. 5b) showed a higher relative content of ethanol, malate and β-glucose for BD table grapes samples with respect to ORG ones. On the contrary, ORG samples contained more arginine, citrate, tartaric acid and α-glucose than BD ones. In the pairwise comparison BD vs INT, the S-line plot for the model (Fig. 5d) indicated a greater abundance of ethanol, citrate, tartaric acid and carbohydrates such as α-/β-glucose in BD samples with respect to INT ones. On the other hand, INT samples were characterized by higher relative content of malate, fructose and amino acids such as arginine, glutamine and leucine. Finally, in the ORG vs INT comparison, the S-line plot for the model (Fig. 5f) pointed out the presence of more citrate, tartaric acid and α-/β-glucose for ORG samples than INT ones. These latter contained, instead, higher quantities of malate, fructose and glutamine.

A complete comparative view of occurring discriminating metabolites’ trends in samples, relatively to all the investigated agricultural managements, is summarized in Fig. 6. Potential significant discriminating metabolites were identified using VIP and p(corr) values (Additional file 1: Tables S1, S2, S3). Only assigned metabolites showing a major contribution to the OPLS-DA models, with high statistical reliability |p(corr)| ≥ 0.5 and strong discrimination power (VIP≥1), were selected. The relative quantification of these significant discriminant metabolites, for BD vs ORG, BD vs INT and ORG vs INT samples classes, was performed by considering the Fold Change (FC) ratio. FCs were calculated from the buckets’ integral values of the relevant NMR signals for leucine, ethanol, arginine, glutamine, malate, α- and β-glucose. Significant (p value < 0.05) log2(FC) values are reported in the graph, as shown in Fig. 6a, b, c.

Fig. 6
figure 6

Graphical representation of discriminant metabolites in each pairwise comparison for the investigated agricultural managements (only metabolites with |p(corr)|≥ 0.5 and VIP ≥ 1 were taken into account). Quantitative comparisons are, respectively, between a BD/ORG, b BD/INT and c ORG/INT samples classes. The x-axis reports significant (p value < 0.05) log2(FC) values. (FC = Fold Change, BD = biodynamic, ORG = organic, INT = integrated treatment)


The present study aimed at analyzing the effects of different viticultural treatments on metabolic profile of cv. Italia table grapes juice. We report, for the first time in this cultivar, a comparison among samples from biodynamic, organic and integrated agriculture, over a 3-year period, by using an 1H NMR-based metabolomics analysis combined with MVA. Besides the sampling method and the analytical protocol, also other experimental aspects were kept constant over the three considered vintages (2020, 2021 and 2022): geographical area of samples origin, supplier farms and agricultural managements of vineyards. About the management, within each type of cultivation (biodynamic, organic and integrated) the used practices in the vineyard were kept the same and replicated year by year. The preliminary unsupervised PCA analysis, performed on the whole NMR data set, showed that both inter-annual variability (Fig. 3a) and different farming methods (Fig. 3c) resulted as significant drivers for samples differentiation. The two supervised PLS-DA analyses (Fig. 4a and c), respectively, set on “harvesting year” and “viticultural treatment” as classification categories, revealed that BD, ORG and INT cultivations determined discrimination among samples at least to a same extent as the vintages. This evidence emerged from the comparison of models’ statistical parameters. The relative loading scatter plots (Fig. 4b, d) highlighted that discriminating metabolites for the observed classes separation were diverse between the two PLS-DA analyses. Thus, vintages and agricultural managements of vineyards seemed to differently affect the cv. Italia grape berries metabolic profiles. This aspect, combined with reasonably sound PLS-DA models parameters, buttressed the need to deepen the specific influences of farming methods. Thus, OPLS-DA pairwise comparisons (BD vs ORG, BD vs INT and ORG vs INT) were performed (Fig. 5a, c, e), along with the S-line for each model (Fig. 5b, d, f). As showed by the descriptiveness (R2) and predictive ability (Q2) values parameters, the highest discrimination was observed between BD vs ORG samples groups, followed by BD vs INT and, then, ORG vs INT comparisons (Table 4) in decreasing discrimination order. These results suggest that BD agriculture could produce clearly different foodstuffs from those obtained by ORG and INT treatments, which appear more similar to each other. Then, the relative quantification of discriminant metabolites (|p(corr)|≥ 0.5 and VIP ≥ 1) was carried out by considering the Fold Change (FC) ratio for each pairwise comparison (Fig. 6a, b, c).

In the case of BD vs ORG comparison (Fig. 6a), significant relative content changes between samples groups from the two considered agricultural practices resulted for β-glucose, ethanol, malate and arginine. In particular, β-glucose, ethanol and malate relative contents were significantly higher in BD samples than ORG ones with 0.05-, 1.30- and 1.14-fold increases, respectively. With the exception of the sugars, similar findings were reported by Picone et al. [32] in cv. Sangiovese winegrapes regarding trends in ethanol and malate contents. The concomitant relative increase of the latter would suggest the activation of glycolysis toward fermentative metabolic pathways mostly in BD products [32]. Furthermore, the resulting greater abundance of arginine (0.87-fold increase) in ORG samples compared to BD ones is an outcome dissimilar to that described by other authors [32], who also observed an increased concentration of some amino acids in BD grapes with respect to ORG ones.

In the BD vs INT comparison (Fig. 6b), similarly as observed in the previous comparison (BD vs ORG), BD samples showed a statistically significant higher relative content of β-glucose and ethanol (0.03- and 1-fold increases, respectively) with respect to INT samples. In the case of β-glucose similar results was obtained, although not statistically meaningful, in the roots of red beet Beta vulgaris from BD agriculture with respect to both INT and ORG [71]. On the other hand, the large relative content of ethanol in BD samples, with one-fold increase respect to INT ones, is noteworthy and may suggest that the abundance of ethanol could be a specific feature of studied BD grapes. By opposite, in samples from INT practices a higher relative content of amino acids arginine, leucine and glutamine resulted. The application of fertilizers (mineral and synthetic, as declared by supplier farmers) in INT vineyards could account for the observed greater relative concentration of amino acids in INT samples [72] rather than in BD ones, which were grown up without these kinds of fertilization treatments.

Finally, in the ORG vs INT comparison (Fig. 6c), the resulting most discriminant metabolites were α-glucose, malate and glutamine. In particular, ORG samples showed a limited—although statistically significant—higher relative content of α-glucose (0.02-fold increase) with respect to INT samples. In accordance with some literature data on the total sugars content from apples produced by ORG and INT farming methods [73], the α-glucose content was found to be slightly higher in ORG samples compared to INT ones. Other authors found, instead, no substantial differences between foodstuffs from ORG and INT cultivations, in terms of sugars quantity (red beet: [71], apples: [74]), and this could justify the only α-glucose increase. On the other hand, malate and glutamine showed higher statistically significant content (1.13- and 1.29-fold increases, respectively) in INT samples than ORG ones. In the case of malate, the obtained result appears consistent with previous findings related to the highest malate and citrate total content in apple fruit from INT vs ORG production treatment [75].

Therefore, a general overview of the obtained data suggests that, also in the here studied systems, sugars content was clearly affected by the agricultural practices in accordance to some literature results [32, 33]. However, although significant, the here reported relative differences in α-/β-glucose content were not particularly pronounced in all the treatments’ pairwise comparisons (Fig. 6). Moreover, apart from glucose, none of the other sugar metabolites appeared as significant discriminating factor in the present study. Hence, the total sum of soluble sugars in the grapes juice could remain approximately unchanged, among the applied different agricultural practices, according to other literature data. A specific review proposed a meta-regression analysis showing that the juice sugar concentration of grapes from ORG and BD managed vineyards was almost the same as grapes from CONV or INT viticulture [23]. Further insights on the issue are required, since other authors found similar outcomes, detecting no statistically significant differences on sugars concentration also in other food products cultivated with different farming methods (red beet: [71], apples: [74]).

Focusing on organic acids, the most discriminant metabolite in agricultural practices pairwise comparisons was malate (Fig. 6). In particular, grapes juice samples from both BD and INT agriculture showed higher content of malate, with similar values of FC increase (respectively, 1.14 and 1.13, Fig. 6a and c), with respect to samples from ORG treatment. This result is in agreement with literature data related to grapes [32] and other food matrices as reed beet [71] and apple fruits [75]. Since malate is involved in diverse aspects of cellular metabolism in grapes, its quantity can be influenced by many factors. In general, the accumulation of malate in grapevines is known to be due to de novo synthesis during the pre-veraison. This occurs through metabolism of assimilates translocated from leaf tissues to berries, as well as photosynthetic activity within the fruit itself [76]. Furthermore, loss of malate is associated with post-veraison period, when it is involved in various catabolic pathways. The switch from net accumulation to degradation of malate occurs just before veraison. Since its content in fruits is extremely variable [76], many reasons could account for the abundance of malate in the studied BD and INT grapes juice samples, with respect to ORG ones, buttressing the need for further insights.

In the case of ethanol, significantly discriminant for BD grapes in both BD vs ORG and BD vs INT comparisons (Fig. 6a, b), it is known to be naturally present in grape berries as metabolism product and it was found to be one of the most farming methods-dependent metabolites [33]. However, as already proposed by Gallo et al. [33], it can be hypothesized that the occurring ethanol could be the sum of the ethanol naturally contained before harvest and that further formed by enzymatic degradation during storage and sample manipulation. The latter portion could be either negligible with respect to the starting amount or variable due to other specific factors (e.g., yeast population) depending on cultivation method [33].

Finally, pairwise comparisons (Fig. 6) highlighted remarkably that INT samples would appear the richest of some amino acid compounds, compared to both BD and ORG samples (Fig. 6b, c). It can be assumed that metabolism in INT vine would invest appreciably in anabolic pathways, as the amino acids’ biosynthesis, boosted by the combined use of organic and mineral/synthetic fertilizers. These latter provide plants with important nutrients as nitrogen (N), which seem to increase the concentration of some amino acids in many crops [72]. The higher abundance of glutamine in INT samples could be an index of the general better nutritional status of INT vines, since glutamine is known to have an important anabolic role as N atoms source for biosynthesis of other amino acids and also biological molecules containing N [77, 78]. In our case, as declared in the questionnaires by the supplier farms, the dissimilar features in the fertilization method between INT and BD viticulture along other treatments (Table 1) may explain different plant’s nutritional conditions and, consequently, the statistically significant differences in the relative quantity of some amino acids shown in the BD vs INT comparison (Fig. 6b). In the case of arginine, its content appears to be strictly dependent on farming methods, in accordance to previous work [33]. Arginine is important for N storage in plants and many mechanisms are involved in the regulation of its levels in plant tissues. Arginine synthesis and catabolism, both are linked to the overall nutritional status of the plant cell and its consumption seems to be coordinated with the sugar starvation status of plants. Thus, arginine catabolism allows mobilization of stored nitrogen and relates to other metabolic pathways, including fine-tuning of polyamines production (essential for development and stress responses of plants) [79]. Although the regulation of the different catabolic pathways concerning arginine requires further research, several regulatory mechanisms of its biosynthesis were identified [79]. Arginine accumulates in grape berries from shortly before veraison until the fruit ripens, when its biosynthesis stops [80]. Therefore, the observed lower arginine content in ripe BD table grapes juice compared to both ORG and INT samples (Fig. 6a, b), besides testifying a major intake of nutrients characterizing ORG and particularly INT agriculture, may also suggest a specific metabolic response by the BD plants. Indeed, BD agriculture could stimulate the natural resistance of the grapevines activating the synthesis of polyamines, which are important in plant defense processes and determine the arginine use, so explaining our findings.


The present metabolomics study demonstrated that cv. Italia table grapes juice composition is significantly affected by the use of different kinds of vineyard managements. The comparison here carried out was among three farming methods: biodynamic (BD), organic (ORG) and integrated (INT). The metabolic profiling of cv. Italia table grapes—from the same geographical area and three different vintages 2020, 2021 and 2022—was performed with the 1H Nuclear Magnetic Resonance spectroscopy as analytical platform and multivariate statistical analysis of the NMR spectra-derived data set. The preliminary unsupervised PCA analysis of the data revealed both inter-annual variability (vintages 2020, 2021, 2022) and different farming methods (BD, ORG, INT) as significant drivers for samples differentiation. In the following supervised PLS-DA data analyses, the inter-vineyards variability due to the diverse managements turned out to discriminate samples at least to a same extent as the harvesting year. Moreover, the observed discriminating metabolites for samples separation were different considering the viticultural practices rather than the harvesting year, as assigned class in the PLS-DA analyses. Noteworthy, OPLS-DA pairwise comparisons between field treatments indicated the biodynamic vs organic samples as the best differentiation, followed by the biodynamic vs integrated and organic vs integrated samples comparisons, in decreasing discrimination order. These results somehow quantified the relations of the overall changes for metabolites composition with different agronomic management. Although variation of product characteristics, according to the different agricultural practices, is an expected outcome, an estimation of its specific effect in a multi-year experiment, for the three different vineyard conductions here focused, is novel. Ethanol, sugars as α-/β-glucose, organic acids as malate and some amino acids as arginine, leucine, glutamine, resulted the chemical species mostly influenced by the viticultural treatments. In particular, besides ethanol, arginine content appears to be strictly dependent on farming methods, in accordance to previous work. The here observed lower arginine content in ripe BD table grapes juice with respect to both ORG and INT samples, besides testifying a major intake of nutrients characterizing ORG and particularly INT agriculture, also suggests a BD specific metabolic response possibly related to plant defense processes.

Availability of data and materials

The data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request.



Nuclear Magnetic Resonance


Multivariate Statistical Analysis










Automatic tuning-matching


Free induction decay


Principal Component Analysis


Partial Least Squares Discriminant Analysis


Orthogonal Partial Least Squares Discriminant Analysis


Variable importance for the projection


Fold Change


  1. Antonacci D, Scienza A. L’uva da tavola: botanica, storia e arte, alimentazione, paesaggio, coltivazione, ricerca, utilizzazione, mondo e mercato. Collana Coltura & Cultura. Ed. Script, Bologna. Milano: Bayer CropScience S.r.l.; 2010. 624 p.

  2. ISTAT. Statistiche Istat. 2023. Accessed 4 Dec 2023.

  3. Ali K, Maltese F, Choi YH, Verpoorte R. Metabolic constituents of grapevine and grape-derived products. Phytochem Rev. 2010;9(3):357–78.

    Article  CAS  PubMed  Google Scholar 

  4. Teixeira A, Eiras-Dias J, Castellarin SD, Gerós H. Berry phenolics of grapevine under challenging environments. Int J Mol Sci. 2013;14(9):18711–39.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Pereira GE, Gaudillere JP, Pieri P, Hilbert G, Maucourt M, Deborde C, et al. Microclimate influence on mineral and metabolic profiles of grape berries. J Agric Food Chem. 2006;54(18):6765–75.

    Article  CAS  PubMed  Google Scholar 

  6. Mulas G, Galaffu MG, Pretti L, Nieddu G, Mercenaro L, Tonelli R, et al. NMR analysis of seven selections of vermentino grape berry: metabolites composition and development. J Agric Food Chem. 2011;59(3):793–802.

    Article  CAS  PubMed  Google Scholar 

  7. van Leeuwen C, Friant P, Choné X, Tregoat O, Koundouras S, Dubourdieu D. Influence of climate, soil, and cultivar on terroir. Am J Enol Vitic. 2004;55(3):207–17.

    Article  Google Scholar 

  8. Pereira GE, Gaudillere JP, Van Leeuwen C, Hilbert G, Lavialle O, Maucourt M, et al. 1H NMR and chemometrics to characterize mature grape berries in four wine-growing areas in Bordeaux. France J Agric Food Chem. 2005;53(16):6382–9.

    Article  CAS  PubMed  Google Scholar 

  9. Pereira GE, Gaudillere JP, van Leeuwen C, Hilbert G, Maucourt M, Deborde C, et al. 1H NMR metabolite fingerprints of grape berry: comparison of vintage and soil effects in Bordeaux grapevine growing areas. Anal Chim Acta. 2006;563(1):346–52.

    Article  CAS  Google Scholar 

  10. Mazzei P, Francesca N, Moschetti G, Piccolo A. NMR spectroscopy evaluation of direct relationship between soils and molecular composition of red wines from Aglianico grapes. Anal Chim Acta. 2010;673(2):167–72.

    Article  CAS  PubMed  Google Scholar 

  11. Food and Agriculture Organization of the United Nations, FAO. “Energy-Smart” Food for People and Climate—issue Paper (2011). Accessed 4 Dec 2023.

  12. Food and Agriculture Organization of the United Nations, FAO. Building a common vision for sustainable food and agriculture. Principles and approaches. (2014) Accessed 4 Dec 2023.

  13. Campbell BM, Vermeulen SJ, Aggarwal PK, Corner-Dolloff C, Girvetz E, Loboguerrero AM, et al. Reducing risks to food security from climate change. Glob Food Sec. 2016;1(11):34–43.

    Article  Google Scholar 

  14. Wright J. A revitalisation of European farming and the promise of the biodynamic worldview. Chem Biol Technol Agric. 2022;9(1):64.

    Article  Google Scholar 

  15. Ponzio C, Gangatharan R, Neri D. Organic and biodynamic agriculture: a review in relation to sustainability. Int J Plant Soil Sci. 2013;2(1):95–110.

    Article  Google Scholar 

  16. Zucca G, Smith DE, Mitry DJ. Sustainable viticulture and winery practices in California: What is it, and do customers care? Int J Wine Res. 2009;1:189–94.

    Google Scholar 

  17. Mann S, Ferjani A, Reissig L. What matters to consumers of organic wine? Br Food J. 2012;114(2):272–84.

    Article  Google Scholar 

  18. Villanueva-Rey P, Vázquez-Rowe I, Moreira MT, Feijoo G. Comparative life cycle assessment in the wine sector: biodynamic vs. conventional viticulture activities in NW Spain. J Cleaner Prod. 2014;65:330–41.

    Article  Google Scholar 

  19. Integrated Pest Management (IPM) Accessed 4 Dec 2023.

  20. Meissner G, Athmann ME, Fritz J, Kauer R, Stoll M, Schultz HR. Conversion to organic and biodynamic viticultural practices: impact on soil, grapevine development and grape quality. OENO One. 2019.

    Article  Google Scholar 

  21. Döring J, Frisch M, Tittmann S, Stoll M, Kauer R. Growth, yield and fruit quality of grapevines under organic and biodynamic management. PLoS ONE. 2015;10(10): e0138445.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Fritz J, Athmann M, Meissner G, Kauer R, Köpke U. Quality characterisation via image forming methods differentiates grape juice produced from integrated, organic or biodynamic vineyards in the first year after conversion. Biol Agric Hortic. 2017;33(3):195–213.

    Article  Google Scholar 

  23. Döring J, Collins C, Frisch M, Kauer R. Organic and biodynamic viticulture affect biodiversity and properties of vine and wine: a systematic quantitative review. Am J Enol Vitic. 2019;70(3):221.

    Article  Google Scholar 

  24. Reeve JR, Carpenter-Boggs L, Reganold JP, York AL, McGourty G, McCloskey LP. Soil and winegrape quality in biodynamically and organically managed vineyards. Am J Enol Vitic. 2005;56(4):367–76.

    Article  CAS  Google Scholar 

  25. Botelho RV, Roberti R, Tessarin P, Garcia-Mina JM, Rombolà AD. Physiological responses of grapevines to biodynamic management. Renew Agric Food Syst. 2016;31(5):402–13.

    Article  Google Scholar 

  26. Ferrara G, Mazzeo A, Netti G, Pacucci C, Matarrese AMS, Cafagna I, et al. Girdling, gibberellic acid, and forchlorfenuron: effects on yield, quality, and metabolic profile of table grape cv. Italia. Am J Enol Viticult. 2014;65(3):381–7.

    Article  CAS  Google Scholar 

  27. Malagoli M, Sut S, Kumar G, Dall’Acqua S. Variations of elements, pigments, amino acids and secondary metabolites in Vitis vinifera (L.) cv Garganega after 501 biodynamic treatment. Chem Biol Technol Agric. 2022;9(1):36.

    Article  CAS  Google Scholar 

  28. Tassoni A, Tango N, Ferri M. Comparison of biogenic amine and polyphenol profiles of grape berries and wines obtained following conventional, organic and biodynamic agricultural and oenological practices. Food Chem. 2013;139(1):405–13.

    Article  CAS  PubMed  Google Scholar 

  29. Tassoni A, Tango N, Ferri M. Polyphenol and biogenic amine profiles of albana and lambrusco grape berries and wines obtained following different agricultural and oenological practices. Food Nutr Sci. 2014;5:9.

    Google Scholar 

  30. De Pascali SA, Coletta A, Del Coco L, Basile T, Gambacorta G, Fanizzi FP. Viticultural practice and winemaking effects on metabolic profile of Negroamaro. Food Chem. 2014;15(161):112–9.

    Article  Google Scholar 

  31. Laghi L, Versari A, Marcolini E, Parpinello GP. Metabonomic investigation by 1H-NMR to discriminate between red wines from organic and biodynamic grapes. Food Nutr Sci. 2014;5:8.

    Google Scholar 

  32. Picone G, Trimigno A, Tessarin P, Donnini S, Rombolà AD, Capozzi F. 1H NMR foodomics reveals that the biodynamic and the organic cultivation managements produce different grape berries (Vitis vinifera L. cv. Sangiovese). Food Chem. 2016;213:187–95.

    Article  CAS  PubMed  Google Scholar 

  33. Gallo V, Mastrorilli P, Cafagna I, Nitti GI, Latronico M, Longobardi F, et al. Effects of agronomical practices on chemical composition of table grapes evaluated by NMR spectroscopy. J Food Compos Anal. 2014;35(1):44–52.

    Article  CAS  Google Scholar 

  34. Qu Q, Jin L. Application of nuclear magnetic resonance in food analysis. Food Sci Technol. 2022.

    Article  Google Scholar 

  35. Trimigno A, Marincola FC, Dellarosa N, Picone G, Laghi L. Definition of food quality by NMR-based foodomics. Curr Opin Food Sci. 2015;1(4):99–104.

    Article  Google Scholar 

  36. Calò F, Girelli CR, Angilè F, Del Coco L, Mazzi L, Barbini D, et al. 1H-NMR profiling shows as specific constituents strongly affect the international EVOO blends characteristics: the case of the Italian oil. Molecules. 2021;26(8):2233.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Son HS, Hwang GS, Ahn HJ, Park WM, Lee CH, Hong YS. Characterization of wines from grape varieties through multivariate statistical analysis of 1H NMR spectroscopic data. Food Res Int. 2009;42(10):1483–91.

    Article  CAS  Google Scholar 

  38. Amargianitaki M, Spyros A. NMR-based metabolomics in wine quality control and authentication. Chem Biol Technol Agri. 2017;4(1):9.

    Article  Google Scholar 

  39. Consonni R, Bernareggi F, Cagliani LR. NMR-based metabolomic approach to differentiate organic and conventional Italian honey. Food Control. 2019;1(98):133–40.

    Article  Google Scholar 

  40. Girelli CR, Schiavone R, Vilella S, Fanizzi FP. Salento Honey (Apulia, South-East Italy): a preliminary characterization by 1H-NMR metabolomic fingerprinting. Sustainability. 2020;12(12):5009.

    Article  CAS  Google Scholar 

  41. Lau H, Laserna AKC, Li SFY. 1H NMR-based metabolomics for the discrimination of celery (Apium graveolens L. var. dulce) from different geographical origins. Food Chem. 2020;332:127424.

    Article  CAS  PubMed  Google Scholar 

  42. Hussain M, Girelli CR, Verweire D, Oehl MC, Avendaño MS, Scortichini M, et al. 1H-NMR metabolomics study after foliar and endo-therapy treatments of Xylella fastidiosa subsp. pauca infected olive trees: medium time monitoring of field experiments. Plants. 2023;12(10):1946.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Belton PS, Colquhoun IJ, Kemsley EK, Delgadillo I, Roma P, Dennis MJ, et al. Application of chemometrics to the 1H NMR spectra of apple juices: discrimination between apple varieties. Food Chem. 1998;61(1):207–13.

    Article  CAS  Google Scholar 

  44. Salvino RA, Colella MF, De Luca G. NMR-based metabolomics analysis of Calabrian citrus fruit juices and its application to industrial process quality control. Food Control. 2021;1(121): 107619.

    Article  Google Scholar 

  45. REGOLAMENTO (UE) 2018/848 DEL PARLAMENTO EUROPEO E DEL CONSIGLIO del 30 maggio 2018 relativo alla produzione biologica e all’etichettatura dei prodotti biologici e che abroga il regolamento (CE) n. 834/2007 del Consiglio.

  46. DEMETER - Biodynamic Federation Demeter International. International Demeter Biodynamic Standard. Accessed 4 Dec 2023.

  47. L’Abate G, Costantini E, Roberto B, Fantappiè M, Lorenzetti R, S. M. Carta dei Suoli d’Italia 1:1.000.000 (Soil map of Italy, scale 1:1.000.000). S.EL.CA. Firenze, Italia; 2012 [cited 2024 Feb 10].

  48. Carrante V, Della Gatta L, Perniola M, Lopez G. I terreni agrari della provincia di Taranto. In: Annali di Sperimentazione Agraria [Internet]. Roma: Fondazione per la sperimentazione agraria, Ministero agricoltura e foreste; 1957. p. 79–145., Map available at

  49. Caliandro A, Lamaddalena N, Stelluti M, Steduto P. Progetto ACLA 2. Caratterizzazione agroecologica della Regione Puglia in funzione della potenzialità produttiva. Opuscolo divulgativo. Bari: UE-Regione Puglia; 2005. 179 p.

  50. Apulian pedological map published on the website of Apulia Region. Accessed 12 Feb 2024.

  51. Calabrese A, Mandrelli L, Loi E, Blonda M. Chemical and microbiological characterization of soil under different agronomical use and practical: first focus on nitrogen cycles. IOSR J Biotechnol Biochem. 2020;6:45–57.

    Google Scholar 

  52. Regione Puglia - Disciplinare Di Produzione Integrata - Anno 2023. Accessed 5 Dec 2023.

  53. Santoni M, Ferretti L, Migliorini P, Vazzana C, Pacini GC. A review of scientific research on biodynamic agriculture. Org Agric. 2022;12(3):373–96.

    Article  Google Scholar 

  54. Girelli CR, Accogli R, Del Coco L, Angilè F, De Bellis L, Fanizzi FP. 1H-NMR-based metabolomic profiles of different sweet melon (Cucumis melo L.) Salento varieties: Analysis and comparison. Food Res Int. 2018;114:81–9.

    Article  CAS  PubMed  Google Scholar 

  55. Girelli CR, De Pascali SA, Del Coco L, Fanizzi FP. Metabolic profile comparison of fruit juice from certified sweet cherry trees (Prunus avium L.) of Ferrovia and Giorgia cultivars: a preliminary study. Food Res Int. 2016;90:281–7.

    Article  CAS  PubMed  Google Scholar 

  56. Ali K, Maltese F, Fortes AM, Pais MS, Choi YH, Verpoorte R. Monitoring biochemical changes during grape berry development in Portuguese cultivars by NMR spectroscopy. Food Chem. 2011;124(4):1760–9.

    Article  CAS  Google Scholar 

  57. Ali K, Maltese F, Zyprian E, Rex M, Choi YH, Verpoorte R. NMR metabolic fingerprinting based identification of grapevine metabolites associated with downy mildew resistance. J Agric Food Chem. 2009;57(20):9599–606.

    Article  CAS  PubMed  Google Scholar 

  58. Savage AK, van Duynhoven JPM, Tucker G, Daykin CA. Enhanced NMR-based profiling of polyphenols in commercially available grape juices using solid-phase extraction. Magn Reson Chem. 2011;49(S1):S27-36.

    Article  CAS  PubMed  Google Scholar 

  59. Consonni R, Cagliani LR, Cogliati C. NMR based geographical characterization of roasted coffee. Talanta. 2012;15(88):420–6.

    Article  Google Scholar 

  60. van den Berg RA, Hoefsloot HC, Westerhuis JA, Smilde AK, van der Werf MJ. Centering, scaling, and transformations: improving the biological information content of metabolomics data. BMC Genomics. 2006;7(1):142.

    Article  PubMed  PubMed Central  Google Scholar 

  61. Jackson JE. A user’s guide to principal components. Vol. 587. John Wiley & Sons; 2005.

  62. Wold S, Eriksson L, Trygg J, Kettaneh N. The PLS method–partial least squares projections to latent structures–and its applications in industrial RDP (research, development, and production). Umeå University: Umeå, Sweden. 2004; 1–44

  63. Trygg J, Wold S. Orthogonal projections to latent structures (O-PLS). J Chemom. 2002;16(3):119–28.

    Article  CAS  Google Scholar 

  64. Eriksson L, Byrne T, Johansson E, Trygg J, Vikström C. Multi-and megavariate data analysis basic principles and applications. Vol. 1. Umetrics Academy; 2013.

  65. Mahadevan S, Shah SL, Marrie TJ, Slupsky CM. Analysis of metabolomic data using support vector machines. Anal Chem. 2008;80(19):7562–70.

    Article  CAS  PubMed  Google Scholar 

  66. Wheelock ÅM, Wheelock CE. Trials and tribulations of ‘omics data analysis: assessing quality of SIMCA-based multivariate models using examples from pulmonary medicine. Mol BioSyst. 2013;9(11):2589–96.

    Article  CAS  PubMed  Google Scholar 

  67. Le Mao I, Da Costa G, Bautista C, de Revel G, Richard T. Application of 1H NMR metabolomics to French sparkling wines. Food Control. 2023;1(145): 109423.

    Article  Google Scholar 

  68. Cocchi M, Biancolillo A, Marini F. Chapter ten—chemometric methods for classification and feature selection. In: Jaumot J, Bedia C, Tauler R, editors. Comprehensive analytical chemistry. Elsevier; 2018. p. 265–99.

    Google Scholar 

  69. Vinholes J, Silva BM, Silva LR. Hydroxycinnamic acids (HCAS): structure, biological properties and health effects. Adv Med Biol. 2015;88(8):1–33.

    Google Scholar 

  70. Girelli CR, Del Coco L, Papadia P, De Pascali SA, Fanizzi FP. Harvest year effects on Apulian EVOOs evaluated by 1H NMR based metabolomics. Peer J. 2016;4:e2740.

    Article  PubMed  PubMed Central  Google Scholar 

  71. Bavec M, Turinek M, Grobelnik-Mlakar S, Slatnar A, Bavec F. Influence of industrial and alternative farming systems on contents of sugars, organic acids, total phenolic content, and the antioxidant activity of red beet (Beta vulgaris L. ssp. vulgaris Rote Kugel). J Agric Food Chem. 2010;58(22):11825–31.

    Article  CAS  PubMed  Google Scholar 

  72. Wang ZH, Li SX, Malhi S. Effects of fertilization and other agronomic measures on nutritional quality of crops. J Sci Food Agric. 2008;88(1):7–23.

    Article  CAS  Google Scholar 

  73. Reganold JP, Glover JD, Andrews PK, Hinman HR. Sustainability of three apple production systems. Nature. 2001;410(6831):926–30.

    Article  ADS  CAS  PubMed  Google Scholar 

  74. Róth E, Berna A, Beullens K, Yarramraju S, Lammertyn J, Schenk A, et al. Postharvest quality of integrated and organically produced apple fruit. Postharvest Biol Technol. 2007;45(1):11–9.

    Article  Google Scholar 

  75. Jakopic J, Simoncic A, Slatnar A, Stampar F, Veberic R. Analysis of selected primary metabolites and phenolic profile of ‘golden delicious’ apples from four production systems. Fruits. 2012;67(5):377–86.

    Article  CAS  Google Scholar 

  76. Sweetman C, Deluc LG, Cramer GR, Ford CM, Soole KL. Regulation of malate metabolism in grape berry and other developing fruits. Phytochemistry. 2009;70(11):1329–44.

    Article  CAS  PubMed  Google Scholar 

  77. Yemm EW. Glutamine in the metabolism of barley plants. New Phytol. 1949;48(3):315–31.

    Article  CAS  Google Scholar 

  78. Roubelakis-Angelakis KA, Kliewer WM. Nitrogen Metabolism in Grapevine. In: Horticultural Reviews. 1992. p. 407–52.

  79. Winter G, Todd CD, Trovato M, Forlani G, Funck D. Physiological implications of arginine metabolism in plants. Front Plant Sci. 2015;6:534.

    Article  PubMed  PubMed Central  Google Scholar 

  80. Stines AP, Grubb J, Gockowiak H, Henschke PA, Pb HJ, van Heeswijck R. Proline and arginine accumulation in developing berries of Vitis vinifera L. in Australian vineyards: influence of vine cultivar, berry maturity and tissue type. Aust J Grape Wine Res. 2000;6(2):150–8.

    Article  CAS  Google Scholar 

Download references


This work was performed in collaboration with Azienda Agricola Lacalamita Rosa (for organic grapes certified Reg. CE 848/18, DEMETER production rules), Azienda D'Alessandro (for organic grapes certified Reg. CE 848/18 production rules) and Azienda Martemucci (for IPM grapes certified GDO production rules).



Author information

Authors and Affiliations



FPF and GC designed the research and choose the methodology; GC collected samples in the field experiment; CSC and MH performed the experiments; CSC and CRG analyzed the data; CSC, CRG and FPF interpreted the results; CSC wrote the manuscript and created the figures; CRG and FPF reviewed the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Francesco Paolo Fanizzi.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Figure S1.

local pedological map; Figure S2. questionnaires supplied to farms; Figure S3. [1H,1H]-cosy spectrum; Figure S4. [1H,13C]-HSQC spectrum; Figure S5. [1H,13C]-HMBC spectrum; Figure S6. PCA loading scatter plot; Figure S7. examples of permutation tests; Table S1. list of discriminating chemical descriptors with corresponding p(corr) and VIP, for BD vs ORG samples comparison in the related OPLS-DA model; Table S2. list of discriminating chemical descriptors with corresponding p(corr) and VIP, for BD vs INT samples comparison in the related OPLS-DA model; Table S3. list of discriminating chemical descriptors with corresponding p(corr) and VIP, for ORG vs INT samples comparison in the related OPLS-DA model.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Colì, C.S., Girelli, C.R., Cesari, G. et al. Biodynamic, organic and integrated agriculture effects on cv. Italia table grapes juice, over a 3-year period experiment: an 1H NMR spectroscopy-based metabolomics study. Chem. Biol. Technol. Agric. 11, 35 (2024).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: