Results

Agronomic and Meteorological Context of the Sampled Material

Fruits and leaves from seven Prunus avium cultivars were collected across nine orchards during the 2014 and 2015 growing seasons, at the stage of commercial fruit maturity as determined in collaboration with orchard managers (Table S1). To provide an environmental context for the sampled material, meteorological data from three weather stations representative of the production area were compiled for the 2014 and 2015 growing seasons (Table S2). These data showed year‐associated meteorological differences, particularly when annual parameters were considered. Overall, 2014 was characterized by higher total rainfall and a higher number of rainy days, whereas 2015 showed lower rainfall, fewer rainy days, and higher solar radiation. However, climatic differences during key phenological periods, including flowering and harvest, were less pronounced than those observed on an annual basis. Thus, the metabolomic dataset was obtained under contrasting annual meteorological conditions. This environmental information was used to contextualize the interpretation of metabolomic patterns across cultivars, years, and orchard locations. Agronomic assessment confirmed that fruits were harvested at comparable physiological maturity, as reflected by similar values of fruit weight, firmness, total soluble solids, and pH across cultivars. As the present study aims to investigate leaf and fruit metabolomes, agronomic traits are not further discussed.

Untargeted Metabolomics Revealed the Specialized Metabolites Occurring in Fruits and Leaves

Untargeted metabolomic profiling revealed a clear and expected organ‐specific differentiation between fruits and leaves of sweet cherry. HPLC–DAD chromatograms showed that fruit extracts were dominated by mildly to highly polar metabolites (Figure 2a), whereas leaf extracts displayed a prevalence of less polar compounds (Figure 2b).

FIGURE 2: HPLC–DAD chromatographic profiles of representative sweet cherry fruit (a) and leaf (b) extracts. Chromatograms were recorded in the 280–600 nm range. The color bar indicates absorbance intensity (mAU, arbitrary units), reflecting metabolite abundance in the extracts. Fruit extracts are dominated by mildly to highly polar metabolites, whereas leaf extracts show a higher contribution of less polar compounds (highlighted area). Peak numbers correspond to metabolites identified by UPLC–HRMS and reported in Table 1.

The major metabolites detected by HPLC–DAD were identified by UPLC–HRMS and mainly belonged to hydroxycinnamic acids, flavonols, flavan‐3‐ols, and anthocyanins (Table 1). While most compounds were detected in both organs, several metabolites showed a marked organ preference, with anthocyanins predominantly detected in fruits and flavonol derivatives, particularly kaempferol‐O‐acetylhexoside and kaempferol‐O‐malonylhexoside, enriched in leaves. In addition, genotype‐dependent accumulation patterns were observed.

 | Leaf
Sandra | Brulat | Romana | Durone Rosso | Ferrovia | Early Bigi | Grace Star
1 | 188.84 ± 30.80a | 95.17 ± 10.15c,d | 102.38 ± 17.84b,c | 68.32 ± 5.78d,e | 68.63 ± 9.87d,e | 49.15 ± 16.45e | 123.22 ± 10.57b
2 | 64.42 ± 11.42a | 34.95 ± 6.79b | 82.04 ± 18.58a | 27.55 ± 8.63b,c | 43.71 ± 17.68b | 12.58 ± 11.77c | 7.81 ± 1.42c
3 | 252.38 ± 62.62a | 341.56 ± 64.77a | 342.56 ± 137.05a | 202.75 ± 34.39a | 247 0.40 ± 33.08a | 318.60 ± 142.10a | 272.46 ± 110.35a
4 | − ± − | − ± − | − ± − | − ± − | − ± − | − ± − | − ± −
5 | − ± − | − ± − | − ± − | − ± − | − ± − | − ± − | − ± −
6 | 59.61 ± 7.85a | 66.78 ± 7.00a | 41.75 ± 4.80b | 34.56 ± 3.56b | 32.60 ± 3.20b | 57.47 ± 8.46a | 56.11 ± 10.10a
7 | 115.39 ± 16.25a | 106.00 ± 12.35ab | 95.63 ± 12.15b | 55.99 ± 5.04c | 53.54 ± 6.40c | 95.03 ± 13.40b | 92.10 ± 11.17B
8 | 28.33 ± 6.20a | 37.11 ± 8.07a | 37.91 ± 14.12a | 22.97 ± 4.48a | 28.39 ± 4.59a | 36.86 ± 15.45a | 30.43 ± 11.60a
9 | 113.67 ± 18.46b | 153.70 ± 14.98a | 69.93 ± 16.46c | 76.09 ± 10.20c | 78.65 ± 8.53c | 118.95 ± 30.62b | 98.03 ± 26.71b,c
10 | 61.87 ± 10.12a,b | 72.39 ± 10.29a | 32.22 ± 8.25c | 43.98 ± 11.01b,c | 59.93 ± 17.26a,b | 54.27 ± 17.39a,b | 44.79 ± 13.06b,c
11 | 160.98 ± 16.50a | 159.85 ± 18.25a | 145.58 ± 27.61a,b,c | 111.12 ± 11.76c | 114.75 ± 6.43bc | 149.13 ± 35.56ab | 127.59 ± 22.85abc
12 | 2.39 ± 0.82c | 1.02 ± 2.87c | 11.31 ± 3.40a,b | 11.49 ± 3.17ab | 15.23 ± 3.53a | 9.88 ± 3.38b | 8.61 ± 1.45b
13 | − ± − | − ± − | − ± − | − ± − | − ± − | − ± − | − ± −
14 | − ± − | ‐ ± − | − ± − | − ± − | − ± − | − ± − | − ± −
 | Fruit
Sandra | Brulat | Romana | Durone Rosso | Ferrovia | Early Bigi | Grace Star
1 | 37.47 ± 5.64a | 15.78 ± 3.83c | 22.80 ± 3.96b | 13.81 ± 1.99c | 17.83 ± 4.38b,c | 6.58 ± 1.67d | 34.41 ± 5.59a
2 | 11.15 ± 2.07b | 18.76 ± 5.27a | 19.30 ± 4.48a | 9.94 ± 1.77b,c | 9.85 ± 2.58b,c | 5.73 ± 2.51c | 5.35 ± 1.68c
3 | 3.03 ± 0.46b,c | 3.71 ± 0.80a,b | 4.30 ± 0.81a | 3.20 ± 0.52a,b,c | 3.45 ± 0.57a,b,c | 2.66 ± 0.73c | 3.67 ± 0.67a,b
4 | 5.47 ± 1.07a | 5.72 ± 1.29a | 3.21 ± 0.95b | 2.42 ± 0.74b | 2.48 ± 0.51b | 3.79 ± 0.78b | 2.47 ± 0.21b
5 | 4.16 ± 1.72c | 7.40 ± 2.08a,b | 8.02 ± 1.61a | 4.69 ± 0.80c | 5.86 ± 0.40a,b,c | 5.27 ± 1.78b,c | 6.21 ± 1.60a,b,c
6 | *16.41 ± 1.04a | *15.66 ± 0.99ab | *13.57 ± 0.68b | *16.56 ± 1.77a | *16.43 ± 2.54a | *15.70 ± 1.73ab | *13.61 ± 0.53b
7 | *6.91 ± 0.73a | *6.38 ± 0.30a | *6.26 ± 0.54a | *7.84 ± 1.23a | *7.54 ± 1.30a | *5.58 ± 0.70b | *5.98 ± 0.73a
8 | *3.37 ± 0.42a | *3.80 ± 0.85a | *3.94 ± 0.80a | *3.31 ± 0.66a | *3.36 ± 0.48a | *3.11 ± 0.77a | *3.86 ± 0.92a
9 | 3.81 ± 0.92a | 3.54 ± 0.77a | 3.29 ± 0.57a | 3.79 ± 0.68a | 3.45 ± 0.76a | 3.82 ± 0.95a | 2.77 ± 0.56a
10 | *20.37 ± 8.15a | *19.65 ± 12.92a | *11.01 ± 0.82a | *12.32 ± 2.30a | *10.91 ± 0.98a | *16.36 ± 4.25a | *9.62 ± 1.85a
11 | *9.65 ± 1.15b | *8.87 ± 0.84b | *15.03 ± 3.41a | *11.85 ± 2.93b | *12.21 ± 2.53b | *8.02 ± 1.92b | *9.86 ± 2.54b
12 | − ± − | − ± − | − ± − | − ± − | − ± − | − ± − | − ± −
13 | 60.69 ± 23.88a,b | 65.10 ± 31.26a,b | 31.67 ± 7.52b | 64.03 ± 17.42a,b | 48.13 ± 18.58ab | 71.39 ± 23.42a | 34.39 ± 11.40b
14 | 1.14 ± 0.43a,b | 1.30 ± 0.64a,b | 1.32 ± 0.65a,b | 1.54 ± 0.74ab | 1.02 ± 0.59b | 2.03 ± 0.65a | 1.54 ± 0.62a,b

In fruits, genotype‐dependent differences mainly involved the accumulation of a small number of phenolic compounds (Table 1). Early Bigi showed the highest levels of cyanidin derivatives, while Romana and Grace Star displayed consistently lower amounts. Conversely, neochlorogenic acid accumulated preferentially in Sandra and Grace Star and was less represented in Early Bigi, whereas coumaroyl quinic acid was enriched in Burlat and Romana and reduced in Early Bigi and Grace Star.

In leaves, genotype‐dependent variation mainly affected the accumulation of hydroxycinnamic acids and flavonols (Table 1). Neochlorogenic acid showed marked cultivar specificity, with the highest levels detected in Sandra and the lowest in Early Bigi. Similarly, kaempferol‐3‐O‐rutinoside was most abundant in Sandra and Burlat and comparatively reduced in Durone Rosso. In contrast, chlorogenic acid did not display significant differences among cultivars, despite being one of the most abundant leaf metabolites.

The observed accumulation patterns were consistent across the two growing seasons (Table S3). Conversion of metabolite content from a fresh‐weight basis to a per‐organ basis resulted in only minor changes in cultivar ranking, and the overall trends remained consistent (Table S4). Across all cultivars, fruits consistently accumulated higher levels of neochlorogenic acid than chlorogenic acid, whereas the opposite trend was observed in leaves.

UPLC–HRMS Profiling Highlights Organ‐Specific Metabolite Class Distribution

UPLC–HRMS‐based untargeted metabolomics further confirmed the strong differentiation between fruits and leaves, generating two distinct base peak chromatographic profiles depending on the analysed organ (Figure 3).

FIGURE 3: UPLC–HRMS base peak chromatograms of representative sweet cherry fruit and leaf extracts. Chromatograms highlight the organ‐specific distribution of major metabolites. Peak identities are reported where available; U.i., unidentified compound. Time is expressed in minutes. BPI, Base Peak Ion.

Data processing resulted in a matrix of 3219 signals, of which 115 m/z features were putatively annotated (File S1). These metabolites encompassed diverse chemical classes, including hydroxycinnamic acids, hydroxybenzoic acids, flavonoids, anthocyanins, flavan‐3‐ols, flavanones, flavones, chalcones, flavonolignans, cyanogenic glycosides, jasmonates, and stilbenoids (Figure 4).

FIGURE 4: Relative distribution of metabolite classes in fruit and leaf extracts of sweet cherry obtained by UPLC–HRMS analysis. Values are expressed as percentage of the total ion signal (set to 100%) and represent the mean ± standard deviation of all cultivars in the two growing seasons (2014 and 2015). HCAs, Hydroxycinnamic acids; HBAs, Hydroxybenzoic acids.

The annotated metabolites accounted for approximately 54%–57% of the total chromatographic signal, and their distribution markedly differed between organs (Figure 4). Leaves were mainly characterized by flavonols (~29% total signal), hydroxycinnamic acids (~21% total signal), and flavan‐3‐ols (~4% total signal), whereas fruits were dominated by anthocyanins (~26% total signal), followed by hydroxycinnamic acids (~13% total signal), flavan‐3‐ols (~8% total signal), and flavonols (~4% total signal). Comparable metabolite class distributions were observed across the two growing seasons.

Multivariate Analysis Discriminates Fruits and Leaves Based on Metabolite Composition

Multivariate analysis of the UPLC–HRMS dataset revealed a clear and expected separation between fruit and leaf samples. Principal component analysis (PCA) showed a strong organ‐driven discrimination, with the first principal component explaining 68.2% of the total variance (Figure 5a).

FIGURE 5: Multivariate statistical analysis of fruit and leaf metabolomic profiles. (a) Score scatter plot of principal component analysis (PCA) performed on the UPLC–HRMS dataset, showing clear separation between fruit and leaf samples along PC1. (b, c) OPLS–DA loading plots obtained by comparing fruit and leaf samples, illustrating metabolites positively correlated with each organ. Triangles represent metabolites, and colors indicate the metabolite chemical class.

Supervised OPLS–DA analysis further highlighted the metabolites most strongly associated with each organ (Figure 5b,c; File S2). Hydroxycinnamic acid derivatives, particularly dicaffeoylquinic acid isomers, and flavonols were mainly associated with leaf samples, whereas anthocyanins and flavan‐3‐ols were preferentially associated with fruits.

Genotype‐Dependent Variation in Fruit and Leaf Metabolome Composition

To investigate genotype effects independently within each organ, fruit and leaf datasets were analysed separately. PCA of fruit samples explained 57.8% of the total variance across the first two components and revealed a cultivar‐associated distribution of samples (Figure 6a). The overall PCA model included six components and showed good descriptive and predictive performance (R 2 X(cum) = 0.869; Q 2(cum) = 0.812). Although the dataset included samples collected for each cultivar across two different growing seasons and from multiple commercial orchards, samples belonging to the same cultivar consistently clustered together.

FIGURE 6: Multivariate statistical analysis of sweet cherry fruit metabolomes. (a) PCA score scatter plot showing cultivar‐specific clustering of fruit samples and the identification of three main cultivar groups. (b) O2PLS–DA loading plot highlighting metabolites associated with each cultivar group. Triangles represent metabolites, and boxes represent classes. Different colors were used to highlight the chemical class of the metabolite. (c) Heat map showing relative abundances of fruit metabolites selected from the O2PLS–DA loading plot as the variables most associated with cultivar‐group discrimination. No hierarchical clustering was applied; metabolites are arranged according to their association with the corresponding cultivar groups. Colors indicate normalized abundance levels (red, high; blue, low).

Based on metabolome similarity, three main cultivar‐groups were identified: Sandra and Grace Star (group 1), Romana and Burlat (group 2), and Durone Rosso, Ferrovia, and Early Bigi (group 3). Subsequent O2PLS–DA analysis identified the metabolites contributing most strongly to the discrimination among cultivar groups (Figure 6b,c). The cultivar‐based model, built using three cultivar groups as class labels, included two predictive and three orthogonal components and showed high descriptive and predictive performance (R 2 X(cum) = 0.829; R 2 Y(cum) = 0.900; Q 2(cum) = 0.888), supporting the presence of a strong cultivar‐associated component in fruit metabolomic profiles. Sandra and Grace Star were positively characterized by feruloyl quinic acid isomers and neochlorogenic acid, whereas Romana and Burlat were associated with flavan‐3‐ols and coumaroyl quinic acid derivatives. The third group, including Durone Rosso, Ferrovia, and Early Bigi, was mainly characterized by anthocyanins, particularly cyanidin‐3‐O‐rutinoside.

To explicitly evaluate whether this distribution was associated with growing season or orchard of origin, the same fruit PCA score plot was additionally visualized by coloring samples according to year and orchard (Figure S1a,d). Compared with the cultivar‐based representation, no clear separation exclusively associated with year or orchard was observed, suggesting that cultivar‐associated structure was more evident than year‐ or orchard‐associated separation in the unsupervised analysis.

Because the two growing seasons differed in meteorological conditions, a supervised OPLS–DA model was built to evaluate the year effect in fruit samples (Figure S1b,c). The model separated samples according to growing season and showed high predictive ability (R 2 X(cum) = 0.902; R 2 Y(cum) = 0.901; Q 2(cum) = 0.847), indicating that inter‐annual environmental variation contributed to the fruit metabolome. However, this year‐associated separation was not evident in the unsupervised PCA, and the predictive component accounted for only a small fraction of the X variance (R 2 X[1] = 0.023), suggesting that the year effect, although detectable by supervised modeling, was not the most evident source of variation in the unsupervised fruit metabolomic dataset.

The possible contribution of orchard location was also evaluated by O2PLS–DA using the nine orchards as class labels (Figure S1e,f). The model explained a large fraction of the X variance (R 2 X(cum) = 0.908), but only a moderate fraction of the Y variance (R 2 Y(cum) = 0.413), with limited predictive ability (Q 2(cum) = 0.310). Consistently, the score plot did not show clear orchard‐specific clustering, and the corresponding loading plot did not reveal strong orchard‐specific metabolite associations. These results suggest that orchard‐related variation was detectable but less clearly structured than cultivar‐associated variation.

PCA of leaf samples revealed a clear cultivar‐dependent clustering (despite, as for fruits, samples of each cultivar originating from different orchards and being collected across two growing seasons), identifying four distinct groups (Figure 7a). Sandra and Romana each formed independent clusters, while Durone Rosso and Ferrovia grouped together, and Burlat, Grace Star, and Early Bigi formed a fourth cluster. The PCA model included six components and showed good descriptive and predictive performance (R 2 X(cum) = 0.799; Q 2(cum) = 0.717), indicating that a substantial fraction of the leaf metabolomic dataset was captured.

FIGURE 7: Multivariate statistical analysis of sweet cherry leaf metabolomes. (a) PCA score scatter plot showing cultivar‐dependent clustering and the identification of four main groups. (b, c) O2PLS–DA loading plots displaying metabolites correlated with specific cultivar groups. Triangles represent metabolites, and boxes represent classes. Different colors were used to highlight the metabolite chemical class. (d) Heat map showing relative abundances of leaf metabolites selected from the O2PLS–DA loading plots as the variables most associated with cultivar‐group discrimination. No hierarchical clustering was applied; metabolites are arranged according to their association with the corresponding cultivar groups. Colors indicate normalized abundance levels (red, high; blue, low).

O2PLS–DA analysis highlighted cultivar‐specific metabolite associations (Figure 7b–d). The cultivar‐based model, built using four cultivar groups as class labels, included three predictive and three orthogonal components and showed high descriptive and predictive performance (R 2 X(cum) = 0.784; R 2 Y(cum) = 0.939; Q 2(cum) = 0.932), supporting the presence of a reproducible cultivar‐associated component in leaf metabolomic profiles. Sandra leaves were characterized by elevated levels of hydroxycinnamic acids, including caffeoyl threonic acid derivatives and neochlorogenic acid, as well as phloridzin and benzyl alcohol hexose hexose. Durone Rosso and Ferrovia showed higher abundance of glycosylated and acylated coumaroyl derivatives, caffeoyl hexose, caffeoyl threonic acid deoxyhexoside, feruloyl quinic acid isoform 2, amygdalin, and the malonylated and acetylated forms of kaempferol‐O‐hexoside. Romana was associated with caffeoyl coumaroyl threonic acid and coumaroyl threonic acid. Burlat, Grace Star, and Early Bigi were characterized by higher levels of dicaffeoyl quinic acid isomers and quercetin glycosides.

Similarly to fruits, the leaf PCA score plot was also represented by coloring samples according to growing season and orchard of origin (Figure S2a,d). Cultivar‐associated grouping remained more evident than year‐ or orchard‐associated separation in the unsupervised analysis. Because the two growing seasons differed in meteorological conditions, a supervised OPLS–DA model was further built to evaluate the year effect in leaf samples (Figure S2b,c). This model separated samples according to growing season and showed good predictive ability (R 2 X(cum) = 0.775; R 2 Y(cum) = 0.856; Q 2(cum) = 0.774), suggesting that inter‐annual environmental variation contributed to the leaf metabolome. However, the year effect was not clearly visible in the unsupervised PCA, and the predictive component accounted for only a small fraction of the X variance (R 2 X[1] = 0.0227). Moreover, the corresponding S‐plot did not reveal metabolites strongly correlated with a specific growing season. These observations suggest that the year effect was detectable at the multivariate level but was broadly distributed across the metabolome rather than driven by strongly year‐specific individual metabolites.

The possible contribution of orchard location was also evaluated by O2PLS–DA using the nine orchards as class labels (Figure S2e,f). The model included eight predictive and two orthogonal components and explained a large fraction of the X variance (R 2 X(cum) = 0.874), but a lower fraction of the Y variance (R 2 Y(cum) = 0.416), with limited predictive ability (Q 2(cum) = 0.286). Consistently, the score plot did not show a clear orchard‐specific clustering, and the corresponding loading plot did not reveal strong orchard‐specific metabolite associations. These results suggest that orchard‐related variation was detectable in leaf samples, but its association with the metabolomic profiles was limited and not uniformly supported across orchard classes.

Overall, the different cultivar‐associated clustering patterns observed in fruits and leaves suggest that genotype‐associated regulation of metabolite biosynthesis and/or accumulation is at least partly organ‐specific. At the same time, the additional year‐ and orchard‐based models indicate that environmental variation contributed to both fruit and leaf metabolomes, particularly at the year level. Within the limits of the present experimental design, cultivar‐associated structure appeared more evident than orchard‐associated variation, while inter‐annual variation represented a detectable additional component.

Cultivar‐Associated Metabolites Are Conserved Across Organs

Overall, leaf and fruit metabolomes did not show similar cultivar‐dependent distribution patterns across the seven Prunus avium cultivars. However, within‐cultivar comparisons identified a subset of metabolites displaying conserved accumulation patterns in both leaves and fruits across cultivars, suggesting a shared genotype‐ (cultivar‐) dependent regulation independent of organ and growing season (Figure 8). The OPLS‐DA analysis revealed that Sandra and Romana accumulated both caffeoyl threonic acid derivatives and coumaroyl threonic acid derivatives in both fruits and leaves, with higher levels observed in fruits for the former and in leaves for the latter (Figure 8a,b). Early Bigi was specifically associated with the flavone apigenin‐O‐hexoside both in fruits and leaves (Figure 8c), while Burlat and Grace Star showed lower levels of dihydroxybenzyl alcohol hexose (Figure 8d).

FIGURE 8: Box plots of metabolites associated with specific sweet cherry cultivars and conserved across organs and growing seasons. Data represent metabolite abundance in fruits and leaves across cultivars. Statistical analysis was performed using one‐way ANOVA followed by Tukeys HSD test (p < 0.05). Different letters indicate statistically significant differences among cultivars.