Olive plants were grown and drought-stressed as previously described (Parri et al., 2023). Briefly, 18-month-old olive trees (Olea europaea L., cultivars Leccino, Maurino, and Giarraffa) were grown in 4-L pots (15x15x20 cm) with a substrate of 50% peat and 50% pumice (Tosca et al., 2019). The plants were then transferred to a growth chamber with LED illumination for flowering and growth (photoperiod of 12 hours light and 12 hours dark), which provided a photosynthetic photon flux of 450-550 µmol m^-2 s^-1 at plants’ leaves level. After one week of acclimation, plants of each cultivar were divided into two groups: a fully irrigated control group (CTRL) and a drought-stressed group (DS) that was water deprived for 4 weeks. The water deficit period was divided into 3 time points (t0, t2, t4) corresponding to the onset of withholding irrigation, the second and the fourth weeks of water deprivation, respectively. Three plants per cultivar were sacrificed for stem analysis at t0. At both time points, t2 and t4, three plants from each of the drought-stressed and control groups were sacrificed for stem analysis. Leaf samples were taken from at least 6 individuals at all time points. Experimental design is shown in Supplementary Figure S1 of the Supplementary Material. Temperature and humidity were recorded hourly (average temperature of 27.5°C and humidity of 51.1%). The pots inside the chamber were rotated weekly to avoid any positional effects.

The growth of the plants was monitored by measuring the difference between the height at t4 and at t0 of 6 plants for each experimental group. The control group of ‘Giarraffa’, ‘Leccino’ and ‘Maurino’ grew by about 2.6, 2.4, 4.3 cm respectively, while the drought stressed groups of ‘Giarraffa’, ‘Leccino’ and ‘Maurino’ had a growth of about 2.2, 1.7, 1.8 cm.

Frozen olive leaves and stems harvested at t0, t2 and t4 from both control and stressed groups of cultivars ‘Giarraffa’, ‘Leccino’ and ‘Maurino’ were dried at 40°C for 7 days. The dried plant material was then finely ground in a mill to obtain a powder for metabolite extraction. The ground material was combined with n-hexane (1:10 w:v) at room temperature with magnetic stirring for 48 hours. The n-hexane was then decanted and a second extraction was performed with fresh n-hexane for a further 24 hours. The combined n-hexane extracts were concentrated to dryness in a rotary evaporator under reduced pressure to give the crude extracts, which were air dried for one week. The pellet resulting from the n-hexane extraction was further air dried and then subjected to extraction with 50 mL of methanol to isolate phenolic compounds. The methanol extraction involved two cycles: a first cycle of 48 hours at room temperature with magnetic stirring, followed by removal of the methanol, and a second cycle of 24 hours with fresh methanol. The methanol extracts from both cycles were pooled and concentrated to dryness in a rotary evaporator under reduced pressure. Finally, the concentrated extract was air dried for two weeks.

The extracted samples from the n-hexane extraction were weighed and prepared for silylation. In a glass tube, a 200 μL aliquot of the extract was mixed with 200 μL of tetracosane solution (0.5 mg mL^-1), 250 μL of pyridine, 250 μL of N,O-bis(trimethylsilyl)trifluoroacetamide and 50 μL of trimethylsilyl chloride. The mixture was then incubated at 70°C for 40 minutes. After the incubation period, 1 μL of the silylated extract was injected into a gas chromatography-mass spectrometry (GC-MS) apparatus (QP2010 Ultra Shimadzu, Kyoto, Japan). The chromatography conditions followed the protocols described in Dias et al. Dias et al. (2019). Data were collected at a rate of one scan s⁻¹ over a range of m/z 33–750, as described in Dias et al. (2019). For identification of lipophilic compounds, chromatographic peaks were compared with entries in mass spectral databases such as the NIST14 Mass Spectral Library and the Wiley Registry^® of Mass Spectral Data. In addition, comparison was made with mass spectra and retention times of pure compounds prepared and analyzed under conditions similar to those of the samples. The identification of some compounds was done using the retention index relative to n-alkanes injected in the same chromatographic conditions as described in Costa et al. (2021). Calibration curves were established for quantification purposes using standard compounds representing the major families of compounds present in the extracts. These standards included maltose for sugars (y = 0. 0416x + 0.0117, R² = 0.99), palmitic acid for fatty acids (y = 0. 0941x + 0.5236, R² = 0.99), octadecane for alkanes (y = 0. 0942x + 0.061, R² = 0.99), octadecanol for alcohols (y = 0. 2322x + 0.0474, R²  = 0.99), and cholesterol for sterols and terpenes (y = 0. 0596x + 0.0447, R² = 0.99). The detection and quantification limits (LOD and LOQ, respectively) were determined from the parameters of the calibration curves. Quantitative results were expressed as grams per kilogram dry weight (g/Kg DW) and are presented as the mean ± standard deviation of three independent analyses.

Methanol residues were weighed and resuspended in methanol to achieve a final concentration of 10 mg/mL. Next, 4 μL of the solution was injected into a Thermo Scientific Ultimate 3000RSLC Dionex (Waltham, MA, USA) equipped with a Dionex UltiMate 3000 RS diode array detector coupled with a mass spectrometer operating in negative ion mode. The analysis was performed using a Hypersil GOLD column (1.9 μm particle diameter, Thermo Scientific, Lenexa, KS, USA) as described in Dias et al. (2019). A mass range of 50.00–2000.00 m/z was covered. Compound identification was also performed as described in the same reference. UV-Vis spectral data were collected between 250 and 500 nm, and the chromatogram profile was recorded at 280 nm. A semi-quantitative analysis was performed using peak integration through the standard external method. The peaks were identified by comparing the retention times, UV-Vis spectra, and spectral data obtained from the reference compounds. The calibration curves were calculated using the reference compounds (quercetin for flavonoids and oleuropein for secoiridoids) to determine the detection and quantification limits. The calibration curves were generated by injecting various concentrations of quercetin and oleuropein. The equation for quercetin was y = 4 × 106x − 390882 with an R² value of 0.99, where x represents the amount of the compound in mg/mL and y represents the peak area obtained in the chromatogram. The equation for oleuropein was y = 106x − 6948 with an R² value of 0.98, where x represents the amount of the compound in mg/mL and y represents the peak area obtained in the chromatogram. The detection and quantification limits (LOD and LOQ, respectively) were determined from the parameters of the calibration curves. The same conditions were used for the sample analysis. The compound concentration was measured in milligrams per gram of tissue dry weight. The mean ± standard deviation of three independent analyses per sampling time and treatment are presented.

A “group” is defined by the cultivar considered (Giarraffa, Leccino and Maurino), the irrigation treatment (control, CTRL or drought stressed, DS) and the time of sampling (beginning of stress, t0; two weeks after the beginning, t2; four weeks after the beginning, t4). Each group was analyzed in triplicate. Data distributions were tested for normality using the Shapiro-Wilk normality test in R studio (ver. 4.2.2, R core team, Vienna, Austria, 2022). Repeated measures ANOVA was used to test the significance of each of the three factors (treatment, cultivar and organ) and their interactions. Post hoc This statistical analysis was conducted using the Systat 11 statistical package (Systat Software Inc., Richmond, CA, USA). Due to the absence of the “MAU DS t4” group of stems, leaf and stem data were processed separately for the following analysis. This allowed not to exclude this group from the leaf data processing. For the following analyses, first, metabolite contents were normalized by Z-score and missing data were replaced by a value of 0.0001. Hierarchical clustering analysis (HCA) was performed on metabolites using Euclidean distance as the similarity metric and the complete linkage method between groups. The resulting heat maps and clustering were generated using Rstudio (ver. 4.2.2, R core team, Vienna, Austria, 2022) with the package “pheatmap” version 1.0.12. The observation inputs correspond to the variance of each group at t2 and t4, when the drought stress occurred. The variables input were the classes of both phenolic and lipophilic compounds. The Principal Component Analysis (PCA) biplots were realized with Rstudio (ver. 4.2.2, R core team, Vienna, Austria, 2022) with the package “Factoextra” version 1.0.7.

As a first general observation, 20 phenolic compounds were identified (16 flavonoids and 4 secoiridoids): 6 in both plant organs, 3 in stems, and 10 in leaves only. Retention times, m/z, MS2 (m/z) fragments and mean ± standard deviations for each experimental group at t0 and t2 are summarized in Supplementary Tables S1 , S2 for leaves, and in Supplementary Tables S3 – S5 for stems ( Supplementary Material ). As part of an ongoing collaboration, data on phenolic compounds from leaves at t4 are published in Cerri et al. (2024) ( Table 1 ); in that manuscript, olive leaf extracts were characterized and tested on human umbilical vein endothelial cells (HUVECs). Tables 1 – 3 show the repeated measures ANOVA results for phenolic compounds found in the stem and leaf, leaf only, and stem only, respectively. Results indicate that cultivar had a significant effect on all compounds examined (p-value < 0.05). Indeed, the cultivars studied contained significantly different amounts of the identified compounds, with some compounds found only in one cultivar (for example, fraxamoside was found only in the stem of ‘Leccino’). Unlike the other two cultivars, in the leaf of ‘Giarraffa’ luteolin-7-O-rutinoside was not detected. Instead, another isomer 1 of luteolin-7-O-glucoside was detected in this cultivar, but not in ‘Leccino’ and ‘Maurino’. In terms of leaves, chrysoeriol-7-O-glucoside and luteolin-7-O-glucoside is.3 were found only in the cultivar Leccino. The “organ” parameter significantly influenced the amount of compounds found in both leaves and stems. For example, luteolin, apigenin, and oleuropein were more abundant in leaves, while luteolin-7-O-glucoside and dihydroquercetin were more abundant in the stems.

It is noteworthy that drought treatment affected most of the compounds found in the leaves (such as oleuropein aglycone, aldehyde form of decarboxyl elenolic acid, luteolin-7-O-rutinoside) as well as the ubiquitous compounds (luteolin-7-O-glucoside, oleuropein, luteolin, and apigenin), but not the compounds found only in the stems. However, for dihydroquercetin, quercetin, apigenin-7-O-rutinoside is.1, and luteolin-7-O-glucoside is.3, which were unaffected by drought stress alone, the interaction between cultivar and treatment was significant. Only luteolin-7-O-glucoside was significantly affected by treatment, with no differences in cultivar-organ interactions with treatment.

A total of 30 lipophilic compounds were detected, belonging to sterols and terpenes, sugars, alcohols, fatty acids, and alkanes. Leaves and stems had distinct profiles, with 12 compounds found in both leaves and stems, 12 in leaves and 6 in stems only. Retention times and mean ± standard deviations for each experimental group at t0, t2 and t4 are summarized in Supplementary Tables S6 – S8 for leaves and in Supplementary Tables S9 – S11 for stems ( Supplementary Material ). Tables 4 – 6 show the results of the repeated measures ANOVA for lipophilic compounds found in both stem and leaf, leaf only, and stem only, respectively. Cultivar had a significant effect on all identified compounds except for sugar turanose in stems. Similarly, alpha-tocopherol was not detected in the leaves of ‘Maurino’, and phytol was only found in the leaves of ‘Giarraffa’. The “organ” parameter significantly influenced the compound profile of both stems and leaves. In fact, the leaves contained more alpha-D-mannopyranose, alpha-D-sorbitol, alpha-D-glucose, palmitic acid, stearic acid, alpha-monopalmitin, alpha-monopalmitin derivatives, lupeol derivatives, and ursolic aldehyde than the stems. Only oleic acid and alpha-linolenic acid were more abundant in the stems. Ursolic acid was the only compound for which neither organ, treatment, or interaction had a significant effect. Other compounds not significantly affected by drought stress included alpha- and beta-amyrin, two long-chain alkanes in leaves, and turanose in stems. However, treatment, cultivar, and organ interactions lose significance for D-glucose, whereas drought becomes significant for alpha-amyrin and the second long chain alkane found in the cultivar by treatment interaction.

HCA analysis of leaf metabolites ( Figure 1 ) revealed two major clusters (clusters 1 and 2), each with two subclusters (1a, 1b, 2a, 2b). The sub-cluster 1a consisted of 8 phenolic compounds that were very abundant in the cultivar Maurino; in particular, the metabolites luteolin-7-O-rutinoside, dihydroquercetin and oleuropein were very abundant in “MAU DS t4”, whereas apigenin-7-O-rutinoside is.3 was highly enriched in “MAU DS t2”. The metabolites of this cluster (especially chrysoeriol-7-O-glucoside and luteolin-7-O-glucoside is.3) were significantly less abundant in ‘Leccino’ regardless of time point or treatment. Furthermore, metabolites accumulated in “MAU DS t4” (including apigenin-7-O-rutinoside is.3) were less abundant in all ‘Giarraffa’ groups compared to the other cultivars. The sub-cluster 1b contained 10 metabolites (mainly lipophilic compounds except for the flavonoid diosmetin) the content of which was significantly lower in ‘Giarraffa’ and in “MAU DS t4” regardless of time point or treatment. “GIA DS t2” had particularly a very low content of D-glucose and D-sorbitol. Metabolites from this cluster accumulated preferentially in all other groups of ‘Leccino’ and ‘Maurino’ (except for “MAU DS t4”).

The sub-cluster 2a contains 6 phenolic compounds but only 2 lipophilic compounds. The metabolites of this sub-cluster were mainly accumulated in the cultivar ‘Leccino’ regardless of the time point and treatment, although ‘GIA DS t2’ had a high amount of alpha-tocopherol. Conversely, the compounds in this subcluster showed a lower content in both ‘Giarraffa’ and ‘Maurino’. The 14 compounds on the left in sub-cluster 2b were all lipophilic, with the exception of the flavonoid luteolin-7-O-glucoside is.1. The metabolites of this cluster accumulated more in ‘Giarraffa’ (regardless of time point or treatment) than in the other two cultivars; in fact, their content was particularly low in “MAU DS t4”. The HCA revealed a clear separation of leaf lipophilic and phenolic compounds, which were mostly grouped into distinct clusters.

Cultivar clustering analysis (including time points and treatments) revealed two major sub-clusters (B1 and B2) of cluster B, as well as an orphan group containing only “MAU DS t4” (cluster A), which behaved very differently from the other groups. In particular, it showed a very low content of all long-chain alkanes and other metabolites belonging to the sterol and terpene class, whereas a huge amount of many phenolic compounds of sub-cluster 1a were found, especially oleuropein, dihydroquercetin and luteolin-7-O-rutinoside. Sub-cluster B1 included the Giarraffa cultivar, which differed from the other two cultivars in the higher content of long-chain alkanes, sterols, and terpenes and the lower amount of some fatty acids such as oleic and linoleic acid. Within sub-cluster B1, the metabolic composition of the ‘Giarraffa’ control differed from the metabolic profile of ‘Giarraffa’ under drought stress at t2 and t4, especially for the lower amount of D-glucose, neophytadiene and D-sorbitol. Sub-cluster B2 included both ‘Maurino’ and ‘Leccino’, which behaved much more similarly to each other than to ‘Giarraffa’, especially in terms of lipophilic compounds (sub-clusters 1b and 2b). The ‘Maurino’ control and drought-stressed samples at t0 were further separated from the drought-stressed ‘Maurino’’ at t2, where there was a higher amount of apigenin-7-O-rutinoside is.3, oleuropein aglicone and luteolin-7-O-glucoside is.2. Similarly, in cultivar Leccino, the control and stressed samples at t0 differed significantly from the drought-stressed samples at t2 and t4, which contained lower levels of diosmetin and dihydroquercetin. As a result, the groups in cluster B were more similar within the cultivar than when treatment conditions were considered. In particular, sub-cluster B1 distinguished ‘Giarraffa’ from the other two cultivars. Within each cultivar cluster, the control groups formed a separate cluster from the stressed groups.

HCA was also used to assess the pattern of stem metabolites ( Figure 2 ). The analysis divided the compounds into two main clusters (1 and 2). Sub-cluster 1a consisted of two flavonoids (luteolin and quercetin-3-O-glucoside). Their amount remained constant or increased in the irrigated samples of ‘Leccino’, “GIA DS t4” and “MAU DS t2”. The sub-cluster 1b included two flavonoids (luteolin-7-O-glucoside is.2 and apigenin) and two lipophilic compounds (stearic acid and ursolic acid). They were more abundant in the Giarraffa cultivar regardless of time or treatment, but especially in “GIA DS t4” and in “MAU DS t4”. The sub-cluster 2a contained oleuropein, quercetin, and chrysoeriol-7-O-glucoside, phenolic compounds whose content decreased in the experimental groups of ‘Giarraffa’ and ‘Leccino’ but increased in the cultivar ‘Maurino’, especially in “MAU DS t2”. The sub-cluster 2b contained 15 lipophilic compounds, with a higher content in the stressed groups of ‘Maurino’ at t2, ‘Giarraffa’ and ‘Leccino’ at t4, as well as “LEC CTRL t4”. Exceptions included neophytadiene and stigmast-5-ene (which decreased significantly in “LEC DS t4”) and LCAlkane 4, which decreased in “MAU DS t2”. The clustering of stem lipophilic and phenolic metabolites was less clear than that of leaf compounds, except for the large clustering of the 15 lipophilic compounds.

The clustering of the experimental group of cultivars showed two distinct clusters. Cluster A was divided into two sub-clusters, the first of which (A1) included ‘Giarraffa’, which had high levels of ursolic acid, luteolin-7-O-glucoside is.2 and apigenin. The second sub-cluster (A2) included ‘Maurino’ and ‘Leccino’. Both showed low levels of many fatty acids, including linoleic acid, sugars and other metabolites belonging to sub-cluster 2b. Finally, cluster B included four experimental groups, three of which were exposed to drought stress. Interestingly, the ‘Maurino’ group, after 2 weeks of stress, was grouped together with the stressed ‘Leccino’ and ‘Giarraffa’ groups, stressed by 4 weeks of drought. All of them had higher amounts of the metabolites from sub-cluster 2 and, overall, of some other compounds (such as luteolin, quercetin-3-O-glucoside, neophytadiene) belonging to the other sub-clusters of metabolites.

PCA biplots were used to highlight differences between the experimental groups and identify the classes of leaf metabolites that contributed most to group separation. Principal components 1 (PC1) and 2 (PC2) together accounted for 74.6% of the variation in the leaf data. As shown in Figure 3A , all experimental groups of the Giarraffa cultivar were distinct from the other two cultivars and distributed in a restricted area, which corresponded to the class of alkanes. On the other hand, the ‘Leccino’ and ‘Maurino’ groups were less distinct than ‘Giarraffa’ because their distributions partially overlapped. However, ‘Leccino’ was more closely related to the classes of alcohols and sugars, while ‘Maurino’ was best associated with the classes of secoiridoids, sterols and terpenes, and fatty acids. However, when considering the treatment exposure of the experimental groups ( Figure 3B ), the control and stressed samples did not have a clearly defined distribution area. This means that drought stress elicited a diverse range of metabolic responses in the experimental groups, resulting in no distinct or few common metabolic responses to drought stress in the leaves of the three cultivars.

Figure 4 shows a PCA biplot, a graphical representation of the stem metabolite classes. PC1 and PC2 accounted for a significant 75.3% of the total variation in the stem data, indicating that these two components capture much of the information in the dataset. Like the leaf results in Figure 3 , the cultivars were separated according to the distribution of stem compounds ( Figure 4A ). This distinction was more pronounced than that observed between the control and drought treatments ( Figure 4B ). PC1 was primarily responsible for determining cultivar distribution. The secoiridoids class was the main contributor to this component, whit flavonoids and alkanes having negative loadings. This suggests that the class of secoiridoids had a significant impact on the distribution of cultivars along PC1. Two cultivars, ‘Leccino’ and ‘Giarraffa’, had similar and partially overlapping distributions. However, ‘Leccino’ was strongly associated with the negative loadings of five lipophilic classes, namely fatty acids, sterols and terpenes, sugars, and alcohols. In contrast, the cultivar Maurino had a different distribution. The distribution area of stem samples from both the stressed and control groups was extensive, with some overlap. This overlap indicated that the stems of the three cultivars did not share a common metabolic response pathway to drought.

Like phenolic compounds, leaf tissue contains more primary metabolites than stems. However, higher levels of unsaturated acids, such as alpha-linolenic acid and oleic acid, were observed. Two palmitic acid derivatives were found only in olive stems, which could be related to the lower amount of palmitic acid (a saturated acid) found in the stem than the leaf. In contrast to the phenolic profiles, the treatment significantly altered the profile of primary metabolites in both the stem and the leaf. Drought mainly increased the content of fatty acids (with the exception of linoleic acid), which can act as membrane reinforcement against peroxidation Piccini et al., 2022 or as an energy source for stress recovery Dias et al., 2021. The lower sugar levels available during stress may be due to a reduction in the photosynthetic process or/and an increased use of energy to cope with stress. Furthermore, drought stress decreased sterols content, which may be related to the conversion of sterols into steryl esters, which were linked to membrane reinforcing in drought-tolerant cultivars Rogowska and Szakiel, 2020, or simply an increase in membrane fluidity due to the stress Dias et al., 2021. Unlike previous studies on the UV-B stress response, drought stress did not significantly increase the content of long-chain alkanes. These compounds are involved in cuticle wax thickness, which can be useful in counteracting solar radiation Piccini et al., 2022 and even water loss Bacelar et al., 2004. This was not the case in our study. Sorbitol, a mannitol isomer, is an important osmoprotectant in olive trees. The increase of the levels of polyols, like D-mannitol and galactinol, were reported by Azri et al. (2024) in olive leaves. In the present study, sorbitol did not accumulate during drought stress, but it was found in high concentrations in the leaf, possibly aiding in cell turgor and/or acting as an antioxidant. Because no differences were found, the antioxidant properties of triterpenes and lupeol derivatives were most likely unnecessary after drought stress. The levels of these molecules were also found to be unaffected by stress Dias et al., 2021, with variations occurring only during recovery. Jiménez-Herrera et al. (2019) did not find changes in pentacyclic terpenes (maslinic acid, oleanolic acid, erythrodiol and uvaol) in olive leaves, suggesting that drought does not change the production of these compounds. In contrast, Azri et al. (2024) found a decrease of ursolic acid and oleanolic acid in response to drought.

Although the data collected allowed for the identification of the most responsive molecules and their different accumulation in the olive leaf and stem, the three cultivars analyzed showed distinct metabolic profiles of drought response. Lipophilic compounds responded more consistently than phenolic compounds within each cultivar, regardless of drought treatment. The two sub-clusters 2b of Figures 1 , 2 of leaf and stem metabolites, which contain most of the lipophilic compounds, clearly distinguish Giarraffa from the other two cultivars, regardless of treatment. ‘Giarraffa’ leaves contained a high concentration of long chain alkanes, which may be linked to the thickness of the epicuticular wax, allowing this cultivar to avoid excessive water loss. Furthermore, the abundance of palmitic and stearic acids, sterols, and terpenes may enable ‘Giarraffa’ to maintain good membrane fluidity while avoiding excessive permeability (Dias et al., 2021). The potential accumulation of wax on the leaf surface, combined with the physiological responses of ‘Giarraffa’ under drought stress Parri et al., 2023, suggests that this is a typical ‘drought-avoiding’ cultivar Fang and Xiong, 2015. In ‘Giarraffa’, water deficiency reduces stomatal conductance relatively early, resulting in increased stem water content and prolonged soil water availability. Another water-saving strategy of ‘Giarraffa’ is the accumulation of the osmoprotectant D-sorbitol in leaves exposed to two weeks of drought stress. In our previous study, ‘Giarraffa’ had the lowest level of lipid peroxidation and membrane damage, as measured by malondialdehyde content and electrolyte leakage assays Parri et al., 2023. Therefore, there was no need to adjust flavonoid and secoiridoid pools in response to drought, indicating a low level of oxidative stress Agati et al., 2012. However, ‘Giarraffa’ showed a significant decrease in D-glucose, indicating stomatal closure and a lower photosynthetic rate. ‘Giarraffa’ also showed higher levels of alkanes and fatty acids under UV-B conditions, but we did not observe a corresponding accumulation of flavonoids Piccini et al., 2022; this is most likely due to the early physiological response implemented by ‘Giarraffa’ Parri et al., 2023, which allows this cultivar to avoid drought and reduce oxidative stress. In contrast, the metabolic profile of ‘Maurino’ after four weeks of drought stress revealed unusual levels of phenolic and lipophilic compounds, putting it in a separate cluster. As suggested by the high levels of malondialdehyde and electrolyte leakage, the reduction in lipophilic compounds in both leaves and stems could be due to damage on cellular components caused by high levels of oxidative stress Parri et al., 2023. Surprisingly, the stressed group of ‘Maurino’ accumulated flavonoids and secoiridoids, which may aid in the regulation of oxidative stress in both the stem and the leaves. However, unlike Amaranthus tricolor Sarker and Oba, 2018 and Zea mays Li et al., 2021, which showed higher MDA and EL levels but lower flavonoid content, the antioxidant response of ‘Maurino’ did not allow it to avoid oxidative stress damage. Agati et al. (2012) proposed one possible explanation: flavonoid accumulation as an oxidative stress response occurs particularly in stress-sensitive individuals under severe stress conditions, when the first line of defense against ROS (antioxidant enzymes) is compromised. The metabolite response of ‘Leccino’ is intermediate. Unlike ‘Giarraffa’, ‘Leccino’ contains few long-chain alkanes, sterols, and terpenes and, like ‘Maurino’, the primary fatty acids are oleic and linoleic. However, under drought stress, the phenolic profile changes in a heterogeneous manner. Secoiridoids accumulate only after four weeks of drought, while changes in the flavonoid pool occurred only for a few of them, such as luteolin and apigenin-7-O-rutinoside and glucoside, at the cost of a decrease in apigenin and luteolin levels. However, the phenolic profile of this cultivar was associated with lower antioxidant capacity than the other two cultivars under drought stress conditions, as shown by Ferric Ion Reducing Antioxidant Power analysis Cerri et al., 2024.

Finally, analyses of stem and leaf phenolic and lipophilic profiles of the three Italian olive cultivars exposed to drought stress revealed a cultivar-specific response to drought. The cultivars Maurino and Leccino responded more similarly than Giarraffa. However, ‘Maurino’ showed the higher antioxidant response and the greater decrease in most of the lipophilic compounds, indicating a “drought stressed” profile, while ‘Giarraffa’ did not increase flavonoid and secoiridoid pools and showed higher levels of cell wall and cuticle wax components than the other cultivars, supporting the “drought avoidance” pattern shown by the physiological analyses (Parri et al., 2023).