Olive oil is one of the main components of the Mediterranean diet due to its nutraceutical properties (Piroddi et al., 2017). Among olive oil constituents, phenols play a very important role in its healthy properties (Serreli and Deiana, 2019). For example, olive phenols reduce chronic inflammation and help in the prevention of some diseases, such as Alzheimer’s (Omar et al., 2017) or different types of cancer (Piroddi et al., 2017).

The phenolic composition of olive oil depends, mainly on the phenolic composition of the fruit upon arriving to the mill factory, which is then modified by enzymatic processes occurring during oil extraction (Lukić et al., 2017). This initial fruit phenolic composition could be affected by many factors including genotype, harvest date, tree environmental conditions and management practices (Servili and Montedoro, 2002).

The wide olive germplasm has shown a high genetic variability for both phenolic content and composition in the evaluation of cultivar collections (Miho et al., 2018), breeding progenies (Pérez et al., 2014) and comparative trials (Pérez et al., 2018; El Riachy et al., 2013).

Several studies have also attempted to evaluate the phenolic variability of single cultivars in different environments of their country of origin, such as ‘Gemlik’ (Ben Ghorbal et al., 2018) in Turkey, ‘Picholine Marrocaine’ in Morocco (Bajoub et al., 2015), ‘Chemlali’ (Bouaziz et al., 2004) in Tunis and ‘Baladi’ in Lebanon (El Riachy et al., 2018). Among the environmental variables, phenol composition is mainly influenced by both abiotic and biotic stresses. Among the first ones, water stress has been reported as one of the main factors influencing phenol content and composition (Dabbou et al., 2015; Gómez-Rico et al., 2006; Gucci et al., 2019). In general, an increase in water stress implies an increase in phenolic content in olive oil (García et al., 2020). However, some individual phenols, such as lignans, might show the opposite pattern (Ovar et al., 2002). Biotic stresses such as olive fruit fly (Medjkouh et al., 2018) and Verticillium Wilt (Landa et al., 2019) can also increase oil phenolic content and composition.

The total phenolic content and composition normally decreases during the harvest season, as shown for different olive cultivars (Alowaiesh et al., 2018; Bengana et al., 2013; Bodoira et al., 2015; Dag et al., 2011). This includes ‘Arbequina’, the most widely planted cultivar in the world (Fernández-Escobar et al., 2013), whose phenolic content has been evaluated in olive oil (Abenoza et al., 2015; Morelló et al., 2004) and fruit (Talhaoui et al., 2015) through maturation in single locations in Spain. On the contrary, some phenols such as lignans and flavones have been reported to increase in oil through maturation (Bengana et al., 2013) as well as verbascoside (Bodoira et al., 2015) and hydroxytyrosol (Bouaziz et al., 2004) in fruits. In other cases, irregular patterns of variation in fruit phenols through maturation have been reported (Ben Ghorbal et al., 2018).

All the mentioned works on olive fruit phenols have been performed in Mediterranean climatic conditions. Few reports exist on olive phenols in other climates as is the case of Argentina (Bodoira et al., 2015). However, olive growing is currently expanding worldwide, in many cases outside the Mediterranean climate. Besides, this expansion is mainly based on few cultivars specially adapted to the new trends of olive growing (Fernández-Escobar et al., 2013), with ‘Arbequina’ being the clearest example. Therefore, it would be of great interest to compare the influence of very different climate conditions (Mediterranean vs. non-Mediterranean) on variation in an important component of olive oil such as phenolic compounds.

In this sense, Tenerife, in the Canary Islands, is one of the non-Mediterranean locations where olives now have some importance (Medina et al., 2018). Its Sub-Tropical climate might be of interest as a natural scenario with climatic conditions similar to those of the typical Mediterranean olive growing area in the likely event of a climate warming (Medina et al., 2020).

In the present work, we compared the variation in phenolic content and components of ‘Arbequina’ fruits through maturation in typical Mediterranean locations from Andalusia in Sothern Iberian Peninsula, with others from the Sub-Tropical climatic conditions in Tenerife, Canary Islands. That comparison was used to compare the test of location, harvest date and their interaction on phenolic content and composition.

The six locations involved in this work showed different climatic conditions during the experimental period, mainly associated with the region level (Figure 2). The two locations in Tenerife, Canary Islands, showed milder temperatures both in summer and winter, with around 7 ºC range of average maximum-minimum temperatures. Temperatures in Los Tomillos (located 200 m.a.s.l.) were 2-3 ºC higher than in El Viso (400 m.a.s.l.). Among the locations in Andalusia, Baena, La Rambla and Úbeda showed quite similar temperature patterns; while only slightly lower temperatures were observed in Antequera. Some differences in rainfall distribution among the Andalusian locations were also observed, particularly due to high rainfall in Antequera in the first week of November. Compared to the Peninsula, locations in Tenerife were characterized by limited rainfall throughout the year, with only exceptional rainfall observed in Los Tomillos in March.

The oil accumulation of ‘Arbequina’ in the six locations sampled showed a similar pattern considering that harvest 1 was done in August for El Viso and Los Tomillos (in Tenerife) and in September for the Andalusian locations (Figure 3).

Among all the identified phenols in ‘Arbequina’ fruits, oleuropein represented around half of the total phenolic content (Table 1), and together with demethyloleuropein, comseogoside, ligstroside and verbascoside constituted 95% of the total phenolic content. All the phenols showed high variability (high coefficient of variance), being especially high for verbascoside and oleuropein.

This high variability in phenolic content and composition was mainly due to location effect for phenols and most of the components (Table 2) except for oleuropein and ligstroside, for which location and harvest date showed comparable variance and for demethyloleuropein, which showed a double variance for the harvest date effect (41.9) compared to location (21.2) and their interaction (18.0). The interaction between location and harvest date represented the main contributor to total variance only for hydroxytyrosol-4-glucoside. Error variance represented more than half of the total variance for luteolin-7-O-glucoside. In any case, location, harvest date and their interaction showed a significant effect on phenolic content and all its components except for rutin and luteolin-7-O-glucoside, for which location was the only significant factor.

The phenolic components described were found in all locations and harvest dates (Table 3). The high variability due to location for total phenols was mainly due to the high values observed for Antequera and Baena (Table 3, Figure 4). While the decrease in total phenols with harvest date was, on average, more similar between harvest 1 and 2 than between harvest 2 and 3. However, for Antequera and Baena, this decrease was more evident between harvest 2 and harvest 3 -- just the opposite for the rest of the locations. Oleouropein, the major phenol identified, showed a similar pattern of variation to that of total phenols. The same could be said for ligstroside.

Different patterns of variation were observed for the rest of phenolic components (Table 3, Figure 4). Demethyloleuropein showed an increase between harvest 1 and harvest 2 in all locations except for Los Tomillos. In Antequera, this increase was also maintained between harvest 2 and harvest 3, maybe associated to a heavy rainfall at that time. Few variations among harvest dates were observed for comselogoside and verbascoside. Only in Antequera and Baena, the two locations with the highest values for both components, a significant decrease was observed in harvest 3.

For the four components with lower contents in ‘Arbequina’ fruits (hydroxytyrosol-4-glucoside, rutin, luteolin-7-O-glucoside and tyrosol-1-glucoside), the highest values were observed in Antequera and Baena samples, probably due to the water stress. The unique exception was hydroxytyrosol-4-glucoside with very high levels in the fruits from La Rambla, and a significantly different accumulation pattern observed in Los Tomillos. This component is probably less influenced by water stress. Few variations with harvest date were observed for rutin, except for the very high values in Baena in harvest 2.

Fruit traits were also evaluated in the six locations and on three harvest dates. Fruit size, moisture and maturity index showed most of the variance due to the location effect. The oil content was mainly influenced by harvest date, as expected (data not shown). Antequera was the location with the smallest fruit size; while Los Tomillos showed higher oil content than the rest of the locations (Table 4). Fruit moisture was much lower in the two dry farming locations (Antequera and Baena), as expected. Maturity index was very delayed in Antequera, and very advanced in Úbeda, although intermediate in the rest of the locations.

A separate analysis of variance was performed for the four irrigated locations to better test differences among the Mediterranean and Subtropical conditions of Andalusia and Tenerife regions (Table 5). Region was the main contributor to the total variance for total phenols, comselogoside, verbascoside, rutin and tyrosol-1-glucoside. This effect was mainly due to the higher values for those components in the two Andalusian locations (La Rambla and Úbeda), especially for the first two harvest dates (Figure 4). While oleuropein and lisgtroside showed higher variance for harvest date mainly due to the sharp decrease in their contents, especially between harvests 2 and 3.

The correlation between phenolic content components and with fruit traits were also studied (Table 6). Total phenolic content showed a high positive correlation with most individual phenolic components except for demethyloleuropein, which increased throughout the harvest season, and hydroxytyrosol-4-glucoside and luteolin-7-O-glucoside which did not show a clear decrease during ripening. Among the phenolic components, the stronger correlation was found between oleuropein and ligstroside. Although comselogoside, verbascoside and rutin seemed to also be highly correlated. Besides, tyrosol-1-glucoside showed very high correlation coefficients with verbascoside and comselogoside. Both have a similar chemical structure in which the glucose is directly linked to the phenolic alcohol moiety, hydroxytyrosol and tyrosol, respectively. No high correlations were found between the total phenolic content or specific phenolic components and fruit traits, with most of them being negative, except for demethyloleuropein and tyrosol-1-glucoside. Ripening index showed low correlations with phenol content and composition and with the fruit traits evaluated. In particular, fruits having a high ripening index (more than 2) showed low phenol contents (less than 20,000 µg/g of fresh olive pulp); while fruits having a ripening index lower than 2 showed a very large range in variation in phenol contents (from 8,180 to 51,767 µg/g of fresh olive pulp).

The nine major phenolic components identified, in all locations and harvest dates, showed high variability (coefficient of variance higher than 50%, except for comselogoside and tyrosol-1-glucoside). Oleuropein was the main phenol as previously reported for olive fruits of other cultivars (Ben Ghorbal et al., 2018; Valente et al., 2020).

The high variability found was mainly due to a location effect for total phenols and for most individual phenolic components. Significant the effect of location was previously reported for phenols in several cultivars. This is the case of ‘Gemlik’ in Turkey (Ben Ghorbal et al., 2018), ‘Arbequina’, ‘Manzanilla’ and ‘Arauco’ in Argentina (Bodoira et al., 2015) and Italian cultivars from a central region of Italy (Mousavi et al., 2019).

Most variability associated with location was caused by the high values observed for Antequera and Baena compared to the other four locations. The most outstanding difference between Antequera and Baena and the rest of the locations is that olive trees are managed in dry farming, while, in Los Tomillos, El Viso, La Rambla and Úbeda, drip irrigation is used. The great influence of water availability on phenolic content and composition of olive fruits and virgin olive oils was previously reported when olive dry farming and irrigation were compared (Gómez-Rico et al., 2006; Cirilli et al., 2017). It seems that water stress periods caused a higher concentration of phenolic compounds in in the dry farming locations of Antequera and Baena, which remained during most of the maturation period. Water availability in the dry farming locations was especially low at the beginning of the oil accumulation period with respect to the four irrigated locations. Previous reports have suggested that water stress in this period has a strong influence on the phenol content (Gucci et al., 2019). This is maybe the reason why rainfall at the beginning of November in Antequera had little influence on phenol content. However, in some other works, a decrease in phenol content with higher stress has been reported (Valente et al., 2020). Probably, in the latter case, the combination of water and heat stress gave a different response of the olives.

When considering only the four irrigated locations, two in Andalusia and two in Tenerife, higher content for total phenols and some components was observed in Andalusia. These differences were especially important for the first harvest date. The higher summer temperatures in Andalusia with respect to Tenerife could produce higher stress which could be the cause of those differences.

Previously, the Sub-Tropical temperatures of Tenerife have shown a great influence on the flowering phenology of the olive tree (Medina-Alonso et al., 2020). This is important, since the Tenerife climatic conditions could help to predict the influence of climate change in the Mediterranean climate. In our case, it seems that the main factor associated with climate change that would impact phenol content is water availability more than changes in air temperature. However, the higher heat stress in summer, predicted in a climate change scenario, could also increase phenol content although have negative influence on other parameters such as oil content (Navas et al., 2019). More experimentation is needed to accurately determine the influence of climate change on phenol content and composition.

Harvest date was also showed to influence the variability in total phenols and phenolic components. As observed in our trials, most of the previous works reported a decrease in total phenolic content with maturation (Abaza et al., 2017; Ferro et al., 2020), including ‘Arbequina’ in North-East Spain (Benito et al., 2013). However, different patterns of variation were observed for individual phenolic components. In fact, the decrease in oleuropein content throughout the ripening process was concomitant with a parallel increase in demethyloleuropein, as previously reported (Gómez-Rico et al., 2008). This was the only component with a significant increase with harvest date. Previous reports have also shown different variation patterns for the different phenolic components in fruit (Ben Ghorbal et al., 2018; Talhaoui et al., 2015; Ferro et al., 2020).

ANOVA analysis revealed the comparison of the relative influence of location and harvest date. For total phenols and some components such as comselogoside, tyrosol-1-glucoside, verbascoside and rutin, location had a much greater influence than harvest date. On the contrary, in the case of demethyloleuropein, harvest date was more important. And for other components such as oleuropein and ligstroside, both factors were equally important. Therefore, the relative influence of harvest date and location very much depend on the phenolic component. No previous study on the comparative variance analysis has been performed for olive phenols, except for the case of ‘Baladi’ in Lebanon (El Riachy et al., 2018), where a greater influence of location was observed with respect to harvest date for total phenol content determined spectrophotometrically using the Folin-Ciocalteu method. Different influences of location and genotype for the different phenolic components were previously reported in a set of breeding trials (Pérez et al., 2018).

As expected, all phenolic components showing similar behavior across locations and harvest dates were highly correlated. The extremely high correlation found for oleuropein and ligstroside suggest a common biosynthetic pathway for both secoiridoid glucosides. While the negative correlation between oleuropein and demethyloleuropein support the hypothesis on the interconversion of oleuropein into demethyloleuropein (Obied et al., 2008). On the contrary, demethyloleuropein, hydroxytyrosol-4-glucoside and luteolin-7-O-glucoside showed no correlation with the rest as having a different pattern of variation. Different correlations than the ones reported here were found in a previous set of breeding trials (Pérez et al., 2018). This is probably due to the fact that the variability here is only due to environmental factors; while in the previous work, the phenol variability is attributable to different environments but also to different genotypes. The correlation of phenols with oil and moisture contents in the fruit was low and negative, which seems to indicate different metabolic pathways.

Ripening index seems to be also negatively correlated with phenolic content and most of the phenolic components. In fact, ripening index has been proposed as an indicator of fruit composition (Sánchez de Medina et al., 2014). However, the correlation coefficients obtained in this work were not very high and varied greatly across locations. Overall, low phenolic content was found in fruits with high ripening index (2.0 to 4.0). However, at lower ripening indexes (0.8 to 2.0), both fruits with high and low phenolic contents were found. These results could explain some contradictory results reported in previous studies. Indeed, Gomez-Rico et al. (2008) and Morelló et al. (2004) showed both high correlation and lack of correlation of ripening index with phenolic content and composition in the ‘Arbequina’ cultivar. Therefore, a general relationship between fruit color and phenolic content should be taken cautiously. A strong association of phenol content with harvest time but not with ripening index could be explained by the fact that there is not a consistent pattern of variation in the ripening index with harvest time in the different locations considered.

In summary, water availability and harvest date seem to play a more important role than air temperature on the phenolic content and composition of ‘Arbequina’ fruits, even when very different climatic conditions such as Sub-Tropical and Mediterranean are compared. Therefore, in order to obtain olive fruits with high phenolic content, low water availability together with an early harvest should be considered. Taking into account that those two factors would also reduce the oil content in fruit. Among the phenol components studied, comsegoloside seems to be the one with the greatest variance among locations. Therefore, future studies might consider this phenol as a good marker for plant stresses, especially water availability. Further research is needed to determine the relative influence of water availability and harvest date on other cultivars with different phenolic profiles than the ‘Arbequina’ one here considered.