Resistance of native oak to recurrent drought conditions simulating predicted climatic changes in the Mediterranean region

The capacity of a Quercus pubescens forest to resist recurrent drought was assessed on an in situ experimental platform through the measurement of a large set of traits (ecophysiological and metabolic) studied under natural drought (ND) and amplified drought (AD) induced by partial rain exclusion. This study was performed during the third and fourth years of AD, which correspond to conditions of moderate AD in 2014 and harsher AD in 2015, respectively. Although water potential (Ψ) and net photosynthesis (Pn) were noticeably reduced under AD in 2015 compared to ND, trees showed similar growth and no oxidative stress. The absence of oxidative damage could be due to a strong accumulation of α-tocopherol, suggesting that this compound is a major component of the Q. pubescens antioxidant system. Other antioxidants were rather stable under AD in 2014, but slight changes started to be observed in 2015 (carotenoids and isoprene) due to harsher conditions. Our results indicate that Q. pubescens could be able to cope with AD, for at least 4 years, likely due to its antioxidant system. However, growth decrease was observed during the fifth year (2016) of AD, suggesting that this resistance could be threatened over longer periods of recurrent drought.


Introduction
Mediterranean forests have high resilience to extreme drought episodes (Lloret et al. 2004;Saura-Mas et al. 2015). However, predictions based on recent scenarios in the Mediterranean area show a reduction in annual precipitation of up to 30% by the end of this century accompanied by longer periods of summer drought stress (i.e. summer drought intensification) compared to current conditions (Giorgi & Lionello 2008;IPCC 2013;Polade et al. 2014). Thus, climate change could consist of recurrent drought periods, extending over years, and the response of Mediterranean forests to such recurrent stress remains largely unknown (Niinemets 2010;Brzostek et al. 2014).
Water deficit eventually leads to stomatal closure reducing leaf gas exchange, especially carbon uptake. In this way, the photosynthetic processes could be disturbed and reactive oxygen species (ROS) produced due to limited ability to dissipate light energy and the use of the water-water cycle, hence increasing the oxidative pressure inside leaves (Chaves et al. 2009;Ehonen et al. 2017). ROS can lead to cell death through damage to proteins, lipids, carbohydrates and/or DNA (Halliwell 2006;Gill & Tuteja 2010).
Plants employ several mechanisms to cope with oxidative stress (Blokhina et al. 2003;Jaleel et al. 2009). Production of antioxidant-related defence metabolites prevails as a way to tolerate oxidative stress and, a fortiori, long-lasting and intense drought periods. A large variety of compounds produced by plants are well known for their antioxidant properties.
Besides these universal antioxidants, plants also produce other types of antioxidants, such as terpenes and phenolic compounds, whose diversity is non-universal or species-specific.
Isoprene, the main volatile terpene emitted by plants, seems to maintain the membrane stability of thylakoids preserving the photosynthetic machinery under oxidative conditions (Affek & Yakir 2002, Velikova et al. 2011. Moreover, phenolic compounds, especially phenolic acids, flavonoids and tannins, feature a strong capacity to quench ROS (Close & McArthur 2002). As a result, both terpenes (including isoprene) and phenolic compounds are expected to be increasingly produced in response to abiotic oxidative pressures, which is in accordance with the unified mechanism of action for volatile terpenes (Vickers et al. 2009) and the oxidative pressure theory for phenolic compounds (Close et al. 2004).
Several studies reviewed in Munné-Bosch (2005) provide evidence that plant protection against oxidative stresses is not due to a single compound, but to the whole bouquet of antioxidants. They also show that compensatory mechanisms can occur between antioxidants as well as between antioxidants and growth. Antioxidant synthesis requires carbon allocation, but this allocation could be limited under intense drought since photosynthetically assimilated carbon is limited (Chaves et al. 2003). This trade-off or compensation is, however, not necessarily observed since primary reserves of plant carbohydrates (starch for instance) can be mobilized to support metabolite production (Brilli et al. 2007;Chaves et al. 2009).
Numerous studies have already focused on gas exchange and/or growth and/or antioxidant concentration or emission under several drought treatments, mostly applied under controlled conditions, and more rarely under natural conditions (Llusià & Peñuelas 1998;Jiang & Zhang 2002;Hernández et al. 2004). However, there is still a lack of knowledge on how plants resist in situ amplified recurrent drought (several years), especially in the Mediterranean area where plants often encounter this type of stress.
Quercus pubescens Willd, our vegetal model, is one of the most widespread species in the Northern part of the Mediterranean basin. This species is a thermophilic and xerophilic species, highly resistant to drought conditions. This study aimed to evaluate if Q. pubescens forest can resist the recurrent and amplified drought expected with climate change in the Mediterranean region. We addressed the aim by measuring a vast array of ecophysiological (pre-dawn and midday water potential, stomatal conductance to water, net photosynthesis, chlorophyll content, chlorophyll fluorescence and lipid peroxidation as a direct proxy of cell oxidation) and metabolic (non-universal and universal antioxidants) traits. We hypothesized that (i) recurrent amplified drought (AD) could increase the oxidative pressure in leaves, leading to modifications of the antioxidant system, and (ii) carbon assimilation decrease with recurrent AD would involve trade-offs or compensations between antioxidants or between antioxidants and growth.

Experimental site
The study was performed at the Oak Observatory at the "Observatoire de Haute Provence (O 3 HP)", located 60 km North of Marseille (5°42'44'' E,43°55'54'' N), at an elevation of 650 m above sea level. The O 3 HP (955 m²), free from human disturbance for more than 70 years, consists of a homogeneous forest mainly composed of Q. pubescens, aged 70 years-old. This forest site features a supra-Mediterranean sub-humid bioclimate with an average annual precipitation of 830 mm, averaged from 1967 to 2000 (Santonja et al. 2015). Hence, this is considered a site with a Mediterranean climate, but with the peculiarity that the dry period naturally takes place for just one or two months per year. Using a rainfall exclusion device mechanically deployed to exclude some rain events (Fig. 1), it was possible to reduce natural annual rain by ~30% over part of the O 3 HP canopy, hereafter referred to as "the amplified drought plot" (AD). This method allowed us to extend the drought period, in an attempt to mimic the current climatic model projections for the end of this century (Giorgi & Lionello 2008;IPCC 2013;Polade et al. 2014). Two plots were present in the site: a plot (300 m 2 ) subjected to amplified drought (AD) and an adjacent plot (232 m 2 ), which received natural precipitation where trees were subjected to natural drought (ND). Amplified drought started in April 2012 and continued each year from then, principally during the growth period, with a 35% exclusion of precipitation in 2012 (from April 29 th to October 27 th ), 33% in 2013 (from July 7 th to December 29 th ), 35.5% in 2014 (from April 8 th to December 8 th ) and 33.9% in 2015 (from April 16 th to November 21 st ). Ombrothermic diagrams -built up using values of precipitation and temperature measured through rain gauges and temperature sensors placed in both plots -indicate that the drought period in the AD plot was extended for 2 months in  Fig. 1) and two additional trees per plot were included for antioxidant analyses (tocochromanols, chlorophylls and carotenoids). Trees were chosen randomly in the heart of each plot (to avoid the buffer zone of 2 m at the verges of the AD plot) while having canopies accessible from the footbridge.

CO 2 and H 2 O exchange
Two identical dynamic branch enclosures were used for sampling CO 2 and H 2 O exchange as well as isoprene emission (see below). The enclosures are fully described in Genard-Zielinski et al. (2015) and only slight modifications were made as described in Saunier et al. (2017).
Branches used for gas exchange sampling were located at the top of the canopy and exposed to the sun. Each branch was enclosed for 2 days in a ≈ 30 L PTFE (polytetrafluoroethylene) frame closed with a 50 µm thick PTFE film. Branch enclosures were installed on the day before the first measurement and were immediately flushed with a continuous flow of purified air at 9 L.min −1 (allowing complete air renewal every 3.33 min) using a PTFE pump (N840.1.2FT.18, KNF, Germany). A PTFE fan was used to ensure a rapid mixing of the chamber air and a slight positive pressure within the enclosure enabled it to be held away from the leaves to avoid biomass damage. Exchange of CO 2 and H 2 O of the enclosed branches were continuously measured (every minute) using infrared gas analyzers (IRGA 840A®, LI-COR, USA) at 0.2 L.min -1 . Net photosynthesis (Pn, µmolCO 2 m −2 s −1 ), stomatal conductance to water (Gw, mmolH 2 O m −2 s −1 ) and transpiration rate (E, mmolH 2 O m −2 s −1 ) were calculated using equations described by Von Caemmerer and Farquhar (1981) as presented in Genard-Zielinski et al. (2015). Transpiration rate data are not presented in this work, but were used to calculate the water use efficiency (WUE, mmolCO 2 .molH 2 O -1 ) as Pn/E. All gas exchange data were averaged between 12:00 and 15:00 (local time).
Relative humidity (RH) and temperature (T) were measured using an RHT probe (HMP60, Vaisala, Finland) placed inside the chamber and photosynthetically active radiation (PAR) was measured using a quantum sensor (PAR-SA 190®, LI-COR, USA) placed outside the chamber. Measurements with the RHT probe and the PAR sensor were recorded every minute using a data acquisition system (data logger, Li-1400-04®, LI-COR, USA). Once CO 2 and H 2 O exchange had been measured, leaves from enclosed branches were collected in order to calculate the surface area of each leaf with a leaf area meter (AM350, ADC Bioscientific Ltd., UK) and to report gas exchange of the whole branch leaf surface. The foliage used for the water potential and biochemical analyses was taken from branches adjacent to those enclosed and used for the gas exchange measurements.

Stem water potential
Stem water potential at predawn (Ψ pd ) and midday (Ψ m ) were measured from detached stems during each field campaign (except Ψ pd in spring 2014 and Ψ m in autumn 2014) with a Scholander pressure chamber (PMS instrument Co., USA, range 0 -7Mpa). Two measurements of Ψ were performed on each tree if the difference between both values did not exceed 0.3 MPa. Otherwise, a third measurement was performed.

Chlorophyll fluorescence
Chlorophyll fluorescence of photosystem II was measured through the maximum quantum efficiency of PS II (Fv/Fm, the ratio of variable to maximum fluorescence). Measurements were performed on the adaxial side of the leaves, with the leaves dark-adapted for at least 30 min, using a portable fluorimeter (FMS2, Hansatech, UK).

Chlorophyll pigment
For all biochemical analyses (i.e. including the metabolic traits studied, see below), leaf samples were frozen in liquid nitrogen and stored at -80°C before measurements. Chlorophyll pigments were extracted from leaves in methanol, quantified by HPLC using authentic standards of chlorophyll a and b (Sigma-Aldrich, USA) and expressed in mg.g DM -1 , as described by Havaux et al. (2007). The ratio between chlorophyll a and b (Chl a/Chl b) was also calculated as an indicator of PSII antenna size and PS stoichiometry.

Estimation of leaf oxidation through lipid peroxidation imaging
Lipid peroxidation was visualized in detached leaves by autoluminescence imaging to assess leaf oxidation due to water stress (Havaux et al. 2006). As detailed in Birtic et al. (2011), this technique is based on the measurement of the weak luminescence associated with the slow decomposition of lipid peroxides using a very sensitive photon-imaging system. The intensity of the luminescence signal is proportional to the amount of lipid peroxides present in leaves.
In order to validate the applicability of this method on Downy oak, detached leaves were exposed to a photooxidative stress treatment by combining high PAR (1500 µmol m -² s -1 ) and low temperature (14 °C). Spontaneous photon emission from detached Downy oak leaves was measured after 3 h of dark adaptation using a liquid N 2 cooled charge-coupled device (CCD) camera. Acquisition time was 20 min and pixel binning was 2 x 2. The luminescence signals were analyzed and quantified with Image J software.

Growth assessment
Growth was estimated in ND and AD plots through 3 parameters: i) litter production (g.m -2 ) as a direct proxy of leaf biomass production, ii) increase of tree trunk diameter at breast height (DBH, cm.yr -1 ) and iii) leaf mass per unit of leaf area (LMA, g.m -2 ) indicator of foliage thickness (Vile et al. 2005). Litter production was measured within 8 litter traps (0.24 m²) placed on the soil of both plots in 2014 and 2015. Litter traps allowed the collection of senescent leaves during leaf abscission from October to April and thus estimation of the leaf biomass production in each plot. The DBH increase was measured as the difference between DBH in 2016 (5 th year of AD) and DBH in 2015 (4 th year of AD), DBH being always measured at the end of the vegetative period (in December). Note that DBH is the only parameter monitored in 2016. These measurements were performed on all trees of each plot (n=146 in ND and n=52 in AD) with a meter (tape measure). Finally, the LMA of enclosed leaves was measured as the ratio between the dry leaf biomass and weight, expressed in g.m -2 . LMA is only presented for autumn 2014 and 2015 for trees followed for gas exchange. We assumed that it was a representative measurement for the whole year as Q. pubescens is a winter-deciduous species.

Metabolic traits
For all metabolic traits studied, leaf samples were collected and immediately frozen in liquid nitrogen and stored at -80°C in a laboratory freezer before extractions.
α-tocopherol standards were obtained from (Sigma Aldrich, USA) and the plastochromanol standard was a kind gift from Dr. J. Kruk (Jagiellonian University, Krakow, Poland).

Carotenoid pigments
Antioxidant carotenoid pigments from the xanthophyll cycle (violaxanthin+antheraxanthin+ zeaxanthin), as well as neoxanthin, lutein and β-carotene were extracted and analyzed using This article is protected by copyright. All rights reserved.
the same method used for chlorophylls (see above, Havaux et al. 2007). Their concentrations (ng.mg DM -1 ) were calculated from calibration curves performed using pigment standards (Extrasynthèse, France). where ERiso is expressed in µgC g DM -1 h −1 , Q 0 is the flow rate of the air introduced into the chamber (L h −1 ), C out and C in are the isoprene concentrations in the outflowing and inflowing air respectively (µgC L −1 ) and B is the total dry biomass matter (g DM ). ERiso are not available for autumn 2014 due to a technical problem. Investment of carbon sequestration into isoprene (C iso ) was calculated with the following formula:

Isoprene
where ERiso and Pn used in this equation were expressed in µgC m -2 h -1 .

Total phenolic compound index
Total phenolic compounds index, were extracted using 0.25 g DM of lyophilized ground leaves in 20 mL of methanol:water solvent (70:30). The mixture was shaken constantly for 1 h at ambient temperature and the extract was then filtered through a membrane with a 0.22 µm pore size. The total amount of water-soluble phenolic compounds was determined based on the Folin-Ciocalteu method used by Santonja et al. (2015). The quantitative values were expressed in mg equivalents of gallic acid per g of leaf dry mass.

Starch
Starch was extracted from 0.1 g DM of lyophilized ground leaves in 100 µL of ethanol/water solvent (80:20). After addition of 3 mL of amylase, each sample was incubated in boiling water for 6 min and 100 µL of amyloglucosidase was then added. The samples were mixed and incubated for 30 min at 50 °C and then transferred into graduated 25 mL vials. Volumes were adjusted up to 25 mL with distilled water. An aliquot was centrifuged at 3000 rpm for 10 min. Then, 50 µL of each sample was mixed with 2 mL of distilled water, 100 µL of buffer solution and 100 µL of NADP + solution. After 3 min, the extract was measured at 340 nm on a Biomate 3 spectrophotometer (ThermoFisher, USA). A calibration curve was made with a D-glucose solution provided with the analytical kit (Megazyme, USA).

Statistical analyses
Statistical analyses were performed with STATGRAPHICS® centurion XV (Statpoint Technologies, USA) and R (3.3.2). To evaluate differences between the factors "drought treatments" and "seasons" on every plant trait, we separated the data set in 2014 and 2015, and performed a two-way repeated measures ANOVA followed by Tukey post hoc tests after having checked the normality and homoscedasticity of the data set. Since significant interactions between drought and seasonality occurred, we performed a one-way repeated ANOVA followed by Tukey post hoc tests to evaluate the effect of season on ND and AD separately, and Student tests to evaluate differences between ND and AD during each season.
Student's tests were also performed to test for differences between both years. Student tests allowed the analysis of differences in growth between AD and ND. In order to check for the occurrence of compensatory mechanisms (increasing the production or emission of some antioxidants as others decrease), Spearman correlations were performed between antioxidants. Spearman correlations were also used to assess the relationship between antioxidants and physiological parameters (Pn and Ψ pd ). Correlations were always carried out for each season with years and treatments pooled together.
Principal component analyses (PCA) were performed on antioxidant data for each season, (previously centered and standardized) in order to highlight whether some compounds were produced more under AD in 2014 and 2015 (ade4 package in R) followed by two-way PERMANOVA to check for significance of the results (vegan package in R).

Water stress level indicators
Amplified drought had a strong effect on Ψ pd with decreases of 22 and 29% in summer and autumn of 2014, respectively, and 12 and 43% in summer and autumn of 2015, respectively (Table 1) Likewise, starch content did not change with drought treatment, irrespective of the year.
Autoluminescence imaging was used to detect lipid peroxidation in Q. pubescens leaves.
Under photooxidative stress conditions, a high leaf autoluminescence was found for all leaves. This indicates, as expected, strong lipid oxidation, and confirms the applicability of the method for this species (Fig. 4, HL). In both 2014 and 2015, leaf autoluminescence was very low throughout the year and did not increase with AD.

Growth
Neither leaf biomass production, nor LMA varied between ND and AD in any year tested (Table 2). However, trunk growth, estimated by the DBH increase between 2015 and 2016, was reduced by 43% under AD compared to ND.

Sensitivity to amplified drought and seasonal course
Leaf α-tocopherol concentration was not sensitive to AD, irrespective of the season or the year ( Fig. 5a and 5b). However, α-tocopherol was sensitive to seasonal changes with a Phenolic compounds index was not sensitive to AD either in 2014 or 2015 (Fig. 8).
Moreover, these compounds did not exhibit a seasonal course, irrespective of the year, with similar concentrations during both years.

Main antioxidants vary according to the season and the year
Antioxidant concentrations mostly remained similar under both drought treatments (Fig. 9a, b, c, PCA, followed by two ways PERMANOVA), except in summer when AD had a marginal effect. By contrast, antioxidant concentrations were strongly dependent on the year and season: i) in spring, β-carotene and, to a lesser extent, isoprene and α-tocopherol were produced in greater amounts in 2015 compared to 2014; ii) in summer, α-tocopherol, plastochromanol and carotenoids belonging to the xanthophyll cycle were produced in greater amounts in 2015. Isoprene and phenolic compounds did not have a high impact on the data repartition for the summer season; iii) in autumn, the xanthophyll cycle carotenoids, αtocopherol and isoprene were produced in greater amounts in 2015 than 2014. Similar to the data for the summer, data repartition in autumn was not highly affected by phenolic compounds.

Compensatory mechanisms
Spearman's correlations between antioxidants and physiological parameters (Pn and Ψ pd ) showed that, in spring, xanthophyll cycle antioxidants were positively correlated with Pn (Table 3 and Tables S1 a and b in supplementary files). In summer, the xanthophyll cycle antioxidants, as well as α-tocopherol, were strongly and positively correlated with the drop in water availability (indicated by Ψ pd ), involving a decrease in carbon assimilation. Moreover, the drop in water availability was also associated with a decrease in carotenoids (β-carotene and neoxanthin), as indicated by their negative correlation. In autumn, just like in summer, αtocopherol and xanthophyll cycle antioxidants were positively correlated with the decrease of Ψ pd , which was negatively correlated with neoxanthin.
In each season, the three carotenoids, lutein, β-carotene and neoxanthin, were positively correlated: i) in spring, only positive correlations were observed between antioxidants, with isoprene positively correlated with neoxanthin and β-carotene positively correlated with αtocopherol; ii) in summer, α-tocopherol was negatively correlated with carotenoids (lutein, βcarotene and neoxanthin) and positively correlated with the xanthophyll cycle.
Plastochromanol was also positively correlated with the xanthophyll cycle; iii) in autumn, αtocopherol was negatively correlated with neoxanthin and the latter was also negatively correlated with isoprene emissions and xanthophyll cycle pigments. Moreover, isoprene, xanthophyll cycle antioxidants and α-tocopherol were positively correlated with each other.
Negative correlations are probably indicators of trade-offs or compensations between antioxidants, whereas positive correlations could reflect synergistic actions of antioxidants.

Amplified drought affects gas exchange by Q. pubescens leaves.
Under mild water deficit in 2014 (Ψ pd close to -1.0 MPa in summer AD), Q. pubescens exhibited partial stomatal closure (50% lower under AD than ND in summer). Although this response resulted in a 40 % decrease of Pn, it had the advantage of avoiding Ψ m drop under AD (conservative strategy). This response reflects a drought-resistance strategy through drought-avoidance (Martínez-Ferri et al. 2000). However, it also suggests that Q. pubescens showed some degree of drought tolerance since, despite the substantial Pn decline, Pn values were still rather high (~9 µmolCO 2 m -2 s -1 in summer) (Damesin and Rambal 1995) during this relatively wet year. It should be noted that Q. pubescens exhibits other droughtavoidance traits, including a well-known high stem hydraulic efficiency, which is also partly responsible for leaf water maintenance (Nardini & Pitt 1999;Poyatos et al. 2008;Genard-Zielinski et al. 2014).
In 2015, when harsher drought conditions occurred (up to Ψ pd ~-3 MPa in summer under AD), Q. pubescens closed its stomata under AD during most of the year (except in spring), which implied a marked limitation of Pn. This strategy seemed, however, more efficient in summer than in autumn since stomatal closure under AD in summer prevented Ψ m from dropping compared to ND (drought-avoidance), while in autumn, stomatal closure implied limited Pn, but could not prevent Ψ m decline, showing the limits of Q. pubescens resistance to drought.
Our study indicates that Q. pubescens features high plasticity when faced with water scarcity, with a shift from combined drought-avoidance and -tolerance responses under mild water deficit (2014) to a drought-avoidance strategy under more severe dryness (2015). This shift between drought-resistance strategies with respect to gas exchange seems to be common in Mediterranean species (Chaves et al. 2009).
It is important to note that the net photosynthesis decrease observed in 2015 is likely to be due to the harsher climatic conditions of that year, but could also be due to the cumulative effect of recurrent amplified drought as expected with climate change (Brzostek et al. 2014).
These results suggest that, besides the drought severity of the year, drought recurrence is also a major factor to take into account when evaluating species resistance to expected climate change.

The antioxidant system preserves Q. pubescens leaves from drought-induced oxidative damage
The low levels of lipid peroxidation observed in this study (indicator of Q. pubescens resistance to water stress) implies that leaves were able to avoid cellular oxidative damage during drought stress, presumably due to several metabolic adjustments and compensations.
In 2015, AD brought about a decrease in lutein, neoxanthin and β-carotene compared to ND, which is concomitant with the Ψ pd drop. This decrease could be due to the oxidative pressure occurring within leaves resulting in carotenoid oxidation, as demonstrated by Ramel et al. (2012). It is possible that this phenomenon was not efficient enough to protect chloroplasts from oxidative damage since a compensatory mechanism between carotenoids and αtocopherol (negative correlation) was highlighted. This loss of carotenoids, together with an increase in chla/chlb, could also come from reduction of the PSII antenna in AD-treated leaves, since both lutein and neoxanthin are components of the light-harvesting antenna of the photosystems, with neoxanthin being specific to the PSII antenna (Croce et al. 1999). This phenomenon is a typical response of plants to excess light energy (Park II et al. 1997;Ballottari et al. 2007). However, the maximal quantum efficiency of PSII (Fv/Fm) was not affected by AD, indicating that there was no photoinhibition and that PSII was still functional despite the loss of pigments (Krause & Weis 1991). Thus, the carotenoid changes could actually reflect an acclimation to light and drought in 2015 and their role as antioxidants through oxidation reactions.
A striking observation was that lipid-soluble antioxidants (α-tocopherol, plastochromanol) displayed a very strong accumulation in Q. pubescens leaves throughout the year, both in 2014 and 2015. Compared to spring, α-tocopherol and plastochromanol contents were increased in autumn by a factor of around 8. Both tocopherol and plastochromanol are known to be located in the chloroplasts (Kruk et al. 2016) and are dependent on light, with an excess of light energy boosting their synthesis (Ksas et al. 2015). Although phenolic compounds are well known to have a role in photoprotection (Dixon & Paiva 1995) and are associated with drought resistance (Hernández et al. 2004;Akula & Ravishankar 2011), they did not respond to drought conditions in our study. Moreover, they did not show compensatory mechanisms or a synergic action with other antioxidants. There is a large variety of phenolic compounds in plants (e.g. flavonols, flavanols, anthocyanins) with many different roles and effects on oxidative stress (Close & McArthur 2002). For example, it has been demonstrated that the monomeric flavan-3-ols (e.g. catechins), a particular group of flavonoids, had an in vitro antioxidant activity five times higher than tocopherol or ascorbic acid (Blokhina et al. 2003). Thus, it is possible that specific groups or individual phenolic compounds rather than the total phenolic content respond to AD (Bernal et al. 2015), which could not be seen with our global analysis.

Growth of Q. pubescens could be affected by amplified drought in the long term
Trees seemed to invest similarly into growth under ND and AD until 2015, at least in terms of leaf biomass production. However, the trunk growth estimated by the increase in DBH started to significantly differ between 2015 and 2016 (between the 4 th and the 5 th year of AD), indicating that a relatively long period of water stress is needed to observe a significant growth decline in this species. Therefore, the effects of AD on oak forest growth could be amplified in the coming years. This result could be due to the limiting conditions in terms of precipitation in 2015 or to the recurrent drought (Smith et al. 2009). Moreover, Q. pubescens produced the same quantities of leaves under ND and AD irrespective of the year, which supports this hypothesis. Likewise, Limousin et al. (2012) showed that growth of Q. ilex, another major Mediterranean species, was impacted only after 7 years of precipitation restriction in the field.
As a general conclusion, the present work shows that the predicted decrease in rainfall of 30% by the end of the century associated with climatic change will not strongly affect Downy oak forest ecosystems in the Mediterranean area in the mid-term (after 4 years). Although Q.
pubescens is a highly drought-resistant species that possesses efficient antioxidant mechanisms (tocopherol, xanthophyll cycle) to cope with oxidative stress, we can anticipate substantial decreases in gas exchange, adjustments of the photosynthetic pigment composition (chlorophylls, carotenoids), impaired growth in the long-term and maybe stronger compensatory mechanisms between antioxidants.
Even though our study highlights the occurrence of a few metabolic and ecophysiological shifts with water stress recurrence and severity, these investigations could be expanded in further studies by integrating other compounds that are indicators of plant protection against water stress (e.g. enzymatic antioxidants, targeted phenolic compounds, sugars), and by considering longer time scales (probably up to 10 years). This would allow the evaluation of whether the observed shifts in this study become more marked over time and/or the more    Table 3: Spearman's coefficients of correlations between antioxidants [α-tocopherol, plastochromanol, lutein, β-carotene, neoxanthin and xanthophyll cycle (violaxanthin+zeaxanthin+antheraxanthin, called V+Z+A)], isoprene, phenolic compounds, and physiological parameters [predawn water potential in absolute value (Ψ pd ) and net photosynthesis (Pn)] in summer with data from the different years and drought treatments merged together. Coefficients are marked with bold font when correlations were significant at the at least 95% of confidence interval (n=20, *: 0.01 < P < 0.05, **: 0.001 < P < 0.01, ***: P < 0.001).