Inter-genotypic differences in drought tolerance of maritime pine are modified by elevated [CO₂]

Abstract

Background and Aims Despite the importance of growth [CO₂] and water availability for tree growth and survival, little information is available on how the interplay of these two factors can shape intraspecific patterns of functional variation in tree species, particularly for conifers. The main objective of the study was to test whether the range of realized drought tolerance within the species can be affected by elevated [CO₂].

Methods Intraspecific variability in leaf gas exchange, growth rate and other leaf functional traits were studied in clones of maritime pine. A factorial experiment including water availability, growth [CO₂] and four different genotypes was conducted in growth rooms. A ‘water deficit’ treatment was imposed by applying a cycle of progressive soil water depletion and recovery at two levels of growth [CO₂]: ‘ambient [CO₂]’ (aCO₂ 400 μmol mol⁻¹) and ‘elevated [CO₂]’ (eCO₂ 800 μmol mol⁻¹).

Key Results eCO2 had a neutral effect on the impact of drought on growth and leaf gas exchange of the most drought-sensitive genotypes while it aggravated the impact of drought on the most drought-tolerant genotypes at aCO2. Thus, eCO2 attenuated genotypic differences in drought tolerance as compared with those observed at aCO2. Genotypic variation at both levels of growth [CO2] was found in specific leaf area and leaf nitrogen content but not in other physiological leaf traits such as intrinsic water use efficiency and leaf osmotic potential. eCO2 increased Δ¹³C but had no significant effect on δ¹⁸O. This effect did not interact with the impact of drought, which increased δ¹⁸O and decreased Δ¹³C. Nevertheless, correlations between Δ¹³C and δ¹⁸O indicated the non-stomatal component of water use efficiency in this species can be particularly sensitive to drought.

Conclusions Evidence from this study suggests elevated [CO₂] can modify current ranges of drought tolerance within tree species.

INTRODUCTION

A fundamental question in plant science and global change research is to understand the interactive effects of multiple co-occurring environmental factors in plant functioning (Wullschleger et al., 2002). For example, the interactive effects of growth [CO₂] and water availability on tree performance and forest functioning have received considerable attention (e.g. Chaves, 1992; Beerling et al., 1996; Wullschleger et al., 2002) and they are still a topic of current interest and active ongoing research (e.g. Koutavas, 2013; Perry et al., 2013; Farrior et al., 2015). It has been a long-standing hypothesis that elevated [CO₂] should ameliorate the adverse impacts of droughts on tree performance (see Kelly et al., 2016) due to a positive fertilization and water-saving effect of elevated [CO₂] on trees. Yet, specific acclimation responses to long-term exposure to elevated [CO₂] such as down-regulation of photosynthesis (Ceulemans et al., 1997) could attenuate CO₂ fertilization and hence this potential ameliorating effect.

Previous studies of this subject have found inconsistent and sometimes conflicting results. For example, elevated [CO₂] can ameliorate (Wertin et al., 2010; Koutavas, 2013), exacerbate (Bobich et al., 2010) or have a neutral effect (Centritto et al., 1999; Perry et al., 2013) on the impact of drought on tree responses. These conflicting results might arise because the effect of one factor can constrain or alter the tree response to the other factor in different ways which can depend on the severity of the stress imposed (Roden and Ball, 1996; Duan et al., 2013, 2014) and be species-specific (Beerling et al., 1996; Leakey et al., 2009). It could also vary across genotypes within the same species, since intraspecific variation in the response to either water availability (e.g. Monclus et al., 2006; López et al., 2009; Sánchez-Gómez et al., 2011) or growth [CO₂] (e.g. Mycroft et al., 2009; Moran and Kubiske, 2013) has been found. In this respect, it could be hypothesized that drought tolerance could be differentially modulated by growth [CO₂] across genotypes of the same species. In other words, genotypic rankings in drought tolerance within a species could be changed by elevated [CO₂] and, consequently, shifts in these rankings could be expected under future environmental scenarios. Despite this, to our knowledge, only the study by Polley et al. (1999), among the few studies evaluating intraspecific variation in the response of trees to the combined effects of elevated [CO₂] and water deficit (Conroy et al., 1990; Johnsen, 1993; Polley et al., 1999; Wang et al., 2000; Lewis et al., 2013), has addressed this particular question. In that study, the impact of elevated [CO₂] on the survival response to increased water deficit was similar across families of the broadleaved Prosopis glandulosa, suggesting that ‘Genetic types with superior ability to survive drought today, however, apparently will maintain that advantage in the future’ (Polley et al., 1999: 365). This finding agrees with the idea that breeding for current environmental scenarios is likely to provide successful outcomes in the future (Ainsworth, 2016), since it implicitly assumes that elevated [CO₂] would not change current genotypic rankings in tree performance within a species. However, studies addressing this specific question in trees are lacking, in particular for conifers.

Maritime pine (Pinus pinaster) is a Mediterranean conifer of high ecological and socio-economic value. It is considered a drought-avoiding species due to its high stomatal sensitivity to soil water deficit (Granier and Loustau, 1994; Picon et al., 1996; Schwanz and Polle, 2001). Nevertheless, previous studies on this species have found intraspecific variation in several functional traits related to drought tolerance (Nguyen-Queyrens et al., 2002; Nguyen-Queyrens and Bouchet-Lannat, 2003; Corcuera et al., 2010; Aranda et al., 2010; Sánchez-Gómez et al., 2010; de Miguel et al., 2012). While important advances have been made in maritime pine to understand the role of water availability in shaping population structure and differentiation (e.g. Aranda et al., 2010; Sánchez-Gómez et al., 2010; Lamy et al., 2011) and the molecular basis of drought tolerance (e.g. Dubos et al., 2003; Eveno et al., 2008; de Miguel et al., 2014), the role of elevated [CO₂] as a potential modulator of the species response to water availability has remained largely overlooked (but see Guehl et al., 1994; Picon et al., 1996; Schwanz et al., 1996; Schwanz and Polle, 2001 for comparative studies with other tree species).

The present study was undertaken to investigate intraspecific variability in the functional response of maritime pine to different levels of growth [CO₂] and water availability. Leaf gas exchange, plant growth and other key leaf traits with potential adaptive value for drought tolerance [i.e. leaf-level water use efficiency (WUE), leaf osmotic potential, specific leaf area (SLA) and leaf nitrogen content] were evaluated. The main objective was to test the hypothesis that variability in drought tolerance within the species can be changed by elevated [CO₂]. Thus, it was explicitly tested whether the impact of elevated [CO₂] on maritime pine’s sensitivity to water deficit in terms of leaf gas exchange and growth is genotype-dependent. It was also tested whether intraspecific variability in other potential drought-adaptive leaf traits (WUE, leaf osmotic potential, SLA and leaf nitrogen content) were related to the observed patterns in leaf gas exchange and growth.

MATERIALS AND METHODS

Plant material and experimental setting

The genotypes of the study (for details see de Miguel et al., 2012) derived from an F₁ full-sib cross of maritime pine (Pinus pinaster Ait.) between a ‘drought-tolerant’ male parent from Oria (south-east Spain: 37°31′N, 2°21′W) and a ‘drought-intolerant’ female parent from Pontevedra (north-west Spain: 42°10′N, 8°30′W). Several ramets (clonal replicates of each clone-genotype) were obtained from coir-rooted cuttings propagated and grown as previously described by de Miguel et al. (2012). Two-year-old second-order ramets were transplanted into 6-L containers with a 3:1 (v/v) mixture of peat moss (Floratorf, 0–7 mm, Floragard Vertriebs GmbH, Oldenburg, Germany) and washed river sand, supplemented with 2 kg m⁻³ Osmocote Plus fertilizer (16-9-12 NPK+2 micronutrients, Scotts, Heerlen, Netherlands). A total of four genotypes with 12 healthy ramets per genotype were selected for the experiment among those displaying contrasting functional response to water deficit in a previous study (de Miguel et al., 2012) to ensure that the interactive effects of growth [CO₂] and water availability are tested at a wide range of drought tolerance within the species. After transplanting, the selected plant material was grown within a single growth (walk-in) room (Fitoclima 10000EHHF, Aralab Ltd, Sintra, Portugal) for 2 months as an establishment phase. The plants were rearranged weekly within the growth room at this stage. This growth room allows controlled levels of [CO₂], photosynthetic photon flux density (PPFD), temperature and humidity by a built-in feedback system including sensors, and microprocessor-based controllers. CO₂ was injected from industrial CO₂ cylinders when needed, to increase CO₂ levels within the cabin. The light was produced by a mixture of fluorescent tubes (Philips, Master TL-D, Super 80, 58W/840) and metal halide lamps (Osram, HQI^®-T, 250 W/D Pro+). The surface of the growth room walls has a reflective material to minimize light variability within the cabin. Humidity was controlled by a dual system. It combined a spray mist system injecting sprayed water into the incoming air coupled with an ultrasonic humidifier for fine adjustments (target RH ± 0·5 %). A computer program (FitoLog THC 849, Aralab Ltd) was used to monitor the environmental conditions within the cabin. Throughout the establishment phase, the plants were submitted to the following environmental conditions (mean±s.d.; registered every minute): 16/8 h photoperiod, 13 h of 799 ± 61 µmol photons m⁻² s⁻¹ PPFD at the top of the plants during the light period, 20·03 ± 0·12 °C night temperatures, 25·00 ± 0·09 °C day temperatures, 64·94 ± 0·81 % relative humidity (RH), 399 ± 52 μmol mol⁻¹ of growth [CO₂] and watered to field capacity when soil volumetric water content (VWC_s) dropped below 20 vol.%. In the following phase, the experimental layout was based on a factorial design with three factors: genotype (a total of four different genotypes), growth [CO₂] [two levels: ‘elevated CO₂’ (eCO₂) and ‘ambient CO₂’ (aCO₂)] and water availability [two levels: ‘well-watered conditions’ (WW) and ‘water deficit’ (WD)]. Half of the ramets (a minimum of six) of each genotype were randomly assigned to either eCO₂ or aCO₂. Two separate but contiguous growth rooms (Fitoclima 10000EHHF, Aralab Ltd) were used, one for each growth [CO₂] level. The rest of the environmental conditions were set up identically in both growth rooms until the end of the experiment, following the same settings as in the establishment phase. Both growth rooms were factory calibrated 1 month before the experiment, thereby ensuring their optimal performance. No significant differences (F=0·04, P=0·84 for temperature, F=0·63, P=0·43 for humidity; and F=0·28, P=0·60 for PPFD) were found between the environmental conditions (registered every minute) between growth rooms except for the growth [CO₂] levels. The plants were submitted, at this stage, to the following environmental conditions (mean±s.d.): 16/8 photoperiod, 13 h of 813 ± 23 µmol photons m⁻² s⁻¹ PPFD at the top of the plants, 20·01 ± 0·33 °C night temperatures, 24·99 ± 0·24 °C day temperatures, 64·33 ± 3·88 % RH. The actual [CO₂] in the cabins was 801 ± 23 μmol mol⁻¹ in the eCO₂-growth room and 398 ± 19 μmol mol⁻¹ in the aCO₂-growth room.

Water deficit was imposed for 2 months to half of the ramets of each clone within each growth [CO₂] level (a minimum of three ramets). The spatial location of the plants within each growth room was assigned by randomization of water availability × genotype treatments.

Watering treatment

VWC_s was individually monitored by soil moisture sensors (CS650-L, Campbell Scientific Inc., Logan, UT, USA) connected to dattaloggers (CR1000, Campbell Scientific). Well-watered plants (WW-plants) were kept at a VWC_s higher than 20 vol.% throughout the experiment while plants submitted to water deficit (WD-plants) were allowed to progressively deplete soil water content down to 5 vol.% from 15 March to 14 May. Intermediate targets (18, 15, 10, 8 vol.%) were established to homogenize the drought imposition rate among different ramets at the same pace whatever the size of plants. Every other day, VWC_s was recorded and the water needed for each plant to reach the target was calculated based on current VWC_s and individual evapotranspiration rates. The target 5 vol.% for WD-plants was reached after 19 d of water deficit was initiated and plants were kept at this stage for the rest of the water deficit period (43 additional days). A final recovery stage was established afterwards so WD-plants were rehydrated back to VWC_s above 20 vol.%. This stage took 2 weeks. The protocol allowed intensity, duration and effectiveness of water stress to be evenly applied regardless of plant size or genotype-specific water consumption rates.

Measurements and studied variables

Gas exchange and morphological measurements were carried out at three sampling times throughout the experiment (T₁, T₂ and T₃). T₁ corresponded to Julian days 71–74, just before the beginning of water deficit. T₂ corresponded to Julian days 127–130, where plants submitted to water deficit reached the peak of stress (minimum VWC_s). T₃ corresponded to Julian days 148–150, after 2 weeks of recovery from the water deficit stage.

A Li-Cor 6400 portable photosynthesis system (Li-Cor, Inc., Lincoln, NE, USA) with the 6400-07 needle chamber and the built-in Li-Cor 6400-01 CO₂ mixer to set the [CO₂] in the cuvette was used to measure gas exchange at the three sampling times. Measurements were carried out for every plant of the experiment at ambient light (811 ± 6 µmol m² s⁻¹ PPFD; mean±s.d.), leaf temperature at 24·05 ± 0·09 °C, RH at 63·97 ± 2·51 % and vapour pressure deficit at 1·22 ± 0·08 kPa and two measurement [CO₂] values (i.e. 400 and 800 µmol mol⁻¹) whichever was the growth [CO₂] of the plant measured. Leaf temperature and RH within the cuvette tend to vary according to sample differences in stomatal conductance. Thereby, slight readjustments were made on block temperature and RH (scrubbing or bypassing the air flow through the desiccant) to keep leaf temperature and RH close to 24 °C and 65 %, respectively. Measurement [CO₂] was first set to match the growth [CO₂] of the plant (400 or 800 μmol mol⁻¹) and afterwards changed to the alternative measurement [CO₂]. Records were taken after stable readings (approx. 2 min for the first measurement [CO₂] and 10 min for the second alternative measurement [CO₂]). The flow rate was set at 400 μmol s⁻¹. The most apical fully developed needles were chosen for gas exchange measurements. For each plant, four needles were aligned and placed across the 6400-07 needle cuvette. The cuvette enclosed the central portion of the needles. Gas exchange measurements began 2 h after the light-day period had started in the growth rooms and lasted for 4 h a day. Four consecutive days were needed to complete the measurements for each sampling time.

Net photosynthetic rates at the growth [CO₂] (A_gr μmol m⁻² s⁻¹), stomatal conductance at the growth [CO₂] (g_gr mol m⁻² s⁻¹) and intrinsic water use efficiency at the growth [CO₂] (iWUE=A_gr/g_gr μmol mol⁻¹) were considered for most of the analyses. Integrated A_gr and g_gr over the whole period of the experiment (A_w and g_w, respectively) were also calculated using a macro (‘area below curves’) implemented in SigmaPlot for Windows v.11·0 (Systat Software, Inc., Chicago, IL, USA).

For each growth [CO₂] level, net photosynthetic rates at measurement [CO₂] of 400 µmol mol⁻¹ (A₄₀₀) and 800 µmol mol⁻¹ (A₈₀₀) were considered to estimate the short-term photosynthetic acclimation response to [CO₂] for each individual (SR_800-400) which was computed as:

$${\mathrm{SR}}_{800-400}\mathrm{ }=\mathrm{ }(A_{800}-A_{400})/A_{400}$$ (1)

Just after gas exchange measurements, chlorophyll fluorescence measurements were taken on the same needles at ambient light conditions (a PPFD of 800 µmol m² s⁻¹). An FMS 2 Pulse Modulated Chlorophyll Fluorometer (Hansatech Instruments Ltd, King’s Lynn, UK) was used for these measurements. Effective quantum efficiency of photosystem II (PSII) (Φ_PSII) was calculated as:

$${\mathrm{\Phi }}_{\mathrm{P}\mathrm{S}\mathrm{I}\mathrm{I}}\mathrm{ }=\mathrm{ }\frac{F_{m\mathrm{\text{'}}}-F_s}{F_{m\mathrm{\text{'}}}}$$ (2)

where F_m′ is the maximum light-adapted fluorescence when a saturating light pulse (intensity ∼4000 μmol m⁻² s⁻¹, 0·8 s duration) is superimposed on the current ambient irradiance and F_s is ‘steady-state’ fluorescence of the light-adapted sample (Genty et al., 1989). Afterwards, in those needles, the height and width of the needles at their middle point were measured with a calliper (± 0·01 mm) and finally oven-dried (65 °C until constant weight) for dry mass determination. The leaf area enclosed in the cuvette (Ag, cm²) was estimated assuming a hemi-elliptical shape of the needle cross-section.

$$Ag\mathrm{ }=\mathrm{ }l\mathrm{ }SPe$$ (3)

where l is the length of the needle portion inside the gasket and SPe is the semi-perimeter of the elliptical shape defined by needle height (semi-minor axis) and needle half width (semi-major axis) at the middle of the needle. The perimeter of this ellipse (Pe=2SPe) was calculated using Ramanujan’s approximation. SLA (m² kg⁻¹) was calculated as: leaf area/leaf dry mass. Additional chlorophyll fluorescence measurements were taken on dark-adapted needles (pre-dawn measurements) close to those used for gas exchange (FMS 2 Pulse Modulated Chlorophyll Fluorometer, Hansatech Instruments). Maximum quantum efficiency of PSII (F_v/F_m) was calculated as an indicator of chronic photoinhibition (Maxwell and Johnson, 2000).

Dry needles used for gas exchange at the peak of the water deficit period (T₂) were ground in a ball mill. The leaf powder obtained was used to determine elemental nitrogen content that was expressed on a mass basis (N_m, cg g⁻¹) and ¹²C and ¹³C abundances at the Stable Isotope Facility of the UC Davis, California. Elemental and isotopic analyses were made with a PDZ Europa ANCA-GSL elemental analyser interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd, Crewe, UK). Leaf nitrogen content on an area basis (N_a, g m⁻²) was derived from N_m and SLA at T₂. Photosynthetic nitrogen-use efficiency (PNUE, μmol mol⁻¹ s⁻¹) was also calculated as A_gr/N_m at T₂. Carbon isotope composition (δ¹³C, ‰) was calculated as:

$${\mathrm{\delta }}^{13}\mathrm{C}\mathrm{ }=\mathrm{ }\left(\frac{Rs}{Rb}-1\right)\times 1000$$ (4)

where Rs and Rb refer to the ¹²C/¹³C isotope ratio in the sample and the Pee Dee belemnite standard respectively. This method had a precision of ± 0·2 ‰. Carbon isotope composition was converted to discrimination values (Δ¹³C, ‰) following Farquhar et al. (1982):

$${\Delta }^{13}\mathrm{C}\mathrm{ }=\mathrm{ }\frac{{\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{a}}-{\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{p}}}{1+\mathrm{ }{\mathrm{\delta }}^{13}{\mathrm{C}}_{\mathrm{p}}/1000}$$ (5)

where δ¹³C_a is carbon isotope composition of the air (source) and δ¹³C_p is carbon isotope composition of the plant (product). Determinations of δ¹³C were made also on five maize plants that were grown in each growth room during the experiment to calculate δ¹³C_a following the empirical formula obtained by Marino and McElroy (1991). Additionally, δ¹⁸O was estimated in sub-samples of the same plant material used for δ¹³C analysis. After pyrolysis in an elementary PyroCube (Elementar Analysensysteme GmbH, Hanau, Germany), the ¹⁸O/¹⁶O ratio was measured with a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon). Values of δ¹⁸O were expressed relative to the V-SMOW standard (Vienna Standard Mean Ocean Water) according to:

$${\mathrm{\delta }}^{18}\mathrm{O}\mathrm{ }=\mathrm{ }\left(\frac{R_{\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}}}{R_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{d}}}-1\right)$$ (6)

where R is the ratio of ¹⁸O/¹⁶O atoms of the sample and standard, respectively. One hour before the day-light period had started in the growth rooms, two-needle samples were taken from each plant for pre-dawn water potential measurements (Ψ_pd, MPa). C-52 psychrometric chambers connected to a Psypro water potential datalogger (Wescor Inc., Logan, UT, USA) were used for determination of Ψ_pd. The C-52 psychrometric chambers were previously calibrated with standard NaCl solutions of known molality. At T₂, two additional needle samples were sampled for leaf osmotic potential measurements (Ψ_π, MPa). These needles were cut into small pieces (3 mm) and frozen in liquid nitrogen. After thawing, and cellular content extraction at 4 °C with a refrigerated centrifuge, osmotic potentials were measured from discs of filter paper soaked with the extracted solution in C-52 psychrometric chambers and connected to a Psypro water potential datalogger (Wescor).

The length and diameters of the main stem (at the root collar and the tip) of each plant were measured throughout the experiment every 7–10 d until the peak of the water deficit. A ruler (± 1 mm) was used to measure lengths while a digital calliper (± 0·01 mm) was used to measure diameters. The volume of the main stem was used as a proxy of plant size and was calculated as the volume of a truncated cone:

$$V\mathrm{ }=\mathrm{ }\frac{1}{12}\pi h\left(D^2+d^2+Dd\right)$$ (7)

where V is the volume of the main stem, h is the height of the main stem, D is the diameter of the main stem at the root collar and d is the diameter of the main stem at the tip. Relative growth rate (RGR) in plant size (V) before the application of water deficit (RGR_b), over the water deficit period (RGR_wd) and over the whole period of the experiment (RGR_w) was obtained from the classical formula of Fisher (1921) as:

$$\mathrm{R}\mathrm{G}\mathrm{R}\mathrm{ }=\mathrm{ }\frac{\mathrm{L}\mathrm{n}\mathrm{ }{V}_{t_j}-\mathrm{L}\mathrm{n}\mathrm{ }{V_t}_i}{\left(t_j-t_i\right)}$$ (8)

where t_j and t_i are two different time points of the experiment that define the time period over which RGR is calculated. The difference between t_j and t_i is provided in days given that t_j > t_i.

Data analyses

Analysis of variance was performed to test for the main effect of the genotype (G), growth [CO₂] (C), water availability (W) and ‘time’ (T, repeated measure factor) on the studied functional traits. Additionally, a covariate (initial size, estimated as the volume of the main stem at the beginning of the experiment) and the interaction terms of different orders between factors (C×W, C×G, W×G and C×W×G) were included in the analyses. Shapiro-Wilk’s and Levene’s tests were previously used to test for normality and homogeneity of variances, respectively. Those variables that failed these tests were log-transformed to meet the assumptions of normality and homoscedasticity. The relationships between specific variables were tested through correlation and regression analysis. STATISTICA v.10 (Statsoft Inc., Tulsa, OK, USA) was used to perform the analyses.

RESULTS

Predawn water potential (Ψ_pd) differed significantly between water availability levels (F=26·59, P < 0·001). Mean values were −0·7 MPa for WW-plants and −1·2 MPa for WD-plants. Water deficit was evenly applied across clones and growth [CO₂] levels (effect of clone for Ψ_pd, F=1·29, P=0·292; effect of growth [CO₂] for Ψ_pd, F=0·72, P=0·545). Soil volumetric water content (VWC_s) reached 4·8 ± 0·2 % (mean±s.d.) in WD-plants at the peak of the water deficit period while it remained at 29 ± 2 % in WW-plants. The repeated measure factor (‘Time’) was significant for A_gr, g_gr, iWUE, SR_800-400 and SLA but not for the photochemical variables (i.e. Φ_PSII and F_v/F_m). The effect of the covariate ‘initial size’ was only significant for A_gr, SLA and PNUE (Table 1).

Table 1.

Summary of the analysis of variance performed to test for the effects of the covariate ‘Initial size’ (I) factors ‘Time’ (T, repeated-measure factor), growth [CO₂] (C), water availability (W), genotype (G) and their interactions on the studied variables

Variable	Factor/covariate
	I	T	C	W	G	C×W	C×G	W×G	C×W×G
	(1/29)	(2/58)	(1/29)	(1/29)	(3/29)	(1/29)	(3/29)	(3/29)	(3/29)
Agr	7·49***	8·74**	251·56***	37·60***	0·88	4·02	0·87	3·35*	0·05
ggr	3·64	10·28***	1·59	51·84***	3·66*	1·58	1·07	1·67	0·33
iWUE	3·19	3·74*	32·96***	30·53***	1·39	6·91*	0·38	0·11	0·40
SR800-400	2·21	7·58**	9·53**	46·84***	0·72	0·29	1·45	2·52	0·65
ΦPSII	0·68	1·42	5·54*	2·16	0·97	0·02	0·40	1·63	0·36
Fv/Fm	0·99	0·51	10·06**	0·77	0·60	0·45	0·47	1·05	0·12
SLA	23·23***	3·17*	25·25***	1·75	4·80**	0·01	1·85	0·30	2·29
RGRb	3·90	–	0·04	0·01	0·13	1·41	1·67	1·61	1·32
RGRwd	1·66	–	0·02	6·34*	1·42	0·47	1·48	2·79	0·54
Ψπ	0·11	–	1·01	54·99***	1·02	1·01	2·81	0·53	0·33
Nm	0·68	–	11·44**	1·44	6·57**	1·40	0·20	0·52	1·36
Na	3·42	–	8·62**	2·21	10·17***	0·63	0·33	0·47	1·97
PNUE	6·70*	–	31·86***	45·07***	6·99**	0·85	2·02	3·88**	1·19
Δ13C	3·02	–	11·86**	22·55***	0·12	0·367	0·22	1·50	1·19
δ18O	0·34	–	1·18	12·37**	2·38	3·61	0·61	0·25	0·54

Fisher’s F values and significance levels (indicated by asterisks) are shown. Degrees of freedom are provided in parentheses.

^*0·01 ≤ P < 0·05;

^**0·001 ≤ P < 0·01;

^***P ≤ 0·001.

Gas exchange and chlorophyll fluorescence

Growth [CO₂] had a significant effect on all gas exchange and chlorophyll fluorescence variables studied except for g_gr which was unaffected (Table 1). The treatment eCO₂ increased A_gr and iWUE (Fig. 1 and Supplementary Data Fig. S1) but caused a 16·1, 7·6 and 1·1 % decrease in SR_800-400, Φ_PSII and F_v/F_m, respectively. Overall, water availability had a significant effect on gas exchange but not on chlorophyll fluorescence (Table 1). A_gr and g_gr decreased under WD while SR_800-400 and iWUE increased under these experimental conditions (Figs 1 and S1). The genotypes studied significantly differed in g_gr (Table 1 and Fig. 1). As a general trend, genotypes 132 and 144 had the lowest g_gr values under WW while genotypes 4 and 147 had the highest values under these conditions (Fig. 1). The reverse pattern arose in WD-plants during the peak of water deficit at aCO₂. Yet, this last pattern can only be taken as a trend since non-significant differences among genotypes were found at the peak of water deficit for WD-plants. The capacity of plants submitted to WD to recover g_gr to control values depended on the genotype. Genotype 4 was unable to recover both g_gr and A_gr to control values at either aCO₂ or eCO₂ (Fig. 1). Genotype 132 recovered g_gr to control values at both growth [CO₂] levels. Genotypes 144 and 147 recovered well at aCO₂ but they did not recover g_gr to control values at eCO₂ (Fig. 1).

Fig. 1.

Net photosynthetic rates (A_gr) and stomatal conductance (g_gr) for each growth [CO₂] level (aCO₂, ambient [CO₂]; eCO₂, elevated [CO₂]) and watering level (WW, well-watered conditions; WD, water deficit) . Mean values and standard errors (error bars) are provided. The number of replicates for each treatment and genotype combination is three to four. *Significant differences (Fisher’s least significant difference post-hoc test) between watering levels within the same growth [CO₂] level; #significant differences between growth [CO₂] levels within the same watering level. The comparisons are made with respect to the following treatments: WW – eCO₂ and WD – aCO₂.

A significant interaction between water availability and genotype (W×G) was found for A_gr (Table 1). This interaction effect indicated significant variability in the sensitivity of this trait to water availability among genotypes. At the peak of the water deficit period, clones 4 and 147 were the most sensitive in response to water deficit for A_gr at aCO₂ as shown by the difference in the mean values between WW and WD levels. They also had the lowest A_gr rates at WD at the peak of the water deficit period, in particular at aCO₂. In contrast, clones 132 and 144 were less sensitive to water deficit (Fig. 1). At eCO₂, the differences among genotypes in sensitivity to water deficit became less obvious for A_gr, with genotypes 132 and 144 showing a sensitivity to water deficit similar to genotypes 147 and 4 (Fig. 1). During the recovery period WD-plants recovered A_gr rates to control values better at aCO₂ than at eCO₂ (Fig. 1). Significant effects of water availability (F_1,29=7·92; P=0·009) and W×G interaction (F_3,29=3·03; P=0. 045) were also observed for Φ_PSII at the peak of the water deficit period, without differences in the other two periods. Differences among genotypes in the sensitivity of Φ_PSII to water deficit followed the same pattern as that observed for A_gr, i.e. high sensitivity to water deficit of genotypes 4 and 147 and lower sensitivity of genotypes 132 and 144 at aCO₂ (Supplementary Data Fig. S2). As found for A_gr, eCO₂ seemed to increase the sensitivity of Φ_PSII to water deficit for the least sensitive genotypes (Fig. S2). At eCO₂ WD-plants had consistently lower Φ_PSII than WW-plants across genotypes at the peak of the water deficit period (Fig. S2), with non-significant differences among genotypes. F_v/F_m at the peak of the water deficit period followed a slightly different pattern to that observed for Φ_PSII. The interaction term W×G was not significant (F_3,29=1·03; P=0·39) for F_v/F_m. Yet, as found for Φ_PSII, F_v/F_m decreased consistently across genotypes with water deficit at eCO₂ at the peak of the water deficit period (Fig. S2). The photochemistry recovered to control values 2 weeks after the peak of the water deficit since no significant effect of water availability (F_1,29 < 0·32; P > 0·57 for both Φ_PSII and F_v/F_m) was observed after the recovery period.

The effect of water availability on iWUE depended on growth [CO₂] (significant C×W interaction, Table 1). In particular, differences between growth [CO₂] levels were higher under WD than under WW at both the peak of the water deficit period and the end of the recovery period (Supplementary Data Fig. S1). iWUE did not recover to WW values either at aCO₂ or at eCO₂ (Fig. S1). The impact of water deficit on SR_800-400 did not disappear completely after the recovery period, in particular at eCO₂ (Fig. S1).

Growth patterns

RGR_b [mean ± 95 % confidence interval (CI): 0·024 ± 0,002 cm³ cm⁻³ d⁻¹] did not differ among genotypes and treatments (Table 1). The effect of growth [CO₂] was not significant for RGR_wd either. In contrast, water availability had a significant effect on RGR_wd (Table 1). RGR_wd of WW-plants (mean ± 95 % CI: 0·023 ± 0·002 cm³ cm⁻³ d⁻¹) did not differ from RGR_b (overlapping 95 % CIs). The interaction term W×G was nearly significant for RGR_wd (F_3,29=2·79; P=0·058) suggesting that sensitivity of RGR_wd to water deficit was dependent on genotype. This trend of genotypic variation in RGR_wd in response to water availability (Supplementary Data Fig. S3) reflected the pattern observed independently for A_gr (Fig. 1). In particular, at aCO₂, water deficit reduced RGR_wd of genotypes 4 and 147 (20·41 and 26·93 %, respectively) while no reduction was found in genotype 132 and only a 0·5 % reduction occurred in genotype 144 (see also Fig. S4). At eCO₂, genotypes 132 and 144 increased their sensitivity of RGR_wd to water deficit, with all the studied genotypes having similar sensitivity to water deficit at this growth [CO₂] level (Fig. S3). Although the genotype effect was not significant, genotypes 4 and 147 displayed a trend of higher absolute RGR_w values than genotypes 132 and 144 under WW (Fig. S3).

Additional leaf functional traits

The effect of the CO₂ treatment was significant for SLA, N_m, N_a, PNUE and Δ¹³C (Table 1). The eCO₂ treatment decreased SLA. Such an impact depended on the measurement time (significant T×C). Thus, growth [CO₂] was not significant at T₁ and T₂, but it was highly significant at T₃ (F_1,29=42·65; P < 0·001) when it accounted for a 17·2 % decrease in SLA (see mean values in Table 2). The eCO₂ treatment also decreased N_m (24·7 %) and N_a (20·2 %), but increased PNUE (46·5 %) and Δ¹³C (5·4 %, see mean values in Table 2). Water deficit significantly decreased Ψ_π, PNUE and Δ¹³C, while it increased δ¹⁸O (Tables 1 and 2). Significant genotypic differences were found for SLA, N_m, N_a and PNUE. Genotype 132 had the highest SLA while genotype 144 had the lowest SLA which did not significantly differ from that of genotype 4 (Table 2). Genotype 132 had also the highest leaf nitrogen content (both N_m and N_a), which differed significantly from that of genotypes 4 and 147. The latter was the genotype with the lowest leaf nitrogen content (Table 2). In general, genotypes 147 and 4 had higher PNUE than genotypes 132 and 144, in particular at eCO₂ and WW. This was not the case at aCO₂ and WD where non-significant differences among genotypes were found (Tables 1 and 2). The interaction W×G displayed by A_gr was also reflected in PNUE (Table 1).

Table 2.

Least square means for the studied traits

Variable	CO2 level	Watering level	Genotype
			4		132		144		147
SLA (m2 kg−1)	aCO2	WW	6·54	ab #	7·15	a #	6·38	bc #	7·62	a #
		WD	6·90	a	6·92	a	5·64	b	6·57	b
	eCO2	WW	5·40	a	5·59	a	5·37	a	5·52	a #
		WD	5·09	b #	5·68	b #	5·61	b	6·23	a
Ψπ (MPa)	aCO2	WW	−1·65	a	−1·65	a *	−1·57	a *	−1·60	a *
		WD	−1·84	a	−1·87	a	−1·90	a	−1·84	a
	eCO2	WW	−1·61	a	−1·49	a	−1·69	a	−1·67	a
		WD	−1·85	b *	−1·80	b *	−2·01	ab *	−2·11	a #*
Nm (cg g−1)	aCO2	WW	1·52	ab	1·99	a	1·19	b *	1·01	b
		WD	1·94	a	1·80	a	1·81	a	1·24	b
	eCO2	WW	1·41	ab	1·56	a	1·08	bc	0·66	c
		WD	1·05	bc #	1·71	a	1·11	ab	0·84	c
Na (g m−2)	aCO2	WW	2·10	ab	2·82	a	1·90	ab	1·35	b
		WD	2·55	ab	2·51	ab	2·96	a	1·52	c
	eCO2	WW	1·87	a	2·22	a	1·88	a	0·89	b
		WD	1·59	bc #	2·61	a	1·82	ab #	1·25	c
PNUE (μmol mol−1 s−1)	aCO2	WW	46·12	ab #*	29·37	b #	49·92	ab *	72·63	a #*
		WD	13·41	a	20·81	a	20·01	a	23·22	a
	eCO2	WW	83·49	b	60·39	bc	58·45	c	134·91	a
		WD	55·84	a #	27·47	b *	42·64	ab	51·74	ab *
Δ13C (‰)	aCO2	WW	22·21	a	22·85	a *	22·28	a	22·74	a
		WD	21·05	a	20·69	a	21·02	a	21·43	a
	eCO2	WW	24·02	a	24·24	a	23·13	a	24·57	a
		WD	21·59	ab *	21·85	ab *	23·54	a	21·34	b *
δ18O (‰)	aCO2	WW	23·88	a	24·44	a	24·48	a	23·82	a
		WD	23·39	b	25·50	a	24·83	ab	25·01	ab
	eCO2	WW	22·55	a	23·32	a	23·84	a	22·89	a
		WD	24·47	a *	24·72	a	25·89	a *	24·74	a *

The values are computed for the covariate (‘initial size’) at its mean. The letter codes denote homogeneous groups (Fisher’s least significant difference post-hoc test) among genotypes of the same treatment. ^#Significant differences between growth [CO₂] levels in the same watering level (WW, well-watered conditions; WD, water deficit). *Significant differences between watering levels in the same growth [CO₂] level (aCO₂, ambient [CO₂]; eCO₂, elevated [CO₂]).

Relationships among traits

iWUE was also negatively related to RGR_wd and Δ¹³C and positively to δ¹⁸O for each treatment combination (data not shown). A significant negative correlation between Δ¹³C and δ¹⁸O was found at both eCO₂ (P=0·008, R²=0·31) and aCO₂ (P=0·011, R²=0·28) for pooled water availability levels. In contrast, the relationship between Δ¹³C and δ¹⁸O differed according to water availability. A significant negative relationship was found at WW while no relationship was observed at WD for pooled growth [CO₂] levels (Fig. 2). Δ¹³C was not correlated with g_w at either WW or WD for pooled growth [CO₂] levels. By contrast, δ¹⁸O displayed a significant correlation with g_w at WW for pooled growth [CO₂] levels, while this relationship was not significant at WD (Fig. 3). The residuals of the linear regression analysis between Δ¹³C and δ¹⁸O were plotted against A_w, resulting in a positive correlation that explained 19 % of the variation at WD but no significant correlation was found at WW for pooled growth [CO₂] levels (Fig. 2).

Fig. 2.

Scatterplot of Δ¹³C versus δ¹⁸O and residuals of the regression analysis between δ¹⁸O and Δ¹³C versus integrated net photosynthetic rate over the whole period of the experiment (A_w). Solid symbols represent well-watered conditions (WW), while open symbols represent the water deficit level (WD). Triangles represent eCO₂ (elevated [CO₂]) while circles represent aCO₂ (ambient [CO₂]) Regression lines are shown only when significant and they are fitted for either WD (dashed line) or WW (solid line) with pooled data from both growth [CO₂] levels.

Fig. 3.

Relationships of Δ¹³C and δ¹⁸O with integrated stomatal conductance over the whole period of the experiment (g_w). Solid symbols represent well-watered conditions (WW), while open symbols represent the water deficit level (WD). Triangles represent eCO₂ (elevated [CO₂]) while circles represent aCO₂ (ambient [CO₂]). Regression lines are shown only when significant and they are fitted for either WD (dashed line) or WW (solid line) with pooled data from both growth [CO₂] levels.

Over the whole experiment, the higher photosynthetic rates reached at eCO₂ did not lead to a subsequent increase in RGR in comparison with plants grown at aCO₂. Moreover, A_w was positively correlated with RGR_w at aCO₂ while this correlation was not significant at eCO₂ (Supplementary Data Fig. S4).

DISCUSSION

Interplay of growth [CO₂] and water availability on photosynthetic rates and growth

The combined effects of elevated [CO₂] and water deficit on leaf gas exchange, growth rate and other key functional traits were evaluated in selected clones of maritime pine. The observed variation in photosynthetic and growth patterns among genotypes at aCO₂ confirmed the expected contrasting drought tolerance of the selected genotypes and agrees with previous findings reporting high inter-genotypic variability in maritime pine’s photosynthetic rates (de Miguel et al., 2012), growth patterns (de La Mata et al., 2014) and survival (Gaspar et al., 2013; Ramírez-Valiente and Robledo-Arnuncio, 2014) in response to water deficit at current atmospheric [CO₂], However, at eCO₂ we observed that drought tolerance (estimated as the variation in photosynthetic and growth patterns in response to water deficit) changed differently across genotypes, leading to an attenuation of the genotypic differences in drought tolerance as compared with those observed at aCO₂. In other words, genotypic differences in drought tolerance at aCO₂ disappeared at eCO₂. This finding suggests strongly that the range of realized drought tolerance within the species can be narrowed under future climatic conditions, implying also that genotypic rankings of drought tolerance within the species are probably not static and can be modified by differential interactive effects of growth [CO₂] and water availability across genotypes. The potential of elevated [CO₂] to change the genetic composition of plant populations has been previously documented (Andalo et al., 2001; Kubiske et al., 2007), which has been attributed to its fertilization effect, exacerbating differential growth responses among genotypes. The results of this study add further insight into a new potential mechanism for elevated [CO₂] to change the genetic composition of populations. This mechanism is the ability of elevated [CO₂] to modulate inter-genotypic differences in the plant response to other selective forces such as water availability. Although plant responses to elevated [CO₂] are generally interpreted in the context of ameliorating the negative impacts of drought (Wullschleger et al., 2002), we did not find such an ameliorating effect but rather a neutral (Centritto et al., 1999; Perry et al., 2013) or aggravating effect (Bobich et al., 2010) depending on the genotypes. Moreover, studies carried out on juvenile trees in growth rooms could overestimate the fertilization effect of elevated [CO₂] (Dawes et al., 2011), and thus the actual aggravating effect of elevated [CO₂] on the response of maritime pine to water deficit could be even more severe. The interaction between elevated [CO₂] and water deficit on tree performance is species-dependent (Leuzinger and Körner, 2007; Dawes et al., 2011), but regarding conifers or pines in particular, evidence is accumulating from free-air CO₂ enrichment (e.g. Ellsworth, 1999; Moore et al., 2006; Domec et al., 2017) and dendroecological studies (e.g. Linares et al., 2009; Hereş et al., 2014; Olano et al., 2014) that elevated [CO₂] does not generally mitigate the impact of drought in this group of trees, in agreement with our study.

Although elevated [CO₂] increased photosynthetic rates, it had no effect on above-ground growth rates throughout the experiment, even under well-watered conditions, suggesting that the stimulating effect of elevated [CO₂] on growth of maritime pine might be rather limited as found previously in this (Guehl et al., 1994) and other pine species (e.g. Dawes et al., 2011). In addition, the relationship between photosynthetic rates and growth was attenuated at eCO₂, suggesting changes in carbon source–sink activity and changes in the balance of carbon allocation (White et al., 2015). This involves mechanisms not explored here (dark respiration, root growth and respiration, changes in biomass partitioning and secondary metabolism, increase of root exudates, etc.), which need further research.

The observed effects of elevated [CO₂] in this study (i.e. low fertilization effect, neutral to aggravating effect on the functional response to water deficit and its potential to narrow the range of drought tolerance within the species) raise concern about the future adaptive response of maritime pine populations. While many studies support the existence of genotypic variation in maritime pine’s drought tolerance through the study of different functional traits at current atmospheric [CO₂] (e.g. Fernández et al., 2000; Nguyen-Queyrens et al., 2002; Corcuera et al., 2010; Aranda et al., 2010; de Miguel et al., 2012; Gaspar et al., 2013; de La Mata et al., 2014; Ramírez-Valiente and Robledo-Arnuncio, 2014), evidence from this study suggests that elevated [CO₂] can modify the range of drought tolerance of maritime pine and hence its adaptive potential to increased aridity. These results have important implications for forecasting the response of natural and/or artificial populations of maritime pine to climate change. Further studies into population genetics and genomics of maritime pine should devote particular attention to growth [CO₂] as a factor that can modulate the species response to water availability and constraint the range of drought tolerance of the species in the future.

Variation of functional traits and their relationship to drought tolerance and the impact of elevated [CO₂]

In contrast to previous studies in maritime pine, leaf osmotic potential (Fernández et al., 1999; Nguyen-Queyrens et al., 2002; Nguyen-Queyrens and Bouchet-Lannat, 2003) and WUE (Aranda et al., 2010; de Miguel et al., 2012; Marguerit et al., 2014) did not differ across genotypes, denoting that significant intraspecific variation in drought sensitivity of photosynthetic and growth patterns can be displayed regardless of genotypic variation in WUE or leaf osmotic potential. WUE is often increased at elevated [CO₂] primarily by reduced stomatal conductance, enhanced photosynthesis or both (Tschaplinski et al., 1995). While the reduction of stomatal conductance by elevated [CO₂] is a general response, it is not universal (Xu et al., 2016). Besides, recent studies report that the contribution of CO₂-mediated downregulation of stomatal conductance to increase WUE can be minor under water deficit conditions (Perry et al., 2013; Duan et al., 2014). In this study, the higher WUE observed at eCO₂ was only the result of increased photosynthetic rates since stomatal conductance was unresponsive to growth [CO₂] at either watering treatment, indicating a species-specific low sensitivity of stomatal conductance to growth [CO₂], as found previously (Guehl et al., 1994; Picon et al., 1996). Thus, the potential effect of elevated [CO₂] to improve the water economy at the leaf level and further increase WUE through reduced stomatal conductance does not seem to operate in this species, at least at the early ontogenetic stages. According to Scheidegger et al. (2000), the relationship between Δ¹³C and δ¹⁸O found in our study denotes the variation in WUE was strongly coupled to stomatal conductance at well-watered conditions but not so at water deficit, where the absence of any relationship indicates that both photosynthetic capacity and stomatal conductance could be simultaneously affected. The increase in the relative importance of photosynthetic capacity in the variation of WUE at water deficit was also confirmed by the fact that photosynthetic rates could explain part of the mismatch (19 %) between patterns of Δ¹³C and δ¹⁸O at water deficit but not at well-watered conditions. These results suggest that the non-stomatal photosynthetic component of WUE is particularly sensitive to drought in the species.

Contrary to the observed patterns for WUE and leaf osmotic potential, intraspecific variability in other key functional traits (e.g. SLA and leaf nitrogen content) was found in this study. Genotypic variation in SLA and leaf nitrogen content can be interpreted in terms of functional advantage under drought. For example, low SLA can provide leaves with improved structural characteristics against desiccation (Ackerly et al., 2002) and high leaf nitrogen content could increase photosynthetic rates at low stomatal conductance occurring under water deficits (Wright et al., 2003). Yet the results of this study do not agree with these expectations. We did not find any significant relationship between estimated drought tolerance and either SLA or leaf nitrogen content. Previous studies on intraspecific variability in broadleaved species support the view that low SLA (Ramírez-Valiente et al., 2010) or high nitrogen content (Sánchez-Gómez et al., 2013) are related to regional variation associated with water availability gradients. Nevertheless, broadleaves differ in form, longevity and phylogeny from conifer needles, which could explain the mismatch among these studies and our findings.

Limitations and further research

The results found in this controlled experiment with juvenile trees should be extrapolated with caution to mature individuals in the field since the observed inter-genotypic variability might change over ontogeny. However, it should also be highlighted that selective pressure is highest at early ontogenetic stages (Reich et al., 2003) due to small size and high vulnerability to environmental stresses as juvenile trees. Thus, tree responses to limiting or stressful conditions at these early stages could weigh heavily in shaping the structure and dynamics of forests in a context of climate change. Another point to be considered is that environmental conditions although homogeneous and precisely controlled in growth room experiments (desirable for accurately testing specific hypothesis on the plant functional responses to environmental gradients or testing relationships among functional traits) do not mimic the variability and heterogeneity of the natural environment. For example, pot size could constrain root growth and change biomass allocation patterns. In this respect, we observed that roots did not fulfil the pot volume at the end of the experiment so we can assume root growth limitation was unlikely in this study. The use of artificial soil media in pots could also lead to patterns of water deficit different from those actually experienced in the field and to nutrient limitation. On the one hand, we think nutrient limitation in the experiment was unlikely since we used a slow-release fertilizer (6-month longevity) at the manufacturer’s recommended concentration and visual symptoms of nutrient deficiency were not evident. On the other, the water deficit imposed was moderate, which could be rather mild in comparison with those in the field. Severe droughts could lead not only to reductions in photosynthesis and growth but also to mortality events. In this case, hydraulic traits (not studied here) could play an important role. A recent review notes that the effect of elevated [CO₂] on these traits might increase the vulnerability to drought of isohydric pines (Domec et al., 2017), which agrees with the results of this study.

Although our findings provide some clues and open new research avenues to advance our understanding of the interactive effects of elevated [CO₂] and water deficit on functional tree responses and its variability at the intraspecific level, we emphasize that this study used juveniles of four selected genotypes under controlled experimental conditions and was not specifically designed to provide wide generalizations on these topics but rather to test a very specific hypothesis. As far as we know, this is the first research study reporting that elevated [CO₂] can change inter-genotypic differences in drought tolerance of a tree species. Complementary studies including more genetic variability, different ontogenetic stages and performed in the field will be valuable to test to what extent the whole range of variation in maritime pine’s drought tolerance could be constrained by future levels of atmospheric [CO₂].

CONCLUSIONS

Evidence from this study indicates that elevated [CO₂] not only can aggravate the effects of water deficit on photosynthetic performance and growth of maritime pine but also it can attenuate genotypic differences in drought tolerance as compared with those observed at current atmospheric [CO₂]. This suggests elevated [CO₂] can constrain the range of drought tolerance of the species and hence its adaptive potential to increased aridity. The observed genotype-dependent modulatory effect of elevated [CO₂] on growth and photosynthetic rates in response to water deficit indicates that genotypic rankings of drought tolerance within the species are probably not static and can be modified by differential interactive effects of growth [CO₂] and water availability across genotypes. Finally, the findings of this study support the view that variability in maritime pine’s drought tolerance can be independent from variation in presupposed drought-adaptive traits at the leaf level, such as WUE, osmotic potential, SLA or leaf nitrogen content, at least at early ontogenetic stages.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and consist of the following. Figure S1: Intrinsic water use efficiency (iWUE) and short-term photosynthetic response to [CO₂] (SR_800-400) at the peak of the water deficit period and at the end of the recovery period for each treatment combination. Figure S2: Effective quantum efficiency of PSII (Φ_PSII) and maximum quantum efficiency of PSII (F_v/F_m) at the peak of the water deficit period. Figure S3: Relative growth rate in plant size for the whole water deficit period (RGR_wd) and over the whole period (RGR_w) for each treatment combination. Figure S4: Relative growth rate over the whole period of the experiment (RGR_w) versus net photosynthetic rate over the whole period of the experiment (A_w).