European beech (Fagus sylvatica L.) saplings 10–20 cm tall and 3–5 years old were collected from 12 beech populations (Arend et al., 2016a) and transplanted to the 16 lysimeter plots of the outdoor model ecosystems at the Swiss Federal Research Institute WSL (47°21′ N, 8°27′ E, 545 m a.s.l.). In each plot, two saplings from each population were grown with a randomized distribution. The plots are 3 m² in area, have a depth of 150 cm and are filled with natural forest soil (acidic Haplic Alisol; loamy sand; pH 4.6). The water regime of each plot is controlled by sliding glass roofs, which close automatically at the onset of rain fall, and an automated irrigation system. During the growing season, the plots were irrigated every second or third day with 50 l m^-2 deionized water enriched with nutrients to simulate the average composition of ambient rainfall (Kuster et al., 2013). After 2 years, when the saplings had reached a height of approximately 150 cm, a summer drought was imposed in 8 of the 16 plots by omitting the irrigation from June to mid-August 2013, while the other eight plots were regularly irrigated. Leaf development was complete when the drought treatment started. After 10 weeks, the plots were intensely re-watered and afterward regularly irrigated. For this particular study, six control and six drought-treated saplings from a mesic beech population (Collombey, Switzerland; 46°16′ N 6°56′ E; annual precipitation 1,055 mm; annual temperature 8.9°C), each growing in a separate plot, were selected.

Soil water content (SWC) was measured volumetrically at 30 cm soil depth using PC-controlled soil moisture probes (Decagon 5TM; Decagon, United States) installed in each plot of the model ecosystem. Pre-dawn leaf water potentials (ψ_pd) were measured (bi)weekly using a Scholander pressure chamber (M 600; Mosler Tech Support, Berlin, Germany) in two leaves collected before sunrise from the outer part of the canopy of each sapling. To minimize the impact of the frequent leaf harvest on the saplings, the same leaf material was used for the extraction of NSC. The leaves were briefly microwaved to stop any metabolic activity and then oven-dried at 60°C for 48 h.

On every sampling date, leaf gas exchange characteristics, i.e., net-photosynthesis (A_N), intercellular CO₂ concentration (c_i) and stomatal conductance (g_S), of two to four sun-exposed leaves per sapling were measured between 11h00 and 16h00. Measurements were performed with a portable photosynthesis system using a broadleaf cuvette (LI-COR 6400; LI-COR, Lincoln, NE, United States). Conditions in the cuvette were controlled during the measurements to maintain a CO₂ concentration of 400 ppm and a photon flux density of 1,000 μmol m^-2 s^-1, while temperature was adjusted to track values outside the cuvette, which ranged from 11.2 to 36.4°C during the measurements. Fast fluorescence kinetics were analyzed pre-dawn using a plant efficiency analyzer (Pocket PEA, Hansatech Instruments, Ltd., Norfolk, United Kingdom). These fluorescence measurements were conducted on each sampling date prior to the analysis of ψ_pd in the dark-adapted state. After a saturating light pulse of 3,500 μmol quanta m^-2 sec^-1 of red light (650 nm) was applied, the increase in fluorescence was recorded at a high resolution for 1 s. Maximum quantum efficiency of photosystem II (F_v/F_m) and the total performance index (PI_tot) were calculated using PEA plus 1.10 (Hansatech Instruments, Ltd., Norfolk, United Kingdom).

Non-structural carbohydrates (starch, glucose, fructose, and sucrose) were extracted according to Critchley et al. (2001). Approximately, 100 mg of oven-dried and powdered leaf material was incubated for 15 min with 1.12 M perchloric acid and then centrifuged at 3000 g for 15 min. The supernatant used for quantification of soluble sugars was adjusted to pH 6 by the addition of 2 M KOH, 0.4 M MES, and 4 M KCl. The precipitated potassium perchlorate was removed by centrifugation at 3000 g for 15 min. Sucrose in the extract was broken down to glucose and fructose by invertase (Roche, Rotkreuz, Switzerland). Free glucose and glucose originating from sucrose were then converted to gluconate-6-phosphate by glucose-6-phosphate dehydrogenase (Roche, Rotkreuz, Switzerland) and determined photometrically with a 96-well microplate reader (ELx800, BioTek, Luzern, Switzerland). Afterward, fructose was converted to glucose by phosphogluco-isomerase (Roche, Rotkreuz, Switzerland) and then measured as described above. For starch quantification, the remaining pellet was thoroughly washed with 80% EtOH, dried at room temperature and re-suspended in water before starch was broken down to glucose monomers via amyloglucosidase and α-amylase for 2 h at 37°C (both Roche, Rotkreuz, Switzerland) and then determined photometrically as described above. Photometric quantification was performed according to Hoch et al. (2003). The NSC concentrations are expressed on a dry matter basis (mg g^-1).

Soluble sugars for δ¹³C analysis were extracted and prepared according to the method established by Wanek et al. (2001) and modified by Göttlicher et al. (2006) and Richter et al. (2009). Briefly, 100 mg of powdered leaf material was extracted with 1 mL MCW (methanol, chloroform, water, 12:3:5, v/v/v) for 30 min at 70°C. After cooling to room temperature, the sample was centrifuged at 10,000 g for 2 min, and 800 μL of the supernatant was used for the extraction of soluble sugars. Phase separation of the supernatant was induced by adding 800 μL water and 250 μL MCW and then vigorously mixing. After every six samples, one blank was processed (800 μL MCW). After centrifugation at 10,000 g for 2 min, 1.2 mL of the upper aqueous phase was mixed with 500 μL chloroform and centrifuged for phase separation. Then, 1 mL of the upper phase was oven-dried for 24 h at 60°C. The sample was re-dissolved in 1 mL water and separated using an ion-exchange cartridge (5 mL syringes, BD Plastipak^TM, Beckon Dickinson S.A., Madrid, Spain) made of 2.2 mL cation-exchange resin (DOWEX 50W X 8, 50–100 mesh, H⁺-form) above 3.2 mL anion-exchange resin (DOWEX 1 X 8, 50–100 mesh, HCOO-form), separated by filter paper. After rinsing the columns with 30 mL water, 35 mL of the eluate, mainly consisting of soluble sugars, was collected. The samples were lyophilized and re-dissolved in 1 mL water. Volumes of 150 μL of each sample were pipetted into tin capsules and oven-dried for 48 h at 60°C before isotopic analysis was conducted.

For the analysis of δ¹³C in leaves, about 0.5 mg of the powdered leaf material or extracted leaf sugars was weighed into tin capsules (Säntis Analytical, Teufen, Switzerland). The samples were combusted to CO₂ with an excess of oxygen at 1,020°C in an elemental analyzer (EA-1110, Carlo Erba Thermoquest, Milan, Italy), which was connected to a Delta S mass spectrometer with a CONFLO II system (both Finnigan MAT, Bremen, Germany), performing in continuous flow mode. The isotope ratio of δ¹³C is given in reference to its international standard, Vienna Pee Dee Belemnite (VPDB), in the delta notation in ‰: δ_sample = (R_sample/R_standard-1), with R_sample being the ¹³C/¹²C ratio of the sample and R_standard being the ratio of VPDB. The standard deviation of laboratory cellulose standards was used as an estimate of analysis precision and was lower than 0.10‰.

All data were analyzed using SPSS 21.0 (SPSS, Inc., Chicago, IL, United States). In order to evaluate differences among the treated and the control saplings, a non-parametric test of variance (Mann–Whitney U-test) was applied and FDR (false discovery rate) corrected to account for multiple comparisons (Benjamini and Hochberg, 1995). Data were analyzed separately for the drought and recovery period to detect days with significant differences between the treatments. The differences between the treatments were considered significant when P ≤ 0.05. The statistical tests were performed with four to six replicates per treatment group.

Irrigation was excluded from the drought-treated plots for 10 weeks, from June to mid-August, while control plots received regular irrigation (Figures 1A,B). Soil water content in the control plots ranged from 24 to 29% throughout the growing season (Figure 1C). In drought-treated plots, it ranged from 18 to 25% before and after treatment but decreased gradually to 12% during the drought period without irrigation. Trees in control plots maintained pre-dawn leaf water potentials (ψ_pd) above -0.5 MPa, except at the end of the growing season in October, when leaf senescence started and ψ_pd decreased to values of -1.0 MPa (Figure 1D). In drought-exposed saplings, ψ_pd started to decrease in mid-July, 7 weeks after irrigation was first withheld, and reached a minimum value of -1.8 MPa (mean of six replicates) at the end of the drought period in mid-August. After the first re-watering event, ψ_pd increased within 1 day to the level observed for control trees and even reached slightly higher ψ_pd values than those of control trees (Figure 1D).

Due to variable weather conditions, with both hot, sunny days and cool, cloudy days, stomatal conductance (g_S) varied strongly throughout the growing season, regardless of the applied drought treatment (Figure 2A). Nevertheless, a clear drought effect on g_S was observed in non-irrigated saplings. Stomatal conductance started to decrease when ψ_pd dropped below -0.6 MPa in mid-July and remained lower than values of the controls until re-watering in mid-August. After re-watering, g_S showed a fast initial increase and recovered fully after 6 days, reaching values comparable to those of controls. A similar pattern was obtained for intercellular CO₂ concentration (c_i), which decreased when ψ_pd dropped below -0.6 MPa and recovered quickly after drought release (Figure 2B).

Net-photosynthesis (A_N) showed a distinct seasonal pattern, with higher rates in summer and lower rates in spring and autumn (Figure 2C). Net-photosynthesis in drought-exposed saplings was comparable to that in control trees as long as ψ_pd remained above -0.6 MPa. When ψ_pd dropped below this critical value in mid-July, A_N started to decrease and was reduced by up to 60% at the end of the drought treatment in mid-August (Figure 2C). Net-photosynthesis responded quickly to re-watering, reaching 85% of the photosynthetic rate in control trees after 4 days and recovering completely after 1 week. Afterward, previously drought-exposed saplings had higher values of A_N compared to control trees, with a post-drought stimulation apparent toward the end of September and the beginning of October. The overall limitation of A_N in drought-treated saplings in relation to control saplings (estimated from the difference of the integrated areas below the curves) was -33%. The post-drought stimulation, integrated over the whole period after full recovery, was +16%.

Photochemical effects on photosynthesis were studied using PSII chlorophyll fluorescence analysis and the derived parameters maximum quantum yield of PSII (F_v/F_m) and total performance index of PSII (PI_tot). While F_v/F_m did not show any seasonal variation or response to the imposed drought, PI_tot showed a clear drought signal (Figures 3A,B). A decrease in PI_tot was observed in mid-August when ψ_pd reached values of -1.8 MPa, more than 3 weeks after the first drought effects on g_S and A_N were detected. After re-watering, PI_tot showed a delayed response, increasing slowly after 2 days and recovering completely 1 week after drought release, at the same time when full recovery of A_N was observed.

Concentrations of NSCs were measured in pre-dawn harvested leaves. In control saplings, the concentration of soluble sugars ranged from 10 to 25 mg g^-1 dry weight throughout the whole growing season, with no clear seasonal trend (Figure 4A). The concentration of starch and of total NSC ranged from 6 to 35 mg g^-1 dry weight and 25 to 58 mg g^-1 dry weight, respectively, throughout the growing season, with a gradual decline occurring from June to October (Figures 4B,C). While total NSC was not affected by the imposed drought, both starch and soluble sugars showed a distinct drought response. The starch concentration in leaves of drought-exposed saplings decreased as ψ_pd dropped to -1.0 MPa, 2 weeks after g_S and A_N showed a drought response. After re-watering in mid-August, the starch concentration increased within 1 day to the level in control saplings. After full recovery, the previously drought-treated saplings had slightly higher starch concentrations than control saplings, although this difference was not significant (Figure 4B). Soluble sugars showed a drought response that was the inverse to that observed for starch, with increasing concentrations during the drought treatment (Figure 4A). The first drought effects on the concentration of soluble sugars were observed when ψ_pd reached a value of -1.0 MPa. Soluble sugar concentrations returned to the same level as observed in control saplings within 1 day of re-watering and subsequently remained at the control level.

The ¹³C signature of the leaf bulk tissue (δ¹³C_L) did not show seasonal variation or a distinct drought response (Figure 5A). For both the control and the drought treatment, δ¹³C_L was around -30‰ at the beginning of the measurements in early June and decreased gradually to -31‰ in early October.

In contrast to δ¹³C_L, the ¹³C signature of soluble sugars (δ¹³C_S) showed a clear seasonal trend and a distinct drought response. δ¹³C_S in control saplings decreased gradually from -27‰ in June to -30‰ in early October. In drought-exposed saplings, an (non-significant) increase of δ¹³C_S by 2.7‰ (P = 0.06) was observed when ψ_pd reached a value of -0.8 MPa (Figure 5B). δ¹³C_S increased further with decreasing ψ_pd and reached a maximum value of -26‰ at the end of the drought period in mid-August. The drought signal in δ¹³C_S declined quickly after re-watering, reaching values close to those in control saplings after 1 day.

Our study on beech saplings revealed a sequence of leaf physiological changes that gradually developed in response to increasingly severe drought conditions (Figure 6). These responses were related to pre-dawn leaf water potential (ψ_pd), a physiological indicator of plant-internal water relations (Cochard et al., 2001; Ehrenberger et al., 2012). Leaf physiological responses were observed first for stomatal conductance (g_S), intercellular CO₂ concentration (c_i) and net-photosynthesis (A_N), which started to decrease when ψ_pd dropped below -0.6 MPa. Previous studies on beech showed similar responses of g_S and A_N to decreasing ψ_pd, suggesting that a ψ_pd of around -0.6 MPa represents a general threshold for drought effects on beech leaf gas exchange (Aranda et al., 2012; Cocozza et al., 2016).

Shortly after A_N started to decrease, the isotopic signature of soluble sugars (δ¹³C_S) showed an increase in drought-treated saplings relative to values in control saplings. This result may indicate limitations of A_N, due either to high leaf resistance to CO₂ diffusion and/or to impaired photochemistry (Farquhar et al., 1989). The latter explanation can be excluded, as chlorophyll fluorescence did not indicate impaired photochemistry and photochemical reactions are resistant to moderate stress (Kaiser, 1987; Havaux, 1992; Epron and Dreyer, 1993; Arend et al., 2013). Thus, the early decrease of A_N must be a result of limited CO₂ diffusion to the sites of photosynthetic reactions in the chloroplasts. Indeed, diffusional limitations are the predominant factors controlling photosynthesis in response to stress, including changes in both stomatal and mesophyll resistance (Centritto et al., 2003, 2009; Galmés et al., 2007; Flexas et al., 2008). In the present study, however, we obtained evidence that high stomatal resistance, i.e., low g_S, was the rate-limiting factor for A_N. In fact, the decrease in A_N was accompanied by a parallel decline in g_S and c_i, suggesting that CO₂ diffusion through the stomata was more affected than subsequent CO₂ diffusion through the mesophyll. This interpretation is consistent with the positive relationship between mesophyll resistance and c_i (Flexas et al., 2008), which in turn supports our assumption that stomatal resistance was the rate-limiting factor in leaf CO₂ diffusion. Reduced c_i has previously been shown to be an early response to water stress and is exclusively related to stomatal limitation of photosynthesis (Sharkey and Seemann, 1989; Grassi and Magnani, 2005). Overall, the initial increase in δ¹³C_S in drought-exposed beech most likely reflects stomatal limitation of A_N, which also suggests higher photosynthetic water use efficiency (Farquhar et al., 1989). With increasingly severe drought, however, when ψ_pd decreased to -1.8 MPa, concurrently measured chlorophyll fluorescence suggested additional limitations of photochemical reactions. The total performance index of PSII (PI_tot), a stress-sensitive fluorescence parameter (Strasser et al., 2010; Albert et al., 2011), indicated impaired PSII photochemistry that may have contributed to the further decline in A_N with drought.

Impaired photosynthesis, as observed in the present study, cause concomitant changes in leaf carbon metabolism (Pinheiro and Chaves, 2011). Reduced availability of recent photo-assimilates, for instance, may lead to greater consumption of transiently stored starch as an alternative source of sugars for metabolic activity, osmoregulation, and cellular stress defense (Chaves, 1991; Hare et al., 1998). In drought-exposed beech, we observed an increase in the concentration of soluble sugars at the expense of starch when drought stress severity increased, as indicated by a decline in ψ_pd to -1.0 MPa. We therefore cannot exclude that carbon partitioning from starch to sugars may have contributed to the observed increase in δ¹³C_S, as sugars derived from transiently stored starch are enriched in ¹³C relative to sugars from recently assimilated carbon (Rossmann et al., 1991; Gleixner and Schmidt, 1997; Gleixner et al., 1998). Indeed, starch concentrations started to decrease when δ¹³C_S increased, while the isotopic signature of leaf bulk material (δ¹³C_L) was less affected. Thus, there remains some uncertainty as to the source of increased δ¹³C_S, which seemingly compromises the interpretation given above linking δ¹³C_S with a diffusional limitation of A_N. Transient starch, however, forms a highly dynamic carbon pool that relies on the daily uptake and release of recently assimilated carbon. The rapid turnover of starch explains why sugars derived from transiently stored starch integrate the same physiological signals into their isotopic signature as sugars derived from recently assimilated carbon (Brugnoli et al., 1988; Blessing et al., 2015). It is therefore very likely that the increase in δ¹³C_S still reflects diffusional limitation of A_N, regardless of the fact that different sugar fractions contribute to this signal. Despite altered sugar and starch concentrations, the total amount of leaf NSC was not affected during drought. Thus, we can exclude the possibility that carbon starvation plays a role in the drought response of beech, which is in line with findings from previous studies on drought-exposed beech and oak trees (Li et al., 2013; Liu et al., 2017).

An important, but not fully explored, aspect of drought tolerance in trees is their ability to recover after drought release and quickly resume full physiological activity (Gallé and Feller, 2007; Arend et al., 2013, 2016b). In the present study, we took advantage of the immediate restoration of tree internal water relations after re-watering, which allowed us to study the recovery of leaf physiological activities without interfering effects of further plant-hydraulic constraints (Figure 6). Indeed, after the first re-irrigation pulse was applied, ψ_pd increased within 1 day to the level of control trees, even though soil re-wetting was not complete after the long-lasting summer drought. From this observation, we conclude that the hydraulic system in beech is fairly resistant toward drought-induced damage, as suggested in previous studies (Wortemann et al., 2011; Choat et al., 2012; Gleason et al., 2016). This interpretation is in line with previous observations that critical values of ψ_pd, which induce recovery failure of tree internal water relations, photosynthesis and other eco-physiological traits, are notably low in beech compared with other angiosperm tree species (Urli et al., 2013).

The immediate restoration of tree internal water relations was followed by a rapid adjustment of the leaf carbon metabolism to that in controls, as indicated by increasing starch and decreasing sugar levels and a fast decay of the drought signal in δ¹³C_S. The latter result suggests a rapid turnover of primary carbon metabolites after drought release, presumably driven by the increased delivery of new photo-assimilates and by an enhanced carbon demand for leaf respiration and export of sugars to respiratory sinks in the roots and rhizosphere (Zang et al., 2014). In fact, leaf carbon metabolism depends on whole tree carbon allocation dynamics, as demonstrated by a close coupling of above and belowground carbon fluxes (Högberg et al., 2001, 2008; Epron et al., 2011). In beech, it has recently been proposed that the drought recovery of respiratory sink activity in the roots and rhizosphere triggers the carbon export from leaves, which implies a lower residence time of recent assimilates in leaves and thus rapid turnover of primary carbon metabolites (Hagedorn et al., 2016).

Although levels of sugars and starch recovered within 1 day, photosynthetic traits remained constrained by the previous drought. After an initial rise, A_N did not increase further and remained for 1 week at lower levels than in controls. Other studies on beech have reported different time scales for a full recovery of A_N, ranging from 3 days to 4 weeks (Gallé and Feller, 2007; Zang et al., 2014; Blessing et al., 2016). However, results of drought experiments are difficult to compare, owing to the heterogeneity of experimental conditions, different stress intensities and ecological diversity of the plant material (Miyashita et al., 2005; Vicca et al., 2012; Arend et al., 2016b). In our study, the post-drought limitation of A_N was attributable to a persistent drought effect on stomatal and photochemical traits of photosynthesis, i.e., g_S and PI_tot, but we cannot exclude that altered mesophyll conductance was also involved (Flexas et al., 2008; Centritto et al., 2009). While g_S responded quickly to the first re-watering pulse, it did not fully recover and instead remained at lower levels than in controls for approximately 1 week. In contrast, PI_tot responded much more slowly to the first re-watering pulse, and full recovery was delayed by 1 week, indicating a down-regulation of PSII photochemistry that may have contributed to the delayed recovery of A_N.

There is increasing evidence that a drought effect is imprinted on trees, which compensates for the previous limitations of metabolic activities (Arend et al., 2016b; Hagedorn et al., 2016). Together with the tree’s capacity to recover from limiting stress, such a post-drought effect may be crucial for the tree’s overall drought resilience. In support of this, we observed a slight, but significant, post-drought stimulation of A_N, which counteracted the previous limitation of A_N under drought conditions. Interestingly, A_N started to increase after PI_tot recovered completely, suggesting that the post-drought stimulation of A_N depends directly or indirectly on full PSII photochemical activity. Altered stomatal regulation could be excluded as a cause for the observed post-drought stimulation of A_N, as g_S and c_i did not show any further change in previously drought-exposed beech saplings after full recovery. Instead, we observed a distinct trend toward increased starch concentrations and higher PI_tot values in previously drought-exposed beech saplings, lending further indirect support for improved leaf metabolic activity after drought recovery.