In November 2018, C. sativa seeds were collected from 16 adult trees per population in three wild populations in Spain (Puebla de Sanabria, Valle de Matamoros, and Paterna del Río). The three populations had contrasting climates in terms of precipitation and absolute maximum temperature (Table 1). Puebla de Sanabria is in northern Spain, close to the Transmontane region of northeast Portugal, known as Cold Land (“Terra Fria”), and has ca. 1,000 mm rainfall per year and belongs to the supra-Mediterranean bioclimatic zone (Table 1). Valle de Matamoros is in the foothills of Sierra Morena (centre-south of Spain), is characterized by 740 mm annual rainfall and belongs to the meso-Mediterranean bioclimatic zone (Table 1). Paterna del Río is in the Alpujarras, on the southern slopes of Sierra Nevada, has ca. 650 mm rainfall per year and belongs to the meso-Mediterranean bioclimatic (Table 1). For clarity, the populations were named northern, central, and southern populations, respectively. In each chestnut population, 16 fruit-bearing trees were selected at least 70 m apart to minimize the chances of sampling intercrossed individuals.

In November 2018, about 100 seeds collected from each of the 16 trees per population were immersed in a fungicide solution (2 g L^–1 Thiram 80GD, ADAMA Inc., Spain) for 5 min, and those that floated were discarded as non-viable. Seeds were rinsed with sterilized water and stored at 4 ^°C for 2 weeks. In December 2018, they were individually weighed and sown at 1 cm depth in 48-cell plastic root trainers with one seed per cell. Individual cells were 330 ml in volume, 18 cm high, 5.3 cm × 5.3 cm upper surface, and contained peat (PKN1 Florava^® Peat Substrate; pH = 4.5). Germination was considered successful when the aerial part emerging from the embryo was green. Aerial emergence of plants was assessed weekly. Plants were kept in natural daylight under greenhouse shade that reduced solar radiation by 50% and hand watered every 4 days to field capacity until they were well established. The greenhouse was at the Faculty of Forestry of Plasencia (40°02′ N, 6°04′ W; 374 m a.s.l., Extremadura region, Spain).

The experiment included seedlings from three populations, obtained from 16 mother trees in each population. Thirty seedlings per mother tree were used, of which half were subjected to ambient temperature (control), and half were subjected to a heat stress treatment. To reduce spatial thermal variability, the plants were arranged following a split-plot random design replicated in three blocks, with the heat treatments acting as the main factor (two categories: control and heat-stressed; whole plots) and the populations as the split factor (three categories: northern, central, and southern, as shown in Table 1; split plots). In the three blocks, the three populations were represented in each whole plot by five seedlings per mother tree. Seedlings were randomly positioned within the blocks. The experiment comprised 1,440 seedlings corresponding to 3 blocks × 2 treatments × 3 populations × 16 mother trees × 5 seedlings.

In May 2019, when the plants were 6 months old, 720 seedlings were subjected to ambient temperature, and 720 seedlings were subjected to the heat treatment. From 11 a.m. to 5 p.m., mean temperatures of control and heat-stressed plants were 30.2 and 42.5°C, respectively. The time frame selected for applying heat stress simulated natural conditions of daily temperature increase during summer in the Mediterranean Basin, with a maximum temperature from midday to early afternoon. The heat stress treatment comprised placing the plants inside a climate chamber with translucent walls in the same greenhouse. During treatments, both groups of plants were well irrigated. Volumetric soil water content (VWC), a relative measure of soil moisture, was maintained at values around 30% by irrigating once a day the control plants and by irrigating twice a day the heat stressed plants. Previous work indicated that 30% VWC was optimal for chestnut growth (Camisón et al., 2020). VWC was verified using a TDR 100 soil moisture meter (Spectrum Technologies Inc., Plainfield, IL, USA) and 12-cm-long rods in 10 cells per block per treatment. The leaves of control plants were green and turgid throughout the experiment (Figure 1A). During treatments, the relative air humidity and light conditions of both groups of plants were similar (≈80% and PAR ≈ 850 μmol photons m^–2 s^–1). The heat treatments lasted 7 days (day 7, “heat,” Supplementary Figure 1). After the heat treatments ended, heat-stressed seedlings were kept under control conditions for 10 days to assess their recovery (day 17, “recovery phase,” Supplementary Figure 1).

Measurements and sampling for biochemical analysis were performed on the first day of treatments (day 1), the last day of treatments (day 7, heat; Supplementary Figure 1), and 10 days after treatments ceased (day 17, recovery phase; Supplementary Figure 1) when plants were in recovery from stress (Figure 1C). On day 1, leaf gas exchange parameters were measured to observe the photosynthetic response of the populations to the increase in temperature. In the heat and recovery phases, the effects of heat on seedlings were evaluated by (i) foliar symptoms, leaf wilting, plant growth, and mortality; (ii) leaf gas exchange; (iii) quantification of primary metabolism parameters (NSC and proline) and N content; and (iv) quantification of secondary metabolism parameters (total phenolic compounds and ortho-phenolic and flavonoid compounds), 2,2-diphenyl-1-picrylhydrazyl (DPPH) free radical scavenging, and non-targeted metabolite profiling.

Leaf gas exchange parameters–net photosynthetic rate (Pn), transpiration rate (E), and stomatal conductance (g_s)–were determined on the first day of treatments (day 1) in 15 seedlings per population (five per block) and treatment. In the heat and recovery phases, leaf gas exchange parameters were measured in 10 seedlings per population using a portable differential infrared gas analyser (IRGA; Li-6400, Li-Cor INC., Lincoln, NE, USA) connected to a broadleaf chamber and a temperature sensor. On day 1, measurements were performed from 9 a.m. to 2 p.m., with temperature ranging from 32 to 48^°C, photosynthetically active radiation (PAR) ranging from 500 to 800 μmol photons m^–2 s^–1 (daylight and variable PAR conditions), and ambient CO₂ concentration ranging from 390 to 410 μmol mol^–1. In the heat phase, temperatures ranged from 23 to 27^°C and 37 to 42^°C in the control and heat treatments, respectively. In the recovery phase, temperatures ranged from 23 to 27^°C for both treatments. To reduce errors associated with CO₂ concentration and PAR changes over time, leaf gas exchange measurements were carried out by systematically alternating seedlings from the three populations (day 1) or by systematically alternating seedlings from the three populations and two treatments (heat and recovery phases). Instantaneous and intrinsic water use efficiencies were calculated based on leaf gas exchange, using the formulae instantaneous WUE = Pn/E and intrinsic WUE = Pn/g_s, respectively.

Eighteen plants per population and treatment were destructively harvested for primary and secondary metabolite analysis. After collection, samples from three different seedlings were pooled to obtain a final sample size of six pools per population and treatment. Samples were immediately frozen in liquid nitrogen and stored at −80°C. To account for the diurnal oscillation of NSC content in plant tissues (Tixier et al., 2018) and to ensure the results were comparable, plants were sampled from 2.00 p.m. to 3.00 p.m. Leaves, aerial woody tissues (stem plus small twigs, “stems”) and roots (taproot plus fine roots, “roots”) of each seedling were used to determine dry biomass and analyse N and NSC. Within 30 min of harvesting, excised organs were subjected to microwave treatment (80 s, 800 W) to inactivate the enzymes and fix the actual NSC content (Hoch et al., 2002). Samples were oven dried (48 h, 60°C), weighed on a precision balance for dry biomass quantification, and ground to a fine powder in a ball mill (Mixer Mill MM 400, Retsch, Germany) to pass through a 0.42 mm mesh screen.

For analyses of proline, secondary metabolites (total phenolic compounds and ortho-phenolic and flavonoid compounds), 2,2-diphenyl-1-picrylhydrazyl (DPPH) free radical scavenging and non-targeted metabolite profiling, leaves, and roots of seedlings from the two treatments were sampled. For leaves, those that were fully developed near the shoot tip were sampled. For roots, the outermost fine roots were collected after carefully lifting the root ball from the tray.

External symptom evaluation comprised visual characterisation of the foliar discolouration observed on seedlings. Wilting percentage was determined by visual estimation of the percentage of wilted leaves in plants subjected to heat stress in 10% intervals following Alcaide et al. (2019). Total plant dry biomass, comprising the dry weight of tissues used for NSC and nitrogen determination, was used as an indicator of plant growth.

Total NSCs included the sum of soluble sugars and starch. Soluble sugars and starch in leaf, stem, and root tissues were determined following the protocol adapted for chestnut by Camisón et al. (2020). Soluble sugar:starch ratios were calculated in each plant tissue and used as a proxy for mobilisation of starch to soluble sugars (Piper, 2011).

Total N in leaf, stem, and root tissues was determined by the Dumas method (DUMATHERM^® CN, C-Gerhardt Analytical Systems) and expressed as a percentage of dry weight. Proline content was determined using the protocol of Bates et al. (1973), with slight modifications for chestnut by Camisón et al. (2021). NSC or N concentration at the whole plant level was calculated by a weighted mean, taking into account the total NSC or N concentration of each tissue and the relative contribution of each tissue’s biomass to plant total dry biomass, following: Whole plant NSC or N (%) = [(X_l × a_l + X_s × b_s + X_r × c_r)/(a_l + b_s + c_r)]; where X_l, X_s, and X_r are the total NSC or N concentrations in leaves, stems, and roots, respectively, and a_l, b_s, and c_r are the proportions in which each tissue contributes to plant dry biomass.

To determine how the increase in temperature affected gas exchange parameters, a global and population-based regression analysis was performed. To assess the effect of heat stress on the physiological and biochemical parameters of C. sativa seedlings, one-way analysis of variance (ANOVA) was performed using treatment as the only factor. To analyse the differential response of populations, two-way ANOVAs were performed, this time including the main effects of treatment, population, and their interaction as factors. Within each date (heat and recovery phase), Tukey’s multiple comparison tests were used to identify significant differences between treatments and populations (P < 0.05). Lack of differences between treatments for physiological and biochemical parameters indicated no phenotypic plasticity in response to heat stress, whereas significant differences between treatments indicated phenotypic plasticity in response to heat stress. A significant treatment × population interaction was interpreted as genetic variation in phenotypic plasticity of a parameter in response to heat stress.

Discriminant function analysis (DFA) was performed with compounds from the non-targeted metabolite profiling to differentiate the three C. sativa populations and their response to heat in the heat and recovery phases. The population × treatment interaction was used as a grouping variable, and the concentrations of the compounds (identified and unidentified) were used as independent variables. The number of compounds initially used as independent variables in the DFA was 18. The independent variables finally selected were obtained from a sequential variable forward stepwise method. The DFA analysis was performed with STATISTICA v10 software (StatSoft Inc, 2011), and the graphical representation was performed with R software (R Core Team, 2018).

Principal component analysis (PCA) was applied to detect patterns of variation in physiological and biochemical parameters due to origin (population), response to heat, and recovery. The PCA was performed using the R software “factoextra” statistical package (R Core Team, 2018). Before the PCA, the data were centred and standardised to reduce scale effects (Martín et al., 2008b). Relationships between leaf and root parameters used to perform PCA were assessed for the heat phase, separated into control, and heat stress conditions, using a Pearson’s correlation matrix performed with the R software “corrplot” package (R Core Team, 2018).

For each parameter and population, phenotypic plasticity in response to heat stress was expressed as a percentage relative to control plants following the equation proposed by Rehschuh et al. (2020) in the heat phase: Phenotypic plasticity (%) = [(x–C_pop)/C_pop) × 100]; where x is the heat-induced value of the parameter of a particular seedling of the population, and C_pop is the mean constitutive (control) value of the parameter of seedlings of the population. To check whether the effect of treatment differed between populations, one-way ANOVAs were performed using the population as the only factor. Before all analyses, data were checked for normality and homogeneity of variances. ANOVAs and Tukey’s tests were performed using STATISTICA v10 software (StatSoft Inc, 2011).

On day 1, Pn and g_s values in seedlings decreased exponentially as temperature increased (R² = −0.90 and R² = −0.59, respectively, P < 0.001; Supplementary Table 1), while E values increased as temperature increased (R² = 0.28, P = 0.037). Gas exchange response to temperature increase depended on population (Figure 2). In seedlings from the northern population, E increased with temperature, while in seedlings from the central and southern populations, E was not significantly altered (Figure 2B). In seedlings from the northern population, g_s did not change, while in seedlings from the central and southern populations, stomata closed as temperature increased (Figure 2C).

During the heat phase, the leaves of stressed seedlings showed brown margins, yellowing, and browning of inner areas, and scorching (Figure 1B). Leaf damage was observed only in about 10% of seedlings, but heat-stressed seedlings showed lower leaf, root, and total biomass compared to control plants (Table 2 and Supplementary Table 2). Heat stress did not lead to plant mortality. In the recovery phase, seedlings that had wilted more started to resprout (Figure 1C), although significant changes in seedling biomass were not detected (Supplementary Table 2).

Heat stress did not affect plant Pn, E, and g_s but induced a significant decrease in instantaneous and intrinsic WUE compared to control seedlings (Table 2 and Supplementary Table 2). During the recovery phase, E and intrinsic WUE remained altered in heat stressed seedlings compared to control seedlings (Table 2 and Supplementary Table 2).

Major alterations observed during the heat phase in primary metabolism parameters occurred mainly in leaves (Table 2 and Supplementary Table 2). Heat-stressed seedlings responded by increasing soluble sugars and decreasing starch in leaves and stems (Table 2 and Supplementary Table 2). Soluble sugar:starch ratio was significantly increased by heat stress in leaves and the whole plant. Total NSC content in leaves decreased significantly with heat stress, while total NSC in stems and roots did not change significantly (Table 2 and Supplementary Table 2). In heat-stressed seedlings, N content significantly increased in all organs and at the whole plant level, and proline content in leaves doubled (Table 2 and Supplementary Table 2). Recovering seedlings continued to have higher soluble sugar concentration, lower starch concentration in leaves and stems and higher soluble sugar:starch ratios in leaves than control seedlings (Table 2 and Supplementary Table 2). In contrast, N and proline concentration returned to the levels of control seedlings (Supplementary Table 2).

Secondary metabolism of heat-stressed seedlings was most altered in the roots (Tables 2–4 and Supplementary Table 2). Chestnut seedlings subjected to heat stress significantly increased total phenolic compounds, ortho-phenolic, and flavonoid compounds only in roots (Table 2 and Supplementary Table 2). Given the heat-induced alterations in leaf and root phenolic compounds, non-targeted metabolite profiling was applied, revealing 18 compounds, 15 of which were identified (Table 3). Six identified compounds increased in heat-stressed plants (Table 4). Five of the six compounds showing variation were in the roots and belonged to the phenolic acid (hydroxybenzoic acid subclass) and flavonoid (flavonol subclass) families (Tables 3, 4). DFA revealed different secondary metabolism profiles between treatments in chestnut (Figure 3A). Isorhamnetin in leaves, and hydroxybenzoic acid, kaempferol 3-O-(6′′-O-acetyl) glucoside-7-O-rhamnoside and hyperoside in roots, were involved in this discrimination (P < 0.001). However, under heat stress, antioxidant capacity increased significantly only in leaves (Table 2 and Supplementary Table 2). During the recovery phase, only flavonoids in leaves remained altered by heat stress (Table 2 and Supplementary Table 2). DFA of non-targeted metabolite profiling showed differences between recovery and control seedlings (Figure 3B). Astragalin and hydroxybenzoic acid decreased significantly in recovering seedlings compared to controls (Table 4 and Supplementary Table 4).

Principal component analysis revealed a clear difference in the physiological and biochemical status of C. sativa seedlings under control and heat treatments (Figure 4A). During the heat phase, the separation between control and heat was mainly associated with starch depletion in leaves (28.4%), high ratios of soluble sugars:starch in leaves (27.9%) and at the whole plant level (8.1%), proline accumulation in leaves and roots (2.6 and 12.1%, respectively), and accumulation of phenols (1.9%), ortho-phenols (2.8%), and flavonoids (3.1%) in roots (Figure 4A). After recovery, these differences mostly disappeared, and groups of plants clustered close to each other (Figure 4B). Correlation analysis between PCA variables revealed a biologically interesting correlation between proline content in leaves and soluble sugars in leaves and stems (Figure 5). This correlation changed from negative under control conditions to positive under heat stress conditions (Figure 5).

During the heat phase, no interpopulation differences were observed in external symptoms and foliar damage. However, seedlings from the northern population were the most adversely affected by heat stress in terms of biomass reduction in leaves, roots and at the whole plant level (Supplementary Table 3). Biomass of seedlings of southern origin was unaltered during and after heat stress (Supplementary Table 3). At recovery, no differences in biomass between heat-stressed and control seedlings were found for any population (Supplementary Table 3). Among leaf gas exchange parameters, E showed the highest interpopulation variability in response to heat stress (Supplementary Figure 2), increasing ca. 50 and 25% in the southern and central populations and decreasing ca. 25% in the northern population compared to controls (Supplementary Figure 2).

Interpopulation differences in primary metabolism were mostly observed in leaves and stems (Figure 6 and Supplementary Table 3). Leaf starch and soluble sugars:starch ratio were the parameters with the greatest interpopulation variability, each contributing approximately 45% to the separation between populations in the PCA (Supplementary Figure 3A). The southern population was the most reactive population to heat stress, reducing starch concentration in leaves by almost 100% and markedly increasing the soluble sugars:starch ratio relative to the control (Figure 6A and Supplementary Table 3). Variations in leaves for starch and the soluble sugars:starch ratio of seedlings from the northern population were low (Supplementary Table 3). In the stem, starch and NSC content decreased in the northern and central populations but increased in the southern population. Soluble sugar changes were similar in the northern and southern populations (Figure 6B).

Central and southern heat-stressed seedlings significantly increased their N content in leaves and stems compared to control seedlings (Supplementary Table 3). Heat-stressed seedlings from the northern population had a fourfold increase in leaf proline content, while no change was observed in the proline content of seedlings from the central and southern populations (Supplementary Table 3). After recovery, only southern seedlings had lower total starch and NSC content in leaves compared to the control (Supplementary Table 3). Both N and proline in recovering seedlings returned to control levels, regardless of the population (Supplementary Table 3).

Regarding secondary metabolism compounds, the populations most adapted to heat stress (the meso-Mediterranean central and southern populations) showed the highest variations in total phenylpropanoids and antioxidant capacity in leaves and roots after heat treatment (Figure 7). Non-targeted metabolite profiling in response to heat showed different changes in the contents of the compounds identified depending on the population of origin (Figure 4A and Table 4). Heat-stressed seedlings from the central population showed the greatest alteration of secondary metabolism, with six compounds (hyperoside, rutin, isorhamnetin, astragalin, ellagic acid, and ellagic acid acetyl-xyloside) accumulating in leaves (Supplementary Table 5). Seedlings from the southern and northern populations showed an altered metabolomic profile in roots (Supplementary Table 4). DFA of the recovery phase showed different metabolite profiles depending on the seedling origin and pretreatment (Figure 3B). PCA analysis also showed a differential interpopulation response during the heat phase (Figure 4A). The greatest response to heat stress by PCA analysis was in seedlings from the southern population, associated with higher starch depletion and a higher soluble sugars:starch ratio in leaves and at the whole plant level, and higher accumulation of proline, total phenolic compounds and ortho-phenolic and flavonoid compounds in roots (Figure 4A). After recovery, only seedlings from the heat-stressed southern population were separated from their respective controls, mainly due to starch depletion in leaves (Figure 4B).

Genetic variation in phenotypic plasticity of C. sativa in response to heat stress was observed for ten physiological and biochemical parameters (Figure 8). The southern population was not responsive in terms of biomass in comparison to the northern and central populations. However, the southern population was the most responsive in terms of soluble sugars: starch ratios and N content (Figure 8). NSC always declined for the central population, but behaved differently for the other populations, being the parameter showing the highest genetic variation in phenotypic plasticity (Figure 8).

Observations reported here in C. sativa seedlings subjected to 42.5°C for 7 days confirm that heat stress caused leaf and branch scorch, leaf senescence and abscission, and inhibition of shoot and root growth (Wahid et al., 2007). However, differentiation across populations in genetic variation of biomass was minimal. Leaf gas exchange parameters were different on the first and seventh days after heat stress. During the first episode of temperature increase, seedlings decreased net photosynthetic ratio, but after a heat wave (7 days of heat stress), this parameter did not alter, agreeing with results obtained by Ameye et al. (2012) in Quercus rubra. This suggests that in chestnut seedlings Pn is more sensitive to daily temperatures than to prolonged heat exposure and could also indicate plant adaptation to elevated temperatures.

Heat stress induced considerable changes in compounds derived from primary and secondary metabolisms. NSC was especially plastic above ground, and at the whole plant level, NSC decreased in response to heat stress except in the southern population. The NSC reserve pool in trees constitutes the primary resource supply for energetic disparities and generally draws down when trees experience abiotic stress (Galiano et al., 2011; Wiley and Helliker, 2012; Hartmann et al., 2013; Dietze et al., 2014; O’Brien et al., 2014). The high genetic variation in phenotypic plasticity of NSC (Figure 8) is in accordance with a recent study performed in Populus trichocarpa trees (Blumstein and Hopkins, 2021). This genetic variation in heat response could provide a key mechanism through which populations can evolve a more adaptive response to future global warming.

Non-structural carbohydrates act as protective molecules against heat stress (Wani and Kumar, 2020). Soluble sugars stabilise proteins and cell membranes, maintain cell turgor, participate in stress signalling pathways, and scavenge excess ROS (Bhattacharya and Kundu, 2020), and starch degrades to aid plant fitness and survival under harsh growing conditions (Thalmann and Santelia, 2017), e.g., a rapid degradation of starch generates organic acids and sugars that maintain stomatal pore guard cell turgor and promote stomatal opening (Horrer et al., 2016). In response to heat stress, glucose, and fructose increase has been reported in conifer needles (Marias et al., 2017). Rapid starch degradation and soluble sugars accumulation has been reported in Populus tremula leaves at elevated temperatures as a strategy of trees to protect respiratory membranes and buffer consumption in respiratory processes (Hüve et al., 2012). In a drought-tolerant Eucalyptus globulus clone subjected to heat stress, a decrease in leaf starch content and an increase in soluble sugars (mannitol, sorbitol, inositol, and several amino acids) were observed (Correia et al., 2018). In our heat-stressed chestnuts, plastic variation in the amount of NSCs also had a genetic component.

A positive phenotypic correlation was observed between soluble sugar content and proline content in leaves. Proline protects the structure of enzymes and proteins by maintaining membrane integrity and eliminating ROS (Hayat et al., 2012). Proline and soluble sugar signalling pathways interact synergistically, forming part of the plant antioxidant system under stress situations (Moustakas et al., 2011). The increase in soluble sugars, N, proline and antioxidant capacity observed here in leaves, together with the absence of changes in phenolic compounds in leaves, suggests chestnut trees use carbohydrates as antioxidant compounds. In addition, the decrease in starch, increase in soluble sugars, and unaltered stomatal conductance between treatments suggest chestnut trees use sugars to keep stomata open despite heat stress.

In relation to the secondary metabolism, an accumulation of total phenolic compounds and ortho-phenolic and flavonoid compounds was observed below ground in chestnut seedlings. We identified and quantified 15 phenolic compounds, some of which (ellagic acid, 4-hydroxyphenylacetic acid, hydroxybenzoic acid, quercetin 3-O-galactoside, kaempferol 3-O-(6′′acetyl) glucoside 7-O-rhamnoside, tyrosol, and 4-hydroxy-3-methoxycinnamaldehyde) are reported here for the first time in chestnut roots. Moreover, hydroxybenzoic acid, kaempferol 3-O-(6′′-O-acetyl) glucoside-7-O-rhamnoside and hyperoside in roots were significantly involved in discriminating between heat-stressed and non-stressed seedlings. Accumulation of phenolic compounds in roots has been associated with plant root biomass loss (Rhodes et al., 1987), as occurred here. Derivatives of two compounds involved in the DFA separation between treatments [kaempferol 3-O-(6′′-O-acetyl) glucoside-7-O-rhamnoside and hyperoside] accumulated in roots and were associated with inhibition of polar auxin transport (PAT) in roots (Yin et al., 2014). Phenolic accumulation, especially of compounds with antioxidant functions, has been reported as frequent when applying abiotic stress and may occur as an adaptive mechanism (Rivero et al., 2001; Zandalinas et al., 2017; Bhattacharya and Kundu, 2020; Carneiro-Carvalho et al., 2021).

Regarding the recovery of chestnut seedlings from heat stress, most parameters returned to initial levels, agreeing with observations in other forest species (Escandón et al., 2016; Correia et al., 2018). However, leaf soluble sugar, starch and flavonoid content, and leaf antioxidant capacity remained altered. The durable accumulation of soluble sugars, flavonoids, and antioxidant capacity concur with the role of these compounds in tolerance, adaptation, and recovery from abiotic stress (Singh et al., 2017; Bhattacharya and Kundu, 2020). Previous studies in chestnut reported durable accumulation of carbohydrates in response to water stress (Camisón et al., 2020).

Physiological and biochemical heat stress responses of trees can vary depending on the climate of origin, and different populations can show varied responses, from high tolerance in warm climates to low tolerance in cold climates (Marias et al., 2017). Knowledge about phenotypic plastic responses and their genetic control is necessary to forecast the full potential of a species to adapt and/or evolve to changing conditions (Zas et al., 2020; Blumstein and Hopkins, 2021). Likewise, because of the increasing impact of global warming on plants, there is increasing interest in quantifying phenotypic plasticity and local adaptation (i.e., adaptive genetic changes) to heat and drought stress (Castellana et al., 2021). We found interpopulation differences in chestnut seedlings in morphological, physiological, and biochemical parameters in response to heat stress. Seedlings whose mother trees originated from less exposed heat conditions in their area of origin, i.e., from the northern population, were the most adversely affected by heat in terms of biomass reduction. During the first hours of heat, plants from the supra-Mediterranean (northern) population increased their transpiration rates exponentially compared to seedlings from the meso-Mediterranean (central and southern) populations. We hypothesise that seedlings from the northern population use immediate transpiration as a strategy to cool leaves and avoid heat stress consequences, as observed in Q. rubra (Ameye et al., 2012) and Acer rubrum (Weston and Bauerle, 2007). In the northern population, the decrease in Pn while g_s remained unchanged suggests an alteration of the photosynthetic capacity of the mesophyll (Camejo et al., 2005). Decreased Pn due to non-stomatic limitations has been observed in plants not adapted to heat (Camejo et al., 2005). However, during the heat phase, no interpopulation differences were observed in gas exchange parameters. This could suggest that regardless of the population, chestnut is able to adapt its photosynthetic apparatus to sustained heat stress.

Different interpopulation responses to heat stress were detected in the primary metabolism of seedlings. Seedlings from the southern population were the most responsive, followed by those from the central population and, finally, by those from the northern population. Seedlings from the southern population were characterised by almost total starch depletion and an increase in soluble sugars and the soluble sugars:starch ratio and, as discussed above, starch mobilisation and accumulation of soluble sugars are associated with tolerance to heat stress (Xiong et al., 2015). Proline accumulation in leaves of heat-stressed plants occurred only in seedlings from the northern population, and proline was the main variable explaining the separation of populations in the PCA of heat-stressed seedlings (Figure 3A). Low proline accumulation in chestnut has been associated with tolerance to heat stress in plants treated with SiK (Carneiro-Carvalho et al., 2021), and heat-induced proline accumulation in Arabidopsis plants has been considered detrimental and responsible for reduced thermotolerance (Lv et al., 2011). Thus, the accumulation of this osmolyte in C. sativa seedlings from the northern population could indicate reduced thermotolerance.

Regarding secondary metabolism, the southern and northern populations also showed the highest and lowest response to heat stress, respectively. The response of seedlings from the southern population was a high accumulation of total phenolic compounds and ortho-phenolic and flavonoid compounds in roots. However, analysis of non-targeted metabolite profiling showed that seedlings from the central population, exposed to maximum temperatures of almost 40^°C in the area of origin, experienced the most pronounced changes, comprising an increase in ellagic acid, ellagic acid acetyl-xyloside, derivatives of quercetin, kaempferol-3-O-glucoside and isorhamnetin in leaves. Targeted and non-targeted approaches provided complementary information and should be used together, when possible, to better characterise tree response to heat stress.

Ten days after the heat stress treatment ceased, seedlings from the three populations reached levels of control seedlings in practically all physiological and biochemical parameters. Only those from the southern population remained separated from the control in the PCA analysis due to starch depletion and mobilisation to soluble sugars. In Spain, differences in the pluviometry of chestnut habitats have resulted in differentiation of two C. sativa ecotypes with different adaptive responses to drought stress (Míguez-Soto and Fernández-López, 2015; Alcaide et al., 2019; Míguez-Soto et al., 2019). These ecotypes were also differentiated by EST-SSRs associated with heat stress using the same populations of the present study (Dorado et al., 2022). Based on the results obtained here, differences in the thermal regime of chestnut habitats may also have resulted in different ecotypes associated with heat stress tolerance. It should be acknowledged that our results were obtained in seedlings. Since tolerance to environmental stress may change throughout the lifetime of trees (Solla et al., 2005; Niinemets, 2010), variation of phenotypic plasticity and acclimatisation to heat stress may also change depending on the age of trees and forests.