Hot and Hungry - High temperatures induce changes in leaf carbon dynamics and sugar isotope fingerprints

Abstract

Predicting vegetation responses to global warming and reconstructing them from stable isotope records requires a clear understanding of how they respond to temperature. We investigated leaf carbon-dynamics (gas exchange, non-structural carbohydrate (NSC)) and the hydrogen (δ²H) and oxygen (δ¹⁸O) isotopic composition of leaf water and sugars in well-watered C₃ trees, forbs, grasses, and one C₄ grass across air temperatures from 10 °C to 40 °C at a low VPD. In C₃ species, temperatures ≥30 °C reduced A_net, increased R_dark, and shifted NSC composition from starch to sugars. Concurrently, apparent δ²H and δ¹⁸O fractionation between leaf water and sugars decreased. δ²H and δ¹⁸O in C₃ plants can be modeled from leaf water isotope composition when accounting for temperature-driven changes in gas exchange and NSC dynamics. These findings reveal heat-induced shifts in carbohydrate metabolism leave distinct isotope signatures, providing a mechanistic basis for improved isotope models to study current and past tree carbon dynamics.

Introduction

Temperature is a key driver of leaf physiological responses in plants. However, under natural conditions, its effects are often confounded by vapor pressure deficit, and technical challenges make it difficult to isolate the impact of temperature, particularly at extreme levels (e.g., >30 °C). As a result, how specific temperature impacts leaf level carbon dynamics (i.e., photosynthesis, respiration, carbohydrate metabolism) and the resulting isotope patterns remain not fully understood. Here we aimed to isoloate the temperature effect (10 to 40 °C in 5 °C steps) for a wide range of species from different plant functional types (i.e., trees, grasses, forbs) to improve the mechanistic understanding on the dynamics of leaf physiology and its impact on hydrogen (δ²H) and oxygen (δ¹⁸O) isotope pattern in plant carbohydrates.

The analysis of stable isotope ratios is a widely applied tool to reconstruct past and study current responses of plants to various environmental drivers^1,2,3. While traditional mechanistic models to predict δ²H and δ¹⁸O in tree ring cellulose⁴ indeed separate photosynthetic (ε_A, i.e., autotrophic) and post-photosynthetic ε_H, i.e., heterotrophic) fractionation processes, they have traditionally been considered to be constant within and among species and in response to environmental drivers^2,5. However, recent research has shown that the fractionation of oxygen and especially hydrogen isotopes being highly variabile among different photosynthetic types (i.e., C₃, C₄, CAM), with up to 20‰ in δ¹⁸O and more than 200‰ in δ²H^6,7. The observed high variability in δ²H mostly derives from the large variability within C₃ species^8,9. Furthermore, the ²H fractionation is strongly responding to environmental drivers^10,11,12, but even δ¹⁸O is known to have a distinct temperature response¹³. Deriving from the available source water, primary assimilates such as sucrose form the base of the isotopic composition of cellulose of leaves¹⁴ and twig xylem^4,8. Therefore, to fully understand the observed δ²H and δ¹⁸O in tree-ring cellulose, it is crucial to understand how the observed apparent autotrophic ²H and ¹⁸O fractionations are responding to environmental drivers. It has been shown that the temperature effect of δ¹⁸O in tree-ring cellulose is closely correlated to the vapor pressure deficit (VPD) of the air¹⁵, however, to our knowledge, no study isolated the pure temperature effect on the authotrophic ¹⁸O fractionation from leaf water to leaf sugar. The fractionation of ²H and ¹⁸O during NSC synthesis and metabolism is influenced by drivers such as temperature¹³, and different biochemical reactions^7,16,17. Recent studies suggest that particular the hydrogen isotope composition of plant carbohydrates strongly responds to changes in leaf gas exchange and NSC pools¹⁰, and are connected to leaf dark respiration (R_dark)¹⁴. Other recent studies show the possible link between a negative plant carbon balance and a ²H enrichment in plant organic matter¹⁸, and a decoupling between δ²H and δ¹⁸O in tree-ring cellulose delivered important information whether the carbon balance of a tree is out of balance^11,12. However, these studies fall short of providing enough physiologically relevant measurements with high resolution under changing environmental conditions to investigate how leaf gas exchange and carbohydrate metabolism are integrated into the stable isotope pattern of plant carbohydrates. Therefore, there is still a lack of comprehensive studies directly connecting the temperature response of plant physiological processes, carbohydrate metabolism, and the fractionation of ²H and ¹⁸O in leaf sugar. To overcome this limitation, we investigate and link the temperature responses of leaf internal carbon dynamics with the apparent photosynthetic ²H and ¹⁸O isotope fractionation.

Photoautotrophic CO₂ fixation, one key process in the global carbon cycle and the primary process that provides the carbohydrates needed for plant growth and productivity, is highly dependent on several environmental factors, of which temperature is one of the most important¹⁹. The optimum temperature for the net CO₂ assimilation (A_net) varies between plant species, typically ranging from 20–30 °C, usually align with the prevailing growth temperatures in their environment²⁰, and can be stable until 46 °C in plant species adapted to heat²¹. Above the optimum temperature, A_net decreases due to several processes, including damage to the photosynthetic machinery^22,23,24. Post-photosynthetic metabolic processes such as mitochondrial respiration provide energy for growth and maintenance, and are thus essential for plant reproduction and survival. Respiratory processes continue to increase until much higher temperatures than photosynthesis²⁵ due to increased enzyme activity²⁶ and maintenance costs²⁷. This can lead to a decrease in energy efficiency if maintenance costs become too high, thereby reducing the plant’s ability to produce a surplus of carbohydrates for growth and storage²⁸.

Since A_net and respiration are temperature-dependent, by extension plant non-structural carbohydrate (NSC, i.e., sugar, starch) dynamics are also temperature-dependent^29,30,31. Assessing the effects of rising temperatures on plant metabolism can be challenging because temperature changes are often associated with concomitant changes in vapor pressure deficit (VPD)^32,33, which can impact plant physiology and biochemistry in complex ways^34,35. For instance, high VPD can lead to a reduction in stomatal conductance (g_s) and therefore CO₂ uptake³⁶, affecting A_net³⁷ and, ultimately, plant metabolism and growth³⁸. However, experiments aiming to isolate temperature from VPD effects on leaf gas exchange and NSC metabolism at plant physiologically relevant temperature ranges are still rare due to technical challenges^33,39,40.

To isolate the effects of rising temperature under constant VPD on plant metabolism, we conducted a climate chamber experiment where we grew six C₃ (including two trees, Quercus pubescens Willd. and Phytolacca dioica L.; two grasses, Hordeum vulgare L. and Oryza sativa L.; and two forbs, Salvia hispanica L. and Solanum cheesmaniae (Riley) Fosburg) and one C₄ (Sorghum bicolor (L.) Moench) plant species. As shown by Schuler et al.⁷, C₄ plants, unlike C₃ plants, did not show a temperature response in their δ²H of leaf sugar and leaf cellulose, and was therefore used as a control. This selection includes one temperate and one subtropical tree species and five agriculturally important crops from different geographical and climatic zones. Similar to Wieloch et al.¹⁷, we reduced the old NSC pool between each temperature cycle by keeping the plants in the dark at 20 °C for 48 h. This reduction of old NSC was used to ensure that the measured δ²H and δ¹⁸O in sugars corresponded to the physiological conditions of the plants at each specific temperature. We exposed the plants to a constant daytime temperature for five days for 18 h each, starting at 10 °C and subsequently increasing to 40 °C in 5 °C steps. This allowed the plants to acclimate photosynthesis and respiration⁴¹ to each tested temperature. Water supply (kept at field capacity), light (PAR = 800 µmol m⁻² s⁻¹ during the daytime hours), and VPD (1 kPa, through regulation of relative humidity) were kept constant. On the fourth day, we sampled leaves for non-structural carbohydrates (NSC) and stable isotope (δ¹⁸O and δ²H of leaf sugars and water) analyses. On the fifth day, we measured leaf gas exchange (A_net, R_dark, g_s) and chlorophyll fluorescence. The gained knowledge will help to better understand the measured oxygen and hydrogen isotope fractionations in plant carbohydrates and to eventually update current isotope models⁴.

Results

Temperature response of the leaf gas exchange and the electron transport rate in PSII

Leaf physiology responded strongly to the temperature treatment from 10 to 40 °C (Fig. 1, Supplementary Tables 1 and 2). A_net (Fig. 1a) of the C₃ species, except Salvia hispanica, reached its maximum between 25 and 30 °C, followed by a steady decline with further temperature increase. On the other hand, A_net of the C₃ plant S. hispanica and the C₄ plant Sorghum bicolor increased until 30 °C and remained largely stable until 40 °C. Leaf dark respiration (R_dark, Fig. 1b) increased strongest in Solanum cheesmaniae with increasing temperature, while R_dark of P. dioica was in general higher between 15 and 35 °C compared to the other species, which exhibited similar and modest increases with rising temperatures. Gross photosynthesis (P, A_net + R_dark, Fig. 1c) increased with increasing temperature in all species but decreased strongly only in Phytolacca dioica above 30 °C. Stomatal conductance (g_s, Fig. 1d), while being highly variable among all tested species, as well as its response to increasing temperatures, generally increased with higher temperatures. While S. hispanica and S. bicolor maintained a stable electron transport rate (ETR, Fig. 1e) above 25 °C, ETR of the other C₃ species showed a strong decrease above 30 °C (ETR measurements of C₄ species does not give reliable results with the used method). A_net was significantly related to g_s in five of the seven species (Fig. 1f), but since g_s kept increasing after A_net reached its optimum, their relationship decoupled or reversed (i.e., became negative) at high temperatures⁴⁰.

Fig. 1: Temperature response of leaf gas exchange, electron transport rate, and stomatal decoupling.

Species-specific temperature response of a the net assimilation rate (A_net), b dark respiration rates (R_dark), c gross photosynthesis (P, i.e., A_net + R_dark), d stomatal conductance (g_s), e electron transport rates (ETR), and f the relationship between A_net and g_s. Colors indicate species, the quadratic models depicting the relationship are shown only for species with a significant temperature response (p ≤ 0.05), and the light shading denotes the 95% confidence level interval for predictions of the quadratic fit.

Changes in non-structural carbohydrate concentration and composition in response to rising temperatures

The concentration and composition of leaf NSC responded significantly to temperature (Fig. 2, Supplementary Tables 3 and 4). Across all species, the total NSC concentration was significantly higher at lower temperatures (10–15 °C) compared to higher temperatures (35–40 °C) (Mann–Whitney U test p < 0.001), with an average of 13.72% of the leaf dry mass at lower temperatures vs. 7.84% at higher temperatures. Similarly, the percentage of starch that contributes to the total leaf NSC pool was significantly higher at lower temperatures (51.44%) than at higher temperatures (21.96%) (Mann-Whitney U test p < 0.001). Two groups can be identified based on their NSC storage strategy: One group consisting of the temperate tree and the grass species (Fig. 2 left panels; Quercus pubescens, Oryza sativa, Hordeum vulgare, and S. bicolor) always stored most of their total NSC pool as sugar. In contrast, the subtropical tree and forbs (Fig. 2, right panels; S. cheesmaniae, P. dioica, and S. hispanica) stored most of the total leaf NSC as starch at moderate temperatures. Lastly, O. sativa was the only species with no significant change in its NSC composition, with 80% of the leaf NSC stored as sugar at all temperatures.

Fig. 2: Temperature response of non-structural carbohydrates in leaves.

Runing mean of species-specific temperature response of the total content of non-structural carbohydrates (NSC in % of leaf dry mass, black, left y-axis), and the percentage of the leaf NSC pools consisting of either starch (% Starch, blue, right y-axis) or sugar (% Sugar, red). Each variable is smoothed using lowess (locally weighted scatterplot smoothing) to show trends across the temperature range.

The dual isotope response to rising temperature

Across all species, the two measured isotope ratios (δ¹⁸O and δ²H) showed distinct responses to the increasing temperature (Fig. 3, Supplementary Figs. 2, 3, Supplementary Table 5). While the observed temperature effect on Δ²H of leaf water (Δ values are normalized to the initial average value for each species at 10 °C) was significant but small (R² = 0.114, p < 0.01), higher temperature led to a non-linear increase in Δ²H of leaf sugar (R² = 0.423, p < 0.001), which was caused by the temperature response of the biological ²H fractionation ε_HA (R² = 0.484, p < 0.001). In contrast, Δ¹⁸O of the leaf sugar decreased linearly with increasing temperature (R² = 0.74, p < 0.001), reflecting a combination of the linear decreases in leaf water Δ¹⁸O and ε_OA (R² = 0.446, p < 0.001 and R² = 0.234, p < 0.001, respectively), leading to an overall decrease of ε_OA by 0.11‰ per 1 °C temperature increase. We found a positive relationship (covariation) between Δ¹⁸O and Δ²H in leaf water within and across all tested temperatures (Supplementary Fig. 4). However, in leaf sugar, we only found a significant and positive relationship between Δ¹⁸O and Δ²H at 25 and 30 °C, while the overall relationship across all temperatures was negative (Supplementary Fig. 4).

Fig. 3: The temperature response of ²H and ¹⁸O.

The isotope response to temperature of Δ²H and Δ¹⁸O in leaf water and leaf sugar, and ε_HA and ε_OA, the apparent ²H and ¹⁸O fractionation between leaf water and leaf sugar. Values are given as Δ; e.g., normalized to average δ values measured at 10 °C individually for each species. The lines represent the quadratic (²H) respectively the linear model (¹⁸O) depicting the relationship between temperature and Δ and ε values, the lighter shading denotes the 95% confidence level interval for predictions of the quadratic and linear fit, R², and p values as well as the corresponding functions for the temperature response (i.e., ‰ at a given temperature) values inform on the temperature relationship of the variables.

Exploring the drivers of leaf sugar δ²H and δ¹⁸O by using generalized additive models (GAM)

The 34 GAM evaluated for δ²H (Supplementary Table 6) and 19 GAM evaluated for δ¹⁸O (Supplementary Table 7) of δ²H and δ¹⁸O of leaf sugar (δ²H_LS, δ¹⁸O_LS) showed a wide range of explanatory power. For δ²H of leaf sugar (δ²H_LS), the species-specific average apparent fractionation between leaf water and leaf sugar (ε_HA25), representing the baseline in all models, explained 58.8% of the variation (model 34: adj. R² = 0.581, p < 0.001). The explanatory power of δ²H of leaf water (δ²H_LW) for δ²H_LS was only significant for two simple versions of the δ²H_LS model (model 33: adj. R² = 0.612, p < 0.01). However, as δ²H_LW forms the basis of the isotopic composition of δ²H_LS, it is expected to have a significant and consistent effect over larger geographic or temporal scales when the δ²H value of the source water differs considerably. Thus, the δ²H_LW was not removed in the more complex models, although its contribution was not significant in this study, it will be necessary to integrate δ²H of the source water into the model to predict leaf sugar δ²H. Assimilation (A_net) and dark respiration (R_dark) were found to have a significant impact on the explanatory power of the models (model 32: adj. R² = 0.631, p < 0.001, and model 29: adj. R² = 0.657, p < 0.001, respectively), and an even greater impact when considered in interaction with each other (model 23: adj. R² = 0.722, p < 0.001). Including the NSC concentration (% dry mass; model 31: adj. R² = 0.655, p < 0.001) as well as the percentage starch contributes to the total NSC pool (% of NSC; model 22: adj. R² = 0.726, p < 0.001) also significantly contributed to the model performance. The best model combined the interactive terms of temperature with R_dark (<0.01), temperature with A_net (<0.001), and the interaction between the NSC concentration and the percentage starch contributes to the total NSC pool (<0.001) and explained 90.7% of the variation (model 1: adj. R² = 0.884).

Plotting the modeled data of model 1 against the measured δ²H_LS (Fig. 4a) further confirmed the largely good reproducibility of the measured data (R = 0.95, p < 0.001, R² = 0.91, m = 0.9). Furthermore, the model slightly overestimates the lowest and underestimates the highest δ²H_LS. However, the model can reproduce the observed δ²H_LS temperature response well (Fig. 4b; RMSE = 10.13, Mean Absolute Error (MAE) = 8.12).

Fig. 4: Modeled vs. measured ²H and ¹⁸O fractionation.

Comparison and evaluations of the modeled vs. the measured δ²H_LS (a, c) and δ¹⁸O_LS (b, d), using the newly developed generalized additive model (GAM). a, b Scatter plots of the measured vs. the modeled δ²H_LS and δ¹⁸O_LS, respectively. The linear regression line (violet) with the 95% confidence intervals (shaded in pink), illustrates the relationship between measured and modeled values, as well as their R, R², and p-values. The dashed 1:1 line is a reference to assess model performance visually. c, d Temperature response of the modeled (dashed line) and measured (straight line) δ²H_LS and δ¹⁸O_LS, respectively, including significance analysis between the predicted and observed temperature response (i.e., n.s. when the precited temperature response did not significantly differ from observed response).

δ¹⁸O_LW was the baseline for the 19 evaluated GAM for δ¹⁸O_LS (Supplementary Table 7) and explained 64.3% of the observed variation (adj. R² = 0.634, p < 0.001). Again, the best model (adj. R² = 0.876, variation explained = 89.2%, GCV = 1.81, Train RMSE = 1.2, Test RMSE = 1.2, AIC = 411, BIC = 469) included the interaction of temperature and R_dark (p ≤ 0.05), temperature and A_net (p < 0.001), and the interaction between the NSC concentration and the percentage starch contributes to the total NSC pool (<0.001).

The relationships of ε_HA and ε_OA with the carbon metabolism

The apparent autotrophic ²H fractionation ε_HA was positively related to R_dark in 5 of 7 species (not in P. dioica and S. bicolor, Supplementary Fig. 5) and positively related in all C₃ but not the C₄ (S. bicolor) species with the percentage R_dark contributed to P (i.e., R_dark + A_netR_dark⁻¹; Supplementary Fig. 6). Furthermore, ε_HA was positively related to the NSC partitioning into sugar and starch, as expressed in the % that sugar contributes to the total leaf NSC pool (Supplementary Fig. 7), in 5 of the 7 species (not in O. sativa and the C₄ species S. bicolor). The observed relationships strongly vary between the different species, with R² ranging, when significant, from 0.37 to 0.86 with R_dark, from 0.22 to 0.70 with the percentage R_dark contributed to P, and from 0.22 to 0.81 with the % that sugar contributes to the total leaf NSC pool (Supplementary Figs. 5–7).

ε_OA was negatively related to R_dark in 4 of 7 species (not in Q. pubescens, S. hispanica, and S. bicolor, Supplementary Fig. 8) and negatively related in 3 of 7 species (not in O. sativa, Q. pubescens, S. hispanica, and S. bicolor) with the percentage R_dark contributed to P (Supplementary Fig. 9. Lastly, ε_OA was negatively related to the NSC partitioning into sugar and starch (Supplementary Fig. 10) in 3 of the 7 species (H. vulgare, P. dioica, and S. cheesmaniae), and positively related in Q. pubescens. As in ε_HA, the observed relationships strongly vary between the different species, with R² ranging, when significant, from 0.37 to 0.72 with R_dark, from 0.30 to 0.59 with the percentage R_dark contributed to P, and from 0.25 to 0.62 with the % that sugar contributes to the total leaf NSC pool (Supplementary Figs. 8, 9 and 10).

Discussion

Mortality during heatwaves is strongly driven by drought-induced hydraulic failure (Rowland et al.⁵⁵; Kono et al.⁵⁶). However, even without soil or atmospheric drought, elevated temperatures may lead to significant shifts and perturbations in carbohydrate dynamics, as shown in this study (Fig. 2, Supplementary Tables 3 and 4). As a result, plants may become in the long-term more susceptible to other stressors such as drought, pests and diseases. A_net of C₃ plants decreased at temperatures above 30 °C (Fig. 1, Supplementary Table 1) due to a combination increased respiration, a reduced photosystem II functionality, as indicated by increasing NPQ and decreasing ETR and ΦPSII (Supplementary Fig. 1, Supplementary Table 2), and increasing photorespiration (Keys et al.⁵⁷). The strongly increased respiration (Fig. 1) at high temperatures is needed to support higher metabolic rates (Criddle et al.²⁵), thus requiring plants to have more of their carbohydrate pool directly available (Scafaro et al.⁵⁸). This was reflected by a shift towards a higher share of leaf sugars in the total leaf NSC pool with increasing temperature (Fig. 2, Supplementary Tables 3 and 4). Furthermore, these changes in assimilation, respiration, and leaf-level carbohydrate dynamics significantly impacted δ¹⁸O_LS, ε_OA, δ²H_LS, and ε_HA.

While some studies found no changes in NSC concentration and composition in response to temperature only at temperatures bellow 20 °C^42,43,44 and a recent review comparing NSC temperature response across biomes⁴⁵, we found that high temperatures above 30 °C alone can significantly impact NSC concentration and composition. Therefore, further research should investigate the long-term response of plants to prolonged exposure to temperatures above 30 °C. If plants cannot adjust their respiration rates to chronically high temperatures above a certain, likely species-specific threshold, they might ultimately face carbon starvation if the imbalance between assimilation and respiration rates persists for too long. If the observed increase in leaf sugar δ²H, potentially accompanied by a simultaneous decrease in δ¹⁸O, is translated into the isotopic signature of tree-ring cellulose, this could be used as a new isotope proxy to identify trees with an unfavorable carbon balance, which would be in agreement with recent studies on the subject^11,12. For instance, this could be used to identify trees that have shifted out of their climatic niche due to climate change. Thus, additional studies to investigate the here observed leaf-level temperature-induced carbohydrate depletion on a whole plant level could contribute to the understanding of the high-temperature response of plants.

The close positive relationship between δ¹⁸O_LW and δ¹⁸O_LS (Fig. 3), explaining explained 64.3% of the observed variation alone in the tested GAM models (Supplementary Table 7), aligns well with the findings of previous studies (Yakir and DeNiro⁵⁹; Roden et al.⁴; Zech et al.⁶⁰). However, we can further demonstrate that the δ¹⁸O_LS and ε_OA are temperature-dependent in C₃ plants (Supplementary Fig. 3), with a smaller ε_OA at high temperatures for all except Q. pubescens and S. bicolor (Supplementary Fig. 3, Supplementary Table 5). A similar temperature response has been demonstrated for cellulose oxygen isotope ratios of aquatic plants (Sternberg and Ellsworth¹³). Our results show that the temperature dependence of ε_OA for leaf sugar, while small in comparison to ε_HA, must be considered to correctly understand and interpret δ¹⁸O in plant organic matter, the overall temperature effect on ε_OA is −0.11‰ °C⁻¹ (Fig. 3).

In contrast, the temperature response of ε_HA and δ²H_LS are non-linear and not driven by changes in leaf water (Fig. 3, Supplementary Fig. 2, Supplementary Table S6), indicating the interplay of different biochemical processes responsible for the apparent photosynthetic ²H fractionation. While CO₂ fixation produces sugar highly depleted in ²H (Zhang et al.⁶¹), higher respiration rates have been found to be related to higher δ²H_LS (Holloway-Phillips et al.⁶²; Lehmann et al.¹⁰). As the isotopic composition of sucrose derived from transitory starch is the same as that of sucrose synthesized during the day⁴⁶, it is unlikely that sugar:starch partitioning during the daytime is responsible for the observed changes in the leaf sugar isotopic composition. An explanation for this process could be a preferential usage of glyceraldehyde 3-phosphate (GAP) or sugar with ¹H instead of ²H for respiration, similar to the equilibrium tritium isotope effect that has been observed between glucose and hexokinase (Lewis and Schramm⁶³), the first enzyme involved in glycolysis. However, other enzymatic reactions, such as glucose-6-phosphate dehydrogenase⁴⁷ or phosphoglucose isomerase^17,48, among others, could be responsible for the observed respiratory ²H enrichment in leaf sugar. The observed ²H enrichment in tree-ring cellulose during a reduced or negative carbon balance of trees in response to defoliation or a reduced water access^11,12,49 supports the findings of this study, where an increase in δ²H in carbohydrates is related to a reduced or overall negative C balance. However, also other reactions such as sugar derived from the photorespiratory pathway might influence the observed ²H enrichment^50,51.

In Fig. 5, we summarize the suggested main links between a leaf-level carbon dynamic with the ¹⁸O and ²H fractionation: (1) The initial photosynthetic ¹⁸O fractionation ε_OA from H₂O to leaf sugar leads to an ¹⁸O-enriched leaf sugar, without species-specific differences. This photosynthetic ¹⁸O enrichment is likely reduced under elevated temperatures. (2) Increased isotopic exchange reactions between leaf sugar and leaf water under elevated metabolic rates further alter the initial leaf sugar δ¹⁸O, leading to leaf sugar less ¹⁸O enriched and therefore (3) alter the apparent autotrophic ¹⁸O fractionation, ε_OA*. (4) The strong photosynthetic ²H fractionation ε_HA from H₂O to leaf sugar leads to a strong ²H depletion compared to the leaf water, with a large amplitude between species, which must be already present in glyceraldehyde 3-phosphate (G3P). During daytime respiration (R), (5) isotopically lighter G3P might be preferably used as a substrate, (6a) leading to a ²H enriched leaf sugar which is synthetized from the remaining ²H enriched G3P pool, leading therefore an altered apparent autotrophic ²H fractionation, ε_HA*. However, (6b) glycolysis might further contribute to a further ²H enrichment in the remaining sugar pool, especially in times of high respiratory demand. (7) This “processed” leaf sugar is eventually exported to the long-term sinks, where its isotopic signature, after being further altered during the transport and in the sink tissues by various processes, influences the δ¹⁸O and δ²H of the tree-ring cellulose.

Fig. 5: The apparent autotrophic ²H and ¹⁸O fractionation.

The leaf-level processes leading to the apparent a ¹⁸O and b ²H fractionation, ε_OA* and ε_HA*, respectively, in leaf sugar.

In conclusion, we can demonstrate that air temperatures above 30 °C substantially alter the leaf-level carbon (i.e., A_net and R_dark) and carbohydrate dynamics C₃ plants. This was associated with a reduction in the total NSC pool size and a higher proportion of the carbohydrates being stored as directly metabolically available sugars and less as starch reserves. This was reflected in increased δ²H and a decreased δ¹⁸O in leaf sugar. The C₄ plant maintained low respiration and high assimilation rates even at temperatures above 35 °C, thus, no temperature response of leaf sugar δ²H was observed. Finally, by using the collected data on gas exchange and NSC concentrationin generalized additive models, we were able to precisely reconstruct the measured O and H isotopic composition of the leaf sugar, and therefore confirm the close relationship between plant carbohydrate metabolism and stable isotope fractionation beyond source water effects, especially for δ²H in leaf sugar. We speculate that the increased respiration rate likely caused a ²H enrichment in the remaining sugar pool, leading to a reduced apparent autotrophic ²H fractionation (ε_HA) from leaf water to leaf sugar when temperatures increased, with leaf water having no impact on ε_HA. Furthermore, since the partitioning between sugar and starch changed in S. bicolor with rising temperatures but ε_HA did not, while the partitioning in O. sativa did not change but ε_HA did, we assume that the partitioning process is likely not involved in the observed changes in ε_HA. This is supported by the finding that the δ²H of sugar deriving from transitory starch during the nighttime does not differ from the δ²H of sugar derifing from G3P during the daytime⁴⁶. However, the metabolic changes for the observed respiratory ²H enrichment of the leaf sugar, at least in some species, coincide with changes in the NSC partitioning. On the other hand, the observed relative depletion of leaf sugar δ¹⁸O with increasing temperatures was caused by a combination of a decreasing leaf water δ¹⁸O and a reduced apparent ¹⁸O fractionation (ε_OA) from leaf water to leaf sugar. Therefore, increasing oxygen isotopic exchange reactions between leaf water and leaf sugar are likely the main driver behind leaf sugar δ¹⁸O.

Position-specific δ²H analysis of sucrose^17,47 and other metabolites will be needed to identify the exact biochemical reactions for the observed respiratory ²H enrichment and ¹⁸O depletion with increasing temperatures in leaf sugar, and how they are later transferred into the tree-ring archive.

Methods

Experimental design and plant growing conditions

To isolate the effects of rising temperature under a constant VPD on leaf physiology, metabolism, and the corresponding triple isotope fractionation, we established a specific experimental and sampling design. We selected six C₃ and one C₄ plant species, with different biochemical and anatomical features as well as temperature adaptations. For the C₃ species, we selected two trees, Quercus pubescens WILLD. and Phytolacca dioica L.; two grasses, Hordeum vulgare L. and Oryza sativa L.; and two forbs, Salvia hispanica L. and Solanum cheesmaniae (RILEY) FOSBURG. For the C₄ plant, we selected the grass Sorghum bicolor (L.) MOENCH. With this species selection, we aimed to make the results of this study relevant to a broad field of plant sciences, including plant ecophysiology, forestry and forest ecology, as well as agriculture. Starting in November 2021, we grew replicates of plants (n = 3 for Quercus pubescens, Phytolacca dioica, Solanum cheesmaniae, Sorghum bicolor, n = 50 for Hordeum vulgare and Oryza sativa) for 2 to 3 months in a climate chamber (Plant Growth Chamber PGR15, Controlled Environments Limited (CONVIRON), Winnipeg, Manitoba, Canada) at the Swiss Federal Institute for Forest, Snow, and Landscape Research WSL, at a temperature of 25 °C, a VPD of 1 kPa, and with a photosynthetic active radiation (PAR) of 800 µmol of photons m⁻² s⁻¹ for 18 h a day, followed by 6 h of nighttime in complete darkness at a temperature of 20 °C. The low VPD was chosen to avoid plant stress due to VPD while changing temperature and to ensure saturated intercellular airspaces in the leaves during gas exchange measurements (Diao et al.⁶⁴). After the initial growth period, the actual treatment period of seven weeks started. To reduce the pool of old leaf NSC between each of the seven temperature cycles, plants were kept in the dark at 20 °C for 48 h, similar to (Wieloch et al.¹⁷). This depletion of old NSC was used to obtain an unadulterated stable isotope (²H, ¹⁸O) signal, reflecting the plant physiological conditions at the respective temperature, and thus avoiding autocorrelation. During the start of each week, we exposed the plants for 5 days to 18 h of a constant daytime temperature, starting at 10 °C and subsequently increasing to 40 °C in 5 °C steps every week (Fig. 1a). This allowed the plants to acclimate to each of the studied temperatures. The nighttime temperature for the daytime 10, 15, and 20 °C treatments was the same as the daytime temperature and a VPD of 1 kPa to avoid chilling damage. Nighttime temperatures for 25, 30, 35, and 40 °C were all set to 20 °C and a VPD of 1 kPa to enable the plants to recover their photosynthetic machinery overnight. After four days of treatment, we sampled leaves for non-structural carbohydrates (NSC) and stable isotope analysis in the early afternoon. After 5 days, we conducted gas exchange and fluorescence measurements. The separation of leaf sampling and gas exchange measurements was done to avoid any influence of introduced unstable conditions during gas exchange measurements. At 40 °C, one of the three replicates of S. cheesmaniae and about two-thirds of the H. vulgare plants died.

Plant physiological measurements

After 5 days of exposure to each temperature, one leaf per plant was dark adapted by gently folding aluminum foil around it for at least 20 min. After that, dark respiration (R_dark) and dark-adapted fluorescence were measured using a Li-6800 (LI-COR Biosciences, Lincoln, NE, USA) at the same conditions (CO₂, RH, and temperature) as present in the climate chamber but without light. After that, photosynthetic active radiation (PAR) of the LI-6800 was set to the value of the climate chamber (CO₂, RH, and temperature still the same as in the climate chamber), and a light-adapted leaf of the same plant in close proximity was fixed into the measuring chamber. The leaf and the chamber were allowed to equilibrate for 15 to 20 min until gas exchange reached steady state before measuring A_net, stomatal conductance (g_s), as well as light-adapted fluorescence. Gross photosynthesis (P) was calculated as the sum of A_net and R_dark. With the dark- and light-adapted fluorescence measurements, the LI-6800 automatically calculated the non-photochemical quenching (NPQ), the photosynthetic efficiency of photosystem II (Fv/Fm), the quantum yield of photosystem II (ΦPSII), and the electron transport rate of photosystem II (ETR). All measured values can be found in Supplementary Tables 1 and 2.

Sampling of leaf material

For each temperature step, three samples each consist of several light-exposed leaves were collected from each plant in the early afternoon using scissors. Leaf material was sampled in excess to make sure there was enough plant material and water (>2 mL of water for all samples) to avoid methodological bias during water extraction (Diao et al.⁶⁵). The fully developed leaves were immediately transferred into individual gas-tight 12 ml glass vials (Prod. No. 738W, Exetainer, Labco, Lampeter, UK, stored on dry ice, and transferred in a −20 °C freezer until leaf water extraction.

Extraction of leaf water and sugars

Leaf water was cryogenically extracted using a hot water bath at 80 °C and liquid nitrogen following established protocols (West et al.⁵²; Diao et al.⁶⁵), and a schematic overview of the extraction unit can be found in Diao et al.⁶⁵. In short, glass vials with the frozen leaf samples, which were blocked by PP fiber filters (Nozzle protection filter, Socorex Isba SA, Ecublens, Switzerland), were attached to the extraction unit. The sample vials were submerged in a water bath at 80 °C, while u-shaped glass sample collection tubes were submerged in liquid nitrogen to instantly trap arriving water vapor. The extraction procedure was conducted under vacuum (<0.02 mbar) for 2 h^52,53. Afterwards, when the extraction was complete, the extraction line was filled with dry nitrogen gas to ambient pressure. The u-shaped glass sample collection tubes were then taken out of the extraction line and closed with rubber plugs. The thawed water samples were transferred to 2 mL glass vials (Infochroma AG, Goldau, Switzerland) with pipettes and stored in glass vials at −20 °C until isotope analysis.

After the water extraction, the dried leaf material was ground (MM400, Retsch, Haan, Germany), and the bulk leaf sugar fraction for isotope analysis was then extracted from 100 mg of leaf powder following established protocols (Rinne et al.⁶⁶; Lehmann et al.⁶⁷). First, the ground leaf material was mixed with deionized water in a 2 ml reaction vial and the water-soluble content was extracted at 85 °C for 30 min. Leaf sugars were then purified from the water-soluble content using ion exchange cartridges (OnGuard II A, H and P, Dionex, Thermo Fisher Scientific, Bremen, Germany). Finally, leaf sugar material was acquired by freeze-drying the purified sugar solutions.

δ²H and δ¹⁸O analyses of leaf water

The δ²H and δ¹⁸O of water (δ²H_LW and δ¹⁸O_LW) samples were measured with a high-temperature conversion elemental analyzer coupled to a DeltaPlus XP isotope ratio mass spectrometer (TC/EA-IRMS; Finnigan MAT, Bremen, Germany), and the isotope ratios are reported in per mille (‰) relative to Vienna Standard Mean Ocean Water (VSMOW). Calibration was done using a range of labortory standard waters with different δ²H and δ¹⁸O values (B2193, δ²H = −98.32‰, δ¹⁸O = −12.34‰; and Fiji Artesian Water, δ²H = −42.85‰, δ¹⁸O = −6.41‰), respectively, resulting in a precision of 2‰ for δ²H and 0.3‰ for δ¹⁸O. All the obtained values can be found in Supplementary Table 5.

δ²H and δ¹⁸O analyses of sugars using a hot water vapor equilibration method

The here used procedure originates mainly from the description in Schuler et al.⁸. δ²H of sugars were analyzed according to the previously developed hot water vapor equilibration method (Schuler et al.⁶⁸). Subsequently, as described by Schuler et al.⁸, dry sugar samples were dissolved in water, with a target concentration of 1 mg sugar per 20 µL water. Two identical sets of each sugar sample, with 1 mg sample material each, were prepared by pipetting 20 µL sugar solution into pre-weighed 5 × 9 mm silver foil capsules (Prod. No. SA76981106, Säntis, Switzerland). Sugar samples for δ¹⁸O measurements were prepared by transferring 20 µL sugar solution of the same solution into pre-weighed 3.3 × 5 mm silver foil capsules (Prod. No. SA76980506, Säntis). All samples were then frozen at −20 °C, freeze-dried with a condenser temperature of −50 °C, and the duplicates for δ²H measurements were packed into a second 5 × 9 mm silver foil capsule. Sugar samples were stored in a desiccator at low relative humidity (2–5%) until δ²H and δ¹⁸O measurements.

For the δ²H measurements, the sets of duplicates were then equilibrated with hot water vapor by evaporating two isotopically distinct waters (δ²H water 1 = −160‰ and δ²H water 2 = −428‰) at 130 °C (Schuler et al.⁶⁸). After 2 h, the samples were dried with dry nitrogen gas (N25.0, Prod. No. 2220912, PanGas AG, Dagmersellen, Switzerland) for 2 h at 130 °C. After that, they were immediately transferred into a Zero Blank Autosampler (N.C. Technologies S.r.l., Milano, Italy), which was installed on a sample port of a high-temperature elemental analyzer system. The latter was coupled via a ConFlo III referencing interface to a Delta^Plus XP IRMS (TC/EA-IRMS, Finnigan MAT, Bremen, Germany). The autosampler was evacuated to 0.01 mbar and filled with dry helium gas. The samples were pyrolysed in a reactor according to Gehre et al.⁶⁹, and carried in a flow of dry helium (150 ml min⁻¹) to the IRMS. Raw δ²H values were offset corrected using polyethylene foil standards (IAEA-CH-7 polyethylene foil, International Atomic Energy Agency, Vienna, Austria; SD < 0.7‰ within one run). δ¹⁸O measurements were done according to established protocols (Weigt et al.⁷⁰; Lehmann et al.⁶⁷) with a PYRO cube (Elementar, Hanau, Germany, precision ~0.3‰).

Calculation of the non-exchangeable hydrogen isotope ratio (δ²H_ne), ε_OA, ε_HA

The here used procedure originates mainly from the description in Schuler et al.⁸. All isotope ratios (δ) were calculated as given in Eq. 1 (Coplen⁷¹):

$${\rm{\delta }}=\,\frac{{{\rm{R}}}_{{\rm{Sample}}}-{{\rm{R}}}_{{\rm{Standard}}}}{{{\rm{R}}}_{{\rm{Standard}}}}$$

(1)

where R = ²H/¹H of the sample (R_Sample) and of Vienna Standard Mean Ocean Water (VSMOW2; R_Standard) as the standard defining the international isotope scale. To express the resulting δ in permil (‰), results were multiplied by 1000.

According to Filot et al.⁷², the %-proportion of exchanged hydrogen during the equilibrations (x_e, Eq. 2) can be calculated as:

$${{\rm{x}}}_{{\rm{e}}}=\,\frac{{{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{e}}1}-\,{{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{e}}2}\,}{{{\rm{\alpha }}}_{{\rm{e}}-{\rm{w}}}\,\cdot \,({{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{w}}1}-\,{{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{w}}2})}\,$$

(2)

where δ²H_e1 and δ²H_e2 are the measured δ²H values of the two equilibrated subsamples, δ²H_w1 and δ²H_w2 are the δ²H values of the two waters used, and α_e-w is the fractionation factor of 1.082, which is the same for sugars and cellulose (Filot et al.⁷²; Schuler et al.⁶⁸). Typical x_e values for pure sugars are between 0.32 and 0.36 (Schuler et al.⁶⁸).

δ²H_ne can then be calculated with Eq. 3 using one of the two equilibrations (equilibration one in this example, δ²H_e1 and δ²H_w1):

$${{\rm{\delta }}}^{2}{{\rm{H}}}_{\mathrm{ne}}=\frac{{{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{e}}1}\,-\,{{\rm{x}}}_{{\rm{e}}}\,\cdot \,{{\rm{\alpha }}}_{{\rm{e}}-{\rm{w}}}\,\cdot \,{{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{w}}1\,}-\,1000\,\cdot \,{{\rm{x}}}_{{\rm{e}}}\,\cdot \,({{\rm{\alpha }}}_{{\rm{e}}-{\rm{w}}}\,-\,1)\,}{1-{{\rm{x}}}_{{\rm{e}}}}$$

(3)

Three sucrose samples for the equilibrations of leaf sugars and three cellulose samples for the equilibrations of the twig xylem cellulose, each measured in triplicates, were used as internal reference material to calibrate the results. For the sake of simplicity, δ²H has been used throughout the manuscript instead of δ²H_ne.

The apparent autotrophic fractionation factors between precursor and product (¹⁸O = ε_OA, and ²H = ε_HA) were calculated with Eq. 4 and Eq. 5, respectively:

$${{\rm{\varepsilon }}}_{{\rm{OA}}}={{\rm{\delta }}}^{18}{{\rm{O}}}_{{\rm{leaf\; sugar}}}-{{\rm{\delta }}}^{18}{{\rm{O}}}_{{\rm{leaf\; water}}}$$

(4)

$${{\rm{\varepsilon }}}_{{\rm{HA}}}={{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{leaf\; sugar}}}-{{\rm{\delta }}}^{2}{{\rm{H}}}_{{\rm{leaf\; water}}}$$

(5)

As in Schuler et al.⁸, the two biological fractionation factors ε_HA and ε_OA were expressed as the actual difference between the δ¹⁸O, and δ²H of leaf sugars and the δ¹⁸O and δ²H of leaf water, respectively. All the obtained values can be found in Supplementary Table 5.

Leaf-level non-structural carbohydrates (NSC) analysis

The leaf material for NSC analysis derives from the material which was dried during the cryogenic water extraction at 80 °C. Then, leaves were ground in fine powder and NSC concentrations were measured following previously established protocols (Hoch et al.⁷³; Schönbeck et al.⁷⁴). Ten to twelve mg of finely ground leaf material were heated in 2 mL distilled water for 30 min. An aliquot of 200 µL was treated with invertase from baker’s yeast (S. cerevisiae, Sigma-Aldrich Chemie GmbH, Germany) for an hour to degrade sucrose and convert fructose into glucose. The sugar concentration was determined at 340 nm in a 96-well plate spectrophotometer (Thermo Fisher Scientific Multiskan GO, Finland) after an enzymatic conversion to gluconate-6-phosphate of about 35 min, using glucose Assay Reagent (Sigma-Aldrich Chemie GmbH, Germany) and Isomerase from baker’s yeast (S. cerevisiae, Sigma-Aldrich Chemie GmbH, Germany). The total amount of NSC was measured by taking an aliquot of 500 µL of the extract (including starch and sugar) and treated for 15 h at 49 °C with Amyloglucosidase from Aspergillus niger (Sigma-Aldrich Chemie GmbH, Germany) to digest starch into glucose. Total glucose (corresponding to total NSC concentration) was determined using a spectrophotometer, as explained above. The starch concentration was calculated as the total NSC subtracted by the sugar concentration. Standard solutions, including pure starch, glucose, fructose, sucrose, and standard plant powder (Orchard leaves, Leco, USA), were used as references for the comparison and reproducibility of the results between runs. All the obtained values can be found in Supplementary Table 3.

Statistical analyses

Statistical analyses were performed using R version 4.1.2 (R.Core.Team⁷⁵). Linear and polynomial models, Mann-Whitney-U-Test, and Welch’s t-test implemented using basic R and ggplot2 (Wickham⁷⁶) were used to determine the leaf-level physiological and NSC temperature response. The final assembly of the graphs was done using the R package patchwork (Pedersen⁷⁷). Least Absolute Shrinkage and Selection Operator (Lasso) regressions to identify important variables were performed using the R package glmnet (Tay et al.⁷⁸). Subsequently, to analyze the relationship between isotope fractionation and leaf physiological variables, we developed generalized additive models (GAMs) using the R package mgcv (Wood⁷⁹), which allowed us to capture any non-linear relationships and thus identify the main drivers behind hydrogen and oxygen isotope fractionation. In total, we tested 34 different models with increasing complexity for δ²H and 19 for δ¹⁸O in leaf sugar. The models included smooth terms for the variables of interest, and interactions were modeled using tensor product smooths and were trained on 80% and tested on 20% of the dataset. Whether the simulated leaf sugar isotope temperature response differed significantly from the measured leaf sugar isotope temperature response by testing the significance of the interaction term in an ANOVA. Model visualization and diagnostics were performed using ggplot2, ggplotify, and cowplot to combine plots. Lastly, we tested the relationships between ε_HA and ε_OA, respectively, with R_dark, the % R_dark contributes to P (R_dark + A_net), as well as the partitioning between sugar and starch using Pearson’s correlation coefficients (cor.test from dplyr⁵⁴) within each species.