Acclimation of Foliar Respiration and Photosynthesis in Response to Experimental Warming in a Temperate Steppe in Northern China

Background Thermal acclimation of foliar respiration and photosynthesis is critical for projection of changes in carbon exchange of terrestrial ecosystems under global warming. Methodology/Principal Findings A field manipulative experiment was conducted to elevate foliar temperature (T leaf) by 2.07°C in a temperate steppe in northern China. R d/T leaf curves (responses of dark respiration to T leaf), A n/T leaf curves (responses of light-saturated net CO2 assimilation rates to T leaf), responses of biochemical limitations and diffusion limitations in gross CO2 assimilation rates (A g) to T leaf, and foliar nitrogen (N) concentration in Stipa krylovii Roshev. were measured in 2010 (a dry year) and 2011 (a wet year). Significant thermal acclimation of R d to 6-year experimental warming was found. However, A n had a limited ability to acclimate to a warmer climate regime. Thermal acclimation of R d was associated with not only the direct effects of warming, but also the changes in foliar N concentration induced by warming. Conclusions/Significance Warming decreased the temperature sensitivity (Q 10) of the response of R d/A g ratio to T leaf. Our findings may have important implications for improving ecosystem models in simulating carbon cycles and advancing understanding on the interactions between climate change and ecosystem functions.


Introduction
The balance between respiration and photosynthesis is critical to the exchange of carbon between the atmosphere and the terrestrial biosphere [1][2][3]. Instantaneous increases in foliar temperature (T leaf ) typically result in an increase in respiration/ photosynthesis (R/A) ratio because the response of respiration to T leaf normally follows an approximate exponential-type curve (at moderate temperatures) while the response of photosynthesis to T leaf often bears a bell-shaped curve [i.e. the thermal optimum (T opt ) of respiration is higher than that of photosynthesis] [4,5]. In contrast, long-term warming experiments have suggested that R/A ratio is often conservative to changes in growth temperature (T growth ) through acclimation, the metabolic adjustment for compensating changes in T growth [6][7][8]. Acclimation could occur via suppression of respiration in response to changes in foliar carbohydrate supplies [4,9]. The thermal acclimation of respiration and photosynthesis is associated with multitudes of signal cascades and networks, which involves the reallocation of resources to achieve and maintain not only optimal R/A ratio but also protective strategies under sustained warming as projected by global climate models [10][11][12]. However, the mechanisms of thermal acclimation of respiration and photosynthesis to climate warming are far from clear, especially in natural ecosystems.
The acclimation of foliar respiration to warmer T growth has been found in numerous studies [8,[13][14][15][16][17], which may also be associated with plant developmental stage and other abiotic factors, such as drought and nutrient availability [18][19][20][21]. Thermal acclimation of respiration might occur via changes in the temperature sensitivity, Q 10 , or the basal respiration, R 10 (respiration at a reference temperature, such as 10uC) [11]. Altered Q 10 partially reflects temperature-mediated changes in energy demand and/or available substrates [1,17,20] whereas changes in R 10 may be associated with temperature-mediated changes in respiratory capacity, reflecting changes in mitochondrial abundance, structure and/or protein composition [22][23][24]. As a result, thermal acclimation of respiration may enhance plant net carbon assimilation by reducing carbon loss under warmer T growth while maintaining basal rates of respiration in colder T growth for subsequent recovery [12,20,25,26].
The thermal acclimation of the foliar net CO 2 assimilation rate (A n ) may involve three primary sets of processes that control the A n /T leaf curves (response of A n versus T leaf ), namely respiratory, biochemical and stomatal processes [27]. First, A n is the difference between gross CO 2 assimilation rate (A g ) and foliar dark respiration (R d ), A n = A g -R d , which requires the decoupling of the two processes because A g and R d feature different thermal dynamic properties and thus involve different thermal acclimation processes [28]. This could result in a shift in T opt and a change in the shape of the A n /T leaf curve. Therefore, R d must be evaluated separately and factored out to understand the acclimation mechanisms of A g in response to global warming [3,18,29]. Second, the acclimation of A g to warmer T growth deals with the changes in Rubisco activity [29][30][31][32][33] and electronic transport processes [34] where T growth affects the thermal dependence of various enzymes in the dark and light reactions [35,36]. Therefore, the temperature sensitivity of the maximum rate of Rubisco carboxylation (V cmax ) and the maximum rate of photosynthetic electron transport (J max ) are associated with the acclimation of A g [36,37]. In addition, the change in the balance between carboxylation and regeneration of RuBP, indicated by J max /V cmax ratio, may also result in the shift of T opt of A g due to nitrogen (N) partitioning in the photosynthetic apparatus [3,31,38,39]. Finally, the temperature-dependent diffusion processes of CO 2 to chloroplasts, such as stomatal conductance (g s ) and mesophyll conductance (g m ), can also affect the thermal acclimation of photosynthesis [36,40]. Kirschbaum and Farquhar [41] showed that higher conductance could cause an increase of CO 2 concentrations in the carboxylation site (C c ) and then resulted in a shift in limitation of A g from Rubisco to electron transport capacity. Since T opt of electron transport-limited A g is higher than that of Rubisco-limited A g , T opt of A g was increased (0.05uC per 1 mmol mol 21 CO 2 ) [36].
Stipa krylovii Roshev. is a keystone species in the temperate steppe in northern China [42,43]. Climate models predict this region will be 4uC warmer by 2100, which may have severe impacts on Stipa krylovii Roshev. [44]. Examining the respiration and photosynthesis of this species is critical to the steppe productivity and the carbon cycle of the ecosystem. The objectives of the current study are to examine: (1) the acclimation capacity of respiration and photosynthesis to experimental warming under field conditions, and (2) the homeostasis of respiration/photosynthesis ratio in response to experimental warming in the steppe ecosystem.

Site Description
The research site (42u029 N, 116u179 E, 1324 m a.s.l.) is a typical temperate steppe located in Duolun County, Inner Mongolia Autonomous Region, China. The experiment has received the permits for the field study from the land owner, Institution of Botany, Chinese Academy of Sciences. The mean Figure 1. Layout of the experiment plots in a temperate steppe in northern China. Infrared heaters were suspended as the heating sources at the warming plots while 'dummy' heaters were suspended to simulate shading effects of the infrared heater at the control plots. doi:10.1371/journal.pone.0056482.g001 annual temperature (MAT) is 2.1uC, with monthly mean temperature ranging from 217.5uC in January to 18.9uC in July. The mean annual precipitation (MAP) is approximately 385 mm with approximately 85% falling from May to September. The soils are chestnut (Chinese classification system) or Haplic Calcisols (FAO classification system), with 62.8% sand, 20.3% silt, and 17.0% clay respectively. The soils are characterized as sandy, slightly alkaline and nutrient poor with pH values around 7.7 and bulk density of 1.3 g cm 23 and soil total organic C and N concentrations of 16.1 and 1.5 g kg 21 respectively. The plant communities in the temperate steppe are dominated by Stipa krylovii Roshev., Artemisia frigid Willd., Potentilla acaulis L., Cleistogenes squarrosa (Trin.) Keng., Allium bidentatum Fisch. ex Prokh., and Agropyroncristatum (L.) Gaertn.

Warming Experiment
The warming experiment was initiated in April 2006 with infrared heaters (MSR-2420, Kalglo Electronics Inc., USA; radiation output is approximately 1600 W) as the heating source ( Fig. 1). Briefly, an infrared heater of 1.65 m in length was suspended at 2.25 m above the ground in each warming plot which features a dimension of 364 m. A reflector associated with the heater can be adjusted so as to generate an evenly distributed radiant input to the plant canopy. In the control plots, a 'dummy' heater with the same shape and size was suspended at the same height to simulate shading effects of the infrared radiator. The effects of warming on T leaf were measured using a portable infrared thermometer (FLUKE 574, Fluke Inc., USA). The mean daytime T leaf in the warming plots was increased by 2.07uC compared to the control plots. The warming experiment was designed for long-term simulation of global change and it featured a complete random block design with multiple treatments (day warming, night warming, diel warming, and N addition) and six replicates. We took advantage of this multi-factor experiment by selecting the diel warming and control plots with all the other factors kept at control levels. The details of the experiment can be found in Wan et al. [44] and Xia et al. [45].

Gas Exchange Measurements
We measured foliar gas exchange using a portable photosynthesis system (LI-6400, LI-COR Inc., USA) in the middle of the growing seasons (late July to early August) in 2010 and 2011 (Fig. 2) to remove the effect of seasonal changes in photosynthetic and respiratory acclimation in Stipa krylovii Roshev. [19]. Four individuals (one individual per plot) were measured in each treatment. Eight days were required to complete all field measurements each year. Light, T leaf , humidity, and CO 2  concentration were independently controlled in a 263 cm cuvette. Given the T leaf control capacity is limited (within 66uC) with the factory setup of the LI-6400 system, we modified the T leaf control system by adding metal blocks with water channels to heat or cool the peltiers, thermoelectric cooling elements. The water channels were connected to a heating/cooling water bath whose temperature was controlled by adding hot water or ice. This modification allows holding T leaf at any level between 10 and 40uC during the summer growing season in the steppe.
The photosynthetically active photon flux density (PPFD) was provided by the red/blue LED light source built in the foliar cuvette calibrated against an internal photodiode (LI-6400-02B, LI-COR Inc.). The vapor pressure deficit (VPD) in the foliar cuvette was controlled by passing the air entering the cuvette through either anhydrous calcium sulfate for the lower T leaf when humidity was high or bubbling air via water at higher T leaf when the air was dry. CO 2 concentrations in the cuvette were controlled using an injector system (LI-6400-01, LI-COR Inc.) which functions with a CO 2 mixer and compressed CO 2 cartridges. Cuvette was sealed with plasticine to prevent leakage. Potential leakage of CO 2 out and into the empty cuvette was determined for each concentration and used to correct the measured foliar fluxes with the equations provided by von Caemmerer and Farquhar [46] and Galmés et al. [47]. The gas exchange system was zeroed using H 2 O and CO 2 free air every day.
Typical A n /C i curves (A n versus calculated intercellular CO 2 concentrations, C i ) were measured at T leaf changing from 10 to 40uC with 5uC increments each. We started with the A n /C i curves at low T leaf (10uC) in the morning around 7:00 am finished at high T leaf around noon. As to the problem of co-variance between the daily cycle and temperature, Luo et al. [48] and Way and Sage [3] suggested that the observed responses in the biochemical parameters resulted mainly from changes in temperature rather than changes in time of day. It usually took c. 5 min for T leaf to reach stability at each step change in temperature. Photosynthesis was induced for 10 min in saturating PPFD (1500 mmol photons m 22 s 21 ) and at ambient CO 2 concentration (C a ) of 380 ppmv. Measurements were made at saturating light (1500 mmol photons m 22 s 21 ), and a leaf VPD between 0.5 and 2.0 kPa, except for 40uC where the VPD was 4.560.05 kPa. A n was measured at cuvette CO 2 partial pressures between 50 and 1200 ppmv CO 2 . The C a was lowered stepwise from 380 to 50 ppmv and then increased again from 380 to 1200 ppmv with the total of 9 points. In total, 112 A n /C i curves were measured and used for the analysis of physiological parameters in this study. A n /T leaf curves (response of light-saturated A n at 380 ppmv versus T leaf ) were obtained based on the A n /C i curves measured from 10 to 40uC. R d was measured by turning off the LED light source for at least 5 minutes in the cuvette after each A n /C i curve was accomplished [49]. All other conditions were the same as A n /C i curve measurements. Measurements of R d on previously illuminated leaves were performed after a period of darkness in order to avoid light-enhanced dark respiration (LEDR) [13,18]. Five data points of R d were logged at a 30 s interval and averaged for R d at a given T leaf . A g was calculated by adding R d to A n at each T leaf . Table 1. Results (P-values) of one-way ANOVA on the effects of warming on the responses of A n (the net CO 2 assimilation rate), R d (dark respiration), A g (the gross CO 2 assimilation rate), V cmax (the maximum rate of Rubisco carboxylation) and J max (the maximum rate of photosynthetic electron transport) expressed per unit foliar area and nitrogen to instantaneous change (10-40uC within a 5 h period) in T leaf (foliar temperature) in 2010 and 2011. Estimation of V cmax , J max , TPU and g m A n /C c curves (A n versus chloroplastic CO 2 concentration) were fitted to estimate V cmax , J max , TPU (triose-phosphate utilization) and g m . The spreadsheet-based software of Sharkey et al. [50] was modified (Appendix S1) to fit the A n /C c curve by fixing the R d value which was measured following the A n /C i curve. This modification will improve the model performance by reducing the number of estimated parameters and thus decreasing the degree of freedom in fitting the model. As in the original software the optimum of V cmax , J max , TPU and g m was obtained by minimizing the root mean square error (RMSE) of each curve [51,52].

Estimation of Dependence of Reaction Rates on Temperature
The responses of R d and V cmax to T leaf were fitted to a nonpeaked model, following Harley et al. [53], due to the fact that the deactivation of R d and V cmax was not observed in our study: where c is a scaling constant, DH a is the activation energy, R is the molar gas constant (0.008314 kJ K 21 mol 21 ) and T k is the absolute T leaf (K) [54]. Q 10 of R d and V cmax were modeled using the following general function: where ref 10 is the estimated basal rate at the reference temperature of 10uC, and T leaf is the leaf temperature (uC). The responses of A n , A g and J max to T leaf were fitted using a peak model in view that the deactivation at high T leaf was substantial: where DH d is a term for deactivation and DS is an entropy term [54,55]. The second derivative of Eqn 3 shows that Topt can be calculated [56] as follows if the parameter includes a peak: Table 2. Warming effects on the responses of A n (the net CO 2 assimilation rate), R d (dark respiration), A g (the gross CO 2 assimilation rate), V cmax (the maximum rate of Rubisco carboxylation) and J max (the maximum rate of photosynthetic electron transport) expressed per unit foliar area and nitrogen to instantaneous change (10-40uC within a 5 h period) in T leaf (foliar temperature) in the dry growing season (2010).
Estimation of Biochemical Limitations to Photosynthesis Temperature dependence of A g limited by RuBP carboxylation (A c ), RuBP regeneration (A j ) and TPU (A p ) were reconstructed as follows: where V cmax , J max and TPU were derived from fitted kinetic parameters (c, DH a , DH d and DS) in our study, K c , K o and ? ? were derived from a general set of kinetic parameters in Sharkey et al. [50]. C c was set at 250.8 ppmv in view that the mean C c /C a ratio was 0.66 at ambient CO 2 concentration (380 ppmv) for all the A n / C i curves measured in the current study, O was the partial pressure of oxygen at Rubisco.

Foliar Characteristics
Foliar N concentration on an area basis was determined using the foliage covered in the cuvette during the gas exchange measurements. The foliage samples were first used to measure the leaf area with an area meter (Li-3100, Li-Cor Inc.) and then biomass where the samples were dried at 65uC for 48 h. Then the dry samples were ground to powder for measuring the total C and N concentrations with a CN analyzer (NA Series 2, CE Inc., Germany).

Data Analyses
The raw data from the gas exchange measurements were cleaned and processed in Excel spreadsheets where the non-linear A n /C c curve fitting was performed as in Sharkey et al. [50]. The fitting was improved by fixing R d with the measured value (Appendix S1). Further statistical analyses were conducted using SPSS (version 17.0, SPSS Inc., USA). One-way ANOVA was used to analyze the effects of warming on (1) the foliar chemical properties (C, N, and C/N ratio) and (2) the thermal dynamic Table 3. Warming effects on the responses of A n , R d , A g , V cmax and J max expressed per unit foliar area and nitrogen to instantaneous change (10-40uC within a 5 h period) in T leaf in the wet growing season (2011).      strongly correlated with mean daytime T air . In addition, T growth in the warming plots were approximatively calculated by adding warming effects (2.07uC) to the mean daytime T air during the 5 d prior to gas exchange measurements in each plot.

Microclimate and Experimental Warming
The meteorological data collected at the experimental site showed that the growing season of 2010 was dry while the growing season of 2011 was wet (Fig. 2). The daily mean T air between 1 May, the onset of plant growth, and the time of the field measurements ( (Fig. 2).
The experimental warming significantly increased daytime T leaf by 2.07uC (P,0.001), on average (Fig. 3). Warming increased daytime T growth in the warming plots reaching 28.59 and 23.14uC in 2010 and 2011, respectively. Meanwhile, the daytime T growth in the control plots was only 25.72 and 21.31uC in 2010 and 2011, respectively. The details of the warming effects on microclimate at the study site can be found in Wan et al. [44] and Xia et al. [45].

Respiration
Warming significantly decreased respiratory temperature sensitivity, Q 10 , in both years (both P,0.05) (Fig. 4, Table 1). Q 10 of R d on a foliar area basis decreased from 1.83 in the control plots to 1.66 in the warming plots in 2010 (P = 0.049) ( Table 2) and from 2.19 to 1.81 in 2011 (P = 0.042) (Table 3) Table 2, 3). Warming effects on the R 10 on a foliar N basis were similar to the area-based R d (Fig. 4).

Photosynthesis
The A n /T leaf curves were typically bell-shaped in both warming and control plots (Fig. 4). Warming had little effect on T opt of A n in both years (both P.0.05) ( Table 1). T opt of A n on a foliar area basis was 22.49 and 23.99uC for the control and the warming plots respectively in 2010, and 24.89 and 26.48uC respectively in 2011 (Fig. 4). T opt of A g on a foliar area basis was 22.53 and 24.30uC for the control and the warming plots respectively in 2010 (P = 0.328), and 25.65 and 27.34uC respectively in 2011 (P = 0.637) (Fig. 4). Warming also had little effects on T opt of A n and A g on a foliar N basis in either 2010 or 2011 (all P.0.05) ( Table 1).

Biochemical Limitations to Photosynthesis
The effects of warming on Q 10 of V cmax were not statistically significant between the warming and the control plots in both years (both P.0.05) ( Table 1), but we found a general decreasing trend from the control to warming plots (Fig. 5). Q 10 of V cmax on a foliar area basis was 1.91 and 1.74 for the control and the warming plots respectively in 2010 (P = 0.062), and 1.87 and 1.83 respectively in 2011 (P = 0.779) (Fig. 5, Table 2, 3). Q 10 of V cmax on a foliar N basis was 1.87 and 1.72 for the control and the warming plots respectively in 2010 (P = 0.174), and 1.90 and 1.78 respectively in 2011 (P = 0.668) (Fig. 5, Table 2, 3). The warming effects on Q 10 of J max were not be detected in 2010 or 2011 (both P.0.05) ( Table 1). In addition, the warming effects on the slope and y-intercept of the temperature-response curves for J max /V cmax ratio were not statistically significant (all P.0.05), though the ratio decreased linearly with the T leaf (Fig. 5).

Diffusion Limitations to Photosynthesis
In 2010, a dry year, g s in the warming plots was marginally greater than that in the control plots (P = 0.137), and T opt for g s was about 17.42uC in the warming plots and less than 10uC in the control plots (Fig. 6). The g m in the warming plots was significantly greater than that in the control plots (P,0.001), and T opt for g m appeared at 37.09uC in the warming plots and 27.86uC in the control plots (Fig. 6). C c in the warming plots was approximately 35 ppmv greater than that in the control plots (P,0.001), but C c was independent of T leaf in both the warming and the control plots (both P.0.05) (Fig. 6). Similarly, C c /C a ratio was constant and independent of T leaf in the warming and the control plots (both P.0.05) (Fig. 6). However, experimental warming significantly increased C c /C a ratio in 2010 (P = 0.001) with an average value of 0.70 in the warming plots and 0.61 in the control plots (Fig. 6).
In 2011, a wet year, Warming had little effect on g s and g m (both P.0.05), which resulted in no difference in C c between the warming and the control plots (P = 0.860) (Fig. 6). Experimental warming also had little effect on C c /C a ratio in 2011 (P = 0.447) with an average value of 0.67 in the warming plots and 0.65 in the control plots (Fig. 6).

Foliar Characteristics
Warming marginally decreased foliar N concentration in 2010 (P = 0.063), but significantly increased that in 2011 (P = 0.002) ( Table 4). Warming had little effect on foliar carbon concentration in both years (both P.0.05). Foliar C/N ratio was significantly higher in the warming plots than in the control plots in 2010 (P,0.001) and the opposite was true in 2011 (Table 4).

Acclimation of Respiration
R d was sensitive to T leaf with the R d /T leaf relationship following a typical exponential curve, but warming reduced the magnitude (Fig. 4, Table S1). Our results are consistent with previous studies [18,20,62] that the temperature sensitivity of R d is negatively related to the T growth (Fig. 7). According to the respiratory acclimation mechanisms proposed by Atkin and Tjoelker [11], the temperature-mediated change in Q 10 is determined by the maximum enzyme activity and/or substrate availability [1,17,20]. Earlier results from the same warming experiment confirmed that day warming significantly reduced foliar starch concentrations (-6.1%, P = 0.009), suggesting the reduction in Q 10 in the current study might be attributed to the lower substrate concentrations.
Foliar N concentrations induced by experimental warming in our study may also affect the temperature sensitivity of R d , Q 10 ( Fig. 7). To date, few studies have examined the role of N in the change in Q 10 . Turnbull et al. [63] found that Q 10 of R d for the trees in a temperate rainforest increased with increasing N availability along a soil chronosequence in New Zealand. However, Ow et al. [64] have reported that N had little or no impact on Q 10 of R d when saplings grown at high and low N availabilities were transferred to a different T growth regime. Here, we found a negative correlation between Q 10 of R d and foliar N concentrations (Fig. 7). The detailed mechanisms are not clear, but the confounding effect of foliar N concentrations with other factors, such as temperature and precipitation, may have played an important role in the ''apparent'' Q 10 [11,65,66].
In the current study we found that experimental warming marginally reduced base respiration rate at 10uC (R 10 ) in 2010 but increased that in 2011 ( Table 2, 3). This could have been attributed to the differential responses of foliar N concentration to warming in the two hydrologically contrasting growing seasons. Warming marginally decreased foliar N concentration in the dry growing season (2010), but increased that in the wet growing season (2011) ( Table 4). A growing number of studies [8,14,17], including our current study, have found that foliar N concentration was strongly related to R 10 (Fig. 7). Therefore, we believed that foliar N concentration played an important role in the diverging responses of R 10 to warming in both years.

Acclimation of Photosynthesis
Photosynthesis has long been known to acclimate to prevailing T growth by shifting the T opt [67]. For example, Gunderson et al. [60] have reported that a 3-year warming of 2-4uC has resulted in a higher T opt of A n for five species of deciduous trees. In the current study we found that a 6-year warming of 2.07uC did not resulted in changes in T opt of A n (Fig. 4, Table S1). We also found that there were not statistically significant differences between the shift in T opt of A n and A g in 2010 (P = 0.896) or 2011 (P = 0.984). This suggests that the instantaneous response of photosynthesis was independent of changes in R d .
It has been proposed that the increase in the temperature sensitivity of V cmax , indicated by DH a of V cmax , contributed to the thermal acclimation of photosynthesis to experimental warming [36,61,68]. However, in the current study we found that warming slightly decreased DH a of V cmax (Fig. 5, Table 1). Biochemically, the change in DH a of V cmax is closely related to the temperature dependence of Rubisco activity [69], Rubisco activation status [70,71], dimorphism of Rubisco [31], and the amount of Rubisco [72]. The lower DH a of V cmax obtained from the warming plots indicated that warming slightly decreased the temperature sensitivity of those processes.
Previous studies found that RuBP regeneration processes may play an important role in the thermal acclimation of photosynthesis [34,39,73]. The increase in the thermal stability of photosystem II, indicated by DH a of J max , has been shown to be related to the thermal acclimation of A g to warming [34][35][36]74]. However, in the current study we found only minor response of DH a of J max to warming (Fig. 5, Table 1). This is also confirmed by our results that the RuBP regeneration seldom limited A g (Fig. 8).
A number of studies have reported that the balance between the carboxylation and the regeneration of RuBP, indicated by J max / V cmax ratio, can also affect the thermal acclimation of photosynthesis [39,75]. In our study, the experimental warming had little effect on the linear trend of J max /V cmax ratio to T leaf (Fig. 5). Nevertheless, in this study we found that J max /V cmax ratio declined sharply and linearly with the instantaneous increase in T leaf (Fig. 5). Many ecosystem models, such as Biome-BGC [76], have set J max / V cmax ratio as a constant (2.1) which is independent of T leaf . Wullschleger [77] analyzed 164 A n /C i curves for 109 C 3 plant species which were measured under T leaf ranging from 13 to 35uC and found the average J max /V cmax ratio was 2.1. Others found that J max /V cmax ratio was not a constant instead varying with T leaf through a linear [51,[78][79][80] or nonlinear relationship [81]. Our current results show that the relationship (between J max and V cmax ) itself is highly temperature dependent, suggesting that photosynthesis models have to consider the temperature dependence of J max /V cmax ratio.
In addition to biochemical limitations, the thermal acclimation of photosynthesis may also relate to CO 2 diffusion processes in leaves and chloroplasts, such as g s and g m , because changes in T growth may affect CO 2 diffusivity, solubility, membrane permeability and stomatal movement [82][83][84][85]. Previous studies have found that increasing g s and/or g m can cause the increase of T opt of A n [36,40,41,67,86]. In the current study we found that warming increased g m (Fig. 6) in 2010 which might contribute to the modest variation in T opt of A g in 2010. However, we found smaller increases in g s and g m (Fig. 6) in 2011, which may explain the weaker acclimation in 2011 (Fig. 4). The differential responses of CO 2 diffusion process to warming in the two hydrologically contrasting growing seasons could have been attributed to changes in soil moisture and N availability induced by warming [87]. It is noted that, so far, no consistent conclusions have been achieved on the warming effect on g s and g m . Some researchers found that warming increased g s [39,[88][89][90] and g m [91], and others found warming decreased g s [92] and g m [61], or no effect on g s [93] and g m [40]. Those various studies suggest that other factors, such as warming-induced water depletion and change in N availability, may have interacting effects on responses of CO 2 diffusion process to warming. These results call for multi-factor experiments, such as the combination of warming with water manipulation and fertilization [21], for understanding the mechanisms of thermal acclimation of photosynthesis under future global change.

Balance between Respiration and Photosynthesis
The acclimation of foliar respiration and photosynthesis is also reflected in R/A ratio which indicates the balance between carbon gain, loss and accumulation [1,2]. Our results show that the instantaneous (,5 h) warming at foliage level has non-linearly increased R d /A g ratio, indicating proportionally more carbon loss through R d as T leaf goes up (Fig. 9). However, the 6-year experimental warming has resulted in thermal acclimation of the grasses as evidenced by the decrease of the curvature of the response curve of R d /A g ratio to T leaf (Fig. 9). It is important to note that though the balance between R d and A g was re-established through the thermal acclimation [6,8,9,18], R d /A g ratio was still increasing with T growth in a wet year (Fig. 9). This means that, at foliage level, acclimation can only partially compensate the negative impact from the global warming.

Supporting Information
Table S1 Results (P-values) of two-way ANOVA on the effects of warming, year, and both interactions on the responses of A n (the net CO 2 assimilation rate), R d (dark respiration), A g (the gross CO 2 assimilation rate), V cmax (the maximum rate of Rubisco carboxylation) and J max (the maximum rate of photosynthetic electron transport) expressed per unit foliar area and nitrogen to instantaneous change (10-406C within a 5 h period) in T leaf (foliar temperature). c is a scaling constant, DH a is the activation energy, DH d is a term for deactivation, DS is an entropy term, T opt is the thermal optimum, Q 10 is the temperature sensitivity and ref 10