Rapid stomatal response to fluctuating light: an under-explored mechanism to improve drought tolerance in rice

Additional keywords: abiotic stress, acclimation, drought stress, photosynthetic rate, stomatal conductance, yield.

Light levels in a canopy fluctuate frequently and rarely stay at constant levels. The passing cloud can create intermittent shade to leaves in a canopy (Knapp and Smith 1988). Leaves at the top of a canopy can shade leaves at lower layers of a canopy. When the wind blows, leaves at lower layers of a canopy experience more fluctuating light (FL) levels or sunflecks (Amtmann 1985; Way and Pearcy 2012). Even if leaves in a canopy keep still, the gradual changes of the sun-earth geometry during a day can create FL on leaves at lower layers of a canopy (Myneni and Impens 1985). Although FL may only last from seconds to minutes, FL can contribute up to 20–80% of the total solar energy incident on understory leaves (Chazdon 1988; Pearcy 1990).

Possibly as a result of long-term plant evolution under dynamic light conditions, plants are equipped with a diverse set of mechanisms to gain increased fitness or to avoid damage under FL levels (Stitt and Zhu 2014). For example, under strong light, state transition can occur to transfer energy away from PSII to avoid photo-damage (Dietzel et al. 2011). Chloroplast movement is another commonly observed mechanism plants utilise to avoid photo-damage (Kasahara et al. 2002). In addition to these relatively slow structural or anatomical changes, plants also use an array of biochemical and biophysical mechanisms to cope with FL. Under strong light, the rapid decrease of lumen pH can lead to increased rate constant for heat dissipation, which can help quickly dissipate light energy that are in excess of the capacity of photochemistry (Ort 2001). When FL occur, the rate constant of heat dissipation decreases and helps gain increased canopy light use efficiency (Zhu et al. 2004). The understory plants are capable of accumulating large amount of 3-phosphate glycerate, which can be used to utilise FL (Pearcy 1990).

Recently, another mechanism that may be important for controlling plant light and WUE in the field has been recognised: the rapid rate of responses of stomata conductance to FL levels (Lawson and Blatt 2014). The theory is that when the light level decreases, there will be a potential decrease of WUE if stomatal conductance (g_s) cannot decrease rapidly; however, when the light levels increase abruptly from low to high light levels, if g_s cannot increase rapidly, photosynthetic light use efficiency will be decreased. There are large variations in both the sensitivity and responsiveness of g_s to changing environments among different species (Lawson 2009; Lawson et al. 2003, 2012).

Applying a proper mathematical modelling to qualify dynamic response in g_s to FL is critical to identifying variations among or inter-species. There are several strategies to estimate dynamic changes in g_s to FL treatments, such as piece-wise linear, logistic models, asymmetric sigmoid function and exponential models (Kirschbaum et al. 1988; Zipperlen and Press 1997; Naumburg et al. 2001; Vico et al. 2011; Vialet-Chabrand et al. 2013). Notably, Vico et al. (2011) first introduced two parameters – τ_cl and τ_op – to represent stomatal response delays during closing and opening phase respectively. Vico et al. (2011) obtained these parameters by exponential fitting on 40 dataset from the public resource. In most species, delays in stomatal opening appear to be shorter than delays in closing, that is τ_op < τ_cl (Ooba and Takahashi 2003; Vico et al. 2011). Time scales of stomatal delays in different species also exhibit a large degree of variation. This variation was reported to be commensurate with FL durations (Cardon et al. 1994; Naumburg et al. 2001) and to depend on the time of day and history of FL occurrence (Kaiser and Kappen 1997).

Laboratory studies have shown that leaves with more rapid responses of g_s exerted higher leaf-level water and light use efficiencies (Lawson and Blatt 2014). Considering the high level of temporal and spatial heterogeneities of light inside a canopy, it is predicted that plants with more rapid response to changes in light levels should have increased water and light use efficiencies in the field. In addition, plant functional types and geographic origins are two important determinants of g_s delays in response to FL (Vico et al. 2011). Therefore, in this work, this hypothesis has been tested with a rice diversity panel that was developed by USDA to represent the genetic diversity of the 18 000 accessions of the USDA-ARS rice (Oryza sativa L.) germplasm collection (Agrama et al. 2009). This collection contains 217 accessions which were originated from 76 countries covering 15 geographic regions. It consists of six groups, indica (35.4%), aus (18.7%), tropical japonica (18.2%), temperate japonica (15.2%), aromatic (3.0%) and their admixtures (9.6%) (Agrama et al. 2009; Li et al. 2010). This mini-core collection, with its large genetic diversity and large diversity of climates of their origins, represents an ideal choice of germplasm to study the natural variation of stomatal delays and its implication to both water and light use efficiencies in the field.

In this study, we aim to address three specific questions. (1) What are the levels of variability in stomatal responses to FL in the rice mini-core diversity panel? (2) Do plants with different stomatal responses to FL have different WUE? (3) Do plants with different stomatal responses to FL have different drought tolerance in the field?

The experiments were conducted in 2013 and 2014. In July 2013, a diversity panel composed of 204 accessions derived from 217 USDA mini-core rice accessions. The remaining 13 accessions in the mini-core accessions were not used due to long growth seasons.

Data of geography such as latitude (LAT), longitude (LONG) and elevation (ELV) for the origin of each accession was collected from Genetic Stock Oryza (http://www.ars.usda.gov/Main/docs.htm? docid˛ = Í8318, accessed 3 May 2014), according to Genetic stock number (GSOR) of each accession. Annual climate information such as mean (MET) and maximum (MAT) annual temperature (°F), mean annual dew point (MEDP) and mean annual precipitation (PRCP) was downloaded from National Oceanic and Atmospheric Administration, NOAA (https://gis.ncdc.noaa.gov/, accessed 15 March 2015). All of these data were collected based on the year that each accession released (1914–97) to determine originated weather condition for each accession. Total photosynthetically active radiation (PAR), and maximum daylength (MDL) during growth season were determined by the solar elevation angle, which further depends on the originated information of LAT and LONG for each accession (Song et al. 2013).

The experiments using the diversity panel with 204 accessions were conducted in the Institute Genetics and Developmental Biology, Chinese Academy Sciences (CAS), Beijing (116.3943°E, 39.9820°N) in 2013. Rice seeds from the each accession were grown for ~30 days in a soil seed bed. Then, from early June, seedlings were transplanted into plastic pots (12 L volume) containing commercial peat soil (Pindstrup Substrate no. 4, Pindstrup Horticulture Ltd, Shanghai, China) and then grown under outdoor conditions. Three plants for each accession were seeded in one pot and each accession used two pots. Measurements were started from early July 2013. During this period, the average lower temperature range was 24~26°C, and the average higher temperature range was 30~31°C, whereas the humidity range was 61~75%. Before the measurements, plants were first moved into an environment-controlled growth room for 30–50 min to ensure full-adaptation of stomata. The growth room was maintained to have an air temperature of 27°C and a photon flux density of around 600 μmol m^–2 s^–1. The experiments were conducted between 0830 to 1630 hours.

Based on the measured stomatal delays in response to fluctuating light (FL) levels (see gas exchange measurements for more details) from 2013, a subset of 204 rice accessions, i.e. 12 accessions with extremely contrasting τ_cl was selected to determine whether different τ_cl is able to affect a drought treatment study in 2014. This selection strategy is based on larger amplitude and longer duration in τ_cl than that τ_op (Ooba and Takahashi 200; Vico et al. 2011). Therefore, in 2014 experiments, our study focussed on τ_cl instead of τ_op. Among these 12 accessions, six were in fast τ_cl group, six were in slow τ_cl group (Table 1). Between two τ_cl groups, the MAT were significantly different (P < 0.05). MAT in fast τ_cl group exhibited a lower value compared with slow τ_cl group.

Table 1.  Geographic and climate information in 12 accessions selected in 2014 Hainan experiments
Stomatal delays (τ_cl) were classified into two subgroups (fast and slow) based on previous observations in Beijing in 2013 experiments. Genetic stock number (GSOR) in Genetic Stock Oryza (http://www.ars.usda.gov/Main/docs.htm?docid˛=Í8318, see text). Abbreviations: ACN, accession number; POP, population; LAT, latitude; LONG, longitude; PAR, total earth surface photosynthetically active radiation(PPFD); MDL, maximum daylength (h); MET, mean annual temperature (°F); MAT, maximum annual temperature (°F). Geographic and geoclimate information derived from the year that individual accession released (1914–97), and geoclimate information was collected from NOAA (https://gis.ncdc.noaa.gov/map/viewer/#apP=cdoandcfg=cdoandtheme=annualandlayers=1, see text)

In 2014, these 12 rice accessions with contrasting τ_cl based on previous investigation in 2013 were grown under two conditions, i.e. an extreme drought stress condition (DS hereafter) and a normal irrigated condition (NORM hereafter) in Hainan (110.0375°E, 18.5060°N), China. For the drought treatment, rice was sheltered during rain. The mean daily air temperature between two locations during growth seasons (Beijing and Hainan) were 25.8 and 26.5°C. Mean daylengths were 12.5 and 12 h respectively. These data was obtained from China Meteorological Administration (http://www.cma.gov.cn/, accessed 15 March 2015) recorded from 2013 to 2014. Two plots were used for the DS conditions and one plot was used for the NORM condition. To decrease the boundary effect, 36 plants were planted (6 × 6) for each accession, with 15 cm between plants within each row and 20 cm between rows. The rice fields were managed according to standard local agronomic practice with the following fertiliser application guideline: 48 kg N ha^–1, 120 kg P₂O₅ ha^–1, and 100 kg K₂O ha^–1 as the basal fertiliser, and additional 86 kg N ha^–1 at the tillering stage and 28 kg N ha^–1 at the booting stage. Weeds in both DS and NORM fields were controlled by a combination of chemical and manual methods and insects were controlled chemically. For DS blocks, soil moisture was monitored with a time domain reflectometry method at a soil depth of 20–30 cm (Ledieu et al. 1986). For the DS blocks, the soil water content was 8~12% (v/v) during photosynthetic measurements.

Experiments were conducted between 50 and 60 days after emergency (DAE) in both 2013 and 2014. The middle segment of the top fully expanded leaf on the main tiller was chosen during experiments. Four portable infrared gas exchange system (Li-6400XT, Li-Cor Inc., Lincoln, NE, USA) were pre-adjusted by manufacturer standard gas and used simultaneously during experiments. The chamber is 2 × 3 cm in dimension and is equipped with a blue–red LED light source. The measurements were conducted on sunny days between 0900 and 1130 hours and between 1300 and 1500 hours. During the measurements, the leaf temperature was maintained at 25°C and the RH was maintained in range of 68 to 87%. The reference CO₂ concentration was set as 400 μmol mol^–1 during the experiment.

During the fluctuating light (FL) treatments, leaf was first exposed to a high light with a PPFD of 1200 μmol m^–2 s^–1 for at least 5 min until it reaches a steady-state. Then the PPFD was switched to be 100 μmol m^–2 s^–1. In 2013, the plants were kept under this low light for 25 min while in 2014 plants were kept under low light for 20 min. The light level of 100 μmol m^–2s^–1 was chosen based on work by Lawson and Blatt (2014). Leaves were kept under low light until g_s stabilised. After the g_s stabilisation, the PPFD was recovered to 1200 μmol m^–2 s^–1. Leaves were kept under high light for 5 min. Four biological replicates from four different plants for each accession were determined. With the measured A and g_s, we calculated the ratio of A (and g_s) between DS and NORM conditions, i.e. A_DS/A_NORM and g_sDS/g_sNORM, where A_DS and A_NORM represents the A under DS and NORM conditions, and g_sDS and g_sNORM represents the g_s under DS and NORM conditions.

In this study, 13 photosynthetic traits during FL were determined including steady-states and instantaneous photosynthetic rates. The initial WUE, g_s and A under high light as well as these three photosynthetic parameters under low light were estimated as the average values of these parameter during the last five measurements before light switching. We estimated the ratio in A under high light over low light (A_loss) as function of light use efficiency of A. Same equation was applied in WUE (WUE_loss) as well. As g_s exhibits an asymmetric response to light changes, we used an exponential fitting via MATLAB R2010a (MathWorks Inc., Natick, MA, USA), to determine amplitude of stomatal delays when light switching according to Vico et al. (2011) using the following equation:

where τ_g takes different values depending on whether the leaf is responding to a sudden increase (τ_op) or decrease (τ_cl) in light. The regions used to calculate τ_cl and τ_op are shown in Fig. 1a. These two parameters represent stomatal delays in minute scale. Goodness of fit (R²) after simulation in stomatal opening phase and stomatal closing phase were 0.88 and 0.95, respectively, which reveals a good fitting algorithm applied in this study. To decrease initial effects of g_s on τ_cl, τ_cl over g_s was calculated. To compare the potential decrease of WUE due to difference in the g_s responses, we estimated WUE_total by obtaining the averaged WUE throughout the light fluctuating experiment. We determined the SPAD value using the SPAD 502 Plus Chlorophyll meter (Spectrum Technologies Inc., Aurora, IL, USA). For each leaf, averaged SPAD values were recorded from five different positions in the middle section, and four replicates from four different plants were determined for each accession.

Fig. 1.  Dynamics of photosynthetic CO₂ uptake rate (A), stomatal conductance (g_s) and water use efficiency (WUE) to changing light levels. (a) Regions used to calculate WUE and also to model τ_cl and τ_op. (b) Measured dynamic changes of A, g_s and WUE under fluctuating light. The error bars of each trait indicate maximum s.e. throughout fluctuating light treatments. The light intensities used in the fluctuating light treatments were 1200 and 100 μmol m^–2s^–1 for high light and low light conditions respectively. (a, b) A, g_s and WUE are depicted in green, red and white respectively.

In 2013 large-scale measurements, four spike related traits, i.e. final tiller number (TIR), stem base diameter (SBD), spike length (SPL) and number per plant (SPN), and three grain weight related traits, i.e. yield per plant (YD), yield per tiller (YD/TIR) and heading time (HEAD) were determined. At least four replicates were conducted for each trait above.

The total aboveground biomass for each accession was measured with a sample size of four while the gas-exchange data were collected. After determining FW, samples were kept at 120°C for 1 h and then under 70°C for at least 24 h until constant weight in baking oven. In 2014 experiments, DW under NORM over drought conditions were calculated to determine biomass accumulation in response to DS at canopy levels.

To evaluate the long-term WUE, we measured the carbon isotope ratio (δ¹³C) of the top fully-expanded leaf at the maturity stage (around 90 DAE). To do this, leaf materials were first dried at 70°C to constant weight, then grounded and analysed using Thermo Delta V advantage. The long-term standard deviation of the equipment is 0.1‰. The ¹³C/¹²C ratio was expressed following the delta (δ) notation in parts per mill (‰) relative to the international V-PDB (Vienna PeeDee Belemnite) standard. The measurements were performed at the Mass Spectrometry Laboratory of Forestry Science Research Institute China. Four replicates were used for δ¹³C measurements.

To analyse the variance of 24 physiological traits and 9 originated climate information in 204 mini-core collection, one group variance analysis was conducted using StatView, ver. 5.0.1 (SAS Institute Inc., Cary, NC, USA). Coefficient of variance and standard deviation were calculated when comparing groups with or without same unit. The correlation between traits was calculated as Pearson’s correlation coefficients among genotype means. Corrplot package in R software (http://www.r-project.org/, accessed 13 June 2015) was employed to calculate Pearson’ coefficient index (r) or goodness of fit (R²).

During fluctuating light (FL), A, g_s and WUE showed dynamic changes (Fig. 1b). Steady-state A were 23.02 ± 2.96 μmol m^–2 s^–1 at 1200 μmol m^–2 s^–1 (Fig. 1b; Table 2). After decreasing PPFD from 1200 to 100 μmol m^–2 s^–1, A decreased immediately by 80% within 1 min and kept 4.87 ± 0.64 μmol m^–2 s^–1, whereas g_s showed an exponential reduction: g_slow across mini-core accession was 0.12 ± 0.04 μmol m^–2 s^–1, which accounts for 25% of the initial g_s (Fig. 1b). However, WUE increased in an opposite pattern as the change in g_s, and reached a maximum at around 13 min after switching to low light. When the light level was increased again, a rapid increase in A, g_s and WUE was observed.

Table 2.  Phenotypic variations among 30 ecophysiological, geographical and biomass related traits for the 204 accessions used in the 2013 Beijing experiment
One-way ANOVA analyses were conducted using StatView, ver. 5.0.1 (see text). Abbreviations: CV, coefficient of variance; FD, fold differences. For further abbreviations see Table 1

In the 2013 experiments, we investigated nine eco-physiological traits and eight biomass related traits. To further analyse whether stomatal responses might be caused by different eco-niches in which plants were originally grown, we obtained the 13 geography and climate factors related to the growing regions of these rice accessions from NOAA database (https://gis.ncdc.noaa.gov, accessed 15 March 2015). The fold differences between the maximum and minimum for each geographical factor were more than those of the ecophysiological and biomass traits. Most of these traits showed a significant difference (P < 0.0001), except for maximum daylength (MDL) according to one-way ANOVA (Table 2). Across these geographical factors, annual precipitation (PRCP) and maximum air temperature (MAT) exhibited 676- and 14.1-fold differences between the maximum and minimum respectively. Among them, latitude (LAT) and longitude (LONG) were in the range from –34.9 ~50.8°, and –99.2 ~178.7° respectively. Annual mean temperature (MET) ranged from 31.2 ~91.8°F. MAT ranged from 9 ~76.5°F.

There were large natural variations in τ_cl and τ_op (Fig. 2). The values in τ_cl ranged from 1.81 to 12.42 with a normal distribution across these 204 rice accessions (Fig. 2; Table 2). Among subpopulation, AUS exhibited a relatively higher median and minimum values of τ_cl and τ_op, compared with two types of japonica (TEJ and TRJ) (Fig. 2). The median of τ_cl and τ_op across indica (IND) showed similar or higher rates than in japonica. Mean of τ_cl was 4.46, which is two times higher than τ_op. Fold differences between minimum and maximum in τ_cl and τ_op were 6.86 and 3.24, respectively, the variance level is significant (P < 0.0001) (Table 2). To eliminate the effects of initial g_s, τ_cl was normalised (τ_cl/g_s) in this study. The τ_cl/g_s ranged from 0.28 to 2.36 with an 8.42-fold difference (Table 2). With respect to fold difference in A, g_s and WUE under both high light and low light, the ranking from high to low is: g_s > WUE > A, i.e. a relatively higher fold difference was observed in g_s under high light and low light, which were 5.5 and 5.2 respectively. Whereas A under high light and low light exhibited a relatively low differences, the fold difference is no more than 2.5. Among ecophysiological traits, SPAD values exhibited highest standard deviation and coefficient of variance.

Fig. 2.  Natural variations of τ_cl (a) and τ_op (c), and distribution of τ_cl (b) and τ_op (d), in mini-core population. Abbreviations: Admix, admixture of subpopulation; ARO, aromatic; AUS, aus; IND, indica; TEJ, temperate japonica; TRJ, tropical japonica. Data from 204 accessions were used for the plot.

Under different geography and climate conditions, biomass related traits exhibited a significant variation as well (P < 0.0001). FW and DW per plant ranged from 48.6 to 350.3 and 9.8 to 80.1 g respectively (Table 2). In terms of spike traits, i.e. SBD, TIR and SPL, the fold differences between the minimum and maximum were 3.96, 5.49 and 1.82 respectively. In contrast, YD and YD/TIR exhibited higher fold differences, i.e. 8.68 and 5.60 respectively. HEAD time ranged from 64 to 150 days. The mean of HEAD time was 113 days.

Among the collected parameters, we calculated the Pearson correlation coefficient (r) between them (Fig. 3; see Table S1, available as Supplementary Material to this paper). Correlation r in photosynthetic traits during FL revealed a robust relationship between these traits. τ_cl was positively correlated with τ_op, τ_cl/g_s, A, g_s, A_low, g_slow and A_loss (P < 0.05), but was negatively correlated with WUE and WUE_total (P < 0.05). Furthermore, τ_op was positively correlated with WUE_loss, WUE_low and WUE_total (P < 0.05), but not significantly correlated with A, g_s and WUE. And τ_op was negatively correlated with SPAD (P < 0.05).

Fig. 3.  Matrix of correlation coefficients between 24 ecophysiologcal and biomass-related traits and 9 geoclimate traits across 204 mini-core accessions. The colour gradients from red to blue represent the range of Pearson correlation coefficient (r) from –1.0 to 1.0. The black lines are used to separate variable categories. The P-values for the correlation between each pair of traits were shown in Table S1, available as Supplementary Material to this paper. Abbreviations: τ_cl, stomatal delays during closing phase; τ_op, stomatal delays during opening phase; A, photosynthesis rates; g_s, stomatal conductance; WUE, water use efficiency; A_low, photosynthesis rates under low light; g_slow, stomatal conductance under low light; WUE_low, water use efficiency under low light; A_loss, photosynthesis rates under low light over high light; WUE_loss, water use efficiency under low light over high light; WUE_total, totally mean WUE throughout FL treatments; SPAD, SPAD records; LAT, latitude; LONG, longitude; ELV, elevation; PAR, total earth surface photosynthetically active radiation; MDL, maximum daylength; MET, mean annual temperature; MAT, maximum annual temperature; MEDP, mean dew point; PRCP, precipitation; SBD, stem diameter; TIR, tiller number; SPL, spike length; YD, yield per plant; YD/TIR, yield per tiller; HEAD, heading time.

The diverse origin, from the perspective of geography and year of release, in the mini-core collection can affect stomatal evolution under FL. As shown in Fig. 3, τ_cl and τ_op displayed an intricate relationship in response to geography factors, τ_cl and τ_cl/g_s was positively correlated with MAT (P < 0.05), whereas τ_op was positively correlated with LAT and MAL (P < 0.05) but negatively correlated with MEDP (P < 0.05) that is typically representing absolute air humidity.

For the biomass related traits, both τ_cl and τ_op were positively correlated with biomass accumulation (DW and FW) (P < 0.05), but no correlation was observed between τ_cl/g_s and biomass (Fig. 3). In contrast, τ_cl, τ_op and τ_cl/g_s were all negatively correlated with SBD. In terms of spike traits, τ_op was positively correlated with TIR and SPL, but negatively correlated with YD/TIR (P < 0.05). In contrast, τ_cl was negatively correlated with HEAD time (P < 0.05).

Because τ_cl was negatively correlated with WUE related traits (Fig. 3) and showed a larger variation compared with τ_op (Fig. 2), we selected 12 accessions for further experiment in 2014 to test whether difference in τ_cl can lead to difference in WUE in the field (see material and methods for details). Among these 12 accessions, six accessions possessed extremely high six τ_cl (Fast group) and six accessions with extremely low six τ_cl (slow group).

In the 2013 experiments, A and g_s within the slow group were 30% and 80% higher than those in the fast group respectively before light was switched from high to low light (Fig. 4a, b). After light was switched to low light, g_s in the fast group declined and gained stable values within 8 min. In contrast, there was a greater lag in g_s in the slow group. A under low light didn’t differ between two groups. When light was increased from low to high light levels, g_s and A in two groups increased rapidly. Before light switching, WUE in the fast group was ~75 μmol mmol^–1, which is 50% higher than that in slow group. After light switching to low light levels, WUE dropped sharply then recovered back after a new balance between A and g_s was formed. WUE_low in both groups were slightly lower compared with WUE under high light. However, WUE in fast group recovered to normal levels at a faster pace than that in the slow group. In terms of C_i/C_a (Fig. 4d), because of stomatal effects, a relatively higher C_i/C_a was induced in the slow group than the fast group after light switching. The dynamics in A under FL was similar between the fast and slow groups; however, the dynamics of g_s differ dramatically between the fast and slow group (Fig. 4, see Fig. S1, available as Supplementary Material to this paper).

Fig. 4.  Dynamic changes of carbon assimilation (a), stomatal conductance (b), water use efficiency (c), and internal CO₂ over leaf surface CO₂ (d) during fluctuating light. Grey area represents low light treatments. Error bars indicate the maximum s.e. for each group. Closed circles in red and black show τ_cl in the fast and slow groups respectively.

The τ_cl measured across 12 accessions under two growth locations (Beijing, BJ and Hainan, HN) were highly correlated (Fig. S3). R² regarding to τ_cl between BJ and HN was 0.65; similarly, R² for τ_cl/g_s between BJ and HN was 0.79 (P < 0.001) (Fig. S6). In addition, τ_cl hold the higher values in slow group than that that of the fast group under two growth conditions (Fig. S3). A co-ordination in τ_cl and a conserved ranking across accessions between NORM and DS condition were observed (Fig. 5, Fig. S3). In contrast, the τ_op showed a slightly less conserved ranking between these two conditions (Fig. 5b). Furthermore, when the τ_cl/τ_op in the DS was plotted against the ratio in NORM (Fig. 5c), a slope <1 was obtained, i.e. we have a smaller τ_cl/τ_op under DS compared with the ratio under NORM.

Fig. 5.  Correlation of τ_cl, τ_op, and τ_cl/τ_op between normal condition and drought stress condition. Closed circles in red and black show τ_cl in the fast and slow groups respectively. Determination of coefficient is depicted as R².

Among accessions, τ_cl appeared to be substantial differences. In particular, T4174 and C4030 showed an extremely contrasting τ_cl with around 3-fold differences under both NORM and DS conditions (Fig. 6). Diverse response across accessions due to DS was observed. Drought treatments induced 2-fold differences in τ_cl in T4174, but 50% fold differences in T4177 (Fig. 6). A positive correlation between τ_cl and biomass (DW) under both NORM and DS conditions was observed (R² = 0.67 and 0.37 respectively) (Fig. 7c and insert in Fig. 7c). Remarkably, both τ_cl was negatively correlated with biomass drought ratio (Fig. 7d). The R² is 0.69, while R² of τ_cl/g_s is 0.43 (Fig. S6). In contrast, A and g_s across 12 subset accessions were not significantly correlated with biomass drought ratio (Fig. S4).We further evaluated the correlation between biomass drought ratio with the changes of A and g_s under drought stress. Specifically, we calculated the ratio of A (and g_s) between DS and NORM conditions, i.e. A_DS/A_NORM and g_sDS/g_sNORM, and the obtained the correlation between biomass drought ratio and A_DS/A_NORM and g_sDS/g_sNORM separately. The biomass drought ratio showed a higher R² with τ_cl than either A, or g_s or A_DS/A_NORM or g_sDS/g_sNORM (Figs 7d, S5). These results suggested that τ_cl, instead of either A or g_s or A_DS/A_NORM or g_sDS/g_sNORM, was closely related to drought tolerance. Biomass between 2013 and 2014 experiments showed a robust correlation (R² = 0.88) (Fig. S2).

Fig. 6.  Comparison of τ_cl among accessions under normal condition and drought stress condition (insert). Red and black bars represent τ_cl of the fast and slow groups respectively. One-way ANOVA was used to compare the significant levels between accessions, and same letter represent no significant differences (P < 0.05).

Fig. 7.  Comparison on biomass (a) and biomass drought ratio (b) among accessions, and correlation of τ_cl with biomass under normal (c) and drought (c insert) and biomass drought ratio (d). One-way ANOVA was used to compare the significant levels between accessions, and different letters represent significant differences (P < 0.05). Determination of coefficient is depicted as R².

Drought induced ~60% reduction in biomass on average across accessions (data not shown), but changes in biomass under DS differed among accessions obviously from each other (Fig. 7b). Among these accessions, biomass in T4174 showed a relatively higher values; in contrast, C4030 showed a lower biomass (Fig. 7a). In terms of biomass drought ratio, N4129 and Y4211 were around 0.75, which are 2-fold higher than T4174 and N4122 (Fig. 7b).

To investigate whether τ_cl can affect long-term leaf water use efficiency, carbon isotope ratio, δ¹³C in 12 accessions were determined. Result showed a negative correlation between τ_cl and δ¹³C (R² = 0.5) (Fig. 8 insert). δ¹³C across accessions were in ranges of –24.9 to –29.4‰. Accessions in fast group exhibited a lower values in δ¹³C, compared with that in slow group (P = 0.05). In particular, δ¹³C of T4174 and J4087 in slow group were averaged around –29.0‰, in contrast, the average δ¹³C of Y4211 and N4216 in fast group were –25.1‰ (Fig. 8). The τ_cl showed a positive correlation with biomass and a negative correlation with the biomass drought ratio, i.e. the ratio of the biomass under drought to the biomass under control conditions (Figs 6, 7c, d).

Fig. 8.  Comparison on δ¹³C and τ_cl for each rice accession. Inserted panel showed correlation between δ¹³C and τ_cl. One-way ANOVA was used to compare the significant levels between accessions, and different letters represent significant differences (P < 0.05). Determination of coefficient is depicted as R².

Stomata always exhibited a lag before getting stabilised during fluctuating light (FL) regimes, whereas carbon assimilation (A) responded to FL at a faster pace (Pfitsch and Pearcy 1989; Vico et al. 2011). This study used a rice mini-core panel consisting of 204 accessions collected by USDA to examine the significance of this phenomena to drought resistance in the field. The results suggest that there were large natural variations in the stomatal delays in response to FL and this variation was closely related to WUE and drought resistance in the field.

There were substantial natural variations in the stomatal delays to changes in light levels in the rice diversity panel (Fig. 2; Table 2), which is consistent with earlier reports (Chazdon 1988; Pearcy 1990; Ooba and Takahashi 2003). Simulation studies (Vico et al. 2011) using different species assembled from public database showed τ_op is less than that τ_cl during FL in graminoids, forbs and angiosperms, and there is a strong coordination between τ_cl and τ_op. Rice in our study showed the same trend as well (Fig. 3; Table 2). The value of τ_cl across these 204 rice accessions varied between 1.81 and 12.42 (Table 2) whereas the values of τ_op ranged between 1.21 and 3.93 (Table 2). This much higher value of τ_cl compared with τ_op (Table 2; Fig. 2) suggested that rice has the tendency of optimising carbon assimilation instead of conserving water under NORM conditions. This might be related to the fact that rice normally is grown in wet environments, therefore, the evolutionary pressure is on selection of faster stomatal opening to optimise the photosynthetic light use efficiency. Consistent with this, a survey of light saturated rate of photosynthesis in 24 accessions of rice (A_max), representing 17 species and 8 genomes, showed that A_max ranged from 25 to 45 μmol m^–2 s^–1 (Giuliani et al. 2013), which is relatively higher compared with A_max of other C₃ plants (Wright et al. 2004).

Compared with the NORM conditions, rice grown under DS conditions showed decreased τ_op and τ_cl (Fig. 5a, b). But the decrease in τ_cl was higher than τ_op, as shown by the decreased ratio of τ_cl/τ_op under DS conditions (Fig. 5c). This indicated that the internal regulatory mechanisms of the control of stomatal conductance conditions to optimise water conservation instead of carbon assimilation under DS. For rice grown in a paddy field, the relatively faster opening of stomata compared with the closing of stomata (i.e. high τ_cl/τ_op) under FL can lead to a relatively less favourable WUE at the top of the canopy where the leaves experience relatively constant high light levels.

The observed variations of both parameters should be related to the long-term adaptation to the climatic conditions of the eco-niches, as demonstrated by the correlation between these parameters and the geographic and climate factors of the origin of rice accessions (Fig. 3). For example, τ_cl was positively correlated with MAT (P < 0.05), but τ_op was negatively correlated with MEDP (P < 0.05) and PRCP. The positive correlation between τ_cl and MAT was in contrast to the expectation that under high temperature regions a lower τ_cl would be desirable to ensure higher WUE. A higher dew point indicates higher humidity in the air. In our study, the negative correlation between τ_op and MEDP suggested that areas with high humidity, i.e. high MEDP, will have a lower τ_op, i.e. faster stomatal opening speed. In line with this, PRCP showed a negative correlation with τ_op, i.e. under high precipitation, the τ_op would be lower. These are reasonable since under conditions without water stress, decrease τ_op, i.e. increasing the stomatal opening speed can help optimise photosynthetic light use efficiency, which is more important than decreasing the stomatal response time for water saving. Therefore, humidity and precipitation are dominant environmental parameters that influence the speed of stomatal dynamics under varying light conditions.

Large variations in τ_cl and τ_op between different species have been observed previously, this study showed that not only there is also a huge variation of τ_cl and τ_op within the same species, but also there are substantial variation between subgroups of rice. For example, the AUS showed relatively higher τ_cl and τ_op while TRJ showed low values of τ_cl and τ_op (Fig. 2). These rice accessions therefore provide a rich source of materials to study the potential driving force and molecular mechanisms controlling the evolution of these different features.

The large variation in τ_cl is closely related to drought resistance in the field. The relationship between τ_cl and WUE has been reported earlier at the leaf level (Lawson and Blatt 2014). However, the potential impacts of different stomatal dynamics on drought resistance in the field will depend the time scale of the dynamic changes of the light environments inside a canopy, which will be further determined by the canopy architecture, wind speed, cloud movement, angle of incident radiation etc. (Chazdon 1988; Pearcy 1990). Data from this study demonstrate that accessions with higher τ_cl, i.e. slower rate of stomatal closure, exerted a lower drought resistance, as demonstrated by the relatively higher biomass drought ratio (Fig. 7b, d). This is further confirmed by the changes of isotope carbon ratio, δ¹³C (Fig. 8), which reflects the long-term WUE of a plant (Farquhar and Richards 1984; Farquhar et al. 1989). We further tested whether the drought resistance might also be linked to the difference in A, g_s, or changes of A or g_s under drought conditions (i.e. A_DS/A_NORM, or g_sDS/g_sNORM). Correlation analysis suggested that in the biomass drought ratio possessed much higher correlation with τ_cl as compared with those with either A, or g_s or A_DS/A_NORM, or g_sDS/g_sNORM (Figs 7, S4, S5). The biomass ratio showed a R² of 0.2 with g_sDS/g_sNORM. Therefore, though the data also indicated that g_s might also be related to the biomass drought ratio, τ_cl is a parameter that are more closely related to drought resistance in the field.

Our findings suggest that τ_cl exhibited a positive correlation with biomass accumulation, though decreased τ_cl was linked to an increased drought resistance. Indeed, both τ_cl and τ_op were positively correlated with biomass accumulation (DW and FW) (Fig. 3). Concurrent with this positive correlation, we also found a positive correlation between τ_cl and g_s and A (Fig. 3), i.e. those lines with a faster stomatal closing dynamics indeed performed on average lower g_s and A values. The observed correlation between τ_cl and biomass therefore may be related to the overall higher A in those lines with higher τ_cl. The molecular mechanism underlying the positive correlation between τ_cl and A and g_s is unknown. However, these data suggest that for regions where drought is relatively a rare event, decreasing τ_cl should not be used as a breeding target. In contrast, for regions with frequent severe drought, it would be more effective for rice productivity (Tuberosa and Salvi 2006). This is clearly demonstrated by the strong inverse relationship between τ_cl and biomass drought ratio (Fig. 7d). For example, the line Y4211, which showed a τ_cl of ~2 min, its decrease of biomass under DS was only ~20%; this is in great contrast with line T4174, which showed a τ_cl of ~2 min of ~8.5 min and had a decrease of biomass by ~70% under DS (Figs 6, 7d). Though a positive correlation between τ_cl and biomass is apparent in our used 12 accessions, we cannot eliminate the possibility that the observed positive correlation between τ_cl and biomass is caused by other factors, such as changed root : shoot ratio or different construction cost for the biomass.

Torres et al. (2013) conducted a large scale drought tolerance using rice genebank germplasm for yield and found that many drought tolerant accessions belong to AUS or IND group. However, three accessions out of the 12 accessions used in 2014 belong to the AUS group, but in general they showed relatively low stomatal closing speed (Table 1). We emphasise here that although faster stomatal closing is shown here to be associated with drought tolerance, rapid stomatal closing is not the only mechanism that plant use to cope with drought. The AUS and IND groups used by Torres et al. (2013) may have other drought escape, recovery, and avoidance mechanisms to gain high drought tolerance. This points to a future direction of improving drought tolerance in AUS or IND cultivars, i.e. the stomatal response dynamics can be optimised in the AUS or IND rice cultivars to gain even higher drought tolerance.

In summary, this study demonstrated that there are large natural variations in τ_cl and τ_op in response to fluctuating light in one single species. We also noted the tendency of stomatal dynamics to optimise either carbon uptake or water conservation depending on water availability and temperature of origin of the rice accession. Our findings revealed that the biomass accumulation had a positive correlation with stomatal closing speed, even though this speed was positively correlated with drought resistance. All these findings suggest that stomatal movement has a complex influence on the fitness of plants in the field and its application in rice breeding needs to consider the severity of the water stress.