Effects of kinetics of light‐induced stomatal responses on photosynthesis and water‐use efficiency

Summary Both photosynthesis (A) and stomatal conductance (g s) respond to changing irradiance, yet stomatal responses are an order of magnitude slower than photosynthesis, resulting in noncoordination between A and g s in dynamic light environments. Infrared gas exchange analysis was used to examine the temporal responses and coordination of A and g s to a step increase and decrease in light in a range of different species, and the impact on intrinsic water use efficiency was evaluated. The temporal responses revealed a large range of strategies to save water or maximize photosynthesis in the different species used in this study but also displayed an uncoupling of A and g s in most of the species. The shape of the guard cells influenced the rapidity of response and the overall g s values achieved, with different impacts on A and W i. The rapidity of g s in dumbbell‐shaped guard cells could be attributed to size, whilst in elliptical‐shaped guard cells features other than anatomy were more important for kinetics. Our findings suggest significant variation in the rapidity of stomatal responses amongst species, providing a novel target for improving photosynthesis and water use.


Introduction
Stomata control the balance of gases between the internal leaf environment and the external atmosphere; regulating CO 2 uptake for photosynthesis and water loss through transpiration (E). Low stomatal conductance to water vapour (g s ) can restrict CO 2 uptake by limiting CO 2 influx and thus net CO 2 assimilation rate (A), whereas high g s facilitates high rates of A, but greater water loss is an inevitable consequence.
This balance between CO 2 limitation and water loss is characterized by intrinsic water use efficiency (W i ), which is the ratio between A and g s . On an instantaneous timescale, maintaining a suitable and appropriate balance is impeded by the temporal stomatal responses, which are a magnitude slower than those of A. Therefore, in response to changing light, the kinetics of g s can greatly impact CO 2 uptake and water loss, which has significant implications water use efficiency (WUE). WUE can be defined as the ratio of net CO 2 uptake relative to water loss through transpiration (E) or as the ratio of biomass or yield accumulation to water use over the growing season. Consequently, WUE is often a target for improving crop performance; however, it should be noted that greater W i is often at the expense of A (Blum, 2009;Lawson et al., 2010;Lawson & Blatt, 2014). The rate of water transpired through the stomata is an order of magnitude greater than the rate of CO 2 uptake for A due to the greater water concentration gradient between the intercellular spaces within the leaf and the external atmosphere (as well as biochemical limitation on A). In order to maintain an optimal balance between A and E, stomatal guard cells are continually adjusting to environmental and intracellular cues (Lawson & Blatt, 2014).
Many previous studies have reported a strong correlation between A and g s (Wong et al., 1979;Farquhar & Sharkey, 1982). This correlation is generally observed because steady-state values are often reported, yet under dynamic conditions g s responses are not always coupled with A (Knapp & Smith, 1987, 1990. In natural environments, photosynthetic photon flux density (PPFD) fluctuates on timescales of seconds to days and seasons (Assmann & Wang, 2001) driven by changes in cloud cover, sun angle and shading from adjacent leaves in the canopy (Pearcy, 1990;Chazdon & Pearcy, 1991;Way & Pearcy, 2012). Plants therefore experience short and long term fluctuations in PPFD creating 'sun' and 'shade' flecks to which A and g s respond. Slower stomatal opening when A responds rapidly to a PPFD increase can limit CO 2 assimilation (Tinoco-Ojanguren & Pearcy, 1993), whilst delayed stomatal closing responses following a decrease in PPFD and photosynthesis result in unnecessary water loss when carbon gain is limited (Lawson et al., 2010;Lawson & Blatt, 2014). Due to the difference in the rate of carbon gain to water loss, any disparity in the response of A and g s also increases the probability of water stress (Condon et al., 2002). For example, slow closing of stomata when A has decreased will result in higher than necessary transpiration rates that will deplete the soil water more rapidly and thus potentially create a soil water deficit.
It has been estimated that stomata can limit A by up to 20%, which can impact substantially on crop yields (Farquhar & Sharkey, 1982;Jones, 1987Jones, , 1998Fischer et al., 1998;Lawson & Blatt, 2014). In order to maximize A and optimize W i , species or cultivars with rapid stomatal responses would be intuitively desirable, as there would be greater synchrony with mesophyll demands for CO 2 . Amplitude and rapidity of stomatal movements are therefore potential targets to improve A and W i . The majority of studies reporting stomatal influences on photosynthesis describe steady-state values and explore the potential of increasing or decreasing g s to enhance A or diminish water loss, however this often results in an overall reduction of A and thus productivity (Blum,2009). We propose here to follow another approach for plant improvement, which exploits stomatal kinetics to facilitate synchronous g s responses with mesophyll demands for CO 2 , thus simultaneously reducing CO 2 limitations as well as avoidable water losses, incidentally enhancing W i .
Only a handful of investigations have focused on dynamic stomatal responses and even fewer of these have explored the effects on A and W i . Most of these studies have examined forest understorey species and the impact of sun-fleck regimes on A and E (Pearcy, 1990;Tinoco-Ojanguren & Pearcy, 1993;Leakey et al., 2005;Way & Pearcy, 2012). In addition, assessing the rapidity of stomatal movements is complicated by variation in both the sensitivity and responsiveness of stomata between different species (Ooba & Takahashi, 2003;Lawson et al., 2010;Vico et al., 2011) and between individuals of the same species grown in different habitats (Drake et al., 2013). After a change in PPFD, the temporal response of g s is usually composed of three steps: an initial lag where the value of g s remains stable for several minutes, followed by an exponential phase during which rapid increases in g s are observed before reaching final steady-state plateau (Naumburg et al., 2001;Vialet-Chabrand et al., 2013). Recently, a dynamic sigmoidal model has been developed by Vialet-Chabrand et al. (2013) to analyse the temporal response of g s by estimating the initial lag time (k), a time constant (k) and a steady-state target (G smax ; see Table 1 for a summary of parameters and units). The time constant was used to describe the rapidity of the exponential phase independently of the amplitude of the g s response, facilitating species or cultivar comparisons as well as proposing a more accurate interpretation.
In order to determine the impact of stomatal responses to increasing and decreasing PPFD on limitation of A, water loss and intrinsic WUE, we have assessed and quantified the rapidity of stomatal movements in a range of plant types, including several major crops. We have selected species with kidney-and dumbbell-shaped guard cells to estimate the influence of anatomical features on rapidity of responses.

Plant material and growth conditions
Thirteen important crop species (including three C 4 species) were selected along with the model plant Arabidopsis thaliana, and the relict gymnosperm species Ginkgo biloba. Eight had kidney-or elliptical-shaped guard cells whilst four had dumbbell-shaped guard-cells that are typically found in grasses.

Research
New Phytologist lycopersicum), Sorgum (Sorghum bicolor), Barly (Hordeum vulgare), wheat (Triticum aestivum), maize (Zea mays), French bean (Phaseolus vulgaris) and broad bean (Vicia faba) were germinated in 650-cm 3 pots containing peat-based compost (Levington F2S). Following germination, plants were grown in a temperature-controlled glasshouse for 4-8 wk before measuring. Established Miscanthus (Miscanthus nepalensis) were supplied in 1-l pots from a commercial nursery (Beth Chatto, Colchester, UK). Solar radiation provided a PPFD of c. 500 lmol m À2 s À1 , supplemented by sodium vapour lamps (600W; Hortilux Schr eder, Monster, the Netherlands) to 300 lmol m À2 s À1 PPFD when external PPFD dropped below 1200 lmol m À2 s À1 over a 10 h period. Air temperature was maintained at 25°C AE 3°C during the day and 18°C AE 3°C at night. Plants were watered daily from below, with any excess water not absorbed by the pot within 2 h removed.
Rice (Oryza sativa) seeds were germinated and transferred to 650-cm 3 pots as described above and grown in a controlled environment with a photoperiod of 12 h : 12 h, light : dark at a PPFD of 500 AE 20 lmol m À2 s À1 , a day temperature of 25°C and VPD of 0.8 AE 0.2 kPa. Plants were measured after 12 wk.

Leaf gas-exchange measurements
Photosynthetic carbon assimilation (A) and stomatal conductance to water (g s ) were measured on the youngest fully expanded leaf using infrared gas analysis (Li-Cor 6400, Lincoln, NB, USA, and CIRAS-1, PP Systems, Amesbury, MA, USA). Light was provided by an integrated LED light source (Li-Cor, PP Systems). Leaves were first equilibrated at a PPFD of 100 lmol m À2 s À1 until both A and g s reached 'steady state', this being defined as a < 2% change in rate during a 10-min period (c. 30-60 min). Once steady state was satisfied, PPFD was increased to 1000 lmol m À2 s À1 for 1 h before returning to 100 lmol m À2 s À1 for 30 min. The leaf cuvette was maintained at 400 lmol mol À1 CO 2 concentration (C a ), a leaf temperature of 20°C (AE2°C) and a VPD of 1 AE 0.05 kPa. A and g s were recorded every 1 min. Intrinsic water use efficiency (WUE) was calculated as W i = A/g s . All measurements were completed before 14:00 h to avoid any unwanted diurnal or circadian effects on photosynthesis.

Leaf anatomical measurements
Stomatal impressions of the ad-and abaxial leaf surfaces were taken of the same area, measured using gas exchange. A negative impression was made using a dental polymer (Xantoprene, Heraesus Kulzer Ltd, Hanau, Germany) following the methods of Weyers & Johansen (1985). Once the impression material had dried and was removed from the leaf, a positive impression was made from this by placing in nail varnish on a microscope slide. Stomatal density, guard cell length (L) and guard cell width were determined using IMAGEJ software (National Institute of Health, Bethesda, MD, USA) from twenty fields of view (size 1250 lm 2 ) captured from each impression using a 5 MP eye-piece camera (MicroCAM 5 MP, Bresser Optics, Rhede, Germany).

Modelling g s , A and W i responses to PPFD
In order to describe the temporal response of g s to a single stepchange in PPFD, an analytical model derived from the model by (Vialet-Chabrand et al., 2013) was used (Fig. 1).
The model described the temporal response of g s using a time constant (k, min), an initial time lag (k, min) and a steady-state g s (G smax , mmol m À2 s À1 ) reached at given PPFD: where time 0 is the point at which PPFD was increased from 100 to 1000 lmol m À2 s À1 ; r 0 (mmol m À2 s À1 ), initial value of stomatal conductance before the change in PPFD). In this equation, the time constant k is a measure of the rapidity of response of g s independent of the amplitude of variation in g s (Eqn 2). To distinguish between the time taken for the stomata to open (increase) and to close (decrease), the abbreviations k i and k d are used. A second parameter combining rapidity and amplitude of the response, the maximum slope (Sl max ), was used to describe the maximal slope of the g s response to the step-change in PPFD: Parameter values were estimated using a Metropolis Hasting algorithm and a Bayesian model. The priors (a priori probability of the parameter values) used were uniform covering a large range of possible values and the initial values were chosen randomly.
The initial values were chosen from observed values (AE 10%) of both r 0 and G. For k, the range of values were selected from between 10 and 60 min, whilst k values were between 0.1 and Fig. 1 Theoretical temporal response of stomatal conductance (g s ; black) and net CO 2 assimilation (A; red) to a step change in PPFD from 100 (shaded area) to 1000 (unshaded area) lmol m À2 s À1 . Where Sl max describes the maximum temporal response of g s (dashed line), k describes the time-lag before g s starts to increase (blue arrow) and G smax describes the steady-state target of g s under 1000 lmol m À2 s À1 PPFD. The dotted lines represented the time and the value were 95% A is reached.

Temporal responses in g s limits A
During a step increase in PPFD, photosynthesis was considered limited by stomatal conductance until 95% A (A 95 ) was reached. Using this assumption, the percentage of limitation of A by g s was estimated by: Eqn 3 ( R t 0 ðA max À AÞ, integral of the difference between the maximum potential A (A max ) and the observed limited A from the beginning of the observed curve to the time t where A reached 95% of the steady state; R 60 0 A tot , maximum integral of A for 1 h period). Calculating the ratio using R 60 0 A tot normalized g s limitation over the 1-h measurement period (see Table 1 for a summary of parameters).

The impact of different g s and A responses on water loss
The nonsynchronous g s and A response influences the temporal W i response and the amount of water lost following a step increase in PPFD. To investigate the impact of g s responses on water use efficiency, we predicted g s from a simple model (g s = A/W i ) using a constant W i during the transient response. The constant value of W i was chosen close to the maximum A (95%), assuming that this would be close to an optimal W i with no limitation of A. On the one hand, when observed values of g s were greater than predicted by the constant W i model, more water was 'lost' than required to maintain optimal A; on the other, when observed values of g s were lower than predicted, water was 'saved', illustrating a close coupling of g s with A. As an investigating tool, this approach allowed us to assess the percentage of water 'lost' and 'saved' by comparing the coupling between A and g s in different species.

Statistical analysis
Statistics were conducted using SPSS (v.16; SPSS Inc., Chicago, IL, USA) and R (http://www.r-project.org/). A Shapiro-Wilk test was used to test for normality and a Levene's test of homogeneity was used to determine if samples had equal variance. Single factor differences were analysed using a one-way ANOVA with a Tukey-Kramer honest significant difference test where more than one group existed or a Student's t-test where only two groups were compared.

Results
Most species measured achieved steady-state g s after 60 min of high PPFD, except Helianthus and Vicia, which had not attained their maximum g s values within this timeframe ( Fig. 2), which might have led to an underestimation of their k i values, which were already high (Fig. 3). Additionally, Ginkgo displayed atypical g s and A behaviour (Fig. 2). The 30-min exposure to low light may not have been sufficient for complete, steady-state stomatal closure for some species; however, this does not greatly impact our estimations of additional water loss compared to instantaneous stomatal responses, because the major part of the water loss can be attributed to the initial rapid opening response of the stomata (Fig. 2).
Quantifying A and g s responses to step changes in PPFD Steady-state g s at the initial PPFD of 100 lmol m À2 s À1 (G smin ) varied significantly among the species (F (14,49) = 5.007, P < 0.0001), with the lowest values recorded for G. biloba Fig. 2 Normalized temporal response of net CO 2 assimilation (A; circles) and stomatal conductance to water vapour (g s ; triangles) of 15 species to an increase in irradiance from 100 (shaded area) to 1000 (unshaded area) lmol m À2 s À1 followed by a decrease to 100 lmol m À2 s À1 (see Table 3 for abbreviations of species nomenclature). The dashed line indicates where 95% maximum A (A 95 ) was achieved. Data are the mean AE SE (n = 3-5). Values were normalized to the initial values at 100 lmol m À2 s À1 PPFD and maximum values at (1000 lmol m À2 s À1 PPFD). (2016)

Research
New Phytologist (13.2 mmol m À2 s À1 ) and highest values for T. aestivum (255.9 mmol m À2 s À1 ) ( Fig. 3a; Supporting Information Fig. S1), whereas A was below 9 lmol m À2 s À1 for all species (Figs 3b, S2). An increase in PPFD to 1000 lmol m À2 s À1 led to an immediate and rapid increase in A compared to g s for all species. After this initial period, the increase in A slowed to a magnitude similar to the concurrent increase in g s . For the majority of species, A reached steady state while g s continued to increase. These different periods of coupled and uncoupled responses of A and g s were species-dependent. Although the majority of species displayed a mainly uncoordinated A and g s temporal response (Fig. 2), final steady-state values of A and G smax were significantly correlated among species (r s(51) = 0.78, P < 0.001 for C 3 species and r s(13) = 0.84, P < 0.001 for C 4 species - Fig. S3). In contrast to the majority of the species examined, S. bicolor, O. sativa and G. biloba all exhibited low g s and an unusually strong coupling between A and g s . The key difference between these three species was that S. bicolor and O. sativa exhibited a faster response of A and g s , whereas G. biloba showed rather slower responses.
The steady-state values of g s (at 1000 lmol m À2 s À1 PPFD) estimated by the model (G smax ) were significantly different among species (P < 0.0001, F (14,49) = 14.469; Fig. 3a), with observed values 10-fold higher in T. aestivum (482.9 AE 18.7 mmol m À2 s À1 ) compared with G. biloba (45.2 AE 1.9 mmol m À2 s À1 ). Values of G smax were positively related to Sl max (r = 0.61, P < 0.01) in elliptical-and dumbbell-shaped guard cells (r = 0.41, P < 0.05), whereas Sl max was related to the total time taken to open to G smax (k i ) in elliptical-(r = À0.54; P < 0.01) and dumbbell-shaped (r = À0.68; P < 0.01) guard cells (Table 2). Six out of the seven species with dumbbell-shaped guard cells showed the highest values of Sl max (Table 3). Applying the Vialet-Chabrand model to the temporal response of g s to an increase from 100 to 1000 lmol m À2 s À1 PPFD, showed that the initial lag time in the g s response (k) was significantly different among species (P < 0.0001, F (14,49) = 5.819) and ranged from 12 s for A. sativa to 6 min for G. biloba (Table 3).
After PPFD was returned to 100 lmol m À2 s À1 , A decreased immediately whereas g s showed a slow, exponential and speciesdependent decrease. Most species showed a temporal response of g s to decreasing light which was an order of magnitude lower than that of A, with some exceptions. Sorghum bicolor and Z. mays demonstrated significantly lower values of k d (< 1 min; Fig. 3c) and thus faster responses compared to other species, approaching the speed of assimilation rate response to light (Fig. 2). When PPFD was returned to 100 lmol m À2 s À1 , no significant differences in k were observed (data not shown), although steady-state g s and Sl max varied significantly amongst species.
Significant differences in the opening (k i ) and closing (k d ) time constants were observed amongst species (Fig. 3c); k i ranged from 0.9 min in O. sativa to 23 min in V. faba, and k d ranged from 0.9 min in S. bicolor to 14 min in P. vulgaris. k i and k d were positively correlated in species with elliptical-(R 2 = 0.29, P < 0.01) and dumbbell-shaped (R 2 = 0.52; P < 0.001) guard cells. Although the majority of species showed tendencies for greater rapidity in stomatal closing than opening (Fig. 3c), this was significant in only six species (Fig. 3c). On average, the species with dumbbell-shaped guard cells were 10 min faster in opening than elliptical species, reaching G smax in significantly shorter periods of time (t (44) = À8.2, P > 0.0001). C 4 species increased g s more rapidly than C 3 species (P < 0.0001). Estimations from the closing response showed that dumbbell-shaped guard cells were also faster than elliptical-shaped guard cells (P < 0.04) and C 4 species closed faster than C 3 species (P < 0.0001) ( Table 3).
In order to assess the extent that A was limited by g s during the increase of PPFD, we determined the time taken to reach 95% of (a) Comparison between species of (a) steady-state g s under 100 lmol m À2 s À1 PPFD (G smin ), g s at 95% maximum net assimilation under 1000 lmol m À2 s À1 PPFD (A 95 ) and steady state (G smax ) under 1000 lmol m À2 s À1 PPFD for 15 species; (b) steady-state A under 100 lmol m À2 s À1 PPFD (A initial) and A 95 ; (c) time constants k i and k d for stomatal opening and closure, respectively; and (d) the time taken to reach A 95 . Data are the mean AE SE (n = 3-5). Asterisks represented a significant asymmetry of k i /k d (P < 0.05). Species in bold have dumbbell-shaped guard cells, underlined species have a C 4 metabolism and species in plain font have elliptical-shaped guard cells and C 3 metabolism (see Table 3 for species name abbreviations). www.newphytologist.com maximum A (A 95 ) at 1000 lmol m À2 s À1 PPFD (Fig. 3d). Further increases in g s after A 95 were substantially greater than the remaining 5% increase in A, suggesting that g s was no longer limiting A (Fig. 2). Stomata opening did not increase significantly after A 95 had been reached for the species with dumbbell-shaped guard cells (Z. mays, S. bicolor, M. nepalensis, O. sativa) and for G. biloba (Fig. 2). The majority of species achieved A 95 within 30 min with values ranging between 5.7 and 30.7 lmol m À2 s À1 (Fig. 3b,d). Zea mays and A. sativa both attained the highest A and achieved A 95 in the least time (12-11 min; Fig. 3d), whereas G. biloba was the slowest, taking on average 48 min to achieve the lowest A 95 . Species with elliptical-shaped guard cells achieve significantly lower A 95 (P < 0.0001) compared with species with dumbbell-shaped guard cells (Fig. 3b). The percentage of stomatal limitation of A during the opening response to A 95 (Fig. 4) demonstrated significant variation among species (P < 0.0001, F (14,49) = 27.368), with values ranging from 6.5 (P. sativum) to 24.3% (G. biloba). With the exception of G. biloba, which was statistically different from all other species, the limitation was 10-15%. Values of A 95 also provided a set point from which to Table 2 Correlation matrix between parameters (grey cells) describing the temporal response of g s during opening and closing of elliptical-(upper triangle of the matrix) and dumbbell-shaped (lower triangle of the matrix) guard cells (see Table 3 G smax , predicted steady-state g s under 1000 lmol m À2 s À1 PPFD; k i, time constant for g s to increase to G smax under 1000 lmol m À2 s À1 PPFD; k d , decrease from G smax to G smin under 100 lmol m À2 s À1 ; k, initial lag in the response time of g s to a step increase in PPFD; Sl max, maximum rate of g s opening to an increase in PPFD from 100 to 1000 lmol m À2 s À1 . Anatomical parameters of stomatal density (SD), pore length (PL) and guard cell width (GCW) were also compared. Significance: *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant.

Research
New Phytologist quantify the g s response after A achieved near maximum steady state (Fig. 2, dotted line) with the majority of species with elliptical-shaped stomata showing an 'overshooting' in stomatal opening, demonstrated by a significant increase in g s (P < 0.01) between 10 and 125 mmol m À2 s À1 (Fig. 3a). Species with dumbbell-shaped stomata (with the exception of T. aestivum; A. sativa) and G. biloba showed no or only a small 'overshoot' (< 3 mmol m À2 s À1 (Fig. 3a)).

Quantifying W i responses to a step change in PPFD
The consequence of the lack of synchrony between the responses of A and g s to a step increase in PPFD can be illustrated by the temporal responses of W i (Figs 5, S4). Following the step increase in PPFD, A rapidly increased compared to g s (Fig. 2) and W i reached a maximum value (W imax ) well before A 95 was achieved. The first 20 min of this response is shown in Fig. S4. The subsequent further increases in g s drove a continuous decrease in W i until both A and g s reached steady state. In most of the C 3 species, W i continued to decrease after A had reached a steady state due to the continued increase in g s . W imax represents the greatest CO 2 uptake for g s ; however, it should be noted that this value occurs earlier in the transient response before A 95 is reached and that W imax is achieved only for an extremely brief period of time. The diversity in W i between species was determined by G smax rather than A max as no correlation between A max with W imax was observed, whereas G smax was negatively correlated with W imax (P < 0.0001, r s = À0.67) and with W i before the decrease in PPFD (P < 0.0001, r s = À0.74). For example, G. biloba, S. bicolour, M. nepalensis and Z. mays achieved and maintained the highest W i (P < 0.001) with vastly different values of A by maintaining a relatively low g s compare to other species. The balance between CO 2 fixation and water loss was different between species, revealed by the four-fold difference in W imax observed Fig. 4 Percentage limitation of net CO 2 assimilation (A) by stomatal conductance (g s ) after the 60 min at 1000 lmol m À2 s À1 PPFD. Data are the mean AE SE (n = 3-5). Species in bold have dumbbell-shaped guard cells, species underlined have a C 4 metabolism and species in plain font have elliptical guard cells and C 3 metabolism (see Table 3 for species name abbreviations).

Fig. 5
Normalized temporal response of intrinsic water-use efficiency (W i ) of 15 species to an increase in irradiance from 100 (shade area) to 1000 (white area) lmol m À2 s À1 . Data are the mean AE SE (n = 3-5). The initial and maximal average values of W i are indicated above the x-axis for each species and a dashed line denotes net CO 2 assimilation rate at 95% of maximum (A 95 ). Values were normalized to the initial values at 100 lmol m À2 s À1 PPFD and maximum values at (1000 lmol m À2 s À1 PPFD) (see Table 3 for species name abbreviations). between the 15 species, with the lowest values observed in H. vulgare (0.054 AE 0.006 lmol mmol À1 m À2 s À1 ). On average, the percentage decrease between W imax and W i at the end of the response under 1000 lmol m À2 s À1 PPFD was significantly less in species with dumbbell-shaped guard cells (P < 0.0001) than in species with elliptical-shaped guard cells. The temporal response of W i is driven by the temporal variation in g s to increasing PPFD that is uncoordinated with A, resulting in unnecessary water loss (Fig. 2). To investigate the theoretical variation of g s required to optimize W i , if coordinated with A, a model with a constant W i chosen at A 95 (W i95 ) was applied and the difference between observed and modelled g s assessed (Fig. 6), with modelled values of g s greater than observed signifying 'water saving' and values less than observed representing a 'water loss'. Figure 6(a,b) show examples for T. aestivum and V.faba. During the first part of the response, the amount of potential 'water saved' was not significantly different between species (0.98-17.3%) (Fig. 6c). However, after W i95 was achieved, a significant difference in 'water loss' between species, in terms of percentage change in g s (P < 0.0001, F (14,49) = 3.454) was observed. For example, in P. vulgaris, g s increased by 57% (AE 23%) for only a 5% gain in A, which illustrates the strong uncoupling of A and g s in this species and the negative impact on W i (Figs 3a, S5). By contrast, the observed response of g s in S. bicolor was close to the modelled optimal g s with minimal increases in g s once A 95 was reached (Figs 3a, 6c).
The results revealed that W imax and A 95 were not reached at the same point during the temporal response (Fig. 5). To reach A 95 (denoted by the dotted line, Fig. 5), the species typically displayed a decrease in W i from W imax . The percentage increase in A from 100 to 1000 lmol m À2 s À1 PPFD was significantly greater (P < 0.01) than the percentage decrease in W i . The highest gains in A were observed for G. biloba and the species with dumbbellshaped guard cells (with the exception of M. nepalensis) which all achieved > 30% increase in A (Fig. S5).

Anatomical features
Stomatal density was significantly different between species with abaxial stomatal densities ranging from 68.5 to 376.3 mm À2 and adaxial densities between 0 and 281.6 mm À2 (Fig. S6) A positive correlation between ad-and abaxial density for species both with elliptical-(R 2 = 0.87) and dumbbell-shaped (R 2 = 0.79 excluding M. nepalenis) guard cells was observed. When considering both types of guard cells, a strong correlation between abaxial and adaxial values for stomatal density (R 2 = 0.76 excluding M. nepalensis), pore length (PL; R 2 = 0.79) and guard cell width (GCW; R 2 = 0.85) was also observed, and therefore mean values were used to correlate with stomatal response traits. A strong correlation between PL and GCW was observed and hence only PL was used for further analyses.
With reference to opening responses of elliptical-shaped guard cells, no significant relationships were found between the anatomical features (PL and SD) and Sl max , k or G, whereas in dumbbell-shaped guard cells, PL and SD correlated significantly with G smax , and k i was correlated with PL but not with SD. The same correlations were observed with reference to closing responses in both guard cell types; however, a significant relationship between SD and k i was also observed in dumbbell-shaped guard cells (Table 2).

Discussion
As light changes rapidly and is often considered the most dynamic and most important environmental variable influencing both stomatal behaviour and photosynthetic rate, we examined the kinetics of photosynthesis (A) and stomatal conductance (g s ) Fig. 6 Examples of observed (dotted lines) and modelled (dashed lines) temporal response of stomatal conductance to water vapour (g s ) for (a) wheat (Triticum aestivum) and (b) broad bean (Vicia faba). The modelled data represent g s at constant water-use efficiency (W i ) achieved at 95% net CO 2 assimilation (A 95 ). The light grey shading represents water loss and the dark grey shading represents water conserved. (c) The percentage change in water loss (light grey) and water conserved (dark grey) for a 5% increase in A derived from the differences in observed and modelled for each species. Data are the mean AE SE (n = 3-5). Species in bold have dumbbell-shaped guard cells, underlined species have a C 4 metabolism and species in plain font have elliptical-shaped guard cells and C 3 metabolism (see Table 3 for species name abbreviations). (2016)

Research
New Phytologist to a step increase followed by a decrease in photosynthetic photon flux density (PPFD), in a number of species; assessing the speeds of the g s response, the amplitude of change, g s limitation of A and the impact of these kinetics on intrinsic water use efficiency (W i ). The temporal dynamics showed clear species-specific differences and noncoordination between A and g s , with g s exhibiting a slower and more varied response than A. Such uncoordinated A and g s responses could have significant implications for cumulative carbon assimilation and transpirational water loss, especially in dynamic light environments. For example, Lawson & Blatt (2014) modelled synchronous g s and A behaviour and calculated a theoretical 20% increase in water use efficiency if g s responded instantaneously to the changes in PPFD and matched mesophyll demands for CO 2 .
A combination of rapid responses and high steady-state values of g s reduce CO 2 diffusional limitations of A, but can also drastically reduce W i , due to the nonlinear relationship between A and g s (Wong et al., 1979). We show for example that the high steady-state values and rapid responses observed in Oryza sativa, Avena sativa and Triticum aestivum facilitated high photosynthetic rates but ultimately resulted in low W i , which may be indicative of traditional breeding and selection practices for high yield at the expense of water loss (Jones, 1987). Although high g s reduces W i , it is also possible that, under well-watered conditions, such stomatal behaviour would increase overall photosynthetic carbon gain by enabling plants to opportunistically use sun flecks in the canopy that can occur on a timescale of seconds to hours (Chazdon & Pearcy, 1991;Kirschbaum et al., 1988;Pearcy, 1990;Way & Pearcy, 2012). Under the measurement conditions used here, when PPFD was raised A immediately (within 1 min) increased in all species, indicating that g s at the lower light level was greater than required. However, a clear stomatal limitation of A was also apparent as all species took > 9 min to reach 95% final A (A 95 ) (Fig. 3d).
A common feature of g s dynamics was the noncoordination in A and g s responses and the continued stomatal opening after A 95 had been reached (or 'overshooting' of g s ; Fig. 3a), resulting in decreases in W i . The observed diversity in responses of A and g s in the species measured questions the mechanisms that coordinate these parameters. Intercellular CO 2 concentration (C i ) was originally proposed as the mediator for the close correlation between g s and A (Wong et al., 1979;Farquhar & Wong, 1984;Mansfield et al., 1990;Buckley et al., 2003). It was assumed that stomata adjust to a steady-state aperture to maintain C i at 2/3 atmospheric [CO 2 ] (Ehleringer & Pearcy, 1983) and therefore, when A is increasing the resulting decrease in C i would cause stomata to open and vice versa. When A reaches steady state, further increases in g s would result in a greater C i that cannot increase A and therefore, following the C i hypothesis, no further increases in g s would be expected once steady-state A has been achieved. However, our results do not fully support this conclusion (e.g. Vicia faba in Fig. 2) and agree with findings from work on transgenic plants with reductions in photosynthesis, which showed increasing g s with light despite high C i (Von Caemmerer et al., 2004;Baroli et al., 2008;Lawson et al., 2008). Many studies support C i -driven stomatal responses (e.g. Roelfsema & Prins, 1995) and we do not argue against CO 2 as a driver; however, our results show that C i is clearly not of high priority in the hierarchy.
Stomata have been a key target for improving plant water use efficiency (WUE) and/or a plant's ability to cope with reductions in water availability. However, improvements of WUE in crop plants often come at the expense of photosynthetic rates (Yoo et al., 2009(Yoo et al., , 2010 and are therefore of limited value, given that current global research efforts focus on increasing crop yield for sustainable food and fuel production (Long et al., 2015). However, as we have illustrated here, W imax does not correspond to maximum assimilation rate (Fig. 5) as maximum WUE can often only be achieved when g s restricts A. Based on these observations of dynamic responses we could suggest a steady-state g s target that would provide a compromise between A and W i and propose that this target should be the lowest g s value that enables A 95 to be achieved. It should be noted that this is an optimal target and that fluctuations in the environment could result in different integrated values of A, g s and therefore W i , highlighting the importance of appreciating the speed of stomatal responses and coordination between A and g s .
It is well known that significant variation in photosynthetic capacity (A max ) exists both within and amongst different species (Lawson et al., 2012) and, as observed here, this is generally correlated with steady-state g s (and G smax ). As may be expected the C 4 species measured in our study were able to achieve a greater A 95 at a lower g s (at A 95 ) compared to C 3 species (Fig. S3), and it is likely that the faster stomatal opening and closing responses observed in C 4 species (Fig. 3c) facilitated this greater level of coordination between A and g s (Fig. 2). This faster response was a common feature not just in C 4 plants, but also C 3 species with dumbbell-shaped guard cells. However, despite the close coupling in C 4 species, the same stomatal limitation on A of c. 10% or greater was observed in both C 4 and C 3 plants (Fig. 4). This illustrates the importance of considering both CO 2 uptake and water loss when evaluating steady-state or transient W i (McAusland et al., 2013), as maximum W i is often not observed at maximum A. Here the decrease in W i (between W imax and A 95 ) with increasing g s was outweighed by a substantial gain in A in all species.
The temporal uncoupling between stomatal behaviour and carbon demand observed in many species can be evaluated by comparing measured g s responses with those modelled assuming a stable W i (at A 95 , which represented a W i value that is achieved without a limitation on A), providing an estimate of the differences between variable and stable W i in terms of water gain and expense for each species. Using this model over the period measured, stomatal behaviour in the majority of species resulted in water expense exceeding water conservation, illustrating that the latter was not the priority. As all the plants in these experiments were maintained under relatively well-watered conditions, this would have led to a higher stomatal conductance than would be observed in plants experiencing water limitation (Comstock & Ehleringer, 1993;Mott & Peak, 2012;Lawson & Blatt, 2014).
In general, C 4 species were an exception to this and either demonstrated a balanced water budget or greater gains than losses, further exemplifying the more synchronous A and g s responses observed in the three species measured. The most likely explanation for the greater loss of water in C 3 species is the substantial overshoot in g s after A 95 had been achieved, which was not apparent in the two C 4 species studied (Figs 3a, 6). However, C 3 species with rapid stomata responses (e.g. O. sativa) also exhibited a positive water balance, whilst species with the slowest stomatal opening (with the exception of G. biloba) demonstrated the most negative water balances (Fig. 6c) hinting at the possible existence interspecific diversity of stomatal control.
The rapidity of response for stomata to open (increase; k i ) and close (decrease; k d ) was positively correlated for species with elliptical-as well as dumbbell-shaped guard cells, suggesting that similar mechanisms or pathways were involved in both opening and closing responses. Overall, significant asymmetry of the stomatal responses revealed a faster closing than opening, which has previously been associated with conserving water (Tinoco-Ojanguren & Pearcy, 1993;Ooba & Takahashi, 2003). In comparison, species with dumbbell-shaped guard cells displayed the fastest responses and most had greater similarity in the rapidity of opening and closing, which is consistent with the fact that these guard cells require fewer solutes and less water to achieve a given unit increase in aperture (Franks & Farquhar, 2007;Raven, 2014). The rapidity of increasing g s impacts on A, with species with high k i taking longer to achieve A 95 , as low g s restricts CO 2 diffusion. Under field conditions with a dynamic light environment, slowly responding stomata could restricted CO 2 uptake and thus have a compound effect on the cumulative A over the growing season and affect yield (Reynolds et al., 1994;Fischer et al., 1998). However, slow stomatal closure would negatively impact on W i when environmental conditions reduce A. It should be noted that under field conditions, changing the light environment also results in changes in leaf temperature. A direct impact of increasing PPFD would be an increased leaf temperature, which would lead to higher leaf-toair vapour pressure difference and thus exacerbate the transpirational losses of 'overshooting' stomata. However, higher transpirational losses would have a cooling effect. Therefore, concomitant temperature variation could have complex effects on the dynamic responses of A, g s and W i , and should be studied in detail using appropriate experimental set-ups.
Variation between species was also observed in the maximum speed of g s response (Sl max ) and the rapidity of opening (k i ) to achieve steady-state conductance. Previous research has associated the speed of stomatal responses with the size of stomata, with smaller stomata facilitating rapid opening and closing (Hetherington & Woodward, 2003;Franks & Farquhar, 2007;Franks & Beerling, 2009;Drake et al., 2013;Raven, 2014).The majority of these studies have used the maximum slope (Sl max ) as a measure of the maximum speed of response; however, this measurement is also dependent on the amplitude of the response (Eqn 2). Additionally, because g s is determined by both stomatal aperture and density, small changes in aperture in plants with smaller, more numerous stomata will have a greater Sl max for the same change in aperture as species with fewer larger stomata.
Therefore, although Sl max may provide a useful comparative measure within species (in which anatomical features and the scales of stomatal responses are similar), it is not a useful parameter to compare speeds of response between species with different anatomical features and magnitudes of change. To address this issue, we used the time constant k to provide a measure of the rapidity of g s, independently of the magnitude of response and the absolute g s values observed. For elliptical-shaped guard cells, we did not detect any significant correlation between stomatal density and Sl max or k, suggesting that on an interspecific basis neither the speed nor the amplitude of the stomatal responses to PPFD were dependent on stomatal density. On the other hand, for dumbbell-shaped guard cells, variation among species in stomatal size impacted on the speed and amplitude of stomatal responses. This hints at the possibility that for elliptical-shaped guard cells attributes other than anatomical features are important contributors to the speed of stomatal responses (Hetherington & Woodward, 2003;Franks & Beerling, 2009) such as membrane permeability due to ion channels number or distribution (see discussion, Lawson & Blatt, 2014). By contrast, for dumbbell-shaped guard cells, anatomical variations seem to impact the rapidity of their response. Additionally, these species were also able to achieve a greater Sl max and tended to be faster. This may be due to the energetic requirements for stomatal movement of the often smaller dumbbell-shaped guard cells (Grantz & Assmann, 1991;Franks & Farquhar, 2007;Raven, 2014). The dumbbell-shaped design means that small changes in width can cause larger changes in stomatal aperture and maximize the potential of these stomata to track changes in environmental conditions (Hetherington & Woodward, 2003). Although transients of leaf-level W i provide insight into potentially optimizing stomatal behaviour, numerous other processes contribute to W i in the field. Manipulation of the speed of g s provides scope for improving carbon acquisition in fluctuating light environments but also enhances drought tolerance through improved conservation of water. Integration of these dynamic responses over daily or seasonal time periods is complex and would require a model that includes respiration (both from leaves and parts including stems and roots) and transpiration as a product of changes in diurnal saturation deficit (Cowan & Farquhar, 1977;Farquhar et al., 1989;Jones, 2004) Identifying varieties or genotypes with more rapid stomatal responses could be used as an optimizing strategy for whole-plant water use over the growing period, potentially improving the ability of the plant to adapt to changing environments (Schulze & Hall, 1982;Campitelli et al., 2016) which could feed forward to maintain or improve yields (Chaerle et al., 2005;Lawson & Blatt, 2014).

Conclusion
This is one of the few studies to investigate temporal responses in A and g s in relation to carbon assimilation and W i , and illustrates significant species-specific variation in the speed of stomatal responses and magnitude of change, as well as coordination with A. Slow stomatal responses can limit A by c. 10%, which could equate to substantial losses in photosynthetic rates, productivity

Research
New Phytologist and reductions in yield. Previous research focusing on improving productivity has shown that by enhancing photosynthesis by only 2-3%, substantial increases in plant growth and biomass can be achieved over the season (Lefebvre et al., 2005;Zhu et al., 2007;Simkin et al., 2015). The work presented here illustrates that similar short-term improvements in A could be gained by improving the rapidity of stomatal responses and coordination with A. Tighter coupling between stomata and A therefore has the potential to achieve a substantial improvement in WUE, as in the present study, overshooting of g s by up to 80% was observed for only a 5% gain in A and fast closing responses resulted in substantial saving in water loss.
Our findings support faster responses in dumbbell-compared with elliptical-shaped guard cells and suggest that photosynthetic type (C 3 /C 4 ) also plays a role. The speed of stomatal responses might not be dependent on the same underlying processes when comparing elliptic-and dumbbell-shaped guard cells, with physiological processes being more important for the former and anatomical features for the latter. Improving the rapidity of stomatal responses could greatly improve productivity and W i but achieving this will require greater knowledge of the physiological and molecular mechanisms that determine the speed of stomata and coordination with mesophyll demands for CO 2 , and further field-based measurements that integrate the dynamics of A, g s and W i over seasons.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information tab for this article: Fig. S1 Response of stomatal conductance to water vapour (g s ) of 15 species to an increase in irradiance from 100 to 1000 lmol m À2 s À1 PPFD.

Fig. S2
Response of net CO 2 assimilation (A) of 15 species to an increase in irradiance from 100 to 1000 lmol m À2 s À1 PPFD.