The impact of stomatal kinetics on diurnal photosynthesis and water use efficiency under fluctuating light

Dynamic light conditions require continuous adjustments of stomatal aperture. As stomatal conductance (gs) kinetics are a magnitude slower than photosynthesis (A), they are hypothesized to be key to plant productivity and water use efficiency. Using step-changes in light intensity, we studied the diversity of light-induced gs kinetics in relation to stomatal anatomy in five banana genotypes (Musa spp.) and modelled the impact on A and intrinsic water use efficiency (iWUE). Banana generally exhibited a strong limitation of A by gs, indicating a priority for water saving. Significant genotypic differences in gs kinetics and gs-based limitations of A were observed. For two contrasting genotypes the impact of differential gs kinetics on A and iWUE was further investigated under realistic diurnally fluctuating light conditions and at whole-plant level. Genotype-specific stomatal kinetics observed at the leaf level were corroborated at whole-plant level, suggesting that despite differences in gs control at different locations in the leaf and across leaves, genotype-specific responses are still maintained. However, under diurnally fluctuating light conditions gs speediness had only a momentary impact on the diurnal iWUE and carbon gain. During the afternoon there was a setback in kinetics: the absolute gs and the gs responses to light were damped, strongly limiting A and the diurnal iWUE. We conclude that the impact of the differential gs kinetics on the limitation of A was dependent on the target light intensity, the magnitude of change, the gs prior to the intensity change and particularly the time of the day. One sentence summary Genotype-specific stomatal rapidity is for the first time validated at whole-plant level, but under fluctuating light the impact of stomatal dynamics depends on other factors like the time of the day.

Introduction In order to survive, plants need to balance CO 2 uptake for photosynthesis with water loss via 48 transpiration. By adjusting their aperture, stomata control gaseous exchange between the leaf 49 interior and the external atmosphere. Stomatal aperture is adjusted by moving solutes into or out 50 of the guard cells. These changes in osmotic potential elicit water movement in or out of the 51 guard cells, altering turgor pressure and subsequently aperture. In general, stomatal opening in 52 well-watered C3 and C4 species is triggered by high light intensity, low VPD and low CO 2 53 concentrations. Opposite environmental conditions (low light, high VPD and high [CO 2 ]) 54 stimulate stomatal closure (Assmann and Shimazaki, 1999;Outlaw, 2003;Lawson and Morison, 55 2004). Therefore, in a dynamic field environment, stomata are continuously adjusting aperture to 56 achieve an appropriate balance between carbon gain and water loss (Pearcy, 1990;Lawson and 57 Blatt, 2014). However, most research has studied stomatal conductance (g s ) and photosynthesis 58 (A) under steady-state conditions. A high g s under steady-state conditions is associated with high 59 A and consequently improved growth (Fischer et al., 1998;Franks, 2006). However, as g s 60 kinetics are a magnitude slower than those of A, the speed in which these steady-state values are 61 reached in a fluctuating environment have a great influence on the growth and water use 62 efficiency (Lawson and Blatt, 2014;Kaiser et al., 2016;McAusland et al., 2016;Taylor and 63 Long, 2017;De Souza et al., 2020;Yamori et al., 2020). In a fluctuating field environment, light 64 intensity is one of the most variable environmental conditions as it changes continuously by 65 moving cloud covers and shading from adjacent plants (Pearcy, 1990;Slattery et al., 2018;66 Morales and Kaiser, 2020). In this way, stomata frequently experience alternating light 67 intensities, inducing stomatal responses that change A, g s and the ratio of these, the intrinsic 68 water use efficiency ( i WUE). The balance between CO 2 gain and H 2 O loss under changing light 69 intensities is disturbed by delayed g s responses (Vialet-Chabrand et al., 2017;Slattery et al., 70 2018). The slower g s increase to increased light intensity limits the CO 2 uptake for A, while the 71 slower g s decrease to decreased light intensity results in unnecessary water loss. The limitation of 72 A by the slower kinetics of g s has been shown to be significant in well-watered C3 species 73 (Farquhar and Sharkey, 1982;Jones, 1998;Lawson and Blatt, 2014;McAusland et al., 2016). 74 Rapid g s kinetics therefore have been hypothesized to maximize A and i WUE, as steady-state 75 values under the new conditions can be rapidly achieved (Lawson and Blatt, 2014;Papanatsiou 76 et al., 2019;Kimura et al., 2020;De Souza et al., 2020). The g s kinetics are, together with the 77 Initial steady-state stomatal conductance at 100 μmol m -2 s -1 (g s,100 ) ranged from 0.016 to 0.032 117 mol m -2 s -1 (Fig. 2a, Supplemental Table S1). Following the step increase in light to 1000 μmol 118 m -2 s -1 for 90 min average g s,1000 ranged between 0.13 and 0.15 mol m -2 s -1 (Fig. 2a, Supplemental 119 Table S1). The differences observed at both g s,100 and g s,1000 were not significant between 120 genotypes (Fig. 2a, Supplemental Table S1). 121 The speed of g s increase varied strongly between the Musa genotypes and the modelled variables 122 differed significantly Supplemental Table S1). The genotype with the slowest g s 123 increase, Cachaco, had an average time constant K i of 17 min, while the fastest genotype, 124 Mbwazirume, had a K i of 6.4 min (Fig. 2b, Supplemental Table S1). The speed of the decrease in 125 g s (K d ) was also genotype-dependent (Fig. 2c, Supplemental Table S1). K d was about 2 fold 126 higher in Cachaco (9.5 min) than in Mbwazirume (4.4 min). Across all genotypes, K i was 127 significantly correlated with K d (R² = 0.41, P < 0.001; Fig. 2d & Supplemental Fig. S1). 128 However, the decrease in g s was significantly faster than the increase (P < 0.001). The maximal 129 slope of g s increase and decrease (Sl max,i and Sl max,d  response to a step increase and decrease in light intensity between 100 and 1000 µmol m -2 s -1 for 140 five different banana genotypes (n = 7-8). (a) Steady-state g s at 100 (g s,100 faded colours) and 141 1000 µmol m -2 s -1 (g s,1000 bright colors). (b) Time constant of g s increase (K i ) for different 142 genotypes. Different letters indicate significant differences between genotypes (P < 0.05; 143 A>B>C). (c) Time constant of g s conductance decrease (K d ) for different genotypes. Different 144 letters indicate significant differences between genotypes (P < 0.05; A>B>C). (d) Significant 145 correlation between K i and K d (R² = 0.41, P < 0.001). K i was significantly higher than K d . The 146 dashed line shows the 1:1 line. Points and error bars represent mean ± SE (n = 7-8). 147 148 Impact of stomatal opening and closing speed on A 149 The speed of the increase in g s following a step-change in PPFD from 100 to 1000 µmol m -2 s -1 150 strongly determines CO 2 uptake during this period. The time to reach 95% of maximum A was greater than 30 min for almost all genotypes and differed significantly between Cachaco (51.8 152 min) and the genotypes Mbwazirume (30.3 min) and Banksii (29.5 min) (Fig. 3a The step increase in light intensity induced an initial increase in A that was relatively larger than 167 the increase in g s . These responsiveness differences increased i WUE, reaching the maximum 168 i WUE ( i WUE max ) during the light period in all cases within 6.5 min (Supplemental Fig. S4). 169 i WUE max was reached earlier than steady-state A, showing that the maximal i WUE and maximal 170 increased (Supplemental Fig. S4). i WUE only stabilized when both A and g s reached steady-state. 172 The genotype Cachaco had a significantly higher cumulative i WUE during the high light period 173 compared to Mbwazirume (Supplemental Fig. S5). The cumulative i WUE was significantly 174 correlated with the time constant K i and Sl max,i with slower g s responses resulting in higher i WUE 175 (R² 0.19 & 0.41, P < 0.01, Supplemental Fig. S1). The reduction in light intensity from 1000 to 176 100 µmol m -2 s -1 instantaneously lowered i WUE as A immediately declined because of light 177 limitation (Supplemental Fig. S4). The cumulative i WUE during this low light period was 178 significantly higher in Kluai Tiparot, than in Leite (Supplemental Fig. S5

182
Banana has elliptical-shaped guard cells surrounded by four to six subsidiary cells (Rudall et al., 183 2017). Abaxial stomatal density, stomatal length, guard cell size and subsidiary cell size were 184 quantified from the leaf part enclosed in the gas exchange cuvette and significant differences 185 between genotypes were observed (Supplemental Fig. S6). Overall these anatomical 186 characteristics were not correlated with any of the modelled light-induced g s kinetics (Fig. 4, 187 Supplemental Fig. S1). However, several correlations between anatomy and g s kinetics were 188 significant if the genotype Cachaco with lowest g s rapidity was not considered (Fig. 4). In this 189 case, stomatal density was significantly correlated with the time constant K as well as the 190 maximum slope of g s response Sl max during both stomatal opening and closing (P < 0.01; R² 191 ranging between 25 and 46 %). after c. 22 min, while in Cachaco this was only after 35 min ( Fig. 5a-b). These results were also 204 reflected in the transpiration rate before and after dawn. The whole plant transpiration rate did 205 not differ significantly between both genotypes pre-dawn, but after the onset of light, the 206 transpiration rate was significantly higher in Mbwazirume (Fig 5c). 207 indicate the time before dawn. '.' for P < 0.1, * for P <0.05, ** for P <0.01, *** for P <0.001. 220

221
To evaluate the impact of g s kinetics on diurnal A and i WUE, plants were subjected to fluctuating 222 light intensities and phenotyped over an entire diurnal period. Similar to the transpiration rate 223 measured at the whole-plant level, the morning increase in g s at leaf-level under gradually 224 increasing light intensity was faster in Mbwazirume compared to Cachaco (Fig. 6a). The time 225 constant for the g s increase (K i ) was significantly higher in Cachaco (P < 0.005, Fig. 6b). 226 However, the faster increase of g s in Mbwazirume, did not result in increased A (Fig. 6c). Throughout the day, g s kinetics were in most cases significantly faster for the genotype 245 Mbwazirume compared to Cachaco (Fig. 8a), again confirming the previously observed kinetics 246 (Fig. 2, Fig. 5). However, under fluctuating light conditions g s kinetics were dependent on the 247 magnitude of light intensity change, g s values prior to the light intensity change and the time of 248 the day (Fig. 8a). During the afternoon there is a setback in kinetics: the absolute g s and the g s 249 responses to light are damped, strongly limiting A (Fig.7, Fig. 8). The limitation of A by g s in the 250 afternoon was three times higher in Cachaco (52.6 %) compared to Mbwazirume (17.5 %) (Fig.  251   7, Fig. 9d). The reduction of g s in the afternoon resulted in a significantly lower average diurnal 252 g s (Fig. 9a) which translated into a greater diurnal i WUE in Cachaco compared to Mbwazirume 253 ( Fig.8c, Fig. 9c). Step-changes in light intensity have shown to induce an uncoupling of A and g s in many of 281 species (Barradas and Jones, 1996;Lawson and Blatt, 2014;McAusland et al., 2016;Faralli et 282 al., 2019a). However, all Musa genotypes maintain a tight coupling between A and g s following a 283 step increase in light intensity (Fig. 1). This indicates a strong stomatal control of photosynthesis 284 and overall high limitation of A by g s with an average limitation of about 13 % (Fig. 3). This 285 behaviour shows that banana strongly controls stomatal aperture, resulting in water conservation 286 at the expense of potential carbon gain. This prioritizing of water conservation in banana can be 287 explained by its intrinsic need to maintain a high leaf water potential (Turner and Thomas, 288 1998). 289

Diversity in light-induced stomatal responses
290 Stomatal responses to changes in light intensity have been shown to vary at an inter-and intra-291 specific level (Vico et al., 2011;Drake et al., 2012;McAusland et al., 2016;Qu et al., 2016;292 Durand et al., 2020;De Souza et al., 2020). A higher steady-state g s has been linked with faster 293 light-induced g s responses (Drake et al., 2012;Kaiser et al., 2016;McAusland et al., 2016;294 Wachendorf and Küppers, 2017). Although the differences observed in steady-state g s values 295 between banana genotypes were not significant, their g s kinetics differed strongly (Fig. 2). These 296 results suggest that other factors such as stomatal anatomy, hydraulic conductance and 297 membrane transporters are involved in determining the rapidity of changes in g s . 298 Within the Musa family we observed significant differences in the speed of increase and 299 decrease in g s (Fig. 2b,c). Differences across genotypes were not explained by genomic 300 constitution, which is in agreement with the wide diversity of transpiration phenotypes observed 301 irrespective of their genomic constitution (van Wesemael et al., 2019). Consistent with previous 302 works in other species (Vico et al., 2011;McAusland et al., 2016;Faralli et al., 2019a), the speed 303 of g s increase and decrease were significantly correlated (Fig. 3d). This correlation could suggest 304 that the genotypic differences in net solute flux across guard and tonoplast membranes are the 305 main bottlenecks and are maintained in both opening and closing response. Differences in these 306 fluxes could be attributed to differences in the number or activity of such membrane transporters. 307 Decreases in g s were faster than opening in all Musa genotypes (Fig. 3d), which is not the case for all crops (McAusland et al., 2016;Qu et al., 2016). The faster g s closure again indicates that 309 Musa prioritises water conservation over maximization of carbon uptake. 310 The two most extreme genotypes Cachaco and Mbwazirume, with the slowest and fastest g s 311 responses respectively, also showed at the whole-plant level, differences in light-induced 312 transpiration rate (Fig. 5). This finding suggests that despite possible differences in g  . 6a). This faster g s increase in Mbwazirume did not result in higher A, indicating that at 320 dawn, under low light intensities, g s was not limiting A and was higher than necessary for 321 maximal A (Fig. 6 & Fig. 7). These results demonstrate that the impact of g s kinetics on A and 322 i WUE depend on the time of the day and the light conditions. The uncoupling of g s and A under 323 realistically increasing light conditions at dawn was not beneficial for carbon uptake. Gosa et al. 324 (2019) called this period after dawn in tomato the golden hour because in dry climates it is the 325 time of the day with the highest g s . Later in the day, VPDs become too high, restricting g s (Gosa 326 et al., 2019). Breeding for an even higher g s during this golden hour was suggested to improve 327 plant productivity. However, care must be taken to breed for an improved morning CO 2 uptake, 328 rather than for a high g s with associated uncoupling of A and g s . Although the absolute water loss 329 resulting from excessive morning g s might be relatively low because of low evaporative demands 330 at dawn (Chaves et al., 2016), it may lead to a crucial decrease in overall plant water status. 331 Despite the confirmed genotypic differences in stomatal kinetics, the impact of g s kinetics on A 332 and i WUE before noon hardly differed between the genotypes Cachaco and Mbwazirume under 333 field-mimicking light conditions (Fig. 7, Fig. 8). This could be explained by lower amplitudes of 334 light switches compared to a single step-change in light intensity and/or g s values not being at 335 steady-state prior to changing light intensity. The genotype-specific speed of the g s response 336 observed under a single step-change in light intensity did not explain the diurnal i WUE, 337 indicating that g s kinetics only partially affect diurnal water use efficiency and carbon gain (Fig. 9b,c). The absolute g s and the g s responses to light decreased strongly in the afternoon, and this 339 effect was more pronounced in the genotype Cachaco (Fig. 7, Fig. 8a). The three times higher 340 afternoon g s limitation of A in the genotype Cachaco compared to Mbwazirume, resulted in a 341 significant higher diurnal i WUE (Fig. 9c,d). The genotype Cachaco with the slowest g s kinetics 342 thus achieved the highest i WUE, showing that not only g s speed but also the diurnal pattern 343 determines the overall water use efficiency and carbon gain. We show for that under fluctuating 344 light conditions this intrinsic diurnal pattern of absolute g s decrease and g s light responsivity 345 reduction is decisive for diurnal i WUE (Fig. 9c). 346 (2019a) reported no or only a weak inter-and intra-specific correlation between stomatal 350 anatomy and light-induced g s kinetics. We confirmed that stomatal density and size were not 351 correlated with the g s kinetics (Fig. 4, Supplemental Fig. S1). Remarkably, the genotype with the 352 slowest increase in g s , Cachaco had the second highest density and the smallest stomata. Without 353 this genotype a significant correlation between density and the speed of g s increase and decrease 354 was observed (Fig. 4). This exception proves that the surface-to-volumes ratios are not always 355 directly related to stomatal speed as this assumes uniform ion transport activity per surface area 356 Our findings prove that there is diversity in g s rapidity to light within closely related genotypes. 360

Impact of stomatal anatomy on responses
The priority of banana for water saving is shown by strong stomatal control of A and faster 361 decrease in g s than increase. The observed diversity in g s rapidity was not related to stomatal 362 density or subsidiary cell sizes and therefore suggests that variation is driven by functional 363 components rather than anatomy. We show here for the first time that the g s rapidity observed at 364 the leaf level can also be found at the whole-plant level. However, under fluctuating light 365 conditions, g s rapidity is only one of the many physiological factors determining overall plant 366 water use efficiency and carbon gain. Leaf gas exchange measurements 382 Photosynthetic rate (A) and stomatal conductance to water (g s ) were measured every 30 s on the 383 middle of the second youngest fully developed leaf using a LI-6400XT infrared gas analysis and 384 dew-point generator model LI-610 (LI-COR, USA). Light was applied by an integrated LED 385 light source. The leaf cuvette maintained a CO 2 concentration of 400 µmol mol -1 , a leaf 386 temperature of 25°C and a VPD of 1 kPa. All measurements were performed before 14:00 h to 387 avoid circadian influences on A by g s . 388 Stomatal response to a step-change in light intensity 389 The light intensity was kept at 100 µmol m -2 s -1 until A and g s were stable for 10 min. Once 390 steady-state was reached, light intensity was increased to 1000 µmol m -2 s -1 for 90 min. Then, 391 light intensity was lowered back to 100 µmol m -2 s -1 for 30 min. 392 The increase in g s after the increase in light intensity and the decrease in g s after the decrease in 393 light intensity followed a sigmoidal pattern and was modelled using the non-linear sigmoidal With g s the stomatal conductance at time t, K the time constant for rapidity of g s response (min), 397 λ the initial time lag before g s increase or decrease (min), g s,100 and g s,1000 (mol m -2 s -1 ) the 398 steady-state g s at 100 and 1000 µmol m -2 s -1 respectively. Parameter values were estimated for 399 each individual plant using non-linear model optimization in R (V 3.4.3). K i indicates the g s 400 increase time constant, K d the g s decrease time constant. The maximum slope of g s during 401 opening and closing was calculated and defined as Sl max . Intrinsic water use efficiency ( i WUE) 402 was calculated as i WUE = A/g s . Outlying values (0.5 % quantile; i WUE < 0 or > 400 µmol mol -403 1 ) caused by low g s were discarded for plotting. Photosynthesis was considered to be limited by 404 stomatal conductance during the light-induced stomatal opening until 95 % of steady-state Stomatal impressions of the abaxial surface of the leaf were made when stomata were completely 413 closed using impression material. Impression were made by applying dental polymer according 414 to the protocol of Weyers & Johansen (1985), followed by covering the polymer with nail 415 varnish and placement on a microscope slide. Stomatal anatomy was quantified using an EVOS 416 digital inverted microscope. Stomatal density was determined in three microscopic field of views 417 of 1.12 mm² captured with a 10 x objective lens (54 to 117 stomata per field of view). Guard cell 418 length (µm), guard cell size (mm²) and lateral subsidiary cell size (mm²) were determined in 419 three microscopic field of views of 0.07 mm² captured with a 40 x magnification respectively 420 (four to seven stomata per field of view). Measurements were performed in ImageJ software 421 The daily plant weight was estimated from the projected leaf area using genotype-specific 437 correlations (n > 50; R² > 0.94). Plants were watered with a nutrient solution during the night and 438 kept at well-watered conditions. Radiation was collected every 5 min via a sensor (Skye 439 instruments, UK) inside the greenhouse. Supplemental lighting of 14 W m -² at plant level was 440 provided when solar radiation was below 250 W m -² during the daytime. Temperature and 441 relative humidity data were collected using six data loggers (Trotec, DE) registering data every 5 442 min. The onset of light was defined as the moment when intensity increased above 2 W m -². 443  increase after the increase in light intensity. Different letters indicate significant differences 507 between genotypes (P < 0.05; n = 7-8; A>B). 508 Fig. 4 Relation between abaxial stomatal density and the time constant describing the speed of 509 stomatal conductance increase (K i ) after the light intensity increase from 100 to 1000 µmol m -2 s -510 1 . There was no significant correlation, caused by the outlying genotype Cachaco. Points and 511 error bars represent mean ± SE (n = 7-8). 512 indicate the time before dawn. '.' for P < 0.1, * for P <0.05, ** for P <0.01, *** for P <0.001. 524 intensity fluctuated throughout the day, ranging up to 1500 µmol m -2 s -1 (black line). The 541 significance of the time constant of g s increase or decrease (K) and the maximal slope of g s 542 increase or decrease (Sl max ) is shown. Throughout the day, g s kinetics were faster for the genotype 543 Mbwazirume compared to Cachaco, but differences were dependent on the target light intensity, 544 the magnitude of change, the g s prior to the intensity change and the time of the day. *for P 545 indicate times of darkness. Green areas indicate the analyzed time frame of the g s rapidity 547 response. PPFD, Photosynthetic Photon Flux Density. Data are the mean ± SE (n = 4). 548  Table S1 Modelled steady-state and light-induced variables of the stomatal 567 conductance (g s ) response to a step increase and decrease in light intensity from 100 to 1000 568 µmol m -2 s -1 for five different banana genotypes (Musa spp.). 569