Variation between rice accessions in photosynthetic induction in flag leaves and underlying mechanisms

Significant natural variation was found in rice flag leaves for photosynthetic traits measured in steady- and non-steady-state conditions, with implications for improving water use efficiency and photosynthesis in this key crop.


Introduction
In cereals, the flag leaf is defined as the last leaf to emerge on a mature flowering stem. The flag leaf has a higher photosynthetic capacity relative to lower canopy leaves due to its position at the top of the canopy, which allows for greater This paper is available online free of all access charges (see https://academic.oup.com/jxb/pages/openaccess for further details) interception of light (Adachi et al., 2017). Additionally, high rice yields are closely correlated with the size and health of the flag leaf as they contribute some 50% of assimilates utilized for grain filling (Yoshida, 1981;Ishii, 1993;Nakano et al., 1995;Li et al., 1998). This large contribution of photosynthates is partially due to the proximity of the flag leaf to the grain, as the sink can more easily attract assimilates from closer sources (Sicher, 1993). The rest of the photosynthates used in grain filling are supplied by the leaf immediately below the flag leaf, and by remobilization of stored carbohydrates in leaf sheaths and older senescing leaves (Li et al., 2017;Lin et al., 2018). Agronomic strategies are typically aimed to protect the flag leaf, since its destruction during grain filling is associated with yield losses of up to 45% in rice (Abou-Khalifa et al., 2008).
Consequently, great effort has been dedicated to understanding and optimizing the flag leaf to improve yields (Li et al., 1998). For example, flag leaf size and area can significantly influence grain yield (Li et al., 1998;Zhang et al., 2015). In rice, flag leaves with larger areas are significantly correlated with greater yields and have become a target for breeding programs seeking to achieve an ideal phenotype or ideotype (Zhang et al., 2015). Quantitative trait loci (QTLs) for larger flag leaves have been identified with the purpose of improving yield (Zhang et al., 2015). Additionally, rice yields have been increased by up-regulating NAL1, a gene that affects both flag leaf area and photosynthetic rates (Fabre et al., 2016). Increasing rice flag leaf photosynthesis, through either delaying senescence or a more acute leaf angle, has led to higher yields (Mantilla-Perez et al., 2017). Early flag leaf senescence significantly reduces seed-setting rate, 100-grain weight, and yield (Lin et al., 2018). Conversely, delaying senescence prolongs the period in which the flag leaf is photosynthetically active, resulting in more photosynthates that can fill grain (Ishii, 1993;Kobata et al., 2015;Leng et al., 2017). In previous studies, more vertical flag leaf angles resulted in 13% higher photosynthetic rates, reduced photoinhibition, delayed leaf senescence, and 15% higher yields (Chen et al., 2002;Mantilla-Perez et al., 2017).
Open-air elevation of [CO 2 ] using free-air CO 2 enrichment technology is an artificial means of enhancing photosynthesis season-long in field-grown rice and observing the yield response. Modern high-yielding rice lines that have larger single-grain sizes and are able to produce larger panicles show a greater response to this elevation of photosynthesis, in contrast to older lines which appear partially sink limited. The fact that even under these circumstances only 70-80% of spikelets ripen-and less than in the older rice lines-suggests that these high-yielding lines are actually strongly source limited (Hasegawa et al., 2013;Zhu et al., 2015;Sakai et al., 2019;Ainsworth and Long, 2020). These findings suggest that increasing the photosynthetic capacity of the flag leaf would be particularly valuable for these high-yielding cultivars.
Photosynthetic induction is the process by which leaves begin to assimilate CO 2 upon a transition from low light into high light, and is characterized by a lag in photosynthetic efficiency.
Previously, examination of rice leaves in the vegetative growth stage found greater variation for photosynthetic traits during non-steady-state lighting compared with steady state (Acevedo-Siaca et al., 2020). Despite the focus on flag leaf improvement, no studies have characterized the response of flag leaf photosynthesis in non-steady-state conditions or explored the possibility that there is variation in this character that could be utilized for improvement. Recently, it was shown that rice flag leaves are subject to different endogenous aging programs compared with other leaves (Lee et al., 2017). Given its proximity to the major sink in the panicle, feedback would also be expected through non-structural carbohydrate-driven 'feast' and 'famine' gene expression response networks (Paul et al., 2017). It is therefore likely that photosynthetic properties of flag leaves will differ from those of leaves formed in the vegetative growth stage.
Non-steady-state photosynthesis is important in a field agricultural setting where light is never constant. At the top of the canopy, flag leaves are subject to light fluctuations due to intermittent cloud cover and shadowing by other flag leaves or panicles as wind and sun angles change (Taylor and Long, 2017;Wang et al., 2020). In wheat, time taken for photosynthetic efficiency to recover from transient shadowing over the course of the day in the field was calculated to cost 21% of potential flag leaf assimilation (Taylor and Long, 2017). Improving the rate of recovery in this rarely measured photosynthetic parameter of flag leaves has the potential to increase yield. Limitations to the speed of photosynthetic induction on such shade to sun transitions are due to four main processes: the photoactivation of enzymes involved in the regeneration and production of ribulose-1,5-bisphosphate (RuBP); the buildup of intermediates of carbon metabolism; the activation of Rubisco by Rubisco activase; and stomatal opening (Pearcy et al., 1994;Mott and Woodrow, 2000;Slattery et al., 2018).
This study analyzed the performance of flag leaf photosynthesis in both steady-state and non-steady-state conditions. The objectives were limited to rice flag leaves and aimed to (i) determine the extent of variation between accessions in nonsteady-state and steady-state photosynthetic parameters relating to productivity and resource use efficiency across rice accessions; (ii) examine the response of photosynthetic induction of rice flag leaves at different [CO 2 ] to understand limitations to induction; and (iii) compare the response of photosynthetic induction in rice flag leaves between accessions and with induction in leaves in the vegetative growth stage of the same accessions as reported in Acevedo-Siaca et al. (2020).

Growing conditions and germplasm
Six accessions were selected from the 3000 Rice Genome Project (3K RGP) held at the Germplasm Resources Center at the International Rice Research Institute (IRRI) in Los Baños, Philippines. Accessions were selected on nucleotide mismatches for the gene encoding Rubisco activase, which plays a central role in photosynthetic induction. Seeds were maintained at 50 °C for 1 week to break dormancy and then sown into soil from the IRRI Upland Farm in small pots (4.5 cm diam-eter×12 cm) and fertilized using 0.4 g -1 Osmocote Plus 15-9-12 (The Scotts Company Ltd, Thorne, UK). Seedlings were transferred to larger individual pots (21.5 cm diameter×21.5 cm, 6 liters) after the emergence of the second leaf. These were then placed in a screen house, a type of greenhouse with a glass roof and screen-meshed walls, with no additional lighting or temperature control at IRRI, during the Philippines dry season from March to May 2017. Each pot was kept flooded using a drip irrigation system to simulate paddy conditions ( Supplementary Fig. S1).

Gas exchange measurements
Photosynthetic induction After anthesis, the flag leaf of the main stem was placed in the cuvette of an open gas exchange system (LI-6400XT, LI-COR, Lincoln, NE, USA). Light was provided by an integrated LED head (2×3 LED, LI-6400-02B). Within the cuvette, air temperature was 28 °C to approximate ideal growing conditions, flow rate was 400 µmol s -1 , and [CO 2 ] was maintained at 400 µmol mol -1 , and water vapor pressure deficit (VPD) at 1.3-1.7 kPa. Prior to measuring photosynthetic induction, rice plants were dark adapted for at least 1 h.
For induction, leaves were first allowed to reach a steady state in low light with a photosynthetically photon flux density (PPFD) of 50 µmol m -2 s -1 ('shade') for 300 s followed by 720 s at 1700 µmol m -2 s -1 ('sun'). Gas exchange measures were recorded every 10 s for the duration of the experiment. Measurements were repeated for all six accessions (n=8 plants per accession) over 4 d to minimize any age effects. Plants were selected by a randomized design and measured from 08.00 h to 12.00 h, to avoid confounding accessions with any diurnal influences. Net CO 2 uptake (A), stomatal conductance (g s ), intercellular CO 2 concentration (C i ), transpiration (E), and intrinsic water use efficiency (iWUE) were calculated following the equations of Farquhar et al. (1981). For a summary of all traits measured, see Table 1.
Two accessions, previously reported to show very different rates of induction for their leaves during the vegetative phase of growth, IR64-21 and AUS 278, were selected for further analysis of flag leaf induction at different [CO 2 ] (Acevedo-Siaca et al., 2020). Induction was measured following the protocol described above for induction, but at a cuvette [CO 2 ] of either 100, 200, 300, 400, 600, or 800 µmol mol -1 during induction. The order of cuvette [CO 2 ] treatments for each individual leaf was randomized to avoid confounding [CO 2 ] with time. Leaves were dark adapted for a minimum of 1 h between measurements at the different [CO 2 ] at which they would later be measured. To determine limitations through induction, A was plotted against C i for different time points, following the procedure of Soleh et al. (2016) and of Acevedo-Siaca et al. (2020). Five time points were selected for further analysis: 60, 180, 300, 360, and 700 s from the initiation of induction by transfer from darkness to high light (1700 µmol m -2 s -1 ). This allowed determination, at each time point, as to whether photosynthesis within the mesophyll was limited by the apparent maximum rates of RuBP regeneration (J max ) or RuBP carboxylation by Rubisco (V c,max ).

Steady-state measurements
For steady-state measurements, flag leaves were allowed to reach constant rates of A and stomatal conductance (g s ) at 1700 µmol m -2 s -1 PPFD. Cuvette conditions for steady-state measurements were as described above for photosynthetic induction.

Calculations
The rate constant of Rubisco activation, time to activation of photosynthesis, and forgone assimilation The rate constant of Rubisco activation (1/τ) was calculated by fitting the slope of the linear phase of the natural log of corrected photosynthetic induction [ln(A f *-A*)] as described in Woodrow and Mott (1989). A* is photosynthesis at a point in time during induction corrected to a C i value of 300 µmol mol -1 to remove limitation from stomata. A f * is the corrected value for photosynthesis at a C i of 300 µmol mol -1 at the end of the induction. The correction to C i was made following the methods of Soleh et al. (2016): The time required to complete the activation of photosynthesis (τ) was calculated by taking the reciprocal of the rate constant of Rubisco activation (Woodrow and Mott, 1989). The integrated amount of CO 2 uptake foregone due to the lower rates through induction compared with steady-state (C Loss t ) was calculated as in Acevedo-Siaca et al. (2020): where A is the steady-state rate of uptake and Ā t the average rate across the measured time period from the start of the induction (t), 300 s and 700 s, respectively.

Statistical analyses
All statistical analyses and model fitting used R (version 3.5.2, R-project) (R Core Team, 2020). Normal distribution and homogeneity of variances were tested by the Shapiro-Wilk test and Brown-Forsythe test, respectively. Assumptions were met and ANOVA was performed followed by Tukey's mean discrimination analysis, using the R-Project: 'agricolae: Statistical Procedures for Agricultural Research' package. Pearson correlation coefficients between different photosynthetic measures were calculated using accession mean values (R; 'corrplot' and 'Hmisc').

Characterizing photosynthesis in flag leaves
Significant variation was found across accessions for photosynthetic traits in both non-steady-state and steady-state conditions (Figs 1, 2). Significant differences were seen in steady-state parameters such as CO 2 uptake in saturating light (A sat , P<0.0001), stomatal conductance (g s , P=0.0003), intercellular CO 2 concentration (C i , P=0.03), and intrinsic water use efficiency (iWUE, P=0.006) (Fig. 2) (Fig. 3). Significant differences were also found among accessions for C Loss 300 and C Loss 700 (Fig. 4).

Understanding biochemical limitations to photosynthetic induction in flag leaves
Photosynthetic induction in flag leaves was limited primarily by biochemistry, rather than stomata, during the first 300 s of induction (Fig. 5). The rate of photosynthesis corrected to remove stomatal limitation (A*) did not differ significantly from the uncorrected A values, suggesting that biochemical limitation dominated for these accessions (Fig. 5). This was also evident by the fact that C i was higher than at steady state through most of the induction (Fig. 1C). A* showed only small differences from the corresponding A (~3-4%), except for AUS 278 which had an average difference of ~13% between A and A* (Fig. 5). Photosynthetic induction was measured at six different [CO 2 ] values in two selected accessions, AUS 278 and IR64-21. As expected, as [CO 2 ] increased, A increased in both accessions ( Supplementary Fig. S3). Also as expected, stomatal conductance decreased with increased [CO 2 ] ( Supplementary  Fig. S3). As time after the beginning of induction increased, the amount of CO 2 assimilated increased as well (Fig. 6). This response to increasing CO 2 assimilation over time was seen in both selected accessions (Fig. 6).
Utilizing these data, the operating point at an ambient [CO 2 ] (C a ) of 400 µmol mol -1 was calculated for each curve.
The operating point fell on the initial slope of the A/C i curve, indicating limitation by Rubisco for both accessions at all time points (Fig. 6) Fig. S7).
Finally, the photosynthetic response curves to PPFD, measured at low [CO 2 ] (≤300 µmol mol -1 ), were used to infer the response of V c,max to a PPFD in vivo. These results indicated a significant difference between AUS 278 and IR64-21, with AUS 278 being more strongly limited by the rate of carboxylation ( Supplementary Fig. S8). This contrasts with induction at ambient [CO 2 ], where AUS 278 and IR64-21 did not vary significantly in photosynthetic induction traits Ā 300 , g s avg , C i avg , iWUE avg , IT 50 , and IT 90 (Fig. 3).

Comparing steady-and non-steady-state photosynthetic performance
The photosynthetic performance of accessions in non-steadystate and steady-state conditions was compared using Pearson's correlation coefficient. Significant (P<0.05) correlations were found between the different photosynthetic traits measured in steady-state and between the different traits measured in non-steady-state conditions (Fig. 7). However, there were no significant correlations between traits measured at steady state and their corresponding traits measured at non-steady state (Fig. 7). Additionally, as was previously found in vegetativephase leaves, there was a significant correlation (P<0.05) between Ā 300 and time at which A reached 50% induction (IT 50 ) (Fig. 7).

Comparing photosynthetic induction performance between flag leaves and vegetative-phase leaves
The photosynthetic traits measured here for flag leaves were compared with those made for leaves in the vegetative growth stage of the same accessions in a previous study (Acevedo-Siaca et al., 2020). Significant correlations were found between steady-state C i in leaves in the vegetative growth stage and iWUE in flag leaves, C i in leaves in the vegetative growth stage and g s in flag leaves, and iWUE in leaves in the vegetative growth phase and flag leaves (Fig. 8A). In non-steady-state conditions, Fig. 3. Mean and variation for flag leaf non-steady-state photosynthetic performance in six rice accessions. (A) CO 2 assimilation during the first 300 s of induction (Ā 300 ), (B) average stomatal conductance during the first 300 s of induction (g s avg ), (C) average intrinsic water use efficiency (iWUE avg =Ā 300 /g s avg ), (D) average intercellular CO 2 concentration (C i avg ), (E and F) time at which A reached 50% and 90% of A 300 (IT 50 and IT 90 , respectively). Accessions are ordered by median performance. Letters are indicative of a significant difference between accessions. significant correlations were found between C i avg in leaves in the vegetative growth stage and both C i avg and iWUE avg in flag leaves (Fig. 8). However, there was no significant correlation between any measure of CO 2 assimilation between flag leaves and leaves during the vegetative growth phase.

Discussion
Currently, no study has aimed to characterize the photosynthetic induction response in rice flag leaves. With respect to the objectives as outlined in the Introduction, the following key answers were obtained. (i) There were substantial differences among the six accessions of ~150% between leaf CO 2 uptake over the course of induction, with smaller differences in the light-saturated steady state (Figs 2, 3). This suggests significant scope for exploiting germplasm to increase rice photosynthesis in this key leaf for grain filling. (ii) Analysis of the responses of CO 2 uptake to intercellular CO 2 concentration showed that the in vivo activity of Rubisco (V c,max ) limited photosynthesis throughout induction (Fig. 6), suggesting that breeding or bioengineering increased Rubisco content and activity as a means to increase rice productivity. (iii) When the results obtained here for flag leaves were compared with those for the same accessions in the earlier study of leaves in the vegetative growth stage (Acevedo-Siaca et al., 2020), there was little correspondence between photosynthetic parameters for the two growth stages, with the exception of iWUE (Fig. 8). This suggests that breeding efforts to improve crop photosynthesis would need to address selection at both growth stages, although improved WUE could be selected at the vegetative stage alone, making this more tractable. The following discussion places these new findings in the context of prior studies and potential for increasing rice photosynthetic capacity, efficiency of resource use, and productivity.

Significant variation for photosynthetic traits in rice flag leaves
Previously, it was suggested that natural variation for photosynthesis was more prevalent in the flag leaves than for leaves in the vegetative growth stage of cereal crops, a finding confirmed here for rice (Dunstone et al., 1973;Bansal et al., 1993). Previous studies have focused on either steady-state or more recently non-steady-state photosynthesis, and demonstrated diversity within crop germplams (Flood et al., 2011;Gu et al., 2012;Driever et al., 2014;Soleh et al., 2016Soleh et al., , 2017Acevedo-Siaca et al., 2020;De Souza et al., 2020;Taniyoshi et al., 2020). Here, we found that significant natural variation exists for photosynthetic traits in both steadyand non-steady-state conditions (A 6.4-25.9 µmol m -2 s -1 ; Ā 300 2.3-16.7 µmol m -2 s -1 ) in the flag leaf. Indeed, 43, 12, and 68% more variation was found here between flag leaves when compared with leaves in the vegetative growth stage of the same accessions (Acevedo-Siaca et al., 2020) for Ā 300 , iWUE avg , and IT 50 , respectively. This is significant given the key role of the flag leaf.

Further evidence for the lack of correlation between steady-and non-steady-state measurements
Consistent with studies of soybean, cowpea, and leaves in the vegetative growth stage of rice, no significant correlations were found between parameters of photosynthetic CO 2 uptake in steady-and non-steady-state light in rice flag leaves (Soleh et al., 2016;Acevedo-Siaca et al., 2020;De Souza et al., 2020). These studies, together with the present study, provide compelling evidence for a reconsideration of when during crop Fig. 4. The integra1 of CO 2 uptake forgone due to the lower than steadystate rates through the first (A) 300 s and (B) 700 s of induction compared with steady state (C Loss 300 and C Loss 700 , respectively). development we measure and how we measure photosynthesis when considering crop improvement. In particular, clear evidence that steady-state measurements do not indicate photosynthetic efficiency under the non-steady-state fluctuating light conditions that can be dominant in the field should be noted. However, the greater between-accession variability of non-steady-state photosynthesis highlights a greater opportunity for increasing net crop CO 2 uptake that could meet the apparent strong source limitation of modern high-yielding cultivars (Hasegawa et al., 2013;Zhu et al., 2015;Sakai et al., 2019).
It is possible that a balance for the distribution of resources between photosynthetic proteins underlies the differences between steady-state and non-steady-state photosynthesis. Previous studies have suggested a trade-off between maximum rates of photosynthesis in steady-state saturating light versus the speed of induction due to limited amounts of nitrogen invested and divided between Rubisco and Rubisco activase content (Woodrow and Mott, 1989;Mott and Woodrow, 2000). It is hypothesized that plants grown in fluctuating light environments that do not experience steady-state conditions frequently would benefit from having higher Rubisco activase content to be able to respond more quickly to changes in light (Mott and Woodrow, 2000;Yamori et al., 2012;Carmo-Silva and Salvucci, 2013;Kaiser et al., 2016). However, leaves that experience fewer sunflecks and have more exposure to direct sunlight would benefit from investing a higher proportion of resources in Rubisco to sustain higher photosynthetic rates at steady state (Mott and Woodrow, 2000). Such a trade-off is supported by the finding that antisense down-regulation of Rubisco activase increased Rubisco content in rice leaves (Jin et al., 2006). This trade-off between Rubisco and Rubisco activase content could help to partially explain the lack of correlation between steady-and non-steady-state photosynthesis, as an increase in the protein that helps the leaf excel in induction acts to the detriment of the protein needed at steady state. The trade-off between steady-and non-steady-state photosynthesis was shown clearly in IR64-21, which had the highest steady-state A sat yet was among the slowest to reach 90% of its steady-state level (IT 90 A ) during induction (Figs 2, 3).

Photosynthetic induction is primarily limited by biochemistry in flag leaves
At ambient [CO 2 ], rice flag leaves are predominantly limited by biochemistry, specifically the maximum activity of Rubisco (Figs 5, 6). Differences between A and A* at ambient [CO 2 ] were generally small, indicating little limitation by stomata (Fig. 5). This is likely to be due to the shape and small size of rice stomata allowing fast responses (McAusland et al., 2016), but also the evolutionary history of rice, which was domesticated from emergent aquatic progenitors and then bred in paddy conditions where water would not be limiting to the plant (Nay-Htoon et al., 2018).
Calculation of the operating point at ambient [CO 2 ] (C a =400 µmol mol -1 ) in AUS 278 and IR64-21 showed that photosynthesis during induction was predominantly limited by Rubisco, and not affected by either the capacity for regeneration of RuBP or triose phosphate utilization, throughout the induction and into steady state. The operating point is on the initial slope of the A/C i curve, throughout (Fig. 6). This parallels the previous findings for rice leaves in the vegetative growth stage (Acevedo-Siaca et al., 2020) and suggests that the predominant limitation to photosynthetic CO 2 assimilation at steady state and non-steady state throughout the life cycle of rice is consistently due to the in vivo capacity and, presumably, amount of Rubisco. This is also consistent with the recent observation that transgenic up-regulation of Rubisco in rice significantly increases paddy yield (Long, 2020;Yoon et al., 2020). One caveat is that this analysis is based on C i and not the CO 2 concentration at Rubisco (C c ). Ease of movement of CO 2 from the intercellular space to Rubisco is governed by mesophyll conductance (g m ), which was not measured here. So, it is possible that activation of g m , as well as Rubisco, could be a limiting factor. However, prior work with other species has suggested that activation of g m is likely to be faster than activation of Rubisco (Deans et al., 2019).

Photosynthetic performance shows little correlation between flag and vegetative-phase leaves
The only significant correlations found between flag and leaves in the vegetative growth stage were for C i and iWUE (Fig. 8). These results suggest that iWUE is not be affected by rice developmental stage and could be consistent throughout the lifetime of these rice plants. Water availability is the biggest limitation to agricultural production worldwide and is expected to pose an even greater limitation with climate change and population growth (Wassmann et al., 2009;Ort and Long, 2014;Oladosu et al., 2019). If iWUE is consistent throughout the life cycle of rice, it could allow breeders to screen for high iWUE early in development, saving time and resources in selecting germplasm in breeding more water use-efficient plants.
Otherwise, no significant correlations were found between the photosynthetic performance of flag leaves and leaves in the vegetative growth stage in both steady-and non-steadystate conditions (Fig. 7). For example, elite cultivar IR64-21 was outperformed by other accessions in measurements of photosynthesis in leaves in the vegetative growth stage during steady-and non-steady-state conditions (Acevedo-Siaca et al., 2020). However, IR64-21 flag leaves had the highest photosynthetic rates during photosynthetic induction and at steady state, significantly outperforming the other accessions (Figs 1, 2; Supplementary Fig. S3). This suggests that increased flag leaf CO 2 uptake may have been inadvertently improved through conventional breeding selection. This corresponds with other evidence that flag leaf photosynthesis has been improved unintentionally through breeding for higher yield potential. Newer and higher yielding rice varieties have flag leaves with higher photosynthetic rates per unit area than older varieties (Ishii, 1993). Additionally, flag leaves in Oryza sativa were found to maintain higher photosynthetic rates for longer relative to wild Oryza species (Ishii, 1993), which is curiously in contrast to wheat (Dunstone et al. 1973). These studies suggest that deliberate breeding for increased flag leaf photosynthesis might be a fertile avenue for a further increase in rice yield potential.
However, while an increased emphasis on flag leaf photosynthesis can lead to higher yields (Fabre et al., 2016), there should still be a focus on improving photosynthetic efficiency throughout the life cycle. Photosynthesis in leaves during the vegetative stage is important in establishing the plant and developing a robust root system and tillers capable of becoming reproductive. Improved photosynthesis in leaves in the vegetative growth stage results in increases in non-storage  Traits at steadystate are: intrinsic water-use efficiency (iWUE=A/g s ), transpiration (E), intercellular CO 2 concentration (C i ), stomatal conductance (g s ), and net CO 2 assimilation in saturating light (A sat ). Traits at non-steady state over the first 300 s of induction are: the time at which A reached 50% and 90% of A 300 (IT 50 and IT 90 , respectively), average C i during first 300 s of induction (C i avg ), average intrinsic water use efficiency (iWUE avg =Ā 300 /g s avg ), average g s , the maximum A during induction (A Max ), A at the end of this period (A 300 ), and the average A (Ā 300 ). A significant R value is marked by a black line on the scale (0.8).
carbohydrates in leaves and stems, which can subsequently be remobilized, for grain filling. The expected ideotype would therefore be an accession which shows high capacity and efficiency during both growth phases.

Non-steady-state photosynthesis-a practical target for breeding?
Non-steady-state photosynthesis has received little attention in breeding, largely because the gas exchange methods required to effectively phenotype traits are not practical for large-scale testing. However, it was recently shown in wheat that large-scale screening of non-steady-state photosynthesis could be achieved effectively with excised leaves using modulated chlorophyll fluorescence imaging (McAusland et al., 2019). This high-throughput technique could make selection and breeding for improved efficiency under nonsteady-state conditions practical, and probably more effective than selecting for improvement under steady-state conditions. Additionally, these methods could allow for the improvement of iWUE under non-steady-state conditions. As noted above, current methods of breeding paddy rice for yield may have lowered WUE. This study has revealed considerable variation-even within the limited germplasm examined-and a correlation in WUE between vegetative and reproductive growth. This finding suggests that selection of improved WUE could be achieved by screening during early growth. Furthermore, the development of integrated thermal and modulated fluorescence imaging of instantaneous WUE would now allow high-throughput phenotyping of this trait (McAusland et al., 2013). Accelerated breeding of rice lines requiring less water would help address the rising pressures on water supplies in many paddy rice-growing regions (Schyns et al., 2019). These benefits that go beyond increased yield are imperative for creating more sustainable agricultural systems that utilize this planet's limited resources with more discretion, especially in the face of global climate change and a growing human population.

Supplementary data
The following supplementary data are available at JXB online. Fig. S1. Growing conditions at the IRRI used in the study. Fig. S2. Corrected photosynthesis for time to activation of photosynthesis (τ). Fig. S3. Induction of leaf CO 2 uptake (A) and stomatal conductance (g s ) at six [CO 2 ] in IR64-21 and AUS 278. Fig. S4. Leaf CO 2 uptake (A) and leaf CO 2 uptake corrected for stomatal limitation (A*) in AUS 278. Fig. S5. Leaf CO 2 uptake (A) and leaf CO 2 uptake corrected for stomatal limitation (A*) in IR64-21. Fig. S6. Corrected photosynthesis for time to activation of photosynthesis (τ) in AUS 278 and IR64-21.  Table S1. Description of accessions used in this study.