Natural genetic variation in photosynthesis: an untapped resource to increase crop yield potential?

Summary Raising crop yield potential is a major goal to ensure food security for the growing global population. Photosynthesis is the primary determinant of crop productivity and any gain in photosynthetic CO2 assimilation per unit of leaf area (A) has the potential to increase yield. Significant intraspecific variation in A is known to exist in various autotrophic organs that represent an unexploited target for crop improvement. However, the large number of factors that influence photosynthetic rates often makes it difficult to measure or estimate A under dynamic field conditions (i.e. fluctuating light intensities or temperatures). This complexity often results in photosynthetic capacity, rather than realized photosynthetic rates being used to assess natural variation in photosynthesis. Here we review the work on natural variation in A, the different factors determining A and their interaction in yield formation. A series of drawbacks and perspectives are presented for the most common analyses generally used to estimate A. The different yield components and their determination based on different photosynthetic organs are discussed with a major focus on potential exploitation of various traits for crop improvement. To conclude, an example of different possibilities to increase yield in wheat through enhancing A is illustrated.


INTRODUCTION
Photosynthesis is the primary determinant of crop productivity and any gain in photosynthetic efficiency has the potential to result in increases in yield (Flood et al., 2011;Lawson et al., 2012;Gu et al., 2014). Free air CO 2 enrichment (FACE) studies have provided substantial evidence that increased photosynthetic rates have translated into greater crop yields, demonstrating the link between photosynthesis and yield (Ainsworth and Long, 2005). The yield potential of a crop can be described by the following equation: where Q is total solar radiation, e i is the interception efficiency, e c is the efficiency for conversion into biomass and e p is the efficiency of partitioning biomass into harvested organs (Long et al., 2015). In the absence of environmental stress, parameters such as harvest index are already close to the theoretical maximum (Foulkes et al., 2010;Reynolds et al., 2012). Additionally, many canopy traits such as canopy architecture (Long et al., 2006), light interception (Murchie et al., 2009), and photosynthetic duration (Shearman et al., 2005) have been optimized. However, empirical analyses of the yield potential components demonstrate significant inefficiency in e c in C 3 crops (Zhu et al., 2008, in which photosynthetic CO 2 assimilation per unit leaf area (A) is the primary process (Kebeish et al., 2007;Maurino and Peterh€ ansel, 2010;Walker et al., 2016). The maximum potential conversion energy efficiency for C 3 plants is 4.6% . However, plants including crops attain much less than this and therefore A is far from optimal and a target for further improvements (Parry et al., 2010;Long et al., 2015). Several studies have explored opportunities to increase energy conversion through genetic manipulation, including manipulating Calvin cycle activity (Simkin et al., 2015;Lopez-Calcagno et al., 2018), RuBisCO properties (Parry et al., 2003) and increasing the kinetics of non-photochemical quenching for photo-protection (Kromdijk et al., 2016), all of which have translated into increased A and greater plant biomass. However, restrictions on growing genetically modified crops in many countries especially in Europe means that alternative methods to achieve increases in photosynthesis must be realized. An undervalued and currently unexploited opportunity to increase yield, not mutually exclusive of genetic engineering approaches, is the extensive natural variation in photosynthetic capacity in different C 3 crops (Rawson et al., 1983;Blum, 1990;Watanabe et al., 1994;Fischer et al., 1998;Herv e et al., 2001;Pettigrew, 2004;Flood et al., 2011;Gu et al., 2012;Lawson et al., 2012;Driever et al., 2014;Gaju et al., 2016;Carmo-Silva et al., 2017;Qu et al., 2017;Pater et al., 2017;Faralli et al., 2019b). A number of studies have explored natural variation in photosynthesis in commercial wheat varieties (often relative to the year of release) (Fischer et al., 1981(Fischer et al., , 1998Blum, 1990;Watanabe et al., 1994;Reynolds et al., 2000;Xue et al., 2002;Chytyk et al., 2011;Sadras et al., 2012), and demonstrated a correlation between photosynthesis and yield (e.g. Blum, 1990;Fischer et al., 1998), although, such a relationship often depended on growth conditions (Xue et al., 2002), or when measurements were taken during the growing season (Reynolds et al., 2000), while others reported no relationship (e.g. Driever et al., 2014). Variation in photosynthesis has been attributed to differences in radiation use efficiency (Sadras et al., 2012), biochemical differences in RuBisCO activation properties (Carmo-Silva and Salvucci, 2013), carboxylation efficiency (Driever et al., 2014) and electron transport capacity (Carmo-Silva et al., 2017). In addition, variations in traits limiting the diffusion of CO 2 to the site of carboxylation including mesophyll conductance (g m ) (Jahan et al., 2014) and stomatal conductance (g s ) (Fischer et al., 1998), which also includes the rapidity of g s responses to changing environmental conditions (Lawson et al., 2010Faralli et al., 2019b) have been reported in several crops.
Here we review natural variation in physiological traits with a focus on: (i) photosynthetic capacity, which is determined by plant acclimatory responses and constrained by genetics; and (ii) dynamic short-term modifications to A (e.g. biochemical factors such as the regulation of enzymes, g s and g m ). The most common methods used to estimate A are evaluated and discussed. To conclude, we will focus on natural variation in A, stomatal physiology and the associated photosynthetic limitation in wheat (i.e. source limitation; lack of photo-assimilates, or sink limitation) with a particular emphasis on the potential exploitation for crop improvement.

NATURAL VARIATION IN PHOTOSYNTHESIS
The biochemical processes of photosynthesis in C 3 crops are considered essentially identical, (although recent metabolite profiling of C 3 species by Arrivault et al. (2019) has reported considerable variation in levels of metabolites), however, significant intraspecific and interspecific variation in photosynthetic rates exists, providing a valuable source of unexploited genetic diversity (Flood et al., 2011) (Table 1a). Furthermore, the physiological or genetic mechanisms underlying these differences in both photosynthetic potential as well as dynamic behaviour may provide valuable information on the performance of different cultivars under specific environments (Driever et al., 2014).
Intraspecific variation in photosynthetic traits in wheat and the potential effect of selection on photosynthesis was shown initially by Rawson et al. (1983) and then by Blum (1990) where breeding in Mediterranean environments had led to an increase in photosynthetic efficiency at saturating light for the modern cultivars compared with older varieties. Furthermore, Watanabe et al. (1994) followed by Fischer et al. (1998) phenotyped historical Australian and Mexican wheat cultivars for photosynthetic traits and reported a strong correlation between increased rates of CO 2 assimilation and yield genetic gain with year of release, demonstrating that breeding has unintentionally selected for higher A. Subsequent research focusing on intraspecific variation in major crops such as cotton (Pettigrew, 2004), canola (Pater et al., 2017), rice (Gu et al., 2012), sunflower (Herv e et al., 2001) and wheat (Reynolds et al., 2000;Sadras et al., 2012;Driever et al., 2014;Carmo-Silva et al., 2017), highlighted a wide diversity of photosynthetic traits [including A sat and the light and CO 2 saturated rate of photosynthesis A max ; the maximum carboxylation capacity V cmax as well as the maximum rate of electron transport (J max )]. In addition, at the leaf level, CO 2 uptake from the atmosphere to the site of carboxylation is subject to two main restrictions, stomatal and mesophyll, both of which therefore determine the rate photosynthesis. g s (the reciprocal of stomatal resistance) controls CO 2 diffusion from the atmosphere into the intercellular air spaces in the gaseous phase (Farquhar and Sharkey, 1982;Sharkey, 1985). Subsequently, g m adds an additional limitation in the liquid phase for the diffusion of CO 2 from the intercellular airspaces to the site of carboxylation in chloroplasts (Flexas et al., 2008). Intraspecific variation exists for both g s and g m (Table 1b) in a series of food crops including wheat (Fischer et al., 1998;Jahan et al., 2014), barley (Gonz alez et al., 1999;Barbour et al., 2010) and rice (Ouyang et al., 2017). Therefore exploiting the existing natural variation in photosynthesis as well as optimizing the components determining A in elite cultivars (Driever et al., 2014), landraces (Gaju et al., 2016) and wild relatives (Prins et al., 2016) could provide novel targets for crop improvement.
However, while Crosbie et al. (1981) showed that leaf photosynthesis of maize can be improved by recurrent selection (i.e. increasing the frequency of favourable alleles for quantitatively inherited traits, in this case for A) five cycles of recurrent phenotypic selection did not produce the expected results in term of productivity, and changes in grain yield were not significant for any of the populations tested (Crosbie and Pearce, 1982). Indeed, correlating photosynthesis with yield is not straight forward, with inconsistencies in the relationship described in the literature, for example positive correlation (Carmo-Silva et al., 2017), no correlation (Ojima, 1974;Driever et al., 2014), or a correlation but only when photosynthesis was measured at particular phenological stages (Gaju et al., 2016). These inconsistences in the relationship between A and yield emphasize the complexity of yield formation in crops that is based on a series of interrelated subcomponents (Miralles and Slafer, 2007), and that is further complicated by the different methodologies used to estimate A and the influence of fluctuating environmental conditions to which the crop is subjected . Individual point measurements of A sat or A max taken either at different times during the crop cycle or on individual leaves within the canopy, often do not correlate with yield (Rawson et al., 1983;Driever et al., 2014). Having said this, in some  Blum et al. (1990) and Fischer et al. (1998) measurements were taken at 345 ll L À1 [CO 2 ], in Herv e et al. (2001) and Watanabe et al. (1994) measurements were taken at a [CO 2 ] of 350 lmol mol À1 and in Gu et al. (2012) measurements were taken at a [CO 2 ] of 380 lmol mol À1 . (b) Variation in stomatal conductance (g s ) and mesophyll conductance (g m ) in different studies.

(a) Reference Crop
Range of A sat (lmol m À2 sec À1 ) Intraspecific variation (lmol m À2 sec À1 ) Relation with yield Rawson et al. (1983) Wheat 20-45 (mg dm À2 h À1 ) 1 2 m g d m À2 h À1 A sat and yield not correlated. Cumulative CO 2 fixation by flag leaf and yield P < 0.001 r 2 = 0.30 Blum (1990) Wheat 9 cases (i.e. Fischer et al., 1998) a significant relationship between A sat and some yield components (i.e. grain number) or the average grain yield (over 5 years) was evident.
In addition, when operational A was measured (i.e. single measurements of A at light intensities similar to those experienced by the crop) in the field at the pre-anthesis and post-anthesis, a strong correlation with yield was reported (Carmo-Silva et al., 2017). Although an instantaneous 'snapshot' analysis of A sat , carried out by Rawson et al. (1983) did not correlate with yield, a significant (P < 0.001) correlation between cumulative carbon assimilation of the flag leaf (measured as several snapshot A sat measurements over the life cycle) and yield was observed. These studies highlight that the different methods used to measure A, the complexity of the relationship between A, plant growth and yield as well as the influence of the environment on these processes, need to be considered for estimating overall crop photosynthesis.

Factors determining the variation in photosynthetic rate per unit leaf area
Biochemical factors and anatomical features. One of the first studies to examine the underlying biochemical function of interspecific variation in photosynthesis was Wullschleger (1993). Using response curves of A as a function of substomatal CO 2 concentration (A/C i ), Wullschleger demonstrated that most of the observed variation in A in the 109 species analyzed was attributed to variation in the underlying biochemistry and photosynthetic capacity with differences in both carboxylation capacity (V cmax ) and electron transport capacity for RuBP regeneration (J max ). Wullschleger (1993) also reported a positive correlation between V cmax and J max suggesting co-ordinated regulation by these two processes. A small number of species (23) was reported to be limited by the utilization of triose phosphates, which ranged from 4.9 to 20.1 lmol m À2 sec À1 , and reflects the short-term interaction between A and starch-sucrose production, which ultimately reflects growth. It is clear from the representative A/C i curves in Wullschleger (1993), that the switch-over point between limitation by carboxylation capacity and capacity for electron transport differed greatly in the four species illustrated, and that the maximum rates of A achieved were vastly different, which may be due to nitrogen allocation between RuBisCO and light harvesting. Nitrogen (N) concentration is a key determinate of A, as the majority of leaf N is invested in the photosynthetic apparatus, in particular RuBisCO (Hikosaka, 2010). Differences in N-use efficiency and N concentration in different crops have suggested these as targets for both increased A and optimization of fertilization input (Guarda et al., 2004;Hirel et al., 2007). Although, there is evidence that within species variation in A, can be explained by differences in V cmax and J max (Driever et al, 2014;Carmo-Silva et al., 2017), Driever et al. (2014) highlighted that the variation in carboxylation capacity was not due to RuBisCO content (or N allocation) but possibly RuBisCO activation, demonstrating further complexity in identifying specific targets for future wheat improvement. Furthermore, the same study also reported that some of the highest V cmax values were found in older species, suggesting that photosynthetic capacity potential has not been fully exploited in past breeding programmes. However, since a major goal of future agriculture is to enhance resource-use efficiency, it has been hypothesized that increasing RuBisCO carboxylation efficiency while reducing N allocation to RuBisCO might be a successful alternative in crops to improve or sustain A (Reynolds et al., 2012). A reduction in RuBisCO content (up to 20%) led to a 10% lower N requirement in wheat, although reductions in A at high light intensities were also present (Reynolds et al., 2012). More recently, Carmo-Silva et al. (2017) found significant genotypic variation for RuBisCO carboxylation efficiency and RuBisCO content in wheat, with the cultivar Gatsby combining a high A and a low RuBisCO content, suggesting the potential of this preferable combination for further exploitation. Genetic engineering approaches have shown that increasing protein abundance (e.g. sedoheptulose1,7-biphosphatase, SBPase) led to a significant increase in A which suggests that although photosynthesis requires a large number of protein-protein interactions, part of the genetic variation can be explained by differences in key protein abundance and activity, that result in improved photosynthetic capacity (Flood et al., 2011;Simkin et al., 2019) and which also might explain variation in metabolite profiles in C 3 species (Arrivault et al., 2019). The potential for exploiting natural variation in photosynthetic capacity has been demonstrated by Gu et al. (2014) who used a simulation analyses to assess the contribution that the natural variation in RuBisCO and electron transport rate could make to photosynthesis in rice and showed that exploiting this could increase rice yield by 22-29%, depending on location and year. Many studies have focused on significant variation in photosynthetic capacity that is determined by acclimation to particular environmental conditions and genetically constrained. However, on a day-to-day basis, plants respond dynamically to changes in the surrounding environmental conditions that introduce a further layer of complexity to variation in photosynthesis as there is significant variation in dynamic responses. These dynamic processes include regulation and expression levels of enzymes Hikosaka, 2010), dynamic regulation in response to environmental change  including changes in non-photochemical quenching of excess energy dissipating mechanisms (K€ ulheim et al., 2002;Lawson et al., 2012), and the rapidity of stomatal responses (Lawson et al., 2010 (Flood et al., 2011;Gilbert et al., 2011).
Other processes, although not directly related to the photosynthetic machinery, also play a role in photosynthetic performance. For instance, sucrose transport from the mesophyll cells to heterotrophic tissues is of pivotal importance to sustain diurnal A, as it is generally accepted that A slowly decreases over the diurnal period due to the accumulation of photosynthates (Ainsworth and Bush, 2011). Recently, Ainsworth and Lemonnier (2018) reported the existence of genetic variation in different phloem loading mechanisms. Apoplastic loading-unloading strategies are typically common in crop species and optimization cannot only help in sustaining A but also enhance sink strength, therefore these are potential targets to further maximize the diurnal integrated A (Ainsworth and Lemonnier, 2018). Furthermore, morphological factors substantially influenced A with long-lived evergreen plants showing thicker leaves, with a higher leaf mass per unit leaf area, lower g m and therefore lower A than herbaceous plants (e.g. grasses) (Flood et al., 2011). Therefore, differences in leaf functionality between species are the result of differences in leaf longevity and subsequent optimization of resource investment into photosynthetic organs.
Mesophyll and stomatal limitations of photosynthesis. Mesophyll conductance is considered a key trait for future improvement in A and yield potential, as lower resistance for CO 2 diffusion to the chloroplast will allow higher substrate availability for carboxylation. Additionally, an attractive property of increasing g m is the potential to increase A without increasing water loss (Nadal and Flexas, 2018), which is not possible if g s is increased. g m can be dissected into three subcomponents: conductance through intercellular air spaces (g ias ), through cell wall (g w ) and through the liquid phase inside cells (g liq ) (Flexas et al., 2008;Terashima et al., 2011). Variation in g m between species has been associated with alterations in all these components: for instance leaf structure may affect mostly g ias and g w , in particular in thick leaves (Evans and von Caemmerer, 1996). In tobacco and soybean however, the most limiting component to g m appeared to be g liq (Evans and von Caemmerer, 1996). The intraspecific variation in g m in crop species (Table 1) suggests that both morphological and metabolic factors are involved in CO 2 diffusion into chloroplasts, with evidence of aquaporin modulation of the g liq component (Gillon and Yakir, 2000;Hanba et al., 2004;Flexas et al., 2006). For instance, overexpression of the aquaporin OsPIP1;2 in rice increased g m by up to 150% compared with the wild type, resulting in greater biomass and yield (Xu et al., 2018) and therefore provided evidence for a major role for aquaporins in the modulation of intracellular CO 2 diffusion (Uehlein et al., 2003;Uehlein et al., 2008). Such studies often introduce the question 'why have such changes not occurred naturally', however it should be borne in mind that survival to reproduce is the plant's ultimate goal, while photosynthesis and biomass may or may not be a part of this process, and therefore resource allocation and adaptive capacity will regulate such changes. g m is generally affected by both light and temperature, therefore g m can have a signficant impact on photosynthetic efficiency under fluctuating environments (e.g. Flexas et al., 2008;Kaiser et al., 2018). However, methodologies to quantify g m are time consuming and subject to high levels of uncertainty (see review by Pons et al., 2009 and references therein), severely limiting high-throughput phenotyping for this trait. In addition, g m is principally dependant on the physical capacity of CO 2 to diffuse into the leaf tissue, and therefore dependent on g s and stomatal dynamics.
Increasing CO 2 diffusion from the atmosphere to the leaf interior increases A (Lawson et al., 2010) and it has been demonstrated in several studies that manipulating stomatal density (Tanaka et al., 2013) or aperture (Lawson and Blatt, 2014;Duan et al., 2015) increases g s, while recent studies have also suggested that stomatal kinetics and the rapidity of g s responses to the changing environment can increase carbon assimilation (McAusland et al., 2016;Papanatsiou et al., 2019). Increasing g s represents a trait already unintentionally included in breeding for high yielding varieties over many decades (Fischer et al., 1998;Lu et al., 1998;De Vita et al., 2007). The positive effects of higher g s are numerous: in particular, under steady-state conditions A is co-related to g s and therefore high g s leads to elevated photosynthetic rates (by limiting the resistance to CO 2 diffusion into intracellular airspaces) and, at the same time, increased evaporative cooling maintains optimal leaf temperature for A (Lawson and Blatt, 2014). As in C 3 crops, a strong limitation of A is the temperature-dependent increase in the oxygenation reaction of RuBisCO, the maintenance of optimal leaf temperature through high transpiration rates may be key in limiting photorespiration (Long et al., 2006). In addition, although high g s may lead to early soil water depletion, it has been shown that the extra assimilates gained early in the growing season may enable greater carbon investment in roots (Blum, 2011), facilitating higher water extraction from the deeper soil layers therefore avoiding drought stress (Venuprasad et al., 2011). It is therefore unarguable that g s is a key trait for improving crop yield potential and stability with substantial natural variation known to exist (Faralli et al., 2019a). Stomatal conductance is determined by the number of stomata per unit leaf area and the pore aperture (which is often dependent on the size of stomata) both of which represent breeding targets for altered g s . For example, Arabidopsis lines lacking the epidermal patterning factor (EPF) 1 and 2, exhibited high stomatal density, greater g s and A when compared with the wild type Col-0 (Franks et al., 2015). Large natural variation in g s has been shown in a number of plants, including crops (i.e. Tich a, 1982;Roche, 2015;Faralli et al., 2019a and Table 1), suggesting g s as a potential target to exploit for increased A, and therefore yield. Stomata open and close in response to changes in environmental cues (i.e. water availability, light, VPD) and depend upon plant hydraulic capacity, which is the plant's ability to take up and distribute water around the plant (Sack and Scoffoni, 2013;Lawson and Blatt, 2014). In the field and inside a crop canopy, light and VPD can vary within minutes or even seconds. Stomatal responses are an order of magnitude slower than the response of A. For example, g s in wheat can take between 5 and 15 min to reach steady state following a shade or sun fleck (Faralli et al., 2019b) and this lag in behaviour can limit A by up to 15%. Both intraspecific and interspecific variation have been shown to exist for stomatal rapidity (McAusland et al., 2016;Faralli et al., 2019b). In addition significant developmental effects on stomatal responses were shown by Faralli et al. (2019b), in which a decrease in stomatal rapidity was reported in wheat during the post-anthesis stage compared with the early booting stage. Therefore, g s and the dynamic response of g s can be potential unexploited targets for future crop improvement.
Measuring photosynthesis. To date, most photosynthetic measurements have been based on two approaches using infrared gas analyzer systems: (i) capacity measurements where photosynthetic CO 2 assimilation is measured as a function of substomatal CO 2 concentration curves (A/C i ) or as a function of light intensity (A/Q); or (ii) 'snapshot' or instantaneous measurements of A at selected times of the day. Additionally, other methods such as carbon isotope discrimination has been successfully used to estimate transpiration efficiency (Rebetzke et al., 2002) and the photosynthetic contribution of different non-foliar organs to grain yield (Sanchez-Bragado et al., 2016). In general, A/C i analysis is a powerful tool from which the biochemical properties under light saturated conditions, a constant leaf temperature and minimal boundary layer resistance can be determined. These conditions, necessary to assess maximum photosynthetic capacity are unlikely to represent those to which a leaf is exposed in the field Driever et al., 2014) (Figure 1). Assessing photosynthesis as a function of light (A/Q analysis) might be considered more representative of field conditions. These measurements can be used to model A rates over the diurnal period if incident light is monitored. However, it should be noted that A/Q curves are usually performed in near optimal environmental conditions, particularly at low vapour pressure deficits and often measured early in the diurnal cycle, both of which promote high g s . Therefore dynamic stomatal behaviour in the field environment could significantly decrease realized A when compared with the 'theoretical maximum'   (Figure 1c). Indeed, in a study on the effect of dynamic light on Arabidopsis by Vialet-Chabrand et al. (2017) continuous diurnal gas exchange measurements of A were compared with those determined from A/Q response curve and incident photosynthetic active radiation (PAR), the latter failed to accurately predict the measured photosynthetic rates due to the limitation imposed by stomata (Figure 1c,1) as well Figure 1. Example of a CO 2 assimilation as a function of substomatal CO 2 concentration curve (A/C i ) and light (A/Q) (a, b respectively) measured on the flag leaf of wheat (cv. Robigus) at booting stage with an infrared gas analyzer (Li-Cor 6400, Li-Cor, USA). The A/Ci was measured at saturating light [1500 µmol m À2 sec À1 photosynthetic active radiation (PAR)] and a leaf temperature of 20°C. RuBisCO carboxylation efficiency (V cmax ), the maximum electron transport rate for RuBP regeneration (J max ) was estimated following Sharkey et al. (2007) and A sat represents the light saturated A at current ambient [CO 2 ]. (c, d) Diurnal measurement of photosynthetic CO 2 assimilation (A) and stomatal conductance (g s ) were measured on the same day as the A/C i and A/Q analysis on an adjacent part of the flag leaf at 20°C leaf temperature following a fluctuating light environment. In (c), black dots represent recorded A values, whereas the red line represents A estimated through the A/Q response. Orange regions represent the discrepancy in A between observed and modelled values from the A/Q. Light brown regions represent the potential overestimation in daily CO 2 uptake if A sat is used to assess total daily photosynthesis for plants growing under a natural fluctuating light regime. In (d), grey regions represent the light regimes (therefore the photosynthetically active radiation, PAR) at which the diurnal measurement with the Li-Cor was carried out (following a simulated sunny-cloudy pattern shown in (c)), while blue dots represent g s .  Matthews et al., 2018). Similar methodological drawbacks are present for simpler (and quicker) analysis of instantaneous or 'snapshot' measurements of photosynthesis that are either captured under natural irradiance, or use a light source to mirror in situ irradiance intensities. Stomatal limitation, enzyme activation states and photoinhibition can greatly influence short-term photosynthesis. Additionally the environmental conditions that the plants have been exposed to before measurements also impact on instantaneous measurements, therefore increasing the complexity for data interpretation (Lawson and Weyers, 1999).
Therefore, it is not surprising that A is not always correlated with yield as analyses are often based on either photosynthetic capacity (e.g. A/C i and A/Q curves) that are not realized in the field or instantaneous measurements that represent a single point measurement of A that fails to characterize the diurnal photosynthetic pattern. In Table 1, the best link between yield and photosynthesis was found when integrated CO 2 uptake was determined over the growing season or 'operational' photosynthesis was measured in situ (Rawson et al., 1983;Carmo-Silva et al., 2017 respectively) suggesting that: (i) the different components defining yield are determined over spatial and temporalspecific phenological stages, and therefore A sat (the most used trait estimated in the literature) may correlate to a particular yield component rather than overall grain yield; (ii) A sat is representative of a steady-state and optimal condition that crop plants hardly ever experience in the field, and more realistic conditions for the analysis (e.g. subsaturating light intensities) are the most appropriate way to evaluate the realized A in natural dynamic environments; and (iii) although technically challenging, time consuming and subject to a high degrees of errors (e.g. time of senescence initiation), integrated CO 2 uptake of the most photosynthetically active leaf (i.e. flag leaf) has the potential to be a representative trait linked to grain yield, at least in wheat (Rawson et al., 1983). Therefore, new instrumentation that would enable diurnal and seasonal measurements of realized photosynthesis to be captured under natural dynamic field conditions and at different layers within the canopy is required (e.g. Salter et al., 2018;Murchie et al., 2018;Vialet-Chabrand and Lawson, 2019). For example, the development of the 'OCTOflux' system by Salter et al. (2018), which is a multiplexed semiportable gas exchange system that enables A max to be measured in eight leaves simultaneously. Furthermore, new tools are needed to facilitate high-throughput measurements of photosynthetic capacity in situ and on large numbers of plants, such as the recent developments in hyperspectral imaging to rapidly measure V cmax in the field (ca. 10 sec) (Meacham-Hensold et al., 2019). Although the approaches mentioned above have made significant advancements in measuring photosynthetic capacity, further developments on instrumentation are necessary to enable diel operational or realized photosynthetic rates to be determined, that are subject to the limitations driven by the growth conditions as well as the kinetics of various processes that a plant is subjected to over the dynamic diurnal period.

EXPLOITING NATURAL VARIATION IN PHOTOSYNTHETIC CAPACITY AND STOMATAL FUNCTION FOR IMPROVING CROP PRODUCTIVITY: A CASE IN WHEAT
In wheat, several physiological traits have been unintentionally selected for to produce high yielding cultivars with increased grain number m À2 (GN) and hence yield (Fischer et al., 1998). In the last few years, however, yield has stagnated in many countries suggesting the need for greater effort and new targets for increasing productivity (Ray et al., 2012). The critical and source-limited phase of stem extension determines GN (Slafer et al., 2015). Two not mutually exclusive possibilities have been proposed to increase GN in wheat: (i) lengthening the duration and rate of growth and (ii) increasing resource availability (i.e. photosynthesis) (Miralles and Slafer, 2007). Indeed, increasing sedoheptulose1,7-biphosphatase activity increased flag leaf photosynthetic capacity and GN per spike in greenhouse-grown wheat , suggesting that elevated flag leaf A can increase spike fertility. Most of the work characterizing photosynthesis in wheat has focused on flag leaf A, however, understanding and assessing earlier canopy photosynthetic efficiency (e.g. early stem extension) might be of greater importance to optimize spikelet and floret fertility. Several studies have already reported significant variation in photosynthetic capacity and light saturated rate of photosynthesis in wheat, suggesting the potential exploitation of diversity for selection and/or gene discovery (Driever et al., 2014;Carmo-Silva et al., 2017). In particular, highthroughput phenotyping approaches can help detect important genomic regions for leaf and/or canopy photosynthetic traits in wheat and speed up the selection of desirable traits. Either large panels of wheat with unknown ancestry or bi and multiparental populations (for quantitative trait loci analysis) can be used for this approach, as already demonstrated in rice (Teng et al., 2004;Gu et al., 2012) and recently reviewed by van Bezouw et al. (2019). In addition the development of single-nucleotide polymorphism platforms in wheat (Wilkinson et al., 2012) and the recently annotated genome of bread wheat (Appels et al., 2018) will ensure a more comprehensive understanding of the genetic control of photosynthetic traits or other A-determining traits such as g s and stomatal dynamics.
As yield generally plateaus at high GN due to the tradeoff with grain weight (GW) (Figure 2, scenarios a and b) (Gamb ın and Borr as, 2010; Quintero et al., 2018), understanding and potentially optimizing the GW component is of major importance for wheat yield improvement. Recent work reported the presence of a potential source limitation during grain filling ( Alvaro et al., 2008;Xie et al., 2015;Quintero et al., 2018). These reports suggest that increased A in post-anthesis would help facilitate the attainment of the potential maximum individual GW, especially if GN is increased (Figure 2, scenario c). GW can rely on three main sources of assimilates: leaf photosynthesis, spike photosynthesis and the remobilization of the water-soluble carbohydrates (WSC) from the stem. While efforts have largely focused on selecting and screening for post-anthesis leaf photosynthetic duration (Blake et al., 2007) and WSC concentration (Rebetzke et al., 2008), spike photosynthesis is an unexplored determining component contributing to GW. When compared with the flag leaf, the spike has shown a higher degree of drought tolerance (Tambussi et al., 2005(Tambussi et al., , 2007 generally explained by a greater intrinsic water-use efficiency (driven by a low g s per unit area and a high degree of re-fixed respiratory CO 2 ) and a more pronounced osmotic adjustment (Tambussi et al., 2005(Tambussi et al., , 2007. This situation suggests that spike photosynthesis has an important role in times of water limitation, possibly compensating the flag leaf during grain filling. Furthermore, the assimilates produced in the spike are directly translocated into the grains (Carr and Wardlaw, 1965) leading to a contribution to GW between 10 and 45% depending on environmental conditions and genotype tested (Maydup et al., 2010;Sanchez-Bragado et al., 2016). Indeed a large variation in gross spike A (calculated as the sum of A and dark respiration (R d ) as a proxy of respiration in the light) has been shown in both durum and bread wheat (Maydup et al., 2010;Molero et al., 2013;Zhou et al., 2016;Sanchez-Bragado et al., 2016), suggesting the existence of natural genetic diversity for exploitation. For instance, the presence of awns (lemma-derived organs) has been considered an important source of external CO 2 assimilation of the spike (Maydup et al., 2010) although other factors such a spike morphology (e.g. photosynthetic surface area of spikelets) seems to drive the observed variation in spike A (Guo and Schnurbusch, 2016). Earlier evidence proposes that, in the UK, a significant genotypic variation for spike gross A and for the contribution of spike A to GW is present (Faralli et al., 2019c) and confirms the importance of spike photosynthetic CO 2 assimilation for grain filling. Additional work is needed to fully understand the underlying mechanism of spike A, as well as the extent of existing natural variation. Further development of high-throughput phenotyping tools focusing on spike A would take full advantage of this unexploited trait for GW improvement.

CONCLUSION
Photosynthesis is a key determinant of crop yield. Large natural variation in A and A-determining traits in different photosynthetic organs exists in a number of crop species that represent a currently unexploited target for crop improvement. Owing to the complexity of the relation between A and yield, improvements in high-throughput, reliable and relevant methodologies will enable the dissection of useful genetic targets for marker-assisted selection. In wheat, enhancing leaf canopy photosynthesis will increase GN although greater yield will only be achieved with a parallel increase in GW, which relies primarily on enhanced spike photosynthesis. With this in mind, screening for high photosynthetic capacity in both organs should be considered a prime target for high yielding wheat cultivars. In summary, genetic manipulation and elevated [CO 2 ] experiments have shown a yield advantage when photosynthesis is increased in food crops; therefore exploiting natural genetic variation in photosynthesis will facilitate the development of cultivars with greater yield potential.