Variation in key leaf photosynthetic traits across wheat wild relatives is accession dependent not species dependent

(cid:1) The wild relatives of modern wheat represent an underutilized source of genetic and phenotypic diversity and are of interest in breeding owing to their wide adaptation to diverse environments. Leaf photosynthetic traits underpin the rate of production of biomass and yield and have not been systematically explored in the wheat relatives. This paper identiﬁes and quantiﬁes the phenotypic variation in photosynthetic, stomatal, and morphological traits in up to 88 wheat wild relative accessions across ﬁve genera. Both steady-state measurements and dynamic responses to step changes in light intensity are assessed. (cid:1) A 2.3-fold variation for ﬂag leaf light and CO 2 -saturated rates of photosynthesis A max was observed. Many accessions showing higher and more variable A max , maximum rates of carboxylation, electron transport, and Rubisco activity when compared with modern genotypes. Variation in dynamic traits was also signiﬁcant; with distinct genus-speciﬁc trends in rates of induction of nonphotochemical quenching and rate of stomatal opening. (cid:1) We conclude that utilization of wild relatives for improvement of photosynthesis is supported by the existence of a high degree of natural variation in key traits and should consider not only genus-level properties but variation between individual accessions.


Introduction
Modern hexaploid bread wheat cultivars are the product of a genetic bottlenecka reduction in genetic diversity brought about by domestication through polyploidization and the intensive selection for agronomically important traits over the past 10 000 yr (Charmet, 2011;Faris, 2014). However, despite breeding efforts in recent years, increases in global yields have slowedaveraging between 0 and 1.1% annually (Dixon et al., 2009). Potential yield gains have been circumvented by increasingly unpredictable environmental conditions, susceptibility to biotic stresses, and agronomic practices (Brisson et al., 2010). During the processes of domestication and selection, modern wheat may have lost key alleles required for adaptive robustness to abiotic and biotic stressa negative side effect resulting from single trait selection. With pressure to raise yields between 1.6 and 2.4% per annum over the next 50 yr (Brisson et al., 2010;Ray et al., 2013), the emphasis falls to increasing the genetic diversity of modern wheat to maintain or improve yields under current environmental conditions (Evans & Lawson, 2020).
The wild, uncultivated relatives of modern wheat provide a global and mostly underutilized source of genetic and phenotypic diversity (King et al., 2017), with over 80 000 wheat accessions documented (Crop Wild Relative Diversity, 2019). The wild relatives represent adaptation to diverse habitats and climates, suggesting the existence of genes that are unavailable in the existing elite wheat germplasm. Exploration of the wild relatives may uncover previously untapped potential for wider and enhanced characteristics to face changing environmental conditions and disease resistance, ultimately aiming to improve the productivity and resilience of future modern varieties.
Despite the high probability of discovering novel traits and genes, the current number of species and accessions already investigated as potential candidates for crop improvement is relatively few. Though the majority of studies have generated lines that are resistant to biotic or abiotic stresses, a smaller proportion have reported the transfer of more complex polygenic traits, including yield stability, yield gains (Villareal et al., 1996), and increases in photosynthetic rates (Austin et al., 1982). Photosynthesis is a complex polygenic trait that fundamentally underpins the rate of production of biomass, and ultimately of yield. The close relationship demonstrated between enhanced photosynthetic CO 2 assimilation, biomass, and yield under elevated, ambient CO 2 supports the assumption that enhancing the capacity of individual leaves to fix carbon (C) will support higher yields though increased biomass (Long et al., 2006;Zhu et al., 2008Zhu et al., , 2010Zhu et al., , 2018Murchie et al., 2009).
The relationship between photosynthesis and biomass accumulation is a ubiquitous and cumulative process involving numerous, interconnected processes. For example, at the canopy level, canopy architecture can drastically change both the amount of intercepted light and the efficiency with which it is converted into biomass and yield (Murchie et al., 2018;Wu et al., 2019). At the leaf level, under optimal environmental conditions, strong positive correlations are observed between stomatal conductance, carboxylation capacity, and the rate of leaf CO 2 uptake (Wong et al., 1979). Under saturating light, the Rubisco carboxylation efficiency and capacity are vital (Carmo-Silva et al., 2015). Under fluctuating environmental conditions, asynchronies arise between the changes in light intensity and the limitations imposed by the components of Rubisco, stomatal conductance (g s ), and photoprotection (Lawson & Blatt, 2014;Kromdijk et al., 2016;McAusland et al., 2016;Taylor & Long, 2017;Murchie & Ruban, 2019), leading to restricted CO 2 assimilation and lowered water-use efficiency. Identifying species-specific rapid responses of stomata, photoprotection, and Rubisco is crucial in maximizing CO 2 uptake and minimizing water loss in response to a fluctuating field environment (Faralli et al., 2019b). In addition, the anatomy of the leaf also has a large impact on the diffusion pathway of CO 2 from traits such as stomatal density and distribution (Pearson et al., 1995;Faralli et al., 2019b), the size of the intracellular spaces (Lundgren et al., 2019), and the distance between veins (Richards, 2000;Terashima et al., 2010;Reynolds et al., 2012;Burgess et al., 2015;Driever et al., 2017).
The convergence of these traits provides a plethora of targets to improve photosynthesis, but natural genetic variation that can be used for breeding is unclear, especially in wheat (Driever et al., 2014;Salter et al., 2019). In some cases the underlying genes are known, along with the mechanism likely to underlie the improvement, and these currently rely on genetic modification to achieve both the proof of concept and the putative improved variety (e.g. Kromdjik et al, 2016). Clearly, the ability to combine discrete improvements will offer the greatest possibilities for yield improvement, and recent work has highlighted the need to improve photosynthesis in dynamic environments and across different climatic conditions (Tanaka et al., 2019;Wu et al., 2019;Faralli et al., 2019a,b).
As wild relatives are increasingly used to introduce genetic diversity into modern cultivars it is key to rapidly identify potential candidates with higher photosynthetic rates and efficiencies, to determine the basis for these improvements, and to link the physiological and/or anatomical traits with gene discovery. These steps are not linear, and one process may inform the discovery of another. One of the limiting factors has been the inability to rapidly genotype large numbers of lines, but this has recently been overcome (King et al., 2017;Devi et al., 2019;Grewal et al., 2020). Another significant bottleneck has been the availability of sophisticated high-throughput photosynthesis phenotyping tools. Recent advances in phenotyping techniques and pipelines provide the means to investigate variation on a larger scalefrom screening populations of thousands to assessing the mechanistic basis behind that variation (McAusland et al., 2015(McAusland et al., , 2019Murchie et al., 2018;Silva-Perez et al., 2018;Araus et al., 2018).
Here, we undertake a large-scale analysis of static and dynamic photosynthesis traits across accessions of wheat wild relatives. Our overall aim was to represent wide background adaptations, therefore we sourced these from a range of climates and continents across South America, western and eastern Europe, the Middle East, South and East Asia (Fig. 1). This work highlights for the first time the phenotypic and genomic diversity in the wild relatives as a source of variation for improving photosynthesis in modern wheat.
We show substantial variation in key traits, with specific wild accessions demonstrating a greater diversity and superior leaf photosynthesis in comparison with cultivated elite lines in both steady-state measurements and dynamic traits. These discoveries are important for developing strategies for selective inclusion into pre-breeding and breeding programmes.

Materials and Methods
Seed was obtained from the collection held by The Nottingham BBSRC Wheat Research Centre (University of Nottingham, Sutton Bonington, UK). A total of 88 accessions were investigated for variation in CO 2 and light-saturated photosynthesis, consisting of 41 species and five genera. A table of all plant material measured is presented in Table S2. Not all measurements were possible on all genotypes owing to the large variation in rates of growth and development. Accessions used were genotyped as a record that will be of use in future work to ensure that the same genotype was used across different/similar studies; these data are available at https://www.cerealsdb.uk.net/cerealgenomics/Cerea lsDB/indexNEW.php. All sample sizes n stated in this paper are the number biological replicates. relatives, the modern cultivars were grown and measured three times during this time period; September 2016-October 2016, November 2017-January 2018, and July-August 2018. The glasshouse was located at Sutton Bonington Campus, University of Nottingham, Leicestershire, UK (52°49 0 41.52″N, 1°14 0 54.60″W). Glasshouse conditions were maintained at 25 AE 2°C : 18 AE 2°C, day : night, under regular mildew, aphid, and thrip control measures applied following the manufacturer's recommendations. Photosynthetic photon flux density (PPFD) was maintained to achieve 16 h of light using supplemental lighting (HPS PL90E+ with Son-T Argo; Philips, Guildford, UK), applying up to 250 μmol m −2 s −1 PPFD at plant height when ambient PPFD fell below 500 µmol m −2 s −1 .
Using the flag leaf (Zadoks growth stage 4.1-4.5; Zadoks et al., 1974), the response of photosynthetic CO 2 assimilation rate A to 11 different external CO 2 concentrations C a was measured. Leaf intercellular [CO 2 ] C i and A were measured using an infrared gas analyser (LI-6400-XT; Li-Cor, Lincoln, NE, USA). Measurements started at an ambient C a of 400 µmol mol -1 before C a was decreased stepwise to a lowest concentration of 50 µmol mol -1 , then increased stepwise to an upper concentration of 1500 µmol mol -1 . Readings were recorded when A had stabilized to the new C a (c. 2 min). To induce rapid responses, the leaf was exposed to a high saturating irradiance of 2000 µmol m −2 s −1 , using a 2 cm 2 leaf chamber with a blue-red LED light source. Leaf temperature and vapour pressure deficit were maintained at 25°C and 1.2 kPa, respectively.
The A-C i response curves were fitted using the Farquhar et al. (1980) model using the R PLANTECOPHYS package (Duursma, 2015). The maximum velocity of Rubisco for carboxylation V c,max and the maximum rate of electron transport demand for ribulose 1,5-biphosphate (RuBP) regeneration J max were estimated. All calculations were done in the R environment, v.3.5.0 (R Core Team, 2016).

Chl fluorescence imaging
Chl fluorescence imaging was performed on a subset of 25 wild relatives that represented good diversity in A-C i responses and on three modern genotypes using a customized FluorCam imaging pulse-amplitude modulated fluorometer (Photon Systems Instruments, Brno, Czech Republic), as in McAusland et al. (2019). The same individual plants were used as those for the A-C i analysis. Shutter time and sensitivity of the charge-coupled device were adjusted in accordance with the sample. The FluorCam was located in a temperature-controlled dark room maintained at 20 AE 2°C. Flag leaves were excised at 09:00 h and allowed to dark-adapt for 1 h in a custom imaging chamber. Fifteen minutes before the initial saturating pulse to determine F v /F m was taken, a pre-mixed gas of 400 µmol mol −1 CO 2 and 2% oxygen (O 2 ) was used to saturate the chamber to ensure consistent CO 2 and O 2 concentrations around the samples. The protocol consisted of three consecutive light steps of 15 min: 500, 100, and 1000 μmol m −2 s −1 PPFD (the latter is the capacity for the device). Saturating pulses were taken every minute throughout the protocol. The values of F v /F m and the responses of F v 0 /F m 0 (maximum efficiency of photosystem II (PSII) in the light), F q 0 /F m 0 (operating efficiency of PSII in the light), photochemical quenching (qP), and nonphotochemical quenching (NPQ) were extracted from each protocol (for an in-depth description of these parameters see Maxwell & Johnson, 2000;Baker, 2008;Murchie & Lawson, 2013).
To determine the rate of NPQ relaxation (under 100 μmol m −2 s −1 PPFD) and induction (1000 μmol m −2 s −1 PPFD), data Fig. 1 A selection of the global origins of the wild relatives presented in this paper with genus indicated by shape: Aegilops (circles), Triticum (triangles), Thinopyrum (squares). The shading represents the 12 wheat mega-environments, typified by CIMMYT as broad, diverse growing environments, allowing targeted development of germplasm for improved yields globally (Rajaram et al., 1995).

Ó2020 The Authors
New Phytologist Ó2020 New Phytologist Trust New Phytologist (2020) 228: 1767-1780 www.newphytologist.com were fitted using a three-factor exponential function (Eqn 1)using the curve-fitting toolbox in MATLAB (R2018a; The MathWorks Inc., Natick, MA, USA) : where a determines the initial value, b is a constant representing the rate of exponential decay or growth and c is a constant describing the vertical shift in NPQ from start to end of the step.
y ¼ ae Àbx Eqn 2 (a, initial value; b, a constant representing the rate of exponential decay or growth). To determine the time t taken to achieve either 50% of the maximum NPQ values I 50 or 50% of the maximum NPQ values R 50 , the equations were solved for b and the following calculations applied: (t, time constant; b, obtained from the rearrangement of either Eqn 1 or Eqn 2).

Leaf properties
Flag leaf adaxial absorbance for the leaves used in the Chl fluorescence screen (25 wild relative accessions and three modern genotypes) was measured using an integrating sphere (Li1800-12; Li-Cor) and spectroradiometer (ASD HandHeld 2: Hand-held VNIR; Malvern Panalytical, Boulder, CO, USA). Multiple flag leaves from the same plant were aligned to cover the measurement window if a single flag leaf was < 2cm 2 . Absorbance was calculated in MATLAB (R2018a). Specific leaf area (SLA, m 2 kg −1 ) was estimated following the protocol of Cornelissen et al. (2003); in brief, after photographing and measuring the fresh weight of the flag leaf, the leaves were placed in an oven at 70°C for 72 h. The leaves were reweighed to determine dry weight and the area calculated to produce SLA (IMAGEJ; Rasband, 1997Rasband, -2018. Ear number per plant was also measured within this subset of plants.

Rubisco total activity and in vitro maximum carboxylation activity
Rubisco total activity was determined in flag leaves of glasshousegrown plants (Sutton Bonington, University of Nottingham) for three modern Triticum cultivars (Triticum aestivum) and 19 wild relatives, from the same individual plants grown for A-C i analysis in 2018 between the phenological Zadoks stages 4.2-5.5 (Zadoks et al., 1974). Leaf segments were snap-frozen in liquid nitrogen (N 2 ) and stored at −80°C. Rubisco was extracted as described by Carmo-Silva et al. (2017), and Rubisco total activity was measured by the incorporation of 14 CO 2 into acid-stable products at 30°C (Parry et al., 1997). The radioactivity was measured by liquid scintillation counting (Packard Tri-Carb; PerkinElmer, Waltham, MA, USA). Total soluble protein (TSP) was quantified by the Bradford assay (Bradford, 1976). Rubisco was extracted from flag leaf tissue of three modern wheat cultivars and six wild relatives. The maximum in vitro carboxylation rate V cmax of fully activated Rubisco was determined (Prins et al., 2016), incorporating the modifications described by Orr et al. (2016). Rubisco was quantified by the 14 C-carboxyarabinitol bisphosphate binding method of Whitney et al. (1999).

Stomatal dynamics
Twenty-one accessions were grown in glasshouses at the University of Essex (Colchester, UK). Seedlings were vernalized as described earlier and potted into 650 cm 3 pots containing peatbased compost (Levington F2S). Solar radiation provided a PPFD of c. 500 μmol m −2 s −1 , supplemented by sodium vapour lamps (600 W; Hortilux Schréder, Monster, the Netherlands) to 300 μmol m −2 s −1 PPFD when external PPFD dropped below 1200 μmol m −2 s −1 over a 10 h period. Air temperature was maintained at 25 AE 3°C during the day and 18 AE 3°C at night. The rapidity of the stomatal and photosynthetic responses was studied on fully expanded fourth leaves (growth stage Z1.4). Leaves were enclosed in a gas-exchange chamber (LCpro-SD; ADC BioScientific Ltd, Hoddesdon, UK) and left to equilibrate under dark conditions (c. 30 min). Gas exchange was recorded every 1 min for 5 min under dark conditions, and light was set to 1200 µmol m −2 s −1 for another 1 h. Leaf temperature was set at 25°C, leaf vapour pressure deficit was maintained at c. 1.2 kPa, and [CO 2 ] was set at 400 ppm. In order to describe the temporal response of stomatal conductance to water vapour g s to a single step-change in PPFD, an analytical model derived from the model by Vialet-Chabrand and co-workers (Vialet-Chabrand et al., 2013;McAusland et al., 2016) was used. In brief, this dynamic model predicts the temporal response of g s to PPFD using an asymmetric sigmoid function parameterized by specific time constants to describe the opening response of stomata.

Statistical analyses
Statistical analyses were conducted in R (http://www.r-project.org/). A Shapiro-Wilk test was used to test for normality, and a Levene test of homogeneity was used to determine if samples had equal variance. Single factor differences were analysed using a one-way ANOVA with a Tukey-Kramer honest significant difference test where more than one group existed or using a Student's t-test where only two groups were compared. For analysing more than two dependent variables, a MANOVA was used.

Diversity of photosynthetic responses
A-C i response curves were used to determine a 2.3-fold difference in light and CO 2 -saturated photosynthetic assimilation A max New Phytologist (2020) 228: 1767-1780 Ó2020 The Authors New Phytologist Ó2020 New Phytologist Trust www.newphytologist.com

Research
New Phytologist between 88 accessions across five genera (Fig. 2). Overall, the modern varieties did not show a higher A max than the wild relatives, and the variation was largely accession dependent, not species dependent. Of the 88 accessions, 11 wild relatives demonstrated significantly higher values of A max (P < 0.05) than at least one of the three modern Triticum cultivars measured. Significant differences were observed between the individual accessions (P < 0.0001, F (88,477) = 5.71) and genera (P = 0.0016, F (5,560) = 3.96; see Fig. S1a). Secale A max values were significantly lower than all other wild relative genera (P < 0.03), but not significantly different to the modern Triticum. It is notable that A max for many accessions exceeded that seen in elite lines (38.4-47.0 μmol m −2 s −1 ; Driever et al., 2014).
The Aegilops genus exhibited the greatest variation in V cmax and J max (5.8 and 4.4-fold, respectively; Fig. S1b,c), whereas the values of the modern Triticum cultivar were more conserved (2.7 and 2.4-fold, respectively). Secale accessions had significantly lower V cmax and J max (P < 0.05) than the Aegilops, Thinopyrum, and wild and modern Triticum genera. Overall, the ratio V cmax : J max (Fig. 3) was significantly (P < 0.05) lower in the modern Triticum cultivars than in Secale, Thinopyrum, Amblyopyrum and Aegilops genera, but not significantly different to the wild relative Triticum (P = 0.23).
V cmax : J max has been suggested to indicate the limiting step of CO 2 assimilation, which is estimated by the transition point at which A shifts from being Rubisco to RuBP limited (C itransition ). C itransition was significantly different between accessions (P < 0.0001) and genera (P < 0.0001) (Fig. S2a). Ranging between 197.73 and 455.36 μmol mol −1 CO 2 , the modern Triticum genus demonstrated the lowest C itransition values (258 AE 65.16 μmol mol −1 ), whereas members of the Secale (307.38 AE 117.74 μmol mol −1 ) were the highest (Fig. S2b).
The amount of TSP and Rubisco total activity (RV t ) were used to quantify the investment of 19 accessions into leaf rubisco (Fig.  4). The ratio of RV t to TSP was broadly genus specific, with modern Triticum showing considerably higher RV t for similar values of TSP. This higher RV t is likely to represent a greater investment of modern Triticum in Rubisco protein, since in vitro V cmax was not significantly different between genera (Fig. S3).
Genus and accession-specific differences were determined for TSP (P < 0.02) and RV t (P < 0.0001), although the modern Triticum cultivars were only found to have significantly higher TSP (P < 0.02) compared with the Amblyopyrum accessions. These modern cultivars had significantly higher RV t (P < 0.003) when compared with the five genera studied. Cultivar T. aestivum 'Paragon' exhibited 3.2-fold greater RV t than Aegilops muticum 2130004 did, which had the lowest activity of the accessions studied (31.9 AE 9.0 μmol CO 2 m −2 s −1 ).

Dynamic photosynthesis
The responses of 27 accessions in Chl fluorescence parameters to step changes in PPFD were analysed ( Fig. S4; Table S3). This technique was utilized to uncover variation in dynamic photosynthetic traits; for example, in the speed of induction on transfer to high light or the kinetics of decay on transfer to low light (Fig.  S4a). For a summary of the significant interactions for each light step, see Table S3. At steady state, the greatest number of significant differences were found under high light (1000 μmol m −2 s −1 PPFD). Maximum PSII efficiency in the light (F v 0 /F m 0 ) under 1000 μmol m −2 s −1 PPFD was significantly higher in the Secale than in the wild relative Triticum, Aegilops and Amblyopyrum genera (P < 0.05), whereas no significant www.newphytologist.com differences were determined between the genera under 500 or 100 μmol m −2 s −1 PPFD. Similarly, no significant differences were determined under any light intensity between the modern Triticum cultivars and the five wild relative genera. In general, accessions from the Amblyopyrum and Aegilops genera achieved the highest values of PSII operating efficiency (F q 0 / F m 0 ; Fig. S4b) and electron transport rate (ETR) under 500 and 1000 μmol m −2 s −1 PPFD. These values were consistently, significantly higher (P < 0.05) than in the accessions from the wild and modern Triticum genera. Under low light (100 μmol m −2 s −1 PPFD), the Secale genus achieved significantly (P < 0.05) higher F q 0 /F m 0 and ETR than the wild Triticum accessions did. By contrast, the wild and modern Triticum accessions achieved significantly higher values of NPQ under 500 and 1000 μmol m −2 s −1 PPFD intensities than the Secale and Amblyopyrum accessions did. The wild and modern Triticum also maintained the highest NPQ at 100 μmol m −2 s −1 PPFD when compared with Secale and Amblyopyrum. It was notable that those genera achieving the highest values of NPQ under 1000 μmol m −2 s −1 PPFD also demonstrated higher NPQ under low light but also obtained the greatest magnitude of change in NPQ from 100 to 1000 μmol m −2 s −1 PPFD (Fig. S5).
The kinetics of NPQ in response to changes in irradiance are indicative of the dynamic equilibrium between photoprotection and photochemistry in response to the fluctuating light environment experienced in the field. The analysis presented here indicates interesting and accession-dependent trends in the speed of NPQ induction and relaxation. For example, the time taken to induce 50% of the maximum NPQ under 1000 μmol m −2 s −1 PPFD (I 50 ; Fig. 5a) was consistently lower than the time taken to relax to 50% minimum NPQ under 100 μmol m −2 s −1 PPFD (R 50 ; Fig. 5b) for all accessions. Interestingly, there was no correlation observed between I 50 and R 50 (R 2 = −0.06, P = 0.80).
In general, although there were significant accession-specific differences in I 50 (P < 0.0001) and R 50 (P < 0.0001), the range of I 50 (1.9-7 s) was much narrower than that observed for R 50 (26-133 s). In general, the wild Triticum species took the least time to induce (2.13 AE 0.43 s), whereas the modern Triticum took the longest (5.34 AE 1.25 s). By contrast, the modern Triticum achieved the fastest R 50 time (79.32 AE 20.61 s), Fig. 3 The correlation between maximum rate of carboxylation V cmax and electron transport J max for 88 accessions across six genus groups. Data are individual points (n = 3-25 biological replicates). The average ratio J max : V cmax was calculated for each genus (micromoles of electrons per micromole of CO 2 ) AE SD (Carmo-Silva et al., 2017). A linear regression was fitted to all data within each genus and a Pearson's correlation coefficient calculated. The mean V cmax and J max for the modern Triticum genus is indicated on each plot by a red circle.

Research
New Phytologist whereas the Thinopyrum species took the longest to relax (111.02 AE 19.73 s). We conclude that modern wheat has faster NPQ relaxation than the wild relatives do with a slower induction. Some genus outliers were noted; for example, Aegilops biuncialis 550945 and Triticum monococcum took respectively 2.7 and 1.5-fold longer to induce NPQ than the other Aegilops and Triticum species did. On average, Triticum timopheevii took 245 s to relax to 50% minimum NPQ, compared with the Triticum average of 70 s. Interestingly, those plants with a faster NPQ induction might have a higher capacity for photosynthesis: a significant negative relationship was observed between A max and I 50 (Fig. 6a; R 2 = −0.46, P < 0.03), whereas no correlation was observed between A max and R 50 ( Fig. 6b; R 2 = 0.01, P = 0.96).

Stomatal dynamics
Stomatal responses are an order of magnitude slower than metabolic processes. To investigate variation in stomatal responses, 24 accessions were subjected to a step increase in PPFD from 100 to 1000 μmol m −2 s −1 PPFD and a model used to determine both magnitude of opening and rapidity of response (Vialet-Chabrand et al., 2013). Substantial variation in dynamics of stomata was observed among the wild relatives, with some faster than the modern wheat. At steady state, no significant correlation (R 2 = −0.05, P = 0.81) was determined between A and g s under 1000 μmol m −2 s −1 PPFD (A 1000 and g s1000 , respectively; Fig. 7a). In addition, no correlation was observed between g s1000 and g s under 100 μmol m −2 s −1 PPFD (R 2 = 0.14) or between A 1000 and A max (R 2 = 0.02, P = 0.93).
A significant positive relationship (R 2 = 0.87, P < 0.0001) was observed for the time taken to achieve 95% A 1000 and g s1000 (Fig. 7b); however, all species took longer to achieve 95% g s1000 compared with 95% A 1000 . Those species that took the longest to achieve 95% A 1000 also achieved higher A 1000 values (R 2 = 0.41, P < 0.05), with the modern Triticum accessions achieving the highest A 1000 values (19.9 AE 2.1 μmol m −2 s −1 ), on average 2.8 AE 0.6 μmol m −2 s −1 greater than the other five genera measured. The species achieving the highest mean A 1000 for the lowest mean g s1000and hence the highest intrinsic water-use efficiency (W i , µmol CO 2 mmol −1 H 2 O m −2 s −1 )was T. monococcum TM01 (0.044 μmol CO 2 mmol −1 H 2 O m −2 s −1 ), whereas the lowest W i was found for Aeglps caudata 2090001 (0.022 μmol CO 2 mmol −1 H 2 O m −2 s −1 ).
On average, the stomata from Thinopyrum accessions were the fastest (6.1 AE 1.3 min) to achieve 95% A 1000 , whereas the Aegilops accessions were the slowest (8.6 AE 1.5 min). To achieve 95% g s1000 , all species opened for an average of 7.0 min (AE 2.3 min) longer than was required to achieve 95% A 1000 . The longest time was found in S. cereale 428373, which achieved 95% g s1000 c. 4 min later than the other genera (11.0 min). There was no correlation between the maximum rate of g s increase to a doubling in PPFD from 100 to 1000 μmol m −2 s −1 (Sl max ) and the time constant used to describe the time taken to achieve steady-state g s (k) and A 1000 (R 2 = −0.14, P = 0.51 and R 2 = 0.19, P = 0.37, respectively). To relate speed with magnitude of opening, Sl max and k were compared with g s1000 ; whereas a positive correlation was observed between Sl max and g s1000 (R 2 = 0.60, P = 0.01), no correlation was observed between k and g s1000 (R 2 = −0.04, P = 0.85).

Whole plant and leaf characteristics
Leaf morphological analysis was carried out on a selection of the accessions (25 wild relative accessions and three modern genotypes; Figs S6-S9). Though we do not expect yield components to be of value in the undomesticated species, we note some interesting relationships relevant to the theme of this paper. There was a genus-specific negative correlation between SLA and A max (Fig. S6), suggesting a functional trade-off between leaf thickness, density of photosynthetic components, and photosynthetic capacity. We also noted negative correlations between SLA (Fig.  S7a), leaf absorbance (Fig. S8, R 2 = −0.50, P = 0.03), and total Rubisco activity (R 2 = −0.49, P = 0.04). A positive correlation was also observed between SLA and ear number (Fig. S9; R 2 = 0.70, P = 0.0001).
The Aegilops accessions were found to have the highest SLA ( Fig. S7a; P < 0.003), whereas Thinopyrum had the lowest (P < Fig. 4 Total soluble protein and Rubisco total activity was determined from the flag leaves of three modern Triticum (black squares) varieties and 19 wild relative species across five genera: Aegilops (red circles), Ambylopyrum (black triangles), Secale (black plus symbols), Triticum (black asterisks), and Thinopyrum (black enclosed crosses). The modern cultivar, Triticum aestivum 'Paragon', is highlighted (red square), and ellipses highlight the largest genus groups modern Triticum, Triticum and Aegilops. Data are the means (n = 2-5 biological replicates).

Discussion
Physiological and genotypic exploration of variation associated with the improvement of photosynthesis is crucial for the successful introduction and identification of interrelated traits that improve biomass acquisition in staple crops such as wheat and rice. Using wild relatives to introduce new genetic diversity into elite cultivars has gathered momentum in recent years (King et al., 2017;Prohens et al., 2017), generating a need for understanding the breadth of physiological variation available to breeding programmes, particularly traits that promote greater C acquisition. Here, we address this for the first time and encompass a wide selection of notable wild relatives, originating from > 14 countries, covering 35 polyploid genomes, six genera, and 37 genetically distinct species ( Fig. 1; Table S2). We have provided a database of photosynthetic traits, shown substantial variation for key photosynthetic traits, and discovered novel patterns among the wild relatives that partly explain the underlying causes of the differences observed.
Our results suggest that highly significant variation exists between the six genera, both for the capacity to fix CO 2 under saturating conditions (Figs 2-4) and for temporal processes that facilitate the achievement of high rates of CO 2 uptake under fluctuating conditions, such as those found in the field. High flag-leaf photosynthesis can be correlated to grain yield in modern wheat cultivars (Fischer et al., 1998;Gaju et al., 2016;Carmo-Silva et al., 2017). This is often linked to Rubisco activity. Curiously, though Rubisco total activity was found to be highest in the modern cultivars, the high A max values observed in the wild relative genera were underpinned by higher and wider ranges of maximum carboxylation V cmax and electron transport J max . In addition, the wild relative genera exhibited greater J max : V cmax ratios, driven by higher values of J max (and hence the electron-transportmediated rate of RuBP regeneration; Fig. 3). Whereas some accessions (such as Aegilops juvenalis 574463 (#1), Thinopyrum ponticum 547312 (#70) and T. dicoccoides P95983.2 (#76)) utilized this relationship for increased rates of C uptake, some accessions (e.g. Aegilops biuncialis 550940 (#29)) still achieved high J max : V cmax ratios without the accompanying increase in A max . Therefore, there is broad lack of tightness between carboxylation (a) (b) Fig. 7 (a) The relationship between maximum stomatal conductance g 1000 and CO 2 assimilation A 1000 under 1000 μmol m −2 s −1 PPFD. (b) To highlight stomatal limitation on CO 2 assimilation between species, the relationship between the times taken to achieve 95% (t 95 ) of A 1000 and g 1000 under 1000 μmol m −2 s −1 photosynthetic photon flux density are plotted and a 1 : 1 relationship is indicated (dashed line). For both plots, data are the mean for each species, with genus indicated by shape: modern Triticum (black squares), Aegilops (black circles), Ambylopyrum (black triangles), Secale (black plus symbols), Triticum (black asterisks), and Thinopyrum (black enclosed crosses). The modern cultivar, Triticum aestivum 'Paragon', is highlighted (red squares). For plot (a), the larger the data point the greater the time taken to achieve 95% A 1000 ; for plot (b), the larger the shape the greater the time constant used to describe the time taken for stomata to open (k) and a selection of accessions are labelled. A linear regression is fitted to all data points, with the shading indicating a 95% confidence interval on the fitted values. A Pearson's correlation coefficient is shown in the bottom right corner of each plot.

Ó2020 The Authors
New Phytologist Ó2020 New Phytologist Trust New Phytologist (2020) 228: 1767-1780 www.newphytologist.com capacity and electron transport capacity across the material analysed here. However, J max values estimated from A vs C i analyses may not always accurately represent maximum electron transport rate (Buckley & Diaz-Espejo, 2015). Calculation of C itransition indicates the likely point for limitation of CO 2 assimilation (i.e. RuBP carboxylation or regeneration) and also indicates N partitioning among photosynthetic components (Yamori et al., 2011). Greater leaf N content may decrease the V cmax : J max (Yamori et al., 2011); therefore, it is possible that some variation could be explained by differences in N investment between electron transport and Rubisco. To support this, modern lines in the current study had the lowest C itransition (Figs 3, S2b). Higher Rubisco activity in modern varieties is not a surprise since it would be expected that these genotypes would take up and accumulate a higher leaf N content for ready mobilization from leaf to the grain during senescence (Havé et al., 2017). We speculate that the higher V cmax in some wild relatives may be related to possession of thick, narrow leaves (a tendency in some lines), which is supported by the negative relationship between A max and SLA (Fig. S6). Quantifying variation in V cmax and J max is vital in modelling C exchange at different scales (Rogers et al., 2017;Bloomfield et al., 2019) and is mediated by the balance of leaf N and phosphate, often evidenced by changes in SLA (Reich et al., 1997;Evans & Poorter, 2001). Though changes in SLA can alter the N content per unit leaf area, they can also change the light absorbance of the leaf, with decreases in J max being somewhat negated by increases in absorbance (Evans & Poorter, 2001). The modern cultivars in this study had some of the lowest electron transport rates but also achieved some of the highest absorbance values, accompanied by low SLA (Figs S1, S7a, S8).
It should also be noted that increases in J max (such as those observed for the wild relatives) are not always directly associated with improvements in CO 2 uptake by the Calvin-Benson cycle. With high electron transport rates increasing the reducing power for other essential pathways in the leaf, such as chloroplastic conversion of nitrate to ammonium (Anderson & Done, 1978;Searles & Bloom, 2003), driving production of isoprene (Morfopoulos et al., 2013), and driving alternative electron sinks, such as the Mehler reaction, reflecting changes in the apportionment of photosynthetic proteins (Yamori et al., 2005). Changes in the V cmax to J max ratio may also reflect the resource allocation bias of the plant to maintain high photosynthetic rates to set down biomass and remain competitive (Bryant et al., 1998), with leaves at the top of the canopy limited by V cmax rather than J max at ambient CO 2 (Quebbeman & Ramirez, 2016). Walker et al. (2014) modelled the instantaneous relationship between J max and V cmax under fluctuating light, hypothesizing that increases in J max compensated for Rubisco carboxylation limitations under high light, thus increasing photoprotection and buffering against photoinhibition (Walker et al., 2014). By contrast, the work presented here showed NPQ 1000 was highest in modern Triticum and wild Triticum genera (Fig. S5), despite the lower J max values. Interestingly, the modern cultivars also demonstrated some of the slowest rates of NPQ induction (I 50 ; Fig. 5a). This suggests that higher rates of electron transport supported faster induction in a handful of wild relatives but not greater magnitudes of photoprotection under high light conditions (Fig.  S5).
Manipulating photoprotection is another route for increasing yield, and through identifying and manipulating the magnitude and response timings of NPQ this has led to improvements in crop productivity (Hubbart et al., 2012;Kromdijk et al., 2016). Interestingly, 18 of the 24 wild relative accessions responded faster than the modern cultivar 'Paragon' to a step increase in PPFD, which suggests there may be some room to improve NPQ induction and relaxation in this modern cultivar. The significant negative correlation between A max and I 50 is a strong example of screening for a temporal trait that can be linked with improvements in photosynthetic capacity. In rice, fast induction of NPQ was linked to an inhibition of the rise in CO 2 assimilation (Hubbart et al., 2012), consistent with this trend. Furthermore, these data suggest that I 50 could be utilized as a proxy for the more time-intensive measurements of A max , allowing greater numbers of plants to be screened more rapidly for variation in photosynthetic capacity (McAusland et al., 2019).
Decreasing the relaxation time could be advantageous for leaves constantly under rapid high light fluctuations, allowing efficient induction of photosynthesis to maximally utilize available PPFD (Murchie & Niyogi, 2011;Murchie & Ruban, 2019). A much greater range of R 50 values was observed between the accessions than with the I 50 values. Though there was no correlation between A max and R 50 , this does not mean that the rate of relaxation is not important in C acquisition. Instead, it has been shown to be important during fluctuations in PPFD, and, if reduced, it could increase C fixation 7-30% during a diurnal time course (Long et al., 1994;Werner et al., 2001;Zhu et al., 2004;Kromdijk et al., 2016). In general, the modern cultivars took the shortest time to relax NPQ, but there were examples of specific wild relative accessions that relaxed more rapidly (Fig.  5b).
Stomatal behaviour is another temporal trait that has been the focus of recent work on optimizing the balance between C gained and water lost . I 50 was found to positively correlate with the magnitude of stomatal opening g s1000 and CO 2 assimilation rates achieved under 1000 µmol m −2 s −1 PPFD (A 1000 ), highlighting that greater opening not only allows higher A 1000 but also facilitates more rapid induction of NPQ under high light. These data suggest that, under high light, the plant mediates a fine balance between maintaining high rates of carbon fixation, with subsequent induction of photoprotection, and minimizing loss of water. Interestingly, there was no correlation between g s1000 and A 1000 , suggesting that the variation observed in g s1000 could be manipulated to reduce water loss without restricting C gainexemplified by the two-fold difference in W i values between the accessions measured (Fig. 7b).
Though large variation in the rate of stomatal opening has been observed between different species (McAusland et al., 2016) and between cultivars of a single species (Faralli et al., 2019a), the total time taken to achieve steady-state g s (k) has been shown not to correlate with maximum opening under high light (Vialet-New Phytologist (2020) 228: 1767-1780 Ó2020 The Authors New Phytologist Ó2020 New Phytologist Trust www.newphytologist.com

Research
New Phytologist Chabrand et al., 2013;McAusland et al., 2016). Maximum rate of opening Sl max is mathematically dependent on the magnitude of change in g s (k) and maximum opening, and therefore it is not unsurprising that those species that achieved greater g s1000 also achieved the highest Sl max . However, the variation in maximum opening is somewhat determined by the stomatal density : size relationship (Franks & Beerling, 2009). These data point to anatomical variability between the wild relative accessions; and though densities in modern cultivars are known to have increased through breeding (Fischer et al., 1998), there may still be an optimal density for minimizing water loss evidenced by a wild relative genus or accession. Maintaining high water-use efficiency is a particularly relevant target for modern cultivars, which typically exhibit a higher leaf water content than wild relative species (Fig.  S7b). In addition, efficient water management will directly contribute to improved tolerance to drought and heat stress, therefore maintaining yields under increasing unpredictable climactic conditions (Bertolino et al., 2019).
As with stomatal density, internal leaf anatomy will also contribute to the photosynthetic variation discovered in this study; with traits such as airspace volume (Lehmeier et al., 2017), mesophyll size (Austin et al., 1982), porosity and conductance (Lundgren et al., 2019), and distance to veins (Brodribb et al., 2007) playing a vital role in efficient C acquisition. Though no anatomical measurements are presented in this study, the significant variation in SLA (Fig. S7a) and absorbance (Fig. S8) suggests that extensive variation occurs at the cellular level.
It is important that both static and temporal responses are measured so that the capacity, dynamic adaptability, and the interrelated nature of the processes that support and maintain high rates of C acquisition are identified (Murchie et al., 2018;Salter et al., 2019). Methodologies such as the Chl fluorescence screen described here (McAusland et al., 2019), and other powerful platforms (e.g. , offer rapid, high detail measurements that can be taken during the lifetime of the plant rather than at single phenological stages. Here, we cultivated plants in glasshouse conditions but note the importance of further studies in variable field-like conditions (Poorter et al., 2016). However, A max values for 'Paragon' are similar to those observed by Driever et al. (2014), a field-based study. Complex emerging data sets may require development of functional models to predict optimal combinations in realistic field environments (e.g. Zhu et al., 2012;Wu et al., 2019).
Though only briefly mentioned here, the wild relatives of modern wheat also demonstrate a wealth of variation in agronomically important traits, such as disease (Table S1) and biotic resistance (Peleg et al., 2005), phenology (e.g. flowering time, perennial or annual), pollen fecundity, and root physiology (de Dorlodot et al., 2007;Atkinson, 2016). Quantifying and understanding the breadth of diversity available, tempered by the ease of introducing the material into modern wheat backgrounds, will not only enable efficient strategic crosses in breeding programmes designed to improve photosynthetic C gain but also provide traits to screen for in the subsequent backcross generations. In turn, this will speed up the identification of direct and related phenotypes associated with increased C assimilation per unit area, leading to greater biomass and improved yield.

Concluding remarks
The wild relatives of crop species represent a way of targeted introduction of beneficial traits to improve yield. Until now, photosynthetic variation in the wild relatives of wheat has not been explored. Here, we analyse relevant features of static and dynamic traits across a broad range of wild accessions and genera. The widest variation was found across individual accessions, suggesting local adaptation to be important when selecting crosses and that genotyping of individual accessions is important. We find key differences between wild relatives and modern lines for dynamic traits, notably that photoprotection relaxation is fast in modern lines but induction is slow. These results highlight fundamental variation in wild species and those that may have indirectly been selected for in breeding. The phenotypic and genotypic data presented are a first step to inform follow-on gene discovery and pre-breeding programmes.

Fig. S1
Light saturated rates of CO 2 assimilation, carboxylation and electron transport in accessions from four wild relative genera and modern wheat cultivars.        Table S1 A summary of the wild relatives and traits successfully identified and/or introduced into modern wheat cultivar populations.

Table S2
Parameters from A vs C i response curves including light-and CO 2 saturated photosynthetic rate (A max ), maximum carboxylation (V cmax ) and maximum electron transport (J max ) for 88 accessions across 5 genera.