Wide variation in the suboptimal distribution of photosynthetic capacity in relation to light across genotypes of wheat

Abstract Suboptimal distribution of photosynthetic capacity in relation to light among leaves reduces potential whole-canopy photosynthesis. We quantified the degree of suboptimality in 160 genotypes of wheat by directly measuring photosynthetic capacity and daily irradiance in flag and penultimate leaves. Capacity per unit daily irradiance was systematically lower in flag than penultimate leaves in most genotypes, but the ratio (γ) of capacity per unit irradiance between flag and penultimate leaves varied widely across genotypes, from less than 0.5 to over 1.2. Variation in γ was most strongly associated with differences in photosynthetic capacity in penultimate leaves, rather than with flag leaf photosynthesis or canopy light penetration. Preliminary genome-wide association analysis identified nine strong marker-trait associations with this trait, which should be validated in future work in other environments and/or materials. Our modelling suggests canopy photosynthesis could be increased by up to 5 % under sunny conditions by harnessing this variation through selective breeding for increased γ.

. Field plot layout. Figure S2. Zadoks stages. Appendix S1. Explanation of simulation methods. Figure S3. Sample photosynthesis simulations. Figure S4. Sample irradiance simulations. Figure S5. Sample timecourses of meteorological conditions. Figure S6. Flag vs penultimate leaf photosynthetic capacity. Figure S7. LAI, canopy transmittance and effective canopy extinction coefficient. Figure S8.  vs kcanopy. Figure S9.  vs Zadoks score.     Appendix S1. Simulations to estimate optimal redistribution of photosynthetic N between penultimate and flag leaves A. Overview We simulated daily photosynthesis in both penultimate and flag leaves of each genotype, using the observed values of photosynthetic capacity (Am) in both leaf layers, and again, using values of Am adjusted so as to maximize the total photosynthesis in both layers combined, but while holding total N constant.

B. Photosynthesis model
We used the photosynthesis model of Farquhar et al (1980) to simulate photosynthesis separately for each leaf layer (flag and penultimate) in each genotype, and for sunlit and shaded fractions of each layer, and summed the sunlit and shaded values using expressions given by de Pury and Farquhar (1997) to give total daily photosynthesis for each layer. Net CO2 assimilation rate (A, mol m -2 s -1 ) is calculated as the lesser of two values, one (Av) limited by RuBP carboxylation and another (Aj) limited by RuBP regeneration: (S1) = min , where Vm (mol m -2 s -1 ) is carboxylation capacity (maximum RuBP carboxylation velocity), ci (mol mol -1 ) is the intercellular CO2 concentration,  * (mol mol -1 ) is photorespiratory CO2 compensation point, K' (mol mol -1 ) is the effective Michaelis constant for RuBP carboxylation, Rd (mol m -2 s -1 ) is the rate of non-photorespiratory CO2 release that continues in the light, and J (mol m -2 s -1 ) is the potential electron transport rate. We assumed that ci = 280 ppm (~70% of ambient; Wong et al., 1979), and calculated  * , K' and Rd from expressions given by de Pury and Farquhar (1997), based on leaf temperature (estimated as described in section F below). We calculated J as the lesser root of the expression where j is a dimensionless curvature factor (assumed to be 0.7), i (mol m -2 s -1 ) is incident PPFD, and  is the initial slope of J vs i (assumed to be 0.3 mol mol -1 ).

C. Calculating photosynthetic parameters that depend on nitrogen (Vm, Jm and Rd)
We assumed that total photosynthetic N for each layer was the sum of components that determine and are each proportional to carboxylation and electron transport capacities (Vm and Jm, respectively), that Jm = 2.1 Vm (Wullschleger, 1993;Medlyn et al., 2002), that Am = 0.25J -Rd (the limit of Eqn S3 at high ci), and that Rd = 0.0089Vm (de Pury and Farquhar, 1997 Eqn S6 leads to a quadratic expression for J, which can be solved for J and the result applied to S5 to give Jm. The assumptions outlined above imply that if total N between both leaf layers is conserved, then the sum of Jm in both layers is also conserved. Thus, for each genotype, we calculated initial Jm in each layer from Am using Eqns S5 and S6 as described above, and then calculated their sum (Jm,total). To maximize total photosynthesis for both leaves combined while ensuring total N for both layers was conserved, we numerically adjusted "final" (post-redistribution) flag leaf Jm (Jmf') and calculated final Jm for the penultimate leaf (Jm2') as Jm2' = Jm,total -Jmf'. Finally, we calculated Vm and Rd in each layer for both pre-and post-redistribution conditions as Jm/2.1 and 0.0089Vm, respectively. Jm, Vm and Rd were corrected for temperature as described by de Pury and Farquhar (de Pury and Farquhar, 1997).
Simulated diurnal timecourses of net CO2 assimilation rate are shown in Figure S2.

D. Simulating incident PPFD (i)
In each leaf layer, we computed the daily timecourse of PPFD in both sunlit and shaded fractions separately, using expressions given by de Pury and Farquhar (1997), with a timestep of 0.2 hours, from sunrise to sunset (~6:36 am -5:24 pm; 55 total timesteps). Those expressions take as an input the cumulative leaf area index (L, m 2 m -2 ) above the layer in question. To calculate L for the flag leaf layer, we numerically adjusted L until the simulated total daily irradiance was 0.841 times the abovecanopy value (0.841 was the mean ratio of daily irradiance above the flag leaf to that above the canopy, measured in the field). To calculate L for the penultimate leaf layer, we adjusted L until the ratio of simulated daily irradiance between the penultimate and flag leaf layers was equal to the ratio measured in the field by ceptometry for the genotype in question (that ratio was denoted as id2/idf in the main text).
The canopy light model also required a value for atmospheric transmissivity (a, dimensionless), which influences the distribution of incoming light between beam (direct) and scattered (diffuse) components. We used expressions of Roderick (1999), which predict the diffuse fraction of total irradiance (fd, dimensionless) as a function of the ratio of sunshine hours to total daytime hours, or equivalently, the sunshine probability (psun, dimensionless), given the latitude. Those expressions predict that for our study site (Narrabri, NSW, -30.283 o latitude), fd = 0.1513 for psun = 1.0 (sunny conditions), and fd = 0.960 for psun = 0.0 (cloudy conditions). We adjusted a iteratively to make simulated daily diffuse irradiance as a fraction of total daily irradiance (above the canopy) match these values of fd, giving a = 0.076 (cloudy) and 0.830 (sunny).
Finally, the canopy light model also required input values for day of the year (254 = 11 September 2017, the middle of our measurement campaign), a longitudinal correction for solar noon (-0.023 hours at Narrabri), and latitude (-30.283 o ). Sample timecourses for incident irradiance are shown in Figure S3.

E. Simulating other environmental parameters
We simulated diurnal trends for other climatic conditions using historical climate records for a weather station near our study site, available from the Australian Bureau of Meteorology (Narrabri West Post Office, site #053030; data available at bom.gov.au). We based our simulations on 9 am and 3 pm mean values for air temperature in early September (Ta: 14.56 and 21.7 o C, respectively), air relative humidity (h: 67 and 39.75%, respectively) and wind speed (vw: 15.5 and 19.4 km h -1 , respectively), and also the mean daily minimum and maximum air temperatures (Tamin = 6.85 o C and Tamax = 22.5 o C, respectively). We assumed that Ta varied sinusoidally during the day, according to the function Ta = Tamin + (Tamax -Tamin)0.5(1 -cos((t -to)/ts)), where  = 3.1415..., t is time of day (hours) and to and ts are empirical parameters (both hours); and we adjusted to and ts iteratively to match the reported mean 9 am and 3 pm values (giving to = 4.417 and ts = 9.241 hours). We assumed that the absolute vapor pressure (pwa, Pa) varied linearly with time (pwa = pwao + mwat, where pwao and mwa are empirical parameters), and adjusted pwao and mwa to match the reported mean values of relative humidity (h(t) = 100pwa(t)/psat(Ta(t)), where psat is the saturation vapor pressure, given by psat(T) = 611.2exp(17.62T/(243.12 + T)), with T in o C and psat in Pa) at 9 am and 3 pm (giving pwao = 1229 Pa and mwa = -13.33 Pa h -1 ). We assumed that the wind speed varied linearly with time (vw = vwo + mvwt, where vwo and mvw are empirical parameers), and adjusted vwo and mvw to match the reported mean values at 9 am and 3 pm (giving vwo = 9.65 km h -1 and mvw = 0.65 km h -2 ). From pwa and Ta we calculated the vapor pressure deficit of the air, Dair(t), as psat(Ta(t)) -pwa(t).
Simulated timecourses of Ta, VPD and vw are shown in Figure S4.

F. Simulating leaf temperature
We simulated leaf temperature in each leaf layer by energy balance: where Q (J m -2 s -1 ) is absorbed shortwave radiation (0.3173*i; this accounts for absorption of both visible and near-infrared radiation; Buckley et al., 2014), fir (unitless) is the fraction of possible infrared exchange that occurs between the leaf in question and the atmosphere, a (unitless) is the atmospheric IR emissivity, L is the leaf IR emissivity (0.98), TaK is air temperature in kelvins, TLK is leaf temperature in kelvins, gbh (mol m -2 s -1 ) is the (2-sided) leaf boundary layer conductance, cpa is the molar heat capacity of the air (29.3 J mol -1 K -1 ), TL is leaf temperature in o C,  is the latent heat of vaporization (4.410 4 J mol -1 ), gtw (mol m -2 s -1 ) is total leaf conductance to water vapor, Dleaf (Pa) is the vapor pressure difference between the leaf intercellular airspaces and the air, and patm (Pa) is atmospheric pressure (9.8910 4 Pa in Narrabri [212 m elevation]). Eqn S7 ignores metabolic energy storage, assumes zero net IR exchange among leaves in the canopy, and assumes stomatal conductances are similar at both leaf surfaces. We assumed that the canopy attenuates IR with an extinction coefficient equal to that for diffuse shortwave radiation (kd' = 0.719; de Pury and Farquhar, 1997), so that fir = exp(-kd'L). a is a function of air temperature and vapor pressure (a = 0.642(pwa/TaK) 1/7 ) (Leuning et al., 1995). We modeled gbh as a function of leaf width (assumed to be 2 cm) and wind speed (gbh = 20.123(vw(L)/0.02 m) 1/2 for vw in m s -1 ; ref), and assumed that wind speed attenuated down through the canopy with an extinction coefficient of 0.5, such that vw(L) = vw(0)exp(-0.5L), where vw(0) is the value above the canopy, calculated as described in section E. We calculated gtw as (rsw + rbw) -1 (where rsw and rbw are stomatal and boundary layer resistances to water vapor, respectively), and computed rsw as (1/1.6)(rtc -rbc) (where rtc and rbc are total and boundary layer resistances for CO2, respectively) and rbw as rbc/1.37. rtc = A/(400 -ci) and rbc = 1.27/gbh. Together this gives gtw = 1.6/(A/(400 -ci) + 0.213/gbh).
Dleaf depends on leaf temperature, but can be estimated from Dair by linear extrapolation: Dleaf  Dair + s(TL -Ta), where s = psat/T = 17.62243.12/(243.12+T) 2 . Similarly TLK 4 can be approximated by expansion of (TaK + [TL -Ta]) 4  TaK 4 + 4TaK 3 (TL -Ta). Applying these approximations to Eqn S7 and rearranging leads to the following expression for TL: Since the estimate of TL given by Eqn S8 depends on TL itself via A (from which gtw is estimated), we used an iterative procedure: first using a leaf temperature of TL,1 = 25 o C to calculate A1, gtw,1 from A1, and TL,2 from gtw,2; then calculating A2 from TL,2, gtw,2 from A2, and TL,3 from gtw,2; and finally recalculating a final value of A from TL,3.   Figure S12. The difference in carbon isotope discrimination ( 13 C) between flag and penultimate leaves was larger in magnitude in genotypes with high , but the relationship was very weak (y = 0.584x + 0.125, r 2 = 0.033, p < 0.05). n = 160 genotypes.