Can exploiting natural genetic variation in leaf photosynthesis contribute to increasing rice productivity? A simulation analysis



    1. Department of Plant Sciences, Centre for Crop Systems Analysis, Wageningen University, Wageningen, The Netherlands
    Current affiliation:
    1. Key Laboratory of Crop Genetics and Physiology of Jiangsu Province, Yangzhou University, Yangzhou, Jiangsu, China
    Search for more papers by this author

    Corresponding author
    1. Department of Plant Sciences, Centre for Crop Systems Analysis, Wageningen University, Wageningen, The Netherlands
    Search for more papers by this author

    1. Department of Plant Sciences, Centre for Crop Systems Analysis, Wageningen University, Wageningen, The Netherlands
    Search for more papers by this author

    1. Department of Plant Sciences, Centre for Crop Systems Analysis, Wageningen University, Wageningen, The Netherlands
    Search for more papers by this author


Rice productivity can be limited by available photosynthetic assimilates from leaves. However, the lack of significant correlation between crop yield and leaf photosynthetic rate (A) is noted frequently. Engineering for improved leaf photosynthesis has been argued to yield little increase in crop productivity because of complicated constraints and feedback mechanisms when moving up from leaf to crop level. Here we examined the extent to which natural genetic variation in A can contribute to increasing rice productivity. Using the mechanistic model GECROS, we analysed the impact of genetic variation in A on crop biomass production, based on the quantitative trait loci for various photosynthetic components within a rice introgression line population. We showed that genetic variation in A of 25% can be scaled up equally to crop level, resulting in an increase in biomass of 22–29% across different locations and years. This was probably because the genetic variation in A resulted not only from Rubisco (ribulose 1,5-bisphosphate carboxylase/oxygenase)-limited photosynthesis but also from electron transport-limited photosynthesis; as a result, photosynthetic rates could be improved for both light-saturated and light-limited leaves in the canopy. Rice productivity could be significantly improved by mining the natural variation in existing germ-plasm, especially the variation in parameters determining light-limited photosynthesis.


Cereal yield is determined by the accumulated photosynthetic assimilates over the entire growing season that are partitioned into the caryopses. Improvements in crop management and genetic gain in harvest index are largely responsible for the increased cereal yields over the last decades (Austin 1999; Peng et al. 2008). However, it has been argued that cereal production is now approaching a plateau and further increases in yield will necessitate an increase in photosynthesis (Austin 1994; Mitchell & Sheehy 2006; Lawson, Kramer & Raines 2012).

Crop photosynthesis accumulated for the entire growing season depends on the ability of the crop to build up and maintain a canopy for capturing incoming light, and also on the photosynthetic capacity and efficiency of leaves. There may be chances to increase the light capture by improving early leaf area growth rate or by introducing ‘stay green’ genes to extend the growing season (Long et al. 2006). For rice, however, leaf area dynamics and canopy architecture may have been effectively optimized for maximum light capture through breeding (Horton 2000). Any further increase in photosynthesis of the rice crop may largely have to come from improved leaf photosynthesis, although there may still be scope to improve canopy architecture. Photosynthesis per unit leaf area seems to have been improved already as suggested by experimental comparisons of old and modern varieties of cereals, including rice, in concert with improvements in harvest index and grain number (Fischer & Edmeades 2010).

Given that crop genetic transformation is becoming increasingly routine, opportunities for improving leaf-level photosynthesis via genetic engineering have been extensively explored, by either experimental approaches or theoretical computation. Approaches include, for example, designing more efficient ribulose 1,5-bisphosphate carboxylase/oxygenase (Rubisco; Mueller-Cajar & Whitney 2008; Whitney & Sharwood 2008); exploiting existing interspecific variation in Rubisco efficiency (Zhu, Portis & Long 2004a); increasing RuBP regeneration and light reaction (Miyagawa, Tamoi & Shigeoka 2001; Peterhansel, Niessen & Kebeish 2008; Rott et al. 2011); increasing mesophyll conductance (Uehlein et al. 2008); introducing CO2-concentrating mechanism into C3 crops (Price et al. 2008); introducing CO2-concentrating mechanism with Kranz anatomy into C3 crops (von Caemmerer, Quick & Furbank 2012); short-circuiting photorespiration (Maurino & Peterhansel 2010); and increasing the rate of transition from photoprotection (Zhu et al. 2004b). Long et al. (2006) estimated that these ambitious approaches, if successful, would need research efforts of 10–30 years, depending on the avenues to be used.

Leaf photosynthesis could be improved not only through transgenic biotechnology, but also through the exploitation of natural variation with a conventional breeding approach. Parry et al. (2011) indicated that mining existing genetic variation could be the most efficient method for short-term improvements (<5 years). Recently, quantitative trait loci (QTLs) related to different photosynthetic parameters have been successfully mapped (Takai et al. 2009; Adachi et al. 2011; Gu et al. 2012a). Furthermore, Gu et al. (2012b), using the biochemical photosynthesis model of Farquhar, von Caemmerer & Berry (1980) as adapted by Yin et al. (2009), successfully dissected genetic variation of leaf photosynthesis present in an introgression line (IL) population into different biophysical and biochemical components. Their analysis showed that by using genetic variation in all components, leaf-level photosynthesis could potentially be increased by ca. 20% through marker-assisted selection.

However, photosynthesis rate per unit area of leaf does not correlate well with crop yield (Evans & Dunstone 1970; Teng et al. 2004). This has led to a common notion that increasing leaf photosynthesis is not a useful strategy to increase crop yield (Richards 2000; Zhao et al. 2008). Actually, this notion was confirmed by our own work on the IL population: among the many physiological parameters examined, leaf photosynthesis was not important in determining the differences in crop yield among the ILs observed in a field experiment, either under drought or under well-watered conditions (Gu 2013). This lack of persistence of variation across scales is probably due to the complex hierarchy from leaf-level photosynthesis to crop yield and to interaction and feedback mechanisms occurring between physiological components within the individual plant, between plants of the same crop and between the crop and the environment. Moreover, relationships between leaf photosynthesis and crop yields may depend on genetic background of plant materials. These complexities might mask the potential contribution of the small within-population variation in leaf photosynthesis to the variation in final crop yield. Therefore, modelling has been a useful tool to investigate the potential of improved photosynthesis on crop productivity (Day & Chalabi 1988; Long et al. 2006).

In this paper, we used the process-based crop model GECROS (Yin & van Laar 2005) to examine the extent to which exploiting the natural genetic variation in leaf photosynthesis components can contribute to variation in canopy photosynthesis and in crop productivity in rice. The GECROS model combines sufficient physiological rigour for complex phenotypic responses with genotype-specific parameters. We used this model to scale up variation in leaf photosynthesis components as detected in our previous study (Gu et al. 2012b) to variation in canopy photosynthesis and in biomass productivity across the entire growing season for contrasting environments. Input parameter values for model simulation are only those derived from our previous results on QTLs for various leaf photosynthesis parameters, while other input parameters of the GECROS model are maintained the same across rice genotypes. In this way, potential confounding effects due to variation in other physiological processes can be avoided to exclusively illustrate the potential impact of natural genetic variation in leaf photosynthesis on crop productivity. We specifically hypothesize for potential larger persistence during scaling up than observed in previous studies from the literature if photosynthesis can be improved irrespective of light level, and test this hypothesis using the IL population segregating for QTLs related to both light-saturated and light-limited photosynthesis parameters.

Materials and Methods

Based on the genetic variation found in an IL population of rice (Gu et al. 2012a,2012b), the crop model GECROS (Yin & van Laar 2005) was used to evaluate the expression of genetic variation in leaf photosynthesis in terms of variation of canopy photosynthesis and crop biomass production.

Crop growth model GECROS

GECROS is a generic model that operates in daily time steps, simulates the growth and development of the crop over time and generates phenotypes for a multitude of traits, based on concepts of the balance, interaction and feedback mechanisms among various contrasting components of crop growth. A detailed description of GECROS and its algorithms can be found in Yin & van Laar (2005). In this section, only key features, related to carbon partitioning, nitrogen (N) demand and partitioning, phenological development and photosynthesis are described; the corresponding algorithms are given in Appendix 1.

Nitrogen demand is the maximum of the deficiency-driven and the growth activity-driven demand (Eqns A1–6). The deficiency-driven demand is the amount of nitrogen required to restore the critical minimum nitrogen concentration. The growth activity-driven demand is based on the optimum nitrogen/carbon ratio for the maximum relative growth rate.

Root:shoot partitioning for nitrogen and carbon responds to environmental factors, based on the root:shoot functional balance theory (Charles-Edwards 1976). The intra-shoot nitrogen partitioning is based on a pre-defined maximum grain nitrogen concentration of a genotype and a minimum nitrogen concentration in the stems. If the nitrogen requirements for the grains and stems are met from the current nitrogen uptake, the remaining shoot nitrogen goes to the leaves, which include the photosynthetically active parts of the stems, sheaths and ears. If the requirements for the grains are not met, remobilization of nitrogen first from the reserves and then from the leaves and the roots takes place, until the reserves are depleted and the nitrogen concentrations in the leaves and roots reach their minimum values. This remobilization stimulates leaf and root senescence. If the grain nitrogen requirements are not met by shoot nitrogen and remobilization, the grain nitrogen concentration declines. Intra-shoot carbon partitioning to the stems (including sheaths) and to the grains is determined according to their expected daily carbon demands, which are described by the differential form, Eqn A7, of a sigmoid function for asymmetric determinate growth (Yin et al. 2003). The remaining shoot-carbon goes either to the leaves, or to the carbon reserve pool in the stems when the green-surface area index (GAI) becomes nitrogen limited. The GAI is calculated according to the principles described by Yin et al. (2000), as either the carbon- or the nitrogen-limited GAI. The carbon reserves, if any, become available to the grains, when current photosynthesis does not satisfy the carbon demand by grains.

In GECROS, phenological development is calculated by Eqns A8–10. Development stages (DS) are defined as 0 at seedling emergence, 1 at start of grain filling and 2 at physiological grain maturity. The intervals from stage 0 to 1 and from 1 to 2 depend on the genotype-specific number of days at optimum temperature. A flexible bell-shaped non-linear function (Yin et al. 1995) is used to describe the temperature response of development rate, which has a value of zero when the hourly temperature is below the base temperature or above the ceiling temperature and one when it is equal to the optimum temperature. In case of photoperiod-sensitive genotypes, development rate is also affected by daylength during the photoperiod sensitive phase of the vegetative interval.

Canopy photosynthesis sub-model of GECROS

To compute canopy photosynthesis as driver of crop growth, GECROS uses the two-leaf approach that divides the canopy into sunlit and shaded fractions, based on solar height; each fraction is modelled separately with a single-layer leaf model (Eqns A11–16; de Pury & Farquhar 1997; Wang & Leuning 1998). For both canopy fractions, the photosynthetically active nitrogen is calculated using a base value of leaf nitrogen (below which photosynthesis is zero) and a leaf nitrogen extinction coefficient to describe an exponential profile in the canopy for vertical decline in nitrogen (Yin et al. 2000). For example, photosynthetically active nitrogen for the entire canopy (Nc), for the sunlit leaf fraction of the canopy (Nc,su) and for the shaded leaf fraction of the canopy (Nc,sh), can be estimated by Eqns A17–19. To estimate the photosynthesis parameters for the entire canopy, we introduced the nitrogen dependency through a linear function (Harley et al. 1992), Eqn A20, in which the base value for leaf photosynthesis nb was assumed to be the same across all genotypes and was obtained from the data of Gu et al. (2012b). The photosynthetic rate of each canopy fraction was then computed using a leaf model as described below. The instantaneous rates of canopy photosynthesis were extended to daily total, using the numerical Gaussian integration method. The canopy photosynthesis model was also decoupled from GECROS in order to simulate the variation among the ILs in canopy photosynthesis alone, without the crop growth feedback loops.

Leaf photosynthesis sub-model

In GECROS, prediction of the rate of photosynthesis at leaf level is based on the models of Farquhar et al. (1980) as modified by Yin et al. (2009; Eqns A21–24). A phenomenological model of the Leuning type (Leuning 1995), Eqn A25, was introduced (Yin & Struik 2009) for quantifying stomatal conductance, gs. A similar equation, Eqn A26, was used to describe mesophyll conductance gm (Yin et al. 2009). Parameters δm and δs in Eqns A25 and A26 can be used to estimate the gm:gs ratio (Gu et al. 2012b). Eqns A27 and A28 were used to predict the effects of vapour pressure difference on the conductances. Leaf temperature, which affects the rates of most biochemical reactions of photosynthesis (Eqns A29–30), is also predicted in GECROS by coupling the gs and leaf photosynthesis models (Yin & Struik 2009) with the Penman–Monteith equation. The kinetic constants (Kmc and Kmo) of Rubisco required for leaf photosynthesis were taken from Bernacchi et al. (2002), and Rubisco specificity (Sc/o) was set at 3.02 mbar μbar−1 (Gu et al. 2012b), and both were assumed to be conservative across all the ILs. In GECROS, day respiration (Rd) was assumed to be scaled with the maximum Rubisco activity (Vcmax). Genetic variation in leaf photosynthesis parameters like Vcmax was based on results from Gu et al. (2012b; also see below).

Genetic input parameters for leaf photosynthesis of ILs

The above leaf model contains six photosynthesis parameters κ2LL, Jmax, θ, δm, δs and Vcmax. Genome regions or QTLs were previously assigned for these parameters, based on 11 representative lines of the IL population (Gu et al. 2012b; also see Supporting Information Fig. S1). The additive effects of the QTLs estimated therein were used to estimate genotype-specific values of the six parameters for individual lines of the IL population, based on the nearest-marker allelic information for these ILs. So, a parameter value X of IL k, containing N QTLs, was presented as

display math

where μ = the intercept; an = the additive effect of the nth QTL; Mk,n = genetic score of the n-th QTL of the individual IL k that takes either the value −1 for the allele coming from recurrent parent ‘Shennong265’ or 1 for that from donor parent ‘Haogelao’. All the calculations were based on the results from Gu et al. (2012b) on the estimates of the additive effect an. For parameters Vcmax and Jmax that depend on leaf nitrogen content as well (see Eqn A20 in Appendix), the effects of leaf nitrogen and genotype were assumed to be independent. Note that while the population contains 96 ILs, only 38 genotypes were identified based on the allelic information of all QTLs for the six photosynthesis parameters.

Other input parameter values and model simulation

All other parameter values were calibrated for rice (Gu 2013) and were used here across all individual ILs. Simulation was performed for two growth seasons in 2008 (seedling emergence on 13 May) and in 2009 (14 May) at Shangzhuang Experimental Station (39°54′N, 116°24′E; elevation of 50 m above sea level) of China Agricultural University, in Beijing, China (coded as BJ08 and BJ09, respectively). To exhibit responses of the same genotypes to different climate conditions, simulations were also carried out for the dry season (from 10 Jan) in Los Baños (14°11′N, 121°15′E; elevation of 21 m above sea level), International Rice Research Institute, Philippines from 2001 to 2005 (PH01, PH02, … , PH05, respectively). The time course of main weather variables in these environments are shown in Supporting Information Fig. S2.

Data analysis

In order to quantitatively evaluate the genetic variation in the IL population, genetic variation was calculated as math formula (%) where Xmax, Xmin and math formula stand for maximum, minimum and the mean value, respectively, for a given model-input parameter or model-output trait, of all the lines in the population. Note that although the simulations were conducted only for the 38 genotypes, the population mean of the simulated output traits was still calculated on the basis of 96 ILs, that is, the weighted mean given the number of IL repeats in each genotype. For each individual IL, the genetic gain or loss was calculated as math formula (%), where Xi stands for the input parameter or output trait value for the ith IL. The correlations and multiple analyses were calculated by PROC CORR, PROC GLM, respectively, in SAS 9.2 (SAS Institute Inc., Cary, NC, USA).

Results and Discussion

Genetic variation in leaf photosynthesis

For this paper, we were only interested in the genetic variation in leaf photosynthesis, so the 96 ILs (including the two parents) were divided into 38 unique genotypes, based on origin of the alleles at the seven loci for leaf photosynthesis parameters reported by Gu et al. (2012b). Associated with these QTLs, all the photosynthesis parameters varied among the ILs. The genetic variation, as defined in Table 1, in photosynthetic efficiency under limiting light (κ2LL, θ), diffusional conductance (δm, δs), maximal rate of electron transport (Jmax) and maximum rate of Rubisco activity (Vcmax), ranged from 20.3 to 49.5% (Table 1).

Table 1. Minimum, maximum and population mean of photosynthetic parameters and of various traits at leaf level, canopy level and crop level, and their genetic variation in the introgression line population. Genetic variation was calculated as math formula, where Xmax and Xmin stands for maximum and minimum value in the population, respectively; math formula stands for the population mean
 TraitMinMaxPopulation meanGenetic variation (%)
  1. κ2LL, value of conversion efficiency of incident light into e transport at the strictly limiting light; Jmax, maximum value of e transport under saturated light; θ, convexity factor for response of e transport to irradiance; δm, δs and δt, parameters related to chloroplast /intercellular [CO2] ratio, intercellular /ambient [CO2] ratio, and chloroplast /ambient [CO2] ratio at saturating light, respectively; Vcmax, maximum rate of Rubisco activity-limited carboxylation; A100, A500 and A2000, leaf photosynthesis at low light (100 μmol m−2 s−1), intermediate light (500 μmol m−2 s−1) and saturated light (2000 μmol m−2 s−1), respectively; Ac,1 and Ac,5, daily gross rate of canopy photosynthesis at high light level for green-surface area index (GAI) = 1 and GAI = 5, respectively; GAI0.25, GAI1.0 and GAI2.0, value of GAI at development stage 0.25 (seedling), 1.0 (flowering), 2.0 (harvest), respectively; Bio0.5, Bio1.0, Bio1.5 and Bio2.0, total biomass at development stage 0.5 (tillering), 1.0 (flowering), 1.5 (grain filling) and 2.0 (maturity) for Beijing in 2009, respectively; BioBJ08, BioPH01, BioPH02, BioPH03, BioPH04 and BioPH05, total harvest biomass at location Beijing year 2008, and location Philippines year 2001 to 2005, respectively.
Leaf levelA1002.94.13.631.4
Canopy levelAc,125.533.228.926.4
Crop levelGAI0.251.862.091.9312.0

Light response curves of leaf photosynthesis for the IL population were constructed based on detected QTLs for each of these six parameters (Fig. 1). The genetic variation in calculated leaf-level photosynthesis at various light intensities was considerable. At low light level (100 μmol m−2 s−1), genetic variation amounted to 31.4%, whereas it was 18.7% at intermediate level (500 μmol m−2 s−1), and 25.4% at saturated level (2000 μmol m−2 s−1; Table 1; Fig. 1).

Figure 1.

The modelled light response curves of net leaf photosynthesis rate (A) to light intensity at ambient CO2 concentration (380 μmol mol−1) and a leaf temperature of 25 °C in a population of 38 introgression lines.

Estimated genetic variation in daily canopy photosynthesis

Canopy photosynthesis (Ac) was spatially integrated from base to the top of the canopy. Some parameters (e.g. Vcmax, Jmax, gs and gm) were adjusted to change with depth in the canopy, based on the modelled profile of leaf nitrogen content in the canopy. Spatially integrated canopy photosynthesis must consider the heterogeneous radiation in canopies and the non-linear response of photosynthesis to irradiance. At the same time, the heterogeneous light condition within the canopy is affected by solar angles, incident photon flux during the day and different GAI values. All these complexities shed doubt whether relations between photosynthetic parameters and leaf photosynthesis apply when scaling up to canopy level photosynthesis. Using the canopy model decoupled from GECROS, we simulated Ac with various GAIs and under different environmental conditions.

As shown by Figs 2 and 3, leaf photosynthesis at saturating light (A2000) correlated well with daily canopy carbon gain at high light intensity when GAI was 1 or 5 (i.e. Ac,1 and Ac,5). Moreover, component parameters of leaf photosynthesis show similar correlations with A2000, Ac,1 and Ac,5. In Fig. 2, we found more or less the same genetic gain or loss at leaf (A2000) and canopy level (Ac,1 and Ac,5) for each genotype. For A2000, genetic variation was 25.6%, which is comparable with 26.5% for Ac,1 and 25.8% for Ac,5 (Table 1). All these results suggest that genetic variation in leaf photosynthesis in this IL population scales up well to canopy level.

Figure 2.

Heat map of genetic gain or loss (%) of 38 genotypes in photosynthetic parameters and output traits at leaf, canopy and crop level, from left to right as separated by the vertical black lines. Each row represents a unique genotype based on alleles at detected loci (Gu et al. 2012b), each column represents a model parameter. The first column shows in total 38 different genotypes out of 96 introgression lines with number of repeats in brackets. The values increase from green via white to red. The genetic gain or loss for each introgression line (%) was calculated as math formula, where math formula stands for the parameter or trait value for ith genotype, and math formula stands for the population mean. In this figure, the vertical direction shows the genetic gain or loss for all genotypes at the leaf, canopy, and crop level; the horizontal direction shows how each line performed with regards to the photosynthetic components at leaf level, the canopy photosynthesis at low or high GAI, the GAI dynamics, the biomass dynamics, and the harvested biomass at different locations and years. The meaning of symbols and abbreviations is the same as in Table 1.

Figure 3.

Pearson correlation coefficient heat map of parameters and traits from leaf level, canopy level to crop level. Correlations are scaled by the colour of the corresponding cell. Parameters are represented in the same order on the x- and y-axes. The meaning of symbols and abbreviations is the same as in Table 1.

We also examined the light response of daily canopy photosynthesis of canopies of different sizes. For that purpose, the genotype-specific light response of daily Ac at various GAIs were calculated for the site with 39°54′N latitude (Beijing) on day 107 with a day length of 13.1 h (a typical day in the season there; Fig. 4; Table 2) assuming a uniform light distribution over the day. Under the environmental conditions that air temperature is 25 °C and vapour pressure is 1.5 kPa, the genetic variation of the daily Ac among the ILs when GAI is 5 was 30.3% at low light level (100 μmol m−2 s−1, ∼PAR = 1.04 MJ m−2 d−1), 31.0% at intermediate level (500 μmol m−2 s−1, ∼PAR = 5.18 MJ m−2 d−1) and 25.8% at high light level (2000 μmol m−2 s−1, ∼PAR = 20.7 MJ m−2 d−1; Table 2). When adjusting canopy size GAI to 3 or 1, canopy photosynthesis changed significantly (Fig. 4A). Still, the genetic variation within the IL population was relatively stable for the different levels of light intensity (range 23.4–30.5%) (Table 2). Similarly, canopy photosynthesis responded to environmental conditions, for example, to vapour pressure (Fig. 4B), and the genetic variation exhibited within the IL population was within a similar range (Table 2).

Figure 4.

Canopy photosynthetic responses to photosynthetically active radiation (PAR) under different green-surface area index (GAI; a) and vapour pressure (VP; b) for the introgression line population at constant air temperature of 25 °C. For clarity, only maximum and minimum values of the introgression line population are shown. The control is for location Beijing, the 107th day of the year 2009, with day length of 13.11 h, and assuming a uniform light distribution over the day, an ambient CO2 concentration of 380 μmol mol−1, a green-surface area index of 5 and a vapour pressure of 1.5 kPa. For each individual simulation, control parameters were used for the whole introgression line population except for the tested parameters that were varied.

Table 2. Genetic variation in daily gross canopy photosynthesis (Ac) at different combinations of green-surface area index (GAI) and vapour pressure (VP), when air temperature was at 25 °C
 PAR (MJ m−2 d−1)Ac (min) (g CO2 m−2 d−1)Ac (max) (g CO2 m−2 d−1)Population meanGenetic variation (%)
Control (GAI = 5, VP = 1.5 kPa)1.0359.1712.4410.8130.3
GAI (VP = 1.5 kPa)     
VP (GAI = 5)     
0.5 kPa1.0358.5311.5710.0230.3
3.0 kPa1.03510.6614.3612.6029.3

Genetic variation in green-surface area index and biomass production

We used GECROS to assess how the spatial and temporal integration of genetic variation in leaf photosynthesis resulted in genetic variation in GAI and total biomass (including dry weight of all living and dead shoot and root materials) in this IL population. Simulated grain yield was not analysed here because of the multiple uncertainties related to the quantification of grain numbers in the model (the model assumes that grain number can be carbon or nitrogen determined; also see Gu 2013).

Fig. 5 illustrates for Beijing, 2009, assuming a total nitrogen uptake of 15 g N m−2, that there was considerable genetic variation in both GAI and total biomass, throughout the growing season. For GAI, there was more genetic variation at flowering (DS = 1.0: 26.9%) than at an early vegetative growth stage (DS = 0.25: 12.0%), or at harvest stage (DS = 2.0: 13.0%; Table 1). The reason for the larger genetic variation at flowering could be the fully developed canopy and the large amount of nitrogen held in the canopy. At early vegetative stage, the canopy was not fully developed yet, and therefore, the genetic variation was not fully expressed. At harvest stage, leaves were senescing and most nitrogenous compounds in the leaves were decomposed and the N translocated to the grains. For biomass, the genetic variation at DS 0.5 (∼ tillering stage), 1.0 (flowering), 1.5 (mid-grain filling) and 2.0 (harvest) was 34.4, 29.3, 26.5 and 26.5%, respectively.

Figure 5.

Time courses of calculated (a) green-surface area index (GAI; m2 m−2) and (b) total biomass (g m−2) of 38 genotypes in Beijing during the year 2009.

There was a negative relationship between GAI and total biomass, especially at flowering stage (see also Fig. 3). The negative relationship could be due to the feedback mechanisms in the model. A high rate of canopy photosynthesis is associated with high Jmax and Vcmax (Fig. 3), which means more nitrogen content per leaf area, whereas in the model GAI is co-determined by available carbon assimilates and canopy nitrogen content (Yin et al. 2000). The feedback mechanism in the GECROS model – that current high photosynthesis may dilute leaf nitrogen in the subsequent days – can lead the model to predict an accelerated leaf senescence, a phenomenon generally also observed experimentally (Fangmeier et al. 2000; Ainsworth & Long 2005).

In view of the fact that environmental factors considerably influence both leaf and canopy photosynthesis, we show the dynamic pattern of GAI and biomass during an entire growing season (Beijing 2009; Fig. 5), and biomass production at different latitudes and in different years (Beijing 2008, 2009 and Los Baños 2001–2005; Table 1). The simulations carried out for both Beijing and Los Baños have a similar tendency across different growth stages growing seasons and locations, despite of variation in the climatic variables across years and locations (Supporting Information Fig. S2). In fact, genetic variation for each level mostly ranged between 20 and 30% (Table 1).

Our model simulation already showed that genetic variation in leaf photosynthesis among the 38 genotypes resulted in an average 25.9% increase in biomass production. Further improvement is still possible. Based on the seven QTL regions, there are in total 27 = 128 possible genotypes. By evaluating the 128 possible genotypes, the potential ideotype that combines all positive alleles for photosynthesis was simulated to have a biomass production advantage of 5.6, 7.4, 8.8, 7.0, 5.1, 7.6 and 8.2%, when compared with the best IL of the 38 genotypes, for BJ08, BJ09, PH01, PH02, PH03, PH04, PH05, respectively.

On contribution of leaf photosynthesis components to biomass productivity

There has been a long-standing controversy as to whether an increase in leaf-level photosynthesis would increase yield (Evans & Dunstone 1970; Borrás, Slafer & Otegui 2004; Long et al. 2006). It is commonly assumed that even when there is an improvement in leaf photosynthesis components, the effects will be diluted through biological hierarchy, resulting in only a small effect at canopy level or crop level. For example, Zhu et al. (2004a) did show that replacing the average Rubisco of terrestrial C3 plants by Rubisco from other photosynthetic organisms would increase canopy photosynthesis significantly; but given a widely observed inverse relationship between maximum catalytic rates of carboxylation per active site (kcc) and Rubisco specificity (Sc/o), they also showed that replacing the average C3 Rubisco with one having an optimal Sc/o would only increase canopy photosynthesis by ca. 3%. Sinclair, Purcell & Sneller (2004) presented an calculation for soybean, starting with an assumed 50% increase in the production of mRNA for synthesis of Rubisco, but ending with only 6% increase or even a 6% decrease in yield depending on whether there is extra nitrogen accumulation possible or not. Yin & Struik (2008) assessed the impact of a successful introduction of the full C4 system into rice. The simulation resulted in ca. 25% yield increase, lower than the originally expected 50% increase.

These results are not surprising as photosynthesis in the canopy can be either light saturated or light limited. Light-limited photosynthesis is electron transport limited, whereas light-saturated photosynthesis is generally Rubisco limited, particularly at lower [CO2]. In Zhu et al.'s calculation, the optimal specificity of Rubisco will increase leaf photosynthesis at high light level (>400 μmol m−2 s−1), this effect was weakened by an opposite effect at low light (because of the negative relationship between Sc/o and kcc). In Sinclair's simulation, 50% more mRNA for synthesis of the subunits of Rubisco only increased light-saturated photosynthesis. If there were no additional N inputs, the required investment in Rubisco will cause less nitrogen being available for subsequent transfer to the seed and the seed becomes nitrogen limited (i.e. 6% decrease in yield). In Yin and Struik's analysis on the potential benefit from introducing the C4 system into C3 rice through the CO2-concentrating mechanism, light-saturated photosynthesis was significantly improved while the light-limited photosynthesis at leaf level remained unchanged. Furthermore, any shortage of the required nitrogen may also hamper the realization of ‘C4 rice’ yield potential. These may explain the much lower increase in yield than originally expected.

Using our simulation results, we performed multiple regression analysis to indicate which parameters are most important for determining final biomass. Multiple regression analysis showed that the leaf-level genetic variation in κ2LL contributed most to the variation in total biomass, followed by δs, θ, δm and Jmax (Table 3), which somewhat differed from the results of the simple correlation analysis (Fig. 3). As κ2LL contributes to electron transport at limiting light, this result was in line with the assumption that in a canopy, most leaves were in the state of electron transport-limited photosynthesis. This is further supported by the fact that Vcmax did not significantly contribute to the explanation of observed variance in any of the simulations (Table 3). This may explain the results of our simulation, which unlike most earlier simulation studies, incorporated the genetic variation in both light-saturated and light-limited photosynthesis parameters. Our results are in line with the report of Day & Chalabi (1988) that a same percentage increase in quantum use efficiency, relative to light-saturated photosynthetic capacity, resulted in a more significant increase of canopy photosynthesis.

Table 3. Regression coefficients (with probability of significance between brackets), intercept and total variation accounted for (R2) for multiple linear regression analysis of total biomass (Bio) as a function of κ2LL, Jmax, θ, δm, Vcmax and δs (i.e. Bio = b0 + b1κ2LL + b2Jmax + b3θ + b4δm + b5Vcmax + b6δs), based on data of an introgression line population. Regressions were made for locations Beijing, 2008 (BJ08); Beijing, 2009 (BJ09) and Los Baños, Philippines from 2001 to 2005 (PH01, PH02, … , PH05, respectively). The meaning of symbols are the same as in Table 1
TraitRegression coefficient (probability of significance)Intercept (b0)R2 (%)
  1. Coefficient values significant at a level of P < 0.05 are in bold.
BioBJ085483.1 (<1.0 × 10−8)11.3 (0.0035)604.3 (2.6 × 10−6)176.9 (0.0012)−3.8 (0.1520)224.3 (<1.0 × 10−8)−1427.699.5
BioBJ095806.6 (<1.0 × 10−8)8.2 (0.0400)774.6 (9.0 × 10−8)259.7 (2.8 × 10−5)−1.9 (0.4813)213.0 (<1.0 × 10−8)−1500.999.4
BioPH015806.2 (<1.0 × 10−8)10.5 (0.0069)864.7 (<1.0 × 10−8)219.7 (0.0001)−2.9 (0.2606)340.5 (<1.0 × 10−8)−1902.099.5
BioPH025433.8 (<1.0 × 10−8)−1.4 (0.7690)791.4 (3.8 × 10−6258.9 (0.0005)5.0 (0.1549)321.2 (<1.0 × 10−8)−692.399.0
BioPH035040.2 (<1.0 × 10−8)−0.5 (0.9356)709.0 (0.0003)274.8 (0.0023)4.2 (0.3296)298.6 (<1.0 × 10−8)−527.998.3
BioPH045572.8 (<1.0 × 10−8)7.8 (0.0597)803.5 (1.0 × 10−7)265.0 (3.7 × 10−5)−1.4 (0.6133)334.2 (<1.0 × 10−8)−1440.899.3
BioPH055820.8 (<1.0 × 10−8)7.0 (0.0414)796.0 (<1.0 × 10−8)186.1 (0.0003)−1.2 (0.6272)359.4 (<1.0 × 10−8)−1503.699.6

Concluding Remarks

In our previous research (Gu et al. 2012b), we not only found QTLs for light-saturated photosynthesis (e.g. QTLs for Vcmax, δm, δs), but also QTLs contributing to light-limited photosynthesis (e.g. QTLs for κ2LL, θ). This could also be the explanation why genetic variation in our genetic population showed to scale up equally from leaf, to canopy to crop level (Fig. 2). Our results showed a very promising approach to increase plant production through conventional marker-assisted breeding. This is in line with Parry et al. (2011), who suggested that improving photosynthesis through mining existing germ plasm is the most efficient way. There have been reports on considerable genetic variation between rice cultivars (Adachi et al. 2011; Gu et al. 2012a) to be utilized in a breeding programme. Further progress could be enhanced through recent advances in genome-wide association studies (Huang et al. 2010) by exploiting a broader genetic background.


This work was supported by the China Scholarship Council and partly by the Dutch research programme ‘BioSolar Cells’.

Appendix: Appendix 1

Main equations used in the GECROS model (Yin & van Laar 2005) to simulate nitrogen demand and carbon partitioning, phenological development and canopy and leaf photosynthesis.

Nitrogen demand and carbon partitioning sub-model

display math(A1)
display math(A2)
display math(A3)
display math(A4)
display math(A5)
display math(A6)
display math(A7)

Crop phenology sub-model

display math(A8)
display math(A9)
display math(A10)

Canopy photosynthesis sub-model

display math(A11)
display math(A12)
display math(A13)
display math(A14)
display math(A15)
display math(A16)
display math(A17)
display math(A18)
display math(A19)
display math(A20)

Leaf photosynthesis sub-model

display math(A21)
display math(A22)
display math(A23)
display math(A24)
display math(A25)
display math(A26)
display math(A27)
display math(A28)
display math(A29)
display math(A30)

List of symbols (with units) used in Eqns (A1–30) in the Appendix

a1An empirical coefficient, see Eqn A28 (−)
ANet photosynthesis rate (μmol m−2 s−1)
AcRubisco activity limited net photosynthesis rate (μmol m−2 s−1)
AjElectron transport limited net photosynthesis rate (μmol m−2 s−1)
ctCurvature factor in Eqn A9 (−)
CTotal carbon in live material of the crop (g C m−2 ground)
CcCO2 concentration at the carboxylation site of Rubisco (μbar)
CmaxMaximum carbon content of stem or seed at the end of its growth (g C m−2 ground)
CRCarbon in live root (g C m−2 ground)
CSCarbon in live shoot (g C m−2 ground)
CϑiCarbon demand for growth of an organ at stage ϑi (g C m−2 ground d−1)
DDeactivation energy (J mol−1)
D0An empirical coefficient, see Eqn A28 (kPa−1)
DlpDaylength for photoperiodic response of phenology (h)
EActivation energy (J mol−1)
g(T)Function for phenological response to temperature (−)
gmMesophyll conductance (mol m−2 s−1)
gmoResidual mesophyll conductance in Eqn A26 (mol m−2 s−1)
gsStomatal conductance for CO2 (mol m−2 s−1)
gsoResidual stomatal conductance for CO2 in Eqn A25 (mol m−2 s−1)
gtDiffusion conductance from ambient air to the site of carboxylation (mol m−2 s−1)
h(Dlp)Function for phenological response to photoperiod (−)
Ib0Incident direct-beam radiation above canopy (J m−2 ground s−1)
Ic,shAbsorbed radiation by shaded leaves of canopy (J m−2 ground s−1)
Ic,suAbsorbed radiation by sunlit leaves of canopy (J m−2 ground s−1)
IcAbsorbed radiation by canopy (J m−2 ground s−1)
Id0Incident diffuse radiation above canopy (J m−2 ground s−1)
IincPhoton flux density incident on leaves (μmol photon m−2 s−1)
JElectron transport rate through PSII used for NADP+ reduction (μmol e m−2 s−1)
κNitrogen-carbon ratio in crop (g N g−1 C)
kbDirect-beam radiation extinction coefficient (m2 ground m−2 leaf)
kbDirect-beam radiation extinction coefficient (m2 ground m−2 leaf)
math formula Scattered-beam radiation extinction coefficient (m2 ground m−2 leaf)
math formula Diffuse radiation extinction coefficient (m2 ground m−2 leaf)
knNitrogen extinction coefficient (m2 ground m−2 leaf)
KmcMichaelis–Menten constant of Rubisco for CO2 (μbar)
KmoMichaelis–Menten constant of Rubisco for O2 (mbar)
LGreen-surface area index of canopy (m2 leaf m−2 ground)
LiL counted from the top to the ith layer of canopy (m2 leaf m−2 ground)
mRMinimum number of days for seed filling phase (d)
mVMinimum number of days for vegetative growth phase (d)
MopMaximum or minimum optimum photoperiod (h)
n0Canopy top-leaf nitrogen (g N m−2 leaf)
nactActual nitrogen concentration in living shoot (g N g−1 dw)
nbMinimum leaf nitrogen for photosynthesis (g N m−2 leaf)
ncriCritical shoot nitrogen concentration (g N g−1 DW)
ncri0Initial critical shoot nitrogen concentration (g N g−1 DW)
NaLeaf nitrogen content per area (g N m−2 leaf)
Nc,shPhotosynthetically active nitrogen in shade leaves of canopy (g N m−2 ground)
Nc,suPhotosynthetically active nitrogen in sunlit leaves of canopy (g N m−2 ground)
NcTotal photosynthetically active nitrogen in canopy (g N m−2 ground)
NmaxupMaximum crop nitrogen uptake (g N m−2 ground d−1)
NRNitrogen in live root (g N m−2 ground)
NSNitrogen in live shoot (g N m−2 ground)
OOxygen partial pressure (mbar)
P25Value for Vcmax or Jmax or Rd (μmol m−2 s−1), or Kmc (μbar), or Kmo (mbar), at 25 °C
PsenPhotoperiod sensitivity of phenological development (h−1)
RUniversal gas constant (=8.314 J K−1 mol−1)
RdDay respiration (respiratory CO2 release in the light other than by photorespiration) (μmol m−2 s−1)
SEntropy term (J K−1 mol−1)
TDiurnal temperature (°C)
TbBase temperature for phenological development (°C)
TCCeiling temperature for phenological development (°C)
TlLeaf temperature (°C)
TOOptimum temperature for phenological development (°C)
VPDLeaf-to-air vapour pressure difference (kPa)
VcmaxMaximum rate of Rubisco carboxylation (μmol m−2 s−1)
WSWeight of live shoot (g DW m−2 ground)
χSlope factor between Vcmax or Jmax at 25 °C and leaf nitrogen (μmol g−1 N s−1)
δmA parameter in the gm model Eqn A26 (−)
δsA parameter in the gs model Eqn A25 (−)
δtA parameter in the gt model Eqn A27 (−)
κ2LLValue of conversion efficiency of incident light into J at the strictly limiting light [mol e (mol photon)−1]
ρcbCanopy beam radiation reflection coefficient (−)
σCRelative shoot activity (g C g−1 C d−1)
ϑDevelopment stage (−)
ϑ1Development stage at which plant starts to become sensitive to photoperiod (−)
ϑ2Development stage at which plant ends to respond to photoperiod (−)
ϑeDevelopment stage at the end of growth of stem or seed (−)
ϑiDevelopment stage during the growth of stem or seed (−)
ϑmDevelopment stage at the time of maximal growth rate of stem or seed (−)
ϕshFraction of shaded leaves in a canopy (−)
ϕsu,iFraction of sunlit leaves at canopy depth Li (−)
ϕsuFraction of sunlit leaves in a canopy (−)
Γ*Cc based CO2 compensation point in the absence of Rd (μbar)
θConvexity factor for response of J to Iinc (−)
σLeaf scattering coefficient (−)