The impact of modifying photosystem antenna size on canopy photosynthetic efficiency—Development of a new canopy photosynthesis model scaling from metabolism to canopy level processes

Abstract Canopy photosynthesis (Ac) describes photosynthesis of an entire crop field and the daily and seasonal integrals of Ac positively correlate with daily and seasonal biomass production. Much effort in crop breeding has focused on improving canopy architecture and hence light distribution inside the canopy. Here, we develop a new integrated canopy photosynthesis model including canopy architecture, a ray tracing algorithm, and C3 photosynthetic metabolism to explore the option of manipulating leaf chlorophyll concentration ([Chl]) for greater Ac and nitrogen use efficiency (NUE). Model simulation results show that (a) efficiency of photosystem II increased when [Chl] was decreased by decreasing antenna size and (b) the light received by leaves at the bottom layers increased when [Chl] throughout the canopy was decreased. Furthermore, the modelling revealed a modest ~3% increase in Ac and an ~14% in NUE was accompanied when [Chl] reduced by 60%. However, if the leaf nitrogen conserved by this decrease in leaf [Chl] were to be optimally allocated to other components of photosynthesis, both Ac and NUE can be increased by over 30%. Optimizing [Chl] coupled with strategic reinvestment of conserved nitrogen is shown to have the potential to support substantial increases in Ac, biomass production, and crop yields.

light use efficiency and hence increase canopy photosynthesis.
Increasing light penetration into bottom layers of a canopy can be realized through manipulating canopy structure, which has been applied in crop breeding. For instance, erect leaves were selected during rice breeding, which increases light penetration into bottom layers of a canopy (Peng, Khush, & Cassman, 1994). Besides canopy architecture, decreasing leaf chlorophyll concentration has also been suggested as another potential method to improve light distribution and hence light use efficiency inside crop canopies (Ort, Zhu, & Melis, 2011;Zhu, Long, & Ort, 2010). However, previous theoretical studies used models in which the microclimatic condition inside a canopy were dramatically simplified. In particular, the light environment inside the canopy was mainly divided into only sunlit and shaded (Norman, 1980). Our theoretical analysis has shown that such a simplification leads to up to 17% bias in the estimated total canopy photosynthetic CO 2 uptake rate (Zhu, Song, & Ort, 2012). In addition, two major advances in recent years now make possible development of a new generation of dynamic systems model of canopy photosynthesis, where both the light environment inside the canopy and also the detailed photosynthetic processes are integrated (Zhu, Wang, Ort, & Long, 2013).
The first major development is the tool to predict the light environment inside a canopy (Song, Zhang, & Zhu, 2013). Light inside a canopy is highly heterogeneous both spatially and temporally (Pearcy, 1990).
The leaves in the lower layers usually have low light levels; however, these low light levels are sporadically interrupted by high light sunflecks (Pearcy, 1990), which make up to a large proportion of the incident solar energy on lower canopy leaves. A lot of previous efforts to model canopy photosynthesis, including the classical big leaf model (Running & Coughlan, 1988;Sellers, Berry, Collatz, Field, & Hall, 1992;Thornley & Johnson, 1990), sunlit/shaded model (Dai, Dickinson, & Wang, 2004; DePury & Farquhar, 1997;Wang & Leuning, 1998), and multilayer model (DeWit, 1965;Lemon, Stewart, & Shawcroft, 1971;Norman, 1979), do not fully consider the high level of spatial and temporal heterogeneities of light inside the canopy. Zhu and colleagues used a reverse ray tracing algorithm combined with a simplified canopy architecture to predict the spatial and temporal heterogeneity inside an idealistic canopy (Zhu, Ort, Whitmarsh, & Long, 2004). Using this model, the potential impact of formation and relaxation of photoprotection inside a canopy was explored, which led to the discovery that the natural slow recovery from photoprotected state could lead to up to 30% loss of A c (Zhu et al., 2004). Recently, algorithms to reconstruct three-dimensional canopy architecture and algorithms for forward ray tracing were developed, enabling a more accurate prediction of light environment of a canopy and allowing for user-defined canopy architecture parameters (Song et al., 2013).
A comprehensive dynamic systems model of leaf photosynthesis, which incorporates description of the detailed processes including both the electron transfer processes and the dynamics of carbon metabolism, has also been developed recently . This model, in comparison to earlier steady state biochemical photosynthesis model (Farquhar, Caemmerer, & Von, & Berry J.A., 1980), can predict the dynamic changes of photosynthesis under varying light and CO 2 levels.
This improved model, known as e-photosynthesis, is also able to predict the potential impacts of manipulation of different components to leaf photosynthetic efficiency. By combining with evolutionary algorithms, we are now able to explore the optimal nitrogen distribution into different enzymes of photosynthetic carbon metabolism to maximize photosynthesis. Combining this advanced dynamic leaf photosynthesis model with modelling of the heterogeneous light environment within canopies enables prediction of dynamic changes of canopy photosynthesis in any canopy of defined architecture.
The e-photosynthesis model, in which each photosynthetic reaction and process is explicitly represented, also enables the study of the nitrogen investment to maximize photosynthesis . Because light varies widely in different layers within canopies, there is photo-acclimation of leaves to irradiance that changes as canopy grows. It is well known that leaves under higher growth light tend towards higher nitrogen content per leaf area as indicated by the observed decline in nitrogen content with light levels inside the canopy (Evans & Poorter, 2001;Field, 1983;Hikosaka, 2005). In addition, nitrogen distribution among photosynthetic enzymes within leaves are different under different growth irradiance (Evans, 1993a;Evans, 1993b;Evans & Poorter, 2001;Hikosaka & Terashima, 1995;Niinemets, Kull, & Tenhunen, 1998). For example, under higher growth light, more nitrogen is partitioned to Rubisco and electron transport chain components, as compared to low growth light where nitrogen investment shifts towards light harvesting (Evans, 1989). Combining a realistic light environment inside a canopy with the e-photosynthesis model offers the opportunity to investigate the optimal nitrogen distribution among photosynthetic enzymes within those leaves.
In this study, we have assembled an integrated canopy photosynthesis model by combining canopy architecture model (Song et al., 2013), ray tracing algorithm (Song et al., 2013), photo-acclimation model (Hikosaka & Terashima, 1995;Kull & Kruijt, 1999;Moreau et al., 2012), and dynamic systems model of C 3 leaf photosynthesis . Using this model, we have systematically evaluated the effects of reducing leaf chlorophyll concentration to light and nitrogen use efficiencies of a rice canopy.

| Plant materials and experiments
Rice cultivar 9522 (Oryza sativa L. japonica) was planted in the experimental station in Shanghai (Latitude 31°N) in 2012 with a planting density 25 × 20 cm 2 (20 plants/m 2 ). Canopy architectural features and the physiological parameters were collected in the booting stage (August 23, 235 DOY). The leaf reflectance and transmittance were measured using integrating sphere and spectrometer (Ocean Optics, Dunedin, Florida, USA). The leaf reflectance (r) and transmittance (t) were then calculated according to the following equations (Equations 1-2), where the r n is reflectance of wave length n and I n is light intensity at wave length n.
r ¼ ∑ 700 n¼400 r n ⋅I n = ∑ 700 n¼400 I n ; (1) t ¼ ∑ 700 n¼400 t n ⋅I n = ∑ 700 n¼400 I n : We measured the SPAD values using a chlorophyll metre SPAD-502Plus (Konica Minolta, Japan) for different leaf segments, that is, the leaf base at 1/6 of the leaf length, leaf middle segment at 1/2 of the leaf length, and leaf tip at 5/6 of the length, of the flag leaf, the second leaf, the third leaf, and the fourth leaf ( Figure 1a). Chlorophylls at these different segments were also extracted with 95% ethanol to measure concentrations using spectrophotometer following Arnon (1949). The chlorophyll concentrations and the corresponding SPAD readings were used to derive a relationship between chlorophyll concentration and single-photon avalanche diode (SPAD) reading (Equation S2). Photosynthesis was measured with the gas exchange method using LI-6400XT (LI-COR, Lincoln, Nebraska, USA). Light response curves of flag leaves were measured under a CO 2 concentration of 400 ppm and the photosynthetic photon flux density (PPFD) was changed stepwise from 2,400 to 50 μmol·m −2 ·s −1 . (Figure 1b). The P max (the maximal light saturated photosynthesis under ambient CO 2 concentration) and ϕ (the initial slope of light response curve) were fitted with a nonrectangular hyperbola model (Thornley, 2002). Leaf nitrogen content of the flag leaf was deter-

| Canopy model and ray tracing algorithm
A 3D canopy structure model representing nine rice plants was constructed using mCanopy (PICB, Shanghai) with parameters collected from rice plants described above using the methods described in Song et al. (2013). A ray tracing algorithm was applied to simulate PPFD distribution in this canopy using the software fastTracer developed in Song et al. (2013) Table S1). t and r were then used to parameterize the ray tracing programme to simulate PPFD distribution inside a canopy.

| Nitrogen distribution in canopy
A model that describes the relationship between vertical nitrogen distribution and light distribution within a canopy (Moreau et al., 2012) was used to estimate the nitrogen profile in different canopies. In the model (Equations 3 and 4), N LA is nitrogen per leaf area, N fl LA is the N LA in flag leaf, and n b (g N m −2 leaf lamina) is the basal leaf nitrogen concentration. of the flag leaf, the maximal photosynthetic CO 2 uptake rate under ambient CO 2 and saturate photosynthetic photon flux density (PPFD; P max ) and the initial slope of the A-Q curve (ϕ)fitted with a nonrectangular hyperbola model. (c) Leaf nitrogen contents for different leaf positions. Flag leaf nitrogen content was measured and other leaves was predicted based on model (Equations 3-4; mean ± std, n = 5) I l is PPFD incident on the flag leaf and I lfl is I l at the middle of a flag leaf layer as N LA of a layer was related to I l /I lfl (Milroy, 2001) and b is equal to the ratio of extinction coefficient of nitrogen and light in canopy. In this study, we calculated b based on green leaf area index (GAI) using the equation used in Moreau et al., 2012.

| Enzymes and proteins concentration calculated with photo-acclimation model
A photo-acclimation model for nitrogen partitioning among major photosynthetic proteins in a leaf was developed to link e-photosynthesis  and leaf nitrogen concentration. The e-photosynthesis model can predict the amount of leaf absorbed PPFD used for photochemistry, heat dissipation, and fluorescence emission . The photo-acclimation model assumes that leaves optimize the distribution of nitrogen among photosynthetic components for maximizing the daily carbon uptake per leaf area (Hikosaka & Terashima, 1995). concentrations of major components and leaf photosynthesis parameters, P max and ϕ, were generated using e-photosynthesis model (Equations 5-8); second, the diurnal PPFD absorbed by a leaf was simulated using fastTracer software for 5 days to generate an average diurnal growth PPFD curve of 5 days; third, the molecular weights and nitrogen contents of these groups were calculated, and at a given nitrogen content, the relationship between P max and ϕ was generated; and finally, for the simulated averaged diurnal growth PPFD, a range of P max and corresponding ϕ were used to calculate daily carbon uptake and the optimal P max and ϕ for maximal daily carbon uptake were selected, then the concentrations of Rubisco, CE, ETCF, and PSII were calculated based on their relationships to P max and ϕ.
2.5 | Leaf photosynthetic CO 2 uptake calculated with e-photosynthesis model The e-photosynthesis model  was parameterized with the enzymes concentrations (c) and catalytic numbers (k cat ). First, the enzymes and proteins in photosynthesis were divided into seven groups as described above. Within each group, the ratios among enzymes were set constant (Table S3) and the concentrations of those groups were calculated from leaf nitrogen content and environmental light according to photo-acclimation model described above. The k cat of all enzymes are for typical C 3 plants as used in Zhu et al., 2013; Table S3). The V max of all enzymes were then calculated by equation (Equation 9).
2.6 | Chlorophylls in antennas of PSII and PSI   The integrated canopy photosynthesis model scales from metabolism to canopy, which provides the capacity for studying the impacts of modification made at the molecular level on leaf and canopy level photosynthetic CO 2 uptake rates. In this study, we explored two options of modifying leaf chlorophyll concentrations using the FIGURE 2 A diagram showing the integrated canopy model of rice that combines 3D canopy structure, ray tracing algorithm, and e-photosynthesis model. The 3D canopy model was constructed based on rice canopy structural parameters. The ray tracing algorithm follows Song et al. (2013) and predicts photosynthetic photon flux density (PPFD) of all leaves in the canopy. The PPFD was used as input of the e-photosynthesis to calculate the leaf photosynthetic rate. Canopy photosynthetic CO 2 uptake rate was calculated as the integral of photosynthetic CO 2 uptake rates of all leaves. The diagram of photosynthesis is adapted from Zhu et al. (2013) with permission e-photosynthesis model. The first one was by changing the number of photosystems units while keeping antenna size for each photosystem constant ( Figure S1), and the second one was by modifying the antenna size while maintaining the number of photosystems units constant ( Figure S1). For the first option, the energy conversion efficiency for one photosystem did not change because the structure of each photosystem was the same. However, for the second option, when decreasing antenna size, the proportion of absorbed PPFD used for photochemistry gradually increased and the heat dissipation and fluorescence emission gradually decreased with decreasing antenna size ( Figure 3). This is because the leaf absorbance and total absorbed PPFD decreased with decreasing antenna size, but the PPFD used for photo-chemistry was almost the same (Figure 3).
To demonstrate the impact of these two options of changing leaf chlorophyll concentration on leaf photosynthesis under different light intensities, we simulated leaf photosynthesis under different absorbed light for leaves with different chlorophyll concentrations by changing antenna size ( Figure S2A) and by changing photosystems number ( Figure S2B). The initial slope of the curve increased when chlorophyll concentration was decreased by changing antenna size ( Figure S2A), but the initial slope decreased when chlorophyll concentration was decreased by changing photosystems number ( Figure S2B). We further simulated leaf photosynthesis under different incident light. Simulation results show that when chlorophyll concentration was changed by changing antenna size, the initial slope were almost the same ( Figure S2C), but the slope decreased when chlorophyll concentration was decreased by changing photosystems number ( Figure S2D).

| Distribution of PPFD in a canopy when leaf chlorophyll concentration ([Chl]) was modified
Modifying leaf chlorophyll content can lead to modified light environments inside a canopy because leaf absorbance positively related to chlorophyll concentration with R-square 0.91 (Figure 4d). We quanti-

| The influence of modified leaf chlorophyll concentration on the optimal distribution of nitrogen into different components of photosystems
For a given investment of nitrogen into the photosynthetic apparatus, there needs to be an optimal allocation to maximize photosynthetic light and hence nitrogen use efficiencies. Thus, as less nitrogen is invested in the photosystems, it matters how conserved nitrogen is re-invested. This is illustrated in two simulated scenarios. In one scenario, the antenna size was decreased without modifying content of other photosynthetic proteins, whereas in another scenario, the antenna size was decreased with increasing content of other proteins to maintain the total nitrogen invested into photosynthetic apparatus to be constant. Figure 5 illustrates these two scenarios when leaf chlorophyll concentration was decreased by 60%, though the LHC decreased dramatically, all the contents of all other enzymes, that is, Rubisco, electron transport chain (ETC), PSII in photosynthesis were increased for all leaves in the canopy ( Figure 5).

| The influence of modifying leaf chlorophyll concentration on leaf and canopy photosynthetic efficiency
To test the hypothesis that decreasing antenna size can improve canopy photosynthesis and nitrogen use efficiency (NUE), we calculated daily canopy photosynthetic CO 2 uptake rate (A c ) and NUE. Our analysis showed that the A c was increased over 3% and NUE increased by   decreasing leaf chlorophyll content has been proposed as a viable option. The potential impacts of modifying leaf chlorophyll concentration on canopy photosynthesis has been explored earlier using a sunlitshaded model , where the leaves inside the canopy was assumed to be either sunlit or shaded (Norman, 1980). In other words, the temporal and spatial heterogeneities of light environments inside the canopy was ignored. Considering that ignoring the heterogeneity of such light environments can potentially bias the estimate of canopy photosynthetic rates, here we study the potential benefits of modifying leaf chlorophyll concentration on rice canopy photosynthetic rates. Our analysis shows that decreasing antenna size in general can increase canopy photosynthetic CO 2 uptake rates (Figures 6 and   8), even though the magnitude of the benefit depends on both the growth latitude and a number of plant architectural parameters ( Figure 8). For example, under low leaf area index, the relative benefit of decreasing antenna size will be lower (Figure 8). When the leaf chlorophyll concentration was decreased, the canopy nitrogen use efficiencies increased dramatically ( Figure 6). This is due to the decreased nitrogen investment while at the same time the increased canopy photosynthetic rates ( Figure 6).
The increased canopy photosynthetic CO 2 uptake rates under decreased antenna size is attributed to two major factors. First, when the antenna size decreases, the proportion of PPFD used for photochemistry increases ( Figure 3) as a result of the decreased proportion of heat dissipation (Zhu et al., 2005). This is also reflected in the increased leaf photosynthetic CO 2 uptake rates under nonsaturated light when the antenna size was smaller ( Figure S2A). Second, when the antenna size decreased, the light distribution inside the canopy was improved, in the sense that the absorbed PPFD of top leaves were slightly decreased while the absorbed PPFD of leaves at bottom layers were increased (Figure 7) due to the decreased extinction coefficient ( Figure 4a). This modified light environments combined with the nonlinearity light response curve of photosynthesis (A-Q curve) together results in a higher A c (Figures 6 and 7).
If the nitrogen saved by decreasing leaf chlorophyll content can be optimally allocated to other components of photosynthesis, much higher increase in total canopy photosynthesis was predicted ( Figure 6a). Our earlier study suggested that the current nitrogen investment into photosynthetic machinery is not optimal, as a result of changed global CO 2 concentrations, which in theory can shift the control over photosynthetic CO 2 uptake from Rubisco to RuBP regeneration (Zhu, de Sturler, & Long, 2007). This is later demonstrated in the field experiment where tobacco with overexpressed SBPase showed greater stimulation in biomass accumulation under elevated atmospheric CO 2 concentration (Rosenthal et al., 2011). Hence, it is desirable to consider the optimal nitrogen allocation patterns together with the decreased antenna size (Zhu et al., 2007). Now the challenge is to identify the optimal option to decrease leaf chlorophyll concentration and also the antenna size. One possibility is to modify chlorophyll a oxidase, which has been reported to be related to antenna size (Masuda, Tanaka, & Melis, 2003). Another possibility is to modify FetZ, which is a major factor involved in the chloroplast division machinery and hence influence mesophyll chloroplast number (TerBush, Yoshida, & Osteryoung, 2013). In theory, decreased expression of FetZ should lead to decreased chloroplast division and hence increased leaf light transmittance, and potentially reflectance as well, due to sieve effect.
The impacts of these modifications on leaf and canopy photosynthesis awaits experimental verification.

| Potential applications of the new dynamic model of canopy photosynthesis and its future developments
Canopy photosynthesis, rather than leaf photosynthesis, should be the target to increase for higher biomass production and crop yield (Zhu et al., 2012), as has been demonstrated in cotton (Wells et al., 1986) and soybean (Harrison & Ashley, 1980). Unfortunately, the complexity of the photosynthetic process, which consists of about 100 proteins, combined with the heterogeneous microclimates, in particular light conditions inside a canopy, make it rather challenging to identify the limiting factors controlling canopy photosynthesis using the traditional transgenic approaches. The model presented here incorporates a realistic three-dimensional plant architecture, detailed prediction of light environments inside the canopy (Song et al., 2013), together with a dynamic systems model of leaf photosynthesis , which enables a direct prediction of the impacts of modifying a particular enzyme or a set of enzymes involved in photosynthesis on canopy photosynthesis and nitrogen use efficiencies of a crop with defined canopy architecture, growth location, and growth densities. Such a newly gained capacity is timely because modern biotechnologies, such as genome editing technologies (Bortesi & Fischer, 2015), now make it possible to engineer any one or combination of genes relatively easily while the challenge is to define the targets to manipulate. The model also enables evaluation of different planting strategies on canopy photosynthesis rate, as demonstrated in our recent study where we show the impact of using different planting systems, that is, even or varied row spacing, on sugarcane production (Wang et al., 2017). Skipping some rows in rice or wheat will potentially lead to decreased leaf area index, which can potentially decrease the potential benefit of decreasing antenna size. As shown in our sensitivity analysis (Figure 8 icle for rice, in particular indica rice, is usually the same as or lower than that of the flag leaf. Therefore, the spikes of rice and wheat influence light canopy microenvironments differently. Most likely, the existence of floral structure decreases light levels inside canopies and hence can magnify the impacts of lower chlorophyll on canopy photosynthetic rate. Second, in the current model, the contribution of leaf sheath photosynthesis is not incorporated. Many evidences suggest that in rice and wheat, photosynthate contributed by sheath photosynthesis is important to grain filling (Guo, He, & Deng, 2013;Zhang, Zhang, Wang, & Wang, 2011) and can be 5-14% of the total final grain yield (Zhang et al., 2011). The predicted detailed light environments at different parts of a leaf sheath make it possible to calculate the contribution of sheath photosynthesis if the biochemical and physiological parameters related to sheath photosynthesis are available. Third, in the current model, the CO 2 gradient inside the canopy is not explicitly simulated. Earlier studies have shown a moderate drawdown of CO 2 concentration from the air immediately above the canopy to the middle of a soybean canopy at midday (Francis & Parks, 1988). Though such a drawdown only has a~4% impact on total canopy CO 2 uptake rate (Zhu et al., 2012); however, for canopies with much higher leaf area index in an environment with still air, the potential CO 2 drawdown and impact on canopy photosynthesis can be greater. Therefore, future models of dynamic canopy photosynthesis also need to incorporate the dynamic changes of CO 2 concentration inside a canopy. Under such conditions, the proportion of leaves performing light-limited photosynthesis in a lower layer of canopies decreases. As a result, the benefit of increasing light availability for lower layer leaves will decrease.
Models with explicit simulation of CO 2 gradients inside a canopy need to be developed to quantify the impacts of decreasing leaf chlorophyll concentration on canopy photosynthesis under such cases.