Canopy nitrogen distribution is optimized to prevent photoinhibition throughout the canopy during sun flecks

As photoinhibition primarily reduces the photosynthetic light use efficiency at low light, sunfleck-induced photoinhibition might result in a fatal loss of carbon gain in the shade leaves within a canopy with barely positive carbon balance. We hypothesized that shade leaves at the lower canopy might retain a certain amount of leaf nitrogen (NL) to maintain energy consumption via electron transport, which contributes to circumventing photoinhibition during sunflecks to keep efficient utilization of low light during the rest period of daytime. We investigated excess energy production, a potential measure of susceptibility to photoinhibition, as a function of NL distribution within a Japanese oak canopy. Optimal NL distribution, which maximizes canopy carbon gain, may lead to a higher risk of photoinhibition in shade leaves during sunflecks. Conversely, uniform NL distribution would cause a higher risk of photoinhibition in sun leaves under the direct sunlight. Actual NL distribution equalized the risk of photoinhibition throughout the canopy indicated by the constant excess energy production at the highest light intensities that the leaves received. Such a homeostatic adjustment as a whole canopy concerning photoinhibition would be a key factor to explain why actual NL distribution does not maximize canopy carbon gain.

leaves may well acclimate to sun flecks (brief and strong irradiance) rather than averaged irradiance, so as to circumvent transient photoinhibition.
Photoinhibition primarily reduces the photosynthetic light use efficiency at low PPFD, i.e. a reduction of the initial slope of the photosynthetic light-response curve rather than the light-saturated photosynthetic rate 17 . In this context, sunfleck-induced photoinhibition may result in a fatal loss of carbon gain in the shade leaves with barely positive carbon balance 16 . To prevent photoinhibition, electron transport, mainly involved in photosynthesis and photorespiration 21 , plays an important role in consuming absorbed light energy as well as thermal energy dissipation 3,22,23 . Excess energy, neither consumed by electron transport nor dissipated as heat, is known to have a close relationship to photoinactivation of PSII under the inhibition of PSII repair 24,25 .
Leaf N participation for the circumvention of photoinhibition is specifically involved in electron transport, but not in xanthophyll-related thermal energy dissipation, since xanthophyll pigments, violaxanthin (C 40 H 56 O 4 ), antheraxanthin (C 40 H 56 O 3 ), and zeaxanthin (C 40 H 56 O 2 ), do not contain any N 26 . As for the fractions of leaf N in Rubisco, and in proteins related to linear electron transport, there can be little differences in the fractions between upper and lower canopy leaves 27 . In this context, leaf N gradient within the canopy can closely reflect the gradient of photosynthesis-related proteins. The capacity of electron transport depends on N L , which is varied through intra-canopy light gradients 28,29 . Thus, shade leaves at the lower canopy may retain a certain amount of N L to circumvent photoinhibition during sunflecks to keep efficient utilization of low PPFD during the rest period of daytime 16 .
We hypothesized that N L should be distributed within a canopy to prevent photoinhibition both in shade leaves during sunflecks and sun leaves under the direct sunlight, leading to the less steep N L distribution than the optimal N L distribution. To test this hypothesis, we determined how excess energy production, which could be a possible measure of the susceptibility to photoinhibition 22,24,25,30 , was distributed within a canopy when N L distribution was changed. We investigated photosynthetic traits in the leaves within a canopy of Japanese oak by gas exchange and chlorophyll fluorescence measurements, as well as light and N distribution within the canopy.

Results
Total leaf area index (LAI), F T , in the forest was estimated to be 5.91 m 2 m −2 based on the stratified clipping method conducted in 2001 and 2003. Based on the LAI cumulated from the canopy top (F), the light extinction coefficient of the canopy (K L ) was 0.739 in Eq. 6. Based on the relationship between the relative light intensity (I/ I 0 ) and F (Eq. 6), we derived F of 15 leaves used for the photosynthetic measurements in 2007 by using relative light intensity at each leaf position. The actual distribution of N L as a function of F in 2007 was revealed as follows: − . + . . . N 222 exp( 0 241F) 0 269(according to Eq 7 in Methods) L Based on the integration of the actual N L distribution, total N per unit ground area (N T ) was calculated as 8.59 g m −2 . Optimal N distribution in the canopy was expressed as follows: − . + . . . N 524 exp( 0 739F) 0 269(according to Eq 8 in Methods) L N L with uniform distribution was estimated to be 1.45 g m −2 , estimated as N T /F T . Actual and optimal N L distributions within the canopy are shown in Fig. 1. Optimal N L was higher than the actual value in the case of F < 1.7 m 2 m −2 , whereas it was lower than the actual value in the case of F > 1.7 m 2 m −2 . F of 1.7 m 2 m −2 corresponds to the average integrated daily quantum flux density in June (Q int ) of 12 mol m −2 day −1 (Eq. 6).
Electron transport rate (ETR) with actual, optimal and uniform N L distribution showed similar relationships as a function of N L (Fig. 3a). Higher ETR in sun leaves but lower ETR in shade leaves was observed with optimal N L distribution when compared with those with actual N L distribution (Table 1). To investigate how ETR changed from the actual to the optimal and uniform N L distribution, we showed ETR opt /ETR act and ETR uni /ETR act as a function of Q int (Fig. 4a). Higher ETR opt /ETR act was observed in leaves grown under higher Q int , compared among leaf types (Fig. 4a, Table 2). When compared with ETR in leaves with actual N L distribution for each leaf type, relatively lower ETR in the deep shade leaves and relatively higher ETR in the typical sun leaves were observed with optimal N L distribution (Table 2). Conversely, with uniform N L distribution, higher ETR uni /ETR act was observed in the deep shade leaves than the other leaves, where relatively higher ETR in the deep shade leaves and lower ETR in sun leaves were observed when compared with those in leaves with actual N L distribution (Fig. 4a, Table 2).
Higher thermal energy dissipation (D) was generally observed in leaves grown under higher Q int with any N L distribution, whereas the slope become less steep in leaves with optimal N L than the actual, and uniform N L distribution (Fig. 3b, Table 1). With optimal N L distribution, the typical sun leaves showed lower D opt /D act than the other leaf types, and relatively lower D than that in the typical sun leaves with actual N L distribution. Conversely, there was no significant differences in D uni /D act among leaf types with uniform N L distribution, and also there was no difference in D in leaves with uniform N L distribution from that in leaves with actual N L distribution within each leaf type (Fig. 4b, Table 2).
There was no difference in excess energy production (E) among leaf types with the actual N L distribution (Fig. 3c, Table 1). A significantly higher E was observed in shade leaves than in sun leaves with optimal N L distribution. Conversely, a significantly higher E was observed in sun leaves than in shade leaves with uniform N L distribution (Fig. 3c, Table 1). Similarly, higher E opt /E act was observed in shade leaves than in sun leaves with Scientific REpoRTS | (2018) 8:503 | DOI:10.1038/s41598-017-18766-0 optimal N L distribution, whereas higher E uni /E act was observed in sun leaves than in shade leaves with uniform N L distribution (Fig. 4c, Table 2). With optimal N L distribution, deep shade leaves showed relatively higher E, but typical sun leaves showed relatively lower E compared with those in actual N L distribution (indicated by the asterisks in Table 2). Conversely, with uniform N L distribution, deep shade leaves showed relatively lower E, but sun leaves showed relatively higher E compared with those in actual N L distribution.
Based on the relationship between variance in E (s 2 ) and the leaf N distribution coefficient (K n ), the K n minimizing s 2 was estimated as 0.291 (Fig. 5), which was very close to the K n in the actual N L distribution (0.241 in Eq. 7). We also found that the values of E at PPFD max with the actual N L distribution within the canopy of Japanese oak were very similar to those with the N L distribution to maintain E constant, minimizing the variance in E (K n = 0.291) (Fig. 6).

Discussion
We investigated the fate of absorbed light energy (ETR, D, and E) in leaves with actual, optimal, and uniform N distribution within the canopy of Japanese oak. Excess energy is known to have a close relationship to photoinactivation of PSII under the inhibition of PSII repair 24,25 . There are two mechanisms involved in photoinhibition (defined as photoinactivation of PSII), i.e., an excess energy mechanism and a two-step mechanism 25 . As excess , and typical sun leaves (circle, 25 < Q int < 35 mol m −2 day −1 ). energy might be predominantly involved in photoinhibition in the field under visible light especially at the range of relatively high irradiance 25,30,31 , c.f. 700-1800 µmol m −2 s −1 during the sunflecks in the present study, we used the chlorophyll fluorescence parameter "excess energy, " as an empirical measure of the sensitivity of photoinhibitory damage based on the former mechanism. It is noteworthy that E during sunflecks for shade leaves or direct sunlight for sun leaves was stable among leaf types with the actual N L distribution at various canopy depths, whereas a significantly higher E was observed in the shade leaves than sun leaves with the optimal N L distribution; a significantly higher E in sun leaves than shade leaves with the uniform N L distribution (Fig. 3c, Table 1). As photoprotective D was not enhanced in the leaves with decreased ETR ( Table 2) 22 , the decreases in ETR in the shade leaves with an optimal N L distribution, and in the sun leaves with a uniform N L distribution both resulted in the increases in E. This also suggests that the photoprotective thermal energy dissipation would not respond enough to set off the changes in ETR with different N L distributions as far as the present model prediction is concerned, partly because of the small variation in D in the fully acclimated leaves to their growth environments (Supplemental Fig. S1) as reported in long-term drought acclimated plants 32,33 .
In the present study, the optimal N distribution was estimated using the simple Beer's law with an assumption that leaves received diffuse light only 6,7 . Hikosaka 10 reported that the optimal N distribution under direct and diffuse light, a more realistic condition than diffuse light only 6,7 , was steeper than that under diffuse light only. However, from the viewpoint of photoinhibitory damage, the actual N distribution in the canopy of Japanese oak, which was less steep than the optimal distribution, might be inherently optimized to circumvent photoinhibitory damage at the whole-canopy level by maintaining excessive energy at a certain level under direct sunlight throughout the leaves grown at various light environments within a canopy.
We estimated canopy C gain based on light response curves built with different N distribution and incident PPFD with 1-min-intervals in July 2007 (details in Supplementary Information: Estimation of canopy C gain, Figs S2-S5). We used coefficients of light response curves measured at a leaf temperature of ≈27 °C, with no correction for the effects of air temperature, and air humidity on photosynthesis. Total canopy C gain was 24.9, 21.3, and 15.8 mol m −2 on the ground surface basis for optimal, even-E, which makes E constant throughout the canopy with K n = 0.291 (Fig. 6), and uniform leaf N distribution, respectively. Even-E N distribution, showed a 15% decrease in canopy C gain integrated in July, during a photosynthetically most vigorous period, compared with optimal N distribution. Conversely, uniform N distribution resulted in a 36% decrease in canopy gain compared with the optimal N distribution. It is noteworthy, not only the sun leaves with higher leaf N L with optimal canopy N distribution, but also the shade leaves with lower N L compared with those with even-E N distribution contributed to the higher C gain as a consequence of reduced nighttime respiration rate (Supplementary Figure S5). In this estimation, a decrease in photosynthetic efficiency caused by photoinhibition under fluctuating light was not taken into account 16,19,20 . An enhanced recovery capacity from photoinhibition by bioengineering lead to an increase in biomass of tobacco plants by up to 20% in fluctuating light 16 . Thus, there is a possibility to reduce the difference in the canopy C gain between even-E and optimal N distribution, when the photoinhibitory effects on photosynthetic efficiency are accounted. In addition to an acclimation to short-term fluctuating light (sunflecks), as Japanese oak is a gap-dependent species, which needs a gap formation for its regeneration 2,23 , keeping relatively higher ETR in shade leaves than optimal would be a pre-conditioning for long-term fluctuating light (gap formation) for this species.
The cost of D1 protein turn-over, as a relevant process of PSII repair, is substantially low at most 0.5% of photosynthetic ATP production 34 , and the cost of relaxation of thermal energy dissipation is also reported to be substantially small, about 0.05% of the total photosynthetic electron flow under saturating light 26 . Therefore, the recovery cost from photoinhibition would be negligible for canopy C balance. Rather, reduced photosynthetic efficiency during the recovery from photoinhibition on transfer from sun flecks to shade may have a more significant adverse effect on canopy C gain 16 . In this context, circumvention of photoinhibition via an enhancement of ETR is essential to maintain the canopy C gain by preventing photoinhibition without any decrease in photosynthetic efficiency, in contrary to xanthophyll-related thermal energy dissipation.
Leaf photosynthetic capacity is closely related to leaf morphology such as leaf mass per area (LMA), where leaves with greater LMA generally have greater leaf area-based nitrogen contents and greater photosynthetic capacity 3,35,36 . During leaf maturation of Japanese oak, LMA reaches its maximum, accompanied with an increase in net photosynthetic rate, several weeks after the leaves are fully expanded and the light environment within a canopy is fixed (Tobita unpublished data 37 ). Although the sun/shade anatomy of deciduous leaves is mainly determined by the light condition in the previous year 38 , less extent but some plasticity exists in leaf morphological change in response to current-year light environment 39 . For Japanese oak, LMA and net photosynthetic rate increased during June, after the leaf expansion had completed at the end of May (Tobita unpublished data). It was for this reason that we chose growth light environments (Q int , PPFD max ) in June as determining factors for photosynthetic traits. The equalized amount of excessive energy production observed across the various light environments within a canopy (Figs 5 and 6) could be explained if the developments of photosynthesis (involved in ETR) and photoprotection (involved in D) along with the increase in LMA would be continued until the excessive energy production during sunflecks or direct sunlight drops below a certain level both in shade and sun leaves 19,20 . The hypothesis of the present study: higher N L than optimal is needed in the shade leaves to prevent photoinhibition during sunflecks, is partly similar to that proposed by Dewar et al. 12 , where a lower bound of LMA in the shade leaves exists, in relation to the limitation in leaf morphological plasticity 23 . Furthermore, the present study also proposes that the upper bound of N L in sun leaves would be regulated as the minimum required to circumvent photoinhibition under direct sunlight, which might be related to the upper-bound constraint on photosynthetic capacity at the top of the canopy proposed by Lloyd et al. 13 to explain the less steep N L decline. In the present study, water stress during leaf development was considered to be negligible because of a considerable amount of spring snowmelt in the forest (mean maximum snow depth was 114 cm) 40 . Peltoniemi et al. 14 proposed that limited hydraulic conductance for the sun leaves may result in a lower N L in sun leaves with actual N L distribution than the optimal one. In contrast, leaves acclimated to long-term drought often show higher N L accompanied with higher ETR, than leaves grown without drought stress, to prevent photoinhibition under lower intercellular CO 2 concentration by stomatal closure 32,33,41,42 . In this context, water stress on sun leaves induced by the limited hydraulic conductance in the upper canopy 43,44 would not necessarily decrease N L in sun leaves, but rather increase N L for preventing photoinhibition. Further investigation is needed on such environmental stresses during leaf development influencing excess energy production to predict leaf N distribution within a canopy. optimal (open) and uniform N L distribution (grey symbols) (ETR opt , D opt and E opt ; ETR uni , D uni and E uni ) to those with actual N L distribution (ETR act , D act and E act , respectively) at PPFD max for a given leaf as a function the daily mean of the integrated photon flux density during leaf development (Q int ). Dashed line indicates the ratios = 1. Symbols are the same as in Fig. 1 Table 2. The ratio of ETR, D and E with optimal and uniform N L distribution (ETR opt , D opt and E opt ; ETR uni , D uni and E uni ) to those with actual N L distribution (ETR act , D act and E act , respectively) at PPFD max in different leaf types of Q. mongolica. Values are means ± SE for each leaf type. Different letters indicate significant differences among the leaf types at P < 0.05, according to Holm's pairwise comparisons. ns indicates non-significant effect of leaf type. *Indicates significant differences in ETR, D and E from those of actual N L distribution in each leaf type at P < 0.05 with t-test.

Conclusion
The present study provides a novel insight into the canopy N L distribution with respect to circumvention of photoinhibitory damage at the whole-canopy level like a homeostatic adjustment, where plants follow a strategy through which N L distribution is not optimized for canopy C gain but rather regulated to prevent photoinhibition during sunflecks for shade leaves or direct sunlight for sun leaves. This regulation of N L distribution contributes to keeping an efficient light utilization capacity during the daytime at the whole-canopy level, by means of minimizing a time lag in the recovery from photoinhibition 16 , which indicates an important biological plasticity in the response to light environment, and may suggest genetically-driven adjustments for successful acclimation in a changing light environment. Revealing such a physiological coordination of individual leaves for the benefit of the whole canopy will contribute to advancing the understanding of carbon-nitrogen interactions 45,46 and energy flow in terrestrial ecosystems. Further studies with different species and at different environmental conditions are needed to demonstrate that the regulation of N L distribution found in this study is an unequivocal biological mechanism occurring across species.

Methods
Study site. The where ISF is the indirect (diffuse) site factor estimated by the WinScanopy software, penetration of the direct radiation is also assessed with the sun track simulated by the software. The average integrated daily quantum flux density at each leaf position in June (Q int ; mol m −2 day −1 ) was calculated as a measure of growth light environment based on 1-min-interval PPFD from the 1st to 30th June, 2007. Average daily peak PPFD in June at each leaf position (PPFD max ) was estimated using the daily five highest values during June 3 .

Leaf area index (LAI), and light gradients within the canopy. We used the relationship between LAI
and light gradients within the canopy of an inventory plot (50 × 50 m) of the experimental forest after Utsugi et al. 48   were conducted under an ambient CO 2 concentration of 360 µmol mol −1 and at various PPFD (0, 50, 100, 200, 300, 600, 1,000, 1,500, and 2,000 µmol m −2 s −1 ) provided by a red/blue LED array (Li-6400-40, Li-Cor), with blue light comprising 10% of the total PPFD. Leaf temperature was ≈27 °C during the measurements. After the measurements, the leaves were sampled and used for determination of N L by the combustion method using an analysis system composed of a N/C determination unit (SUMIGRAPH, NC 800, Sumika Chem. Anal. Service, Osaka, Japan), a gas chromatograph (GC 8 A, Shimadzu, Kyoto, Japan), and a data processor (Chromatopac, C R6A, Shimadzu). Electron transport rate (ETR) was calculated as ETR = Φ PSII × leaf absorptance × light intensity × 0.5 22,49 . Leaf absorptance was calculated from a calibration curve between SPAD readings (measured with a SPAD chlorophyll meter, SPAD 502, Minolta, Osaka, Japan) and leaf absorptance 3 . The responses of electron transport to incident irradiance were fitted by the convexity equations 50 as below: where φ is the initial slope (maximum quantum yield), θ the convexity of the curve, and ETR max the maximum rate of electron transport. The responses of Fv′/Fm′ and (1 − qP) Fv′/Fm′ to incident irradiance for each leaf were fitted as follows: Analogous to ETR, thermal dissipation (D) and excess energy production (E) were estimated from (1 − Fv′/ Fm′) × leaf absorptance × light intensity × 0.5 and ((1 − qP) Fv′/Fm′) × leaf absorptance × light intensity × 0.5, respectively 22,24 . Each coefficient in Eqs 2-4 was fitted as a function of N F defined as (N L -N b ) of the leaves used for the chlorophyll fluorescence measurements, where N b is the x-intercept of the maximum rate of photosynthesis (A max ) − N L relationship, which can be regarded as non-photosynthetic leaf structural N content. Conversely, N F can be regarded as photosynthetically-functional leaf N content 6,51 . A max was estimated from the light response of A n based on the convexity equations 50 , as below: where R d is the dark respiration rate, φ the initial slope (maximum quantum yield), and θ is the convexity of the curve. The relationships between N F and photosynthesis-related coefficients are summarized in Supplemental Fig. S1. Based on the coefficients of light responses in ETR, Fv′/Fm′, and (1 − qP) Fv′/Fm′ as a function of N F , we extrapolated the light responses in leaves that have various N L content with optimal and uniform N L distribution.
Actual, optimal and uniform leaf N distribution. We derived the light extinction coefficient of the canopy, K L 52 as follows: where F is the LAI cumulated from the canopy top, and I and I 0 are PPFD at F and the canopy top, respectively. Actual N L distribution at depth F in the canopy is also described by an exponential function of F with a photosynthetic leaf N distribution coefficient (K n ): where N 0 is the N L of the leaves at the canopy top. Optimal N L distribution was derived after Anten et al. 6 as follows: where N TF is the total free nitrogen in the canopy and N T and F T are the total leaf N pool and LAI per unit ground area, respectively. N T is calculated based on the integration of Eq. 2 11 . A N L distribution is defined as uniform when all leaves have the same N L irrespective of leaf position. Therefore, N L with uniform distribution is determined as: L T T Based on the estimated N L with optimal and uniform distribution, we calculated ETR, D, and E under the maximum PPFD during sunflecks for each leaf by using the N L -dependent photosynthetic coefficients (Supplemental Fig. S1) and Eqs 2-4. We hypothesized that canopy N might be distributed to maintain E constant throughout the canopy during sunflecks, thus to keep an efficient light utilization capacity by preventing photoinhibition for all leaves across sun and shade environments. Therefore, we estimated the leaf N distribution coefficient (K n ) that minimized the variance in E at PPFD max across the different leaf types within the canopy, using Eq. 8 with various K n (0.1 to 1.0 with an interval of 0.1) instead of K L as follows: Statistical analysis. We expediently classified the leaves into four types based on their growth light environments: (1) deep shade leaves (0 < Q int < 3 mol m −2 day −1 , n = 5); (2) moderate shade leaves (3 < Q int < 15 mol m −2 day −1 , n = 4); (3) moderate sun leaves (15 < Q int < 25 mol m −2 day −1 , n = 3); and (4) typical sun leaves (25 < Q int < 35 mol m −2 day −1 , n = 3). One-way analysis of variance (ANOVA) was used to test the differences among leaf types in ETR, D, and E with actual, optimal, or uniform N L distribution, and the ratio of ETR, D, and E with optimal or uniform to actual N L distribution 53 . Furthermore, when the ANOVA returned an overall significant effect of leaf type, the means were further tested using Holm's pairwise comparisons. The threshold for statistical significance was predefined at an alpha level of 0.05.
Data availability. All data used in this manuscript are present in the manuscript and its supplementary information.