Effects of soil nitrogen (N) deficiency on photosynthetic N-use efficiency in N-fixing and non-N-fixing tree seedlings in subtropical China

Soil nitrogen (N) deficiencies can affect the photosynthetic N-use efficiency (PNUE), mesophyll conductance (gm), and leaf N allocation. However, lack of information about how these physiological characteristics in N-fixing trees could be affected by soil N deficiency and the difference between N-fixing and non-N-fixing trees. In this study, we chose seedlings of two N-fixing (Dalbergia odorifera and Erythrophleum fordii) and two non-N-fixing trees (Castanopsis hystrix and Betula alnoides) as study objects, and we conducted a pot experiment with three levels of soil N treatments (high nitrogen, set as Control; medium nitrogen, MN; and low nitrogen, LN). Our results showed that soil N deficiency significantly decreased the leaf N concentration and photosynthesis ability of the two non-N-fixing trees, but it had less influence on two N-fixing trees. The LN treatment had lower gm in D. odorifera and lower leaf N allocated to Rubisco (PR), leaf N allocated to bioenergetics (PB), and gm in B. alnoides, eventually resulting in low PNUE values. Our findings suggested that the D. odorifera and E. fordii seedlings could grow well in N-deficient soil, and adding N may increase the growth rates of B. alnoides and C. hystrix seedlings and promote the growth of artificial forests.

www.nature.com/scientificreports www.nature.com/scientificreports/ treatment (Fig. 4). The V cmax and J max of the two N-fixing tree seedlings were less affected by the soil N deficiency (More details see Supplementary Table S3).
Effects of soil N treatments on leaf N allocation proportion of the Rubisco (P R ), bioenergetics (P B ), light-harvesting components (P L ), photosynthetic system (P p ), cell wall (P CW ), and other parts (P Other ). The P R , P B , P P , and P CW values of C. hystrix were higher than the corresponding values obtained for the other three species under any soil N treatments (Fig. 5). No significant change was observed in the P R , P B , P L , P P , and P Other values of D. odorifera under any N treatment; the P CW of D. odorifera in the LN treatment was 71.4% higher than that in the Control treatment. No significant change was observed in the P R , P B , P P , P CW , and P Other values of E. fordii under any N treatments, and the P L of E. fordii was 33.3% higher in the LN treatment than in the Control treatment. The LN treatment significantly decreased the P B (−28.6%) and P Other (−41.2%), and it increased the P CW (+66.7%) of C. hystrix when compared with the corresponding values obtained under the Control conditions. The LN treatment significantly decreased the P R (−38.5%), P B (−42.9%), P L (−33.3%), Figure 1. Light-saturated photosynthesis (A max ′), leaf N content per area (N area ), leaf N content per mass (N mass ), leaf mass per area (LMA), and photosynthetic-N use efficiency (PNUE) in the seedling leaves from the four studied tree species after exposure to different soil nitrogen (N) treatments. The statistical differences between each characteristic of the different species under three N treatments (mean ± SE) are the results of a one-way analysis of variance (ANOVA) (n = 7). The lowercase letters indicate significant differences at the 0.05 level between different N treatments, and the uppercase letters indicate significant differences at the 0.05 level between the species under the same N treatment. Control, high N; MN, medium N; and LN, low N.

Figure 2.
Regression analysis of the leaf nitrogen (N) concentration per area (N area ) and light-saturated photosynthesis (A max ′) of the seedling leaves from the four studied tree species. The determination coefficients (R 2 ) and P-values are shown. The lines fitted for N-fixing and non-N-fixing trees are significantly different (P < 0.05) according to the result of a one-way analysis of covariance (ANCOVA) with A max ′ as a dependent variable, whether it could fix N as a fixed factor, and N area as a covariate Relationships between parameters. The P R , P B , and P P values showed a significant positive correlation with the PNUE in these tree seedling leaves (P < 0.01; Fig. 6a,b,d). No significant correlation was observed between the P L and PNUE in these trees (Fig. 6c). Significant positive relationships were observed between the g m and PNUE in these tree seedling leaves (P ≤ 0.001; Fig. 7). The changes in P R , P B , and g m were important physiological factors influencing the PNUE.
Significant negative relationships were found between the P CW and g m in D. odorifera, E. fordii, and C. hystrix (P < 0.001; Fig. 8a,c,d); no significant relationships were observed in B. alnoides (Fig. 8b). Significant positive relationships were observed between P CW and C i -C c in D. odorifera (P = 0.002; Fig. 9a). Significant negative relationships were noted between the P CW and C i -C c in E. fordii (P = 0.004; Fig. 9b), and no significant relationships were observed in C. hystrix and B. alnoides (Fig. 9c,d). The improved P CW in D. odorifera might relate to its thicker cell walls, but in E. fordii, it might relate to the higher cell wall density. . Stomatal conductance (g s ), mesophyll conductance (g m ), CO 2 concentration in substomatal cavities (C i ), CO 2 concentration at the carboxylation site (C c ), and C i -C c in the seedling leaves of the four tree species after exposure to different soil nitrogen (N) treatments. The statistical differences between each characteristic of the different species under the three N treatments (mean ± SE) are the results of a one-way analysis of variance (ANOVA) (n = 7). The CO 2 conductance data were measured under light saturated conditions, and the leaf chamber CO 2 concentration was 380 μmol mol −1 . The lowercase letters indicate significant differences at the 0.05 level between different N treatments, and the uppercase letters indicate significant differences at the 0.05 level between the species under the same N treatment. Control, high N; MN, medium N; and LN, low N. . Maximum carboxylation rate (V cmax ) and maximum electron transport rate (J max ) in the seedling leaves of the four tree species after exposure to different soil nitrogen (N) treatments. The statistical differences between each characteristic of the different species under the three N treatments (mean ± SE) are the results of a one-way analysis of variance (ANOVA) (n = 7). The lowercase letters indicate significant differences at the 0.05 level between different N treatments, and the uppercase letters indicate significant differences at the 0.05 level between the species under the same N treatment. Control, high N; MN, medium N; and LN, low N. www.nature.com/scientificreports www.nature.com/scientificreports/ No significant relationships were observed between the P CW and P R in D. odorifera and E. fordii, but significant negative relationships were observed in B. alnoides and C. hystrix (P ≤ 0.002). The cell wall N might influence the variation in N in the Rubisco, thus influencing the photosynthetic capacity in these two non-N-fixing tree seedlings. A regression analysis of the P CW with P R in the B. alnoides seedling leaves under the LN treatment was obtained within the shaded zone. Most Control and MN treatment parameters for B. alnoides and C. hystrix . Nitrogen (N) allocation proportion of the Rubisco (P R ), bioenergetics (P B ), light-harvesting components (P L ), photosynthetic system (P P ), cell wall (P CW ), and other parts (P Other ) in the seedling leaves of the four tree species following exposure to different soil N treatments. The statistical differences between each characteristic of the different species under three N treatments (mean ± SE) are the results of a one-way analysis of variance (ANOVA) (n = 7). The lowercase letters indicate significant differences at the 0.05 level between different N treatments, and the uppercase letters indicate significant differences at the 0.05 level between the species under the same N treatment. Control, high N; MN, medium N; and LN, low N. Figure 6. Regression analysis of nitrogen (N) allocation proportions in the photosynthetic system (P P ), lightharvesting components (P L ), Rubisco (P R ), and bioenergetics (P B ) with the photosynthetic N use efficiency (PNUE) in the seedling leaves of the four tree species after exposure to different soil N treatments. The determination coefficients (R 2 ) and P-values are shown. The lines fitted for the N-fixing and non-N-fixing trees are significantly different (P < 0.05) according to the results of a one-way analysis of covariance (ANCOVA) with the PNUE as a dependent variable, whether it could fix nitrogen as a fixed factor, and P P , P R , P B , and P L as covariates.
www.nature.com/scientificreports www.nature.com/scientificreports/ were in the shaded zone, and D. odorifera and E. fordii were found under the shaded zone (Fig. 10). Low soil N increased the competition between the Rubisco and cell wall N.

Discussion
The leaf N contents of two non-N-fixing tree seedlings, B. alnoides and C. hystrix, were significantly affected by the soil N content (Fig. 1, Supplementary Table S5), which was consistent with previously published studies 1,11,12,27 . However, the leaf N content of E. fordii was not significantly affected by the soil N content. This finding might be due to its strong N fixation capacity and its maintenance of the N content stability in leaves. Different N treatments significantly affected the N mass of D. odorifera seedling leaves, but the N area of D. odorifera was not affected by the soil N content (Fig. 1). Because the N area was influenced by the N mass and LMA, the LMA of D. odorifera changed with the soil N gradient (Fig. 1); the maintenance of the N area at a steady state showed good leaf morphological plasticity. The low soil N content decreased the A max ′ in D. odorifera, B. alnoides, and C. hystrix ( Fig. 1) for different reasons. In D. odorifera, the low soil N content primarily decreased its C c (Fig. 3), which is one of the important raw materials for photosynthesis 44 , and the CO 2 partial pressure is important for Rubisco activity because O 2 is a competitive inhibitor of the C assimilatory reaction of Rubisco for promoting the Rubisco  www.nature.com/scientificreports www.nature.com/scientificreports/ oxidation reaction 12 . For the two non-N-fixing tree seedlings, the low soil N content decreased their V cmax and J max values (Fig. 4), which are the key biochemical parameters of the photosynthetic capacity 14,45 .
The fraction of the total leaf N allocated to the photosynthetic apparatus 46 , especially to Rubisco and bioenergetics, could influence the variation in the PNUE 1,3,16 . The g m could also influence the PNUE 32,47 by affecting the C c 11,12 . In this study, the P R and P B showed a significant positive correlation with the PNUE (P < 0.001, Fig. 6a,b), and the g m significantly affected the PNUE in the seedling leaves of the four studied tree species (Fig. 7), although the effect of the g m on the PNUE was different among the species 48 . The LN treatment significantly decreased the g m in D. odorifera and the P R , P B , and g m in B. alnoides (Figs 1 and 5), leading to lower PNUEs in the LN treatment. It has been reported that low soil N could decrease the g m 12,49 and N allocation 3,29 . However, Chen et al. (2014) found an improvement in the P R and P B of female Populus cathayana with improved soil N, but the P R and P B of the males decreased 1 . Warren (2004) also found that an improvement in the soil N could decrease the P R in Eucalyptus globulus. Some plants might have a different strategy for adapting to the soil N 11 .  www.nature.com/scientificreports www.nature.com/scientificreports/ The PNUEs of the two non-N-fixing tree seedlings were significantly higher than those of the two N-fixing tree seedlings under any soil treatment (Fig. 1, Supplementary Table S5), which was first attributed to their relatively low N area and N mass (Figs 1, 2, Table S7). The N-fixing species, which could gain N from air through legume bacteria, usually have a higher leaf N content than the non-N-fixing species 43,50 . High P R and P B (Fig. 4, Supplementary Table S5) were the primary biochemical factors leading to their higher PNUEs. These results were also consistent with other studies [40][41][42] . The leaves are the photosynthetic organs of plants, and plants have roughly two survival strategies, namely, quick investment-return and slow investment-return 51 . Two N-fixing trees might belong to the slow investment-return species and use a different strategy to use N, such as compensation for their low productivity through a long leaf lifespan 18 and storing N for other processes, such as reproduction 1 . Two N-fixing tree seedlings might grow well in N-deficient soil, and applying N could increase the growth rates of the two non-N-fixing tree seedlings and promote the growth of artificial forests. Of course, some N-fixing trees have the same N utilization and distribution strategies as non-N-fixing trees, such as Acacia mangium 52 .
A decrease was observed in the g m of the D. odorifera, C. hystrix, and B. alnoides seedlings under the LN treatment, but the reasons for this decline were different. The changes in A max ′ or C i -C c could influence the value of g m . In these tree seedlings, the A max ′ decreased under the LN treatment, but the changes in the C i -C c were different. D. odorifera and C. hystrix showed an increased C i -C c in the LN treatment, but B. alnoides showed no change in its C i -C c value (Fig. 3). After entering through the stomata, the CO 2 diffuses through air spaces, cell walls, cytosol, and chloroplast envelopes and finally reaches the chloroplast stroma, where it is fixed by Rubisco 26,53 . Generally, cell walls account for >50% of the total cell CO 2 diffusion resistance and a variable proportion of respiration 26 . D. odorifera, C. hystrix, and B. alnoides showed improved P CW values in the LN treatment (Fig. 5). Mu et al. (2016) also found an increase in the P CW of maize growing under low-N stress 29 . D. odorifera showed no significant reduction in its N area in the LN treatment, and thus there was an increase in the N contents in the cell wall (Q CWarea ) of D. odorifera (+62.4%, Supplementary Table S6). The percentage of N in the cell wall showed a slight variation in the same species 16 . An improvement in the N CW of D. odorifera under the LN treatment indicates the high dry mass of the cell wall, resulting in improved LMA 16,54 , and it might improve the thickness of the cell wall, thereby improving its C i -C c value 16 . However, B. alnoides and C. hystrix showed a reduction in their N area values in the LN treatment, leading to a smaller change in the Q CWarea (+5.9% and +29.6%, respectively, Supplementary  Table S6). Thus, there were no significant changes in their LMA and C i -C c values. An improvement in the P CW of D. odorifera therefore significantly decreased its C i -C c and g m , and no significant relationship was observed between the P CW and C i -C c in B. alnoides and C. hystrix (Figs 8, 9).
The P CW did not influence the variation in the C i -C c , but it showed a significant negative correlation with the g m in two non-N-fixing trees (Fig. 8) 20,21 . B. alnoides and C. hystrix showed high P R and P CW values (Fig. 5), and a part of the distribution area in or on the shade zone ( Fig. 10; for a further explanation of the shade zone, please see Harrison et al. 20 ) indicates that the free amino acid, NO 3 − , and NH 4 + in the leaves were not sufficient (appearing as low P Other ) to supply N to both Rubisco and the cell wall 20 , which explained the existence of a trade-off between the P R and P CW (Fig. 10). It is important to note that the regression analysis of the P CW with the P R in the B. alnoides seedling leaves exposed to the LN treatment was found in the shaded zone; most Control and MN treatments of B. alnoides and C. hystrix were in the shaded zone (Fig. 10). Low soil N increased the competition between the Rubisco and cell wall N.
The two non-N-fixing tree seedling leaves showed improved P CW , and the D. odorifera seedling leaves improved both the LMA and P CW values under the LN treatment (Figs 1, 5). The LMA is the product of leaf thickness and density, and it is positively correlated with leaf toughness 55 and is a fundamental defensive trait of plants 56,57 . The cell wall also directly functions as a defense organ 58 . We observed that the N mass values of these trees were affected by the soil N content (Fig. 1). Low nutrient availability limits the growth rate of seedlings and might damage the seedlings during the growing season 59 . The LN treatment might pose a threat to these seedling leaves; thus, plants need to have tougher leaves to survive 16 , as shown by the relatively high P CW and LMA in D. odorifera seedling leaves and high P CW in the B. alnoides and C. hystrix seedling leaves. Givnish (2002) hypothesized that soil fertility is the primary driver of the leaf lifespan 60 , and a high LMA leads to a long leaf lifespan 51 . Therefore, an improvement in the LMA might also increase the leaf lifespan of D. odorifera seedling leaves, ultimately maximizing the carbon assimilation per unit of nutrient over the lifespan of the leaf 61-63 . Different species have different response characteristics to the soil N conditions.
To understand the changes in the various parameters under low soil N in the four species, we drew a process diagram (Fig. 11). Generally, we found fewer parameter changes in the two N-fixing tree seedlings and more parameter changes in the two non-N-fixing tree seedlings. The physiological and ecological characteristics of these two N-fixing tree seedlings are more stable, and these two N-fixing tree seedlings could be good tree species for afforestation in N-poor areas. We also performed Between-Subjects effects tests on the tree varieties and N treatments for the variables in the four species (Supplementary Table S8). In general, varieties of the trees were more important than the N treatment interaction effect, but the N treatment interaction effect was more important in influencing the A max ′ and g m . More trees and more variables must be further studied. (

Conclusions
In revisiting our questions, we concluded that (1) soil N deficiency significantly decreased the leaf N concentration and photosynthesis ability in two non-N-fixing trees, but it had less influence on these indices in the two N-fixing trees. (2) The LN treatment had a lower g m in D. odorifera and had lower P R , P B and g m in B. alnoides, eventually resulting in their low PNUE values. (3) D. odorifera, B. alnoides, and C. hystrix seedling leaves showed improved P CW and (or) LMA to adapt to a low-N soil environment. These findings were important for understanding the ecophysiological changes in plants under low soil N conditions. Our findings suggested that the two N-fixing tree seedlings could grow well in N-deficient soil, and they could be good tree species for the afforestation of N-poor areas. Adding N may increase the growth rates for the two non-N-fixing tree seedlings and promote the growth of artificial forests. Because these species live in the same area, it is possible to mix non-N-fixing with N-fixing tree seedlings for afforestation, and mix N-fixing trees in non-N-fixing pure forest after intermediate cutting or selective cutting in non-N-fixing pure forest, which could improve soil N utilization efficiency.

Materials and Methods
Study area and plant material. This study was performed in the Experimental Center of Tropical Forestry (22°7′19″-22°7′22″N, 106°44′40″-106°44′44″E) at the Chinese Academy of Forestry located in Pingxiang, Guangxi Province, China. This location has a subtropical monsoon climate with distinct dry and wet periods, and the mean annual temperature is 21 °C. The mean monthly minimum and maximum temperatures are 12.1 and 26.3 °C, respectively. The mean annual precipitation, which takes place primarily from April to September, is 1400 mm. The active accumulated temperature above 10 °C is 6000-7600 °C. The total annual sunshine duration is 1419 hours 64,65 .
The seeds of D. odorifera, E. fordii, and C. hystrix were collected separately from the mother trees, and the B. alnoides seedlings were somaclones. The D. odorifera, E. fordii, and C. hystrix seeds were germinated in a seedbed in February of 2014, and B. alnoides was budding at the same time. When the seedlings were approximately 20 cm tall, 90 similarly sized seedlings per species were transplanted to pots (5.4 L, filled with washed river sand) and established in an open site at the Experimental Center of Tropical Forestry in March, 2014.
From April to June, three levels of soil N treatments were set up (Hyponex M. Scott & Sons, Marysville, OH, USA, dissolved in the water from the aqueous solution preparation). Nitrogen fertilizer was applied ten times, once per week. A total of 0.2 (low nitrogen, LN), 0.7 (medium nitrogen, MN), and 1.5 g (set as Control) of available N were applied per pot, with each treatment including 30 seedlings per species. The forms of N that were applied in this study were mixed N (both NH 4 + and NO 3 − ), and the NH 4 + to NO 3 − ratio was 1:1. We chose these forms because we used washed river sand as a culture substrate with a pH value of approximately 7, and only using NH 4 + or NO 3 − might cause the soil to become more acid or alkaline, respectively, affecting the plant growth. Wu et al. (2012) found that the proper amounts of N applications for D. odorifera seedlings were 1.74-2.15 g N per pot 66 . Li et al. (2003) found that the appropriate N applications for E. fordii seedlings were approximately 1.39-1.86 g N per pot 67 . Although the purpose of this research is to understand the effects of soil N deficiency on plant metabolism, we also want to explore the plant physiological process from a comparatively www.nature.com/scientificreports www.nature.com/scientificreports/ sufficient to a lack of soil N, because non-N-fixing woody species might be more sensitive to changes in the soil N gradient, and the different ecophysiological processes between a comparatively sufficient to a lack of soil N could help us to understand the effects of soil N deficiency on plant metabolism. Therefore, we set up a high N treatment as Control. The seedlings in each treatment were watered every day to keep the soil moist. Natural light (100% light in the field) was used for illumination.

Determination of gas exchange parameters.
Fifteen days after the last N fertilization, on sunny days from 9:00 to 11:00 h in July and August of 2014, seven healthy and similarly sized seedlings were chosen per treatment, per species. One healthy and mature leaf per seedling that was exposed to the sun was chosen to determine the gas exchange parameters. These parameters were determined with a LiCor-6400 portable photosynthesis system (LI-COR, Lincoln Nebraska, USA), and the photosynthetic response to the photosynthetic photon flux density (PPFD, µmol m −2 s −1 ) and C i (μmol mol −1 ) were determined. Under 380 μmol mol −1 of leaf chamber CO 2 concentration (the average air CO 2 concentration in the day time), the photosynthetic rates were measured under photon flux densities of 1500, 1200, 1000, 800, 600, 400, 200, 150, 100, 80, 50, 30, 20, 10 and 0 μmol m −2 s −1 . Under a saturated PPFD, the photosynthetic rates were detected using the same leaf-under leaf chamber CO 2 concentrations of 380, 200, 150, 100, 80, 50, 380, 600, 800, 1000, 1200, 1500, 1800 and 2000 μmol mol - 1 28,47 . We started at a 380 μmol mol −1 concentration because this is the average air CO 2 concentration during the day time that could reduce the plant activation time 28 . The relative humidity of the air in the leaf chamber was maintained at 60-70%, and the leaf temperature was set to 30 °C. The values for the following data or parameters were determined: the net photosynthetic rate (A n , μmol m −2 s −1 ), A max ′ (μmol m −2 s −1 ), g s (mol CO 2 m −2 s −1 ), and dark respiration (R n , μmol m −2 s −1 ). The light-and CO 2 -saturated net CO 2 assimilation rate (A max , μmol m −2 s −1 ) was calculated according to Farquhar et al. 14 . The relative humidity of the air in the leaf chamber was maintained at 60-70%, and the leaf temperature was set to 30 °C.
Determination of the chlorophyll fluorescence, mesophyll conductance, V cmax , and J max . The fluorescence yield was measured using a LiCor-6400 leaf chamber fluorometer (6400-40, LI-COR, Lincoln, Nebraska, USA) on the same leaf and with seven repetitions for each species. The chamber relative humidity and leaf temperature were controlled under the same conditions as described in the gas exchange parameters. The leaf chamber CO 2 concentration was set to 380 μmol mol −1 . The fluorescence yield (ΔF/F m ′) was subsequently determined. The photosynthetic electron transport rate (J f , μmol m −2 s −1 ) was calculated according to the equation described by Loreto et al. 68 as follows: where PPFD is the photosynthetic photon flux density; Leafreflu is the leaf absorptance valued between 0.82-0.85 69 (we used 0.85 in this paper); and PARDistPhotosys is the fraction of quanta absorbed by photosystem II (valued as 0.5) 68 . The mesophyll conductance (g m , mol CO 2 m −2 s −1 ) was calculated using three different methods to obtain a more accurate value. The variable J method was described by Harley et al. 70 , and it has been commonly used in recent years [71][72][73] . The A-C i curve fitting method was described by Ethier and Livingston 74 , and Sharkey et al. 75 developed a software package to estimate the g m and other parameters based on this method. The exhaustive dual optimization (EDO) method described by Gu et al. 76 could estimate up to eight parameters, including the g m , and we obtained an automated analysis of A-C i curves through a website (http://www.leafweb.org) by uploading our data to determine the value of the g m . Subsequently, the g m calculated by these three methods was used to calculate C c (μmol mol −1 ) as follows: The C c and g m calculated using the three methods are shown in Supplementary Table S9. The mean value of C c was used to fit the A n -C c curve, followed by the calculation of V cmax (μmol m −2 s −1 ) according to Farquhar et al. 14 and the J max (μmol m −2 s −1 ) according to Loustau et al. 77 . The running fitting model used in the in vivo Rubisco kinetics parameters (i.e., K o , K c , and their activation energy) was measured according to Niinemets and Tenhunen 13 .
Determination of additional leaf traits. After the gas exchange parameters and fluorescence yield were determined, the leaf samples and nearby leaves (30-50 leaves per seedling in total, the sizes of which were similar to those of the leaves used to determine the photosynthesis, healthy and mature characteristics, and sun-exposed parameters) were collected from each pot. The surface areas of 10-20 leaves were measured using a scanner (Perfection v700 Photo, Epson, Nagano-ken, Japan). The leaves were subsequently oven-dried to a constant weight at 80 °C for 48 h. The dry weight was measured using an analytic balance, and then the LMA (g m −2 ) was calculated. The dried leaf samples were ground into dry flour. The organic carbon (C) concentration was determined by potassium dichromate-sulfuric acid oxidation method (C mass mg g −1 , Supplementary Table S10). The N concentration was determined using a VELP automatic Kjeldahl N determination apparatus (UDK-139, Milano, Italy), and then the N mass (mg g −1 ) and N area (g m −2 ) values were calculated. Then, PNUE (μmol mol −1 s −1 ) was calculated using the following formula: www.nature.com/scientificreports www.nature.com/scientificreports/ where 14 is the atomic mass of nitrogen.
The remaining 20-30 leaves were frozen and kept for laboratory analysis. The frozen leaves (0.2 g, 5-10 leaves) were cut into small 5-10-mg pieces. The leaves were placed in a volumetric flask and brought to a consistenttant volume of 25 mL using 95% (v/v) alcohol. The volumetric flask was protected from light for 24 h, and then the chlorophyll contents were determined using a Shimadzu ultraviolet-visible spectrophotometer (UV 2250, Fukuoka, Japan). For the chlorophyll contents, please see Supplementary Table S10.
The remaining frozen leaves were used to determine the cell wall N content according to the method of Onoda et al. 16 as follows: 1 g of leaves was powdered in liquid N and suspended in sodium phosphate buffer (pH 7.5, 25 mL), the homogenate was centrifuged at 2500 g for 5 min, and the supernatant was discarded. The pellet was washed with 3% (w/v) SDS, amyloglucosidase (35 units ml −1 , Rhizopus mold, Sigma, St Louis, USA) and 0.2 M KOH and then heated and centrifuged, and the remaining pellet was washed with distilled water and ethanol and then dried in an oven (75 °C) for 2 days (for more details see Onoda et al.) 16 . The nitrogen content of the rest of the pellet (cell wall N) was determined using a VELP automatic Kjeldahl N determination apparatus. The P CW represents the ratio of the cell wall N content to the total N content.
Calculation of the N allocation in the photosynthetic apparatus. The N allocation fractions of each component in the photosynthetic apparatus were calculated according to Niinemets and Tenhunen 13 , which has been widely used in recent years 1,45,78 . where C Chl is the chlorophyll concentration (mmol g −1 ), V cr is the specific activity of Rubisco (μmol CO 2 g −1 Rubisco s −1 ), J mc is the potential rate of photosynthetic electron transport (μmol electrons μmol −1 Cyt f s −1 ), and C B is the ratio of leaf chlorophyll to leaf N during light-harvesting (mmol Chl (g N) −1 ). The V cr , J mc , and C B were calculated according to Niinemets and Tenhunen 13 .
Statistical analysis. The differences between the seedling leaves of the four tree species, the N-fixing and non-N-fixing tree seedlings, and the three levels of soil N were analyzed by performing a one-way analysis of variance (ANOVA), and a post-hoc test (Tukey's test) was conducted to determine if the differences were significant. The effects of the tree varieties and N treatments on the variables in the four species were analyzed by two-way ANOVA and Tukey's test. The significance of the correlation between each pair of variables was tested with a Pearson's correlation (two-tailed). All the analyses were performed using the Statistical Product and Service Solutions 17.0 program (SPSS17.0, Chicago, USA).