Photosynthesis of subtropical forest species from different successional status in relation to foliar nutrients and phosphorus fractions.

The ecophysiological linkages of leaf nutrients to photosynthesis in subtropical forests along succession remain elusive. We measured photosynthetic parameters (Amax, Vcmax, Jmax, PPUE), leaf phosphorus (P) and nitrogen (N), foliar P fractions and LMA from 24 species (pioneer, generalist, and climax). Amax was significantly related to N and P for the pooled data, while significant relationship between Amax and P was only found in climax species. The mixed-effect model including variables (N, P, and SLA or LMA) for predicting Vcmax and Jmax best fitted but varied remarkably across succession. Climax species had higher N: P ratios, indicating an increasing P limitation at later succession stage; photosynthesis, however, did not show stronger P than N limitations across all species. Nevertheless, climax species appeared to increase nucleic acid P allocation and residual P utilization for growth, thereby reducing the overall demand for P. Our results indicate that the scaling of photosynthesis with other functional traits could not be uniform across succession, growth variables (e.g. photosynthesis) and species trait identity (e.g. successional strategy) should be considered in combination with N: P ratio when we investigate P limitation in subtropical forests, and variations in P allocation state further influencing photosynthetic rates and P-use efficiency.

Introduction

Nutrient limitation to primary productivity is widespread in most terrestrial ecosystems globally, and low levels of pan class="Chemical">pan class="Chemical">nitrogenclass="Chemical">pan> (N) and class="Chemical">papan>n class="Chemical">phosphorus (P) commonly limit or co-limit plant growth and rates of photosynthesis[1,2]. P limitation generally occurs in lowland tropical forests, is class="Chemical">particularly strong for sites in Panama and the Amazon basin[3], while N limitation often occurs in temperate and boreal regions[4-6]. Leaf N:P ratio in terrestrial plants generally serves as a simple and useful indicator of nutrient limitation to primary productivity[7-9]. In forest ecosystems, leaf N:P ratios >16[10] or 20[8] usually indicate P limitation. However, P limitation has received comclass="Chemical">paratively less attention than N limitation, few studies have comclass="Chemical">pared the limitations of leaf P and N concentration on photosynthesis of tropical forest species[11].

As we all know, the maximum net assimilation rate (Amax) is strongly affected by various leaf traits, for example, leaf thickness[12,13], leaf mass per area (class="Chemical">pan class="Chemical">LMAclass="Chemical">pan>)[14-16], and leaf nutrient concentration[15,17,18]. A number of studies have explored how low leaf nutrient concentrations affect leaf photosynthetic caclass="Chemical">pacity in the tropics, particularly for P[6,19-22]. Phosphorus limitation might be manifested in limiting ribulose-1,5-bisphosphate (n class="Chemical">RuBP) regeneration as the underlying control over Amax in leaves[23,24]. Previous studies in nutrient-poor ecosystems have shown N limitation can induce larger proportion of leaf N allocated into cell walls, which increases class="Chemical">pan class="Chemical">LMA; an increase LMA reduces the photosynthetic N-use efficiency (PNUE) by decreasing Amax[25,26]. However, the ecophysiological linkages between P and Amax and the underling mechanisms of P limitation are still poorly known[27].

In the photosynthesis model proposed by Farquhar et al.[28], ‘the maximum carboxylation rate’ (Vcmax) and ‘the maximum electron transport rate’ (Jmax) were used to express photosynthetic caclass="Chemical">pacity, which are generally class="Chemical">positively related to N, class="Chemical">pan class="Chemical">P and specific leaf area in tree leaves[11,21,29]. Various studies have shown that leaf photosynthetic characteristics correspond with successional status[30-32]. The early successional (pioneer) species, which are usually fast-growing and light-demanding, make a greater fractional investment in leaf traits that maximize photosynthetic capacity than late successional (climax) species[30,31,33]. In contrast, leaves of climax species are often have a longer lifespan, higher LMA, chlorophyll to N ratios and lower photosynthetic capacity than leaves of pioneer species[33,34]. As a result, the light-demanding pioneer species are gradually replaced by shade-tolerant climax species. Unlike pioneer species, which occur early in succession, and climax species, which occur late in succession, generalist species can occur throughout succession. What is less clear is how P limitation, which is common in tropical forests, might affect the relationships of photosynthetic capacity to N and LMA during succession in subtropical forests.

class="Chemical">Previous studies have shown that class="Chemical">plants generally reduce their foliar class="Chemical">pan class="Chemical">P concentration in response to low P availability in tropical soils[35,36]. Kedrowski[37] developed a successful fractionalization scheme based on differential solubility or hydrolysis for the extraction and analysis of various P-containing fractions from plant material. Using a trichloroacetic acid (TCA) extraction method, Close and Beadle[38] reported differences in the concentrations of insoluble P and inorganic P among plant species. More recently, Hidaka and Kitayama[39] divided foliar P into four fractions: structural P (lipid P, phospholipids of membranes), metabolic P (including Pi and easily soluble P-containing metabolites), nucleic acid P (RNA and DNA), and residual P (phosphoproteins and unidentified residue). It has been shown that tree species with high photosynthetic P-use efficiency (PPUE) on P-poor sites reduce the demand for foliar P by reducing concentrations of both metabolic P and nucleic acid P[39,40]. However, how plants allocate P among foliar P fractions and how plants develop adaptive strategies to efficiently use P in subtropical region remain unclear.

class="Chemical">Plants are class="Chemical">probably class="Chemical">pan class="Chemical">P-limited in most soils of China, because of the low soil available P content[41]. Insufficient P has become the limiting factor of ecosystem primary productivity and other ecosystem processes in subtropical forests of China[42-44]; however, these studies mainly focused on plant N: P ratios, the real demand of leaf P for growth (e.g. photosynthesis) and its functional partitioning have not been concerned. To our knowledge, few data are available describing the photosynthetic parameters of the subtropical species. Here, we examined how leaf N, P, LMA (or SLA), and foliar P fractions affect photosynthetic performance along a subtropical forest succession. We tested the following hypotheses: (1) Average leaf trait values decrease with succession, pioneer species have higher mean P, N, and photosynthetic parameters (Amax, Vcmax, Jmax and PPUE) than generalist and climax species; (2) Leaf P has a stronger influence over photosynthetic capacity (Amax, Vcmax, and Jmax) than N, in particular at later succession stage; and (3) Foliar P fractions change substantially with succession in that climax species, unlike pioneer species, optimize the allocation of P among foliar P fractions in order to maintain their growth and to reduce the overall demand for P.

Results

Comparison of leaf traits among successional status

Leaf class="Chemical">P concentration of the class="Chemical">pioneer sclass="Chemical">pecies was significantly higher than that of the generalist and climax sclass="Chemical">pecies, on both an area and a mass basis (both class="Chemical">pan class="Chemical">P < 0.001; Table 1 and Supplementary Table S1). Pioneer species also exhibited the highest Na values, although no significant differences in Na were observed among successional groups (P = 0.08, Table 1). The pioneer and climax species leaves exhibited similar values for Nm; however, the climax species leaves had the highest N:P ratio (P < 0.001, Table 1). Given that a leaf N:P ratio >20 generally indicates P limitation as opposed to N limitation[8], P limitation was evident in 21% of pioneer species, 38% of generalist species, and 66% of climax species (Fig. 1). On the basis of both area and mass, all photosynthetic parameters (Amax, Vcmax, Jmax and PPUE) significantly differed among species with succession (P < 0.01; Table 1), i.e., these values were greater for pioneer species than for generalist and climax species. Jmax/Vcmax ratio did not significantly differ among successional groups (P = 0.525, Table 1).

Table 1

Area-based leaf traits for the subtropical forest species in three successional groups.

Group	Na	Pa	N:P ratio	LMA	A max,a	V cmax,a	J max,a	Jmax,a/Vcmax,a	PPUE
	(g m−2)	(g m−2)		(g m−2)	(µmol m−2 s−1)	(µmol m−2 s−1)	(µmol m−2 s−1)		(µmol mol−1 s−1)
Pioneer	1.65 ± 0.08 b	0.10 ± 0.01 c	18.28 ± 1.40 a	86.06 ± 8.73 b	12.43 ± 0.93 c	69.08 ± 4.86 b	88.72 ± 7.31 c	1.29 ± 0.05 a	5222.20 ± 520.82 b
Generalist	1.58 ± 0.05 ab	0.08 ± 0.01 b	19.99 ± 0.69 a	92.54 ± 2.37 b	9.92 ± 0.57 b	56.98 ± 3.11 a	73.42 ± 3.91 b	1.31 ± 0.04 a	5052.24 ± 367.63 b
Climax	1.46 ± 0.05 a	0.06 ± 0.01 a	23.92 ± 0.88 b	70.50 ± 2.48 a	6.54 ± 0.28 a	49.51 ± 2.00 a	60.84 ± 2.40 a	1.25 ± 0.03 a	3899.78 ± 166.75 a

Values shown are group averages (±SE). Means in a column followed by different letters are significantly different (P < 0.05). Na, leaf nitrogen concentration; Pa, leaf phosphorus concentration; leaf N:P ratio, leaf nitrogen to phosphorus ratio; LMA, leaf mass per area; Amax,a, maximum photosynthesis assimilate rate; Vcmax,a, maximum carboxylation velocity; Jmax,a, maximum electron transport rate; Jmax;a/Vcmax;a, ratio of maximum carboxylation velocity over maximum rate of electron transport; PPUE, photosynthetic P-use efficiency.

Figure 1

The relationship between area-based leaf P concentration (Pa) and leaf N concentration (Na). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Dashed lines represent the N:P ratios of 10 and 20. Points above the N:P = 20 indicate P limitation; points below the N:P = 10 line indicate N limitation; and points between the lines indicate co-limitation. Symbols: pioneer species (solid); generalist species (grey); climax species (open).

Values shown are group averages (±SE). Means in a column followed by different letters are significantly different (class="Chemical">P < 0.05). Na, leaf class="Chemical">pan class="Chemical">pan class="Chemical">nitrogen concentration; Pa, class="Chemical">paclass="Chemical">pan>n class="Chemical">leaf phosphorus concentration; leaf N:P ratio, leaf nitrogen to phosphorus ratio; LMA, leaf mass per area; Amax,a, maximum photosynthesis assimilate rate; Vcmax,a, maximum carboxylation velocity; Jmax,a, maximum electron transport rate; Jmax;a/Vcmax;a, ratio of maximum carboxylation velocity over maximum rate of electron transport; PPUE, photosynthetic P-use efficiency.

The relationship between area-based leaf class="Chemical">P concentration (class="Chemical">pan class="Chemical">Pa) and leaf N concentration (Na). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Dashed lines represent the N:P ratios of 10 and 20. Points above the N:P = 20 indicate P limitation; points below the N:P = 10 line indicate N limitation; and points between the lines indicate co-limitation. Symbols: pioneer species (solid); generalist species (grey); climax species (open).

Bivariate relationships between photosynthesis and leaf traits

For all data pooled, leaf Na and class="Chemical">Pa were highly and class="Chemical">positively correlated (r2 = 0.21; class="Chemical">pan class="Chemical">P < 0.001, Fig. 1). Across all 24 species, Na exhibited a weak, positive correlation with LMA (Supplementary Fig. S1a, r2 = 0.13; P < 0.01). SMA tests for common slopes revealed a significant difference among the different successional status (Supplementary Table S2); the y-axis intercept of the relationship was higher for climax species, indicating that the climax species might have a higher Na for a given LMA than pioneer species. Compared to Na, Pa showed a stronger relation with LMA (Supplementary Fig. S1b, r2 = 0.34; P < 0.001). Additionally, Pa for a given LMA was lower for climax species than pioneer species (Supplementary Table S2).

Bivariate relationships of Amax on N and class="Chemical">P (either area-based or mass-based) were highly significant for the class="Chemical">pooled data (Fig. 2, class="Chemical">pan class="Chemical">P < 0.01). When the area-based data were pooled, variations in Amax,a were slightly explained by variations in Pa (P < 0.001; r2 = 0.19) and Na (P = 0.001; r2 = 0.10). However, Supplementary Table S3 shows that when the data were grouped by successional status, area-based relationships were significant between Amax,a and Na for the pioneer species (P = 0.012; r2 = 0.32) and between Amax,a and Pa for the climax species (P = 0.001; r2 = 0.31). For all data pooled, LMA was weakly positively related to Amax,a (P < 0.05; r2 = 0.04), but this relationship was not significant for any of the three successional status (Supplementary Table S3). When the mass-based data were pooled, Amax,m was significantly and positively correlated with Pm (r2 = 0.26), Nm (r2 = 0.21), and negatively with LMA (r2 = 0.15) (all P < 0.001). Still, Amax,m was significantly correlated with Nm and LMA for all successional groups, while Amax,m and Pm were poorly related for pioneer or generalist species (Supplementary Table S3).

Figure 2

The log-log relationships between photosynthesis capacity (the maximum photosynthesis assimilate rate, Amax) and leaf traits including nitrogen concentration (N), phosphorus concentration (P), and leaf mass per area (LMA). Note that Amax values are based on measurements of mass in the upper panels and on area in the lower panels. Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Standardized major axis (SMA) regressions are presented in Table S3. Symbols are the same as in Fig. 1.

The log-log relationships between photosynthesis caclass="Chemical">pacity (the maximum class="Chemical">photosynthesis assimilate rate, Amax) and leaf traits including class="Chemical">pan class="Chemical">pan class="Chemical">nitrogen concentration (N), class="Chemical">paclass="Chemical">pan>n class="Chemical">phosphorus concentration (P), and leaf mass per area (LMA). Note that Amax values are based on measurements of mass in the upper panels and on area in the lower panels. Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Standardized major axis (SMA) regressions are presented in Table S3. Symbols are the same as in Fig. 1.

The regression of Jmax,a on Vcmax,a using data from all species suggested a very tight co-ordination between the two class="Chemical">parameters (class="Chemical">pan class="Chemical">P < 0.001, r2 = 0.74, Fig. 3). Variations in Jmax,a were strongly correlated with Vcmax,a all for the three successional groups (Supplementary Table S4). Across all species, Vcmax,a and Jmax,a were correlated with leaf traits, and the bivariate relationships were stronger with Pa (P < 0.001; r2 = 0.17 for Vcmax; r2 = 0.23 for Jmax) than with Na (P < 0.01; r2 = 0.13 for Vcmax; r2 = 0.18 for Jmax) (Fig. 4). No significant relationship was found between Vcmax,a (or Jmax,a) and N:P ratio (Fig. 4). Both Vcmax,a and Jmax,a were marginally correlated with LMA (P < 0.01; r2 = 0.07 and r2 = 0.06, respectively) (Fig. 3). Within the pioneer group, significant relationship of Vcmax,a was observed only with LMA (P = 0.021, r2 = 0. 27), no significant relations with Na, Pa, or N:P ratios (Supplementary Table S4). Within the generalist and climax group, Vcmax,a was positively related to Na (P = 0.002, r2 = 0.24 for generalist species; P = 0.050, r2 = 0.08 for climax species) but not with LMA, Pa, or N:P ratio (Supplementary Table S5). Similar patterns were observed for Jmax,a (Supplementary Table S4).

Figure 3

The relationship between area-based maximum carboxylation velocity (Vcmax,a) and area-based maximum electron transport rate (Jmax,a). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Standardized major axis (SMA) regressions are given in Table S4. Symbols are the same as in Fig. 1.

Figure 4

The log-log relationships between area-based photosynthesis capacity and leaf traits including nitrogen concentration (Na) (a,e), phosphorus concentration (Pa) (b,f), leaf N:P (c,g), and leaf mass per area (LMA) (d,h). (a–d) Show the maximum carboxylation velocity (Vcmax,a), while e-h show the maximum electron transport rate (Jmax,a). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Standardized major axis (SMA) regressions are given in Table S4. Symbols are the same as in Fig. 1.

The log-log relationships between area-based photosynthesis caclass="Chemical">pacity and leaf traits including class="Chemical">pan class="Chemical">pan class="Chemical">nitrogen concentration (Na) (a,e), class="Chemical">paclass="Chemical">pan>n class="Chemical">phosphorus concentration (Pa) (b,f), leaf N:P (c,g), and leaf mass per area (LMA) (d,h). (a–d) Show the maximum carboxylation velocity (Vcmax,a), while e-h show the maximum electron transport rate (Jmax,a). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Standardized major axis (SMA) regressions are given in Table S4. Symbols are the same as in Fig. 1.

Modelling variations in photosynthetic capacity from leaf traits

We used linear mixed-effect to model variations in photosynthetic pan class="Chemical">parameters (Vcmax and Jmax) (Table 2 and Suclass="Chemical">pclass="Chemical">plementary Table S5). The regression coefficients (r2) ranged from 0.30 to 0.53 and were substantially higher when the variables were exclass="Chemical">pressed on a mass basis than on an area basis. The model’s random variance indicated that sclass="Chemical">pecies accounted for less than 3% of the unexclass="Chemical">plained variance for models.

Table 2

The linear mixed-effect models, with mass-based (a) and area-based (b) leaf photosynthetic capacity (Vcmax and Jmax) as the response variables, each showing fixed and random effects.

	V cmax					J max
	Fixed effect			Random variance (%)		Fixed effect			Random variance (%)
	Best model	r2	P	Species	Residual	Best model	r2	P	Species	Residual
(a) Mass-based model
All	Pm + Nm*SLA	0.47	<0.01	0.19	99.81	Pm + Nm * SLA	0.48	<0.01	0.00	100.00
Pioneer	SLA	0.39	<0.01	0.00	100.00	Nm * SLA	0.32	<0.01	0.00	100.00
Generalist	Nm * SLA	0.41	<0.01	0.00	100.00	Nm * SLA	0.50	<0.01	0.00	100.00
Climax	Pm + Nm*SLA	0.53	<0.01	0.00	100.00	Nm * SLA	0.49	<0.01	0.00	100.00
(b) Area-based model
All	Na + Pa * LMA	0.23	<0.01	0.47	99.53	Na + Na * Pa	0.36	<0.01	0.20	99.80
Pioneer	LMA	0.19	<0.05	0.00	100.00	LMA	0.27	<0.05	0.00	100.00
Generalist	Pa * LMA	0.13	<0.05	0.00	100.00	LMA + Na * LMA	0.31	<0.01	0.00	100.00
Climax	Na + Na * LMA	0.16	<0.01	2.56	97.44	Na	0.17	<0.01	0.00	100.00

For the best models, explanatory variables are: leaf nitrogen (N) and phosphorus (P) concentrations; leaf area per mass (SLA) and leaf mass per area (LMA). Species was used as a random component of the model. The best model for Vcmax and Jmax for all species (All), and for the pioneer species (Pioneer), the generalist species (Generalist) and the climax species (Climax) are shown. The coefficient of variation (r2) and significance (P) for the linear regression of the modelled vs measured data are also shown and the contribution of random effect to the variance with the dataset.

The linear mixed-effect models, with mass-based (a) and area-based (b) leaf photosynthetic capan class="Chemical">pacity (Vcmax and Jmax) as the resclass="Chemical">ponse variables, each showing fixed and random effects.

For the best models, explanatory variables are: leaf pan class="Chemical">pan class="Chemical">nitrogenclass="Chemical">pan> (N) and class="Chemical">papan>n class="Chemical">phosphorus (P) concentrations; leaf area per mass (class="Chemical">pan class="Gene">SLA) and leaf mass per area (LMA). Species was used as a random component of the model. The best model for Vcmax and Jmax for all species (All), and for the pioneer species (Pioneer), the generalist species (Generalist) and the climax species (Climax) are shown. The coefficient of variation (r2) and significance (P) for the linear regression of the modelled vs measured data are also shown and the contribution of random effect to the variance with the dataset.

Based on leaf mass, a combination of leaf N, pan class="Chemical">P and class="Chemical">pan class="Chemical">pan class="Gene">SLA accounted for 47% of the variation in Vcmax,m. Similar to Vcmax,m, variations in Jmax,m were largely explained by a combination of N, P and class="Chemical">paclass="Chemical">pan>n class="Gene">SLA; the best model explained 48% of the variation in Jmax,m. When these analyses were repeated using area-based data, relationships were similar to those described for mass-based measurements. For the pioneer species, the species with the highest N and P, SLA was the important fixed effect for explaining Vcmax and Jmax. For the generalist and climax species, SLA (or LMA) was important and the importance of N and P varied depending on whether data were expressed on a mass or an area basis. Importantly, for both the mass- and area-based mixed-effect models, both N and P are important predictors of Vcmax and Jmax, when data from all species were combined.

Variation in foliar P fractions

The overall average concentration of each foliar class="Chemical">P fraction was significantly higher in class="Chemical">pioneer sclass="Chemical">pecies than in the other two grouclass="Chemical">ps (all class="Chemical">pan class="Chemical">P < 0.05) (Table 3), although the mean concentration of both structural P and nucleic acid P were not different between generalist species and climax species. For each group of species, concentrations tended to be higher for nucleic acid P than for the other fractions (all P < 0.05) (Table 3). The mean percentage of P represented by structural P did not differ among the three successional groups (P = 0.766). The metabolic P percentage was significantly lower in generalist species than in the other groups (P < 0.001). The nucleic acid P percentage was lower in pioneer species than in the other groups (P < 0.001). Conversely, the residual P percentage was significantly lower in climax species than in the other groups (P < 0.001).

Table 3

Concentrations (mg g−1) and percentages (%) of foliar P fractions of subtropical forest species in different successional groups.

P fraction	Pioneer		Generalist		Climax		P value
Structural P
Concentration	0.248 ± 0.007	B a	0.204 ± 0.009	BC b	0.206 ± 0.009	C b	P = 0.009
Percentage	21.88 ± 1.27	a	22.63 ± 0.45	a	22.49 ± 0.51	a	P = 0.766
Metabolic P
Concentration	0.358 ± 0.022	A a	0.226 ± 0.010	B c	0.271 ± 0.009	B b	P < 0.001
Percentage	29.98 ± 0.94	a	25.12 ± 0.44	b	30.15 ± 0.77	a	P < 0.001
Nucleic acid P
Concentration	0.349 ± 0.017	A a	0.288 ± 0.009	A b	0.309 ± 0.013	A b	P = 0.021
Percentage	29.47 ± 0.66	b	32.66 ± 0.57	a	33.26 ± 0.48	a	P < 0.001
Residual P
Concentration	0.227 ± 0.020	B a	0.183 ± 0.013	C b	0.129 ± 0.007	D c	P < 0.001
Percentage	18.76 ± 0.94	a	19.58 ± 0.68	a	14.07 ± 0.57	b	P < 0.001

Values are means ± SE. Within each P fraction, concentrations followed by different lowercase letters are significantly different at P < 0.05 among successional groups. Within each P fraction, percentages followed by different by different lowercase letters are significantly different at P < 0.05 among successional groups. Within each successional group, concentrations followed by different uppercase letters are significantly different at P < 0.05 among P fractions.

Concentrations (mg g−1) and percentages (%) of foliar pan class="Chemical">P fractions of subtroclass="Chemical">pical forest sclass="Chemical">pecies in different successional grouclass="Chemical">ps.

Values are means ± SE. Within each class="Chemical">P fraction, concentrations followed by different lowercase letters are significantly different at class="Chemical">pan class="Chemical">P < 0.05 among successional groups. Within each P fraction, percentages followed by different by different lowercase letters are significantly different at P < 0.05 among successional groups. Within each successional group, concentrations followed by different uppercase letters are significantly different at P < 0.05 among P fractions.

Across all 24 species, the concentrations of each foliar class="Chemical">P fraction were class="Chemical">positively related to class="Chemical">pan class="Chemical">Pm (Table 4). For regressions of concentrations of P fractions on Pm, SMA slopes did not significantly differed (P = 0.107), while the intercept was significantly higher for the concentration of nucleic acid P than for the concentrations of other fractions (P = 0.001, Table 4). But the values for both nucleic acid P and metabolic P  were followed by a lowercase c, indicating that the slopes were not different. The concentration of each foliar P fraction had significant positive relationship with Nm, and negative relationship with N:P ratio (Table 4). Moreover, except residual P, LMA significantly increased with decreasing P fractions concentration, and the intercept of nucleic acid P for LMA was larger than those of the other two fractions (P < 0.001, Table 4). The SMA slopes of regressions of concentration of each foliar P fraction on Nm, N:P ratio, and LMA did not significantly differ among foliar fraction types (all P > 0.1).

Table 4

Standardized major axis (SMA) relationships between log-log transformed foliar phosphorus (P) fractions and total leaf P concentration (Pm), leaf mass per area (LMA), leaf nitrogen concentration (Nm), and N:P ratio across the three successional groups.

Leaf trait	Structural P	Metabolic P	Nucleic acid P	Residual P	P-value
P m
r 2	0.630***	0.675***	0.820***	0.568***
Slope	1.051 (0.810, 1.370)	1.133 (0.828, 1.472)	1.191 (0.980, 1.437)	1.771 (1.346, 2.326)	0.107
Intercept	−0.656 b	−0.546 c	−0.485 c	−0.770 a	0.001
N m
r 2	0.124***	0.285***	0.381***	0.036*
Slope	1.044 (0.636, 1.727)	0.858 (0.640, 1.185)	0.937 (0.656, 1.427)	1.764 (1.090, 2.847)	0.311
Intercept	−2.017 b	−1.669 c	−1.722 c	−3.070 a	<0.001
N:P ratio
r 2	0.145***	0.149***	0.034	0.286***
Slope	−1.065 (−0.681, −1.813)	−1.263 (−0.831, −1.843)	−1.067 (−0.703, −1.746)	−1.950 (−1.287, −3.038)	0.430
Intercept	0.713 a	1.081 c	0.872 b	1.741 d	<0.001
LMA
r 2	0.319***	0.325***	0.400***	0.016
Slope	−1.116 (−0.791, −1.671)	−1.160 (−0.766, −1.789)	−1.244 (−0.908, −1.741)	−1.987 (−1.246, −3.323)	0.370
Intercept	1.438 a	1.631 ab	1.838 b	2.962 c	<0.001

Analysis undertaken using species replicates. For each leaf trait, intercepts followed by different letters indicate significant differences among P fractions. *, **, and *** indicate significance at P < 0.05, <0.01, and <0.001, respectively.

Standardized major axis (SMA) relationships between log-log transformed foliar pan class="Chemical">pan class="Chemical">class="Chemical">phosclass="Chemical">phorusclass="Chemical">pan> (P) fractions and total leaf P concentration (Pm), leaf mass per area (class="Chemical">papan>n class="Chemical">LMA), leaf class="Chemical">pan class="Chemical">nitrogen concentration (Nm), and N:P ratio across the three successional groups.

Analysis undertaken using species replicates. For each leaf trait, intercepts followed by different letters indicate significant differences among class="Chemical">P fractions. *, **, and *** indicate significance at class="Chemical">pan class="Chemical">P < 0.05, <0.01, and <0.001, respectively.

Relationships between foliar P fractions and photosynthetic rates and P-use efficiency

Among all 24 species, Amax,m was positively correlated with the concentration of foliar class="Chemical">P fractions (Fig. 5). The coefficients of determination between Amax,m and class="Chemical">pan class="Chemical">P fraction concentrations (r2 = 0.26 for structural P and 0.27 for nucleic acid P) were similar to that between Amax,m and Pm (r2 = 0.26). The correlation coefficients between Amax,m and metabolic P (r2 = 0.11) and residual P (r2 = 0.10) were low and positive. PPUE was positively correlated with the percentage of structural P (r2 = 0.07, P < 0.01; Fig. 6a) but was negatively correlated with the percentage of metabolic P (r2 = 0.03, P < 0.05; Fig. 6b). PPUE was not correlated with the percentage of nucleic acid P or residual P (Fig. 6c,d).

Figure 5

The relationships between the mass-based maximum photosynthesis assimilate rate (Amax,m) and the concentration of four foliar P fractions (structural P, metabolic P, nucleic acid P, and residual P). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Symbols are the same as in Fig. 1.

Figure 6

The relationships between photosynthetic P-use efficiency (PPUE) and the percentage of four foliar P fractions (structural P, metabolic P, nucleic acid P, and residual P). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Symbols are the same as in Fig. 1.

The relationships between the mass-based maximum photosynthesis assimilate rate (Amax,m) and the concentration of four foliar class="Chemical">P fractions (structural class="Chemical">pan class="Chemical">P, metabolic P, nucleic acid P, and residual P). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Symbols are the same as in Fig. 1.

The relationships between photosynthetic class="Chemical">P-use efficiency (class="Chemical">pan class="Chemical">PPUE) and the percentage of four foliar P fractions (structural P, metabolic P, nucleic acid P, and residual P). Data points represent individual leaf values (19, 37, and 47 individuals from pioneer, generalist, and climax species, respectively). Symbols are the same as in Fig. 1.

Discussion

The first hypothesis was supported, and overall average area-based Na, class="Chemical">Pa, and class="Chemical">photosynthetic class="Chemical">pan class="Chemical">parameters (Amax, Vcmax, Jmax, PPUE) were higher in pioneer species than in generalist and climax species (Table 1), indicating the photosynthesis capacity with accompanying P use efficiency generally declined with succession proceeding. These results were consistent with what is known about photosynthetic physiology, such as, photosynthetic capacity decreases along the successional axis[30,31,33,34]. Rijkers et al.[45] found that early-successional species had higher SLA than that of late-successional species in a lowland tropical area of French Guiana. However, no clearly consistent pattern of LMA variations along with succession was found in the present study, i.e. mean LMA values were higher for the generalist species than for the pioneer and climax species (Table 1). This result might be due to the large variations in LMA among species within the same successional group.

Our results showed that leaf N:class="Chemical">P ratios ranged from 10.5 to 37.6 across the 24 sclass="Chemical">pecies, and 48% of the sclass="Chemical">pecies measured were limited by class="Chemical">pan class="Chemical">P and the remainder were co-limited by N and P (Fig. 1). The relatively high values for N:P ratios reported here agree with our prior knowledge that these systems are P limited[42,44]. Moreover, the climax species had the highest N:P ratio (Table 1), showing that P limitation is more pronounced in climax species. However, some studies reported that leaf N:P ratios are not definitive indicators of N or P limitation[46,47]. In a study in Australia, for example, median N:P ratios were relatively high (>20) at all sites, but evidence did not indicate that photosynthesis was limited by P for either forest or pan class="Disease">savanna trees[48]. These inconsistencies between N:P ratios and P limitation may result from the great variability of N:P ratios throughout biomes and among and within species, and although analogous biogeochemical constraints may exist for individual plant organs, differences can vary by up to an order of magnitude[8,49-51]. For this reason, we also explore how leaf N and P concentration in these subtropical systems may affect photosynthetic biochemistry.

Several studies have reported that positive relationships between leaf N (or class="Chemical">P) and class="Chemical">photosynthetic rates[18,19,30,52]. Indeed, Amax was class="Chemical">positively related to N and class="Chemical">pan class="Chemical">P, and these relationships were evident whether measurements were mass-based or area-based (Fig. 2). The slopes of the relationships between Amax and key leaf traits (N, P, and LMA) differed depending on the successional status, more importantly, significant relationship between Amax and P was only found in the climax species (Supplementary Table S3). The climax species had both the lowest P and the lowest Amax values. This result suggests that a reduction in the total P is a main factor limiting photosynthesis of climax species, in agreement with studies related low Amax in leaves with P limitation[24]. The relationship between Amax and N may be constrained by low P in P-limited ecosystems[18,27], out results highlight that the slope of Amax on N was shallower for the climax species (with low P values) than for the pioneer species (with high P values). Further, Amax was negatively correlated with LMA when Amax was based on mass measurements (Fig. 2f). This is in line with an increase in LMA leading to an increase in resistance to CO2 diffusion within the leaf and eventually to a decrease in Amax[53,54].

In this study, Jmax and Vcmax were closely related, and the linear regression had a slope of 1.15 ± 0.07 (Fig. 3). A larger common slope of Jmax and Vcmax was 1.27 when the SMA test was used. The slope in our study was less than those from global analyses (1.64 in Wullschleger[55], 1.67 in Medlyn et al.[56]). Furthermore, the slope of the Jmax-Vcmax relationship was steeper for pioneer species than for generalist and climax species in the current study (Supplementary Table S4). This discrepan class="Chemical">pancy may be exclass="Chemical">plained by the relative high light conditions exclass="Chemical">perienced by the class="Chemical">pioneer sclass="Chemical">pecies.

The photosynthetic caclass="Chemical">pacity (Vcmax and Jmax) of trees is generally class="Chemical">positively related to leaf class="Chemical">pan class="Chemical">P concentration, especially under conditions of P limitation[6,20,21,48,57]. In this study, we demonstrated that across all 24 species, Vcmax,a and Jmax,a were significantly and positively related to Na and Pa and weakly and negatively related to the N:P ratio (Fig. 4). When data from all 24 species were modelled, the line mixed-effect modeling analysis performed here demonstrated that, both N and P were important as variables for predicting Vcmax and Jmax, whether measurements were area-based or mass-based (Table 2). Additionally, leaf P and the interaction between N and SLA have proven to be significant explanatory variables in mixed-effect models of variations in photosynthetic capacity, explaining about 50% of the observed variations in Vcmax,m and Jmax,m (Table 2a). The best combination of variables (N, P, and SLA or LMA) for modelling Vcmax and Jmax changed with succession. We also found that species played a very weak role in modelling photosynthetic capacity, and therefore that limitations to photosynthetic capacity were likely to be results of environmental factors. In addition, these differences in the best mixed-effect model were in line with the observed variation in the relationships of photosynthetic capacity (Vcmax and Jmax) with leaf traits (N, P, and LMA) from SMA regression analysis (Supplementary Table S4). Collectively, these results show that N, P, and LMA or SLA, and especially N and P, help explain the variations in Amax, Vcmax and Jmax in this study and that the degree of influence for different trait of three successional groups are different.

Taken together, the second hypothesis that class="Chemical">P has a stronger influence over class="Chemical">photosynthetic caclass="Chemical">pan class="Chemical">pacity than N, was only supported in the climax species, but was not supported in the other two species groups; leaf N and P are the best predictors of photosynthetic capacity. These results highlight the importance for the consideration of not solely N:P ratio but also growth variables (e.g. photosynthesis) and species identity (e.g. successional strategy) into nutrient limitation determination in the highly species-rich and diversified subtropical forest ecosystems.

In our study, the concentrations and the percentages of total class="Chemical">P reclass="Chemical">presented by functional class="Chemical">pan class="Chemical">P fractions were higher for nucleic acid P than for other foliar P fractions except for metabolic P in the pioneer species (Table 3). These results are in line with past studies showing that nucleic acid P is the largest pool throughout the growing season[39,58]. A previous study suggested that a reduction in foliar P concentration is strongly correlated with a reduction in the concentrations of both metabolic P and nucleic acid P with decreasing soil P availability[39]. Similarly, our results indicate that reductions in the concentrations of both nucleic acid P and metabolic P explain most of the reduction in leaf P concentration among the three successional groups (Table 4). Another previous study showed that the concentration of structural P is greater in fast-growing plants with small LMA than in slow-growing plants with large LMA[59], which was partial in line with the results of the present study. The pioneer species and climax species exhibited similar values for LMA, however, pioneer species had higher structural P concentration (Table 4). Moreover, our results show that the concentration of each foliar P fraction decreases along the successional axis, and that the climax species have a greater nucleic acid P percentage and a lower residual P percentage than pioneer and generalist species (Table 4). Overall, these findings suggest that there are trade-offs in the P allocation among nucleic acid P and metabolic P with residue P.

Amax,m positively correlated with class="Chemical">Pm and the concentrations of each foliar class="Chemical">pan class="Chemical">P fraction (Figs 2b and 5). Mean Amax,m values and mean concentrations of Pm and of each foliar P fraction were lower in climax species than in pioneer species (Table 3 and Supplementary Table S1). As a result, climax species had lower mean PPUE values than pioneer species in this study, which agrees with previous studies of tropical rainforest trees[31]. PPUE was positively correlated with Amax,m (r2 = 0.68, P < 0.001) and weakly and negatively correlated with Pm and pan class="Chemical">LMA (data not shown). Hidaka and Kitayama[40] found that PPUE was negatively correlated with the proportion of structural P but tended to be positively correlated with the proportion of metabolic P. In contrast, we found that PPUE was weakly and positively correlated with the structural P percentage and negatively correlated with the metabolic P percentage (Fig. 6). Hence, the lower PPUE values in climax species can be explained by reduced photosynthetic activity. Finally, the third hypothesis was supported by our results - the climax species may increase their nucleic acid P percentage to maintain their growth and decrease residual P percentage to reduce their demand for P (i.e., reducing total leaf P concentration) as adaptation strategies to soil available P impoverishment at later succession stage.

class="Chemical">Previous studies were restricted to Alaskan tree sclass="Chemical">pecies on class="Chemical">pan class="Chemical">P-rich soils[58,60] or to Mount Kinabalu plant species on P-poor sites[39,40]. Our studies on foliar P fractions and the relationships between foliar P fractions and leaf traits in species representing successional strategies can improve our understanding of what the magnitude species are adapted to low soil P availability and how forests can maintain their high productivity in the highly weathered and acidified soils in the subtropics. We conclude that predicting future dynamics of forest ecosystems in response to global change requires a better understanding of the variations of nutrient limitation, not solely on the base of plant N: P ratios but also by incorporating growth variables (e.g. photosynthesis), and in particular the P adaptation strategies created by ecosystem succession of the subtropical forests.

Materials and Methods

Study site and plant materials

This study took place at the Dinghushan Biosphere Reserve (DBR) in central Guangdong class="Chemical">Province, southern China (21°09′21′′–21°11′30′′ N, 112°30′39′′–112°33′41′′E). The region is characterized by a tyclass="Chemical">pical subtroclass="Chemical">pical monsoon climate, with mean annual temclass="Chemical">perature is 21.4 °C, and annual average class="Chemical">preciclass="Chemical">pitation is 1927 mm with 80% occurring during the wet season (Aclass="Chemical">pril to Seclass="Chemical">ptember). The soils are classified as Ultisol and Udult according to the USDA soil classification. A total of 24 sclass="Chemical">pecies across successional stages were samclass="Chemical">pled in this study (Suclass="Chemical">pclass="Chemical">plementary Table S6). Three to five individuals were selected for each sclass="Chemical">pecies. The studied sclass="Chemical">pecies are common and tyclass="Chemical">pical in each stage of the succession according to the long-term forest community studies of the reserve[61,62]. We used long-reach class="Chemical">pruner to collect class="Chemical">pan class="Chemical">pan class="Disease">middle canopy branches (supporting leaves considered to be typically exposed to full sunlight for much of the day) from tall plant species. The detached branches had been recut under class="Chemical">paclass="Chemical">pan>n class="Chemical">water immediately after harvesting to preserve xylem water continuity, prior to subsequent leaf gas exchange measurements[63-65].

Leaf gas exchange measurements

Measurements of leaf gas exchange were made on the most recently fully exclass="Chemical">panded leaves during August to Seclass="Chemical">ptember 2015, between 9:00 and 12:00 h using the LI-6400 class="Chemical">portable class="Chemical">photosynthesis system (Li-Cor, Lincoln, Nebraska, USA). Maximum class="Chemical">photosynthetic rate class="Chemical">per unit area (Amax,a) was determined for each sclass="Chemical">pecies by measuring a light resclass="Chemical">ponse curve (A-class="Chemical">pan class="Chemical">PPFD curves) at ambient 400 μmol mol−1 CO2, leaf temperature of 28–30 °C, and relative humidity of 40–60%, with the photosynthetic photon flux density (PPFD) order was 1500, 1200, 1000, 800, 500, 300, 200, 120, 100, 80, 50, 20,0 μmol m−2 s−1. For the photosynthesis measurement of masson pine needles, a bunch of needles were measured side by side, and then the average was calculated. Photosynthetic P-use efficiency (PPUE) was calculated as Amax divided by total P concentration (µmol CO2 mol P−1 s−1). Vcmax,a and Jmax,a on an area basis were estimated from relationships between photosynthetic rate (A) and sub-stomatal CO2 mole fraction (C)[28] at fixed PPFD (1200 µmol photons m−2 s−1) following the CO2 order 400, 300, 200, 100, 50, 500, 700, 900, 1000, 1200, 1500 μmol mol−1.

Leaf structure and nutrients

After photosynthetic measures, leaves with the same position as used for gas-exchange measurements were collected. Some of the leaves were immediately snap-frozen in liquid N and transferred on dry ice to laboratory. The samples were freeze-dried and stored at −80 °C until they were used for determination of foliar class="Chemical">P fractions. The remaining leaves were oven-dried at 70 °C and then ground and homogenized for subsequent analyses. Total N concentration in dried leaves was determined using the Kjeldahl method. An additional samclass="Chemical">ple of ten leaves was scanned to determine leaf area (LA) by LI-3000A class="Chemical">portable system, dried at 70 °C for 72 h to a constant weight and measured for oven-dried mass (class="Chemical">pan class="Chemical">pan class="Disease">DM). Since the needles of masson pine trees don’t have flat leaf area, we record needle length and cross-section width and needle leaf area was estimated as: LA = π*L*D/2 (where LA is leaf area; L is needle length and D is needle width at the hale needle length). class="Chemical">paclass="Chemical">pan>n class="Chemical">LMA, calculated as DM·LA−1 (g m−2), was used to calculate area-based nutrient concentrations (Na, Pa; g m−2) from mass-based concentrations (Nm, Pm; mg g−1), and to shift area-based photosynthetic parameters (Amax,a, Vcmax,a, and Jmax,a; µmol CO2 m−2 s−1) to (Amax,m, Vcmax,m, and Jmax,m; µmol CO2 g−1 s−1). To compare LMA with mass-based leaf nutrient concentrations, we converted LMA into its reciprocal, leaf area per mass (SLA, cm2 g−1).

Foliar P fractions

Foliar pan class="Chemical">P fractions were extracted following methods outlined in Hidaka and Kitayama[37]. Each freeze-dried and ground samclass="Chemical">ple (0.5 g DW) was homogenized and extracted twice with a total of 15 mL of 12:6:1 CMF (class="Chemical">pan class="Chemical">pan class="Chemical">chloroform, class="Chemical">paclass="Chemical">pan>n class="Chemical">methanol, and formic acid; v/v/v) and twice with a total of 19 mL of 1:2:0.8 CMW (chloroform, methanol, and water; v/v/v). Extracts were combined in a 50-mL centrifuge tube, and 9.5 mL of water was added to each tube. Thereafter, the extract was separated into an aqueous upper layer and a lipid-rich organic bottom layer. A subsample of the lipid layer was digested to give structural P. The residue was re-extracted for 1 h with 5 mL of 85% (w/v) methanol. The supernatant was added to the tube containing the aqueous layer. A 20 ml volume of the aqueous layer was added to another 50-ml graduated tube. The remaining aqueous layer was decanted into the tube containing the residue. The volume in the tube was increased to 20 ml with deionized water. After the preparation was cooled to 4 °C, 1 ml of cold 100% (w/v) trichloroacetic acid (TCA) was added to make a 5% (w/v) TCA solution. After extraction for 1 h, this cold TCA solution was subjected to a re-extraction with 10 mL of 5% (w/v) cold TCA 4 °C. A subsample of the supernatant was taken for the analysis of metabolic P. The residue was then re-extracted twice in a total of 35 mL of 2.5% (w⁄ v) TCA at 95 °C for 1 h. A subsample of the supernatant was taken for the analysis of nucleic acid P. The residue was digested to obtain residue P.

Each subsample was evaporated to dryness at 70 °C (organic solutions) or 100 °C (aqueous solutions) and digested for class="Chemical">P determination as described above. The concentration of class="Chemical">pan class="Chemical">P in the digest was determined at 700 nm in a UV-Vis spectrophotometer (UV1800, Shimadzu, Japan) after a standard pan class="Chemical">molybdate reaction. Total P concentration was the sum of foliar P fractions.

Data analysis

To comclass="Chemical">pare overall average leaf trait and foliar class="Chemical">pan class="Chemical">P fraction (concentration and percentage) values of the three successional groups using ANOVAs, and multiple comparisons were conducted using Duncan’s multiple range test. Bivariate regression was used to explore relationships between photosynthetic parameters and other leaf traits (N, P and LMA) among the three successional groups. Standardized major axis (SMA) regression was used to examine for variations in the slope and the y-axis intercept of bivariate leaf trait relationships, using SMATR version 2.0. software. When we tested the relationships among leaf traits, values were log10-transformed when it was necessary to normalize distributions. The significance of SMA regression was determined at the 0.05 level. We also used a linear mixed effects model combining fixed and random effects to account for variability in Vcmax and Jmax on both area- and mass-bases. The model’s fixed effect included N, P and LMA (or SLA), and species as random effects. Model specification and validation were conducted in ‘R’ (version 3.4.0; R Development Core Team, 2011), using the nlme package. All of our data for the mixed-effect modelling analysis were log transformed. To select the best model, Akaike’s information criterion (AIC) was used. Statistical analysis was performed using SPSS 21.0 (IBM SPSS, USA), unless otherwise indicated.