Impact of growth temperature on scaling relationships linking photosynthetic metabolism to leaf functional traits

Summary

1. Scaling relationships linking photosynthesis (A) to leaf traits are important for predicting vegetation patterns and plant-atmosphere carbon fluxes. Here, we investigated the impact of growth temperature on such scaling relationships.

2. We assessed whether changes in growth temperature systematically altered the slope and/or intercepts of log–log plots of A vs leaf mass per unit leaf area (LMA), nitrogen and phosphorus concentrations for 19 contrasting plant species grown hydroponically at four temperatures (7, 14, 21 and 28 °C) in controlled environment cabinets. Responses of 21 °C-grown pre-existing (PE) leaves experiencing a 10 day growth temperature (7, 14, 21 and 28 °C) treatment, and newly-developed (ND) leaves formed at each of the four new growth temperatures, were quantified. Irrespective of the growth temperature treatment, rates of light-saturated photosynthesis (A) were measured at 21 °C.

3. Changes in growth temperature altered the scaling between A and leaf traits in pre-existing (PE) leaves, with thermal history accounting for up to 17% and 31% of the variation on a mass and area basis, respectively. However, growth temperature played almost no role in accounting for scatter when comparisons were made of newly-developed (ND) leaves that form at each growth temperature.

4. Photosynthetic nitrogen and phosphorus use efficiency (PNUE and PPUE, respectively) decreased with increasing LMA. No systematic differences in temperature-mediated reductions in PNUE or PPUE of PE leaves were found among species.

5. Overall, these results highlight the importance of leaf development in determining the effects of sustained changes in growth temperature on scaling relationships linking photosynthesis to other leaf traits.

Introduction

In almost all biomes, variations in temperature play a major role in determining rates of photosynthetic CO₂ uptake (Sage & Kubien 2007) and the magnitude of related leaf traits (Atkin et al. 2006a; Poorter et al. 2009). Rates of photosynthesis (A) are typically near maximal at moderate temperatures, with most temperate species exhibiting a broad temperature optimum (T_opt) in the range of 15–30 °C (Atwell, Kriedemann & Turnbull 1999; Larcher 2004). Exposure to chilling or high temperatures results in an immediate reduction in net CO₂ exchange in the light (e.g. chilling, Powles, Berry & Björkman 1980; high temperatures, Badger, Björkman & Armond 1982). However, in response to sustained exposure to changes in growth temperature, rates of A often thermally acclimate (Berry & Björkman 1980; Huner et al. 1993; Hurry et al. 1994; Sage & Kubien 2007). Acclimation can result in recovery of rates of A at the new growth temperature, a shift in the T_opt, and ultimately complete homeostasis [i.e. identical rates of A in plants growing at contrasting temperatures (Berry & Björkman 1980; Hurry et al. 1994; Strand et al. 1997, 2003; Hikosaka, Murakami & Hirose 1999; Atkin, Scheurwater & Pons 2006b]. Given the likely importance of thermal acclimation for predicting shifts in vegetation patterns (Jump, Mátyás & Peñuelas 2009) and changes in carbon fluxes between plants and the atmosphere (Reich, Wright & Lusk 2007; Sitch et al. 2008), it is essential that the impacts on growth temperature on A, and the scaling between A and related leaf traits, be thoroughly characterized in a wide range of contrasting plant species.

Using comparisons of A measured at moderate common temperature, past studies have shown that acclimation is maximal following development of new leaves at the new growth temperature (Strand et al. 1997; Campbell et al. 2007). Thermal acclimation of A is underpinned by changes in a range of factors, including a change in the abundance of photosynthetic proteins (Hurry et al. 1994, 1995b; Hikosaka, Murakami & Hirose 1999; Strand et al. 1999; Savitch et al. 2001; Stitt & Hurry 2002; Onoda, Hikosaka & Hirose 2005b; Yamori, Noguchi & Terashima 2005), the extent of limitations imposed by triose phosphate use (TPU) (Sharkey et al. 1986; Sage & Sharkey 1987) and the redistribution of inorganic phosphate between cellular compartments (Hurry, Gardeström & Öquist 1993a; Hurry et al. 2000; Stitt & Hurry 2002). Such changes could alter the scaling relationships used to predict rates of A [e.g. log–log A-LMA-[N] relationships (Reich, Walters & Ellsworth 1997; Reich et al. 1999; Wright et al. 2004, 2005a)], particularly if sustained changes in growth temperature have a different impact on rates of A relative to growth-temperature mediated changes in leaf [N] and LMA. We know that LMA typically increases when leaves develop in the cold (Atkin et al. 2006a; Poorter et al. 2009; Gorsuch, Pandey & Atkin 2010), and that leaf [N] is often greater in cold-developed leaves (compared to their warm-grown counterparts: Körner & Larcher 1988; Tjoelker, Reich & Oleksyn 1999b; Campbell et al. 2007). Moreover, variations in log–log scaling relationships have been observed in comparisons of plants grown in contrasting environments [e.g. sites grouped according to levels of nutrients and rainfall (Wright, Reich & Westoby 2001) and rainfall and mean annual temperature (Wright et al. 2005b)] and when comparing plants grown under ambient and elevated atmospheric CO₂ concentration (Tjoelker, Oleksyn & Reich 1999a). However, to date, no study has quantified the impact of growth temperature per se on scaling relationships linking rates of A to variations in [N] and/or LMA values.

Another important unknown is the impact of growth temperature on scaling relationships linking rates of A with leaf phosphorus (P). P plays important roles in regulating photosynthetic carboxylation capacity (de Groot et al. 2003) and variations in P concentration ([P]) can alter scaling between A and related leaf traits (Reich, Oleksyn & Wright 2009). Typically, P declines as LMA increases (Wright et al. 2004), with N and P concentration usually scaling together and with RGR (Gusewell 2004). Given the pivotal role of P supply in cold acclimation (Hurry et al. 2000; Stitt & Hurry 2002), it is important that the impact of sustained changes in growth temperature on log–log scaling relationships linking [P] to A (and related traits) be established (Reich, Oleksyn & Wright 2009).

In addition to altering scaling relationships, changes in photosynthesis and tissue chemistry resulting from changes in growth temperature may also have an impact on the efficiency of nutrient use. There is a strong relationship between inherent leaf functional traits and the rate of photosynthesis per unit N (PNUE), with PNUE increasing as LMA decreases (e.g. Poorter & Evans 1998; Pons & Westbeek 2004; Wright et al. 2005a) and increasing with RGR (Poorter, Remkes & Lambers 1990; Westbeek et al. 1999). Importantly, PNUE may also be influenced by growth irradiance (Poorter & Evans 1998) and nutrient availability (Pons & Westbeek 2004). Temperature-mediated changes in [N] and N allocation (see references cited above) are also likely to impact on the efficiency of N use in photosynthesis. Yamori et al. (2009) have shown that PNUE was lower in leaves of 11 crop species grown at low temperature. However, it is not known whether the impacts of growth temperature on PNUE differ among contrasting plant species from opposite ends of the ‘leaf economic spectrum’ (LES; Wright et al. 2004). Similarly, the impacts of growth temperature on rates of photosynthesis per unit P (PPUE) on contrasting species is unknown.

In the current study, we ask whether log–log scaling relationships linking A with related leaf traits (LMA, [N] and [P]) are altered when plants experience a broad range of growth temperatures (7, 14, 21 and 28 °C) under controlled environment conditions, both in pre-existing leaves that experience sustained changes in growth temperature, and in newly-developed leaves that form in each new thermal environment. Rates of light-saturated photosynthesis (A_sat) for each growth temperature treatment were measured at a common temperature (21 °C), following Campbell et al. (2007). While comparisons of A_sat at a common temperature could be confounded by growth-temperature dependent changes in T_opt (Berry & Björkman 1980), past studies have shown that such changes in the T_opt of A_sat do not alter the relative difference in rates of A_sat exhibited by cold and warm grown plants, when rates of A are measured at moderate temperatures (Atkin, Scheurwater & Pons 2006b). Moreover, a recent study by Dillaway & Kruger (2010) reported that growth temperature had little impact on the T_opt of photosynthesis in a range of tree species. Specifically, we address the questions: (i) do sustained differences in growth temperature alter the slope and/or intercepts of the log–log scaling relationships linking rates of A to related leaf traits; and (2) is the efficiency of N and P use in A in a wide range of contrasting species covering a wide range of traits from the LES systematically altered by growth temperature?

Materials and methods

Plant material

Data from 19 species were used, representing three functional groups: grasses (Bromus ramosus, B. erectus, Poa trivialis L., P. costiniana J. Vickery); forbs [Achillea millefolium, A. ptarmica, Plantago major, P. euryphylla, Silene dioica, S. uniflora, and Arabidopsis thaliana (ecotype Ost-0)]; and evergreen shrubs and trees (Acacia melanoxylon R. Br., A. aneura R. Muell Ex Benth, Cistus ladanifer L., C. laurifolius L., Eucalyptus dumosa, E. delegatensis, Quercus suber and Q. ilex ssp. ballota). Details of the origin and natural distribution of some of these species can be found in Loveys et al. (2002); the rest can be found in Campbell et al. (2007). Campbell et al. (2007) assessed inter-functional group differences in long-term temperature responses of photosynthesis for the same species, but did not analyze the effect of growth temperature on log–log scaling relationships linking rates of A to related leaf traits (LMA, [N], and [P]), or the impact of growth temperature on PNUE and PPUE.

Growth conditions

Seeds of grasses and forbs were initially sown on trays of John Innes F2 compost. Seeds of Eucalyptus species were vernalized at 4 °C for 6 weeks before sowing. Seeds of Acacia species were pre-treated as described by Atkin et al. (1998) and transferred to trays of 50 : 50 sand : vermiculite once the radicle had emerged. Quercus acorns were planted 2·5 cm deep in a 50 : 50 sand : vermiculite mixture with five acorns per 20 cm diameter pot. Seeds of Cistus spp were heated for 5 min. in an oven at 100 °C in order to break dormancy and then soaked in water for 24 h before being sown on trays of John Innes F2 compost. All species were placed in a greenhouse maintained at 25 ± 2 °C (day) and 20 ± 2 °C (night) with a 16 h day using supplementary lighting (400 W high pressure sodium bulbs). Plants were carefully removed from the compost and the roots washed thoroughly with water once the roots had reached at least 3 cm in length. They were transferred to 17 L hydroponics tanks containing fully-aerated modified Hoagland’s nutrient solution (Poorter & Remkes 1990). The solution was maintained at a pH of 5·8 and replaced weekly.

Experimental procedure

The tanks were placed in growth cabinets (Snijders Microclima 1750, Snijders Scientific BV, Netherlands) at a constant 21 °C with a 16 h day and 300 ± 10 μmol m⁻² s⁻¹ PPFD provided by a combination of fluorescent tubes (54W TF, 24W WTF, 58W Brite Gro 2023, 58W Brite Gro 2084, 18W Contour 840TL; Sylvania, Shipley, W. Yorkshire, UK). After establishment at 21 °C, plants of each species were transferred to cabinets maintained at 7, 14, 21 and 28 °C and 70% RH (Day 0); three replicate plants were transferred to each new growth temperature treatment. At day 10 photosynthesis measurements were made on the last fully-expanded, pre-existing (PE) leaf that had developed before the transfer. The three replicate plants were maintained in these conditions until new fully-expanded leaves had developed, at which point gas exchange measurements were carried out on these fully-expanded, newly-developed (ND) leaves.

Gas exchange measurements

Photosynthesis measurements were made at a reference CO₂ level of 400 ppm and at a set temperature of 21 °C using a Li-Cor 6400 portable infra-red gas analyser (Li-Cor Inc., Lincoln, NE, USA). Measurements were not begun until 2 h into the photoperiod. The plants were transferred to 21 °C 1 h before measurements were made. A was measured at saturating irradiance (1500 μmol m⁻² s⁻¹) and then at ambient irradiance (300 μmol m⁻² s⁻¹), each time allowing the readings to stabilize. The leaf (or phyllode of Acacia species) area used for the photosynthesis measurements was determined using a Li-Cor LI-3000A leaf area meter (Li-Cor Inc., Lincoln, NE, USA). Samples were weighed and then frozen in liquid N₂ and stored at −20 °C.

Nitrogen and phosphorus analysis

Leaf samples for each species/treatment combination were freeze-dried (Edwards Modulyo Freeze Drier, York, UK), pooled and ground using a hammer mill (31–700 Hammer Mill; Glen Creston, Stanmore, UK). Total N concentrations were analyzed using pooled leaf samples (i.e. combining leaves from three replicate plants) as described previously (Loveys et al. 2003; Campbell et al. 2007). Measurements of P were also carried out using the pooled leaf samples. We used a triple digestion and measured P concentrations using the molybdenum blue method (Allen 1974).

Statistical analyses

Data from the IRGA was managed in a custom-designed MICROSOFT ACCESS (Microsoft Inc.) database in which gasket diffusion correction calculations (Bruhn, Mikkelsen & Atkin 2002) were performed. Statistical analyses were carried out using spss v.13, (SPSS Science, Birmingham, UK), sigmaplot v.10 (Systat Software Inc., Chicago, IL, USA) and MICROSOFT EXCEL 2000 (Microsoft Inc.). Where noted in the Results, log₁₀ transformations were carried out on the data and linear regressions then fitted. Hierarchical multiple regression and Analysis of Covariance (ancova) were carried out using spss v.13. Homogeneity of variance in the ancovas was checked using Levene’s test. Where appropriate regression lines were differentiated using Bonferroni’s post hoc test in spss v.13.

Results

Scaling relationships irrespective of growth temperature

When considered collectively (i.e. all species, leaf types and temperature treatments together), pair-wise comparisons showed that most, but not all, traits were significantly correlated (Fig. 1, Table 1). Log–log A_mass-N and log–log A_mass-LMA correlations were similar to those reported previously by Wright et al. (2004) (see Fig. S1 in Supporting Information). However, area-based relationships were much weaker, and there was no significant correlation between LMA and A_area (Fig. 1b, Table 1). A hierarchical multiple regression on the mass-based data indicated that N_mass accounted for 40% of the variability in A_mass and LMA accounted for an additional c. 15%. There was a significant relationship between P content and A on a mass and an area basis although the correlations were low (Table 1). P_mass scaled negatively with LMA but positively with N_mass (Fig. 1c, Table 1).

Figure 1.  Three-way log–log relationships among light-saturated leaf photosynthetic assimilation (A) and leaf phosphorus (P) concentration, and nitrogen (N) concentration with reference to leaf mass per unit leaf area (LMA). Data shown are for the 19 species exposed to 7–28 °C under controlled environment conditions, both for 10-day temperature-treated pre-existing leaves and newly-developed leaves formed at each growth temperature. The direction of the data cloud in three-dimensional space can be determined from the shadows projected on the floor and walls of the three dimensional space: (a) mass-based leaf photosynthetic assimilation (A_mass), LMA, and mass-based leaf nitrogen concentration (N_mass); (b) area-based photosynthetic assimilation (A_area), LMA, and N_area; (c) mass-based leaf phosphorus concentration (P_mass), LMA, and N_mass. Table 1 shows linear regression bivariate relationships between each set of traits.

Table 1.   Bivariate relationships between leaf traits. All species, growth temperature treatments and leaf types (pre-existing and newly-developed) are combined in this linear regression analysis applied to log–log plots of leaf traits (see Supporting Information). Coefficients of determination (r²), y-axis intercepts and standardized major axis slopes are given, with the y-axis intercept values being shown a log₁₀ and untransformed basis. Significance of the slopes of log–log plots compared to zero are shown (NS denotes non-significant relationships)

x-axis	y-axis	r^2	y-axis intercept	Slope	P
*P ≤ 0·05, **P ≤ 0·01 and ***P ≤ 0·001. The regression values given for the relationships between LMA and N are those reported in Atkin et al. (2008). LMA, leaf mass per unit leaf area; Nmass and Narea, N concentration per unit dry mass and per unit leaf area, respectively; Pmass and Parea, P concentration per unit dry mass and per unit leaf area, respectively; Amass and Aarea, light-saturated rates of photosynthesis per unit dry mass and per unit leaf area, respectively.
LMA (g m^−2)	Amass (nmol CO2 g DM^−1 s^−1)	0·49	4·21	−1·14	***
Nmass (%)	Amass (nmol CO2 g DM^−1 s^−1)	0·40	1·46	1·45	***
LMA (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	1 × 10^−3	1·10	−0·05	NS
Narea (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	0·03	0·95	0·29	*
Pmass (%)	Amass (nmol CO2 g DM^−1 s^−1)	0·19	2·30	0·57	***
Parea (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	0·07	1·13	0·23	**
Nmass (%)	Pmass (%)	0·36	−0·68	1·01	***
LMA (g m^−2)	Pmass (%)	0·13	0·64	−0·45	***
Nmass (%)	LMA (g m^−2)	0·42	2·20	−0·90	***
Narea (g m^−2)	LMA (g m^−2)	0·49	1·48	0·91	***

Impact of leaf type and thermal history on scaling relationships

Figure 2 shows the impact of growth temperature for pre-existing (PE; Fig. 2a,c,e) and newly-developed (ND; Fig. 2b,d,e) leaves on the bivariate relationships for LMA, N_mass, P_mass and A_mass. Regression statistics for each relationship/temperature combination are provided in Table S1. When considering leaf type (i.e. PE and ND) with all temperature treatments combined, we found that while the relationship between A_mass and N_mass or LMA was stronger in PE leaves than in ND leaves (as indicated by the coefficient of determination, r²), overall there was no significant difference between PE and ND leaves in bivariate scaling relationships linking A_mass to each covariate (Table 2).

Figure 2.  Rates of mass-based light-saturated leaf photosynthetic assimilation (A) (measured at 21 °C in all cases) in relation to leaf mass per unit leaf area (LMA; a and b), leaf nitrogen (N) concentration (c and d), and leaf phosphorus (P) concentration (e and f), with the x and y-axes shown on a log₁₀ scale. Data shown are for the 19 species exposed to 7–28 °C (7 °C – yellow circles; 14 °C – green triangles; 21 °C – black squares; 28 °C – red diamonds) under controlled environment conditions, both for 10-day temperature-treated pre-existing leaves (a, c and e) and newly-developed leaves formed at each growth temperature (b, d and f). Regression lines fitted to each growth temperature shown; see Supporting Information for details on individual regressions for each growth temperature/leaf type combination.

Table 2.   Analysis of covariance and linear regression values for log–log leaf-trait data for the comparison of pre-existing and newly-developed leaves. Values represent the coefficient of determination (r²), y-axis intercept (A), scaling relationship slope (b) and the significance levels for (a) the relationship between the covariate (x) and the dependent variable (y), (b) the effect of leaf type (pre-existing, PE, or newly-developed, ND) and (c) the interaction between the covariate and leaf type. Linear regressions were calculated on all PE or ND data combined across temperature treatments. NS denotes non-significant relationships

x-axis	y-axis	Leaf type	r^2	y- axis intercept	Slope	(a)	(b)	(c)
*P ≤ 0·05, **P ≤ 0·01 and ***P ≤ 0·001. LMA, leaf mass per unit leaf area; Nmass, N concentration per unit dry mass; Pmass, P concentration per unit dry mass; Amass, light-saturated rate of photosynthesis per unit dry mass. Levene’s test was not significant on either an Nmass or LMA basis, indicating homogeneity of variance between PE and ND leaves.
Nmass (%)	Amass (nmol CO2 g^−1 s^−1)	PE	0·50	1·50	1·44	***	NS	NS
		ND	0·30	1·40	1·52
LMA (g m^−2)	Amass (nmol CO2 g^−1 s^−1)	PE	0·55	4·36	−1·26	***	NS	NS
		ND	0·46	4·16	−1·09
Pmass (%)	Amass (nmol CO2 g^−1 s^−1)	PE	0·29	2·33	0·63	***	NS	NS
		ND	0·09	2·27	0·48

In PE leaves, when considered separately from ND leaves, there was a significant relationship between P and A (Fig. 2e) on both an area and mass basis (Table 3). This relationship was however not significant in ND leaves (Fig. 2f). The positive scaling relationship between N and P concentration (r² = 0·38 and 0·43 for PE and ND leaves on a mass basis; r² = 0·17 and 0·59 for PE and ND leaves on an area basis) was not significantly affected by growth temperature in PE and ND leaves on a mass or area basis (Table 3).

Table 3.   Analysis of covariance for log–log leaf-trait data. Values represent significance levels for (a) the relationship between the covariate (x) and the dependent variable (y), (b) the effect of temperature treatment and (c) the interaction between the covariate and temperature for 10-day treated pre-existing leaves and for newly-developed leaves formed under each new temperature regime. NS denotes non-significant relationships

x-axis	y-axis	Leaf type
		Pre-existing			Newly-developed
		(a)	(b)	(c)	(a)	(b)	(c)
*P ≤ 0·05, **P ≤ 0·01 and ***P ≤ 0·001. LMA, leaf mass per unit leaf area; Nmass and Narea, N concentration per unit dry mass and per unit leaf area, respectively; Amass and Aarea, light-saturated rate of photosynthesis per unit dry mass and per unit leaf area, respectively; Pmass and Parea, P concentration per unit dry mass and per unit leaf area, respectively.
LMA (g m^−2)	Amass (nmol CO2 g^−1 s^−1)	***	NS	**	***	NS	NS
Nmass (%)	Amass (nmol CO2 g^−1 s^−1)	***	*	NS	***	NS	NS
LMA (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	NS	NS	*	NS	NS	NS
Narea (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	NS	NS	*	NS	NS	NS
Pmass (%)	Amass (nmol CO2 g^−1 s^−1)	***	**	NS	NS	NS	NS
Parea (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	**	**	NS	NS	NS	NS
Nmass (%)	Pmass (%)	***	NS	NS	***	NS	NS
Narea (g m^−2)	Parea (g m^−2)	**	NS	NS	***	NS	NS

In PE leaves, the slopes of the log–log A_mass-LMA relationship varied with growth temperature (as shown by the significant interaction between the covariate and temperature in Table 3). One consequence of this temperature effect was that there was a greater differential between rates of A_mass among growth temperatures at high LMA in PE leaves (Fig. 2a). Conversely, there was little difference in A_mass between temperature regimes at low LMA values (Fig. 2a). By contrast, growth temperature had no significant effect on generalized log–log relationships between LMA and A (either on a mass or area basis) in ND leaves indicating that a single scaling relationship is re-established when new leaves subsequently develop at each new growth temperature (Fig. 2b, Table 3).

For PE leaves, generalized log–log relationships between A_mass and the covariates N_mass and P_mass were also affected by growth temperature (Fig. 2c,e, Table 3). However, unlike the impact of growth temperature on A_mass-LMA relationships in PE leaves, the slopes of the log–log A_mass-N_mass and A_mass-P_mass relationships in PE leaves were similar for all temperature treatments, as shown by the lack of interaction between the covariate and temperature in Table 3. Rather, temperature altered these scaling relationships via a significant effect on the elevation of the slopes (i.e. the intercept differed; Table S1). The overall significant difference between treatments was largely driven by the response of the 7 °C treated plants which exhibited lower rates of A_mass per unit N_mass than for plants at the other three growth temperatures (Fig. 2c) and also lower rates per unit P_mass than 21 and 28 °C treated plants (Fig. 2e) as indicated by Bonferroni’s post hoc test (data not shown). Values did not differ significantly between 14, 21 and 28 °C for N_mass or between 21 and 28 °C for P_mass. By contrast, temperature had no effect on the slope or intercept of log–log A_mass-N_mass and A_mass-P_mass relationships in ND leaves (Table 3). Thus, although changes in growth temperature alter scaling between A_mass and N_mass and P_mass in PE leaves, single scaling relationships are, as with LMA, re-established in newly-developed leaves.

Using hierarchical multiple regression and knowledge of the recent thermal history of mature leaves, we can account for some of the scatter in data around the predicted log–log scaling relationships for PE and ND leaves (Fig. 2, Table 4). In PE leaves, LMA accounted for c. 55% of the variability in A_mass and temperature accounted for an additional c. 14% (Fig. 2a). N_mass accounted for 50% of the variability in A_mass in PE leaves (Fig. 2c) and in this case temperature accounted for an additional 17% (Table 4). By contrast, temperature accounted for little additional variability in A_mass in ND leaves (Fig. 2b,d, Table 4). P_mass accounted for 29% of the variability in A_mass in PE leaves and temperature accounted for an additional 16·2%. The relationship between P_mass and A_mass was weaker in ND leaves, with temperature explaining c. 3% of A_mass (Table 4). Hence, temperature is important in explaining variability in these scaling relationships in PE leaves exposed to a change in temperature (particularly in leaves shifted from 21 to 7 °C) but not for leaves that are developed at each respective growth temperature.

Table 4.   Hierarchical multiple regression for log–log leaf-trait data for each leaf type. Values of r² representing (a) the variability in y accounted for by the covariate, x, (b) the additional variability in y accounted for by temperature and (c) the total variability in y accounted for by both predictors, are shown

x	y	Leaf type
		Pre-existing			Newly-developed
		(a) Covariate	(b) Temp	(c) Total	(a) Covariate	(b) Temp	(c) Total
LMA, leaf mass per unit leaf area; Nmass, N concentration per unit dry mass; Amass, light-saturated photosynthesis per unit dry mass.
LMA (g m^−2)	Amass (nmol CO2 g^−1 s^−1)	0·55	0·14	0·69	0·46	0·01	0·47
Nmass (%)	Amass (nmol CO2 g^−1 s^−1)	0·50	0·17	0·67	0·30	0·01	0·34
LMA (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	0·01	0·21	0·22	1 × 10^−3	0·02	0·02
Narea (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	0·04	0·24	0·28	0·01	0·02	0·03
Pmass (%)	Amass (nmol CO2 g^−1 s^−1)	0·29	0·16	0·45	0·09	0·03	0·11
Parea (g m^−2)	Aarea (μmol CO2 m^−2 s^−1)	0·10	0·31	0·40	0·01	0·01	0·02

Scaling between net CO₂ uptake per unit leaf N or P and LMA

A negative relationship was found between rates of light-saturated A per unit leaf N (i.e. light-saturated PNUE) and LMA in both leaf types (PE leaves, P ≤ 0·001; ND leaves, P ≤ 0·001; Fig. 3, Table 5), with r² values of 0·45 and 0·45 for PE and ND leaves respectively. The relationship between PNUE and LMA was also significant at growth irradiance (Table 5), with r² values of 0·61 and 0·62 for PE and ND leaves respectively. There was no overall significant effect of temperature treatment on PNUE, or a significant interaction between growth temperature and LMA at saturating irradiance, although values of PNUE (under saturating and growth irradiance) were numerically lower in the 7 °C treatment than for the 21 and 28 °C treatments in PE leaves (Fig. 3, Table 5). At growth irradiance there was a significant interaction between temperature and LMA effects on PNUE in PE leaves (Table 5), with a greater reduction in PNUE at low temperatures in high LMA species. There was also a negative relationship between PPUE and LMA in PE and ND leaves (PE leaves, P ≤ 0·01; ND leaves, P ≤ 0·001; Fig. 3, Table 5). There were no significant differences between temperatures, or interactions, in either leaf type (Table 5).

Figure 3.  Photosynthetic nitrogen use efficiency (PNUE; a and b) and photosynthetic phosphorus efficiency (PPUE; c and d) in relation to leaf mass per unit leaf area (LMA) with the x and y-axes shown on a log₁₀ scale. Photosynthesis measured at 21 °C in all cases. Data shown are for the 19 species exposed to 7–28 °C under controlled environment conditions, both for 10-day temperature-treated pre-existing leaves (a and c) and newly-developed leaves formed at each growth temperature (b and d). Regression lines fitted to each growth temperature are shown.

Table 5.   Analysis of covariance for log–log photosynthesis per unit leaf nitrogen (PNUE) and photosynthesis per unit leaf phosphorus (PPUE) data. Values represent significance levels for (a) the relationship between the covariate (log₁₀x) and the dependent variable (log₁₀y), (b) the effect of temperature treatment and (c) the interaction between the covariate and temperature for 10-day treated pre-existing leaves and for newly-developed leaves formed under each new temperature regime. NS denotes non-significant relationships

x-axis	y-axis	Leaf type
		Pre-existing				Newly-developed
		(a)	(b)	(c)	n	(a)	(b)	(c)	n
*P ≤ 0·05, **P ≤ 0·01 and ***P ≤ 0·001. Data for PNUE calculated at saturating (1500 μmol photons m^−2 s^−1) and at growth irradiance (300 μmol photons m^−2 s^−1). LMA, leaf mass per unit leaf area.
LMA (g m^−2)	Light saturated PNUE (μmol CO2 g N^−1 s^−1)	***	NS	NS	75	***	NS	NS	67
LMA (g m^−2)	Growth irradiance PNUE (μmol CO2 g N^−1 s^−1)	***	NS	*	47	***	NS	NS	45
LMA (g m^−2)	Light saturated PPUE (μmol CO2 g N^−1 s^−1)	**	NS	NS	60	***	NS	NS	53
LMA (g m^−2)	Growth irradiance PPUE (μmol CO2 g N^−1 s^−1)	***	NS	NS	34	***	NS	NS	34

Discussion

Growth temperature and photosynthesis-leaf trait scaling relationships

Our study sought to quantify the impact of growth temperature on scaling relationships linking photosynthetic CO₂ assimilation to a range of leaf traits. Key to addressing this objective was quantifying how much of the scatter in log–log scaling relationships linking light-saturated A (measured at a common temperature of 21 °C, irrespective of the growth temperature treatment) to related leaf traits (Figs 1 and 2) is due to differences in growth temperature, and whether there is a consistent effect of growth temperature on the slope and intercept of such scaling relationships. We measured A_sat at the common temperature of 21 °C because: (i) this temperature is likely to be close to the T_opt of A_sat of the selected species (Campbell et al. 2007); (ii) growth temperature-mediated changes in the T_opt of A_sat are unlikely to alter the relative difference in rates of A_sat exhibited by cold and warm grown plants, when rates of A are measured at moderate temperatures (Atkin, Scheurwater & Pons 2006b); and, (iii) comparison of rates of A in ND leaves of 7 °C and 21 °C grown plants measured at each respective growth temperature revealed similar trends to those observed when using 21 °C measured values (Campbell et al. 2007). Overall, our results highlight the importance of developing leaves at the new growth temperature in determining the effects of sustained changes in growth temperature on log–log scaling relationships linking photosynthesis to other leaf traits.

We found that the temperature experienced in the 10 days prior to measurement accounted for 14–17% and 21–31% of the variation in A_sat on a mass and area basis, respectively, in pre-existing (PE) leaves that had fully developed at a moderate temperature prior to experiencing a change in growth temperature. Exposure of PE leaves to chilling temperatures for 10 days resulted in reduced rates of A (measured at a common temperature), resulting in changes in the intercept and/or slope of scaling relationships linking A to the other leaf traits, including LMA, N_mass and P_mass. Thus, prolonged changes in growth temperature have the potential to change the fundamental scaling between A and related leaf traits in PE leaves, and account for a substantial proportion of the scatter in log–log plots. By contrast, growth temperature had no significant effect on the slope or intercept of log–log A-LMA-N-P scaling relationships in newly-developed (ND) leaves that formed at each respective growth temperature. Thermal history also played almost no role in accounting for scatter in the log–log A-LMA-N-P plots in ND leaves. This finding contrasts with the strong impact that thermal history has on the elevation of scaling relationships used to predict rates of leaf R in leaves that develop under steady-state contrasting temperatures (Atkin et al. 2008), further highlighting the asynchronous effects of growth temperature on respiratory and photosynthetic metabolism (Campbell et al. 2007).

Why are A-LMA-N-P scaling relationships altered in PE leaves that experience prolonged chilling, whereas ND leaves developed at different temperatures exhibit similar overall A-LMA-N-P scaling relationships? In the case of A_mass-LMA relationships in PE leaves (Fig. 2a), we found that there was a greater divergence in A_mass values between temperature regimes at high LMA values. This could be due to: (i) accumulation of total non-structural carbohydrates (TNC) in the cold (Campbell et al. 2007) resulting in proportional changes in A_mass that were greater than the changes in LMA (i.e. reduced rates of A_mass at any given LMA); and (ii) greater proportional declines in A_mass in species exhibiting inherent high LMA values than their low LMA counterparts. Inter-specific differences in temperature-induced changes in LMA may also have contributed (Atkin, Scheurwater & Pons 2006b). The observed disparity in variance between the data sets from different growth temperatures in the A_sat× N_mass relationship suggests that there was a greater variation in A between species in response to a drop in temperature than to an increase.

Underpinning the similar log–log A-LMA-N-P scaling relationships in ND leaves that developed at different temperatures was a recovery (i.e. acclimation) of A in leaves that developed at 7 °C (compared to PE leaves that were cold-treated for 10 days). Acclimation of A in cold-developed leaves was particularly evident when rates were expressed on a leaf area basis, with acclimation being associated with a recovery in F_v/F_m (Campbell et al. 2007). The latter is likely underpinned by increases in triose phosphate use (TPU) (Sharkey et al. 1986; Sage & Sharkey 1987), which is in turn dependent on changes in the concentration of proteins that limit photosynthetic capacity and assimilate processing (Hurry et al. 1995a; Savitch et al. 2001; Stitt & Hurry 2002; Yamori, Noguchi & Terashima 2005) and redistribution of inorganic phosphate between cellular compartments (Hurry et al. 1993b, 2000; Stitt & Hurry 2002; Strand et al. 2003). Changes in photosynthetic electron transport capacity relative to carboxylation capacity (Hikosaka, Murakami & Hirose 1999; Onoda, Hikosaka & Hirose 2005a; Yamori, Noguchi & Terashima 2005; Atkin, Scheurwater & Pons 2006b) and the degree of unsaturation of membrane lipids (Raison, Roberts & Berry 1982; Nishida & Murata 1996) may also play a role. In addition to such changes (which mediate recovery of A), cold-acclimated ND leaves also exhibit large changes in N concentration and LMA (Körner & Larcher 1988; Tjoelker, Reich & Oleksyn 1999b; Atkin et al. 2006a; Campbell et al. 2007; Poorter et al. 2009; Gorsuch, Pandey & Atkin 2010). Collectively, such changes result in similar log–log A-LMA-N-P scaling relationships remaining constant in ND leaves that developed at different temperatures.

In studies that investigate the effect of environmental gradients on leaf log–log scaling relationships, Wright, Reich & Westoby (2001) outlined several possible outcomes: (i) no impact of the environment on either the x or y-axis parameter, with the result that there is no change in the slope or elevation of the log–log scaling relationship; (ii) the environmental factor does affect the magnitude of the x and y-axis parameters – however, in cases where the proportional change in the x or y-axis parameters is similar, no change in the slope or elevation of the scaling relationship would occur, even though the spread of the data cloud might have shifted [(i.e. the ‘shift 1’ scenario shown in Fig. 1 in Wright, Reich & Westoby (2001)]; (3) environment-dependent changes in leaf traits occur, but with the proportional change in x-axis parameters being different from the proportional change in y-axis parameter, with the result that there is a change in the elevation and/or slope of the scaling relationship [i.e. ‘shift 2’ scenario; Wright, Reich & Westoby (2001)]. The significant effect of growth temperature on the elevation of log–log A-LMA-N-P relationships in PE leaves (Fig. 2) indicates that ‘shift 2’ changes in species-trait relationships occurred when 21 °C grown plants were shifted to 7 °C for 10 days, underpinned by an 65% decline in average rate of A_mass and a 27% increase in average LMA values. By contrast, we observed no temperature-induced change in the elevation or slope of the log–log A-LMA-N-P relationships in ND leaves (Fig. 2), indicating either that either there was no effect of growth temperature on x or y-axis parameters in ND leaves, or that a ‘shift 1’ response occurred. The latter seems likely, as we have previously shown that there are substantial temperature-induced changes in LMA and A_mass in ND leaves (Campbell et al. 2007), and that there is no systematic difference among species in plasticity/acclimation of A_mass or LMA when plants experience contrasting growth temperatures (Atkin et al. 2006a). The net result is that growth temperature has no overall effect on log–log scaling relationships in ND leaves (Fig. 2), even though growth temperature has substantial effects on leaf traits of those leaves.

Efficiency of nutrient use

Our study also sought to determine whether photosynthetic nutrient use efficiency is systematically altered by growth temperature across a wide range of contrasting species differing in LMA values. When all data were considered together irrespective of growth temperature, we found a negative relationship between PNUE and LMA, and PPUE and LMA, as has been reported previously in comparisons of photosynthetic nutrient use efficiency in contrasting plants species (Poorter & Evans 1998; Pons & Westbeek 2004; He et al. 2006; Hidaka & Kitayama 2009). Several factors can influence inter-specific variations in PNUE, including the proportion of leaf N allocated to the photosynthetic apparatus (Pons & Pearcy 1994) and/or the allocation of N within the photosynthetic apparatus (e.g. light harvesting, carbon reduction and electron transport); (Hikosaka & Shigeno 2009). The extent to which N is allocated to cell walls can also play a role in some cases (Hikosaka 2004; Onoda, Hikosaka & Hirose 2004; Feng et al. 2009; Harrison et al. 2009) but not others (Hikosaka & Shigeno 2009). Although the factors responsible for inter-specific differences in PPUE are not known, it seems likely that a similar range of factors will be involved, particularly the leaf P allocated to photosynthesis.

Given the negative effect that 10 days exposure to 7 °C had on rates of A_area and A_mass in PE leaves (Fig. 2), and reduced values of F_v/F_m values exhibited by the cold-treated plants (Campbell et al. 2007), we had expected that PNUE would also be lower in 7 °C treated PE leaves. Although there was no overall significant effect of temperature treatment on PNUE (Fig. 3) or a significant interaction between growth temperature and LMA (Fig. 3, Table 5), values of PNUE (under saturating and growth irradiance) appeared to be lower in the 7 °C treatment than for the 21 and 28 °C treatments in PE leaves, with some recovery being observed in ND leaves (Fig. 3). Moreover, a recent study using 11 crop species reported that low growth temperatures does reduce PNUE (Yamori et al. 2009). Thus, it seems likely that a general response of PE leaves to sustained chilling might be a reduction in the photosynthetic nitrogen use efficiency. Importantly, however, our results strongly suggest that there is unlikely to be any systematic difference in temperature-mediated reductions in PNUE among species along the ‘leaf economic spectrum’ (Wright et al. 2004), as evidenced by the lack of interaction between LMA and temperature (Fig. 3, Table 5) and the similar slopes in log–log A-N plots (Fig. 2).

Our data set showed that the relationship between N and P was unaffected by temperature (Table 3). This would infer that as N increased at lower temperatures (Campbell et al. 2007) then P would also increase. However, due to substantial scatter in P values no detectable effect of temperature on P content was found. As our data was collected under conditions of nutrient sufficiency in controlled conditions this might indicate that in field conditions any effect of temperature on levels of leaf N and P might be through impacts on nutrient availability (and hence on plant growth), rather than through the plants’ physiological responses. Other studies have shown that N : P ratios are determined mainly by variation between species and between sites (He et al. 2008; Kerkhoff et al. 2006) or soil fertility (Ordonez et al. 2009) rather than by climatic variables. Alternatively, luxury storage of N and P is a possible explanation; NO₃⁻ was typically less than 5–10% of total N (data not shown) in our leaves and thus luxury consumption of N is unlikely to have played a substantive role. However, we cannot rule out P accumulation as a factor, particularly given that Broadley et al. (2004) demonstrated that angiosperm species grown in hydroponics had lower N : P ratios than field-grown plants. In addition, inorganic P accumulation can occur at low temperatures (Hurry et al. 2000). Further work is needed to determine the extent of P accumulation [e.g. using the P fractionation method of Kedrowski (1983)] and the impact that it would have on the balance between N and P.

Conclusions

Our analysis has shown that accounting for the growth temperature experienced by leaves is important in understanding the factors determining scaling relationships linking photosynthesis to other leaf functional traits, particularly when considering the effects of large, sustained changes in growth temperature on already fully-expanded leaves. Importantly, log–log A-LMA-N-P scaling relationships in different temperature regimes converge once new leaves form at a new growth temperature, highlighting the crucial role development plays in determining impacts of the environment on scaling relationships. Our results also demonstrate for the first time that there is no systematic variation among a wide range of contrasting plant species in the impact of growth temperature on photosynthetic nutrient use efficiency. This finding, if more widespread, could have important implications for our understanding how of how thermal and nutrient gradients may impact on the efficiency of nitrogen and phosphorus use in plant species representative of opposite ends of the ‘leaf economic spectrum’ (Wright et al. 2004).

Acknowledgements

This work was supported by a Natural Environment Research Council (NERC) research grant to OKA and VH (NER/A/S/2001/01186). We would like to thank David Sherlock for expert technical assistance.