Leaf photosynthetic, economics and hydraulic traits are decoupled among genotypes of a widespread species of eucalypt grown under ambient and elevated CO₂

Summary

1. Leaf economics and hydraulic traits strongly influence photosynthesis. While the level of coordination among these traits can differ between sets of species, leaf functional trait coordination within species remains poorly understood. Furthermore, elevated concentrations of atmospheric CO₂ commonly influence the expression of leaf photosynthetic, economics and hydraulic traits in contrasting ways, yet the effect of variable concentrations of atmospheric CO₂ on patterns of trait coordination within species remains largely untested.
2. We examined the relationships among key leaf photosynthetic (e.g. net photosynthesis and photosynthetic biochemistry), economics and water-use (e.g. leaf mass per unit area and stomatal conductance) and hydraulic traits (e.g. vein density) in 14 genotypes of Eucalyptus camaldulensis grown in ambient (aCO₂) and elevated (eCO₂) [CO₂]. We examined the level of coordination among leaf traits in aCO₂ and then assessed whether growth in eCO₂ altered that coordination.
3. We found that leaf traits related to photosynthetic capacity, economics and water-use, and hydraulics were decoupled among genotypes grown in aCO₂, yet strong relationships were generally observed among suites of traits within each ‘functional group’.
4. Significant responses to growth in eCO₂ were observed for most leaf photosynthetic and economics and water-use traits, with the magnitude and direction of the response varying among traits. In contrast, leaf hydraulics traits were unaffected by variable growth CO₂. Despite this, growth in eCO₂ did not substantially alter patterns of leaf trait coordination observed in aCO₂.
5. These results suggest suites of leaf traits associated with photosynthetic capacity, economics and water-use and hydraulics, respectively, can form independent axes of variation among genotypes of a single species, regardless of growth CO₂. Although growth in eCO₂ did not substantially alter patterns of trait coordination, decoupling of leaf functional traits among genotypes may allow genetically distinct populations to produce novel combinations of traits that may be adaptive in response to changes in their local environment.

Introduction

Studies of functional trait coordination in leaves have greatly advanced our understanding of leaf structure and function (Givnish 1987; Sack et al. 2003; Brodribb, Feild & Sack 2010) and helped identify the biophysical constraints and trade-offs that underlie key ecological strategies (Westoby et al. 2002). From an economics perspective, consistency in the relationships among many leaf functional traits within and across biomes is suggestive of a unified plant traits spectrum, whereby all species are characterized along a single axis of variation in terms of fast, medium or slow plant function (Reich 2014). Across species, evidence for a single axis of trait variation is drawn from correlations among leaf economics traits that define plant ecological strategies with fast to slow photosynthetic payback relative to leaf investment (Wright et al. 2004), and from evidence that leaves with more efficient hydraulic systems tend to have higher photosynthetic rates (Sack et al. 2003; Brodribb, Feild & Jordan 2007). The common connection between leaf economics and leaf hydraulics in serving photosynthesis has also led to proposals that hydraulic traits such as vein density (VD) are predictive of the leaf economics spectrum (Blonder et al. 2011). However, leaf functional traits may not always vary in such a coordinated manner. Recent studies in diverse forest systems have observed trait decoupling among drought tolerance and economics traits (Marechaux et al. 2015), and among leaf hydraulic and leaf economics traits (Li et al. 2015), suggesting that different functional trait combinations are possible within leaves. Functional trait coordination in leaves might also be decoupled in response to environmental change such as elevated atmospheric [CO₂] (eCO₂), which commonly induces differential responses in leaf functional traits.

A common response of C₃ plants to eCO₂ is increased net photosynthesis and decreased stomatal conductance. Longer term growth under eCO₂ may also lead to photosynthetic down-regulation (Ainsworth & Rogers 2007) and changes in a suite of functional and structural traits associated with leaf gas exchange and water transport (hydraulic) systems (Franks et al. 2013; Rico et al. 2013). Plant growth in eCO₂ may reduce stomatal density (Woodward & Kelly 1995) via feedback loops that optimize water loss relative to long-term carbon gain (Franks et al. 2012). Plant growth in eCO₂ may also reduce leaf traits associated with water supply, such as VD, so that investment in water transport capacity is also optimized relative to photosynthetic carbon gain (Brodribb & Jordan 2011). However, experimental studies suggest that leaf hydraulics can be decoupled from gas-exchange traits in eCO₂ (Uhl & Mosbrugger 1999; Locke et al. 2013). Leaf-level responses to eCO₂ may also include increased leaf mass per area (LMA), due to increases in leaf thickness and carbohydrate accumulation, and a reduction in leaf nitrogen content (Ainsworth & Long 2005; Leakey et al. 2009). Taken together, such potentially differential responses to growth in eCO₂ may act to alter patterns of trait coordination within leaves.

Species often vary substantially in their physiological response to eCO₂ (Ainsworth & Rogers 2007; Hovenden & Williams 2010). However, intraspecific (i.e. genotypic) variation in trait responsiveness may also be significant (Bernacchi et al. 2003; Davey et al. 2006) and important in determining population and species performance under climate change (Moran, Hartig & Bell 2016). Despite this, the linkages among leaf functional traits are rarely examined within a given species (but see Aspinwall et al. 2013; Niinemets 2015), and it is unclear the degree to which variable CO₂ may alter coordination of leaf functional traits within species. Determining leaf functional trait coordination among genotypes could provide insight into patterns of physiological divergence within species. Moreover, determining how eCO₂ influences leaf trait coordination among genotypes could provide an integrative view of genotypic variation in leaf functional trait responses to eCO₂.

Widely distributed species of Eucalyptus represent potentially useful and important models for studying patterns of leaf functional trait coordination within species. Eucalypts dominate forest and woodland communities throughout Australia and are economically important as major forestry species worldwide. Numerous studies have examined growth and physiological traits among Eucalyptus provenances in response to changes in soil moisture, salinity, light and temperature (Bedon et al. 2012; Bush et al. 2013; Drake et al. 2015), while other studies have examined changes in leaf structure and function in response to variable growth [CO₂] in individual eucalypt species (Smith et al. 2012). However, to the best of our knowledge, there are no studies that have examined the relationships between leaf functional traits among Eucalyptus genotypes and how these relationships are potentially altered under eCO₂.

Here, we assessed the linkages among key leaf photosynthetic, economics and hydraulic traits (see Table 2) across 14 Eucalyptus camaldulensis (subsp. camaldulensis) genotypes, representing six provenances, grown under non-limiting water and nutrient conditions at ambient [CO₂]. We tested the specific hypothesis that genetic variation in leaf photosynthetic, economic and hydraulic traits would occur along a single axis of variation (Reich 2014) on the basis that (i) leaf hydraulic traits (especially VD) underpin key aspects of the leaf economics spectrum (Blonder et al. 2011); and (ii) both leaf hydraulic and economics traits influence rates of leaf photosynthesis (Wright et al. 2004; Brodribb, Feild & Jordan 2007; Wyka et al. 2012). We also examined genetic variation in the response of each trait to growth in eCO₂. Because the magnitude and direction of the eCO₂ response can vary among leaf traits, we also hypothesized that growth in eCO₂ will alter patterns of coordination of leaf functional traits among genotypes.

Materials and methods

Species and Experimental Conditions

Fourteen genotypes of Eucalyptus camaldulensis subsp. camaldulensis (River Red Gum) were included in this study. Eucalyptus camaldulensis is distributed widely across mainland Australia and is an important component of floodplain and riparian environments. The distribution of subspecies, camaldulensis, encompasses much of the Murray – Darling River basin of New South Wales and Victoria and extends into parts of Queensland and South Australia (Atlas of Living Australia). The subspecies’ distribution extends across significant gradients of temperature (mean annual temperature range 13·6–19·2 °C) and precipitation (mean annual precipitation range is 275–750 mm). Previous work has found limited genetic structure among populations within this subspecies, indicating high levels of gene flow and genetic diversity (Dillon et al. 2015).

In our study, clonal plants were propagated from each genotype that had been grown from seed collected from an open-pollinated mother tree (i.e. half-sib family). The mother trees originated from six provenances representing different geographic and climatic origins throughout the subspecies distribution (Table 1). The clones of each genotype were prepared via vegetative propagation of rooted cuttings. The cuttings were established in 38 mm diameter × 210 mm tall containers filled with potting mix, and grown in a common shade house for roughly 2 months. The cuttings were transplanted into 6·9 L cylindrical pots with several ~1 cm diameter drainage holes and grown in a naturally lit glasshouse facility at Western Sydney University in Richmond NSW, Australia (see Resco de Dios et al. 2015), once average stem height (H) and basal diameter (D) reached 25 cm and 2 mm respectively. Each pot contained 7·5 kg of coarse textured soil (Australian Native Landscape Pty, Ltd, Bagerys Creek, NSW, Australia), with a pH of 6·5. To ensure that no nutrient limitations occurred, the plants were fertilized every fortnight with a commercial liquid fertilizer (500 mL Aquasol, at 1·6 g L⁻¹; 23% N, 4% P, 18% K, 0·05% Zn, 0·06% Cu, 0·013% Mo, 0·15% Mn, 0·06% Fe, 0·011% B; Yates Australia, Padstow, NSW, Australia).

Table 1. Geographic and climatic origin of the 14 Eucalyptus camaldulensis sub sp. camaldulensis genotypes included in this study

Provenance	Lat (°S)	Long (°E)	MAT (°C)	MAP (mm)	ET (mm)	G
Coonawarra	37·2	140·42	14·5	646	1029	2
Yass River	34·53	149·02	13·9	395	1141	2
Ovens Valley	36·36	146·47	15	331	1214	2
Barmah	35·5	145·07	16·3	644	1331	2
Condobolin	33·06	147·09	17·6	393	1440	2
Nyngan	31·33	147·11	19·2	345	1565	4

• MAT, mean annual temperature; MAP, mean annual precipitation; ET, evaporation, G, number of genotypes in the study from each provenance.

Eight to 16 replicate clones of each genotype were grown in two CO₂ concentrations; ambient (aCO₂, 400 μmol mol⁻¹) and elevated (eCO₂, 640 μmol mol⁻¹). The study design was a randomized split-plot design with two glasshouse rooms per CO₂ treatment. For each CO₂ treatment, replicate plants of each genotype were split into two groups and randomly assigned to one of the two rooms. Each group (glasshouse room) of plants was considered a block, and the unit of replication was the individual plant. Within each room, plants were rotated weekly to reduce potential location effects on plant performance. Mean daytime and night-time air temperature and relative humidity within the glasshouse were 26 and 17 °C, and 45 and 60% respectively, which is representative of average summer values in Richmond. All pots were hand-watered daily (using charcoal-filtered tap-water) to field capacity which ensured that soil moisture was not limiting.

Gas-Exchange Parameters

Leaf-level net photosynthesis CO₂ response (A-C_i) curve measurements were made on all replicates of each genotype in each growth CO₂ treatment using several cross-calibrated Li-Cor 6400-XT portable infrared gas analysers (Li-Cor Inc., Lincoln, NE, USA) with a 2 × 3 cm² cuvette head and a red and blue LED light source. A-C_i curve measurements were made on mature, fully expanded, upper canopy leaves after roughly 60 days of growth under the CO₂ treatments. Measurements were conducted on consecutive days (~9), and gas-exchange analysers were randomly assigned to glasshouse bays each day. The order in which replicates of each genotype within each CO₂ treatment were measured was random. For all measurements, light conditions within the cuvette were controlled at a photosynthetic photon flux density of 1800 μmol m⁻² s⁻¹ and block temperature was fixed at the daytime growth temperature of 26 °C. The Li-Cor 6400 desiccant tubes were used to control water vapour inside the cuvette, so that vapour pressure deficit (VPD) was 1·4 ± 0·02 (standard error) kPa. Prior to beginning each A-C_i curve, steady-state rates of leaf net photosynthesis (A_net, μmol m⁻² s⁻¹) and stomatal conductance to water vapour (g_s, mol m⁻² s⁻¹) were measured at the growth [CO₂]. Following steady-state measurements, A-C_i curves were produced by measuring A_net at 10 reference CO₂ (C_a) concentrations, ranging from 0 to 1000 μmol mol⁻¹. A biochemical model of photosynthesis (Farquhar & von Caemmerer 1982) was used to parameterize the A-C_i curve data for each plant. The model estimates the maximum rate of carboxylation (V_cmax; μmol m⁻² s⁻¹) and the potential rate of electron transport for RuBP regeneration at 1000 μmol m⁻² s⁻¹ (J₁₀₀₀; μmol m⁻² s⁻¹). Because we did not measure leaf mesophyll conductance to CO₂ (g_m), which is known to show variable responses to growth under eCO₂ across species (Flexas et al. 2012), V_cmax and J₁₀₀₀ are apparent values based on intercellular [CO₂] (C_i) rather than [CO₂] concentration at the chloroplast. The model was fit using the plantecophys package (Duursma 2015) in r v3.2 (R Core Team, 2015).

VD and Stomatal Traits

Vein density and stomatal traits were measured on fully expanded leaves sampled from the upper canopy of individual plants after roughly 70 days of growth under the CO₂ treatments. To determine VD, we sampled one leaf from each of three replicate plants of each genotype within each CO₂ treatment. Veins located centrally in the lamina of each leaf were exposed by removing an area of ~1 cm² of epidermis and top layers of palisade mesophyll with a razor blade. These ‘windows’ into the leaf venation were cut out and chemically cleared using 20% household bleach. Cleared leaf sections were stained with saffranine-O and counter stained with fast-green and semi-permanently mounted on glass slides using glycerol jelly. VD was calculated as the total length of vein (mm) per unit area (mm²) of image taken at 10× magnification using a digital camera (Leica DCF 500, Cambridge, UK) mounted to a light microscope (Olympus BX60, Center Valley, PA, USA). Each image focused on an area of leaf venation away from 2nd order veins, which do not contribute substantially to total vein length compared to higher order veins (Sack & Scoffoni 2013). Image analysis was performed using imagej (http://imagej.nih.gov/ij/).

Stomatal traits were measured from one upper canopy leaf sampled from each of four to five plants per genotype. Epidermal impressions were made in the midsection of abaxial and adaxial leaf surfaces using clear nail varnish. The dried layers of nail varnish were peeled off the leaf surface using clear tape and mounted on glass slides. Three images were taken at 20× magnification using a digital camera mounted to a compound light microscope. Stomatal density (SD) was calculated as the number of stomata per image area (mm⁻²) averaged across both leaf surfaces. Stomatal aperture (μm) and single guard cell width (μm) were also measured on between 10 and 30 well-defined stoma on each leaf surface, respectively, and then averaged. Measurements of stomatal size and density were used to calculate the theoretical maximum stomatal conductance to water vapour (g_wmax) derived from Franks & Farquhar (2001):

where d is the diffusivity of water vapour in air (m² s⁻¹); ʋ, the molar volume of air (m³ mol⁻¹); D, stomatal density (m⁻²); a_max, the maximum area of the open stomatal pore (m²); l, the depth of the stomatal chamber (equivalent to the width of a single guard cell). The a_max was estimated as π (p/2)², where p is the length of the stomatal pore.

Leaf Economics Traits

Plants were harvested within a week, after approximately 80 days of growth under the CO₂ treatments. Total tree leaf area (LA, cm²) of each plant was determined by measuring the LA of all leaves using a LA meter (LI-3100C; Li-Cor Inc, Lincoln, Nebraska). Total leaf dry mass was determined after oven drying at 70 °C to a constant mass, and a specific leaf area (SLA, cm² g⁻¹) was calculated as the ratio of LA to total leaf dry mass. Leaf carbon (C) and nitrogen (N) content were determined on a subsample of dried leaf material from each tree collected at the harvest. Leaves were ground into a fine powder using a ball grinder, stored under desiccation, and leaf N concentration were determined using a combustion elemental analyser (CE Instruments, Wigan, UK). Nitrogen per unit LA (N_area; g [N] m⁻²) was obtained by the quotient of N per unit leaf mass (g kg⁻¹) and adjusted to a m² basis. Following determination of non-structural carbohydrates (see below), we calculated TNC-free LMA (LMA_noTNC).

Leaf starch (L_starch; g kg⁻¹) and leaf soluble sugars (L_sugars; g kg⁻¹) were determined on recently mature, sunlit leaves, sampled from the upper canopy. Leaf collections were made mid-morning, after 11 weeks of plant growth under the treatment conditions. The leaf samples were flash frozen in liquid N, freeze-dried, ground to a fine powder in a ball mill, and then extracted three times with 2 mL of a methanol-chloroform-water (12 : 5 : 3 v/v) solution to separate the soluble sugars from the starch (pellet fraction). Starch was hydrolyzed from the pellet with 5 mL of perchloric acid (35% v/v) for 30 min. L_starch and L_sugars concentrations were determined using a colorimetric phenol-sulphuric method (Tissue & Wright 1995).

Leaf ¹³C composition (δ¹³C_leaf) was measured using an Isochrom continuous-flow mass spectrometer (Micromass, Manchester, UK) following combustion in the elemental analyser. δ¹³C_leaf was converted to discrimination values using mean values for previously measured ¹³C composition in the aCO₂ and eCO₂ treatments in the glasshouse: δ¹³C_air; −9·83‰ and −17·02‰ for aCO₂ and eCO₂, respectively, relative to Vienna Pee Dee Belemnite (Lewis et al. 2015). Leaf carbon isotope discrimination (Δ¹³C) is inversely related to δ¹³C_leaf, and is calculated as: Δ¹³C = (δ¹³C_air – δ¹³C_leaf)/(1 + δ¹³C_leaf/1000) (Farquhar & Richards 1984).

Data Analysis

Linear mixed effects models were conducted using the ‘lme4’ package (Bates et al. 2015) in r v3.2 (R Core Team, 2015) to test the main effects of growth treatment CO₂ concentration (CO₂), genotype (G) and their interactions (CO₂ × G) on all response variables, with the exception of SD and g_wmax which had not been assigned to individual replicates at the time of sampling (a Student's t-test was used to compare SD and g_wmax across CO₂ treatments). In the mixed effects models, glasshouse bay nested within CO₂ treatment was treated as a random effect. We assessed homogeneity of variance by examining plots of the residuals (Y-axis) vs. the predicted values (X-axis) from each model and ensured that the residuals were normally distributed around ‘0’ and variance in the residuals did not vary with changes in the predicted values. If the CO₂ × G term was significant in the model, we interpreted this as an indication that genotypes differed in their response to eCO₂ (indicating intraspecific variation in phenotypic plasticity). Genotype mean CO₂ response ratios were also calculated for each trait [i.e. mean (eCO₂)/mean (aCO₂)].

We did not test the effect of ‘provenance’ for several reasons. First, our primary objective was to examine patterns of leaf trait coordination across as many genotypes as possible. Secondly, a related objective was to determine the degree of genetic differentiation in leaf traits, and examine relationships between genotype leaf trait means and responsiveness to eCO₂ (i.e. plasticity). With six provenances, neither analysis is feasible at the provenance level. In addition, the number of genotypes and replicate plants from within each provenance was unbalanced, making tests of provenance differences and provenance variation in responses to eCO₂ (i.e. CO₂ × provenance) problematic to interpret.

Pearson correlation analysis was performed for all pairwise trait comparisons of genotype means within each CO₂ treatment. We also examined covariation among genotype means within each CO₂ treatment for leaf photosynthetic, economics and hydraulic traits (with the exception of L_starch and L_sugars) using principal components analysis (PCA) with orthogonal (varimax) factor rotation in sas v9.3 (PROC FACTOR, SAS Institute Inc. 2010 Cary, NC). We excluded L_starch and L_sugars from the PCA because these traits had not been measured across all 14 genotypes. Parallel analysis was used to determine the number of principal components (PC's) to retain (Franklin et al. 1995; Peres-Neto, Jackson & Somers 2005).

Results

Genotypes showed substantial differentiation in many leaf physiological, economics and hydraulic traits (Table 2). For example, averaged across both CO₂ treatments, N_area and LMA varied by 42% and 41%, respectively, among genotypes. In comparison, genotypes showed comparatively less variation in A_net (18%) and VD (16%). Stomatal traits such as SD (37%) and g_wmax (43%) also varied substantially among genotypes, as did the ratio of SD displayed on each side of the leaf (i.e. the degree of amphistomy; 31%). With the exception of L_sugars, genotypes showed no significant (P > 0·05) variation in leaf trait responses to variable growth [CO₂], i.e. there was no evidence of G × CO₂ interactions (Table 2).

Table 2. Leaf photosynthetic, leaf economics and water-use, and leaf hydraulic traits of 14 genotypes of Eucalyptus camaldulensis sub. camaldulensis grown in ambient (aCO₂) and elevated eCO₂ [CO₂], expressed as means (±standard errors). Analysis of Variance (anova) statistics for CO₂, genotype and CO₂ × genotype effects are presented for all leaf traits, with the exception of SD and g_wmax

Trait	Symbol	Unit	n	aCO2	eCO2	anova results
						CO2		Genotype (G)		G × CO2
						d.f.	χ^2	d.f.	χ^2	d.f.	χ^2
Photosynthetic
Net photosynthesis at growth CO2	A net	μmol m^−2 s^−1	14 (154)	26·0 ± 0·4	33·4 ± 0·5	1	199***	13	42·9***	13	21·5
Max. rate of Rubisco carboxylation	V cmax	μmol m^−2 s^−1	14 (153)	109 ± 2	102 ± 2	1	9·10**	13	36·6***	13	17·2
Potential rate of electron transport	J 1000	μmol m^−2 s^−1	14 (153)	198 ± 3	198 ± 4	1	0·0008	13	22·4*	13	6·55
Leaf economics and water-use
Stomatal conductance	g s	mol m^−2 s^−1	14 (144)	0·72 ± 0·03	0·58 ± 0·03	1	19·7***	13	22·2	13	15·6
Leaf mass per unit area	LMA	g m^−2	14 (156)	69·4 ± 1·6	83·2 ± 1·8	1	67·1***	13	165***	13	8·02
Nitrogen per unit leaf area	N area	g [N] m^−2	14 (156)	1·80 ± 0·05	1·71 ± 0·04	1	6·09*	13	183***	13	6·67
Leaf starch concentration	L starch	g kg^−1	9 (98)	81·7 ± 4·4	110 ± 5	1	21·0***	8	37·2***	8	1·63
Leaf soluble sugar concentration	L sugars	g kg^−1	9 (98)	49·9 ± 1·7	49·5 ± 1·4	1	0·05	8	32·4***	8	38·8***
TNC-free LMA	LMA noTNC	g m^−2	9 (98)	60·2 ± 1·8	68·7 ± 1·9	1	30·5***	8	196***	8	3·9
Leaf carbon isotope iscrimination	Δ^13C	‰	14 (156)	24·0 ± 0·1	20·7 ± 0·1	1	912***	13	75·4***	13	5·19
Leaf hydraulics
Vein density	VD	mm mm^−2	14 (111)	10·5 ± 0·1	10·6 ± 0·1	1	0·9	13	73·8***	13	18·9
Stomatal density	SD	no. mm^−2	14 (136)	311 ± 7	306 ± 7
Maximum stomatal conductance to water vapour	g wmax	mol m^−2 s^−1	14 (136)	3·03 ± 0·12	2·91 ± 0·08

• LMA, leaf mass per area.
• Chi-squared values denoted with ***, **, and *are significant at P < 0·0001, P < 0·01 and P ≤ 0·05 respectively.

Averaged across genotypes, growth in eCO₂ significantly increased A_net, LMA and L_starch and decreased g_s, V_cmax and N_area, while there was no significant [CO₂] treatment effect on J₁₀₀₀, L_sugars and VD (Fig. 1; Table 2). In addition, LMA_noTNC significantly decreased in eCO₂ indicating that eCO₂-induced increases in LMA were due to leaf structural (leaf thickness) and carbohydrate level changes. Student's t-test comparison of stomatal traits, which were not included in the anova indicated that eCO₂ did not affect SD (t = 0·51, P = 0·6) or g_wmax (t = 0·89, P = 0·38).

Figure 1

Mean CO₂ response ratios for all leaf photosynthetic, leaf economics and leaf hydraulic traits measured in 14 Eucalyptus camaldulensis genotypes grown in aCO₂ (400 p.p.m.) and eCO₂ (640 p.p.m.). Values are averages calculated across genotype mean response ratios (eCO₂/aCO₂). Error bars indicate the standard error across genotype mean response ratios.

Trait Coordination in aCO₂

Under aCO₂, PCA of genotype means identified three independent axes which cumulatively explained 78% of the total variation in genotype mean leaf trait values (Table 3; Fig. 2). The first PC accounted for 35% of the total variation and was correlated with leaf economics and water-use traits (positively with LMA, N_area and g_wmax, and negatively with g_s and Δ¹³C). The second PC (32% of total variation) was positively correlated with photosynthetic traits A_net, V_cmax and J₁₀₀₀. Finally, the third PC (12% of total variation) was positively correlated with hydraulic traits VD, SD and g_wmax.

Table 3. Principal component (PC) factor loadings for selected leaf photosynthetic, leaf economics and leaf hydraulic traits across genotype means in Eucalyptus camaldulensis subsp. camaldulensis measured in aCO₂ and eCO₂

Growth CO2	aCO2			eCO2
	PC 1	PC 2	PC 3	PC 1	PC 2	PC 3
Eigenvalue	3·51	3·15	1·15	3·86	2·2	1·68
Variance explained	35%	32%	12%	38%	22%	17%
Photosynthetic
A net	−0·25	0·90	−0·28	−0·25	0·85	0·37
V cmax	0·01	0·91	−0·17	0·17	0·90	0·06
J 1000	0·14	0·91	−0·10	−0·07	0·95	−0·01
Leaf economics
g s	−0·83	−0·09	−0·06	0·12	0·30	0·68
Leaf mass per area	0·87	−0·14	−0·22	0·94	−0·17	−0·01
N area	0·93	0·04	−0·13	0·84	−0·12	0·01
Δ^13C	−0·81	0·0	0·07	−0·74	0·13	0·22
Leaf hydraulics
Vein density	−0·09	−0·14	0·83	−0·01	0·15	0·91
SD	−0·13	−0·34	0·74	0·67	−0·24	0·43
g wmax	−0·65	−0·07	0·51	−0·19	−0·07	0·83

• Bolded values are those with Eigen scores >0·5, indicating significant association with corresponding PC axes. Trait abbreviations are provided in Table 2.

Figure 2

Principal component analysis (PCA) highlighting independent axes of leaf functional traits among 14 Eucalyptus camaldulensis ssp. camaldulensis genotypes grown in ambient (plots a, c, e) and elevated (plots b, d, f) [CO₂]. Trait abbreviations are given in Table 2. The PCA identified three major axes of variation associated with separate ‘functional sets’ of traits (Table 3): leaf economics and water-use traits (rectangles; PC 1), leaf photosynthetic traits (triangles; PC 2) and leaf hydraulic traits (ovals; PC 3). The trait loading plots (a–f) represent bivariate comparisons of each of the three major axes of trait variation at each CO₂ level: PC 1 vs. PC 2 (plots a–b), PC 1 vs. PC 3 (plots c–d) and PC 2 vs. PC 3 (plots e–f). Bolded line coloring and ‘functional set’ symbols are shown only when leaf traits are significantly associated with either axis (photosynthetic traits are shown as red lines, leaf economics and water-use traits as green lines, and hydraulic traits as blue lines).

The separation of co-related traits into three functional groups was broadly supported by univariate correlations between genotype means (Table 4). Among leaf traits measured under aCO₂, A_net was positively correlated with V_cmax (r = 0·87, P < 0·0001) and J₁₀₀₀ (r = 0·79, P < 0·0001), but was unrelated to g_s and all leaf economics (e.g. LMA and N_area) and hydraulic traits (e.g. VD) respectively (Table 4). Under aCO₂, VD was not significantly correlated with SD or g_wmax, and was unrelated to g_s and all leaf economics traits respectively (Table 4). In contrast, significant relationships were observed among leaf economics and water-use traits; g_s was negatively correlated with LMA and N_area, and positively correlated with Δ¹³C; while Δ¹³C was negatively correlated with LMA and N_area (Table 4).

Table 4. Pearson correlation r values among selected leaf traits measured in 14 genotypes (n = 14) of Eucalyptus camaldulensis under aCO₂ (lower diagonal/shaded) and eCO₂ (upper diagonal/non-shaded) respectively

	A net	V cmax	J 1000	g s	LMA	N area	L starch	L sugars	Δ^13C	Vein density	SD	g wmax
A net		0·85	0·75	0·41	−0·35	−0·26	−0·21	0·62	0·32	0·46	0·23	0·26
V cmax	0·87		0·78	0·15	−0·21	−0·03	−0·21	0·49	0·41	0·12	0·06	0·01
J 1000	0·79	0·74		0·23	−0·11	−0·03	0·05	0·03	−0·03	0·09	−0·28	−0·08
g s	0·24	−0·06	−0·22		−0·16	−0·36	0·37	0·49	0·20	0·67	0·17	0·34
LMA	−0·25	−0·07	−0·02	−0·59		0·91	0·27	0·28	−0·57	−0·18	−0·23	0·33
N area	−0·14	0·07	0·18	−0·73	0·91		−0·20	−0·13	−0·41	−0·13	−0·37	−0·01
L starch	−0·21	−0·19	−0·51	0·29	0·27	0·08		−0·27	−0·34	0·01	−0·68	−0·52
L sugars	−0·05	−0·04	0·03	−0·30	0·28	0·38	−0·46		0·63	0·74	0·72	0·67
Δ^13C	0·16	0·04	−0·12	0·65	−0·57	−0·66	−0·43	0·08		0·15	0·51	0·38
Vein density	−0·33	−0·31	−0·25	0·15	−0·18	−0·21	0·05	−0·06	0·05		0·30	0·64
SD	−0·46	−0·35	−0·37	−0·06	−0·23	−0·12	−0·48	0·34	0·21	0·45		0·57
g wmax	−0·37	−0·13	−0·08	−0·45	0·33	0·36	−0·13	−0·01	−0·41	0·32	0·25

• LMA, leaf mass per area.
• All traits were log-transformed to improve normality. Bolded values indicate significant correlations (P < 0·05) between pairs of independent traits. Trait abbreviations are provided in Table 2.

Trait Coordination in eCO₂

Growth in eCO₂ did not substantially alter (with two exceptions) the patterns of functional coordination among leaf traits observed under aCO₂ (Table 3; Fig. 2). The PCA identified three independent axes of trait variation, with each PC broadly associated with leaf economics and water-use traits (PC1, 38% of total variation), leaf photosynthetic traits (PC2, 22% of total variation) and leaf hydraulics traits (PC3, 17% of total variation) respectively. Two traits were observed to shift their PC grouping under eCO₂: g_s shifted from the leaf economics PC to the leaf hydraulics PC, while SD shifted from the leaf hydraulics PC to the leaf economics PC.

In eCO₂, strong pairwise trait correlations were observed among photosynthetic traits, but fewer correlations were observed among leaf economics and water-use traits compared to aCO₂ (Table 4). In agreement with the PCA, variation in g_s was significantly correlated with VD under eCO₂, which in turn was correlated with g_wmax. In addition, a significant positive correlation was observed between VD and L_sugars (r = 0·74, P = 0·02).

Discussion

Genotypes of E. camaldulensis subsp. camaldulensis showed substantial differentiation across CO₂ treatments in nearly all leaf traits, despite evidence of limited genetic structure among populations of this subspecies (Dillon et al. 2015). This level of genotypic variation allowed us to examine patterns of coordination among leaf functional traits. Across genotypes grown in aCO₂, we found that leaf traits were clustered into three separate syndromes associated with leaf photosynthetic capacity, leaf economics and water-use, and leaf hydraulics respectively. This suggests that key sets of leaf functional traits do not align themselves along a single axis of variation across genotypes of E. camaldulensis, as recently proposed for functional trait variation across species (Reich 2014). Instead, it suggests independent functional trait dimensions may exist within leaves (Li et al. 2015; Marechaux et al. 2015), even within a single species. Furthermore, we found that growth in eCO₂ did not substantially alter these functional groupings, despite differences in the magnitude and direction of eCO₂ effects among the trait variables.

Decoupling of Leaf Functional Traits

Across genotypes, we expected leaf photosynthetic, economics and hydraulic traits to covary along a single axis of trait variation (Reich 2014). This expectation was based on theory that hydraulic traits such as VD underlie key aspects of the leaf economics spectrum (Blonder et al. 2011), and evidence that leaf hydraulics and economics traits both influence rates of photosynthesis (Wright et al. 2004; Brodribb, Feild & Jordan 2007). Our within-species findings add to evidence that sets of leaf functional traits can vary independently from each other, and are consistent with studies that have observed independence between leaf economics and hydraulic traits across diverse sets of species (Sack et al. 2013; Li et al. 2015).

We found that leaf hydraulic traits such as VD were unrelated to net photosynthesis and stomatal conductance among E. camaldulensis genotypes. These results support emerging reports of weak coordination between VD and leaf gas exchange across some sets of angiosperms (Walls 2011; Gleason et al. 2016) and suggest that the evolutionary drivers linking hydraulics and gas exchange are not completely understood. We also found that photosynthetic traits were unrelated to both leaf investment (LMA) and leaf N, indicating that photosynthetic capacity can vary among E. camaldulensis genotypes without covariation in key aspects of leaf structure and nitrogen concentration. These results contrast with previous studies that have observed positive relationships among photosynthesis and leaf economics traits on an area basis (Wright et al. 2004), especially among species with structurally similar leaves (Wyka et al. 2012). Finally, stomatal conductance and ΔC¹³ were most closely related to leaf economics traits, indicating a potentially strong influence of leaf tissue structure on CO₂-water exchange dynamics (Zwieniecki & Boyce 2014).

Our study was conducted under non-limiting nutrient and soil moisture conditions in an environmentally controlled glasshouse. This approach is somewhat similar to a common garden experiment and allowed us to assess the genetic basis of trait variation in E. camaldulensis subsp. camaldulensis. In comparison, most studies of leaf functional traits examine patterns of cross-species variation along gradients of resource availability (light, water, nutrients and temperature) in the field (Westoby et al. 2002; Poorter et al. 2009). In these studies, the range of trait values is often very large, making it easier to detect adaptive patterns of coordination within and across habitats (e.g. Reich et al. 2003). However, it is not always possible to distinguish between genetic and environmental (plastic) sources of trait variation in these cross-species comparisons, the latter of which may alter leaf trait relationships within species (Niinemets 2015). In our study, leaf traits were strongly differentiated among genotypes; however, the range of values across genotypes for each trait was relatively small, thereby making it difficult to detect significant trait–trait correlations.

In cross-species comparisons, trait decoupling may be due to evolutionary divergence among plant lineages in different aspects of leaf structure and function (Li et al. 2015; Marechaux et al. 2015). Such evolutionary divergences are relatively constrained among genotypes of a single species. However, our within-species results provide insight into the eco-evolutionary potential for divergence in plant functional traits that lead to multiple trait dimensionalities and which are important for explaining community assembly theory across biomes (Laughlin 2014). Across species, independent trait dimensions may arise as a result of the physical separation of leaf structures associated with different sets of traits. Li et al. (2015) utilized this argument to explain why leaf economics and hydraulic traits were unrelated across a diverse group of tropical-subtropical angiosperms. Central to their argument was that leaves typically contain two functional sublayers; the upper palisade- (associated with traits such as leaf N that influence photosynthetic capacity; V_cmax) and the lower spongy-mesophyll (associated with traits such vein and stomatal density that influence hydraulic capacity). Processes within each sublayer can vary independently in different growth situations, while traits such as LMA integrate tissue structures of the entire leaf. Although this framework is appealing for understanding how multiple trait combinations could arise, it breaks down to some extent in the current study. Eucalyptus camaldulensis leaves are iso-bilateral (i.e. they do not have a clearly defined ‘upper’ and ‘lower’ surface), suggesting both sides of the leaf are functionally equivalent. Nevertheless, genotypes did vary significantly in the degree of leaf amphistomy they displayed (the ratio of stomata on each side of the leaf), which may have altered the relationships between leaf structural, hydraulic and gas-exchange traits (Brodribb, Jordan & Carpenter 2013). It might also help explain the lack of significant correlation between VD and stomatal traits across genotypes.

Although the magnitude and direction of eCO₂ effects differed among the trait variables, eCO₂ did not substantially alter patterns of leaf functional coordination among genotypes. PCA again revealed three independent axes of trait variation in eCO₂, with individual traits generally adhering to the same ‘functional groupings’ defined in aCO₂. There were, however, two notable exceptions. Among genotypes grown in eCO₂, stomatal density shifted towards the leaf economics syndrome, consistent with previous reports of coordination between stomatal density, LMA and cell size (Loranger & Shipley 2010; Brodribb, Jordan & Carpenter 2013), while stomatal conductance shifted towards the hydraulic traits syndrome, and was in fact strongly correlated with VD across genotypes. This correlation between VD and stomatal conductance suggests some degree of coordination among genotypes between hydraulic supply and leaf water-use in response to growth in eCO₂, even though VD was overall unresponsive to eCO₂ while stomatal conductance declined significantly. VD was also significantly correlated with leaf soluble sugars (P < 0·05) and weakly correlated with net photosynthesis (P < 0·1) across genotypes, but only when grown in eCO₂. Taken together, these results are intriguing and may highlight a possible role of leaf venation in determining the level of the eCO₂ response in leaf water-use and carbon traits, in terms of its influence on water transport capacity via the xylem and sugar export capacity via the phloem. It may also help explain the overall lack of adjustment in VD in response to eCO₂ on the basis of contrasting functional demand between reductions in leaf water-use and increases in photosynthetic activity (Muller et al. 2014).

Trait-Mean Responses to eCO₂

Growth in eCO₂ significantly increased net photosynthesis, presumably as a result of increased [CO₂] at the sites of carbon fixation, which increases rates of Rubisco carboxylation and decreases photorespiration (Long et al. 2004). Importantly, this stimulation in photosynthesis occurred despite an overall reduction in photosynthetic capacity (V_cmax), leaf N content and stomatal conductance, and an increase in LMA. In E. camaldulensis, we found that eCO₂-induced increases in LMA were due to increased leaf thickness and increased foliar concentrations of non-structural carbohydrates, consistent with eCO₂ responses in other tree seedlings (Tjoelker, Oleksyn & Reich 1998; Smith et al. 2012; Duan et al. 2013). However, there was no evidence to suggest these changes in leaf structure and carbohydrate levels inhibited the stimulation of photosynthesis in E. camaldulensis in eCO₂. Similar findings have been reported in other fast-growing species (Davey et al. 2006), including eucalypts (Crous et al. 2013), and may indicate that E. camaldulensis maintains, at least initially, high carbon sink demand in response to growth in eCO₂.

Averaged across genotypes, we found little difference in VD between CO₂ treatments, suggesting leaf hydraulics do not acclimate to growth in eCO₂ in E. camaldulensis. These findings do not support expectations that VD would be reduced in eCO₂ in accordance with reductions in stomatal conductance and hydraulic demand. It is possible that the relatively short duration of the experiment (3 months) did not allow sufficient time for leaf structural traits related to the venation. They are consistent, however, with previous studies reporting a lack of response in VD (Uhl & Mosbrugger 1999) and leaf hydraulic capacity (Locke et al. 2013) in response to eCO₂. Importantly, our results indicate that hydraulic capacity did not inhibit increased photosynthesis in this species in eCO₂. Stomatal size and density were also unaffected by eCO₂. While this lack of response in stomatal traits in E. camaldulensis is in contrast to the response in many woody species in variable CO₂ (Woodward & Kelly 1995), it is consistent with a recent study in a closely related species of Melaleuca that showed a lack of response in stomatal traits to historical changes in atmospheric CO₂ (Hill, Hill & Watling 2015). It could be hypothesized that the overall lack of eCO₂ responsiveness in leaf venation and stomatal traits may have been due to the relatively short duration of the experiment (3 months). However, other studies have clearly shown that hydraulic traits can respond to changes in CO₂ (Phillips et al. 2011) and light conditions (Carins Murphy, Jordan & Brodribb 2012) within a similar time frame. Furthermore, because hydraulic traits were measured on recently developed canopy leaves, there was opportunity for older leaves to detect and transmit potential signals to induce an appropriate developmental response by stomata (Lake et al. 2001) and leaf veins (Sack et al. 2012) of new leaves.

Conclusion

We provide evidence that key leaf functional traits can vary independently among genotypes of a single species grown under common conditions. Such trait decoupling may allow different ecotypes or populations to produce novel combinations of adaptive traits in response to changing environmental conditions, without potential constraints of covariation with other leaf traits. Over evolutionary time, this process may underpin patterns of multiple trait dimensionalities, similar to observations across diverse sets of species (Laughlin 2014; Li et al. 2015). Nevertheless, our results suggest that elevated concentrations of atmospheric [CO₂] may not act strongly in altering patterns of trait coordination already apparent under ambient conditions. We recommend that future work in this area should consider the roles of environmental (plastic) and genetic sources of variation in explaining patterns of leaf trait coordination.

Acknowledgements

We thank Danielle Creek, John Drake, Michael Loik, Sebastian Psfautch, Paul Rymer, Anita Wesolowski and many others from HIE who helped with data collection. We appreciate technical assistance in light microscopy by Liz Kabanov. We also thank Graham Farquhar, Michael Battaglia, Libby Pinkard, Tony O'Grady and Florian Bush for their useful comments on the manuscript. This study was funded by a Science Industry Endowment Fund (SIEF; project code RP04-122), and supported by the Hawkesbury Institute for the Environment (HIE), Western Sydney University. VRD acknowledges a Ramón y Cajal Fellowship (RYC-2012-10970).

Data accessibility

The data are deposited in the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.5tm76 (Blackman et al. 2016) and will be available 3 years from publication.