Bundle-sheath leakiness and intrinsic water use efficiency of a perennial C₄ grass are increased at high vapour pressure deficit during growth

Bundle-sheath leakiness of a perennial C₄ grass responds dynamically to short-term variation of atmospheric CO₂ concentration, and is altered by long-term changes of vapour pressure deficit.

Abstract

Bundle-sheath leakiness (ϕ) is a key parameter of the CO₂-concentrating mechanism of C₄ photosynthesis and is related to leaf-level intrinsic water use efficiency (WUE_i). This work studied short-term dynamic responses of ϕ to alterations of atmospheric CO₂ concentration in Cleistogenes squarrosa, a perennial grass, grown at high (1.6 kPa) or low (0.6 kPa) vapour pressure deficit (VPD) combined with high or low N supply in controlled environment experiments. ϕ was determined by concurrent measurements of photosynthetic gas exchange and on-line carbon isotope discrimination, using a new protocol. Growth at high VPD led to an increase of ϕ by 0.13 and a concurrent increase of WUE_i by 14%, with similar effects at both N levels. ϕ responded dynamically to intercellular CO₂ concentration (C_i), increasing with C_i. Across treatments, ϕ was negatively correlated to the ratio of CO₂ saturated assimilation rate to carboxylation efficiency (a proxy of the relative activities of Rubisco and phosphoenolpyruvate carboxylase) indicating that the long-term environmental effect on ϕ was related to the balance between C₃ and C₄ cycles. Our study revealed considerable dynamic and long-term variation in ϕ of C. squarrosa, suggesting that ϕ should be determined when carbon isotope discrimination is used to assess WUE_i. Also, the data indicate a trade-off between WUE_i and energetic efficiency in C. squarrosa.

Introduction

The CO₂ concentrating mechanism (CCM) is a specialized feature of C₄ photosynthesis and enables the maintenance of a very high CO₂ partial pressure at the site of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), thus effectively minimizing photorespiration (Hatch, 1987). The CCM underlies the higher quantum yield, and higher water- and nitrogen-use efficiency of C₄ plants relative to C₃ plants under high temperature and/or low intercellular CO₂ (Ehleringer et al., 1997; Long, 1999). In C₄ photosynthesis, CO₂ is first fixed by phosphoenolpyruvate carboxylase (PEPc) in mesophyll cells to form C₄ acids (the C₄ cycle), which are transported into bundle-sheath cells, where the acids are decarboxylated and CO₂ is finally fixed by Rubisco (the C₃ cycle). Effective CCM requires a strong ‘CO₂ pump’ (a high rate of the C₄ cycle). However, leakage of CO₂ from bundle-sheath cells back to mesophyll cells represents an energetic inefficiency that is related to the cost of the regeneration of phosphoenolpyruvate (PEP; Hatch, 1987; Furbank et al., 1990). The ratio of this leakage rate to the C₄ cycle rate is termed bundle-sheath leakiness (Farquhar, 1983), or leakiness (ϕ), and is a key parameter of C₄ photosynthesis.

Leakiness is generally quantified using a carbon isotope approach based on the model of carbon isotope discrimination (Δ) of C₄ photosynthesis (Farquhar, 1983). The simplified version of the C₄ discrimination model indicates that Δ is determined by ϕ and C_i/C_a, the ratio of intercellular CO₂ concentration (C_i) to atmospheric CO₂ concentration (C_a):

$$\Delta =a+(b_{\text{4}}+\varphi \left(b_{\text{3}}-s\right)-a)C_{\text{i}}/C_{\text{a}},$$ (1)

with a the fractionation during diffusion of CO₂ in air, b₄ the combined fractionation of PEP carboxylation and the preceding fractionation associated with dissolution of CO₂ and conversion to HCO₃⁻, b₃ the fractionation by Rubisco and s the fractionation during leakage of inorganic C out of the bundle sheath. Since leaf-level intrinsic water use efficiency (WUE_i; net CO₂ assimilation rate/stomatal conductance) is a function of C_i and C_a, combined with Eqn 1 we have:

$${\text{WUE}}_{\text{i}}=\frac{C_{\text{a}}}{1.6}\left(1-\frac{\Delta -a}{b_4+\left(b_3-\text{s}\right)\varphi -a}\right),$$ (2)

showing that WUE_i can be estimated from Δ if ϕ is known. Similar to the application of Δ to quantify WUE_i of C₃ plants (Farquhar and Richards, 1984; Yang et al., 2016b), Δ was suggested as a potentially promising screening tool for breeding and improvement of C₄ crops (Henderson et al., 1998; Gresset et al., 2014; von Caemmerer et al., 2014). However, such endeavours require knowledge of the magnitude and variability of ϕ and relevant environmental drivers. Early studies led to the notion that leakiness is relatively constant within a species with moderate short-term environmental variation, e.g. in response to light, CO₂, and temperature (Henderson et al., 1992). A constant leakiness would largely simplify the relationship between WUE_i and Δ. However, more recent studies suggest dynamic variation of leakiness in response to short-term variation of irradiation (Kromdijk et al., 2010; Pengelly et al., 2010; Ubierna et al., 2013; Bellasio and Griffiths, 2014) or temperature (von Caemmerer et al., 2014).

Analysis of dynamic changes of leakiness in response to short-term variation of atmospheric CO₂ provides an opportunity for studying the functioning of the CCM. However, model analyses suggested an increase of leakiness with increasing C_i (von Caemmerer and Furbank, 1999; Kiirats et al., 2002; Ghannoum, 2009; Yin et al., 2011), while (the few) experimental studies using the carbon isotope approach showed no clear response of leakiness to C_i (Henderson et al., 1992; Pengelly et al., 2012).

Long-term effects of environmental drivers on leakiness have been observed. Thus, drought stress was shown to increase leakiness of several species (Saliendra et al., 1996; Williams et al., 2001; Fravolini et al., 2002), and this effect was attributed to the relative activities of Rubisco/PEPc (Saliendra et al., 1996). N stress was reported to affect leakiness of several sugarcane species (Meinzer and Zhu, 1998), but another study with different species found no effect of N fertilizer supply (Fravolini et al., 2002). Increased CO₂ concentration was shown to increase (Watling et al., 2000; Fravolini et al., 2002) or have no effect (Williams et al., 2001) on leakiness. Currently, it is unknown if vapour pressure deficit (VPD) during plant growth has an effect on leakiness. Studies on long-term effects of N nutrition, water supply, or CO₂ concentration during growth on leakiness using biomass-based Δ have considerable uncertainty related to post-photosynthetic fractionation and the temporal mismatch between biomass formation and gas exchange measurements as discussed by many authors (Henderson et al., 1992; Cousins et al., 2008; von Caemmerer et al., 2014).

Scarcity of experimental evidence is partially due to the fact that quantifying leakiness using combined measurements of gas exchange and ∆ (‘on-line’ ∆) is a laborious task. Measurement artefacts, i.e. CO₂ leak fluxes between the leaf cuvette and the surrounding air and isotopic disequilibria between photosynthetic and respiratory CO₂ fluxes, can lead to errors of measured ∆ of several permil (Gong et al., 2015). This technical issue is particularly relevant for C₄ species, as the ∆ of C₄ plants is generally much lower than that of C₃ plants (Farquhar, 1983).

A previous study in our lab demonstrated that Cleistogenes squarrosa (Trin.) Keng—a perennial C₄ (NAD-ME) grass and co-dominant species in the semiarid steppe of Inner Mongolia—had a large and variable biomass-based Δ of leaves (6–9‰; Yang et al., 2011), potentially indicating a high and variable ϕ. As ϕ is related to WUE_i (Eqn 2) and potentially driven by environmental factors, we were interested to unravel the mechanisms governing Δ in this C₄ species. Furthermore, understanding the physiological response of C. squarrosa to N nutrition and VPD may provide new insight into the adaptation of this species to its habitat given that: (i) water and N availability are the major limiting factors that determine primary production and resource use efficiency of plants in the semiarid steppes of Inner Monglia (Gong et al., 2011) and (ii) the abundance of C₄ plants has increased substantially during the past several decades in the Mongolian plateau (Wittmer et al., 2010). Given the lack of knowledge on long-term effects of VPD or N nutrition and short-term dynamic effects of CO₂ on ϕ, we performed controlled environment experiments with C. squarrosa grown under a low or high level of VPD combined with a low or high level of N fertilizer supply. Leakiness was determined using combined measurements of gas exchange and ∆ (on-line ∆) following the protocols of Gong et al. (2015).

Materials and methods

Plant material and growth conditions

Cleistogenes squarrosa is an NAD-ME type (determined by ¹⁴C labelling) C₄ grass, with a ‘chloridoid’ type of Kranz anatomy (Pyankov et al., 2000). Stands of C. squarrosa were grown from seed in plastic pots (5 cm diameter, 35 cm depth) filled with quartz sand. Pots were placed in growth chambers (PGR15, Conviron, Winnipeg, Canada) at a density of 236 plants m⁻². A photosynthetic photon flux density (PPFD) of 800 µmol m⁻² s⁻¹, provided by cool white fluorescent tubes, was maintained at the top of the canopies during the 16 h-long photoperiods. Air temperature was maintained at 25 ºC during photo- and dark periods. A CO₂ concentration ([CO₂]) of 386±3 µmol mol⁻¹ (mean±SD, n>400) in chamber air was maintained during the photoperiods (cf. Schnyder et al., 2003). A modified Hoagland nutrient solution was supplied three times per day by an automatic irrigation system throughout the entire experiment similar to Lehmeier et al. (2008). The time of first watering is referred to as imbibition of seeds. Before germination, all chambers were run with a VPD of 0.63 kPa (corresponding to a relative humidity (RH) of 80%).

N nutrition and VPD treatments

The study had a 2 × 2 factorial design, with N fertilizer supply and VPD as factors, two levels (low and high, see below) for each factor, and four replicates (four replicate stands) for each combination of N and VPD level. About 1 week after the imbibition of seeds (i.e. first watering), VPD and N treatment were implemented until the end of the experiment. The nutrient solution with 7.5 mM N in the form of equimolar concentrations of calcium nitrate and potassium nitrate was supplied to chambers of the low N treatment (N1), while another nutrient solution with 22.5 mM N was supplied to chambers of the high N treatment (N2). The concentration of other nutrients was the same in both nutrient solutions: 1.0 mM MgSO₄, 0.5 mM KH₂PO₄, 1 mM NaCl, 125 µM Fe-EDTA, 46 µM H₃BO₃, 9 µM MnSO₄, 1 µM ZnSO₄, 0.3 µM CuSO₄, 0.1 µM Na₂MoO₄. VPD (difference between actual vapour pressure and the saturation vapour pressure) was maintained at 0.63 kPa (corresponding to an RH of 80%) in the chambers of the low VPD treatment (V1) or at 1.58 kPa (corresponding to an RH of 50%) in the chambers of the high VPD treatment (V2). Each chamber (or each stand) was assigned to one of the treatments: N1V1, N1V2, N2V1, and N2V2. As we had a total of four chambers, in order to repeat each treatment four times, all replicates were accommodated in four sequential experimental runs (in total, 16 stands were grown) (cf. Table S1 in Liu et al. 2016). In each treatment, two replicate chambers were supplied with CO₂ from a mineral source (δ ¹³C_CO2 of −6‰) and the other two chambers with CO₂ from an organic source (δ ¹³C_CO2 of −33‰). Carbon isotope discrimination by the canopies led to some ¹³C-enrichment of the CO₂ inside the growth chambers. However, that effect was small due to a high rate of air flow through the chambers: quasi-continuous on-line measurements of δ ¹³C_CO2 inside the growth chambers demonstrated a δ ¹³C_CO2 inside the chambers with the ¹³C-enriched CO₂ source of −5.2±0.1‰ (mean±SD, n=23) and −31.7±0.1‰ (n=23) inside the chambers with the ¹³C-depleted CO₂.

Gas exchange measurement system

The method reported by Gong et al. (2015) was used for leaf-level combined measurements of gas exchange and ¹³C discrimination (on-line ∆). Measurements were performed using a leaf-level ¹³CO₂/¹²CO₂ gas exchange and labelling system, which included a portable CO₂ exchange system (LI-6400, LI-COR Inc., Lincoln, NE, USA) housed in a gas exchange mesocosm (Gong et al., 2015). The air supply to the mesocosm and LI-6400 was mixed from CO₂-free, dry air, and CO₂ of know δ¹³C_CO2 (cf. Schnyder et al., 2003). [CO₂] inside the mesocosm (growth chamber) was monitored with an infrared gas analyser (LI-6262, LI-COR Inc.). During the measurement, the plant to be measured and the sensor head of the LI-6400 were placed inside the mesocosm, with concentration and δ¹³C of CO₂ in both gas exchange facilities monitored and controlled. Using this set-up, we separately controlled the concentration and δ¹³C in both facilities (sensor head and growth chamber) for purposes of gas exchange measurement or ¹³C labelling (see below).

The mesocosm and leaf cuvette systems were coupled to a continuous-flow isotope ratio mass spectrometer (IRMS; Delta^plus Advantage equipped with GasBench II, ThermoFinnigan, Bremen, Germany) via a stainless steel capillary. Sample air was drawn through the capillary via a peristaltic pump and passed through a 0.25 ml sample loop attached to the 8-port Valco valve of the GasBench II. Sample air in the loop was introduced into the IRMS via an open split after passage of a dryer (Nafion) and a GC column (25 m×0.32 mm Poraplot Q; Chrompack, Middleburg, The Netherlands). After every second sample a VPDB-gauged CO₂ reference gas was injected into the IRMS via the Open Split. The whole-system precision (SD) of repeated measurements was 0.10‰ (n=50). For further details of the method see Gong et al. (2015).

Gas exchange and concurrent on-line ¹³C discrimination measurements

Gas exchange and on-line ∆ measurements followed the optimized protocols suggested by Gong et al. (2015). In order to minimize the effects of any CO₂ leak artefact across the gaskets of the clamp-on leaf cuvette and the surrounding (mesocosm) air, [CO₂] inside the mesocosm was maintained close to that of the leaf cuvette of the LI-6400 (with the difference <30 μmol mol⁻¹), and the same CO₂ source was supplied to the mesocosm and LI-6400. By these means, the diffusive gradient of ¹³CO₂ or ¹²CO₂ between leaf cuvette and the surrounding air was effectively minimized (Gong et al., 2015). In order to also minimize the isotopic disequilibrium artefact between the photosynthetic and respiratory CO₂ flux, the same CO₂ source was used for growing plants and for on-line leaf-level ∆ measurements (Gong et al., 2015).

Gas exchange and on-line ∆ measurements were performed on the fully expanded young leaves near the tip of major tillers (the second youngest fully expanded leaf), started at about 38 days and ended at about 53 days after the imbibition of seeds. We used the standard 2 × 3 cm cuvette of LI-6400. Two leaf blades from separate major tillers of the same plant were carefully placed in the cuvette. This was done to compensate for the narrow leaf width of C. squarrosa (about 3 mm). The enclosed leaf area was determined after gas exchange measurements by photographing the leaf blades and analysis with ImageJ software (National Institutes of Health, Bethesda, MD, USA). All gas exchange parameters were recomputed using the measured enclosed leaf area.

Photosynthetic CO₂ response curves were taken with the successive CO₂ steps of 390, 800, 390, 260, 180, 120, 90, 60, and 30 μmol mol⁻¹ ([CO₂] in the cuvette) on at least two plants randomly chosen from each growth chamber. In total, 42 plants were measured. The data of five plants were eliminated from further analysis due to extremely low leaf N contents. Finally, each treatment had nine to ten replicates of individual plants. During the gas exchange measurements, the environmental parameters were maintained close to the growth conditions: leaf temperature was maintained at 25 °C, PPFD at 800 µmol m⁻² s⁻¹, leaf-to-air VPD at 0.76–0.96 kPa (V1 treatment) or 1.4–1.6 kPa (V2 treatment; Supplementary Fig. S1 at JXB online) at all CO₂ levels. To generate the appropriate VPD for gas exchange measurements, the inflow air stream (well-mixed dry air) was passed through a gas washing bottle placed in a temperature-controlled water bath. For each step of the CO₂ response, the [CO₂] in the mesocosm was also adjusted to be close to that of the cuvette. Leaves were acclimated for at least 30 min before the start of gas exchange measurements. After CO₂ exchange rates reached a steady state at each step of the CO₂ response, gas exchange parameters (measured by LI-6400), e.g. net CO₂ assimilation rate (A), transpiration rate (E), and stomatal conductance to vapour (g_s) were recorded. Instantaneous water use efficiency was calculated as WUE_ins=A/E; intrinsic water use efficiency was calculated as WUE_i=A/g_s.

Among the 37 CO₂ response curves, 22 plants were concurrently measured for on-line ¹³C discrimination. For this, the [CO₂] and δ¹³C of the incoming (C_in and δ_in) and outgoing cuvette air (C_out and δ_out) were measured on five to six plants per treatment combination. All measurements of δ¹³C were corrected for the non-linearity effect of the IRMS (Elsig and Leuenberger, 2010). ¹³C discrimination during net CO₂ assimilation (∆) was calculated according to Evans et al. (1986):

$$\Delta =\frac{\xi ({\delta }_{out}-{\delta }_{in})}{1+{\delta }_{out}-\xi ({\delta }_{out}-{\delta }_{in})},$$ (3)

where ξ=C_in/(C_in−C_out).

Determination of leaf respiration rate in light

Additional measurements were performed to determine leaf respiration rate in light (R_L). This was done with leaves of the same age class as used for CO₂ response measurements using the ¹³C labelling method reported by Gong et al. (2015). All measurements were performed under growth conditions (i.e. at C_a of 390 μmol mol⁻¹, PPFD of 800 µmol m⁻² s⁻¹). Briefly, ∆ of each leaf was measured first with the same CO₂ source as used for growing plants, then the same measurement was repeated with the other CO₂ source (i.e. the ¹³C-enriched or -depleted CO₂ source). Making use of isotopic disequilibria between photosynthetic and respiratory CO₂ fluxes derived from the two sets of on-line ∆ measurements, the ratio of R_L to net CO₂ assimilation rate (A) was solved as:

$$R_{\text{L}}/A=\left({\delta }_{\text{AE}}-{\delta }_{\text{AD}}\right)/\left({\delta }_{\text{outE}}-{\delta }_{\text{outD}}\right)-\text{1},$$ (4)

$${\delta }_{\text{A}}=\left({\delta }_{\text{in}}{C}_{\text{in}}-{\delta }_{\text{out}}{C}_{\text{out}}\right)/\left(C_{\text{in}}-C_{\text{out}}\right),$$ (5)

where δ_A and δ_out are the δ¹³C of net CO₂ assimilation and that of CO₂ in the cuvette, respectively, and subscripts E and D indicate the relevant parameters measured with the ¹³C-enriched or the ¹³C-depleted sources of CO₂, respectively. R_L was measured on six individuals in each treatment. R_L of other leaves of the same treatment was estimated by multiplying A of that leaf by the treatment-specific mean of R_L/A. For further details of the methodology see Gong et al. (2015).

Bundle-sheath leakiness (ϕ), C₃ cycle rate and C₄ cycle rate

Bundle-sheath leakiness was determined using the complete version of the Farquhar C₄ discrimination model (Farquhar, 1983), modified to include the ternary correction (Farquhar and Cernusak, 2012):

$$\begin{array}{l}\Delta =\frac{1}{1-t}[a_{\text{b}}\frac{C_{\text{a}}-C_{\text{L}}}{C_{\text{a}}}+a\frac{C_{\text{L}}-C_{\text{i}}}{C_{\text{a}}}]+\frac{1+t}{1-t}[\left(e_{\text{s}}+a_{\text{l}}\right)\frac{C_{\text{i}}-C_{\text{m}}}{C_{\text{a}}}\\ +\frac{b_4+\varphi \left(\frac{b_3{C}_{\text{bs}}}{C_{\text{bs}}-C_{\text{m}}}-s\right)}{1+\frac{\varphi C_{\text{m}}}{C_{\text{bs}}-C_{\text{m}}}}\frac{C_{\text{m}}}{C_{\text{a}}}],\end{array}$$ (6)

where C_a, C_L, C_i, C_m, and C_bs are the CO₂ concentrations in air (C_a=C_out during gas exchange measurements), at the leaf surface, in the intercellular airspace, in the mesophyll cell, and in the bundle-sheath cell, respectively. C_L was calculated as C_L=C_a−1.37A/g_BL, with g_BL the boundary layer conductance to vapour calculated by the LI-6400 software. a_b=2.9‰, a=4.4‰, e_s=1.1‰, a_l=0.7‰, and s=1.8‰ denote fractionation constants associated with CO₂ diffusion across boundary layer, across stomata, dissolution in water, diffusion in liquid phase, and leaking out of bundle sheath cells, respectively. t represents the ternary correction factor (Farquhar and Cernusak, 2012):

$$t=\left(\text{1}+\overline{a}\right)E/\left(\text{2}{g}_{\text{sc}}\right),$$ (7)

where E is the transpiration rate and g_sc is the total conductance to CO₂: g_sc=1/(1.6/g_s+1.37/g_BL), where g_s is the stomatal conductance to vapour. ā is the weighted fractionation across boundary layer (a_b=2.9‰) and stomata (a=4.4‰): ā=[a_b(C_a−C_L)+a(C_L−C_i)]/(C_a−C_i). b₃ (¹³C fractionation during carboxylation by Rubisco including respiratory fractionation) and b₄ (¹³C fractionation by CO₂ dissolution, hydration, and PEPc carboxylation including respiratory fractionation) were calculated according to Farquhar (1983):

$$b_{\text{3}}=b{\prime }_{\text{3}}-e{R}_{\text{L}}/V_{\text{c}}-0.\text{5}f{V}_{\text{o}}/V_{\text{c}},\text{ and}$$ (8)

$$b_{\text{4}}=b{\prime }_{\text{4}}-0.\text{5}e{R}_{\text{L}}\left(\text{1}-\varphi \right)/\left(A+0.\text{5}{R}_{\text{L}}\right),$$ (9)

where b′₃=30‰, b′₄=–5.7‰, e=–6‰ (Ghashghaie et al., 2003; Kromdijk et al., 2010), f=11.6‰ (Ghashghaie et al., 2003; Lanigan et al., 2008), and V_c is the rate of Rubisco carboxylation and V_o is the rate of oxygenation. In our experiments, the same source CO₂ was used for growing plants and for online ∆ measurements; therefore an additional term (cf. e′ in Ubierna et al. 2013) accounting for the discrepancy of δ¹³C_CO2 between growth environment and gas exchange measurements was not included. Here, it is assumed that respiration in mesophyll cells (R_m) and bundle-sheath cells (R_bs) both equal 0.5R_L (von Caemmerer and Furbank, 1999; Pengelly et al., 2010).

To solve ϕ using Eqns 7–9, C_m, C_bs, V_o, and V_c must be estimated. Those parameters were determined using the equations of the C₄ photosynthesis model of von Caemmerer and Furbank (1999). C_m can be calculated from:

$$g_{\text{m}}=A/\left(C_{\text{i}}-C_{\text{m}}\right),$$ (10)

if mesophyll conductance to CO₂ is known. It has been generally assumed that g_m is high and non-limiting for CO₂ assimilation in C₄ photosynthesis. Thus, in many published works on ϕ, g_m was assumed to be infinite, i.e. C_i=C_m. However, a recent study using a new oxygen isotope approach found that g_m of C₄ species ranged between 0.66 and 1.8 mol m⁻² s⁻¹ measured at a leaf temperature of about 32 °C (Barbour et al., 2016). We calculated the minimum g_m as: g_mmin=A/C_i (von Caemmerer et al., 2014). These results showed that g_mmin decreased with increasing C_i (see Supplementary Fig. S2). Further, we assumed that the young leaves of C. squarrosa had a g_m of 0.66 mol m⁻² s⁻¹ at the growth C_a of 390 μmol mol⁻¹. Appling a constant ratio of g_m/g_mmin across the CO₂ response curve, we estimated the response of g_m to C_i. The estimates of g_m ranged between 0.2 and 2.4 mol m⁻² s⁻¹ and decreased with increasing C_i (Supplementary Fig. S2); similarly, decrease of g_m with increasing C_i was found in other C₄ species using the oxygen isotope approach (Osborn et al., 2016) or an ‘in vitro V_pmax’ method (N. Ubierna, personal communication).

A simplified version of the C₄ photosynthesis model for enzyme-limited CO₂ assimilation rate (von Caemmerer and Furbank, 1999) can be used to model the photosynthetic CO₂ response curve: A=min[(V_p−R_m+g_bsC_m), (V_cmax−R_L)], where V_P is the rate of PEP carboxylation, V_cmax is the maximum rate of Rubisco carboxylation, and g_bs is the bundle-sheath conductance to CO₂. A is determined by the term V_cmax−R_L at high CO₂ concentrations. According to this model, we assumed that V_cmax=A_max+R_L, where A_max is the maximum net assimilation rate observed in the photosynthetic light response curves measured on leaves of the same age at ambient CO₂ (data shown in Supplementary Fig. S3). Treatment-specific V_cmax of C. squarrosa ranged between 16 and 19 μmol m⁻² s⁻¹, which is in the range of V_cmax of C₄ grass used in Earth System Models (13–33 μmol m⁻² s⁻¹; Rogers, 2014). C_bs was estimated using equations of von Caemmerer and Furbank (1999):

$$C_{\text{bs}}=\frac{{\gamma }^{*}{O}_{\text{bs}}+K_{\text{c}}(1+\frac{O_{\text{bs}}}{K_{\text{o}}})(A+R_{\text{L}})/V_{\text{cmax}}}{1-(A+R_{\text{L}})/V_{\text{cmax}}},$$ (11)

$$O_{\text{bs}}=\alpha A/\left(0.0\text{47}{g}_{\text{bs}}\right)+O_{\text{m}},$$ (12)

where O_bs is the bundle-sheath oxygen concentration, and O_m (210 mmol mol⁻¹) is the oxygen concentration in mesophyll cells; γ* (0.000193) is half of the reciprocal of Rubisco specificity; K_c and K_o are the Michaelis–Menten constants of Rubisco for CO₂ (650 μmol mol⁻¹) and O₂ (450 μmol mol⁻¹) at 25 °C, respectively; α (0.1) is the fraction of PSII activity in the bundle-sheath; g_bs=0.003 mol m⁻² s⁻¹. Furthermore, V_o/V_c was calculated as V_o/V_c=2γ*O_bs/C_bs. Knowing V_o/V_c, V_o and V_c were calculated using the equation: V_c=A+R_L+0.5V_o. The values of model parameters and equations were derived from von Caemmerer and Furbank (1999). Knowing ϕ and applying the assumption that R_m=R_bs=0.5R_L (von Caemmerer and Furbank, 1999; Pengelly et al., 2010), the rate of PEP carboxylation (V_p) or Rubisco carboxylation (V_c) was estimated (see Supplementary Fig. S4). As C_m and C_bs cannot be directly measured, the determination of both parameters in this study was associated with some uncertainty. Another commonly used assumption in calculating leakiness using the complete or simplified version of Eqn 6 is that CO₂ is in equilibrium with HCO₃⁻ in the mesophyll cytoplasm. The extent of this equilibrium is determined by the relative rate of PEP carboxylation and hydration of CO₂ (V_h). If CO₂ is not in equilibrium with HCO₃⁻, b′₄ of –5.7‰ in Eqn 9 should be replaced by b′₄=–5.7 + 7.9V_p/V_h (Farquhar 1983). Thus, we calculated ϕ using C₄ discrimination models with different assumptions on g_m, C_bs, and V_p/V_h (Supplementary Fig. S5). These results are compared and discussed below.

Fitting of CO₂ response curves

For each plant, measured A was plotted against the respective intercellular CO₂ concentration. A model of C₄ photosynthesis was used to fit the CO₂ response curves (Wang et al., 2012):

$$A=a\prime (\text{1}-{\text{e}}^{-b}{}^{\prime x})+c\prime$$ (13)

where x is C_i and c′ is R_L. Using the fitted parameters (a′ and b′) of each CO₂ response curve, CO₂-saturated net assimilation rate (A_sat) was obtained as (a′+c′), and carboxylation efficiency (CE) was calculated as b′(a′+c′) (the initial slope of the A–C_i curve). For CO₂ response curves of C₄ leaves, A_sat is proportional to Rubisco activity, while CE is proportional to the PEP carboxylase activity (von Caemmerer and Furbank, 1999; Yin et al., 2011). The ratio of A_sat to CE was calculated to indicate the ratio of Rubisco to PEPc activities.

Leaf trait parameters and nitrogen nutrition index

N content in dry mass of leaves previously used for gas exchange measurements was measured using an elemental analyser (NA 1110, Carlo Erba, Milan, Italy). For sample preparation, leaves were dried at 60 °C for 48 h, weighed and then milled. N content on a mass (N_mass, %DM) and leaf area basis (N_area, g m⁻²) were calculated. Specific leaf area (SLA, cm² mg⁻¹) was obtained as SLA=leaf area/leaf dry mass. In each experiment, a set of individual plants (four plants per chamber) were sampled four or five times (twice per week) after canopy closure (when leaf area index was >2) for the determination of standing aboveground biomass, and N content in total aboveground dry mass. Those data were used to calculate nitrogen nutrition index (NNI) according to Lemaire et al. (2008).

Statistical analysis

Two-way ANOVA was used to test the effects of N supply, VPD, and their interactions on leaf traits and gas exchange parameters, using general linear models of SAS (SAS 9.1, SAS Institute, USA). The effects of N nutrition and VPD on the relationship between lnC_i and leakiness were analysed using a dummy regression approach; the parallelism and coincidence of linear regressions of different treatments were tested according to Berenson et al. (1983). Non-linear regression analysis on photosynthetic CO₂ response curve of individual plants was performed using the Nlin procedure of SAS.

Results

Leaf traits and photosynthetic CO₂ assimilation

N supply increased N_mass of fully expanded young leaves by 25% (P<0.05) and also tended to increase N_area (P=0.06), but had no effect on SLA (Table 1). N supply also significantly increased nitrogen nutrition index (NNI; determined from N content and aboveground standing biomass of each stand): NNI was 0.88 ± 0.05 (mean±SD, n=4 replicate stands) for the N1V1 treatment, 0.71±0.06 for the N1V2 treatment, 1.25±0.06 for the N2V1 treatment, and 1.32±0.05 for the N2V2 treatment. The effects of VPD and the interaction of VPD and N supply on N_mass, N_area and SLA were all non-significant (P>0.05) (Table 1).

Table 1.

Leaf trait and gas exchange parameters of C. squarrosa measured under the same environmental conditions as during growth (leaf temperature 25 °C, PPFD 800 μmol m⁻²s⁻¹, [CO₂] 390 μmol mol⁻¹, VPD 0.8 kPa for V1 and 1.6 kPa for V2)

Plants were grown at low or high N fertilizer supply (N1 or N2) combined with low or high VPD (V1 or V2). Leaf trait parameters include N content per unit dry mass (N_mass, % DM) and per leaf area (N_area, g m⁻²), and specific leaf area (SLA, cm² mg⁻¹). Gas exchange parameters: net CO₂ assimilation rate (A, μmol m⁻² s⁻¹), transpiration rate (E, mmol m⁻² s⁻¹), stomatal conductance to water vapor (g_s, mol m⁻² s⁻¹), respiration rate in light (R_L, μmol m⁻² s⁻¹), the ratio of internal to atmospheric CO₂ concentration (C_i/C_a), instantaneous water use efficiency (WUE_ins=A/E, mmol mol⁻¹), and intrinsic water use efficiency (WUE_i=A/g_s, μmol mol⁻¹). Data are shown as mean±SE (n=6 for R_L, n=9–10 for the other parameters). Significant treatment effects (P<0.05) are shown in bold.

	N1		N2		P-value
	V1	V2	V1	V2	N	VPD	N×VPD
Nmass	2.4 ± 0.1	2.4 ± 0.1	3.0 ± 0.1	3.0 ± 0.2	0.01	0.81	0.79
Narea	1.2 ± 0.2	1.3 ± 0.2	1.6 ± 0.2	1.6 ± 0.2	0.06	0.82	0.92
SLA	0.21 ± 0.02	0.20 ± 0.02	0.21 ± 0.02	0.21 ± 0.02	0.91	0.82	0.85
A	14.8 ± 0.9	13.1 ± 0.9	14.9 ± 0.9	12.3 ± 1.0	0.81	0.03	0.58
E	0.93 ± 0.08	1.14 ± 0.09	0.93 ± 0.08	1.23 ± 0.09	0.69	0.01	0.60
g s	0.10 ± 0.01	0.08 ± 0.01	0.11 ± 0.01	0.07 ± 0.01	0.69	0.01	0.45
R L	0.81 ± 0.26	0.62 ± 0.30	0.53 ± 0.10	0.51 ± 0.10	0.36	0.62	0.71
C i/Ca	0.35 ± 0.03	0.26 ± 0.04	0.35 ± 0.03	0.25 ± 0.04	0.93	0.01	0.91
WUEins	15.9 ± 0.6	11.7 ± 0.6	16.8 ± 0.6	10.0 ± 0.6	0.82	0.01	0.04
WUEi	153 ± 9	173 ± 9	154 ± 9	178 ± 10	0.79	0.02	0.82

The effects of N supply on N_mass and N_area were not associated with parallel effects on gas exchange parameters measured at growth CO₂ level (390 μmol mol⁻¹). Thus, N supply had no effect on A, E, g_s, C_i/C_a, and WUE_ins, and there was only a small interactive effect of N and VPD on WUE_i (Table 1). Conversely, VPD during growth had significant effects on all these parameters, but none of these effects demonstrated an interaction with N supply except for WUE_ins. High VPD decreased A by 14%, g_s by 29%, C_i/C_a by 27% at both low and high N, and WUE_ins by 26% at low N and 40% at high N. Conversely, high VPD increased E by 27% and WUE_i by 14% independently of N levels (Table 1). The rate of leaf respiration in light (R_L) did not differ between treatments.

The response of photosynthetic gas exchange parameters to dynamic changes of CO₂ concentration (Fig. 1) showed typical CO₂ responses of net CO₂ assimilation rate. VPD had clear effects on gas exchange parameters: compared with low VPD, A was 20% lower (Fig. 1A, B), E was 21% higher (Fig. 1C, D), and g_s was 32% lower (Fig. 1E, F) at high VPD averaged over N levels and CO₂ concentrations.

Fig. 1.

Net CO₂ assimilation rate (A; A, B), transpiration rate (E; C, D) and stomatal conductance to water vapour (g_s; E, F) in response to short-term variation of intercellular CO₂ (C_i) under low (N1, circles; A, C, E) or high N supply (N2, triangles; B, D, F) combined with low (V1, open symbols) or high VPD (V2, filled symbols). Data are shown as the mean±SE (n=9–10). Operating conditions of gas exchange measurements were the same as conditions in growth chambers (leaf temperature 25 °C, PPFD 800 μmol m⁻² s⁻¹, VPD 0.8 kPa for V1 and 1.6 kPa for V2). The corresponding C_i values at growth CO₂ concentration (390 μmol mol⁻¹) are indicated by grey (low VPD) and black arrows (high VPD) in (E) and (F).

Carbon isotope discrimination and bundle-sheath leakiness

Carbon isotope discrimination (∆) during net CO₂ exchange revealed strong dynamic responses to short-term variation of CO₂ concentration; ∆ increased with increasing C_i in all treatments, and ranged from 4 to 8‰ (Fig. 2A, B). Compared at the same C_i, ∆ was higher at high VPD than at low VPD. Estimates of ϕ also increased with increasing C_i in all treatments, and ranged from 0.23 to 0.81 (Fig. 2C, D). The relationship between C_i and ϕ followed a logarithmic function. Regression analyses between lnC_i and ϕ including VPD as a dummy variable indicated a significant VPD effect on ϕ: ϕ was higher by 0.13 at high VPD than at low VPD at both N levels (Fig. 3, Table 2). Plants grown with high N supply had a higher ϕ at moderate to low C_i (C_a≤390) than plants with low N supply (Fig. 3). Accordingly, the slope of the linear regression between lnC_i and ϕ was significantly lower for the high N treatment than for the low N treatment (Table 2). The increase of ϕ along the gradient of C_i was apparently related to the discrepancy between rates of C₃ and C₄ cycles: V_c already reached the maximum at a C_i<100 μmol mol⁻¹, while V_p increased throughout the range of measured C_i (Supplementary Fig. S4). We also calculated ϕ using C₄ discrimination models with different assumptions on g_m, C_bs, and V_p/V_h, and our conclusions on short-term response to CO₂ and long-term response to VPD and N were not altered (see Supplementary Fig. S5).

Fig. 2.

Carbon isotope discrimination (∆) during net CO₂ exchange (A, B) and bundle sheath leakiness (C, D) of Cleistogenes squarrosa leaves in response to short-term variation of intercellular CO₂ (C_i). Plants were grown under low (N1, circles; A, C) or high N supply (N2, triangles; B, D) combined with low (V1, open symbols, dashed lines) or high VPD (V2, filled symbols, solid lines). Operating conditions of gas exchange measurements were the same as conditions in growth chambers (leaf temperature 25 °C, PPFD 800 μmol m⁻² s⁻¹, VPD 0.8 kPa for V1 and 1.6 kPa for V2). Data are shown as the mean±SE (n=5–6). The regressions were fitted using a function of y=y₀+aln(x); all regressions have r²>0.8.

Fig. 3.

Relationships between bundle sheath leakiness and ln(C_i) under low (N1, circles; A) or high N supply (N2, triangles; B) combined with low (V1, open symbols, dashed lines) or high VPD (V2, filled symbols, solid lines). Operating conditions of gas exchange measurements were the same as conditions in growth chambers (leaf temperature 25 °C, PPFD 800 μmol m⁻² s⁻¹, VPD 0.8 kPa for V1 and 1.6 kPa for V2). Data are shown as the mean±SE (n=5–6). The data of each N level were fitted by linear regression including a dummy variable indicating VPD treatments (V) after the confirmation of parallelism of the two regressions of VPD levels: with V=0 for low VPD and V=1 for high VPD (for equations see Table 2).

Table 2.

Results of regression analysis on the relationship between C_iand leakiness (ϕ)

The data for each N level were fitted by linear regression including a dummy variable indicating VPD treatments (V) after the confirmation of parallelism of the two regressions of VPD levels, with V=0 for low VPD and V=1 for high VPD; 95% confidence intervals of slope and coefficient of dummy variable (V) are shown. The coefficient of V quantifies the mean difference in ϕ between high and low VPD treatments across C_i levels. Both regressions are significant (P<0.001).

N level	equation	r 2	95% CI slope	95% CI coeff. V
N1	ϕ =−0.30+0.16lnCi+0.15V	0.94	0.13~0.20	0.09~0.21
N2	ϕ =0.001+0.12lnCi+0.12V	0.91	0.09~0.15	0.05~0.18

Relationships between leaf traits and gas exchange parameters

VPD had a clear effect on A_sat (Fig. 4A): averaged over N treatments, high VPD caused a 15% reduction of A_sat. N supply and N×VPD interactions had no effect on A_sat. CE of leaves seemed to be slightly higher at high (0.71 mol m⁻² s⁻¹) than low N (0.53 mol m⁻² s⁻¹), although this effect was only significant at P=0.10 (Fig. 4B). A_sat/CE was calculated to indicate the ratio of Rubisco to PEPc activities. Treatment-specific mean A_sat/CE was negatively correlated with ϕ (P<0.05, Fig. 4C). Across all treatments, we found no correlation between A_sat and N_mass (P=0.41, Fig. 5A), but a positive correlation between CE and N_mass (r²=0.43, P<0.001, Fig. 5B) and a negative correlation between A_sat/CE and N_mass (r²=0.29, P<0.001, Fig. 5C). WUE_i was negatively correlated with A_sat (r²=0.34, P<0.001, Fig. 6A) and A_sat/CE (r²=0.49, P<0.001, Fig. 6C), but positively with CE (r²=0.19, P<0.001, Fig. 6B) and leakiness (r²=0.19, P<0.05, Fig. 6D). Furthermore, a positive correlation between WUE_i and leakiness was also found during short-term response to CO₂ (r²>0.9, P<0.01, Supplementary Fig. S6).

Fig. 4.

A _sat (A), carboxylation efficiency (CE; B) and the relationship between leakiness measured under growth conditions (390 μmol mol⁻¹) and A_sat/CE (μmol mol⁻¹; C) under low or high N fertilizer supply (N1, circles; or N2, triangles) combined with low or high VPD (V1, open symbols; or V2, filled symbols). Data are shown as the mean±SE (n=9–10 for A_sat and CE, n=5–6 for leakiness). * indicates the treatment effect was significant at the P-level of 0.05, while ns indicates no significant effect.

Fig. 5.

Correlations between N_mass and A_sat (A), CE (B) and A_sat/CE (C) of leaves grown at low or high N fertilizer supply (N1, circles; or N2, triangles) combined with low or high VPD (V1, open symbols; or V2, filled symbols). Each symbol represents a data point of an individual plant.

Fig. 6.

Correlations between WUE_i and A_sat (A), CE (B), A_sat/CE (C), and leakiness (D) of leaves grown at low or high N fertilizer supply (N1, circles; or N2, triangles) combined with low or high VPD (V1, open symbols; or V2, filled symbols). Each symbol represents a data point of an individual plant.

Discussion

This study indicates an effect of VPD of the growth environment on bundle-sheath leakiness: ϕ was higher at high VPD than at low VPD with an absolute difference of 0.13 at both N supply levels. This effect appeared to be constitutive, as it was manifest throughout a wide range of short-term variations of CO₂ concentration. Remarkably, the higher leakiness at high VPD was associated with a 14% improved WUE_i. Also, we observed an N effect on ϕ that responded to short-term variation of CO₂ levels: high N supply increased ϕ of leaves at ambient to low CO₂ levels (C_a≤390 μmol mol⁻¹). Further, ϕ responded dynamically to short-term changes of [CO₂], in support of theoretical predictions of C₄ photosynthesis models (von Caemmerer and Furbank, 1999; Yin et al., 2011). Lastly, across treatments, a positive correlation between WUE_i and leakiness was evident, pointing to a trade-off between WUE_i and energy use efficiency of this species.

Unfortunately, leakiness cannot be measured directly, and the online ∆¹³C approach for estimations of leakiness in conjunction with the C₄ photosynthetic discrimination model of Farquhar (1983) can lead to inaccurate estimates of leakiness if improper simplifications are made (Ubierna et al., 2011, 2013; Gong et al., 2015). One commonly used assumption is that g_m of C₄ plants is infinite, and thus C_i=C_m. However, a recent study indicated that g_m of C₄ plants is finite, although close to the high-end values of g_m of C₃ plants (Barbour et al., 2016). In our study, assuming a lower range of g_m across C_i levels (see Supplementary Fig. S5), or a constant g_m of 1.8 mol m⁻² s⁻¹ (data not shown) or an infinite g_m (Supplementary Fig. S5) did not modify observed treatment effects. Only if g_m was much lower at low compared with high VPD (e.g. g_m at N1V1 was 60% lower than at N1V2 treatment), would VPD effects on leakiness disappear. However, it is unlikely that C. squarrosa had lower g_m at low than at high VPD, as studies with C₃ species generally showed a decrease of g_m by high VPD or water stress (Flexas et al., 2008). Thus, our estimates of treatment effects on leakiness appeared to be largely insensitive to assumptions on g_m.

Many studies using Eqn 6 have assumed that CO₂ is in equilibrium with HCO₃⁻ in the mesophyll cytoplasm. However, according to Cousins et al. (2008) that assumption might not be proper, especially for C₄ grasses, which have lower carbonic anhydrase (CA) activity than C₄ dicots. Our observations with the same species in the same experiment suggested that, under growth conditions, assimilated CO₂ was in equilibrium with HCO₃⁻, as indicated by the fact that δ¹⁸O of leaf cellulose did not differ between plants growing in the presence of CO₂ with contrasting δ¹⁸O (Liu et al., 2016). Thus the observed treatment effects on leakiness of C. squarrosa seemed not to be related to CA activity. Further, we calculated leakiness using different assumptions on V_p/V_h: V_p/V_h=0 or V_p/V_h=0.23 (mean of NAD-ME grasses in Cousins et al., 2008) and observed a minor effect on leakiness (less than 0.06). Thus, the short-term response of leakiness to CO₂ was not explained by changes in V_p/V_h. Applying another commonly used simplification that C_bs is sufficiently higher than C_m, so that C_bs/(C_bs−C_m)=1 and C_m/(C_bs−C_m)=0, also had no influence on our conclusions on effects of environmental factors on leakiness.

Effects of VPD and N nutrition on bundle-sheath leakiness

The study demonstrated a higher leakiness of plants grown at high VPD than at low VPD, and this effect was not modified by N nutrition. The present effect was not related to drought stress, as the elongation rate of phytomers ( Yang et al., 2016a) and the growth rate of individuals (Liu et al., 2016) were not influenced by VPD. The long-term response of leakiness to VPD was related to the decrease of C_i/C_a and the increase of ∆ (0.55‰ at ambient CO₂ level) at high VPD, although the direct and indirect influences cannot be clearly distinguished due to the intrinsic correlation between C_i/C_a and ∆. The latter observation was also supported by the ¹³C discrimination of leaf dry mass, which was significantly higher at high VPD (5.0±0.2‰, averaged over N levels) than at low VPD treatment (4.4±0.2‰). Likely, the higher leakiness at high VPD resulted from an enhanced capacity of the C₄ cycle relative to the C₃ cycle at low g_s, as was supported by the lower A_sat/CE ratio and the lower C_i/C_a at high VPD relative to low VPD. This interpretation is in line with the findings that the relative activities of Rubisco/PEPc decreased at low g_s under water stress, with the adjustment achieved by increasing activity of PEPc (Saliendra et al., 1996; Foyer et al., 1998) or decreasing sensitivity of PEPc to the inhibitor malate (Foyer et al., 1998). Also, the data may shed light on the mechanism that underlies the increased leakiness under limited water supply (Saliendra et al., 1996; Williams et al., 2001; Fravolini et al., 2002): that response may have been triggered by decreased stomatal conductance at high VPD. Although a VPD effect on this relationship has not yet been reported, the adjustment of Rubisco/PEPc activities seems to be an important mechanism for plants to counteract the reduced g_s under long-term low water supply (Saliendra et al., 1996) or high VPD. Maintaining a powerful C₄ pump (at the expense of increased leakiness) might reduce the quantum yield due to the additional ATP requirement for PEP regeneration (Hatch, 1987; Furbank et al., 1990). However, this mechanism may be beneficial for plants to maintain a high WUE_i under high VPD or drought stress.

In this work, high N nutrition increased leakiness in short-term exposures to subambient to ambient [CO₂], but not to elevated [CO₂] of 800 μmol mol⁻¹. This result is at variance with previous reports of a higher leakiness of plants growing with a low N supply (Meinzer and Zhu, 1998; Fravolini et al., 2002). Other studies also reported non-significant effects of N supply on leakiness (Fravolini et al., 2002; Wang et al., 2012). In the present study, the plants of the low N treatment had a nitrogen nutrition index of ~0.8, which indicated that N was just growth limiting. High N supply led to a NNI of 1.3, and increased plant growth rate via stimulated tiller production (data not shown), but leaf-level net assimilation rates of young leaves were not affected. These results indicate that our low N and high N treatment represented rather ‘nearly-adequate’ and ‘copious supply’ of N for leaf-level photosynthesis. High N supply increased leakiness, in correspondence with a lower A_sat/CE ratio, again indicating the involvement of the relative activities of Rubisco/PEPc. This interpretation was further supported by the positive correlation between N_mass and CE, and simultaneous absence of such a relationship between N_mass and A_sat. These results suggest that under copious N supply to C. squarrosa, the activity of PEPc was more strongly enhanced relative to that of Rubisco, in line with experimental results on Amaranthus retroflexus (Sage et al., 1987). Thus, our results are also in agreement with the hypothesis that C₄ plants have less plasticity in increasing the amount of Rubisco compared with C₃ plants, due to spatial restrictions in bundle-sheath cells (Sage and McKown, 2006).

Another potential origin of the variance in leakiness is related to the bundle-sheath conductance (g_bs=ϕV_P/(C_bs−C_m)). However, current understanding on g_bs is limited due to the lack of a methodology for quantification of g_bs that is independent of gas exchange measurements. Furthermore, it is unknown how VPD may affect the properties of bundle-sheath cell walls, limiting to some extent discussions of potential mechanisms. A recent study demonstrated that Miscanthus×giganteus plants grown at high N supply had higher bundle-sheath area per unit leaf area, which might increase g_bs; however, estimates of leakiness did not differ between N levels (Ma et al., 2016). Such results highlight the potential complexity of influences on leakiness by biochemical/physiological and anatomical features. In our study, treatments had little effect on individual leaf area and no effect on thickness (Liu et al., 2016), indicating that effects of leaf anatomy may have been minor. Across all treatments, leakiness measured under growth conditions was negatively related to A_sat/CE, again suggesting that variations of leakiness across treatments was related to the balance between C₃ and C₄ cycles.

Response of bundle-sheath leakiness to short-term variation of CO₂

The response of ϕ to short-term variation of [CO₂] has been predicted by model analyses (von Caemmerer and Furbank, 1999; Kiirats et al., 2002; Ghannoum, 2009; Yin et al., 2011). Although these models were designed for estimating gas exchange rate rather than leakiness, the present results supported the theoretical predictions. So far, only a small number of studies investigated the response of ϕ to dynamic changes of [CO₂]. A study using the on-line ∆ method showed no response of ϕ of F. bidentis to dynamic changes of CO₂ (Pengelly et al., 2012), in agreement with the finding of an earlier study on several species (Henderson et al., 1992). However, in the latter study, each leaf was measured at only two or three levels of CO₂. Those results suggested a rapid regulation of C₃ and C₄ cycle rates in response to changing CO₂ concentrations. The present results indicate that this kind of regulation did not occur to the same extent during short-term CO₂ responses in C. squarrosa. Clearly, the response of ϕ to short-term variation of CO₂ should be studied in detail with a greater number of species.

Implications on the ecophysiology of C. squarrosa

This work questions the interpretation of leakiness as a simple indicator for efficiency of C₄ photosynthesis or physiological fitness of C₄ plants, as increased leakiness was related to the enhancement of WUE_i in C. squarrosa. The latter observation was consistent across N and VPD treatments (long-term response) and short-term variation of CO₂ levels. This result was in line with the theoretical prediction of the simplified model of C₄ discrimination: WUE_i is positively correlated to leakiness when ∆>4.4‰, i.e. when ϕ>0.37. This relationship indicates a potential trade-off between quantum yield and water use efficiency: high leakiness under VPD stress might decrease the efficiency of energy use, but may be a condition for maintaining the effectiveness of CCM and minimizing photorespiration at a very low g_s in this species. This mechanism is meaningful for C₄ species in arid or semi-arid grasslands, where shading is uncommon but high VPD or drought stress is a common attribute of the habitat. Considering the reduction of Rubisco specificity for CO₂ relative to O₂ by the increase of leaf temperature, this trade-off may be important for C₄ plants in the context of global warming. Observational evidence has indeed shown that this C₄ grass has increased its relative abundance in vast areas of Inner Mongolian grasslands during regional warming in the last decades (Wittmer et al., 2010).

Supplementary data

Supplementary data are available at JXB online.

Fig. S1. Relative humidity and leaf-to-air vapour pressure deficit in the leaf cuvette during the measurement of CO₂ response curves.

Fig. S2. Minimum mesophyll conductance to CO₂ (g_mmin=A/C_i) and estimated g_m in response to C_i.

Fig. S3. Photosynthetic light response curves of leaves of C. squarrosa.

Fig. S4. V_p and V_c in response to C_i.

Fig. S5. Bundle-sheath leakiness in response to C_i, calculated using different models of ¹³C discrimination.

Fig. S6. Correlations between WUE_i and bundle-sheath leakiness of C. squarrosa.