Elevated CO₂-induced changes in mesophyll conductance and anatomical traits in wild type and carbohydrate-metabolism mutants of Arabidopsis

Abstract

Decreases in photosynthetic rate, stomatal conductance (g_s), and mesophyll conductance (g_m) are often observed under elevated CO₂ conditions. However, which anatomical and/or physiological factors contribute to the decrease in g_m is not fully understood. Arabidopsis thaliana wild-type and carbon-metabolism mutants (gwd1, pgm1, and cfbp1) with different accumulation patterns of non-structural carbohydrates were grown at ambient (400 ppm) and elevated (800 ppm) CO₂. Anatomical and physiological traits of leaves were measured to investigate factors causing the changes in g_m and in the mesophyll resistance (expressed as the reciprocal of mesophyll conductance per unit chloroplast surface area facing to intercellular space, S_c/g_m). When grown at elevated CO₂, all the lines showed increases in cell wall mass, cell wall thickness, and starch content, but not in leaf thickness. g_m measured at 800 ppm CO₂ was significantly lower than at 400 ppm CO₂ in all the lines. Changes in S_c/g_m were associated with thicker cell walls rather than with excess starch content. The results indicate that the changes in g_m and S_c/g_m that occur in response to elevated CO₂ are independent of non-structural carbohydrates, and the cell wall represents a greater limitation factor for g_m than starch.

Changes in mesophyll conductance of Arabidopsis wild-type and carbohydrate-metabolism mutants grown at elevated CO₂ are explained by changes in cell wall mass and thickness rather than by non-structural carbohydrates.

Introduction

Mesophyll conductance (g_m) describes the diffusivity of CO₂ from the intercellular space to the site of CO₂ fixation in the chloroplast stroma as it passes through the cell wall, cell membrane, cytosol, and the chloroplast envelope. Three major methods have been used to estimate g_m; A–C_i curve-fitting (Ethier and Livingston, 2004), chlorophyll fluorescence (Harley et al., 1992), and carbon isotope discrimination (Evans et al., 1986). The magnitude of the resulting value of g_m expressed on a leaf-area basis is comparable to that of the stomatal conductance (g_s), the measure of CO₂ diffusion from the ambient air to the intercellular space (e.g. Tazoe et al., 2011; von Caemmerer and Evans 2015). Thus, g_m represents a significant limitation to photosynthesis. Both g_s and g_m are known to change considerably depending on environmental conditions and plant growth status (Farquhar and Sharkey, 1982; von Caemmerer and Evans, 2015). Determining how the factors underlying g_m respond to environmental conditions and their differences between species will increase our understanding of plant biomass production.

Among the components that influence g_m, the cumulative surface area of the chloroplasts positioned along the cell membrane facing the intercellular space (S_c) and the cell wall thickness are the most crucial anatomical factors. Tholen et al. (2008) demonstrated that g_m increases with an increase in S_c in photoreceptor mutants of Arabidopsis that showed marked differences in chloroplast positioning. Adachi et al. (2013) also found close relationships between S_c, g_m, and maximum photosynthetic rate in inbred lines of Oryza sativa. In C₃ plants, it is estimated that cell wall resistance accounts for 50% of the mesophyll resistance, the inverse of g_m (Evans et al., 1994, 2009; Terashima et al., 2011). Scafaro et al. (2011) found that two wild Oryza relatives that had thicker cell walls than O. sativa showed greater draw-down of CO₂ from the intercellular space to the chloroplast stroma. It has also been argued that extreme enlargement of starch grains would distort the chloroplasts (Cave et al., 1981; Delucia et al., 1985; Pritchard et al., 1997), causing suppression of photosynthesis by the hindrance of CO₂ diffusion in the liquid phase (Nafziger and Koller, 1976; Nakano et al., 2000; Sawada et al., 2001).

Some species show decreases in g_m in response to instantaneous increases in ambient CO₂ concentration in the short-term (Flexas et al., 2007a; Tazoe et al., 2011); however, in other species including soybean and wheat g_m remains virtually unchanged (Bernacchi et al., 2005; Tazoe et al., 2009). We have previously examined responses to elevated CO₂ in Arabidopsis wild-type and stomatal-function mutants, and found that g_s and g_m respond independently to changes in CO₂ concentration(Mizokami et al., 2019). The mechanisms associated with the decrease in g_m in response to CO₂ differ from those in response to water deficit, in which ABA plays a key role (Mizokami et al., 2015).

There are some reports that growth at elevated CO₂ results in a decrease in g_m in some species. It has been debated whether such interspecific differences in g_m responses to the growth CO₂ concentration can be explained by changes in cell wall thickness and/or starch content. For example, Zhu et al. (2012) reported that under free-air CO₂ enrichment conditions, cell wall thickness significantly increased in rice but not in wheat, and a decrease in g_m was found only in rice. Kitao et al. (2015) found a negative correlation between g_m and starch content in the leaves of Betula platyphylla (Japanese white birch) grown under ambient or elevated CO₂. Thus, an increase in cell wall thickness and/or starch accumulation may cause a decrease in g_m at elevated CO₂. However, the contribution of each of these factors to the increase in resistance to CO₂ diffusion has not yet been fully elucidated since they usually occur simultaneously.

Teng et al. (2006) found that growth of Arabidopsis at elevated CO₂ resulted in cell wall thickening, although the effects on S_c, g_m, and photosynthesis were not reported. Given this fact, it should be possible to separately evaluate the effects of wall thickness and non-structural carbohydrates on g_m by using Arabidopsis carbohydrate-metabolism mutants grown at ambient or elevated CO₂. In this study, selected the Col-0 wild-type and three mutants of that are deficient in α-glucan/water dikinase (GWD1), plastidic phosphoglucomutase (PGM1), and cytosolic fructose 1, 6-bisphosphatase (CFBP1), which are key enzymes in starch breakdown, starch synthesis, and sucrose synthesis, respectively. Leaves of gwd1 show excess starch accumulation (Zeeman et al., 2004), leaves of pgm1 are starchless but show relatively high sugar concentrations (Periappuram et al., 2000), and leaves of cfbp1 are expected to show reduced sucrose content. The photosynthetic capacity of gwd1 is comparable to that of Col-0, whereas that of pgm1 and cfbp1 are significantly lower, especially when plants are grown at elevated CO₂ (C.K.A. Watanabe et al., unpublished results). In addition to these mutants, we grew wild-type plants for an extended period with either low or high nitrogen availability at either ambient or elevated CO₂ in order to obtain a wide range of control data. We also considered the reciprocal of mesophyll conductance per unit chloroplast surface area, i.e. mesophyll resistance per unit chloroplast surface area (S_c/g_m), which might be directly related to the resistance associated with cell wall thickness (Evans et al., 2009). Sugars and starch tend to accumulate at elevated CO₂, which may cause the down-regulation of photosynthesis due to a decrease in Rubisco content and activity (Sheen, 1994; Krapp and Stitt, 1995). We paid particular attention to the down-regulation of photosynthesis due to accumulation of non-structural carbohydrates.

Materials and methods

Plant material and growth conditions

We used Arabidopsis thaliana (L.) Heynh. accessions Col-0 (wild-type, WT), and the mutants CS3093 (At1g10760; gwd1/sex1), CS210 (At5g51820; pgm1), and SALK_064456C (At1g43670; cfbp1/fins1). cfbp1 corresponds to fins1 (Cho and Yoo, 2011). These mutants are each deficient in one of the enzymes related to carbohydrate metabolism, as follows: chloroplastic glucan water dikinase 1 (gwd1), chloroplastic phosphoglucomutase 1 (pgm1), and cytosolic fructose-1, 6-bisphosphatase 1 (cfbp1).

Seeds of all the accessions were sown in a mixture of autoclaved Metro Mix 350 (Sun Gro Horticulture, Bellevue, WA, USA) and vermiculite (1:1, v/v) in 200-ml plastic pots. The pots were placed in a cold room at 4 °C for two nights and transferred to two CO₂-controlled growth chambers (LPH-0.5P-SH, Nippon Medical & Chemical Instruments).

Plants were grown at a photosynthetically active photon flux density (PFD) of 200 μmol m⁻² s⁻¹ provided by fluorescent lamps during an 8-h light period, with day/night temperatures of 23/21 °C, and a relative humidity of 60%. The CO₂ concentrations in the growth chambers were controlled at either 400 ppm (ambient, aCO₂) or 800 ppm (elevated, eCO₂).

For the first week, plants were irrigated with deionized water. From the second week, they were irrigated with modified Hoagland solution 2–3 times a week. The solution contained 1.5 mM MgSO₄, 1.35 mM NaH₂PO₄, 0.1 mM NaCl, 0.05 mM Fe-EDTA, 0.05 mM H₃BO₃, 0.01 mM MnSO₄, 0.001 mM ZnSO₄, 1 μM CuSO₄, 0.5 μM Na₂MoO₄, 0. 2 μM CoSO₄. High- and low-nitrogen solutions of 8 mM N and 1 mM N, respectively, with 5 mM K⁺ and 5 mM Ca²⁺ were obtained by adding KNO₃, Ca(NO₃)₂, KCl, and CaCl₂.

WT plants irrigated with the high-nitrogen solution and grown for a total of 39–42 d are hereafter referred to as Col_HN, whilst WT plants irrigated with the low-nitrogen solution and grown for 38–42 d are referred to as Col_LN. To obtain leaves with different photosynthetic traits, some Col_HN plants were grown for a further 2 weeks (a total of 50–52 d) and these are referred to as Col₅₀. The gwd1, pgm1, and cfbp1 plants were irrigated with the high-nitrogen solution. Col_HN, Col_LN, and cfbp1 were grown for 38–42 d after sowing, whilst gwd1 and pgm1 were grown for 50–52 d after sowing due to their slower growth rates. Measurements of photosynthesis and the various traits were conducted on young mature leaves, which were the 11th–14th leaves to emerge.

Gas exchange and isotope measurements

Photosynthesis measurements were conducted during the second half of the light period. Gas exchange measurements were performed using a laboratory-made leaf cuvette (50×55×20 mm) designed for a single leaf of Arabidopsis (Tholen et al., 2008; Mizokami et al., 2015, 2019). All gas exchange measurements to evaluate photosynthetic traits were conducted at 1% O₂ to minimize the effect of photorespiration on g_m, using cylinders containing 100% N₂, 100% O₂, and 1% CO₂, which were mixed using mass-flow controllers (MM-3102L-NN, LINTEC, Tokyo, Japan). The O₂ concentration of the gas was checked using an oxygen sensor (3080-O_2, Walz, Effeltrich, Germany).

Light was provided by a metal halide lamp (PCS-UMX250, NPI, Tokyo, Japan). The PFD at the leaf level was monitored using a GaAs photodiode (G1738, Hamamatsu Photonics, Hamamatsu, Japan) placed in the chamber during the measurements. The GaAs sensor was calibrated against a quantum sensor (LI-190SA, LI-COR, Lincoln, NE, USA). The concentrations of CO₂ and H₂O in the air entering and leaving the cuvette were monitored using an infrared gas analyser (LI-7000, Li-Cor). The O₂ effect on the sensitivity of the infrared CO₂/H₂O analyser was corrected following Mizokami et al. (2015).

To measure photosynthesis at different levels of CO₂, we placed two leaves from two different plants together in the leaf cuvette. The leaf temperature and vapor pressure deficit (VPD) were kept at 22.5 °C and 0.85 kPa, respectively, and PFD was set to 1000 µmol photons m^–2 s^–1. For the plants grown at aCO₂, the first measurement was taken at an ambient CO₂ concentration (C_a) of 400 ppm, and then C_a was switched to 800 ppm. For the plants grown at eCO₂, the first measurement was taken at 800 ppm, and then C_a was switched to 400 ppm. Gas exchange parameters were recorded when the photosynthesis attained a steady-state rate, and the photosynthetic rates and stomatal conductance measured at C_a of 400 ppm (A₄₀₀ and g_s400) and 800 ppm (A₈₀₀ and g_s800) were obtained. The air entering and leaving the cuvette was collected in 30-ml Pyrex bottles with two stopcocks, and the ¹³C/¹²C ratios were measured with a mass spectrometer (IsoPrime 100, IsoPrime Ltd, Manchester, UK) to obtain mesophyll conductance at the two C_a levels (g_m400 and g_m800).

Calculation of mesophyll conductance

Mesophyll conductance was calculated as described by Tazoe et al. (2011):

$${\displaystyle \begin{array}{l}{\displaystyle g_m=\frac{(b-a_i-\frac{e{R}_d}{A+R_d})\frac{A}{C_a}}{a_b+(a_s-a_b)\frac{C_s}{C_a}+(b-a_s)\frac{C_i}{C_a}-\frac{e{R}_d(C_i-{\mathrm{\Gamma }}^{\ast })}{C_a(A+R_d)}-\frac{f{\mathrm{\Gamma }}^{\ast }}{C_a}-\mathrm{\Delta }}}\end{array}}$$

where C_a is the ambient CO₂ concentration, C_s is the CO₂ at the leaf surface, C_i is the intercellular CO₂, C_c is the CO₂ in the chloroplast stroma, a_b and a_s are the carbon isotope discriminations caused by diffusion through the boundary layer (2.9‰) and stomata (4.4‰), respectively, a_i is the carbon isotope discrimination during CO₂ diffusion/dissolution through water (1.8‰), b is the carbon isotope discrimination caused by the carboxylation reaction by Rubisco and phosphoenolpyruvate carboxylase (30‰), e is the discrimination in respiration (calculated as described by Tazoe et al., 2009, 2011), and Δ is the measured carbon isotope discrimination (Evans et al., 1986). The factor e required a correction because δ¹³C during plant growth was different to that during measurement (Wingate et al., 2007), its values were set to –24.4‰ and –17.5‰ for plants grown at aCO₂ and eCO₂, respectively, as described by Mizokami et al. (2019). f is the carbon isotope discrimination during photorespiration (11.6‰) and Γ* is the CO₂ compensation point without day respiration (Lanigan et al., 2008). R_d is the day respiration rate. For Γ* and R_d at 1% O₂, we assumed values for Arabidopsis Col-0 grown at aCO₂ and eCO₂ previously estimated by (Mizokami et al., 2019) using the Laisk method (Laisk, 1977). For the plants grown at aCO₂, Γ* and R_d were 11.3 µmol mol⁻¹ and 0.53 µmol m⁻² s⁻¹, respectively, whilst for the plants grown at eCO₂, Γ* and R_d were 5.4 µmol mol⁻¹ and 0.45 µmol m⁻² s⁻¹, respectively. Although the values of Γ* at 1% O₂ were higher than those estimated by Walker and Cousins (2013) in a previous study, a sensitivity analysis has shown that variations in Γ* have a minor impact on g_m (Mizokami et al., 2019). We did not apply the ternary correction (Farquhar and Cernusak, 2012) since the VPD (0.872±0.083 kPa) and transpiration rate (0.0018±0.0005 mol m⁻² s⁻¹) were sufficiently low in our present study: for Col-0 grown at aCO₂ we determined that g_m400 calculated with or without the ternary correction differed by only 0.005 mol m⁻² s⁻¹, and g_m800 differed by only 0.0031 mol m⁻² s⁻¹.

Sampling

The leaves from the two plants that had been used for the photosynthesis measurements were subsequently sampled for microscopic and biochemical analyses. Two or three similarly aged leaves including leaves used for the photosynthesis measurements were sampled to determine leaf morphological traits, leaf nitrogen content, δ¹³C, and the content of non-structural carbohydrates (starch and soluble sugars). Discs were taken from the leaves and measurements were taken after drying. The other two or three similarly aged leaves were stored at –80 °C for later determination of Rubisco content and cell wall content. In total, 4–6 leaves were sampled from the two plants and the set of data obtained from them was dealt with as one biological replicate.

Light and transmission electron microscope analyses

Small lamina segments were cut with a razor blade, immediately immersed in 0.2 M sodium cacodylate buffer (pH 7.0) containing 2.5% paraformaldehyde and 2% glutaraldehyde, and vacuum-infiltrated until most of the segments sank. They were then stored at 4 °C overnight. The segments were post-fixed in 1% OsO₄ for 1 h and dehydrated in an ethanol series. Some segments were embedded in Technovit 7100 (Heraus Holding, Hanau, Germany) and used for light-microscope analysis. The other segments were further dehydrated in a series of propylene oxide and embedded in Spurr’s resin for TEM analysis.

For light microscopy, leaf transverse sections of 1 µm thick were cut on an ultramicrotome (Reichert Ultracut S, Leica, Vienna, Austria) with a glass knife and stained with a 0.1% (w/v) Toluidine Blue solution in 1% (w/v) sodium borate. Leaf thickness, the surface area of mesophyll cell walls exposed to the intercellular space (S_mes, m² m⁻²), and the surface area of chloroplasts facing the intercellular space (S_c, m² m⁻²) were determined for each replicate. S_mes and S_c were calculated using the curvature correction factor following Thain (1983) and Evans et al. (1994). The mean value of the correction factor was 1.24 in the present study. For TEM, ultrathin sections of 70 nm thick were cut on the ultramicrotome with a diamond knife (Ultra 45°, Diatome AG, Switzerland) and placed on a 150-mesh copper grid. The grids were stained with a 2% uranyl acetate solution followed by a lead citrate solution. The sections were examined on a JEM-1010 TEM (JEOL, Japan). The wall thickness of mesophyll cells was measured on the ultrathin sections using the ImageJ software (Schneider et al., 2012). The thickness was calculated for randomly selected cells by dividing the cross-sectional area of the cell wall by its length. On average, 58 μm of cell wall length was analysed for each plant.

Leaf mass per unit area and non-structural carbohydrates

The 2–3 dried leaf discs were weighed to determine leaf mass per area (LMA, g m⁻²) and then ground with a Multi-beads Shocker (Yasui Kikai, Osaka, Japan). The ground samples (3–5 mg each) were then used to determine the contents of glucose, sucrose, and starch according to Araya et al. (2006). Soluble sugars were extracted with 80% ethanol, and sucrose was hydrolysed to glucose and fructose with an invertase solution (Wako Chemical, Osaka, Japan). The precipitate was treated with amyloglucosidase (A-9228, Sigma-Aldrich, St. Louis, MO) to break down starch into glucose. Finally, glucose, and glucose equivalents of sucrose and starch, were quantified using a Glucose CII test kit (Wako Chemicals). The content of sugars and starch were expressed on a leaf-area basis. Structural LMA (sLMA, g m⁻²) consisting of proteins, minerals, lipids, and soluble and insoluble phenolics (Poorter et al., 2006) was calculated by subtracting the content of non-structural carbohydrates from LMA (Bertin et al., 1999).

Nitrogen and δ¹³C

The nitrogen content (Nmass, g N g^–1) and δ¹³C of the ground leaf samples were determined with a CN analyser (Vario Micro, Elementar Analyzensysteme GmbH, Hanau, Germany) connected to an isotopic ratio mass spectrometer (IsoPrime100, IsoPrime, Manchester, UK). N content per area (N_area, g N m^–2) was calculated as a product of Nmass and LMA. δ¹³C (‰) as follows:

$${\displaystyle \begin{array}{l}{\displaystyle {\delta }^{13}\mathrm{C}=\frac{R_{sample}-R_{standard}}{R_{standard}}\times 1000}\end{array}}$$

where R_sample and R_standard are the ¹³C/¹²C ratios of the sample and the standard (PDB, 0.011180).

Rubisco and cell wall contents

The two or three frozen leaf discs were used to determine the contents of Rubisco and cell wall materials as described previously (Mizokami et al., 2015; Sugiura et al., 2017). The frozen discs were homogenized in Tris–HCl buffer (62.5 mM, pH 6.8) using a Multi-beads Shocker (Yasui Kikai) to extract Rubisco. After extraction, SDS–PAGE, Coomassie Brilliant Blue staining, and formamide (NACALAI TESQUE, Kyoto, Japan) extraction were conducted and the Rubisco content was determined by measuring absorbance of the extract at 595 nm. From the residual pellet after the Rubisco extraction, starch was removed using amyloglucosidase and the cytoplasmic protein was removed using 1 M NaCl, and the resultant pellet was assumed to contain the cell wall materials. The pellet was weighed after drying, and cell wall mass per area (CMA, g m⁻²) was calculated.

Δ(S_c/g_m400), Δcell wall thickness, and Δstarch

Because g_m is proportional to the surface area available for diffusion, it is sometimes expressed as per unit chloroplast surface area adjacent to the intercellular airspaces (g_m/S_c; Terashima et al., 2011). Because it is the resistance that is expected to be linearly related to cell wall thickness, we used the reciprocal of the conductance per S_c in our analyses, i.e. S_c/g_m400. S_c/g_m400 and the draw-down of CO₂ from the intercellular space to the site of carboxylation in the chloroplast (C_i–C_c) were then related to cell wall thickness and starch content. To quantify the effect of growth CO₂ on S_c/g_m400, we used ΔX (%), the percent change in the trait X between the plants grown at eCO₂ and those grown at aCO₂ normalized to those grown at aCO₂, as follows:

$${\displaystyle \begin{array}{l}{\displaystyle X=\frac{X_{eC{O}_2}-X_{aC{O}_2}}{X_{aC{O}_2}}\times 100}\end{array}}$$

ΔX was calculated for S_c/g_m400, cell wall thickness, and starch content.

Statistical analysis

Statistical tests were performed using Systat13 (Systat Software, Richmond, CA, USA). The effects of growth CO₂ on anatomical, morphological, physiological, and photosynthetic traits were evaluated by ANOVA followed by Tukey’s multiple comparison test. δ¹³C of the leaves was compared only by Tukey’s test at each growth CO₂ level since δ¹³C in the air differed depending on the growth chambers. Photosynthetic traits were also compared between the CO₂ concentrations during the measurements by Student’s t-test.

Results

Leaf anatomical traits

Leaf anatomical traits of the different Arabidopsis lines were analysed using light and electron micrographs of leaf transverse sections (Supplementary Figs S1, S2 at JXB online) and were found to differ markedly (Table 1). Only the cell wall thickness was increased significantly by growth at eCO₂. The thickness was greatest in Col₅₀ and lowest in cfbp1 at both aCO₂ and eCO₂ (Table 1, Supplementary Fig. S2). Thickness was greater in eCO₂ by ~20–35% in the WT and pgm1, whereas it did not differ in gwd1 or cfbp1. Leaf thickness, S_c, S_m, and S_c/S_m were greatest in pgm1, but the differences between the mutants and the WT were small.

Table 1.

Anatomical and physiological traits of Arabidopsis Col-0 and carbohydrate-mutants grown under ambient CO₂ (aCO₂, 400 ppm) or elevated CO₂ (eCO₂, 800 ppm)

	Growth CO2	ColHN	Col50	ColLN	gwd1	pgm1	cfbp1
Leaf thickness (μm)	aCO2 (ns)	201.8±28.3b	216.6±32.7b	221.6±13.8b	223.6±20.7b	306.0±39.6a	184.4±6.0b
	eCO2	202.4±24.0B	191.9±19.5B	215.5±23.0B	233.1±17.4B	355.2±69.4A	193.2±25.0B
Cell wall thickness (μm)	aCO2**	0.233±0.023ab	0.286±0.032a	0.221±0.032ab	0.247±0.012ab	0.246±0.012ab	0.210±0.059b
	eCO2	0.317±0.033A	0.336±0.063A	0.257±0.024AB	0.246±0.052AB	0.295±0.059AB	0.195±0.015B
S c (m2 m−2)	aCO2 (ns)	6.52b±0.49b	8.29±1.26b	7.59±0.52b	7.76±1.52b	12.82±1.22b	6.54±1.12b
	eCO2	7.15±1.21B	8.19±0.80B	8.22±0.85B	7.88±2.09B	14.63±3.21A	6.78±0.34B
S m (m2 m−2)	aCO2 (ns)	8.41±0.58b	10.13±1.72b	9.53±0.91b	10.19±1.47b	14.15±1.63a	8.37±1.54b
	eCO2	9.19±1.43B	10.29±1.00B	9.98±1.34B	9.90±2.24B	16.17±3.57A	8.31±0.30B
S c/Sm (m2 m−2)	aCO2 (ns)	0.776±0.054b	0.820±0.038ab	0.798±0.022b	0.757±0.057b	0.908±0.032a	0.784±0.056b
	eCO2	0.777±0.017B	0.797±0.044B	0.828±0.065AB	0.790±0.050B	0.904±0.031A	0.816±0.033AB
Narea (g N m−2)	aCO2**	1.167±0.078c	1.391±0.060b	0.881±0.139d	1.194±0.039c	1.731±0.085a	1.071±0.070c
	eCO2	1.227±0.066BC	1.427±0.133B	0.813±0.123D	1.342±0.049B	1.979±0.110A	1.120±0.033C
Rubisco (g m−2)	aCO2**	1.050±0.278a	1.061±0.111a	0.852±0.112a	1.015±0.045a	1.183±0.076a	1.175±0.148a
	eCO2	1.271±0.078A	1.091±0.148A	0.695±0.110B	1.128±0.044A	1.151±0.093A	1.158±0.149A
δ13C (‰)	aCO2	–32.5±0.34b	–33.0±0.19ab	–32.5±0.29bc	–32.0±0.19c	–33.5±0.14a	–32.6±0.23b
	eCO2	–40.4±0.48AB	–39.3±0.24C	–39.6±0.10BC	–38.9±0.28C	–41.2±0.56A	–40.9±0.65A

S _c is the chloroplast surface area exposed to the intercellular space; S_m is the mesophyll surface area exposed to the intercellular space; and N_area is leaf nitrogen content per area. Values are means (±SD), n=4. The overall effect of growth CO₂ was evaluated by ANOVA (**P<0.01; ns, P>0.05), except for δ¹³C (see Methods). Significant differences within each CO₂ treatment were determined using ANOVA followed by Tukey’s test (P<0.05), and are indicated by different letters (lower-case for aCO₂, upper-case for eCO₂).

Leaf morphological and physiological traits

Leaf morphological and physiological traits were determined for the leaves that had been used for the photosynthesis measurements. LMA, sLMA, and CMA were greater in the eCO₂ samples, and they differed significantly among the lines (Fig. 1). In both aCO₂ and eCO₂ plants LMA was highest in gwd1 whilst sLMA and CMA were highest in pgm1. These traits tended to be greater in Col₅₀ than Col_HN.

Fig. 1.

Leaf mass per area (LMA), structural leaf mass per area (sLMA), and cell wall mass per area (CMA) in leaves of Arabidopsis Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1 grown under either ambient CO₂ (aCO₂, 400 ppm) or elevated CO₂ (eCO₂, 800 ppm). Data are means (±SD), n=4. The overall effect of growth CO₂ was evaluated by ANOVA and is indicated above the graph (**P<0.01). Significant differences among the lines within each CO₂ treatment were determined using ANOVA followed by Tukey’s test (P<0.05), and are indicated by different letters (lower-case for aCO₂, upper-case for eCO₂).

The level of non-structural carbohydrates was also increased by elevated growth CO₂. The starch content was highest in gwd1 (the starch-excess mutant) and lowest in pgm1 (Fig. 2A). Within the WT, the starch content was highest in Col_LN in both aCO₂ and eCO₂. Soluble sugars were highest in pgm1, followed by gwd1 (Fig. 2B). Chloroplasts with excess starch were clearly visible in light and electron micrographs of gwd1, whilst starchless chloroplasts could be seen in pgm1 (Supplementary Figs S1, S2).

Fig. 2.

(A) Starch and (B) soluble sugars in the leaves of Arabidopsis Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1 grown under ambient CO₂ (aCO₂, 400 ppm) or elevated CO₂ (eCO₂, 800 ppm). Data are means (±SD), n=4. The overall effect of growth CO₂ was evaluated by ANOVA and is indicated above the graph (**P<0.01). Significant differences among the lines within each CO₂ treatment were determined using ANOVA followed by Tukey’s test (P<0.05), and are indicated by different letters (lower-case for aCO₂, upper-case for eCO₂).

N_area was slightly increased by elevated growth CO₂, whereas Rubisco content was not (Table 1). N_area was highest in pgm1 and lowest in Col_LN in both aCO₂ and eCO₂. The Rubisco content in Col_LN grown at eCO₂ was significantly lower than in the other lines, but otherwise no other differences were observed. There was a positive correlation between the Rubisco content and N_area among the WT and mutants gwd1 and cfbp1 grown at aCO₂ and eCO₂ (Supplementary Fig. S3). However, pgm1 did not fit this pattern, having relatively low Rubisco content despite having the highest N_area.

δ¹³C differed significantly among the lines (Table 1), with the highest value in gwd1 and lowest in pgm1 in both aCO₂ and eCO₂. The WT showed intermediate values for both aCO₂ and eCO₂. There were negative relationships between δ¹³C and C_c (Supplementary Fig. S4A, B) and positive correlations between δ¹³C value and starch content for all the lines (Supplementary Fig. S4C, D).

Leaf photosynthetic characteristics

Photosynthesis measurements were conducted at 400 ppm CO₂ (A₄₀₀, g_s400, and g_m400) and 800 ppm (A₈₀₀, g_s800, and g_m800) for plants grown at aCO₂ and eCO₂.

A ₄₀₀ and A₈₀₀ were higher in the WT and gwd1 than in pgm1 and cfbp1 at both aCO₂ and eCO₂ (Fig. 3A, B). A₈₀₀ was significantly lower than A₄₀₀ in Col_HN, Col₅₀, and Col_LN at aCO₂ (Fig. 3A), and also significantly lower in Col_HN, Col₅₀, pgm1, and cfbp1 at eCO₂ (Fig. 3B). g_s400 and g_s800 were lowest in pgm1 at both aCO₂ and eCO₂ (Fig. 3C, D). g_s800 was significantly lower than g_s400 in all the lines at both aCO₂ and eCO₂. g_m400 and g_m800 were lowest in cfbp1 at both aCO₂ and eCO₂ (Fig. 3E, F). g_m800 was significantly lower than g_m400 in all the lines at both aCO₂ and eCO₂ (Fig. 3C, D) except for cfbp1 at aCO₂ and pgm1 at eCO₂. Although the photosynthetic rates (A₄₀₀ and A₈₀₀) and mesophyll conductances (g_m400 and g_m800) were not significantly affected by growth CO₂ conditions, g_m400 was slightly lower in the WT and pgm1 grown at eCO₂. Stomatal conductance (g_s400 and g_s800) was significantly increased by elevated growth CO₂, especially in gwd1, pgm1, and cfbp1.

Fig. 3.

Photosynthetic characteristics of Arabidopsis Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1 grown under ambient CO₂ (aCO₂, 400 ppm) or elevated CO₂ (eCO₂, 800 ppm). (A, B) Photosynthetic rates measured at 400 ppm CO₂ (A₄₀₀) and 800 ppm CO₂ (A₈₀₀) for plants grown in (A) aCO₂ and (B) eCO₂. (C, D) Stomatal conductance measured at 400 ppm CO₂ (g_s400) and 800 ppm CO₂ (g_s800) for plants grown in (A) aCO₂ and (B) eCO₂. (E, F) Mesophyll conductance measured at 400 ppm CO₂ (g_m400) and 800 ppm CO₂ (g_m800) for plants grown in (A) aCO₂ and (B) eCO₂. Data are means (±SD), n=4. The overall effect of growth CO₂ was evaluated by ANOVA (**P<0.01; ns, P>0.05). Significant differences among the lines within each growth CO₂ treatment were determined using ANOVA followed by Tukey’s test (P<0.05), and are indicated by different letters (lower-case for aCO₂, upper-case for eCO₂). Significant differences between measurements at 400 ppm CO₂ and 800 ppm CO₂ within each line were determined using Student’s t-test (**P<0.01; *P<0.05; ns, P>0.05).

Factors that determine photosynthetic characteristics

Relationships among photosynthetic characteristics and various morphological and physiological traits were analysed to determine which traits were involved in the responses to growth CO₂ conditions. Here, we focus on the photosynthetic characteristics measured at 400 ppm CO₂.

There was a positive correlation between A₄₀₀ and Rubisco content for the WT and gwd1, whereas pgm1 and cfbp1 showed lower photosynthetic rates despite having high Rubisco contents (Fig. 4A). The relationship between A₄₀₀ and g_s400 was not significant (Fig. 4B); however, A₄₀₀ was positively correlated with g_m400 for all the lines except for pgm1 (Fig. 4C).

Fig. 4.

Relationships between photosynthetic rate, Rubisco content per area, stomatal conductance, and mesophyll conductance measured at 400 ppm CO₂ for Arabidopsis lines grown under ambient CO₂ (aCO₂, 400 ppm) or elevated CO₂ (eCO₂, 800 ppm). The lines, as indicated in the key in (A), were Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1. (A) Photosynthetic rate (A₄₀₀) and Rubisco content per area, (B) A₄₀₀ and stomatal conductance (g_s400), and (C) A₄₀₀ and mesophyll conductance (g_m400). Data are means (±SD), n=4. Regression lines are shown in (A) for Col_HN, Col₅₀, Col_LN, and gwd1 (R²=0.68), and in (C) for Col_HN, Col₅₀, Col_LN, gwd1, and cfbp1 (R²=0.60).

Relationships among photosynthetic characteristics and accumulation of sugars and starch were investigated to determine whether feedback regulation was occurring (Fig. 5). No negative correlations with the soluble sugar or starch contents were found for either Rubisco content (Fig. 5A, B), A₄₀₀ (Fig. 5C, D), or g_m400 (Fig. 5E, F).

Fig. 5.

Relationships between photosynthetic characteristics measured at 400 ppm CO₂ and content of soluble sugars (glucose and sucrose) and starch for Arabidopsis lines grown under ambient CO₂ (400 ppm) or elevated CO₂ (800 ppm). The lines, as indicated in the key in (A), were Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1. Rubisco versus contents of glucose and sucrose (A) and versus starch (B), photosynthetic rate (A₄₀₀) versus contents of glucose and sucrose (C) and versus starch (D), and mesophyll conductance (g_m400) versus contents of glucose and sucrose (E) and versus starch (F). Data are means (±SD), n=4.

Relationships between the mesophyll conductance and morphological and anatomical features were investigated. There was no overall relationship between S_c/g_m400 and cell wall thickness among the lines (Fig. 6A); however, within each line S_c/g_m400 increased with the increase in cell wall thickness. No relationship was observed between S_c/g_m400 and starch content among the lines (Fig. 6B). The draw-down of CO₂ from the intercellular space to the site of carboxylation in the chloroplast (C_i–C_c) was positively correlated with cell wall thickness for all the lines except for pgm1 (Fig. 6C). There was no clear relationship between C_i–C_c and starch content among the lines (Fig. 6D). Cell wall thickness was positively correlated with cell wall mass per area (CMA) for all the lines and CMA was also positively correlated with sLMA (Supplementary Fig. S5B, C). We also expected a positive correlation between mesophyll conductance and S_c, but no clear relationship was found (Supplementary Fig. S6).

Fig. 6.

Relationships between mesophyll resistance and draw-down of CO₂ from intercellular space to the chloroplast (C_i–C_c) measured at 400 ppm CO₂ and cell wall thickness and starch content for Arabidopsis lines grown under ambient CO₂ (400 ppm) or elevated CO₂ (800 ppm). The lines, as indicated in the key in (A), were Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1. Mesophyll resistance (S_c/g_m400) versus cell wall thickness (A) and versus starch content (B), and draw-down of CO₂ versus cell wall thickness (C) and versus starch content (D). Data are means (±SD), n=4. The regression line for Col_HN, Col₅₀, Col_LN, gwd1, and cfbp1 is shown in (C) (R²=0.65).

We examined how the changes in the cell wall thickness and starch content in response to growth CO₂ conditions affected mesophyll resistance using ΔS_c/g_m400, the percentage change in a given line between the plants grown at eCO₂ and those grown at aCO₂ normalised to aCO₂. ΔS_c/g_m400 showed a strong positive correlation with ΔCell wall thickness but there was no clear correlation between Δ S_c/g_m400 and ΔStarch content (Fig. 7A, B).

Fig. 7.

Relationships between changes in mesophyll resistance and changes in cell wall thickness and starch content in response to CO₂ growth conditions for Arabidopsis lines grown under ambient CO₂ (400 ppm) or elevated CO₂ (800 ppm). The lines, as indicated in the key in (A), were Col-0 grown with high nitrogen for 39–42 d (Col_HN), with low nitrogen for 38–42 d (Col_LN), or with high nitrogen for 50–52 d (Col₅₀), and the carbohydrate-metabolism mutants gwd1, pgm1, and cfbp1. Percentage change in mesophyll resistance (ΔS_c/g_m400) measured at 400 ppm CO₂ versus (A) percentage change in cell wall thickness (ΔCell wall thickness) and versus percentage change in starch content (ΔStarch). For calculation of the percentage changes see Methods. Data are means (±SD), n=4. The regression line for all the data is shown in (A) (R²=0.97).

Discussion

The cell wall is a greater limiting factor for g_m than starch content

Our results indicated that among the changes in various morphological, anatomical, and physiological traits that we observed, the increases in cell wall mass and thickness in Arabidopsis had the major impact on mesophyll conductance. The excess accumulation of starch was not linked to a decrease in g_m (Fig. 5F), whereas within each line the increase in cell wall thickness was associated with an increase in mesophyll resistance, S_c/g_m (Fig. 6A), and a greater draw-down of CO₂ from the intercellular space to the chloroplast, C_i−C_c (Fig. 6C). It was clear that the percentage change in mesophyll resistance between eCO₂ and aCO₂ plants (ΔS_c/g_m400) was associated with the percentage change in cell wall thickness (ΔCell wall thickness) across all lines (Fig. 7A). The fact that there was no clear correlation between ΔS_c/g_m400 and ΔStarch indicated the lack of effect of the latter (Fig. 7B). The measurement of cell wall thickness requires observations with an electron microscope and is therefore laborious, whereas measurement of cell wall mass is relatively easy. Since cell wall thickness was highly correlated with cell wall mass per area (CMA, Supplementary Fig. S5A), the latter could be an alternative indicator of cell wall thickness when investigating the environmental responses of g_m in the same species. As also shown in a previous study (Sugiura et al., 2017), there was a strong positive correlation between CMA and sLMA (Supplementary Fig. S5B), indicating that sLMA could also be an efficient parameter within a plant species that reflects the thickness and mass of the cell wall.

When plants are grown under elevated CO₂ conditions, increases are observed in the thickness and mass of cell walls (Teng et al., 2006) and in the accumulation of non-structural carbohydrates (Kitao et al., 2015; Sugiura et al., 2017), whilst decreases are observed in g_m (Zhu et al., 2012). It has been proposed that excess accumulation of starch in mesophyll cells could hinder CO₂ diffusion and result in decreased g_m (Nafziger and Koller, 1976; Nakano et al., 2000; Sawada et al., 2001). Zhu et al. (2012) reported a decrease in g_m and an increase in starch content in flag-leaves of rice grown with free-air CO₂ enrichment); however, inter-relationships among the traits were not examined in detail. In our present study, although the starch-excess mutant gwd1 accumulated starch at both aCO₂ and eCO₂ (Fig. 2A), g_m did not decrease (Fig. 3E, F). Our results suggested that direct effects of non-structural carbohydrates on CO₂ diffusion and g_m were minor, whereas the increased cell wall mass and thickness caused a decrease in g_m and an increase in S_c/g_m under elevated CO₂ conditions (Table 1, Figs 1, 7A). Thus, we conclude that the increases in cell wall mass and thickness contributed more to the decrease in g_m and the increase in S_c/g_m than starch has been observed to do in previous studies.

Although we found a simultaneous increase in S_c/g_m and cell wall thickness in each individual line, there was not a positive correlation between them overall across all lines (Fig. 6A). The fact that Col_LN, pgm1, and cfbp1 showed different patterns suggests that variation in other physiological processes and anatomical features could also be involved. Mesophyll conductance can be separated into gaseous and liquid phases, and the liquid phase can be further separated into five phases, namely cell wall, plasma membrane, cytosol, chloroplast envelope, and stroma (Evans et al., 2009). It is considered that CO₂ diffusion across the cell wall and across the plasma membrane are the major limiting steps (Terashima et al., 2011). In recent years, it has been suggested that aquaporins—proteins involved in water transport in the plasma membrane—may affect the permeability of CO₂ through membranes (Terashima et al., 2006; Mori et al., 2014; Groszmann et al., 2017). Therefore, it is possible that Col_LN grown under low N availability showed higher S_c/g_m400 regardless of the thinner cell wall due to a lower expression of aquaporins in the plasma membranes. The expression levels of aquaporins are significantly decreased in roots of Oryza sativa grown under low N availability (Ishikawa-Sakurai et al., 2014), so it is possible that a similar effect may occur in the leaf mesophyll cells in Arabidopsis. In cfbp1, in which the sucrose synthesis pathway is impaired, the photosynthetic rate, A₄₀₀, was limited by the low g_m (Figs 3, 4A) even though this line showed a high value of N_area, high Rubisco content, and the thinnest cell walls (Table 1). Therefore, factors other than cell wall thickness, such as the plasma membrane, cytoplasm, chloroplast envelope, might be markedly changed in cfbp1. Furthermore, since the composition of the cell wall is markedly different among carbohydrate-metabolism mutants of Arabidopsis, including pgm1 and gwd1 (Engelsdorf et al., 2017), it is possible that differences in the porosity of the cell wall affected g_m in these mutants, as discussed by Evans et al. (2009). Future metabolomic, transcriptomic, and micro-anatomical studies will be required to elucidate the mechanisms underlying the changes in g_m in these mutants.

It is reported that g_m in plants within the same functional group, such as annuals, broad-leaved deciduous trees, and broad-leaved evergreen trees, is roughly proportional to S_c (Terashima et al., 2006). Such a positive correlation between g_m and S_c has been found for WT and chloroplast-positioning mutants of Arabidopsis (Tholen et al., 2008). However, there was no correlation between g_m400 and S_c for the lines used in our present study (Supplementary Fig. S6), possibly due to the fact that the positioning of chloroplasts was normal in the lines that we used and that the range of variation in S_c was not large enough.

Short-term changes in g_m in response to elevated CO₂ are independent of non-structural carbohydrates

Short-term changes in g_m in response to the increase in the CO₂ concentration from 400 ppm to 800 ppm during our photosynthesis measurements were observed in all the carbohydrate-metabolism mutants as well as in the WT (Fig. 3E, F). This suggests that the amount of non-structural carbohydrates was not involved in the short-term changes in g_m. A previous study had shown that g_s and g_m were independently regulated in response to CO₂ concentration during gas-exchange measurements in Arabidopsis WT plants and two mutants of stomatal function, open stomata 1 (ost1) and slow-type anion channel 1–2 (slac1-2) (Mizokami et al., 2019). Since these plants showed stable g_s and g_m over an hour, the observed changes when the ambient CO₂ concentration was changed from either 400 ppm to 800 ppm or vice versa were not caused by the stress of long-term measurements. In the present study, pgm1 showed the lowest g_s whereas cfbp1 showed the lowest g_m. This also suggests that g_s and g_m are not necessarily coordinated well in Arabidopsis. In addition to the short-term changes in g_m, the positive correlations between A₄₀₀ and Rubisco and g_m400 that we observed were consistent with previous reports (Fig. 4; Flexas et al., 2007b; Tholen et al., 2008; José Javier et al., 2017).

Ecophysiological significance of the down-regulation of photosynthesis

The photosynthetic rate measured at 800 ppm (A₈₀₀) was significantly lower than that measured at 400 ppm (A₄₀₀) (Fig. 3A, B), which can be explained by a limitation in triose phosphate use (Sharkey, 1985; Harley and Sharkey, 1991). According to these studies, CO₂ assimilation rate measured at low oxygen concentration reaches a maximum at lower C_i than that measured at normal concentration. In our present study, all the gas-exchange measurements were performed at 1% O₂, which led to a decrease in photosynthetic rate with the increase in C_a and C_i (Woo and Wong, 1983).

The decrease in Rubisco was not observed in gwd1, which showed the highest starch content, and in pgm1, which showed the highest soluble sugars (Table 1, Fig. 5A, B). This strongly supports the idea that accumulation of soluble sugars and starch does not cause the down-regulation of photosynthesis through the decrease in Rubisco content in Arabidopsis. Although an increase in the content of starch and decreases in Rubisco and A₄₀₀ were found simultaneously only in Col_LN (Fig. 5A–D), this would be related to the promotion of leaf senescence and nitrogen remobilization to other organs. Ludewig and Sonnewald (2000) reported that the down-regulation of photosynthesis in tobacco was found in senescing leaves only, and there was no correlation between the transcript levels of photosynthesis-related genes and soluble sugar contents.

It has been argued that the sensitivity of maximum photosynthetic capacity to carbohydrate varies greatly among plant species (Sugiura et al., 2018). Sugiura et al. (2015, 2017) showed that Raphanus sativus, which in common with Arabidopsis belongs to Brassicaceae, can be classified as a carbohydrate insensitive species. They reciprocally grafted plants with different sink activities and found that neither the maximum photosynthesis nor the Rubisco content was down-regulated even though excessive non-structural carbohydrates were accumulated in the source leaves in shoots grafted to the stock of a variety with a low sink activity. Therefore, it is possible that Brassicaceae can be classified as having carbohydrate-insensitive species. Since carbohydrate-insensitive species such as R. sativus and Glycine max show increases in CMA in response to accumulation of non-structural carbohydrates, it is possible that the cell wall is also an important carbohydrate sink in the response to changes in the sink–source balance (Sugiura et al., 2017, 2018). Thus, these plants might down-regulated photosynthetic capacity through the suppression of CO₂ conductance by increasing cell wall thickness in order to avoid excess accumulation of non-structural carbohydrates.

Since Rubisco discriminates less against ¹³C at low concentrations of CO₂ in the chloroplast stroma (C_c), we expected that the δ¹³C values of the leaves would reflect C_c, and indeed we found negative correlations between them (Supplementary Fig. S4A, B). Meanwhile, positive correlations were found between δ¹³C and starch content for all the lines at both aCO₂ and eCO₂ (Supplementary Fig. S4C, D). Brugnoli et al. (1988) found higher δ¹³C values in starch than in soluble sugars. They considered that the slower turnover rate of starch is a possible reason, since it would not be available for isotopic exchange for several days, thus resulting in a higher concentration of ¹³C. Hence, attention should be paid when treating the δ¹³C value as an indicator of C_c when non-structural carbohydrates are accumulated excessively.

The Arabidopsis lines that we grew under aCO₂ or eCO₂ showed cell wall thicknesses ranging from 0.195–0.336 μm (Table 1). To confirm our findings, it would be necessary to investigate the relationship between g_m and cell wall thickness in plants over a wider range of wall thickness. It would be possible to obtain plants with thinner cell walls by growing them under low-light conditions (Conn et al., 2011; Lehmeier et al., 2017) and with thicker cell walls by growing them under high-light or more elevated CO₂ conditions (Teng et al., 2006). It would also be interesting to determine the extent to which cell walls become thickened depending on the CO₂ growth conditions. Another issue is to determine interspecific differences in how plastically cell wall thickness and g_m are regulated in the field in response to micro-environmental changes such as temperature (von Caemmerer and Evans, 2015), light intensity, and CO₂ concentration (Tazoe et al., 2009). This would reveal the ecological significance of the down-regulation of photosynthesis.

Conclusions

Our results suggest that elevated CO₂ conditions could decrease mesophyll conductance, g_m, and increase mesophyll resistance, S_c/g_m, through increases in cell wall mass and thickness in Arabidopsis in the long-term. On the other hand, excess starch accumulation had minor effects on g_m and S_c/g_m. We also demonstrated that short-term changes in g_m in response to elevated CO₂ are independent of non-structural carbohydrates in Arabidopsis. Our study provides clues to the ecophysiological significance of the down-regulation of photosynthesis that are observed under elevated CO₂ conditions.

Supplementary data

Supplementary data are available at JXB online.

Fig. S1. Cross-sections of leaves of Arabidopsis grown under 400 ppm or 800 ppm CO₂.

Fig. S2. Electron microscopy of leaves of Arabidopsis grown under 400 ppm or 800 ppm CO₂.

Fig. S3. Relationship between Rubisco content per area and leaf nitrogen content per area in leaves of Arabidopsis grown under 400 ppm or 800 ppm CO₂.

Fig. S4. Relationships between δ¹³C and CO₂ concentration in the chloroplast stroma and starch content in leaves of Arabidopsis grown under 400 ppm or 800 ppm CO₂.

Fig. S5. Relationships between cell wall thickness and cell wall mass per area (CMA) and that between CMA and structural leaf mass per area in leaves of Arabidopsis grown under 400 ppm or 800 ppm CO₂.

Fig. S6. Relationship between mesophyll conductance measured at 400 ppm CO₂ and the chloroplast surface area exposed to intercellular space in leaves of Arabidopsis grown under 400 ppm or 800 ppm CO₂.