Importance of leaf anatomy in determining mesophyll diffusion conductance to CO2 across species: quantitative limitations and scaling up by models.

Foliage photosynthetic and structural traits were studied in 15 species with a wide range of foliage anatomies to gain insight into the importance of key anatomical traits in the limitation of diffusion of CO2 from substomatal cavities to chloroplasts. The relative importance of different anatomical traits in constraining CO2 diffusion was evaluated using a quantitative model. Mesophyll conductance (g m) was most strongly correlated with chloroplast exposed surface to leaf area ratio (S c/S) and cell wall thickness (T cw), but, depending on foliage structure, the overall importance of g m in constraining photosynthesis and the importance of different anatomical traits in the restriction of CO2 diffusion varied. In species with mesophytic leaves, membrane permeabilities and cytosol and stromal conductance dominated the variation in g m. However, in species with sclerophytic leaves, g m was mostly limited by T cw. These results demonstrate the major role of anatomy in constraining mesophyll diffusion conductance and, consequently, in determining the variability in photosynthetic capacity among species.

Introduction

Leaf anatomical characteristics are key functional and adaptive traits determining plant capacity to thrive in specific environments, in particular, because these traits also have important implications for foliage potential photosynthesis (Niinemets ; Scafaro ; Terashima ). Analysis of global variations in leaf functional traits—the leaf economics spectrum—has established that the variation in leaf dry mass per area (M A) is strongly associated with other key leaf traits such as maximum photosynthetic capacity per dry mass (A mass), leaf life span, nitrogen and phosphorous contents per dry mass, and respiration (Wright ). Species with lower M A present short leaf life spans, high photosynthetic capacities and nutrient contents, and low leaf area construction costs, resulting in fast growth in environments with high availability of resources. In contrast, species with higher M A and lower A mass present the opposite suite of traits and have higher cost for leaf area formation, particularly due to investment in vasculature and cell walls (Niinemets ; Hikosaka & Shigeno, 2009) and overall improved resistance to low fertility and drought, but low growth rates (Niinemets, 2001; Wright ). It has been hypothesized that the negative relationship between M A and photosynthetic capacity is partly because of greater biomass investment in support tissues and cell wall thickening involving stronger CO2 diffusion limitations to photosynthesis (Niinemets, 1999; Wright ; Niinemets )

Mesophyll conductance to CO2 (g m) is the measure of the CO2 diffusion facility from substomatal cavities to the sites of carboxylation in the chloroplasts (Flexas , 2012) Mesophyll conductance is finite and variable and plays a major role in constraining photosynthetic productivity (Niinemets ). Large differences in g m have been shown between and within species with different leaf forms and habits (Flexas ; Warren, 2008; Niinemets , 2011). Whilst rapid changes of g m in response to environmental drivers probably depend on biochemical factors such as changes in the permeability of membranes to CO2 facilitated by cooporins (Hanba ; Flexas , 2012), maximum values of g m for a given species or genotype are suggested to be related to leaf anatomical properties (Niinemets ; Tosens ). In particular, it has been shown that leaves with a more robust structure and higher M A exhibit lower photosynthetic rates due to large CO2 drawdown from substomatal cavities (C i) to chloroplasts (C c), C i-C c, demonstrating that the photosynthetic capacity is limited by g m (Flexas , Niinemets ). Therefore, understanding the structural and physiological basis of variation in g m is crucial for understanding photosynthetic controls in natural ecosystems and for breeding of plant cultivars with improved photosynthetic characteristics.

At the leaf level, two components of M A—leaf thickness and density—have been proposed to exert opposite effects on setting the maximum g m, with increases in thickness increasing g m and increases in density reducing it (Niinemets , Hassiotou ). Inside leaves, the CO2 diffusion pathway consists of two phases, an intercellular gas phase and a cellular liquid phase, the latter consisting of aqueous and lipid components(Niinemets and Reichstein, 2003b; Evans ). The gas phase pathway through intercellular air spaces is assumed to have a smaller effect on the overall diffusion limitations than the components of the liquid phase (Evans ). This was confirmed in several studies comparing CO2 diffusion in air and helox—air where helium replaces nitrogen to increase diffusivity—showing that the diffusion in the intercellular gas phase had little effect on photosynthesis (Parkhurst and Mott, 1990) The cellular phase is composed of the cell wall, plasma membrane, cytosol, and chloroplast envelopes and stroma. Among these components, the cell walls and chloroplast envelope have been suggested to limit g m most severely (Terashima ). Accordingly, several reports have shown positive correlations between g m and the surface of chloroplasts adjacent to intercellular air spaces (S c/S), which is sometimes considered as the most important anatomical trait affecting g m (Evans ; Terashima ; Tholen ). However, some estimates suggest that differences in cell wall thickness (T cw) can explain as much as 25–50% of the variability in g m (Evans ; Terashima ; Tosens ). Negative correlations between g m and T cw have been shown when comparing Australian Banksia species (Hassiotou ), rice relatives (Scafaro ), Eastern Australian species with varying anatomy (Tosens ), and Mediterranean Abies species (Peguero-Pina ). Recently, Terashima showed that g m/(S c/S) decreases with increasing T cw, i.e. the relative influence of the exposed chloroplast surface in setting the maximum g m is variable, and that this variation can potentially be explained by variations in T cw.

Few previous studies have quantitatively addressed the influence of leaf anatomical traits on the diffusion of CO2, and these studies have focused only on a few species and specific parts of the CO2 diffusion pathway (Evans ; Terashima ; Hassiotou ; Scafaro ; Peguero-Pina ; Tosens ). Hence, the whole diffusion pathway of CO2 from the substomatal cavities to the chloroplasts has not been quantitatively linked to g m in plants with widely varying leaf structures and photosynthetic capacities. Furthermore, the overall importance of g m in constraining the photosynthetic rate in species with different foliage architecture has not been characterized. To fill this gap, we aimed with the present study: (i) to analyse the role of different components of the diffusion pathway across a wide range of foliage architectures and leaf photosynthetic capacities; (ii) to associate the interspecific differences in leaf anatomy with the integrated leaf architectural traits such as M A and g m; (iii) to quantify the distribution of overall photosynthetic limitation among biochemistry, mesophyll diffusion, and stomata; and (iv) to quantify the resistance that each anatomical component exerts on the diffusion of CO2 inside the leaf.

Material and methods

Plant material

Fifteen taxa of different growth form and leaf longevity were selected for the study to obtain an extensive range of variation in leaf morphology and anatomy (Supplementary Table S1 at JXB online). Five species were annual herbs (Capsicum annuum, Helianthus annuus, Phaseolus vulgaris, Spinacea oleracea, Ocimum basilicum) and the rest were broad-leaved trees: four deciduous (Acer negundo, Alnus subcordata, Betula pubescens, Catalpa speciosa), one semi-deciduous (Quercus brantii) and five evergreens (Quercus ilex, Citrus reticulata, Ficus elastica, Pittosporum tobira, Washingtonia filifera). All species were dicots, except for the palm Washingtonia filifera.

All plants were grown either from commercial seed or from seeds collected in the field, except for F. elastica where rooted cuttings of a single mother plant were used. The plants were grown in a growth room with a 10h photoperiod, a day/night temperature of 24/18 ºC, 60% air humidity, and a constant photon flux density of 350 µmol m−2 s−1 at plant level provided by Philips HPI-T Plus 400W metal halide lamps. The daily integrated incident quantum flux density was 12.6mol m–2 d–1. The growth substrate was a 1/1 mix of quartz sand and standard potting soil (Biolan Oy, Finland) including slow-release NPK (3/1/2 ratio) fertilizer with microelements, and the plants were irrigated daily to soil field capacity. The size of the pots varied between 1 and 5 l depending on plant age and size. In all cases, fully developed young (current-season leaves in evergreens) leaves were used for the measurements. In herbs, the plants were measured 1 month after seed germination, whilst woody species were measured on the second growing year. All physiological and structural analyses were replicated with at least three independent plants per taxa.

Foliage gas exchange and fluorescence measurements

Attached leaves were used for simultaneous leaf gas-exchange and chlorophyll-fluorescence measurements using a portable gas exchange fluorescence system GFS-3000 (Walz, Effeltrich, Germany) equipped with a leaf chamber fluorometer with an 8cm2 cuvette window area. Light was provided by the LED light source of the leaf chamber fluorometer (10% blue and 90% red light) and the humidity was controlled by a built-in GFS-3000 humidifier. Use of a certain fraction of blue light is routinely used in portable photosynthesis devices to induce stomatal opening. Although blue light is absorbed more strongly by the upper leaf layers and may lead to discrepancies among photosynthesis and fluorescence profiles (Evans and Vogelmann, 2006), thereby altering g m estimations by the combined gas-exchange/fluorescence techniques (Loreto ), the amount of blue light used in our study was small and the expected effect minor.

The standard conditions for leaf stabilization in the cuvette were: leaf temperature of 25 ºC, saturating quantum flux density of 1500 µmol m–2 s–1, and CO2 concentration in the cuvette (C a) of 385 µmol CO2 mol air–1. Once the steady-state conditions were reached, typically 15–20min after clamping the leaf in the cuvette, CO2 response curves of net assimilation (A N) were measured. First, C a was lowered stepwise from 385 to 50 µmol CO2 mol air–1 and then raised again to 385 µmol CO2 mol air–1, and the leaf was kept at this C a until the original A N value was achieved. Next, C a was increased stepwise from 385 to 1500 µmol CO2 mol air–1 and returned again to 385 µmol CO2 mol air–1. In all cases, measurements of A N and steady-state fluorescence yield (F s) were recorded after the gas-exchange rates stabilized at the given C a. After recording the A N value, a flash of saturating white light was given to determine the maximum fluorescence yield in light-adapted state (F m’). After completion of the CO2 response curves, the light was switched off and respiration rate in the dark (R d) was determined. In calculations of A N, R d, and intercellular CO2 concentration (C i), corrections for the diffusion leakage of CO2 into and out of the leaf chamber were included as described by Flexas .

Measurements of leaf optical properties

Leaf transmittance and reflectance measurements were conducted with a spectrometer (AvaSpec-2048-2; Avantes, Apeldoorn, The Netherlands) using an integrating sphere (ISP-80-8-R; Ocean Optics, Dunedin, FL, USA). Leaf absorptance (α) was calculated as 1 minus the sum of reflectance and transmittance. Three leaves of each species were measured, and within each leaf, three replicate measurements were made. Average absorptance across the 400–700nm region was used to characterize the fraction of incident photosynthetically active radiation absorbed by the leaf.

Anatomical measurements

After the gas-exchange measurements, 1×1mm pieces were cut between the main veins from the same leaves for anatomical measurements. Leaf material was quickly fixed under vacuum with 4% glutaraldehyde and 2% paraformaldehyde in 0.1M phosphate buffer (pH 7.2). Afterwards, the samples were fixed in 1% osmium tetroxide for 1h and dehydrated in a graded ethanol series, followed by washing three times in propylene oxide. The dehydrated segments were embedded in Spurr’s resin (Monocomp Instrumentación, Madrid, Spain) and cured in an oven at 60 ºC for 48h. Semi-thin (0.8 µm) and ultrathin (90nm) cross-sections were cut with an ultramicrotome (Reichert & Jung model Ultracut E). Semi-thin cross-sections were stained with 1% toluidine blue and viewed under an Olympus BX60light microscope. Photos were taken at 200× and 500× magnification with a digital camera (U-TVO.5XC; Olympus) to measure the leaf thickness and thickness of the palisade and spongy tissue layers (Supplementary Fig. S1A–C). Ultrathin cross-sections for transmission electron microscopy (TEM H600; Hitachi) were contrasted with uranyl acetate and lead citrate. Photos were taken at 2000× magnification (Supplementary Fig. S1D–F) to measure the length of mesophyll cells and chloroplasts adjacent to intercellular air spaces and chloroplast width and thickness, and the volume fraction of intercellular air space calculated as:

where ΣS s is the total cross-sectional area of mesophyll cells, W is the width of the section, and t mes is the mesophyll thickness between the two epidermises. Mesophyll (S m/S) and chloroplast (S c/S) surface area exposed to intercellular air spaces per leaf area were calculated separately for spongy and palisade tissues as described by Evans and Syvertsen . The curvature correction factor was measured and calculated for each species according to Thain (1983) for palisade and spongy cells by measuring their width and height and calculating an average width/height ratio. The curvature factor correction ranged from 1.16 to 1.4 for spongy cells and from 1.4 to 1.5 for palisade cells. All parameters were analysed at least in four different fields of view and at three different sections. Weighted averages based on tissue volume fractions were calculated for S m/S and S c/S.

T cw and cytoplasm thickness (T cyt) were measured at 20 000–40 000× magnification depending on the species (Supplementary Fig. S1G–I). Three different sections per species and four to six different fields of view were measured for each anatomical characteristic. Micrographs were selected randomly in each section and T cw was measured for two to three cells per micrograph. Ten measurements for spongy tissue and ten for palisade parenchyma cells were made for each anatomical trait, and weighted averages based on tissue volume fractions were calculated. All images were analysed with Image analysis software (ImageJ; Wayne Rasband/NIH, Bethesda, MD, USA).

MA and leaf density

The leaves were scanned at 300 dpi, and then oven dried at 70 C for 48h and their dry mass was estimated. Leaf area was determined from the images with Image J. From these measurements, M A was calculated. Using the estimates of leaf thickness from anatomical measurements, leaf density (D L) was calculated as M A per unit leaf thickness (Niinemets, 1999).

Estimation of gm and model parameters Farquhar et al. (1980) by combined gas-exchange/fluorescence method

The chloroplastic hypothetical CO2 compensation point (Γ*) in the absence of R d was calculated from the Rubisco specificity factor (S C/O) as:

using the average values for S C/O reported by Galmés for each different leaf habit (Supplementary Table S2 at JXB online). A sensitivity analysis showed that the precise value of Γ* within the reported range did not significantly affect the g m estimates (Supplementary Table S3A at JXB online).

From chlorophyll fluorescence measurements, the actual photochemical efficiency of photosystem II (ФPSII) was determined from F and the maximum fluorescence yield during a light-saturating pulse of 4500 µmol m–2 s–1 (F m’) following the method of Genty :

The linear electron transport rate on the basis of chlorophyll fluorescence (J F) was then calculated as:

where Q is the photosynthetically active quantum flux density, α is the leaf absorptance, and εPSII is the fraction of light absorbed by PSII. As routinely assumed, εPSII was taken as 0.5 (Loreto ; Niinemets ).

Furthermore, the g m to CO2 was estimated according to Harley as:

where R L is the respiration rate in the light. In this study, R d was used as a proxy for R L (Gallé ). In other studies, half R d has been used (Piel ; Niinemets ). However, as shown in Supplementary Table S3B, no significant differences in g m were found when using the proxy for R L, and hence we concluded that selection of the appropriate value for R L is not a critical issue for our g m estimates, confirming a previous sensitivity analysis (Niinemets ).

The obtained values of g m were used to transform the A N-C i response curves into A N versus C c response curves as C c=C i – A N/g m. Finally, Farquhar model parameters, the maximum velocity of carboxylation (V cmax) and the capacity for photosynthetic electron transport (J max) on the basis of C c were calculated according to Bernacchi . Three replicates estimates of g m were available for every species.

Estimation of gm from gas exchange measurements only: the curve-fitting method

The curve-fitting method introduced by Ethier and Livingston (2004) as applied by Niinemets was used to obtain an alternative estimate of g m. This method is based on changes in the curvature of A N versus C i response curves due to finite g m such that the Farquhar model based on C i imperfectly fits the data (Ethier and Livingston 2004). Thus, including g m as a fitted parameter significantly improves the model fit. Estimates of J max, V cmax, and g m were derived from fitting A N-C i curves as previously described. Values of the Michaelis–Menten constant for CO2 (K c), and oxygen (K o) and their temperature responses used for these estimations were from Bernacchi . Γ* was calculated according to Eqn 2, and R d by gas exchange measurements at 385 µmol CO2 mol air–1. At least three plants per species were used to estimate g m. The same leaves were used for estimation of g m by the Ethier and Livingston (2004) and Harley methods.

gm modelled from anatomical characteristics

The one-dimensional gas diffusion model of Niinemets and Reichstein (2003a) as applied by Tosens was employed to estimate the share of different leaf anatomical characteristics in determining g m. g m as a composite conductance for within-leaf gas and liquid components is given as:

where g ias is the gas phase conductance inside the leaf from substomatal cavities to outer surface of cell walls, g liq is the conductance in liquid and lipid phases from outer surface of cell walls to chloroplasts, R is the gas constant (Pa m3 K–1 mol–1), T k is the absolute temperature (K), and H is the Henry’s law constant (Pa m3 mol–1). g m is defined as a gas-phase conductance, and thus H/(RT k), the dimensionless form of Henry’s law constant, is needed to convert g liq to corresponding gas-phase equivalent conductance (Niinemets and Reichstein, 2003a). In the model, the gas-phase conductance (and the reciprocal term, r ias) is determined by average gas-phase thickness, ΔL ias, and gas-phase porosity, f ias (fraction of leaf air space):

where is the diffusion path tortuosity (m m–1) and D a (m2 s–1) is the diffusion coefficient for CO2 in the gas phase (1.51×10–5 at 25 °C). ΔL ias was taken as half the mesophyll thickness. The partial determinants of the liquid-phase diffusion pathway (the reciprocal term r i, where i stands either for cell wall, cytosol, or stroma conductance) were calculated as:

where ΔL i (m) is the diffusion path length in the corresponding component of the diffusion pathway, p i (m3 m–3) is its effective porosity, and D w is the aqueous phase diffusion coefficient for CO2 (1.79×10–9 m2 s–1 at 25 °C). The dimensionless factor r f,i accounts for the reduction of D w compared with free diffusion in water, and was taken as 1.0 for cell walls (Rondeau-Mouro ) and 0.3 for cytosol and stroma (Niinemets and Reichstein, 2003b). In addition, r f,i values for cytosol and stroma were estimated using a least-squares iterative analysis to assess the sensitivity of g m to values of r f,i (Supplementary Figs S2 and S3 at JXB online). In this analysis, r f,i was allowed to vary between 1 and 0.05, and the values of r f,i were varied within this range to minimize the difference between measured and modelled g m. Whilst this approach improved the agreement between modelled and measured g m, the extreme values obtained for r f,i seemed unrealistic (Supplementary Figs S2 and S3). p i was set to 1.0 for cytosol and stroma. There are no direct measurements of cell wall porosity, but it has been suggested that this parameter might vary with T cw among species (Terashima ; Evans ; Tosens ). Therefore, given the heterogeneous series of species used in this experiment, p i was estimated using a least-squares iterative analysis assuming a hypothetical relationship between porosity and T cw as described by Tosens . Again, a least-squares iterative approach was employed to get the best fit between measured and modelled g m. The p i range in this analysis was fixed at 0.028 (Tosens ) for the thickest to 0.3 (Nobel, 1991) for the thinnest cell walls (Supplementary Table S5 at JXB online). We used an estimate of 0.0035 m s–1 for both plasma membrane conductance (g pl) and chloroplast envelope conductance (g env) as previously suggested (Evans ; Tosens ).

Carbonic anhydrase in cytosol and chloroplasts could facilitate the diffusion of CO2 through the liquid phase. However, there is little evidence for the involvement of carbonic anhydrase in g m and A N (reviewed by Flexas , 2012). Therefore, following Tosens , we did not include the potential effect of carbonic anhydrase in our analysis.

In previous studies, we scaled the total liquid-phase diffusion conductance by S c/S ratio (Tosens ) that determines the number of parallel diffusion pathways from outer surfaces of cell walls to chloroplasts.

Although, cell wall and plasmalemma resistances actually scale with the S m/S ratio, use of S c/S has been deemed to be appropriate, as S c/S is generally close to the S m/S ratio (Scafaro ; Peguero-Pina ), i.e. there is little cell wall area free of chloroplasts. Even if S c/S is significantly smaller than S m/S, the cytosolic distance between the neighbouring chloroplasts is generally large and this can still constrain the diffusion flux in interchloroplastial areas of cell wall (locally large cytosol conductance, g cyt; Fig. 1). However, the significance of the r cyt depends on the other parts of the diffusion pathway as well.

Fig. 1.

Illustration of the diffusion pathway in exposed cell wall areas lined with chloroplasts (path 1) and interchloroplastial areas (path 2). The diffusion pathway in leaf lipid and liquid phases includes cell wall, plasmalemma, cytosol (shown by red arrows), chloroplast envelope membranes, and chloroplast stroma (shown by dark green arrows). The effective diffusion path length in cytosol along path 1 is taken as the average distance of chloroplasts from the cell wall, ΔL cyt,1, whilst the diffusion pathway length in interplastidial areas is determined by the distance between the chloroplasts and ΔL cyt,1 (Eqn 12).

To explicitly assess the importance of diffusion of CO2 through interchloroplastial areas, we considered two different pathways of CO2 inside the cell, one for cell wall parts with chloroplasts and the other for interchloroplastial areas as described by Tholen . For exposed cell wall portions lined with chloroplasts, the partial liquid phase conductance, g cel,1, inside the cell is given as:

where r cyt,1 and r st,1 are cytosolic resistance from the plasmalemma inner surface to the outer surface of chloroplasts and the stromal resistance in the direction perpendicular to cell wall (Fig. 1), respectively, both calculated by Eqn 8. For r cyt,1, the diffusion pathway length, ΔL cyt,1, is given as the average distance between the chloroplasts and cell wall in cell wall areas lined by chloroplasts (Fig. 1), whilst for r st,1, ΔL i, was taken as half of the chloroplast thickness, ΔT chl/2. For the cell wall portions without chloroplasts, the partial conductance, g cel,2, is given analogously as:

where r cyt,2 is the cytosolic resistance from interchloroplastic cell wall portions towards the chloroplast and r st,2 is the stromal conductance in a direction parallel with the cell wall (Fig. 1). The diffusion path length for r cyt,2 (Eqn ), ΔL cyt,2, is driven both by the distance between the neighbouring chloroplasts, chloroplast thickness, and chloroplast distance from the cell wall and was approximated as:

where ΔL chl is the distance between the neighbouring chloroplasts. ΔL cyt,2 was calculated as a harmonic average, which more correctly represents the diffusion pathway of r cyt,2. Finally, the diffusion pathway length for r st,2 was taken as a quarter of the chloroplast length.

Considering further that the fraction of exposed cell wall area lined with chloroplasts is given by S c/S m and the fraction free of chloroplasts as 1 – S c/S m, the total cellular conductance (sum of parallel conductances) is given as:

Total liquid phase conductance from the outer surface of cell walls to carboxylation sites in the chloroplasts is the sum of serial conductances in the cell wall, plasmalemma, and inside the cell:

Alternatively, the total cellular diffusion pathway can be considered to consist of two parallel pathways from the outer surface of the cell walls to the chloroplasts, one pathway corresponding to the diffusion flux through cell wall areas lines with chloroplasts and the other without chloroplasts:

Although Eqns 14 and 15 are conceptually different, the values of conductances calculated by both equations were very similar, differing at most by 4%. In the current study, we have used Eqn 14.

Analysis of quantitative limitations on AN

To separate the relative controls on A N resulting from limited stomatal conductance (l s), mesophyll diffusion (l m), and limited biochemical capacity (l b) (l s+l m+l b=1), we used the quantitative limitation analysis of Jones (1985) and implemented by Grassi and Magnani, (2005). The limitations of the different components were calculated as:

where g s is the stomatal conductance to CO2, g m was according to Harley , Eqn 5), and g tot is the total conductance to CO2 from ambient air to chloroplasts (sum of the inverse serial conductances g s and g m). δA N/δC c was calculated as the slope of A N-C c response curves over a C c range of 50–100 µmol mol–1. At least three curves per species were used, and average estimates of the limitations were calculated.

Quantitative analysis of partial limitations of gm

The determinants of g m were divided between the component parts of the diffusion pathway (Eqns 6–8). The proportion of g m determined by limited gas-phase conductance (l ias) was calculated as:

The share of g m by different components of the cellular phase conductances (l i) was determined as:

where l i is the component limitation in the cell walls, the plasmalemma, and inside the cells, and g i refers to the component diffusion conductances of the corresponding diffusion pathways. To determine the limitations derived from the different components inside the cell (cytoplasm, chloroplast envelope, and stroma), weighted limitations of both pathways, the fraction of exposed cell wall area lined with chloroplasts and the fraction free of chloroplasts, were used.

Statistical analyses

Regression and correlation analyses were conducted using the Sigma Plot 10.0 software package (SPSS; Chicago, IL, USA). Univariate analysis of variance was performed to reveal differences between species in the studied characteristics. Differences between means were revealed by Tukey analyses (P <0.05). These analyses were performed with the IBM SPSS statistics 19.0 software package (SPSS).

Results

Leaf structural and anatomical traits

M A varied sixfold (20–123g m–2) (Supplementary Table S4 at JXB online). The variation in leaf thickness was 3.7-fold with Acer negundo having the thinnest (123 µm) and F. elastica the thickest (459 µm) leaves. Spongy mesophyll thickness varied 5.2-fold, and palisade mesophyll thickness 2.5-fold (Supplementary Table S4). Generally, the palisade tissue comprised approximately 40%, and spongy tissue approximately 60% of total mesophyll, except for some species as F. elastica with 75% and W. filifera with 100% of spongy tissue. The variation in D L was 6.4-fold with Phaseolus vulgaris having the least dense (0.11g cm–3) and Q. ilex the most dense (0.70g cm–3) leaves. M A exhibited a significant positive correlation with D L (Supplementary Fig. S4 at JXB online), but was weakly correlated with leaf thickness (r 2=0.27, P <0.05; data not shown). Therefore, the variation in M A was mainly attributed to the leaf density.

Among the leaf ultrastructural characteristics estimated from transmission electron micrographs (Supplementary Tables S4 and S5, and Supplementary Fig. S1D–I), S m/S varied 3.3-fold across all species (14.4–40 m2 m–2) and S c/S varied 2.7-fold (6–19.7 m2 m–2). S c/S m varied between 0.31 (Citrus reticulata) and 0.74 (O. basilicum). For T cw (Supplementary Fig. S1G–I), 4.8-fold variation was observed between all species (113.6–543.7nm). Herbaceous species exhibited the thinnest cell walls together with Catalpa speciosa, whilst evergreens had the thickest cell walls with the maximum value of 543.7nm observed in Pittosporum tobira.

Estimation of gm with different methods

The values of g m calculated according to the methods of Harley and Ethier and Livingston (2004) were strongly correlated (Supplementary Fig. S5 at JXB online, r 2=0.80). However, the Harley et al.-based estimates exhibited the smallest average coefficient of variation for independent estimates within a species and therefore we report the data obtained with this method only.

Mesophyll conductance calculated by the method of Harley varied 24-fold across all species. H. annuus showed the maximum values and Citrus reticulata the minimum values of g m. The minimum value for the coefficient of variation in g m was 1.9% (Pittosporum tobira), whilst the maximum value was 32.9% (Q. ilex). The average of the coefficient of variation for all species was 16.5%.

gm in relation to physiological characteristics

Net assimilation rate correlated positively with g s and g m (Supplementary Fig. S6 at JXB online). C i-C c ranged from 240 to 112 µmol mol–1 in woody deciduous and evergreen species, and had lower values (40–67 µmol mol–1) in herbs. C i-C c decreased with increasing g m (Supplementary Fig. S7 at JXB online). This relationship was qualitatively identical when g m was expressed on the leaf area or dry mass basis (data not shown).

gm in relation to leaf structural and anatomical traits

g m per dry mass was negatively associated with M A (r 2 = 0.85, P <0.0005; data not shown). g m per unit leaf area or per unit dry mass (data not shown) was not correlated with S m/S, reflecting the circumstance that S m/S was almost invariable, between 16 and 24 m2 m–2 across the species. S c/S was not significantly correlated with g m (Fig. 2A, P >0.13). However, a positive correlation between g m and S c/S was observed when the species with the largest T cw (Pittosporum tobira and Q. brantii) were not included in the correlation (r 2=0.77, P <0.0001; data not shown).

Fig. 2.

Correlations of mesophyll diffusion conductance (g m) determined according to Harley with the surface area of chloroplasts exposed to intercellular airspaces per unit leaf area (S c /S) (A), mesophyll diffusion conductance with chloroplast surface area per leaf density ((S c/S)/D L) (B), and mesophyll diffusion conductance per S c/S (g m /(S c/S)) with T cw (C). In the main panels, the data were fitted by linear (B) and non-linear (C) regressions in the form y=ae –. In the inset, the data were fitted by linear regression. Different species are represented as: herbs (circles), woody deciduous and semi-deciduous species (triangles), and woody evergreen species (squares). Values are means ±standard error (SE) of three to four replicates per species.

The positive and significant correlation (r 2=0.84, P <0.001) between g m and (S c/S)/D L suggested the importance of the anatomical components to the internal diffusion of CO2 (Fig. 2B). Moreover, the negative and significant relationship observed between g m/(S c/S) and T cw showed the importance of T cw in affecting g m (Fig. 2C).

gm calculated from anatomical variables

Using the leaf anatomical traits measured, g m was modelled and compared with g m measured by the method of Harley . A good positive linear relationship between modelled and measured g m was observed (r 2=0.90, P <0.0001; Fig. 3). However, the slope was different from unity, so that the g m modelled tended to be overestimated in species with high M A and underestimated in species with low M A. g m values calculated by the model based on leaf anatomy ranged between 0.217 and 0.056mol m–2 s–1. H. annuus showed the largest and W. filifera the smallest values of g m. The coefficient of intraspecific variation in g m estimates for different replicates was lower than for the experimental estimations, being between 1.2% (Betula pubescens) and 22% (Pittosporum tobira).

Fig. 3.

The relationship between mesophyll diffusion conductance (g m) measured with Harley et al. method and g m modelled with anatomical parameters (Eqn 6–15). Values are means ±SE of three replicates per species. Symbols are the same as in Fig. 2. The data were fitted by linear regression. Broken lines correspond to the 1:1 relationship.

Overall importance of gm

According to quantitative limitations analysis of A N, stomatal openness and g m restricted the photosynthetic capacity to a similar percentage, 19–65% and 13–64%, respectively. However, the biochemical limitations were lower than the stomatal and mesophyll limitations, being between 6 and 33% (Fig. 4A–C). Both the stomatal and biochemical components tended to be more important in species with non-sclerophytic leaves (low M A), whilst mesophyll diffusion limitation was most significant in species with high M A (Fig. 4). Thus, herbaceous plants showed the maximum values for stomatal limitations, whilst the maximum mesophyll limitations were observed in evergreen species with more robust foliage structure.

Fig. 4.

Quantitative limitation analysis of photosynthetic CO2 assimilation and mesophyll conductance to CO2 (g m) in relation to the leaf dry mass per area (M A). The stomatal (A), mesophyll (B), and biochemical (C) limitations of photosynthetic assimilation were calculated according to Eqns 16–18. Limitations analyses were based on a chloroplastic CO2 concentration (C c) range of 50–100 µmol mol–1. Quantitative limitations of g m due to different anatomical components of the diffusion pathway were calculated using leaf anatomical characteristics (Eqns 19 and 20). The relative CO2 diffusion limitations separated were: intercellular spaces (D), cell wall thickness (E), cytoplasm (F), plasmalemma (G), chloroplast envelope (H), and chloroplast stroma (I).

Limitation of gm due to individual components of the diffusion pathway

From the different components of the whole diffusion pathway of CO2, the percentage limitations of g m were estimated (Fig. 4D–I). Intercellular air spaces represented a smaller resistance to the CO2 diffusion (4–22%) than the cellular phase, because the rate of CO2 diffusion in air was larger than in water. In the cellular phase, the cell walls appeared to be the most important factor that limited the internal diffusion of CO2 in the species that presented a high M A. However, the plants with low M A that presented a low percentage of limitation of g m by the cell wall revealed a higher limitation by the stroma of around 43%. On the other hand, the plasmalemma and chloroplast envelope accounted for only up to 8% of the limitation.

Discussion

Values of gm in a range of species exhibiting different foliage morphologies

The range of g m values observed in our study is representative of the whole range of g m values described so far in large literature-based datasets, except that the maximum g m values found in the present study were somewhat lower than reported previously (Flexas ; Warren, 2008, Niinemets ). These relatively low maximum values were possibly due to moderate growth light intensity compared with full sun (Piel ; Niinemets ). This explanation is consistent with the observations of a significant number of chloroplasts not closely facing the cell walls (Fig. 2A) and relatively low ratios of chloroplast exposed to mesophyll exposed cell wall surfaces (S c/S m) (Supplementary Table S4), both being traits that depend on the growth light environment (Terashima ).

Relationship of gm to leaf anatomy and its importance in limiting photosynthesis

As in previous studies, g m showed a high degree of correlation with several leaf anatomic characteristics, notably a negative correlation with M A (Flexas ; Niinemets ,b) and a positive correlation with S c/S (Evans , 2009). The M A effect on g m supports the idea that g m depends on species differences in leaf density, as density was positively correlated with M A, whilst leaf thickness showed a weak correlation. The photosynthetic capacity was also significantly and positively correlated with g m as demonstrated previously (reviewed by Flexas ; Niinemets ).Overall, these results suggest that, in species with high M A, photosynthesis is more limited by g m, as indirectly supported by the negative effect of leaf density on g m (Niinemets, 1999) and more directly evidenced by the fact that they present higher values of C i-C c (Warren, 2008; Niinemets ).

The relative contribution of g s, g m, and photosynthetic biochemistry to total photosynthesis limitation (following Grassi and Magnani, 2005) was variable and depended on leaf structural characteristics, i.e. M A (Fig.4A–C). At a typical operating CO2 concentration, the biochemical limitations of photosynthesis decreased from a maximum of approximately 33% at low M A to minimum values as M A increased, whilst, in parallel, mesophyll diffusion limitations increased from a minimum of approximately 15% to maximum values up to 65%. Stomatal limitations showed a less clear variation with M A. Overall, these data demonstrated that species with low M A showed a notable coordination of the limiting factors for photosynthesis, i.e. they were similarly co-limited by stomatal, mesophyll, and biochemical limitations. In contrast, species with high M A were mostly limited by mesophyll (on average by 57%) and stomatal (30%) diffusion, and were less limited by biochemistry (13%). This is consistent with the idea that species with thicker and denser leaves, e.g. evergreen trees, are more limited by g m than species with thinner leaves (Galmés ; Niinemets ).

Key structural factors regulating differences in gm between distant leaf structures

The fact that g m and mesophyll diffusion limitations were strongly correlated with M A suggested that interspecific variations in g m are driven by leaf structural characteristics. Among the key structural traits suggested to limit CO2 diffusion the most are the traits that alter effective diffusion path length and area for diffusion, in particular T cw, and chloroplast distribution along the exposed mesophyll cell wall (Fig. 2; Evans ), although the role of other variables, such as leaf porosity, and the path lengths for CO2 through the plasmalemma and chloroplast envelope membranes, cytosol, and stroma cannot be ruled out (Evans ). In the present study, we modelled g m considering all major leaf structural traits as described by Tosens . A high significant positive correlation between measured and modelled g m estimates was found (Fig. 3). This correlation supports the view that at least a significant proportion of the interspecific variations in g m is somehow related to differences in the thickness of the structures involved in CO2 diffusion, as well as to the number of parallel CO2 diffusion pathways determined by S c/S.

Despite the high correlation, the slope of the relationship was not unity, so that the biggest discrepancies between measured and modelled estimates of g m were found at the higher and lower ends of g m. A similar discrepancy was observed in different Australian sclerophyll species occurring in the field under different soil nutrients and water availabilities, especially at high values of g m (Tosens ).

This strong discrepancy between measured and modelled values may arise from the inherent uncertainties associated with both estimates. As for the Harley approach, besides the small variability in the estimates associated with uncertainties in the exact values of R L and Γ* (Supplementary Table S3), it has recently been shown that g m cannot be considered as a purely diffusional component, but instead intrinsically includes a flux-weighted quantity related to the amount of respiratory and photorespiratory CO2 from the mitochondria diffusing towards the chloroplasts (Tholen ). Concerning the anatomically based model used here, the precise outputs largely depend on a number of variables assumed as constants or inferred indirectly. For instance, the reduction in D w compared with free diffusion in water (r f,i) was considered constant for all species, although different for cell wall and intercellular components. Both g pl and g env were also taken as constant, whilst cell wall porosity (p i) was indirectly estimated from T cw using an empirical equation. There is not sufficient knowledge for the actual values of all these parameters, and they may vary among species, hence contributing to most of the observed slope discrepancy. It can be seen, for instance, that the difference between measured and modelled g m almost disappeared when the r f,i values were calculated using a least-squares iterative analysis (Supplementary Fig. S2). However, this ‘perfect correspondence’ is bound to some probably non-realistic r f,i values as low as 0.05. Moreover, values of g m have been modelled considering CO2 diffusivities in the different media involved—assumed to be either ‘pure’ air, lipid, or aqueous phases with fixed thicknesses, whilst, in most cases, determination of the thickness of the given phase is not that straightforward. Also, we assumed no facilitation mechanism that could improve the diffusivities in lipid and aqueous phases. Among these, membrane-bound aquaporins (Uehlein , 2008; Hanba ; Flexas ) and cytosol and stromal forms of carbonic anhydrases (Price ; Gillon and Yakir, 2000) are likely candidates (Terashima ). For instance, allowing g pl and/or g env to vary within the range of published values (Evans ) also results in a better agreement between the measured and modelled values (Supplementary Fig. S8 at JXB online). In summary, current uncertainties about the actual values of these parameters and their variability among species preclude the development of a truly predictive anatomically based model for g m. However, the good correlation, despite the divergent slope, can be taken as strong evidence that a substantial part of g m is indeed dependent on a series of leaf anatomical features.

Despite the discussed limitations of the model approach used here, the results suggest that chloroplast distribution and T cw are the most influential leaf structural characteristics in setting the limits for g m (Evans ; Terashima ). In particular, a significant positive correlation was found between g m and S c/S only when species with very large T cw (Q. brantii and Pittosporum tobira) were excluded, highlighting the fact that the impact of chloroplast distribution on g m became less important as T cw increased, in agreement with past suggestions (Terashima , 2011).

In addition, a highly significant negative relationship was observed between the ratio g m/(S c/S) and T cw considering all species, similar to that obtained by Terashima pooling literature data. Using a limitation analysis to separate the contributions of the components of g m (Eqns 13 and 14) revealed that, globally, the limitation imposed by T cw spanned the most, ranging from approximately 4 to 70% (Fig. 4E). This was followed by chloroplast stroma, which ranged from 4 to 46% (Fig. 4F, I). However, the limitations inside the cell (cytosol and stroma) could be underestimated, especially in species with high M A, as g m was modelled assuming that cytosolic and stromal viscosity (r f,i) was constant in all species.

The limitations imposed by intercellular air spaces, the plasmalemma, and the chloroplast envelope were much smaller than the rest of the diffusion pathway components as was observed by Tosens ,b). The fact that the latter two components had only a moderate effect on limiting g m is in conflict with the observed larger g m changes observed in aquaporin mutant plants without any appreciable differences in S c/S or any other leaf structural characteristic (Flexas ). This could be due to the fact that the assumed values for g pl and g env are constant among species, which may not necessarily be the case. Differences of up to four orders of magnitude have been reported for CO2 permeabilities of biological membranes. For instance, if the permeability for a given species was 0.00002 m s–1, as found for chloroplast envelopes by Uehlein , instead of the 0.0035 m s–1 used in the present simulation, the combined limitation to g m imposed could be larger than 40% (data not shown). At the other extreme, if values were closer to the 0.016 m s–1 reported by Missner for lipid bilayers, the maximum modelled g m values will be closer to estimates based by the Harley approach (data not shown). Clearly, improved knowledge on the actual permeability to CO2 of biological membranes is required to fully understand the basis for the regulation of g m.

Despite these general tendencies, the impact of each specific leaf component on g m changed with M A. Specifically, the limitations imposed by T cw strongly increased with increasing M A, whilst the limitations associated with all the other components decreased with increasing M A. Thus, in species with low M A, like annual herbs, about 60% of the total limitation to g m is imposed by cytoplasm and stroma, whilst another 12% is accounted for by the plasmalemma and chloroplast envelope. Moreover, in species with thinner leaves, the fraction of exposed cell wall lined with chloroplasts (g cel,1) was higher, whilst limitations inside the cell through interchoroplastial areas (g cel,2) were more important in species with higher M A (Fig. S9 at JXB online). This suggests that it is in such species where facilitating mechanisms (aquaporins, carbonic anhydrases, chloroplast movements, and others) have the strongest influence on g m. In contrast, in species with high M A, like evergreen sclerophylls, g m is mostly (up to 70%) limited by T cw, which is likely to be less variable in the short term, and may explain the low photosynthetic capacity displayed by these plants even under non-limiting conditions. Possible interspecific variation in the role of aquaporins in limiting g m is clearly a topic that deserves high priority in future studies.

In conclusion, the present study showed that mesophyll limitations are crucial in determining the maximum photosynthetic capacity when a large range of leaf types are analysed collectively. These limitations are variable depending on the leaf structural properties, i.e. M A and associated structural traits such as leaf density. The variability in mesophyll diffusion limitations was explained mainly by variations in the rate of CO2 diffusion pathways through cell walls, as well as the area for diffusion determined by the chloroplast distribution. However, the impact of each component of the diffusion pathway largely depended on M A, so that CO2 diffusion in species with thin leaves (e.g. herbs) depends more on membranes and aqueous compartments—and is probably more influenced by aquaporins and carbonic anhydrases. In contrast, diffusion in species with thick leaves is almost fully determined by cell wall conductance. Altogether, the variability in g m with M A helps explain the worldwide leaf economics spectrum showing a negative dependency between photosynthetic capacity and M A.

Supplementary data

Supplementary data can be found at JXB online.

Supplementary Table S1. List of studied species, species origin, life form, and leaf longevity.

Supplementary Table S2. Physiological characteristics measured in all studied species.

Supplementary Table S3. Sensitivity analysis of the influence of uncertainties in chloroplastic hypothetical CO2 compensation point (Γ*) and day respiration on the estimation of mesophyll conductance (g m).

Supplementary Table S4. Leaf dry mass per unit area (M A), leaf thickness (T L), leaf density (D L), thickness of mesophyll layers, number of palisade cell layers, mesophyll surface area exposed to intercellular airspace (S m/S), chloroplast surface area exposed to intercellular airspace (S c/S), and the ratio S c/S m in all studied species.

Supplementary Table S5. Cell wall thickness (T cw), cytoplasm thickness (T cyt), chloroplasts length (L chl), chloroplasts thickness (T chl), and effective porosity of the cell wall (p i).

Supplementary Fig. S1. Representative light micrographs at 200× magnification for Phaseolus vulgaris, Ficus elastica, and Washingtonia filifera, and representative transmission electron micrographs at 2000× magnification for Helianthus annuus, Acer negundo, and Washingtonia filifera and at 20 000× magnification for H. annuus, Alnus subcordata and Pittosporum tobira.

Supplementary Fig. S2. The relationship between mesophyll diffusion conductance (g m) measured with the Harley Harley method and g m modelled with anatomical parameters using least-squares iterative analysis for the r f,i parameter.

Supplementary Fig. S3. Effects of the parameter r f,i of the cytosol and chloroplast stroma on g m modelled from anatomical characteristics.

Supplementary Fig. S4. Correlation between leaf density (D L) and leaf dry mass per unit area (M A).

Supplementary Fig. S5. Relationship between g m measured according to Harley method versus the Ethier and Livingston (2004) method.

Supplementary Fig. S6. Net photosynthesis rate (A N) in relation to stomatal (g s) and mesophyll (g m) conductance.

Supplementary Fig. S7. The relationship between g m and CO2 drawdown (C i -C c).

Supplementary Fig. S8. The relationship between mesophyll diffusion conductance (g m) measured with the Harley method and g m modelled with anatomical parameters using different values for the membrane permeabilities of plasmalemma (g pl) and chloroplast membrane (g env) conductances.

Supplementary Fig. S9. Quantitative limitation analysis of conductance to CO2 inside the cell (g cel,tot) calculated on the basis of leaf anatomical characteristics.