Mesophyll conductance response to short‐term changes in p CO2 is related to leaf anatomy and biochemistry in diverse C4 grasses

Summary Mesophyll CO2 conductance (g m) in C3 species responds to short‐term (minutes) changes in environment potentially due to changes in leaf anatomical and biochemical properties and measurement artefacts. Compared with C3 species, there is less information on g m responses to short‐term changes in environmental conditions such as partial pressure of CO2 (pCO2) across diverse C4 species and the potential determinants of these responses. Using 16 C4 grasses we investigated the response of g m to short‐term changes in pCO2 and its relationship with leaf anatomy and biochemistry. In general, g m increased as pCO2 decreased (statistically significant increase in 12 species), with percentage increases in g m ranging from +13% to +250%. Greater increase in g m at low pCO2 was observed in species exhibiting relatively thinner mesophyll cell walls along with greater mesophyll surface area exposed to intercellular air spaces, leaf N, photosynthetic capacity and activities of phosphoenolpyruvate carboxylase and Rubisco. Species with greater CO2 responses of g m were also able to maintain their leaf water‐use efficiencies (TEi) under low CO2. Our study advances understanding of CO2 response of g m in diverse C4 species, identifies the key leaf traits related to this response and has implications for improving C4 photosynthetic models and TEi through modification of g m.


Introduction
Mesophyll CO 2 conductance (g m ) describes the movement of CO 2 from substomatal cavities across the intercellular air space, cell walls and membranes to the site of first carboxylation. This carboxylation occurs in the mesophyll chloroplast in species with the C 3 photosynthetic pathway and mesophyll cytosol in species with the C 4 photosynthetic pathway (Evans & von Caemmerer, 1996). g m varies across plant groups due to variation in leaf anatomy and biochemistry, changes dynamically in response to environmental stimuli and has a significant impact on the plant and ecosystem-level photosynthetic CO 2 uptake and water-use efficiency Flexas et al., 2014;Knauer et al., 2019b;Pathare et al., 2020a). Despite its importance for photosynthetic CO 2 uptake and water-use at both plant and ecosystem levels and its variation across diverse plants groups, g m is only beginning to be explicitly implemented into global models that upscale leaf-scale photosynthetic processes to canopy and global scales (Rogers et al., 2017;Knauer et al., 2019a,b;von Caemmerer, 2021). The implementation of g m is complicated because there is a lack of detailed information on responses of g m to short and long-term changes in environmental conditions (e.g. light, precipitation, temperature and CO 2 concentration) across diverse plant groups (Rogers et al., 2017;Knauer et al., 2019a,b). Most investigations on the responses of g m to changes in environmental conditions have focussed on C 3 species (von Caemmerer & Evans, 2015;Xiong et al., 2015;Carriqui et al., 2018;Shrestha et al., 2018), but there is less information on the response of g m to changing environmental conditions in diverse C 4 species (Ubierna et al., 2017(Ubierna et al., , 2018Kolbe & Cousins, 2018;Sonawane et al., 2021). A better understanding of how C 4 -g m responds to changing environmental conditions is essential for improving the models of C 4 photosynthesis at both the leaf and global scales (Rogers et al., 2017;Knauer et al., 2019a,b;von Caemmerer, 2021) as well as for potentially increasing water-use efficiency of C 4 crops through manipulation of g m Pathare et al., 2020a).
The influence of g m on C 3 photosynthesis has been well studied for the past several years and there has been a significant understanding of C 3 -g m and its responses to changes in longterm growth conditions and short-term measurements conditions. In general, g m varies greatly among diverse C 3 species and limits C 3 photosynthesis as much as stomatal conductance (g sw ) (Flexas et al., 2014;Muir et al., 2014;Barbour & Kaiser, 2016;Veromann-Jürgenson et al., 2017). Variation in C 3 -g m is influenced by leaf ontogenic and anatomical traits such as leaf development and senescence (Grassi & Magnani, 2005;, surface area of chloroplasts appressed to intercellular air space (S c ) (Tosens et al., 2012;Peguero-Pina et al., 2016), mesophyll cell wall thickness (T CW ) (Veromann-Jürgenson et al., 2017;Ellsworth et al., 2018;Evans, 2021) and leaf thickness (Flexas et al., 2008;Muir et al., 2014). In terms of responses to long-term growth conditions, C 3 -g m is influenced by water stress, elevated CO 2 , salinity, nutrient supplement and growth latitude (Flexas et al., 2008;Momayyezi & Guy, 2017;Mizokami et al., 2018;Shrestha et al., 2018). Additionally, rapid responses of g m (within minutes) have been observed in response to short-term changes in leaf temperature, quantity and quality of light, relative humidity and CO 2 concentrations (Hassiotou et al., 2009;von Caemmerer & Evans, 2015;Xiong et al., 2015), although not always (Loreto et al., 1992;Tazoe et al., 2009). However, there is no consensus on the exact cause of rapid responses of g m to changes in environmental conditions such as CO 2 .
Some studies have suggested that rapid changes in g m in C 3 plants can be explained by changes in chloroplast position and movement that could lead to short-term change in S c (Oguchi et al., 2005;Tholen et al., 2008;Terashima et al., 2011). However, mesophyll cell wall thickness (T CW ) and composition are considered invariable in the short term (minutes) and may not explain the rapid responses of g m Terashima et al., 2011;Carriqui et al., 2018). Alternatively, the rapid responses of C 3 -g m have also been attributed to changes in activities of key photosynthetic enzymes such as carbonic anhydrase (CA), which catalyses the conversion of CO 2 to HCO 3 − Momayyezi & Guy, 2017) and the facilitation effect of CO 2 -permeable aquaporins (Uehlein et al., 2008;Flexas et al., 2012;Kaldenhoff, 2012;Groszmann et al., 2017; but please refer to Kromdijk et al., 2019;Huang et al., 2021). Still, other groups have suggested that the rapid responses of C 3 -g m could be the result of systematic methodological errors or the use of oversimplified models. These include mathematical dependency of g m on other variables (such as A net and C i ) used to calculate it in fluorescence and Δ 13 C methods, neglecting the contribution of respiratory and photorespiratory CO 2 release to the total CO 2 pool in the leaf and inaccurate measurement of day respiration or estimates of the Rubisco fractionation factor in the Δ 13 C method (Tholen & Zhu, 2011;Gu & Sun, 2014;Carriqui et al., 2018;Ubierna et al., 2019). The exact mechanism underlying the CO 2 responses of C 3 -g m therefore remains controversial. However, considerable evidence based on diverse species and methods of estimating g m have suggested that the C 3 -g m values increase at low partial pressures of CO 2 (pCO 2 ).
There has been a recent increase in research on the short-term and long-term variability of g m in diverse C 4 species and in the leaf traits that could explain this variability Cano et al., 2019;Pathare et al., 2020a,b). Our recent study demonstrated that, as for C 3 species, g m varied significantly among diverse C 4 grasses and had significant effects on photosynthetic rates and leaf water-use efficiencies (Pathare et al., 2020a). This variation in C 4 -g m was correlated with leaf-level traits such as leaf thickness, stomatal ratio (SR), adaxial stomatal densities and S mes . We also demonstrated that C 4 grasses adapted to low precipitation habitats exhibited traits related to greater g m but lower leaf hydraulic conductance compared with grasses from habitats with relatively high precipitation (Pathare et al., 2020b). These studies have advanced our understanding of the variability of g m in C 4 grasses, as well as how g m is influenced by leaf-level traits and is affected by long-term growth conditions such as precipitation. However, there is still a limited understanding of how g m in diverse C 4 grasses responds to short-term changes in environmental conditions such as CO 2 . The few previous studies have largely focussed on a few species such as sorghum, maize and Setaria Kolbe & Cousins, 2018;Ubierna et al., 2018;. To the best of our knowledge, no studies to date have explored if g m responses to short-term changes in pCO 2 vary among diverse C 4 grasses and what potential anatomical and biochemical traits could explain this variation in CO 2 response of C 4 -g m .
The overall objectives of the current study were (1) to investigate the response of g m to short-term changes in pCO 2 in diverse C 4 species, (2) identify the anatomical and biochemical traits that may explain the variable CO 2 response of C 4 -g m , and (3) evaluate the impact of varying CO 2 responses of g m on changes in photosynthesis and leaf water-use efficiency. To address the above objectives, we estimated g m under changing pCO 2 in 16 diverse C 4 grasses (please refer to Pathare et al., 2020a for details), using the most recent method described by Ogee et al. (2018). We also investigated the relationship of leaf anatomical traits, previously known to influence C 4 -g m , with variable CO 2 responses of g m in the 16 C 4 grasses. Furthermore, we investigated the impacts of photosynthetic capacity on the CO 2 response of g m in these C 4 grasses through measurements of maximum photosynthetic rates (A max ), leaf nitrogen content (N area ), activities of key enzymes of C 4 photosynthetic pathway such as CA, phosphoenolpyruvate carboxylase (PEPC) and ribulose-1,5-bisphosphate carboxylase/ oxygenase (Rubisco) and PEPC affinity for its substrate HCO 3 − (K m ). N area is integral to the proteins of photosynthetic machinery (such as PEPC, Rubisco and CA) that, along with the leaf structure, are responsible for the drawdown of CO 2 inside the leaf (Parkhurst, 1994;Wright et al., 2004;Evans et al., 2009;DiMario et al., 2018).

Plant growth
Sixteen C 4 grasses (Table 1) were selected for this study. In the graphs, each species is identified by four letter word that combines the first letter of the genus and first three letters of the species name. The plants were raised from seeds and grown in 3-l free drainage pots in a controlled environment growth chamber (model GC-16; Enconair Ecological Chambers Inc., Winnipeg, MB, Canada). The photoperiod was 14 h including a 2 h ramp at the beginning and end of the light period. Light and dark temperatures were maintained at 26 and 22°C, respectively. Light was provided by 400-W metal halide and high-pressure sodium lamps with maximum photosynthetic photon flux density (PPFD) of c. 1000 μmol photons m −2 s −1 at plant height. One individual per species was grown per pot in a Sunshine mix LC-1 soil (Sun Gro Horticulture, Agawam, MA, USA) with five or six replicate pots per species. The plants were irrigated daily to pot saturation and fertilised twice a week with Peters 20-20-20 (2.5 g l −1 ). Pots were randomised daily within the growth chamber.
Gas exchange measurements and estimation of g m The measurements of net photosynthetic rates (A net , μmol CO 2 m −2 s −1 ), stomatal conductance to water vapour (g sw , mol water m −2 s −1 ), intercellular CO 2 concentrations (C i , Pa) and mesophyll conductance to CO 2 (g m , μmol m −2 s −1 Pa −1 ) were performed at four different CO 2 levels inside the chamber or C a (34, 27, 20 and 14 Pa) using an LI-6400XT infrared gas analyser (Li-Cor, Lincoln, NE, USA). Intrinsic water-use efficiency (TE i , μmol CO 2 mol −1 water) was calculated as A net /g sw for each CO 2 level. Maximum photosynthetic rates (A max , μmol CO 2 m −2 s −1 ) were measured at saturating light of c. 1200 μmol photons m −2 s −1 and pCO 2 of c. 1500 μmol mol −1 . Several methods have been used to estimate C 4 -g m . Pfeffer & Peisker (1998) calculated the g m from the initial slope of a photosynthetic CO 2 response curve and assumed no CO 2 dependence of g m . However, C 4 -g m is sensitive to pCO 2 (Kolbe & Cousins, 2018;Ubierna et al., 2018) and therefore the initial slope method may be problematic. Using anatomical traits such as S mes for estimating C 4 -g m could also be subject to errors due to assumptions made for values of T CW , porosity and membrane permeability (Pengelly et al., 2010). The Δ 13 C in vitro V pmax method (Ubierna et al., 2017) estimates C 4 -g m by retrofitting models of C 4 photosynthesis model (von Caemmerer, 2000) and the Δ 13 C (Farquhar & Cernusak, 2012) with gas exchange, kinetic constants and in vitro PEPC activities. The Δ 13 C in vitro V pmax method may also result in inaccurate estimates of C 4 -g m due to errors associated with the estimation of in vitro PEPC and CA activities and enzyme kinetic parameters. The Δ 18 O method (Gillon & Yakir, 2000;Barbour & Kaiser, 2016) utilises simultaneous measurements of the oxygen isotope composition (δ 18 O) of transpired H 2 O and oxygen isotope discrimination (Δ 18 O) in CO 2 to calculate the CO 2 concentration at the site of isotope exchange by CA. The Δ 18 O method assumes full isotopic equilibrium between CO 2 and H 2 O at the site of CA (θ = 1), which may not always be true and therefore g m could be underestimated. In the current study we used the method described by Ogee et al. (2018) that builds upon the Δ 18 O method Ogee et al., 2018). This method utilises a newly developed model of C 4 photosynthetic discrimination that provides an estimate of the isotopic equilibrium between CO 2 and H 2 O inside the leaf and g m and accounts for the physical separation between mesophyll and bundle sheath cells in C 4 leaves and the contribution of respiratory fluxes (Ogee et al., 2018). For estimating g m , isotopologs of CO 2 and H 2 O were measured using a LI-6400XT infrared gas analyser (Li-Cor) coupled to a tunable diode laser absorption spectroscope (TDLAS, model TGA 200A; Campbell Scientific, Logan, UT, USA) and a cavity-ring down absorption spectroscope (Picarro, Sunnyvale, CA, USA), as described previously (Ubierna et al., 2017;Kolbe & Cousins, 2018;Pathare et al., 2020a). The entire LI6400XT, the 2 cm × 6 cm leaf chamber (6400-11; Li-Cor) and LI-6400-18-RGB light source were placed in a growth cabinet (model EF7, Conviron; Controlled Environments Inc., North Branch, MN, USA) with fluorescence lamps (F48T12/CW/VHO; Sylvania, Wilmington, MA, USA) set at a PPFD of c. 250 μmol photons m −2 s −1 and air temperature was maintained at 25°C. For each species, four-point CO 2 response curve of g m , g sw , A net and TE i was performed. During measurements CO 2 sample (CO 2 S) was set to c. 34, 27, 20 and 14 Pa. At each CO 2 level, the leaves were allowed to adjust for at least 30 min or until stable values of A net and g sw were achieved. Data for isotopologs of CO 2 and H 2 O and physiological parameters (A net , g sw , C i , TE i ) were collected and averaged over the next 20-30 min for each CO 2 level (c. 10-15 cycles of TDLAS) with the Li6400XT set to log data only when the TDLAS analysed the sample line. At each CO 2 level, three biological replicates were measured for the species eari and five biological replicates were measured for species' pvir and eser. For the remaining 13 species, measurements were conducted on four biological replicates. After the completion of above measurements, lights were tuned off and the leaves were allowed to stabilise for 15 min before logging the rates of dark respiration (R n , μmol CO 2 m −2 s −1 ). Mesophyll conductance was estimated for each species at each of the four pCO 2 levels using the Ogee et al. (2018) method. Key input parameters used in calculation of isotope parameters and estimation of g m are given in Supporting Information Table  S1. Further details of equations and calculations of fractionation factors can be found in Ogee et al. (2018); Ubierna et al. (2017). While estimating g m , we assumed that ϕr or the fraction of respired CO 2 not produced in the bundle sheath cells of C 4 plants = 0.5 (von Caemmerer, 2000) and analysed the impacts of changing ϕr values (0-1) on the calculation of g m (Fig. S1). We also performed a sensitivity analysis of g m to changes in leaf temperature (ranging from 23 to 28°C) (Fig.  S2). For C 4 plants, photorespiration was assumed to be negligible and therefore not accounted for while estimating g m . While estimating g m at different pCO 2 (34, 27, 20 and 14 Pa), it was assumed that the pH of the mesophyll cytosol was constant due to the small micromolar shifts in dissolved CO 2 at different pCO 2 (DiMario et al., 2018). We assumed that the day respiration (Vr, μmol CO 2 m −2 s −1 ) was equal to the dark respiration (R n ) for the C 4 species.

Measurement of anatomical traits and habitat mean annual precipitation
Light and electron microscopy techniques were used to measure important structural and anatomical traits such as adaxial stomatal density (SD ada , number mm −2 ), abaxial stomatal density (SD aba , number mm −2 ), SR (unitless) expressed as ratio of SD ada : SD aba , leaf thickness (μm), mesophyll surface area exposed to intercellular air spaces (S mes , μm 2 μm −2 ) and T CW (μm). The details of sample preparation for microscopy and New Phytologist (2022)

Research
New Phytologist measurements are presented in Pathare et al. (2020a) and in the Methods S1. Values for mean annual precipitation (MAP) for habitats where the C 4 grasses commonly occur were obtained as indicated in Pathare et al. (2020b).

Enzyme assays and measurement of leaf nitrogen content
Immediately following gas exchange measurements, leaf samples were taken from the same leaf and frozen in liquid nitrogen for enzyme assays. Measurements of CA, phosphoenolpyruvate carboxylase (PEPC, μmol m −2 s −1 ) and ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco, μmol m −2 s −1 ) activities were performed at 25°C, as described previously (Sharwood et al., 2016;Sonawane & Cousins, 2019;Pathare et al., 2020a). Carbonic anhydrase activities were expressed as the first-order rate constant (k CA , μmol m −2 s −1 Pa −1 ). In addition to enzyme activities, PEPC affinity for HCO − 3 (K m , μM HCO 3 −1 ) values were derived using a membrane-inlet mass spectrometer. K m values have been published in DiMario et al. (2021). For measuring leaf nitrogen content, leaf samples were taken from the same leaf on which gas exchange measurements were performed. Samples were dried in a hot air oven at 60°C for 72 h. Leaf nitrogen content was measured using a Eurovector elemental analyser and expressed on a leaf area basis (N area , g m −2 ).
Estimating CO 2 responses of physiological traits (g m , g sw , A net and TE i ) Changes in g m in response to decrease in pCO 2 (from 34 to 14 Pa) were analysed using two different methods. First, the percentage change in g m in response to a decrease in pCO 2 (from 34 to 14 Pa) was calculated for each biological replicate for the 16 species as follows: trait value at 14 Pa À trait value at 34 Pa ð Þ Â 100= trait value at 34 Pa ð Þ Similar to g m , we also calculated percentage change in A net , g sw , C i and TE i in response to the decrease in pCO 2 (from 34 to 14 Pa). pCO 2 levels of 34 and 14 Pa were chosen, as these two pCO 2 levels were the highest and lowest levels, respectively, used in current study. Relationships of percentage change in physiological traits with key leaf anatomical and biochemical traits are given in the Figs S3-S8.
In a second method, we used an equation (y = a × (34/ CO 2 ) b ) to model the changes in g m in response to changes in CO 2 (here C a ). In the equation, y indicates the g m , coefficient a (μmol m −2 s −1 Pa −1 ) is the value of g m at 34 Pa pCO 2 , CO 2 indicates the pCO 2 inside the leaf chamber (C a ) and coefficient b (unitless) indicates the sensitivity of g m to changes in C a . For further analysis (Figs 2-4 please refer to later paragraphs), we considered coefficient b as a proxy for the CO 2 response of g m , with a relatively greater value of b indicating a greater increase in g m with a decrease in C a . Mean AE SE values for model coefficients a and b for the 16 C 4 grasses are given in Table S3 (please refer to later paragraphs). Average values for three biochemical subtypes (NAD-ME, NADP-ME and PCK) are also included in Table  S3. Fig. 1 shows the relationship of g m and C a for 16 C 4 grasses along with the model line, whereas Fig. S9 shows relationship of g m and C a along with the model line for the three C 4 subtypes.

Statistical analyses
All statistical analyses were performed using R software (v.4.1.0, R Foundation for Statistical Computing, Vienna, Austria). Normality and equal variances were tested and, when necessary, square root or log transformations were used to improve the data homoscedasticity (Zar, 2007). One-way ANOVA with post-hoc Tukey's test was used to examine differences in leaf-level anatomical and biochemical traits among the 16 diverse C 4 grasses. Results for post-hoc Tukey's test are given in Table 1. Results of one-way ANOVA for traits used in the current study are given in Table S2. For key physiological traits such as g m , Δ 18 O, A net , g sw , C i and TE i , two-way ANOVA was performed with species and pCO 2 as the main effects using the aov function in R (R Core Team, 2018). Results of two-way ANOVA are given in Table 2 and mean values are presented in Figs S1-S6. For the ANOVAs, Pvalues ≤ 0.05 were considered as statistically significant. We grouped the species into biochemical subtypes and analysed whether CO 2 responses of g m varied among the subtypes (Table S3; Fig. S9).
Regression analyses were performed to examine the relationship of coefficient b with the key anatomical and biochemical traits such as T CW , S mes , ratio of T CW : S mes , SR, SD ada , leaf thickness, PEPC activity, Rubisco activity, k CA , K m , A max and N area . To account for the combined effect of anatomy and biochemistry on the CO 2 responses of g m , we derived ratios of T CW to biochemical traits (activities of PEPC, Rubisco and CA, K m , A max and N area ). We examined the relationship between coefficient b and the ratio of T CW to biochemical traits. Similarly, we also examined the relationships between percentage change in g m and key anatomical and biochemical traits and the ratio of T CW to the biochemical traits mentioned above. Both, coefficient b and percentage change in g m showed similar relationships with anatomical and biochemical traits. In the main text we used coefficient b as a proxy for CO 2 response of g m (Figs 2-4). The relationship of percentage change in physiological traits with key leaf anatomical and biochemical traits are given in Figs S3-S8. For the regression analysis, P-values ≤ 0.05 were considered as statistically significant. The function outlierTest from the R package CAR (Fox & Weisberg, 2019) was used to identify any potential influential points in the relationships. Influential data points (with Bonferroni P-value ≤ 0.05) were removed while deriving regression statistics if required (Figs 3c,f, S3, S5, S7). To complement the regression analysis, we also performed a principal component analysis (PCA; Methods S2; Fig. S8; Table  S4) with leaf traits, percentage changes in g m , A net , g sw and TE i and habitat MAP (R package FACTOMINER; Le et al., 2008).

CO 2 response of physiological traits
The 16 C 4 grasses (Table 1) with previously demonstrated variation in leaf-level anatomical traits and g m (Pathare et al., 2020a,b) were chosen to explore the potential variation in the CO 2 response of g m and its relationship with leaf traits and TE i .
Responses of g m to changes in C a and C i are given in Figs 1 and S10, respectively. Responses of other physiological traits (Δ 18 O, g sw , A net , C i and TE i ) to changes in C a are given in Figs S11-S15. Species differed significantly in the CO 2 response of g m (Figs 1, S10) and g sw (Fig. S12). In general, across the 16 C 4 grasses, g m increased as C a and C i decreased with percentage increase at lowest C a , ranging from +13% to +250%. The increase in g m was statistically significant in 12 out of 16 species. As with g m , g sw Fig. 1 Response of mesophyll conductance (g m ) to changes in the partial pressure of CO 2 (pCO 2 ) inside the leaf chamber (C a ) in 16 diverse C 4 grasses. Data for each of the species are shown separately from panels (a) to (p) along with the CO 2 response of g m (black solid line) modelled using the equation, g m = a × (34/C a ) b , where coefficient 'a' is the value of g m at 34 Pa pCO 2 and coefficient 'b' is the rate of change in g m with change in C a . Mean AE SE values for the model constants (a and b) for each species are shown in Supporting Information Table S3. Measurements were performed at constant light (photosynthetic photon flux density (PPFD) = 1200 μmol m −2 s −1 ) and leaf temperature (25°C). Values in each panel represent the mean AE SE (green colour) with n = 3-6. Grey points indicate the replicate values for each species and CO 2 level. Response of g m to changes in pCO 2 for each species is plotted in separate panel of Fig. 1. Species code has been indicated as the first letter of the genus and first three letters of the species (please refer to Table 1 for full names of species). P-values from one-way ANOVA along with Tukey's letters are shown.

New Phytologist
increased with decreasing C a (Fig. S12). However, unlike g m , the magnitude of increase in g sw was lower. Specifically, the percentage increase in g sw ranged from c. 40-80% when C a decreased from 34 to 14 Pa. Alternatively, both A net and TE i decreased with decreases in C a (Figs S13, S15). Particularly, A net decreased by c. 23-40% and TE i decreased by c. 53-64%.

Variation in leaf anatomical and biochemical traits
The results of one-way ANOVA suggested that the 16 C 4 grasses varied significantly (P < 0.001; Table 1) in all the leaflevel anatomical and biochemical traits. T CW showed a significant 2.6-fold variation with values ranging from c. 0.08 to 0.21 μm. S mes showed a 2.4-fold variation with values ranging from c. 8.5 to 19 μm 2 μm −2 . The C 4 grasses measured here also showed a highly significant variation in stomatal traits such as SR and SD ada (P < 0.001; Tables 1, S2). SR varied by 4.7-fold, with values ranging from 0.5 to 2.4, whereas SD ada showed a 13-fold variation with values ranging from 12 to 160 number mm −2 . Leaf thickness showed a significant two-fold variation with values ranging from 120 to 240 μm (P < 0.001; Tables 1, S2). Similar to the variation in anatomical traits mentioned above, the biochemical traits (that is leaf-level activities of PEPC, Rubisco and k CA ) also varied significantly among the 16 C 4 grasses included in the current study (P < 0.001; Tables 1, S2). PEPC activities showed a 7.5-fold variation with values from 96 to 721 μmol m −2 s −1 . Rubisco activities showed a 2.6-fold variation with values from 30 to 80 μmol m −2 s −1 . k CA showed an 18-fold variation with values from 9 to 165 μmol m −2 s −1 Pa −1 . PEPC affinity for HCO 3 − (K m ) showed a 1.6-fold variation across the 16 C 4 grasses with values ranging from c. 29 to 46 μM HCO 3 − (P < 0.001; Tables 1, S2). There was a 1.7-fold variation in maximum photosynthetic rates (A max ) with values ranging from c. 25 to 42.5 μmol CO 2 m −2 s −1 (P < 0.001; Table 1). N area also varied significantly among the grasses (P < 0.001; Table 1) with values ranging from 1.24 to 2.26 g m −2 .

Relationships of CO 2 response of g m with anatomical and biochemical traits
We used two proxies to account for the changes in g m in response to change in C a . First, percentage change in g m in response to decrease in C a (from 34 to 14 Pa) (Figs S3-S8) and second, the coefficient b derived from an equation used to model the changes in g m in response to changes in C a . In the main text, we used coefficient b as a proxy for CO 2 response of g m , (Figs 2-4). Coefficient b showed a strong negative relationship with T CW (R 2 = −0.74, P < 0.001; Fig. 2a) and ratio of T CW : S mes (R 2 = −0.64, P < 0.001; Fig. 2c) and a nonsignificant positive relationship with S mes (R 2 = 0.14, P = 0.11; Fig. 2b). In addition, coefficient b showed a significant positive relationship with SR (R 2 = 0.34, P < 0.05; Fig. 2d), SD ada (R 2 = 0.32, P < 0.01; Fig. 2e) and leaf thickness (R 2 = 0.38, P < 0.01; Fig. 2f). Coefficient b also showed a nonsignificant positive relationship with biochemical traits such as PEPC activity (R 2 = 0.19, P = 0.12; Fig. 3a) and A max (R 2 = 0.20, P = 0.08; Fig. 3e) and a significant positive relationship with Rubisco activity (R 2 = 0.24, P < 0.05; Fig. 3b). However, coefficient b did not show a statistically significant relationship with K m (Fig. 3c,  d). Coefficient b showed significant positive relationships with k CA (R 2 = 0.26, P < 0.05; Fig. 3c) and N area (R 2 = 0.31, P < 0.05; Fig. 3f) after removing the influential species (pvir and udic, respectively).
To account for the combined effects of leaf anatomy (specifically T CW that acts as a resistance in series to CO 2 diffusion) and biochemistry (that can have a facilitating effect) on CO 2 response of C 4 -g m , we derived the ratio of T CW with PEPC activity, Rubisco activity, k CA , K m , A max and N area and investigated the relationship between coefficient b and these ratios (Fig. 4). Coefficient b showed a significant negative relationship with T CW / PEPC (R 2 = −0.44, P < 0.01; Fig. 4a), T CW /Rubisco (R 2 = −0.57, P < 0.01; Fig. 4b), T CW /K m (R 2 = −0.47, P < 0.01; Fig. 4c), T CW /k CA (R 2 = −0.47, P < 0.05; Fig. 4d), T CW /A max (R 2 = −0.62, P < 0.001; Fig. 4e) and T CW /N area (R 2 = −0.75, P < 0.001; Fig. 4f). In summary, C 4 grasses with thinner cell walls combined with greater S mes , N area , A max and activities of key enzymes were able to achieve greater increases in g m at lower C a . Similar results were observed when we plotted percentage change in g m against key anatomical and biochemical traits mentioned above (Figs S3-S8).
Relationship of CO 2 response of g m with CO 2 response of TE i , g sw and A net We investigated the impacts of changes in g m with C a on corresponding changes in TE i , g sw and A net (Fig. S7). For this we used percentage change in the trait's value when C a decreased from 34 to 14 Pa. As g m and g sw increase with decreases in C a , percentage change for both conductances is reported in the manuscript as percentage increase. Alternatively, A net and TE i decrease with decreases in CO 2 S, therefore the percentage change for these two traits have been reported in the manuscript as percentage decrease. Percentage change in TE i showed a strong negative correlation with percentage change in g m (R 2 = −0.54, P < 0.001; Fig. S7a). Particularly, C 4 grasses that were able to achieve a greater increase in g m at lower C a showed a lesser decrease in TE i (indicated by a less negative value for TE i in Fig. S11a). In addition, percentage change in g sw showed a negative relationship with percentage change in g m (R 2 = −0.28, P < 0.05; Fig. S7b), that is species that showed a greater increase in g m at lower C a also showed a lesser increase in g sw . A significant relationship was not observed between percentage change in g m and A net (Fig. S7c).

CO 2 response of g m in the C 4 biochemical subtypes
We also analysed the CO 2 responses of g m among the subtypes (Table S3; Fig. S9). NADP-ME (coefficient b = 0.618) and PCK (coefficient b = 0.0.73) subtypes showed 20% and 43% greater sensitivity of g m to C a respectively compared with NAD-ME (coefficient b = 0.512). However, there were no statistically significant differences in coefficient b among the subtypes (Fig.  S9e), which could be due to the low replication. With only seven NADP-ME, five PCK and four NAD-ME replicate species our study does not provide the statistical power to discuss subtype effects. Therefore, here we focus on species-level differences.

Discussion
In general C 3 -g m increases under short-term decreases in pCO 2 (Flexas et al., 2007;Bunce, 2010;Douthe et al., 2011), although not always (von Caemmerer & Evans, 1991;Loreto et al., 1992;Tazoe et al., 2009). The main candidates suggested to affect the CO 2 response of C 3 -g m include chloroplast movement and therefore changes in S c , changes in activities of CA (Evans et al., (e) (f) Fig. 2 Relationship of coefficient b (or sensitivity of g m , unitless) with (a) mesophyll cell wall thickness (T CW ), (b) mesophyll surface area exposed to intercellular air spaces (S mes ), (c) ratio of T CW : S mes , (d) stomatal ratio (SR), (e) stomatal density adaxial (SD ada ) and (f) leaf thickness among the 16 C 4 grasses measured in current study. Linear models were used for deriving regression coefficients (R 2 ) in all panels, except panels (a) and (c) for which we used a polynomial model. Significance of R 2 : *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. Each circle represents the mean AE SE value for each species (n = 3-6). Species names are indicated by the codes given in Table 1.
New Phytologist (2022) Momayyezi & Guy, 2017) and the facilitation effect of CO 2 -permeable aquaporins (Uehlein et al., 2008;Flexas et al., 2012;Kaldenhoff, 2012). However, the C 3 -g m response to pCO 2 may also result from systematic errors associated with the use of methods such as gas exchange, chlorophyll fluorescence and discrimination against 13 C or the use of oversimplified models (Pons et al., 2009;Yin & Struik, 2009;Tholen & Zhu, 2011;Gu & Sun, 2014). For instance, the apparent response of C 3 -g m to short-term changes in pCO 2 has partly been attributed to changes in photorespiratory and nonphotorespiratory release of CO 2 (or day respiration) to the total CO 2 pool in the leaf, particularly under low pCO 2 . Ignoring these CO 2 pools while estimating C 3 -g m also overlooks the effects of spatial distribution of mitochondria and chloroplast on pathlength of CO 2 movement. Inaccurate measurements of day respiration or estimates of the Rubisco fractionation factor in the Δ 13 C method and measurements conducted at low [O 2 ] have also been suggested as potential sources of artefacts in determining the C 3 -g m response to pCO 2 (Pons et al., 2009;Gu & Sun, 2014).
Alternatively, in C 4 species, the photorespiratory release of CO 2 is relatively low and may not contribute significantly to estimates of g m , even at low pCO 2 . Also, in current study we estimated g m in diverse C 4 grasses by using a method based on modelling of Δ 18 O that considers the contribution of respiratory fluxes (Ogee et al., 2018). Despite some uncertainties associated with the day respiration or water isotope gradient between mesophyll and bundle sheath cells, the Ogee et al. (2018) method provides robust estimates of C 4 -g m . Also, comparing many diverse C 4 species, as done in the current study, using a method that is not subject to the same limitations as those previously used for C 3 species is a reasonable approach to identify the CO 2 responses of C 4 -g m . Here, we discuss the CO 2 response of g m for 16 diverse C 4 grasses and its relationships with leaf anatomical and biochemical traits. photosynthetic capacity (A max ) and (f) leaf N content (N area ) among the 16 C 4 grasses measured in current study. Linear models were used for deriving regression coefficients (R 2 ) in all panels. In panel (c) R 2 = 0.22 + after removing influential species pvir. Significance of R 2 : +, marginally significant; *, P ≤ 0.05. Each circle represents the mean AE SE value for each species (n = 3-6). Species names are indicated by codes given in Table 1.
New Phytologist (2022) 236: 1281-1295 www.newphytologist.com CO 2 response of g m varied among diverse C 4 grasses Very few studies investigating the CO 2 response of C 4 -g m , primarily maize and Setaria viridis Kolbe & Cousins, 2018;Ubierna et al., 2018), have reported an increase in C 4 -g m with a decrease in pCO 2 . Similarly, we demonstrated that g m increases with a decrease in pCO 2 across the 16 diverse C 4 grasses. However, the magnitude of the increase in g m varied greatly across the 16 C 4 grasses (+13% to +250%; Fig. 1). In the following sections we discuss the potential factors that could explain this variability in the CO 2 response of C 4 -g m .
CO 2 response of C 4 -g m is related with T CW, S mes and photosynthetic capacity g m in C 3 and C 4 species is constrained by several anatomical and biochemical parameters such as T CW , S mes , S c , CA activities and aquaporins Evans, 2021). However, the potential implication of leaf anatomy and biochemistry for the CO 2 response of g m has not been studied for C 4 species. Most of the anatomical parameters remain unchanged under short-term changes in environmental conditions such as pCO 2 Terashima et al., 2011). Therefore, the only anatomical trait suggested to influence the CO 2 response of g m is chloroplast movement that can affect g m by changing S c (Terashima et al., 2006;Tholen et al., 2008;Carriqui et al., 2018). However, chloroplast movement is unlikely to explain the variability of the CO 2 response of C 4 -g m observed in the current study, because S mes and not S c is a more accurate determinant of C 4 -g m , in which a greater S mes results in greater g m in C 4 species Pathare et al., 2020a). In contrast with S c , S mes should remain unchanged under short-term changes in pCO 2 . Therefore, although we observed a nonsignificant positive relationship between the CO 2 response of C 4 -g m and S mes (Fig. 2b), S mes may provide only a partial explanation for this variable response. activity expressed as k CA , (e) maximum photosynthetic rates (A max ) and (f) leaf N content (N area ) among the 16 C 4 grasses measured in current study. Polynomial models were used to derive regression coefficients (R 2 ) in all panels. Significance of R 2 : *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. Each circle represents the mean AE SE value for each species (n = 3-6). Species names are indicated by codes given in Table 1.

New Phytologist
T CW could account for > 50% of the total resistance to CO 2 diffusion inside leaves . Therefore, we further investigated the influence of T CW on the variability of the CO 2 response of C 4 -g m . In general, C 3 species with relatively lower T CW showed greater g m (Onoda et al., 2017;Veromann-Jürgenson et al., 2017;Evans, 2021). We did not observe a strong relationship between g m and T CW for the C 4 grasses under ambient CO 2 levels (34 Pa) (Pathare et al., 2020a). However, in the current study, the CO 2 response of C 4 -g m was related to T CW (Fig. 2a). The CO 2 response of C 4 -g m also showed a strong relationship with T CW after accounting for S mes (as T CW /S mes ; Fig. 2c). Particularly, C 4 grasses with relatively lower T CW and greater S mes , showed a greater increase in g m at low pCO 2 . It is unclear why the g m of C 4 grasses with lower T CW is more responsive to changes in pCO 2 . As for S mes , T CW is unlikely to change under short-term changes in pCO 2 and may not be the sole reason for the observed variable CO 2 response of C 4 -g m . However, T CW may still provide a partial explanation for variability in the CO 2 response of C 4 -g m . Resistances to CO 2 diffusion through the liquid phase (comprised of apoplastic and cellular components from mesophyll cell wall to site of carboxylation) are greater compared with the gaseous phase Flexas et al., 2021). CO 2 must dissolve in the water-filled pores of the mesophyll cell walls and then diffuse to the plasma membrane and eventually to the site of CO 2 fixation. Because cell walls represent a significant proportion of liquid phase resistance, C 4 species with a lower T CW may have the potential to achieve a greater change in g m in response to changing pCO 2 . However, the influence of lower T CW on the CO 2 response of C 4 -g m may have also been augmented by greater S mes , the facilitation effect of CO 2 transporting aquaporins and leaf N content and proteins of photosynthetic machinery that determine drawdown and fixation of CO 2 and therefore photosynthetic capacity (Parkhurst, 1994;Wright et al., 2004;Evans et al., 2009;Xiong & Flexas, 2021).
The role of CO 2 -permeable aquaporins in enhancing g m has been well characterised in C 3 species (Uehlein et al., 2008). Only recently, it has been demonstrated that overexpressing a CO 2 -permeable aquaporin in plasma membranes of Setaria viridis (C 4 grass) can enhance C 4 -g m (Ermakova et al., 2021). We did not investigate the role of aquaporins in affecting the CO 2 response of C 4 -g m . However, we observed a greater CO 2 response of g m in C 4 grasses with relatively lower T CW . This indicates that g m in plants with lower T CW may be more influenced by aquaporins (Evans, 2021). There is still a need to further investigate the role of aquaporins in the variable CO 2 response of C 4 -g m .
Here, we investigated the relationship between the CO 2 response of C 4 -g m and photosynthetic capacity as indicated by A max , N area and activities of key C 4 photosynthetic enzymes (PEPC, CA and Rubisco). C 4 species with a greater CO 2 response of C 4 -g m tended to show greater activities of PEPC and Rubisco and greater A max and N area . Also, the CO 2 response of C 4 -g m was strongly related to the ratio of T CW /photosynthetic capacity traits (Fig. 4), in which a greater CO 2 response of C 4g m was evident in species with lower T CW and greater photosynthetic capacity. This suggests that lower T CW may have decreased the resistance to the movement of CO 2 into the mesophyll cells (Evans, 2021;Flexas et al., 2021), whereas the greater photosynthetic capacity may have increased the demand for CO 2 , resulting in greater drawdown of CO 2 and the necessity of maintaining a greater CO 2 supply through an increase in g m at low pCO 2 (Wright et al., 2004;Evans et al., 2009). Furthermore, the enzyme CA catalyses the conversion of CO 2 /HCO 3 − in the cytosol and ensures sufficient HCO 3 − substrate supply to PEPC (Studer et al., 2014;DiMario et al., 2018). Because CO 2 diffuses much faster in liquid phase in HCO 3 − form, greater CA activities in combination with lower T CW could have maintained a rapid supply of HCO 3 − substrate to PEPC and further enhanced g m at low pCO 2 in some C 4 grasses (Fig. 4d).
We also investigated the relationship between the CO 2 response of C 4 -g m and K m , a kinetic constant indicating PEPC affinity for HCO 3 − (DiMario et al., 2021). In general, lower K m values (or high affinity of PEPC for HCO 3 − ) are expected to provide a selective advantage by maintaining high rates of C 4 photosynthesis, particularly under conditions such as drought when CO 2 availability is low due to restricted stomatal conductance. Here, we observed that K m alone did not show a strong relationship with CO 2 response of C 4 -g m (Fig. 3d). However, after accounting for T CW , we observed that T CW /K m showed a significant negative relationship with the CO 2 response of C 4 -g m (Fig. 4c), in which species with relatively lower T CW and higher K m (or lower affinity of PEPC for HCO 3 − ) exhibited a greater CO 2 response of g m . This contrasts with the general expectation and could be explained by the lower T CW leading to higher g m and with the corresponding greater CA activities leading to higher HCO 3 − in mesophyll cells under low C i conditions. The higher HCO 3 − concentration in the mesophyll cells of species with lower T CW and greater CA can reduce the selective pressure on PEPC for lower K m values.
Relationship of CO 2 response of g m with CO 2 response of g sw and TE i Previously, we demonstrated that C 4 -g m is positively related to TE i under ambient pCO 2 (Pathare et al., 2020a). Our current study further suggests that, for C 4 grasses, g m may also influence TE i under short-term changes in pCO 2 . In general, both g m and g sw increased (Figs 1, S12) and TE i decreased at low pCO 2 (Fig.  S15). However, the magnitude of increase in g m at low pCO 2 was greater (values from 13% to 250%) compared with the increase in g sw (values from 40% to 80%). Also, C 4 grasses showing greatest increase in g m at low pCO 2 also showed the lowest increase in g sw (Fig. S7b). Consequently, although TE i decreased at low pCO 2 , the decrease was less in the species showing the greater CO 2 response of g m (Fig. S7a). These types of species with greater CO 2 response of g m may benefit in terms of maintaining TE i under low CO 2 conditions, such as drought, compared with species whose g m was less responsive to changes in pCO 2 . Also, the PCA (Fig. S8) suggests that a greater CO 2 response of g m is generally observed in C 4 grasses adapted to habitats with relatively low MAP.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.   relatively lower b values indicate lower rate of change in g m with C a ) and relationship between coefficient b and percentage change in g m .

Fig. S4
Relationship of percent increase in g m with mesophyll cell wall thickness (T CW ) mesophyll surface area exposed to intercellular air spaces (S mes ) ratio of T CW : S mes , stomatal ratio (SR), stomatal density adaxial (SD ada ) and leaf thickness among the 16 C 4 grasses measured in current study.

Fig. S5
Relationship of percent increase in g m with PEPC activity Rubisco activity CA activity expressed as k CA , PEPC affinity for HCO 3 − (K m ), maximum photosynthetic capacity (A max ) and leaf N content (N area ) among the 16 C 4 grasses measured in current study.

Fig. S6
Relationship of percent increase in g m with ratio of mesophyll cell wall thickness (T CW ) to PEPC activity Rubisco activity PEPC affinity for HCO 3 − (K m ), CA activity expressed as k CA , maximum photosynthetic rates (A max ) and leaf N content (N area ) among the 16 C 4 grasses measured in current study.

Fig. S7
Relationship of percent increase in g m with percent decrease in leaf-level water-use efficiency (TE i ) expressed as A net / g sw (higher negative value indicates greater decrease in TE i ), percent increase in stomatal conductance to water (g sw ) and percent decrease in net photosynthetic rates (A net ) (higher negative value indicates greater decrease in A net ) for the 16 C 4 grasses measured in current study.

Fig. S8
PCA biplot showing major axes of variation in important leaf-level anatomical and biochemical traits and percent change (increase or decrease) in response to CO 2 in physiological traits such as g m , A net , g sw and TE i for the 16 diverse C 4 grasses measured in current study.

Fig. S9
Response of mesophyll conductance (g m ) to changes in pCO 2 inside leaf chamber (C a ) in three C 4 biochemical subtypes (NAD-ME, NADP-ME, PCK).

Fig. S10
Response of mesophyll conductance to CO 2 (g m ) to changes in intercellular CO 2 (C i ) in 16 diverse C 4 grasses measured in current study.

New Phytologist
Fig. S14 Response of leaf intercellular CO 2 concentration (C i ) to changes in pCO 2 inside leaf chamber (C a ) in 16 diverse C 4 grasses measured in current study.

Fig. S15
Response of leaf-level water-use efficiency (TE i = A net / g sw ) to changes in pCO 2 inside leaf chamber (C a ) in 16 diverse C 4 grasses measured in current study.
Methods S1 Measurement of anatomical traits.
Methods S2 Principal component analysis.

Table S1
Key input parameters used in calculation of isotope parameters and estimation of mesophyll conductance (g m ) for the 16 C 4 grasses at four pCO 2 levels (34, 27, 20 and 14 Pa) and a temperature of 25°C.

Table S2
Results of one-way ANOVA with species as main effects for all the leaf-level anatomical and biochemical traits measured for 16 C 4 grasses in current study.
Table S3 C 4 grasses used in current study along with their biochemical subtype and mean AE SE values for the equation (g m = a × (34/C a ) b ) constants derived for the CO 2 response of g m in the 16 C 4 grasses.