The strategies of water – carbon regulation of plants in a subtropical primary forest on karst soils in China

Coexisting plant species in a karst ecosystem may use diverse strategies of trade off between carbon gain and water loss to adopt to the low soil nutrient and low water availability conditions. An understanding of the impact of CO2 diffusion and maximum carboxylase activity of Rubisco (Vcmax) on the light-saturated net photosynthesis (A) and intrinsic water use efficiency (iWUE) can provide insight into physiological strategies of the water–carbon regulation of coexisting plant species used in adaptation to karst environments at the leaf scale. We selected 63 dominant species (across 6 life forms) in a subtropical karst primary forest in southwestern China, measured their CO2 response curves, and calculated the corresponding stomatal conductance to CO2 (gs), mesophyll conductance to CO2 (gm), and Vcmax. The results showed that gs and gm varied about 7.6and 34.5fold, respectively, and that gs was positively related to gm. The contribution of gm to the leaf CO2 gradient was similar to that of gs. gs /A, gm /A and gt /A was negatively related to Vcmax /A. The relative limitations of gs (ls), gm (lm), and Vcmax (lb) toA for the whole group (combined six life forms) were significantly different from each other (P < 0.05). lm was the largest (0.38± 0.12), followed by lb (0.34± 0.14), and ls (0.28± 0.07). No significant difference was found between ls, lm, and lb for trees and tree/shrubs, while lm was the largest, followed by lb and ls for shrubs, grasses, vines and ferns (P < 0.05). iWUE varied about 3-fold (from 29.52 to 88.92 μmol CO2 mol−1 H2O) across all species, and was significantly correlated with gs, Vcmax, gm / gs, and Vcmax / gs. These results indicated that karst plants maintained relatively high A and low iWUE through the covariation of gs, gm, and Vcmax as an adaptation to a karst environment.


Introduction
Diverse strategies of trade off between carbon gain and water loss are critical for the survival of coexisting plant species.In order to adapt to harsh environments, coexisting plant species develop distinct patterns of strategies for carbonwater regulation (light-saturated net photosynthesis (A) and intrinsic water use efficiency; iWUE) (Sullivan et al., 2017).iWUE is the ratio of A to stomatal conductance to H 2 O (g sw ) (Moreno-Gutierrez et al., 2012).Plants with high iWUE are better able to adapt to nutrient-and water-limited environments (Flexas et al., 2016).Due to the greater hydraulic erosion and complex underground drainage network (Nie et al., 2014;Chen et al., 2015), karst soils cannot retain enough nutrients and water for plant growth even though precipitation is high (1000-2000 mm) (Liu et al., 2011;Fu et al., 2012;Chen et al., 2015).An understanding of the impact of CO 2 diffusion and the maximum carboxylase activity of Rubisco (V cmax ) on A and iWUE in karst plants can provide insight into the physiological strategies of water-carbon regulation of plants used in adaptation to karst environments at the leaf scale.Until now, variability in A and iWUE has only been reported in 13 co-occurring trees and 12 vines (Chen et al., 2015), and 12 co-occurring tree species (Fu et al., 2012) in two tropical karst forests in southwestern China.
Based on Fick's first law, A has been shown to be only limited by leaf stomatal conductance to CO 2 (g s = g sw /1.6) and V cmax (Flexas et al., 2012;Buckley and Warren, 2014); originally, mesophyll conductance to CO 2 (g m ) was proposed to be infinite, i.e., CO 2 concentration in chloroplast (C c ) was equal to the CO 2 concentration in intercellular air space (C i ).However, g m varies greatly among species (Warren and Adams, 2006;Flexas et al., 2013).Recent studies have confirmed that A was jointly constrained by g s , g m , and V cmax , and that their relative contribution to A was speciesdependent and site-specific (Carriqui et al., 2015;Tosens et al., 2016;Galmes et al., 2017;Peguero-Pina et al., 2017a, b;Veromann-Jurgenson et al., 2017).
Variation in iWUE (= A / g sw ) depends on the relative changes in A (g s , g m , V cmax ) and g sw (g sw = 1.6 g s ) (Flexas et al., 2013;Gago et al., 2014).Theoretical relationships between iWUE and g s , g m , and V cmax have been deduced using two approaches.First, based on Fick's first law of CO 2 diffusion, Flexas et al. (2013) deduced that iWUE was a function of g m / g s and CO 2 gradients (the ambient CO 2 concentration (C a )-the CO 2 concentration at the sites of carboxylation (C c )) within a leaf.And second, combining Fick's first law of CO 2 diffusion and the Farquhar biochemical model (Farquhar and Sharkey, 1982), Flexas et al. (2016) deduced that iWUE was a function of V cmax / g s , C c , CO 2 compensation point of photosynthesis ( * ), and the effective Michaelis-Menten constant of Rubisco for CO 2 (K m ).Until now, most previous studies have focused on the role of CO 2 diffusion in limiting iWUE, and have suggested that iWUE was negatively related to g s , and positively related to g m / g s (Flexas et al., 2013).Gago et al. (2014) used a meta-analysis with 239 species, and were the first to confirm that iWUE was positively related to V cmax / g s .Although both g m / g s and V cmax / g s were positively correlated with iWUE, there was only a weak correlation between g m / g s and V cmax / g s , which indicates that iWUE can be improved by increasing V cmax or g m (proportionally higher than g s ), not both (Gago et al., 2014).
It is noteworthy that Flexas et al. (2016) and Gago et al. (2014) found that most of the previous work on constraints ofg s , g m , and V cmax on A were conducted in crops or saplings, and only a few studies were in natural ecosystems.For example, g m was the main factor limiting A in two Antarctic vascular grasses (Saez et al., 2017), and in 35 Australian sclerophylls (Niinemets et al., 2009b) in different habitats.The A of two closely related Mediterranean Abies species growing in two different habitats was mainly constrained by g m in one, and by g s in the other habitat (Peguero-Pina et al., 2012).Beyond that, it still remains unknown how g s , g m , and V cmax regulate A and iWUE across species in natural ecosystems.
In this study, we selected 63 dominant plant species, including 6 life forms (tree (n = 29), tree/shrub (n = 11), shrub (n = 11), grass (n = 11), vine (n = 5), and fern (n = 3)), from a subtropical primary forest in the "karst critical zone (karst CZ)" of southwestern China, and measured their A and CO 2 response curves.g m was calculated using the curve-fitting method (Ethier and Livingston, 2004).The obtained g m was used to transform the A-C i into A-C c response curves, and then to calculate the A and V cmax .Our objective was to determine and distinguish the limitations of CO 2 diffusion (g s and g m ) and V cmax on A and iWUE in different life forms in this karst primary forest, and to understand the patterns of strategies for carbon-water regulation in karst plants.

Site information
This study was conducted in a subtropical primary forest (26 • 14 48 N, 105 • 45 51 E; elevation, 1460 m), located in the karst CZ of southwestern China.This region has a typical subtropical monsoon climate, with a mean annual precipitation of 1255 mm, and mean annual air temperature of 15.1 • C (Zeng et al., 2016).The soils are characterized by a high ratio of exposed rock, shallow and nonhomogeneous soil cover, and complex underground drainage networks, e.g., grooves, channels, and depressions (Chen et al., 2010;Zhang et al., 2011;Wen et al., 2016).Soils and soil water are easily leached into underground drainage networks.Soil texture is silt clay loam, and soil PH is 6.80 ± 0.16 (Chang et al., 2018).The total nitrogen and phosphorus content in soil is 7.30 ± 0.66 and 1.18 ± 0.35 g Kg −1 , respectively, which was similar with that of non-karst CZs (Wang et al., 2018).However, the soil quantities (16.04-61.89Kg m −2 ) and nitrogen and phosphorus storage (12.04 and 1.68 t hm −2 ) is much lower than that of non-karst CZs, due to the thin and heterogeneous soil layer (Jobbagy and Jackson, 2000;He et al., 2008;Li et al., 2007;Lu et al., 2010).The typical vegetation type is mixed evergreen and broadleaf deciduous primary forest, dominated by Itea yunnanensis Franch., Carpinus pubescens Burk., and Lithocarpus confinis Huang., etc. (Wang et al., 2018).

Leaf gas-exchange measurements
In July and August 2016, 63 species (Table S1 in the Supplement) were selected for measurements of the A and CO 2 response curves.The species sampled were selected according to their abundance in the study site.They are the main components of this forest, including 55 woody species (46 deciduous and 10 evergreen species) and 5 herb species.To distinguish the strategies of water-carbon regulation of plants among different life forms, the species were grouped into 6 life forms, including (1) tree (n = 29), (2) tree/shrub (n = 11), (3) shrub (n = 11), (4) grass (n = 11), ( 5) vine (n = 5), and ( 6) fern (n = 3)."Tree/shrub" is a kind of low wood plant between tree and shrub.Ferns grow in the understory; vines climb up to the shrub canopy to get light.Details of leaf sampling and measurements of the CO 2 response curve are briefly described as follows.Branches exposed to the sun were excised from the upper part of the crown (trees, tree/shrubs, shrubs and vines) or aboveground portion (grasses, ferns), and immediately recut under water to maintain xylem water continuity.Back in the laboratory, branches and aboveground portions were kept at 25 • C for 30 min.Fully expanded and mature leaves were induced for 30 min at a saturating light density (1500 µmol m −2 s −1 ).CO 2 response curve measurements were performed when A andg s was stable.Three leaves per species were collected and measured.A total of 189 leaves were collected from adult individuals of 63 species.
The CO 2 response curves were measured with 11 CO 2 concentration gradients in the chamber following the procedural guidelines described by Long and Bernacchi (2003).The photosynthetic photon flux density was 1500 µmol m −2 s −1 .The leaf temperature was 25 • C, controlled by the block temperature.The humidity in the leaf chamber was maintained at ambient conditions.Leaf area, thickness (LT), and dry mass were measured after the CO 2 response measurements.Leaf mass per area (LMA) was calculated by dividing the corresponding dry mass by leaf area.Finally, leaf density (LD) was calculated by dividing the corresponding LMA by LT.More details are given in Wang et al. (2018).

Response curve analyses
A and the corresponding g sw (g s = g sw /1.6), C a , and C i were extracted from the CO 2 response curve under saturating light (1500 µmol m −2 s −1 ) conditions, with the CO 2 concentration inside the cuvette set to 400 µmol mol −1 (Domingues et al., 2010).V cmax was estimated by fitting A-C c curves (Ethier and Livingston, 2004).The obtained values of g m were used to transform the Three methods are most commonly used for g m estimation.Those methods have been reviewed by Warren (2006) and Pons et al. (2009).Briefly, g m can be calculated by the stable isotope method (Evans, 1983;Sharkey et al., 1991;Loreto et al., 1992), the J method (Bongi and Loreto, 1989;Dimarco et al., 1990;Harley et al., 1992;Epron et al., 1995;Laisk et al., 2005), and the "curve-fitting" method (Ethier and Livingston, 2004;Sharkey et al., 2007).All of these methods are based on gas-exchange measurements (Pons et al., 2009), and some common assumptions (Warren, 2006).Thus, the accuracy of each method is unknown to some extent (Warren, 2006).
g m was estimated by the "curve-fitting" method in this study.Although the "curve-fitting" method is less precise than the stable isotope method, it is much more readily available and has been used for several decades (Warren, 2006;Sharkey, 2012).Accurate measurements of A and C i are prerequisites for estimating g m using the "curve-fitting" method (Pons et al., 2009).Warren (2006) pointed out that highly accurate measurements require small leaf areas and low flow rates.We confirmed that the calculated C c and the initial slope of the A-C c curves were positive, suggesting that the measured g m was reliable (Warren, 2006).

Theory of trade-off between carbon and water at leaf scale
The exchange of H 2 O and CO 2 between the leaf and the atmosphere is regulated by stomata (Gago et al., 2014).According to Fick's first law of diffusion, A and g s are related as follows: where A is the photosynthetic rate (µmol CO 2 m −2 s −1 ), C a is the ambient CO 2 concentration (µmol mol −1 ), and C i is the intercellular CO 2 concentration (µmol mol −1 ).
Mesophyll is the barrier for CO 2 inside the leaf.A and mesophyll conductance to CO 2 (g m ) are related as follows: where C c is the CO 2 concentration at the sites of carboxylation (µmol mol −1 ).C c not only depends on CO 2 supply by g m , but also on CO 2 demand (the maximum carboxylase activity of Rubisco, V cmax ).

2.4.1
The relationship between iWUE and g m / g s iWUE is a function of CO 2 diffusion conductances (e.g., g s and g m ) and leaf CO 2 concentration gradients.We can express A as the product of the total CO 2 diffusion conductance (g t ) from ambient air to chloroplasts, and the corresponding CO 2 concentration gradients by combining Eq. ( 1) and (2) (Flexas et al., 2013): where g t = 1/(1/g s + 1/g m ).This equation demonstrates that CO 2 concentration gradients in leaves are constrained by stomatal and mesophyll resistance to CO 2 .Therefore, iWUE can be expressed as Equation ( 4) means that iWUE is positively related to g m / g s , but not to g m itself (Warren and Adams, 2006;Flexas et al., 2013;Buckley and Warren, 2014;Cano et al., 2014).

2.4.2
The relationship between iWUE and V cmax / g s When Fick's first law and the Farquhar biochemical model (Farquhar and Sharkey, 1982) are combined, iWUE is also a function of V cmax .Based on the Farquhar biochemical model (Farquhar and Sharkey, 1982), when A is limited by Rubisco, it can be expressed by the following equation (Sharkey et al., 2007): where * is the CO 2 compensation point of photosynthesis in the absence of non-photorespiratory respiration in light (R d ), and K m is the effective Michaelis-Menten constant of Rubisco for CO 2 .Combining Eqs. ( 1) and ( 5) (Flexas et al., 2016), we obtain the following: Because R d is much smaller than A in actively photosynthesizing leaves, V cmax / g s can be approximated as Consequently, iWUE can be expressed as 2.5 Statistical analysis

Quantitative analysis of limitations on A
The relative contribution of g s (l s ), g m (l m ), and V cmax (l b ) to A can be separated by a quantitative limitation model introduced by Jones (1985) and further developed by Grassi and Magnani (2005).The sum of l s , l m , and l b is one.l s , l m and l b can be calculated as follows: where ∂A / ∂C c was calculated as the slope of A-C c response curves over a C c range of 50-100 µmol mol −1 .l s , l m , and l b have no units.A is co-limited by the three factors when l s ≈ 0.3, l m ≈ 0.3, and l b ≈ 0.4 (Galmes et al., 2017).

Data analysis
Data were analyzed either as a whole group (six life forms combined) or by individual life forms.The bivariate linear regressions of leaf gas exchange parameters were performed using the standardized major axis (SMA) regression fits, and all of the data were made on log e -transformed data (Table S2).
To test for differences among life forms, SMA regression fits were used to compare the slope of regression lines where significant relationships had already been obtained.Note that grass, vine, and fern were not considered due to the small sample size.A similar trend was obtained, and no significant difference was found between life forms although significant relationships were not obtained for some bivariate linear regressions.Accordingly, six life forms were grouped together to analyze the strategy of water-carbon regulation of plants.
The difference of relative limitation of g s , g m , and V cmax to A for life forms or as a whole group were performed using one-way ANOVA and Duncan's multiple comparison.The probability of significance was defined at p < 0.05.V is the ratio of photosynthetic capacity (V cmax ) to lightsaturated net photosynthesis (A, µmol CO 2 m −2 s −1 ); G s is the ratio of stomatal conductance to CO 2 (g s , mol CO 2 m −2 s −1 ) to A; G m is the ratio of mesophyll conductance to CO 2 (g m , mol CO 2 m −2 s −1 ) to A; and G t is the ratio of total conductance to CO 2 (g t , mol CO 2 m −2 s −1 ) to A. Lines refer to regression line for 63 species.T, TS, S, G, V, and F represent tree, tree/shrub, shrub, grass, vine, and fern, respectively.

Results
3.1 Interrelation among g s , g m , g t , and V cmax CO 2 concentration gradients in leaves were controlled by CO 2 diffusion conductance and V cmax .Figure 1 shows the relationship between CO 2 gradients (C a -C i , C i -C c , and C a -C c ) in leaves and the corresponding CO 2 diffusion conductance (g s , g m , and g t ) (Fig. 1a-c), and between C a -C c and V cmax (Fig. 1d).CO 2 concentration gradients (C a -C i , C i -C c , and C a -C c ) were significantly negatively associated with the corresponding CO 2 diffusion conductance (g s , g m , and g t ) (P < 0.001).V cmax was positively associated with C a -C c (P < 0.001).
g s , g m , and g t were significantly positively related to each other (P < 0.001) (Fig. S1 in the Supplement).The contribution of g m to the leaf CO 2 gradient was similar to that of g s .The contribution of g s (57.51-155.13µmol mol −1 ) to C a -C c (98.50-282.94µmol mol −1 ) varied from 28 to 86 %, and the contribution of g m (18.15-179.36µmol mol −1 ) to C a -C c varied from 14 to 72 %.But the variation range of g m (0.02-0.69 mol CO 2 m −2 s −1 ) was 4.5 times that of g s (0.05-0.38 mol CO 2 m −2 s −1 ).
No relationship was found between the CO 2 diffusion conductance (g s , g m , and g t ) and V cmax (Fig. S2).However, after normalization of g s , g m , g t , and V cmax for A (normalized parameters are hereafter called G S = g S / A, G m = g m / A, G t = g t / A, and V = V cmax / A), V was significantly positively correlated with G m and G t (P < 0.001) (Fig. 2b, c), and was slightly positively correlated with G s (P < 0.05) (Fig. 2a), which represented the trade-off between CO 2 supply and demand.
3.2 Contribution of g s , g m , and V cmax to A The variation in A was attributed to variation in g s , g m , g t , and V cmax .A was positively correlated with g s (Fig. 3a), g m (Fig. 3b), and V cmax (Fig. 3c).We used the quantitative limitation model (Eqs.9, 10 and 11) to separate g s (l s ), g m (l m ), and V cmax (l b ) limitations to A. l s , l m , and l b were negatively associated with g s , g m , and V cmax , respectively (Fig. 4).The contributions by g s , g m , and V cmax to limiting A were different for each species (Fig. S3).l s varied 2.6-fold (from 0.17 to 0.45), l m varied 10.5-fold (from 0.05 to 0.55), and l b varied 6.2-fold ( from 0.11 to 0.68) across species.Overall, l m (0.38 ± 0.12) was significantly larger than l b (0.34 ± 0.14) and l s (0.28 ± 0.07) (P < 0.05).
To further understand how A was limited by g s , g m , and V cmax among life forms, we grouped the 63 species into 6 life forms: tree, tree/shrub, shrub, grass, vine, and fern.The results showed that there was no significant difference between l s , l m , and l b for trees and tree/shrubs.l m of shrubs and grasses was significantly higher than that of l s and l b (P < 0.05).l m of vines and ferns was significantly higher than that of l s (P < 0.05) (Fig. 5).
3.3 Effect of g s , g m , and V cmax on iWUE iWUE varied from 29.52 to 88.92 µmol CO 2 mol −1 H 2 O.In theory, iWUE is regulated by g s (g sw = 1.6 g s ), g m , and V cmax .However, a simple correlation analysis showed that iWUE was negatively related to g s (Fig. 6b), and not related to A (Fig. 6a), g m (Fig. 6c), or V cmax (Fig. 6d).
A correlation analysis was used to test how g m / g s and V cmax / g s affected iWUE.The results showed that iWUE was positively correlated with g m / g s (Fig. 7a) and V cmax / g s (Fig. 7b).However, there was no significant relationship between g m / g s and V cmax / g s .iWUE was regulated by covariation between g s , g m , and V cmax .A was constrained by g s , g m , and V cmax acting together; however, variability in the relative contribution of these three factors depended on species and habitats (Tosens et al., 2016;Galmes et al., 2017;Peguero-Pina et al., 2017a;Veromann-Jurgenson et al., 2017).A was significantly correlated with g s , g m , and V cmax (Fig. 3a-c).g s was positively related to g m (Fig. S1c), while no relationship was found between the CO 2 diffusion conductance (g s and g m ) and V cmax (Fig. S2).The relative limitations of g s , g m , and V cmax were separated by a quantitative limitation model (Jones, 1985;Grassi and Magnani, 2005).The results showed that l s , l m , and l b of the 63 species varied over a large range (Fig. S3), indicating plants have diverse strategies to coordinate CO 2 diffusion (g s and g m ) and V cmax to maintain relative high A. The order of factor limitations to A was l m > l b >l s (P < 0.05) (Fig. S3).Fur-thermore, we tested the relationship between the relative limitations and the corresponding limitation factors.The results showed that l s , l m , and l b were negatively associated with g s , g m , and V cmax , respectively (Fig. 4).The relationship was also stronger for g m -l m (r 2 = 0.65) than V cmax -l b (r 2 = 0.27) and g s -l s (r 2 = 0.19).g s was more highly correlated with A, while the results showed that A was more limited by g m .This could be due to two possible reasons.Firstly, compared to the linear relationship between A and g s , a nonlinear trend was found between A and g m when g m > 0.4 (Fig. 3a, b).Secondly, leaf structure plays an important role in regulating g m and V cmax and, consequently, in determining A (Veromann-Jurgenson et al., 2017).Negative relationships between A / LMA and LT (r 2 = 0.16, p = 0.002), and A / LMA and LT (r 2 = 0.3, p < 0.001)were also observed (Fig. S4c, d), while A was not correlated with LT and LD (Fig. S4a, b).   ) and (a) the ratio of mesophyll conductance to CO 2 (g m ) to (g s ) (g m / g s ), and (b) the ratio of the maximum carboxylase activity of Rubisco (V cmax ) to g s (V cmax / g s ).
Lines refer to regression line for 63 species.T, TS, S, G, V, and F represent tree, tree/shrub, shrub, grass, vine, and fern, respectively.
The importance of g m in constraining A was variable, and depended on leaf structural traits; only LMA, LT, and LD were analyzed in this study.Large variability in g m has been shown both between and within species with different life forms and habits (Gago et al., 2014;Flexas et al., 2016).Variability in g m in this study is similar to that in global datasets (Gago et al., 2014;Flexas et al., 2016).There was no significant difference among life forms (P > 0.05).Previous studies have confirmed that LMA (Tomas et al., 2013), thickness of leaf cell wall (Peguero-Pina et al., 2017b), liquid phase of mesophyll (Veromann-Jurgenson et al., 2017), cell wall thickness of mesophyll (Tosens et al., 2016;Terashima et al., 2011), and surface area of mesophyll and chloroplast exposed to intercellular space (Veromann-Jurgenson et al., 2017) were the main limitations for g m .The wide variability of g m between different species and life forms in the same ecosystem seems to be related to the diversity of leaf anatomical traits.
No significant difference of LMA, LT, and LD was found among life forms (P < 0.05).Negative correlations of g m (Terashima et al., 2005) or g m / LMA (Niinemets et al., 2009a;Veromann-Jurgenson et al., 2017) with LMA have been reported.In this study, there was a significant relationship between g m / LMA with LMA (P < 0.01); however, no relationship was found between g m and LMA.g m / LMA was significantly negatively related to LD (p < 0.01) (Fig. S5c), and there was also a weak negative relationship with LT (p = 0.06) (Fig. S5d), demonstrating the negative effect of cell wall thickness on g m (Terashima et al., 2006;Niinemets et al., 2009a).The strong investment in supportive structures was the main reason for the limitation of g m on A (Veromann-Jurgenson et al., 2017).However, it is still unknown how leaf anatomical traits affect g m and A, and this should be further explored.
g s is responsible for the CO 2 exchange between atmosphere and leaves, and regulates the CO 2 fixation (A) and water loss (Lawson and Blatt, 2014).The variability of g s was controlled by stomatal anatomy, i.e., stomata density and size, and mesophyll demands for CO 2 (Lawson and Blatt, 2014).However, the stomatal anatomy was not analyzed in this study.We only focused on how the relationship between g s and g m regulates A. A positive relationship has been observed between g s and g m (Flexas et al., 2013).For example, the restricted CO 2 diffusion from the ambient air to chloroplast is the main reason for a decreased A under water stress conditions, due to both the stomatal and mesophyll limitations (Olsovska et al., 2016).g s was significantly positively related to g m for 63 species (P < 0.001, Fig. S1) in this study, and no difference in the slopes of the regression lines between g s and g m was found among life forms, demonstrating that A was regulated by the covariation of g s and g m .However, the variability of g m and l m was larger than g s and l s , respectively (Figs. 1 and S3).
The wide variation in the range of l b (0.11-0.68) highlighted the important role of V cmax in regulating A. V cmax was used to represent the CO 2 demand in the photosynthetic process in this study.The relative contribution of V cmax to A not only depends on C a -C c , but also on leaf nutrient levels.A positive relationship was found between C a -C c and V cmax (Fig. 1d).Furthermore, V cmax / LMA was also co-regulated by leaf nitrogen, phosphorus, and magnesium content (Jing et al. 2018).In addition, V cmax / LMA was negatively related to LT (p < 0.05) (Fig. S6c) and LD (p < 0.05) (Fig. S6d), while V cmax was not correlated with LT and LD (Fig. S6a,b); this demonstrated that leaf structure plays an important role in regulating V cmax .
The trade-off between CO 2 supply (g s and g m ) and demand (carboxylation capacity of Rubisco) can help maintain a relatively high A (Galmes et al., 2017;Saez et al., 2017).In this study, we used V cmax as a proxy for the carboxylation capacity of Rubisco, and the normalized V cmax by A (V = V cmax / A) was significantly negatively correlated with the normalized g t by A (G t = g t / A) (P < 0.001) (Fig. 2c).This indicated that the trade-off between CO 2 supply and demand also existed among different species in the same ecosystems.For the Limonium genus(flowering plants) (Galmes et al., 2017), g t was significantly positively related to Rubisco carboxylase specific activity, and significantly negatively related to Rubisco specificity factor to CO 2 .In the case of Antarctic vascular (Saez et al., 2017) and Mediterranean plants (Flexas et al., 2014), A was mainly limited by low g m , but it could be partially counterbalanced by a highly efficient Rubisco through high specificity for CO 2 .This highlights the importance of the trade-off between CO 2 supply and demand in plant adaptation to karst environment.However, it is still unknown how leaf anatomical traits affect g m , V cmax , and A, and this should be further explored.(Fu et al., 2012;Chen et al., 2015).The average iWUE of 12 vines and 13 trees in the karst tropical primary forest was 41.23 ± 13.21 µmol CO 2 mol −1 H 2 O (Chen et al., 2015), while that of six evergreen and six deciduous trees was 66.7 ± 4.9 and 49.7 ± 2.0 µmol CO 2 mol −1 H 2 O, respectively (Fu et al., 2012).The results demonstrated that karst plants use a diverse strategies of carbon-water regulation to adopt to the harsh karst environment.

Covariation of
Coexisting species have a diverse strategies of carbonwater regulation, ranging from "profligate/opportunistic" to "conservative", which means that their ecophysiological niches are separate (Moreno-Gutierrez et al., 2012;Nie et al., 2014;Prentice et al., 2014).Species with high g s , and low iWUE were defined to have a "profligate/opportunistic" water use strategy, and species with low g s and high iWUE were defined to exhibit a "conservative" water use strategy (Moreno-Gutierrez et al., 2012).Consistent with previous research (Moreno-Gutierrez et al., 2012), coexisting plant species growing in the karst ecosystem had diverse water use strategies.However, karst plants tended to lose more water to gain more carbon, i.e., karst plants used a "profligate/opportunistic" water use strategy to adapt to the low nutrient availability and water stress conditions.Prentice et al. (2014) studied the trade-off between carbon gain and water loss of woody species in contrasting climates, and found that species in hot and wet regions tend to lose more water in order to fix more carbon (high g s / A, low V cmax_Ci / A), and vice versa.Although karst soils cannot contain enough water for plant growth, the trade-off between carbon gain and water loss (high g s / A and low V cmax_Ci / A) was similar to that shown for plants growing in hot and wet regions (Prentice et al., 2014).
iWUE is regulated by the covariation of g s , g m , and V cmax .In theory, water loss is regulated by g s only, while carbon gain (A) is regulated by g s , g m , and V cmax (Fig. 3) (Lawson and Blatt, 2014).However, in this study, iWUE was negatively related to g s (R 2 = 0.30) and V cmax (R 2 = 0.09), and not related to A, g m (Fig. 6).
CO 2 diffusion and the Farquhar biochemical model indicated that iWUE is affected by g m / g s and V cmax / g s (Gago et al., 2014;Flexas et al., 2016).There was a hyperbolic dependency of iWUE on g m / g s due to the roles of g s and g m in C i and C c , and of C c in A (Flexas et al., 2016).In meta-analyses, both Gago et al. (2014) and Flexas et al. (2016) found that iWUE was significantly positively related to g m / g s and V cmax / g s .The results of this study are consistent with the meta-analyses (Fig. 7), demonstrating that plant species with relatively high g m / g s or V cmax / g s had relatively high iWUE.The relationship between iWUE and V cmax / g s (R 2 = 0.50) was stronger than the relationship between iWUE and g m / g s (R 2 = 0.20), demonstrating that iWUE was mainly regulated by V cmax / g s .The reason for this may be that iWUE was correlated to g s and V cmax , and g s was positively related to g m .
However, plants cannot simultaneously have high g m / g s and high V cmax / g s .Similarly to the study by Gago et al. (2014), we found no relationship between g m / g s and V cmax / g s .Gago et al. (2014) thought that the poor relationship between g m / g s and V cmax / g s indicated that the iWUE may be improved by g m / g s or V cmax / g s separately; if both of them were simultaneously improved, an enhanced effect on iWUE could be anticipated.In addition, Flexas et al. (2016) showed in a simulation that the increase in iWUE caused by overinvestment in photosynthetic capacity would progressively lead to inefficiency in the trade-off between carbon gain and water use, causing an imbalance between CO 2 supply and demand.

Conclusions
This study provides information regarding the limitations of A and iWUE by g s , g m , and V cmax in 63 species across 6 life forms in the field.The results showed that plants growing in karst CZs used diverse strategies of carbon-water regulation, but no difference was found among life forms.The covariation of CO 2 supply (g s and g m ) and demand (V cmax ) regulated A, indicating that species maintained a relatively high A through covaring their leaf anatomical structure and V cmax .iWUE was relatively low, but ranged widely, indicating that plants used the "profligate/opportunistic" water use strategy to maintain the survival, growth, and structure of the community.iWUE was regulated by g s , V cmax , g m / g s , and V cmax / g s , indicating that species with high g m / g s or V cmax / g s will have to be much more competitive to respond to the ongoing rapid warming and drought in the karst CZs.

Figure 1 .
Figure1.Relationships between (a) CO 2 gradient between ambient air and intercellular air space (C a -C i , µmol mol −1 ) and stomatal conductance to CO 2 (g s , mol CO 2 m −2 s −1 ); (b) CO 2 gradient between intercellular air space and chloroplasts (C i -C c , µmol mol −1 ) and mesophyll conductance to CO 2 (g m , mol CO 2 m −2 s −1 ); (c) CO 2 concentration gradient between ambient air and chloroplasts (C a -C c , µmol mol −1 ) and total conductance to CO 2 (g t , mol CO 2 m −2 s −1 ); and (d) C a -C c and the maximum carboxylase activity of Rubisco (V cmax , µmol CO 2 m −2 s −1 ).Lines refer to regression line for 63 species.T, TS, S, G, V, and F represent tree, tree/shrub, shrub, grass, vine, and fern, respectively.

Figure 2 .
Figure 2. Relationships between (a) V and G s ; (b) V and G m ; and (c) V and G t .V is the ratio of photosynthetic capacity (V cmax ) to lightsaturated net photosynthesis (A, µmol CO 2 m −2 s −1 ); G s is the ratio of stomatal conductance to CO 2 (g s , mol CO 2 m −2 s −1 ) to A; G m is the ratio of mesophyll conductance to CO 2 (g m , mol CO 2 m −2 s −1 ) to A; and G t is the ratio of total conductance to CO 2 (g t , mol CO 2 m −2 s −1 ) to A. Lines refer to regression line for 63 species.T, TS, S, G, V, and F represent tree, tree/shrub, shrub, grass, vine, and fern, respectively.

Figure 4 .
Figure 4. Relationships between (a) stomatal conductance to CO 2 (g s , mol CO 2 m −2 s −1 ) and l s (g s limitation on light-saturated net photosynthesis; A); (b) mesophyll conductance to CO 2 (g m , mol CO 2 m −2 s −1 ) and l m (g m limitation on A); and (c) the maximum carboxylase activity of Rubisco (V cmax , µmol CO 2 m −2 s −1 ) and l b (V cmax limitation on A).Lines refer to regression line for 63 species.T, TS, S, G, V, and F represent tree, tree/shrub, shrub, grass, vine, and fern, respectively.

Figure 5 .
Figure 5. Limitation to light-saturated net photosynthesis (A) in six life forms by stomatal conductance to CO 2 (l s ), mesophyll conductance to CO 2 (l m ), and the maximum carboxylase activity of Rubisco (l b ).Error bars denominate standard deviation (1σ ).

Figure 7 .
Figure 7.The relationships of the intrinsic water use efficiency (iWUE, µmol CO 2 mol −1 H 2 O) and (a) the ratio of mesophyll conductance to CO 2 (g m ) to (g s ) (g m / g s ), and (b) the ratio of the maximum carboxylase activity of Rubisco (V cmax ) to g s (V cmax / g s ).Lines refer to regression line for 63 species.T, TS, S, G, V, and F represent tree, tree/shrub, shrub, grass, vine, and fern, respectively.
g s , g m , and V cmax in regulating iWUE Compared with the global dataset under well-watered conditions (19.27-171.88µmol CO 2 mol −1 H 2 O) (Flexas et al., 2016), iWUE (52.85 ± 13.08 µmol CO 2 mol −1 H 2 O) was somewhat lower in this study.iWUE varied from 29.53 to 88.91 µmol CO 2 mol −1 H 2 O, and the variability of iWUE was larger than in the karst tropical primary forest