The physiological basis for genetic variation in water use efficiency and carbon isotope composition in Arabidopsis thaliana

Ecologists and physiologists have documented extensive variation in water use efficiency (WUE) in Arabidopsis thaliana, as well as association of WUE with climatic variation. Here, we demonstrate correlations of whole-plant transpiration efficiency and carbon isotope composition (δ13C) among life history classes of A. thaliana. We also use a whole-plant cuvette to examine patterns of co-variation in component traits of WUE and δ13C. We find that stomatal conductance (g s) explains more variation in WUE than does A. Overall, there was a strong genetic correlation between A and g s, consistent with selection acting on the ratio of these traits. At a more detailed level, genetic variation in A was due to underlying variation in both maximal rate of carboxylation (V cmax) and maximum electron transport rate (Jmax). We also found strong effects of leaf anatomy, where lines with lower WUE had higher leaf water content (LWC) and specific leaf area (SLA), suggesting a role for mesophyll conductance (g m) in variation of WUE. We hypothesize that this is due to an effect through g m, and test this hypothesis using the abi4 mutant. We show that mutants of ABI4 have higher SLA, LWC, and g m than wild-type, consistent with variation in leaf anatomy causing variation in g m and δ13C. These functional data also add further support to the central, integrative role of ABI4 in simultaneously altering ABA sensitivity, sugar signaling, and CO2 assimilation. Together our results highlight the need for a more holistic approach in functional studies, both for more accurate annotation of gene function and to understand co-limitations to plant growth and productivity.

Abstract Ecologists and physiologists have documented extensive variation in water use efficiency (WUE) in Arabidopsis thaliana, as well as association of WUE with climatic variation. Here, we demonstrate correlations of whole-plant transpiration efficiency and carbon isotope composition (d 13 C) among life history classes of A. thaliana. We also use a whole-plant cuvette to examine patterns of co-variation in component traits of WUE and d 13 C. We find that stomatal conductance (g s ) explains more variation in WUE than does A. Overall, there was a strong genetic correlation between A and g s , consistent with selection acting on the ratio of these traits. At a more detailed level, genetic variation in A was due to underlying variation in both maximal rate of carboxylation (V c max) and maximum electron transport rate (Jmax). We also found strong effects of leaf anatomy, where lines with lower WUE had higher leaf water content (LWC) and specific leaf area (SLA), suggesting a role for mesophyll conductance (g m ) in variation of WUE. We hypothesize that this is due to an effect through g m , and test this hypothesis using the abi4 mutant. We show that mutants of ABI4 have higher SLA, LWC, and g m than wild-type, consistent with variation in leaf anatomy causing variation in g m and d 13 C. These functional data also add further support to the central, integrative role of ABI4 in simultaneously altering ABA sensitivity, sugar signaling, and CO 2 assimilation. Together our results highlight the need for a more holistic approach in functional studies, both for more accurate annotation of gene function and to understand co-limitations to plant growth and productivity.

Introduction
The efficiency with which plants fix CO 2 relative to their rate of H 2 O loss is called water use efficiency (WUE), and when high, WUE can mitigate the tradeoff between CO 2 uptake and H 2 O loss. In C 3 plants, low stomatal conductance (g s ) minimizes water loss (transpiration, E) and can be a rapid and effective strategy; however, it results in reduced CO 2 uptake (A) and growth (Schulze 1986;Geber and Dawson 1997;Condon et al. 2002). Genetically based variation in WUE has been documented in both crops and non-cultivated species (McKay et al. 2003;Hall et al. 2005). Physiologists are interested in intrinsic WUE (A/g s ) as a tool for studying how the fundamental trade-off of losing water for gaining CO 2 is regulated by stomatal and other physiological adjustments (Buckley and Mott 2002;Comstock 2002). Evolutionary biologists have studied variation in WUE as it is likely an important component of local adaptation (Donovan and Ehleringer 1994;Heschel et al. 2002;Geber and Griffen 2003;Caruso et al. 2005). Likewise, plant breeders have long considered WUE an important target (Passioura 1977).
WUE can be estimated in a variety of ways at various spatio-temporal scales, including with lysimeter studies, gas exchange measurements, or stable carbon isotope composition. Tissue carbon isotope composition is an increasingly popular approach, and its advantages include integration over long periods of gas exchange and development, amenability to high throughput sampling, relatively low cost, and high heritability. Stable carbon isotope composition of leaves (d 13 C) (the ratio of the amount of 13 C to 12 C isotopes in a sample relative to a standard), provides a time-integrated estimate of intrinsic WUE (Farquhar et al. 1989;Dawson et al. 2002). In Arabidopsis thaliana (here after Arabidopsis), common garden experiments have identified substantial variation in d 13 C among natural accessions and some of this variation likely represents local adaptation to climate (McKay et al. 2003Juenger et al. 2005Juenger et al. , 2010Christman et al. 2008;Monda et al. 2011;Des Marais et al. 2012;Lasky et al. 2012). In addition, QTL have been identified for d 13 C (Juenger et al. 2005;Masle et al. 2005;McKay et al. 2008).
In plant breeding, WUE is an important target of selection, although the complexity of the trait, and difficulty of phenotyping has prevented many breeding programs from attempting to select on WUE directly (Araus et al. 2002). Many studies have shown variation in d 13 C among cultivars. In crops, one particularly successful example is an Australian wheat breeding program, where selection on d 13 C in a greenhouse environment led to new varieties that had increased yield in semiarid rainfed conditions . Conversely, in conditions where water is not limiting, selection for reduced WUE may lead to greater yields (Passioura 1977;Fischer et al. 1998).
Although it is heritable, appears to be under selection in nature, and may correlate with yield in C 3 crops (Condon et al. 1987), the mechanistic basis of genetic variation in d 13 C is still unclear. Variation in d 13 C can be due to variation in photosynthetic biochemistry, conductance of CO 2 to the leaf interior and chloroplast, or a combination of these (Seibt et al. 2008). Thus, similar leaf d 13 C and similar WUE can evolve via mutations that cause low A with low conductance or mutations that cause high A with proportionally higher conductance (Farquhar et al. 1989). This is further complicated because conductance from ambient air to the interior of the leaf is influenced both by g s and additional variability of conductance into leaf mesophyll cells and chloroplasts (g m ), which can change over the long-term with leaf morphology (von Caemmerer and Evans 1991;Evans et al. 1994Evans et al. , 2009Tosens et al. 2012) and over the short-term through changes in protein-mediated chloroplast membrane permeability (Flexas et al. 2006;Uehlein et al. 2008;Heckwolf et al. 2011). When examining the combined effects of g s and g m , it is important to recognize that they operate in series rather than in parallel and that the regulation of g m is poorly understood. Within a genotype, g s and g m usually respond in a correlated way to environmental stimuli (Flexas et al. 2007(Flexas et al. , 2008Warren 2008;Barbour et al. 2010) although, opposite responses have also been observed (Galle et al. 2012). Patterns of genetic covariation of g s and g m have not been investigated. However, it is known that variation in g m contributes to leaf carbon isotope discrimination, further increasing the importance of considering g s and g m in interpretations of d 13 C (Warren and Adams 2006; Barbour et al. 2010).
Understanding the physiological basis of variation in d 13 C and intrinsic WUE is important for improving plant productivity and understanding the evolution of wild species. Here, we report a series of experiments designed to investigate a mechanistic understanding of the physiological basis of variation in intrinsic WUE in Arabidopsis. At the coarse level, we can ask if variation in intrinsic WUE is primarily due to variation in A or g s . For example, threefold variation in g s and twofold variation in leaf N concentration among natural accessions of Arabidopsis suggest substantial variation in g s and A may separately or in concert be responsible for the observed variation in d 13 C (Christman et al. 2008;Des Marais et al. 2012). Des  found large differences in physiology between life history classes in Arabidopsis. Although, the Des Marais study focused on variation in gene expression, they also reported constitutive variation in leaf structural traits between life history classes. Winter annual types had higher intrinsic WUE. This is consistent with coordinated selection on WUE, A, and g s and life history observed in other species (Geber and Dawson 1997). Higher WUE was associated with lower leaf water content (LWC) and specific leaf area (SLA) (Des . Taken together, these results suggest that increased leaf density is associated with higher photosynthetic capacity (Terashima et al. 2011), but may come at the cost of lower stomatal and mesophyll conductance to CO 2 (Parkhurst and Mott 1990;Evans et al. 1994;Syvertsen et al. 1995;Kogami et al. 2001).
Studies in Arabidopsis have identified extensive natural variation in plant-water relations and gas exchange physiology (Juenger et al. 2005(Juenger et al. , 2010Masle et al. 2005;Bouchabke et al. 2008;Christman et al. 2008;McKay et al. 2008;Monda et al. 2011;Des Marais et al. 2012;Pons 2012). The present study was undertaken to examine natural variation in leaf physiological traits that are the likely cause of the observed variation in d 13 C and associated WUE parameters in natural accessions of Arabidopsis, and to determine if these traits vary independently or co-vary in a coordinated and predictable manner. First, we tested if the expected relationship between transpiration efficiency (shoot dry mass/transpiration; TE) and leaf d 13 C was present in 96 natural accessions of Arabidopsis. In a smaller set of 18 natural accessions spanning the range of variation in d 13 C, we measured rosette A, g s , and intercellular CO 2 concentration (C i ) and examined the relationship of C i and d 13 C. To further characterize natural variation in A, we examined maximal carboxylation rate (V c max) and photosynthetic electron transport rate (Jmax) in three accessions using photosynthetic carbon dioxide response curves (Sharkey et al. 2007). Additionally, we used gas exchange measurements coupled with online isotopic measurements to determine instantaneous carbon isotope discrimination using tunable diode laser spectroscopy (TDL) (Flexas et al. 2006;Barbour et al. 2007;Heckwolf et al. 2011) to estimate g m in stomatal regulation mutants to investigate the relationship of these mechanistically related traits (Warren et al. 2003;Yamori et al. 2006).

Materials and methods
d 13 C and transpiration efficiency (Experiment 1) Our first goal was to use a relatively high throughput approach to look for variation and co-variation across the species range. 96 natural accessions were selected from the native range of Arabidopsis to evaluate plant biomass production and water use (Nordborg et al. 2005). Individual plants were grown in 250-mL plastic cups, each filled with a standard mass of 1:1 fritted clay and Promix BT potting soil mix. We measured field capacity of the soil mix following a 24-h gravitational drain of saturated soil. Each cup was covered with parafilm and sealed with a plastic lid that had a 6-mm diameter hole. Two replicates of each of 96 ecotypes were planted and cold stratified in the dark for 7 days at 4°C. Plants were grown in two independent growth chambers at 200 lmol m -2 s -1 PPFD in a randomized block design. Photoperiod was 12 h light/12 h dark and the temperature cycled 23/18°C (light/dark). Every 2 days, each container was weighed and additional water was added with a syringe to bring the soil in each container to 90 % field capacity. Total transpiration (E total ) was summed for the 35 days growing period for each experimental plant. Plants were harvested, and aboveground material was oven dried and weighed (DW). We assessed evaporative loss from the containers using ''blanks'' lacking an Arabidopsis plant. Total evaporation from the blank containers was \4 % of the average E total from pots in the experiment. Transpiration efficiency (TE) of each plant was calculated as DW/E total . Dried leaves were ground to a fine powder and d 13 C was determined at the UC Davis Stable Isotope Facility (http://stableisotope facility.ucdavis.edu/). When grown outside in free air, the use of carbon isotope discrimination, D, is preferred ), but when growth chamber and greenhouse studies are included the value of air d 13 C is uncertain and variable, thus requiring the use of leaf d 13 C instead of D. Differences in d 13 C within the same experiment indicate differences in intercellular CO 2 concentration, but d 13 C must be viewed with caution when comparing different experimental conditions.

Whole-shoot gas exchange (Experiment 2)
To follow up on the patterns from the 96 accessions, 18 natural accessions of Arabidopsis were used in whole-shoot gas exchange experiments to evaluate the physiological basis of variation in d 13 C. Eleven of the accessions were spring annuals, and seven were winter annuals. Four replicates of each genotype were grown in a growth chamber in a randomized block design. Each plant was grown in a pot constructed from a 50-mL centrifuge tube with the bottom cut off and ''planted'' in a 164-mL Conetainer TM pots (Stuewe and Sons, Corvallis, OR) filled with a 1:1 mixture of potting mix (Sunshine mix, Sun Gro Horticulture, Bellevue, WA) and fritted clay. After planting, pots were cold stratified at 4°C for 7 days, then transferred to a growth chamber. Photoperiod was 12 h with 350 lmol m -2 s -1 PPFD and temperature was cycled 23/20°C (light/dark).
Instantaneous whole-canopy gas exchange rate was measured using a LI-6400 (Li-Cor Inc., Lincoln, NE, USA) with a custom-made whole-shoot Arabidopsis cuvette (Fig. 1). Cuvette PPFD was maintained at 350 lmol m -2 s -1 PPFD, CO 2 was maintained at 400 lmol mol -1 , and temperature Photosynth Res (2014) 119:119-129  121 and relative humidity were set to growth chamber conditions. Each block was measured on a different day, 28-31 days after sowing. Following measurements for each plant, leaf area was determined from digital photographs of the rosette using Scion Image (Scion Corporation, Frederick, MD, USA). A:C i responses were measured for three accessions (Tsu-1, SQ-8, and Kas-1) which differed in A and d 13 C. Cuvette conditions were the same as above, but light was increased to 1,000 lmol m -2 s -1 PPFD. Photosynthetic carbon dioxide response curves were measured on four rosettes of each accession. The number of replications of A:C i measurements were limited by chamber environment equilibration time at each CO 2 set point. The least squares iterative curve-fitting procedure (Sharkey et al. 2007) model was used to fit Farquhar et al.'s (1980) biochemical model of photosynthesis and obtain maximal carboxylation rate (V c max) and maximal photosynthetic electron transport rate (Jmax).
Leaf water content (Experiment 3) 39 natural accessions from the native range of Arabidopsis previously used in Mckay et al. (2003) were measured for LWC and leaf d 13 C. Four replicates of each ecotype were grown in a greenhouse at UC Davis in a randomized block design. Seeds were sown in 250-mL pots in peat-based potting mix with slow-release fertilizer and vernalized at 4°C for 5 days. Day length was extended to 16 h using supplemental lighting at 350 lmol m -2 s -1 PPFD. Greenhouse mean relative humidity and air temperature were 44 % and 23°C, respectively. Shoots were harvested at the initiation of flowering and shoot fresh weight (FW) was determined, leaf area was determined from scans of dissected rosettes using Scion Image (Scion Corporation, Frederick, MD, USA), and shoots were dried and weighed (DW). Entire dried shoots were ground and processed for carbon isotope analysis at the UC Davis Stable Isotope Facility (http://stableisotopefacility.ucdavis.edu/). LWC (%) was calculated as 100 9 (FW -DW)/DW.

Mesophyll conductance (Experiment 4)
Arabidopsis seeds of ecotype Columbia and the abi4 mutant provided by the Arabidopsis Biological Resource Center (Columbus, OH, USA) were used for leaf mesophyll conductance to CO 2 (g m ) experiments. Seven replicates of each genotype were grown in a growth chamber in a randomized block design. Photoperiod was 12 h with 350 lmol m -2 s -1 PPFD and temperature was cycled 23/20°C (light/dark). A LI-6400 (Li-Cor Inc., Lincoln, NE, USA) with wholeshoot Arabidopsis cuvette ( Fig. 1) was coupled with online isotopic measurements of CO 2 entering and leaving the shoot chamber to determine instantaneous carbon isotope discrimination and g m using TDL (Flexas et al. 2006;Barbour et al. 2007;Heckwolf et al. 2011). Calculations for g m were based on whole-shoot gas exchange measurements at 350, 700, and 175 (lmol m -2 s -1 ) PPFD using the slope-based approach given in Evans et al. (1986). Shoots were harvested after gas exchange, leaf area was determined from rosette photographs using Scion Image (Scion Corporation, Frederick, MD, USA), and shoots were dried and weighed (DW). LWC (%) was calculated as above and SLA was calculated as rosette area/DW.

Statistical analysis
We analyzed phenotypic data for physiological traits using standard fixed effect ANOVAs with the Proc GLM in SAS The cuvette is mounted on the LI-6400 IRGA and cuvette control system (gold-plated panel, fan and aluminum box, upper photograph). This system allows accurate, rapid measurement of CO 2 (A) and H 2 O (E) exchange of whole shoots of Arabidopsis plants. The whole-plant cuvette incorporates a leaf temperature thermocouple that interfaces directly with the LI-6400. Intrinsic WUE (A/g s ), stomatal conductance (g s ), internal CO 2 concentration (C i ), and other variables can be calculated from these measurements. All interior surfaces are Teflon coated or Ni-plated, the cuvette has extremely low leak rates when operated in lab conditions with high external CO 2 , and the circular design provides excellent mixing using the LI-6400 fans. Plants can be rapidly changed using multiple inserts (lower photo) (SAS Institute 1999). We estimated correlations among physiological traits as the standard Pearson product-moment correlation between genotype means.
In the case of the TE experiment, we analyzed phenotypic data for physiological traits using a linear mixed model analysis with the Proc Mixed procedure in SAS (SAS Institute 1999). We fit a model including accessions as a random effect and chamber, experiment, and their interaction as fixed effects. The variance component for the random effect was estimated using restricted maximum likelihood (REML) and assessments of significance were based on likelihood ratio tests (Little et al. 1996). We obtained empirical best linear unbiased predictors (BLUPs) associated with the random effects and consider these breeding values for each accessions. BLUPs are robust estimates of the impact of a particular accession on the measured trait while controlling for the fixed effects (chamber and experimental run). For TE, we fit a model that included both chamber and experimental run as a fixed effect. For d 13 C, we fit a simpler model including accession as a random variable and experimental run as a fixed effect. In this case, factors associated with chamber could not be included because replicates within each experimental run were pooled for mass spectroscopy analysis. All subsequent analyses involving TE and d 13 C rely on BLUP estimates. The TE and d 13 C values were normally distributed and residuals from analyses did not exhibit heteroscedasticity.
We estimated broad-sense heritability by computing the ratio V G /V P , where V G equals the among-accession variance component and V P equals the total phenotypic variance for the study phenotypes. We estimated genetic correlations (r G ) among TE and d 13 C as the standard Pearson product-moment correlation between genotype means or BLUPs.

Results and discussion
Variation in TE and d 13 C The 96 natural accessions of Arabidopsis in experiment 1 (Table 1) exhibited considerable variation in time-integrated measures of water use efficiency, i.e., whole-plant TE and d 13 C. We observed a 3.33 g kg -1 and 5.12 % range of variation in TE and d 13 C among accessions, respectively, (TE mean = 2.02 ± 0.28 g kg -1 ) (d 13 C mean = -30.64 ± 0.90 %). In both cases, we observed significant broad-sense heritability (TE, H 2 = 0.09, accession P = 0.031; d 13 C, H 2 = 0.667, accession P = 0.001). For the experiment 1, we found replication block, growth chamber, and their interaction were significant sources of environmental variation in TE (in all cases, P \ 0.005). Likewise, we found that the replication block was a significant source of environmental variation for d 13 C (P \ 0.0001). Despite the low heritability of the TE data, our experimental design and analysis allowed us to estimate breeding values as BLUPs. Spring accessions fit the expected positive relationship between TE and d 13 C (r G 2 = 0.265, P \ 0.0001, Fig. 2). The winter annuals had greater intrinsic WUE as indicated by d 13 C than the spring annuals, but this was not related to TE (r G 2 = 0.011, P = 0.531, Fig. 2). Together these data suggest that variation in d 13 C is likely due to stomatal limitations (on C i ) in the spring accessions, but in winter accessions, other mechanisms (like g m ) not affecting water loss may be leading to variation in d 13 C (Seibt et al. 2008). Alternatively, variation in root carbon allocation unaccounted for in TE may explain the observed pattern in winter accessions. In principle, the greater belowground allocation in winter accessions could result in lower TE without affecting d 13 C, but this hypothesis remains to be tested.

Variation in components of WUE
The 18 natural accessions of Arabidopsis in experiment 2 were selected to represent a wide range of intrinsic WUE as indicated by d 13 C (Table 1). Whole-plant gas exchange measurements in a custom cuvette (Fig. 1) showed that these lines also exhibit considerable variation in whole rosette A and g s in a common environment (Fig. 3). Accession mean whole rosette A ranged between 10 and 16 lmol m -2 s -1 , but the heritability was not significantly different from zero (P = 0.137). g s showed significant genetic variation, ranging between 0.17 and 0.45 mol m -2 s -1 with a heritability of H 2 = 0.33 (accession P value = 0.002). In addition, g s was a better predictor of variation in d 13 C than A. We found a significant negative correlation between d 13 C and g s among accessions (r 2 = 0.40, P = 0.0027), and a weaker correlation between d 13 C and A (r 2 = 0.25, P = 0.036). In general, the high conductance lines had low intrinsic WUE, as indicated by d 13 C, but there was a wide range of d 13 C in the low conductance lines, suggesting additional sources of variation. The expected negative correlation between d 13 C and g s was largely caused by the spring accessions. The winter accessions tended to show the opposite pattern (not significant), with the exception of Tamm-2, an accession from Finland that had the highest g s of all.
Despite the lack of heritability of A and the weak correlation of A with d 13 C, we did find a significant positive correlation between g s and A among accessions (r 2 = 0.78, P = 0.00001). This is consistent with the optimization of stomatal regulation to maximize carbon gain while minimizing the water loss (Katul et al. 2010). Accessions that have high conductance should be under selection for increased biochemical capacity (Bloom et al. 1985).
Although, it is not formally stated, such optimality approaches interpret consistent patterns of correlation in physiological traits (Reich et al. 1997) as evidence of selection optimizing their ratios or covariances (Donovan et al. 2011). Under such a scenario, selection would favor mutations that lead to a co-limitation of g s and RuBP utilization and regeneration.
In general, winter Arabidopsis accessions had lower g s and A than spring Arabidopsis accessions. Across accessions there was large variation in C i /C a , but it was only weakly related to d 13 C (Fig. 4). No consistent difference in C i /C a was seen between the winter and spring annuals.
The overall finding of experiment 2 was that accessions with low g s and high d 13 C had lower A compared to low d 13 C accessions. Overall, these data are consistent with large effects of g s on d 13 C, but the weaker correlation of C i and d 13 C suggest a more complex mechanism than predicted by theory. To better understand processes limiting photosynthesis in Arabidopsis accessions, we conducted detailed CO 2 response curves of assimilation for low and high WUE spring accessions Tsu-1 and SQ-8 and high WUE winter accession Kas-1. Maximum carboxylation rate of rubisco (V c max) was higher in low WUE Tsu-1 (d 13 C = -29.7) than Sq-8 (d 13 C = -28.6) (P = 0.01), as expected (Fig. 5). Similar, maximal photosynthetic electron transport (Jmax) was also higher in Tsu-1 than Sq-8 or Kas-1 (d 13 C = -28.8) (P = 0.002, P = 0.002).
The major biochemical limitations to photosynthesis, V c max and Jmax, appeared optimized to accessions' C i as indicated by d 13 C. V c max and Jmax were lower in low g s , high WUE accessions operating at lower C i . The higher ratio of V c max to Jmax in Kas-1 compared to Sq-8 suggests a lack of limitation by Jmax under the low g s typical of Kas-1. Simultaneous changes in V c max and Jmax are consistent with a limitation of photosynthesis by RuBP utilization and regeneration (Farquhar and Sharkey 1982). Likewise, proportional changes in components of photosynthetic apparatus and g s suggest acclimation of these processes are closely coupled (Cowan 1986).   In experiment 3, we examined 39 natural accessions of Arabidopsis for variation in d 13 C and LWC (Table 1). We found a significant negative correlation between d 13 C and LWC among accessions (r 2 = 0.6, P \ 0.0001). Spring accessions tended to have higher LWC and lower WUE, as indicated by d 13 C, than winter accessions. Accession differences in LWC most likely result from the effect of mesophyll cell wall thickness on leaf density and not  Fig. 5 Maximum carboxylation rate of rubisco (V c max) and maximal photosynthetic electron transport (Jmax) obtained from photosynthetic carbon dioxide response curves in three accessions (Tsu-1, Sq-8, and Kas-1) which differed in A. Each bar represents the mean ± SE (n = 4) for each accession. Letters represent significant differences among accessions. Genotype F-ratio = 12.14 and P = 0.0078 for V c max. Genotype F-ratio = 11.01 and P = 0.0098 for Jmax Photosynth Res (2014) 119: 119-129 125 differences in water potential as plants in experiment 3 were not water stressed (Garnier and Laurent 1994;Evans et al. 1994). Leaf anatomical traits such as leaf and cell wall thickness, surface area of mesophyll cells exposed to internal air spaces, and the location of chloroplasts within those cells was initially shown to correlate with g m several decades ago (von Caemmerer and Evans 1991;Evans et al. 1994). In particular, mesophyll cell wall thickness was shown to negatively affect g m . Therefore, high LWC accessions should have thinner mesophyll cell walls resulting in high g m and more negative d 13 C (Evans et al. 1994), which is consistent with our data. These ideas have been revisited recently and the importance of the cell wall properties (thickness and water content) and the coverage of air exposed surfaces of mesophyll cells by chloroplasts is receiving more attention (Evans et al. 2009;Tholen and Zhu 2011;Tosens et al. 2012). Direct measurement of leaf thickness and density may explain some of the variation in g m and d 13 C among plants with similar LWC values (Fig. 6). Alternatively, variation in COO-porin content or activity could be responsible for the g m and d 13 C variation in plants with LWC. Recent studies have found a significant role for chloroplast membrane CO 2 transporting aquaporins (COO-porin) has been demonstrated and provides a clearly heritable mechanism for both rapid and sustained adjustment of g m (Flexas et al. 2006;Uehlein et al. 2008Uehlein et al. , 2012Heckwolf et al. 2011). We have found strong correlations between LWC, A, and g s , so focusing on plants with similar LWC should limit the influence of those factors on variation in d 13 C and increase the relative influence of g m from cell wall properties or COO-porin content or activity on d 13 C variation.
The ABI4 transcription factor causes changes in leaf anatomy and mesophyll conductance To further test for a causal effect of leaf anatomy on gas exchange (experiment 4 in Table 1), we used abi4, a mutant of locus AT2G40220, which is an AP2/ERF transcription factor (TF). ABI4 is closely related to the DREB2 TFs and the mutant was initially described as ABA insensitive based on a germination screen (Finkelstein 1994). Subsequent work has shown that the transcript is expressed in seedlings (Soderman et al. 2000) and fully developed rosette leaves (Finkelstein et al. 1998). In addition to its key role in ABA signaling, further characterization of this transcription factor has proposed a large and diverse set of functions including sugar signaling and response (Husijer et al. 2000;Bossi et al. 2009), and root development (Signora et al. 2001;Shkolnik-Inbar and Bar-Zvi 2011). There are hundreds of loci whose expression is altered in the ABI4 mutant (Kerchev et al. 2011). Given that it is a transcription factor, this is not surprising, but does illustrate the challenge of functional annotation of such pleiotropic loci. abi4 had higher SLA and LWC than wildtype, revealing a novel effect of this TF on leaf anatomy. In addition, abi4 had increased g m and more negative d 13 C, consistent with the idea that SLA causes variation in d 13 C via effects on g m (Fig. 7). The correlation of SLA, A, and g s with LWC helps to explain why LWC is strongly correlated with leaf gas exchange, i.e., LWC appears to be an inverse proxy for cell wall thickness. When taken together, our data show that Arabidopsis leaves trade-off high WUE for low A, by trading off leaf anatomy based diffusional CO 2 limitation with water loss through stomata. Essentially, plants with the highest A achieve this via the combination of high g s and thin leaves (high SLA). High g s keeps C i high and the thin leaves have cells with thin walls. Thin walls increase g m and keeps CO 2 concentration at the sites of carboxylation (C c ) high (Evans et al. 1994). Conversely, when photosynthesis is directly limited by the combination of cool winter temperatures and high light through effects on electron transport, then low g s would be selected for to improve WUE. We hypothesize that thicker leaves would provide more internal shading and more efficient light use, further decreasing g m and C c explaining the winter annual phenotype.
Although, a few of the AP2/ERF transcription factors in Arabidopsis have been the subject of detailed study, there are 122 of these loci in Arabidopsis (Nakano et al. 2006) and much remains unknown about their function. Recent studies have revealed increasingly complex roles for members of this transcription factor family. For example, a recent study identified eight AP2/ERFs induced by photorespiration (Foyer et al. 2012). This, combined with the  Fig. 6 Relationship between leaf water content (LWC) and leaf carbon isotope composition (d 13 C) among 39 accessions of Arabidopsis thaliana. Open and filled symbols represent spring and winter accession means, respectively. Line represents linear regression; r 2 and P values are given known roles of ABI4 in sugar signaling to photosynthesis including repression of RBCS (Van Oosten et al. 1997;Teng et al. 2008), and our results showing effects on leaf density and g m , are expanding this picture.

Conclusions
Detailed measurements on a diverse set of accessions detail the traits underlying natural variation in intrinsic WUE and carbon isotope composition. Previous studies have shown that spring accessions have lower intrinsic WUE than accessions with winter life histories. Proportional changes in A, g s , V c max, and Jmax suggest acclimation of these processes are closely coupled. We also show strong covariation between LWC and d 13 C, where spring annuals tend to have higher LWC and lower intrinsic WUE. We hypothesize that this is due to an effect through g m , and test this hypothesis using the abi4 mutant. The abi4 mutant shows increased SLA and reduced g m compared to the wildtype, consistent with the pattern of covariance found in the natural accessions.
Previous separate studies in Arabidopsis have addressed variation in d 13 C, plant-water relations, leaf anatomy, and photosynthetic capacity and limitations, including g m . Here, we use a whole canopy approach to examine variation and covariation in all of these components. As predicted by optimality, these traits are not independent, but instead covary as would be expected if selection and photosynthetic acclimation favors states of colimitation. In addition, we show that perturbation of a single transcription factor leads to this trait covariance. This emphasizes the need for whole plant approaches and high dimensional phenotyping to accurately annotate the gene function.  Fig. 7 Comparison of specific leaf area (SLA), leaf water content (LWC), mesophyll conductance (g m ), and leaf carbon isotope composition (d 13 C) between abi4-1 and Columbia (Col) wildtype. Each bar represents the mean ± SE (n = 7) for each genetic line. P \ 0.05 for g m , SLA, LWC, and d 13 C Photosynth Res (2014) 119:119-129 127