CO2 availability influences hydraulic function of C3 and C4 grass leaves

Carbon limitation and water use characteristics impact the relative performance of C3 and C4 grasses at past and future CO2


Introduction
C 4 photosynthetic pathways have evolved as solutions to photosynthetic inefficiencies linked with the oxygenation reaction of Rubisco (Sage, 2004;Sage et al., 2012). Because of its potential for greater efficiency, engineered C 4 photosynthesis has been proposed as a potential solution for improving global food security (von Caemmerer et al., 2012), and C 4 crops are leading contenders as sources of renewable biomass energy (Byrt et al., 2011). Our understanding of C 4 photosynthesis as an ecological adaptation is continuing to develop rapidly (Edwards and Smith, 2010;Lundgren et al., 2015;Atkinson et al., 2016;Watcharamongkol et al., 2018). New insights into the timing and sequence of C 4 evolution from phylogenetic studies have renewed debate about its expected physiological advantages (Edwards and Still, 2008;Christin et al., 2011bSage et al., 2011). Some C 4 lineages probably arose during the Oligocene (~30 million years ago) but most arose over the last 20 million years, during the subsequent Miocene (Christin et al., 2008(Christin et al., , 2011aVicentini et al., 2008;Besnard et al., 2009; but see Kadereit et al., 2012). During this period, 'icehouse' conditions of globally cooler temperatures and drier climates were linked with atmospheric CO 2 concentrations (c a ) lower than present day (Pagani et al., 2005). Atmospheric CO 2 has increased since the last glacial period, and consequent increases in photosynthetic water use efficiency have been associated with declines in water stress and improvements in plant productivity (Polley et al., 1993;Mayeux et al., 1997). It has been speculated that in addition to impacts on photosynthetic performance (Ehleringer et al., 1997), hydraulic function in C 3 and C 4 plants was differentially affected at low atmospheric CO 2 (Osborne and Sack, 2012) because of greater stomatal opening in C 3 plants resulting in greater water stress (Polley et al., 1993). It is also expected that under future, high CO 2 climates, the combination of CO 2 fertilization and improved water use efficiency will continue to influence the relative performance of C 3 and C 4 plants (Ghannoum et al., 2000;Ainsworth and Long, 2005;Leakey, 2009). To establish whether C 4 plants gain hydraulic advantages because of relatively small increases in stomatal conductance (g s ) responding to c a (Osborne and Sack, 2012), it is important to verify relative stomatal responses experimentally and investigate their impact on physiological function, including hydraulic properties.
Photosynthesis in C 4 leaves is characterized by biochemical pumps that initially combine phosphoenolpyruvate (PEP) and CO 2 to form C 4 acids and subsequently transfer those acids, release CO 2 in the presence of Rubisco, and recycle PEP (Edwards et al., 2001;Sage, 2004). The initial biochemical step used to form C 4 acids is highly efficient, and a high CO 2 concentration at the site of Rubisco carboxylation minimizes photorespiration in C 4 plants. Therefore, leaf internal CO 2 concentrations (c i ) are lower for CO 2 compensation and photosynthetic saturation, and quantum yield can be greater (Pearcy and Ehleringer, 1984). Importantly for plant hydraulics, photosynthetic water use efficiency is consequently high (Pearcy and Ehleringer, 1984;Long, 1999). A central question has been whether this improved water use efficiency provides advantages for C 4 plants over C 3 plants in habitats with restricted water availability (Osmond et al., 1982;Hattersley, 1983;Pearcy and Ehleringer, 1984). Recent comparative studies of the numerous C 4 lineages in the grass family have supported the idea that their evolution and maintenance were often linked with improved performance in drier or more open habitats compared with C 3 sister groups (Osborne and Freckleton, 2009;Edwards and Smith, 2010;. Osborne and Sack (2012) proposed that improved hydraulic safety, afforded by the evolution of lower g s among C 4 species (Osmond et al., 1982;Taylor et al., 2010Taylor et al., , 2012, might have increased the potential of C 4 grasses to colonize drier habitats when c a was lower than it is today. They also noted that g s is usually higher at glacial c a compared with ambient c a , but the increase in g s is less at glacial c a in C 4 plants than in C 3 plants. Using steadystate models of coupled photosynthesis and plant hydraulics, they showed that lower g s could have protected C 4 plants from loss of hydraulic conductivity and allowed net CO 2 assimilation (A) to be maintained as soil dried at low c a .
They therefore proposed that in addition to biochemical advantages supporting higher A at low c i , protection of hydraulic function was an important advantage to C 4 grasses at low c a .
Importantly, the models that Osborne and Sack (2012) used to predict hydraulic performance in C 3 and C 4 species at glacial c a did not predict potential adjustments to co-ordination of leaf gas exchange and hydraulic function at low c a . Although evidence suggests that in non-woody species, decreased g s at elevated c a is associated with less negative leaf water potentials, lower hydraulic conductivity, and greater resistance to embolism, little is known about the influence of c a on co-ordination between photosynthetic capacity and hydraulic function (Domec et al., 2017). Changing irradiance results in parallel changes in leaf hydraulic conductance (K leaf ) and photosynthetic capacity of woody C 3 plants, optimizing leaf hydraulic function (Brodribb and Jordan, 2011;Carins Murphy et al., 2012). In contrast, adjustment to high vapour pressure deficit (VPD) is linked with closure of stomata to protect hydraulic function (Carins Murphy et al., 2014). In the case of c a , hydraulic demand is influenced by changes in g s that compensate for carbon availability . Because the economics of leaf structure-function relationships may depend on c a , it is likely that c a has complex effects on co-ordination between K leaf and g s . For instance, smaller, more densely packed stomata are sometimes observed at low c a (Woodward, 1987;Woodward and Bazzaz, 1988;Franks and Beerling, 2009b), which may increase the sensitivity of g s to VPD (Franks and Beerling, 2009a;Drake et al., 2013), serving a protective function. Conversely, higher anatomical maxima for g s observed at low c a in sunflower, which were a result of larger, more densely packed stomata, were linked with greater xylem-specific conductivity, but the phloem ratio and hydraulic safety were decreased (Rico et al., 2013). Photosynthetic type may further affect the impact of c a on the relationships between hydraulic supply and demand because the carbon assimilation advantage provided by C 4 photosynthesis may support additional flexibility in hydraulic adjustment. At ambient CO 2 , relative to C 3 species, C 4 dicots maintain A at relatively lower g s , and either increase hydraulic safety by decreasing xylem conduit diameter, or display greater leaf area for similar investments in stem xylem supply Sage, 2003, 2004;Kocacinar et al., 2008).
In grasses, leaf hydraulic performance is particularly important: leaves contribute 50-72% of resistance along whole-plant hydraulic pathways Martre et al., 2001). The relative sensitivities of K leaf and g s are also crucial in determining water use strategies among grasses. Both C 3 and C 4 grasses have been reported to show routine diurnal declines in leaf hydraulic conductivity when stomata do not close sufficiently to protect hydraulic function Holloway-Phillips and Brodribb, 2011b). Susceptibility to declines in conductivity is variable both among species and among cultivars (Holloway-Phillips and Brodribb, 2011a), and nocturnal root pressure and refilling of embolized vessels facilitates recovery from diurnal stress in some grass species (McCully, 1999;Holloway-Phillips and Brodribb, 2011b;Cao et al., 2012;Gleason et al., 2017). Protection against runaway declines in K leaf can be provided by stomatal closure (Brodribb and Holbrook, 2003), and fast stomatal responses are considered a key characteristic of grasses (Hetherington and Woodward, 2003;Franks and Farquhar, 2007). Faster stomatal responses to light can improve intrinsic water use efficiency (iWUE=A/g sw , where g sw is g s for water) by producing a better match between rapid photosynthetic responses and the slower stomata, which may improve overall water use efficiency, resulting in greater conservation of soil water and thereby decreased hydraulic stress (Lawson and Blatt, 2014).
Our goal was to determine whether growth c a had different impacts on leaf function in selected C 3 and C 4 annual grasses comparable with crop species. We predicted that to support increased transpiration at low c a , K leaf would increase and turgor loss points would decrease to compensate for increased hydraulic demand. In addition, we determined whether rates of stomatal closure, responding to low light, increased at low c a . We anticipated that leaf mass per area (LMA) would decrease in plants with carbon limitation at low c a , and that decreases in iWUE and the extent of carbon limitation imposed by low c a would be greater for C 3 than for C 4 species (Osborne and Sack, 2012). We therefore expected that leaf physiological responses to a range of c a would be larger in C 3 than in C 4 grasses. Plants were grown in CO 2 concentrations that represented: some of the lowest c a conditions (~200 µmol mol −1 ) that occurred in the glacial period during which C 4 grass lineages diversified ; ambient c a (400 µmol mol −1 ); and superambient c a (640 µmol mol −1 ).

Growth conditions
Plants were grown in walk-in climate-controlled growth chambers (Biochambers, Winnipeg, Manitoba) equipped with additive CO 2 , and CO 2 scrubber equipment. Three c a treatments were imposed: glacial (c GLA ), 204 ± 27 μmol mol −1 ; ambient (c AMB ), 408 ± 11 μmol mol −1 ; and super-ambient (c SUP ) 640 ± 2 μmol mol −1 (mean ±SD; 72 daily means). The c AMB and c SUP treatments were rotated between cabinets 1 week prior to the first measurements, during the fourth week after sowing. The c GLA treatment was maintained in a single cabinet throughout the experiment because of the technical demands of obtaining a stable CO 2 concentration at glacial c a . Growing conditions were set to a night-time temperature of 18 °C and a daytime temperature of 26 °C, resulting in a daily mean temperature of 22 °C (mean ±SD for 72 daily means: c GLA , 21.9 ± 0.23; c AMB , 22.1 ± 0.21; and c SUP , 22.1 ± 0.28).
Temperatures and light levels were ramped daily in two even steps between 06.00 h and 08.00 h and between 18.00 h and 20.00 h (14 h light:10 h dark). Light was supplied by HID lamps, which provided photosynthetic photon flux density (PPFD) at the top of the canopy that varied within each cabinet between 300 μmol m −2 s −1 and 650 μmol m −2 s −1 ; daily quantum inputs were ~21 mol m −2 d −1 (mean ±SD for 72 daily means: c GLA , 21.3 ± 1.8; c AMB , 21.5 ± 1.9; and c SUP , 21.5 ± 2.0). Mean daily values for relative humidity ranged from 60% to 83% (mean ±SD for 72 daily means: c GLA , 77 ± 8; c AMB , 78 ± 1; and c SUP , 76 ± 6), providing VPDs of ~0.74 kPa under daytime conditions, and ~0.46 kPa during the night.

Plant material
Our study plants were eight annual grass species, four C 3 and four C 4 , used as food crops or close relatives of species used as food crops. In Poacaeae, all C 4 grasses belong to a clade referred to as PACMAD (Aliscioni et al., 2012); within PACMAD two C 4 crop species have been domesticated from wild relatives in the Chloridoideae (teff and finger millet), and several from the Panicoideae subfamily . Grasses with C 3 photosynthesis used as grain crops originate in the subfamilies Pooideae and Oryzoideae, which belong to a separate clade currently referred to as BEP (Kellogg, 1998;Aliscioni et al., 2012). Relevant Chloridoideae species could not be obtained, so we only used C 4 grasses from the Panicoideae. Sorghum bicolor (great millet), Setaria italica (foxtail millet), and Digitaria exilis (fonio millet) represent independent evolutionary origins of the NADP-malic enzyme (NADP-ME) subtype of C 4 photosynthesis (Aliscioni et al., 2012); Panicum miliaceum (proso millet) represents the NAD-ME C 4 subtype (Giussani et al., 2001;Aliscioni et al., 2012). C 3 species were Panicum bisulcatum and Steinchisma laxa (two wild relatives from Panicoideae), Triticum turgidum (durum wheat, Pooideae), and Oryza sativa ssp. japonica (rice, Ehrhartoideae; Table 1).
Plants were grown from seed in Osmocote Professional Seed Raising & Cutting Mix (Scotts Australia Pty Ltd, Bella Vista, NSW) in 0.55 litre plastic square tubes (Garden City Plastics, Somersby NSW: top dimension 70 × 70 mm, 160 mm deep). Seeds were sown directly into six pots and germinated under the different CO 2 treatments; the number of plants per pot and the size of plants varied depending on the species. To allow for balanced sampling and to account for within-cabinet variability, at germination, pots were arranged into a fully randomized blocked design with one pot from every species in each block. The pots were checked daily and watered as necessary to prevent surface drying. To minimize root binding, roots were allowed to grow out of pots into a layer of wetted Scoria. To minimize nutrient limitation, plants were fed with a complete fertilizer (Thrive All Purpose Soluble Plant Food, Yates, Auckland, New Zealand) every 2-3 weeks during the course of the experiment.

Steady-state gas exchange and stomatal response to PPFD
We measured gas exchange using six LI-6400XT photosynthesis systems (LI-COR Inc., Lincoln NE, USA) equipped with CO 2 mixers (LI-6400-01) and 2 × 3 cm red-blue LED light sources (LI-6400-02B). Pairs of LI-6400XT machines were randomly allocated to the three c a treatments and were rotated every 2 d: each pair of machines was used to measure two of the six blocks in every cabinet over the course of the experiment. Measurements were made under the growth conditions. To minimize disruption of c a treatments, the cuvette and integrated gas analysers of the LI-6400XT were placed inside the growth chambers and consoles outside the growth chambers (growth chambers were opened briefly before and after switching leaves). Measurements were conducted on the mid-section of individual, recently expanded leaves inserted parallel to the long axis of the 2 × 3 cm chamber, and leaf areas were calculated as cuvette length×average leaf width, measured to the nearest 0.5 mm with a ruler. Leaves were allowed to come to steady state [showing no systematic trends with a coefficient of variation (CV) <0.1 over a 5 min period] at a PPFD of 500 μmol m −2 s −1 (growth light levels) and cuvette CO 2 concentrations matched to c a at the time of measurement: (c GLA , 184 ± 4 μmol mol −1 ; c AMB , 406 ± 5 μmol mol −1 ; and c SUP , 647 ± 6 μmol mol −1 ; mean ±SD ≥43 leaves, CV for individual leaves <5%). Relative humidity was maintained at ~70% and block temperature at 26 °C, resulting in leaf VPDs of 1 ± 0.07 kPa (mean ±SD, n=137 leaves; CV for individual leaves <8.1%). An auto-program (logging every 10 s) was used to record initial steady-state values for gas exchange (A, g sw , and iWUE), followed by the response of g sw to a step-change decrease in light availability from 500 μmol m −2 s −1 to 100 μmol m −2 s −1 PPFD. The rate of stomatal response to PPFD (Δg sw /Δt) was characterized based on the magnitude (Δg sw ) and duration (Δt) of the initial decrease in g sw .

Leaf hydraulic conductance and LMA
Because C 3 and C 4 species differed in the response of A to c a between c GLA and c AMB , we determined K leaf in those treatments using the evaporative flux method (Sack et al., 2002). Cut stems were transported to the lab in water, where flag or second-leaf laminas were excised and, using parafilm, were sealed onto parafilm-wrapped, cylindrical plastic rods. The rod and leaf were submerged, and the leaf re-cut and positioned in waterfilled Tygon tubing linked to a reservoir of de-gassed Milli-Q water on a balance (CPA225D, Sartorius, Göttingen, Germany; 10 µg accuracy). The seal was tested by pressure from a 100 ml syringe that was used to fill and empty the system and was attached to the Tygon tubing using Luer fittings. Leaves were supported using fishing line stretched across a wooden frame, with their adaxial surface uppermost and parallel with the meniscus in the reservoir. Transpiration was induced using a desk fan and a lamp (leaf surface PPFD 100-150 μmol m −2 s −1 ). Every second, output from the balance was logged to a computer and plotted to determine when transpiration (E, mol m −2 s −1 ) obtained a steady state for 5 min. At steady state, the temperature of the abaxial surface of the leaf (T leaf ) and air (T air , shaded) were established using two type-K thermocouples and a Pico TC-08, then the leaf was immediately sheathed in plastic and cut at its base. After 15 min equilibration, a Scholander pressure bomb (PMS 1505D with grass compression gland; PMS Instrument Company, Albany, OR, USA) was used to determine water potential, which estimated the leaf hydrostatic gradient (ΔΨ). Leaf area for these leaves (and an extra set collected from the c SUP treatment) was determined using a Canon LiDE 510 flatbed scanner and ImageJ software (Abràmoff et al., 2004). We calculated K leaf as E/(area×ΔΨ). LMAs (g m −2 ) were calculated using dry masses determined after a minimum of 48 h drying at 65 °C.

Pressure-volume relationships
Pressure-volume (P-V) relationships were determined using bench-drying. On the morning of measurement, attached flag leaves were sealed into plastic bags containing exhaled breath and were allowed to equilibrate for a minimum of 40 min to quench transpiration and ensure high turgor. Leaves sheathed in this manner were subsequently excised at the base of the lamina and moved to the laboratory. Initially, leaves remained sealed in plastic between measurements of fresh mass (FM, g) and water potential (Ψ; Scholander pressure bomb). As water potential declined, leaves were occasionally removed from the plastic for short periods to increase the rate of drying. A minimum of 20 min equilibration was ensured between pressure bomb measurements. At the conclusion of FM and Ψ measurements, leaves were dried for a minimum of 48 h at 65 °C to determine dry mass (DM, g). The turgid mass (TM, g) was estimated by extrapolation of the initial linear FM-Ψ relationship (Kubiske and Abrams, 1990) and used to calculate relative water contents (RWC S ) for entire leaves as (FM−DM)/(TM−DM), and leaf dry matter content (LDMC=DM/TM).
We optimized parameter selection for P-V relationships of individual leaves by minimizing the absolute difference between estimates of osmotic potential at full turgor (π 0 , MPa) obtained below (π 0,1 ) and above (π 0,2 ) turgor loss, comparing all possible combinations that could be fit for each leaf within our data set. First, below-turgor loss fits for 1/Ψ=a(1− RWC)+π 0,1 (linear regression with slope a, and y-intercept π 0,1 ) were obtained from all sequences representing at least three of the smallest RWC values and excluding two or more of the highest RWC values. Next, the x-intercept of the below-turgor loss relationship (apoplastic fraction, a f ) was used to establish the RWC of the symplasm [RWC S =(RWC−a f )/ (1−a f )]. Then the osmotic potential, π=1/[a(1−RWC S )+π 0,1 ] (MPa) and turgor pressure, Ψ P =Ψ−π (MPa) were derived for the complementary above-turgor loss data. Finally, the bulk modulus of elasticity (ε, MPa) and π 0,2 were obtained from linear regression of Ψ P = −ε(1-RWC S )−π 0,2 . Turgor loss point characteristics were calculated for the pair of linear relationships where [π 0,1 −π 0,2 ] was smallest, and 0<a f <1. Using π 0,1 to estimate π 0 , the RWC at turgor loss (RWC TLP ) was established by determining the RWC S at which Ψ P =0, and was used to predict osmotic potential at turgor loss (π TLP , MPa) from the equation for π.

Statistical analysis
Statistics were calculated using R Language and Environment (R Core Team, 2016; https://www.R-project.org/). We fit linear models (lm) of species×c a responses, which we minimized using the Akaike information criterion (AIC). Where species effects were significant (or marginally so), we used linear contrasts to compare species means by photosynthetic type. Tests of photosynthetic type effects are approximate because the C 3 /C 4 comparisons were not phylogenetically independent, and only a small number of species were included in our experiment. To adjust for heteroskedasticity, log e transformation was applied to values for g sw , iWUE, and K leaf . Heteroskedasticity in Δg sw /Δt and P-V parameters could not be eliminated using transformation. For these parameters, we applied non-parametric Kruskal-Wallis tests, and present median values.

Results
Impact of c a on steady-state leaf gas exchange iWUE was more responsive to the difference between c GLA and c AMB than the difference between c AMB and c SUP (Fig. 3A; especially on a relative scale, Table 2). At c GLA , iWUE was 33-74% lower (mean 56%) than at c AMB , while iWUE at c SUP was 4-56% higher (mean 30%) than at c AMB . Differences in iWUE were significant for comparisons between c GLA and c SUP in every species (Tukey's HSD, P≤0.0016), and between c GLA and c AMB in every species except T. turgidum (Tukey's HSD, P≤0.0001; T. turgidum, P=0.11); differences between c AMB and c SUP were not significant (P≥0.077).
C 4 species showed much larger absolute decreases in iWUE than C 3 species (Fig. 1A), linked with a marginally significant contrasts term for photosynthetic type×c a (t 113 , P=0.048). At every c a , iWUE was always higher in C 4 species (range 51-232 μmol mol −1 ) compared with C 3 species (13-68 μmol mol −1 ; Fig. 1A). Relative changes in iWUE were not significantly different between C 3 and C 4 grasses, but this comparison was strongly influenced by C 3 T. turgidum (Table 2). Triticum turgidum showed only a small reduction in iWUE from c AMB to c GLA and, surprisingly, decreased g sw at c GLA compared with c AMB (Fig. 1B). The remaining three C 3 grasses showed larger relative decreases in iWUE (63-74%) than any of the four C 4 grasses (47-60%; Table 2).
Under the light conditions provided by our controlledenvironment cabinets, which were non-saturating for photosynthesis (maximum 650 μmol m −2 s −1 PPFD), A was not higher in C 4 species compared with C 3 species at c AMB and c SUP (Fig. 1B), so the higher iWUE of C 4 species at those CO 2 concentrations was primarily due to lower g sw (Fig. 1C). Higher iWUE among C 4 grasses at c GLA was primarily due to smaller relative reductions in A between c AMB and c GLA for C 4 grasses (5-13%) compared with C 3 grasses (36-49%); over the same range, relative increases in g sw were comparable for C 3 and C 4 species (C 3 excluding T. turgidum, +72-121%; T. turgidum −23%; C 4 , +65-134%; Table 2).

Impact of c a on dynamic leaf gas exchange
The rate of decrease in g sw responding to a step decrease in PPFD from 500 μmol m −2 s −1 to 100 μmol m −2 s −1 (Δg sw / Δt) was generally greater for C 3 than C 4 grasses (Fig. 2). This was broadly consistent with the higher steady-state g sw of C 3 species at a PPFD of 500 μmol m −2 s −1 (Fig. 1C). A significant c a ×species interaction was detected (F 14,113 , P<0.0001), associated with significant photosynthetic-type effects on the average response between c GLA and c AMB (t 113 , P=0.041), and between c GLA and c SUP (t 113 , P<0.0001). All C 4 species increased Δg sw /Δt with decreasing c a , but only one C 3 species (P. bisulcatum) showed a similar trend (Fig. 2). Pairwise tests for responses of Δg sw /Δt within species were significant only when comparing c GLA and c SUP of three species: P. bisulcatum (C 3 , P<0.001), S. bicolor (C 4 NADP-ME, P<0.001), and P. miliaceum (C 4 NAD-ME, P=0.004). Within their respective photosynthetic types, these three species showed the greatest values for iWUE and lowest values for g sw at c SUP , and the greatest decreases in iWUE from c SUP to c GLA (Fig. 1A, C).

Impact of c a on LMA
At c SUP , LMA values for flag leaves of C 3 and C 4 species were similar (Fig. 3); however, the response of LMA to c a differed among the eight species ( Fig. 3; F 14,102 , P=0.0004). None of the C 4 species exhibited significant changes in LMA in response to c a (Tukey's HSD, P>0.72). In the C 3 species, LMA was similar across the three c a treatments for S. laxa, but T. turgidum, P. bisulcatum, and O. sativa all showed significant reductions in LMA from either c SUP to c GLA (T. turgidum and P. bisulcatum, Tukey's HSD P≤0.034) or from c AMB to c GLA (O. sativa, P=0.039; Fig. 3). The contrasts term for photosynthetic type×c a , which was statistically significant (t-test 102 , P=0.014), was therefore broadly associated with less sensitivity of LMA to c a among the C 4 species. Conservation of LMA across CO 2 treatments in most C 4 species was linked with proportionate decreases in mass and area of the flag leaves as c a was reduced. Among C 3 species, decreases in LMA arose because flag leaf mass decreased with c a from c SUP to c GLA , and flag leaf area decreased from c SUP to c AMB , but leaf areas were often similar at c GLA and c AMB (Fig. 4).

Response of K leaf and P-V characteristics to decreases in c a from c AMB to c GLA
There were no significant species×c a effects on K leaf (species×c a F 7,64 , P=0.814); however, on average, K leaf was lower in plants grown at c GLA (F 1,64 , P=0.005). The exception was D. exilis, a C 4 species with small leaves, for which measurement errors were large (Fig. 5A).
In the P-V analysis, the response of LDMC to c a was consistent with that of LMA measured during determination of K leaf . LDMC was not significantly different between the photosynthetic types at either c a , but was lower at c GLA among C 3 and not C 4 leaves (Table 3). LDMC decreased by 4-11% in C 3 grasses grown at c GLA , but C 4 species showed no adjustment to c GLA or increased LDMC by ≤3% at c GLA . This difference in average LDMC responses to c a was statistically significant when comparing C 3 and C 4 species (Table 3). Despite these differences in LDMC responses between C 3 and C 4 species, we found no evidence for significant effects of photosynthetic type on the response of ε or RWC TLP to c a . In contrast, the median π TLP differed between C 3 and C 4 grasses at c AMB but not at c GLA , linked Table 2. Impact of photosynthetic type on leaf gas exchange relative to ambient CO 2 (c AMB ~400 μmol mol −1 ), at glacial CO 2 (c GLA ~200 μmol mol −1 ), and at super-ambient CO 2 (c SUP ~640 μmol mol −1 ) with a significant effect of photosynthetic type (Table 3). At c AMB , π TLP was less negative among C 4 species (C 4 , −0.72 MPa to −0.87 MPa; C 3 , −0.94 MPa to −1.36 MPa). This difference was eliminated at c GLA because only C 4 grasses decreased π TLP to more negative values (C 4 , −0.79 MPa to −1.27 MPa; C 3 , −0.91 MPa to −1.21 MPa; Table 2).

Discussion
We exposed C 3 and C 4 grasses to atmospheric CO 2 concentrations ranging from levels that occurred during the last 30 million years, when C 4 lineages evolved and diversified, to those that could be experienced in the coming centuries. Across the range of c a , we expected that C 4 species would maintain an iWUE advantage and show smaller physiological adjustments. Our results broadly support this expectation: the absolute response of g sw to c a was greater among C 3 than among C 4 grasses; and, as c a decreased from c AMB to c GLA , A, LDMC, and LMA declined more among C 3 than among C 4 species. Investigation of leaf hydraulic function at c AMB and c GLA showed that at c GLA , K leaf decreased in both C 3 and C 4 species;  and π TLP of C 4 leaves became more negative, hence more similar to π TLP of C 3 leaves, which did not adjust. Assaying the stomatal response to shade showed that higher steady-state g sw of C 3 species was linked with more rapid adjustment of g sw to match A. Rates of stomatal closure were slightly more similar for C 3 and C 4 species at low c a , driven by strong responses of species that achieved high iWUE at elevated c a . These new findings are consistent with the hypothesis that carbon limitation is an important factor influencing leaf hydraulic function at different atmospheric [CO 2 ]. Although there was substantial variation among species, photosynthetic type affected how leaf dry matter was deployed and how leaf turgor characteristics responded to c GLA .

Gas exchange responses to c a
Steady-state gas exchange measurements provided the expected outcomes: g s usually increased as c a decreased (Osborne and Sack, 2012;Franks et al., 2013); high g s of C 3 grasses was associated with greater g s responses to c a ; and low g sw of C 4 leaves resulted in higher iWUE at all levels of c a . Importantly, A declined for C 3 but not C 4 grasses at c GLA . Greater A among some C 3 species compared with C 4 species at c AMB and c SUP suggested that C 4 photosynthetic performance may have been limited by PPFD, so the iWUE advantage to C 4 species may underestimate advantages to C 4 species that could arise at higher irradiances (Osmond et al., 1982). C 3 grass leaves generally closed their stomata more quickly than C 4 leaves in response to shade. The higher steady-state g sw of C 3 leaves may partially explain this difference between the photosynthetic types, but closer inspection of the data shows that Δg sw /Δt did not parallel the steady-state g sw for species within each photosynthetic type. Interestingly, among C 4 species, the rate of g sw responses to light was slightly, but consistently, greater at c GLA compared with c AMB and c SUP . This decreased the difference in Δg sw /Δt between C 3 and C 4 species. However, a more striking trend, that probably underpinned the subtle difference in relative performance based on photosynthetic type, was that species with higher iWUEs showed greater changes in Δg sw /Δt in response to decreasing c a . At c SUP , species with high iWUE showed some of the slowest stomatal responses to shade. Because faster stomatal responses are consistent with improved water use efficiency (Lawson and Blatt, 2014), this suggests that transpiration is regulated less tightly at high c a , supporting the overarching hypothesis that increasing c a minimizes the costs associated with hydraulic stress (Polley et al., 1993). It also suggests that characteristics producing high iWUE in the steady state may be costly in low-c a -like scenarios that increase transpiration. For example, high iWUE is likely to be facilitated by Fig. 4. Effect of c GLA (black symbols, versus grey for c AMB and c SUP ) on proportionality between leaf mass and area for individual upper canopy leaves (log-log allocation plots). Crop and crop-related annual grass species are plotted by photosynthetic type: C 3 (filled: circles, Triticum turgidum; squares, Oryza sativa; diamonds, Steinchisma laxa; triangles, Panicum bisulcatum) or C 4 (open: circles, Sorghum bicolor; squares, Digitaria exilis; diamonds, Setaria italica; triangles, Panicum miliaceum). Points are means (n=4-6); SE is omitted for clarity. Lines have slope=1, and the intersect is the mean value at c AMB . high rates of internal diffusion, which are linked with decreases in cell wall dry matter (Onoda et al., 2017) and might increase vulnerability to changes in leaf water status.
Among-species variation was an important feature of our gas exchange results. This is consistent with previous studies, which have indicated that the degree to which grass stomata protect against decreases in hydraulic conductance varies even among genotypes Holloway-Phillips and Brodribb, 2011a). Among C 3 species in our study, only that with the highest iWUE, P. bisulcatum, showed a clear negative association between c a and the stomatal response to shade. At the other extreme, T. turgidum showed exceptionally high steady-state g sw and slow stomatal responses to shade in all three c a treatments, suggesting high transpiration irrespective of leaf water status, a strategy that can maximize CO 2 uptake at a cost to hydraulic conductance (Holloway-Phillips and Brodribb, 2011b). The apparent lack of stomatal regulation in T. turgidum compared with other C 3 species is important to note because iWUE for this species did not decrease at c GLA , contradicting the otherwise consistent trend towards greater decreases in iWUE among C 3 compared with C 4 species.

Impact of glacial c a on LMA and hydraulic characteristics
LMA decreased at c GLA among C 3 but not C 4 grasses. This finding is consistent with observed differences in A, results from a meta-analysis addressing variation in LMA (Poorter et al., 2009), and more recent comparisons using species and c a treatments similar to those chosen for our experiment (Pinto et al., 2014). Further evidence is needed, however, before this result can be generalized as a photosynthetic type effect. LMA responses can, for example, be modified by temperature (Pinto et al., 2011). It is also important to note that the C 4 and two of the wild C 3 species included in our experiment were drawn from one subfamily of the Poaceae: Panicoideae, a broadly mesic-adapted clade (Taub, 2000;Osborne, 2008;Edwards and Smith, 2010;Visser et al., 2014). We expect leaf functional traits to reflect adaptations to habitat, and some major C 4 lineages are adapted to drier environments than those favoured by the Panicoideae (Taub, 2000;Edwards and Smith, 2010). In addition, LMA responses to c a (Pinto et al., 2016) and leaf size (Liu et al., 2012) differ between the Chloridoideae and Panicoideae grass subfamilies. While further work will be needed to establish whether the patterns we observed are general across grass lineages, our findings are directly relevant to crop and croprelated annual grass species from mesic habitats. Taken together with the gas exchange results, differences in LMA indicate that c GLA was linked with greater carbon limitation in C 3 grasses compared with their C 4 relatives. This is important because differences in carbon supply affecting plant size and allocation at the whole-plant level have previously been highlighted as central to functional contrasts between C 3 and C 4 plants (Long, 1999;Atkinson et al., 2016), and influence the mechanisms by which plants acclimate to hydraulic stress (Maseda and Fernández, 2006).
C 3 and C 4 grasses showed similar K leaf , and K leaf decreased at c GLA . The finding that there was no clear difference in K leaf between photosynthetic types is consistent with a previous comparison using the high pressure flow meter technique, applied to predominantly perennial, North American prairie grasses (Ocheltree et al., 2014b). Both of these results are surprising because the clearest anatomical differences between C 3 and C 4 grass lineages are in the ratio of bundle sheath to mesophyll (Hattersley, 1984;Dengler et al., 1994;Christin et al., 2013;Griffiths et al., 2013;Lundgren et al., 2014). Increases in this ratio should decrease hydraulic resistance external to the xylem (Buckley et al., 2015), supporting the hypothesis that differences in leaf hydraulic properties could affect responses to stress imposed by low c a and/or water availability (Osborne and Sack, 2012;Griffiths et al., 2013). It is possible that other aspects of C 4 leaf anatomy or function counteract positive effects of Table 3. Impact of growth c a on pressure-volume curve characteristics and leaf dry matter content (medians, n=3-5; c GLA , glacial CO 2 ~200 μmol mol −1 ; c AMB , ambient CO 2 ~400 μmol mol −1 ) increased bundle sheath ratios on K leaf in C 4 grasses. It is also important to note that C 3 and C 4 grasses often show similar average mesophyll cell sizes at ambient CO 2 (Lundgren et al., 2014), and the cross-sectional area of vascular relative to chlorenchyma tissues does not necessarily change with photosynthetic type (Dengler et al., 1994). The evidence we found for decreased K leaf at c GLA was surprising, because xylem conductivity generally increases with declining c a to support increased g sw (Rico et al., 2013;Domec et al., 2017). Previous in situ measurements of transpiration and leaf water potential in sunflower plants grown at c a similar to c GLA and c AMB showed the expected result: that K leaf measured at ambient CO 2 increased for plants grown at c GLA , minimizing the impact of increased g sw on ΔΨ (Simonin et al., 2015). A decrease in xylem conductivity, linked with smaller conduits in water-stressed tissue that would increase redundancy among conducting elements (Comstock and Sperry, 2000), might contribute to decreases in K leaf for leaves grown at c GLA . However, this is not consistent with the decrease in LMA that we observed and, since transpiration was driven using moderate levels of light, we expect that the primary source of hydraulic resistance was exterior to the xylem (Ocheltree et al., 2014a).
The values of K leaf were low compared with other recently published estimates for similar species [S. bicolor, 19-38 mmol m −2 s −2 MPa −1 (Ocheltree et al., 2014a); O. sativa cultivars, 7.1-8.7 mmol m −2 s −2 MPa −1 (Xiong et al., 2015)], but are within the range reported in the literature for grasses (~0.44-51 mmol m −2 s −2 MPa −1 ; Holloway-Phillips and Brodribb, 2011a;Ocheltree et al., 2014a, b;Liu and Osborne, 2015;Xiong et al., 2015) and may be a consequence of moderate PPFD during growth and measurements (Cochard et al., 2007;Ocheltree et al., 2014a). Further experimentation and comparison of methods is needed for measurements of K leaf in grasses. We need to understand why measurements of K leaf produce similar values for C 3 and C 4 species; to establish whether K leaf responses to c a correspond to changes in hydraulic vulnerability; and to determine the anatomical basis of adjustments to K leaf , especially given evidence for declining LDMC and LMA among C 3 species at c GLA . It will also be important to measure K leaf at different [CO 2 ]; as in the study of K leaf responses to c a that used sunflower (Simonin et al., 2015), we measured K leaf at ambient [CO 2 ].
Effects of c a on P-V characteristics also provide motivation for further investigation of photosynthetic type×c a responses. As leaf size decreased at c GLA , C 4 grasses maintained LDMC and C 3 grasses did not. In parallel, π TLP of C 4 grasses became more negative at c GLA , while π TLP of C 3 grasses did not change. This is an important result because π TLP is a powerful indicator of physiological responses that is expected to integrate smaller changes in, for example, π 0 and ε ( Bartlett et al., 2012). The decrease in LDMC shown by C 3 leaves grown at c GLA is consistent with both lower A and LMA, and previous evidence that C 3 leaves decrease mesophyll cell volume and total non-structural carbohydrates as c a declines (Poorter et al., 2009). Maintenance of LDMC and more negative π TLP in C 4 grasses therefore might be linked with solute accumulation at c GLA . Presumably, decreases in π TLP of C 4 leaves at low c a would support maintenance of turgor in the presence of larger ΔΨ induced by higher g sw (Franks, 2006;Simonin et al., 2015); however, we do not know how leaf-level changes were integrated with adjustments in root and stem properties. The lack of an adjustment in π TLP by C 3 grasses grown at c GLA might be associated with maintenance of leaf water status if root and stem xylem hydraulic conductivity increased or xylem solute concentrations decreased.

Conclusions
We predicted that gas exchange would show greater absolute responses to c a in C 3 compared with C 4 grass leaves, especially in terms of the positive relationship between iWUE and c a . We also predicted that low iWUE at c GLA would be linked with changes in leaf hydraulic properties. We found that while the iWUE advantage of some C 4 grass leaves increased in absolute terms at c SUP , co-ordination among leaf traits was more strongly affected by c GLA than by c SUP . These experimental results broadly support predicted smaller impacts of c GLA on performance of C 4 grasses (Osborne and Sack, 2012), and suggest that iWUE advantages to C 4 species will continue to be important in future. A finding with potential importance for crop improvement programmes is that as c a increases, pressure on plants to improve iWUE through rapid stomatal responses to shade may be reduced, particularly for species capable of achieving high iWUE. These results highlight the need for continued efforts to establish how hydraulics and photosynthetic performance are co-ordinated, both within leaves and at the scale of whole plants. The mechanistic basis of these responses still needs to be better understood to predict the physiological implications of C 4 photosynthesis, both under past glacial climates and as they will affect performance in a future high CO 2 world.