Drought limitation of photosynthesis differs between C3 and C4 grass species in a comparative experiment


C. P. Osborne. Fax: +44 114 222 0002; e-mail: c.p.osborne@sheffield.ac.uk


Phylogenetic analyses show that C4 grasses typically occupy drier habitats than their C3 relatives, but recent experiments comparing the physiology of closely related C3 and C4 species have shown that advantages of C4 photosynthesis can be lost under drought. We tested the generality of these paradoxical findings in grass species representing the known evolutionary diversity of C4 NADP-me and C3 photosynthetic types. Our experiment investigated the effects of drought on leaf photosynthesis, water potential, nitrogen, chlorophyll content and mortality. C4 grasses in control treatments were characterized by higher CO2 assimilation rates and water potential, but lower stomatal conductance and nitrogen content. Under drought, stomatal conductance declined more dramatically in C3 than C4 species, and photosynthetic water-use and nitrogen-use efficiency advantages held by C4 species under control conditions were each diminished by 40%. Leaf mortality was slightly higher in C4 than C3 grasses, but leaf condition under drought otherwise showed no dependence on photosynthetic-type. This phylogenetically controlled experiment suggested that a drought-induced reduction in the photosynthetic performance advantages of C4 NADP-me relative to C3 grasses is a general phenomenon.


Grasses utilizing the C4 photosynthetic pathway have evolved repeatedly over the last ∼32 Ma (Christin et al. 2007, 2008; Vicentini et al. 2008; Bouchenak-Khelladi et al. 2009). These species play a major ecological role at the global scale, dominating warm climate grassland ecosystems (Still et al. 2003), and are important as agricultural crops (e.g. millets, maize, sugarcane), forage (Brown 1999) and biofuel feedstocks (Heaton, Dohleman & Long 2008). The potential importance of contrasts between C3 and C4 photosynthesis in determining ecological patterns, at scales up to and including the continental and global, has long been recognized and debated (Hatch, Osmond & Slatyer 1971; Osmond, Winter & Ziegler 1982; Pearcy & Ehleringer 1984).

Controlling for phylogeny is crucial when comparing the ecophysiological traits of C3 and C4 grasses (Edwards, Still & Donoghue 2007; Edwards & Still 2008; Taylor et al. 2010). Molecular phylogenies place most commonly studied C3 grasses from temperate climates into a clade known as BEP (three subfamilies, Bambusoideae, Ehrhartoideae and Pooideae, exclusively C3), whilst C4 photosynthesis has arisen only in its largely tropical sister clade known as PACMAD (six subfamilies, Panicoideae, Aristidoideae, Chloridoideae, Micrairoideae, Arundinoideae and Danthonioideae, including both C3 and C4 photosynthetic types). These two clades diverged more than 50 Ma ago (Christin et al. 2008; Vicentini et al. 2008; Bouchenak-Khelladi et al. 2009), and recent work has shown that evolutionary divergences both between and within these clades may explain ecophysiological differences that were previously attributed to differences between C3 and C4 photosynthetic types and subtypes (Taub 2000; Edwards et al. 2007; Cabido et al. 2008; Edwards & Still 2008; Edwards & Smith 2010).

Comparative analyses based on large molecular phylogenies indicate that C4 grasses tend to occupy a drier niche than their C3 relatives, and that the evolution of C4 photosynthesis facilitated ecological transitions into drier, open habitats (Edwards & Still 2008; Osborne & Freckleton 2009; Edwards & Smith 2010). These results are consistent with the experimental observation that, in mesic high irradiance conditions, C4 grasses typically achieve higher rates of net leaf photosynthesis (A) than their closest C3 relatives, whilst their stomatal conductance to water vapour (gs) is markedly lower (Taylor et al. 2010), i.e. C4 grasses exhibit higher intrinsic water-use efficiency (A/gs = iWUE). The ratio of A to leaf nitrogen content per unit area, photosynthetic nitrogen-use efficiency (A/Narea = PNUE), also tends to be greater for C4 grasses. However, recent experiments have suggested that such physiological advantages of C4 photosynthesis may not persist under drought. In two independent comparisons of C3 and C4 grasses from the subfamily Panicoideae, drought eliminated differences in A that were observed between C3 and C4 species under well-watered control conditions (Ibrahim et al. 2008; Ripley, Frole & Gilbert 2010). In pot-based studies, under well-watered conditions, gs in C3 species was initially higher, but under drought declined to values similar to or smaller than those observed in their C4 relatives (Ripley et al. 2007, 2010). Whilst stomatal limitation explained a large proportion of the total decline in A in C3 species, metabolic limitation was proposed to be the dominant effect on A in their C4 relatives. Although these experiments considered only a limited range of C4 species, their results undermine the hypothesis that the iWUE advantage of C4 grasses under mesic conditions, which is associated with high A and PNUE (Long 1999), persists under drought.

We tested the generality of these findings with a comparative experimental approach, using a grass phylogeny to sample species representing multiple comparisons between independent C4 lineages and C3 sister groups. We concentrated on comparisons between C4 NADP-me and C3 photosynthetic types, which contribute the majority of phylogenetic diversity in photosynthetic types within the Poaceae (Christin et al. 2009). Our design did not control for habitat or life history, but relied on random sampling of species to represent the ecological diversity of the photosynthetic types. We imposed a drought treatment that consisted of a controlled decline in soil water content and addressed the key question: does drought have differential effects on A and gs in C3 and C4 species? In addition, to investigate the extent to which plants differing in photosynthetic type were tolerant of drought, we measured the response to drought of: (1) photosynthetic resource use efficiency (iWUE and PNUE); and (2) leaf senescence and leaf water potential.


Species selection

To provide a phylogenetically controlled sample of the diversity of C4 NADP-me types and their C3 relatives within the grasses, species were selected from genera placed on two recently published phylogenies (Christin et al. 2008; Vicentini et al. 2008). Genera known to be polyphyletic or paraphyletic (e.g. Setaria), were avoided. A species list was drawn up based on existing collections of plants and seeds, and availability from several suppliers. We categorized the list according to C4 origins and C3 sister clades (Christin et al. 2008), and chose species at random from within each category. Inferred evolutionary relationships at genus level for the species used are shown in Fig. 1.

Figure 1.

Inferred evolutionary relationships, at genus level, for species included in the experiment, after Christin et al. (2008). Species with C4 photosynthesis are indicated by black labels. We assume that the genera are monophyletic. Branch lengths are proportional to time. All genera except Stipa belong to the PACMAD clade.

Plant material and growth conditions

Species and sources for plant material are listed in Table 1. Most species were grown from seed; however, Cortaderia selloana and Stipa gigantea were obtained as adult plants and Phragmites australis was propagated from rhizome segments. Seeds were surface-sterilized before germination on unlit shelves in a growth room (MTPS 120, Conviron, Winnipeg, Manitoba, Canada; conditions 30 °C, 0800–1500 h, declining to 25 °C 1800–0500 h, relative humidity 80%). Germinated seedlings were transferred into a 3:1 mix of John Innes no. 3 compost : perlite, under moderate illumination [mean ± SEM photosynthetic photon flux density (PPFD), 761 ± 25 µmol mol−1 0600–1700 h, ramping from darkness, 1800–0500 h] but otherwise similar conditions. Rhizome segments were treated similarly. When large enough, seedlings were transferred to a separate walk-in growth room (BDW 160, Conviron), and transplanted into 4.5 L pots with an 185 mm top diameter (LBS Group, Colne, Lancashire, UK), filled with a 1:1 mix of John Innes no. 3 compost : washed silica sand (Chelford 52, WBB Minerals, Sandbach, Cheshire, UK).

Table 1.  Species and sources
SpeciesPhotosynthetic typeLife-historySourceaAccession ID
  • a

    Sources: A) Ferndale Garden Centre, Dronfield, Derbyshire, UK; B) USDA National Plant Germplasm System (NPGS), BARC-West Beltsville, MD, USA; C) Millenium Seed Bank Project, Royal Botanical Gardens, Kew, UK; D) Australian Plant Genetic Resource Information Service (AusPGRIS), Queensland, Australia; E) B & T World Seeds, Paguignan, France; F) Seed from a field collection made near Grahamstown, South Africa.

Stipa giganteaC3PerennialA
Aristida adoensisC4PerennialBPI 385318
Cortaderia selloanaC3PerennialA
Phragmites australisC3PerennialC29212
Eriachne aristideaC4Annual/perennialDAusTRCF 322433
Chasmanthium latifoliumC3PerennialE40528
Digitaria ciliarisC4AnnualDKP 5148
Cenchrus ciliarisC4PerennialBPI 147685 701SD
Sacciolepis vilvoidesC3AnnualBPI 338609 01SD
Echinochloa frumentaceaC4AnnualE436583
Oplismenus hirtellusC3PerennialF
Ischaemum afrumC4PerennialBPI 364924 02SD
Paspalum malacophyllumC4PerennialDCPI 27690

Maximum PPFD at mid-canopy height was 1014 ± 17 µmol m−2 s−1 (mean ± SEM). Each day, lighting was ramped from darkness 1800–0500 h, to maximum light 0700–1600 h. Temperature was 25 °C in darkness, ramping to 30 °C between 0800 and 1500 h. Relative humidity was 80%. The two species (C. selloana and S. gigantea) obtained as adult plants were transferred directly into the growing media and final growth conditions 2 weeks before measurements began. Plants were arranged into five blocks, each on a 1.5 × 0.75 m tray, including a pair of plants for each species. To minimize shading, species within each block were separated according to stature, but within the two groups (tall/short), species order was randomized.

The plants were fertilized four times in the course of the experiment; after week one, during weeks four and five, and after week six (except C. latifolium, which received an additional feed during week two). A commercially available plant food (Bayer Lawnfood and Tonic, N : P : K 38:5:5) was diluted into the appropriate volume of water being added to each treatment.

Soil-drying treatment

We created a controlled decline in gravimetric soil water content (w, g H2O g dry matter−1). The saturated water content of our compost mix was first determined. Compost mix in five 4.5 L pots was watered to drip point and left to stand in a tray of water overnight before being allowed to drain freely for 24 h. Samples were collected using a 15-mm-diameter soil core and w determined based on measurements of fresh mass and dry mass (after drying to constant weight at 100 °C). The mean saturated water content was used to calculate an expected dry mass for the contents of each pot in the experiment based on their saturated weights (plant included) at the start of the experiment, assuming that plants would contribute only a small fraction of the total mass. Expected dry masses were used to predict target masses for each pot during the drying phase of the experiment (weeks two to four), to obtain a decrease in w of approximately 2.4 g g−1 d−1 (comparable with rates of soil drying previously observed in natural mixed C3/C4 grassland in South Africa; Ripley et al. 2010).

During the drying phase, pots were weighed each morning and watered with an appropriate volume of water to obtain the target mass. This was continued until substantial wilting and discolouration of leaves was observed in the most susceptible species, at the end of week four. During week five, w was maintained at the level reached at the end of week four, and in week six, the pots were rewatered to drip point for a 2 week recovery period. Pots in the control treatment were saturated by watering to drip point every day, as were pots in the drought treatment during weeks one (prior to drying), six and seven (recovery from drying).

To estimate soil water potential (Ψsoil) from w, a moisture characteristic curve was produced for the compost mix. The dry mass of subsamples of soil was determined after drying at 100 °C. Samples were placed into pre-weighed polystyrene vials (60 mL, 125AP, Sterilin, Caerphilly, UK), and deionized water added to obtain a range of values for w, augmented by repeated measurements as samples dried or were re-wetted. To determine Ψsoil, soil psychrometers (PST-55-15 thermocouple psychrometer/hygrometer, Psypro, Wescor Inc., Logan, UT, USA) sealed into the vials and calibrated against standard solutions of 0.1 to 1 molal NaCl, were used with a datalogger (Psypro, Wescor Inc.). The non-linear relationship between Ψsoil and w was estimated as Ψsoil = aw−b (Campbell & Norman 1998) where a = −0.399, b = −0.642.

Leaf physiology

Leaf gas exchange was measured for every plant in each week during the experiment, using a portable open system (LI-6400; LI-COR, Inc., Lincoln, NE, USA), equipped with a CO2 mixer, 30 mm × 20 mm chamber and red–blue light-emitting diode light source (LI-6400-02B). Spot measurements were made between 1000 and 1530 h, aiming to determine the maximum A at the operating gs of youngest mature leaves on randomly selected tillers. Within each block, a full set of species were measured each morning, followed by the second plant for each species pair in the afternoon (28 plants in total per day). Plants from both drought and control treatments were measured in the morning and afternoon sessions, and for each species, the first treatment to be measured (drought or control) was rotated between each set of measurements.

For fine-leaved species, to increase the leaf area and size of fluxes measured, gas exchange was measured on a pair of leaves from adjacent tillers. A PPFD of 2500 µmol m−2 s−1 was used. Preliminary measurements established that this PPFD was required to saturate photosynthesis in the C4 species and caused no decline in A in shade tolerant species. Plants were acclimated to high light conditions by raising them towards the growth chamber lights for 10–15 min prior to measurement. Leaf chamber conditions were matched to those of the growth environment; temperature was controlled at 30 °C and chamber humidity was maintained between 60 and 85%, to allow minimal adjustment when switching rapidly between leaves with different rates of water efflux (different leaf areas and drought/control plants). This compromise was demanded by the large numbers of samples required to satisfy the experimental design and resulted in values for mean chamber vapour pressure deficit (VPD) ranging between 0.79 and 1.55 kPa. Leaf temperature was estimated using an energy balance calculation.

Additional measurements were taken alongside measurements of gas exchange in week five, when w was at its lowest. Chlorophyll content was estimated using a SPAD meter (Konica Minolta Sensing Inc., Osaka, Japan). Meter readings were calibrated against total chlorophyll content in the leaves of eight grass species following the method of Porra, Thompson & Kriedemann (1989), and were related to SPAD readings using the conversion inline image (Markwell, Osterman & Mitchell 1995); where x = SPAD reading and a is a fitted coefficient.

A 30 mm segment from the centre of the leaf was collected following gas exchange. The area of segments was calculated based on their width at either end and masses were determined after drying at 80 °C for at least 48 h, allowing the calculation of specific leaf area (SLA, cm2 g−1). Dried leaf segments were stored in an airtight container over silica gel prior to analysis for nitrogen concentration. Leaves were ground using a ball mill (TissueLyser, Quiagen, Crawley, West Sussex, UK), and nitrogen concentrations were determined using a stable isotope ratio mass spectrometer (PDZ Europa 20-20, PDZ Europa Ltd, Cheshire, UK).

During the same period in which gas exchange measurements were made, operating leaf water potential (Ψop) was measured for a youngest fully emerged leaf from each plant, using a Scholander pressure-chamber (model 1000, PMS Instrument Company, Albany, OR, USA) and following the methods described by Turner (1981).


Where summaries are presented for C3 and C4 groups, values are weighted means (Gelman & Hill 2007), i.e. the average of species means, weighted by the proportion of the total number of individuals used to estimate them. Standard errors for weighted means are the square root of the summed variances of the species means, weighted by sample size.

Statistical analyses were carried out using the R Language and Environment (R Development Core Team 2005). A generalized least squares approach (gls function, ‘nlme’ package in R (Pinheiro et al. 2009), was used to estimate the responses of species to treatments within each week of the experiment, allowing models to be fitted to untransformed data whilst accounting for differences in variance between treatment groupings. Count data for the number of leaves on a stem were loge transformed for analysis.

Models estimated the effects of species, soil drying, and species × soil drying interactions and were simplified by step-wise deletion according to the Aikake's information criterion (AIC) criterion. Linear contrasts (Crawley 2007) were formulated using tools in the R package ‘gmodels’ (Warnes et al. 2009) and used to address the a priori hypothesis that the C3 and C4 groups would differ in their average response to treatments. To address differences between C3 and C4 species in the control treatment, when the overall model indicated a significant effect of species, contrasts were calculated between predicted species means. To address the effect of photosynthetic type on species responses to drought, in models for which the interaction term (species × soil drying) was statistically significant, contrasts were calculated comparing the average size of interaction coefficients between photosynthetic types.

Scaling of leaf nitrogen with SLA was estimated by using loge transformed data and the package ‘smatr’ in R (Warton & Ormerod 2007), to apply a standardized major axis regression with adjustment for measurement error. Differences in slopes were tested using a likelihood ratio test and differences in elevation using a Wald statistic.


Progress of the soil-drying treatment

During the drying phase of the drought treatment, which began in week two, the average daily decline in mass prior to watering was 0.02 ± 0.05 kg (mean ± SEM) for C4 species, and 0.03 ± 0.04 kg for C3 species (i.e. less than 1% of average saturated pot mass at the start of the experiment). Between-species variation in the rate of soil drying at the diurnal scale was therefore high, and the difference contrast between C3 and C4 groups was not significant. Overall, the drying phase resulted in a decrease in w from a mean value of greater than 0.6 g g−1 in week one, to 0.19 g g−1 in week five (Fig. 2a). Rewatering from the start of week six returned w to values above 0.6 g g−1. Estimated Ψsoil ranged between −0.42 and −1.77 MPa, with a mean value in week five of −1.22 MPa (Fig. 2b).

Figure 2.

Response of soil water availability to experimental drought treatment. Mean ± SEM for 65 pots (SEM < 5% of mean in all cases). (a) Gravimetric soil water content (w, g H2O g dry matter−1). (b) Soil water potential (Ψsoil, MPa) calculated using a psychrometer soil suction curve. Control ○; soil drying treatment inline image.

Response of leaf gas exchange to drought

In the control treatment, C4 grasses showed a 41% decline in A over the course of the experiment (Fig. 3a). A 32% decrease in A was also observed for the C3 grasses. Despite the greater decline in A for C4 grasses compared with C3 grasses, a significant difference between the two groups was supported at every time point and A remained higher for the C4 group throughout (Fig. 3a). The response of A to dry soil, modelled as the difference between control and drought treatments, differed significantly between species in week five, but the species responses did not differ between the photosynthetic types (Fig. 3a).

Figure 3.

Response of leaf gas exchange to declining soil water availability (weeks one to five), and rewatering (weeks six and seven), summarized according to photosynthetic type. Weighted mean ± SEM, n = 3 to 5, for seven C4 and six C3 species; C4 control ○, C4 drought inline image, C3 control ●, C3 drought ▴. (a) net CO2 assimilation (A, µmol m−2 s−1). (b) stomatal conductance (gs, mol m−2 s−1). (c) leaf internal CO2 concentration (ci, µmol mol−1). In the control treatments, between-species effects were significant (P < 0.001) for all comparisons, as were a priori contrasts between photosynthetic types (P < 0.001, except for A in week five, P = 0.041). Comparisons for which the species × drought interaction was significant (P < 0.01) are indicated according to the statistical significance of between-photosynthetic type contrasts: ***P < 0.001, *0.01 < P < 0.05, NSP > 0.05. When there was no significant species × drought interaction, the soil-drying treatment had significant effects (P < 0.05) on; A, weeks three–six; gs, weeks three–six; ci, weeks two–six.

In the control treatment, and under high soil water availability in the dry-down treatment, the gs of C3 grasses was three times that for C4 species (Fig. 3b). There was a significant difference in control values between the two groups at every time point (Fig. 3b). During weeks four, five and six, the response of gs to the drought treatment differed between species, and the average size of the responses differed significantly between photosynthetic types (Fig. 3b). In week five, when the drought was at its most extreme, the decline in gs relative to controls was 76% for C3 species, but only 37% for C4 species. Although the mean for C4 species remained 20% lower than that for C3 species in the drought treatment, the difference was 69% in the control treatment.

The lack of a difference in the response of A, combined with the evidence for a larger response of gs to drought in the C3 species, suggested that the effect of drought on the relationship between A and CO2 supply differed between C3 and C4 species. This was supported by the response of ci (Fig. 3c), which, if the relationship between photosynthesis and CO2 supply was unaffected under drought, should have declined as well. Overall, a decline in ci was observed for the C3 species, but not for the C4 species under drought (Fig. 3c). For each week in which the responses of gs to drought differed between species, this was also the case for estimates of ci (Fig. 3c). However, the average response of ci to the drought treatment only showed a significant difference between photosynthetic types during the drying phase of the experiment. Differences in the response of ci between photosynthetic types were lost immediately upon rewatering (Fig. 3c), although the size of responses to drought continued to differ between species.

An examination of the response of both iWUE and PNUE to drought during week five (Fig. 4) indicated that the extent of variation in iWUE both within- and between-species increased under drought, and that there was greater overlap in values of iWUE between C3 and C4 species (Fig. 4; drought × species F12,102 = 7.4, P < 0.001). This increased overlap occurred against the background of constant atmospheric VPD and was driven primarily by increases in iWUE in C3 species that resulted in a significant interaction between photosynthetic type and drought (t102 = 2.9, P = 0.004) and a 40% reduction in the C4iWUE advantage. Both C3 and C4 species tended to show reduced PNUE under drought (Fig. 4). The extent of this response differed between species (drought × species F12,87 = 5.1, P < 0.001) and the average size of reductions in PNUE was greater for C4 species, resulting in a 41% decline in the C4PNUE advantage (t87 = 2.0, P = 0.045).

Figure 4.

Response of iWUE and PNUE to drought in 13 grasses differing in photosynthetic type. Measurements taken during week five of drought treatment. Mean ± SEM, n = 3 to 5, except C3S. vilvoides in the drought treatment, n = 2 for PNUE; C4 control ○, C4 drought inline image, C3 control ●, C3 drought inline image. In response to the drought treatment, iWUE showed a significant (P < 0.05) decline in one of six C3 species, and increased significantly in four of six C3 and one of seven C4 species; PNUE showed a significant decline in five of six C3 and four of seven C4 species.

Rewatering produced a general recovery of gas exchange (Figs 3 & 4), with no significant difference in A or ci between plants in the drought and control treatments by week seven (Fig. 3a,c). The effect of the drying treatment remained significant for gs (Fig. 3b) in week seven; however, the size of this effect was small, mean gs being 0.03 mol m−2 s−1 greater in the drought treatment than in the control treatment (F1,112 = 5.36, P = 0.022).

Response of leaf water potential to drought

The difference in mean Ψop between drought and control was similar to or greater than the change in Ψsoil for eleven of the 13 species (Fig. 5a). Two C4 species (Digitaria ciliaris and Echinochloa frumentacea) were excluded from the analysis of Ψ, because values for Ψop in the drought treatment were substantially less negative than estimated Ψsoil. The ‘false’ end points detected for these two species may be explained as a result of water contained in non-vascular mesophyll and parenchyma cells being squeezed out of the leaf by the chamber gasket as pressure was applied. There was a significant difference in Ψop between drought and control (P < 0.02) for all species except Sacciolepis vilvoides (t84 = −1.53, P = 0.131). There was also a significant difference in Ψop between C3 and C4 species in the control treatment (t84 = −4.40, P < 0.001), the mean for C3 species being 23% more negative than for the C4 species. However, although the size of the significant responses to drought differed between species (F10,84 = 6.4 P < 0.001), they did not differ significantly between the photosynthetic types (t84 = 1.94, P = 0.055). Two species showed an unusually large response of Ψop to drought (>1.5 MPa, Fig. 5a), which was greater in Ischaemum afrum (C4) than in S. gigantea (C3).

Figure 5.

Response of leaf water potential to drought in eleven grasses differing in photosynthetic type. Measurements taken during week five of drought treatment; (a) Operating leaf water potential (Ψop); (b) Leaf-soil gradient in water potential (ΔΨ). Mean ± SEM, n = 4 to 5; C4 control ○, C4 drought inline image, C3 control ●, C3 drought ▴. Reference lines in (a) are mean Ψsoil, for control (dashed line) and drought (dotted line). Within each photosynthetic type, species order follows mean Ψop under control conditions.

The average size of gradients in water potential between leaf and soil, ΔΨ (Ψop − Ψsoil), showed a significant difference between C3 and C4 species in the control treatment (t84 = −3.878, P < 0.001), the average gradient for C3 species being 0.28 MPa larger than for C4 species. The effect of the drought treatment on ΔΨ was strongly dependent upon species (F10,84 = 5.89 P < 0.001), including a significant effect of photosynthetic type (t84 = 2.08, P = 0.041), that eliminated the difference in average ΔΨ between C3 and C4 species. The importance of this effect is unclear, because a significant difference in ΔΨ between treatments (P < 0.009, Fig. 5b) was detected for only 3 of the 11 species, I. afrum (C4), S. gigantea (C3) and Chasmanthium latifolium (C3). Thus, over the range of Ψsoil generated, most species maintained a relatively constant ΔΨ in both control and drought treatments; however, a small number of highly responsive species drove an increase in between-species differences in ΔΨ under drought (Fig. 5b).

Leaf and whole plant condition under drought

Comparisons between species during week five, at the most extreme phase of the drying treatment, showed that foliar nitrogen content (Narea, Table 2) differed between species in the control treatment (F12,91 = 32.0, P < 0.001) and that mean values for C4 species were significantly lower than for C3 species (t91 = 29.37, P < 0.001). Although pooled means for Narea increased under drought, and the effect of drought was significant (F1,91 = 21.945, P < 0.001), it was underlain by differences in the size of response between species (drought × species F12,91 = 3.25, P < 0.001) that were independent of photosynthetic type (t91 = 1.10, P = 0.273).

Table 2.  Response of leaf condition to drought treatment, for grass species of C3 and C4 (NADP-me) photosynthetic types
SubtypeSpeciesNarea (mmol N m−2)Leaf (Chl a + b) (µmol Chl m−2)Stem (Chl a + b) (µmol Chl m−2)aNumber of leaves on stembNumber of leaves livingb
Mean ± SEMean ± SEMean ± SEMean ± SEMean ± SEMean ± SEMean ± SEMean ± SEMean ± SEMean ± SE
  • Sample size per species in each treatment was four to five, except where shown in parentheses. Pooled sample sizes for weighted averages were 27 to 30 for C3, and 32 to 35 for C4 NADP-me species.

  • a

    Mean chlorophyll content calculated across all leaves on a stem, with number of living leaves included in model as a covariate.

  • b

    Back-transformed from loge(x).

C3C. latifolium106 ± 1077 ± 9759 ± 61490 ± 112505 ± 87532 ± 856.7 ± 5.2/8.67.6 ± 6.5/8.94.4 ± 3.2/6.26.1 ± 5.2/7.1
C. richardii97 ± 8144 ± 13585 ± 60646 ± 48558 ± 44633 ± 626.7 ± 6.1/7.35.2 ± 4.3/6.35.9 ± 5.5/6.44.6 ± 4.0/5.3
O. hirtellus46 ± 2 (3)66 ± 4 (3)287 ± 79452 ± 59383 ± 38515 ± 413.4 ± 2.3/5.14.4 ± 3.3/5.92.2 ± 1.5/3.12.8 ± 2.4/3.3
P. australis166 ± 14196 ± 20733 ± 51997 ± 77472 ± 117515 ± 1256.3 ± 3.8/10.36.1 ± 3.8/9.74.7 ± 3.0/7.44.2 ± 3.1/5.8
S. vilvoides77 ± 1079 ± 22 (2)664 ± 78750 ± 50404 ± 99529 ± 825.4 ± 4.9/6.05.2 ± 4.6/5.94.4 ± 3.8/5.14.0 ± 3.5/4.6
S. gigantea86 ± 4119 ± 9377 ± 115286 ± 78433 ± 102413 ± 1263.4 ± 2.1/5.53.7 ± 2.5/5.62.6 ± 1.8/3.71.9 ± 1.4/2.8
Weighted average100 ± 9122 ± 14564 ± 77604 ± 74461 ± 86523 ± 925.1 ± 3.7/6.95.2 ± 4.0/6.93.8 ± 2.8/5.13.7 ± 3.0/4.6
C4 NADP-meA. adoensis58 ± 652 ± 3201 ± 8206 ± 20320 ± 125420 ± 1143.7 ± 2.8/5.03.5 ± 2.9/4.33.6 ± 2.8/4.73.2 ± 2.8/3.8
C. ciliaris45 ± 679 ± 11461 ± 50791 ± 78476 ± 107537 ± 955.4 ± 3.3/8.97.6 ± 5.1/11.24.9 ± 3.1/7.65.1 ± 3.5/7.5
D. ciliaris56 ± 862 ± 6599 ± 54855 ± 56509 ± 86553 ± 915.1 ± 4.7/5.64.2 ± 3.3/5.34.3 ± 3.9/4.82.9 ± 2.1/3.9
E. frumentacea79 ± 1052 ± 17673 ± 112339 ± 90672 ± 86620 ± 622.2 ± 1.4/3.62.1 ± 1.4/3.32.0 ± 1.3/2.92.1 ± 1.4/3.3
E. aristidea43 ± 968 ± 9399 ± 80471 ± 122540 ± 162609 ± 974.2 ± 2.7/6.54.5 ± 3.3/6.13.8 ± 2.6/5.73.1 ± 2.7/3.5
I. afrum47 ± 1269 ± 9218 ± 17368 ± 102389 ± 159296 ± 833.9 ± 2.5/6.05.1 ± 3.7/6.93.1 ± 2.2/4.42.9 ± 2.2/3.9
P. malacophyllum37 ± 739 ± 3341 ± 46438 ± 144358 ± 95592 ± 1429.2 ± 7.6/11.27.8 ± 6.0/10.07.8 ± 7.0/8.83.5 ± 2.2/5.5
Weighted average52 ± 861 ± 9415 ± 62496 ± 95468 ± 123513 ± 1004.5 ± 3.2/6.34.6 ± 3.4/6.23.9 ± 2.9/5.33.2 ± 2.4/4.3

The pattern in leaf chlorophyll content (Table 2) matched that seen for Narea. In the control treatment, differences between species (F12,102 = 61.7, P < 0.001) were consistent with significantly lower chlorophyll contents in the leaves of C4 species than in C3 species (t102 = 3.94, P < 0.001, Table 2). As for Narea, although the drought treatment resulted in a significant increase in average values for chlorophyll content (F1,102 = 6.2, P = 0.015), this was underlain by changes in chlorophyll content that depended upon species (drought × species F12,102 = 3.58, P < 0.001), but were independent of photosynthetic type (t102 = −0.75, P = 0.449).

The average total number of leaves (live and dead) on a tiller differed between species within the control treatment (loge-transformed data, F12,111 = 3.3, P < 0.001). Contrasts indicated no significant difference in mean values between the C3 and C4 groups (t111 = 0.14, P = 0.256) and, based on the AIC criterion, the effect of drought was dropped from the model. Thus, at the most acute phase of the drought treatment, there was no indication that drought had induced a reduction in the rate of leaf accumulation, and tillers possessed similar numbers of leaves regardless of treatment.

The number of living leaves per tiller showed a significant relationship with the total number of leaves (both axes loge-transformed, F1,109 = 1497.9, P < 0.001) that was consistent across species and treatments (interaction effects were dropped from the minimal model). On average, a doubling of the total number of leaves on a stem resulted in only a 77% increase in the number of living leaves, showing that as stem length increased, further stem extension was associated with leaf mortality and suggesting that for species with more leaves per stem, leaf turnover was higher. The interaction between drought and species was dropped from the minimal model, but between species differences in intercepts suggested that, on average, C3 species (t109 = −0.11, P = 0.035) had 10% fewer living leaves per stem than C4 species. Drought also produced a small but significant reduction in the number of living leaves per stem (average 7% decline; F1,72 = 6.1, P = 0.016), i.e. leaf mortality increased. On the original scale, these differences in log-scaled relationships corresponded to an increasing difference in the number of living leaves with stem length, both between C3 and C4, and between drought and control plants within each photosynthetic type. This effect was, therefore, consistent with a greater decline in the number of living leaves per stem in C4 species, which had a greater number of living leaves in the control condition, than in C3 species: at the average stem length of ∼6 leaves, the difference in the number of living leaves between C3 and C4 species was reduced by ∼7% in the drought treatment relative to the control treatment. However, the size of these effects was small when compared with between species differences (C3/C4 difference ∼26% of between-species range in control treatment, 3.0 to 4.6, for predicted number of living leaves for a stem with six leaves) and they were only meaningful, in terms of predicted whole-leaf counts, for plants whose stems had a large total number of leaves.

To determine whether differences in the leaf chlorophyll content in response to drought were influenced by changes in whole stem properties, mean chlorophyll content was calculated over all of the living leaves on each stem (Table 2) and the total number of leaves per stem (loge-transformed) was tested as a covariate with species and drought effects. The latter were both dropped from the minimal model, and mean chlorophyll content in the living leaves showed a significant positive relationship with the number of leaves on a stem (F1,122 = 5.3, P = 0.023).

Scaling of leaf N with SLA

The scaling of Nmass against SLA (Fig. 6) showed significant differences between the C3 and C4 types, but drought did not affect the relationship in either photosynthetic type. The scaling slope was >1 for the C4 species (slope, +/−95% c.i. control 1.54, 1.15/2.06; drought 1.56, 1.15/2.12), but was not significantly different from 1 for the C3 species (control 1.02, 0.77/1.35; drought 0.95, 0.68/1.32). Differences in Nmass for the C3 species were, therefore, in proportion with differences in SLA, but for C4 species, differences in Nmass were proportionally greater than inter-specific differences in SLA.

Figure 6.

Scaling relationship between foliar nitrogen concentration (Nmass) and specific leaf area (SLA) in week five, loge-scaled axes. Mean ± SEM by species, n = 2 to 5; C4 control inline image, C4 drought inline image, C3 control inline image, C3 drought inline image. Best-fit lines are standardized major axis relationships, accounting for average measurement error. The slope between C3 species differed from that between C4 species (L3 = 8.89, P = 0.031). There was no significant difference in slope between treatments within each photosynthetic type (C3, L1 = 0.11, P = 0.742; C4, L1 = 0.01, P = 0.934). There was a significant difference in the intercept between treatments for the C4 species (w1 = 4.14, P = 0.042), but not the C3 species (w1 = 2.24, P = 0.135).

Although mean SLA values overlapped between C3 and C4 species in both drought and control treatments, they were always greater than 150 cm2 g−1 for C4, whereas some C3 species showed values between 70 and 100 cm2 g−1. In contrast, the range of Nmass values showed strong overlap (Fig. 6). The range of values for Nmass was therefore conserved, whilst for SLA, the range differed between the two photosynthetic types. If the scaling relationship for C4Nmass/SLA were to be projected over the entire 6-fold range of species means for SLA (i.e. across both C3 and C4 species), it would result in a 14-fold range of Nmass C4 species, rather than the observed four-fold difference and mean Nmass values for species at the bottom of this range would be approximately 0.25 mmol g−1, i.e. 0.35%.


We found that the photosynthetic advantage of C4 over C3 species under mesic conditions was reduced under drought. High iWUE in C4 species under mesic conditions resulted from higher A and lower gs than in C3 grasses, consistent with the well-known capacity of C4 photosynthesis to draw down ci to a greater extent than C3 photosynthesis (Long 1999). Under most conditions, the exchange of CO2 and H2O between atmosphere and leaf occurs predominantly via the stomatal pathway, and the capacity to draw down ci through photosynthesis is equivalent to iWUE. Drought forced a decrease in A in both C3 and C4 grasses, but its most striking effect on gas exchange was the near elimination of differences in gs between the C3 and C4 species. The resultant increase in iWUE observed for several C3 species closed the gap between them and their C4 counterparts under drought conditions.

Under adequate water supply, the initial slopes of A-ci relationships for C4 species are steeper than those of C3 species (Pearcy & Ehleringer 1984). Consequently, with no change in the shape of this relationship, reduced CO2 supply due to decreased stomatal aperture would force a greater decline in A for C4 than C3 species (Björkman 1971). The mismatch observed between the size of the decreases in gs (smaller in C4) and in A (equivalent in C3 and C4) is consistent with this theoretical expectation. However, declining gs under drought should also drive down ci, and we found that overall, this was not the case for C4 species, an effect that implies a change in the shapes of Aci relationships, i.e. declines in A relative to CO2 supply. Our findings are, therefore, consistent with those of previous studies restricted to Panicoideae species (Ripley et al. 2007, 2010; Ibrahim et al. 2008) and support the hypothesis that drought-limitation of photosynthetic capacity is greater in C4 NADP-me than C3 grasses. Here, we find that this effect can be generalized to C4 NADP-me species across the PACMAD clade.

However, certain factors must be considered when interpreting trends in ci values as indicators of photosynthetic performance under drought (Lawlor & Cornic 2002). Firstly, calculation of ci requires an accurate description of gs, which, due to the need to obtain accurate estimates of boundary layer conductance (gbl), may be difficult to achieve for narrow leaved species. Secondly, when estimating leaf conductance to CO2, it is normally assumed that cuticular conductance represents a negligible component of leaf gas exchange, an assumption which may be violated under drought (Boyer, Wong & Farquhar 1997). Finally, the value of ci is a spatially averaged mean, weighted by leaf conductance, that behaves as a strong index of photosynthetic performance when the spatial distributions of gs and A are homogeneous across the entire leaf, but may perform poorly in response to heterogeneity (Downton, Loveys & Grant 1988; Terashima et al. 1988; Farquhar 1989). Heterogeneity may result from barriers to lateral (e.g. bundle sheath extensions, Metcalfe 1960) and dorso-ventral diffusion of CO2 (Long et al. 1989) that are known to occur in the leaves of grasses, and heterogeneous gas exchange has been reported for maize (Terashima 1992). Differences in the occurrence of these effects between C3 and C4 species could plausibly explain our observation of a differential effect of drought on ci.

Alternatively, effects on ci may result from drought limitation of photosynthetic capacity. Under mild to severe water shortage, photosynthetic capacity of C3 species is maintained and reductions in gs force reversible declines in ci (Lauer & Boyer 1992; Brodribb 1996). However, decreases in leaf water status (measured as relative water content or leaf water potential) will eventually cause metabolic limitations of photosynthetic capacity (Lawlor & Cornic 2002), and the mechanisms underlying effects of leaf water status on photosynthetic metabolism are expected to be qualitatively different between C3 and C4 plants (Ghannoum 2009). We found no strong evidence for a larger response of leaf water potential to drought in C4 plants when compared with C3 species, and the latter had more negative leaf water potentials in our control treatment. Any proposed metabolic effects on photosynthesis in C4 species would, therefore, have occurred despite similar or less negative leaf water potentials than were experienced by C3 species.

It has been suggested that lower gs and E associated with C4 photosynthesis may result in adaptation of plant hydraulics (Kocacinar & Sage 2003). Our observation of less negative Ψop and smaller ΔΨ in C4 than C3 species in control treatments is consistent with other recent work implying a higher capacity for hydraulic supply relative to demand in the C4 species (Taylor et al. 2010). Few species showed large changes in ΔΨ in response to our drought treatment, suggesting anisohydric-isohydrodynamic behaviour (Franks, Drake & Froend 2007). Assuming that large changes in ΔΨ were representative of impaired hydraulic function, our results show that this was no more common in C3 grasses than in their C4 relatives. The observation that ΔΨ was frequently maintained under drought is interesting from two perspectives. Firstly, it suggests a link between Ψsoil and Ψop that, during the progressive development of drought in natural soils, might result in a progressive onset of metabolic effects of declining Ψop. Secondly, it occurred in the context of much larger decreases in gs for C3 than for C4 species. Assuming a strong relationship between gs and E, in addition to differences in whole-plant conductance (kplant = EΨ) observed between C3 and C4 species under mesic conditions, the response of kplant to drought will be greater in C3 grasses than in their C4 relatives.

We found no strong evidence that our drought treatment reduced the chlorophyll and nitrogen content of leaves measured for gas exchange. This suggests that senescence did not drive observed decreases in A or PNUE in response to drought. However, at the whole plant scale, drought-induced leaf mortality was detected, and was most important in those species with longer-stems or C4 photosynthesis, with the result that in the drought treatment, differences between C3 and C4 species in the number of living leaves per stem were reduced. Thus, changes in leaf area available for photosynthesis may also contribute to effects of drought on whole-plant photosynthesis and transpiration in C4 species compared with C3 relatives under drought.

We have previously found evidence that the range of values for Nmass largely overlapped when comparing PACMAD C3 and C4 grasses, and differences in Narea between photosynthetic types were influenced by SLA (Taylor et al. 2010). The results of the present experiment further support that result. Conservation of the range of Nmass values across species, paired with greater conservatism in SLA between C4 species, provided a strong contribution to differences in Narea between photosynthetic types. Should conservatism in SLA between C4 species be a general trend, the extent of its relationship with C4 Kranz anatomy will be of interest. For example, for a comparison made between Paniceae species, C3 grasses had thinner leaves at a given vein density and reduced vein density in wider leaves, whilst their C4 relatives showed no relationship between vein density and leaf width (Oguro, Hinata & Tsunoda 1985).

By concentrating on the NADP-me biochemical subtype of C4 photosynthesis, we excluded several subtypes of C4 photosynthesis from our comparisons. Drought tolerance has been shown to differ between C4 subtypes (Ghannoum, von Caemmerer & Conroy 2002), an important effect to be considered in developing truly global conclusions on the drought tolerance of C4 photosynthesis. Nonetheless, we provide evidence for a differential response of photosynthetic types to drought. A difference in Ψop when water supply is abundant is coupled with lower gs, and smaller changes in gs in response to changing soil-water conditions in NADP-me grasses than in their C3 relatives. Drought reduced both the PNUE and iWUE advantages commonly observed in C4 species, and produced a slight increase in leaf mortality in C4 species. Our results therefore indicated that despite conservative regulation of gs, C4 NADP-me grass species are no more robust to low water availability than their C3 relatives. In fact, drought-induced limitations of photosynthesis may become evident at less negative leaf water potentials in C4 NADP-me species than in C3 species. These results provide experimental evidence that complements comparative studies of ecological adaptation in C4 grasses (Osborne & Freckleton 2009; Edwards & Smith 2010). Our finding that C4 species are no more tolerant of a ‘drought’ treatment imposed under controlled growing conditions, and the impacts that this has upon their photosynthetic advantages over their C3 relatives, emphasize the need for studies addressing physiological contrasts between related C3 and C4 species in native situations.


The authors thank Peter Franks, Andrew Leakey and Matthew Gilbert for discussions of gas exchange results; Mark Rees for advice on statistical analyses; Pascal-Antoine Christin for his phylogenetic tree; Irene Johnson for advice on plant rearing, and Hui Liu, Stephen Hulme and Sarah Wilkinson for technical assistance. Research was funded by the NERC grant NE/DO13062/1 awarded to CPO and FIW, and a Royal Society University Research Fellowship awarded to CPO.