Gene discovery in cereals through quantitative trait loci and expression analysis in water-use efficiency measured by carbon isotope discrimination



    1. Department of Renewable Resources, 442 Earth Sciences Building, University of Alberta, Edmonton, Alberta, Canada T6G 2E3
    2. Department of Landscape Studies, College of Architecture and Urban Planning, Tongji University, #1239 Siping Road, Shanghai, China, 20092
    Search for more papers by this author

    1. Department of Renewable Resources, 442 Earth Sciences Building, University of Alberta, Edmonton, Alberta, Canada T6G 2E3
    Search for more papers by this author

    Corresponding author
    1. Alberta Innovates – Technology Futures, Vegreville, Alberta, Canada T9C 1T4
    Search for more papers by this author

A. O. Anyia. Fax: +1 780 632 8620; e-mail:; S. X. Chang. Fax: +1 780 492 1767; e-mail:


Drought continues to be a major constraint on cereal production in many areas, and the frequency of drought is likely to increase in most arid and semi-arid regions under future climate change scenarios. Considerable research and breeding efforts have been devoted to investigating crop responses to drought at various levels and producing drought-resistant genotypes. Plant physiology has provided new insights to yield improvement in drought-prone environments. Crop performance could be improved through increases in water use, water-use efficiency (WUE) and harvest index. Greater WUE can be achieved by coordination between photosynthesis and transpiration. Carbon isotope discrimination (Δ13C) has been demonstrated to be a simple but reliable measure of WUE, and negative correlation between them has been used to indirectly estimate WUE under selected environments. New tools, such as quantitative trait loci (QTL) mapping and gene expression profiling, are playing vital roles in dissecting drought resistance-related traits. The combination of gene expression and association mapping could help identify candidate genes underlying the QTL of interest and complement map-based cloning and marker-assisted selection. Eventually, improved cultivars can be produced through genetic engineering. Future efficient and effective breeding progress in cereals under targeted drought environments will come from the integrated knowledge of physiology and genomics.


Water deficit or drought continues to be one of the major factors limiting crop production and productivity in many regions and a potential threat to food security in the 21st century (Tuberosa, Gill & Quarrie 2002). As the world population continues to grow with future global climate change, how to make good use of limited water resources for crop production has become a worldwide concern (Polley 2002; Barnabás, Jäger & Fehér 2008). Moreover, the frequency, duration and severity of drought are variable and unpredictable across seasons and locations, especially in arid and semi-arid regions, causing highly unstable crop yields (CYs) between years and locations. To ensure sustainable food supply, breeding efforts should not only focus on maintaining yield, but more importantly on developing high yielding varieties under water-limited environments, that is, to achieve more production per unit of available water, in other words, ‘more crop per drop’ (Passioura 2006).

Breeding for drought-resistant and water-use efficient crop varieties has been a critical area of agricultural research worldwide. Substantial efforts have been devoted to identifying and selecting for morphophysiological traits that increase water-use efficiency (WUE) and yield under rain-fed conditions (Blum 1996; Richards 1996; Richards et al. 2002). Nevertheless, the complex underlying mechanisms of drought resistance and our lack of knowledge of the genetic and physiological bases of yield have hindered the breeding process in drought environments (Passioura 2002; Tuberosa & Salvi 2006). In the last decade, plant breeding has been greatly advanced by progress in stress physiology and genomics (Araus et al. 2003; Morgante & Salamini 2003; Habash, Kehel & Nachit 2009). For example, the biomass and WUE of transgenic wheat were improved under water deficit conditions by introducing a Group 3 late embryogenic abundant (LEA) proteins encoded by the barely HVA1 gene (Sivamani et al. 2000). A gene (HARDY) identified from Arabidopsis improved WUE in rice by enhancing photosynthetic assimilation and reducing transpiration (Karaba et al. 2007). Another example is represented by the studies in WUE using carbon isotope discrimination (Δ13C). The Δ13C has been demonstrated to be a simple but reliable measure of WUE, and the negative correlation between them has been used as an indirect method for the selection of C3 crops with improved WUE under selected environments (Cattivelli et al. 2008).

Drought resistance is a complex trait that is governed by quantitative trait loci (QTL). Selection efficiency of drought-resistant traits could be enhanced with a better understanding of its genetic control. The advancement of DNA-based molecular markers and computational methods in the late 1980s and 1990s has revolutionized the dissection of quantitative trait inheritance and genetic improvement of yield in dry environments (Baum et al. 2007). The QTL mapping and analysis provides unprecedented opportunities to identify and locate chromosome regions controlling adaptive traits such as Δ13C during plant growth in water-limited conditions. To further unravel the molecular basis of a QTL, two approaches are usually deployed: positional cloning (also termed map-based cloning or recombinational mapping) and association mapping (Tuberosa & Salvi 2004). Comparative analysis of QTL results also provides valuable opportunities for positional cloning and to identify candidate genes for WUE in cereals through exploration of related species using sequence co-linearity from small (such as rice, Oryza sativa L.) to large genomes (such as wheat, Triticum aestivum L.) (Sorrells, Diab & Nachit 2000); for example, QTL for Δ13C was identified near the semi-dwarf gene sdw1 in different barley mapping populations (Ellis et al. 2002; Teulat et al. 2002), which is homoeologous to rice chromosome 1 where Δ13C is located. However, QTL mapping only represents the first step towards detecting genes affecting traits of interest (Frary et al. 2003). The ultimate goal is to assign functions to genes and manipulate either a single gene or pyramiding beneficial QTL alleles through mark-assisted selection (MAS) to derive improved varieties and speed up plant breeding under water-limited environments. The most promising breeding strategy of improving WUE in cereals lies in the integration of QTL mapping, genetic engineering, and conventional breeding. Current advances and future research directions in the improvement of WUE are discussed in this literature review.

Physiological traits related to drought adaptation

Periods of soil and/or atmospheric water deficits pose critical constraints on plant survival and productivity (Boyer 1982), especially under future climate-change scenarios (Petit et al. 1999). Generally, for coping with water scarcity plants have evolved different mechanisms, such as escape, avoidance and tolerance, to ensure their survival and reproduction (Turner 1986; Chaves, Maroco & Pereira 2003). The escape strategy could be attained by changing the phenology and growth rate before the onset of water deficit, such as shorter life cycles (Chaves et al. 2003; Barnabás et al. 2008). Dehydration avoidance is achieved by maintaining relatively high tissue water potential during drought stress, either by minimizing water loss (e.g. stomatal closure, reduced leaf area, leaf rolling, senescence of older leaves, etc.) or maximizing water uptake (e.g. increased root growth, increased hydraulic conductance, etc.) (Jackson, Sperry & Dawson 2000; Mitra 2001). Dehydration tolerance is the plant's ability to withstand severe water deficit, maintain function at low water potential, recover water status and resume full function once water becomes available; this mechanism involves osmotic adjustment (increased solute concentration), increased cell elasticity or decreased cell size (Mitra 2001). Associated with those strategies, plants exhibit multiple adaptations or changes at developmental, phenological, morphological, biochemical and physiological levels. Physiological traits that have been intensively investigated include osmotic adjustment, osmotic potential, water soluble carbohydrates, stomatal conductance, canopy temperature, relative water content, transpiration efficiency, WUE, Δ13C, early vigor, leaf area index, stomatal density, flowering time, root characteristics and so on (Baum et al. 2007).

Genetic gain is slow when directly selecting for increased grain yield under water-limited environments because of a large interaction between genotype and environment and a low heritability of yield components. However, the encouraging heritability of physiological or secondary traits that are highly correlated with yield presents a good opportunity for plant breeding in drought-prone regions (Stiller et al. 2005). WUE is a trait that has been proposed as a criterion for yield improvement under drought conditions (Condon & Richards 1992; Condon et al. 2002; Rebetzke et al. 2002; Richards et al. 2002).

WUE in crops

The term WUE may be defined in different ways (Table 1). At the photosynthetic scale (leaf-level), instantaneous WUE (WUEinstantaneous or WUEis), also referred to as transpiration efficiency (TE), is generally measured as the net amount of carbon assimilated (A) per unit water transpired (transpiration rate, E) during the same period (Farquhar & Richards 1984; Farquhar, Ehleringer & Hubick 1989; Condon et al. 2002; Polley 2002; Bacon 2004). A similar parameter, intrinsic WUE (WUEintrinsic or WUEic) is defined as the ratio between A and stomatal conductance (gs) (Hall, Mutters & Farquhar 1992; Choi, Chang & Bhatti 2007), which is thought to be more closely associated with physiological responses caused by independency of specific environmental conditions. The following equation gives the relationship between A/E and A/gs (Farquhar & Richards 1984).

Table 1.  Several common definitions of water-use efficiency (WUE)
LevelTime scaleNumeratorDenominatorEquation
Leaf (photosynthetic scale)Minutes or hoursNet assimilation rate (A) µmol CO2 m−2 s−1Transpiration (E) mmol H2O m−2 s−1WUEInstantaneous = A/E
A µmol CO2 m−2 s−1Stomatal conductance (gs) mmol H2O m−2 s−1WUEIntrinsic = A/gs
Crop (agronomic scale)Weeks to months or growing seasonAboveground biomassSeasonal evapotranspirationWUEintegrative = dry matter/transpired water
Grain yieldSeasonal evapotranspirationWater productivity or WUEeconomic = dry matter/transpired water

where ca and ci are atmospheric and leaf intercellular CO2 concentrations, gc and gw are stomatal conductance to diffusion of CO2 and water vapour and (wi − wa) or Δw and (ca − ci) refer to the concentration gradients of water vapour and CO2 between the outside (wa, ca) and the inside (wi, ci) of leaf, and 1.6 is the relative diffusivities of water vapour to CO2 in air.

Equation 1 indicates that WUEis or TE and WUEic are negatively correlated with the ratio of ci to ca. A greater WUE at the leaf-level can be achieved through a lower value of ci/ca either by decreasing gs, or increasing A, or a combination of both (Condon et al. 2002; Polley 2002).

For agronomists and plant breeders, WUE is typically calculated as the accumulated dry matter produced divided by the amount of water consumed by the crop during the whole growth cycle (WUEintegrative or WUEie) (Condon et al. 2004; Tuberosa et al. 2007).


where TE is the transpiration efficiency (above-ground dry matter/transpiration), Es is the water lost by evaporation from the soil surface and T is water lost through transpiration by the crop.

In Eqn 2, the denominator could include water loss from run-off during heavy rainfall events and deep drainage below the root zone, depending on the cropping system (Condon, Richards & Farquhar 1993; Gregory, Simmonds & Warren 1997). From Eqn 2, WUEie could be improved either by increasing TE or maximizing T by reducing Es through agronomic management practices (Gregory et al. 1997; Richards et al. 2002; Passioura 2006).

Usually the real-time leaf parameters of WUE are measured by gas exchange methods, that is, infrared gas analysers and porometers (Long, Farage & Garcia 1996). Estimations of WUEie might be performed in greenhouse and growth chamber pot studies by minimizing soil evaporation (Lambrides, Chapman & Shorter 2004; Turner et al. 2007). Accurate water budget and crop biomass measurements required for WUE estimation is labour intensive, time-consuming and expensive, and therefore unattractive to plant breeders, especially under field conditions for large populations. The WUE calculated as grain yield divided by water supply (kg ha−1 mm−1 or kg m−3) may be underestimated because it assumes that the same amount of water is transpired by each genotype.

Carbon isotope discrimination as a measure of WUE

While it is widely recognized that improved WUE can enhance yield in certain environments, the use of the WUE trait in breeding programs has largely been limited because of the lack of a suitable screening method in large populations. A promising screening method for WUE came into the picture in the 1980s when carbon isotopic techniques were introduced (O'Leary 1981; Farquhar, O'Leary & Berry 1982).

Approximately 1.1% of the carbon in the biosphere naturally occurs in the form of the stable isotope 13C (O'Leary 1981; Farquhar et al. 1989; Condon et al. 2002). However, the molar abundance ratio of 13C/12C (R) in plant tissues usually is less than that in atmospheric CO2 because of discrimination against the ‘heavier’13C (lower reactivity) during photosynthesis (O'Leary 1981; Farquhar et al. 1982; Farquhar & Richards 1984; Farquhar et al. 1989). Plants with different metabolic pathways of carbon assimilation exhibit characteristically different discrimination against 13C when incorporating CO2 into plant tissues. The isotope discrimination commonly involves diffusion of CO2 across atmospheric boundary layer and stomata into leaf mesophyll, its inter-conversion of dissolved CO2 and HCO3-, and enzymatic incorporation of CO2 into carbohydrates by ribulose 1,5-bisphosphate carboxylase-oxygenase (Rubisco) in plants with conventional (C3) pathway (O'Leary 1981; Farquhar et al. 1982; Hattersley 1982).

Isotope compositions are usually determined by an online continuous-flow stable isotope ratio mass spectrometer (IRMS). The ratio of 13C/12C in a sample of plant is converted to δ13C (carbon isotope composition) commonly compared with a reference material, the belemnite carbonate standard from the Pee Dee Formation (PDB) in South Carolina (Craig 1953; O'Leary 1981; Ober et al. 2005). The stable carbon isotope composition of plant samples is calculated as:


where Rsample and Rstandard are the 13C/12C ratios measured in the plant material and the standard.

On the PDB scale, the δ13C value for free atmospheric CO2 currently is approximately −8‰ (Farquhar et al. 1989), and this value is becoming slightly more negative over time because of the increasing combustion of fossil fuels (O'Leary 1981). According to Feng (1998), δ13Cair has decreased from −7.4 to −8.2‰ from 1976 to 2003. The δ13C of plants is a negative value because the 13C/12C ratio in the atmosphere is less than that in PDB, and also there is a net discrimination against 13CO2 during diffusion and carboxylation by plants (Farquhar et al. 1982; Condon et al. 2002). The amount of carbon isotope discrimination is generally expressed as:


where Rair and Rplant refer to 13C/12C ratios of the atmosphere and plant samples.

Plants show positive values of Δ13C and typically C3 plants have a discrimination rate around 20‰ during photosynthesis (O'Leary 1981; Farquhar et al. 1982; Farquhar et al. 1989).

Ignoring dark respiration and photorespiration, assuming that the major components contributing to the final discrimination are gaseous diffusivities across the boundary layer and stomata (a) and the fractionation by Rubisco during the carboxylation (b), an approximate expression for the overall Δ13C in leaves for C3 plants during photosynthesis has been developed and described as (Farquhar et al. 1989):


where a and b in Eqn 5 are as mentioned earlier.

Equation 5 shows that Δ13C is positively related to the ci/ca ratio, which was therefore negatively correlated with WUEis or TE as expected from Eqn 1 (O'Leary 1981; Farquhar et al. 1982; Farquhar & Richards 1984; Farquhar et al. 1989; Hubick & Farquhar 1989). Farquhar & Richards (1984) first reported that the extent to which plants discriminate against the 13C during gas exchange was negatively correlated to WUE in a greenhouse experiment with wheat grown in large pots.

Currently Δ13C is widely used as an indirect assessment of WUE in C3 crops under water-limited conditions. Extensive studies in C3 species have been reported and have confirmed the negative relationship between Δ13C and WUE (Hubick, Farquhar & Shorter 1986; Hubick & Farquhar 1989; Condon, Farquhar & Richards 1990; Hall et al. 1992; Ismail & Hall 1992; Rebetzke et al. 2002; Lambrides et al. 2004; Rytter 2005; Khan et al. 2007). This relationship in C3 plants has opened up the prospect of utilizing differences in 13C discrimination for selecting crops that have high WUE under specific environments.

Sufficient genotypic variation, stability across environments, and a high broad-sense heritability (H2) in Δ13C indicate that it is a promising surrogate for WUE that can be applied in breeding programs (Condon & Richards 1992; Schuster et al. 1992; Johnson & Rumbaugh 1995; Araus et al. 1998; Merah et al. 2001; Rebetzke et al. 2008). The high H2of Δ13C has been previously reported in barley (Hordeum vulgare L.; Voltas et al. 1998; Çağirgan et al. 2005), cotton (Gossypium hirsutum L.; Stiller et al. 2005), cowpea (Vigna unguiculata L. Walp.; Hall et al. 1990), peanut (Arachis hypogaea L.; Hubick, Shorter & Farquhar 1988), soybean (Glycine max L.; Specht et al. 2001), bread wheat (Ehdaie et al. 1991; Condon & Richards 1992; Rebetzke et al. 2008) and durum wheat (Triticum durum Desf.; Araus et al. 1998). Non-significant interactions between genotype and environment for Δ13C have been reported for peanut (Hubick et al. 1988), sugar beet (Beta vulgaris L.; Rajabi, Ober & Griffiths 2009) and wheat (Triticum turgidum ssp. dicoccoides; Peleg et al. 2009).

However, Δ13C is not widely applicable to species with C4 and crassulacean acid metabolism (CAM). CAM plants exhibit either C3 or time-separated C4 fixations, showing large variations in isotopic compositions (O'Leary 1981). The mechanisms for Δ13C changes in C4 plants are complex, involving HCO3- fixation in oxaloacetate by the phosphoenolpyruvate carboxylase enzyme (PEPC), CO2 release in the bundle-sheath cells and refixation by Rubisco (Farquhar et al. 1989). The major difference in 13C/12C ratio between C3 and C4 plants is the isotopic fractionation activity between Rubisco and PEPC. PEPC fixes HCO3-, which is 13C-enriched compared with CO2 and a proportion of carbon fixed by PEP carboxylation that subsequently leaks out of the bundle sheath, which reduces the Rubisco discrimination against 13C. Average Δ13C is around 4‰ in C4 plants (Deléens, Ferhi & Queiroz 1983). Monneveux et al. (2007) explored the possibility of using Δ13C as a selection criterion for yield under drought in maize (Zea mays L.), with Δ13C analysed in different organs at flowering stage under both drought and irrigated conditions. However, they did not find any correlation between Δ13C and grain yield within tolerant hybrids, probably because variation of Δ13C is less affected because ci/ca is more stable under drought conditions. Cabrera-Bosquet, Sánchez & Araus (2009b) reported that there was no significant relationship between either leaf or kernel Δ13C and grain yield. Further research is needed for understanding the relationship between Δ13C and WUE in species with C4 metabolism.

Although Δ13C has been intensively exploited as an integrated criterion for screening improved WUE and thus greater productivity in C3 crops under water-limited environments, there are some challenges associated with the application of Δ13C in plant breeding programs.

  • 1Δ13C provides a long-term average estimate of cumulative WUE integrating in time and space without giving any information about the change in WUE as a result of altered A or gs or both (Condon et al. 2002).
  • 2The relationship between Δ13C and grain yield or biomass is either positive, negative or neutral, depending on the season, location and species (Condon et al. 2004; Jiang, Roche & Hole 2006; Anyia et al. 2007; Tambussi, Bort & Araus 2007; Chen, Chang & Anyia 2011). Positive or neutral relationships between Δ13C and grain yield or biomass are often reported in environments characterized with plentiful within-season rainfall or supplemental irrigation, such as wheat and barley grown under Mediterranean climates (Condon et al. 1993; Merah, Deléens & Monneveux 1999; Voltas et al. 1999; Teulat, Merah & This 2001; Teulat et al. 2002; Araus et al. 2003; Jiang et al. 2006), while negative relationships are found in environments where crop relies heavily on stored soil moisture (Condon et al. 1993; Rebetzke et al. 2002; Anyia et al. 2007). However, Δ13C values were not a reliable predictor under severe stress (Jiang et al. 2006).
  • 3Timing for sampling can be difficult to determine. Plant materials could be collected during different developmental stages, such as vegetative phase or maturity. Age and stress can affect Δ13C variation in plants at different times (Francey et al. 1985), such as Δ13C of rice was reduced greatest at tillering than at flowering and maturity when subjected to water stress (Zhao et al. 2004). Condon & Richards (1992) proposed that it would be most effective to assess Δ13C at early stages in plant development under well-watered conditions.
  • 4Which part of the plant should be collected for Δ13C analysis? Plant samples can be collected from root, leaf, sheath, awn or grain, each characterized with its own Δ13C value. Jiang et al. (2006) reported that Δ13C were the highest in flag leaf, intermediate in awn, and lowest in grain in barley. Zhao et al. (2004) also found that in rice root and grain had the lowest Δ13C values, and stem the highest. Different parts have their own potential advantages. Leaves sampled for Δ13C at the stem elongation stage, when there is usually little drought stress and low vapour pressure deficit, could better reflect the integrated WUE during vegetative development and formation of yield potential (Condon & Richards 1992; Anyia et al. 2008; Chen et al. 2011). In addition, leaf Δ13C can be measured before maturity enabling selection and crosses to be made within the same season thereby speeding up the breeding process. However, grain Δ13C was preferred in many studies under Mediterranean-type environments because of its positive relationship with yield (Condon et al. 2004).
  • 5Location conditions can have a marked effect on Δ13C. The δ13C values can be more negative in greenhouse than in field studies because of the contribution of respired CO2 (O'Leary 1981).
  • 6Genotypic difference introduces another level of complexity in the use of Δ13C for evaluating WUE. Most of the studies carried out in cereals showed substantial Δ13C variation in genotypes that differed in flowering time and plant height, and those two characteristics strongly affected yield and generated complex associations between Δ13C and productivity (Condon et al. 2004).
  • 7Large amounts of carbon lost by photorespiration in C3 plants could affect the final Δ13C value in plant tissues. Leaf WUE usually was underestimated without including respiration rates (Tambussi et al. 2007). Water isotopes such as oxygen and hydrogen isotope compositions (expressed as δ18O and δ D) in plant tissues might be alternative indicators of TE (Farquhar, Cernusak & Barnes 2007). The complementary measurement ‘packages’ of stable isotopes such as hydrogen, carbon and oxygen in plant substrates and material could provide more insights to the physiological and biochemical responses of plants to water deficit, such as combining Δ18O and Δ13C to assess plant growth and total transpiration (Cabrera-Bosquet et al. 2009a).

Breeding for improved WUE

There is an increasing urgency in plant breeding for improved CY potential and better adaptation to current and future prolonged aridity (Araus et al. 2002). Great progress in major cereals has been made through empirical (also termed conventional or traditional) breeding programs during the last 50 years by directly selecting a primary trait (such as grain yield); however, progress in traditional breeding has been slow because of the variable nature of drought and the complexity of drought resistance mechanisms (Araus et al. 2008). Selection for a secondary trait that is putatively related to a higher yield potential or a limiting yield factor is called analytical, physiological or indirect breeding, which has been very popular (Baum et al. 2007).

A framework based on maximizing grain yield instead of survival was proposed by Passioura (1977). Under water-limited environments, CY is a function of water use (WU; evapotranspiration), WUE (water truly transpired for biomass growth) and harvest index (HI; i.e. the ratio of grain yield to aboveground biomass):


WU or evapotranspiration includes crop transpiration and soil evaporation. Accordingly, CY in dry environments can be improved by increasing: (1) the capacity to capture more water, either through improving soil water uptake ability or decreasing soil evaporation; (2) the ability to produce more dry matter per unit of water used; and (3) the ability to deliver more assimilates into economic yield (Turner 2001; Araus et al. 2002). None of these components is completely independent, and improvement in any of the components could potentially increase CY (Araus et al. 2002; Richards et al. 2002; Condon et al. 2004; Tambussi et al. 2007).

Richards et al. (2002) pointed out that WU is a function of evaporative demand and leaf area. Mediterranean environments are typically characterized by frequent rainfall during vegetative growth and terminal drought during grain filling, and therefore reducing soil evaporation can provide benefit. Early germination, rapid seedling establishment, and good canopy cover (i.e. early or seedling vigor) together with higher specific leaf area (SLA) have been suggested to play an important role in reducing soil evaporation under Mediterranean-type environments (Tambussi et al. 2007). Genotypes with good early vigor tend to have deep rooting systems and exhibit great soil water extraction capacity (Turner & Asseng 2005), which is an attractive trait for effective use of water under most drought conditions (Blum 2009). For winter-grown crops that rely on stored soil water, restricted leaf area reduces transpiration and conserves soil moisture thereby contributes more water for late grain filling (Rebetzke et al. 2002).

Basically a high HI for a given genotype sets its genetic potential for high yield (Cattivelli et al. 2008). According to Richards et al. (2002), the achievement of a high HI depends on the balance between pre-anthesis and post-anthesis WU under rain-fed conditions. Cereals have achieved great yield improvement through increasing HI by reducing plant height, a process primarily caused by the introgression of semi-dwarfing genes since the second half of the last century (Slafer, Satorre & Andrade 1994; Richards et al. 2002; Zhang 2007). However, the ceiling for genetic increases in CYs based on HI is likely being approached (Richards et al. 1993; Mann 1999; Richards et al. 2002).

Target at genetic improvement of WUE has long been attractive because it is potentially related to the other two components (Condon et al. 2004; Tambussi et al. 2007). Richards et al. (2002) also pointed out that selection for specific physiological and morphological traits in water-limited environments, such as a high WUE, could increase the rate of yield improvement. For example, two wheat cultivars, ‘Drysdale’ and ‘Rees’, have been commercially released in Australia during 2002 and 2003 that were selected for improved WUE based on their low Δ13C, and they have demonstrated a yield advantage compared with high Δ13C lines in environments with lower rainfalls (Rebetzke et al. 2002; Richards 2006). Breeding of hybrids in sunflower (Helianthus annuus L.) for high-yielding cultivars using Δ13C is currently in progress (Lambrides et al. 2004; Richards 2006). Studies with sunflower by Lambrides et al. (2004) suggest that there is a good potential for breeders to develop sunflower germplasm with improved WUE using Δ13C as a selection tool. In field evaluations, low-Δ13C hybrids of sunflower significantly out-yielded high-Δ13C hybrids in three of the four environments studied (Condon et al. 2004).

Although improved WUE and drought resistance without yield penalty offers a promising way to sustainable agricultural production and land use (Karaba et al. 2007), the application of WUE in plant breeding has been a subject of controversy. Blum (2005) has argued that selection for high WUE will result in small or early flowering plants, which achieved high WUE mainly through reducing WU without increasing yield. However, WUE is a ratio of yield to WU, and the different rates of reduction in these two components provide chances for manipulation under drought conditions (Blum 2005). A proportional change in both A and gs might have no effect on WUEic, while a comparable change in A with gs remaining constant would cause a substantial variation in WUEic, and vice versa. As proposed by Flexas et al. (2010), improved WUE could potentially be achieved through two possible approaches: (1) to increase CO2 diffusion to the carboxylation sites by maintaining gs, which could be attained by increasing mesophyll conductance to CO2 (gm); and (2) to improve the Rubisco carboxylation efficiency, which could be realized by introducing carboxylase enzyme from other species. Centritto et al. (2009) reported that rice genotypes with inherently higher gm were capable of maintaining higher A under water-deficit conditions. Galmés et al. (2011) reported that a tomato cultivar ‘Tomàtiga de Ramellet’ with drought resistance displayed higher WUEic under water-deficit conditions, which was positively correlated with gm/gs. So far, there are no reports about increasing WUE by genetically improving the biochemistry of photosynthesis, which is still possible given the rapid development of biotechnology and genetic engineering tools.

Although the relationship between Δ13C and grain yield is not consistent across seasons, sites and species (Condon et al. 2004; Jiang et al. 2006; Anyia et al. 2007; Tambussi et al. 2007; Chen et al. 2011), there are still plenty of opportunities for WUE selection. High Δ13C or low WUE cereal genotypes could be beneficial in Mediterranean, terminal-drought environments (Condon et al. 2004). In these types of environments, plants rely on current rainfall and wide stomatal opening is needed to transpire as much water as possible and to maintain growth when there is abundant rainfall. For stored-moisture environments such as eastern Australia and the Canadian Prairies, yield improvements through a combination of high WUE and greater early vigor are suggested (Condon et al. 2002; Anyia et al. 2008). Under these environments, genotypes achieve high WUE mainly by reductions in stomatal conductance, and thus conserve soil moisture during the vegetative growth stage for use in post-anthesis growth (Turner & Asseng 2005).

QTL analysis for WUE

In general, continuous genetic variation underlying quantitative traits, such as yield, plant height, flowering time, WUE and so on that are generally under considerable environmental influence, is governed by QTL (Hall et al. 1994; Li et al. 1995; Austin & Lee 1996; Juenger et al. 2005). QTL mapping usually provides a starting point for statistically exploiting and identifying the chromosomal regions contributing to genetic variation in agronomically important traits in breeding programs (Zhang 2007).

Understanding the genetic basis of WUE is important for crop improvement under water-limited environments. The first QTL identified for Δ13C was reported in tomato (Lycopersicon esculentum and L. pennellii) by Martin & Nienhuis (1989) and subsequently QTL for Δ13C have been reported in Arabidopsis thaliana (Hausmann et al. 2005; Juenger et al. 2005), barley (Ellis et al. 1997; Ellis et al. 2002; Teulat et al. 2002; Diab et al. 2004), cotton (Gossypium hirsutum and G. barbadense) (Saranga et al. 2001), rice (Laza et al. 2006; Takai et al. 2006; Xu et al. 2009; This et al. 2010), soybean (Specht et al. 2001), tomato (Xu et al. 2008) and wheat (Rebetzke et al. 2008; Peleg et al. 2009). Five QTL affecting δ13C were mapped in Arabidopsis using 162 recombinant inbred lines (RILs), and two QTL were co-located with QTL controlling flowering time, which suggested a potential pleiotropic relationship, and QTL interactions for the above two traits were also observed (Juenger et al. 2005). Teulat et al. (2002) has identified 10 QTL associated with grain Δ13C using 167 barley RILs grown in three Mediterranean environments in a field study. Among the 10 QTL, one was specific to one environment, two exhibited interaction with the environment, six showed main effects across two or three environments and one presented both main effect and QTL by environment interaction. Results also showed that eight QTL for Δ13C were co-located with QTL for several physiological traits related to plant water status and/or osmotic adjustment, and/or for agronomic traits previously measured on the same population, and heading date did not contribute to the effects of environment and interaction between genotype and environment on Δ13C (Teulat et al. 2002). Takai et al. (2009) found that a QTL controlling leaf Δ13C on the long arm of chromosome 3 in rice was associated with gs. Diab et al. (2008) reported that QTL for Δ13C and transpiration were on the same locus (gwm389). However, no single QTL for Δ13C with large effect have been identified in cereals, and most QTL identified for Δ13C have small effects. Table 2 summarizes the efforts to locate QTL for WUE measured as Δ13C in cereals.

Table 2.  Quantitative trait loci (QTL) for water-use efficiency (WUE) measured as carbon isotope discrimination (Δ13C) and/or carbon isotope ratio (δ13C)
SpeciesPopulationNumber of QTL detectedEnvironmentR2 (%)aReferences
  • a

    R2: proportion of the phenotypic variance explained by each QTL.

  • RIL, recombinant inbred line; BIL, backcross inbred line; DH, doubled haploid; NA, not available.

BarleyDHs1569Hydroponic tanks14 and 15Ellis et al. 2002
RILs16710Field (rain fed and irrigated)8.9 and 9.6Teulat et al. 2002
RILs1674Field (rain fed and irrigated)4∼17Diab et al. 2004
DHs571 and 9Glasshouse (hydroponic)14∼80Ellis et al. 1997
RiceBILs986Field (rain fed)8.5∼13.2Ishimaru et al. 2001
RILs2059Field (irrigated)NAPrice et al. 2002
RILs1012∼4Field (irrigated)7.6∼19Laza et al. 2006
RILs1265 and 7Field (rain fed)5.9∼14.3Takai et al. 2006
BILs987Glasshouse7.6∼22.2Xu et al. 2009
DHs9111Glasshouse8–19This et al. 2010
WheatDHs161∼1909∼13Field (rain fed and irrigation)2∼10Rebetzke et al. 2008
RILs15212Field (well-watered and water-limited)0.8∼30Peleg et al. 2009
RILs11029Field (rain fed)NADiab et al. 2008

The marker–QTL–trait association has been regarded as a promising way to develop improved cultivars, that is, MAS and marker-assisted breeding (Thomas 2000). Although the QTL analysis has provided unprecedented opportunities to identify chromosome regions harbouring genes/QTL controlling WUE in cereals, three major issues must be clarified towards efficient and effective implementation of MAS. Firstly, the linkage between random molecular markers and the target gene or QTL can be broken by recombination unless the markers are completely linked to the target allele or generated from gene sequence data (Araus et al. 2008). Secondly, the intrinsic nature of polygene underlying WUE, small size of individual QTL for Δ13C, and the interaction with the environment make MAS for WUE or Δ13C extremely difficult. The ultimate goal of QTL mapping is to transfer QTL for WUE into elite breeding lines to improve their performance when drought happens. Generally, that polygene controlling Δ13C are multiple genes each with small effects, implies that several QTL must be manipulated simultaneously to obtain a major impact (Cattivelli et al. 2008). However, MAS will not be effective when more than three QTL are considered (Araus et al. 2008). Theoretically, it is preferable to target QTL with a major effect that is consistent across environments and populations and also independent of the genetic background. So far, most QTL research on Δ13C were conducted in a single population, and common or repeatable QTL for Δ13C across environments and genetic pools only have been reported in wheat (Rebetzke et al. 2008). Third, most QTL for Δ13C have been reported to be co-located with QTL for yield components and heading date and/or plant height (Forster et al. 2004; Juenger et al. 2005). Favourable alleles for Δ13C and yield components could stem from the two contrasting parents (Lanceras et al. 2004), which may lead to yield penalty by selecting for reduced Δ13C. Furthermore, genotypic variation in Δ13C is usually associated with heading date and/or plant height; for example, QTL for shoot δ13C and grain yield in barley were associated with a semi-dwarf gene ari-e.GP on chromosome 5H near marker Bmac113 (Ellis et al. 2002), which further confound and compromise the relationship between Δ13C and grain yield (Rebetzke et al. 2008); in this case covariance analysis can separate Δ13C effects on yield from plant height and development influence (Rebetzke et al. 2008).

Phenotyping of WUE

Routine QTL analysis comprises four basic components: a segregating population, sufficient segregation markers, accurate phenotypic data for target trait(s), and a sound statistical approach. Accurate and precise phenotyping is a prerequisite for QTL mapping. For cost and labour considerations, the number of replicates and sites is often limited for phenotypic screening, hence reducing the sensitivity of the detection and analysis of QTL. If the evaluation of target trait is biased, the subsequent QTL mapping steps will be worthless. A good understanding of the ecological and physiological basis of trait under investigation, a proper and consistent measurement, and a careful experimental design are crucial to the detection of valid QTL. However, fast and accurate measurements of WUE remain a major bottleneck. Especially for the leaf-level WUE, the phenotyping process of gas exchange measurements using a portable photosynthesis system (e.g. Li-Cor 6400) requires relatively stable and consistent environmental conditions across the populations (such as during the period of maximum rates of net photosynthesis at similar humidity, temperature and daylight conditions), which limits its use in breeding programs for large segregating populations, especially under field conditions.

Although Δ13C has been demonstrated to be a simple and reliable measure of WUE, and it is easy to sample and store plant materials for carbon isotopic analysis, the screening of large breeding populations for Δ13C by IRMS remains costly (typically over US$10 per sample) (Araus et al. 2008; Lopes et al. 2011). As a genetically complex trait, the expression of Δ13C in different plant tissues and organs varies with water-supply (Rebetzke et al. 2008). To maximize genetic variance and heritability and improve QTL detection for Δ13C, screening of populations is suggested to be conducted under favourable and well-watered conditions (Rebetzke et al. 2008).

Validation and fine-mapping of detected QTL

One of the major shortcomings of QTL studies is that the number, location and estimated effects of identified QTL are often inconsistent in different genetic background of the mapping population (Bernardo 2008). The estimated effects of detected QTL are actually overestimated because of limited segregating progenies, a phenomenon called the ‘Beavis effect’ (Xu 2003). Beavis (1994) suggested that when population size is less than 150–200, only modest fractions of QTL are identified and the effect of each single QTL are usually overestimated. In order to provide a stable and reliable prediction of QTL positions and effects, reasonable population size, and replicated field trials, from multi-sites and across seasons are usually required. A QTL validation approach has also been suggested (Tuberosa & Salvi 2004). Association mapping has been recently proposed for QTL discovery and candidate gene validation in plants (Flint-Garcia et al. 2003; Salvi & Tuberosa 2005), which examines a collection of diverse accessions (e.g. varieties, landraces and breeding lines) without generating large mapping populations. Germplasm collections of diverse genetic backgrounds and with different selection histories likely differ in their QTL alleles (Bernardo 2008), and the same QTL would be expected to be present in different populations, assuming that the particular QTL is stable or consistent. Maccaferri et al. (2011) conducted association mapping to indentify QTL controlling the agronomic performance in durum wheat across a broad range of water availability regimes. In their study, the presence of major QTL at key chromosome regions (such as Ppd-A1 alleles, photoperiod-responsive gene) previously identified with biparental mapping were validated, and highly heritable traits (such as heading date and kernel weight) were found to be less affected by environmental conditions as compared with low heritability traits such as yield. Therefore, association mapping may be suitable for validating Δ13C as a high heritability trait. Moreover, QTL can be confirmed at a low density of markers (coarse mapping) by choosing a population with a high level of linkage disequilibrium (Abdurakhmonov & Abdukarimov 2008; Yan, Warburton & Crouch 2011), which describes the non-random association of alleles or alleles and markers at different loci (Yu & Buckler 2006).

As a valuable complementary tool in detecting marker–trait associations, association mapping has been extensively utilized in cereals (Garris, McCouch & Kresovich 2003; Kraakman et al. 2004; Skøt et al. 2005; Breseghello & Sorrells 2006; Ravel et al. 2006; Yu & Buckler 2006; Cockram et al. 2008; Horvath et al. 2009; Stracke et al. 2009; Waugh et al. 2009; Neumann et al. 2011; Yan et al. 2011). By utilizing all the historic recombination events from germplasm development, association mapping can provide a high resolution genetic map (fine mapping) and provide more precise locations of individual QTL (Oraguzie et al. 2007; Maccaferri et al. 2011; Neumann et al. 2011), or a step towards positional cloning (Rafalski 2010), which is more challenging but more rewarding for quantitative traits (Sorkheh et al. 2008). Major factors that affect association mapping include the level of linkage disequilibrium, population structure and stratification, familial relatedness and complexity of target traits (Abdurakhmonov & Abdukarimov 2008; Rafalski 2010).

Sufficient genotypic variation, stability across environments and a high H2 in Δ13C (Condon & Richards 1992; Schuster et al. 1992; Johnson & Rumbaugh 1995; Araus et al. 1998) indicate that Δ13C is a promising phenotype which can be tracked by molecular markers through QTL mapping, and eventually may improve the selection efficiency of WUE through marker-assisted breeding. Knowledge of loci underlying natural variation in WUE and Δ13C would be valuable and beneficial for breeding programs. QTL mapping is an initial step towards unraveling the molecular basis of WUE. The next step towards the application of molecular markers in breeding for high WUE is to fine-map the candidate gene regions to reduce QTL size and clone DNA sequences underlying QTL. Many efforts have been dedicated to understanding the genetic basis of Δ13C. According to Sanchez et al. (2002), the average marker interval of 0.5 cM or 350 kb is appropriate for fine mapping genes and QTL. Xu et al. (2008) fine-mapped a dominant QTL for δ13C (designated QWUE5.1) in tomato to an interval about 2.2 cM long, and located markers SSR 49 and SSR 590 less than 2.2 cM from QWUE5.1, which were valuable for cloning the genes underlying QWUE5.1 and can be effectively used in MAS of QWUE5.1 in tomato breeding programs. To date, map-based cloning is still more challenging to large genome species such as wheat and barley as compared with small genome species such as Arabidopsis, rice or tomato (Diab et al. 2008). However, the high heritability of flowering time has allowed a major QTL Vgt1 isolated via map-based cloning in maize, which has a moderate genome size (Salvi et al. 2007). There are high hopes for cloning major QTL for Δ13C in rice with its available resource such as the genome sequence.

Gene discovery in WUE studies

In the past 10 years tremendous progress has been made in identifying the genetic determinants of physiological responses to abiotic and biotic stress. Candidate gene can be defined as all genes involved in the expression of a given trait by physiologists, or only polymorphic genes putatively engaged in the trait variation by geneticists (Pflieger, Lefebvre & Causse 2001). Candidate genes can be approached either through functional or positional way (Pflieger et al. 2001; Varshney, Graner & Sorrells 2005; Christiane et al. 2007; Zhu & Zhao 2007). Functional candidate genes are shown or suspected to play a functional role in the phenotype under investigation, while positional candidate genes are closely linked to QTL or co-localized with a QTL. Positional candidate genes can be reached through QTL mapping, map-based cloning approaches, and information from closely related species (Li et al. 2010). For cereals, the full genome sequence of rice has been available to researchers (International Rice Genome Sequencing Project 2005), which provides a valuable shortcut for identifying candidate genes in related species currently lacking high density maps, such as wheat and barley. Functional genomic approaches such as transcriptomics and expression genetics holds great promise for identifying functional candidate genes. Plant responses to stress to a large extent are under transcriptional control (Cattivelli et al. 2008). Theoretically, a difference at the mRNA level could potentially contribute to genotypic variation in the target trait. The microarray technique, also called gene expression profiling or transcriptional profiling, has greatly contributed to gene function analysis in plant species since 1995 (Schena & Shalon 1995), which can monitor in parallel the expression of thousands of transcripts simultaneously. With careful experimental design and appropriate data analysis, the expression profiling of large-scale genes can detect certain up- or down-regulated genes, which would help to identify candidate genes, reveal important clues about gene function, provide detailed insights into physiological processes underlying targets of interest and eventually clone and manipulate candidate genes through genetic engineering (Schulze & Downward 2001; Stéphanie et al. 2001; Varshney et al. 2005; Xue et al. 2006; Zhu & Zhao 2007).

So far, many drought-related genes have been isolated and characterized (Jain, Basha & Holbrook 2001; Cattivelli et al. 2002; Ramanjulu & Bartels 2002; Hazen et al. 2005; Masle, Gilmore & Farquhar 2005). Differential gene expression of wheat progeny with contrasting levels of TE conducted in a controlled growth room revealed that 11 genes were positively correlated with high TE trait measured as Δ13C, which was confirmed by quantitative real-time polymerase chain reaction (RT-PCR) analysis (Xue et al. 2006). Results from above experiments suggested that those differentially expressed genes can serve as candidates for further investigation or be used as expression QTL (eQTL) for mapping TE traits. However, the major shortcoming associated with expression profiling is that it does not necessarily provide any information related to post-transcriptional and post-translational modifications (Tuberosa et al. 2002). Diab et al. (2008) reported that QTL for Δ13C co-segregated with two differentially expressed sequence tags (dESTs) through combining positional and functional candidate gene approaches, but the role of these dESTs have not yet been elucidated.

Most research have indicated that plants tend to increase drought resistance and WUE through decreasing transpiration via stomatal closure, reducing stomatal pore size or number (density) rather than increasing CO2 assimilation. ERECTA, the first TE gene that has been isolated in the extensively-studied model plant Arabidopsis thaliana, influences the coordination between transpiration and photosynthesis through several ways (e.g. stomatal density and epidermal cell expansion), and similar genes have been found in rice, sorghum and wheat (Masle et al. 2005). The expression of Arabidopsis HARDY (HRD) gene in rice improves WUE by enhancing photosynthetic assimilation and reducing transpiration (Karaba et al. 2007). The GT-2 LIKE 1 (GTL1) in Arabidopsis thaliana functions as a transcription factor of drought tolerance and WUE by regulating stomatal density and transpiration, and gtl1 plants had high WUE and low transpiration without reduction in CO2 assimilation (Yoo et al. 2010). A transcription factor DREB1A (the Dehydration Responsive Element Binding protein) driven by a rd29A promoter from Arabidopsis thaliana was transferred into a peanut cultivar, which achieved higher TE at a lower gs than the untransformed control under well-watered conditions (Bhatnagar-Mathur et al. 2007). The GPA1 in Arabidopsis thaliana has also been found as a regulator of TE through control of stomatal density (epidermal cell size) and stomatal development (Nilson & Assmann 2010).

However, there are few reports on improving WUE through increasing photosynthesis. Jeanneau et al. (2002) examined the effect of photosynthesis regulation via the ectopic expression of the Sorghum bicolor C4-PEPC gene in transgenic maize. The C4-PEPC over-expressed lines increased WUEic and dry weight by 30 and 20%, respectively, under moderate drought conditions. The observed improvement in WUE and biomass production in over-expressing transgenic lines was achieved by decreased gs (reduced stomatal aperture and density) and improved ability of CO2 fixation (a higher C4-PEPC activity). Eventually, these genes will lead to improved TE in both dry and well-watered conditions through the so called strategy of ‘breeding by design’ (Masle et al. 2005; Peleman & van der Voort 2003).

After minimizing some factors in the population structure such as flowering time and plant height, which may confound the phenotypic expression of target traits, association mapping can increase the probability of novel gene discovery (Reynolds et al. 2009). Mohan (2010) conducted association mapping for drought tolerant traits using a set of 151 rice germplasm accessions. Twenty-six out of 113 simple sequence repeat (SSR) markers were found to be associated with root traits and Δ13C, among those, marker ‘RM 224’ on chromosome 11 co-segregated with carbon isotope ratios and root volume, explaining 73.7 and 73.1% of the phenotypic variance, respectively, indicating that there was a potential major gene around this marker.

Several kinds of tools and approaches are available for the introgression of genes and genomic regions into elite crop varieties, which are susceptible to water scarcity (Bajaj et al. 1999; Varshney et al. 2005; Karaba et al. 2007). Various strategies have been used to produce transgenic plants, including overproduction of enzymes responsible for osmolytes biosynthesis, LEA proteins and detoxification enzymes (Bajaj et al. 1999). For example, a Group 3 LEA proteins encoded by the barely HVA1 gene improved the biomass and WUE of transgenic wheat under water deficit conditions (Sivamani et al. 2000). However, caution needs to be taken so that the over-expression of transcription factors or constitutive expression of selected genes may not activate adverse or deleterious effects on target traits thus having a negative effect on yield (Liu et al. 1998; Bajaj et al. 1999; Wang, Vinocur & Altman 2003; Ito et al. 2006; Araus et al. 2008). Although a large number of molecular markers or candidate genes for WUE or Δ13C have been reported, few are effective in MAS and the engineering of improved WUE with single genes in major crops has not been achieved so far (Karaba et al. 2007).


It is not easy to achieve large genetic gains in WUE and yield concurrently in dry-land environments, even though genetic improvement holds opportunities in increasing CY especially in drought-prone regions. Research from plant physiology, genomics, to molecular biology on drought in the past several decades has produced good techniques and accumulated experiences for future studies. However, routine cloning of genes underlying the QTL for WUE is a long way to go.

In addition to relying on markers from a highly dense map, candidate genes from the outcome of microarray can also be used as markers. Polymorphic markers from functional genes identified in microarrays experiments or physically close to these could be used to identify candidate genes for positional cloning and enrich linkage maps. Association mapping is an alternative way to identify new useful alleles through exploration in wild germplasm (Cattivelli et al. 2008), which is becoming popular with the development of high-throughput genotyping and genome sequencing technologies as well as advanced statistical approaches. The ultimate goal is to tag and isolate genes or pyramid beneficial QTL alleles controlling WUE. With the application of MAS, favourable alleles can be introduced into elite germplasm to derive improved cultivars and speed up plant breeding process in water-limited environments.

However, water deficit combined with other unfavourable environments (such as diseases, insects, weeds, infertile soils, etc.) make genetic selection more complicated. Although breeding for high WUE varieties is a promising strategy via the use of Δ13C, particularly in C3 cereals, the greatest gains still come from the integrated knowledge of physiology and genomics, and the combined strategies of traditional breeding, MAS and appropriate management practices (e.g. reducing water loss from soil surface evaporation, deep drainage and surface runoff) to achieve the yield potential under targeted environments.


The authors are grateful to two anonymous reviewers for their valuable comments and suggestions that improved an earlier version of this review. This work was funded by Alberta Agriculture Research Institute (AARI), Alberta Crop Industry Development Fund (ACIDF), Alberta Barley Commission (ABC), the University of Alberta and the Natural Sciences and Engineering Research Council of Canada (NSERC).