Integrated genomics, physiology and breeding approaches for improving drought tolerance in crops

The use of physiological traits (PTs) in a breeding programme, either by direct selection or through a surrogate such as molecular markers, depends on their relative genetic correlation with yield, extent of genetic variation, heritability and genotype × environment interactions. For instance, in drought environments, osmotic adjustment, accumulation and remobilization of stem reserves, superior photosynthesis, heat- and desiccation-tolerant enzymes, etc. are important PTs. However, it is important to establish their heritability and genetic correlation with yield in target environments. Identification of drought-adaptive PTs and mechanisms is time consuming and costly; however, if successful, the benefits are likely to be substantial. The information on important PTs can be collected on potential parental lines involving screening of entire crossing block, or a set of commonly used parents, thus producing a catalogue of useful PTs. This information can be used strategically in designing crosses, thereby increasing the likelihood of transgressive segregation events, which bring together desirable traits. However, if enough resources are available, screening for PTs could be applied to segregating generations in yield trials, or any intermediate stage, depending on when genetic gains from selection are optimal (Reynolds 2002).

It is important to note that using specific traits, breeding strategies are effective only when these traits are properly defined in terms of the stage of crop development at which they are relevant, the specific attributes of the target environment for which they are adaptive and their potential contribution to yield (Reynolds and Trethowan 2007). The early escape from progressively intensifying moisture stress, through the manipulation of plant phenology, is the most commonly exploited genetic strategy used to ensure relatively stable yields under terminal drought conditions (Richards 1991). When significant genetic diversity for a physiological trait in a germplasm collection for the given species is established, it is imperative that the relevance of the trait as a selection criterion be determined.

The conceptual framework for yield drought adaptation by Passioura (1977) has three important drivers: (1) water uptake (WU), (2) water-use efficiency (WUE) and (3) harvest index (HI). These drivers stimulate trait-based breeding and genetic dissection of drought-adaptive mechanisms. Several traits have been found to be associated with the above yield component drivers. For WU, direct selection for variation in root characteristics is unfeasible; therefore, measurements associated with stomatal conductance like that of canopy temperature (CT) provide indirect indicators of water uptake by roots (see Reynolds and Tuberosa 2008). In addition, validation studies indicated that CT during peak stress periods was associated with ~50 % of the variation in water extraction in deep soil profiles and also with root length density (Reynolds et al. 2007a). For WUE, carbon isotope discrimination seems to be the best estimate and is based on higher affinity of the carbon-fixing enzyme (Rubisco) for the more common ¹²C isotope over the less common ¹³C. A lower discrimination value indicated higher WUE. Some other traits associated with WUE included spike photosynthesis in cereals, photoprotective mechanisms including antioxidant systems, regulation of water flow via aquaporins and signalling molecules such as abscisic acid (ABA) (see Reynolds 2002; Reynolds and Tuberosa 2008). Similarly, for HI, the extreme sensitivity of reproductive processes to drought may result in the reproductive failure, which is associated with low HI, and may eliminate benefits associated with favourable WU or WUE. Considering the overall contributions of these three yield drivers, WU is the most important for improving the yield potential (i.e. biomass) in drought environments, while stable HI is associated with higher yield potential (Blum 2009; Salekdeh et al. 2009). Storage of water-soluble carbohydrates (WSC) in the stem of small grain cereals and their subsequent remobilization to grain can directly influence HI, especially under post-anthesis stress. Translocation of soluble stem carbohydrates to the grain is one of the drought-adaptive traits that relates specifically to improved partitioning, though not to reproductive growth. Remobilization of stem reserves is associated with increased levels of ABA, which presumably is involved in the triggering of enzymes prerequisite to remobilization (Reynolds 2002). The yield potential (YP), expressed as a function of the light intercepted (LI) and radiation-use efficiency (RUE) (whose product is biomass), the partitioning of biomass to yield (the HI) and the focus of improving all the three components will be undertaken through complex physiological trait (PT)-based breeding.

A general model for drought adaptation of wheat was developed by the physiologists and breeders at CIMMYT that encompasses traits which possess a potential role in dry environments (Reynolds et al. 2005). In this model, some of the important traits included: (1) pre-anthesis growth, (2) access to water as a result of rooting depth or intensity that would be expressed by a relatively cool canopy (Reynolds et al. 2005), (3) water-use efficiency (WUE) as indicated by relatively higher biomass/mm of water extracted from the soil, transpiration efficiency of growth (TE = biomass/mm water transpired) indicated by C-isotope discrimination (Δ¹³C) of leaves, and WUE of spike photosynthesis associated with refixation of respiratory CO₂, (4) photoprotection including energy dissipation, anti-oxidant systems and anatomical traits such as leaf wax. The model is used to assist in taking breeding decisions by permitting a strategic approach of accumulating drought-adaptive alleles by crossing parents with contrasting drought-adaptive mechanisms. Accumulation of soluble stem carbohydrates and their remobilization during the post-anthesis drought period help to supply surplus assimilates for grain growth during grain filling (Blum 1998). Similarly, root architecture that helps to have better access to soil moisture under drought enables heat-stressed crop canopies to meet high evaporative demand associated with hot, low-relative humidity environments, thus resulting in cooler canopies (Reynolds et al. 2000). Other traits impact either WUE or RUE depending on the environmental conditions (Reynolds and Trethowan 2007).

It is therefore crucial to target specific physiological mechanisms and to identify those traits most relevant to the patterns of drought stress found in the target environment. For example, in crops grown with residual soil moisture that experience terminal drought, such as chickpea (Cicer arietinum), genotypes with deeper, more profuse roots have an advantage through better water extraction deeper in the soil profile (Kashiwagi et al. 2005). In other crops also, deeper/profuse roots were found to increase plant access to water from deeper soil layers and support greater crop growth under drought conditions (Price et al. 2002a; Sinclair 2011). Therefore in several crops such as chickpea (Silim and Saxena 1993), wheat (Reynolds et al. 2007a) and rice (Yadav et al. 1997; Price et al. 2002a), deeper/profuse roots are targeted to improve grain yield under rainfed conditions. However, some recent studies (Zaman-Allah et al. 2011a, b) reported that selection for yield under terminal drought conditions was not essentially dependent on deeper/profuse root systems, but rather on several other critical traits that contribute to soil moisture conservation during late season water deficits. These traits include: (1) low leaf conductance under non-limited water conditions during the vegetative stage, which could be measured by a warmer canopy, (2) a low leaf expansion rate when soil moisture is still non-limiting for plant growth and a restriction of plant growth under progressive exposure to stress and (3) a higher fraction of transpirable soil water (FTSW) thresholds that reduce transpiration, thus avoiding rapid soil water depletion (Fig. 2). Several studies have shown that FTSW can be linked to variables describing plant water status such as midday leaf water potential, leaf relative water content and stomatal conductance (Sinclair and Ludlow 1986; Pellegrino et al. 2007), which are known to contribute to drought adaptation.

Open image in new window

In addition to the above positive effects of the stay-green trait, enhanced remobilization of stored carbohydrates will lead to identify the important targets for enhancing seed sink strength under drought, thus helping to achieve yield stability under drought (for further details refer reviews by Sreenivasulu et al. 2007; Mittler and Blumwald 2010). Although, in general, photosynthesis is markedly reduced under drought stress, many dicot species are dependent on assimilates produced from current photosynthesis under drought and, therefore, exploring genotypes possessing efficient mechanisms of stay green will be beneficial. On the contrary, monocarpic cereal species seem to prefer assimilates produced prior to flowering (pre-anthesis assimilate), which is stored in the vegetative tissue, mainly in the stem in the form of various soluble sugars translocated through the trigger of remobilization events where ABA plays an important role (Yang and Zhang 2006; Zhang et al. 2009; Seiler et al. 2011). These traits were not all present in a single genotype, reflecting the complexity of drought tolerance and the need to pyramid several beneficial traits through plant breeding.

The conceptual models of drought-adaptive traits have been found useful for accumulating complementary PTs in selected progeny. The key steps in this type of PT breeding include: (1) characterization of crossing block lines for stress-adaptive mechanisms, (2) strategic crossing among parents with different but potentially complementary PT expression, thus ensuring cumulative gene action in selected progeny, and (3) early generation selection (EGS) of bulks for canopy temperature (CT). This type of physiological characterization is used to assess variation and thereby increases the rates of genetic gains (Reynolds and Tuberosa 2008; Reynolds et al. 2009a). The main objective of strategic trait-based crossing is to accumulate traits that will be complementary for a given target environment. Under water-limited situations, traits that improve water uptake, water use efficiency and partitioning to yield, respectively, are likely to work synergistically to maximize productivity in the target environment (Passioura 1977; see Reynolds et al. 2009a). This has resulted in the distribution of advanced lines to rain-fed environments worldwide by the International Maize and Wheat Improvement Center (CIMMYT) and it has been confirmed that PT crossing results in cumulative gene action in selected progeny, resulting in increased yield under drought environments (Reynolds et al. 2005, 2007b; Reynolds and Tuberosa 2008). Characterization of candidate parents for better targeted crossing should have the highest priority in terms of physiological interventions in breeding for a number of reasons including: (1) since a large investment is needed in trait measurement and the information obtained can be used for many generations of crossing once the initial characterization has been made, (2) the number of lines in a crossing block are relatively small (~100/target environment) and the detailed characterization is even possible for traits which are relatively time-consuming, e.g. for traits like soil moisture depletion or stem carbohydrates.

The genetic gains in yield can be accelerated by incorporating complex PTs deterministically in modern plant breeding in addition to simply agronomic inherited traits like plant height, flowering time, resistance to prevalent diseases, quality parameters and yield based on multilocation trials (Braun et al. 2010; Reynolds et al. 2011). PT-based breeding approaches have been already implemented successfully in Australian breeding programmes (Rebetzke et al. 2009) as well as by CIMMYT, leading to international distribution of a new generation of elite drought-adapted lines (Reynolds et al. 2009b).

This method provides spectral information corresponding to the field plot in a single near-infrared spectrum, where physical and chemical characteristics of the harvested seed material are captured. By using calibration models (i.e. mathematical and computational operations that relate the spectral information with phenotypic values), several traits can be determined on the basis of a single spectrum (dry matter, protein, nitrogen, starch and oil content, grain texture and grain weight, etc.; Montes et al. 2007; Wiley et al. 2009; Hacisalihoglu et al. 2010). The use of NIR spectroscopy on agricultural harvesters provides indexing of grain characteristics. In contrast to conventional sample-based methods, NIR spectroscopy on agricultural harvesters secures a good distribution of measurements within plots and covers substantially larger amounts of plot material (Welle et al. 2003), thus reducing sampling error and providing more representative measurements of the plot material in terms of homogeneity.

Spectral reflectance of plant canopy is a non-invasive phenotyping technique that enables several dynamic complex traits, such as biomass accumulation, to be monitored with high temporal resolution (Montes et al. 2007). It has many advantages including easy and quick measurements, integration at the canopy level and additional parameters can also be measured simultaneously via a series of diverse spectral indices like photosynthetic capacity, leaf area index, intercepted radiation and chlorophyll content. Therefore, canopy reflectance is considered as one of the valuable tools for high-throughput phenotyping (Montes et al. 2007; Chapman 2008; Gutierrez et al. 2010). In soybean, canopy reflectance indices have been already used with great promise for measuring the effects of increasing atmospheric CO₂ and O₃ on soybean canopies (Gray et al. 2010).

Investigations at the individual plant level under well-controlled environmental conditions showed that spectral reflectance could be used to: (1) estimate the effects of environmental perturbations, such as changing atmospheric composition, on canopy structure and function (Gray et al. 2010), (2) monitor plant photosynthetic pigment composition and (3) assess plant water status and detect abiotic or biotic plant stresses (Chaerle and van der Straeten 2000; Gutierrez et al. 2010).

Plant water status as determined by plant water content or water potential (Jones 2007; Jones et al. 2009) integrates the effects of several drought-adaptive traits. Several methods are used to determine crop water content, including leaf water potential, leaf stomatal conductance and canopy temperature (CT), which is the relative measure of water flow associated with water absorption from the soil under water deficit (Reynolds et al. 2007a, b).

In addition to the above, one of the most commonly used indirect techniques for measurement of these variables is thermal infrared imaging, or infrared thermography, which involves the measurement of leaf or canopy temperature. Plant canopy temperature is a widely measured variable that is closely related to canopy conductance at the vegetative stage (Zaman-Allah et al. 2011a) and therefore provides insight into plant water status. In any given environment, stomatal variation is the dominant cause of changes in canopy temperature (Jones 2004). Although thermal imaging does not directly measure stomatal conductance, it has become a high-throughput tool for estimating differences in stomatal conductance (Merlot et al. 2002). Thermal infrared imaging for estimating conductance can be used at the whole plant or canopy level over time.

These two methods are used at the Jülich Plant Phenotyping Centre, Germany (Heike Schneider, personal communication) to investigate root/shoot systems growing in sand or soil which allow to assess structure, transport routes and the translocation dynamics of recently fixed photoassimilates labelled with short-lived radioactive carbon isotope (δ¹¹C). Quantitative MRI and PET data not only help to study differences between species, but also provide a phenotype within a species, the growth pattern, water relations and/or translocation properties of assimilates (Jahnke et al. 2009). Therefore, the MRI–PET combination can provide new insights into structure–function relationships of intact plants. It also allows monitoring of dynamic changes in plant properties, which have previously not been possible to assess systematically, thus improving our understanding of plant performance (such as resource use efficiency or biomass production).

The short half-life of δ¹¹C (only 20 min) limits the utility of MRI to study source–sink relationships in minute structures such as developing seeds (Jahnke et al. 2009). NMR provides an alternative in vivo detection platform using ¹H NMR and utilizes the signal emitted by protons associated with carbon nuclei, thereby sucrose and water movement may be imaged and quantified (Sardans et al. 2010; Melkus et al. 2011). Therefore, NMR technology is employed with ¹³C/¹H double-resonant high-resolution coil to achieve better resolution for monitoring the structure of tissues like seeds through non-invasive visualization, mapping water movements and monitoring of sucrose allocation using ¹³C-labelled sucrose (Neuberger et al. 2008; Melkus et al. 2011).

One of the high-throughput integrated phenotyping platforms that includes the pipeline of imaging, image processing automatization and data handling modules was developed by LemnaTec, a German company (http://www.lemnatec.com). The platform has the capacity to measure almost unlimited sets of parameters easily, allows comprehensive screening and provides statistics on various plant traits in a dynamic way. Depending on the degree of automatization, plants are manually placed in the Scanalyzer 3D or transported on conveyor belts directly from the greenhouses to the imaging chambers. Such chambers provide top and side imaging of both shoot and root systems to quantify plant height/width, biomass and plant architecture. Application of different camera and acquisition modes—from visual light to near infrared (NIR/SWIR), infrared (IR) and fluorescence imaging—opens new perspectives for visualization using non-destructive quantification. The key application is in the fast developing domain of plant functional genomics. These automated systems will increase our understanding of plant growth kinetics and help improve plant models for systems biology or breeding programmes.

In summary, the techniques and platforms mentioned above will greatly improve the phenotyping accuracy and throughput, thus contributing to a better elucidation of the genetic control of complex drought tolerance traits in plants. However, many of the techniques discussed above are applied to plants grown under controlled conditions that may not reflect field environment or can only be used to assess a limited number of genotypes due to high costs and/or practicality. Therefore, to overcome this problem, multi-tiered selection screens, where a simple but less accurate screen allows large number of genotypes to be evaluated (first screen), followed by tiers of more sophisticated screens of decreasing numbers of genotypes have been proposed (Sinclair 2011, Fig. 3). A three-tiered sequence of physiological screens have been already used to identify candidate parental genotypes for use as parents in breeding programs for some key traits like nitrogen fixation activity during soil water deficit in soybean (Sinclair et al. 2000). Furthermore, bringing integrative phenotyping technology, such as that developed by LemnaTec, from the controlled environments to the field will improve the assessment of plant responses to drought while enabling high-throughput screening and generating comprehensive and accurate phenotypic data.

Open image in new window

Traditional QTL mapping involves: (1) development of mapping populations segregating for drought tolerance-related traits, (2) identification of polymorphic markers, (3) genotyping of the mapping populations with polymorphic markers, (4) construction of genetic maps, (5) precise phenotyping for drought tolerance-related traits, as mentioned above, and (6) QTL mapping using both genotypic and phenotypic data. This process is commonly called linkage mapping/linkage analysis-based QTL mapping (see Chamarthi et al. 2011). During the past decade, a large number of studies involving linkage mapping have been conducted in several crops to identify QTLs linked to drought tolerance (for reviews see, Cattivelli et al. 2008; Fleury et al. 2010). However, linkage mapping-based QTL mapping does not provide precise information on QTLs because of inherent limitations associated with each mapping population. Some of these limitations are summarized by Myles et al. (2009) and include: (1) insufficient time for recombination to occur and shuffle the genome into small fragments, and as a result the QTLs identified are generally localized to large genomic regions/chromosomal segments, (2) insufficient phenotypic variation for the trait present in the mapping population and (3) segregation of different QTLs for the same trait in different mapping populations.

To overcome some of above constraints, linkage disequilibrium (LD)-based association mapping, initially used in human genetics, has been suggested as an alternative approach for QTL mapping in crop species (e.g. Myles et al. 2009; Rafalski 2010). The association mapping (AM) approach involves: (1) selection of a diverse association panel/group of individuals from a natural population/germplasm collection, (2) precise recording of phenotypic data on the panel, (3) candidate gene sequencing or high-density marker genotyping of the panel, (4) study of population structure (the level of genetic differentiation among groups within the selected population) and kinship (coefficient of relatedness between pairs of each individual within the population) and (5) association analysis based on information gained through population structure, kinship, and correlation of phenotypic and genotypic/haplotypic data. AM offers several advantages over bi-parental linkage mapping and these include: (1) exploitation of all the recombination events that took place during the evolutionary history of a crop species, resulting in much higher mapping resolution, (2) less time required in mapping QTL as there is no need to develop a specialized mapping population, rather a natural germplasm collection of a crop species is sufficient, (3) cost-effectiveness because the same AM panel and genotyping data can be used for mapping of different traits, (4) populations can be structured to avoid randomly generated lines (recombinant inbred lines; RILs), many of which express substandard agronomic type and (5) a higher number of alleles can be sampled compared to linkage mapping where only two alleles are usually surveyed (in apple or potato where heterozygous parental lines are used, more than two alleles/locus can be present). Markers linked to drought tolerance traits, identified using AM, have been reported in wheat (Sanguineti et al. 2007; Maccaferri et al. 2011), barley (Ivandic et al. 2003; Baum et al. 2007; Varshney et al. 2012) and maize (Lua et al. 2010). However, obtaining a clean set of reproducible phenotypic data of drought tolerance from a larger germplasm collection for AM studies remains an open challenge even in the era of phenomics-driven technology.

In general, QTLs identified through linkage mapping-based approaches have low resolution and have been located in 10–20 cM intervals. The support interval of the QTL may also span several hundreds of genes and identifying the right candidate gene(s) with causal effect on the trait is like finding a ‘needle’ in the ‘genomic haystack’. Therefore, to identify the causal gene(s), positional cloning of QTLs have been undertaken in several crop species (Salvi and Tuberosa 2005; Tuberosa and Salvi 2006). QTL cloning, in general, involves the following steps: (1) delimiting the QTL region by using a large mapping population (>1,500 plants) derived from a cross between two NILs for the target QTL, (2) identifying the contig covering the QTL region by screening the closely linked molecular markers with a large insert library like BAC (bacterial artificial chromosome) library, (3) sequencing the contig and candidate gene identification based on sequence data and (4) validating the effect of candidate gene(s) on phenotype.

Although many reports are available on cloning of QTLs associated with different traits (see Salvi and Tuberosa 2007), there are few reports addressing QTL cloning for drought tolerance traits. For instance, a major flowering time QTL “Vgt1” associated with drought tolerance has been cloned in maize (Salvi et al. 2002, 2007). Recently, the gene encoding ATP-binding cassette (ABC) subfamily G (HvABCG31) full transporter was cloned from eibi1 mutation responsible for leaf water conservation in wild barley and rice (Chen et al. 2011). Newer genomics approaches like association mapping and next-generation sequencing (NGS) hold great promise for accelerating QTL cloning of drought tolerance-related traits. The cloning of drought tolerance QTLs provides an opportunity to validate candidate genes that can be used to develop transgenic plants, not only in the original crop species but also in other crop species.

The significant advances made in the model plant systems of major crop species provide an opportunity to identify candidate genes associated with drought tolerance. Some approaches are discussed in the following section:

The relatively low success of MABC for improving drought tolerance can be attributed to the complex nature of drought. In many instances, the expression of drought tolerance is controlled by minor main-effect QTLs or epistatic QTLs. For instance, QTLs with ~10 % phenotypic variation for drought tolerance have been identified in maize (Xu et al. 2009), groundnut (Ravi et al. 2011), etc. These studies highlight the need to transfer several QTLs/genes to achieve a significant impact, assuming additive variance, and this may require unmanageable population sizes (Ribaut et al. 2010).

To overcome the limitations of MABC, particularly when multiple QTLs control the expression of a complex trait, the MARS approach, which involves intermating selected individuals in each selection cycle, has been recommended (Eathington et al. 2007; Ribaut and Ragot 2007; Bernardo 2008). It generally involves the use of an F₂ base population, and can be used in self-pollinated crops like wheat, barley and chickpea for developing pure lines with superior per se performance (for more details, see Bernardo 2008). MARS has the additional advantage of overcoming the limitation of inadequate improvement in the frequency of superior alleles in F₂ enrichment, since MAS is practised in each cycle following intermating to improve the frequency of favourable alleles (Eathington et al. 2007). The successful use of MARS has been reported in sweet corn (Edwards and Jonson 1994), sunflower and soybean (Eathington et al. 2007). In case of wheat, MARS for water use efficiency is being exercised under an Indo-Australian project involving partners from DWR, Karnal, PAU Ludhiana, IARI, New Delhi and Australia. Generation Challenge Programme (GCP) also launched a challenge initiative to improve heat/drought tolerance in wheat through MARS approach involving the Indian Agricultural Research Institute (IARI), New Delhi, India, Chinese Academy of Agricultural Sciences (CAAS), China, and partners from Australia (http://www.generationcp.org/ci_feb_2010_launch_meeting_feature). Similar MARS breeding programmes are being conducted at several other international institutes including ICRISAT, the French Centre for International Agricultural Research (CIRAD) and University of California-Riverside, USA for improving drought tolerance in chickpea, sorghum and cowpea, respectively (see Kulwal et al. 2011).

Genome-wide selection (GWS) or genomic selection (GS) is another important approach to develop superior germplasm lines with overall excellent performance in a target environment.

Genome-wide marker genotyping is used for GWS rather than selected markers showing significant associations (as in case of MARS) with the traits of interest. In summary, individuals in a phenotyped population (generally referred to as the ‘training population’) are genotyped using genome-wide markers and breeding values of alternative alleles of all the markers are fitted as random effects in a linear model. Individuals in subsequent recurrent selection generations are then selected based purely on the sum of those breeding values [genomic estimated breeding value (GEBV); Meuwissen et al. 2001]. Therefore, GWS reduces the frequency of phenotyping and similarly also increases annual gains from selection by reducing cycle time (Rutkoski et al. 2010). Several groups have recently started exploring the GWS approach in both self- and cross-pollinated crops with some modifications for both types of crops (Bernardo 2010). The success of the GWS approach is dependent on the availability of a diverse and representative training population. Furthermore, the phenotyping of the training population is crucial and additional lines should be integrated over time to increase the effectiveness and relevance of the gene effect estimates. This approach has been recently used to improve durable stem rust resistance in wheat (Rutkoski et al. 2010) and eventually could be systematically explored to bring different components of multigenic drought tolerance using the GWS approach.