Physiological and Molecular Approaches to Improve Drought Resistance in Soybean

drought resistance in soybean.


Introduction
Soybean is the world's leading economic oilseed crop. Processed soybeans are also the largest source of vegetable oil and protein feed. In addition to being a source of macronutrients and minerals, soybeans contain secondary metabolites such as isofl avones ( Sakai and Kogiso 2008 ), saponins, phytic acid, oligosaccharides, goitrogens ( Liener 1994 ) and phytoestrogens ( Ososki and Kennelly 2003 ). Global production of soybean in 2007 was around 219.8 million metric tonnes (mmt). The country with the greatest output was the USA, producing 70.4 mmt, followed by Brazil (61 mmt), Argentina (47 mmt) and China (14.3 mmt) ( Soystats 2008 ). While soybean has long been important in Japan for the production of traditional foods such as tofu, miso, shoyu and vegetable oil, the consumption of soybean-based products is increasing worldwide because of the reported benefi cial effects including lowering of cholesterol, prevention of cancer, diabetes and obesity, and protection against bowel and kidney diseases ( Friedman and Brandon 2001 ). Soybean is also viewed as an attractive crop for the production of biodiesel ( Pimentel and Patzek 2008 ). It also has the ability to fi x atmospheric nitrogen ( Burris and Roberts 1993 ) and therefore requires minimal input of nitrogen fertilizer which often accounts for the single largest energy input in agriculture.
Plants are subjected to a range of abiotic and biotic stresses that affect their growth and development. In particular, it is predicted that water defi cit will continue to be a major abiotic factor affecting global crop yields ( Sharma and Lavanya 2002 ). One-third of the world's population resides in water-stressed regions, and with elevated CO 2 levels in the atmosphere and climatic changes predicted in the future, drought could become more frequent and severe. The resilience of legume crops against present-day weather extremes, such as drought, excess water, heat, cool weather during grain fi lling, and early frost, is considered to predict their adaptation to future climate change (Cutforth et al. 2007) . In soybean, drought reduces yield by about 40 % ( Specht et al. 1999 ). Depending on hybrid characteristics, soybeans use about 450-700 mm of water during the growing season ( Dogan et al. 2007 ). The most critical period for water stress in soybean has been reported to be during the fl owering stage and the period following fl owering ( Meckel et al. 1984 ).
Plants use various mechanisms to cope with drought stress. These may be classifi ed into three groups: drought escape, drought avoidance and drought tolerance ( Turner et al. 2001 ). Drought escape allows the plant to complete its life cycle during the period of suffi cient water supply before the onset of drought. Normally the life cycle is shorter and plants set some seeds instead of complete crop failure. An example of drought escape is the Early Soybean Planting System, now used widely in the southern USA. In this system, short season cultivars are planted during March-April in regions where later maturing cultivars have previously been grown. The early maturing cultivars start fl owering in late April to early May and set pods in late May, thus completing the reproductive stage before the period of possible drought in July-August ( Heatherly and Elmore 2004 ). The second mechanism, drought avoidance, involves strategies which help the plant maintain high water status during periods of stress, either by effi cient water absorption from roots or by reducing evapotranspiration from aerial parts. The third mechanism, drought tolerance, allows the plant to maintain turgor and continue metabolism even at low water potential, e.g. by protoplasmic tolerance or synthesis of osmoprotectants, osmolytes or compatible solutes ( Nguyen et al. 1997 ). In soybean, a widely accepted equation for grain yield (Y) under water-limited conditions is a function of three components, namely the amount of water transpired (T), water-use effi ciency (WUE) and harvest index (HI); Y = T × WUE × HI ( Turner et al. 2001 ). Maintenance of optimum transpiration, leading to increased WUE, is one of the strategies to improve yield in soybean. Nine secondary traits have been reported to be associated with the likelihood of increasing or maintaining T during drought. These traits are phenology, photoperiod sensitivity, developmental plasticity, leaf area maintenance, heat tolerance, osmotic adjustment, early vigor, rooting depth and rooting density. Additionally, transpiration effi ciency and leaf refl ectance are the other two traits related to WUE ( Purcell and Specht 2004 ).
Signifi cant progress in breeding for drought resistance cannot be made unless the stress is imposed year after year to allow direct selection for yield. For this reason, a fi eld with poor moisture-holding capacity, soil uniformity and a reasonable drought each year is important, but is unfortunately very diffi cult to achieve . In order to enable high throughput and reliable gravimetric measurements of water relation traits and water use in soybean, it is also important to include modern facilities such as a rainout shelter in a breeding program. By using a rainout shelter with a movable roof, researchers at the National Center for Plant Gene Research (NCPGR) in Wuhan, China, have succeeded in dissecting the genetics of complex traits such as drought resistance (Pennisi 2008) . The plant-wise drought treatment strategy developed at the NCPGR provides a useful method for independent evaluation of the individual components of drought resistance (Yue et al. 2006) . In addition, current advances, such as automated plant phenotyping platforms to study the plant responses to soil water defi cit under controlled conditions (Granier et al. 2006) , an automated rotating lysimeter system for evapotranspiration and plant growth measurements ( Lazarovitch et al. 2006 ), non-destructive measurement of plant water status over time using portable unilateral nuclear magnetic resonance ( Capitani et al. 2009 ), and other precision equipments to quantify plant water use should be exploited. Screening of genotypes in such controlled facilities has the advantage of quantifying the morpho-physiological traits that infl uence drought adaptation, which in turn will complement yield selection criteria.
In this review, we fi rst concentrate on the major physiological traits related to the mechanisms of drought resistance in soybean ( Fig. 1 ). Although the emphasis of this article is on soybean, the basic concepts are applicable to drought resistance improvement in other crops as well. Recent advances in marker-assisted selection, functional genomics and genetic engineering approaches to achieve drought resistance in soybean will also be discussed. adapt to drought by developing a longer taproot which helps reach the lower soil layers where water is available ( Taylor et al. 1978 ). In addition, an extensive fi brous root system can be useful for foraging subsoil surface moisture and nutrients such as phosphorus.
One of the major factors infl uencing soybean rooting depth is taproot elongation rate. As taproots are the fi rst formed roots in soybean, identifi cation of genotypes with rapidly elongating taproots under non-stress conditions may allow the determination of deeper rooting ability. Genotypic variation in the taproot elongation rate in soybean cultivars has been well documented in greenhouse studies. Kaspar et al. (1984) evaluated 105 diverse soybean genotypes from different maturity groups. Taproot elongation within a maturity group varied by as much as 1.3 cm d −1 , and contrasting cultivars were selected for fi eld study. The results of these studies indicated that the group of cultivars with faster elongation rates was also more able to deplete soil water 120 cm below the surface. Limited research has been reported regarding genetic variability in producing lateral roots. Under stress conditions, the number of lateral roots per unit of taproot length signifi cantly increased, but no promotion of taproot length or diameter was observed ( Read and Bartlett 1972 ).
Because root growth and water deployment are critical to maintaining function in different environmental conditions, plasticity for root traits is a critical factor to acquire resources. Water stress tends to increase biomass partitioning to roots, increasing the root : shoot ratio. Non-irrigated soybeans showed an increase in root length when compared with irrigated plants, especially in the subsoil ( Huck et al. 1983 ). Signifi cant correlations exist in soybean between drought resistance and various root traits such as dry weight, total length, and volume and number of lateral roots Bartlett 1972 , Liu et al. 2005a ). These authors proposed using these traits as indicators for drought resistance. Soybean plant growth and development are classifi ed into vegetative (V) and reproductive (R) stages. The V stage comprises stages VE to Vn, where VE is vegetative cotyledon emergence and Vn are numbered according to how many fully developed trifoliate leaves are present. The R stage comprises fl owering in stages R1 and R2, pod formation in stages R3 and R4, and seed fi lling and maturation development in stages R5-R8. In soybean, when drought was imposed during the later stages of vegetative and/or the early stages of reproductive growth (R1-R2), a large increase in root growth was observed in non-irrigated plants, especially in the deeper soil layers. Root growth was less affected when drought was imposed at  the R4 stage and ceased at the R5 stage ( Hoogenboom et al. 1987 ). In another study, it was demonstrated that soybean plants which experienced drought before fl owering had higher seed yields than those stressed after fl owering, because they had already developed a large root system before fl owering ( Hirasawa et al. 1994 ). This shows that if a plant develops a large root system during its early vegetative growth, it would be in an excellent position to maintain turgor under drought conditions.
All of these early studies were based on popular cultivars/ genotypes, and less is known about the rooting ability and/ or plasticity of root growth in exotic plant introduction (PIs) lines. The line 'PI416937' has an exceptionally branched fi brous root system Rufty 1993 , Hudak andPaterson 1996 ) which has shown drought resistance in the fi eld by limiting transpiration during the high evaporative demand period (Fletcher et al. 2007) . The fi brous root trait of 'PI416937' also conferred increased nitrogen fi xation through increased root surface area. 'PI416937' possesses the genetic capability to continue root growth in compacted soils ( Busscher et al. 2000 ). However, no correlation was demonstrated between the rooting ability and drought resistance of this PI line, although a positive correlation between rooting ability and seed protein concentration was found ( Pantalone et al. 1996 ).
The vast diversity of PI lines and cultivars worldwide provides valuable resources for identifi cation of germplasm which can be used in breeding for root traits. However, the laborious procedures involved in separation of roots and the lack of rapid screening procedures are the major constraints for genetic improvement of root traits. Since a deep taproot system associated with a moderate number of lateral roots to extract soil water is the most promising genetic mechanism for improving soybean drought resistance, research in this direction is imperative.
Overall, root traits show strong potential for improvement of drought resistance through breeding. However, it would be extremely diffi cult to perform selection based on measuring the root phenotype. Instead, molecular tagging will facilitate breeding for root-related traits. The candidate gene approach is another alternative methodology, which involves choosing a candidate gene for root traits from public data, obtaining primer sequences to amplify the gene, uncovering polymorphisms, developing a convenient procedure for large-scale genotyping, identifying a population for association studies, carrying out an association study of the candidate gene with trait phenotype and fi nally verifying the uncovered associations. This approach was successful in fi nding candidate genes associated with root number in rice under low moisture conditions (Vinod et al. 2006) .
Understanding the physiological mechanisms and genetic regulation of root adaptation to drought will help to identify specifi c genes and metabolic pathways for either gene-based marker selection or genetic engineering to develop soybean with better root-related traits.
Nitrogen fi xation under drought . Nitrogen (N 2 ) fi xation in legumes is very sensitive to soil drying. In drought conditions, soybean not only suffers losses of CO 2 accumulation and reduced leaf area development, but its symbiotic N 2 fi xation is also especially vulnerable ( Sinclair and Serraj 1995 ). In dry soils, this results in a reduced supply of N 2 for protein production, which is the critical seed product of the plant, and consequently lower crop yields ( Purcell and King 1996 ). Several factors have been related to the inhibition of N 2 fi xation under drought, including reduced oxygen availability, reduction in carbon fl ux to nodules, decline in nodule sucrose synthase activity and an increase in ureides and free amino acids ( Durand et al. 1987, Gonzalez et al. 1995, Arrese-Igor et al. 1999. In particular, a close relationship was found between leaf and nodule water potential ( Durand et al. 1987 ). Nitrogenase activity showed a 70 % decrease during the fi rst 4 d of drought, while photosynthesis only declined by 5 % . This implies that water stress exerts an infl uence on nitrogenase activity that is independent of rate of photosynthesis. It was also found that water defi cit directly affected nodule activity through increased resistance to oxygen diffusion to the bacteroid. The increase in oxygen diffusion resistance, decrease in nitrogenase-linked respiration and enzyme activity, accumulation of respiratory substrates and oxidized lipids, and the up-regulation of antioxidant genes together demonstrate that the respiratory activity of bacteroids is impaired in drought conditions, and that oxidative damage occurs in nodules prior to any detectable effect on sucrose synthesis or leghemoglobin (Naya et al. 2007) . Soil drying also leads to the accumulation of ureides in soybean leaves and is thought to be an inhibitor of nodulation ( Sinclair and Serraj 1995 ).
Substantial genetic variation was found in the sensitivity of N 2 fi xation in response to soil drying ( Sall and Sinclair 1991 ). When comparing soybean germplasm with or without the capacity to maintain tissue turgidity, and therefore leaf and nodule function, under drought conditions at the reproductive stage, the germplasm able to maintain tissue turgity showed the least reduction in grain yield . Screening for petiole ureide levels is an effective initial approach to identify soybean lines whose N 2 fi xation is more tolerant of soil drying. By screening a large collection ( > 3,000) of soybean PI lines in this way, eight lines were identifi ed that exhibited N 2 fi xation drought tolerance ( Sinclair et al. 2000 ). In a further study, Sinclair et al. (2003) found that ureide catabolism independent of manganese was active in six of the eight PI lines. The variety 'Jackson', which is more tolerant of N 2 fi xation in drying soils, has been used as a parent for developing high yielding lines for drought conditions. Two high yielding progeny lines for non-irrigated conditions that exploit the N 2 fi xation drought tolerance trait of 'Jackson' are considered a potential genetic resource for improved symbiotic N 2 fi xation and increased yield under rainfed conditions  .
It is unclear whether N 2 fi xation under drought is only regulated at the whole-plant level (e.g. by a systemic nitrogen feedback mechanism) or whether it can also occur at a local nodule level. There is evidence that signaling for biological N 2 fi xation regulation can be provided by an N 2 feedback mechanism involving shoot N 2 status ( Serraj et al. 2001 ). King and Purcell (2005) suggested that a combination of ureide and aspartic acid levels in nodules, plus the transport of several amino acids from the leaves, might be involved in such a feedback inhibition mechanism in soybean. A recent study using a partially droughted split-root system indicated that N 2 fi xation activity under drought stress is mainly controlled at the local level rather than by a systemic N 2 signal ( Marino et al. 2007 ). More research should be directed toward understanding the molecular genetics of the factors limiting and regulating the response of N 2 fi xation under drought.

Shoot-related traits
Stomatal conductance . One of the shoot-related physiological traits that may affect drought tolerance is the decline in whole-plant water use during a soil water defi cit event. As a soil water defi cit develops, plants undergo a transition from the water-saturated phase, in which whole-plant water use is not dependent on the soil water content, to a second phase where water use is directly related to the availability of soil water ( Sinclair and Ludlow 1986 ). This transition is associated with a reduction in the average stomatal conductance, and can occur at different soil water contents in different species (Earl 2003) . Stomatal conductance is a key variable infl uencing leaf gas and water vapor exchange. The effects of drought stress on leaf growth, stomatal conductance and plant water relations have been well addressed in soybean ( Muchow 1985, Bennett et al. 1987, Ray and Sinclair 1998. It was shown that drought stress decreases relative leaf expansion rate, stomatal conductance and leaf turgor, whereas it increases the content of ABA in the leaf and xylem ( Liu et al. 2003 ). In this study, decreased stomatal conductance coincided with an increase in xylem ABA and occurred before any signifi cant change in leaf turgor was detected, indicating that chemical signals (root-originated ABA) control stomatal behavior at moderate soil water defi cit. Relative stomatal conductance decreased exponentially with increasing relative xylem ABA ( r 2 = 0.98) ( Liu et al. 2003 ). Genotypic differences in the ability to keep stomata open despite internal water stress have been reported in soybean (Vignes et al. 1986) . When compared with soybean, other legumes such as green gram ( Vigna radiata ), black gram ( Vigna mungo ) and cowpea ( Vigna unguiculata ) exhibited better stomatal control over water loss ( Vignes et al. 1986 ). Previous research has clearly established that decreases in stomatal conductance are associated with severe plant water defi cits in soybean ( Bennett et al. 1987, Liu et al. 2005b ). Further research is needed to establish the physiological and biochemical basis for stomatal closure during intermittent drought conditions as well as mild soil water defi cits, particularly under fi eld conditions. Control of stomatal conductance under drought is a promising physiological trait for developing drought resistance in soybean, considering the non-destructive nature of the measurement and availability of precise instruments for the measurement of conductance. However, a better understanding of the quantitative inheritance of this trait is vital if it is to be used for this purpose.
Epidermal conductance . A second shoot-related physiological trait which may help identify cultivars with increased drought tolerance and prolonged crop survival during severe water stress is low leaf epidermal conductance (g e ). Total leaf conductance to water vapor is the sum of the stomatal and cuticular diffusive pathways acting in parallel ( van Gardingen and Grace 1992 ). When the stomata are open, cuticular conductance is generally a negligible fraction of total conductance. However, in water-stressed or dark-adapted leaves in which the stomata tend to be closed, the cuticular component of leaf epidermal conductance may exceed the stomatal conductance (Boyer et al. 1997) . Under severe water defi cit when stomatal closure is maximal, g e (the sum of cuticular conductance and any residual stomatal conductance) determines the rate of water loss from leaf tissues and therefore the rate of progression toward critically low (injurious) leaf water content. Species adapted to arid environments tend to have low g e ( Riederer and Schreiber 2001 ), and crop species or varieties with low g e are often those that survive the longest under severe soil water defi cits Ludlow 1986 , Jovanovi et al. 1996 ). Signifi cant variation in epidermal conductance was found among 74 soybean accessions. Expression of genotypic differences in epidermal conductance among the selected accessions was consistent across environments ( Paje et al. 1988 ). Interestingly, these differences in epidermal conductance were not associated with differences in stomatal density. Selection for lower epidermal conductance could allow improved survival of leaves under intermittent water stress, so that suffi cient leaf area remains available for growth after the stress is relieved ( Paje et al. 1988 ). Recently, a signifi cant negative correlation ( r = -0.74) was reported between g e and WUE under drought, which again supports the hypothesis that lower epidermal conductance is a desirable trait for drought resistance (Hufstetler et al. 2007) . Perennial wild-type soybean exhibited lower values of g e compared with various soybean cultivars and G. soja . Among the soybean genotypes, tropically adapted genotypes exhibited the lowest epidermal conductance and more negative osmotic potential when compared with genotypes adapted to temperate regions, suggesting that the adaptability of tropical genotypes might have resulted from selection for survival under drier environments (James et al. 2008a) . Broad-sense heritability for epidermal conductance was signifi cantly different from zero in 10 hybrid populations and ranged from 60 to 93 % ( James et al. 2008b ). The magnitude of the heritability estimates suggested that the trait was highly heritable. However, to validate the usefulness of these traits in breeding for drought resistance, more research is needed to characterize the interrelationship between g e , conductance, maintenance of leaf turgor and plant survival.
Leaf pubescence density . Leaf pubescence is a common feature of xerophytic plants, as well as some crop plants including soybean. Generally, leaf pubescence density increases refl ectance from the leaf, resulting in lower leaf temperatures under high irradiance. Leaf pubescence density is an important adaptive trait for soybean under waterstressed conditions. Densely pubescent lines had increased vegetative vigor, greater root density and deeper root extension (Garay and Wilhelm 1983) . Increased leaf pubescence density may also increase leaf boundary layer resistance by up to 50 % . Reduced leaf temperature, restricted transpiration water loss and enhanced photosynthesis due to lower radiation penetration into the canopy were also reported to be associated with the dense pubescence trait  . Moreover, extra-dense pubescence can signifi cantly reduce the incidence of soybean mosaic virus infection. However, increased pubescence density controlled by the Pd1 and Pd2 alleles produces both positive and negative agronomic effects ( Pfeiffer et al. 2003 ). In particular, these alleles have been shown to reduce seed yield, delay maturity and increase plant height, with a concomitant increase in lodging . Thus, the negative linkage between the leaf pubescence density and seed yield should be considered before introgressing this trait into elite germplasm.
Water use effi ciency . The WUE of a plant is generally defi ned as the amount of biomass accumulated per unit of water used. Genetic variation in WUE has been reported in recent years for many fi eld crops, including soybean ( Mian et al. 1996 ). High WUE can contribute to crop productivity under drought. The positive association between WUE and total biomass yield in a drought environment suggests that improvement of the WUE of a crop plant should result in superior yield performance if a high HI can be maintained ( Wright 1996 ). Soybean cultivar 'Young' exhibited considerably higher WUE (4.4 g DW l −1 ) than did 'PI416937' (3.7 g DW l -1 ) in greenhouse conditions ( Mian et al. 1996 ). In another greenhouse study, soybean cultivar 'Jackson' accumulated more biomass and total nitrogen than 'PI416937' with similar transpirational losses ( Purcell et al. 1997 ). Signifi cant genetic differences were observed for β , which is an empirical seed yield based on estimation of genotypic WUE. The close association between genotypic β and genotypic mean yield ( r = 0.71) provides evidence that soybean WUE can potentially be improved simply by selecting for genotypic mean yield ( Specht et al. 2001 ). In a study using 24 soybean genotypes, WUE ranged from 2.7 g dry matter kg −1 water to 3.4 g dry matter kg −1 water ( Hufstetler et al. 2007 ). It has been proposed that the aperture of stomata could be regulated in such a way that a partial closure of stomata at a certain level of soil water defi cit might lead to an increase in WUE (Liu et al. 2005b) . While promising results were observed, many uncertainties still remain regarding the implementation of irrigation techniques in different crop species under varying environmental conditions. Therefore, a deep understanding of the physiological basis for stomatal regulation and improved WUE in drought-stressed plants is required ( Liu et al. 2005a ).
Another relevant trait related to WUE is transpiration effi ciency (TE), which is the assimilation or dry matter accumulation per unit of transpiration ( Fischer 1981 ). TE is under genetic control and excludes amount of water lost by soil evaporation, and hence should be considered as a potential trait.
Osmotic adjustment . Osmotic adjustment (OA) is defi ned as the active accumulation of solutes that occurs in plant tissues in response to an increasing water defi cit. OA is considered a useful measure because it provides a means for maintaining cell turgor when tissue water potential declines. OA has been shown to maintain stomatal conductance and photosynthesis at lower water potentials, delayed leaf senescence and death, reduced fl ower abortion, improved root growth and increased water extraction from the soil as water defi cit develops (Turner et al. 2001) . The capacity for OA in soybean ranges from 0.3 to 1.0 MPa, which is not as wide as the ranges observed in chickpea ( Cicer arietinum ) and pigeon pea ( Cajanus cajan ) (0.1-1.3 MPa) ( Cortes andSinclair 1986 , Turner et al. 2001 ). Among six common soybean cultivars tested, leaf OA ranged from 0.3 to 0.5 MPa under prolonged drought control conditions (S. Guttikonda, unpublished). In a recent study, the leaf osmotic potential of cultivated and wild soybean accessions was measured in order to assess the capacity for OA. Leaf osmotic potential at 70 % relative water content (RWC) ranged from -2.3 to -3.5 MPa in soybean cultivars and -4.3 MPa in the wild species G. latifolia and G. canescens ( James et al. 2008a ). There was a highly signifi cation correlation ( r = 0.98) between rate of decline in RWC and relative OA among seven different genotypes tested, suggesting that genotypes with higher osmotic potential showed a slower rate of decline in RWC, indicating the ability of these genotypes to maintain turgor for longer periods ( James et al. 2008b ).
Most of the published studies on OA in crops have used only leaf measurements, and few examined the potential benefi ts of OA upon crop yield ( Serraj and Sinclair 2002 ).
One study in wheat ( Morgan 1995 ) claimed putative benefi ts of OA on crop yield. In chickpea, however, the relationship between OA and yield under water-limited conditions was inconsistent ( Leport et al. 1999, Moinuddin and Khanna-Chopra 2004, Turner et al. 2007 ). Turner et al. (2007) studied advanced breeding lines of chickpea over different continents and demonstrated genotypic differences in OA, but the phenotypic expression of OA was not stable and varied from year to year. It was also found that the difference in OA has no effect on yield under terminal drought when grown under rainfed conditions. These inconsistent results could be due to differences in growth habitat, stress level and phenological characteristics of the tested cultivars. An exception where OA may have an important benefi t during drought is when it is directed to roots, where turgor maintenance might lead to an increased rooting depth and would thus increase the amount of water available to the crop. Emphasis should be given to studying the role of OA in root tips to improve crop growth under water defi cit conditions (Serraj and Sinclair 2002) . Another problem with selecting for OA is the limited number of genotypes that could be screened using current water relation traits. Use of multiple psychrometers ( Santamaria et al. 1990 ) and complex relationships between RWC and OA has limited screening to 50 genotypes per study. As suggested by Turner et al (2001) , the development and identifi cation of molecular markers, as in rice and wheat, would enable selection for OA in early generation populations of soybean.

Soybean seed and grain fi lling during drought stress
Yield as measured by the weight of seeds is reduced most by stress occurring during early formation and pod-fi lling stages. Water stress during early reproductive growth (fl owering and pod set) reduces yield, usually as a result of fewer pods and seeds per unit area. It has been shown that root-originated xylem sap ABA can move to reproductive structures causing kernel/pod abortion, presumably via inhibition of cell division in young ovaries ( Liu et al. 2003 ). Stress during seed fi lling, and after pod and seed numbers are fi xed, could affect the seed growth rate by reducing photosynthesis and, therefore, the supply of assimilate available to the seed, ultimately affecting seed size. At the pro-embryo developmental stage, young ovules are undergoing active division, which coincides with rapid pod expansion and is specifi cally sensitive to water defi cit during seed fi lling (Westgate and Peterson 1993) . In a rainout shelter study, the drought tolerance based on seed yield varied among Brazilian soybean cultivars. Cultivars with high drought tolerance maintained a high crop growth rate during the early reproductive stage, associated with high seed yield ( Oya et al. 2004 ). The length of the seedfi lling period is another important determinant of yield in all grain crops, including soybean. Seed fi ll duration is under genetic control ( Smith and Nelson 1987 ) and is sensitive to water defi cit (Meckel et al. 1984) . When soybean plants were subjected to continuous water stress from the beginning of growth stage R6 (early in seed fi lling) until maturity, the plants lost nitrogen and chlorophyll from their leaves more rapidly than non-stressed plants ( de Souza et al. 1997 ). Stress also shortened the seed-fi lling period (R7 occurred up to 7 d earlier), resulting in smaller seeds (maximum reduction of 32 % ) and lower yields (up to 44 % ). Water stress can induce senescence which cannot be stopped by reapplying water, and even short periods of water stress during seed fi lling might decrease yield by up to 23 % ( Brevedan and Egli 2003 ). By applying large quantities of nitrogen fertilizer to supplant the dependency on nitrogen fi xation during water defi cit, Purcell and King (1996) reported an 18 % increase in soybean yield. However, while inherently delayed senescence may be advantageous for yield under optimal growing conditions, it may be of no consequence under post-anthesis stress because then the over-riding stress factor will impose accelerated senescence or leaf killing.
Interestingly, water stress at the fl ower induction, fl owering, pod formation and pod-fi lling stages of growth did not signifi cantly affect the oil or protein content of soybean seeds ( Dornbos and Mullen 1992 ). Furthermore, drought had little effect on the fatty acid composition of the oil, but high air temperature reduced the proportion of polyunsaturated components. A recent study also showed that irrigation generally had no signifi cant infl uence on unsaturated fatty acid accumulation in seed oil in soybean genotypes with altered fatty acid profi les ( Lee et al. 2008 ).
In conclusion, the key to a complete understanding of the relationship between water stress and soybean seed growth lies in determining the effect of stress on the supply of assimilate to individual seeds.

Marker-assisted selection for drought resistance
In environments where water is limited, genetic improvement of a crop for drought resistance is an economically attractive option ( Blum 2002 ). However, despite the large resources committed to soybean breeding, progress in improving drought resistance has been slow for a number of reasons. (i) Identifying lines with the highest yield potential under optimum moisture conditions is an important selection criterion in soybean. Conversely, evaluating lines from low yielding environments under drought conditions is often not considered, because small yield differences among lines do not separate high yielding genotypes from low yielding genotypes. (ii) Historically, the emphasis in soybean breeding was upon resistance to biotic stress rather than abiotic stress such as drought, due to the complexity of trait evaluation. This unfortunately resulted in a narrow genetic base for initiating drought resistance breeding programs. (iii) Drought resistance is governed by several complex factors including genotype, environment and the interaction of the two ( Carter et al. 1999 ). In the future, marker-assisted selection based on genotype will greatly increase breeding effi ciency.
Identifi cation of quantitative trait loci (QTLs) related to drought resistance is one of the promising approaches using marker-assisted selection  ). In soybean, a large number of QTLs have been identifi ed for traits related to agronomic, physiological, seed composition, and both biotic and abiotic stress parameters ( www.soybase.ncgr.org ). However, to date, only a few QTLs have been associated with drought resistance traits ( Table 1 ). Reported QTLs also explain <10 % of the phenotypic variation for those traits. Most studies for QTL identifi cation have included single and small size populations. The QTL markers identifi ed by a single population may not be automatically applicable to unrelated populations without marker validation and/or fi ne mapping ( Nicholas 2006 ).
As discussed above, WUE is one of the important traits associated with drought resistance in soybean. Since water consumption and its effi cient use by crops are related to yield, carbon isotope discrimination (CID) ( ∆ , a measure of the 13 C/ 12 C ratio in plant tissues compared with the air) may provide a useful indirect measure of genetic variation in TE, i.e. the ratio of net assimilation to water transpired, in C3 species ( Araus et al. 1998 ). The expensive equipment and high cost of analyses restrict the use of ∆ in breeding programs. However, ash content and K concentration have been proposed as cheap and easily determined surrogates of ∆ (Tsialtas et al. 2002). Due to its inherent analytical simplicity, estimation of ash content in vegetative organs became an attractive alternative to ∆ for a preliminary screening of large, genetically diverse populations ( Masle et al . 1992 ). Ash concentration (leaf and stem) of crested wheatgrass ( Agropyron desertorum ) was negatively associated with WUE (Mayland et al. 1993) . The relationship between ash content and WUE depends on the maintenance of a constant concentration of minerals in the transpiration stream. While environmental conditions could cause variation in the mineral concentration, comparison of different genotypes under the same environmental conditions would nonetheless provide repeatable responses. A molecular investigation may provide some insight into the genetic relationships between WUE and leaf ash (LASH) ( Mian et al. 1996 ).
To identify QTLs associated with WUE (total plant dry weight/amount of water used in liters) and LASH in 36-dayold, greenhouse-grown plants, Mian et al. (1996) constructed a restriction fragment length polymorphism (RFLP) map from a soybean [ G. max (L.) Merr] population of 120 F 4derived lines from a cross of' Young × PI416937'. Signifi cant ( P < 0.01) phenotypic differences were detected among the lines for both traits. A total of four and six independent RFLP markers were associated with WUE and LASH, respectively, and, if combined, each group of markers would explain 38 and 53 % of the variability in the respective traits. One marker locus (cr497-1), on USDA Linkage Group (LG) J, explained 13.2 % of the variation in WUE, indicating the presence of a major QTL. The authors also found that LASH was negatively correlated with WUE ( r = -0.40), and two QTLs were associated with both WUE and LASH. For each of these QTLs, the allele for increased WUE was associated with reduced LASH.
one QTL was identifi ed for CID on LG-C2, with a phenotypic contribution of <10 % and with no effect on β ( Specht et al. 2001 ). So far QTLs in soybean under water defi cit conditions have been reported only for WUE and LASH. More studies are needed to identify QTLs that infl uence root architecture and shoot turgor maintenance. Mapping for new QTL(s)/ gene(s) and determination of gene action under drought will probably provide key resources to improve tolerance to drought stress in soybean. High-density genetic maps and confi rmed QTLs/genes, which are screened across the various environments and across genetic backgrounds, are the most important criteria for developing drought-resistant soybeans through marker-assisted selection ( Vuong et al. 2007 ).

Functional genomics resources and tools for drought research in soybean
Plant functional genomics has emerged as a new and rapidly evolving scientifi c discipline to study the functions of genes. In the last decade, considerable progress has been made in developing genomic resources for soybean, including the sequencing of the entire soybean genome of approximately 975 Mb ( http://www.phytozome.net/soybean #C). For the genetic and genomic analyses of the soybean genome, precise genetic and physical maps are important. Various types of physical maps have so far been reported using RFLP, random amplifi ed polymorphic DNA (RAPD), simple sequence repeat (SSR) and amplifi ed fragment length polymorphism (AFLP) markers ( Shinozaki 2007 ). A high-density genetic linkage map of soybean using expressed sequence tag (EST)-derived microsatellite markers was generated using a hybrid between the Japanese cultivar 'Misuzudaizu' and the Chinese line 'Moshidou Gong 503' by Japanese researchers ( Hisano et al. 2007, Xia et al. 2007). The possibility of segmental duplications in the previously suggested regions of the soybean genome was confi rmed by inspecting the co-linearity between the genomes of soybean and Lotus japonicus . This comparison, combined with information relating to regulatory networks deduced by transcriptional profi ling, may provide clues toward understanding the evolutionary process of the paleopolyploid soybean genome ( Hisano et al. 2007 ). Microsatellite markers have been discovered from bacterial artifi cial chromosome (BAC) libraries to create detailed genetic and physical maps of the soybean genome, which together cover the soybean genome by > 35-fold ( Shoemaker et al. 2008 ). The constructed Williams 82 Bsty I library contains 92,160 clones with an average insert size of 150 kb and covers approximately 12 genome equivalents ( http://www.soybeangenomics.org/genome_mapping. php ). A genetic map was then integrated into this physical map by anchoring approximately 1,000 SSR and sequence tagged site (STS) markers, developed from ESTs, which are associated with drought responses, disease resistance, seed development and composition traits ( Wu et al. 2008 ). In addition, a six-dimensional pool array has been recently developed from 49,152 Bsty I soybean clones (∼6 × genome equivalents), comprising 208 BAC pools ( Wu et al. 2008 ). The integrated genetic and physical map will be useful for comparative genetic analysis, map-based cloning of QTLs of desired traits, and genomic sequencing. In addition, yeast artifi cial chromosomes have also been created for the purpose of chromosome walking and in situ hybridization ( Santra et al. 2003 ). Recently, the entire soybean genomic sequence has been released with 66,153 protein-coding loci ( http://www.phytozome.net/soybean #C).
EST sequencing projects in soybean have resulted in a wealth of DNA sequence information. A spotted cDNA microarray is available containing 36,000 elements constructed from soybean cDNAs, which were derived from a variety of EST libraries representing a wide source of tissues and organs, developmental stages and stress-related or pathogen-infected plants ( Vodkin et al. 2004 ). The ESTs isolated from the subtracted library of drought-stressed soybean root tips and submitted to GenBank will greatly contribute to the stress-specifi c unigenes for further functional genomics work aimed at better understanding of the drought stress response of the soybean root system ( Valliyodan and Nguyen 2008 ). This result will provide an important resource for revealing the transcriptional regulatory networks which will illuminate our understanding of the drought stress responses of soybean at different developmental stages.
Full-length cDNAs are important tools for functional analysis of genes, and provide a valuable resource for the improvement of soybean productivity and drought resistance by biotechnology. Umezawa et al (2008) obtained 6570 new full-length sequences of soybean cDNAs derived from tissues treated with different abiotic stresses, in order to analyze gene functions and structures. About 88 % of the soybean cDNA clones contained complete coding sequences in addition to 5 ′ -and 3 ′ -untranslated regions.
Although the gene expression profi ling data created by microarray analysis have been commonly used as valuable information in functional genomics or in the selection of candidate genes for genetic engineering of soybean plants, the integration of transcriptomics, proteomics and metabolomics represents the frontier of integrated functional genomics. It provides an unprecedented opportunity to gain a whole picture of the physiological and biochemical responses to drought stress. Taking advantage of available proteomic technology, a proteomic analysis of droughttreated soybean roots, leaves and seeds has revealed regionspecifi c regulation of metabolism in the growth zone of water-stressed soybean roots (M. Yamaguchi, unpublished). While transcriptomic data and proteomic analyses do not tell the whole story of what might be happening in a cell, metabolic profi ling can give an instantaneous snapshot of the physiology and biochemical changes in the cell (Morgenthal et al. 2007). It has long been known that in addition to ABA, a variety of osmolytes, including certain sugars (such as raffi nose family oligosaccharides, sucrose, trehalose and sorbitol), sugar alcohols (such as mannitol), amino acids (such as proline) and amines (such as glycine betaine and polyamines), accumulate in plants in response to drought stress ( Umezawa et al. 2006, Seki et al. 2007 ). However, the degree to which the metabolome is altered in response to drought stress has remained a mystery. The recent development of highly sensitive metabolite profi ling technologies, such as gas chromatography or liquid chromatography coupled with mass spectrometry [GC-time-of-fl ight (TOF)/MS or UPLC-Q-TOF-MS] makes such comprehensive analysis possible. Recently, Urano et al (2009) used GC-TOF/ MS to detail the changes that occur in the metabolome of Arabidopsis in response to drought stress, and to assess the extent to which the dehydrated metabolome is reconfi gured by inactivation of the NCED3 gene which is involved in the dehydration-inducible biosynthesis of ABA. Although the role of a number of accumulated metabolites remained unknown, this study has revealed new molecular mechanisms of dynamic metabolic networks in response to drought stress. Additionally, genetic engineering using genes encoding components of drought stress-related metabolic pathways has shown the potential to enhance drought resistance in important crops, including rice, wheat and soybean ( de Ronde et al. 2004a, de Ronde et al. 2004b, Umezawa et al. 2006, Seki et al. 2007 ). Thus, metabolomics, in addition to transcriptomics and proteomics, provides a major tool for characterizing the drought stress response in soybean in the post-genomics era.
Both gain-of-function and loss-of-function approaches can be used to elucidate gene function, as well as provide novel tools and mutant resources for improving drought resistance in plants ( Timpte et al. 1994, Klein et al. 2004. Antisense-and RNA interference (RNAi)-mediated transcriptional or post-transcriptional gene silencing has been used to study soybean gene function ( Buhr et al. 2002, Nunes et al. 2005, Subramanian et al. 2005. Alternatively, fast neutrons, X-rays, γ -rays and chemical mutagenesis are effective methods to generate mutants in plants for drought research, including legumes such as Medicago trunculata , L. japonicus and soybean. LegumeBase ( http://www.legumebase.agr. miyazaki-u.ac.jp ) provides a list of mutant resources in these legumes. Chemical mutagenesis can be applied to soybean followed by screening for mutations in a target of interest, using a strategy known as Targeting Induced Local Lesions IN Genomes (TILLING). Recently, TILLING has been applied to four mutagenized soybean populations in two backgrounds ( Cooper et al. 2008 ). Radiation has been also used to induce genomic lesions, and deletion mutants have been developed with γ -rays (Bhatia et al. 1999) , X-rays and fast neutrons ( Men et al. 2002 ).
Insertional mutagenesis is another method by which gene expression can be modifi ed in plants, creating mutants with either gain or loss of function for drought studies. T-DNA-tagged populations in rice  ) and maize ( Cowperthwaite et al. 2002 ) proved to be very useful resources for functional genomics, including drought functional genomics. However, the development of T-DNA mutants takes a lot of time for soybean transformation (5-7 months), and considerable labor to generate an adequate number of insertions (Parrott and Clemente 2004) . To overcome this problem, it has been suggested to couple T-DNA regions with transposon-based elements ( Brutnell 2001 ). Isolation of loss-of-function mutants may be hindered by the lack of an obvious phenotype due to gene redundancy, especially in a paleopolyploid such as soybean ( Mathieu et al. 2009 ). To circumvent this problem, a gain-of-function approach can be adopted to generate mutants by transforming plants with multiple enhancer sequences ( Jeong et al. 2006 ). By using the Ds-transposon system, Mathieu et al. (2009) have created activation-tagging, gene and enhancer trap elements in soybean to establish a transposon-based mutagenesis repository which currently houses around 900 soybean events, with fl anking sequence data derived from 200 of these events ( http://digbio.missouri. edu/gmgenedb/index.php ).
The availability of these functional genomics tools and resources, coupled with the forthcoming completely annotated genome sequence, represents an exciting era in soybean genetics. With the availability of mutant populations and major achievements in marker-assisted selection and soybean transformation, it is now possible to study and characterize the genes related to drought resistance, leading ultimately to better soybean productivity.

Genetic engineering for drought resistance
In response to stress, plants activate a number of defense mechanisms that function to increase tolerance to adverse conditions. A large array of genes is activated and consequently a number of proteins are produced to contribute to the metabolic pathways that lead synergistically to the enhancement of stress tolerance ( Shinozaki and Yamaguchi-Shinozaki 2000, Seki et al. 2003, Valliyodan and Nguyen 2006. The response to drought stress, which involves a number of biochemical-molecular mechanisms, is complex, as schematically represented in Fig. 2 . Although the role of the vast majority of these genes in plant adaptation to environmental stresses remains to be elucidated, the function of some of regulatory genes, which encode regulatory transcription factors (TFs), and that of functional genes, which encode metabolic components, have been described ( Yamaguchi- Shinozaki andShinozaki 2006 , Tran et al. 2007a , b ).
The application of this emerging understanding to the genetic engineering of food crops has already led to examples of improved drought tolerance and increased yield under drought ( Hu et al. 2006, Nakashima et al. 2007, Nelson et al. 2007, Hu et al. 2008. Introduction of the SNAC1 and ZmNF-YB2 TFs into rice and maize, respectively, enhanced the drought tolerance of transgenic plants as demonstrated by fi eld studies. Transgenic rice overexpressing the SNAC1 gene had 22-34 % higher seed setting in the fi eld than the negative control under severe drought stress conditions at the reproductive stage. SNAC1 transgenic rice plants are drought tolerant not only due to the increased expression of genes encoding proteins functioning in the production of osmolytes, detoxifi cation and redox homeostasis, and in protection of macromolecules, but also because of the increased stomatal closure in transgenic leaves which prevents water loss from the plant ( Hu et al. 2006 ). Similarly, SNAC2/OsNAC6 transgenic rice plants have improved drought tolerance because of the enhanced expression of a large number of genes encoding proteins with predicted stress tolerance functions such as detoxifi cation, redox homeostasis and proteolytic degradation ( Nakashima et al. 2007, Hu et al. 2008. Transgenic maize overexpressing the ZmNF-YB2 gene produced an ∼50 % increase in yield relative to controls when water was withheld from the planted fi eld area during the late vegetative stage ( Nelson et al. 2007 ).
A number of important successes have also been reported in creating drought tolerance in economically important plants, by genetic engineering of functional genes. For example, trehalose, a non-reducing disaccharide that functions as a compatible solute to protect biological structures under stress, has been used to increase drought tolerance in rice and tomato ( Garg et al. 2002 , Cortina andMarcia 2005 ). Transgenic rice plants were developed with a trehalose-6-phosphate synthase/phosphatase gene using an ABAinducible promoter. Under drought stress, these transgenic rice plants showed increased amounts of trehalose, leading to higher levels of tolerance not only to drought, but also to salt and low temperature stresses, without any negative effect on plant growth or grain yield ( Garg et al. 2002 ). Transgenic canola was generated using a farnesyltransferase ( ERA1 ) antisense construct driven by a drought-inducible rd29A promoter. The developed transgenic canola plants showed increased ABA sensitivity, a signifi cant decrease in stomatal conductance and decreased transpiration when tested under drought conditions (Wang et al. 2005) . Recently, transgenic tobacco was engineered to overexpress an isopentenyl transferase gene ( IPT ) from Agrobacterium tumefaciens driven by the SARK promoter, which was isolated from bean ( Phaseolus vulgaris ) and shown to be induced by both drought stress and maturation. The transgenic plants showed improved drought tolerance by maintaining high water content and photosynthetic activity ( Rivero et al. 2007 ).
Although a considerable body of literature now supports the development of stress-resistant transgenic crops by genetic engineering as a promising approach, only one study has been reported to date in soybean. Transgenic soybean plants overexpressing the Arabidopsis ∆ 1 -pyrroline-5carboxylate synthase gene, P5CR , showed greater tolerance to drought stress due to an increased free proline level and RWC and a reduced level of reactive oxygen species, particularly hydrogen peroxide ( de Ronde et al. 2004a, de Ronde et al. 2004b, Kocsy et al. 2005. Since this report, much effort has been directed toward isolating drought stress-related genes in all physiological and biochemical aspects of drought stress, TFs and their respective promoters, which will in turn provide novel tools and resources for the development of engineered soybean with improved drought resistance. The identifi ed soybean candidate genes are usually tested for their ability to enhance drought tolerance in Arabidopsis before pursuing their engineering into soybean. Chen et al. (2007) isolated an AtDREB homologous gene GmDREB2 from soybean. Constitutive overexpression of GmDREB2 increased the survival rate of transgenic plants without growth retardation under water defi cit and salinity stress conditions ) . In addition, a few basic-leucine zipper genes encoding bZIP TFs were cloned from soybean, and found to be induced by drought and high salt treatments. By overexpressing some of these genes in Arabidopsis , the researchers found that transgenic plants have increased freezing and salt tolerance, but no signifi cant difference in drought tolerance ( Liao et al. 2008a, Liao et al. 2008b ). In two other recent studies it was found that overexpression of a GmERF TF family member, the GmERF089 gene, and a chillinginducible GMCHI gene promoted enhanced tolerance to drought stress in tobacco and Arabidopis , respectively ( Zhang et al. 2008, Cheng et al. 2009 ). Recently, 31 GmNAC genes, including the six genes previously identifi ed by Meng et al. (2007) , were identifi ed and cloned from soybean. Systematic expression analysis of these 31 GmNAC genes demonstrated that nine genes are dehydration inducible (Tran et al. 2009). These nine GmNAC genes and their respective promoters are promising tools for genetic engineering to improve drought resistance of soybean, as the NAC family was reported to be a major group of TFs that play a role in root development and stress tolerance in plants ( Tran et al. 2004, Hu et al. 2006, Nakashima et al. 2007, Hu et al. 2008.
To develop soybean plants with enhanced tolerance to drought stress, an understanding of the physiological and biochemical responses and gene regulatory networks is essential. Recent advances in soybean research, ranging from breeding programs to genome sequencing and genomics technologies, provide unprecedented opportunities to understand global patterns of gene expression and their association with the development of specifi c phenotypes, as well as promising tools for the genetic improvement of plants cultivated in adverse environments by molecular breeding or transgenic approaches.

Concluding remarks
Increased demand for soybean products over the past decade has continued to raise global soybean production steadily upward. Indeed, rising demand has prompted producers to shift acres from wheat to soybean. Although there are numerous cultivated and wild soybean accessions in the soybean germplasm collections of the world, only a few germplasm have been screened for drought tolerance. Combining genes from both wild and cultivated species shows promise to obtain genotypes with higher levels of tolerance. During our survey of the literature, we found that there have been only limited efforts to study soybean genotypic variation for important physiological traits related to drought tolerance, such as cell membrane thermostability, osmotic adjustment, canopy temperature depression and metabolic traits such as antioxidants and ABA. Considering the importance of soybean to the world economy, the identifi cation of suitable traits and genotypes associated with drought tolerance in soybean is absolutely necessary. More concentrated efforts are needed to screen for resistant germplasm, discover new candidate genes and combine these genes for higher levels of drought tolerance in soybean. While combining genes/traits, preference should be given to pyramid a set of genes/traits contributing to a whole regulatory pathway. With the recent availability of the entire soybean genome sequence and genome-wide expression profi ling data, it is possible to identify key genes regulating drought tolerance. However, the identifi cation and use of traits related to drought tolerance and the development of suitable screening techniques are the prime criteria for cultivar development. Integration of genomics, proteomics and systems biology platforms will undoubtedly lead to the discovery of novel candidate genes and pathways involved in drought resistance.
Transgenic approaches have been shown to be powerful tools to help understand and manipulate the responses of plants to stress, but this can be achieved only when studied by precise physiological and biochemical investigation of transgenic plants under stress conditions. Moreover, to ensure that the responses of the transgenic plants to water stress treatments are comparable with those in fi eld conditions, the plants should be subjected to the same or comparable drought regime that crops experience in the fi eld. Greater emphasis should be placed on the severity and duration of the stress, the method of imposing the stress and the Approaches to improve soybean drought resistance parameters which will be measured. To facilitate accurate evaluation in the fi eld of the genotypes for drought resistance, screening should be performed under managed drought environments over multiple locations and years. An interdisciplinary approach combining the knowledge of plant breeders, crop physiologists and molecular biologists will be apposite to evaluate the complex response of plants to develop better drought-resistant soybeans in future.

Funding
United States Department of Agriculture (USDA)-Cooperative State Research, Education, and Extension Service (CSREES) and the Missouri Soybean Merchandising Council (MSMC) (to the soybean genomics drought research program at the University of Missouri).