**6. Molecular mechanisms of drought tolerance in soybean**

In higher plants, the drought stimuli are presumably perceived by osmosensors (that are yet to be identified) and then transduced down the signaling pathways, which activate down‐ stream drought responsive genes to display tolerance effects [52]. The tolerance involves not only the activities of protein receptors, kinases, transcription factors, and effectors but also the production of metabolites as messengers for transducing the signals. Drought tolerance is of multigenic nature, involving complex molecular mechanisms and genetic networks.

The signaling pathway of drought stress is largely overlapping with the signaling pathway of osmotic stresses which has been extensively reviewed [52]. Here, we provide a summary directly related to drought tolerance and include updated information when appropriate.

### **6.1. Searching for osmosenors**

The perception of drought stimulus is presumably via unknown osmosensors. It is speculat‐ ed that these sensors are associated with alterations in membrane porosity, integrity [53], and turgor pressure [54]. From the spatial perspective, membrane proteins, cell wall recep‐ tors, and cytosolic enzymes are all potential sensors for osmotic stress [55, 56]. For example, the families of THESEUS 1 and FERONIA receptor-like kinases (RLKs) in *A. thaliana* are pu‐ tative stress sensors in cell wall to perceive changes in cell wall integrity and turgor pressure [57-59]. On the other hand, from the functional point of view, calcium ion (Ca2+) channels, Ca2+ binding proteins, two-component histidine kinases, receptor-like protein kinases, Gprotein coupled receptors are also potential candidates of osmosensors [60-63]. For instance, AHK1 has been postulated as a cell surface sensor that activates the high-osmolarity glycer‐ ol response 1 (HOG1) mitogen-activated kinase (MAPK) cascade in transgenic yeast. [64]

In soybean, two-component histidine kinases (GmHK07, GmHK08, GmHK09, GmHK14, GmHK15, GmHK16 and GmHK17) and receptor-like protein kinases (GmCLV1A, GmCLV1B, GmRLK1, GmRLK2, GmRLK3 and GmRLK4) have been identified as candidates of osmosensors [65-67]. However, direct evidence for their functions to perceive stress sig‐ nals in soybean is still missing.

### **6.2. Signal transduction under drought stress**

eliminated in plant cells as "by-products" of photosynthesis, photorespiration, and respira‐ tion in chloroplast and mitochondria [44]. Under drought stress, ROS accumulates when the production outweighs the removal [45]. The over-produced ROS will attack cellular compo‐ nents including nucleic acids, protein, and lipid and eventually leads to cell death [46].

A Comprehensive Survey of International Soybean Research - Genetics, Physiology, Agronomy and Nitrogen

ROS scavenging enzymatic activities of superoxide dismutase, catalase, and glutathione per‐ oxidase increased in 5 soybean germplasms under drought stress [24]. The tested germ‐ plasms displayed different basal and treatment-induced level of ROS scavenging enzymatic activities, which were correlated positively to the final seed yield [24]. The study on GmPAP3 from soybean provides another example for the correlation between enhanced ROS scavenging activity and the adaptation to osmotic stress. GmPAP3 is a mitochondria localized purple acid phosphatase [47]. Ectopic expression of the *GmPAP3* gene significantly

Adverse environmental conditions can bring forth the misfolding of proteins that will accu‐ mulate in endoplasmic reticulum (ER) [49]. The resulting ER stress will activate unfolded protein response [49]. By global expression-profiling analyses on soybean leaves exposed to ER stress inducers and polyethylene glycol, a number of genes were identified as candidate regulatory components integrating ER stress signaling and osmotic stress responses [50]. Moreover, overexpression of soybean BiP (binding protein), an ER-resident molecular chap‐ erone, can enhance drought tolerance in soybean [51]. This evidence tightens the link be‐

In higher plants, the drought stimuli are presumably perceived by osmosensors (that are yet to be identified) and then transduced down the signaling pathways, which activate down‐ stream drought responsive genes to display tolerance effects [52]. The tolerance involves not only the activities of protein receptors, kinases, transcription factors, and effectors but also the production of metabolites as messengers for transducing the signals. Drought tolerance is of multigenic nature, involving complex molecular mechanisms and genetic networks.

The signaling pathway of drought stress is largely overlapping with the signaling pathway of osmotic stresses which has been extensively reviewed [52]. Here, we provide a summary directly related to drought tolerance and include updated information when appropriate.

The perception of drought stimulus is presumably via unknown osmosensors. It is speculat‐ ed that these sensors are associated with alterations in membrane porosity, integrity [53], and turgor pressure [54]. From the spatial perspective, membrane proteins, cell wall recep‐ tors, and cytosolic enzymes are all potential sensors for osmotic stress [55, 56]. For example, the families of THESEUS 1 and FERONIA receptor-like kinases (RLKs) in *A. thaliana* are pu‐ tative stress sensors in cell wall to perceive changes in cell wall integrity and turgor pressure

reduces ROS accumulation and thereby alleviates osmotic stress [48].

tween ER stress and drought response through the activity of chaperones.

**6. Molecular mechanisms of drought tolerance in soybean**

**6.1. Searching for osmosenors**

Relationships

218

Abscisic acid (ABA) regulates the physiology (e.g. closure of stomata) and metabolism of plants (e.g. expression of enzymes) to rapidly cope with environmental challenges [68]. Bio‐ synthesis, accumulation, and catabolism of ABA are all crucial for the transduction of ABAmediated signals. The accumulation of ABA in response to drought is associated with the changes in the levels of Ca2+ and ROS [60, 69]. *In planta*, ABA is synthesized in various cell types including root cells, parenchyma cells, and mesophyll cells. Under drought stress, ABA is transported to guard cells to control stomatal aperture [70]. ABA reaching the target tissues and cells will be recognized and the signals will be transduced down the ABA signal‐ osome [71], including ABA receptors (PYR/PYL/RCAR), negative regulators (e.g. group A protein phosphatases 2C), and positive regulators (e.g. SnRK-type kinases).

Components of this system have been discovered in soybean. For example, GsAPK is a SnRK-type kinase from wild soybean that is up-regulated by drought stress in both leaves and roots, but down-regulated by ABA treatment in roots [72]. *In vivo* assay revealed that the phosphorylation activities of GsAPK is activated by ABA in a Ca2+-independent manner, suggesting that GsAPK may play a role in the ABA-mediated signal transduction [72]. Acti‐ vated SnRK-type kinases in rice and *A. thaliana* will phosphorylate target proteins including bZIP transcription factors and membrane ion channels [72].

Perceived stress signals may trigger transient changes in the cytosolic Ca2+ level which acts as a second messenger [73]. Ca2+ sensors in turn transmit and activate the signaling path‐ ways for downstream stress responses [60]. Ca2+ sensors include various types of Ca2+-bind‐ ing proteins: CaMs (calmodulins), CMLs (CaM-like proteins), CDPKs (Ca2+-dependent protein kinases), and CBLs (calcineurin B-like proteins) [74]. Among these Ca2+ sensors, all are plant and protist-specific with the exception of CaM.

Expression of the soybean CaM (GmCaM4) in transgenic *A. thaliana* activated a R2R3 type MYB transcription factor which in turn up-regulated several drought-responsive genes, in‐ cluding *P5CS* (encoding a proline anabolic enzyme) [75]. While the application of Ca2+ af‐ fects the nodulation of soybean [76], the gene encoding a soybean CaM binding protein was found to be differentially expressed in soybean nodules under drought stress [77].

The drought tolerance related CDPK family is well-studied in rice and *A. thaliana* [78, 79]. In isolated soybean symbiosome membrane, a CDPK was demonstrated to phosphorylate an aquaporin called nodulin 26 and hence enhance the water permeability of the membrane. I was hypothesized that this is an integral part of the drought tolerance mechanism [80, 81].

Besides Ca2+, phosphatidic acid (PA) and the intermediates of inositol metabolism are also second messengers for signal transduction [82-84]. However, there are only very limited evi‐ dence supporting the involvement of phospholipid signaling in drought stress response of soybean. The soybean nodulin gene *G93* encoding a ZR1 homologue was down-regulated under drought stress [85]. Plant ZR1 homologue such as RARF-1 in *A. thaliana* may involve in lipid signaling via interaction with phosphatidylinositol 3-phosphate [86].

When plants are subjected to drought stress, accumulation of cellular ROS will trigger the generation of hydrogen peroxide, a signaling molecule that will activate ROS scavenging mechanisms [87]. In soybean, exogenous application of hydrogen sulphide alleviates symp‐ toms of drought stress, probably via triggering an antioxidant signaling mechanism [88].

Many studies support the roles of protein kinases in stress signaling [89, 90]. In plants, the drought responsive signal transduction of the MAPK family (MAPK, MAPKK/MEKK, MAPKKK/MKK) as well as the MAPK phosphatases (MKP) family have been relatively well-studied in *A. thaliana* and rice [89], but remained under-explored in soybean, although a PA-responsive MAPK has been identified in soybean [91].

On the other hand, some non-MAPK type protein kinases found in soybean may be related to drought responses. The soybean gene encoding a serine/threonine ABA-activated protein kinase was found to be up-regulated by ABA, Ca2+, and polyethylene glycol treatments [92]. The With No Lysine protein kinase 1 of soybean is another serine/threonine protein kinase that is a putative osmoregulator [93].

The ubiquitin-mediated protein degradation pathway is also an integral part of the signal transduction network [94]. This pathway directs the degradation of target proteins by the 26S proteasome and is responsive to drought stress. Two ubiquitin genes and one gene en‐ coding ubiquitin conjugating enzyme were identified as differentially expressed genes in nodulated soybean under drought stress [77]. Overexpression of the ubiquitin ligase gene *GmUBC2* enhances drought tolerance in *A. thaliana*, via up-regulating the expression of genes encoding ion transporters (AtNHX1 and AtCLCa), a proline biosynthetic enzyme (AtP5CS), and a copper chaperone (AtCCS) [94].

### **6.3. Drought-responsive transcription factors**

Transcription regulation plays an important role in drought stress response. For instance, using oligo microarray analysis, transcriptions of 4,433 and 5,098 soybean genes were found to be significantly up-regulated and down-regulated respectively when subjected to a no-ir‐ rigation period for 4 days [95]. The signal transduction pathways can ultimately regulate the expression of drought-responsive genes through diverse transcription factors. Transcription factors often target the corresponding *cis*-acting promoter elements, such as the drought stress related elements DRE, ABRE, Gbox, and T/Gbox [95, 96].

In the soybean genome, ~500 transcription factors were *in silico* annotated [96]. Increas‐ ing efforts have been placed to characterize their importance and functions in relation to drought [97-101]. Soybean transcription factors that confer drought tolerance are sum‐ marized in Table 4.


a +: up-regulated; -: down-regulated; nc: no change; nt: not tested

The drought tolerance related CDPK family is well-studied in rice and *A. thaliana* [78, 79]. In isolated soybean symbiosome membrane, a CDPK was demonstrated to phosphorylate an aquaporin called nodulin 26 and hence enhance the water permeability of the membrane. I was hypothesized that this is an integral part of the drought tolerance mechanism [80, 81]. Besides Ca2+, phosphatidic acid (PA) and the intermediates of inositol metabolism are also second messengers for signal transduction [82-84]. However, there are only very limited evi‐ dence supporting the involvement of phospholipid signaling in drought stress response of soybean. The soybean nodulin gene *G93* encoding a ZR1 homologue was down-regulated under drought stress [85]. Plant ZR1 homologue such as RARF-1 in *A. thaliana* may involve

A Comprehensive Survey of International Soybean Research - Genetics, Physiology, Agronomy and Nitrogen

When plants are subjected to drought stress, accumulation of cellular ROS will trigger the generation of hydrogen peroxide, a signaling molecule that will activate ROS scavenging mechanisms [87]. In soybean, exogenous application of hydrogen sulphide alleviates symp‐ toms of drought stress, probably via triggering an antioxidant signaling mechanism [88].

Many studies support the roles of protein kinases in stress signaling [89, 90]. In plants, the drought responsive signal transduction of the MAPK family (MAPK, MAPKK/MEKK, MAPKKK/MKK) as well as the MAPK phosphatases (MKP) family have been relatively well-studied in *A. thaliana* and rice [89], but remained under-explored in soybean, although

On the other hand, some non-MAPK type protein kinases found in soybean may be related to drought responses. The soybean gene encoding a serine/threonine ABA-activated protein kinase was found to be up-regulated by ABA, Ca2+, and polyethylene glycol treatments [92]. The With No Lysine protein kinase 1 of soybean is another serine/threonine protein kinase

The ubiquitin-mediated protein degradation pathway is also an integral part of the signal transduction network [94]. This pathway directs the degradation of target proteins by the 26S proteasome and is responsive to drought stress. Two ubiquitin genes and one gene en‐ coding ubiquitin conjugating enzyme were identified as differentially expressed genes in nodulated soybean under drought stress [77]. Overexpression of the ubiquitin ligase gene *GmUBC2* enhances drought tolerance in *A. thaliana*, via up-regulating the expression of genes encoding ion transporters (AtNHX1 and AtCLCa), a proline biosynthetic enzyme

Transcription regulation plays an important role in drought stress response. For instance, using oligo microarray analysis, transcriptions of 4,433 and 5,098 soybean genes were found to be significantly up-regulated and down-regulated respectively when subjected to a no-ir‐ rigation period for 4 days [95]. The signal transduction pathways can ultimately regulate the expression of drought-responsive genes through diverse transcription factors. Transcription factors often target the corresponding *cis*-acting promoter elements, such as the drought

in lipid signaling via interaction with phosphatidylinositol 3-phosphate [86].

a PA-responsive MAPK has been identified in soybean [91].

that is a putative osmoregulator [93].

Relationships

220

(AtP5CS), and a copper chaperone (AtCCS) [94].

**6.3. Drought-responsive transcription factors**

stress related elements DRE, ABRE, Gbox, and T/Gbox [95, 96].

b The transcription activity was verified by transactivation tests in yeast.

**Table 4.** Soybean transcription factors that exhibit protective function against drought in transgenic plant systems.

### **7. Strategies for breeding drought tolerant soybean cultivars**

To combat water deficit, one of the most effective ways is to breed for new cultivars that ex‐ hibit durable drought tolerance. A combination of conventional breeding, marker-assisted breeding, and transgenic approaches will shed light on the crop improvement program of drought tolerance in soybean.

## **7.1. Conventional breeding**

The high biodiversity nature of soybean allows the stacking of desirable traits through breeding. Since the genetic background of soybean germplasms varies due to spatial adapta‐ tions to diverse habitats, breeding with soybean germplasms from different origins can ef‐ fectively accelerate crop improvement. A recent study suggested that wild soybean exhibited higher allelic diversity compared to cultivated soybean [113]. Since they are sexu‐ ally compatible, wild soybean can potentially serve as a good genetic source in the breeding programs.

Conventional breeding could be a long and tedious process. For example, the breeding of the drought tolerant soybean cultivar Jindou 21 started by breeding Lín Xiàn White Soybean (an old cultivar of higher drought tolerance but lower yield) against Jindou 2 (drought tolerant, high yield, and early maturation). After the selective breeding for six generations, the resulting drought tolerance line was used as a parent of the next selec‐ tion breeding and crossed with Jindou 14. The final selection breeding of Jindou 21 was carried out in the arid region of western Shanxi for seven years (1987 – 1993). Compar‐ ing to its parent Jindou 14, Jindou 21 exhibited increased yield and enhanced drought tolerant. Since then, Jindu 21 has become one of the most popularized drought tolerant soybean cultivars grown in semi-arid regions of Gansu, Ningxia, and Shanxi Provinces of China, particularly in regions where irrigated agriculture is not practical. The total cultivation area is over 3.75 million hectare [114].

### **7.2. Marker-assisted breeding**

Drought tolerance in crops may involve different mechanisms depending on the nature of drought, making it difficult for phenotypic selection and screening through conventional breeding. A recent genomic study showed that soybean is a species of exceptionally high linkage disequilibrium (low recombination frequency) and hence marker-assisted breeding is a promising approach. The same study also identified more than 200,000 tagged SNPs for this purpose [113].

Marker-assisted breeding makes use of DNA markers that are closely linked to the target QTLs, to expedite the selection of progeny lines by replacing some time consuming pheno‐ typic characterizations [115]. For example, delayed wilting response of canopy is associated with drought tolerance [116, 117]. Four QTLs that are associated with this trait were mapped [118], which are significantly associated with 16 SSR markers. One of the identified QTLs was identified in all tested environments which is therefore a promising candidate for mark‐ er-assisted breeding for delayed canopy wilting trait in different environments, including those with the soil type and moisture level inadequately characterized [119, 120].

### **7.3. Genetic engineering**

With the advancement of biotechnology and availability of genomic sequence information, germplasm resources, and increasing genomic tools available for soybean research, trans‐ genic approach has become an attractive alternative strategy in breeding. One critical hurdle of this approach is to identify ideal candidate genes that can improve drought tolerance but do not have a yield penalty when introduced into the soybean genome.

**7.1. Conventional breeding**

cultivation area is over 3.75 million hectare [114].

**7.2. Marker-assisted breeding**

this purpose [113].

**7.3. Genetic engineering**

programs.

Relationships

222

The high biodiversity nature of soybean allows the stacking of desirable traits through breeding. Since the genetic background of soybean germplasms varies due to spatial adapta‐ tions to diverse habitats, breeding with soybean germplasms from different origins can ef‐ fectively accelerate crop improvement. A recent study suggested that wild soybean exhibited higher allelic diversity compared to cultivated soybean [113]. Since they are sexu‐ ally compatible, wild soybean can potentially serve as a good genetic source in the breeding

A Comprehensive Survey of International Soybean Research - Genetics, Physiology, Agronomy and Nitrogen

Conventional breeding could be a long and tedious process. For example, the breeding of the drought tolerant soybean cultivar Jindou 21 started by breeding Lín Xiàn White Soybean (an old cultivar of higher drought tolerance but lower yield) against Jindou 2 (drought tolerant, high yield, and early maturation). After the selective breeding for six generations, the resulting drought tolerance line was used as a parent of the next selec‐ tion breeding and crossed with Jindou 14. The final selection breeding of Jindou 21 was carried out in the arid region of western Shanxi for seven years (1987 – 1993). Compar‐ ing to its parent Jindou 14, Jindou 21 exhibited increased yield and enhanced drought tolerant. Since then, Jindu 21 has become one of the most popularized drought tolerant soybean cultivars grown in semi-arid regions of Gansu, Ningxia, and Shanxi Provinces of China, particularly in regions where irrigated agriculture is not practical. The total

Drought tolerance in crops may involve different mechanisms depending on the nature of drought, making it difficult for phenotypic selection and screening through conventional breeding. A recent genomic study showed that soybean is a species of exceptionally high linkage disequilibrium (low recombination frequency) and hence marker-assisted breeding is a promising approach. The same study also identified more than 200,000 tagged SNPs for

Marker-assisted breeding makes use of DNA markers that are closely linked to the target QTLs, to expedite the selection of progeny lines by replacing some time consuming pheno‐ typic characterizations [115]. For example, delayed wilting response of canopy is associated with drought tolerance [116, 117]. Four QTLs that are associated with this trait were mapped [118], which are significantly associated with 16 SSR markers. One of the identified QTLs was identified in all tested environments which is therefore a promising candidate for mark‐ er-assisted breeding for delayed canopy wilting trait in different environments, including

With the advancement of biotechnology and availability of genomic sequence information, germplasm resources, and increasing genomic tools available for soybean research, trans‐ genic approach has become an attractive alternative strategy in breeding. One critical hurdle

those with the soil type and moisture level inadequately characterized [119, 120].

Rapid gain-of-function experiments using heterologous model plant systems (tobacco, *A. thaliana* and rice) have been employed to screen for potential candidates. Although of lower efficiency, there are established systems of soybean transformation [121-124], allowing direct assessments of the protective functions of both native and heterologous genes in soybean.

Some promising results using this approach have been obtained, although they are all at the experimental stage. For example, AtMYB44 is a R2R3-type MYB transcription factor from *A. thaliana* that participates in the ABA-mediated abiotic stress signaling [125]. Ectopic expres‐ sion of AtMYB44 in soybean led to improved drought tolerance and yet suffered from re‐ duced growth phenotype under normal conditions [126]. Transgenic soybean expressing the *AtP5CR* gene (encoding L-Δ1-Pyrroline-5-carboxylate reductase) resulted in enhanced toler‐ ance toward drought stress with significantly higher relative water content [127]. Introduc‐ ing the *NTR1* gene from *Brassica campestris* (encoding a jasmonic acid carboxyl methyltransferase) into soybean led to increased accumulation of methyl jasmonate and en‐ hanced tolerance toward dehydration during seed germination [128]. Overexpression of the soybean gene *GmDREB3* (encoding a dehydration-responsive element-binding transcription factor) also enhances drought tolerance, in parallel to the accumulation of proline [129].
