Genetic components of root architecture and anatomy adjustments to water‐deficit stress in spring barley

Funding information Deutsche Forschungsgemeinschaft e.V. (DFG) and Institute of Bioand Geosciences, Plant Sciences (IBG-2), Jülich., Grant/Award Number: PAK 770 Abstract Roots perform vital roles for adaptation and productivity under water-deficit stress, even though their specific functions are poorly understood. In this study, the genetic control of the nodal-root architectural and anatomical response to water deficit were investigated among diverse spring barley accessions. Water deficit induced substantial variations in the nodal root traits. The cortical, stele, and total root cross-sectional areas of the main-shoot nodal roots decreased under water deficit, but increased in the tiller nodal roots. Root xylem density and arrested nodal roots increased under water deficit, with the formation of root suberization/lignification and large cortical aerenchyma. Genome-wide association study implicated 11 QTL intervals in the architectural and anatomical nodal root response to water deficit. Among them, three and four QTL intervals had strong effects across seasons and on both root architectural and anatomical traits, respectively. Genome-wide epistasis analysis revealed 44 epistatically interacting SNP loci. Further analyses showed that these QTL intervals contain important candidate genes, including ZIFL2, MATE, and PPIB, whose functions are shown to be related to the root adaptive response to water deprivation in plants. These results give novel insight into the genetic architectures of barley nodal root response to soil water deficit stress in the fields, and thus offer useful resources for root-targeted marker-assisted selection.

and Sinclair, 2011). Studies indicate that roots play a crucial role in waterstress perception (Janiak et al., 2015;Ksouri et al., 2016), water acquisition (Ehdaie et al., 2012;Palta and Yang, 2014;Lynch et al., 2014;Paez-Garcia et al., 2015), as well as adaptation and tolerance to water-deficit stress (Geng et al., 2018). Considerable variation in root traits that are regulated by multiple genes have been observed in many crop species (Steele et al., 2007;Bernier et al., 2009;Manavalan et al., 2012;Pacheco-Villalobos and Hardtke, 2012), indicating that water-stress adaptation and yield in crops can be improved via selection for root traits in breeding programs.
Crops with deeper and thinner root systems are more favorably adapted to soils undergoing scenarios of water-deficit stress than those with shallow and thick rooting systems (Ram, 2014;Lynch, 2014). Important root traits for water-stress adaptation include: greater primary root elongation, deeper root systems, suppression of the lateral root branching, redistribution of branch root density from surface to depth, and elongation of root hairs (Jovanovic et al., 2007;Wasson et al., 2012, Uga et al., 2013, Lynch, 2013Smith and De Smet, 2012;Comas et al., 2013). These traits may be controlled by several molecular networks (Kulkarni et al., 2017) that regulate gene expression and induce the accumulation of stress proteins to modulate plant-water balance. The suppression of root growth after initiation (arrested roots) has been observed in plants under water-deficit stress (Xiong, Li, & Zhang, 2006;Xiong, Wang, Mao, & Koczan, 2006;Sebastian et al., 2016;Zhan et al., 2015).
Despite the critical and adaptive roles of roots (Tuberosa et al., 2003;Lynch, 2007;Tron et al., 2015), our understanding of their genetic basis of adaptations against water-deficit stress is poor, in part because of the challenge of root phenotyping in the field (Wasson et al., 2012;Topp et al., 2013;Burton et al., 2015;Khan et al., 2016;Merchuk-Ovnat et al., 2017) and the uncertainty about which traits to target. Root system characteristics, including elongation, growth angle, and branching pattern/density, are determined by genetic and environmental factors . The evaluation of desirable root traits would enhance the dissection of root adaptive response to water-deficit stress.
Thus, would facilitate the development of high-yielding and water-deficit stress resilient cultivars that are capable of accessing water in deeper soil layers. The genetics of root adaptations to water deficit involves the integration of several aspects of root biology and models that analyze the root morpho-anatomical changes. Selecting genotypes with appropriate root architecture may increase yields (Lynch and Beebe, 1995). Lynch (2013) and Lynch and Wojciechowski (2015) have proposed using root architectural and anatomical traits to accelerate understanding of the root subsoil water exploration and acquisition in plants. The emergence of DNA-markers, efforts to sequence the whole barley genome, and the availability of powerful biometric methods have made the identification of quantitative trait loci (QTL)/genes associated with yield and waterstress tolerance possible. These water stress-related QTL can be used as markers in breeding programs for developing drought-tolerant (Farooq et al., 2009(Farooq et al., , 2010Ashraf, 2010) and high-yielding genotypes.
Genome-wide association study (GWAS) based on linkage disequilibrium (LD) has been applied in crops to dissect the genetic and molecular basis of several highly complex quantitative traits (Jighly et al., 2015; Contreras-Soto et al., 2017; Fang et al., 2017;Oyiga et al., 2018Oyiga et al., , 2019Ogbonnaya et al., 2017;Thoen et al., 2017). Decades of research have led to the uncovering of genes involved in root growth and development, such as Deeper Rooting 1 (Uga et al., 2013;Arai-Sanoh et al., 2014), Retarded Root Growth (Zhou et al., 2011), Roothairless5 (Nestler et al., 2014), Root Systems Architecture 1 (Rosas et al., 2013), and Crown Root-less1 (Coudert et al., 2015). However, most of these studies were performed in the genetic backgrounds of plants such as Arabidopsis and tropical cereals such as rice, and maize, but not in temperate cereals such as barley and wheat. Moreover, few studies have analyzed the genetic control of the nodal root architectural and anatomical changes during episodes of soil water shortages via GWAS in plant species (Zaidi et al., 2016;Kadam et al., 2017); but none has so far been performed in barley to elucidate for these components, which could act as a model for other temperate cereals.
This study aims to explore the phenotypic variation existing among 192 diverse barley genotypes to identify QTL involved in both additive and epistatic effects of architectural and anatomical nodal root response against soil water deprivation stress in barley and to provide insights into their genetic control. This study provides an important useful resource for undertaking further fine mapping studies, and finally, proof-offunction of the causative genes.

| Plant material
The plant material used in this study consists of 192 genotypes of a barley diversity panel. They were constructed from the barley core collection and the barley gene bank collection at IPK Gatersleben, Germany, including 111 two-rowed and 81 six-rowed spring barleys originating from Europe and Russia (96) (Table S1).

| Field evaluation trials
The field evaluations were conducted in 2013, 2014, and 2015 at Campus Klein-Altendorf Research Facility (50 37 0 N, 6 59 0 E), University of Bonn, Germany, under rainfed (control) and water-deficit stress (rain-out shelter) conditions. The rain-out shelter is made up of an electrical motorized system for rolling part of the roof cover. The roof cover opens to equilibrate with the external ambient conditions and closes during rainfall to exclude rain water. In both control and water-deficit conditions, the GWAS panel was grown in a lattice square design of 0.8 m long rows and 0.21 m between row plot size. The plots were irrigated by moveable overhead sprinklers which were programmed to deliver 5.00 mm/day water per day. Water stress was introduced by withholding water to the plants at BBCH20 (tiller-initiation stage) and continued until data collection at heading (BBCH51). All plots were maintained by adopting all standard agronomic practices. The root architectural and anatomical traits were collected from two (in 2013) and three (in 2014 and 2015) replicates per genotype in control and water-deficit stress conditions. Figure S1 shows the soil moisture content (0-30 cm) of the experimental plots in 2013, 2014, and 2015 under control and waterdeficit stress conditions. 2.3 | Phenotyping root architecture (morphology) by "Shovelomics" The diversity panel was root phenotyped by "Shovelomics" (Trachsel et al., 2011). In brief, genotypes at BBCH51 (in control and waterstressed plots) were excavated with a shovel at a distance of~0.2 m away from the plant base to avoid root destruction. The lumps of excavated soil containing the roots were dissolved by submerging in a bucket of fresh water for~5 min. Thereafter, roots were gently washed to remove the remaining soil debris and rinsed with clean water. The nodal root growth angles of roots from the main shoots and tillers were measured by (a) placing clean roots on a phenotyping board fitted with a large protractor (2013) and (b) taken to the imaging station "field photo box" for photo-image acquisition (2014 and 2015). In both cases, root growth angles were determined by measuring the angle between the soil surface (horizontal line) and the shallowest nodal roots. Since waterdeficit stress inhibits root growth after initiation, the number of nodal and stagnated/arrested nodal roots were obtained by counting from the main shoot and tillers under control and water-deficit conditions.

| Phenotyping of root anatomical traits using laser ablation tomography (LAT)
To investigate the impact of water-deficit stress on the nodal root anatomy, sections (1-2 cm length) of nodal roots ablated from the main-shoot and tillers at a position 1 cm from the root base were obtained: two (in 2013) and three (2014 and 2015) replicates for each genotype/treatment. The harvested root sections were immediately placed in Eppendorf safe-lock tubes containing 75% ethanol for preservation until analysis. Anatomical images were obtained from the root segments via ablation using LAT (Chimungu et al., ,, 2015Strock et al., 2019;Hall and Lanba, 2019) at Penn State University. LAT is a semi-automated system that uses a pulsed laser beam (Avia 7000, 355 nm pulsed laser) to ablate root tissue at the camera focal plane ahead of an imaging stage. Here, the root samples were incrementally extended into the beam, vaporized or sublimated, and imaged simultaneously using a Canon T3i camera (Canon Inc. Tokyo, Japan) and 5× micro lens (MP-E 65 mm) on the laser-illuminated surface. The root images obtained were analyzed using RootScan, an image-analysis tool developed for analyzing root anatomy (Burton et al., 2012). The primary and secondary anatomical root traits obtained via pixel counting (in square millimeters) in control and water-deficit stress conditions include (Table 1): mCRA, mRXSA, mXVA, mTSA, mNXV, tNXV, tCRA, tRXSA, tTSA, tXVA, mAXVA, tAXVA, mXVA/mRXSA, and tXVA/tRXSA.

| Statistical analyses of the phenotypic traits
Year-specific analyses of variance (ANOVA) were performed on the replicated root data obtained using restricted maximum likelihood (REML) model, to investigate the effects of genotypes, water deficit, and their interactions across seasons. Effects due to planting positions (row-andcolumn effects) in the field plots were accounted for by including "Replication/Row*Column" (Gilmour Thompson, & Cullis, 1995): rows crossed with columns nested within replication in the REML as random effects; whereas, the genotype and water-deficit stress treatment effects were considered to be fixed. Significant differences (P ≤ .05) among genotypes, between water-deficit stress treatments, and their interactions were determined using WALD statistics. The best linear unbiased estimates T A B L E 1 Description of root morphological and anatomical traits evaluated in this study

Root Traits Description
Root Architectural Nodal Root Traits  (Fernandez, 1992), where Y p = BLUEs of genotypes under control, Y S = BLUEs of genotypes under water deficit, and Y p = mean BLUEs of genotypes under the control. DTI index is suitable for discriminating genotypes on the basis of drought tolerance status and yield potential (Fernandez, 1992;Sio-SeMardeh, Ahmadi, Poustini, & Mohammadi, 2006;Mohammadi et al., 2010).
The broad-sense heritability (H 2 ) of the traits were also calculated as a ratio of the components of genetic and phenotypic variances, as implemented in GENSTAT 16 for REML (O'Neill, 2010). Pearson correlation and summary statistics were obtained for the traits genotypic means using SPSS software (SPSS version-16, Inc, Chicago, IL, USA).

| Genotyping of the barley diversity panel
The diversity panel was genotyped with 9K iSELECT SNP chip (Jia et al., 2019) and genotyping-by-sequencing (GBS) assays. The DNA was extracted from seedlings at the two-leaf stage as described by Stein et al. (2001). Thereafter, the GBS analysis and enzymatic digestion were performed and after library preparation the enriched DNA fragments were pair end sequenced using NGS technology PstI (manuscript in preparation). All GBS SNP markers were aligned against the reference barley genome sequence "150831 barley pseudomolecules.

| Population structure and linkage disequilibrium (LD) analysis
The population structure of the panel was analyzed based on a Bayesian clustering method as implemented in STRUCTURE v.2.3.4 (Pritchard et al., 2000) using 785 unlinked markers (MAF > 5%; <2% missing data and spaced approximately 2 cM apart). The admixture model was applied with no previous population information. The number of subpopulation (K) tested ranged from 1 to 9, with 20 replications per K. The burn-in period and the number of Markov chain Monte Carlo (MCMC) iterations were 1000,0000 and 1000,0000, respectively. Number of K was determined following the procedure described by Evanno et al. (2005). Thereafter, we plotted the genetic relationships among the genotypes via principal coordinates analysis (PCoA) in GenAlEx 6.5 (Peakall & Smouse, 2012). The pairwise LD was calculated with 6,272 SNPs (with known genetic position) as implemented in TASSEL 5.0 (Bradbury et al., 2007). To investigate the population LD decay rate, the r 2 values obtained for each chromosome were plotted against the genetic distance (cM) between SNP pairs and a cut off of r 2 = 0.1 was chosen as the critical distance up to which a gene locus extends.

| Identification of QTL for barley nodal root responses to water deficit
The QTL associated with changes in the nodal root architectural and anatomical traits due to water-deficit stress were identified using GWAS mixed linear (MLM-PK) approach. Here, the root DTIs were included as phenotype, and the confounding effects of population stratification in the panel was accounted for by including kinship (K-matrix) and population structure (P-matrix) (Price et al., 2010;Kang et al., 2010) as covariates. The K-matrix and P-matrix (principal component analysis) were generated using TASSEL 5.0 (Bradbury et al., 2007). GWAS were performed in TAS-SEL 5.0, and the results obtained were verified using PROC MIXED macro program. The model used is described as: y = Xβ + Sα + Qv + Zu + e, where y is the vector of DTIs; β is the fixed effects other than SNP or population structure; α represents the vector of SNP effects; v is the vector of population effects; u is the vector of polygene background effects; α is a vector of SNP effects; v is a vector of population effects; u is a vector of polygene background effects; e is a vector of residual effects; Q is the matrix from STRUCTURE relating y to v; X, S, Z are incidence matrices (0/1) relating y to β, α, and u, respectively (Yu et al., 2006). The FDR adjusted p-value (qvalue) of 0.01 was estimated with Q-VALUE (Storey and Tibshirani, 2003) and used to correct for the multiple testing. Only significant marker-trait associations (MTAs) with q-values below the FDR ≤ 0.01 threshold were reported. All the associated SNPs in high chromosomal LD with each other were considered to be linked (SNP-clusters).

| Detection of epistatically interacting loci involved in root water stress adaptation
Genome-wide two-locus epistatic interactions were surveyed using the "interactions" function of PROC MIXED procedure in SAS 9.4 (SAS Institute, Cary, NC, USA); by fitting a linear model with P + K variables, the additive effects of the markers and their interactions.
The P-value cutoff was set at 1 × 10 −6 for both total effects and gene-gene interaction effects. Only loci that met these statistical criteria were examined and reported. The significance threshold was determined by cross validation and 1,000 random permutations. The interaction graph was drawn using Circos 0.63-4 (Krzywinski et al. 2009).
2.10 | Identification of candidate genes in the vicinity of the significant SNP markers

| Nodal root architectural traits
Analysis of variance (ANOVA) performed revealed substantial genotypic variability for nodal root architectural traits ( Figure 1; Table 2).
Except in a few cases, the effects of water stress and the interactions

| Nodal root anatomical traits
The ANOVA showed that there were significant (P < .01) differences among genotypes for anatomical nodal root traits in response to water deficit. Significant water-deficit stress and genotype × water-deficit stress interactions were also observed in most of the anatomical root traits. Except in a few cases, the H 2 were moderate, ranging from 10.0 to 52.0% (2013) and 9.0 to 37.0% (2014).
However, highest negative correlations were found for tNXV with tAXVA 3.2 | Population structure, linkage disequilibrium, and SNP marker statistics Analysis of the population structure showed that the maximum ΔK occurred at K = 2, which means that the likely number of sub-populations in this panel is two (Figure 3a). With membership coefficient allotments of <0.6,~44, and~48% of the genotypes were inferred to belong to sub-population 1 and 2, respectively; whereas, 8% of the genotypes were considered hybrids ( Figure S2).
The PCoA plot (Figure 3b) revealed that the panel can be optimally delineated into two groups based on the barley row type (two-and six-row barley), with the first two principal coordinates contributing 21.93% of the genetic variations. We investigated whether there is phenotypic differences in the nodal root traits between the two-and six-row barley ( Figure 3c) and found out that the two groups did not differ significantly for the evaluated nodal root traits, except for mXVA/mRXSA in the 2013 planting.
Summary statistics of the genotypic data (  Figure S3), we concluded that the number of SNPs required for adequate genome coverage and detection of causal QTL was met.
3.3 | QTL associated with nodal root response to water-deficit stress in barley GWAS was performed using 22 root traits and 8987 SNPs. A total of 83 SNPs comprising of 58 and 25 SNPs for anatomical and morphological traits, respectively, were associated with the root response to water deprivation in the diversity panel (Table S3; Figure S4) cosegregate with genes whose gene ontology (GO) terms are related to F I G U R E 3 (a) Population structure analysis inferred using the using the Evanno ΔK method (Evanno et al., 2005) and based on 20 independent runs and K ranging from 2 to 9. The maximal ΔK occurred at K = 2; (b) principal coordinate analysis based on 8,987 SNP/GTBS markers showing a scatter plot of PCo1 (explaining 13.09% of the variance) versus PCo2 (explaining 8.03% of the variance). Colors are according to the barley row-type (two-row, red; six-row, green); and (c) phenotypic variations in root anatomical trait response to drought between two-row and six-row barley. The error bars are presented for each barley row-type root development and stress responses in several plant species (Table S4).
Using the chromosomal LD of the panel ( Most of these intervals were found to be in proximity to QTL previously reported for drought tolerance in wheat (Table 4). Analysis of SNP effect on QTL-5H_3 locus showed that genotypes carrying the "C" allele had higher root drought tolerance values than those with "T" alleles in 2013 and 2014 ( Figure 5).
Among them are 13 loci on 3H, 4H, 5H, and 7H that were also detected via GWAS to exert additive main effects on the nodal root variations under water-deficit stress. The largest number of epistatic QTL were found on 5H (14) and 6H (9) 3.5 | Identification of candidate genes associated with QTL for nodal root responses to water deficit To explore candidate genes at the vicinity of the detected QTL, searches were made with the core sequences of the significant SNPs. BLAST results indicated that some of the SNPs are colocated or cosegregated with genes whose biological functions are related to root response to water-deficit stress (Table S4)

| DISCUSSION
A comprehensive understanding of how roots adapt to stress due to water deficit remains a valuable goal as roots act as sensors for detecting changes of soil water status. Barley roots are composed of axes arising first from: (a) primordia in the seed (seminal/primary roots) and (b) nodes (nodal/crown roots) of the main shoot and tillers. Both primary and nodal root responded differently to soil water deprivation, and the number and length of nodal roots are governed by environmental factors (Kuhlmann and Barraclough, 1987;Rostamza et al., 2013). In this study, the variability in nodal root architecture and anatomy were exploited to elucidate the genetic basis of nodal root response to water-deficit stress in a barley diversity panel. There was wide genotypic variation in the evaluated root phenotypes in response to water-deficit stress, with moderate H 2 and high R for the root traits. This suggests that the

| Effects of water-deficit stress on root architecture
Water-deficit stress induced larger root growth angles of the main shoot (mRGA) and tillers (tRGA) in 2013 and 2015. Increase in root growth angles enhances the ability of plants to avoid drought stress (Uga et al., 2013). Subsequently, this may translate to steeper and deeper root systems that allow plants to access water in deeper soil strata. The number of arrested nodal roots increased, while the number of nodal root per plant decreased during water-deficit stress, an indication that the soil water depletion suppresses nodal root postemergence, root growth (Sebastian et al., 2016), and decreases the number of nodal/lateral roots (Zhan et al., 2015;Gao and Lynch, 2016). The reduction in the number of nodal root during water-deficit stress scenarios may be connected to the plant`s adaptive response to improve the drought stress tolerance by reducing the metabolic costs of soil exploration, permitting greater axial root elongation, greater rooting depth, and hence greater water acquisition from drying soil Saengwilai et al., 2014;Zhan et al., 2015).
Reports have shown that increased number of arrested roots (which is a direct consequence of reduced supply of carbohydrate) in plants under water-deficit stress would reduce metabolic energy costs and conserve water (Fujita et al., 2006;Szalai et al., 2010;De Smet et al., 2003;Ristova et al., 2017).  Lynch (1995) and Strock et al. (2018) have reported that smaller root cross-sectional area induced by plants adaptive response of decreased root secondary growth increases root length (for greater soil exploration) and improves the consumption of growth-limiting resources. Growth allometry is induced by multiple cryptic genetic factors associated with local climate and abiotic stress response (Vasseur et al., 2018), suggesting that the contrasting response of tiller and main-shoot nodal roots may be an adaptive response to water deficit. Thus, may warrants further investigation as a potential root breeding target. Even more so as the presence of smaller diameter roots under soil water scarcity is considered a strategy to maximize absorptive surfaces and increase rates of water and nutrient uptake (Eissenstat, 1992). Richards and Passioura (1989) indicated that root metaxylem vessel regulates crop WUE if water is available in the subsoil, but the top soil is dry. Our findings indicate that water-defcit stress increased the number of xylem vessels, which is in line with reports in rice and wheat (Kadam et al., 2015) and maguey (Peña-Valdivia & Sánchez-Urdaneta, 2009). Increase in the number and thickness of xylem vessels improve tolerance to cavitation, thus would confer resistance to drought (Arend and Fromm, 2007;Awad et al., 2010).
The formation of thick-walled and suberized cell layers at the periphery of the root and around the stele was evident in the water stressed roots. This is an adaptive response to water deficit (Lo Gullo et al., 1998), to enable plants regulate the flux of water from the root to the soil (Hose et al., 2001) and prevent the desiccation of meristematic tissues that is, pericycle and other tissues inside the stele (North and Nobel, 1992). The suberization and lignification of roots affect radial water conductance and may help reduce water loss from mature roots into the dry soil . Root cortical cell size (CCS) and root cortical aerenchyma (RCA) increased under water-deficit stress, as has been observed in maize by Chimungu et al. (2014) and Zhu et al. (2010), respectively.
Larger CCS and RCA are beneficial to plants under water stress because it reduces respiration, nutrient content of root tissues, and the metabolic cost of soil exploration (Zhu et al., 2010;Postma and Lynch, 2011;Chimungu et al., 2014;Lynch et al., 2014;Saengwilai et al., 2014;York et al., 2015) to support increased rooting depth by reducing the proportion of cortical tissue occupied by cytoplasm and by transforming living cortical cells into air space.

| Population structure and relationships
The bayesian clustering algorithm identified two (K = 2) subpopulations in the studied panel, which is in concordance with the PCoA result. The clustering pattern of the panel was based on two-and six-row barley type. The LD analyses revealed that the panel extends over short distances from 2.03 (in 4H) to 4.91 cM (in 7H) when compared to the LD decay of over 10 cM reported in barley (Hamblin et al., 2010;Mezaka et al., 2013;Bellucci et al., 2017) (Pasam et al., 2012), suggesting that QTL-3H_3 may be linked to drought avoidance traits. The expression of circadian clock genes is induced by osmotic stress in the barley root systems (Habte et al., 2014). The QTL-5H_2 associated with tXVA and tAXVA was found in the vicinity of Q5HA reported for relative water content and osmotic adjustment (Teulat et al., 2001). In silico analysis of the QTL-5H_3 showed that it cosegregates with a diagnostic DArT-marker (bPb-5529) detected for heat stress, stay green (Gous et al., 2016), and root length/ root-shoot ratio (Arifuzzaman et al., 2014) and proximal to QTL detected for water content, WUE, and net photosynthetic rate (Gudys et al., 2018;Wójcik-Jagła et al., 2013;. The coincidence of the significant QTLs detected in this study with those previously reported for drought stress adaptive response strongly suggest that they may be linked to genes involved water-deficit response, thus can be exploited to unravel the genetic control and molecular players responsible for root variable responses to soil water depletion.

| Candidate genes in the detected QTL regions for root water-deficit response
Because major responses of plants to water-deficit stress occur at the molecular level via the induction of water stress-responsive genes (Chen & Xiong, 2012), BLAST search was performed in the IPK barley database to identify the genes cosegregating with the significant SNPs detected in this study. The pleiotropic locus at 52.03 cM on QTL-3H_2 associated with root angle and main shoot nodal root Xylem vessel is physically located in the ZIFL2 (HORVU3Hr1G043300) domain. In Arabidopsis, ZIFL superfamilies play key roles in auxin transport, root gravitropism, regulation of stomatal closure, response to karrikin/water deprivation, and root growth and development (Nelson et al., 2010;Remy et al., 2013). Genes controlling traits related to stomatal development and guard cell movements strongly impact the WUE in plants (Ruggiero et al. (2017) and may serve as a potential target for molecular breeding programs. Another pleiotropic SNP at 109.65 cM on 3H is cosegregating with MATE efflux family (MATE; HORVU5Hr1G086830). MATE modulates abscisic acid efflux and ABA sensitivity responses to drought stress (Takanashi et al., 2014;Jarzyniak and Jasinski, 2014;Zhang et al., 2014). SNP (BOPA2_12_11044) at 86 cM on 7H associated with the observed variations in the number of nodal roots is physically located in the domain of PPIB (HORVU7Hr1G095720). PPIB is catalyzed by LAT-ERAL ROOTLESS2 to regulate root gravitropism, lateral root development, and induce thermo-tolerance (Xi et al., 2016;Kaur et al., 2016).
Root gravitropism is a physiological drought response that redirects root growth by gravitational pull toward the down water sources via auxin transport. The mechanisms of auxin transport have also been implicated in this process (Blancaflor, 2013;Sato et al., 2014;Shin et al., 2005). The associated QTL intervals: QTL-3H_2, QTL-5H_3, and QTL-7H were further scanned for detection of other possible candidate genes by searching 1-5 Mbp up-and down-stream away from the genes cosegregating with the significant SNPs (de Koning and Haley, 2005;Ge et al., 2009). We identified two (QTL-3H_2), seven (QTL-5H_3), and six (QTL-7H) additional candidate genes whose molecular functions determine the outcome of root adaptive responses to root water stress response in these regions.

| Epistatic interactions are involved in root trait responses to water deprivation
Epistasis may play an essential role in trait improvement and improves the selection efficiency (Jannink & Wu 2003;Jannink, Moreau, Charmet & Charcosset, 2009). In this study, 13 out of the 44 identified epistatic QTLs also had additive main effect on the nodal root anatomical response to water-deficit stress, an indication that the nonadditive contributions of these loci should not be neglected in the barley root breeding program. Some of the interacting loci detected are cosegregating with genes for drought stress responses. The locus at 118.48 cM on 1H cosegregate with ARF15 and interacted epistatically with 113.32 cM (1H) and 96.73 cM (5H) SNP loci whose sequences are domiciled in the SAUR and PBS gene domains, respectively. ARF15 has been implicated in the activation and repression of early/primary auxin response genes such as Aux/IAA and SAUR gene families (Hagen and Guilfoyle, 2002;Ulmasov et al., 1997Ulmasov et al., , 1999, especially during water-deficit stress. Reports have also shown that the overexpression of SAUR under salt and drought results in higher root length, survival rate, and improved drought/salt tolerance in Arabidopsis plants (Guo et al., 2018). The

PBS gene families positively regulate drought stress in plants via
ABA pathways, stomatal responses, and root growth (Wang et al., 2016;Cui et al., 2018). Two loci at 132.58 cM on 2H coding for RBG1 and at 69.11 cM on 7H interacted epistatically with each other. Reports have shown that RBG1 regulates tolerance to salt and drought stress (Ambrosone et al., 2015) and root growth (Shida et al., 2015). The main-effect QTL at 50.0 cM is in the vicinity of major facilitator superfamily (MFS) protein on 5H. Our result indicated that it is epistatically interacting with eight loci on 2H, 4H, 5H, 6H, and 7H that code for several genes, including locus at 152.36 cM domiciled by Receptor-like protein kinase (RLK) on 5H.
MFS plays a vital role in polar auxin transport and drought stress tolerance (Remy et al., 2013), while Wei and Li (2018) have implicated RLK in the regulation and controlling of root hair development.
In conclusion, this study identified important chromosomal regions harboring candidate genes that might be involved in the architectural and anatomical nodal root response to water-deficit stress in Barley. Going forward, the QTL and genetic variants identified are potential resources for root-targeted breeding for important traits, like drought tolerance improvement in barley.