Utilize of genetic diversity and marker-trait association to improve drought tolerance in rice (Oryza sativa L.)

Drought stress in one of the main problems for rice crop as it reduces the production and productivity of the grain yield signicantly. In Egypt, many restrictions were made on the cultivation of rice due to its high water demand. Producing promising drought-tolerant rice cultivars in a combination with high yielding is one of the main targets for rice breeders. To address this challenge, a set of 22 highly diverse rice genotypes were evaluated under normal and drought conditions. Morphological, physiological, yield traits were recorded on each genotype. Drought susceptibility index (DSI) was estimated for six yield traits to identify the most drought-tolerant rice genotypes. High genetic variation was found among genotypes tested in the experiment. Under normal conditions, the highest phenotypic correlation was found between grain yield (GY) and sterility percentage (SP) (- 0.73**), while, it was among GY and chlorophyll content (CC) (0.82**) under drought conditions. To identify quantitative trait loci (QTL) controlling yielding traits under drought and normal, single marker analysis was performed between all yield traits under both condition and a set of 106 simple sequence repeat (SSR) marker alleles. The genetic association analysis revealed 14 and 17 QTL under drought and normal conditions, respectively. The most drought-tolerant genotypes were selected based on DSI, the number of QTL in each selected genotypes, and the level of genetic diversity. As a result, ve genotypes (Giza 178, IET1444, GZ1368-S-5-4, Nahda, Giza 14) were identied as the most promising drought-tolerant rice genotypes. Eight QTL controlling drought tolerance were identied in Giza 178, Nahda, and GZ1368-S-5-4, while, four QTL were found in IET1444. The number of different QTLs were estimated among the ve selected genotypes. Giza 178 and GZ1368-S-5-4 shared the same QTLs. Seven different QTLs were found among Nahda, IET1444, GZ1368-S-5-4, and Giza 14. Combining information from phenotypic traits, genetic diversity analysis, and QTL identication was very useful in identifying the true drought-tolerant rice genotypes that can be used for crossing in the future breeding program. and NOP (r = 0.91**), the lowest positive signicant correlation was between LR and Ph; and GY and NOP and GY with r of 0.54*. a negative signicant correlation with CC, NOT, PL, NOP and GY with a range extending from − 0.49* (NOP) to – 0.73** (GY). the highest positive signicant correlation between NOT and NOP (r = 0.96**) the signicant FLA and NOP (r = 0.45*). negative signicant correlations with FLA, PL, NOP and GY. and signicantly with FLA and GY. negative signicant correlation was between and LR Compared to normal conditions, GY had signicant correlations with a higher number of traits. GY had positive signicant correlations with CC, NOT, PL and NOP, while it had a negative signicant correlation with SP. Under drought stress, GY had new signicant positive correlations with FLA and HGW while, it was negatively and signicantly correlated with LR. Moreover, the sizes of correlations between GY and the other traits were higher than those under normal conditions. This result indicates that rice plants make substantial physiological and morphological changes to alleviate the effect of drought. For example, GY was signicantly and negatively correlated with LR under drought (r = -0.57**), while, this correlation was non-signicant under normal conditions. This indicates that the tightly rolled genotypes had a signicant reduction in grain yield and low drought tolerance. Radiation on the rice leaf is reduced due to the leaf rolling resulting in the low aspiration that leads to stomatal closure. As a result, the ag leaf area is decreased. This can be observable from the positive and signicant correlation between LR and FLA under only drought stress (r = 0.69**). Therefore, the favorable genotypes tend to have a delay in leaf rolling and higher recovery rate after drought removal Remarkably, there some QTLs reported in this study were previously reported in some drought studies in rice. RM144-228 which associated with increased NOP under both conditions and GY under drought stress was found to be associated with an increased number of branches under well-watered conditions in Liu et al. (2010). Also, the same marker was reported with its association with early owering under drought conditions by Yue et al. (2005). Liu et al. (2010) reported the signicant association between RM161 and an increased number of branches under well-watered conditions, the same marker was found to signicantly associated with increased NOP in our study. RM215 was reported with its association with early owering under drought and drought response index by Yue et al. (2005), decreased plant height under drought by Wang et al. (2016), increased root length Ikeda et al. (2006). This marker was associated with decreased Ph and SP under drought in the current study. Xu et al. (2005) and Shamsudin et al. (2016) reported a signicant association between RM154 grain yield under normal and drought stress, respectively. In our plant material, RM154 was found to be associated with decreased LR under both conditions. Bearing in mind that LR was signicantly correlated with GY (r = − 0.57**). Furthermore, we found that the same marker was associated with early owering under drought stress. The same signicant association between the same markers and early owering under drought stress was reported by Shamsudin et al. (2016). RM285-405 was found to be associated with increased grain yield under drought stress. The same ndings were reported by Venuprasad et al. (2012). Increased CC under control and drought conditions was controlled by RM307-128 in our study. The same marker was found to be associated with a decreased number of days to leaf rolling by Yue et al. (2005). A signicant negative correlation was found between leaf rolling and CC (r = − 0.50*). The same marker associated with increased grain yield under drought stress. In the study of Yue et al. (2005), a signicant association between RM307 and four root traits under drought was reported. Moreover, RM307 was signicantly associated with increased grain yield under drought (Yadav et al. 2019). Genetic diversity of germplasm plays an important role in producing verities with target traits. It is very crucial to understand the level of genetic diversity of germplasm before designing a breeding program to improve target traits such as tolerance to biotic stresses, abiotic stresses, high yielding attributes, etc. (Eltaher et al. 2018; Onde et al. 2019). Here, the 20 genotypes can be considered a diverse collection from different countries and their genetic diversity level was extensively studied by Salem and Sallam (2016). The information from genetic diversity was incorporated with the results of drought tolerance obtained from this study on the same set of genotypes. According to Salem and Sallam (2016), the genotypes were divided into two possible subpopulations namely, G1 and G2 (Table 6). Remarkably, as an average, the genotypes in G2 had a higher level in drought tolerance than those in G1. The genotypes in G2 was more drought tolerant in ve DSIs than genotypes in G1. Moreover, G2 subpopulation had better yield attributes control condition than genotypes in G1. Under drought stress, the performance of both groups was nearly the same except NOP which could be affected by the environmental conditions. Therefore, the genotypes in both subpopulations have a good level in drought tolerance with a little bit of bias for the G2 genotypes. High genetic distance was found among all genotypes in both subpopulations ranging from 0.48 to 0.90 (Salem and Sallam 2016). Therefore, choosing genotypes from both subpopulations could be very fruitful in improving drought tolerance in rice through breeding programs. Furthermore, the genotypes of this population are indica and tropical japonica. Genotypes from both types were identied as drought tolerance in a combination with high yielding by Onde et al. (2019).


Introduction
Rice is one of the most important food crops in the world and major food crop of about half of the world's population. It is the staple diet for most people living in south and south east of Asia, Which is grown in 154 million hectares world-wide in a wide range of environments, (IRRI, 2004). Egypt is facing a major challenge i.e. food security strongly and directly affected by water shortage as second challenge.
Where water shortage is one of the most limiting factors in more than 30% of paddy elds in Egypt, particularly in the terminal areas and the developed varieties cannot perform well under water shortage (AbdAllah 2010). Egypt has restricted the cultivation of rice areas due to its huge water demand. Thus, producing drought-tolerant genotypes is urgently needed to keep food security safe and alleviate poverty.
Environmental stresses, such as water de cit and temperature rises are major factors limiting plant growth and productivity. Yield insurance can only be attained depending on the processes determining plant development and its responses to stress. Among the crops, rice as a submerged crop, is probably more susceptible to water de cit stress than most other plant species. The shortage of irrigation water is one of the major obstacles for increasing rice production not only in Egypt but also worldwide (AbdAllah et al. 2013). Most of the commonly, young seedling and reproductive stages are particularly sensitive to water de cit. It has been established that water de cit stress is a very important limiting factor at the initial phase of plant growth and establishment. It affects both elongation and expansion growth (Kusaka et al. 2005 andShao et al. 2008).
Drought tolerance is a complex trait controlled by many genes. Therefore, the genetic improvement of drought tolerance is a big challenge (Sallam et al. 2019a). Drought tolerance can be assessed by scoring traits that have a relation to water de cit and by which tolerant and susceptible genotypes can be discriminated. These traits can be morphological, physiological, yield traits, etc. Scoring more than trait related to drought tolerance is very helpful in identifying the true drought tolerant genotypes. Stress susceptibility index is also one of the most effective tools to discriminate the tolerant and susceptible genotypes (Lazar et al. 1995 Drought-tolerant rice genotypes can be produced by breeding research through screening potential diverse germplasm which has genetic variation in drought tolerance. Furthermore, improving rice yield should be combined with high tolerance to drought as farmers are interested to pro tably produce their crops in dry environments. The only drawback of breeding research is the cost and time consuming to achieve its goals. The recent advances in DNA technology and genomics paved the way to identify genomic regions controlling important traits such as drought tolerance in crops (Sandhu et  The high polymorphic markers can be used in genetic diversity, QTL, gene identi cation studies, etc. In addition to the previous point, phenetic selection only can be misleading due to human errors or environmental factors. Instead, identifying how many QTL in each selected genotype contains can be used in deciding the best parents for crossing Therefore, combining breeding with genetic studies such as genetic diversity and QTL analyses can be very fruitful in accelerating cultivar production with target traits through genomic selection. Extensive genetic diversity and population structure were analyzed by Salem and Sallam (2016) in a rice diverse population that was phenotyped for drought tolerance in this study. The level of genetic diversity found among the genotypes was utilized along with QTL identi cation and phenotypic screening for several physiological, morphological and yield traits to genetically identify the true-promising drought-tolerant rice genotypes which can be used in future rice breeding program.

Plant material:
A set of 22 local and exotic rice genotypes (varieties and/or lines) representing a wide range of diversity for several agronomic and physiological characters were selected for the study. The names, pedigree and origin of these genotypes were previously described in Supplementary Each genotype was planted in seven rows of ve meters length each. Adopting spacing of 20 × 20 cm spaced plants, and seedlings of each genotype were individually transplanted in the permanent eld. These materials were replicated three times in a randomized complete block design (RCBD) with the same set of genotypes in two experiments (under normal and drought conditions). Drought stress was imposed by using ush irrigation ( ush irrigation is one of the surface irrigations without standing water after irrigation) every 12 days were applied after two weeks from transplanting to harvesting and all recommended cultural practices were followed.
Data of agronomic and yield and its component traits were recorded from ve randomly selected reprehensive plants in all genotypes in each replication. The methods adopted for recording the observation for each of all studied traits are presented below.

1) Agro-physiological traits
The following phisological and agronomic traits Agronomic were measured on each genotype; days to 50% owering (DTF, days) was calculated as the average between numbers of days from sowing to the rst panicle exertion and number of days from sowing to the last panicle emergence; plant height (Ph, cm) was measured from the soil surface to the tip of the main panicle at maturity; panicle length (PL, cm) the main panicle of each plant was measured from the base to the tip of the panicle excluding owns at complete maturity; leaf rolling (LR) was estimated by visual estimation based on methods proposed by (De Data et al., 1988), ag leaf area (FLA, cm 2 ) the FLA of 20 leaves were measured using leaf area meter (model L1-3000 A) and chlorophyll content (CC) was measured by using chlorophyll mater (SPAD-502) Minnolta Camera Co. Ltd., Japan).

2) Yield and its components traits:
Moreover, Yield and its components such as the number of panicles/plant (NOP) was counted the number of panicles/plant when all plants at the ripening stage; sterility percentage (SP, %); the un lled grains of the main panicle were separated and counted and sterility percentage was calculated; 100-grain weight ( HGW, g) was recorded as the weight of 100 random rice grains/plant and grain yield/plant (GY, g) was recorded as the weight of the individual plant grain yield and adjusted to 14% moisture content.
The drought stress was fully monitored, and the total amount of water consumed was estimated using water counters. Soil moisture content was gravimetrically determined on an oven-dry basis. At each sampling date, duplicate soil samples were taken to a depth of 60 cm using an auger. The samples were immediately transferred in tightly closed aluminum cans to the laboratory where they were weighted, oven-dried at 105 °C for 24 h, and reweighed after which their moisture content was determined. Field capacity and permanent wilting percentage were determined. The bulk density was determined using the core method to a depth of 60 cm (Klute, 1986). The soil was

Results
Genetic variation in drought tolerance in rice: The analysis of variance for all traits scored in this study presented in Table 1. High signi cant differences between the treatments for all traits. Signi cant differences among replications were only found for HGW. High genetic variation was observed among all genotypes for all traits. Among all traits, PL did not have signi cant differences in G × T. The minimum, maximum, mean, and coe cient of variation (CV) values for each traits under normal and drought conditions are presented in Table 2. The main average of all genotypes under drought conditions was lower than nourmal conditions for DTF, Ph, CC, FLA, PL, NOP, HGW and GY. Under nourmal conditions. On the other hand, clayey in texture, whereas particle size distribution was 57.1% clay, 31.5% silt, and 12.7% sand. Soil pH was 8.2, and the electrical conductivity of soil and irrigation water was 2.1.

Statistical analyses of phenotyping data:
The analysis of variation (ANOVA) was carried out by PLABSTAT software (Utz 1991) using the following model Y ijt = µ + g i + r j + t k + gt ik +grt ijk where Y ijk is an observation of a genotype i in a replication j tested in a treatment k (normal vs drought stress) k, µ is the general average; g i , t k , and r j , refer to the effects of genotypes, treatment and replications, respectively. gt ik is genotype × treatment interaction. grt ijk is genotype × replications × treatment interaction (error) Heritability of the genotypes for each trait was calculated as (h 2 = genotypic variance / phenotypic variance). The Spearman rank correlation coe cient was calculated for the phenotypic correlation between traits. All graphical presentations were performed by R software and Microsoft Excel 2016.
The drought susceptibility index was calculated for Ph, CC, FLA, HGW, SP and GY. Drought intensity (DI) was calculated according to (Lazar et al. 1995

) as follow
Where Y D is the average all genotypes for the respected trait under drought stress, while, the Y N is the average of all genotypes for the same respected trait under normal conditions. The drought susceptibility index (DSI) was estimated for each genotype and calculated according to (Lazar et al. 1995

) as follow
Where X D is the mean performance of each genotype for the respected trait under drought stress, while, the X N is the mean performance of each genotype for the same respected trait under normal conditions Genotyping and genetic analyses: DNA extraction and simple sequence repeat marker analysis were extensively described in detail by Salem and Sallam (2016). The analysis of SSR markers on the 22 genotypes revealed 106 polymorphic alleles.
Single marker analysis (SMA) was performed using R software to test the marker-trait association between the 106 SSR markers alleles and yield traits under normal and drought conditions. The phenotypic variation explained by a marker (R 2 ) and allele effects of the present allele were estimated using R software. Allele effects of the present allele loci were calculated as differences between the means of the marker classes (1 vs. 0). The positive signs indicate that the present allele increases the trait, while negative signs are associated with a decrease in the trait.
Population structure and genetic diversity were analyzed by Salem and Sallam (2016) and utilized in this study to genetically select the most promising drought-tolerant genotypes. the main average under normal conditions was lower than drought conditions in LR and SP traits. The highest CV (%) was found for LR (34.19%) under N, while SP had tad the highest CV under D. The performance of each genotype in each trait under both treatments are presented in Supplementary Table 2. The lowest effect due to drought stress was found for CC (14.06%), while, the highest reduction occurred in GY with 39.84%. The drought had a great effect on sterility percentage with an increase of sterility reach to 79%.  Drought susceptibility index (DSI) was estimated for each genotype based on six important yield traits (Ph, CC, FLA, HGW, SP, and GY) to determine the most tolerant and susceptible genotypes (Supplementary Table 3). The distribution of all genotypes for each DSI is presented in Fig. 1 (Fig. 2). Two genotypes ETT1444 and Giza 178 were found to be tolerant based on all DSIs. Moreover, three genotypes (GZ1368-S-5-4, Nahda, Giza 14) were found to be tolerant to drought in ve DSI. Therefore, we considered these ve genotypes as the drought tolerant genotypes (Supplementary Table 3). On the other hand, three genotypes (IR74, Giza 177, and YabaniM7) were found to be tolerant in only one DSI and susceptible to ve DSI. These genotypes were considered susceptible to drought stress. Furthermore, among the tolerant genotypes for each DSI, the most three tolerant (the lowest DSI values) and three susceptible genotypes (the highest DSI values) for each DSI were selected (Supplementary Table 3 Phenotypic correlations among traits: The phenotypic correlation among traits under normal and drought stress are presented in Fig. 3 and Fig. 4 The phenotypic correlation among the DSIs for the seven traits is presented in Fig. 1 Single marker analysis for drought tolerance: Marker-trait association was tested between the SSR markers and all phenotypic traits scored in this study (normal conditions and drought stress). The summary of QTL associated with traits under normal and drought stress conditions is presented in Table 3. As a total, the number of QTLs found under normal (16) was higher than those detected under drought stress (14).  The details of the SMA for all traits are presented in Table 4.   (1) Maker allele refers to the name of the SSR primer followed by the band size.
(2) Chromosome number in rice genome (3) Phenotypic variation explained by marker alleles (4) allele effect of the present allele. AE referee to the effect of visible band (marker allele) of the SSR marker on the trait. Positive values increase the trait, while negative values decrease the trait.
Interestingly, a set of eight markers were found to be associated with more than one trait under normal and drought conditions (Table 4).
Noticeably, some SSR markers were associated with the same trait under normal and drought conditions. For example, By looking for the number of QTL existed in the most ve drought tolerant genotypes, the highest number of total detected QTL (15)   Use of genetic diversity analysis to improve drought tolerance: The genetic analysis and population structure were performed on the 22 )see material and methods). The analysis of genetic diversity and population structure divided the 22 genotypes into two possible subgroups (subpopulations). The rst subpopulation (G1) included AgamiM1, Arabi, Bala, IET1444, IR20, IR22, IR24, IR50, IR64, IR74, Nahda, YabaniM1, while, the second subpopulation (G2) included Yabani15, Yabanilulu, YabaniM7, Giza14, Giza171, Giza172, Giza177, Giza178, Giza181, and GZ1368-S-5-4. The mean of all genotypes in each group was calculated for each trait to study the properties of genetic diversity between the two groups in terms of yield attributes and drought tolerance (Table 7). Interestingly, genotypes in G2 performed better in six and ve agronomic traits under normal and drought than those in G1, respectively. Under control and drought stress conditions and compared with genotypes in G1, genotypes in G2 were late in owering, taller stems, higher in CC, higher in PL, higher in NOP. Moreover, genotypes in G2 were lower in SP under normal, and higher in GY under drought conditions. For DSI, genotypes in G2 had a higher level of drought tolerance ( ve DSIs; CC, FLA, HGW, SP, and GY) compared with genotypes in G1. (1) All genotypes were divdied into two subpopulation according to Salem and Sallam (2016) Discussion Genetic variation in drought tolerance: The analysis of variance revealed highly signi cant differences between the two treatments (p < 0.01) for all traits scored in this study, indicating that there was an observable effect of drought on the tested genotypes. This result can be supported by estimating the effect of drought stress on the genotypes in Supplementary Table 1. Sterility percentage as the most trait affected by drought stress. At the vegetative stage, drought stress increased the sterility of spikelets by disrupting the initiation of owers leading to a reduction of grain lling followed by a signi cant reduction in grain yield (Akram et al. 2013). The severe reduction in GY (40%) further supports the previous conclusion as GY was the second trait after SP affected much by drought stress. Chlorophyll content (CC) was the most trait that less affected by drought stress with a reduction reached 14%. Although drought tolerant genotypes tend to increase chlorophyll content under drought stress (Mukahirwa 2015), all genotypes in this study had a reduction in CC. The high genetic variation found among the genotypes in all trait is very useful for rice breeders to select the most promising drought tolerant genotypes.
To characterize each genotype for its ability to tolerate drought, drought susceptibility index (DTI) was calculated for six different yield traits (Ph, CC, FLA, SP, HGW and GY). Most of the earlier studies focused on calculating DSI for GY or HGW to distinguish the tolerant and susceptible genotypes. In our study, DSI for six yield traits was estimated to identify the most drought tolerant genotypes as Ph, CC, FLA and SP are also very important traits that affect the nal grain yield. Giza 14 was the most tolerant for Ph, while IET4444 was the most tolerant genotypes for CC, SP, HGW and GY. As the main objective was to select the most promising drought tolerant genotypes, we selected the genotypes that were drought tolerant according to the calculated DSIs for the six traits. Five genotypes were considered the most promising drought tolerant genotypes; Giza 178, IET1444, GZ1368-S-6-4, Nahda and Giza 14. Considering more than a trait to determine the most promising genotypes are highly recommended in the selection process. Selection of the most promising genotypes based on multiple traits was also reported in wheat  however, the characterization was at the seedling and germination stage. Here, we tested them under eld conditions, con rming that these genotypes can be used in a breeding program to improve drought tolerant at different growth stages.

Page 11/18
Phenotypic correlations: Compared to normal conditions, GY had signi cant correlations with a higher number of traits. GY had positive signi cant correlations with CC, NOT, PL and NOP, while it had a negative signi cant correlation with SP. Under drought stress, GY had new signi cant positive correlations with FLA and HGW while, it was negatively and signi cantly correlated with LR. Moreover, the sizes of correlations between GY and the other traits were higher than those under normal conditions. This result indicates that rice plants make substantial physiological and morphological changes to alleviate the effect of drought. For example, GY was signi cantly and negatively correlated with LR under drought (r = -0.57**), while, this correlation was non-signi cant under normal conditions. This indicates that the tightly rolled genotypes had a signi cant reduction in grain yield and low drought tolerance. Radiation on the rice leaf is reduced due to the leaf rolling resulting in the low aspiration that leads to stomatal closure. As a result, the ag leaf area is decreased. This can be observable from the positive and signi cant correlation between LR and FLA under only drought stress (r = 0.69**). Therefore, the favorable genotypes tend to have a delay in leaf rolling and higher recovery rate after drought removal Analysis of marker-trait association: Identifying QTL associated with drought tolerance is considered one of the most important steps to genetically improve drought tolerance in rice. Many studies depend on phenotypic selection for improving drought, however, phenotypic selection could be misleading due to environmental effects or/and human errors. Therefore, it would be more useful to use the advances in DNA molecular markers and their ability to allow identifying genomic regions associated with important traits (e.g. drought tolerance). Drought tolerance is a complex trait controlled by many genes. Therefore, identifying QTL associated with target traits under drought stress environment is still the main interest for breeders and geneticists to make faster genetic improvement in rice through advanced breeding programs and marker-assisted selection. In this study, several QTLs associated with yield and its attributes under normal and drought conditions were identi ed using SMA. SMA is one of the QTL analyses which is used to test marker-trait association and it provides valuable information on the genetic architecture of complex traits such as a number of possible QTL, additive effects of the markers, and genetic variation explained by markers Ramchander et al. 2016). In the current study, the results of SMA found a set of 16 and 14 QTL associated with yield and its traits under normal and drought conditions, respectively.
Interestingly, eight SSR markers were associated with more than one trait under drought and control conditions (Table 5). This con rms the feature of MAS choosing highly SSR promising markers for improving yield and its traits under both conditions. These QTLs have a pleiotropic effect at one genetic locus. This will help in improving many traits by using this marker in MAS. Amar et al. (2008) and Basunanda et al. (2010) reported that QTLs governing many traits frequently occur at closely linked loci in the genome. Of the eight markers, six were associated with two QTLs under drought, while, one marker was associated with two QTLs under normal conditions. The present allele of each marker had one, and consistent, the trend regarding its effects on drought tolerance. Moreover, the allele effects pattern in different traits follows the pattern of the correlations between the traits. For instance, the correlation between GY and CC under drought stress was 0.82**. In our study, these two traits shared two signi cant markers (RM285 and RM307) with positive allele effects.
Under control, RM307-122 increased GY and decreased SP. There was a negative signi cant correlation between SP and GY under normal conditions (r = -0.73**) Table 5 Markers having signi cant association with more than one trait under normal and drought conditions.

Marker allele Traits
Control Drought yield under normal and drought stress, respectively. In our plant material, RM154 was found to be associated with decreased LR under both conditions. Bearing in mind that LR was signi cantly correlated with GY (r = − 0.57**). Furthermore, we found that the same marker was associated with early owering under drought stress.  (2016), the genotypes were divided into two possible subpopulations namely, G1 and G2 (Table 6). Remarkably, as an average, the genotypes in G2 had a higher level in drought tolerance than those in G1. The genotypes in G2 was more drought tolerant in ve DSIs than genotypes in G1. Moreover, G2 subpopulation had better yield attributes control condition than genotypes in G1. Under drought stress, the performance of both groups was nearly the same except NOP which could be affected by the environmental conditions. Therefore, the genotypes in both subpopulations have a good level in drought tolerance with a little bit of bias for the G2 genotypes. Promising drought-genotypes as candidate parents for a future breeding program: The main objective of any plant breeders is to select suitable parents with target traits for crossing to nally produce cultivars having a combination of all target traits from the two parents. Phenotypic selection only maybe not enough and misleading due to the environmental conditions which affect the quality of phenotyping. The recent advances of DNA technology provided useful molecular markers (e.g. SSR and SNP) by which genetic diversity and QTL controlling target traits can be explored and utilized in selecting the promising parents possessing target genes controlling target traits. In the current study, we incorporate the results obtained from the phenotypic data with results of the genomic data including QTLs and genetic diversity studies to generally and phenotypically select the true promising drought tolerant genotypes. Four criteria were used, rst; phenotypic selection based on DSIs for six important yield traits. As a result, two genotypes were tolerant in all DSIs and three in ve DSIs (supplementary Table S3), second; a number of different QTLs each selected genotype contained (supplementary Table S4)., third; the number of different QTLs between each pair of selected genotypes (Table 7), fourth; the level of genetic diversity among the ve genotypes. By considering the four criteria together, Nahda is highly recommended be a candidate parent that can be crossed with Giza 14, Giza 178 and GZ1368-S-5-4 because Nahad and the other candidate genotypes are from different subpopulations meaning that the genetic similarity is very low. Moreover, Nahda had 7 different QTLs from the other genotypes. Therefore, crossing Nahda with these genotypes can be resulted in producing rice cultivars having combination of the different QTLs from both parents. Crossing between GZ1368-S-5-4 and Giza 178 may be less e cient as both parents have the same QTLs and belong to the same subpopulation. Identi ed drought-tolerant rice genotypes based on phenotypic and genomic levels can be successfully used to increase the genetic diversity of drought-tolerant rice cultivars.

Conclusion
Genetic improvement of drought tolerance in rice can be achieved based on results obtained at phenotypic levels. Extensive genetic analyses including QTLs and genetic diversity were very useful in selecting the truly promising drought tolerant genotypes that can be used for crossing to produce cultivars that have higher levels of drought tolerance. Genetic identi cation of the number of QTLs in each selected parent con rms the results of phenotypic selection as it can be vague due to environmental conditions or human errors. Incorporating genetic analysis with phenotypic data is very important to accelerate breeding programs by selecting suitable genotypes as parents to improve target traits. Moreover, phenotypic selection based on more than one trait could also help in excluding bad performance genotypes.   Table 3 Page 17/18 Phenotypic correlation among all traits scored under normal and drought conditions