Screening for Drought Tolerance in Maize (Zea mays L.) Germplasm Using Germination and Seedling Traits under Simulated Drought Conditions

Maize is known to be susceptible to drought stress, which negatively affects vegetative growth and biomass production, as well as the formation of reproductive organs and yield parameters. In this study, 27 responsive traits of germination (G) and seedlings growth were evaluated for 40 accessions of the Leibniz Institute of Plant Genetics and Crop Plant Research (IPK) germplasm collection, under no stress and simulated drought stress treatments by 10%, 15%, and 20% of polyethylene glycol (PEG). The three treatments significantly reduced G% and retarded seedlings growth, particularly the 15% and 20% PEG treatments; these two treatments also resulted in a significant increase of abnormal seedlings (AS). The heritability (H2) and correlations of the traits were estimated, and drought tolerance indices (DTIs) were calculated for traits and accessions. The H2 of G% values were reduced, and H2 for AS% increased as the PEG stress increased. Positive correlations were found between most trait pairs, particularly shoot and root traits, with 48 highly significant correlations under no stress and 25 highly significant correlations under the 10% PEG treatments, particularly for shoot and root traits. The medium to high heritability of shoot and root seedling traits provides a sound basis for further genetic analyses. PCA analysis clearly grouped accessions with high DTIs together and the accessions with low DTIs together, indicating that the DTI indicates the stress tolerance level of maize germplasm. However, the resemblance in DTI values does not clearly reflect the origin or taxonomic assignments to subspecies and varieties of the examined accessions.


Introduction
Plants are occasionally exposed to a changing adverse biotic and/or abiotic factors, which may prevent plants from performing their maximum potential performance and can threaten their survival [1]. Drought is a primary abiotic constraint affecting crop production worldwide, due to shortages of fresh water. Drought stress on plants occurs when the available water lags continuous plant loss of water by transpiration [2]. With the weather expected to become generally drier and warmer, the situation may be further exacerbated as competition for water intensifies between people and crops [3]. Global climatic change will reduce the productivity of the most valuable crops and induce a detrimental impact on the ecological fitness of cultivated crops [4]. Webber et al. [5] predicted that climate change would lead to yield losses of maize and winter wheat, but drought stress would be more intensive for maize. In low-yielding years, drought stress persisted as the main driver of losses for both crops, with the elevated CO 2 offering no yield benefit [5]. Maintaining crop productivity for  Table 2. Germination and seedling's traits description and abbreviations under control and polyethylene glycol (PEG) drought treatments and the drought tolerance indices (DTIs) used to evaluate traits response to drought treatments.

Trait Abbreviation Description/Methodology
Germination % G% Calculated as G% = n ÷ N × 100, where n is the number of germinated seeds (radicle ≥3 mm) and N is the total number of sown seeds Abnormal Seedling % AS% Seedlings that failed to develop into healthy seedlings after two weeks of sowing

Data Analyses
Box and Whisker charts illustrating the variation of the G%, AS%, and seedling traits under control and drought stress treatments were constructed using Excel 2016 for Windows. In addition, drought tolerance indices (DTIs) were calculated for germination (G-DTI) as the ratio of germination percentage of seeds exposed to each of the PEG treatments compared to the germination percentage of the control seeds. Similar DTIs expressing the change in the root, shoot, and leaf traits were calculated, as described in Table 2. The top 10% accessions scoring best performance expressed as highest means of the examined traits AS% and the bottom 10% accessions scoring the least performance in these traits were determined using Excel 2016 for Windows under control and PEG treatments.
Analysis of variance (ANOVA) was conducted to compare accessions and traits using GenStat Ver. 18 (VSN International, Hemel Hempstead, UK) for the germination and the abnormal seedlings data and the seedling shoot and root traits after 9 days for the control and the 10% PEG treatment, and after 16 days, for the control, and 10% and 15% PEG treatments. The 20% PEG treatment was excluded from the shoot and root data analysis because the germination and seedlings growth rates were too slow. The ANOVA analysis for the leaf measurements and the leaf water content was performed for the control plants and the plants exposed to 10% PEG treatment only after 21 days of sowing. The probability of significance in ANOVA (p < 0.05) was used to indicate significant differences among genotypes, treatments, and interaction effects. Means were separated according to the Fisher's least significant difference (LSD) at 0.05 levels of probability.
Correlations of the studied traits of maize accessions under control and PEG stress treatments were calculated using the GenStat 18. The degree of significance was indicated as p 0.05, p, 0.01, or p, 0.001. Broad-sense heritability was calculated according to Hallauer et al. [34] as follows: H 2 = σ2g / (σ2g + σ2g × t/e + σ2e/re), where σ2g is genotype variance; σ2g × t is the variance of the interaction genotype × treatment, r is the replicates, and e is the error.
The DTIs were used as variables to construct a principal component scatter diagram using the software PAST Version 3.22 based on the Paleontological Statistics software tht wa developed by Hammer et al. in 2001 [35]. The PCA is applied to assign the variables to genotypes and to classify accession based on their sensitivity or tolerance to drought stress. The PCA utilizes orthogonal transformation to convert a set of possibly correlated variables into a set of linearly uncorrelated variables called principal components. This transformation is defined in such a way that the first principal component has the largest possible variance. PCA is sensitive to the relative scaling of the original variables in the PCA scatter plotting visualization. Eigenvectors generated by PCA were used to rank the accessions for their drought tolerance [30]. The grand average of the DTIs of all traits was calculated and used as a measure for the drought tolerance of accessions.

Variation in Germination and Abnormal Seedlings Percentage
The germination percentage (G%) of all accessions varied significantly, as indicated by the ANOVA analysis under both control and PEG stress treatments and showed significant reductions as the PEG concentration increased (Table 3). The box and whisker charts for the G% and G-DTIs ( Figure 1A,B) illustrate substantial variations between accessions and treatments, as indicated by the lower and upper limits of the boxplots for each trait. The G% is less affected by the 10% PEG treatment as compared to the 15% and 20% PEG concentrations. The Zea 3244, the outlier accession in the control G% boxplot ( Figure 1A), showed the lowest G% under the control and the PEG stress treatments (73.3%, 63.33%, 58.33%, and 46.67% for the control, and 10%, 15%, and 20% PEG treatments respectively). Other accessions that showed low G% under control and stress treatments were Zea 677, Zea 3324, and Zea 3244; the latter accession was the only outlier observed for the G% under the 20% PEG treatment. The sensitivity of germination to PEG treatments is clearly indicated by the reduction of G-DTI values as the PEG concentration increased from 10% to 15% and 20%, respectively ( Figure 1B).

Variation in Seedling's Traits
The PEG treatments retarded the seedling growth of all accessions to the extent that it was not possible to evaluate variation in the seedling traits under the 20% PEG treatments. Figure 2 is a photograph showing the retardation of seedling's growth by the 10% and 15% PEG treatments. Box  Table 2. The mean AS% for all accessions showed a successive increase from a value of 7.41% under control conditions to 17.68% under 10% PEG, 32.44% under 15% PEG, and 48.68% under 20% PEG treatments, respectively ( Figure 1C; Table 3). Significant differences (≤0.001) between accessions were recorded under both the control conditions and the PEG treatments. Unlike the drought tolerance indices (DTIs) of all other traits, the AS-DTI increased as the percentage of abnormal seedlings increased. It ranges from a low value of 0.67 for Zea 1224 to the highest value of 14.0 for Zea 1062. Three AS-DTIs of 14.0, 8.0, and 6.33 were scored as outliers for accessions Zea 1062, Zea 711, and Zea 3582, respectively, under 10% PEG. Four AS-DTIs of 27.0, 23.1, 14.8, and 12.0 were observed as outliers in the boxplot for AS-DTI of accessions Zea 711, Zea 1062, Zea 3576, and Zea 1102, respectively, under the 15% PEG treatment and five outliers were observed for AS-DTI values following exposure to 20% PEG, including the above-mentioned accessions plus Zea 323 ( Figure 1D). The G%-DTI and AS-DTI values are given in Table A1.

Variation in Seedling's Traits
The PEG treatments retarded the seedling growth of all accessions to the extent that it was not possible to evaluate variation in the seedling traits under the 20% PEG treatments. Figure 2 is a photograph showing the retardation of seedling's growth by the 10% and 15% PEG treatments. Box and whisker charts show the variation in seedling traits for all accessions, measured for control seedlings and seedlings exposed to 10% PEG after nine days of sowing and for the control, and 10% and 15% PEG treatments after 16 days of sowing ( Figure 3). The mean of the measured traits showed successive reductions as the PEG concentration increased at the two seedling stages of growth, i.e., 9 and 16 days after seed sowing. For the nine days old seedlings, the accessions revealed a highly significant variation (p ≤ 0.001) of the examined traits under control and 10% PEG treatment. Zea 3244. was an outlier in the 16-day-old control seedlings shoot length (C-ShL2) in Figure 3A and shoot dry weight (C-ShDW2) in Figure 2C. In seedlings exposed to 10% PEG, the same accession was the outlier for the shoot dry weight (10%-ShDW2) and root dry weight (10%-RDW2) in Figure 3C,F. The Zea 3244 was also an outlier in seedlings exposed to 15% PEG for shoot dry weight (15%-ShDW2) root fresh weight (15%-RFW2), and root dry weight (15%-RDW2) in Figure 3B,D,F. Highly significant variations (p ≤ 0.001) were found under control, and 10% and 15% PEG treatments for all the 16 days old seedling's traits ( Table 3). The significance and LSD values of ANOVA analysis of control vs. drought treatments for all traits indicated significant variations for all accessions (Table 4). and whisker charts show the variation in seedling traits for all accessions, measured for control seedlings and seedlings exposed to 10% PEG after nine days of sowing and for the control, and 10% and 15% PEG treatments after 16 days of sowing ( Figure 3). The mean of the measured traits showed successive reductions as the PEG concentration increased at the two seedling stages of growth, i.e., 9 and 16 days after seed sowing. For the nine days old seedlings, the accessions revealed a highly significant variation (p ≤ 0.001) of the examined traits under control and 10% PEG treatment, except for the slow-germinating and slow-growing accession Zea 3244. This accession is an outlier in some measurements of the 16-day-old control seedlings shoot length (C-ShL2) in Figure 3A and shoot dry weight (C-ShDW2) in Figure 2C. In seedlings exposed to 10% PEG, the same accession was the outlier for the shoot dry weight (10%-ShDW2) and root dry weight (10%-RDW2) in Figures 3C,F. The Zea 3244 was also an outlier in seedlings exposed to 15% PEG for shoot dry weight (15%-ShDW2) root fresh weight (15%-RFW2), and root dry weight (15%-RDW2) in Figure 3B,D,F. Highly significant variations (p ≤ 0.001) were found under control, and 10% and 15% PEG treatments for all the 16 days old seedling's traits ( Table 3). The significance and LSD values of ANOVA analysis of control vs. drought treatments for all traits indicated significant variations for all accessions (Table 4).    The variation in seedling traits DTIs under 10% PEG for the 9-day-old seedlings and under the 10% and 15% PEG treatments for the 16-day-old seedlings is illustrated in Figure 4 by the lower and upper values of each DTI boxplot. In the 9-day-old seedlings ( Figure 4A), the ShL1-DTI, ShFW1-DTI, RFW1-DTI, and ShDW1-DTI were substantially lower than the RL1-DTI and the RDW1-DTI. In the 16-day-old seedlings, the value of the shoot and root traits DTIs under 10% PEG ( Figure 4B) were generally higher compared to the corresponding values for the 9-day-old seedlings except for RL2-DTI. The ShDW2-DTI for Zea 3582 and Zea 3244 scored much lower values compared to other accessions and appeared as outliers ( Figure 4B). In seedlings exposed to 15% PEG, the DTIs for the examined traits were generally lower compared to seedlings exposed to 10% PEG ( Figure 4B, 4C), but ShDW2-DTI scored higher value compared to the DTIs for other traits. In brief, DTIs for the 16day-old seedlings exposed to 10% PEG treatment were generally higher than their corresponding values in seedlings exposed to 15% PEG. The range of variation is particularly large for ShFW2, RL2, and RFW2. The DTIs for shoot and root traits of 16-day-old seedlings in all accessions are given in Table A2.  The variation in seedling traits DTIs under 10% PEG for the 9-day-old seedlings and under the 10% and 15% PEG treatments for the 16-day-old seedlings is illustrated in Figure 4 by the lower and upper values of each DTI boxplot. In the 9-day-old seedlings ( Figure 4A), the ShL1-DTI, ShFW1-DTI, RFW1-DTI, and ShDW1-DTI were substantially lower than the RL1-DTI and the RDW1-DTI. In the 16-day-old seedlings, the value of the shoot and root traits DTIs under 10% PEG ( Figure 4B) were generally higher compared to the corresponding values for the 9-day-old seedlings except for RL2-DTI. The ShDW2-DTI for Zea 3582 and Zea 3244 scored much lower values compared to other accessions and appeared as outliers ( Figure 4B). In seedlings exposed to 15% PEG, the DTIs for the examined traits were generally lower compared to seedlings exposed to 10% PEG ( Figure 4B,C), but ShDW2-DTI scored higher value compared to the DTIs for other traits. In brief, DTIs for the 16-day-old seedlings exposed to 10% PEG treatment were generally higher than their corresponding values in seedlings exposed to 15% PEG. The range of variation is particularly large for ShFW2, RL2, and RFW2. The DTIs for shoot and root traits of 16-day-old seedlings in all accessions are given in Table A2.    Table 2.

Variation in Leaf Length, Width, and RWC
The variation in leaf length and width and in the RWC values are illustrated in Figure 5 for 21-day-old seedlings under normal conditions and 10% PEG treatments. The calculated means for LL, LW, and RWC under the 10% PEG are significantly reduced compared to the control. This is strongly supported by the highly significant values obtained by the ANOVA analysis of data (Table 3) and the LSD values for the control vs. 10% PEG treatment given in Table 4. However, the scale indicating the lower and upper limits of variation in LL and LW boxplots of mean values is greater than the scale for the RWC ( Figure 5D). The lower and upper values for each DTI also indicate narrower variation among accessions in the RWC-DTI. It is evident from the values and the boxplots in Figure 5 that LW and RWC have higher DTIs than LL. Values of the DTIs of these three traits in all accessions are given in the Appendix A, Table A3.

Heritability of Traits in Control and PEG-Stressed Traits
The calculated heritability (H 2 ) values of G% are generally similar under control conditions and the 10% PEG treatment (Table 3). Higher concentrations of PEG drastically reduced the value of G% H 2 from 0.72 for the control to 0.58 for both the 15% and 20% PEG treatments. However, the H 2 of the AS% is low for the control (0.36) and increased as the PEG concentration increased to 0.66, 0.77, and 0.84 under the 10%, 15%, and 20% PEG treatments, respectively. The calculated H 2 values of the shoot and root traits of the 9-day-old and 16-day-old seedlings are generally similar for the control seedlings and seedlings stressed with the 10% PEG treatments. However, particularly low H 2 values are recorded for the ShDW1 (0.28) and RDW1 (0.37) in 9-day-old control seedlings and in seedlings exposed to 10% PEG for ShDW1 (0.43) and RDW1 (0.46). For the 16-day-old seedlings, H 2 values are slightly lower for all traits in seedlings exposed to 15% PEG treatments. The H 2 values of leaf traits are also given in Table 3 and are slightly higher under control conditions compared to the 10% PEG.
LSD values for the control vs. 10% PEG treatment given in Table 4. However, the scale indicating the lower and upper limits of variation in LL and LW boxplots of mean values is greater than the scale for the RWC ( Figure 5D). The lower and upper values for each DTI also indicate narrower variation among accessions in the RWC-DTI. It is evident from the values and the boxplots in Figure 5 that LW and RWC have higher DTIs than LL. Values of the DTIs of these three traits in all accessions are given in the Appendix Table A3

Heritability of Traits in Control and PEG-Stressed Traits
The calculated heritability (H 2 ) values of G% are generally similar under control conditions and the 10% PEG treatment (Table 3). Higher concentrations of PEG drastically reduced the value of G% H 2 from 0.72 for the control to 0.58 for both the 15% and 20% PEG treatments. However, the H 2 of the AS% is low for the control (0.36) and increased as the PEG concentration increased to 0.66, 0.77, and 0.84 under the 10%, 15%, and 20% PEG treatments, respectively. The calculated H 2 values of the shoot and root traits of the 9-day-old and 16-day-old seedlings are generally similar for the control seedlings and seedlings stressed with the 10% PEG treatments. However, particularly low H 2 values are recorded for the ShDW1 (0.28) and RDW1 (0.37) in 9-day-old control seedlings and in seedlings exposed to 10% PEG for ShDW1 (0.43) and RDW1 (0.46). For the 16-day-old seedlings, H 2 values are slightly lower for all traits in seedlings exposed to 15% PEG treatments. The H 2 values of leaf traits are also given in Table 3 and are slightly higher under control conditions compared to the 10% PEG.

Correlations of Traits under Control and PEG Stress
Correlations (r-value) of the studied 17 traits under control conditions and under the 10% PEG are presented in Figure 6. Under the non-stressed conditions, 14 highly significant r values ≥ 0.70 *** have been recorded, 3 for SHFW1, SHDW1 and RFW1 of the 9-day-old seedlings with each other and 11 for the shoot and root traits of the 16-day-old seedlings (ShL2, ShFW2, ShDW2, RL2, RFW2, and RDW2). Additionally, 34 highly significant positive r values ≥ 0.50 *** were scored for ShFW1, ShDW1, RFW1, and RDW1 of the 9-day-old seedlings and all shoot and root traits of the 16-day-old seedlings, except AS and RWC ( Figure 6A). The LL and LW are also significantly correlated with the six shoot and root traits of the 16 days old seedlings (RL2, ShFW2, ShDW2, RL2, RFW2, and RDW2). Low r values were scored for the control G% and RDW1, while ShFW2 and RFW2 are significantly correlated at the 0.05 significance. On the other hand, the r coefficient values are mostly negative or low and insignificant for the traits AS%, ShL1, RL1, and RWC of the control.
Positive correlations were also scored for the majority of the same 17 traits under the 10% PEG treatment, as indicated by the red and yellow cells in the correlation triangle ( Figure 6B). However, the r values are generally low compared to their corresponding values under the control condition; only 25 highly significant r values are ≥0.5 *** for six shoot and root traits of the 9-day-old seedlings. Most of the shoot and root traits of the 16-day-old seedlings, i.e., ShL2, ShFW2, ShDW2, RL2, RFW2, and RDW2, are significantly correlated with each other. The LL and LW are also mostly significantly correlated with each other but at a lower significance level. On the other hand, negative and insignificantly positive r-values were recorded between the shoot and root traits for the 9-day-old seedlings and the 16-day-old seedlings and for the G%, AS% and RWC. The RWC is relatively higher correlated with the G% and AS% under the 10% PEG treatment compared to the control conditions ( Figure 6A,B). RDW2). Additionally, 34 highly significant positive r values ≥0.50*** were scored for ShFW1, ShDW1, RFW1, and RDW1 of the 9-day-old seedlings and all shoot and root traits of the 16-day-old seedlings, except AS and RWC ( Figure 6A). The LL and LW are also significantly correlated with the six shoot and root traits of the 16 days old seedlings (RL2, ShFW2, ShDW2, RL2, RFW2, and RDW2). Low r values were scored for the control G% and RDW1, while ShFW2 and RFW2 are significantly correlated at the 0.05 significance. On the other hand, the r coefficient values are mostly negative or low and insignificant for the traits AS%, ShL1, RL1, and RWC of the control.  Table 2.
Positive correlations were also scored for the majority of the same 17 traits under the 10% PEG treatment, as indicated by the red and yellow cells in the correlation triangle ( Figure 6B). However, the r values are generally low compared to their corresponding values under the control condition; only 25 highly significant r values are ≥0.5*** for six shoot and root traits of the 9-day-old seedlings. Most of the shoot and root traits of the 16-day-old seedlings, i.e., ShL2, ShFW2, ShDW2, RL2, RFW2, and RDW2, are significantly correlated with each other. The LL and LW are also mostly significantly correlated with each other but at a lower significance level. On the other hand, negative and insignificantly positive r-values were recorded between the shoot and root traits for the 9-day-old seedlings and the 16-day-old seedlings and for the G%, AS% and RWC. The RWC is relatively higher correlated with the G% and AS% under the 10% PEG treatment compared to the control conditions ( Figure 6A,B).
Correlations of 27 traits, including the above 17 traits, and 10 other traits, including G% and AS% under 15% and 20% PEG treatments and shoot and root traits in seedling exposed to 15% PEG treatments for 16 days, are illustrated in Figure 7. In general, traits of 9-day-old seedlings are often positively correlated with each other but negatively or weakly correlated with traits of the 16-dayold seedlings. Traits of 16-day-old seedlings exposed to 15% PEG, and LL and LW are also often positively correlated with each other. Weak or negative r-values are common for the AS% and RWC. Correlation coefficients of 27 germination, seedlings and leaf traits gown under PEG stress treatments  Table 2.
Correlations of 27 traits, including the above 17 traits, and 10 other traits, including G% and AS% under 15% and 20% PEG treatments and shoot and root traits in seedling exposed to 15% PEG treatments for 16 days, are illustrated in Figure 7. In general, traits of 9-day-old seedlings are often positively correlated with each other but negatively or weakly correlated with traits of the 16-day-old seedlings. Traits of 16-day-old seedlings exposed to 15% PEG, and LL and LW are also often positively correlated with each other. Weak or negative r-values are common for the AS% and RWC. Correlation coefficients of 27 germination, seedlings and leaf traits gown under PEG stress treatments are given in Table S1. The correlation of DTIs of 27 traits of maize accessions has been measured and is illustrated in Figure S1. Most of the DTIs of germination and shoot traits of the 9-day-old seedlings are positively correlated with each other and with most shoot and root DTIs of 16-day-old seedlings. The r values and significance values for the correlation of traits DTIs are given in Tables S2 and S3.

Screening for Drought-Tolerant Traits and Accessions
To screen for the most and least tolerant traits and accessions, the frequency of the best performing accessions in the 5% top traits and the least performing accessions in the 5% bottom traits under control and stress treatments is shown in Table 5. Detailed inspection of this table shows that three accessions scored the best performance in ≥10 traits; these are Zea 1062 (16 traits (Table 5). Table S2 lists the top 10% accessions scoring best performance estimated as the maximum mean of G% and shoot, root and leaf traits, and minimum AS% under control condition and 10% PEG treatment. Table S3 lists the bottom 10% accessions scoring lowest performance estimated as the minimum mean of G% and shoot, root and leaf traits and maximum AS% under control condition and 10% PEG treatment.
Plants 2020, 9, x FOR PEER REVIEW 12 of 21 are given in table S1. The correlation of DTIs of 27 traits of maize accessions has been measured and is illustrated in Figure S1. Most of the DTIs of germination and shoot traits of the 9-day-old seedlings are positively correlated with each other and with most shoot and root DTIs of 16-day-old seedlings.
The r values and significance values for the correlation of traits DTIs are given in Tables S2 and S3.  Table 2.

Screening for drought-tolerant traits and accessions
To screen for the most and least tolerant traits and accessions, the frequency of the best performing accessions in the 5% top traits and the least performing accessions in the 5% bottom traits under control and stress treatments is shown in Table 5. Detailed inspection of this table shows that three accessions scored the best performance in ≥10 traits; these are Zea 1062 (16 traits (Table 5). Table S2 lists the top 10% accessions scoring best performance estimated as the maximum mean of G% and shoot, root and leaf traits, and minimum AS% under control condition and 10% PEG treatment. Table S3 lists the bottom 10% accessions scoring lowest performance estimated as the minimum mean of G% and shoot, root and leaf traits and maximum AS% under control condition and 10% PEG treatment.  Table 2. Table 5. Frequency of the best performing accessions in the 5% top traits and least performing accessions in the 5% bottom traits under control and stress treatments (10%, 15%, and 20% PEG).

Serial
Accession ID

Number of Best Traits Total Number of Least Traits Total
Control 10% 15% 20% Top 5% Control 10% 15% 20% Bottom 5%    (Table 1).  (Figure 8). The display of accessions in the PCA scatter diagram clearly demonstrates the resemblance of accessions having similar DTIs. However, resemblance in DTI values for accessions does not clearly reflect their origin or their assignments to subspecies and varieties as identified in the IPK collection (Table 1).

Discussion
The performance of maize germplasm for stress-tolerant traits may be best analyzed by effective screening for discriminating between drought-tolerant and drought-susceptible genotypes by easily measured and evaluated traits. The applied drought stress treatments clearly exerted a negative impact on germination and seedling performance of all maize accessions by retarding shoot and rootrelated traits. Moreover, significant reductions in seedling's traits under the 10% PEG treatment were evident for all traits, after 9 and 16 days of sowing. Another result that demonstrates the low capacity of maize to tolerate drought stress is the high proportion of abnormal seedlings under the 15% and 20% PEG treatments. The retarded emergence of radicles and plumules of seeds exposed to 15% and 20% PEG and the slow growth of seedlings under these two treatments confirm the view that maize is a drought non-tolerant cereal compared to barley [9] and wheat [10]. It is widely accepted that the first action of moisture deficit imposed by drought is impaired germination, resulting in poor plant stand at the early seedling phase and hampering early crop establishment [8,18,36]. The genetics of germination under abiotic stress is not well understood, but recent studies on the genetic variation for the studied traits by GWAS analysis identified several adaptive genes associated with G% and G%-DTI, on different chromosomes under drought, but no genes were identified for G% under control [9,37].
In maize, as in other cereals, seminal roots are responsible for the initial absorption of moisture and nutrients, but selection for an extended root system reaching larger depths is equally important

Discussion
The performance of maize germplasm for stress-tolerant traits may be best analyzed by effective screening for discriminating between drought-tolerant and drought-susceptible genotypes by easily measured and evaluated traits. The applied drought stress treatments clearly exerted a negative impact on germination and seedling performance of all maize accessions by retarding shoot and root-related traits. Moreover, significant reductions in seedling's traits under the 10% PEG treatment were evident for all traits, after 9 and 16 days of sowing. Another result that demonstrates the low capacity of maize to tolerate drought stress is the high proportion of abnormal seedlings under the 15% and 20% PEG treatments. The retarded emergence of radicles and plumules of seeds exposed to 15% and 20% PEG and the slow growth of seedlings under these two treatments confirm the view that maize is a drought non-tolerant cereal compared to barley [9] and wheat [10]. It is widely accepted that the first action of moisture deficit imposed by drought is impaired germination, resulting in poor plant stand at the early seedling phase and hampering early crop establishment [8,18,36]. The genetics of germination under abiotic stress is not well understood, but recent studies on the genetic variation for the studied traits by GWAS analysis identified several adaptive genes associated with G% and G%-DTI, on different chromosomes under drought, but no genes were identified for G% under control [9,37].
In maize, as in other cereals, seminal roots are responsible for the initial absorption of moisture and nutrients, but selection for an extended root system reaching larger depths is equally important for efficient acquisition of nutrients [18]. In addition to root characters, drought stress reduces the phenotypic expression of all the seedling traits such as shoot length and the fresh and dry weight of shoot and root [36]. Reduction in seedling growth is the result of restricted cell division and enlargement, as drought stress directly reduces growth by decreasing cell division and elongation [38,39]. Reduction in shoot length is due to less water absorption and a decrease water deficit created by external osmotic potential [36,40]. In cereals, plant growth performance was found to be positively associated with well-developed root systems, as well as early seedling traits [23,27,41,42], both of which can help to improve stress tolerance. However, significant reductions in root length and root fresh and dry mass under simulated drought occurred in most accessions.
The broad-sense heritability (H 2 ) was estimated under both control and drought conditions. The H 2 of G% was reduced, and H 2 for AS% increased as the PEG concentration increased. The H 2 for seedlings traits are generally similar and values are generally high. However, particularly low H 2 values are recorded for the ShDW1 and RDW1 in the control seedlings and in seedlings exposed to 10% PEG for the 9-day-old seedlings and for the RDW2 in control seedlings and seedlings exposed to 10% and 15% PEG treatments for the 16-day-old seedlings. The higher values of H 2 among traits indicate that selection of maize tolerant genotypes may be based on shoot length and shoot and root fresh and dry weight as well as leaf length and width. Similar heritability values in seedling traits, across nitrogen level applications, were reported in maize, whereas more variation was found in adult plants [32]. This result agrees with the estimates that heritability and genotypic correlation coefficients were significantly high for most of the seedling traits in maize [43].
One important objective of this study is to elucidate correlations of seedling traits with a view to identifying novel traits for measuring drought tolerance at seedling stages among accessions. Comparison of the correlation values under the control condition and the applied stress treatments indicated significantly lower r-values of G% and AS% with increased stress levels. However, the r-values under stress are generally lower compared to their corresponding values under the control condition, but highly significant r-values were scored for most shoot and root trait pairs under 10% PEG stress for the 9-day-old and 16-day-old seedlings. This confirms the view that the effect of stress, as an environmental variable, on the correlations of the studied traits is small [34]. Under the 15% PEG stress, the r-value for trait pairs is generally lower compared to the corresponding r-values under 10% PEG and the control. At the 15% PEG stress level, RL1 and RL2 were not correlated with other traits. In view of positive correlations of shoot and root trait pairs, it may be concluded that selection for shoot and root weight traits would be effective in identifying genotypes for better performance under moderate drought stress conditions.
In maize, significant negative correlations for seedling traits in early and extra-early maturing maize hybrids were reported [24], particularly for fresh shoot weight, shoot moisture content, root-shoot dry weight ratio, and total fresh biomass. Correlations of seedling traits were also used for selection in wheat and barley [8,9,27]. Phenotypic correlation describes the variance that two traits share based on phenotypic measurements; it includes genetic components that are the proportion of variance that pairs of traits tested share due to genetic factors and environmental correlation imposed by external conditions. The high correlations for shoot and root biomass trait pairs and leaf length and width, recorded in this study, under normal conditions and under stress, indicate that such traits are, to a large extent, genetically controlled. Thus, focusing on these traits would provide information to evaluate genetic variability for seedling traits in maize accessions to effectively screen a large number of accessions in a short period of time.
Another major objective of this study is the classification of maize accessions based on their response to drought stress. The 40 accessions were displayed in a PCA scatter diagram based on the calculated DTI values. The grouping of accessions in the PCA based on the contribution of the DTIs of the examined traits are demonstrated in a PCA biplot, which indicated that the five accessions having highest DTIs and the five accessions having the least DTIs are grouped as two distinct groups from other accessions, as in Figure 8. The most contributing DTIs are those concerned with the shoot and root traits and LL, which are often significantly positively correlated, as indicated in Figure 5. Drought tolerance in maize hybrids has been evaluated using the PCA analysis [29]. Similar results on the selection of drought-tolerant genotypes of durum wheat, based on the combination of indices by the biplot method, were reported [44], thereby this method is better than one index alone to identify superior genotypes for drought conditions. More recently, maize-inbred lines and their hybrid responses to a range of macro and micro-environmental stresses were characterized in terms of water use efficiency (WUE), grain yield, and environmental index [30]. Water use efficiency for drought-tolerant hybrids was significantly greater than for non-drought tolerant hybrids [45].
In the current study, accessions with contrasting response to induced drought stress at the seedling stage (most tolerant vs. most susceptible) can be used for additional experiments to determine how well a seedling's drought tolerance can predict the stability of yield under drought in different environments and genetic backgrounds [21,32] in order to identify accessions with potential for higher grain yield for selection of genotypes for breeding commercial lines. For this major objective, evaluation of physiological and biochemical responses are necessary, such as, chlorophyll content photosynthesis rate, chlorophyll fluorescence as well as stomatal conductance, ROS production and osmolytes accumulatiins [46]. In addition, genome-wide association mapping (GWAS) may be applied to identify QTL controlling the variation of traits associated with drought tolerance and seedling development. In this respect, Xu et al. [47] identified candidate genes for drought tolerance in 15 maize inbred lines by whole-genome resequencing. The identification of candidate genes which have roles in the biological pathways of desired traits may be confirmed by finding an association between these trits and their genes by GWAS [48].

Conclusions
Evaluation of germination and seedling root, shoot and leaf traits were performed under induced osmotic stress simulated by PEG treatments as a profound base for drought tolerance of selected accessions. All PEG treatments significantly reduced germination and retarded seedling early growth; the 15% and 20% PEG treatments resulted in a significant proportion of abnormal seedlings. Positive correlations were found between most trait pairs under control and the 10% PEG treatment, particularly shoot and root traits. Medium to high heritability of shoot and root seedling traits were calculated, providing a sound basis for further genetic analyses. The DTI values were most useful in the differentiation of traits and accessions; PCA analysis based on variation in DTIs clearly grouped the accessions with high DTIs together and the accessions with low DTIs together, indicating resemblance between accessions with similar DTIs. In brief, using seedling traits is a cost-effective approach in achieving rapid screening for tolerant or sensitive maize germplasm in a short time.
Supplementary Materials: The following are available online at http://www.mdpi.com/2223-7747/9/5/565/s1, Figure S1: Correlations of 27 traits DTIs of maize accessions, under PEG stress treatments; the correlation r values are plottedon the cells of the correlation triangle produced by GENSTAT and names of traits are abbreviated as a in Table 2. Table S1. Correlation coefficient values of 27 germination, seedlings and leaf traits gown under PEG stress treatments. Table S2. Correlation coefficients of 27 germination, seedlings, and leaf trait's DTIs for seedlings grown under PEG stress treatments. Table S3. Significance values for the correlation coefficients of 27 germination and seedlings, and leaf traits for seedling under stressed conditions.  Appendix A Table A1. Drought tolerance indices (DTIs) for germination percentage and abnormal seedlings proportion under 10%, 15%, and 20% PEG treatments, as well as the root and shoot traits of the 9-day-old seedlings for all 40 maize accessions. Names of traits DTIs are abbreviated as in Table 2