Introduction

Drought stress is one of the most important abiotic stresses that affect growth and yield of small grains especially wheat (Triticum aestivum L.). Although wheat can be successfully grown in a wide range of environments (Sallam et al. 2015a), drought stress is still a serious problem to wheat growers, breeders, and researchers (Nezhadahmadi et al. 2013). Therefore, improving drought tolerance in wheat is required. Drought stress has an enormous effect on wheat plants including changes in protein, photosynthesis efficiency, chlorophyll content, water content, and growth inhibition (Yordanov et al. 2000; Lawlor and Cornic 2002; Zhu 2002). Moreover, it affects plant height and yield traits such as grain weight, grain yield, and spike length (Rivero et al. 2007; Sallam et al. 2015a).

Drought stress can occur in any growth stage and it depends on the geographical region. Breeding for improving drought tolerance in wheat at terminal stage tends to receive more attention, due to the importance of grain yield, than breeding drought tolerance at seedling stage. However, stress at either time can be very important. Therefore, drought stress tolerance in wheat during the vegetative growth stage should be studied (Abid et al. 2016). Moreover, the vegetative stage affects the grain yield at the final stage of growth because the photosynthetic reserves accumulated until flowering provided approximately 57% of the final grain yield (Gallagher et al. 1976).

Drought tolerance in wheat has been evaluated by studying the response of physiological parameters such as changes in water content (Ritchie et al. 1990), the efficiency of transpiration (Sankar et al. 2008), and osmotic adjustment (Maathuis et al. 2003). Other studies focused on the biochemistry of drought tolerance in wheat such as changes in photochemical, antioxidative enzymes, ascorbate peroxidase, and changes in ROS accumulation (Scandalios 1993; van Rensburg and Krüger 1994; Li and van Staden 1998). Drought tolerance also can be assessed in field conditions by estimating reduction due to drought stress in grain yield and traits such as 1000-kernel weight, plant height, grain yield per spike (Sallam et al. 2014, 2015a). Recently, the use of high-throughput phenotyping technology with automated systems has been used by plant researchers due to its fast, accurate large-scale quantification, and controlling of many phenotypic traits (Andrade-Sanchez et al. 2014; Araus and Cairns 2014). This technology has been used to study drought tolerance in field, greenhouses, and growth chambers in wheat (Awlia et al. 2016; Bai et al. 2016; Barmeier and Schmidhalter 2016). However, this technology requires high costs. Drought tolerance in wheat has been studied by integrating phenomic, transcriptomic, metabolomic, and proteomic analyses which shed light on drought-responsive genes and signaling pathways (Mwadzingeni et al. 2016).

Drought experiments under controlled conditions in greenhouses or growth chambers can be very effective because environmental factors (light, humidity, irrigation, and temperature) can be controlled and the plants are exposed primarily to drought stress. Generally, the most important challenge is to define the drought stress you want to measure and then do an accurate phenotyping to identify the most promising drought-tolerant genotypes. Imprecise phenotyping for drought tolerance will reduce heritability and affect our molecular biology understanding (Reynolds et al. 2001).

Most of the earlier studies focused on targeting one or two traits to define and address drought tolerance in wheat through scoring relevant traits such as leaf wilting (Bowne et al. 2012), stay-green (Christopher et al. 2016), and leaf water content (Tahara et al. 1990). However, studying a single trait rarely identified drought tolerant genotypes because drought tolerance is a complex trait controlled by different mechanisms. Leaf wilting and stay-green indicate the ability of the plant to tolerate drought by reducing water losses and continuing photosynthesis under drought stress. In periods of severe drought, the ability of a plant to survive provides important information on its capability to recover and continue its growth when moisture is present after drought. More importantly, combining both information from scoring traits associated with tolerance and recovery will undoubtedly help in more broadly identifying drought tolerant genotypes. Finally, a selection index for drought tolerance could be a very effective tool to select the best genotypes because it can include more than one target trait (Gebre-Mariam and Larter 1996).

As mentioned earlier, drought stress can occur at any growth stage. Therefore, the stage at which genotypes are screened for drought tolerance should be carefully considered. For instance, improving a survival trait under drought condition at seedling stage may be appropriate if the drought stress occurs during seedling growth and development, but inappropriate if drought stress only occurs around heading or grain filling stage. Another possible reason for failure is that the breeding for drought tolerance is often affected by other factors in dry environments (e.g. erratic weather patterns, soil-borne diseases, soil mineral nutrition. etc.). Therefore, running drought experiments in controlled conditions in greenhouses or growth chambers to study the genetic variation in drought tolerance is desirable to augment working in the field where many conditions are not controlled. Therefore, it is important to test the same material under controlled and field experiments to select the promising genotypes for target traits (Sallam et al. 2016). In Nebraska, a low rainfall has been noted in last 10 years during planting periods, September to March, which coincides with the seedling stage of winter wheat.

The objectives of this study were (1) to study the genetic variation in tolerance and survival traits associated with drought tolerance in bi-parental population (2) to select the most promising drought-tolerant RILs with a combination of high yielding to be utilized in further breeding program to improve drought tolerance in Nebraska winter wheat at seedling stage.

Materials and methods

Plant material

The plant materials consisted of 146 recombinant inbred lines (RILs, F9) derived by crossing ‘Harry’ with ‘Wesley’. The 146 genotypes were randomly chosen from 204 RILs of the biparental population. The production scheme and pedigree information on this biparental population was extensively described in Hussain et al. (2018). Two checks were used: ‘Anton’ a drought tolerant cultivar developed by the USDA-ARS and ‘Gimmiza-9’, a drought susceptible cultivar from Egypt. In addition, a set of 10 accessions from Egypt and five accessions from USA were used as a verification set. The ten Egyptian accessions are spring wheats widely sown in Egypt due to their high agronomic performance; American accessions were winter wheats: ‘Anton’ (drought tolerant), ‘Harry’ (drought tolerant), ‘Wesley’ (drought susceptible), ‘Goodstreak’ (drought susceptible), and ‘Settler-CL’ (moderate drought tolerance). These accessions were used to validate findings from the RILs population. Pedigrees of these accessions are listed in supplementary Table 1. Anton, Harry, Wesley, and the RIL population have been deposited in the USDA National Small Grains Collection, Aberdeen ID and are available for distribution from the USDA National Small Grains Collection.

Experimental layout and drought treatment

Eight seeds from each genotype were sown in pots in two rows of four seeds each. The soil in pots (13 × 13 × 13 cm) was filled with a 2000 g of compost soil and sand (1:1, respectively). The experimental layout was a randomized complete block design with three replications. The field capacity was estimated according to Grewal et al. (1990). Pots were irrigated to keep them at 50% soil water capacity. The experiments were conducted in a controlled greenhouse at the Department of Agronomy & Horticulture, University of Nebraska-Lincoln. The photoperiod’s regime was prepared as 16 h/8 h light–dark at 25 °C/20 °C day/night temperatures and 65–70% air humidity.

Drought treatment was applied when the majority of the seedling plants reached three fully expanded leaves (Zadoks et al. 1974). The pots were not watered for 21 days. The end of drought treatment was determined when the leaves of 50% of all plants had turned from green to yellow (after 20 days). On the last day of drought treatment, half of the plants (four plants in one row) from each genotype were cut at the first internode. The other four plants (in the other row) from each genotype were not cut. Subsequently, each genotype was regularly watered for 15 days (100% soil water capacity). The drought regime used in this study is illustrated in Fig. 1.

Fig. 1
figure 1

Stages of drought regime applied on the seedling of HW population. Red abbreviation refer to the names of traits scored in this study. PH plant height, LW leaf wilting, DTW days to wilting, SG stay green, RAI recovery after drought, RB regrowth biomass, DTR days to recovery. (Color figure online)

Full size image

Before drought treatment, plant height (cm) was measured (one measurement/pot). Two types of traits were scored namely: drought tolerance traits which were scored during the drought treatment until the end of drought treatment and recovery traits which were scored after drought treatment ended for the next 15 days. The drought tolerance traits included:

  1. 1.

    Days to wilting (DTW). DTW was visually scored when 50% of plants/genotype started to wilt.

  2. 2.

    Leaf wilting (LW). LW was visually scored using scale ranged from 1 (no wilting) to 9 (fully wilted). This trait was scored every 2 days after the plants started to wilt. As a total, LW was scored 6 times.

  3. 3.

    Sum of leaf wilting (S_LW). All scores of LW were summed up to form one trait reflecting the symptoms of drought stress. This trait ranged from 6 (no wilting) to 54 (fully wilted).

  4. 4.

    Stay-green (SG). SG was visually scored at the end of drought treatment using a scale ranging from 1 (yellow leaves) to 9 (green leaves).

    The recovery traits included

  5. 5.

    Days to regrowth (DTR). DTR was scored on each cut plant (N = 4) within a pot as a number of days from cutting plant to when the first leaf regrowth was visible. Then the number of days for each genotype was transformed to the disposition to survive [°; from 0° to 90°] as described by Roth and Link (2010) and Sallam et al. (2015b)with some modifications:

    $${\text{DTR}} = \arctan {\text{ x}}_{\text{i}} /\upmu_{\text{x}}$$

    where xi refers to a number of days for each plant from cutting to first leaf regrowth and µx = mean number of days of those surviving plant produced leaves after drought. Dead plants were scored as 90°. All DTR values for each genotype were averaged.

  6. 6.

    Regrowth biomass (RB). At the end of the experiment, the leaves and stems of the all surviving, which were cut after drought and before re-irrigation, and hence re-growing plants for each genotype were cut again to measure this regrowth (g) to observe the difference in the vigor of the surviving plants. All RB values were multiplied by 10 due to little regrowth of the most genotypes.

  7. 7.

    Survival rate (SR) was determined by dividing the number of surviving plants by the number of cut plants.

  8. 8.

    Recovery after irrigation (RAI) was scored on those plants that were not cut. RAI ranged from 1 (no recovery = dead) to 9 (fully recovered = the plant could replace the yellow leaves to green leaves).

Selection indices for drought tolerance

Three optimum selection indices (Falconer and Mackay 1996) were calculated to work towards better define drought tolerance (Fig. 2). Tolerance index (TI) was used to better describe DTW (X1) using one auxiliary trait S_LW(X2) as:

$${\text{TI}} = {\text{b}}_{1} {\text{X}}_{1} + {\text{b}}_{2} {\text{X}}_{2} ,{\text{ where b}}_{1} = 0.6029{\text{ and b}}_{2} = - 0.071.$$
Fig. 2
figure 2

Creation of selection indices for drought tolerance

Full size image

Recovery index (RI) was used to augment RB (X1) using two auxiliary traits: SR (X2) and DTR (X3) as

$${\text{RI}} = {\text{b}}_{1} {\text{X}}_{1} + {\text{b}}_{2} {\text{X}}_{2} + {\text{b}}_{3} {\text{X}}_{3} ,{\text{where}}\,{\text{b}}_{1} = 194.0257,{\text{b}}_{2} = 0.2578\,{\text{and}}\,{\text{b}}_{3} = 0.9957.$$

where b1, b2, and b3 are the index coefficients. The vector of Smith–Hazel index coefficient b was calculated as described in (Baker 1986)

b = P − 1 G, where P − 1 is the inverse of the phenotypic variance–covariance matrix for the traits; G is a matrix including the estimates of genotypic and covariance.

Drought tolerant index (DTI) was calculated from TI and RI as follow:

$${\text{DTI}} = \raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} \, \left[ {\left( {{\text{TI}}/{\text{SD}}_{\text{TI}} } \right) + \left( {{\text{RI}}/{\text{SD}}_{\text{ST}} } \right)} \right]$$

where SDTI and SDRI are the phenotypic standard deviation of the TI and RI, respectively.

Field experiments

A set of 204 RILs from the biparental were sown in two low rain-fall environments in Nebraska; Grant and Sidney (Peterson 1992; Lyon and Hergert 2012) for two successive seasons 2015/2016 and 2016/2017. Moreover, the same genotypes were sown in a well rainfall environment; Lincoln, Nebraska to test and compare grain yield between low and well rainfall environments. The experimental layout was augmented incomplete randomized block design with one replication in each environment as described in Hussain et al. (2017). All genotypes along with three different checks (Wesley, Harry and ‘Overland’) were tested in 12 incomplete blocks where each block included 17 RILs (un-replicated) and three checks which were replicated in each block. The seeds of genotypes were sown in a four-row plot of 1.2 m width and 3 m length. Grain yield (g) for each genotype was measured for genotype. The grain yield (g) scores of the 146 genotypes, which were evaluated in seedling experiments, were used to identify the high yielding genotypes under dry environments. In each dry environment, all genotypes were sorted from high to low yielding.

Statistical analyses

This analysis of variance was performed with PLABSTAT software (Utz 1991) and R software using the following equation:

$${\text{Y}}_{\text{ij}} =\upmu + {\text{g}}_{\text{i}} + {\text{r}}_{\text{j}} + {\text{error}}$$

where Yij is an observation of genotype i in replication j, μ is the general mean; gi, rj are the main effects of genotypes and replications, respectively; the error is genotype × replication interaction of genotype i with replication j. Genotypes were considered as fixed effects, while, replications were considered random effects. Broad-sense heritability of the genotypes for each trait was calculated as (H2 = genotypic variance/phenotypic variance). The Spearman rank correlation coefficients were calculated for the phenotypic correlation between traits. The variance–covariance analysis was carried out using GENOT-command with PLABSTAT software to estimate the genetic correlation coefficient and to allow the building of the optimum selection indices.

Analysis of variance (ANOVA) for grain yield was conducted using ASReml R package (Butler et al. 2009). The analysis of variance (ANOVA) model was:

$${\text{Y}} = {\text{E}} + {\text{G}} + {\text{Y}}\left( {\text{E}} \right) + {\text{Pcol }}\left( {\text{Iblock}} \right) + {\text{Prow}} + {\text{E}} \times {\text{Y}} + {\text{G}} \times {\text{E}} + {\text{Error}}$$

where Y is the observation of genotype, E is environment, Y(E) is year within environment, Pcol (Iblock) is the number of columns nested within Iblock, Prow is the row number, G is genotypes and G × E is genotype × environment interaction, E × Y environment × year interaction, and Error is G × E × Y genotype × environment × year interaction

Results

Genetic variation in drought tolerance at seedling stage

The analysis of variance revealed highly significant differences (p < 0.01) among genotypes for all traits (Table 1). All traits showed high heritability estimates (H2). For tolerance traits, the heritability estimates ranged from 0.53 (SG) to 0.82 (S_LW), while, it ranged from 0.74 (RB) to 0.88 (DTR) in the recovery traits. Drought tolerance index (DTI) showed the highest heritability among selection indices with H2 = 0.84. The susceptible check Gimmiza-9 and parent Wesley (Fig. 3) did not survive after the 21 days drought in any replication, while, the tolerant check Anton and parent Harry (Fig. 3) showed regrowth and survived after drought treatment in all replications.

Table 1 Minimum, maximum, mean, least significant differences (LSD), F-value (between genotypes), and heritability estimates for all traits scored in RILs population
Full size table
Fig. 3
figure 3

Phenotypic variation in REG and RAI between Wesley (a, b) Harry after drought treatment and irrigation

Full size image

The phenotypic variation among genotypes for TI, RI, and DTI is presented in Fig. 4a–c. Wesley was more susceptible than the susceptible check Gimmiza-9 for all indices. The American cultivar Anton showed a higher level of drought tolerance than the drought tolerant parent Harry. For each index, many RILs exceeded the parental mean. Only a few RILs showed higher susceptibility to drought than Wesley with four, five, and seven RILs for DTI, TI, and RI (Fig. 4a), respectively. Only two RILs were susceptible to drought in all the three indices. On the other hand, more RILs had higher DTI (19 RILs), TI (24 RILs), and RI (37 RILs) than the drought tolerant parent Harry (Fig. 4b). Four RILs in all the three indices had higher values than Harry.

Fig. 4
figure 4

Phenotypic variation between genotypes for a TI, b RI, and c DTI. A refers to Anton (tolerant check), G refers to Gimmiza/9 (susceptible check), W refers to Wesley (susceptible parent), and H refers to Harry (tolerant parent)

Full size image

In order to identify the most tolerant RILs having a high level of drought tolerance at the seedling stage, the best 20 RILs for each trait scored in this study were identified by sorting them according to their drought tolerance (high to low). Then, a genotype was selected if it was among the best 20 genotypes in at least two traits. A total of 47 RILs were selected (Supplementary Table 2). Two RILs, HW_031 and HW_121, were among the best 20 RILs for eight traits including both tolerance and recovery traits, while, ten RILs were among the best 20 RILs for only two traits. The tolerant check Anton was among the best 20 RILs in six traits and Harry in only two traits.

For the v-set, a high genetic variation was found between all genotypes for DTW, SG, and FLW (Supplementary Table 3). Broad-sense heritability estimates ranged from 0.49 for FLW to 0.86 for S_LW.

Phenotypic and genotypic correlation among drought-tolerant traits

The phenotypic and genotypic correlation among all traits is presented in Table 2. Plant height (Ph) was positively correlated with FLW and S_LW and negatively correlated with RAI. In terms of genotypic correlation, Ph had the most significant correlations with drought-tolerant traits. It was significantly and positively correlated a with FLW, S_LW, and DTR, and negatively and significantly correlated with DTW, SG, RAI, and SR.

Table 2 Spearman rank (normal font) and genotypic (bold font) correlations among all traits scored in this study on RILs population
Full size table

For tolerance traits, DTW had significant phenotypic (rp) and genotypic (rg) correlations with FLW (rphenotypic(p) = − 0.35**; p < 0.01, rgenotypic(g) = 0.43++) and S_LW (rp = − 0.64**; p < 0.01, rg = − 0.82++). Final leaf wilting scorewas positively correlated with S_LWS (rp = 0.70**; p < 0.01). Stay-green (SG) showed low correlation with leaf wilting and DTW.

Among the recovery traits, SR was negatively correlated with DTR (rp = − 0.94**; p < 0.01, rg = − 0.98++). Regrowth biomass was positively correlated with SR (rp = 0.90**; p < 0.01, rg = 0.77++) and a negatively correlated with DTR (r = − 0.86**; p < 0.01, rg = − 0.83++). Recovery after irrigation was positively and significantly correlated with RB (rp = 0.28**; p < 0.01, rg = 0.30++) and SR (rp = 0.23**; p < 0.01, rg = 0.37++), while, it was negatively and significantly correlated with DTR(rp = − 0.21**; p < 0.01, rg = − 0.37++).

By looking at the phenotypic and genotypic correlation between tolerance and recovery traits, it was observed that no or very low correlations were found between the two groups of traits. In terms of phenotypic correlation, FLW and S_LW showed a positive significant correlation with DTR and a negative significant correlation with RAI and RB. Stay green showed a negative significant correlation with DTR (rp = − 0.19*) and a positive significant correlation with RB (rp = 0.20*).

Three selection indices were calculated to better describe drought tolerance in wheat. TI (including DTW and S_LW) had a high phenotypic and genotypic correlation with all tolerance traits and no significant correlation with recovery traits. Likewise, RI (including RB, DTR, and SR) had a high phenotypic and genotypic correlation with all recovery traits and no significant correlation with tolerance traits. No significant correlation was found between RI and TI (rp = 0.03, rg = 0.03). Interestingly, DTI (including RI and TI) had positive significant phenotypic and genotypic correlations with DTW, SG, RAI, RB, and SR. It had negative significant phenotypic and genotypic correlations with FLW, S_LW, and DTR. A highly significant correlation was found between DTI and TI (rp = 0.64**; p < 0.01, rg = 0.62) and RI (rp = 0.75**; p < 0.01, rg = 0.76).

The phenotypic and genotypic correlations between recovery and tolerance traits in v-set are presented in supplementary Table 4. DTW had significant negative phenotypic and genotypic correlations with FLW (rp = − 0.61*, rg = − 1.21++) and S_LW (rp = − 0.91**; p < 0.01, rg = − 1.02++). No significant correlation was found between DTW and recovery traits (RB, SR, and RAI). Final leaf wilting (FLW) had positive phenotypic and genotypic correlations with S_LW and no significant correlation with recovery traits. S_LW was negatively and significantly correlated with RAI (rp = − 0.53*, rg = − 0.65). Stay green had only positive phenotypic and genotypic correlations with RAI. Among recovery traits, positive phenotypic and genotypic correlations were found between SR and RAI.

Genetic variation in grain yield under low rain-fall environments

High genetic variation was found among all genotypes for grain yield in the two dry environments over the two seasons. The genotype-environment interaction (G × E) was highly significant (Table 3). The genotype × environment × years interaction was also highly significant. The distribution of all genotypes for grain yield in the two environments over the two seasons is illustrated in Fig. 5a. The average of grain yield in Lincoln was higher than the average in Grant and Sidney in both growing seasons. The grain yield as an average in Grant was higher than the average in Sidney in both seasons. Moreover, season 2016/2017 had a higher average in grain yield for the three environments competed to the grain yield in season 2015/2016. The distribution of grain yield for all genotypes in the three environments is illustrated in Fig. 5b, c.

Table 3 Analysis of variance for grain yield sown in Lincoln, Sidney, and Grant, NE for the 2015/2016 and 2016/2017 seasons
Full size table
Fig. 5
figure 5

a Average of grain yield of all genotypes in Lincoln, Grant, and Sidney locations over the two growing seasons. b, c Refer to the distribution of grain yield for all genotypes in the three locations in 2015/2016 and 2016/2017 growing seasons, respectively

Full size image

The 45 highest-yielding genotypes in each low-rainfall environment were selected. The selection was based on the number of genotypes (Harry and Anton were excluded) which were identified as drought tolerant in seedling experiments (n = 45 genotypes). Then, the common genotypes in the two environments over the two seasons and seedling experiments were identified (Fig. 6, Supplementary Table 5). As a result, five genotypes were found to be drought tolerant (seedling experiments) and had the highest yield among the 45 selected genotypes in three low rainfall environments. Out of the five genotypes, HW_121was determined as one of the most drought tolerant genotypes (n = 2) at the seedling stage with a combination of high yield in Grant and Sidney (season 2015/2016). Moreover, three genotypes were identified as drought tolerant at the seedling stage with high yield at only Grant over the two seasons. Among these three genotypes, HW_031, which was determined as the most drought tolerant genotypes (in eight traits), had a high grain yield in Grant environment over the two seasons. On the other hand, five genotypes were found to be drought tolerant genotypes at seedling stage with high yield in only Sidney over the two seasons. Five genotypes had drought tolerance at the seedling stage and high yield only in Grant and Sidney 2015/2016. The correlation analysis revealed a high and significant correlation (data not shown) for grain yield in Sidney between the two seasons (r = 0.35**; p < 0.01). No significant correlation for grain yield was found between the two seasons in Grant or between Grant and Sidney.

Fig. 6
figure 6

Number of genotypes with high drought tolerance, at seedling stage, and high yielding in low rainfall environments (N = 4). HW_121 and HW_013 were marked as the highest drought tolerant genotypes at seedling stage

Full size image

Discussion

Genetic variation in drought tolerance

The analysis of variance revealed highly significant (p < 0.01) differences among genotypes in the RILs and the v-set for all traits scored under drought stress. All traits are considered morphological traits that can be rapidly and reliably measured for screening large numbers of genotypes at the seedling stage. The range of trait heritability for the RILs in recovery traits (0.74–0.88) was higher than those reported in tolerance traits (0.53–0.82). Similar heritability estimates (0.57–0.86) were found for most of the traits in v-set. For example, the H2 of DTW was 0.71 in RILs and 0.72 in the v-set. S_LW showed a heritability of 0.82 in RILs and 0.86 in v-set. Regrowth biomass and SR showed lower heritability estimates in the v-set (0.53 and 0.57, respectively) than in RILs (0.74 and 0.84, respectively). High heritability estimates suggest that selections for drought tolerant lines could be effective for improving seedling drought tolerance in winter wheat. Selection for specific traits in more controlled environments is believed to be more useful than using field assays to improve drought tolerance in wheat (Reynolds et al. 2001). The two parents had high differences in their drought tolerance. Harry was considered drought tolerant, while, Wesley was susceptible to drought (Fig. 3).

In the current study, we scored two different types of traits associated with drought tolerance. Tolerance traits included DTW, FLW, S_LW, and SG. Leaf wilting and stay-green have been studied before for assessing drought tolerance in wheat (Izanloo et al. 2008; Bowne et al. 2012; Muir and Thomas-Huebner 2015; Drira et al. 2016; Mart et al. 2016; Rebetzke et al. 2016). Both traits provide different information on drought tolerance mechanisms. Leaf wilting indicates the deficiency of moisture in the soil and a subsequent water uptake and transport to shoots (Sanad et al. 2016). Plants having the stay-green phenotype are likely to preserve green leaf area for a long period, thus continuing maintenance of photosynthesis (Thomas and Howarth 2000). The relationship between drought tolerance and DTW was extensively published in many crops such as Solanum lycopersicum (Neuman et al. 2014; Muir and Thomas-Huebner 2015) and Ipomoea batatas (Laurie et al. 2013; Omotobora 2013; Omotobora et al. 2014). In these experiments, DTW provided an important information on how many days for each genotype to response to the water deficit. However, no study considered this trait for drought tolerance in wheat. On average, Harry started to wilt after 7 days, while Wesley wilted after 5 days after water withholding. For checks, the tolerant check Anton started to wilt after 11 days, while, the susceptible check Gimmiza-9 started to wilt after 6 days. In the RILs population, HW_250 started to wilt after 17 days, while, HW_240 started to wilt after 4 days from water withholding.

The second type of traits scored were recovery traits. One of the basic definitions of drought tolerance is the ability of the plant to recover and continue to grow after a prolonged drought. Recovery traits included DTR, RB, SR, and RAI. These traits were scored after re-watering all genotypes, which were exposed to drought stress for 21 days, then given 15 days to recover with ample watering. We stopped the experiment after 15 days because no new seedlings recovered or regrew. Survival rate and RAI have been studied in wheat previously (Izanloo et al. 2008; Timmusk et al. 2014; Salgado et al. 2015). Regrowth after drought in wheat has been scored at seedling stage. Some of previous studies such as Pearce (1985) scored regrowth after drought by scoring measurements of leaf height before re-watering and after watering within a few hours and on subsequent days. In our study, we tested regrowth in a different way by cutting the plants at the first internode before watering which challenged the plant to reproduce new parts of stem and leaves after drought treatment. The advantages of scoring this trait that is not only for noting dead versus survival for each plant after drought but also for recording differences in vigor and health status of surviving plants. Any regrowth, of course, considers the recovery of the individual in question. Days to regrowth after the drought was introduced in this study in wheat for the first time. This trait reflects the plant response to re-watering after drought and the ability to produce new shoots. Some genotypes responded quickly to re-watering and started to produce shoots after only 2 days, but in subsequent days the plants of these genotypes died. This trait was first introduced as a “disposition to survive” after frost stress in winter faba bean (Vicia faba L.) by Roth and Link (2010). Breeding for improving recovery during drought stress at seedling stage is only effective and useful if drought occurs frequently around this stage (Reynolds et al. 2001).

As discussed above, each group of traits (tolerance and recovery) described drought tolerance in wheat in a different way with different information on the mechanism that the plant uses to tolerate drought. Based on the phenotypic correlation, genotypic correlation, and heritability estimates, a selection index for each group was calculated to produce RI and TI (Fig. 2). Instead of single-trait selection, a selection index provides an effective tool for improving a group of traits in a breeding program. Then, both indices were combined to form a drought tolerance index (DTI) to finally describe drought tolerance in wheat. A transgressive segregation (TS) was observed in the three indices (Fig. 4). This result indicates that alleles with positive or negative effects on drought tolerance were contributed by both parents. Transgressive segregation for positive effects is very useful and should be exploited to improve drought tolerance in wheat. By looking at the rainfall data in the last 16 years (from 2000 to 2016) in Lincoln, Nebraska, the monthly average of the rainfall precipitation from November to March, which coincides with the seedling stage of winter wheat, was 2.62 cm on average, while, it was 3.30 cm from April to August (http://snr.unl.edu/lincolnweather/data/monthly-precipitation.asp). In Sidney (western Nebraska), the total precipitation (cm) in the same period (2000–2016) ranged from 0.00 (2010) to 0.93 (2015) cm. Therefore, breeding for recovery traits should be taken into account to improve drought tolerance in winter wheat according to Nebraska climate conditions.

The superior genotypes (those ranked first) differed for each trait: HW_174 (DTR), HW_31 (SR), HW_175 (RAI), HW_093 (S_LWS), and HW_026 (SG). Only one genotype was ranked first for two traits: HW_250 (DTW and TI) and one was ranked first for three traits: HW_238 (RB, SI, and DTI). Although each selection index was created from a group of traits, it is highly recommended to also look on the superior genotypes for each trait. In order to select the most promising genotypes for drought tolerance, the best 20 drought-tolerant genotypes were sorted based on their drought tolerance for each trait and the genotypes which were among the best 20 genotypes for more than one trait was selected (Supplementary Table 2, Supplementary Figure 1). Two genotypes HW_031 and HW_121 were among the best 20 genotypes in eight traits. HW_031 started to regrow after very few days from cutting and watering (DTR = 39.19°), produced a high shoots weight (RB = 1.20 g), a high recovery rate (SR = 100%), fewer symptoms of wilting (S_LW = 23.67), started to wilt after many days from water withholding (DTW = 10), and showed a high value of TI, SI, and DTI. In addition, HW_121 was among the best genotypes in DTR, RB, SR, RAI, S_LW, SG, SI, and DTI. These two genotypes are thought to have high levels of both tolerance and recovery. Therefore, they can be intergraded into breeding programs to improve drought tolerance in Nebraska winter wheat. Generally, the genotypes having a high recovery rate after drought (recovery traits) are different from those showing a high tolerance to drought (tolerance traits), therefore, combining information from both types of traits should help identify the most promising genotypes for drought tolerance.

Phenotypic and genotypic correlations among drought-tolerant traits

The phenotypic and genotypic correlations among drought-tolerant traits in this study shed light on the different mechanisms involved in drought tolerance in wheat. The phenotypic and genotypic correlations were calculated among all traits scored in the RILs population. To verify these correlations, the traits were also scored on v-set (diverse genotypes:10 spring and five winter wheat).

Plant height (Ph) is the only trait which scored before water withholding. The significant correlation between Ph and FLW, S_LW, and RAI indicate that the short-height genotypes seemed to be drought tolerant. However, the size of correlations was small and direct selection for shorter plants may not be useful to improve drought tolerance. The correlation analysis was divided into three sections: the correlation between tolerance traits, the correlation among recovery traits, and the correlation between recovery and tolerance traits.

Correlation among tolerance traits

Days to wilting (DTW) had higher negative phenotypic and genotypic correlations with S_LW than with FLW. S_LW is an accumulative trait that reflected all drought symptoms of each genotype from water withholding until the end of drought treatment. Hence, this trait is probably more informative than FLW which was scored only at the end of the drought experiment. We noticed that some genotypes reached full wilting after 14 days, while, others reached full wilting at the end of drought treatment. For example, Wesley and HW_212 were scored nine for FLW at the end of drought treatment. However, HW_212 reached to a full wilting after 17 days, while, Wesley became fully wilted only after 20 days (Supplementary Figure 2). Although both scored as susceptible at the end of drought treatment, Wesley was less affected by drought stress than HW_212 based on S_LW which was scored as 33.33 and 35.66 for Wesley and HW_212, respectively. This difference could be a possible reason for the high negative significant correlation between S_LW and DTW. Days to wilting, FLW, and S_LW describe approximately the same mechanism to resist drought. Broadly speaking, genotypes having a low leaf wilting score might have better access to soil water through their root system or better osmotic adjustment (Blum 2011). Therefore, a genotype exhibiting a delayed leaf wilting (low wilting score and many days to wilt) is the preferred phenotype under drought stress. The low or no significant correlation between leaf wilting traits (DTW, FLW, and S_LW) and SG is illustrated in Fig. 7a–c. At the end of drought treatment, some genotypes were fullly wilted, but their leaves were almost green (Fig. 7a), while, other genotypes exhibited few symptoms of leaf wilting, but had yellow leaves. Leaf wilting is a result of plant dehydration and water loss, while, loss of SG is a consequence of chlorophyll breakdown followed by leaf desiccation (Blum 2011). As plant water deficit increases, wilting is followed by leaf desiccation and loss of chlorophyll. Therefore, the low correlation between leaf wilting traits indicates that these traits are probably controlled by independent mechanisms involving different genes for drought tolerance. The lack of correlation between both traits was found also in the v-set. Harb et al. (2010) studied gene expression for genes controlling photosynthesis under drought and they reported that many photosynthesis genes were significantly repressed under wilting drought in Arabidopsis.

Fig. 7
figure 7

Different response of genotypes to drought stress before irrigation (a–c). a A genotype showing a full leaf wilting and green leaves, b a genotype showing a small leaf wilting and yellow leaves, c a genotype showing a little leaf wilting and green leaves

Full size image

Correlation among recovery traits

The main three recovery traits RB, DTR, and SR were highly correlated. However, there are some considerations that should be taken into account in improving drought recovery based on these traits. Days to regrowth and RB described the ability of plants/genotype to produce new shoots after drought treatment and after cutting them. Some genotypes started to regrow after 3 or 5 days from re-watering, but after a while, the plants of these genotypes died, and they were not able to survive until the end of the experiment. The possible reason could be that the roots of those plants supported the regrowth temporarily after cutting with the stored carbohydrates (root reserves) produced by photosynthesis (Mueller and Teuber 2007).

Correlation between recovery and tolerance traits

No or very low significant correlations were found between recovery and tolerance traits. Even after estimating the selection index for each trait, there was no promising correlation between TI and recovery traits or between RI and tolerance traits. The non-significant correlation between SI and TI offers further support that both types of traits are controlled by different mechanisms. In the tolerance mechanisms, which was studied through DTW, FLW, S_LW, and SG, the main consequence of drought on these traits is plant dehydration by which the plant starts to lose its water content from the leaves due to the inability of roots to absorb the water and the essential nutrients from the soil. According to the recovery mechanism, which was studied through RB, SR, DTR, and DAI, the results of drought on these traits is death versus survive/regrowth. The plant tries to survive after re-watering using its ability to produce new shoots using the functionality of its leaves and roots to support the plants with the stored and new sugars and carbohydrates to recover and regrow after a severe drought. Therefore, it seems that both types of traits are controlled by different genes. Being under the control of different genes (independent control) is useful because it means that drought tolerance and recovery can be genetically combined.

The two indices (TI and RI) were combined to calculate DTI which was found to be significantly associated with both groups of traits. Drought tolerance in terms of both tolerance and recovery traits can be improved by selection using the DTI value. The advantage of DTI was that it facilitated selecting the best genotypes showing a combination of high drought tolerance and recovery. As the correlation between traits is an important analysis, looking at the genotypes which had lower correlation values between TI and RI could be also advantageous. As discussed, the selected drought tolerant genotypes may logically have high drought tolerance in a combination with a high recovery after drought. In the current study, two genotypes were among the best drought tolerant genotypes for eight traits from both trait groups, hence should be used in breeding. This approach was used before in wheat to improve kernel weight in combination with high protein content and high yield where a negative correlation was found among them (Gebre-Mariam and Larter 1996).

Genetic variation in grain yield in low rain-fall environments

The grain yield was tested and scored for all genotypes in two low rainfall environments (Grant and Sidney, NE) over two seasons to identify highest grain yield of the most drought-tolerant genotypes, which were tested at seedling stage. The same set of genotypes was tested in Lincoln (high rainfall environment) for two seasons. Compared to Lincoln, there was a reduction of 9.2 and 4.4% in grain yield due to low rainfall in Grant for 2015/2016 and 2016/2017, respectively. In Sidney, the reduction in grain yield was more severe than in Grant with 21 and 43.4% in the two seasons, respectively. This reduction in grain yield confirms the occurrence of drought in these environments compared to the irrigated environment (Lincoln). The low and/or weak correlations for grain yield between the two seasons for the same environment and between the two environments may be due to the highly significant genotype × environment × year interaction. The genotypes having high drought tolerance at seedling stage in a combination with high grain yield in low rainfall environments were identified. The most drought tolerant genotypes identified at seedling stage (HW_121 and HW_031) were found to be among the highest grain yield in Grant over the two seasons and Sidney (for only HW_121 in 2015/2016 season). Notably, some drought tolerant genotypes, at the seedling stage, were found to be among the highest grain yield genotypes in Grant (three genotypes) or Sidney (five genotypes) over the two seasons. Therefore, a breeding program to improve drought tolerance in winter Nebraska wheat may be differentiate lines based on the environment due to the strong interaction of genotype × environment × year. For example, HW_121 and HW_031 can be used to improve drought tolerance suitable for Grant, whereas, the breeding program including crosses to any of HW_249, HW_179, HW_245, HW_113, and HW_206 could be useful to improve drought tolerance for Sidney. However, the selected genotypes (Supplementary Table 5) should be tested for further years due to the high genotype × environment × year interaction. The genotypes which had high grain yield in one environment in 1 year are ineffective for a breeding program. No significant correlations were found among seedling traits and grain yield (in any environment. data not shown). Therefore, breeding for improving drought tolerance in a combination with high grain yield should be carefully studied in both types of experiments (controlled and field conditions) in order to identify the most promising genotypes for target traits.

Conclusions

The protocol used in this study provided useful methods for screening wheat genotypes (in our case RILs) for drought tolerance at the seedling stage. It is highly recommended to consider both types of traits (drought tolerance and recovery) to identify the most promising genotypes that can tolerate and survive drought because both traits shed light on different drought tolerance mechanisms. Selection indices were a very effective way to combine different measurements representing different mechanisms into one selection index. A set of highly recommended drought tolerant and high yielding genotypes were identified. These genotypes can be integrated for breeding program (based on environment) to improve drought tolerance in Nebraska winter wheat after testing for future years. The high genetic variation found in drought tolerance in the RILs population could be very useful for QTL mapping studies to detect genomic regions associated with tolerance traits, recovery traits, and grain yield in low rainfall environments e.g. (Hussain et al. 2017).