Bread wheat genotype designated as LM43, was selected from a panel of germplasm obtained from CIMMYT. The genotype was selected after prior evaluation for its drought tolerance and yield potential (Mwadzingeni et al., 2016). A preliminary study to establish optimal conditions for effective mutagenesis with minimum biological damage was conducted prior to embarking on a large-scale mutagenesis (OlaOlorun et al., 2019).

The selection procedure across generations is illustrated in Figure 1. Preliminary phenotypic variation analyses showed that EMS mutagenesis was effective on genotype LM43 (OlaOlorun et al., 2019). Hence this genotype was selected for large-scale mutagenesis under three EMS treatment conditions. Breeding populations were developed under four generations based on the three EMS treatment conditions (OlaOlorun et al., 2021). Fresh EMS treated M₁ seeds were planted in the field between March and August 2018. The first breeding population (Population 1) was developed after the treatment of seeds at 0.1% v/v EMS for 1 h at 25°C. The second breeding population (Population 2) was derived after seeds were treated under 0.1% v/v EMS for 1 h at 30°C while the third breeding population (Population 3) involved seeds exposed to 0.7% v/v EMS for 1.5 h at 25°C. In addition, an untreated seed of the genotype LM43 was included as Population 4 and as a comparative control. M₁ plants were grown to maturity and the grains were harvested and bulked according to their respective treatments and developed into populations. The M₂ seed harvested from M₁ plants were grown out as M₂ plants. During the M₂ generation, 180 individual plants were purposefully selected based on high biomass and yield potential and further evaluated at M₃ and M₄ generations. Selections made in the M₃ and M₄ generations were for improved agronomic performance, drought tolerance and biomass allocation under drought-stressed and non-stressed conditions.

The M₁ and M₂ generations were evaluated at Ukulinga Research Farm of the University of KwaZulu-Natal (UKZN) (29⁰ 40'S, 30⁰ 24'E; 806 m above sea level) during the 2018/2019 cropping season. The M₃ and M₄ generations were established both at Ukulinga Research Farm and under greenhouse condition at the Controlled Environment Facility (CEF) at UKZN during the 2019/2020 cropping season. The meteorological data during the growing period and soil physiochemical properties at both sites are provided in Tables 1, 2, respectively. The M₁ and M₂ generations were planted under normal growing conditions with irrigation up to maturity, while the M₃ and M₄ were screened under drought-stressed and non-stressed conditions. Under field conditions, seeds were planted on 12 m long rows. The spacing between rows was 60 cm and plants were spaced 10 cm apart within a row. The spacing was consistent with normal field planting but was selected to optimize planting density within the available space and customized planting conditions. The experiments used custom-made plastic mulch as rain-out selection strategy. In the greenhouse, seeds were planted in 10-liter capacity plastic pots filled with pine bark. All experiments were laid out in a randomized complete block design with two replications. Drought stress was imposed by withholding irrigation water to 35% field capacity at anthesis, while the non-stressed treatment was well watered up to physiological maturity.

Quantitative data from 10 selected and tagged plants was collected during each generation to summarize the genetic variation and aid selection. The following data were collected during the M₁ through M₄: days to 90% maturity (DTM), plant height (PH), shoot biomass (SB), spike length (SL), 1000-seed weight (TSW) and grain yield (GY). In addition, percentage germination (%G) and number of spikelets per spike (SPS) were collected at M₁ and M₂ generations, while root biomass (RB) and root-shoot ratio (RSR) were measured at M₃ and M₄ generations. Data collection procedures were adapted from Mathew et al. (2019). The data were subjected to analysis of variance (ANOVA) and vital descriptive statistics were computed using GenStat 18th edition (Payne et al., 2017). The relationships among traits were quantified under each stress treatment using the Pearson correlations coefficient with the SPSS version 24 (IBM SPSS I, 2016). Trait correlation strengths were categorized into weak, moderate and strong following Zou et al. (2003).

The analysis of variance for M₁ and M₂ generations showed that the population × generation interaction effects were significant (p < 0.01) for SB, TSW, and GY (Table 3). Significant (p < 0.001) differences across the mutant generations were observed for all traits measured, while the population main effect showed significant (p < 0.05) impact on PH, RB, and GY.

There were significant (p < 0.05) differences in PH and SB in response to the three-way population × generation × water regime interaction effects at M₃ and M₄ generations (Table 4). The effects of the interaction involving generation and population were significant (p < 0.01) for SB and TSW only. The generation × water regime, and population × water regime interaction effects resulted in significant (p < 0.05) differences in SB, SL, TSW, and GY among the M₃ and M₄ mutants. Significant (p < 0.05) differences were observed among the M₃ and M₄ mutants for most traits due to the main effect of the mutant generations and water regimes, while only SB and GY varied significantly (p < 0.05) among the populations.

Summaries of quantitative traits measured at each generation and from various breeding populations were presented in Figures 2, 3. M₁ mutants from Population 2 recorded the shortest plant height (96.23 cm), highest shoot biomass (66.06 g/m²) and grain yield (21.75 g) compared with other breeding populations (Figure 2A). At the M₂, the mutants from Population 3 recorded the highest SB (61.82 g/m²) while mutant plants developed in Population 2 maintained the highest GY (19.48 g). Mutants from Population 1 recorded the shortest PH (83.71 cm) and highest TSW (47.29 g) (Figure 2B).

At M₃, mutant plants developed in Population 2 produced the highest grain yield of 11.58 g under drought-stress condition (Figure 3A). The highest shoot biomass was produced under non-stress and water stressed conditions at 80.04 and 71.51 g/m², respectively for mutants in Population 1. Mutants from Population 2 recorded the highest root biomass under non-stress and water stress conditions at 14.36 and 13.37 g/m², respectively. During the M₄ generation, mutant plants established in Population 1 produced the highest root biomass (9.38 g/m²) under non-stressed condition, while Population 2 recorded the highest RB (7.87 g/m²) under water stress. The highest GY (23.51 g) under non-stressed condition was recorded for mutants from population 3 while mutants from Population 1 had the highest GY of 14.53 g under water stressed conditions. Under water stress, mutants from Population 2 had the highest SB (32.93 g/m²) while mutant plants from Population 3 recorded the shortest PH of 87.33 cm (Figure 3B).

The mean performance of the three EMS-treated populations and the untreated control across four generations are presented in Figure 4. Mutants developed from Population 3 had the highest SB of 55.43 g/m² while the highest GY (18.39 g) was recorded for mutant plants in Population 2. The SL and TSW were the highest for mutants from Population 2 (13.64 cm and 61.61 g, respectively) across the four generations. Figure 5 summarizes the differences among the M₄ wheat Populations under water stressed and non-stressed conditions in two planting sites.

Many individual plants in the M₃ generation were available for selection based on their breeding population and observed variation in spike and awn morphology (Figure 6). Individual plants with variable tiller number (Figure 7), plant height and shoot biomass production (Figure 8) and, biomass partitioning into roots and shoots (Figure 9) were also observed. Qualitative traits had limited variation in M₃ generation when compared with the M₂. However, segregation at M₃ generation produced a wider range of variation (Figures 8, 9) making selection more efficient. Various spike mutants with high number of seeds from each breeding population were selected. Subsequently, abnormal, and deformed spikes with low number of seeds were discarded. Mutants with high root and shoot biomass and number of tillers were identified and advanced to M₄ generation.

Grain yield showed positive and significant associations with SL (r = 0.71; p < 0.001), TSW (r = 0.41; p < 0.05), and PH (r = 0.49; p < 0.01). Plant height was positively associated with all traits measured in all the four mutant generations (Table 5). Shoot biomass exhibited positive and moderate correlations with DTM (r = 0.46; p < 0.01), PH (r = 0.43; p < 0.05), and TSW (r = 0.40; p < 0.05). Strong and positive correlations existed between TSW and PH (r = 0.79; p < 0.001), and between TSW and SL (r = 0.77; p < 0.001). Likewise, DTM had moderate correlation with TSW (r = 0.46; p < 0.01).

In Table 6, the upper diagonal shows correlations recorded under non-stressed conditions. There were strong correlations between GY and RSR (r = −0.72, p < 0.001), SL (r = 0.77, p < 0.001), and TSW (r = 0.65, p < 0.01). The secondary traits also exhibited interdependent associations. The RSR exhibited a negative and strong association with SL (r = −0.72, p < 0.001) while SB and RB (r = 0.83), SB and RSR (r = 0.62), and RB and RSR (r = 0.79) were significantly (p < 0.05) correlated. The correlations among traits measured under water stressed conditions were different. Root biomass exhibited stronger correlation with GY (r = 0.55, p < 0.05) than the correlation between SB and GY (r = 0.30, p < 0.05) under water stressed conditions (Table 6, lower diagonal). SB was correlated to all the other traits, while RB was only correlated to RSR, SB, and GY. Grain yield exhibited significant association (p <0.05) with all traits except PH. The RSR exhibited moderately to strong correlations with GY (r = 0.67, p < 0.05), SB (r = 0.74, p < 0.001), and RB (r = 0.94, p < 0.001).

The significant (p < 0.05) effects of generations, breeding populations and their interaction for most agronomic traits probably resulted from genetic segregation or cumulative mutagenic effects in subsequent generations. Each generation was self-pollinated to generate the subsequent generation and the variation in subsequent generations could be due to segregation at heterozygous loci caused by mutations in M₁ generation. Similarly, Shorinola et al. (2019) found both superior and inferior mutants in later generations of wheat and supposed that the variation emanated from segregating heterozygous mutant phenotypes from the initial population. In other studies, the phenotypic variation between early and subsequent populations was attributed to the cumulative effects of the EMS. Hussain et al. (2018) asserted that the variation in subsequent generations is induced by non-lethal cumulative mutagenic effects. Singh et al. (2006) reported significant variation between M₁ and M₂ generations with reduced variation in M₃ generation, which was attributed to homozygosity even at mutated loci in advanced generations. Expectedly, phenotypic expression in mutant generations was significantly (p < 0.05) affected by drought-stress. Traits such as SB, SL, TSW, and GY were significantly reduced under drought stress, which corroborated previous studies (Marchin et al., 2020). Soil moisture is vital for biological process and nutrient transport, and inadequate water supply interferes with essential processes leading to poor growth and development (Daryanto et al., 2016). Grain yield production under drought stress was likely supported by families that were able to maintain high shoot biomass production. It is reported that shoot-related traits influence grain production under water-limited environments by translocation of assimilates previously synthesized in the shoot before the onset of the detrimental drought stress (Abdolshahi et al., 2015).

The lack of definite trends in the pattern of variation among the EMS-treated wheat populations point to the random nature of mutations induced by EMS and the wide variation created in subsequent segregating generations. The superior agronomic performance of EMS mutagenized populations compared to the untreated control for biomass, yield and yield-related traits measured under water stress during M₃ and M₄ generation indicates that EMS is efficient in creating potentially useful variation. It can be assumed that genetic modification through mutations induced by EMS improved drought tolerance. EMS is a potent mutagen and widely used in plant breeding programs (Talebi et al., 2012; Luz et al., 2016). Mutagenesis has potential to create genetic variation for exploitation in breeding for improved biomass and yield-related traits under water-limited environments (Addai and Salifu, 2016; Luz et al., 2016). Population 2 had the highest average biomass and yield performance across selection generations under the two water regimes. Selected individual mutants from Population 2 can be recommended for further screening for enhanced biomass and yield stability in water-limited environments.

Morphological variations reported in this study revealed the usefulness of chemical mutagenesis in wheat breeding. Detectable mutations result in traits that are morphologically distinct showing that such traits would be underpinned by heritable genetic changes (Gnanamurthy et al., 2012). The various types of spikes observed at the M₃ generation suggested that genetic changes in the spikes were attributable to EMS mutagenesis. Mutations can occur as chromosomal breakage, disturbed auxin synthesis, disruption of mineral metabolism and accumulation of free amino acids leading to variation in spike morphology (Goyal and Khan, 2010). Reportedly, spike morphology affects the extent of seed set and threshability of wheat grains. Minor and major mutations were reported to affect spike morphology in hexaploid wheat (Nalam et al., 2007). Mutants with favorable spike morphology (i.e., longer spikes, adequate seed set, and heavier seed weight) are useful genetic resources to improve grain yield potential and enhance grain threshability (Sharma et al., 2019). Variations in spike mutants generated from an EMS mutagenized wheat population study were reported by Dhaliwal et al. (2015). The positive effect of EMS mutagen was also confirmed by the wide range of variation in biomass traits. Variation in biomass is important to develop a larger breeding parental population for subsequent drought improvement programs, since evaluating and optimizing biomass partitioning will indirectly improve yield especially for water-limited environments. Variation in agronomic performance is a product of genotype, environment, and genotype × environmental interaction effects. The genetic constitution of individuals is modified due to EMS mutagenesis leading to heritable genetic changes. Mutational events amongst induced plants vary due to inherent genotype differences, EMS dose and permeability, and reshuffling of genes on mutated genomic regions (Hussain et al., 2018). Therefore, variation present in agronomic performance is underpinned by genetic changes introduced through EMS mutagenesis. This will condition substantial changes in the genetic homeostasis and physiological functions including in the synthesis of hormones, growth regulators, protein coding and cell division. As a result, heritable and desirable genetic changes in agronomic performance and physiological responses are introduced in functional mutants for selection. Detection of mutations caused by major genes is feasible using diagnostic molecular markers in the early mutant generations. However, changes due to minor genetic effects are linked to involvement of a larger numbers of genes that condition phenotypic segregations in the early selection generation. Hence, marker-assisted selection should be made in the advanced selection generation when homozygous and homogenous stable mutants are identified. This will identify genetically stable and homozygous individuals due to the inherent self-fertilization system in wheat. Previous studies used molecular markers to select mutant individuals in the M₃ or M₄ generations (Kenzhebayeva et al., 2014; Hussain et al., 2018).

The significant (p < 0.05) correlations observed among the measured traits suggest that the traits were interdependent and provide opportunities for simultaneous selection. The positive and significant association exhibited by GY and SB with the other yield related traits indicate the strong linkage between above ground traits. These traits can easily be selected simultaneously during yield improvement. Taller plants may be able to accumulate adequate photosynthates for attaining higher above ground biomass, which can directly increase grain yield (Zhang et al., 2009). Previously, the influence of above ground traits such as biomass production, spike morphology and kernel weight on grain yield was established (Reynolds et al., 2007; Kandić et al., 2009; Rahman et al., 2016). However, genotypes that accumulate excessive above ground biomass at the expense of developing extensive root systems may be susceptible to drought stress, especially in sub-Sahara Africa where wheat is grown under residual moisture and the rainfall is increasingly becoming erratic and inadequate (Haque et al., 2016). The stronger associations between the biomass traits and grain yield under water stressed conditions shows that biomass partitioning under drought is more critical for plant survival and attaining reasonable yield. For instance, a slight decrease in rooting capacity is likely to have higher influence on grain yield under drought stressed conditions compared to non-stressed conditions. Genotypes with potential to accumulate higher above ground biomass before the onset of drought stress have comparative advantage under terminal drought as they can translocate assimilates from shoot biomass to grains during grain filling (Kandić et al., 2009). The positive and significant correlations of RB and SB are favorable to develop cultivars with high extensive root biomass for water and nutrient extraction and shoot biomass for building adequate above ground biomass to support grain filling. Palta et al. (2011) asserted that a direct and positive relationship between root and shoot biomass is necessary for grain yield improvement. The negative association between RSR and GY regardless of moisture availability conditions indicates that there should be a balance between biomass allocation to above and below ground parts to avoid compromising grain production. Excessively large root systems have high maintenance costs that will limit amount of assimilates available for biomass accumulation in shoots or grain. On the other hand, shallow rooted plants with disproportionately large shoots have higher risk for lodging at anthesis, which increases chances of susceptibility to diseases and pests and reduces grain quantity and quality (Berry, 2013; Dahiya et al., 2018).