Sorghum [Sorghum bicolor (L.) Moench] is the fifth most important cereal crop globally in terms of area coverage and total production after wheat, maize, rice and barley. It has a predominant role in the food and fodder security for millions of rural families in arid and semi-arid regions of the world. Globally, sorghum is cultivated on 43.69 million ha, from which 66 million tons of grain is annually produced;  the   average  productivity  is  1.5 t ha^-1  (FAO, 2017).  In Ethiopia, sorghum is among the most important cereal crops, particularly in areas where rainfall is unreliable and crop failures due to recurrent drought are frequently observed. It plays a significant role for millions of food-insecure people living in such environments. Currently, sorghum is covering a total land area of 1.9 million ha from which 5.2 million tons of grain is annually produced (MoA, 2018; CSA, 2018).  The  major  sorghum producing regions are Oromia, Amhara, and Tigray that contribute 38.8, 35.5 and 13.4% of the area coverage and 40.5, 35.5, and 14.05% of the total production, respectively (CSA, 2018). Despite the multiple importances, the average national yield of the crop has remained very low largely due to drought (Amelework et al., 2015; Mera, 2018; Teshome and Zhang, 2019) and Striga (Ejeta, 2007; Abate et al., 2014).

Drought is a major constraint to sorghum production worldwide, although sorghum by its nature is considered as a highly drought tolerant cereal crop (Kassahun et al., 2010; Sabadin et al., 2012; Reddy et al., 2014; Amelework et al., 2015; Mera, 2018; Teshome and Zhang, 2019). Yield loss due to drought in the tropics alone exceeds 17% and reaches up to 60% in severely affected regions (Ribaut et al., 2002). In Ethiopia, where more than 50% of the total area is drought-prone, insufficient, unevenly distributed, and unpredictable rainfall is usually experienced in drier parts of the country (Amelework et al., 2015; Mera, 2018; Teshome and Zhang, 2019) in which nearly 40% of the population lives (EMA, 1988). It is manifested by either of the delay in onset, dry spell after sowing, and drought during critical crop growth stage such as flowering and grain filling (early withdrawal of rain). Moisture stress during later growth stages (grain filling) is the common phenomenon facing subsistence farmers in the country. It is frequently observed that drought is occurring at more frequent intervals-every two years during recent years. For instance, between 1960 and 1990 there were six drought episodes in the country, but between 1990 and 2014 the episodes increased to nine (USGS, 2017; Mera, 2018) causing as much as complete loss of sorghum and other crops affecting millions of people. This shows drought is becoming very challenging for production and productivity of sorghum and many other crops, possibly due to changing and variable climates. The large loss of sorghum yield is also related to the poor drought tolerance level of the available cultivars/varieties. Hence, control of drought through different options remain an important factor with priority geared towards ensuring food security in Africa as a whole and Ethiopia in particular.

In addition to the agronomic moisture conservation methods like tie-ridging, rainwater harvesting and soil-water conservation, breeding for more productive crop cultivars is one of the sound strategies in increasing crop yields in drought-prone environments. This is because better environmental manipulation with moisture-conserving agronomic practices alone may not lead to better yields from inferior genotypes unless they are integrated with crop genotypes that are capable of efficiently exploiting the limited moisture conserved (Singh, 2002). Therefore, the use of resistant/tolerant varieties could be one of the feasible alternatives to further increase its productivity, stabilize production and contribute  to  food  security  in  areas  where drought is a regular feature of most sorghum growing environments. To this end, conventional breeding has been contributing immensely towards genetically insulating sorghum from various abiotic stresses such as drought for the last many decades. Nonetheless, in the current scenario of crop production wherein multiple and new threats have arisen, conventional breeding alone does not seem to be an effective approach because of technical difficulties encountered in making major advances such as long crossing and backcrossing cycles, costs, and influence of genotype by environment interaction (Bartels and Sunkar, 2005; Khera et al., 2013).

Experiences elsewhere show that when modern biotechnological tools are properly applied with the conventional breeding system, it is obvious that the long backcrossing cycles to transfer specific genes of interest would be shortened, gene pyramiding would be simpler and the release of high yielding varieties and their subsequent use as improved seeds would be enhanced and hastened. The conventional sorghum breeding efforts supported by molecular assisted tools have scored remarkable successes in identifying and incorporating genes for tolerance to drought (Subudhi et al., 2000; Tao et al., 2000; Xu et al., 2000a; Haussmann et al., 2002; Sanchez et al., 2002; Kassahun et al., 2010). The development of drought-tolerant varieties has been dominantly focused on two distinct stages: pre-flowering and post-flowering. The best-characterized form of drought tolerance during the later stage of crop growth is the so-called stay-green, which is the ability to resist premature plant senescence (retain green leaf area), resist lodging and fill grain normally (Rosenow et al., 1983). Maintaining the greenness of leaves for a longer period is a principal strategy for increasing crop production, particularly under water-limited conditions (Tao et al., 2000; Xu et al., 2000a; Haussmann et al., 2002; Sanchez et al., 2002;  Kassahun et al., 2010; Abdelrahman et al., 2017).

Considerable work has been done on the identification of stay-green genotypes, mapping and identification of quantitative trait loci (QTLs) associated with the trait (Xu et al., 2000a; Haussmann et al., 2002; Sanchez et al., 2002). Therefore, it is advisable to validate, refine and adopt molecular markers already developed elsewhere for drought tolerant to better serve the needs in Ethiopia. On the other hand, local sorghum cultivars are highly preferred by the farming communities mostly for their yield, biomass and other morpho-agronomic attributes despite their susceptibility to terminal moisture stress. To this end, limited works have been made so far to improve the major limitations (such as vulnerability to drought) of these cultivars. Thus, conversion of popular and farmer’s preferred cultivars into their drought tolerant versions through incorporation of the responsible genes employing marker-assisted backcrossing (MABC) seems to be the best strategy in terms of time saving, effectiveness and efficiency. The present study was therefore, conducted to introgress drought tolerant genes/QTLs into popular and farmer preferred cultivars through MABC and assess the stay-green expression and associated agronomic performance.

Plant materials

The parental sorghum lines used for this backcrossing program were one donor parent “B35” and eight recurrent parents which are released varieties and known farmers’ cultivars (Table 1). The donor parent is known for post-flowering drought tolerant and it has been used as a source of tolerant genes to drought by the national sorghum-breeding program. B35 is a 3-gene dwarf genotype, BC₁ derivative of IS12555 accession, a durra from Ethiopian and is known for its stay green behaviour (Rosenow et al., 1983), more specifically a type-A stay-green-delayed onset of leaf senescence (Thomas and Smart, 1993; Thomas and Howarth, 2000). As characterized by several research groups (Crasta et al., 1999; Subudhi et al., 2000; Xu et al., 2000b; Sanchez et al., 2002), it was identified as a source of a number of stay green QTLs involving B35. B35 is also known for a number of other characteristics including early maturing, long in stature, has short compact panicle with copious number of infertile branches; purple genotype with small seeds covered by glumes, dry leaf midrib and relatively low yield potential (Srinivas et al., 2009; Kassahun et al., 2010). The recurrent parents are generally high yielding and biomass under optimum moisture conditions (MoA, 2018) and popular amongst the farmers but susceptible to terminal drought.

Development of backcross progeny

The popular and farmers preferred Ethiopia sorghum pure lines (improved and/or local) were crossed with B35 (with stay-green genes). The crossings were made using hand pollination method to generate F₁ progeny and subsequent generations at Melkassa Agricultural Research Center, Ethiopia. Crossing is done by emasculation of selected plant panicles (recurrent parents) and dusting of pollen from identified plants (donor parent). After analysis for the presence of the desired donor parent alleles and recurrent parents’ genome, selected heterozygous F₁ plants were backcrossed with respective recurrent parents to generate BC₁F₁ progeny. Thereafter, the individuals selected based on desired marker(s) were backcrossed to generate BC₂F₁. After each series of   backcrossing,   marker-assisted  foreground  (donor  allele)  and background (recurrent parent’s recovery potential) selections were made to fix through twice selfing and generated 61 BC₂F₃. In this study, five QTL (Stg1and Stg2 (on SBI-03), Stg3a and Stg3b (on SBI-02), and Stg4 (on SBI-05) associated with the stay green character was targeted (Xu et al., 2000a; Crasta et al., 1999; Subudhi et al., 2000; Tao et al. 2000; Haussmann et al., 2002; Sanchez et al., 2002).

Evaluation of backcrossed and parental lines for drought tolerance

Description of study area

Field experiment was conducted in Rama Kebele of Mereblekhe district in central zone of Tigray, Ethiopia. The location was selected based on the potential of sorghum growing and availability of irrigation for imposing a managed level of stress. Rama kebele is situated at 14°23’39″ N latitude and 038°48’90″ E longitude. Rama is found at an altitude of 1389 m above sea level, with average minimum and maximum temperatures ranging from 22 to 38°C, respectively, during the study period (December 2018 to May 2019). The district is characterized by eutric cambisols, haplic xerosols, orthic solonchaks, calcic xerosols, chromic cambisols, eutric nitisols, and orthic luvisols soil types in order of their importance. The specific site was characterized by eutric cambisols soil type.

Experimental setup and treatment combinations

The field trials consisted of 61 BC₂F₃, one donor parent and eight recurrent parents (Table 2) which were evaluated under well-watered and water-limited conditions arranged in -lattice design with three replications. The limited irrigation (stress) trial was irrigated well during the early growth stages but irrigation was withheld after anthesis. Meanwhile, the well-watered trial was fully-irrigated, so that, essentially, no moisture stress occurred at any stage of the crop development. The trials were planted on the same date and the same field in adjacent blocks. The mean traits obtained from the full-irrigation trial were only used to determine the relative mean trait relative reductions and expressed in percentage. The experimental units were two-rows of 4 m long with 0.15 m plant to plant spacing and 0.75 m row to row spacing. Fertilizer (NPS) was applied at the rate of 100 kg ha^-1 at planting and urea at rate of 50 kg ha^-1 split two times, half at planting and the remaining half as knee height. All other agronomic management and protection practices  were  applied   uniformly   to  all  plots  as recommended.

Data collection

Data were collected on important morpho-agronomic and physiological parameters on either pre-tagged random sample plants or whole plot basis depending on the trait studied.

Agronomic traits

The important agronomic traits recorded in this study include: plant height (PLHT, in centimeter, the height of the plant from the bottom to the tip of the panicle at maturity), days to flowering (DTF, number of days from emergence to 50% flowering), days to maturity (DTM, number of days from emergence to form a black tip on seed at the junction between seed and plant at the base of the head), biological yield (BM, in kg, sun dried weight of all above ground part from a plot and later converted to kg ha^-1) , grain yield (YLD, in kg, the grain yield harvested on hectare basis), panicle length (PL, in centimeters, measured from the bottom to tip of the panicle), panicle width (PW, in centimeters, measured at the middle panicle diameter), panicle weight (PWt, in grams, measured from five heads) and thousand seed weight (TSW, weight of 100 seeds in grams and later converted to thousand seed weight).

Physiological or stay-green characters

The leaf senescence expression of individual introgressed and their parental lines were estimated visually on a scale of 1 to 5 based on the degree of premature leaf and plant death at physiological maturity from five pre-tagged plants, hat is, 1 = very slight senescent, 2 = 25% leaves senescent, 3 = 50%  leaves  senescent, 4 = 75% leaves senescent, and 5 = 100% or complete senescent as suggested by Wanous et al. (1991). The total chlorophyll contents were measured with a Minolta Chlorophyll Meter SPAD-502 (Konica-Minolta Camera Co., Ltd Tokyo, Japan) at booting (SPADB) and physiological maturity (SPADM). The SPAD readings were taken from the middle of the leaf lamina of the second and fourth leaves from the top on five random pre-tagged sample plants at three places and averaged for analysis (Xu et al., 2000b). The total number of green leaves at booting (NGLB) and maturity (NGLM) were counted and used to determine percent of green leaves retained at maturity (PGLM), obtained as ratio between NGLM to NGLB expressed in percentage (Srinivas et al., 2009). Green leaf area at booting (GLAB in cm²) and maturity (GLAM in cm²) were measured from the length and the width of five green leaves from the top to bottom five pre-tagged plants and the area of each leaf was estimated using a correction factor of 0.70 (Mahalakshmi, 2002; Srinivas et al., 2009) as:

Leaf area = leaf length × leaf width × 0.70

The total green leaf area of each tagged plant was calculated as the sum of all the measured leaves from that particular plant. The upper six leaves were considered for measuring the green leaf area (Haussmann et al., 2002) as the upper leaves are photosynthetically active and directly assimilate mostly to the grain (Joshi et al., 2003). The average percentage green leaf area preserved at maturity (PGLAM) from each plot was calculated by dividing the total green leaf area of each plot at maturity (GLAM) by the total green leaf area of that plot at anthesis (GLAB) (Srinivas et al., 2009). The rate of leaf senescence (RLS in cm² day⁻¹) was determined as: RLS = [GLAB - GLAM]/number of days taken from booting and maturity (Reddy et al., 2014).

Estimation of relative traits reduction due to water stress

The relative traits reduction (RR) was calculated from traits obtained under full-irrigation (Yp) and water-limited (Ys) conditions as follows:

Data analysis

Data were subjected to statistical analysis using R software version 3.6.1 (R Core Team, 2019). Genotype differences in agronomic and physiological characters were analysed by residual maximum likelihood algorithm (ReML) as suggested by Patterson and Thompson (1971).

Estimation of heritability in broad sense

Broad sense heritability (H²b) was estimated as described by Allard (1960) as follows:

where  genotypic variance, = environmental variance, and r= number of replications.

Genetic advance from selection

Genetic advance (GA) was calculated with the method suggested (Allard, 1960; Falconer, 1989), assuming the selection intensity of 5%, as:

GA = K× σph×H²_b

Where, K= the constant differential (K=2.063 at 5% selection intensity), σph = square root of phenotypic variance and H²b = broad-sense heritability.

The genetic advance as percentage of the mean (GA%) was calculated as described by Johnson et al. (1955) and Falconer (1989) as follow:

x = Grand mean of a character.

Phenotypic trait performances

Differences among the genotypes were significant (P<0.05) for a number of characters (Table 3). The comparison of the developed progeny with their parents showed superior performances for many agronomic attributes. The overall mean of days to flowering (DTF) was 82 days. It is believed that the difference in DTF was attributed to the genetic background as  it  was  subjected to uniform irrigation until the induction of stress after flowering. The mean plant height of the genotypes was 136.4 cm. Plant height of the converted progeny and recurrent parents ranged from 100.3 to 192.6 cm and 104.1 to 163.8 cm, respectively. This showed that the converted progeny performed well revealing the amalgamation of the targeted genes from their parents. The shortest plant height was recorded from B35 with values of 93.5 cm, as it was expected (Kassahun et al., 2010). Days to maturity (DTM) were 118.7 days. The longest DTM was recorded for the developed progeny than either of the parents. The mean DTM of the backcrossed progeny and recurrent parents varied from 114 to 126.7 and 114.8 to 119.9 and that of B35 was 116.9 days, respectively. About 23% of the converted progeny showed significant delay in days to maturity. This might be due to high vegetative growth and devouring of relatively longer part of their reproductive growth to an end-of-season, which is the behavior of the donor parent ‘stay-green’. The maintenance of grain filling in the last stage of plant maturity has been considered as a key to the success of stay green genotypes (Luche et al., 2015). When looking at the yield components such as panicle length, panicle width, and panicle weight ranged from 14.6 to 27.1 cm, 4 to 7 cm, and 116.7 to 465.4 g, respectively. About 6.6, 9.8, and 13.1% of the backcrossed lines showed good performance for the aforementioned traits, respectively, than their parents (Table 3).

The score of leaf senescence (LS) of the genotypes ranged from lowest (stay-green) 1.85 to 4.6 (leaf drying). Among the developed progeny, 34.4% exhibit delayed/reduced LS with values ranged from 1.85 to 3. The converted progeny with stg1+stg2+stg3a+stg4, stg1+stg2+stg3a+stg3b and stg1+stg3a+stg3b+stg4 QTLs showed relatively delayed LS. The donor parent (B35) had LS score of 2.25. The mean rate of leaf senescence (RLS) was 1.1 cm² day^-1. Among the converted progeny, 15 had lower RLS ranged from 0.35 to 0.85 cm² day^-1 and comparable with B35. In most cases, those genotypes with stay-green (LS) trait also had lower RLS.

The results showed that there were six progeny that exhibit good stay-green characters, which can be recommended for further evaluation as potentially released, particularly in environments in which available water during grain filling is not adequate to support potential transpiration. It is a fact that LS is associated with the balance between hormones such as cytokinins and ethylene, and the over expression or suppression of these hormones show changes in the timing of senescence, accelerating and retarding the process (Buchanan-Wollaston et al., 2003; Gregersen et al., 2013). Genotypes with stay-green characteristics have been found to contain higher cytokinin levels (Reguera et al., 2013; Ambler et al., 1987); more stem sugars (Duncan et al., 1981; Dahlberg, 1992; Borrell et al., 1999, 2000b; Zwack and Rashotte, 2013) and more nitrogen possibly associated with a higher transpiration efficiency (Borrell and Hammer, 2000; Borrell et al., 2001; Mahalakshmi and Bidinger, 2002) than senescent genotypes. In addition, drought increases the C/N ratio and this C/N imbalance is associated with various senescence-related symptoms, including decreases in photosystem II efficiency and chlorophyll content, along with up-regulation of senescence-related genes (Reguera et al., 2013; Chen et al., 2015).

Furthermore, the stay-green phenotype may be achieved via the modification of root architecture (nodal root angel) (Mace et al., 2012), canopy development (Borrell et al., 2000a), or both. Mace et al. (2012) reported that nodal root angle in sorghum influences vertical and horizontal root distribution in the soil profile and is thus relevant to drought adaptation. The same report also indicates colocation of the QTLs between nodal root angle and the stay-green drought response in sorghum. Generally, characters such as GLAM, LS, and subsequent RLS are important factors determining greater green leaf area during grain-filling (Van Oosterom et al., 1996; Borrell et al., 2000a; Mahalakshmi and Bidinger, 2002). It is believed that stay-green plants photosynthesize for a longer period (Hörtensteiner, 2006; Tian et al., 2013; Borrell et al., 2014; Abdelrahman et al., 2017) though C-N transition point is delayed, or the transition occurs on time but subsequent yellowing and N remobilization run slowly (Yoo et al., 2007; Thomas and Ougham, 2014).

Relative trait mean performance reduction due to water-stress

Drought stress affects all phenological growth stages, reduces the normal growth and development periods, dry matter production and final yield. The relative reduction was determined on population basis, implying that the parents (recurrent and donor) and developed introgressed lines (whole and 10% selected based on the high yield). All the traits considered in this study were affected by terminal water-stress, although at different magnitude. In general, the mean relative traits performance reduction ranged from 5.7 to 38.9%. The mean relative reduction of plant height, days to maturity, panicle length, panicle width, panicle weight, biomass, thousand seed weight, and grain yield were 13.6, 5.7, 6.9, 16.7, 34.1, 43.3, 14.7, and 38.9%, respectively (Table 4). The relative reduction of grain yield ranged from 34.8 to 43.2% with a mean of 38.9%. The highest relative reductions in grain yield were recorded for the overall developed progeny (43.2%) followed by the recurrent parent (40%) depending on the genetic background and QTLs/genes expression. The relative reduction for the 10% selected converted progeny was intermediate (34.8%), indicating the presence of some promising lines that better tolerate the effects of water- stress as compared to the parents. In general, panicle weight, dry biomass weight and grain yield was amidst the severely affected by the terminal drought or stress.

Estimates of broad sense heritability

The majority of characters showed medium to moderately high   H²b  (40-74.67%)   except  days   to  flowering  and maturity (Table 5). As a bench mark, heritability values greater than 80% were grouped as very high, values from 60-79% were moderately high, values from 40-59% were medium and values less than 40% were low (Johnson et al., 1955; Singh, 2002). Accordingly, most of the characters were categorized as medium or moderately high H²b. The characters having very high heritability (≥80%) indicated that the relative small contribution of the environmental factors to the phenotype and selection for such characters could be effective. Conversely, a trait with low broad-sense heritability (below 40%) indicated that selection could be difficult or virtually impractical due to the environment, concealing genotypic effects (Vinodhana et al., 2009; Keneni, 2012).

Genetic advance from selection

It is obvious that  heritability  in  conjunction  with  genetic advance has a greater role to play in determining the effectiveness of selection of a character. In this study, genetic advance were high for plant height (23.86%), panicle length (13.87%), panicle width (34.11%), biomass (44.1%), yield (42.53%), SPADM (14.69%), GLAB (13.58%), and LS (22.6%) as described by Johnson et al. (1955). Thus, characters with both high H²b and GA indicate selection based on these traits could be effective (Table 5).

The results demonstrated that stay-green QTL from the donor parent have been successfully introgressed into the recurrent parents and are expressed in the developed backcrossed lines. This was exemplified by the presence of more green leaves, greater green leaf area, and high chlorophyll  content  especially  at  physiological  maturity and consequently enhanced grain yield. This shows the potential of MABC in building up the existing cultivars profile in enhancing drought tolerance, which might have limited success with only phenotypic selection.

The authors have not declared any conflict of interests.

The authors greatly appreciate the support by the Ethiopian Institute Agricultural Research (EIAR), Melkassa Agricultural Research Center (MARC), Addis Ababa University and Agricultural Growth Program-II (AGP-II).