Agromorphological Characterization of Introgression Lines Derived from Wild and Exotic Sorghum Germplasm to Climate Change Challenges

Sorghum (Sorghum bicolor (L.) Moench) is vitally important for food security in semi-arid West Africa. The introgression of wild- and high transpiration efficiency genotypes may contribute genes for abiotic stress tolerance or novel traits. Characterizing the introgressed lines can help researchers in the process of varietal improvement. The objectives of this study are (i) To learn if introgression of wild sorghum (CWR) and high transpiration efficiency (HTE) lines into a very early maturing (photoperiod insensitive) local land-race variety can create useful variation for grain yield in drought-prone Sahelian environments, (ii) To determine if grain yield advantages (and putative drought tolerance traits) obtained from introgression of CWR and HTE germplasm are expressed differently in moderately to highly moisture-limited environments, and (iii) To understand the relationship between putative drought tolerance traits. These populations were developed from crosses between the CSM63E (Jakumbe, 75% of the progenies of all populations indicated a chlorophyll content at physiological maturity (>30 SPAD) higher than CSM63E (20 SPAD) chlorophyll content. The estimated broad sense heritability was generally high (≥0.50) for all traits. The results also showed a very significant interaction between genotype and environment for grain yield, flowering time, chlorophyll content, stay green and lodging. There was a strong correlation between stay green and chlorophyll content (0.60) on the one hand and between stay green and lodging (0.48) on the other. This great variability could help improve the grain yield of sorghum for farmers in the Sahelian zone.


Introduction
Sorghum [Sorghum bicolor (L.) Moench] is a staple crop the oldest cultivated in the arid and semi-arid tropics of West Africa [1] [2]. Sorghum is a main food crop of the daily diet of several million people in the semi-arid tropics of West Africa [3]. It is also used for animal feed, energy production and as housetops building material [4]. Despite its importance, sorghum improvement has not seen the same progress as maize, wheat and rice [5]. Indeed, most sorghum improvement work around the world has focused heavily on two of the five main races identified (guinea, bicolor, caudatum, durra and kafir), although the least studied races are vital for the majority. Smallholder farmers depend on sorghum for food security and resilience to climate change [6]. Contrary to the cotton-growing areas located in southern Mali, several new varieties-lines and hybrids are released [7] [8], only one improved variety, Jakumbè (CSM63E), is cultivated by farmers in the arid region of Mali [9].
In addition, in Africa, the average yield of sorghum is still very low. This low productivity of sorghum in West Africa is linked to climate variability, in particular the low and poor distribution of rainfall and the poor exploitation of sorghum genetic resources for improvement [10].
Thus, the development of improved sorghum varieties combining diverse traits and better drought tolerance with adaptation to local conditions could contribute to improving food security in drought prone areas, through higher yield and more stability. Sustainable gains in plant breeding rely on variation in the gene pool of crops. This involves exploring new genetic diversity as well as understanding the genetic and physiological basis of traits contributing to higher yields. Several studies have indeed reported the possibilities of improving many stress adaptation traits through the use of related wild and weed species [6] [11].
However, the transfer of useful traits from wild relatives and weeds into new varieties has only rarely been attempted in sorghum. Therefore, harnessing the genes of wild and weed sorghum is a promising approach to improve the genetic diversity of sorghum for adaptive traits and provide a basis for the selection of high performing, more resilient and widely adapted varieties. Thus, the objectives of this study are 1) To learn if introgression of wild sorghum (CWR) and high transpiration efficiency (HTE) lines into a very early maturing (photoperiod insensitive) local land-race variety can create useful variation for grain yield in drought-prone Sahelian environments, 2) To determine if grain yield advantages (and putative drought tolerance traits) obtained from introgression of CWR and HTE germplasm are expressed differently in moderately to highly moisture-limited environments, and 3) To understand the relationship between putative drought tolerance traits.

Plant Material
Backcross nested association mapping (BCNAM) design was used to developed populations for this study using seven donor parents (two wild and five cultivated) and 1 recurrent parent. The recurrent parent CSM63E is an elite high yielding cultivar with grain quality, largely grown in the Sahelian zones and progressively in the Sudanian zone experiencing climate variability. The wild sorghums (BBISS-08 and BBISS-09) have the stay green character with also many tillers but have very small size seeds and other undesirable characteristics while the five cultivated donor parent were originated from Cameroon, Nigeria and Sudan (IS15428, IS16044, IS10876, IS16044 and IS3585) and differ from each other in their high transpiration efficiency (TE) and their racial and geographic diversity.

Site and Experimental Design
The agro-morphological characterization was carried out during the 2017 rainy season in two environments including Cinzana and the Bema. The Cinzana experimental station has a clay soil, richer in nutrients and more water-retaining, but heavier to work, especially when dry. It has an average temperature of 27.7˚C. While the station of Bema has a sandy-limous soil and an average temperature of 30˚C. Figure 2 showed the distribution of rainfall during this characterization period. A set off 669 BC1F5 progenies including the donors, the recurrent parent CSM63E and a local control Seguifa were evaluated in two environments using a randomized incomplete block design with three replications. The randomization was done with the BMS software. Experimental plots consisted of a single 3-m row with 75-cm distance between rows and 30 cm between hills within rows.

Phenotypic Evaluation and Management
The parental line, the CSM63E, the check and the 669 BC1F5 progenies were evaluated for grain yield and components, flowering time and physiological traits. For each environment the seeds of each of lines were sown in to each row after treatment with Apron Star. Three weeks after emergence, the seedlings were thinned to 2 plants per hill. Mineral fertilization was done according to the required standards, DAP was applied before ridging at a rate of 100 kg/ha and 50 kg/ha of urea (46% N) after 45 days after sowing. Two weedings were done if need and the plot was protected from birds to minimize their impact on the plants. Phenotyping was carried out for agronomic traits of grain yield, flowering time and for physiological traits such as chlorophyll content (SPAD), visual score of green leaf area (STG) and lodging (Ver) were also estimated. A total of eleven (11) quantitative and qualitative traits were used for agro-morphological  ing data were recorded by counting the number of days from sowing until 50% of the plants in each plot had flowered.

Phenotypic Data Analysis
Single environment analysis was done for each trait. The BMS (Breeding View) software was used to obtain the best linear unbiased estimate (BLUES) and the best linear unbiased prediction estimate (BLUP). Two models were used to run the analysis to obtain the BLUEs and BLUPs. In the first model, the genotype was taken as a random factor and the environment as a fixed factor and in the second model, the genotype was considered as fixed. The following formulas were used: Model: where Y ijk is the observed value; μ is the population mean; G i is the effect due to the i-th genotype tested; R j is the effect due to the j-th replicate; B k is the effect due to the k-th block; B(R) = interaction effect of the block in the replicate; E ijk is the effect due to the random error.
The genotype factor that was taken as random was used to calculate the predicted values and standard error for each genotype. R software was used for analysis of correlation, variance (ANOVA) and graphs.
The grain yield was calculated as follows: where 0.75 = line spacing; 3 = length of the sowing row; 0.6 = 2 borders corresponding to the inter-hill, GY is the grain yield and PGr is the grain weight. Coefficients of genotypic variation (CGV) were calculated according to the method proposed by [13] to allow comparisons of genetic variability of all traits between trials. To obtain the average variance of a difference (aVD) between genotypes, the square root of the standard error of difference (SEd) was estimated and normalized by dividing it by the grand mean of the trial. The strict heritability h² was calculated based on the formula of Cullis et al. [14].  where Y ij is the adjusted mean of the ith genotype in the jth environment, μ is the grand mean, G i is the effect of genotype i, E j is the effect of level j of the environment, GE ij is the effect of the interaction between genotype i and level j of the environment, ER jk is the effect of the interaction between level j of the environment and repetition k, ERB jkl is the effect of the interaction of the blocks in the repetition and in the level of the environment and e ijkl is the residual.
The broad heritability (H 2 ) was calculated using the method described by Piepho and Möhring [15]: where: Vg = genotypic variance, nE = number of environments, Vge = variance of the genotype-environment interaction, nR = number of repetitions and Ve = variance error. The genetic correlation (r g ) between the different putative drought tolerance traits was calculated as follows: where: Cov (trait1; trait2) is the covariance of the genotype means of trait1 compared to trait2; σ (trait1) and σ (trait2) are the genotypic standard deviation of trait1 and trait2 respectively. whereas at the Cinzana, grain yield was ranged from 80 to 500 g/m 2 . The best grain yield was obtained in the BC10 population (240 g/m 2 ) at Bema, while the BC15 population (500 g/m 2 ) had the highest grain yield at Cinzana compared to the recurrent parent CSM63E. Thus, at Bema, 50% of the progenies of populations had a higher grain yield (>100 g/m 2 ) than CSM63E (39 g/m 2 ), while at

Progenies Performance for Grain Yield
Cinzana only the progenies of two populations (BC09 and BC15) had a higher grain yield (>300 g/m 2 ) than CSM63E (>245 g/m 2 ). The progenies of the BC16 population gave the lowest grain yields at Cinzana compared to the recurrent parent.  environments. At the Bema, flowering time ranged from 51 to 95 days, while at Cinzana it ranged from 54 to 84 days. With the exception of progenies from the cross between CSM63E and the wild parent BC08 (Population BC08), most of the progenies were late maturing compared to CSM63E which was 65 days in both environments. The earliest progeny from the BC08 population reached 50% flowering 10 days before the flowering period of CSM63E. Progeny from the BC15 population tended to have the same flowering dates regardless of site.

Progenies Performance for Chlorophyll Content
The BC1F5 progenies exhibited a large variability in terms of chlorophyll con-  superior to more than 75% of the progenies. The chlorophyll content of the seven populations at the Bema is higher than those of the populations at the Cinzana site ( Figure 5).

Estimates of Genotypic Variances in Populations
Significant phenotypic variation among progenies was observed for morphological, physiological and grain yield traits in each environment shown by the range between the minimum and maximum of the traits. The differences were similar in both environments. Except the average grain yield at Bema, the average value of all traits was lower than the value of CSM63E ( Table 1). The overall genotypic coefficients of variation (GCV_G) for the traits analyzed ranged from 6.8% to 28.4% at Cinzana and 5.1% to 65.3% at Bema. The genotypic coefficients of variation of CWR progenies (GCV_CWR) and HTE progenies (GCV_HTE) are generally higher than the overall genotypic coefficient for all traits. The CWR-GCVs for the traits stay green and grain yield at Bema are higher compared to the HTE-GCVs at Cinzana. These coefficients of variation are higher for the characters chlorophyll content at maturity, stay green, lodging, vitreosity and grain yield. They are medium for the traits plant height, panicle length and hundred kernel weight and low for the traits flowering date and chlorophyll content at flowering. Heritability is high for all traits except chlorophyll content and lodging. Table 2 indicates the results of the combined analysis of variance of grain yield and its components. Analyzes showed a significant effect of genotype on the expression of agronomic traits panicle length, plant height, and grain yield in both environments. However, it was not significant (p > 00.5) on the chlorophyll content, the one hundred grain weight and the vitreousness at Cinzana. Regarding the effect of the interaction between the genotype and the environment, it was highly significant for grain yield. The genotype × environment interaction also had a significant effect on grain weight (p < 0.005) and vitreousness (p < 0.001).

Analysis of Putative Drought Tolerance Traits
The analysis results of putative drought tolerance traits for the parental lines and the BC1F5 was presented in Table 3. Most of the traits showed a heritability between 0.2 and 0.8. The score of the stay green trait for the donor parents was higher than that of the recurrent parent CSM63E in both environments, with an average of 3.2 for Bema and 2.5 for Cinzana, while the average score for the recurrent parent was 2. The average score of the stay green trait for the BC1F5 was  both with an average value of 30.2 SPAD and 18.5 SPAD respectively in Bema and Cinzana. The broad sense of heritability of the different putative drought tolerance traits ranged from 0.2 to 0.8. Heritability was high for the flowering time trait in both environments (Table 3).
Regarding genotype × environment interaction, it revealed significant differences among the BC1F5 progenies for each putative drought tolerance trait across the two environments. Estimates for the genotypic variance 2 g σ indicated significant effects for all traits evaluated. The estimated broad heritability was higher (>0.5) ( Table 4).

Correlations between Traits
The estimated phenotypic correlation coefficients for the different traits were given in Table 5. A significant but weak correlation between the flowering time and stay green traits, the flowering time and lodging and finally between the flowering time and chlorophyll content at 30 days after sowing was observed. Stay green was strongly correlated with lodging and stay green was also correlated with chlorophyll content at maturity. Lodging showed a very significant     correlation with the content of chlorophyll at maturity (r = 0.3***). The grain yield showed a significant correlation with lodging and the grain yield was also correlated with the content of chlorophyll at maturity.

Discussion
The progenies of the different populations expressed superior performance to the donor parents and to the recurrent parent CSM63E, especially in terms of grain yield, chlorophyll content and flowering time in both types of environments, showing the positive effect of the introgression. In addition, the superiority of the BC09 population over other populations and over the recurrent parent CSM63E could be explained by the fact that the wild relatives were generally found in a wide range of environments providing a source of genotypes adapted to new or changing environments in agricultural production areas [16]. Wild relatives of crops are a source of genes that have enabled significant advances in crop productivity [17] [18]. All populations provided a varied range of flowering time for populations in both types of environments. This could suggest that the material developed can be used for different agroecological zones. The variation in the flowering date was indeed a key feature in the adaptation of cultivated plant populations to environmental conditions [19]. The progenies developed showed adaptation to Cinzana as well to Bema. Quinby [20] showed that the flowering date has major impacts on crop performance during periods of terminal drought. Hanmier [21] indicated that a difference of 5 days between two varieties can express very large differences in grain yield.
The significant level of genetic variation in grain yield, flowering period, chlorophyll content and stay green observed in the seven populations could be due on the one hand to the possibilities of improving the genetic material of sorghum through the use of both high transpiration efficiency sorghum cultivars and wild sorghum and the other hand due to the method of development of the populations. Thus, differences in mean and frequency were noted among the best performing genotypes in progenies derived from the cross between the recurrent parent (CSM63E) and the wild parents and in one of the BC15 populations (cross between CSM63E and IS15428) regarding the grain yield, flowering period, chlorophyll content and stay green traits. Exploitation of wild species and cultivars with high transpiration efficiency could therefore contribute to increasing the diversity of genetic material. According to Gill et al. [22], wild plants have been exposed to extreme climatic conditions and have evolved in terms of adaptive traits and therefore constitute an important source of new traits. Bramel-Cox and Cox [6] also reported the possibility of increasing sorghum yields through the transfer of genes from wild races of sorghum.
Knowledge of genetic variability is essential in variety selection. Demonstrating this genetic variability for certain morphological traits is the first essential step in the description of genetic resources [23] [24]. The study of the coeffi-cients of genotypic variation of progenies from wild parents (CWR) and parents with high transpiration efficiency (HTE) showed high variability among progenies and within the populations. These variations were probably due to differences between the genotypes and especially between the donor parents and the recurrent parent. For the stay green and grain yield traits, the genotypic coefficient of variation of progenies from wild parents (CVG-CWR) showed a greater value than the genotypic coefficient of variation of the progenies derived from the high transpiration efficiency donors (THE), this indicated a significant inter-progeny variability of the progenies derived from crop wild relatives (CWRs). Schreiber et al. [25] reported significant genetic diversity in agronomic traits using wild relatives.
The high significant variability observed in grain yield has also been reported by several previous studies on sorghum [3] [26] [27] [28]. However, the 100 grains weight trait which was used to estimate seed size and grain yield before harvest, according to Deffan et al. [24] showed low variability. Knowledge of genetic variability allows breeding programs to rely on morphological variability [29].
The combined analysis of variance revealed significant effects for the environment, genotypes and genotype-environment interaction. This could confirm a phenotypic diversity within the progeny evaluated and showed different response of the progenies to the environments in terms of chlorophyll content at maturity, stay green, grain yield and flowering date. Previous studies have reported similar results [30]. The significant morphological diversity of the progenies could be attributed on the one hand to the genetic, racial and geographic diversity of the donors and the race of the recurrent parent and on the other hand to backcrossing methods.
The broad sense of heritability in the two environments was high for most of the traits. The higher the broad sense of heritability is, selection will be more effective for the traits [31]. Falconer and MacKay [32] reported that, the values of heritability below 30% are considered low, they are average between 30% -60% and very high, above 60%. The broad sense of heritability values that were high for most traits could indicate little environmental influence on the traits. In this case, the phenotype is a good predictor of the genotype [33]. The broad sense of heritability is an indicator to predict the response to selection, the more heritable is a trait, the more efficient and rapid the selection is to change the population mean [34] [35]. The broad heritability values of traits in the Cinzana environment were generally higher than those of traits in the Bema environment. This was explained by the fact that heritabilities appear to be stronger in conditions of high water and high fertility levels [36]. These results were similar to those reported by Leiser et al. [37] on the sorghum. The joint estimate of the coefficient of genotypic variation (GCV) and heritability (h 2 ) provides the best information for the choice of parents for hybrid development in order to obtain the preferred traits [13].
The genetic correlations between stay green and lodging and between stay green and chlorophyll content were the highest. Xu et al. [38] reported the relationship between the visual score of stay green and the concentration of leaf chlorophyll in sorghum. Wanous et al. [39] reported that visual assessments of green leaf area and green leaf number correlate strongly with measured values of green leaf area for chlorophyll content under drought stress. Blackmer and Schepers [40] reported a positive correlation of chlorophyll concentration with leaf nitrogen concentration and green leaf coloration in maize. The stay green trait protects the leaf from chlorophyll degradation [41]. The results indicated that visual stay green scores were a reliable indicator of leaf senescence and should be useful for sorghum breeders to assess progeny when breeding for drought tolerance.
The negative correlation between SPADIII and grain yield could be explained by the fact that at maturity there was competition between leaves and grains for chlorophyll. Thus, there was loss of chlorophyll from the leaves to the benefit of grain filling causing leaf senescence in most progenies.

Conclusion
Seven populations of 669 progenies were developed by introgression (BC1) including five populations from HTE (High Transpiration Efficiency) and two populations from CWR (Crop Wild Relatives). This study highlighted the potential of wild sorghum and high transpiration efficiency landrace to enhance breeding leading to the identification of superior genotypes. It also suggested that wild sorghums are useful as donors for improving yield for farmer-preferred varieties. The characterization showed the existence of a large phenotypic variability for the flowering period, stay green, chlorophyll content, lodging and grain yield within the 669 BC1F5 progenies. This should enable identification of higher yielding varieties for the drought prone areas in Sahel. The heritability estimates indicated that selection should be effective for grain yield and putative drought tolerance traits. The existence of a correlation between putative drought tolerance traits should be useful to breeders evaluating progenies for drought tolerance. The best selected genotypes were assessed in the farm environment and on-station in 2018 and 2019 using a participatory farmer approach to improve farmers' resilience to climate variability.