Introduction
Knowledge on the genetic diversity among genotypes is fundamental in planning crosses and for precise identification of genotypes for cultivar development and conservation (Geleta et al., Reference Geleta, Labuschagne and Viljoen2006; Perumal et al., Reference Perumal, Krishnaramanujam, Menz, Katile, Dahlberg, Magill and Rooney2007). In eastern Africa, more than 70% of sorghum is cultivated in the dry and marginal areas (Mukuru, Reference Mukuru and Byth1993) reflecting its ability to grow in adverse environments. A wide genetic diversity of sorghum has been developed over the years as a result of continued farmers’ selection (Ayana et al., Reference Ayana, Bryngelsson and Bekele2001). The heterogeneous nature of sorghum landraces aid selection for resistance or tolerance to drought and other abiotic and biotic stresses (Dar et al., Reference Dar, Reddy, Gowda and Ramesh2006). This offers the breeders greater opportunity to select superior parental genotypes for hybrid production.
Hybrid sorghum cultivars have been demonstrated to be more productive than pure line varieties (Kenga et al., Reference Kenga, Alabi and Gupta2004; Kamau, Reference Kamau2007). Significant heterosis for grain yield and other agronomic traits has been reported in sorghum (Blum et al., Reference Blum, Ramaiah, Kanemasu and Paulsen1990; Haussmann et al., Reference Haussmann, Obilana, Ayiecho, Blum, Schipprack and Geiger1999). It has also been reported that F1 hybrids have superior buffering capacity across variable environments than pure lines in sorghum (Reddy et al., Reference Reddy, Sharma, Thakur, Ramesh and Kumar2007). Consequently, breeding for hybrid cultivars is a better option than pure lines varieties while improving sorghum grain yield in water limited environments. However, the concept of heterosis and its patterns with respect to hybrid breeding in sorghum is not well-developed as that of maize (Blum, Reference Blum2013).
Different predictive methods have been employed in plant breeding to identify potential hybrids with superior yield performance. Parental selection based on individual performance, adaptability and yield stability have been used as the major selection criteria (Bertan et al., Reference Bertan, Carvalho and Oliveira2007). Combining ability tests are the traditional methods used to predict the hybrid contributions of sorghum parental lines (Bhatnagar et al., Reference Bhatnagar, Betran and Rooney2004; Fan et al., Reference Fan, Tan, Yang and Chen2004). However, the two methods demand field evaluations of large numbers of inbreds and hybrids in multi-environment trials. Consequently, more robust hybrid prediction approaches are required. The use of molecular markers has been proposed as a more efficient method of selecting inbred lines and superior hybrid combinations, and can reduce the number of multi-location trials of potential hybrids (Menkir et al., Reference Menkir, Melake-Berhan, The, Ingelbrecht and Adepoju2004; Barata and Carena, Reference Barata and Carena2006).
In Ethiopia sorghum breeding has been predominantly restricted to gemplasm characterization using phenotypic traits and combining ability tests to assess the heterotic patterns of exotic parental lines (Adugna and Tesso, Reference Adugna and Tesso2006; Degu et al., Reference Degu, Debello and Belete2009). Although, high level of genetic diversity was reported (Ejeta, Reference Ejeta, Ejeta and Gressel2007), the potential of local landraces for hybrid cultivar development has not yet been exhaustively assessed. Several researches have reported on the effectiveness of morphological and molecular markers for estimating genetic diversity of sorghum germplasm, however, very few studies have been conducted on the association of phenotypic and molecular markers with combining ability and heterosis for yield in sorghum. Therefore, this study was aimed to identify unique sorghum genotypes using phenotypic and simple sequence repeat (SSR) markers; and to determine the relationship between combining ability and heterosis for grain yield in hybrid breeding.
Materials and methods
Plant materials, crosses and study site
A total of 32 sorghum landrace lines and 4 cytoplasmic male sterile (CMS) lines were used in this study (Supplementary Table 1). The landraces were sourced from the Institute of Biodiversity Conservation of Ethiopia (IBC) and the cytoplasmic male sterile lines were provided by the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), Patancheru, India. The landraces were originally collected from three lowland drought prone zones in Ethiopia. They were selected based on prior study, their relatively superior yield performance and adaptability under moisture stress conditions. Landraces were kept homogenous through continued selfing and selection. The CMS lines (ICSA 101, ICSA 749, ICSA 743 and ICSA 756) were selected from different cytoplasmic sources (A1, A2, A3 and A4), respectively (Supplementary Table 1).
The 32 lowland sorghum landraces were crossed with the four CMS lines using a line x tester mating design. Crosses were performed at Sirinka Agricultural Research Centre (SARC). The 128 single-cross hybrids, the parental genotypes and four check varieties were evaluated. A local check variety, ‘Jigurty’, an early maturing, tall plant (~290 cm) with a compact head type, is commonly grown by farmers during moisture stress. Two check varieties, ‘Raya’ and ‘Miskir’, are early maturing, relatively short (120–200 cm) with semi loose head types that were released by the SARC for their moisture stress tolerance. A fourth check variety, ‘ICSV 111’ was developed for terminal moisture stress condition and was bred by ICRISAT.
The phenotyping and hybrid performance evaluation was conducted at the Kobo Research Station in Ethiopia under stressed and non-stressed condition. The Kobo site is situated at 11°56′ to 120°18′N and 39°23′ to 39°47′E at an altitude of 1400 m above sea level. The mean annual rainfall received during the crop growing period was 616 mm. Supplementary irrigation was applied every week from planting to the grain filling stage. Non-stressed plots received 10 supplementary irrigations (~250 mm of water) 4 d before and 6 d after anthesis. The supplementary water requirement was calculated using CWR = KC × EVT, where CWR is crop water requirement, KC is crop coefficient determined at different growth and developmental stages and EVT is evapo-transpiration rate. EVT of the area is computed by using the CROPWAT 4 Window model developed by Food and Agriculture Organization (FAO, 1992).
Phenotyping
The 32 lowland sorghum landraces and the four maintainer B lines were planted in 6 × 6 alpha lattice designs in three replications under rainfed and irrigated conditions. Each genotype was planted in single row of 5 m long with an inter-row spacing of 75 cm. Morphological evaluation was carried out on 10 individual plants of each genotype in the field under stressed and non-stressed conditions. The morphological descriptors used in this study were based on sorghum descriptors recommended by the International Board for Plant Genetic Resource (IBPGR and ICRISAT, 1993). A total of 25 descriptors were used, for 13 quantitative and 12 qualitative traits (Supplementary Table 2).
Genotyping
A set of 30 SSRs were used, which were selected from an SSR diversity kit (Billot et al., Reference Billot, Rivallan, Sall, Fonceka, Deu, Glaszmann, Noyer, Rami, Risterucci, Wincker, Ramu and Hash2012) derived from all the linkage groups of sorghum. Genotyping was conducted by DNA LandMarks Inc., Canada. For each genotype, a bulk sample of leaf tips from five seedlings was harvested 4 weeks after germination. The leaf samples were dried using a silica gel protocol (Rogstad, Reference Rogstad2003). The samples were arranged in 96-well plates. Genomic DNA was extracted from the dried leaf samples using the protocol of Xin and Chen (Reference Xin and Chen2012) and the quality of the extracted DNA was evaluated on 1% agarose gel.
Field evaluation of hybrids
A total of 128 single-cross hybrids and their 36 parental lines and the four check varieties were evaluated in the field using a 12 × 14 alpha lattice design with three replications. The B lines were used in place of their male sterile counterparts. Each genotype was planted in three rows of 3 m long with an inter-row spacing of 75 cm. The plots were separated by 1 m paths. Seeds were manually drill planted into the rows at a seed rate of 10 kg/ha. Planting for the irrigation experiment was carried out on 18 June 2012 and irrigation was applied two times a week from planting to the grain filling stage. Planting was delayed by 20 d for the rainfed experiment to ensure that the materials were exposed to terminal drought stress. Fertilizer was applied at the rate of 100 kg/ha diammonium phosphate (DAP) and 50 kg/ha urea as recommended for sorghum in the lowland of Ethiopia. All the DAP was applied at the time of planting, while urea was applied in a split application: at planting and 45 d after planting. To minimize variability and maximize genetic expression of the genotypes, the field was kept weed free by regular hand weeding. Standard agronomic practices were followed as required. For data recording five plants from the middle rows were tagged in every plot and covered with bird proof bags to protect them from bird damage. Grain yield was measured in all genotypes in the two environments.
Data collection and analysis
For the quantitative phenotypic traits, data were based on 10 measurements for each variety. Quantitative data were subjected to analysis of variance (ANOVA) using the alpha lattice procedure using GenStat for Windows, 17th edition (Payne et al., Reference Payne, Murray, Harding, Baird and Soutar2014). Homogeneity of variances among the two water treatments was examined using Bartlett's test. The quantitative data were converted into classes (Supplementary Table 2) and combined with the qualitative data, which were already in the form of classes. Diversity indices and principal component analysis (PCA) were calculated using GenAlex version 6.5 (Peakall and Smouse, Reference Peakall and Smouse2007) and SAS 9.1.3 (SAS Institute Inc., 2003), respectively.
Genotypic data were subjected to various measures of the genetic diversity of within and among accessions using GenAlex version 6.5 (Peakall and Smouse, Reference Peakall and Smouse2007). The χ2 test was performed to determine the differences in allele frequencies among SSR markers. Genetic diversity parameters, such as the number of alleles per locus (N a), the number of effective alleles per locus (N e), allele richness (A r) and average gene diversity (H e) were determined using the protocol of Nei and Li (Reference Nei and Li1979). Furthermore, the fixation index (F) and per cent polymorphic loci (PIC) were estimated for each locus and pre-determined group, based on collection zones. Genotypic and phenotypic relationships within and among genotypes were assayed with a neighbour-joining algorithm using the unweighted pair group method (UWPGM) in DARwin 5.0 software (Perrier and Jacquemoud-Collet, Reference Perrier and Jacquemoud-Collet2006). A dendrogram was then generated based on the dissimilarity matrix. Bootstrap analysis was performed for node construction using 10,000 bootstrap values.
Heterosis, general combining ability (GCA) effects of the parents and specific combining ability (SCA) effects of hybrids, their corresponding standard error and their mean square were estimated using the line x tester analysis described by Kempthorne (Reference Kempthorne1957). Combining ability effects were performed using SAS statistical software systems (SAS Institute Inc., 2003). The two water treatments were considered as two environments. The combined analysis over environment was done after homogeneity test of variances using Bartlett's test. Adjusted means for grain yield was obtained using the lsmeans option in the PROC GLM model procedure in SAS 9.1.3 (SAS Institute Inc., 2003). The adjusted means were then used, to the subsequent calculation of GCA, SCA and heterosis. Mid- and best-parent heterosis (MPH and BPH) were calculated according to Alam et al. (Reference Alam, Khan, Nuruzzaman, Parvez, Swaraz, Alam and Ahsan2004).
Results
Genetic diversity analysis using phenotypic traits
Table 1 summarizes the PCA and diversity index estimates among the tested sorghum landraces. Seven principal components (PCs) accounted for 80.4% of the total variability. The PC1 variations were mainly due to the average effects of phenological traits such as day to 50% anthesis and maturity and morphological traits such as leaf number, leaf width, leaf length and biomass. PC2 was the contrast between the mean values of panicle weight, leaf rolling, ear-head compactness and shape and average weight of panicle length and awn. Grain yield, panicle weight and 1000 seed weight averages mainly constituted variations in PC3. PC4 was mainly measured by plant height and panicle exsertion. Glume cover and grain colour mainly measured the average variations in PC5. In PC6 and PC7, leaf colour and stay-green traits, leaf orientation and midrib colour had significant contribution, respectively.
Table 1. Diversity index estimates and principal component analysis involving 13 quantitative and 12 qualitative traits in sorghum
DTF, days to 50% flowering; DTM, days to 50% maturity; GFD, grain filling duration; PH, plant height; PL, panicle length; PW, panicle weight; LL, leaf length, LN, leaf number; LW, leaf width; GY, grain yield; BIO, biomass; HSW, 100 seed weight; PE, panicle exsertion; AW, awn; GLcol, glume colour; GLcov, glume cover; LO, leaf orientation; LC, leaf colour; MRC, midrib colour; GC, grain colour, LR, leaf rolling; SG, stay-green; EHO, ear-head orientation; EHC, ear-head compaction; EHS, ear-hear shape; I, Shannon–Weaver diversity index; SE, standard deviation; λ, loading scores; % V; total variance accounted for each principal component (PC).
a Bold faced figures under each component signify traits contributing the most variation for each PC.
The magnitude of phenotypic diversity is computed based on Shannon–Weaver diversity index and presented in Table 1. The overall mean phenotypic diversity index of the lines was 0.84 showing high variability with respect to all phenotypic character classes. The accessions also showed very high polymorphism for panicle weight, glume colour, glume cover, midrib colour, grain colour, leaf rolling and stay-green traits. The mean gene phenotypic diversity was 0.57, with maximum and minimum values recorded by ear-head compactness (0.14) and glume colour (0.79). The genotypes also showed high phenotypic gene diversity for glume colour, midrib colour, grain colour and stay-green traits.
Genetic variation revealed by SSR markers
The statistics of genetic diversity parameters among genotypes using 30 SSR markers are given in Table 2. All the SSR loci investigated in this study were polymorphic, and a total of 206 putative alleles were detected among the 36 sorghum genotypes. The number of alleles observed per locus ranged from 2 (markers mSbCIR223, Xcup61 and Xtxp040) to 14 (Xtxp145). Based on the 30 SSR markers, on average there were 6.8 alleles per locus and at least three alleles were shared by a minimum of four parents. The effective numbers of alleles (Ne) for six loci were more than four, with a mean of 2.9 alleles per locus. The PIC values for SSR loci ranged from 0.11 for Xtxp012 to 0.88 for Xtxp145, with a mean value of 0.50. Half of the markers used had a PIC value of greater than 0.50. The results of χ2 test showed significant differences in allele frequencies at all loci for all the genotype sets. The mean gene diversity (H e) was observed to be 0.60, with maximum and minimum values recorded by the SSR markers mSbCIR223 (0.14) and Xtxp145 (0.88). The R lines used in the present study were genetically pure lines with mean fixation index of F = 0.99 and observed heterozygousity H 0 = 0.01. The genetic distance among the lines ranged from 0.40 to 0.80, with overall mean of 0.63.
Table 2. Summery statistics for the 30 SSR loci screened across 36 sorghum genotypes
LG, linkage groups; N a , total number of alleles per locus; N e , numbers of effective alleles per locus; A r , allelic richness, H 0, observed heterozygosity; H e , gene diversity within genotypes; F, fixation index; PIC, polymorphic information content; SD, standard deviation.
Genotypic and phenotypic relationships within and among genotypes
Cluster analysis based on genetic dissimilarity using the neighbour-joining method in DARwin 5.0 based on SSR markers and phenotypic traits classified the 32 landraces into three distinct groups (Fig. 1). The morphological dendrogram that demonstrated the clustering patterns of genotypes based on 25 phenotypic traits is presented in Fig. 1(a). The first main cluster contained 15 lines that were characterized by the heaviest panicle, white midrib colour and lowest panicle length. Lines in this cluster revealed the highest positive MPH in a cross-combination with ICSA 749 for grain yield. Cluster II consisted of 11 genotypes, of which seven were landraces and four CMS lines. This cluster was characterized by long semi-compact panicles, high panicle exertion, highest thousand seed weight as well as grain yield with stay-green trait. The genotypes in this cluster showed a maximum positive MPH for grain yield in a cross-combination with ICSA 743.The third cluster consisted of 11 genotypes, mainly landraces characterized by late maturity with the longest and highest number of leaves and highest above ground biomass. The lines clustered in this group revealed the highest MPH for grain yield in a cross-combination with ICSA 756. A dendrogram based on 30 SSR markers also revealed three main clusters (Fig. 1(b)). Cluster I consisted of a large number of landraces (14) and it was further sub-divided into two sub-clusters. The second cluster (II) comprised of seven landraces and three CMS lines while the third cluster was composed of 11 landraces and one CMS line. Clustering of genotypes based on dissimilarity, SSR and morphological markers revealed that 10 genotypes were clustered in the same group. In both types of diversity analysis, lines grouped in cluster I, II and III revealed the highest and positive MPH in a cross-combination with ICSA 749, ICSA 743 and ICSA 756, respectively (Fig. 1(a) and (b)).
Fig. 1. Dendograms generated using neighbour-joining based on unweighted pair group method of arithmetic averages (UPGMA) genetic dissimilarity depicting genetic relationship between 32 sorghum genotypes and 4 CMS lines: (a) classification based on agro-morphological markers; (b) classification based on SSR markers.
Combining ability effects and heterosis of grain yield
Combined ANOVA for grain yield measured across stressed and non-stressed environments are presented in Table 3. The coefficient of variation (CV) of yield for the combined environments was 16.3%. The CV in stressed environment was 20.4%, which was higher than in non-stressed environment (12.8%). Environment had a significant effect on the expression of grain yield. Highly significant (P < 0.001) environment by line and line x tester interactions were observed. However, the testers also showed significant (P = 0.05) interaction with environment (Table 3). The observed highly significant main effects due to hybrids for grain yield, suggests that further partitioning of the genetic variance into paternal and maternal combining ability effects could give more information about the superiority of the lines and testers. The mean squares due to GCA of the lines and testers and the SCA of the crosses were highly significant (P < 0.001) for grain yield (Table 3).
Table 3. Analysis of variance (ANOVA) and variance components of grain yield of sorghum genotypes measured across two environments
SE, standard error; R 2, coefficient of determination; CV, coefficient of variations.
GCA effects
The variance component estimates due to the SCA effects were larger than the GCA effects for grain yield. The ratio of the mean square components associated with variance of GCA and SCA was lower than unity for grain yield (Table 3) suggesting that the genetic variation observed among crosses on grain yield was mainly due to non-additive gene effects. Similar results were reported by Kenga et al. (Reference Kenga, Alabi and Gupta2004) for grain yield in sorghum. The female parents such as ICSA 743 (A3) and ICSA 756 (A4) had high positive GCA estimates for grain yield. However, among the male parents, 72572, 75454, 200654, 239175, 239208, 244735A and 242039B had high and positive GCA effects for grain yield (Table 4).
Table 4. GCA effects of 32 lines and four testers, and SCA effects and heterosis of 128 hybrids for grain yield in sorghum
SCA, specific combining ability measured in ton/ha; MPH, mid-parent heterosis and BPH, better parent heterosis measured in percentage; F1, mean grain yield performance per se measured in ton/ha; GCA, general combining ability.
SCA effects
Superior cross-combinations were selected based on both hybrid performance and SCA effects of hybrids. The estimates of SCA effects for grain yield are presented in Table 3. Among 128 crosses 48.5% had significant SCA effects for grain yield in a desirable direction. Twenty-one hybrids revealed the largest and statistically significant SCA, with values larger than 1.0 (grain yield t/ha) and the maximum SCA effect was observed in cross ICSA 756 × 244735A (4.3 t/ha) followed by ICSA 743 × 242039B (3.6 t/ha). Mean grain yields of hybrids varied from 2.2 and 10.0 t/ha. Crosses ICSA 756 × 244735A, ICAS 743 × 242039B, ICSA 749 × 75454, ICSA 749 × 73059, ICSA 749 × 214855, ICSA 756 × 242049A, ICSA 749 × 242049A, ICSA 756 × 239208 and ICSA 101 × 237260 exhibited the highest positive SCA effects and also provided the maximum grain yields across the two environments. In the SSR based dendogram (Fig. 1(b)) five of the eight parents (242039B, 237260, 244735A, 214855 and 73059) that revealed the highest positive significant SCA were grouped in cluster I and the other three (242049A, 239208 and 75454) were found in cluster II. However, in the morphological based dendogram (Fig. 1(a)) the eight parents clustered across the three separate groups, i.e. three parents (73059, 244735A and 75454), two (242039B and 214855) and three (242049A, 239208 and 237260) were grouped in cluster I, II and III, respectively.
Heterosis
The MPH expressed as the per cent increase of the F1 hybrid over the mean of the two parents for grain yield ranged from –51.1% to 85.4%. Crosses ICSA 756 × 244735A (85.4%), ICSA 743 × 242039A (73.7%), ICSA 749 × 75454 (48.3%), ICSA 749 × 73059 (43.6%), ICSA 756 × 242049A (43.4%) and ICSA 749 × 214855 (41.9%) showed significant and positive MPH. The BPH, which was computed as a per cent deviation from the best value of the parents, ranged from −60.0 to 74.5% (Table 4). The highest positive heterosis was observed in crosses ICSA 756 × 244735A (74.5%) and ICSA 743 × 242039B (64.3%).
Three heterotic patterns were observed on the basis of the distribution of the 32 landraces using PCA on SCA data of grain yield (Fig. 2). The grouping patterns were mainly based on the female parents where the genotypes assigned in cluster I had high and positive SCA effects in a cross-combination with ICSA 749/ICSA 756. Cluster II comprised of nine genotypes that showed negative SCA values in a cross-combination with at least three of the female parents. Cluster III was mainly represented by genotypes that revealed high and positive SCA effects in a cross-combination with ICSA 743.
Fig. 2. PCA using SCA effects of grain yield based on crosses involving 32 sorghum landraces grown in two environments.
The percentages of the total variance for the first 2 PCs used to obtain Fig. 2 were 37.7 and 31.3% with a total of 69.0%. There were significant grain yield variations among the 128 F1 hybrids and 36 parental lines, and significant SCA and heterosis for grain yield. However, no significant association was observed between genetic distance estimated by SSR and morphological markers (r = 0.122, P = 0.070) and between genetic distance measured by SSR and morphological marker and grain yield heterosis (r = −0.06, P = 0.671, r = −0.057, P = 0.064), respectively.
Discussion
Relatively low phenotypic diversity was observed for panicle related traits and high diversity for grain and glume colour, resulting from farmer's successive selection of compact head types. In addition, compact head types were characterized by short growth duration, larger seeds, the heaviest panicles and the high yielding, which is the distinctive characteristic of durra sorghum. The compact durra sorghum races are highly adapted to extremely dry conditions and the most frequent grown types in such environments (Grenier et al., Reference Grenier, Bramel, Dahlberg, El-Ahmadi, Mahmoud, Peterson, Rosenow and Ejeta2004). The PCA suggested that anthesis and maturity time, leaf rolling, awn, biomass, panicle weight, panicle length, leaf length, width and number were important traits in differentiating genotypes. This implies that these traits are distinct features of genetic variation in sorghum parental genotypes. These traits have also been reported to be crucial for moisture stress tolerance (Samarah et al., Reference Samarah, Alqudah, Amayreh and McAndrews2009; Taketa et al., Reference Taketa, Yuo, Sakurai, Miyake and Ichii2011). The phenotypic performances and structural variations of the genotypes are determined by the prevailing environmental factors (Rao et al., Reference Rao, Rao, Mengesha and Reddy1996).
In this study, no significant relationship was found between molecular and/or morphological distances, and hybrid performance. In studies on rice (Hua et al., Reference Hua, Xing, Xu, Sun, Yu and Zhang2002), wheat (Corbellini et al., Reference Corbellini, Perenzin, Accerbi, Vaccino and Borghi2002) and grain sorghum (Jordan et al., Reference Jordan, Tao, Godwin, Henzell, Cooper and McIntyre2003) there were also non-significant relationships between whole genome-based genetic distance and hybrid performance. However, Boppenmaier et al. (Reference Boppenmaier, Melchinger, Brunklaus-Jung, Geiger and Herrmann1992) and Mosar and Lee (Reference Mosar and Lee1994) reported significant genetic relationship between genetic distance and hybrid performance of maize and oats, respectively. The prediction power of genetic distance has been inconsistent in many studies using different species and different germplasm (Yu et al., Reference Yu, Hu, Zhao, Guo and Sun2005). This may be because of the peculiarities of many agronomic traits and lack of common phenotypic assaying methods across environments. In the case of SSR markers, the genetic distance estimates can be affected by several factors such as the distribution of markers in the genome, the number of markers used and the nature of the evolutionary mechanism underlying the variation measured (Powell et al., Reference Powell, Morgante, Andre, Hanafey, Vogel, Dingey and Rafalski1996). Additionally, the basic assumption for molecular diversity to predict hybrid performance is the existence of high levels of gametic phase linkage disequilibrium between yield quantitative trait loci (QTLs) and marker alleles (Jordan et al., Reference Jordan, Tao, Godwin, Henzell, Cooper and McIntyre2003; Sorensen et al., Reference Sorensen, Stuurman, Van der Voort, Peleman, Varshney and Tuberosa2007). QTLs influencing heterosis in grain yield are located in certain chromosomal regions, and are unevenly distributed over the genome. To improve prediction efficiency of molecular markers, dissecting the diversity of individual linkage groups will be exploited, as proposed by Jordan et al. (Reference Jordan, Tao, Godwin, Henzell, Cooper and McIntyre2003).
Although no significant association was observed between SSR-based genetic distances and grain yield heterosis, some patterns were detected in the distribution of some traits in each cluster. Cluster I in SSR-based tree was dominated by late flowering and high biomass genotypes that had better heterosis with ICSA 749. The high biomass was, in turn, expressed as large number of leaf and leaf width per plant. However, cluster II was dominated by early flowering with small panicle and high 100 seed weight genotypes that showed better heterosis with ICSA 756. It was reported that heterosis in sorghum is expressed as a high plant or crop growth rate as compared with the parents (Blum, Reference Blum2013).
Information on the GCA effects of parents helps breeders to estimate the genetic potential of a breeding material for many desired traits. The differences in GCA among lines are mainly due to additive genetic effects and higher order additive interactions (Falconer and Mackay, Reference Falconer and Mackay1996). The GCA effects for grain yield among the female parents were larger than those of male parents. This is probably because the female parents had undergone excessive selection for many years. Environment by lines, testers and line x tester interactions were significant (Table 3) indicating that the genotypes were diverse in performance and in their responses to the environments. This suggests that crops adapted to arid and semi-arid tropical environments, such as sorghum, have genetic plasticity that adjusts their grain yield in response to the prevailing environmental conditions (Mukuru, Reference Mukuru and Byth1993).
Genetic diversity is estimated based on the pedigree, phenotypic trait or molecular data. It has been suggested that the genetic distance between parents is positively correlated with heterosis of F1 hybrids. Therefore, the extent of genetic diversity between the two parents has been proposed as a possible measure of the prediction of heterosis (Zhang et al., Reference Zhang, Gao, Yang, Ragab, Saghai Maroof and Li1994). However, strong association has rarely been observed between heterosis and genetic distance between parents (Rao et al., Reference Rao, Reddy, Kulkarni, Ramesh and Lalitha2004). However, studies in different crops have shown moderate to strong correlation between combining ability and per se performance (Bertan et al., Reference Bertan, Carvalho and Oliveira2007). Even though this method is extensively used for prediction of heterosis, it is hypothetical and relies heavily on field evaluation. Consequently, a molecular marker approach was used to assess the diversity, to improve prediction of hybrid performance in a more reliable way and as a basis for designing and carrying out combining ability tests in the field (Menkir et al., Reference Menkir, Melake-Berhan, The, Ingelbrecht and Adepoju2004; Barata and Carena, Reference Barata and Carena2006). Jordan et al. (Reference Jordan, Tao, Godwin, Henzell, Cooper and McIntyre2003) reported that GCA-based prediction is more efficient in identifying superior hybrids than marker-based predictions. Thus, use of parents with high and favourable GCA effects may increase the concentration of desirable alleles; this in turn may increase the chance of creating the best performing cross-combinations (Kenga et al., Reference Kenga, Alabi and Gupta2004).
Based on the present study, lines 72572, 75454, 200654, 239175, 239208, 244735A and 242039B, and testers ICSA 743 and ICSA 756 were found to be the best combiners and would be useful for a future hybrid development programme for moisture stress agro-ecologies. There have been fewer studies on the mechanism of heterosis, heterotic grouping and the use of molecular markers as selection criteria for parents in sorghum when compared with other crops such as maize. Heterotic groups comprise sets of genotypes that perform well when crossed with genotypes from a different heterotic group (Haullauer et al., Reference Haullauer, Russell, Lamkey, Sprange and Dudley1988). Heterotic groups in sorghum have been defined by the milo-kafir cytoplasmic genetic male-sterility system where lines are grouped either as A/B-lines or R-lines (Quinby and Martin, Reference Quinby and Martin1954). However, in this study four different CMS systems were used and they classified the 32 genotypes into three heterotic groups. The availability of these different cytoplasmic sources provides a means to exploit the cytoplasmic diversity available in different races and species of sorghum from diverse geographic locations (Pedersen et al., Reference Pedersen, Kaeppler, Andrews, Lee, Banga and Banga1998). More than 60% of the hybrids with positive MPH involved had either ICSA 743 (A3) or ICSA 756 (A4) as the female parents. However, in most of the breeding programmes, the A1 system is the most commonly used as compared with the other cytoplasmic systems (Sleper and Poehlman, Reference Sleper and Poehlman2006). These CMS systems have distinct genetic features within and among species. Fertility restoration in sorghum is mainly conferred by a nuclear-encoded fertility restorer gene but in some cases additional modifier genes are required for full expression of the restorer gene (Rooney and Smith, Reference Rooney, Smith, Smith and Fredericksen2000; Sleper and Poehlman, Reference Sleper and Poehlman2006).
In the present study, the A3 and A4 CMS systems were the most stable and displayed highly significant positive SCA effects and MPH, suggesting their value in future sorghum hybrid programmes in drought prone areas. The high SCA and heterosis estimates may be associated with the high genetic diversity available among the parental lines. Previous reports by Li and Li (Reference Li and Li1998) and Rattunde et al. (Reference Rattunde, Zerbini, Chandra and Flower2001) indicated that high level of heterosis have been attained from diverse germplasm of sorghum.
Cluster analysis using morphological and SSR markers revealed the existence of three groups of lowland sorghum genotypes with distinct genetic and morphological profiles. Assessments of genetic variability within and between crop species and prediction of performance of hybrids based on a combination of methods were employed. None of these methods per se were found to be effective in predicting hybrid performance. Hybrid mean values coupled with favourable SCA and GCA effects should be used for meaningful selection of parental lines for the development of drought tolerant hybrids. In addition, the prediction efficiency of sorghum hybrid performance could be improved by the use of combined analysis of diversity using phenotypic and marker data. Heterotic potential in sorghum can be improved and patterns can be measured by exploiting the different CMS sources currently available.
In the current study most of the hybrids obtained from crosses among selected lowland landraces showed superior performances than their parents. Thus, it can be concluded that there is an opportunity to develop commercial hybrids with superior grain yield and improved drought tolerance. Male parents 72572, 75454, 200654, 239175, 239208, 244735A, and 242039B and female parents ICSA 749 and ICSA 756 displayed positive and significant GCA effects for grain yield. Crosses including ICSA 756 × 244735A, ICAS 743 × 242039B, ICSA 749 × 75454, ICSA 749 × 73059, ICSA 749 × 214855, ICSA 756 × 242049A, ICSA 749 × 242049A, ICSA 756 × 239208 and ICSA 101 × 237260 were identified as promising hybrids with highly significant SCA effects and positive heterosis for grain yield.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1479262115000696
Acknowledgements
The authors sincerely thank Alliance for a Green Revolution in Africa, the Institute of Biodiversity Conservation of Ethiopia, Generation Challenge Program, DNA Landmark Laboratory and the University of KwaZulu-Natal for providing technical and financial support for this study.
Conflict of interest/Ethical statement
The authors have not declared any conflict of interest.
