In the present study, a wide range of grain yield variation (31–322.7 g/ m2) with an average yield of 115.6 g/m2 was obtained among finger millet genotypes grown in environments with acidic soils. The observed variation in grain yield is in contrast with previously reported grain yield of finger millet grown at Adet and Finoteselam sites in Ethiopia, which yielded 239.7 g/m2 and 238.6 g/m2, respectively (Marefia et al. 2022). Likely, the difference in grain yields obtained in our study compared to the above report may be due to the acidic soils in Gute and Bako, while the soils at Adet and Finoteselam are not acidic and are therefore expected to produce higher grain yields. Interestingly, the top five high-yielding finger millet genotypes were originally collected from Wollega and Gojam (western Ethiopia), whereas most of the low-yielding genotypes were originally collected from Tigray and Gonder (northern Ethiopia). The genotypes originally collected from areas with acidic soils were high yielding, while those from areas with non-acidic soils were low yielding, and, hence, can be regarded as Al-tolerant and Al-susceptible, respectively. This could be mainly due to enhanced Al-tolerance of genotypes acquired through long-term exposure to acidic soils. Tolerance mechanisms could be by exclusion of Al from the root, sequestration of Al to different plant parts once it enters the plant cell, or both (Barcelo and Poschenrieder 2002, Kochian et al. 2005). Future research based on well-designed experiments will shed light on the tolerance mechanisms of the genotypes investigated in this current study. The consistently high-yielding genotypes across the two environments Ec-215829 (243.5 g), Ec-203322 (228.0 g), and Ec-100093 (213.5 g) should be considered in the finger millet breeding program for developing improved cultivars, particularly targeting acidic soils.
Heritability measures the transmission of characters from parents to their offspring (Falconer 1981). In this study, days to emergence, days to heading, days to maturity, and number of grains per spikelet showed higher broad sense heritability (H2) compared to plant height, ear length, and number of fingers. A similar heritability pattern among diverse finger millet germplasm was previously reported (Marefia et al. 2022). The moderate and high heritability observed in most of the targeted traits indicates that genotypes have higher effects than environments on their phenotypic variation, unlike traits with low heritability where environmental effects are high.
The correlations between most traits obtained in this study were in line with previous reports (Bharathi 2011, Sharma et al. 2018), where significant positive correlations of DH with EL, DM, and PH were reported. Likewise, the positive correlation between PH and GY obtained in this study agreed with previous findings (Sivagurunathan 2004, Sharma et al. 2018). These investigations showed that finger millet genotypes that grow taller in acidic soils produce higher grain yields. Interestingly, this study revealed genotypes with desirable characteristics in major target traits including grain yield, plant height as well as tolerance to soil acidity. Hence, crossbreeding of genotypes with these desirable characteristics (high grain yield, tall height, and Al-tolerant) will increase the likelihood of recombination of alleles with positive effects on the phenotypes, and eventually developing superior cultivars.
The DAPC and population structure analysis showed that the genotypes originated from six genetic populations (K = 6), which is in line with previous findings in finger millet (Brhane et al. 2022a). The detailed analysis of genetic diversity and population structure of the genotypes used in this study, including their potential significance for finger millet improvement, was previously published (Brhane et al. 2022a). They represent genotypes from diverse agro-ecozones in Ethiopia and Zimbabwe, as well as cultivars released in Ethiopia. Overall, the genotypic and phenotypic variation, and population structure determined in this set of genotypes make it suitable for GWAS to identify genomic regions and candidate genes potentially associated with Al-tolerance.
Determination of LD between markers is an important step in GWAS, as it facilitates the selection of a set of markers suitable for efficient marker-trait association analysis. The LD between markers varies within a genome as well as between genomes of different crops, as it is affected by various factors, such as reproductive mechanism, rate of recombination, selection pressure, genetic drift, physical linkage, and population structure (Puranik et al. 2020). Since the LD analysis in the present study was based on SNP markers across a large number of scaffolds, it was difficult to estimate the markers’ rate of LD decay accurately, as was also indicated in previous research (Sharma et al. 2018). In this study, about 19.2% of the SNP marker pairs showed significant LD (r2 > 0.1; P < 0.05). A similar finding was reported by Sharma et al. (2018), who reported that 17.9% of finger millet SNP marker pairs were in LD (r2 >; p 0.05). Another study on finger millet by Puranik et al. (2020) indicated that about 16.8% of finger millet SNP pairs were at LD (r2 ≥ 0.2; p < 0.05), and the maximum LD was dropped to its half (LD half-decay) at a distance of 28 kbp. Comparatively, genome-wide LD decay in foxtail millet ranged from 100 to 177 kb (Jaiswal et al. 2019).
In this study, 5226 GBS-derived SNP markers and eight phenotypic traits were used to identify significant marker-trait associations (MTA). The analysis revealed 73 novel SNP markers associated with one or more target phenotypic traits at highly significant levels. Sharma et al. (2018) reported 109 SNP markers associated with 14 finger millet agro-morphological traits including grain yield using SLST = 20, MLMM = 36, and MTMM = 53 models. Another GWAS study on finger millet identified 418 SNP markers associated with nutrition-related traits using GLM and mixed MLM models (Puranik et al. 2020). The relatively low number of MTAs identified in this study could be due to factors, such as the number of phenotypic traits studied and the stringency of the GWAS model used. Overall, this study showed that the GBS-derived SNP marker set used is suitable to identify potential markers suitable for marker-aided selection of genotypes for breeding finger millet for cultivation on acidic soils.
Among SNP loci associated with DH, an SNP marker located at position 1528 in scaffold LXGH01485647.1 explained the highest phenotypic variance in this trait. The alignment of this scaffold to the recently released finger millet chromosome-level assembly Devos et al. (2023) showed that this SNP is located on chromosome 5A within the ELECO.r07.5AG0380680 gene, which codes for Carboxyspermidine synthase /Carboxyspermidine dehydrogenase in finger millet. Carboxyspermidine synthase is an enzyme used for biofilm formation and cell viability in microbes through cationic polyamine spermidine (Ko et al. 2022). Polyamine spermidine promote plant growth in part by the induction of systemic resistance (ISR) to biotic and abiotic stresses (Melnyk et al. 2019). The genotypes at this locus are AA, AT, and TT, which showed significant phenotypic variation in days to heading, and AA genotypes started heading later than TT and AT genotypes.
Among SNP loci significantly associated with GY, LXGH01052700.1_28450 was the most significant, as it explained 30.8% of grain yield variation. The locus is located within the ELECO.r07.1AG0026860 gene on chromosome 1A as per the recently published finger millet genome by (Devos et al. 2023). ELECO.r07.1AG0026860 is an orthologue of the Pgl_GLEAN_10002072 gene in pearl millet. Pgl_GLEAN_10002072 showed significant association with mass features mostly engaged in hydrolase activity, especially acting on ester bonds to promote carotenoid accumulation in pearl millet (Yadav et al. 2021). This suggests the gene plays a key role in pathways that promote the accumulation of antioxidant-related flavonoid compounds. The internal detoxification work in a variety of ways, such as hormone regulation, antioxidant defense system, and vacuole compartmentalization. As an important condition to maintain plant growth and development, the antioxidant defense system (enzyme and non-enzyme system) produces stress response to eliminate the excessive accumulation of reactive oxygen species in the plants (Sharma and Dubey 2007). This ultimately improve grain yield.
The LXGH01346160.1_21363 locus showed a significant association with PH although it explained only 0.94% of the variation in plant height, which is the highest compared to that of other loci associated with this trait. This locus is located within the ELECO.r07.7AG0569730 gene on chromosome 7A. The ELECO.r07.7AG0569730 gene codes for late embryogenesis abundant protein (LEA_2) in finger millet. Late embryogenesis abundant protein, which is also characterized in various other crops, such as O. sativa (OsLEAs), Z. mays (ZmLEAs), and S. bicolor (SbLEAs) was shown to play an important role in plant stress adaptation (Wang et al. 2007, Li and Cao 2016, Nagaraju et al. 2019). When under abiotic stresses such as drought, cold, and salinity, these genes are expressed and play a protective role (Ingram and Bartels 1996). This suggests that the late embryogenesis abundant gene may be induced during Al-stress in finger millet to protect and promote plant growth.