Pearl millet (Pennisetum glaucum) is a nutrient-rich cereal crop cultivated in arid and semi-arid regions, particularly in sub-Saharan Africa, where it serves as a vital source of grain and fodder for millions of smallholder farmers. This crop is known to produce in hot, dry climates and nutrient-poor soils, making it a strategic crop for enhancing agricultural resilience in the face of climate change. Despite its impressive adaptability, pearl millet yield in sub-Saharan Africa remains low. In order to enhance productivity and nutritional quality, it is important to better understand the physiological and genetic mechanisms that regulate the uptake, accumulation, transport and utilisation of nutrients because they are critical for sustaining growth, development, and resistance to biotic and abiotic stresses.
In this project, we aimed at studying the diversity for leaf ion content in pearl millet and its genetic control. For this, a diverse panel of 165 pearl millet inbred lines from the pearl millet association panel (PMiGAP, Sehgal et al., 2015) was analyzed for leaf ion content under irrigated and vegetative drought stress conditions during the 2021 and 2022 growing seasons in the field in Senegal (Affortit et al., in preparation).
The last ligulated leaf from the main tiller was sampled from three plants at 49 days after sowing in 2021 and 42 days after sowing in 2022. Leaves were washed in a 0.1% Triton X-100 solution, rinsed with deionized water, and stored in paper bags before drying at 60°C in an oven for three days. Leaf disks were sampled from dry leaves harvested from the field at around 5 cm from the ligule. Three leaf disks (5 cm diameter) from the three plants harvested in the same plot were pooled. Ion content was also studied in soils sampled at different locations and depths in the field at the beginning of the experiment. Plants remaining in the field were subjected to agro-morphological measurements (Day to flowering in particular; Affortit et al., in preparation).
Ion content in leaves and soils were measured at the ionomic platform of the University of Nottingham using Inductively Coupled Plasma Mass Spectrometers (ICP-MS) following a similar procedure to Danku et al (2013). Samples were analyzed for the content of a number of ions including As, Cr, Li, Na, Mg, P, Pb, Se, S, K, Ca, Mn, Fe, Co, Ni, Cu, Zn, Rb, Sr, Mo and Cd. For ion content in leaves, Best Linear Unbiased Estimators (BLUEs) were calculated for each ion using StatgenSTA package considering a resolvable incomplete block design for the analysis (Rossum, 2023).
Inbred lines were genotyped using tGBS Genotyping by Sequencing technology conducted with the restriction enzyme Bsp1286I (Freedom Markers, USA). Samples were sequenced using an Illumina HiSeq X instrument, and reads were aligned to the Cenchrus Americanus ASM217483v2 (Varshney et al., 2017) reference genome after debarcoding and trimming of reads. Lines with high missing data were removed and SNPs were filtered based on missing percentage (< 50%) and minor allele frequency (MAF > 5%). A total of 269,248 SNP were used in GWAS. The missing data were imputed based on a matrix factorization approach using the "impute" function of the LEA package (Frichot and François 2015).
GWAS was performed on BLUEs, Box-Cox transformed BLUEs to achieve normality and residues from the linear regression between individual ion concentrations and flowering time. Residuals of BLUEs were calculated to correct for potential confounding effect of phenology on ion content, as correlation between flowering time and ion content was observed. GWAS was performed using four mixed models: Latent Factor Mixed Models (LFMM, Frichot et al., 2013), Efficient Mixed-Model Association (EMMA, Kang et al., 2008), Bayesian information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK, Huang et al., 2019), and Compressed Mixed Linear Model (CMLM, Zhang et al., 2010). GWAS analyses were initially conducted separately for each year using BLUEs, and p-values for each method were then combined across two years independently using the Fisher's method (Cubry et al., 2020). False Discovery Rate (FDR) estimation was performed for each trait to correct for multiple testing. To calculate a p-values threshold based on the number of independent SNPs, a pruning process was implemented with Plink v1.9 to exclude highly correlated SNPs. These thresholds were used to select significant SNP-trait associations. Only associations identified by at least two methods in each individual GWAS and further validated by the Fisher's combined method were considered for further analyses.
In total, 161 significant associations were identified and delimited into 73 QTL regions associated with beneficial ions such as magnesium or potassium and toxic heavy metals such as cadmium.
The findings offer novel QTLs associated with grain ion content and candidate genes that are potentially valuable for breeding programs aiming at improving pearl millet leaf ion content and ultimately grain biofortification.
