Building a core collection of Senegalese pearl millet landraces
Using a heuristic approach with phenotypic and genetic data, we built a core collection of Senegalese pearl millets. This subset of landraces is well distributed across cultivated areas and captures wide genetic diversity and phenotypic variations in the pearl millet morphotypes. The spatial pattern of diversity is correlated with the genetic structure, as previously reported [8] and matches the cultivation areas of morphotypes across the Senegal.
Allogamous species often have fewer redundancies due to gene flow between individuals. The great diversity of pearl millet was associated with dynamic gene flow and admixture between wild and cultivated pearl millet [14]. This is certainly the explanation for the 22% of the core collection we defined, higher than in other self-pollinating species like wheat (Triticum aestivum) [15] or rice (Oryza sativa) [16]. Our results showed there are no duplicates of landraces in our core collection of Senegalese pearl millet. However, they confirm the previously reported two genetic pools [8–10]: an early-flowering pool subdivided into three clusters differentiated according to yield, spike thickness, tillering, and flag leaf length, and a late-flowering pool differentiated into three clusters according to yield, and tillering traits (Figure S2).
Differentiation of early- and late-flowering morphotypes may have occurred after the domestication of pearl millet (at least 4,900 years ago), as a consequence of a center of specialization developing across the Sahel belt. This hypothesis is supported by previous reports of migrations, exchanges and gene flow that led to wider genetic diversity as an adaptive mechanism [4, 14]. These findings and this assumption mean the clusters identified among the early-flowering Souna and late-flowering Sanio are genuine heterotic groups that can be used to breed for earliness and biomass, respectively.
Phenology and phenotypic traits featuring early- and late-flowering millets
Among the 16 quantitative agro-morphological traits evaluated at many sites and across many years, six were highly discriminating between early- and late-flowering morphotypes with high heritability (h2 ³ 0.5). There are two phenology features, i.e. heading and flowering time, that are correlated between themselves and also correlated with photoperiod sensitivity in pearl millet. The late-flowering morphotype is more sensitive to photoperiod. The phenology and photoperiod sensitivity are the two mechanisms help pearl millet to adapt to climate variability. Early flowering in pearl millet has been associated with a population adaptation mechanism [7], whereas photoperiod has been considered as an individual adaptation mechanism [5].
In parallel, four traits, biomass yield, plant height, nodal tillering and tillering, associated with the vegetative stage, featured in early- and late-flowering millets with high heritability (h2 ³ 0.5) (Table 1). Biomass, plant height, and nodal tillering contribute to stover yield, while tillering is also a yield component to maximize grain harvested. This is consistent with observations made by farmers who grow both morphotypes based on their own preferences and agro-systems. For example, some Senegalese farmers located in Niakhar and Bambey villages intercrop both early-flowering Souna and late-flowering Sanio in the rainy season to cope with long dry spells (personal communication).
Chromosome rearrangements as a source of diversity
Sequencing at the genome-wide scale revealed features at LG3 and LG6 in early-flowering millets, suggesting specific independent rearrangements at these regions for heading and flowering earliness. Chromosomes 3 and 6 might have undergone variations with breakpoints of 89.7 Mb and 68.1 Mb, respectively, large enough to look closely at gene insertion, deletion or synteny (Figure 3b-c). Indeed, chromosome rearrangements occur as a source of diversity through (i) standing rearrangements that play a role in evolutionary change and adaptive evolution, (ii) rearrangement of transposon elements, is a major mechanism behind the rearrangements we identified, could be catalysts for changes in expression of genes that are altered in association with rearrangements, (iii) variations in transcripts (de novo or level of expression via tandem duplication) [17]. In allogamous crops like pearl millet, extensive chromosomal rearrangements have occurred in its genome since it diverged from a common ancestor with related species [11, 13]. The two later possible sources of variation require more investigation at the gene expression level to test the assumption of independent rearrangements of chromosomes. Rather, we favor the explanation of standing rearrangements, as such variation would provide the genetic diversity needed for a population to rapidly adapt to different environments. This explanation supports evidence for rearrangements in the pearl millet genome revealed by synteny analysis with foxtail millet and sorghum [13].
Genes underlying allelic and trait diversity
The panel assembled is strongly structured between early-flowering Souna and late-flowering Sanio. Strongly structured diversity between early- and late-flowering millets has been considered to be a way of correcting for population structure and kinship. Correction through structure might be an obstacle to the identification of causal loci that are strongly differentiated between population. However, our results pinpoint some candidates for causal SNPs associated with the measured traits. Consequently, the power of GWAS is mainly identifying SNPs and phenotypic variation within each group. A panel with a more admixed genotype would have been preferable to identify phenotypic differences between early- and late-flowering millets. Biomass led to the identification of a large number of scattered SNPs with no standard peak in the P-value around the most significant markers. It is not clear why such a pattern was only observed for biomass. A more classical pattern clearly appeared in other traits, for example, plant height. Correlated genes, PgAAO and PgHK4, are involved in regulating the plant development response (plant height and nodal tillering) induced by abscisic acid [18] and in the synthesis of phylloquinone that is indispensable for photosynthesis [19], respectively. This suggests that during the vegetative phase, late-flowering Sanio millet produces more tillers, grows taller and captures more light through the hormonal and photosynthesis pathways than early-flowering Souna millet. Allelic variation of these genes might enable these specific phenotypes. On the other hand, during the transition to the flowering phase, the expression of genes may be at the origin of phenotype, as repressors or activators of signaling pathways leading to these features. Orthologs of the PPR gene play a role in delaying flowering by mediating the expression of several genes during plant growth, or by repressing genes involved in the transition to the reproductive phase [20]. The PgPPR gene we identified could be a putative candidate gene involved in the control of flowering time between early- and late-flowering millets.
Key features for breeding
The main traits that differentiate the morphotypes in the core collection are yield-related components, biomass, and flowering time. A north-south gradient of early- and late-flowering was identified across the Senegal [21]. There are more early-flowering morphotypes in the north of the country and more late-flowering morphotypes in the south. The special distribution of early- and late-flowering morphotypes also follows a rainfall gradient where the central part of Senegal was on average 500 to 600mm drier than the south (average 1,200mm) between the 1990s and 2014. Not to mention the fact that certain diseases including downy mildew are more prevalent in the agro-ecological zones that range from the south to the center than from the center toward the north of Senegal [22]. In some areas, early-flowering millet would cope better with drought, while late-flowering millet would adapt to cope with more humid environments. Therefore, harnessing the diversity based on flowering earliness to address climate variability in agro-ecosystems would be a step toward breeding early maturing varieties. Our results showed that the length of the flag leaf and the thickness of the middle axis of the panicles are phenotypes that differentiate the subsets of early-flowering morphotypes with high heritability. These may thus be advantageous traits to target under hotter and drier conditions. The nodal tillering, a character that is consistently associated with fodder yield in pearl millet [23], is also involved in the differentiation of subgroups of both early- and late-flowering morphotypes with high heritability. Based on our results, traits featuring early- or late-flowering millet could be targeted for breeding for dual-purpose varieties (yield and fodder). In summary, most of the characters that differentiate the genetic pools are involved in pearl millet performance in the agro-systems of Senegal.