Adaptive introgression in Pearl Millet – PEMILADAPT

Beyond its agricultural importance, the choice of pearl millet (Pennisetum glaucum) as a study model rests on several advantages. This cereal is among the most drought-tolerant, making it a highly relevant system for analyzing mechanisms of adaptation to extreme climates. Furthermore, the documented gene flow between wild and cultivated forms provides a natural framework to study introgression and quantify its evolutionary and agronomic effects.

A critical prerequisite for the project was the improvement of the reference genome, essential to establish robust links between introgressed regions and structural variants. We combined third-generation sequencing (ONT) with optical mapping to produce high-quality chromosome-level assemblies. This strategy enhanced genome contiguity and chromosomal anchoring, enabling reliable identification of structural rearrangements.

Detecting structural variants at the population scale posed a major challenge, as approaches based on complete individual assemblies were prohibitively expensive for large sample sizes. We therefore tested and validated an alternative approach based on so-called local PCA, which allows detection of regions with very low recombination that may correspond to chromosomal inversions.

To achieve fine-scale mapping of introgression, over 100 complete genomes were sequenced at 20–30X coverage. Data were analyzed using classical population genomics statistics, complemented by machine learning approaches aimed at identifying atypical diversity patterns along the genome. The goal was to generate demographic simulations based on the domestication history of pearl millet to train Random Forest algorithms capable of discriminating different evolutionary signatures.

However, preliminary results revealed extensive non-recombining and introgressed plateaus, rendering these approaches unsuitable. The strategy was then redirected using methodological developments from a co-supervised PhD project within a related study, in order to identify the most robust and appropriate approaches for detecting adaptive introgression in this system.

Finally, phenotypic and functional validation relied on genome–phenotype association studies (GWAS) and two years of field trials, despite constraints imposed by the COVID-19 pandemic and political instability in the Sahel. Phenology and reproductive success measurements allowed evaluation of the adaptive impact of regions identified as introgressed. However, the planned transcriptomic analysis could not be performed due to a lack of dedicated personnel, representing a recognized limitation of the project.

The project has led to major advances in pearl millet genomics and, more broadly, in understanding adaptive mechanisms in cultivated plants. One key result concerns the production of a new version of the pearl millet reference genome, published in G3 (Salson et al., 2023). Compared to the previous version, approximately 200 Mb of additional sequences were integrated and anchored at the chromosomal level. Gene completeness now reaches 98.4%, and the proportion of unknown bases has been reduced from 13% to less than 0.3%. This new assembly provides a critical foundation for the analysis of structural variants and introgression.

From an evolutionary biology perspective, the project revealed the existence and persistence of very large low-recombination regions covering approximately 17% of the genome, a result published in Nature Communications (Salson et al., 2025). While such regions have only recently been suggested in other species, their prevalence, exceptional size, and especially their maintenance had never been empirically demonstrated. Our analyses show that these regions maintain high genetic diversity despite low recombination rates and an accumulation of deleterious variants. Many of these regions are maintained under positive selection in the heterozygous state, providing the first empirical demonstration of pseudo-overdominance, a mechanism previously supported only by theoretical models and simulations.

A close link between these regions and introgression was also established. Analysis of a region on chromosome 3 revealed highly divergent haplotypes derived from wild parental introgression. One of these introgressed haplotypes was nearly lethal in the homozygous state but persisted in the population due to a marked selective advantage in heterozygotes. This case illustrates, for the first time, how wild introgression can generate and maintain adaptive diversity through pseudo-overdominance.

Finally, our most recent work (Duranton et al., in preparation) highlights strong heterogeneity of introgression across the cultivated genome, characterized by “deserts” and “islands” of wild fragment enrichment. Two complementary dynamics emerge: localized adaptive introgression, corresponding to the introduction of specific beneficial alleles in certain regions, and compensatory introgression, contributing to the reduction of genetic load in low-recombination regions. Genome-wide association analyses indicate that regions involved in the domestication syndrome, notably traits related to the panicle, remain largely resistant to introgression. In contrast, regions associated with environmental stress responses show enrichment in introgressed segments, confirming the role of wild parents as a potential lever for climate resilience.

The PEMILADAPT project has profoundly advanced our understanding of pearl millet adaptation by highlighting the central role of wild introgression and pseudo-overdominance in maintaining genomic diversity. These findings open several scientific and applied perspectives for the coming years.

A first avenue concerns the development of an integrative pan-genomic approach. Our analyses have highlighted the limitations of a single reference genome for capturing the structural diversity of pearl millet, particularly inversions and deletions. Constructing a pan-genome that incorporates both wild and cultivated compartments would allow identification of alleles currently absent from available references and better characterization of specific genomic regions, notably highly heterozygous pericentromeric regions where pseudo-overdominance appears particularly active.

A second direction involves deepening the functional genomics of introgressed regions. The project identified several wild-derived genomic segments associated with fitness-related traits such as flowering time or tillering. The underlying molecular mechanisms remain to be clarified. Transcriptomic analyses, notably via RNA-seq on reciprocal hybrids, would enable exploration of the role of cis-regulatory polymorphism and testing of the hypothesis that adaptation may partially rely on changes in gene expression.

Furthermore, the results open an important field of research in evolutionary biology. PEMILADAPT provides one of the first empirical demonstrations of pseudo-overdominance in a domestication system. However, the conditions for its emergence and stability remain poorly understood. Future work should combine empirical analyses with demographic simulations integrating gene flow, domestication-related bottlenecks, and recombination landscape.

Finally, these findings have direct implications for crop improvement. The project identified low-recombination genomic regions that may limit the introgression of exogenous diversity in breeding schemes. Characterizing these regions will help guide pre-selection strategies and optimize the use of wild diversity in pearl millet improvement programs, a species expected to play an increasingly important role in adapting agriculture to climate change.

Understanding how species adapt is paramount to biodiversity. Species/populations adaptive potential largely depends on their genetic variability, which relies either on standing genetic variation or on de novo mutation. However, if rates of environmental changes are too rapid, occurrence of de novo mutations might be too slow of a process. When gene flow between species/populations occurs through recurrent backcrosses over generations, allowing foreign DNA to become part of recipient species’ genetic pool, it is named genetic introgression. If the genetic introgression from donor species/population increases the fitness of the recipient species/population then it is called “Adaptive Introgression" (AI). Despite the occurrence of hybridization in nature, relatively limited evidences supporting AI have been gathered until recently. Striking evidences of AI were found between archaic and modern humans. For crops, wild relatives represent a reservoir of adaptations that could refuel crops genetic diversity. With the advance of genomics, it is becoming evident that wild-crops gene flow was far more complex and protracted than previously considered. A compelling example of the potential adaptive outcome of introgression is the adaptation to altitude acquired by highland maize landraces from wild populations. Nonetheless, conclusive evidence for AI is restricted to a limited number of cases. For AI to take place, we will first need crosses and introgression to be successful. When selection against hybrids is weak and backcrosses are possible, genomic heterogeneity in terms of permeability to gene flow will affect introgression rates. Gaps knowledge on the genomic heterogeneity to introgression need to be fulfil particularly for crops. In addition, most evidence for AI comes from hard-selective sweeps, while soft-selective sweeps might be the dominant mode of adaptation. The main objective of the PEMILADAPT project is to understand how introgression from wild relatives can favour crops adaptation. The PEMILADAPT projects aims at responding the following specific questions:
1) How frequent wild introgression is and how is it distributed along the cultivated genome?
2) What modes (hard vs. soft) of selection are acting on introgressed alleles?
3) What are the functional implications of wild introgression in the cultivated genome?
To answer the above questions, we will focus on the pearl millet crop (Pennisetum glaucum). Pearl millet is the sixth most important cereal grain worldwide and it is a key cereal in arid and semi-arid regions where it is a staple food for over 90 million small farmers. Wild and cultivated pearl millet are found in sympatry and gene flow between the two forms is well recognized. By taking the opportunities coming from cutting edge technologies for high-throughout genomic sequencing and for artificial intelligence for powerful machine learning approaches, the project PEMILADAPT is expected to be at the leading front of evolutionary research questions. In addition to the fundamental knowledge on evolution that PEMILADAPT will produce, it may have significant impacts for on–going worldwide environmental challenges and sustainable development. Understanding selection forces and genomic landscape of introgression will likely help breeders in improving efficiency of their breeding strategies to adapt future varieties to on-going climate changes.