Développer de nouvelles variétés de maïs pour une agriculture durable: une approche intégrée de la génomique à la sélection

AMAIZING supported the competitiveness of maize breeding in France and met societal demand for sustainability and quality. It developed innovative breakthroughs in germplasm characterization, breeding methods, ecophysiology of adaptation and underlying genetic factors, for the production of high yielding varieties with improved environmental values. One key to success was a strong public/private partnership between 24 key players of maize research and development in France, convening complementary expertise in Genetics, Genomics, Ecophysiology, Agronomy, Bioinformatics and Statistics. Most initial aims were reached at the expected level or beyond.

WP2 managed data storage and availability for "open science". A new tool was developed for collecting phenotyping data in the field and ThaliaDB and GNPis databases were improved. Genotypic and phenotypic datasets (field and platforms) were collected, organized and stored in information systems (ThaliaDB,  GNPiS, PHIS) and synthesized in a dataverse (https://data.inrae.fr/dataverse/Amaizing). A considerable effort was dedicated to data quality and validation via a FAIR approach, e.g. use of standards and ontologies developed in Phénome.Emphasis and Elixir. BioMercator software now makes it possible to perform visualization of numerous Genome-wide association studies (GWAS) and to place them in the context of genomic information (genes and functions). Training sessions and seminars publicized these tools.

WP3 characterized the genomic and epigenomic variation in inbred lines, to provide information on their origin and role in adaptation. It generated de novo whole genome assemblies and annotations for 7 maize lines, and built a pangenome for two of them. A transcriptomic gene atlas was also added and is  a highly valuable resource for new projects. Two software tools were developed and used to automatically build high-density linkage maps and detect copy number variation from mapping population data. Methylome variation analysis between maize lines highlighted extensive variation, which could be typed on larger panels. Cold showed limited impact on methylome. 22M SNPs were detected by whole genome sequencing lines of an American/European historical panel (67 lines, 18X coverage), leading to new insights on the routes of introduction, admixture and selective history of European maize. Whole genome sequencing of an association mapping panel (40 lines, 5X coverage) was passed to WP4 for imputation.

WP4 provided optimized genetic materials, genotypic data and statistical approaches for GWAS and Genomic Selection (GS). A total of 1500 flint and dent diverse temperate inbred lines were complemented by 1930 new Doubled-Haploid lines recombining different origins (admixture). All lines were evaluated per se and over 1200 hybrids were produced for WP5. All panels were genotyped with array and Genotyping by Sequencing strategies, reaching 1M SNP and 85k structural variants for the dent panel. A new method to estimate SNP frequencies in landraces provided insights into European maize origins and diversity in South-West France. Diversity, population structure and relatedness were analysed for all panels. Impact of population structure on GWAS power and GS was analysed and statistical methods were improved. New GWAS and GS models were built to account for hybrid mating designs and admixture. A mixed linear model software, MM4LMM, was developed to run these efficiently.

WP5 analysed maize responses to abiotic stresses and investigated heterosis. Four hybrid panels from WP4 were phenotyped in multi-site field experiments clustered in environmental scenarios (34,600 plots in 58 environments). Experiments in indoor phenotypic platforms allowing detailed measurements in well-controlled scenarios of water deficit, temperature (cold or hot) and nitrogen. Proteomics and metabolomics profiling was conducted in field and indoor experiments. Coupling these data with ecophysiological approaches highlighted traits, proteins or metabolites that are specific of environmental scenarios and allow predicting yield and searching for related polymorphisms. GWAS were performed and genomic prediction equations were calibrated using 1M SNP detected in WP4. Key regions identified were further transferred to WP6. Trait-associated SNPs and prediction equations were transferred to WP8 for testing in breeding programs, and datasets were transferred to WP7 for integrated studies.

WP6 confirmed the effect of 15 of the 30 genomic regions selected by WP5 and provided insights into underlying mechanisms. Major resources were established including 100 introgression populations, from which near isogenic lines (NILs) were extracted for 30 genomic regions, as well as 25,000 sequence-indexed insertional mutants and 100 independent transformation events for functional analyses. Major results were obtained using these tools. 5 chromosomic regions were validated for flowering time, and 3 for chilling tolerance (CT). A multi-omics analysis of NILs for 3 CT candidates provided mechanistic insight and additional candidates. Three regions were validated for drought with variable effects assessed in 16 fields across Europe representing 6 climatic scenarios. An integrated multi-omics approach demonstrated the role of a transcription factor in the response to water deficit. Four genomic regions were validated for heterosis of grain yield and components. Linked markers for all validated regions were provided to breeders (see WP8).

WP7 provided an integrated view of mechanisms underlying plant adaptation and performance across environmental scenarios, including climate change. Integration of traits collected in multi-site field experiments  and in phenotyping platforms showed that most QTLs had conditional effects on yield, in particular those related to responses to high temperature and water deficit. A considerable genetic gain was observed for grain yield in commercial hybrids from 1960 to 2016, in all tested environments. It was essentially linked to plant architecture and phenology, but largely left aside alleles and traits involved in responses to environmental conditions. A method was proposed to predict the effects of these alleles via the combination of phenomics, modelling and genomic prediction of responses to environmental condition. Process-based modelling highlighted how farmers currently exploit the genetic variability of flowering time to cope with temperature and water deficit across Europe. Simulations showed that their strategy may maintain maize yield in spite of climate change.

WP8 validated the results of other WPs in elite germplasm and sped up the integration of potential results in industrial projects. It also addressed socio-economic issues which could hinder the success of transferring the results into the industrial world. Major results transferred and used in WP8 have been the GS equations, important loci and markers  for drought, cold tolerance and flowering time, the different data produced in the project: genomic sequences, SNPs, structural variation phenotypic data, proteomics and metabolomics, expression data, specific traits extracted form platforms and the different methods developed by the WPs that have been proposed, explained or transferred to the private companies.

WP9 organized results dissemination towards the different stakeholders (scientists, seed companies, farmers, broad audience). 57 scientific papers were published and 168 oral presentations were given in congresses. A website (https://amaizing.fr/) was developed and updated regularly. A leaflet was distributed. A synthetic publication targeting a broad audience was written for Perspectives Agricoles. Nine training sessions on new methodologies were organized and 20 webinars were given on specific topics. Two international conferences were organized in 2015 (Montpellier) and 2018 (Saint-Malo). The final meeting was organized in 2021 with an international open conference and a stakeholder round table. These events were highly attended and appreciated.

AMAIZING has contributed important ressources, results and conceptual advances to better address climatic adaptation. Beyond reaching its initial goals, it has established a community and led to the development of new collaborative projects such as EVA maize, SEqOccin, Dromamed, and C4Future.