Dentre as diversas questões envolvendo a tecnologia duplo haploides (DH) em milho, uma que tem sido pouco discutida é a geração em que se deve induzir haploides, no caso, F1 ou F2. Destas, a F1 tem sido a mais utilizada. No entanto, o seu uso constante pode levar a perdas de ganhos genéticos, devido ao menor número de recombinações. Com isso, alguns autores aconselham o uso da F2 na indução, o que possibilitaria maiores ganhos em variabilidade genética. Desse modo, os objetivos foram verificar o efeito das gerações F1 e F2 na expressão do R1-navajo em germoplasma tropical, na eficiência relativa de cada uma das etapas da metodologia e na variabilidade genética das linhagens DH obtidas. Para isso, cinco fontes de germoplasma, em gerações F1 e F2, foram cruzadas com o indutor de haploidia tropicalizado LI-ESALQ. As sementes deste cruzamento foram agrupadas por meio do marcador R1-navajo em três classes: haploides putativos, diploides e inibidas. Após esta etapa, as sementes dos haploides putativos foram submetidas à duplicação cromossômica e as plantas duplicadas foram transplantadas a campo para a obtenção de linhagens DH. As unidades de sementes, plântulas e plantas em cada etapa da metodologia foram quantificadas para o estudo da eficiência relativa na obtenção de DH. Também foram coletadas amostras foliares das linhagens DH para genotipagem por meio de marcadores SNP (Single nucleotide polymorphism). As taxas de indução de haploides (HIR), sementes diploides (DSR) e inibidas (ISR) foram analisadas por meio de um modelo linear generalizado misto considerando distribuição logit multinomial. As eficiências relativas das fontes de germoplasma e gerações em cada etapa da metodologia DH foram estimadas por meio de porcentagem. Os marcadores SNP foram utilizados em estudos de diversidade genética, estrutura populacional e desequilíbrio de ligação. Os valores médios observados para as taxas de HIR, ISR e DSR foram, respectivamente, 1,23%, 23,4% e 75,2% para a geração F1 e 1,78%, 19,6% e 82,3% para a geração F2. O maior valor de HIR na F2 ocorreu devido à segregação dos genes que inibem o marcador R1-navajo durante a indução de haploides. Entretanto, apesar da geração F2 apresentar maior HIR, ela não deve substituir a F1, uma vez que se perde tempo com um ciclo adicional. A eficiência relativa na obtenção de linhagens DH apresentou o mesmo valor (0,4%) para as gerações F1 e F2, indicando que a escolha da geração não interfere na quantidade de DH produzidos. As estimativas dos parâmetros populacionais para as linhagens DH obtidas de geração F1 apresentaram valores para variância genética (VG) de 700,55, tamanho efetivo populacional (Ne) de 43,1, diversidade genética de Nei (GD) de 0,28 e conteúdo de informação polimórfica (PIC) de 0,23. Para a geração F2 as estimativas foram de VG=648,88, Ne=39,61, GD=0,26 e PIC=0,22. Os valores de desequilíbrio de ligação foram de 0,069 na geração F1 e de 0,067 na geração F2. Ou seja, as linhagens DH oriundas destas duas gerações apresentaram magnitudes semelhantes de diversidade genética e de desequilíbrio de ligação. O uso da geração F2 teoricamente permitiria obter maior variabilidade genética, devido à recombinação adicional. Entretanto, neste trabalho esta tendência não foi observada. Com isto, em milho tropical, recomenda-se o uso da geração F1 para a obtenção de DH, por apresentar o melhor balanço entre tempo e variabilidade genética. / Among the several questions involving doubled haploid technology (DH), one that has been little discussed is the generation in which one should induce haploids, in the F1 or F2. Overall, the F1 generation has been the most used. However, their constant use can lead to losses of genetic gains with the selection cycles due to the lower number of recombination. Thereby, some authors advise the use of F2 in inductions, which would allow greater gains in genetic variability. Thus, the objectives were to check the effect of the F1 and F2 generations on the expression of the R1-navajo in tropical germplasm, on the relative efficiency of each step of the methodology and the genetic variability of the DH lines obtained. For this purpose, five germplasm sources, in F1 and F2 generations, were crossed with the tropicalized haploid inducer LI-ESALQ. The seeds from this cross were grouped using the R1-navajo marker into three classes: putative haploids, diploid and inhibited. Then, putative haploid seeds were submitted to chromosome duplication, and the duplicate seedlings were transplanted to the field in order to obtain DH lines. Hence, seed, seedling, and plant at each stage of the methodology were quantified to study the relative efficiency in developing DH lines. Moreover, leaf samples from the D0 lines were collected for genotyping using SNP (Single nucleotide polymorphism) markers. Finally, the haploid inducer rate (HIR), diploid seed rate (DSR) and inhibited seed rate (ISR) were analyzed using a generalized linear mixed model considering multinomial logit distribution. The relative efficiency of germplasm sources and generation in each stage of the DH methodology was estimated by percentage. SNP markers were used in studies of genetic diversity, population structure, and linkage disequilibrium. The observed mean values of HIR, ISR and DSR were, respectively, 1.23%, 23.4% and 75.2% for the F1 generation and 1.78%, 19.6% and 82.3% for the F2 generation. The higher value of HIR in F2 occurred due to the segregation of genes, which inhibit the R1-navajo marker during haploid induction. However, in spite of the higher value of HIR for F2 generation, it should not replace F1 since time is lost with an additional cycle. The relative efficiency observed in the obtention of DH lines was the same value (0.4%) for generations F1 and F2, indicating that the choice between those generations does not interfere with the quantity of produced DH. Estimates of the population parameters for the DH lines obtained from F1 generation presented values for genetic variance (VG) of 700.55, effective population size (Ne) of 43.1, Nei\'s genetic diversity (DG) of 0.28 and polymorphic information content (PIC) of 0.23. For F2 generation the estimates were VG = 648.88, Ne = 39.61, GD = 0.26 and PIC = 0.22. Linkage disequilibrium values were 0.069 in the F1 generation and 0.067 in the F2 generation. Thus, DH lines from these two generations showed similar values of genetic diversity and linkage disequilibrium. Nonetheless, the use of the F2 generation theoretically would allow obtaining greater genetic variability, due to the additional cycle of crossing-overs. However, this trend was not observed in this study. Thus, in tropical maize, the use of the F1 generation to obtain DH lines is recommended, because it performs the best balance between time and genetic variability.

Chez le maïs, la création de nouvelles variétés repose largement sur la technologie des plantes Haploïdes Doublées. Celle-ci s’appuie sur l’utilisation du pollen de lignées particulières, dites «inductrices d’haploïdes», qui déclenchent un processus de gynogenèse in vivo. Ce mécanisme permet la production d’embryons haploïdes contenant uniquement les informations génétiques maternelles. Après doublement des stocks chromosomiques des plantules haploïdes générées, il est possible d’obtenir des lignées fixées en seulement quelques générations, réduisant ainsi considérablement les schémas de sélection. Bien que les déterminismes génétiques à l’origine de la gynogenèse in vivo soient restés énigmatiques pendant près de soixante ans, nos travaux ont permis le clonage positionnel et l’identification du gène majeur causal de l’induction d’haploïdes. Ce gène, exprimé spécifiquement dans les grains de pollen, code pour une phospholipase A2, nommée NOT LIKE DAD (NLD). Il s’avère que chez l’ensemble des lignées inductrices le gène NLD présente une insertion de quatre paires de bases, conduisant à la formation d’une protéine tronquée responsable de la gynogenèse in vivo. Cette mutation n’entraine pas un défaut spatial ou temporel de l’expression du gène mais une délocalisation de la protéine. La protéine sauvage localise au niveau de l’unité mâle germinative, sur la membrane d’origine végétative entourant les deux cellules spermatiques, alors que la protéine tronquée perd cette localisation. Des approches pharmacologiques et de mutagenèses dirigées ont permis de révéler un adressage de NLD à la membrane grâce à des ancrages lipidiques et des interactions électrostatiques entre la membrane et la protéine. En parallèle, il a pu être observé que la troncation de NLD, en plus de conduire à la formation d’haploïdes maternels, entraînait de nombreux phénotypes de grains anormaux. Le génotypage des produits de fécondations (albumens et embryons) résultant de croisements avec des lignées inductrices ont permis de mettre en évidence l’existence d’anomalies chromosomiques paternelles, pouvant expliquer ces différents phénotypes de grains anormaux. L’ensemble de ces résultats indiquent que la phospholipase NLD intacte est nécessaire pour le bon déroulement de la reproduction sexuée. / In maize, plant breeding mainly relies on double haploid technology. The main principle of this technology is the use of special lines called “inducer lines”, which induce in vivo gynogenesis. This mechanism allows haploid embryos production which only contain maternal genetics information. After chromosome doubling of the haploid seedlings, it is possible to obtain fixed lines in only few generations, thus considerably reducing breeding cycles. Although genetic determinants behind in vivo gynogenesis remained enigmatic for nearly sixty years, our work allowed the positional cloning and the identification of the major gene controlling haploid induction. This gene, specifically expressed in pollen grain, encodes for a phospholipase A2 called NOT LIKE DAD (NLD). It turns out that in all the inducer lines the NLD gene has a 4-bp insertion, leading to the formation of a truncated protein responsible for gynogenesis in vivo. This mutation did not result in a spatial or temporal defect in the expression of the gene but in the delocalization of the protein. The wild type protein localizes at the male germ unit on the vegetative membrane surrounding the two sperm cells, while the truncated protein loses this location. Pharmacological and mutagenic approaches revealed that NLD is targeted to the membrane through lipid anchoring and electrostatic interactions between membrane and protein. In parallel, it has been observed that the truncation of NLD; in addition to leading to the formation of maternal haploids; led to many pleiotropic phenotypes of abnormal kernels. The genotyping of fertilization products (endosperms and embryos), shows the existence of paternal chromosomal anomalies, which can explain these different phenotypes of abnormal kernels. All these results indicate that the intact NLD phospholipase is therefore necessary for the proper course of sexual reproduction.