Manipulation de la recombinaison chez une plante cultivée, le riz.L’accroissement prévisible de la population mondiale ainsi que les conséquences du changement climatique obligent les sélectionneurs à créer de nouvelles variétés plus productives et plus résilientes. Les nouvelles combinaisons d’allèles favorables sont issues de la recombinaison génétique entre chromosomes homologues dont le siège est la prophase de première division de méiose. De récentes avancées chez la plante modèle Arabidopsis ont montré que l'inactivation de certains gènes permet de manipuler la méiose pour abolir ou au contraire augmenter très significativement la recombinaison. Les mécanismes de la méiose étant relativement bien conservés chez les eucaryotes, l’objectif de cette thèse était de transposer ces avancées chez une plante cultivée importante, le riz. Abolir la recombinaison méiotique permettrait de propager de façon clonale par grain des formules variétales hybrides F1 dont le rendement est de 20% supérieur à celui des lignées pures chez le riz mais dont les semences restent peu utilisées par les riziculteurs de subsistance. Les travaux réalisés dans une première partie de la thèse ont montré que le cumul de trois mutations Ososd1, pair1 et Osrec8, permettait d’obtenir des gamètes clonaux diploïdes mâles et femelles. Le phénotype apoméiotique obtenu, appelé MiMe (Mitosis instead of meiosis) chez Arabidopsis, peut être utilisé pour tester différentes stratégies d’induction de la parthénogenèse afin de produire des grains formant des plantes diploïdes clonales apomictiques. Une optimisation du mécanisme permettrait d'envisager l'utilisation de l'apomixie pour fixer l'hétérosis dans les semences hybrides F1. Par ailleurs, une augmentation globale ou locale de la recombinaison méiotique est recherchée car elle permettrait de diminuer la taille des populations de sélection et de réduire la taille des segments chromosomiques introduits dans les variétés élite de riz. Nous avons montré dans une seconde partie, que la mutation du gène OsRECQl4 codant pour une hélicase permet d'augmenter le taux de recombinaison d'un facteur de 3,3 fois faisant passer la taille de la carte génétique de 1670 cM à 5538 cM sans affecter la fertilité de la plante ni le déroulement de la méiose. Chez les plantes affectées dans la fonction d’une autre hélicase, OsFANCM, le taux de recombinaison a été également augmenté mais dans une moindre mesure (x 2,2). L’augmentation de la recombinaison s’opère sur l'ensemble des bras chromosomiques sauf au niveau des centromères. Ces résultats confirment ceux obtenus chez A. thaliana qui ont montré le rôle de régulateur négatif des crossing-overs (CO) des protéines RECQ4 et FANCM. La combinaison en cours de ces mutations entre elles ou avec celle affectant l’AAA-ATPase FIGL1 permet d’espérer une augmentation de la recombinaison encore supérieure. Ces résultats ouvrent la voie à l'utilisation des gènes anti-COs pour augmenter de façon globale le nombre de recombinants dans les croisements chez le riz et sans doute chez les autres céréales. Pour offrir une possibilité concrète aux sélectionneurs d'utiliser les gènes anti-CO, nous avons montré que la technologie CRISPR/cas9 permet d'éteindre l'expression de OsFANCM OsRECQl4 et OsFIGL1. / Manipulation of recombination in a crop, rice.The forecasted increase of world population as well as the consequences of global climate change oblige plant breeders to develop new varieties that are both more productive and resilient. Novel combinations of favourable alleles are generated through genetic recombination between homologous chromosomes, which occurs during the prophase of the first division of meiosis. Recent advances in the model plant Arabidopsis have demonstrated that the inactivation of some genes allows meiosis manipulation resulting in either an abolishment or in contrast, a significant enhancement of meiotic recombination. The meiosis mechanisms being relatively conserved across eucaryotes, the overall objective of this thesis was to transfer these advances to a crop of crucial importance, rice. To abolish meiotic recombination would allow the clonal propagation by seeds of F hybrids, which exhibit a 20% yield enhancement compared to that of pure lines in rice but remain rarely used in subsistence farming. In a first part, we showed that rice plants cumulating 3 mutations inactivating Ososd1, pair1 and Osrec8, formed clonal diploid male and female gametes. This apomeiotic phenotype, called MiMe (Mitosis instead of meiosis) in Arabidopsis, can serve as material to assay several strategies of parthenogenetic induction that would result in seed forming diploid clonal plants. Further optimization of the mechanisms would allow the use of apomixis to fix heterosis in hybrid seeds. Global and local enhancement of recombination is another desirable goal since it would allow a reduction in breeding population size and a downsizing of the introgressed chromosomal segments in elite plant materials. In a second part, we showed that mutation in the DNA helicase gene OsRECQl4 conducted to a 3.3 fold increase of recombination and inflated the genetic map size from de 1670 cM to 5538 cM, without altering plant fertility nor meiosis progression. Plants altered in a second DNA helicase, OsFANCM, exhibited a more modest 2.2 fold recombination enhancement. Recombination increase operated along the whole chromosome arms except at the centromere level. These results confirms the negative regulator role of RECQ4 and FANCM on crossing overs (CO), previously reported in Arabidopsis. On going combination of these mutations together with that altering the l’AAA-ATPase FIGL1 should conduct to an even higher recombination enhancement. These results pave the way to the use of anti-CO genes to enhance recombinant recovery in crosses of rice and possibly of other cereals. To provide breeders with a workable anti-CO system, we eventually showed that the CRISPR/Cas9 technology can be used to abolish OsFANCM, OsRECQl4 and OsFIGL1 expression.

La recombinaison méiotique n'est pas distribué de manière aléatoire le long des chromosomes, mais est caractérisée par des domaines froids et chauds qui limitent la diversité génétique transmise par les gamètes. Cependant, le profil de la recombinaison méiotique peut être modifiée, étant donné que la fusion de l’ endonucléase Spo11 au domaine de liaison à l'ADN de Gal4 est suffisante pour favoriser la formation des cassures double brin (CDB) et la recombinaison à proximité des sites de liaison de Gal4, dans la levure et dans les souris. Ici, dans la levure Saccharomyces cerevisiae, nous avons étudié l'effet de la fusion de Spo11 à 8 protéines de liaison à l'ADN lors de la méiose. Comme modules de ciblage, nous avons utilisé des facteurs de transcription de levure et des protéines artificiels de liaison à l'ADN (TALEs et ZFs), qui sont apparus comme des outils efficaces pour faire varier la position et / ou le nombre de sites ciblés. Lors de l'expression de chacun des fusions Spo11, nous avons examiné la progression de la méiose, la formation des CDB dans les sites naturels et ciblées ainsi que le niveau relatif de la recombinaison méiotique. Ce travail dans l’organisme modèle levure ouvre de nouvelles voies pour modifier la recombinaison méiotique chez d'autres organismes, tels que des mammifères et des plantes, pour augmenter la diversité génétique dans les sites d'intérêt et disséquer l'information génétique, en surmontant les limitations dues à la liaison génétique. / Meiotic recombination is not randomly distributed along the chromosomes, but is characterized by hot and cold domains that limit the genetic diversity transmitted by the gametes. However, the recombination profile can be modified, since the tethering of Spo11 endonuclease, upon fusion to the Gal4 DNA-binding domain, is sufficient to enhance DSB formation and recombination near several Gal4 consensus binding sites, in yeast and in mouse. Here, in the yeast Saccharomyces cerevisiae, we studied the effect of Spo11 fusions to 8 different DNA-binding proteins during meiosis. As targeting modules, we used yeast full-length transcription factors and artificial DNA-binding modules (TALEs and ZFs), which emerged to be efficient tools to vary the location and /or the number of targeted sites. Upon expression of each of the Spo11 fusions, we examined meiotic progression, DSB formation at natural and targeted sites as well as the relative level of meiotic recombination. This work in the yeast model opens new avenues to modify meiotic recombination in other organisms, such as mammals and plants, to boost genetic diversity at sites of interest and to dissect the genetic information, overcoming the restrictions due to the genetic linkage.

Chez les organismes à reproduction sexuée, les gamètes (cellules sexuelles) sont produits par un processus comprenant deux divisions successives appelé méiose. Durant la première division, la recombinaison méiotique permet un contact physique et un échange de matériel génétique entre les chromosomes homologues. Elle résulte de la réparation, par recombinaison homologue, de cassures double-brin de l’ADN générées par la protéine SPO11 au début de la prophase de la première division. Chez les mammifères, les évènements de recombinaison se situent dans des régions de 1-2 kb appelées points chauds de recombinaison. La protéine PRDM9, qui contient un domaine PR/SET et des doigts de zinc, détermine la position des points chauds en ciblant des séquences spécifiques d’ADN par ses doigts de zinc. Son domaine PR/SET porte une activité lysine méthyltransférase, corrélée avec un enrichissement de H3K4me3 au niveau des points chauds, dans les spermatocytes.Les objectifs de mon travail étaient de caractériser l’activité catalytique de PRDM9 et d’étudier son rôle dans l’initiation de la recombinaison chez la souris. La structure cristallisée du domaine PR/SET de PRDM9 en complexe avec un peptide de l’histone H3 nous a permis de montrer que ce domaine adopte une structure similaire aux domaines SET canoniques portés par d’autres méthyltransférases, et d’identifier des résidus clés pour son activité. Nous montrons que le domaine PR/SET de PRDM9 méthyle in vitro non seulement H3K4, mais aussi H3K9 et H3K36. Nous confirmons in vivo la triméthylation de H3K36 dépendante de PRDM9 dans les spermatocytes. Utilisant deux allèles différents de PRDM9, Prdm9b et Prdm9wm7, qui activent des points chauds différents grâce à leur spécificité de séquence, nous avons généré des lignées de souris exprimant des allèles mutés du domaine PR/SET dont l’activité catalytique est abolie, Prdm9wm7G278A ou Prdm9wm7Y357F. La protéine mutante PRDM9wm7Y357F se fixe à ses cibles, mais n’y permet in vivo ni la triméthylation de H3K4, ni celle de H3K36. Enfin, nous montrons que l’activité catalytique de PRDM9 est requise pour promouvoir la recombinaison aux points chauds. Chez les souris exprimant uniquement un allèle Prdm9 muté, les spermatocytes présentent des défauts d’appariement des chromosomes homologues et de réparation des cassures double-brin de l’ADN, ainsi qu’un arrêt de la progression en méiose en milieu de prophase I, phénotype similaire à celui de la souris KO pour Prdm9 (Prdm9-/-). L’ensemble de nos résultats met en évidence le rôle primordial de l’activité méthyltransférase de PRDM9 pour la détermination des sites de recombinaison méiotique et plus généralement pour la progression de la méiose et finalement la formation de gamètes chez la souris. / In sexually reproducing organisms, gametes are produced by a process comprising two successive division, called meiosis. During the first division, meiotic recombination enables a physical contact and an exchange of genetic material between homologous chromosomes. Meiotic recombination results from the repair, by homologous recombination, of programmed DNA double-strand breaks (DSBs) catalyzed by the SPO11 protein at the beginning of prophase I. In mammals, recombination events are localized in 1 to 2 kb-long regions called recombination hotspots. PRDM9, a PR/SET domain and zinc finger-containing protein, determines hotspot localization by targeting specific DNA sequences through its zinc finger array. Notably, PRDM9 PR/SET-domain possesses an H3K4 methyltransferase activity, while PRDM9-dependent H3K4me3 enrichment is found at hotspots in spermatocytes.We aimed at characterizing PRDM9 methyltransferase activity and studying its role in meiotic recombination initiation in mouse. The crystal structure of PRDM9 PR/SET domain, which we generated in complex with a histone H3 peptide, shows that this domain adopts a similar topology to that of classical SET domains and allowed us to identify key residues for its catalytic activity. PRDM9 PR/SET domain catalyzes not only mono-, di- and trimethylation of H3K4, but also of H3K9 and H3K36. We confirmed PRDM9 dependent H3K36 trimethylation in spermatocytes. Taking advantage of the distinct DNA binding specificity of two Prdm9 alleles, Prdm9b and Prdm9wm7, each activating its own set of hotspots, we generated transgenic mouse lines expressing either Prdm9wm7G278A or Prdm9wm7Y357F mutant allele together with the endogenous wild-type Prdm9b allele. Both G278A and Y357F mutations abolish PRDM9 catalytic activity. We show that PRDM9wm7Y357F binds normally to its genomic targets, but is not able to promote H3K4 nor H3K36 trimethylation at these sites. In addition, PRDM9wm7Y357F does not promote recombination at one Prdm9wm7-dependent hotspot, showing that PRDM9 catalytic activity is required for promoting recombination at hotspots. In mice expressing only the mutant allele (Prdm9wm7G278A or Prdm9wm7Y357F), spermatocytes display defects in homologous chromosome synapsis and DSBs repair, as well as an arrest of meiosis at the mid-prophase I. This phenotype is similar to that of Prdm9 KO mice. Overall, our results demonstrate the role of PRDM9 methyltransferase activity in determining recombination hotspots and more generally for meiotic progression and gametes formation.

While the introduction of intracytoplasmic sperm injection (ICSI) has revolutionized the treatment of male infertility, concerns have been raised regarding the risk of chromosomal abnormalities in pregnancies derived from ICSI. Studies on sperm from infertile men have suggested that this population may produce higher rates of aneuploid sperm. Thus, we hypothesized that defects in early meiotic events may contribute to both male infertility and the production of aneuploid sperm. We used immunofluorescent techniques to observe the synapsis and recombination of chromosomes during meiosis, and fluorescent in-situ hybridization (FISH) to assess sperm aneuploidy. We analyzed testicular tissue from thirty-one men (10 fertile and 21 infertile men). We observed that ~36% (5/14) of men with impaired spermatogenesis displayed reduced genome-wide recombination. When all men were pooled, we observed an inverse correlation between the frequency of sex chromosome recombination and XY disomy in the sperm. We combined immunofluorescent and FISH techniques to study recombination patterns on chromosomes 13, 18 and 21 in fifteen men (5 fertile and 10 infertile men). Four of the infertile men displayed altered recombination distributions on at least one of the chromosome arms studied. Finally, we examined early meiotic events in two biopsies from an azoospermic t(8;13) carrier. While global recombination rates were not altered, recombination frequencies were reduced specifically on the rearranged chromosomes. Asynapsed quadrivalents were observed in 90% and 87% of pachytene nuclei from the first and second biopsies, respectively, and were frequently associated with the sex chromosomes. BRCA1 and γH2AX, two proteins implicated in meiotic sex chromosome inactivation, localized along asynapsed regions regardless of whether or not they were associated with the sex chromosomes, suggesting that regions of autosomal chromosomes that fail to synapse undergo transcriptional silencing in humans. In summary, we observed that a subset of infertile men display alterations in the number and position of meiotic crossovers, which may contribute to both infertility and an increased risk of sperm aneuploidy. The fidelity of synapsis is also a critical factor in determining the outcome of gametogenesis in humans, as the transcriptional inactivation of asynapsed regions may silence meiotic genes, leading to meiotic arrest and infertility. / Medicine, Faculty of / Obstetrics and Gynaecology, Department of / Graduate

Polar body emission (PBE), a form of asymmetric division, occurs twice during vertebrate oocyte maturation and is required to produce a haploid egg for sexual reproduction. Our lab elucidated parts of the mechanism that regulates PBE in Xenopus laevis oocytes. Cdc42 and RhoA, two GTPases, were shown to mediate membrane protrusion and the contractile ring, respectively. It is believed that cdc42 is mediating the protrusion by regulating actin polymerization. However, it is not clear what upstream signaling pathway regulates cdc42 activation during PBE. One possibility is calcium signaling, which occurs at fertilization, and is required for second PBE. Interestingly, the fertilization calcium transient also regulates cortical granule exocytosis/membrane retrieval, a process that also involves cdc42-mediated actin assembly. Furthermore, active cdc42 and RhoA are found in non-overlapping concentric zones in single-cell wound healing; their activation requires calcium signaling. To determine possible calcium transients during polar body emission, we employed the calcium-binding C2 domain of PKCβ in live cell imaging. Surprisingly, the most prominent C2 signal was seen after cdc42 activation and membrane protrusion. Co-localization experiments indicated that the C2 signal appeared at the cortical area marked by the contractile ring component anillin, and after partial constriction of the ring. Injection of the calcium chelator, dibromo-BAPTA, abolished the C2 signal, suggesting that it is indeed depicting a calcium transient. Dibromo-BAPTA injection also inhibited polar body abscission, as assessed by a novel abscission assay developed in our lab. We have for the first time detected a calcium signal during PBE that is essential to the last step of cytokinesis—abscission.

La méiose ovocytaire comprend une succession de deux divisions cellulaires, sans phase intermédiaire de réplication de l'ADN, permettant l'haploïdisation du gamète femelle en vue de la fusion des génomes parentaux lors de la fécondation. Le caractère fortement asymétrique de ces divisions permet l'expulsion du matériel génétique surnuméraire, dans de petits globules polaires, tout en conservant l'essentiel des ressources cytoplasmiques qui seront nécessaires au développement précoce de l'embryon. De nombreuses études réalisées sur l'ovocyte de souris ont mis en évidence les capacités intrinsèques du gamète à rompre sa symétrie en positionnant son fuseau de manière excentrée à proximité du cortex. En se positionnant de la sorte le fuseau induit, via un gradient de Ran-GTP porté par les chromosomes, une polarisation du cortex ovocytaire qui permettra de restreindre le site d'émission des futurs globules polaires. Cette polarisation se caractérise notamment par une forte accumulation de filaments d'actines dépendante du facteur de nucléation Arp2/3. Nos travaux nous ont permis de mettre en évidence le rôle de Cdc42-GTP, via l'activation de N-WASP, comme intermédiaire entre le gradient de Ran-GTP et la polymérisation polarisée des filaments d'actine. Nous nous sommes également intéressés à la localisation des protéines ERM (Ezrin Radixin Moesin), connues pour favoriser la formation des microvillosités membranaires. Dans l'ovocyte, les microvillosités et les ERM sont toutes deux exclues du cortex polarisé et nous avons pu démontrer le rôle de Ran-GTP dans ce processus. Enfin, nous avons étudié la localisation du réseau d'acto-myosine cortical lors de la deuxième division méiotique qui nécessite la rotation du fuseau de l'ovocyte de souris. Nos résultats révèlent l'existence de deux sous-populations de myosine 2 corticale, l'une dépendante de la chromatine (Ran-GTP/Cdc42-GTP) et l'autre dépendante du fuseau central (Ect2/RhoA). / Oocyte meiosis is accomplished through two successive rounds of cellular divisions, without DNA replication, allowing for gamete haploidization necessary for parental genome fusion after fertilization. These divisions are highly asymmetric and allow extra-DNA expulsion, in small polar bodies, while retaining most of the cytoplasmic resources needed for early embryo development. Studies in mouse oocyte have demonstrated the capabilities of the gamete to autonomously break his symmetry by positioning the spindle near the cortex. By doing so, the spindle is able to induce a cortical polarization that is dependent on a Ran-GTP gradient emanating from the chromosomes. This polarization will be necessary for delimiting extrusion sites of the future polar bodies. A polarized accumulation of Arp2/3 actin filaments is one of the most evident features of oocyte polarization. We have shown that polarization of Cdc42-GTP, trough N-WASP activation, is an essential intermediate between Ran-GTP and the polarized polymerization of actin filaments. We also investigated ERM (Ezrin Radixin Moesin) proteins localization that are known to promote microvilli assembly. According to our data, microvilli and ERM are excluded from the polarized cortex in a Ran-GTP dependent manner. Finally, we studied cortical acto-myosin dynamics during the second meiotic division which requires spindle rotation. We demonstrated the existence of two cortical myosin 2 sub-populations which depend either on chromosomes (Ran-GTP/Cdc42-GTP) or on the central spindle (Ect2/RhoA).

Meiosis is necessary for sexual reproduction in eukaryotes. Genetic recombination between non-sister homologous chromosomes is needed in most organisms for successful completion of the first meiotic division. Proteins that function during meiotic recombination have been studied extensively in model organisms. However, less is known about the evolution of these proteins, especially among protists. We searched the genomes of diverse eukaryotes, representing all currently recognized supergroups, for 26 genes encoding proteins important for different stages of interhomolog recombination. We also performed phylogenetic analyses to determine the evolutionary relationships of gene homologs. At least 23 of the genes tested (nine that are known to function only during meiosis in model organisms) are likely to have been present in the Last Eukaryotic Common Ancestor (LECA). These genes encode products that function during: i) synaptonemal complex formation; ii) interhomolog DNA strand exchange; iii) Holliday junction resolution; and iv) sister-chromatid cohesion. These data strongly suggest that the LECA was capable of these distinct and important functions during meiosis. We also determined that several genes whose products function during both mitosis and meiosis are paralogs of genes whose products are known to function only during meiosis. Therefore, these meiotic genes likely arose by duplication events that occurred prior to the LECA. The Rad51 protein catalyzes DNA strand exchange during both mitosis and meiosis, while Dmc1 catalyzes interhomolog DNA strand exchange only during meiosis. To study the evolution of these important proteins, we performed degenerate PCR and extensive nucleotide and protein sequence database searches to obtain data from representatives of all available eukaryotic supergroups. We also performed phylogenetic analyses on the Rad51 and Dmc1 protein sequence data obtained to evaluate their utility as phylogenetic markers. We determined that evolutionary relationships of five of the six currently recognized eukaryotic supergroups are supported with Bayesian phylogenetic analyses. Using this dataset, we also identified ten amino acid residues that are highly conserved among Rad51 and Dmc1 protein sequences and, therefore, are likely to confer protein-specific functions. Due to the distributions of these residues, they are likely to have been present in the Rad51 and Dmc1 proteins of the LECA. To address an important issue with the gene inventory method of scientific inquiry, we developed a heuristic metric for determining whether apparent gene absences are due to limitations of the sequence search regimen or represent true losses of genes from genomes. We collected RNA polymerase I (Pol I), Replication Protein A (RPA), and DNA strand exchange (SE) sequence data from 47 diverse eukaryotes. We then compared the numbers of apparent absences to a single measure of protein sequence length and sequence conservation (Smith-Waterman pairwise alignment (S-W) scores) obtained by comparing yeast and human protein sequence data. Using Poisson correlation regression to analyze the Pol I and RPA subunit datasets, we confirmed that S-W scores and apparent gene absences are correlated. We also determined that genes encoding products that are critical for interhomolog SE in model organisms (Rad52, Rad51, Dmc1, Rad54, and Rdh54) have been lost frequently during eukaryotic evolution. Saccharomyces cerevisiae null rad52, dmc1, rad54, and rdh54 mutant phenotypes are suppressed by rad51 overexpression or mutation. If rad51 overexpression or mutation affects other eukaryotes in a similar fashion, this phenomenon may account for frequent losses of genes whose products are critical for the completion of meiosis in model organisms. Finally, we place this work into greater context with a review of hypotheses for the selective forces and mechanisms that resulted in the origin of meiosis. The review and the data presented in this thesis provide the basis for a model of the origin of meiotic genes in which meiosis arose from mitosis by large-scale gene duplication, following a preadaptation that served to reduce increased numbers of chromosomes (from diploid to haploid) caused by erroneous eukaryotic cell-cell fusions.

La recombinaison méiotique est au cœur de l'hérédité Mendélienne, de l'évolution et de l'amélioration des plantes, car elle assure, grâce aux crossovers, une transmission fidèle des chromosomes et le brassage de l’information génétique au fil des générations. Deux voies de formation des crossovers coexistent chez les plantes. La voie principale (voie I) dépend de la protéine MSH4 (et de quelques autres). La voie secondaire ne produit que quelques crossovers (dits de voie II) au cours de la méiose d’une plante de type sauvage ; ils sont indépendants de MSH4 et leur nombre est limité par des protéines telles que FANCM. Si ces deux voies ont été bien décrites chez des espèces diploïdes, ce n’est pas le cas chez des plantes allopolyploïdes, pourtant très fréquentes parmi les plantes cultivées. Il s'agit là d'une lacune importante, car la présence de plusieurs jeux de chromosomes apparentés conduit à augmenter le nombre de partenaires susceptibles de former des crossovers et le nombre de copies de tous les gènes méiotiques, rendant la recombinaison méiotique plus complexe. Cette thèse vise à explorer l'interaction entre les voies de formation des crossovers et la polyploïdie en utilisant des mutants de colza (Brassica napus; AACC) et d’un de ces parents diploïdes (B. rapa; AA) pour deux gènes de la recombinaison méiotique.J'ai tout d'abord testé dans quelle mesure la formation de crossovers entre chromosomes homologues et entre homéologues (chromosomes A et C) est tributaire des voies I et II en évaluant l’effet d’une diminution du nombre de copies fonctionnelles de MSH4 sur le nombre de crossovers. J'ai montré que ce dernier n'est altéré que lorsque les deux copies MSH4 sont inactivées, toute autre combinaison de mutations conduisant au même nombre de crossovers inter-homologues que chez le sauvage. J'ai également montré que la proportion de crossovers de voie II chez des mutants msh4 de colza est bien supérieure à celles observées chez d’autres plantes mutantes pour msh4. Cette observation reste vraie chez des mutants msh4 de B. rapa, suggérant que la proportion accrue de crossovers de voie II n’est pas spécifique au colza, mais probablement une caractéristique des Brassicaceae. Chez des plantes allohaploïdes (AC) de colza, chez lesquelles les crossovers ne peuvent se former qu’entre homéologues, les copies MSH4 ne se compensent plus complétement ; le nombre de crossovers de voie I fluctue au contraire proportionnellement au dosage de MSH4, devenant presque nul lorsque toutes les copies sont inactivées. Mes résultats illustrent deux nouvelles propriétés spécifiques des crossovers entre homéologues: une plus grande sensibilité vis-à-vis du dosage MSH4 pour les crossovers de voie I et une plus faible efficacité des crossovers de voie II.Dans un second temps, j'ai caractérisé cytologiquement des mutants fancm de colza pour vérifier que l'augmentation des crossovers de voie II ne nuit pas à au bon déroulement de sa méiose. Cette question est restée en suspens, les mutants fancm de colza n’étant pas complètement nuls. Cet écueil m'a incité à développer une approche de TILLING par séquençage pour identifier de nouveaux mutants de recombinaison chez le colza. J'ai alors combiné les mutations fancm et msh4 chez B. rapa pour vérifier si les premières suffisent à corriger les défauts méiotiques induits par les secondes. J'ai montré que, conformément à ce qui avait été observé chez Arabidopsis thaliana, la mutation fancm augmente le nombre de crossovers à un point tel qu’elle restaure la formation de bivalents dans un mutant msh4. La fonction de FANCM est donc conservée chez B. rapa.Mes résultats ont fait progresser la compréhension des voies de formation des crossovers lors d’une méiose allopolyploïde. Ils indiquent que la transmission des chromosomes chez ces espèces implique principalement des crossovers de voie I, et qu’elle pourrait être assurée en limitant l’efficacité de cette voie (e.g. en diminuant le nombre de copies de gènes). / Meiotic recombination ensures, through the formation of crossovers (COs), both faithful chromosome transmission and allelic shuffling over generations; it is at the heart of Mendelian heredity, evolution and plant breeding. Two crossover pathways co-exist in plants. The main pathway (class I) is dependent on MSH4 (and additional proteins). The secondary pathway produces only a few MSH4-independent (class II) crossovers during wild-type meiosis that are limited in number by anti-crossover proteins such as FANCM. These pathways have been extensively described in diploid species, disregarding one of the most pervasive features of crop genomes: polyploidy. This is a major gap in our understanding because the presence of more than two related sets of chromosomes leads both to extra partners for crossover formation and additional copies for all meiotic genes, which make meiotic recombination more intricate. This thesis aims at exploring the interplay between meiotic recombination pathways and polyploidy using mutants for two recombination genes in allotetraploid Brassica napus (AACC; 2n=38) and its diploid progenitor, B. rapa (AA; 2n=20). I have first tested the extent to which class I and class II pathways contribute to inter-homolog and inter-homoeolog (between A and C chromosomes) crossover formation by analyzing how crossovers are affected as the number of functional MSH4 copies decreases. I showed that inter-homolog crossover formation is impaired only when the two MSH4 copies are lost, any other combination of msh4 mutations resulting in wild-type crossover numbers. I also observed that, when class I crossovers are completely abolished in B. napus, the highest frequency of class II crossover ever reported among plant msh4 mutants is observed. I reproduced this result using B. rapa msh4 mutants, thereby demonstrating that increased class II crossover frequencies is not specific to B.napus, but could instead be a general feature of the Brassicaceae. In B. napus allohaploids (AC), where crossovers are forced to occur between homeologs, MSH4 copies no longer complement each other perfectly; counter to the situation in euploids, the number of MSH4-dependent crossovers formed between homoeologs fluctuates with MSH4 dosage in these plants, and approximate zero when all MSH4 copies are depleted. Altogether, my results illustrate two novel specific properties of inter-homeolog crossovers: a greater sensitivity to MSH4 dosage for class I pathway and a lower efficiency for class II.Next, I characterized cytologically B. napus fancm mutants to confirm that boosting class II crossovers would not be detrimental to B. napus meiosis. However, a prudential interpretation of these results is demanded since the B. napus fancm alleles retained residual anti-crossover activity. This has prompted me to set up a TILLING-by-sequencing procedure in order to produce new recombination mutants in B. napus. I also combined the B. rapa fancm and msh4 mutations to test whether the former is sufficient to fix the meiotic defects resulting from the latter. I showed that, similarly to what had been observed in Arabidopsis thaliana, fancm mutation boost COs to such a point that it restores bivalent formation in B. rapa msh4 background. My results therefore confirmed that the function of FANCM is conserved in B. rapa. Overall, the findings and achievements of this thesis make a step forward dissection of CO pathways during allopolyploid meiosis. They indicate that meiotic adaptation to allopolyploidy mainly involve the class I crossover pathway and could be achieved by limiting its efficiency (e.g. by decreasing gene copy number).

Meiotic recombination is a fundamental genetic process in all sexually reproducing eukaryotes, ultimately responsible for the generation of new combinations of alleles upon which natural selection can act. It begins with the formation of programmed double stranded breaks along the genome, and ends with their repair as non-crossover or crossover recombination events. The localization of such events along the genome has important evolutionary consequences for genome structure, base composition, patterns of genetic diversity, linkage disequilibrium and introgression, along the genome, as well as in the evolution of post-zygotic hybrid sterility and speciation. Understanding how meiotic recombination events are localized is thus crucial to the proper interpretation of observed genetic variation, and to the field of population genetics as a whole. However, little is known about how most species localize recombination events. While some species localize meiotic recombination events fairly evenly along the genome (e.g., Caenorhabditis elegans or Drosophila), most species studied to date, including all yeasts, plants and vertebrates, localize the vast majority of meiotic recombination events to narrow intervals of the genome known as recombination hotspots. Within such species, there appear to be at least two general mechanisms underlying the localization of hotspots. First, in many species, including baker’s yeast, canids, birds, and plants, the vast majority of hotspots are found in close proximity with promoter-like features of the genome, such as transcriptional start sites and CpG-islands. Recombination landscapes in these species tend to be highly conserved between closely related species. Second, in mice, primates and cattle, the vast majority of hotspots are found away from promoter-like features of the genome, and at sites bound by the PRDM9 protein, which has a rapidly evolving DNA-binding specificity. Concordantly, the recombination landscapes in these species tends to be rapidly evolving. The aim of Chapter 2 of this dissertation is to characterize the distribution of mechanisms across vertebrates indirectly, by leveraging what is known about their genetic and molecular underpinnings. In particular, I consider what is known about the molecular mechanisms and evolutionary consequences of using PRDM9 to localize recombination events, and attempt to infer which vertebrate species are or are not likely to be using PRDM9 in an analogous manner. I find that PRDM9 has been lost repeatedly within vertebrates, and, moreover, that many species carry partial PRDM9 orthologs lacking one or more feature believed to be important for its role in recombination. In Chapter 3, I demonstrate that swordtail fish, which have such a partial PRDM9 ortholog, do not use PRDM9 to localize recombination events. Instead, they use promoter-like features of the genome, similar to species lacking PRDM9 altogether. This work suggests that only species carrying complete PRDM9 orthologs are likely to use them to localize recombination events, and that upon the partial or complete loss of PRDM9, species typically default to the use of promoter-like features. Beyond more immediately practical insight, understanding the phylogenetic distribution of mechanisms by which meiotic recombination events are localized along the genome will shed light on why different species employ different mechanisms. The repeated losses of PRDM9-directed recombination across vertebrates suggests that selective pressures are not always strong enough to justify the evolutionary maintenance of PRDM9. Notably, theory suggests that PRDM9’s DNA-binding specificity has to be continually evolving in order for it to localize recombination events to hotspots. This is a consequence of gene conversion acting to remove PRDM9 binding sites from the population over time. Models have been proposed in which selection favors younger PRDM9 alleles because their binding sites have experienced less erosion due to gene conversion. Nonetheless, it has remained unclear how the loss of PRDM9 binding sites might cause a reduction in fitness, principally because it has remained unclear what the evolutionary benefit of having hotspots is more generally. Recently, however, a number of studies investigating the role of PRDM9 in mediating hybrid sterility in certain crosses of musculus subspecies have implicated the erosion of its binding sites in this process. In particular, the lineage specific erosion of PRDM9 binding sites causes, in the F1 generation, the PRDM9 alleles from each parental lineage to bind primarily to the non-parental genetic background, where its binding sites have not yet been eroded. These studies suggest that there is a benefit to the symmetric binding of PRDM9 across homologous chromosomes, and that fitness is reduced as a consequence of asymmetry in PRDM9 binding. In Chapter 4 of this dissertation I develop a population genetics based model of the co-evolution of PRDM9 and its binding sites taking into consideration these recent findings. In particular, I model competition between PRDM9 binding sites and define fitness as a function of PRDM9 binding symmetry. This model demonstrates that PRDM9 binding symmetry will decrease over time in randomly mating populations, and that selection for symmetric binding is sufficient to drive the rapid turnover of PRDM9 alleles. Importantly, the requirement for symmetry in this model shapes the recombination landscape by favoring highly skewed binding distributions. This model thus provides theoretical support for the hypothesis that a requirement for symmetry might underlie the evolutionary advantage of recombination hotspots.

Replication Factor A (RFA) is an essential heterotrimeric single-stranded DNA (ssDNA) binding complex, comprised of Rfa1, Rfa2, and Rfa3 in Saccharomyces cerevisiae. RFA is required for DNA replication, repair, recombination, and cell cycle regulation. RFA acts as a sensor of ssDNA, a common intermediate of these processes, and coordinates these processes through recruitment of proteins. For example, during the DNA damage response (DDR), RFA-coated ssDNA is necessary for the recruitment and activation of the sensor kinase Mec1. Additional checkpoint proteins, also recruited by RFA, are necessary for the downstream recruitment and activation of the effector kinase Rad53 that ultimately leads to cell cycle arrest. Thus, RFA acts as a bridge to recruit the proteins required for checkpoint regulation in response to DNA damage. Importantly, cell cycle resumption is contingent on Rad53 deactivation. There are two known scenarios in which Rad53 is deactivated: (1) checkpoint recovery, in which cells resume the cell cycle after DNA repair or (2) checkpoint adaptation, in which cells proceed with the cell cycle despite the continued presence of irreparable DNA damage. Previous work has demonstrated that cells undergoing checkpoint adaptation display late Rfa2 N-terminal (NT) phosphorylation that is correlated with the inactivation (dephosphorylation) of Rad53. Additionally, the use of rfa2 NT mutations consistently demonstrate that a negatively charged NT promotes adaptation in all adaptation-deficient strain backgrounds investigated. Interestingly, Rfa2 NT phosphorylation also occurs early during meiosis. This work demonstrates that: (1) Rfa1-DBD-F participates in protein-protein interactions that are sensitive to DNA damage, (2) Rfa2 phosphorylation increases the DNA damage sensitivity of mutants with deficient DNA damage checkpoints, (3) the Rfa2 NT is required for proper progression through meiosis that appears to be unrelated to RFA functions in replication or DNA repair by homologous recombination (HR), and (4) Rfa2 phosphorylation may regulate Mec1 checkpoint signaling during the DDR to control checkpoint exit and cell cycle resumption. A mechanism is proposed that considers both Rfa1 DBD-F and the Rfa2 NT involvement to initiate HR repair that essentially allows for the continuation of the cell cycle by the delocalization of Mec1.