We demonstrate here, for the first time, that female fertility can be lost in vitro in M. oryzae rice strains grown asexually. This loss could be observed in 10 to 19 asexual generations, representing at least 20 000 to 38 000 mitotic generations. All strains tested displayed a loss of female fertility for at least one replicate. The loss of fecundity in laboratory conditions has already been documented in other fungal species, such as Blastomyces dermatidis and Histoplasma capsulatum . The specific loss of female fertility has also been described before, in an M. oryzae strain from finger millet , in which the female-sterile phenotype was found to be controlled by a single gene. In our experiment, the female-sterile phenotype that appeared during the experiment segregated in the first-generation crosses and in backcrosses. This result is consistent with the hypothesis that the loss of female fertility is probably due to genetic mechanisms, such as mutations, gene deletion or major chromosomic rearrangements , rather than epigenetic mechanisms. We were expecting a 1:1 segregation of female-fertile:female-sterile strains in the offspring, indicative of the involvement of a single gene. The 1:3 ratio observed in most crosses suggests the involvement of two genes. Independent mutations in two unlinked genes would have occurred twice, once in strain S1-B12 and once in strain S3-A10. However, we cannot rule out the possibility of mutations in a single gene and distortion of the segregation of this gene. Additional crosses would be required to distinguish between these two hypotheses. Segregation distortion has been documented in many Eukaryotes, sometimes resulting in a bias in sex ratio [44, 45]. Theoretical studies suggest that segregation distortion may play an important role in the evolution of sexual reproduction, acting, for instance, on sex determination . Allelism tests would be a suitable method for determining whether the same locus is affected in all strains and all replicates. However, it is not possible to cross two female-sterile strains. Genetic mapping of the mutations would make it possible to compare gene positions and to determine whether different loci were involved. Alternatively, complementation experiments could be performed by smashing the mycelia of two female-sterile evolved strains together to restore female fertility. We performed this experiment with two strains (S2-A5 and S4-A5) after the single-spore isolation of female-sterile conidia, but no restoration of female fertility was observed. We could perform complementation experiments on all combinations of pairs of female-sterile evolved strains, to test the hypothesis that there are several mutations at different loci.
We provide the first demonstration of the loss of such an important life-history trait as female fertility in M. oryzae rice strains. In an experiment on M. oryzae on artificial medium and on plants over 10 generations, Park et al.  observed no change in the pathogenicity or genome sequences of avirulence genes. They concluded that the genome was highly stable. However, each generation resulted from the transfer of only one or two conidia after two weeks of growth. So, even if mutants appeared at a normal rate, the sample studied by Park was too small for their detection. By contrast, we observed a gradual decrease in perithecium production until complete female sterility was attained. This may be explained by the transfer of at least of 27,000 (± 24,000) conidia -- constituting a population -- between Petri dishes at each generation. We therefore observed the process by which female-sterile mutants occur and colonise within a population. Consequently, the number of perithecia reflected the proportion of wild-type female-fertile individuals remaining in the population. Monospore isolation at an intermediate generation confirmed that the evolved strains consisted of mixtures of female-fertile and female-sterile individuals. The transfer of several thousands of conidia between asexual generations resulted in a smaller bottleneck than the transfer of single spores and favoured invasion by mutants and, thus, their detection.
The detection of female-sterile mutants was also probably favoured because female-sterile mutants had a selective advantage, allowing them to increase in the population. We first checked that the mycelium of the female-sterile mutants did not induce necrosis of the mycelium of female-fertile strains when grown together (data not shown). Fitness differences were revealed by several experiments in which traits relating to fitness were compared between female-fertile and female-sterile strains grown either individually or in competition. These differences in fitness were not accounted for by differences in growth rates, because growth rates (measured individually) were no higher for female-sterile than for female-fertile strains, either in vitro or in planta. For one of the four mutants tested, grown either separately or in competition in the exact selective conditions of the evolutionary experiment, we detected no selective advantage. This may be related to complex dynamics of fitness gain . However, for the other three mutants, we showed that the conidia of female-sterile mutants were transferred significantly more efficiently than those of female-fertile strains. This greater efficiency of transfer was probably linked to differences in the conidial attachment process. This alteration favoured the release of conidia from female-sterile mutants over that of conidia from wild-type strains. This unexpected change in asexual reproduction phenotype may account for the dynamics of the loss of sexual reproduction observed in our study: the decrease in the number of perithecia, reflecting the invasion of the population by female-sterile mutants, was initially rapid and then gradually slowed down until female fertility was completely lost.
We also showed that the loss of female fertility, and hence of sexual reproduction, was associated with modifications to asexual spore production in vitro. We expected to see a trade-off between sexual and asexual reproduction, in the form of an increase in asexual spore production with decreasing female fertility. However, we actually observed the opposite: asexual sporulation rates were lower in experimental strains that had lost female fertility, at least in vitro, and also, indirectly, in planta, due to a decrease in the number of lesions produced after infection. This finding can be accounted for by independent mutations or by mutations of a single locus with positive pleiotropic effects on sexual and asexual reproduction. The existence of a positive correlation between sexual and asexual reproduction capacity has already been observed in various pathosystems. For example, in an experiment on a finger millet strain of M. oryzae, Tharreau et al.  showed that female sterility was associated with higher levels of asexual sporulation in vitro but lower levels of sporulation in planta. Ali et al.  found a positive correlation between sexual spore production and asexual spore production. Zeyl et al.  found a positive correlation between sexual fitness and asexual fitness in wild-type strains in Saccharomyces cerevisiae when sexual selection was applied. A decrease in asexual reproduction capacity associated with a decrease or even a loss of sexual reproduction capacity in vitro has also been observed in Cryptococcus neoformans . Hill and Otto  showed positive pleiotropic effects between sexual sporulation and asexual sporulation in the experimental evolution of S. cerevisiae. Mycelial growth rate and sporulation rate are often used to assess fitness in fungi . However, we found, in the progenies of two crosses, that the decrease in asexual sporulation was independent of female fertility (data not shown). This study therefore provides no evidence of a pleiotropic effect of mutations conferring sterility on asexual sporulation rate.
Another interesting result was the decrease in male fertility. About half the eight strains had completely lost the ability to induce the production of perithecia by another strain by the end of the experiment. The correlation between the decrease in male fertility and the loss of female fertility observed for some of the evolved strains cases may reflect pleiotropic effects of mutations.
We also found that, at certain times, evolved strains were able to produce perithecia in the absence of strains of the opposite mating type, but this trait was not observed at all time points. The autoformation of sterile perithecia has already been reported in one M. oryzae isolate from finger millet  and the gene responsible for female fertility has negative epistatic effects on the gene responsible for the autoformation of sterile perithecia. The occurrence of such a phenotype may also result from mutants that rapidly died out due to their lower fitness.
As female fertility and sexual reproduction seem to be easily lost, it remains unclear why some M. oryzae strains have remained female-fertile in Asia. Sexual reproduction presumably confers some advantage on these strains in this environment. In some fungi, organs of sexual reproduction constitute resting structures allowing the fungus to survive in adverse conditions. However, M. oryzae ascospores have a short lifetime and perithecia are thus unlikely to constitute a resting stage in natura. As Asia is the centre of domestication of rice , the varieties in this area are likely to be more diverse than elsewhere in the world. In such a heterogeneous environment, there may be selection for strains able to reproduce sexually [31, 52].