Résumé
The genetic structure of populations of Mycosphaerella fijiensis was studied by sampling 13 populations from banana plantations in Latin America and Caribbean region and Africa. The isolates were analysed using CAPS (Cleaved Amplified Polymorphic Sequences) and microsatellite molecular markers. The highest levels of genetic diversity in the Latin America and Caribbean region were found in the populations from Costa Rica and Honduras and the lowest levels were observed in the Caribbean populations. In Africa, genetic diversity levels were similar between countries, with the exception of Côte d'Ivoire and the Comoros where the levels were approximately half. A high level of genetic diversity and random mating populations were maintained even at the plant scale. Overall estimates of Fst were 0.19 and 0.30 in African and Latin America-Caribbean regions respectively, indicating a high and significant (p<0.001) level of genetic differentiation. A high and significant level of genetic differentiation was detected between most pairs of populations within each region. The founder effects observed in the genetic structure of M. fijiensis in the African and Latin American and Caribbean regions is consistent with a stochastic spread of the disease at continental and country scales rather than a steady advance of an epidemic front. Genetic and epidemiological evidence support single-step invasions of the disease in the different regions of the world through the movement of infected plant materials rather than ascopore dispersal between continents. Expansion of the range of M. fijiensis populations within continents may result either from limited ascospore dispersal over a few hundred kilometers or from the movement of infected plant material. Although the relative importance of these two dispersal processes could not be determined, some evidence of movement of infected plant material strongly supports the improvement of quarantine measures to limit the risk of disease introduction in new areas and exchanges between existing populations from different countries. The sustainability of disease resistance management strategies will first depend on the effectiveness of quarantine measures. Disease management should also try to limit gene flow between pathogen populations via natural dispersal. Parameters corresponding to these evolutionary and epidemiological processes should therefore be estimated accurately. (Texte intégral)