Vineyards and wineries are ecological habitats that house a diverse community of yeast and bacteria. Grapes and cellars are two sources of S. cerevisiae strains involved in the winemaking process, but the relationship between both S. cerevisiae populations is still poorly understood. Indeed strains in vineyards samples are rarely the same as those isolated from vats. In order to better apprehend this issue, 1374 S. cerevisiae isolates were collected from 193 samples of Merlot grapes obtained across 5 Bordeaux regions and from 11 spontaneously fermenting must of 7 cellars. We obtained 402 different genotypes using 17 microsatellite markers. A first analysis of genotypes indicated that approximately ¼thof isolates presented more than 75% of similarity with commercial yeast starters, suggesting that they escaped from the cellar environment, but presented variations higher that could be detected from the analysis of their industrial batch production. The resulting S. cerevisiae populations of 302 grapes-associated and 225 cellar-associated unique profiles revealed a global low differentiation (Fst=0.036), but with differences from sites to sites. To limit potential geographic or sampling effects, a subset of 72 individuals were selected among the vineyard and cellar isolates, representing 5 cellars and the vines located in their immediate environment. Again a low differentiation was noticed (Fst= 0.03 +/- 0.001). The geneflow between vines and cellars was inferred, using a likelihood approach implemented in the software MIGRATE. Inferences showed that cellar and grape metapopulations present similar theoretical sizes and are connected by asymmetric geneflow: almost 4 times higher in the direction "grapes-to-cellar" than for "cellar-to-grapes" (number of migrants per generation 191 [166 -226] versus 55 [25 - 83], respectively). This reveals that vines and cellars are two compartments of the same ecosystem, which has deep ecological significance.< Réduire