Microbes living in the rhizosphere influence plant fitness by altering nutrition, resistance to pathogens and development. In the other way, plants influence the microbiota through root exudates. However, there is no conclusive empirical evidence that the assemblage of the plant and its microbiota evolve as a single coherent unit of selection; an idea at the core of the holobiont concept (Zilber-Rosenberg & Rosenberg, 2008). Our objective was to test the hypothesis that the plant-microbiota assemblage is a unit of selection, justifying the holobiont concept. We pursue to determine if the evolution of plant-microbiota interactions was mandatory to understand the evolutionary dynamics of the plant-microbiota association, or if the plant-microbiota response was merely the addition of the individual response of the plant and the microbiota. In a five-generation artificial selection experiment aiming to increase the number of siliques in Arabidospis thaliana, we selected either the plant (by using the seeds of selected individuals for the next generation), the microbiota (by using an inoculum made of the root­-adhering soil of selected plants) or both (seeds and microbiota of selected plants). Plant selection led to an increased number of siliques over generations (33.5%), but unexpectedly, microbiota selection led to a decrease in the number of siliques (-46.2%), likely due to a change in resource allocation from vegetative to reproductive organs. When both plant and microbiota were selected, the negative effect of microbiota selection was dominant over the positive effect of plant selection (-33.6%). From the plant perspective, the response to selection of both plant and microbiota was additive in terms of silique number, but a significant plant*microbiota interaction response was observed for the relative silique number per vegetative biomass unit (+40.6%), suggesting that the holobiont may be considered as a selection unit depending on the considered trait. From the microbiota perspective, the plant*microbiota selection resulted in non-additive effects on the microbiota structure compared to the sole microbiota selection. A second objective was to estimate the microbial contribution the parental seeds and rhizosphere to the assembly of the offspring rhizosphere microbiota. The seed was poorly contributing, while the soil contribution was important. We observed a clear shift in rhizosphere composition and we have identified the taxa potentially driving these effects, likely an enrichment of Apiotrichum, Trichoderma, and an unclassified Chitinophagaceae. Hence, we conclude microbiota has a stronger role in the plant holobiont.