The relevance of biodiversity to human health is an increasing international political issue as it causes concern for ethical and aesthetic reasons, but also has a strong impact on ecosystem properties and ecological goods and services utilised by humanity. However, humans have applied increasing pressure on worldwide biodiversity through pollution, land-use and climate change, over exploitation and the introduction of invasive species which can lead to major alterations of biological communities and a consequent decrease in biodiversity. When communities are assembled at random from a pool of species, more diverse mixtures have a higher probability to contain species or species-groups with high capacity to drive ecosystem processes. Therefore, it is relevant to ask "how may we classify species in a community in terms of functionality?", and "which species-group is important to maintain the productivity and stability of ecosystems?" In this thesis, I chose to classify species according their frequency and their cumulative relative cover in a plant community (i.e. plant hierarchy) and to differentiate three species-groups: dominant, subordinate and transient, which contribute towards biodiversity. Dominant species are clearly very important for ecosystem functioning, due to the large amount of biomass they produce, but there is growing evidence that subordinate species, which represent a low amount of plant biomass in grassland ecosystems, are also of considerable importance for ecosystem functioning. Given this recognition, a further objective was to explore the ecosystem level effects of these subordinate plant species, which is currently not well known. This research focused on two aspects: the persistence of subordinate species in the community, and their role in grassland ecosystems. The approach to satisfy these objectives encompassed four years of experimental field studies and glasshouse microcosm experiments. In order to explain the persistence of subordinate species in semi-natural grasslands, we performed two greenhouse experiments to test the effects of root competition exclusion and the inoculation of arbuscular mycorrhizal fungi (AMF) on the competitiveness of dominant and subordinate species, directly measured by their biomass production. The effects of subordinate species on ecosystem functioning were assessed through long-term field experiment in two sites in the Swiss Jura Mountains, where dominant, subordinate and transient species were preliminarily determined. At both site (Les Amburnex, La Frétaz), a removal experiment was carried out in randomly replicated plots with three different treatments: control without perturbations, removal of all subordinate species and partial removal of dominant biomass. In one site, a summer drought treatment was added using rainout shelters to simulate an extreme climatic event. Biomass production, litter decomposition and soil respiration were monitored during four years in each plot and complemented in some cases by measurements of soil dissolved inorganic nitrogen concentrations, soil microbial communities (bacteria and AMF) and carbon and nitrogen isotopes in plant leaves. The results of the different experiments conducted in this thesis highlight the importance of cattle activity and AMF on the persistence of subordinate plant species in semi-natural grasslands. Indeed, cattle activities (i.e. trampling) provide spatial heterogeneity, through gap creation (areas of root competition reduction), and thus favours the growth of less competitive subordinate species. While AMF acted as a parasite in the greenhouse experiment, the current findings suggest that the competitiveness of dominant species was reduced through the action of the fungus, which indirectly enhanced subordinate species. The interactions between subordinate species and soil organisms were confirmed in the field as bacterial and AMF communities shifted in plots where subordinate species were removed compared to the control treatments. Moreover, the absence of subordinate species had many negative effects on ecosystem functioning by reducing litter decomposition, soil respiration, nitrogen mineralization and community above-ground production. These findings suggest that plant-soil feedbacks explain the importance of subordinate species on ecosystem functioning despite only representing a very low quantity of biomass in plant communities. During the summer drought, subordinate species increased the resistance of the plant community and maintained productivity. Therefore, these outcomes confirm the role of drought-resistant subordinates in the functioning of grassland ecosystems threatened by climate change. The synthesis of this thesis has important implications for the way grasslands should be managed for biodiversity and confirms that extensive grazing must be maintained as it promotes species coexistence and the persistence of key subordinate species. Moreover, this work demonstrates the important role of subordinate species in ecosystem functioning, showing that not only dominant species determine ecosystem properties. This thesis serves as a contribution to the advances in our understanding of ecosystem functioning and above and below-ground linkages, and provides basis for future research in this domain.