Previous studies that have considered the ocean circulation on Enceladus have generally assumed the salinity to be Earth-like. However, according to observations and geochemical constraints, the salinity of Enceladus's ocean is likely to be lower, and importantly, it is probably low enough to reverse the sign of thermal expansivity. We investigate the ocean circulation and stratification of Enceladus's ocean using a combination of theoretical arguments and simulations using the MITgcm. We find that if the salinity is high, the whole ocean is unstratified, and convection dominates the entire ocean. However, if the salinity is low enough, there exists a stratified layer in the upper ocean, whose thickness depends on the magnitude of the turbulent vertical diffusivity, which remains poorly constrained. Such a layer can suppress the vertical flux of heat and tracers, thereby affecting the heat flux to the ice shell and leading to a vertical tracer mixing timescale across the stratified layer of at least hundreds of years. This timescale is inconsistent with a previous estimate of vertical ocean mixing of several years, based on the size of detected silica nanoparticles in the plumes, leading us to conclude that either the salinity of Enceladus's ocean is higher than previously suggested or the interpretation of silica nanoparticle observations has to be reconsidered.