The multilevel and historical comparison of spatial and hierarchical organisation of urban systems of the world highlights some universal properties (rank-size law, centre-periphery structure) but also an evolution toward more complex patterns (polycentrism, sprawl, spatial repartition of densities and prices). More specifically, European and North American cities have long opposed in terms of density and rent landscapes. Despite recent shifts toward more entangled forms, when considering these evolutions over a period of several centuries, it is still relevant to claim that they are representative of two distinct classes of urban systems. This presentation will focus on the morphological differences of cities in Europe and in the US. We argue that the intensity of the centre-periphery gradient can be partly explained by the geometry of the transport networks. We study the validity of this hypothesis by the modelling and simulation of generic localization mechanisms of urban functions inside the city, over a period of two centuries. The tools developed include an agent-based model named simpopNano. It allows testing the effect of generic localization mechanisms in a city that transport network is growing over time. This generic model has been tested with simple, “caricature” streets networks (grid-based, radio concentric). The first simulations tend to reinforce the idea that the difference of topologies of the streets networks, largely due to historical reasons, could be sufficient to generate some more diluted spatial repartitions, as observed in US cities when compared to European ones.