This paper investigates demand-originated reversible lane design plans for transportation networks, where the tidal and asymmetrical characteristics of the demand structure are taken into consideration. Compared to traditional network design methods, this reversible lane technique has some advantages. It not only can improve the performance of the considered network by taking full advantage of the potential resources and existing infrastructures but is also easy to operate, returns rapidly, is less costly and requires no additional land. Specifically, this study first defines a coupling measure to quantify the relationship between network structure and demand structure. Then, focusing on maximizing the coupling measure, a nonlinear bilevel mixed-integer programming model is presented to find the optimal lane combination strategy in the considered network from the viewpoint of systematology. For solving the proposed model, we design heuristic algorithms to obtain its approximate optimal solution. Finally, the effectiveness of the proposed approaches is examined on the Sioux Falls network and Beijing main road network.