This paper studies a fundamental problem in dynamic spectrum access (DSA) networks: given a set of detected spectrum bands that can be temporarily used by each node in a DSA network, how to form a topology by selecting spectrum bands for each radio interface of each node, called topology formation in this paper. We propose a novel layered graph to model the temporarily available spectrum bands, called spectrum opportunities (SOPs) in this paper, and use this layered graph model to develop effective and efficient routing and interface assignment algorithms to form near-optimal topologies for DSA networks. We have evaluated the performance of our layered graph approach and compared it to a sequential interface assignment algorithm. The numerical results show that the layered graph approach significantly outperforms the sequential interface assignment.