Reconfigurable Intelligent Surface (RIS) technique is effective to improve the capacity and coverage of various networks. However, the practical RIS has dissimilar responses to signals in different frequencies due to its structure. Therefore, directly applying existing schemes to multi-band heterogeneous networks leads to beam misalignment and performance degra- dation issues. This paper considers RIS-assisted multi-band heterogeneous networks where base stations (BSs) use different frequencies. We formulate the problem for maximizing the sum rate of all users (SR) in every frequency band, and aim to jointly design BSs precoding vectors and RIS parameters while taking user choices into account. Based on the closed-form relationship of phase shifts between frequencies, an iterative algo- rithm is proposed to solve the challenging non-convex problems. In particular, we use fractional programming to decouple the problem, and solve the subproblems with Quadratic Transform (QT) and Genetic Algorithm (GA). Simulation results show that the proposed scheme can significantly improve SR among all frequency bands, which could instruct the coexistence of mul- tiple communication systems in future heterogeneous networks. What’s more, results also prove that deploying RISs recklessly will result in a degradation of the network.