Coprime arrays have holes in the difference coarray, which significantly degrade the performance of the coarray-based estimators. In this paper, we propose a multi-layer concept to fill the holes and resultantly improve the direction of arrival (DOA) estimation of coprime arrays. Specifically, we first construct a coprime array by directly duplicating the subarray of coprime array with displaced subarray (CADiS), named as coprime array with multiple displaced subarrays (CAMDiS). Subsequently, we propose a multi-layer concept and classify the holes in the difference coarray of CAMDiS into multiple layers, which reveal that the holes are the results of two missing interlayer subarrays. Furthermore, we propose a coprime array with multi-layer subarrays (CAMLS) by incorporating the interlayer subarrays into CAMDiS. It is noteworthy that the proposed coprime array outperforms the family of nested arrays in mitigating mutual coupling and especially in the scale of uniform difference coarray. Simulation results validate the superiorities of the proposed CAMLS over existing sparse arrays in terms of DOF and DOA estimation performance.