Traditional live attenuated vaccines (LAVs) are typically developed through serial passaging or genetic engineering to introduce specific mutations or deletions. While viral RNA secondary or tertiary structures have been well-documented for their multiple functions, including binding with specific host proteins, their potential for LAV design remains largely unexplored. Herein, using Zika virus (ZIKV) as a model, we demonstrate that targeted disruption of the primary sequence or tertiary structure of a specific viral RNA element responsible for Musashi-1 (MSI1) binding leads to a tissue-specific attenuation phenotype in multiple animal models. The engineered MSI1-binding-deficient ZIKV mutants (MBD) maintained full competence in MSI1-deficient tissues but were significantly restricted in ZIKV-vulnerable tissues (brain, testis, eye and placenta) and exhibited substantially reduced vertical transmission in mice. Importantly, a single immunization with MBD ZIKV induced robust immune responses and conferred protection against ZIKV challenge in both mice and non-human primates. Thus, our study demonstrates that manipulating viral RNA structures that interact with host proteins represents a powerful platform for developing the next generation of LAVs against emerging viruses.