High-Throughput In Vivo Screening of Tissue-Specific Regulatory Elements for Next-Generation Gene Therapy Vectors

Abstract

Engineering robust and safe gene delivery vehicles is critical to clinical translation of gene therapies. Adeno-associated virus (AAV) is well-established for in vivo delivery of gene therapies due to its widespread biodistribution and natural tissue tropism. However, transgene expression in unwanted tissues can result in toxicities that have led to adverse consequences in many studies. Tropism conferred by natural or engineered viral capsids provides some control over tissue expression, but greater specificity is needed for many applications. Regulatory elements within AAV vectors can be engineered to increase specificity of expression downstream of cell entry. Massively parallel reporter assays (MPRAs) have been used to screen for tissue-specific regulatory elements. AAV vectors containing a library of regulatory elements can be screened via MPRAs to identify elements that drive tissue-specific expression. This dissertation establishes a framework for the identification, characterization, and combinatorial optimization of tissue-specific transgene regulatory elements (TREs) for gene therapy applications, focusing on distinguishing between closely related muscle types. Using genomics-guided selection informed by chromatin accessibility, histone modification profiles, gene expression data, and convolutional neural network–based prioritization, we constructed a library enriched for candidate elements predicted to differentiate skeletal muscle, heart, and liver expression. We first utilized the lentiMPRA in vitro platform to screen TRE libraries in immortalized and iPSC-derived cell lines and further adapted that platform for AAV-MPRA screening in mouse tissues. In vivo AAV-MPRA screening identified TREs capable of driving robust, tissue-specific expression in either skeletal muscle or heart while minimizing liver activity. Notably, human-derived TRE sequences retained functional specificity in murine tissues, supporting conservation of regulatory grammar across species. We next characterized genomic and transcription factor binding features associated with tissue-specific activity and identified likely contributing factors, such as GATA family factors in the heart and AP-1, MYOD, and SIX family factors in the skeletal muscle. Individual validation in mice using AAV9-luciferase constructs confirmed that MPRA-identified TREs produced selective transgene expression across multiple tissues. Finally, we selected a subset of TREs to screen in tandem pairs to evaluate whether regulatory elements can be rationally assembled to improve expression profiles. Most combinations exhibited multiplicative behavior, however, deviations from multiplicativity revealed asymmetric regulatory logic between cardiac and skeletal muscle elements, with cardiac-targeted TREs displaying greater positional autonomy and skeletal muscle–targeted TREs exhibiting increased reliance on cooperative interactions. In summary, we developed an in vivo AAV-MPRA platform to identify transgene regulatory elements that drive muscle-type-specific expression. This approach enabled the discovery and validation of TREs that distinguish expression between cardiac and skeletal muscle and revealed potential mechanisms for TRE specificity. We further demonstrate that combining tissue-specific TREs largely yields multiplicative effects on transcriptional activity, providing a modular strategy to control expression strength and specificity. Together, these results highlight the potential of this screening strategy to discover and engineer regulatory elements for improving precision of AAV-based gene therapies. While our study focused on muscle tissues, this approach can be adapted for other tissues or differential conditions to advance gene therapy applications.