In-Depth Parallel Profiling of Tissue and Cell-Type Tropism of AAV Variants by Single-Cell RNA Sequencing

Abstract

Adeno-associated viruses (AAVs) are popular gene delivery vehicles and there is a continuing high demand for AAV variants with improved transduction efficiency and specificity. Directed evolution and/or rational design have been used extensively to engineer the capsid of naturally occurring AAVs in order to better customize their properties for research and clinical work. While these engineering approaches are scalable and have generated useful variants, the subsequent transduction profiling of these variants remains either low throughput or lacks resolution across the many relevant cell and tissue types. Single-cell RNA sequencing (scRNA-seq) via droplet-based methods allows in-depth profiling of gene expression of several thousand individual cells. We established a tissue processing and data analysis pipeline that leverages the capabilities of scRNA-seq to achieve simultaneous characterization of AAV variants across multiplexed tissue cell types. To verify our approach, we retro-orbitally co-injected C57Bl/6 mice with PHP.eB (Chan et al., Nat. Neurosci., 2017) and a neuron-biased PHP.eB-evolved variant (Flytzanis*, Goeden* et al., ASGCT, 2019), each packaging a construct expressing different fluorophores. After two weeks of expression we harvested the brain and used one hemisphere for characterization by traditional immunohistochemistry and one hemisphere for characterization by scRNA-seq. Single-cell libraries were prepared with the Chromium Single Cell Kit by 10x Genomics and analyzed with multiplexed Illumina sequencing. As a proof of concept, we compared the two characterization methods by analyzing the infection rate of neurons (NeuN), astrocytes (S100b) or oligodendrocytes (Olig2). For immunohistochemistry, a cell was classified as infected based on expression of fluorophores while in scRNA-seq transduced cells were identified based on the presence of defining viral transcripts. Louvain community detection method (Blondel et al., J. Stat. Mech., 2008) followed by analysis of significantly differentially expressed genes was used to identify cell types in the scRNA-seq data set. Given the differences in RNA and protein abundance and detection thresholds between imaging and sequencing, the two characterization methods detect different absolute numbers of infection rates; however, the cell type transduction biases are consistent among the three different cell types we tested. After verifying our method, we further explored the data set beyond major cell types and discovered previously unnoticed sub-cell type enrichments in, for example, cortical inhibitory neurons. These findings are being confirmed by mapping mRNA expression using in situ hybridization chain reaction. Besides sub-cell type tropism characterization, we are analyzing the transcriptome of infected and non-infected cells in search of mechanistic insights into AAV transduction that could facilitate rational design of recombinant AAVs with disease-relevant cell-type specificity. Our approach will aid the gene therapy field to both characterize more thoroughly existing recombinant AAVs and guide engineering of novel AAV variants.

Additional details