Common AAV gene therapy vectors show indiscriminate transduction of living human brain cell types

Abstract

The development of cell-type-specific gene therapy vectors for treating neurological diseases holds great promise, but has relied on animal models with limited translational utility. We have adapted an ex vivo organotypic model to evaluate adeno-associated virus (AAV) transduction properties in living slices of human brain tissue. Using fluorescent reporter expression and single-nucleus RNA sequencing, we found that common AAV vectors show broad transduction of normal cell types, with protein expression most apparent in astrocytes; this work introduces a pipeline for identifying and optimizing AAV gene therapy vectors in human brain samples.

Figure 1 |. Human brain organotypic slice model facilitates characterization of AAV tropism by immunofluorescence.

a, Workflow of the human brain organotypic model: example MRIs with location of tissue resected, resected samples, vibratome slicing, NMDG-aCSF recovery, culture plates, and immunostaining or single-nucleus sequencing. b, The AAV capsid variants studied and the CAG-eGFP construct used for immunostaining and single-nucleus sequencing. c, The anatomical locations of patients’ samples, pathologies involved, and approximate ages. d, Representative images of entire sections of PHP.S and AAV9 from adult frontal cortex overlying tumor. Scale bars = 200 μm. e, Magnified images of PHP.S with DAPI, NeuN (neurons), GFAP (astrocytes), and transduced cells (GFP). f, Representative magnified images showing PHP.S and astrocytes, AAV9 and neurons, and AAV2 and microglia in pediatric temporal lobectomy samples. Scale bars = 200 μm. g, Percent transduction of all cells present in the tissue (measured as percent of DAPI-based ROIs that show GFP signal higher than the brightest control ROIs). (‘A’ indicates significant differences from the other bar labeled ‘A’, as does ‘B’. Kruskal-Wallis test with Dunn correction for multiple comparisons, adjusted p < 0.05) h, Percent of NeuN-positive, GFAP-positive, and Iba1-positive cells that co-express GFP, indicating successful transduction by the respective capsid variants. (* indicates p < 0.05 for post-hoc adjusted comparison of astrocytes versus microglia for the indicated variants except AAV2-retro, where the post-hoc significant comparison was astrocytes versus neurons. Kruskal-Wallis test with Dunn post-hoc test.)

Figure 2 |. Single-nucleus RNA sequencing identifies human brain cell types transduced with AAV capsid variants.

a, Samples from the pediatric temporal lobe and adult frontal lobe were sequenced and aligned independently and then combined for cluster analysis. Panels separated by day. Clusters assigned using SingleR and the Allen Institute’s middle temporal gyrus transcriptomic data. b, Correlation matrix for log2 fold change by cell type, comparing day 0 tissue to day 14 transduced (virus added) tissue. Each cell contains the Pearson correlation coefficient. c, UMAP of the Day 14 with virus sample (15,540 cells), with dark blue cells indicating the presence of at least one viral transcript identified (1,432 cells). d, Bar graphs showing total number of cells transduced by each AAV capsid variant for the pediatric and adult samples (pediatric sample missing Sch9). e, Heatmap showing percent of each cell type transduced by AAV capsid variant.