PDF, English Download (4MB) | Terms of use

Abstract

Viral vectors based on Adeno-associated viruses (AAV) have a broad application spectrum including gene therapy and basic research. However, because naturally occurring AAV capsids are rarely sufficiently efficient and/or specific for a given application, techniques were developed to broaden the existing capsid repertoire. A prototype technology is DNA family shuffling where, in a first step, homologous cap genes encoding capsid subunits are fragmented and recombined, yielding a viral library which can then be subjected to selection in order to enrich promising variants. The aim of the present study was to dissect and improve four critical steps along this procedure. Firstly (1), two different methodologies for production of cap gene fragments were compared, resulting in the identification of DNase I based fragmentation as the most robust approach. Interestingly, cap DNA concatamer formation during nested PCR was observed, leading to amendment of the PCR purification protocol. Next (2), we studied the impact of chimerism on the essential assembly-activating protein (AAP) that is encoded in an alternative open reading frame within cap and is recombined as well during DNA family shuffling. Importantly, by performing a battery of complementary experiments, we were able to show that shuffling of AAP is not impairing its function, i.e. the support of particle assembly. Furthermore, no influence on titers was observed for wild-type and most chimeric vector productions, altogether relieving long-standing concerns about a potential rate-limiting role of AAP for AAV vector generation and evolution. Thirdly (3), we established a pioneering in vivo AAV library selection strategy in which, unlike most previously reported schemes, we selected novel capsids in specific cell types within an organ instead of the organ as a whole. Specifically, we were motivated by the facts that liver disease is wide-spread in humans and that hepatic stellate cells (HSC) are known to drive liver fibrosis, thus contributing to disease progression. Alas, tools to genetically manipulate HSCs are limited. Therefore, a library encompassing 10 capsid variants was selected in HSC by AAV injection into mice, HSC isolation and PCR rescue using purified total DNA. Following multiple selection rounds, in vivo bulk validation was performed based on next-generation sequencing. In total, 157 capsid variants were screened in parallel and again, the liver was segregated into the single cell types, i.e. hepatocytes, HSC, Kupffer cells and liver sinusoidal endothelial cells. Notably, this revealed that the selection was successful as hepatocyte-detargeted vectors were identified that showed a strong co-transduction of HSC and Kupffer cells. Intriguingly, we noted differences in vector specificity and efficiency on the DNA versus the RNA level. In order to even further restrict the new vectors to a given cell type, vector cassettes were generated bearing cell-type specific promoters and miRNA binding sites to suppress off-targeting in cells expressing these miRNAs. Testing of these constructs in vitro gave promising results especially for the miRNA-based detargeting strategy. Finally (4), we implemented improvements during the selection and analysis steps, including the use of PacBio/SMRT sequencing technology to monitor AAV sequence enrichments throughout the course of selection. Additionally, we managed to increase the stringency of the PCR rescue of cap genes, by incorporating sample-specific barcodes, i.e., short, unique nucleotide stretches, into the AAV library genomes. By using these barcodes as a primer during sample recovery, we could isolate single libraries out of a complex library mixture, as validated in vitro. In the future, this original strategy could be exploited to track individual libraries in vivo upon injection of a mixture of libraries, which should in turn help to accelerate the identification of top-performing variants for validation studies. In summary, different steps along the powerful methodology of DNA family shuffling were improved advancing future vector development and the lingering concern about AAP impairment upon shuffling was dispersed.