Preview

PDF, English Download (29MB) | Terms of use

Abstract

Over the last years, ground-breaking success has been achieved in the field of human gene therapy (i.e., the correction of diseases at the genetic level), culminating in the approval of the first gene therapeutic drug Glybera® in the Western world in 2012. Glybera® is based on Adeno-associated viruses (AAV) which currently form the basis for one of the most promising therapeutic vectors. Their advantages over other vectors are their apathogenicity in humans, the ability to achieve long-term gene expression in non-dividing cells due to episomal persistence, and the limited immune response after vector administration. Moreover, they exhibit a broad natural tropism for a variety of cell types. Still, some tissues cannot be transduced by AAVs and are thus refractory to AAV gene therapies. Vice versa, the broad tropism can lead to non-specific off-target expression of the delivered transgene. Consequently, there is a strong need for new AAV capsid variants which can overcome these hurdles. One way to improve AAV properties is to modify the virus capsid via DNA family shuffling (DFS). This method is based on partial fragmentation of cap genes of several AAV serotypes and their homology-based re-assembly into chimeras. Since this process is highly error-prone, a first goal in this work was to optimize DFS on the technical and molecular levels. To this end, we aimed to improve the critical step of DNA fragmentation and hence compared standard DNaseI digestion with DNA shearing by ultrasonication. Of note, albeit the latter was more robust, it was outperformed by DNaseI in terms of shuffling efficiency and crossover formation. Secondly, we noted a dependency of recombination efficiencies on input DNA homologies that was most evident for the diverse cap genes of AAV4 and 5. We thus re-synthesized these variants and adapted their sequence to that of AAV2. The resulting 13 % homology increase translated into 2.5 x more crossovers per clone in an AAV2-4-5 cap library containing the new sequences as compared to its wild-type counterpart. These benefits were further validated with higher complexity libraries including larger numbers of AAV serotypes. In addition, we investigated the role of the recently discovered, essential assembly-activating protein AAP on the outcome of the DFS method. As AAP is fully encoded in the AAV cap gene and was shown to be necessary for capsid assembly, we hypothesized that AAV shuffling may disrupt AAP integrity. To investigate this theory, we created assembly-deficient AAP knock-out mutants of AAV serotypes 1 – 9 and rh10. Notably, generation of functional particles could be restored in all cases by supplying AAP expression plasmids in trans during virus production. We furthermore tested 46 shuffled and randomly chosen AAPs from different AAV cap libraries in our new model system. Remarkably, 39 (84.8 %) were able to rescue the AAP knock-out, to extents varying from 5 to 100 %. Of note, most of the deficient AAPs originated from shuffled libraries containing AAV4 or 5 capsid genes. We thus produced virus particles from a shuffled capsid library of AAV2-4-5-8-9 in the presence or absence of a mixture of all corresponding AAP expression plasmids. Strikingly, AAP addition had no effect on virus titers albeit 60 % of the encoded AAPs were predicted to be non-functional due to inadvertent recombination or disruption during shuffling. As a whole, our data strongly suggest that fortunately, unintended AAP shuffling is no major limiting factor for DFS-based generation of new AAV capsids. Finally, our AAP knock-out constructs and a newly produced anti-AAP2 polyclonal antibody allowed us to study AAP biology in greater detail. Surprisingly, we detected an influence of AAP2 on viral cap protein (VP) stability, indicating the occurrence of post-translational VP modifications. Moreover, we studied AAP2 expression by microscopy and compared it to AAPs of other serotypes. Conspicuously, cellular localization varied between the AAPs, implying unique hotspots of viral capsid assembly. Collectively, the work presented here has substantially improved the method of AAV DFS and should thus contribute to a much wider use of this powerful technology in the future. Concomitantly, our data provide novel insights into AAP biology and, together with our new constructs and models, open up a multitude of avenues for future research into the fundamental and applied aspects of AAV.