E Pluribus Unum: 50 Years of Research, Millions of Viruses, and One Goal--Tailored Acceleration of AAV Evolution

Abstract

Fifty years ago, a Science paper by Atchison et al. reported a newly discovered virus that would soon become known as adeno-associated virus (AAV) and that would subsequently emerge as one of the most versatile and most auspicious vectors for human gene therapy. A large part of its attraction stems from the ease with which the viral capsid can be engineered for particle retargeting to cell types of choice, evasion from neutralizing antibodies or other desirable properties. Particularly powerful and in the focus of the current review are high-throughput methods aimed at expanding the repertoire of AAV vectors by means of directed molecular evolution, such as random mutagenesis, DNA family shuffling, in silico reconstruction of ancestral capsids, or peptide display. Here, unlike the wealth of prior reviews on this topic, we especially emphasize and critically discuss the practical aspects of the different procedures that affect the ultimate outcome, including diversification protocols, combinatorial library complexity, and selection strategies. Our overall aim is to provide general guidance that should help users at any level, from novice to expert, to safely navigate through the rugged space of directed AAV evolution while avoiding the pitfalls that are associated with these challenging but promising technologies.

Figure 1

Schematic diagram depicting one round of directed molecular evolution which involves three basic steps: diversification, selection, and amplification.

Figure 2

Methods for AAV capsid diversification. Depicted are the three strategies that are predominantly found in the literature—DNA family shuffling (DFS), peptide display (PD), and error-prone PCR (epPCR). Also indicated are typical results or recent optimizations, respectively, of each technology. DFS: (1) “good” capsid combining properties of the parental viruses; (2) “bad” capsid in which too many diverse fragments have disrupted functionality; (3) “superior” capsid which has gained novel useful features (pink) not found in any parent. PD: (1) depletion of library from heparin binders (outside the circle), to improve retargeting; (2) use of a chimeric capsid as scaffold for peptide display; (3) same as (2), but using a different AAV wild type. epPCR: (1) rare example of a “winner” capsid which has been improved by a few point mutations; (2) typical wealth of nonfunctional capsids (“losers”) resulting from disruptive mutations; (3) focused PCR randomization in a surface-exposed capsid region.

Figure 3

Typical workflow for directed AAV capsid evolution. Depicted from top to bottom are the main steps, from cap gene diversification via different methods (see also Figure 2) and packaging of the resulting mutant pool as a viral library, to selection in cells or animals (positive selection pressure) and, if desired, in the presence of neutralizing anti-AAV antibodies (negative selection pressure). Also indicated is that the ensuing enriched library can be subjected to further rounds of selection after amplification (dashed line) and, again if desired, after additional diversification (dotted line). Eventually, up to five of these iterative cycles will result in one (ideal outcome) or a few mutant capsids that fulfill all requirements and are best tailored for a given application. This figure contains clipart from Servier Medical Art.

Figure 4

Schematic depiction of four possible 3D fitness landscapes resulting from plotting of AAV capsid sequence diversity (horizontal axes) against particle fitness (vertical axis). Indicated on the left bottom in each panel is a starting AAV library that is subjected to positive and negative selection pressures, in order to enrich a lead candidate (always circled in red) that managed to climb the highest fitness peak. (a–d) From a to d, the complexity of the landscapes increases which in turn augments chances that other particles may also start to become enriched (such as those in yellow or green), whereas others may be rapidly lost (the one in blue in panel d). An example for panel a is a simple library selection in a single cell type in culture without any additional pressure, whereas the setting in panel d is representative for in vivo biopanning in a live animal. Obviously, such a complex landscape with its many peaks and pits, and hence numerous potential outcomes, puts significant demands on the experimenter to foster the eventual selection of desired capsids. See main text for possible strategies.