Abstract
Directed evolution of viral vectors involves the generation of randomized libraries followed by artificial selection of improved variants. Directed evolution only yielded limited results in adenovirus (AdV) engineering until now, mainly due to insufficient complexities of randomized libraries. Meanwhile, clinical applications of AdVs as gene therapy or oncolytic vectors are still hampered by the predetermined tropism of natural types. To overcome this challenge, we hypothesized that randomized peptide insertions on the capsid surface can be incorporated into the AdV bioengineering toolbox for retargeting. Here we developed AdV-directed EVOlution protocols based on fiber knob peptide display. Human AdV-C5-derived libraries were constructed following three distinct protocols and selected on a panel of cancer cell lines, with the goal of identifying variants able to infect and lyse these tumor cells more efficiently. All protocols enabled the construction of high complexity libraries with up to 9.6 × 105 unique variants, an approximate 100-fold improvement compared with previously published AdV libraries. After selection, the most enriched variants, which were robustly selected in various cancer cell lines, did not display enhanced infectivity but rather more efficient replication and cell lysis. Selected inserts also conferred enhanced lysis ability to oncolytic AdVs restricted to telomerase-expressing cell lines.