Directed evolution of a thermostable family-A DNA polymerase into a reverse transcriptase for 2'-modified RNA

RNA-based therapeutics increasingly rely on nuclease-resistant xeno-nucleic acids (XNAs), but engineering polymerases capable of reading and amplifying these non-canonical substrates remains a major challenge. Natural DNA polymerases cannot process C2' sugar modifications, and the complete structural determinants governing this limitation remain poorly understood, limiting rational design strategies. In this study, we apply directed evolution methods to Thermus thermophilus DNA polymerase using a microfluidic implementation of the Compartmentalized Self-Replication (CSR). We identify an engineered variant, termed E5, capable of reverse transcribing C2'-modified RNA, a previously inaccessible activity for this enzyme. Next-generation sequencing allow us to track evolutionary trajectories across nine rounds of evolution. Through an integrated analysis combining statistical methods, protein language models, and crystallographic data, we reveal unknown epistatic networks controlling polymerase function. This large-scale experimental dataset (>196,000 thermostable enzyme variants) provides a robust basis for predicting sequence-to-function relationships, accelerating development of enzymes for aptamer discovery and synthetic biology applications.