Long-stranded XNA-cssDNA hybrids for robust data storage

DNA is a promising medium for data storage due to its high density and low energy cost. Long-stranded DNA with minimal indexing improves storage density but suffers from poor stability. To address this, we introduce a long-stranded xeno nucleic acid-circular single-stranded DNA (XNA-cssDNA) hybrid strategy to enhance storage robustness. We evaluated multiple XNAs under extreme conditions and identified 2'-fluoro-arabinonucleic acid (FANA) as the optimal storage material. To overcome XNA synthesis length and speed limitations, we used a temperature-guided language model to evolve a FANA polymerase Tgo^mut, which is ~4.4-fold faster than Tgo^D4K and enables synthesis of strands exceeding 7500 nucleotides. Data-encoded cssDNA were produced using M13 bacteriophage. The resulting FANA-cssDNA hybrids show strong resistance to chemical and enzymatic degradation. Using a dual-strategy framework, we achieved reliable data writing and reading with 100% data recovery, even after DNA degradation, and successfully visualized encoded digital data and an enhanced green fluorescent protein gene in mammalian cells.