Pentose Sugars Encode Sequence-Dependent DNA-RNA Segregation for Programmable Engineering of Multiphase Condensates

Abstract

The nucleus segregates DNA and RNA with high precision to safeguard genome stability and gene regulation despite their homologous structures, yet the underlying molecular principles have long evaded both understanding and synthetic translation. Here we show that a single atomic difference between the pentose sugars, a 2-OH in RNA versus a 2-H in DNA, suffices to drive their phase segregation with cationic peptides. This subtle chemical distinction makes RNA bind peptides more strongly than sequence-identical DNA, generating an asymmetry that, when modulated by sequence-encoded homotypic interactions, drives the formation of core-shell multiphase condensates. Using a supervised machine-learning approach, we distill these molecular principles into a predictive design rule and harness it to program a library of oligonucleotide pairs, termed SEGREGamers, that self-assemble into droplets with coexisting DNA- and RNA-rich domains. These synthetic condensates recapitulate nuclear compartment functions, including selective molecular partitioning and enhanced RNA catalysis. Our results establish a chemically encoded platform for engineering synthetic nuclear mimics and programmable biomolecular condensates, and suggest that sugar identity may have served as an ancient physical mechanism for organizing nucleic acids.