In the cell, biochemical reactions are organized not only by membrane-bound organelles but also by membraneless condensates that form through phase separation. Although the initial stages of droplet formation are well studied, their long-term evolution, known as aging, remains poorly understood. Aging describes the transition from liquid-like droplets to viscoelastic, gel-like, or even solid states over timescales ranging from minutes to months. This slows down the diffusion of molecules, strips droplets of their ability to fuse, and gives rise to conformational heterogeneity, much as in glasses. A moderate increase in viscosity can stabilize functional assemblies, but excessive hardening is linked to neurodegenerative disorders and cancer. This work sits alongside studies of cell aging and droplet organization in the cell, but it focuses on molecular mechanisms that had previously only been hypothesized.
The authors used a coarse-grained RNA model in which each nucleotide is represented by a single particle that retains the identity of its base. The model includes harmonic bonds along the backbone, excluded-volume repulsion, and a many-body potential for Watson-Crick base pairs. Simulations were run for 64 chains of (CAG)31 in a periodic box at a concentration of 400 µM, using Langevin dynamics with a 10 fs time step at 300 K. To analyze aging, they tracked only the molecules that stayed inside the droplet the whole time and split the trajectory into 27 windows. Dynamics were quantified through the overlap function, the pairwise root-mean-square displacement, and relaxation times extracted from a biexponential fit. Spatial heterogeneity was probed by dividing the droplet into concentric shells 5 nm thick. The topology of the entanglement network was analyzed using the primitive path method. For comparison, they simulated RNA with a random sequence of the same composition.
The simulations showed that the (CAG)31 condensate ages spontaneously: the overlap function does not decay to zero, and relaxation follows a power law, indicating incomplete ergodicity. A biexponential fit revealed fast (τ1) and slow (τ2) relaxation modes that grow with droplet age, with τ1 and τ2 differing by two orders of magnitude. The fraction of the frozen population, A0, increases over time. Shell-by-shell analysis showed that aging is spatially uneven: in the core, diffusion slows down fastest (aging exponent αw,core = 0.53 ± 0.11), while at the periphery it slows more gradually (αw,outer = 0.26 ± 0.15). In the core, the fraction of non-relaxing contacts reaches 30%, and relaxation times are six times longer than at the surface. RNA chains in the core are swollen (radius of gyration about 7.0-7.5 nm versus 5.5-6.0 nm at the periphery), have a Flory exponent ν > 0.5, and are partially oriented radially. Primitive path analysis showed the number of entanglement nodes rising from 5 to 8 over the course of the simulation, with node density in the core five times higher than in the shell. The density of Watson-Crick base pairs in the core is four times higher than at the periphery and consists mainly of interchain pairs (19-20 pairs per chain). The lifetimes of pairs and chain contacts increase with age. Comparison with a random sequence showed that it ages 4-5 times more slowly, has a nearly uniform distribution of base pairs, and is less entangled, despite a similar density.
The work shows that the aging of RNA condensates requires no external triggers or special tuning of parameters: it is encoded in the sequence and arises from interchain interactions. This sets it apart from neighboring studies, which often looked at the aging of protein condensates or used models with fitting parameters. Here, for the first time, spatial heterogeneity of aging is shown to arise from a gradient in base pairing rather than from a preordained structure. The results connect molecular interactions with mesoscopic material transitions and offer a general principle: a feedback loop in which base pairing swells the chains, swelling increases entanglement, entanglement slows relaxation, and slowing strengthens the network.
The field could move toward studying longer sequences, the effects of RNA modifications, helicase activity, and changing ionic conditions on aging. Multiscale approaches that combine coarse-grained modeling with experiments will help test the predictions on biologically relevant timescales. It is also important to investigate whether cells can regulate aging through sequence design or active processes to keep condensates in a functional state.
The results matter for understanding neurodegenerative diseases linked to CAG repeat expansion, where RNA foci may harden irreversibly. They also provide a foundation for designing synthetic RNA condensates with tailored properties.
Test the predictions in experiments with RNA droplets, for instance using FRAP and microrheology. Investigate how other repeat sequences and environmental conditions affect aging.
In the field of biomolecular condensates, the question of how sequence determines material properties and their evolution remains open. This work links aging to specific intermolecular interactions, but it does not solve the problem of reversibility of aging in the cell or the role of active processes such as the action of ATP-dependent helicases.
In the core of the droplet, relaxation times are six times longer than at the periphery, and the density of entanglement nodes is five times higher. Yet both sequences have the same nucleotide composition but a different order, and that alone is enough for one to age five times faster than the other.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.