How does environmental context shape FUS protein phase separation?

Protonuclei recreate a crowded, confined nucleus

The paper's central problem is that test-tube assays strip away the conditions that define the cell nucleus: crowding, many weak interactions at once, and a mechanically active interior. To reintroduce those conditions under control, the authors built programmable DNA-based protonuclei (PN), compartments assembled from DNA strands that form a shell and a DNA-rich core. FUS, a neurodegeneration-linked protein, is loaded into these compartments and its phase separation is triggered by cleaving an MBP solubility tag with TEV protease.

The PN design makes the environment itself an experimental variable. The core can carry different DNA or RNA sequences, the shell confines the interior, and crosslinking strands can stiffen the core. That lets the study ask whether FUS condensation follows the protein's intrinsic binding preferences or the physical context in which those interactions occur.

Binding affinity does not predict FUS partitioning

The most decision-relevant comparison is between conventional affinity measurements and behavior inside the PN. Electrophoretic mobility shift assays (EMSAs) measure how well FUS binds a given nucleic acid sequence in dilute solution. Partitioning instead measures how much FUS concentrates inside the PN relative to outside it. The two measures diverged sharply.

Two duplexes with similar EMSA affinities produced very different partitioning: the T20-PN sample reached a partition coefficient of 743, while the m/m* sample reached 36, roughly a 20-fold difference. The m/m* core also showed high binding affinity in EMSA yet the lowest partitioning at the PN level. The authors interpret this as evidence that multivalency, crowding and confinement reshape protein-nucleic acid interactions, so a dilute-solution affinity map cannot be used as a proxy for behavior in a nucleus-like compartment.

FUS loading ratio sets three condensate morphologies

Within pristine PN, the ratio of FUS-GFP to repeating units of the core DNA strand, called the loading ratio, determined which morphology appeared. At low loading ratios of about 0.30 and below, FUS formed droplets inside the PN, a pattern the authors describe as binodal phase separation. At intermediate ratios of roughly 0.60 to 0.90, the condensate became co-continuous, with a wetting layer at the core-shell interface and a depletion layer, resembling spinodal decomposition. At high ratios of about 1.20 and above, droplets of the dilute phase appeared within the dense phase, the inverted region of the phase diagram.

A continuum model treating the PN and its contents as order-parameter fields reproduced this sequence, including surface-directed phase separation when a mild attraction between DNA and the inner shell wall was assumed. The agreement matters because it links the observed morphologies to a physical mechanism rather than to an unexplained list of outcomes.

Core dynamics control condensate fate

Changing the core sequence changed both the speed and the shape of FUS condensation. In pristine PN, FUS-GFP droplets appeared within five minutes of MBP cleavage and later wet the inner shell surface. Adding a dangling RNA strand kept condensates in the PN center, prevented leakage and droplet formation outside the shell, and improved FUS-GFP retention after five hours. Fluorescence recovery after photobleaching showed that FUS-GFP in those condensates partially recovered after one day while the core channel remained non-dynamic, indicating that FUS can exchange RNA partners within the core.

Other core modifications produced interconnected FUS networks rather than droplets. The authors attribute this to viscoelastic phase separation: the slowly relaxing DNA core sustains mechanical stress, so the minority FUS phase forms a network instead of discrete droplets. In the T20-PN sample the arrested structure persisted for two days, while in the U20-RNA and T20-m samples the network relaxed slowly over the same span.

Crosslinking suppresses the liquid-to-solid transition

The clearest test of the physical-environment argument comes from tuning DNA crosslinking in the core. With 25% crosslinked p(A20-m-XL), FUS phase separation appeared within five minutes but stayed grainy and only slightly coarsened over three days, forming a spongy arrested network instead of a coarsened FUS-rich phase. At 50% crosslinking, FUS phase separation was barely visible after six hours and progressed little over the next three days. At 75% crosslinking and above, no microscopically visible FUS phase separation occurred over three days.

The authors explain this through dynamic asymmetry: crosslinks raise the energetic and kinetic cost of redistributing the DNA core during FUS condensation, restricting domain growth and relaxation. In their interpretation, the stiffened core suppresses the liquid-to-solid transition that is a hallmark of disease-linked FUS condensates. This is a statement about the PN model system, not a demonstrated therapeutic effect in cells.

Where the protonuclei result applies

The PN platform sits deliberately between dilute test-tube assays and live cells. It captures nucleic acid crowding, compositional heterogeneity and tunable viscoelasticity, but it is a synthetic reconstruction, not the nucleus itself. The FUS constructs are MBP-FUS-GFP and MBP-FUS, and phase separation is initiated by TEV cleavage, so the observed kinetics and morphologies belong to this reconstituted setup. The crosslinking experiments likewise describe how an artificial DNA core's mechanics alter FUS behavior, not how chromatin stiffness acts in a living cell.

Within those bounds, the paper's practical priority is clear: when judging whether a nucleic acid sequence will promote FUS condensation, dilute-solution affinity is an unreliable guide. The variables that mattered here were the sequence and structure of the core nucleic acid, the loading ratio of FUS to core repeating units, and the crosslink-dependent dynamics of the core. The 20-fold partitioning gap between two similarly binding duplexes, and the disappearance of visible FUS phase separation at 75% crosslinking, are the comparisons most worth carrying forward.

Sources used in this answer