Recent experiments have established that condensate interfaces are not passive bystanders in amyloid formation. In condensates formed by the low-complexity domain of the RNA-binding protein hnRNPA1, and in droplets of the ALS-linked protein FUS, fibril formation is strongly enhanced at the droplet surface rather than in the interior. Other condensate systems suppress aggregation by sequestering proteins inside, or redirect it by altering the structural polymorphism of amyloid-forming proteins like Tau and Aβ. What has remained unclear is which molecular and physical variables select among these outcomes. The new work tackles this question with a purpose-built simulation framework called flux-driven molecular dynamics, or FD-MD, which allows the researchers to independently tune chain architecture, chain rigidity, and the transport of molecules toward a condensate interface.

The model is deliberately minimal. Each protein is represented as a 25-bead coarse-grained polymer with two bead types arranged in a repeating block pattern, analogous to a simple block copolymer. Like beads attract one another through Lennard-Jones potentials, while unlike beads repel. Crucially, the two internal blocks of cohesive beads are stiffened by harmonic angular potentials, making them behave like rigid rods that stand in for β-prone, aggregation-prone motifs, while the intervening linkers remain floppy. This design separates cohesion from stiffness: a chain can be made more rigid without increasing its attraction strength, and the placement of its sticky motifs can be rearranged without changing the total amount of cohesive material. The simulations do not resolve residue-level chemistry, strand registry, or hydrogen-bond directionality, so the filaments they produce are coarse-grained analogs of amyloid rather than atomically resolved structures.

The simulation geometry mimics a driven cellular environment. Two semi-arrested condensate surfaces, built from 8,000 chains of the same molecular species, face each other across a dilute region. Up to 20,000 additional chains are then introduced at a prescribed deposition rate, each given an initial velocity directed toward the nearest interface—an initial kick that mimics the chemical-potential gradient dragging molecules toward the droplet surface in a living cell. Without this bias, inserted molecules would simply diffuse and aggregate in the dilute phase. Together, the deposition rate and the initial drift velocity define a minimal flux protocol for probing how non-equilibrium delivery shapes interfacial assembly.

The first major finding concerns rigidity. When chains with stiff β-prone segments were deposited onto the condensate interface, they formed anisotropic protrusions that extended away from the surface, elongated over time, bundled together, and eventually formed mesh-like structures capable of bridging the two opposing interfaces. Chains with the identical bead pattern but floppy backbones accumulated at the interface just as readily, yet relaxed into laterally spreading, disordered surface aggregates with no fibrillar character. Because the two simulations differed only in angular stiffness, the comparison identifies backbone rigidity as the variable that converts ordinary interfacial adsorption into directional fibrillar growth. When the researchers scrambled the placement of cohesive beads at fixed composition, the chains failed to form filaments altogether and instead promoted the growth of extended planar surfaces, underscoring that motif patterning matters as much as motif count.

Why does stiffness matter so much? The answer lies in the orientational field imposed by the interface itself. By measuring a nematic order parameter as a function of distance from the surface, the researchers found that condensate interfaces organize arriving chains into a layered structure before any large-scale assembly begins. Segments that directly contact the surface lie flat in the interfacial plane, approaching the theoretical planar limit of the order parameter within the first few bead diameters. Beyond this contact layer, the order parameter does not decay monotonically; instead, it alternates with a characteristic spacing of roughly eight bead diameters, with positive maxima indicating segments preferentially aligned along the interface normal. This spacing matches the separation between successive rod-like units along an extended chain, meaning that outer segments of surface-anchored chains project into the dilute region while their leading segments remain adsorbed. The result is a layered architecture—a planar contact layer followed by segments oriented outward—that is precisely the local arrangement a protruding fibril needs in order to elongate.

The second major axis of control is molecular flux. Material arriving at a condensate interface can be incorporated in two geometrically distinct ways: it can spread laterally, increasing surface coverage, or it can extend outward, increasing the height and volume of the growing assembly. Tracking the exposed area and convex-hull volume of the largest adsorbed cluster, the researchers showed that increasing the initial drift velocity dramatically accelerates surface coverage—reducing the time to reach a coverage shoulder from roughly 1,000 chain insertions at the slowest drift to about 300 at the fastest. But faster arrival came with a cost. When the researchers measured the elongation rate of individual filaments along their principal axis, they found that the longitudinal growth rate decreased systematically as the initial drift velocity increased, while the transverse thickening of the filaments depended only weakly on it. Faster molecules, in other words, cover the surface quickly but are less efficiently incorporated into the growing fibril tip.

The likely explanation is a competition between two fates for incoming chains. Chains that attach along the lateral surface of an existing protrusion mainly increase its radial extent, whereas chains captured near the filament tip contribute directly to axial elongation. A large initial drift velocity promotes rapid interfacial capture and surface coverage, starving the protrusion tip of material that could otherwise drive extension. The deposition rate and drift velocity act together to select the growth trajectory: frequent deposition favors protrusive growth when the initial drift is weak, whereas large drift velocities bias the system toward lateral spreading and suppress out-of-plane extension. Aging in this model, the authors conclude, is not a single pathway but a family of non-equilibrium trajectories selected by how material is delivered to, and captured by, the condensate interface.

The broader implications are considerable. Because sequence-encoded rigidity and molecular delivery are, in principle, independently tunable—through protein engineering on one hand and through expression level, exchange kinetics, transport, or chemical remodeling on the other—the framework suggests that aging trajectories of condensates could be redirected by modulating molecular delivery alone, without altering the protein sequence itself. The authors also propose mesoscale observables that bridge their simulations to experiments: convex-hull geometry to separate lateral coverage from anisotropic extension, a molecular loading ratio that reports how accumulated material distributes over the exposed aggregate area, and the gyration-tensor spectrum that identifies persistent symmetry breaking along the filament axis. These quantities do not require residue-level resolution and can be compared directly with microscopy-based morphology measurements.

The researchers are careful to note the limits of their abstraction. The model does not define a critical nucleus or measure nucleation rates and barriers, so they use the term interfacial fibril initiation rather than claiming to resolve a specific nucleation pathway. Each point in their morphology map corresponds to a single trajectory, making the map descriptive rather than a thermodynamic phase diagram, and the fibrils themselves are coarse-grained assemblies stabilized by cohesion and rigidity rather than residue-resolved amyloid. Still, the core message is clear and testable: the same protein can age along entirely different morphological routes simply because the rate and manner of molecular delivery change. If this separation between structural competence and transport control holds in cellular condensates, it opens a practical route to intervening in the earliest stages of amyloid disease—not by redesigning the proteins involved, but by changing how fast they arrive at the droplet surface where fibrils are born.

Subject of Research: Simulation study of how chain rigidity, motif patterning, and molecular flux control fibril growth at biomolecular condensate interfaces

Article Title: β motifs and molecular flux promote fibril growth at condensate interfaces

Article References: Biswas, S., & Potoyan, D. A. (2026). β motifs and molecular flux promote fibril growth at condensate interfaces. iScience, 29(11), Article 117855. https://doi.org/10.1016/j.isci.2026.117855

Image Credits: AI Generated

DOI: Not provided

Keywords: biomolecular condensates, amyloid fibrils, phase separation, molecular dynamics, protein aggregation, condensate aging, β-sheet motifs, molecular flux, hnRNPA1, FUS, neurodegenerative disease, coarse-grained simulation