Protein tangles change the pace at which a gel gives way

A gel’s softness answers only part of the question of how it feels. Hold it at a fixed deformation and the force it exerts can diminish as its molecular network rearranges. That loss of stress over time is stress relaxation. Two materials that initially resist a push can therefore behave differently when the push lasts.[1]

Puqing Deng and colleagues set out to design one molecular source of that timing: pairs of protein chains that intertwine. Their TangleDiff system, reported in Nature Communications on 21 September, generates candidate protein sequences with specified ranges of binding energy. The team made selected proteins and tested the gels they formed, connecting an AI design choice to a material property measured in the laboratory.[1]

A tangle that can rearrange

A hydrogel is a water-rich molecular network. Its connections restrict the movement of its chains, while rearrangements allow mechanical energy to dissipate. A permanent covalent junction and a pair of physically entangled protein domains provide different kinds of connection. The latter can rearrange as the chains and their contacts move, giving the material a route to relax stress.[1]

TangleDiff works in sequence space. The researchers first screened natural protein sequences computationally, using predicted structures to identify candidate entangled pairs. A dataset of 79,890 sequences supplied training material. The generative model uses a protein-language representation and a diffusion process that progressively removes noise, with a binding-energy condition steering the sequences it generates. Those energies are calculated design targets, not direct measurements of the final gel’s behaviour.[1]

For the experimental search, the team generated 4,000 sequences in each of three binding-energy groups. Computational filters checked predicted folding, solubility and entanglement before three designs from each group were chosen for laboratory work. Eight of the nine tested designs formed dimers, meaning pairs of protein molecules. Seven of those eight produced hydrogels in both of the network arrangements the researchers tested.[1]

That distinction matters because generation and material formation are separate achievements. A sequence can look suitable computationally yet fail to form the required pair, or form a pair that does not produce the intended network. The experimental result concerns the nine selected designs after screening, rather than a measured success rate for all 12,000 generated sequences.[1]

Two ways to connect the same proteins

In one arrangement, complementary protein components joined covalently into long, linear chains. The designed entangled domains supplied physical connections between those chains. In the other, three-way components added permanent branches to the network. The seven successful designs therefore yielded fourteen gel formulations, allowing the same entangled domain to be examined with and without those extra junctions.[1]

The researchers oscillated the gels and measured how much mechanical energy they stored and dissipated. They then inferred continuous relaxation spectra from those frequency measurements. A peak identifies a characteristic relaxation timescale; it is not the instant at which every part of the gel has finished relaxing. The gels contained 8% protein by weight, and the frequency sweeps used 5% strain at room temperature. The paper reports three independent repetitions of gel preparation and rheological testing.[1]

The added permanent branches moved both characteristic peaks towards longer times in the published curves. Yet C4_370 remained faster-relaxing than C2_1020 in either architecture. The choice of protein and the way it is connected both contribute to the material’s timing.[1][2]

Permanent junctions also increased the elastic contribution: the branched networks stored more energy, and their storage modulus stayed above their loss modulus throughout the measured frequency range. They could dissipate energy while retaining a connected, solid-like framework. The linear networks made with C4_370 and C2_1020 crossed towards liquid-like behaviour at low frequencies.[1]

Binding energy is only part of the design

The strongest-binding group showed no clear relaxation peaks within the tested window. Three other designs, including members of the weaker and intermediate groups, also lacked the expected peaks. The authors suggest that structural differences and additional intermolecular interactions could explain those exceptions. Binding energy helped organise the search, but it did not uniquely determine the material’s response.[1]

A useful control was C4_1771, the selected design that remained monomeric. Both attempted network arrangements made from it stayed as viscous liquids, with much lower elastic response and no relaxation peaks. That result supports a mechanical role for the paired, entangled domains. A separate molecular-linking and cleavage experiment on C4_370 also supplied evidence that its two chains were physically entangled.[1][3]

The achievement is a route from generated sequence to a gel with a different response to sustained force. Its timing comes from the interactions of the designed proteins with the wider network, so the architecture of the material remains part of the design. A useful protein for this job has to do more than fold: it has to let a connected material rearrange at a useful pace.[1][2]

Report an error or suggest a correction