Proteins no longer fold one way: after terminal remodeling, two or three mutations shift the conformational balance

Protein design can now target two distinct folded states of one sequence rather than a single static structure. Computational design previously produced mainly single stable conformations, and the few dynamic designs relied on external triggers such as ligands or ions and involved only minor rearrangements. Starting from 93 native domains, terminal remodeling and filtering left 58 folds and 2,291 native-alternative pairs; about 30% of alternative-state designs were predicted by AlphaFold2 to fold as designed with high confidence, and most designs retained more than 50% sequence identity to the parent.

Two complementary sequence-design routes were used to find amino-acid sequences compatible with both backbones. One route constrained the sequence to the space between the two target-state sequences, while the other placed no restriction on amino-acid identity and optimized only against the two target backbones. The chimeric route used ProteinMPNN in binary mode over RFdiffusion transition trajectories, and the full-redesign route used iterative ProteinMPNN exchanging a logits bias until convergence to a single sequence; designs were expressed in E. coli, confirmed folded by SEC-MALS and circular dichroism, and probed by NMR, MD and BioEmu.

The conformational equilibrium can be shifted in a directed way by a few mutations, stabilizing states that were hard to observe for structure determination. The designed metamorphic sequences were conformationally heterogeneous with broadened crosspeaks, making either state difficult to resolve directly. Mutants selected with AlphaFold3 and BioEmu improved HSQC dispersion and peak counts; the H2H-chimeric alternative-state crystal structure matched the design model at 1.8 Å backbone RMSD, the B2H-full alternative-state NMR structure was 2.2 Å from the AF3 model, the B2H-full native-state crystal structure was 2.6 Å from the AF3 model, and MD retained the helical conformation of a stabilized mutant up to 15 μs.

The demonstrated scope is generating alternative conformations from small native domains, designing sequences that sample both states, and using state-stabilizing mutations to shift the equilibrium toward either state so that crystal or NMR structures can be obtained. It is intended for researchers working on protein design, conformational dynamics and NMR characterization. The source proposes as next steps controlling the relative free energies of the two states to tune populations and controlling the barrier between them to tune exchange rates.

The designed metamorphic proteins are themselves highly dynamic, with broadened and partly missing NMR crosspeaks; only about 50% of residues in H2H-chimeric could be assigned, and B2H-full showed broad, heterogeneous resonances. The redesigned segment of H2H-full was predominantly disordered rather than sampling the intended folded state. BioEmu did not show two clearly defined minima for B2H-full but favored the alternative state, and structure predictors were not developed as quantitative predictors of equilibrium populations, so their predictions still require experimental validation.