Nine-residue peptides tile into a honeycomb lattice with 5-nanometre water channels

Lateral growth is the part that has resisted design. Short peptides elongate readily into one-dimensional fibrils, but their side interfaces are polymorphic, which has prevented controlled tiling into higher-dimensional lattices ^[12]. Controlled two- or three-dimensional lattices built from minimal linear beta-sheet peptides are rare ^[16]. The routes that did work leaned on more scaffolding: coiled-coil proteins and collagen mimetics with large pre-organised interfaces ^[13], and peptide nanochannels from cyclic scaffolds held in shape by covalent backbone pre-organisation ^[14].

DILT1 gets both interfaces out of one short chain. The N-terminal K1-V2-K3-V4 stretch pairs face to face into a cross-beta dimer with hydrophobic packing, which sets the hydrophobic contact surface for lateral organisation and, through backbone hydrogen bonding, the axial stacking interface ^[8]. A trimeric junction at the other end encodes lateral connectivity and symmetry, and a central inversion residue orients the two motifs relative to each other ^[11]. The prototype sequence is K1-V2-K3-V4-S5-Q6-I7-N8-M9, and the family name stands for dimer-inversion-lock trimer ^[9]. Four residues to the dimer, one to the inversion, four to the junction ^[15]. The cross-beta segment was inspired by the MAX1 oligopeptide family ^[10].

The rules come from the perturbation series. Changing positions one at a time let the authors tie residue position to lattice symmetry, to lattice propagation and to channel topology ^[6]. Cryo-EM resolved the assemblies and found symmetry and pore geometry conserved across the variants ^[5]. Those two results point in slightly different directions: the sequence governs whether and how the honeycomb propagates, and the channel it leaves behind comes out around 5 nm across each time ^[4].

The water in those channels was established by molecular dynamics simulation and by vibrational spectroscopy ^[7]. Both speak to whether water occupies the pore; a permeation or rejection measurement is a different experiment.

In my view the work establishes the first half of a design cycle: a nine-residue family whose supramolecular symmetry and connectivity follow from stated residue positions, checked in the hydrated state by cryo-EM ^[5][6]. Writing a chosen channel width into that sequence is a second claim, and the variants reported here all came out with the same pore ^[5].