While deep learning has revolutionized protein structure prediction, almost all experimentally characterized de novo protein designs have been generated using physically based approaches such as Rosetta. Here we describe a deep learning–based protein sequence design method, ProteinMPNN, with outstanding performance in both in silico and experimental tests. On native protein backbones, ProteinMPNN has a sequence recovery of 52.4%, compared to 32.9% for Rosetta. The amino acid sequence at different positions can be coupled between single or multiple chains, enabling application to a wide range of current protein design challenges. We demonstrate the broad utility and high accuracy of ProteinMPNN using X-ray crystallography, cryoEM and functional studies by rescuing previously failed designs, made using Rosetta or AlphaFold, of protein monomers, cyclic homo-oligomers, tetrahedral nanoparticles, and target binding proteins.

P-glycoprotein (Pgp or multidrug-resistance protein) shows drug-stimulated ATPase activity. The catalytic sites are known to be of low affinity and specificity for nucleotides. From the sequence, two nucleotide sites are predicted per Pgp molecule. Using plasma membranes from a multidrug-resistant Chinese hamster ovary cell line, which are highly enriched in Pgp, we show that vanadate-induced trapping of nucleotide at a single catalytic site produces stably inhibited Pgp, with t 1/2 for reactivation of ATPase activity of 84 min at 37 degrees C and >30 h at 4 degrees C. Reactivation of ATPase correlated with release of trapped nucleotide. Concentrations of MgATP and MgADP required to produce 50% inhibition were 9 and 15 microM, respectively, thus the apparent affinity for nucleotide is greatly increased by vanadate-trapping. The trapped nucleotide species was ADP. Divalent Cation was required, with magnesium, manganese, and cobalt all effective: cobalt yielded a very stable inhibited species, t1/2 at 37 degrees C = 18 h. No photocleavage of Pgp was observed after vanadate trapping with MgATP, nor was UV-induced photolabeling of Pgp by trapped adenine nucleotide observed. Vanadate-trapping with 8-azido-ATP followed by UV irradiation caused permanent inactivation and specific labeling of Pgp. Vanadate-induced inhibition was also shown with pure, reconstituted Pgp, with similar characteristics to those in plasma membranes. Vanadate trapping overcomes technical difficulties posed by lack of high affinity nucleotide-binding site(s) or a covalent enzyme-phosphate catalytic intermediate in Pgp. The finding that vanadate trapping of nucleotide at just one site/Pgp is sufficient to give full inhibition at ATPase activity shows that the two predicted nucleotide sites can not function independently as catalytic sites.

The regular arrangements of β-strands around a central axis in β-barrels and of α-helices in coiled coils contrast with the irregular tertiary structures of most globular proteins, and have fascinated structural biologists since they were first discovered. Simple parametric models have been used to design a wide range of α-helical coiled-coil structures, but to date there has been no success with β-barrels. Here we show that accurate de novo design of β-barrels requires considerable symmetry-breaking to achieve continuous hydrogen-bond connectivity and eliminate backbone strain. We then build ensembles of β-barrel backbone models with cavity shapes that match the fluorogenic compound DFHBI, and use a hierarchical grid-based search method to simultaneously optimize the rigid-body placement of DFHBI in these cavities and the identities of the surrounding amino acids to achieve high shape and chemical complementarity. The designs have high structural accuracy and bind and fluorescently activate DFHBI in vitro and in Escherichia coli, yeast and mammalian cells. This de novo design of small-molecule binding activity, using backbones custom-built to bind the ligand, should enable the design of increasingly sophisticated ligand-binding proteins, sensors and catalysts that are not limited by the backbone geometries available in known protein structures.

In nature, structural specificity in DNA and proteins is encoded quite differently: in DNA, specificity arises from modular hydrogen bonds in the core of the double helix, whereas in proteins, specificity arises largely from buried hydrophobic packing complemented by irregular peripheral polar interactions. Here we describe a general approach for designing a wide range of protein homo-oligomers with specificity determined by modular arrays of central hydrogen bond networks. We use the approach to design dimers, trimers, and tetramers consisting of two concentric rings of helices, including previously not seen triangular, square, and supercoiled topologies. X-ray crystallography confirms that the structures overall, and the hydrogen bond networks in particular, are nearly identical to the design models, and the networks confer interaction specificity in vivo. The ability to design extensive hydrogen bond networks with atomic accuracy is a milestone for protein design and enables the programming of protein interaction specificity for a broad range of synthetic biology applications.

No abstract