Accurate computational design of three-dimensional protein crystals

Abstract

Protein crystallization plays a central role in structural biology. Despite this, the process of crystallization remains poorly understood and highly empirical, with crystal contacts, lattice packing arrangements and space group preferences being largely unpredictable. Programming protein crystallization through precisely engineered side-chain-side-chain interactions across protein-protein interfaces is an outstanding challenge. Here we develop a general computational approach for designing three-dimensional protein crystals with prespecified lattice architectures at atomic accuracy that hierarchically constrains the overall number of degrees of freedom of the system. We design three pairs of oligomers that can be individually purified, and upon mixing, spontaneously self-assemble into >100 µm three-dimensional crystals. The structures of these crystals are nearly identical to the computational design models, closely corresponding in both overall architecture and the specific protein-protein interactions. The dimensions of the crystal unit cell can be systematically redesigned while retaining the space group symmetry and overall architecture, and the crystals are extremely porous and highly stable. Our approach enables the computational design of protein crystals with high accuracy, and the designed protein crystals, which have both structural and assembly information encoded in their primary sequences, provide a powerful platform for biological materials engineering.

Extended Data Fig. 1 |. Design rules of 3D protein crystals.

a, Constrained degree of freedom (DOF): The angle of rotation at the designed dihedral crystal interface (Fig. 1a, right panel) must be precisely specified by the design process, where the C2 axis of the dihedral needs to coincide with the C2 axis of the space group. In this example, the disruptive effect (highlighted in red) of a 15-degree error in alignment on crystal assembly is illustrated; similar crystal lattice breakdowns occur with all deviations from the target alignment angle. b, Accessible secondary structure (SS): Dihedral interfaces with helices perpendicular to the symmetry axis (docked from T33–15 cage) are more designable than those with helices parallel to the symmetry axis (docked from T33–21 cage). Interacting secondary structures are highlighted in red. c, Affinity and Specificity: Working interfaces have sufficient hydrophobic packing with specific polar interactions at the boundary. Highly hydrophobic interfaces destruct the designed self-assembly, including insoluble components and off-target assemblies.

Extended Data Fig. 2 |. Characterizations of the constituent cages of designed crystals.

a-d, T33–15-D3–4, e–h, T32–15, i-l, O43–2. a,e,i, SEC chromatograms of two oligomeric components (green and orange) and cages assembled via in-vitro mixing of components (blue). b,f,j, nsEM images (scale bars, 50 nm). c,g,k, overlay of the design model with 3D reconstructed nsEM density map/cryoEM model (scale bars, 5 nm). d,h,l, SAXS profile and simulation results of cages.

Extended Data Fig. 3 |. Characterizations of new tetrahedral cages for crystal design.

a–e, from left to right, computational model, SEC chromatogram, SAXS profile, and nsEM images.

Extended Data Fig. 4 |. Characterizations of new octahedral cages for crystal design.

a–d, from left to right, computational model, SEC chromatogram, SAXS profile, and nsEM images.

Extended Data Fig. 5 |. Symmetric dockings of tetrahedral and octahedral cages into crystal lattices.

a, Two tetrahedral cages are docked along their C3 axis for crystal contacts of D3 dihedrals, which allow them to crystallize in the F4₁32 space group. b, Two octahedral cages are docked along their C3 axis for crystal contacts of D3 dihedrals, which allow them to crystallize in the I432 space group. See methods for detailed docking protocol.

Extended Data Fig. 6 |. Optical microscopy and cryoEM characterization of designed protein crystals.

a, Optical micrograph of F4₁32–1-0 crystals. b, Optical micrograph of F4₁32–1 crystals. c, CryoEM image of F4₁32–1 crystals. d, Optical micrograph of F4₁32–2-6H crystals. e, Optical micrograph of F4₁32–2 crystals. f, CryoEM image of F4₁32–2 crystals. g, Optical micrograph of I432–1 crystals. h, Optical micrograph of I432–1-CC crystals. i, CryoEM image of I432–1 crystals.

Extended Data Fig. 7 |. CryoEM data of the T32–15 cage.

a, Representative 2D class averages of the T32–15 cage. b, CryoEM local resolution map of the T32–15 cage (top) and built atomic model (bottom). Local resolution estimates range from ~2.5 Å at the core to ~4 Å along the crystal-contact forming helices. c, Map-to-model comparison within a low-resolution region (top) and a high-resolution region (bottom). d, Global FSC. e, Orientational distribution plot demonstrating full angular sampling.

Extended Data Fig. 8 |. Tuning the crystallization behavior of designed crystals by mutagenesis.

a, Mutations to the F4₁32–1 crystals. b, Mutations of F4₁32–2 crystals. c, Mutations and redesigns (orange) of I432–1 crystals. Top panels, crystal interface models based on X-ray structure. Interface side chains are hypothetically placed to demonstrate mutation sites. Bottom panels: optical micrographs of representative crystallization results. Scale bars, 100 µm.

Extended Data Fig. 9 |. Design pipeline for engineering crystal unit cell dimension.

The crystal contact of the F4₁32–2 crystal was redesigned with different DHR arm fusion. See Methods for the details of step a-g.

Fig. 1 |. Hierarchical crystal design strategy.

a, Schematic illustration of the three-step design hierarchy for a diamond lattice (F4₁32 space group) formed from a tetrahedral polyhedron built from a C2 dimer (grey) and a C3 trimer (cyan). Monomers (first column) are docked into cyclic dimers and trimers (second column), which are docked into a two-component cage (third column), which is then arrayed in a 3D lattice (fourth column). b, Interfaces driving crystal assembly. For the designed crystal example in a, the three designed interfaces on the trimer component are shown that drive the assembly of the cyclic oligomer (blue), the tetrahedron (orange) and the crystal lattice (red). c, Interfaces mapped to the monomer in b are shown between interacting partners for the trimer (left), tetrahedral cage (middle) and crystal (right). To maximize assembly cooperativity, the interface size (BSA, buried surface area) and affinity (Rosetta calculated ddG) decrease through the design hierarchy; the number of system degrees of freedom available for sampling at each step (Methods) decreases in parallel, making the design challenge more difficult.

Fig. 2 |. Computational design and experimental characterization of F4₁32–1, F4₁32–2 and I432–1 crystals.

a, Construction of F4₁32–1 crystals from cyclic oligomers. In the second step, symmetry elements of the cage are superimposed with corresponding symmetry elements of the unit cell. b, CryoEM images of F4₁32–1 crystals (scale bar, 100 nm) and optical micrographs of single crystals (inset, scale bar, 100 µm). c, SAXS profiles of F4₁32–1 microcrystals (orange) compared to simulated profiles computed from the design model (grey). d, Computational design model of F4₁32–1 crystals (left, orange and grey) spliced to experimentally determined crystal structure (right, sky blue and light blue). The flanking panels are enlargements of the four designed interfaces, with the design model superimposed on the crystal structure. e, Construction of F4₁32–2 crystals from cyclic oligomers. In the second step, symmetry elements of the cage are superimposed with corresponding symmetry elements of the unit cell. f, CryoEM images of F4₁32–2 crystals (scale bar, 100 nm) and optical micrographs of single crystals (inset, scale bar, 100 µm). g, SAXS profiles of F4₁32–2 microcrystals (teal) compared to simulated profiles computed from the design model (grey). h, Computational design model of F4₁32–2 crystals (left, teal and grey) spliced to experimentally determined crystal structure (right, sky blue and light blue). The flanking panels are enlargements of the four designed interfaces, with the design model superimposed on the crystal structure. i, Construction of I432–1 crystals from cyclic oligomers. In the second step, symmetry elements of the cage are superimposed with corresponding symmetry elements of the unit cell. j, CryoEM images of I432–1 crystals (scale bar, 100 nm) and optical micrographs of single crystals (inset, scale bar, 100 µm). k, SAXS profiles of I432–1 microcrystals (pink) compared to simulated profiles computed from the design model (grey). l, Computational design model of I432–1-CC crystals (left, pink and grey) spliced to experimentally determined crystal structure (right, sky blue and light blue). The flanking panels are enlargements of the four designed interfaces, with the design model superimposed on the crystal structure. There is a close agreement in all three cases. d,h,l, Left top and bottom: the two cyclic oligomer interfaces. Top right: interface between cyclic oligomers that generates the polyhedral cage. Bottom right: interface between polyhedra that generates the crystal. All-atom RMSDs are calculated for each symmetry unit and summarized in Supplementary Table 5. Exp., experimental; Sim., simulated.

Fig. 3 |. Engineering crystal properties.

a,b, Tuning the unit cell dimensions of the F4₁32–2 crystals by design. a, Design models and optical micrographs. Scale bars, 50 µm. b, SAXS profile. Peaks of the same index are connected by dashed lines. c, Space group distribution over all crystals in the PDB. Inset, distribution of the crystal solvent content with an enlargement of the high solvent content region versus crystal resolution. Crystals designed in this paper are highlighted in orange. d, I432–1-CC crystals incubated at 95 °C for 1 h (left panel) and autoclaved at 121 °C, 13 psi for 40 min (right panel). Scale bars, 100 µm. e, I432–1-CC crystals formed overnight by mixing of E. coli lysates. Scale bar, 50 µm. Des., designed.

Fig. 4 |. Scaffolding of 3D AuNP superlattices using the designed protein crystals.

a, 5 nm AuNPs are encapsulated inside the I432–1-CC cages by a metal coordination interaction (Supplementary Fig. 16). The AuNP-encapsulated cages further self-assembly into designed I432–1 crystal lattices. b, AuNP superlattice single crystals exhibit reversible contraction and expansion during drying and rehydration cycles. c, Occupancy can be controlled for AuNP superlattice single crystals. 100% and 13% AuNP encapsulations are shown. Crystal scaffolds are in grey and AuNPs are drawn as golden spheres. d, Optical micrograph of crystals with AuNP encapsulation. Scale bar, 50 µm. Inset, representative 2D nsEM class average of AuNP-encapsulated O43–2 cage. Scale bar, 10 nm. e, NsEM micrographs (top panel) and 2 × 2 × 2 unit cell crystal model (bottom panel) of different crystals facets: ⟨100⟩ (left), ⟨111⟩ (middle) and ⟨110⟩ (right). Scale bars, 50 nm. f, Dehydration and rehydration cycles of AuNP superlattice single crystals. The size changes of seven crystals were measured and plotted.Representative optical micrographs of the crystals are shown for each hydration state. g, Experimental SAXS profiles of hydrated (solid red line) and dried (dashed red line) crystals with 100% AuNP encapsulation. Simulated patterns (solid grey line) of a superlattice formed by 5.6 nm diameter AuNPs with a lattice parameter of 236.3 Å closely match the experimental results. h, Optical micrographs of crystals with different AuNP encapsulation ratios (13%, 28%, 50% and 100%) in hydrated (solid lines) and dried states (dashed lines). Scale bars, 20 µm. Representative scanning electron microscope (SEM) images of dried crystals are shown on the right. Scale bars, 50 nm. i, Representative hyperspectral dark-field scattering spectra for crystals with AuNPs. The orange arrow shows the redshift of the scattering peak with increasing AuNP encapsulation ratio. The blue arrow illustrates the reversible shift of the scattering peak between dried crystals and hydrated crystals (dried crystals are redshifted). Colours and line styles correspond to samples in g.