De novo design of immunoglobulin-like domains

Abstract

Antibodies, and antibody derivatives such as nanobodies, contain immunoglobulin-like (Ig) β-sandwich scaffolds which anchor the hypervariable antigen-binding loops and constitute the largest growing class of drugs. Current engineering strategies for this class of compounds rely on naturally existing Ig frameworks, which can be hard to modify and have limitations in manufacturability, designability and range of action. Here, we develop design rules for the central feature of the Ig fold architecture-the non-local cross-β structure connecting the two β-sheets-and use these to design highly stable Ig domains de novo, confirm their structures through X-ray crystallography, and show they can correctly scaffold functional loops. Our approach opens the door to the design of antibody-like scaffolds with tailored structures and superior biophysical properties.

Fig. 1. Topology of immunoglobulin-like domains.

a Three-dimensional cartoon representation of an Ig structure formed by seven β-strands (left); and backbone hydrogen bond patterns (annotated thin lines) between paired β-strands along the sequence (right). Cross-β interactions have higher sequence separation (and higher contact order) than β-hairpins, which slows down folding. b β-arches of the cross-β motif belong to two contiguous and distinct Greek key motifs: with 2 β-strands in each β-sheet (left); and with 3 β-strands in one β-sheet and 1 β-strand in the other (right). From the folding and design perspective, the main limiting factor for correctly assembling the Ig structure is formation of the cross-β motif, since the three β-hairpins can form independently of one another.

Fig. 2. Design rules for cross-β motifs in β-sandwiches.

a Cartoon representation of a 7-stranded immunoglobulin-like domain model formed by two β-sheets packing face-to-face, and the corresponding cross-β motif, which generates translations and rotations between the two opposing β-sheets. b Topology diagram of a cross-β motif with circles and arrows representing β-strand residue positions and connections, respectively. Dark- and light-colored circles correspond to residues with sidechains pointing inwards or outwards from the β-sandwich, respectively. c Efficiency of pairs of common β-arch loop geometries (described with ABEGO backbone torsions) in forming cross-β motifs obtained from Rosetta folding simulations (gray shaded squares). Loop geometries were classified in four groups according to the sidechain directions of the adjacent residues. Colored squares group pairs of loops that, due to their sidechain orientations, have different requirements in β-strand length: in red or blue, if all β-strands need an odd or even number of residues, respectively; in green, if the β-strands of the first and second sheet need an odd and even number of residues, respectively; and in yellow for the opposite case (even and odd number of residues for the first and second sheet, respectively). Black-outlined boxes highlight loop combinations observed in natural Ig domains. On the right, examples of changes in cross-β motif geometry linked to β-arch loop geometry. d β-arch helices are formed by a short ɑ-helix connected to the adjacent β-strands with short loops, and are complementary to β-arch loops for connecting cross-β motifs. e Topology diagram of a 7-stranded Ig domain. β-strands and β-arch loops are indicated as S_i and L_i, respectively, where i is the corresponding number. f Examples of de novo designed Ig backbones generated with different geometries and β-arch connections following the described rules, colored from N-terminus (blue) to C-terminus (red).

Fig. 3. Folding and stability of designed proteins.

a Examples of design models. b Simulated folding energy landscapes, with each dot representing the lowest energy structure obtained from ab initio folding trajectories starting from an extended chain (red dots) or local relaxation of the designed structure (green dots). The x-axis depicts the Cα-RMSD from the designed model and the y-axis, the Rosetta all-atom energy. c Far-ultraviolet circular dichroism spectra (blue: 25 °C; green: 55 °C; red: 75 °C; black: 95 °C). d Far-ultraviolet circular dichroism spectra at different guanidine hydrochloride concentrations and at 25 °C (blue: 0 M; green: 1 M; red: 2 M; cyan: 3 M; yellow: 4 M; magenta: 5 M; gray: 6 M; black: 7 M).

Fig. 4. Crystal structure of the dIG14 dimer.

a SEC-MALS analysis of dIG14 estimates a molecular weight corresponding to a dimer (M_w monomer = 9.7 kDa). b Crystal structure of the homodimer interface formed by antiparallel pairing between β-strands 1 and 6 enabled by flipping out of the C-terminal β-strand; the monomer core becomes more accessible and the interface is primarily formed by hydrophobic contacts (right inset). PDB accession code of the dIG14 crystal structure: 7SKP. c dIG14 design model (green) in comparison with the crystal structure (gray, chain B). Sidechain packing interactions in the non-terminal edge β-strands were well recapitulated in the crystal structure (left inset). A shift in β-strand pairing register observed in the crystal structure is highlighted by the two colored arrows (right inset). d The AlphaFold monomer prediction (left) superimposes well with the design model (Cα-RMSD 1.0 Å); while AlphaFold-Multimer (right) correctly predicts the monomer subunits in the crystal structure (Cα-RMSD 0.6 Å, except for the C-terminal disordered β-strand).

Fig. 5. Crystal structure of dIG8-CC and functional loop scaffolding.

a Design model of dIG8-CC with a disulfide bridge (spheres) between β-strands 3 and 6. b SEC-MALS analysis of dIG8-CC estimates a molecular weight between monomer (8.3 kDa) and dimer (16.6 kDa). c Design model (green) in comparison with the crystal structure with PDB accession code 7SKP (gray, chain C). d Cross-β motif connections and core sidechain interactions in the design and the crystal structure. The β-arch helix and loop conformations are well preserved across monomer copies in the crystal asymmetric units (insets). e Crystal homodimer interface by parallel pairing between the two terminal β-strands, which are stabilized through hydrophobic and salt-bridge interactions (inset). f Computational model of dIG8-CC with a grafted EF-hand motif (design EF61_dIG8-CC, cartoon), showing Tb³⁺ (sphere) bound to EF-hand motif residues (sticks). Tb³⁺ luminescence is sensitized by absorption of light (purple) by a proximal tyrosine residue on the EF-hand motif with subsequent fluorescence resonance energy transfer (FRET) to Tb³⁺, resulting in Tb³⁺ luminescence (green). g Far-ultraviolet circular dichroism spectra of EF61_dIG8-CC without Tb³⁺ (blue: 25 °C; green: 55 °C; red: 75 °C; black: 95 °C). h Time-resolved luminescence emission spectra in 100 μM Tb³⁺ final concentrations for EF61_dIG8-CC (blue) and dIG8-CC (red) at 20 µM. Time-resolved luminescence intensity is given in relative fluorescence units (RFU). i Tb³⁺ concentration-dependent time-resolved luminescence intensity of 20 µM EF61_dIG8-CC using excitation wavelength λ_ex  = 280 nm and emission wavelength λ_em  = 544 nm. Normalized intensities are fit to a one-site binding model by non-linear least squares regression (K_d = 267 μM).