Unintended specificity of an engineered ligand-binding protein facilitated by unpredicted plasticity of the protein fold

Abstract

Attempts to create novel ligand-binding proteins often focus on formation of a binding pocket with shape complementarity against the desired ligand (particularly for compounds that lack distinct polar moieties). Although designed proteins often exhibit binding of the desired ligand, in some cases they display unintended recognition behavior. One such designed protein, that was originally intended to bind tetrahydrocannabinol (THC), was found instead to display binding of 25-hydroxy-cholecalciferol (25-D3) and was subjected to biochemical characterization, further selections for enhanced 25-D3 binding affinity and crystallographic analyses. The deviation in specificity is due in part to unexpected altertion of its conformation, corresponding to a significant change of the orientation of an α-helix and an equally large movement of a loop, both of which flank the designed ligand-binding pocket. Those changes led to engineered protein constructs that exhibit significantly more contacts and complementarity towards the 25-D3 ligand than the initial designed protein had been predicted to form towards its intended THC ligand. Molecular dynamics simulations imply that the initial computationally designed mutations may contribute to the movement of the helix. These analyses collectively indicate that accurate prediction and control of backbone dynamics conformation, through a combination of improved conformational sampling and/or de novo structure design, represents a key area of further development for the design and optimization of engineered ligand-binding proteins.

Keywords: affinity versus specificity; crystal structure; ligand binding; protein engineering.

Fig. 1

ChemDraw and space fill representations of (a) 25-hydroxy-cholecalciferol (25-D3) and (b) tetrahydrocannabinol (THC).

Fig. 2

Score comparison of designs vs a representative set of randomized scaffolds. Shape complementarity (x-axis) and Rosetta interface energy (y-axis) for all ordered designs targeting the ligands (a) 25-hydroxy-cholecalciferol (25-D3) and (b) tetrahydrocannabinol (THC). Each plot compares the top twenty dock energies for the generated designs with each ligand (red) vs a random set of native protein structures (black). A naturally occuring 25-D3 binder (PDB ID: 1DB1) is also included in the random set of wild-type protein scaffolds (indicated by arrow). Its predicted energy places it among the best 25-D3 designs and helps validate the design metrics.

Fig. 3

Yeast surface display titrations for affinity and specificity estimation of designed binders CDL1 and THC1. Yeast surface display titrations for both initial designs (solid lines) and evolved variants (dashed lines). (a) Designs CDL1 and CDL1.1 targeting 25-D3 (black lines) were tested for specificity against similar ligand D3 (red lines). (b) Designs THC1 and THC1.1 targeting THC (black lines) were tested for specificity against similar ligand CBD (red lines). Approximate K_D values are 2 μM for CDL1 versus 25-D3, 1 μM for CDL1 versus D3, 200 nM for CDL1.1 versus 25-D3, 400 nM for CDL1.1 versus D3, ~30 μM for THC1 versus THC, >10 μM for THC1 versus CBD, ~5 μM for THC1.1 versus THC and >10 μM for THC1.1 versus CBD.

Fig. 4

(a) Sequence alignment of the wild-type protein scaffold used for engineering (PDB ID: 3HX8), the original computationally redesigned variant of that protein scaffold intended to bind THC, but instead displayed binding signal against 25-D3 (‘CDL2’), and a series of subsequent variants produced through a combination of epPCR and redesign steps that iteratively display enhanced binding of 25-D3 (‘CDL2.1’, ‘CDL2.2’, ‘CDL2.3a’ and ‘CDL2.3b’). The individual mutations relative to the starting protein scaffold that found in each step of design and selection are listed below the alignment. (b) Cartoon representation of the wild-type protein scaffold with the residues subjected to mutagenesis indicated by side chain sticks (colored corresponding to the highlighted residue positions in the sequence alignment, indicating their first appearance during the engineering process). The position of bound 25-D3, extracted from the crystal structure of the engineered CDL2.3a construct, is shown in spheres to illustrate the location and size of the designed ligand-binding pocket.

Fig. 5

Comparison between design model CDL2 and its evolved variant CDL2.1. Fluorescence polarization binding data for 25-D3 binder CDL2 (red) versus its evolved variant CDL2.1 (black) binding a fluorescently labeled 25-D3 molecule. CDL2 has an approximate K_D = 2 μM. CDL2.1 has an approximate K_D = 200 nM

Fig. 6

Crystallographic structures of engineered constructs. Electron density are unbiased omit maps. Left: Fo–Fc difference maps calculated in the absence of modeled ligand. Right: 2Fo–Fc difference maps contoured across the bound ligand and nearest contacting side chains. (a) CDL2.2, (b) CDL2.3a and (c) CDL2.3b.

Fig. 7

Superposition of starting wild-type protein scaffold against the original computationally designed model (CDL2) and against the crystallographic structure of CDL2.2. The wild-type (non-engineered) starting protein is blue in all superpositions. The original computationally designed model (CDL2) and the crystal structure of the first laboratory-evolved variant of that design model (CDL2.2) are light green and dark green, respectively. (a) Superposition of the starting protein and CDL2 design model and corresponding fit of the intended THC ligand into the designed binding pocket in that computational model. (b) Superposition of the same starting protein and the CDL2.2 crystal structure and corresponding fit of the observed 25-D3 ligand into the binding pocket in that structure. The largest backbone differences between the original protein scaffold and the engineered and laboratory-evolved construct are indicated with highlighted arrows in the upper (helix motion) and lower (loop motion) panels.

Fig. 8

Molecular dynamics simulations of PDB 3HX8 and design CDL2. A small shearing movement of the N-terminal helix, in the same direction as that observed in our crystal structures (Fig. 8) is observed for the CDL2 starting model, with peaks in RMSD backbone shifts corresponding to residues near the two ends of the helix.

Fig. 9

Calculated energetic docking funnels. For all docking plots, the y-axis represents the calculated Rosetta interface energy and the x-axis represents the root mean squared deviation (RMSD) of various docked ligand positions within the protein model binding site. (a) 25-D3 docked in the original CDL2 design model. (b) 25-D3 docked in the original CDL2 design model, after the addition of five amino acid substitutions found in the CDL2.2 construct. The surrounding backbone conformation is unchanged from the original design. (c) 25-D3 docked in the actual CDL2.2 crystal structure.

Fig. 10

Computed versus observed ligand-binding side chain contacts. (a) Computationally designed CDL2/THC complex. (b) Crystallographic structure of CDL2.2/25-D3 complex.