Design of High Affinity Binders to Convex Protein Target Sites

Abstract

While there has been progress in the de novo design of small globular miniproteins (50-65 residues) to bind to primarily concave regions of a target protein surface, computational design of minibinders to convex binding sites remains an outstanding challenge due to low level of overall shape complementarity. Here, we describe a general approach to generate computationally designed proteins which bind to convex target sites that employ geometrically matching concave scaffolds. We used this approach to design proteins binding to TGFβRII, CTLA-4 and PD-L1 which following experimental optimization have low nanomolar to picomolar affinities and potent biological activity. Co-crystal structures of the TGFβRII and CTLA-4 binders in complex with the receptors are in close agreement with the design models. Our approach provides a general route to generating very high affinity binders to convex protein target sites.

Figure 1. Design of 5HCS scaffolds to target convex interfaces.

a, Distribution of protein-protein interface curvatures from the PDB and designed protein binders. Blue dots: previously designed protein binders (for these, the designed binders are partner 1 and the targets, partner 2). Previously designed protein binders have been limited to binding to flat or concave interfaces (receptor convexity <=0). Orange dots: examples of native protein complexes, v: PDB ID, 5XXB; vi: TGFβIII/TGFβRII complex, PDB ID, 1KTZ, vii: CD86/CTLA-4 complex, PDB ID, 1I85, viii, PD-1/PD-L1,PDB ID, 3bik. The TGFβRII and CTLA-4 functional interfaces showed high convexity, which we used as case studies to design concave binders. Green dot: The 5HCS scaffolds described in this paper can target convex binding sites. The distribution of convexity of the 5HCS scaffolds (upper part of panel a) shows that the 5HCS scaffolds are diverse enough to cover most of the naturally existing convex interfaces. b. Design models of complexes highlighted in panel a. i,ii,ii are PDGFR, 1GF1R, H3 in complex with corresponding de novo minibinders; iv, 5HCS binder in complex with TGFβRII; v, PDB ID: 5XXB; vi, TGFβIII/TGFβRII complex, PDB ID:1KTZ. Binders and receptors are shown as blue and green cartoons, respectively. Interfacial heavy atoms from binders are shown as yellow solid spheres. Fitted spherical surfaces are shown as blue transparent spheres. c, Design workflow. Column 1: 5HCS concave scaffolds with a wide range of curvatures were designed with three helices (blue) forming the concave surfaces (Cbeta labeled as spheres ) and two helices (orange) buttressing at the back side. Column 2: Docking of 5HCS scaffolds to target binding sites. Column 3: Following docking, the interface sequencing is optimized for high affinity binding.

Figure 2. Concave 5HCS binder to TGFβRII.

a. Design model of 5HCS_TGFBR2_1 (cartoon) binding to TGFβRII (PDB ID: 1KTZ). 5HCS_TGFBR2_1 is colored by Shannon entropy from the site saturation mutagenesis results at each position in blue (low entropy, conserved) to red (high entropy, not conserved). b. Circular dichroism spectra from 25 °C to 95 °C for 5HCS_TGFBR2_1. c. Biolayer interferometry characterization of 5HCS_TGFBR2_1. Biotinylated TGFβRII were loaded to Streptavidin (SA) tips and incubated with 2.7 nM, 0.9 nM and 0.3 nM of 5HCS_TGFBR2_1 to measure the binding affinity. The binding responses are shown in solid lines and fitted curves shown in dotted lines. d. Dose-dependent inhibition of TGF-β3 (10 pM) signaling in HEK293 cells. The mean values were calculated from triplicates for the cell signaling inhibition assays measured in parallel, and error bars represent standard deviations. IC₅₀ values were fitted using four parameter logistic regression by python scripts. e. Heat map of the log enrichments for the 5HCS_TGFBR2_1 SSM library selected with 1.6 nM TGFβRII at representative positions. Enriched mutations are shown in red and depleted in blue. The annotated amino acid in each column indicates the residue from the parent sequence. f,g. Crystal structure of 5HCS_TGFBR2_1 in complex with TGFβRII. Left are top and side views of the crystal (blue and gray) superimposed on the design models (green and white). In the middle, TGFβRII is shown in surface view and colored by electrostatic potential (using ChimeraX; red negative, blue positive). On the right, detailed interactions between 5HCS_TGFBR2_1 (blue, green) and TGFβRII (gray, white) are shown.

Figure 3. Designed 5HCS CTLA-4 binder.

a. Model of 5HCS_CTLA4_1 (cartoon) binding to CTLA-4 (PDB ID:1l85 ) colored by Shannon entropy from site saturation mutagenesis results. b. Circular dichroism spectra from 25 °C to 95 °C for 5HCS_CTLA4_1. c. Biolayer interferometry characterization of 5HCS_CTLA4_1. Biotinylated CTLA-4 was loaded to Streptavidin (SA) tips and these were incubated with 2.7 nM, 0.9 nM and 0.3 nM of 5HCS_CTLA4_1 to measure the binding affinity. d. Increase of TCR activation induced signal (via NFAT pathway) from engineered CTLA-4 effector cells lines by 5HCS_CTLA4_1 (green), lpilimumab (gold) and 5HCS_CTLA4_1_c6 (blue) is shown. EC₅₀ values were fitted using four parameter logistic regression by python scripts. Color schemes and experimental details are as in Fig 2. f. Designed interactions between 5HCS_CTLA4_1 (green) and CTLA-4 (white). e. Log enrichments for the 5HCS_CTLA4_1 SSM library selected with 10 nM CTLA-4 at representative positions. The annotated amino acid in each column indicates the residue from the parent sequence. f,g. Crystal structure of 5HCS_CTLA4_1 in complex with CTLA-4. Color schemes are the same as Fig. 2.

Figure 4. Designed 5HCS binder to PD-L1.

a. Model of 5HCS_PDL1_1 (cartoon) binding to PD-L1 (PDB ID: 3BIK), with 5HCS_PDL1_1 colored by Shannon entropy from site saturation mutagenesis results. b. Circular dichroism spectra from 25 °C to 95 °C for 5HCS_PDL1_1. c. Biolayer interferometry characterization of 5HCS_PDL1_1. Biotinylated PD-L1 was loaded to Streptavidin (SA) tips and these were incubated with 8 nM, 2.7 nM and 0.9 nM of 5HCS_PDL1_1 to measure the binding affinity. d. The increase of TCR activation induced signal (via NFAT pathway) from engineered PD-1 effector cells lines by 5HCS_PDL1_1 (green), control antibody (gold) is shown. The mean values were calculated from triplicates for the cell signaling inhibition assays measured in parallel, and error bars represent standard deviations. Color schemes and experimental details are as in Fig3. e. Heat map representing the log enrichments for the 5HCS_PDL1_1 SSM library selected with 6 nM PD-L1 at representative positions. The annotated amino acid in each column indicates the residue from the parent sequence. f. WT A431 and PD-L1 KO A431 cell lines were stained with fluorophore labled 5HCS_PDL1_1 and anti-PD-L1 antibody respectively and then analyzed through FACS. g,h. Unbound crystal structure of 5HCS_PDL1_1 and designed interactions between 5HCS_PDL1_1 (green) and PD-L1 (white). Color schemes are the same as Fig. 2.