De novo design of highly selective miniprotein inhibitors of integrins αvβ6 and αvβ8

Abstract

The RGD (Arg-Gly-Asp)-binding integrins αvβ6 and αvβ8 are clinically validated cancer and fibrosis targets of considerable therapeutic importance. Compounds that can discriminate between the two closely related integrin proteins and other RGD integrins, stabilize specific conformational states, and have sufficient stability enabling tissue restricted administration could have considerable therapeutic utility. Existing small molecules and antibody inhibitors do not have all of these properties, and hence there is a need for new approaches. Here we describe a method for computationally designing hyperstable RGD-containing miniproteins that are highly selective for a single RGD integrin heterodimer and conformational state, and use this strategy to design inhibitors of αvβ6 and αvβ8 with high selectivity. The αvβ6 and αvβ8 inhibitors have picomolar affinities for their targets, and >1000-fold selectivity over other RGD integrins. CryoEM structures are within 0.6-0.7Å root-mean-square deviation (RMSD) to the computational design models; the designed αvβ6 inhibitor and native ligand stabilize the open conformation in contrast to the therapeutic anti-αvβ6 antibody BG00011 that stabilizes the bent-closed conformation and caused on-target toxicity in patients with lung fibrosis, and the αvβ8 inhibitor maintains the constitutively fixed extended-closed αvβ8 conformation. In a mouse model of bleomycin-induced lung fibrosis, the αvβ6 inhibitor potently reduced fibrotic burden and improved overall lung mechanics when delivered via oropharyngeal administration mimicking inhalation, demonstrating the therapeutic potential of de novo designed integrin binding proteins with high selectivity.

Figure 1.. Computational design of αvβ6 and αvβ8 selective minibinders.

a) Crystal structure of β8 (PDB ID 6OM2) overlaid on the structure of αvβ6 integrin in complex with the L-TGF-β3 peptide RGDLXX(L/I) (PDB ID 4UM9). Inset highlights the zoomed-in regions shown in panels b-f. The αv subunit is shown in green, β6 is in lavender, β8 is in blue, and the RGDLXX(L/I) peptide is in orange. b,c) Shared polar interactions between the RGD motif and (b) αvβ6 and (c) αvβ8 integrin (L-TGF-β3/αvβ6 complex PDB ID 4UM9, L-TGF-β1/αvβ8 PDB ID 6OM2). d) Hydrophobic packing of LXX(L/I) motif of L-TGF-β3 peptide with SDL2 of αvβ6 (PDB ID 4UM9). Leu247 packs optimally against Y185 from SDL2 of β6. e) Hydrophobic packing of LXX(L/I) of L-TGF-β1 peptide with SDL2 of αvβ8 (PDB ID 6OM2). I221 of L-TGF-β1 peptide packs less tightly against L174 on SDL2 of β8 compared to the homologous interactions in panel d. f) Charge reversal on β subunit: β8 contains K304 whereas the equivalent position on β6 is E316. g) Surface structure of the αvβ6 integrin in complex with L-TGF-β3 peptide (red cartoon representation, PDB ID 4UM9). h) Low RMSD matches to the L-TGF-β3 peptide bound to αvβ6 were harvested from the PDB database (orange stick representations). i) Non-clashing fragments with αvβ6 were then incorporated in the α/β ferredoxin folds (orange ribbon representation) using Rosetta. j) Loop extension strategy to design an αvβ8 selective minibinder: to make more extensive contacts to the β8 subunit the β-loop was resampled by one residue insertion (blue surface representation for β8 subunit, PDB ID 6OM2). In addition to the loop extension, the LXX(L/I) motif was allowed to be redesigned using Rosetta. k) Partial sequence alignment of SDL2 of the β6/β8 subunits is shown highlighting two key positions packing against the LXX(L/I) motif of the L-TGF-β ligand (I183 and Y185 in SLD2-β6, and Y172 and L174 in SDL2-β8).

Figure 2.. Selectivity of designed binders for αvβ6 and αvβ8.

a) The A39K mutation confers selectivity towards αvβ6 compared to αvβ8 where there is a charge reversal (Glu316 for β6 shown as a gray stick, Lys304 for β8 shown as a blue stick). b) Cell surface titration of B6B8_BP and B6_BP against K562 cells stably transfected with αvβ8. B6B8_BP lacking the A39K mutation binds to αvβ8 with a Kd of ~7.3 nM whereas B6_BP containing the A39K mutation binds to αvβ8 >500 nM. c) Cell surface titration of AlexaFluor-488-labelled B6B8_BP and B6_BP using αvβ6 (+) human epidermoid A431 carcinoma cells. B6_BP binds to A431 cells with higher potency than B6B8_BP (30 pM vs 167 pM). d) B6_BP_dslf selectively inhibits αvβ6-mediated TGF-β1 activation. αvβ6 and αvβ8 transfectants were co-incubated with CAGA-reporter cells and GARP/TGF-β1 transfectants and inhibitors. B6_BP_dslf inhibits TGF-β activation with an IC50 of 32.8 nM. e) Binding affinities (Kd) of B8_BP_dslf (MAVY) point mutants to integrins αvβ6 and αvβ8, determined by BLI. The LATI motif is in native L-TGF-β1. f, g) Competitive inhibition of h-LAP1 binding to (f) αvβ6 and (g) αvβ8 by designed inhibitors and control small molecule PLN-74809. h) Heatmap of IC50 values for h-LAP1 binding assays in f and g. i) Binding affinities (Kd) and fold-selectivity values of B6_BP_dslf and B8_BP-LATY to all eight RGD integrins compared to small molecules PLN-74809 and GSK3008348. Binding data for PLN-74809 and GSK3008348 are taken from Decaris et al. 2021. Rows shaded in gray indicate the RGD integrin(s) for which each molecule is selective (i.e., B6_BP_dslf and B8_BP-LATY are both mono-selective whereas PLN-74809 and GSK3008348 are dual- and tri-selective, respectively). n/a, not available.

Figure 3.. Structural characterization.

a) Representative 2D class averages of integrin with and without minibinder. For αvβ6, both closed and open headpiece conformations are present in the unbound state, but in the presence of the minibinder the open conformation is dominant. For αvβ8, no open headpieces were observed, with or without minibinder. b) CryoEM density map of αvβ6 bound to minibinder B6_BP_dslf. B6_BP_dslf (goldenrod) binds the integrin ligand binding cleft between the αv (green) and β6 (light blue) subunits and induces or stabilizes the open conformation. The sharpened, locally refined cryoEM map is shown in color, superimposed with the unsharpened map showing all domains of the αvβ6 headpiece in semi-transparent white. c,d) Overlay of the designed αvβ6 + B6_BP_dslf model (gray) and the experimentally determined cryoEM model (colors). Although the overall angle of the minibinder is shifted, the RGD loop positioning is as predicted. Insets in c) and d) are magnified in panels h) and i), respectively. e) CryoEM density map of αvβ8 bound to minibinder B8_BP_dslf. Similar to B6_BP_dslf, B8_BP_dslf (brown) binds the integrin ligand binding cleft between the αv (green) and β8 (blue) subunits, however the conformation of αvβ8 remains in the closed headpiece conformation. The sharpened, locally refined cryoEM map is shown in color, superimposed with the unsharpened map showing all domains in the αvβ8 ectodomain construct in semi-transparent white. f,g) An overlay of the designed αvβ8 + B8_BP_dslf model (gray) and the experimentally determined model. Although the overall angle of the minibinder is shifted, the RGD loop positioning is as predicted. Insets in f) and g) are magnified in panels i) and k), respectively. h,i) Key designed interactions between β-loop and β6/β8 subunit are observed in the cryoEM structure: K41 from B6_BP_dslf forms a salt bridge with E316 from the β6 subunit (panel h). E41 from β-loop makes backbone level hydrogen bond with I216 from β8 subunit and D40 makes salt bridge interaction with K304 from β8 subunit (panel i). j) Experimental vs designed (gray) packing pattern of the LXXL motif and SDL2 of αvβ6. k) Experimental vs designed (gray) packing pattern of the MAVY motif and SDL2 of αvβ8.

Figure 4:. In vivo imaging of αvβ6 (+) A431 tumors using fluorescently-labeled B6_BP.

a) Athymic nude mice were injected with αvβ6 (+) A431 cells on the left shoulder and αvβ6 (−) HEK293T cells on the right shoulder. AlexaFluor-680-labelled B6_BP (AF680-B6_BP) was injected via the tail vein to image the tumors over time as indicated (see Extended Data for additional images, n=5). b) Semiquantitative ex vivo biodistribution assay of AF-680-B6_BP at 6 hours post-tail vein injection. B6_BP selectively accumulates in αvβ6 (+) tumors and primarily clears via glomerular filtration in the kidneys.

Figure 5:. In vivo efficacy of OA-administered B6_BP_dslf in bleomycin-induced IPF.

a) Three dimensional renderings of HR-uCT scans (top panel), representative HR-uCT scans (middle panel), and representative Masson-trichrome images for nontreated (NT), bleomycin treated (BLM) and inhaled B6_BP_dslf group show improvements in tissue health following B6_BP_dslf OA treatment. b) Average Ashcroft Scoring of Masson-trichrome images (mean ± SEM, NT n=5, BLM n=3, B6_BP_dslf 46.3 ug/kg n=3, B6_BP_dslf 185.2 ug/kg n=4) c) Forced vital capacity as measured by SCIREQ flexiVent FX (mean ± SEM, NT n=9, BLM n=10, B6_BP_dslf 46.3 ug/kg n=4, B6_BP_dslf 185.2 ug/kg n=6). d) Pressure-Volume curves measured by SCIREQ flexiVent with peak volumes in the inset graph (NT n=20, BLM n=14, B6_BP_dslf 46.3 ug/kg n=4, B6_BP_dslf 185.2 ug/kg n=7). e) Whole lung tissue homogenate western blot analysis of Collagen1 and f) p-SMAD2 show a dose-dependent reduction of these pro-fibrotic markers following B6_BP_dslf OA treatment (mean ± SD NT n=3, BLM n=3, B6_BP_dslf 46.3 ug/kg n=3, B6_BP_dslf 185.2 ug/kg n=3). g) Soluble and h) Insoluble collagen levels are lower following B6_BP_dslf OA treatment (mean ± SD NT n=3, BLM n=3, B6_BP_dslf 46.3 ug/kg n=3, B6_BP_dslf 185.2 ug/kg n=3). i) Cytokine array analysis of common cytokines implicated in inflammation and IPF (all NT vs. BLM are significant, p-value<.05, mean ± SD, NT n=4, BLM n=4, B6_BP_dslf 46.3 ug/kg n=4, B6_BP_dslf 185.2 ug/kg n=4). j) Time course of hFLO organoid growth, bleomycin induction, and 10 nM B6_BP_dslf treatment. k) Fluorescent confocal microscopy imaging of hFLO sections immunostained with pro-fibrotic markers αSMA, fibronectin, and PDGFRα. Volumetric analysis of l) αSMA, m) fibronectin, and n) PDGFRα normalized to DAPI signal (mean ± SD, N=15 cell aggregates were used per treatment, *P<0.05 determined using two-tailed Welch’s t-test).