A novel class of inhibitors that disrupts the stability of integrin heterodimers identified by CRISPR-tiling-instructed genetic screens

Abstract

The plasma membrane is enriched for receptors and signaling proteins that are accessible from the extracellular space for pharmacological intervention. Here we conducted a series of CRISPR screens using human cell surface proteome and integrin family libraries in multiple cancer models. Our results identified ITGAV (integrin αV) and its heterodimer partner ITGB5 (integrin β5) as the essential integrin α/β pair for cancer cell expansion. High-density CRISPR gene tiling further pinpointed the integral pocket within the β-propeller domain of ITGAV for integrin αVβ5 dimerization. Combined with in silico compound docking, we developed a CRISPR-Tiling-Instructed Computer-Aided (CRISPR-TICA) pipeline for drug discovery and identified Cpd_AV2 as a lead inhibitor targeting the β-propeller central pocket of ITGAV. Cpd_AV2 treatment led to rapid uncoupling of integrin αVβ5 and cellular apoptosis, providing a unique class of therapeutic action that eliminates the integrin signaling via heterodimer dissociation. We also foresee the CRISPR-TICA approach to be an accessible method for future drug discovery studies.

Fig. 1. Cell surface proteome CRISPR screens identify the essential role of ITGAV in cancer cells.

a, Schematic outline of cell surface proteome CRISPR screens (2,973 sgRNAs) in Cas9-expressing cancer cell models. b,c, Gene rankings for cell surface proteome CRISPR screens in five individual cell models (b) and the combined analysis (c) as calculated by the MAGeCK algorithm. The rankings of ITGAV (red), positive controls (yellow), negative controls (green) and the total library (gray) are indicated. d, Western blot of ITGAV and β-actin in MDA231-Cas9⁺ cells transduced with sgCtrl (n = 2 independent sgRNA sequences) and sgITGAV (n = 3 independent sgRNA sequences) for 3 days. e, Growth competition assay of MDA231-Cas9⁺ and PANC1-Cas9⁺ cells transduced with RFP-labeled sgCtrl (gray lines; two independent sgRNA sequences) and sgITGAV (red lines; three independent sgRNA sequences). Asterisk indicates that all three sgITGAV groups were significantly different (P < 0.01) from the two sgCtrl groups (n = 3 for each group). f,g, Cellular apoptosis detected by Annexin V⁺/DAPI⁻ (f) and cell cycle monitored by EdU incorporation (g) in MDA231-Cas9⁺ cells transduced with sgCtrl and sgITGAV for 3 days (n = 3 for each group). h, Survival curves for cancer patients with high (top quartile; n = 927) versus low (bottom quartile; n = 927) ITGAV expression (data source: GEPIA). Data are presented as the mean ± s.e.m. P values were calculated by two-sided Student’s t-test. FC, fold change; GBM, glioblastoma multiforme. Source data

Fig. 2. ITGAV supports cancer cell expansion through small GTPase RAC1.

a, Gene ranking based on the Pearson coefficient (r) of CERES scores between ITGAV and RAC1 (purple) compared with CDC42 and RHOA (green) in the 769 tested cell models (Extended Data Fig. 5a). b, RNA sequencing analysis and GSEA showing changes in expression of the ‘RAC1_GTPase_Cycle’ gene set in MDA231-Cas9⁺ cells transduced with sgCtrl and sgITGAV for 3 days (n = 3 independent sgRNA sequences per group). c, Western blot of RAC1 and β-actin in MDA231-Cas9⁺ cells transduced with sgCtrl (n = 2 independent sgRNA sequences) and sgRAC1 for 3 days (n = 3 independent sgRNA sequences). d, Growth competition assay of MDA231-Cas9⁺ cells transduced with RFP-labeled sgCtrl (gray lines; two independent sgRNA sequences) and sgRAC1 (purple lines; three independent sgRNA sequences). Asterisk indicates that all three sgRAC1 groups were significantly different (P < 0.01) from the two sgCtrl groups (n = 3 for each group). e,f, Cellular apoptosis as detected by Annexin V⁺/DAPI⁻ (e) and cell cycle monitored by EdU incorporation (f) in MDA231-Cas9⁺ cells transduced with sgCtrl and sgRAC1 for 3 days (n = 3 for each group). g, Representative fluorescence images of F-actin (fluorescein isothiocyanate (FITC), green) and nucleus (DAPI, blue) staining in MDA231-Cas9⁺ cells transduced with sgCtrl, sgITGAV and sgRAC1. Scale bars, 20 µm. h, Violin plot showing distribution of cell size (µm²) in MDA231-Cas9⁺ cells transduced with sgCtrl, sgITGAV and sgRAC1. Data are presented as the mean ± s.e.m. P values were calculated by two-sided Student’s t-test. NS, not significant. Source data

Fig. 3. Integrin family CRISPR screens reveal the critical role of integrin αVβ5 in cancer cell expansion.

a, Model of integrin α (red) and β (blue) subunits and domain structures. The binding site of the extracellular ligand (yellow) is assembled upon heterodimerization of the α/β subunits. b, Schematic outline of integrin family CRISPR screens (712 sgRNAs) in Cas9-expressing MDA231 and PANC1 cells. c, Fold change of each sgRNA from day 0 to day 24 in MDA231-Cas9⁺ (x axis) and PANC1-Cas9⁺ (y axis) cells. The sgRNAs targeting ITGAV (red dots), ITGB5 (blue dots), positive controls (yellow triangles), negative controls (green triangles) and the total library (gray dots) are indicated. d, Heatmap showing CRISPR impact scores (median log₁₀ fold change of 25 sgRNAs) of each integrin subunit in the integrin network consisting of 24 distinct integrin α/β heterodimers. The solid lines indicate the integrin α/β pairs forming the RGD receptors (yellow), collagen receptors (pink), laminin receptors (brown) and leukocyte-specific receptors (green). The red dotted circle highlights αVβ5 as the top essential integrin heterodimer in cancer cells. e, Growth competition assay of MDA231-Cas9⁺ cells transduced with RFP-labeled sgCtrl (gray lines; two independent sgRNA sequences) and sgITGB1/3/5/6/8 (blue lines; three independent sgRNA sequences for each gene). Asterisk indicates that all three sgRNAs for each ITGB gene group were significantly different (P < 0.01) from the two sgCtrl groups (n = 3 for each group). f, Western blot of ITGB5 and β-actin in MDA231-Cas9⁺ cells transduced with sgCtrl (n = 2 independent sgRNA sequences) and sgITGB5 (n = 3 independent sgRNA sequences) for 3 days. g,h, Cellular apoptosis detected by Annexin V⁺/DAPI⁻ (g) and cell cycle monitored by EdU incorporation (h) in MDA231-Cas9⁺ cells transduced with sgCtrl and sgITGB5 for 3 days (n = 3 for each group). i, Gene ranking based on the Pearson coefficient (r) of CERES scores between ITGAV and ITGB5 (blue) compared with other ITGAV partner β subunit genes ITGB1/3/6/8 (yellow) in the 769 tested cell models (Extended Data Fig. 5b). Data are presented as the mean ± s.e.m. P values were calculated by two-sided Student’s t-test. Source data

Fig. 4. High-density CRISPR tiling identifies a critical pocket in the β-propeller domain of ITGAV.

a, Schematic outline of the ITGAV high-density CRISPR-tiling scan (412 sgRNAs) in MDA231-Cas9⁺ cells. b, 2D annotation of ITGAV CRISPR scan. The red line indicates the smoothened model of NCS derived from 348 sgRNAs (dots) targeting the coding exons of ITGAV. The median NCS of the positive control (gray dotted line; defined as −1.0) and negative control (defined as 0) sgRNAs are highlighted. The brown dashed box contains the β-propeller domain. The numbers 1–7 pinpoint the CRISPR-hypersensitive regions within the β-propeller domain. c, 3D annotation of ITGAV CRISPR scan NCS relative to AlphaFold structural modeling of ITGAV (AlphaFold ID: P06756). d, Enlarged view of the β-propeller domain showing the CRISPR-hypersensitive regions (numbers 1–7 as indicated in b) pointing to the center cavity of the β-propeller HIP. The residues contributing to this aromatic-enriched pocket are highlighted. e, Schematic outline of the NanoBRET reporter system for detecting the ITGAV–ITGB5 interaction in living cells. f, Effect of alanine substitution of the ITGAV β-propeller HIP residues (brown; n = 3 for each group) on the NanoBRET signal compared with the wild-type ITGAV (gray; n = 3 for each group). Data are represented as mean ± s.e.m. P values were calculated by two-sided Student’s t-test. Ex, exon; TSS, transcription start site. TM, transmembrane. Source data

Fig. 5. Identification of compounds targeting ITGAV β-propeller domain by CRISPR-TICA pipeline.

a, 3D ‘docking box’ (cube) defined by the CRISPR-hypersensitive regions (numbers 1–7) within the ITGAV β-propeller domain. b, Compound (Cpd) ranking based on free binding energy (ΔG°) to the ‘docking box’ within the β-propeller domain predicted by AutoDock Vina. c,d, Heatmap showing relative CellTiter Glo (left) and CCK8 (right) signals (percentage of the signal for dimethyl sulfoxide; DMSO) in MDA231 cells incubated with 10 µM of 500 selected compounds (c) and the top nine effective compounds (d) for 3 days. Effective cell killing was defined as less than 10% relative signals for both CellTiter Glo and CCK8 assays. e, Schematic outline of flow cytometric measurement of cell surface integrin αVβ5 using a monoclonal antibody against integrin αVβ5 heterodimers. f, Effects of the top nine candidate compounds on cell surface integrin αVβ5 levels upon 1 h compound treatments (n = 4 for each condition). g,h, Cellular apoptosis detected by Annexin V⁺/DAPI⁻ (g) and cell cycle monitored by EdU incorporation (h) in MDA231 cells treated with Cpd_AV2 (40 µM) for 0 to 3 h (n = 3 for each time point). i, Representative fluorescence images of F-actin (FITC, green) and nucleus (DAPI, blue) staining in MDA231 cells treated with control (DMSO) and Cpd_AV2 (40 µM) for 10 min. Scale bars, 20 µm. j, Violin plot showing the distribution of cell size (µm²) in MDA231 cells treated with control (DMSO) and Cpd_AV2 (40 µM) for 10 min. Data are presented as the mean ± s.e.m. P values were calculated by two-sided Student’s t-test. Source data

Fig. 6. Characterization of ITGAV β-propeller domain inhibitor Cpd_AV2.

a, Purification of bacterial-expressed recombinant ITGAV β-propeller domain (peptide region 31–492 aa; N-terminal His⁶-tagged) using immobilized metal affinity chromatography (IMAC) and anion exchange chromatography (IEX). The input and purified ITGAV β-propeller domain samples were visualized by gel electrophoresis and silver staining (right; gel representative of two independent protein purification experiments). b, Protein thermal stability as estimated by fluorescent dye incorporation of the purified ITGAV β-propeller domain under control (DMSO) and Cpd_AV2 (40 µM) conditions. c, Protein surface model (left) showing a docking simulation of the ITGAV β-propeller domain (colored by NCS) interacting with Cpd_AV2 (yellow). Protein ribbon model (right) illustrates an overlap of ITGB5 lysine 287 (within βA loop; cyan) and Cpd_AV2 (yellow) binding on the β-propeller HIP of ITGAV. d,e, Effects of cilengitide and Cpd_AV2 on cell surface integrin αVβ5 levels after 1 h treatment (d) and cell expansion after 72 h treatment in MDA231 cells (e) (n = 3 for each group). f, Effects of 72 h Cpd_AV2 treatment on expansion of six cancer cell models (n = 3 for each group). g, Chemical structure of Cpd_AV2 (source: NCI/DTP Open Chemicals Repository). h, Model showing distinct mechanisms of action between Cpd_AV2 (left) and cilengitide (right) for suppressing ECM-to-integrin αVβ5 signaling (middle). Data are presented as the mean ± s.e.m. P values were calculated by two-sided Student’s t-test. NSCLC, nonsmall-cell lung cancer. Source data

Extended Data Fig. 1. CRISPR genetic screen libraries used in this study.

(a) Map of the ipUSEPR vector expressing a sgRNA together with a puromycin-resistant gene (PuroR) and a TagRFP fluorescent protein. Primers for Sanger (hU6-F_seq) and Illumina (DCF01 and DCR03) sequencing are listed. (b–d) Design and distribution of individual sgRNA frequencies RPMR (reads per million reads) in the CRISPR libraries targeting (B) cell surface proteome genes (n = 2,973 sgRNAs), (C) integrin family genes (n = 714 sgRNAs), and (D) coding regions of ITGAV (n = 412 sgRNAs). (B) 90.1%, (C) 97.7%, and (D) 96.3% of sgRNA in these libraries passed the QC by exhibiting RPMR ≥ 10. Data are represented as median ± interquartile range. Source data

Extended Data Fig. 2. Analyses of the surface proteome CRISPR library screens.

(a) Combined gene ranking of the cell surface proteome CRISPR screens was calculated by the MAGeCK algorithm. The ranking of ITGAV (red), positive controls (yellow; target common essential genes), negative controls (green; target non-essential sequences), and total library (grey) are indicated. The pink box highlights the leading-edge essential genes with a combined Log2FC below -1.0. (b) Distribution of the positive (n = 12 genes) and negative (n = 5 genes) controls in the screen. Data are represented as median ± interquartile range. P value was calculated by two-sided Student’s t-test. (c) Three surface protein genes (ITGAV, ATP6AP2, and TFRC) were identified as the leading-edge essential genes. (d) Correlation of the CERES scores (computational method to estimate gene-dependency levels from CRISPR-Cas9 essentiality screens) and gene expression of ITGAV (left panel), ATP6AP2 (middle panel), and TFRC (right panel)(source: https://depmap.org/portal/; BROAD Institute). The cancer cell dependency on ITGAV is correlated with its expression. (e) Top ten candidate hits and (f) an overlap plot of the surface proteome CRISPR screens in five cell models. Red (ITGAV and ATP6AP2) indicates the common essential surface proteins in all screened cell types. Other colors (blue, orange, green, cyan, and pink) highlight the cell type-specific candidate genes. Source data

Extended Data Fig. 3. Schematic outline of the flow cytometric analysis.

(a) RFP growth competition assay (used in Figs. 1e, 2d, 3e): The ipUSEPR vector expresses a sgRNA together with a puromycin-resistant gene (PuroR) and a TagRFP fluorescent protein. The RFP fluorescent signal of live (DAPI^–) singlet cells was detected by an Attune NxT flow cytometer with an HTS autosampler. The sgRNA targeting a functionally important gene will result in a reduced RFP⁺ population in the culture. (b) Gating strategy for detecting the cell surface αVβ5 expression (used in Fig. 5e, f).

Extended Data Fig. 4. The effect of sgITGAV can be reversed by the exogenous ITGAV cDNA.

(a) Schematic outline of the ITGAV gene coding region. The recognition sites of sgITGAV#2 and sgITGAV#3 span across the exon-intron junctions. These sgITGAVs only target the endogenous ITGAV coding sequence (with introns) but cannot recognize the ITGAV cDNA sequence (w/o introns), thus allowing the reconstitution of ITGAV through cDNA transduction. (b) Transduction of exogenous ITGAV cDNA in MDA231 cells significantly reversed the impact of sgITGAV on cell survival (n = 3 for each group). Data are represented as mean ± s.e.m. P values were calculated by two-sided Student’s t-test. Source data

Extended Data Fig. 5. Correlation of the CERES scores between ITGAV and other genes.

CERES score is a computational method to estimate gene-dependency levels from CRISPR-Cas9 essentiality screens. The CERES scores of ITGAV (x-axis) and (a) Rho small GTPase genes RAC1, CDC42, and RHOA (y-axes; left, middle, and right respectively) and (b) ITGB1/3/5/6/8 (y-axes; top-left, top-middle, top-right, bottom-middle, bottom-right, respectively) in 769 cell models (dots) were obtained from the DepMap CRISPR screen consortium database (source: https://depmap.org/portal/; BROAD Institute). A higher Pearson coefficient (r) of the CERES scores between two genes indicates a higher likelihood the two genes are co-regulated in the tested cell models. The gene rank number is based on the Pearson coefficient (r) of the CERES scores between ITGAV and a total of 17,709 genes tested in the genome-wide CRISPR library screens. Source data

Extended Data Fig. 6. Effect of alanine substitution of non-HIP aromatic residues on the NanoBRET assay.

(a) The location of four aromatic residues (W234, F507, W790, and F938) outside of HIP were highlighted. (b) Alanine substitution of these non-HIP residues (purple; n = 3 for each group) exhibits minimal impact on the NanoBRET signal compared to the wild-type ITGAV (gray; n = 3). Data are represented as mean ± s.e.m. P values were calculated by two-sided Student’s t-test. Source data

Extended Data Fig. 7. Information of the candidate ITGAV targeting compounds.

The NCI/DTP NSC identifier, the predicted binding free energy (ΔG°) to ITGAV’s β-propeller HIP, and the chemical structure of the top 9 candidate compounds are indicated. The identity (up-right; within one ppm of theoretical value) of Cpd_AV2 was validated by an Orbitrap Fusion Tribrid Mass Spectrometer (Thermo Scientific) at the City of Hope Integrated Mass Spectrometry Shared Resource.

Extended Data Fig. 8. Modeling of ITGAV/ITGB5 interaction.

(a) 3D structure of the extracellular domain of ITGB5 was modeled by AlphaFold2 (cyan) and overlaid with the ITGB3 portion of integrin αVβ3 structure resolved by Xiong et al. (PDB ID: 3IJE, chain B; blue). Overall, we observed high concordance of the 3D structures between ITGB3 and ITGB5, including the highly conserved basic amino acid (ITGB3’s R287 or ITGB5’s K287) in the loop motif of the βA domain highlighted in (b). (c) Modeling of ITGAV/ITGB5 interaction using the AlphaFold2 predicted ITGB5 structure (cyan) and the ITGAV portion of integrin αVβ3 structure (PDB ID: 3IJE, chain A; red). (d and e) Molecular dynamics simulation using GROMACS 2022 with CHARMM36m force field indicates (d) a close contact between ITGB5’s K287 and ITGAV’s β-propeller HIP pocket (purple box), and (e) the occupancy of Cpd_AV2 (yellow) into ITGAV’s HIP pocket disengaged the side chain of ITGB5’s K287 from stably interacting with ITGAV. (f) Substitution of ITGB5’s K287 with an alanine (K287A) significantly attenuated the ITGAV/ITGB5 NanoBRET signal, highlighting an essential role of this basic residue in integrin αVβ5 assembly (n = 3 for each group). Data are represented as mean ± s.e.m. P value was calculated by two-sided Student’s t-test. Source data

Extended Data Fig. 9. Potential impact of Cpd_AV2 on additional ITGAV integrin pairs.

(a) Sequence alignment of ITGAV heterodimer partners ITGB1/3/5/6/8 at the loop motif of their βA domain. The highly conserved basic amino acid (K/R287) encapsulated in ITGAV’s β-propeller is labeled. (b) Cpd_AV2 treatment attenuates the integrin αVβ6-mediated adhesion to fibronectin (Fn) in HT-29 colorectal carcinoma cells (n = 3 for each group). Source data

Extended Data Fig. 10. CRISPR-TICA evaluation of well-defined drug-targeting pockets.

The smoothened CRISPR tiling data (left panel; blue lines) of (a) BRD4, (b) AURKB, (c) CDK1, and (d) WEE1 were obtained from Munoz et al. On the right panels, the green boxes highlight the CRISPR-TICA region of interest based on the CRISPR sensitivity. The yellow arrows indicate the previously reported inhibitors for these proteins. Source data