Designed endocytosis-inducing proteins degrade targets and amplify signals

Abstract

Endocytosis and lysosomal trafficking of cell surface receptors can be triggered by endogenous ligands. Therapeutic approaches such as lysosome-targeting chimaeras^1,2 (LYTACs) and cytokine receptor-targeting chimeras³ (KineTACs) have used this to target specific proteins for degradation by fusing modified native ligands to target binding proteins. Although powerful, these approaches can be limited by competition with native ligands and requirements for chemical modification that limit genetic encodability and can complicate manufacturing, and, more generally, there may be no native ligands that stimulate endocytosis through a given receptor. Here we describe computational design approaches for endocytosis-triggering binding proteins (EndoTags) that overcome these challenges. We present EndoTags for insulin-like growth factor 2 receptor (IGF2R) and asialoglycoprotein receptor (ASGPR), sortilin and transferrin receptors, and show that fusing these tags to soluble or transmembrane target protein binders leads to lysosomal trafficking and target degradation. As these receptors have different tissue distributions, the different EndoTags could enable targeting of degradation to different tissues. EndoTag fusion to a PD-L1 antibody considerably increases efficacy in a mouse tumour model compared to antibody alone. The modularity and genetic encodability of EndoTags enables AND gate control for higher-specificity targeted degradation, and the localized secretion of degraders from engineered cells. By promoting endocytosis, EndoTag fusion increases signalling through an engineered ligand-receptor system by nearly 100-fold. EndoTags have considerable therapeutic potential as targeted degradation inducers, signalling activators for endocytosis-dependent pathways, and cellular uptake inducers for targeted antibody-drug and antibody-RNA conjugates.

Fig. 1. Design strategies for endocytosis-triggering EndoTags.

a, Schema of designed endocytosis mechanisms. Top, design of binding to constitutively cycling receptors at sites that do not overlap with binding sites for natural ligands to avoid competition. Middle, design of binders that trigger endocytosis by eliciting conformational changes in the receptor. The EndoTag binds at two distinct epitopes on the target and actively triggers the conformational change. Bottom, designed endocytosis via receptor clustering. The multivalent EndoTag clusters multiple copies of the target receptor and induces endocytosis. b, Design strategy for sortilin and TfR EndoTags. c, Cellular uptake of 100 nM AF647-labelled Sort_EndoTags, TfR-EndoTags or LHDB scaffold control for 2 h in U-251MG cells. Data were normalized to the 100 nM AF647-labelled LHDB group (no endocytosis). MFI, mean fluorescence intensity. d, Confocal imaging of Sort_EndoTag (red) and lysosomal marker (green, AF488-labelled LysoTracker) after 24 h incubation in U-251MG cells. e, Design strategy for IGF_EndoTags. f, Cellular uptake of IGF_EndoTags in Jurkat cells with biotinylated 100 nM IGF_EndoTags or IGF2 and 33 nM Streptavidin–AF647 for 24 h. Data were normalized with the control group treated with 33 nM Streptavidin–AF647 alone. g, Fluorescence microscopy showing IGF_EndoTag1 (pink) co-localization with lysosomes (green, anti-LAMP2A) in HeLa cells. h, Design strategy for ASGPR EndoTags. i, Cellular uptake of ASGPR EndoTags. Hep3B cells were treated with 100 nM AF647-conjugated ASGPR EndoTags for 24 h followed with flow cytometry. Data were normalized to 100 nM AF647-labelled LHDB group (no endocytosis). j, Confocal imaging of AS_EndoTag (red) with lysosome (green, AF488-labelled LysoTracker); 500 nM AF647-labelled AS_EndoTag/AS_EndoTag-2C/AS_EndoTag-3C was incubated with Hep3B cells for 24 h. In c,f,i, data are mean ± s.e.m. of three biological replicates. In d,g,j, images are representative of three independently replicated samples. Scale bars, 20 µm.

Fig. 2. Surface receptor degradation with tissue-specific pLYTACs.

a, Schema of tissue-specific pLYTACs for receptor degradation. POI, protein of interest. b, Western blot analysis of total EGFR in Hep3B cells after treatment with 200 nM EGFRn or EGFRn–AS_EndoTag for 48 h. c, Western blot analysis of total EGFR levels in wild-type (WT) or TfR-knockout (KO) HeLa cells after treatment with 200 nM EGFRn or EGFRn–TfR_EndoTag for 48 h. d, Western blot analysis of total EGFR in wild-type or sortilin-knockout HeLa cells after treatment with 200 nM EGFRn or EGFRn–Sort_EndoTag for 48 h. e, Western blot analysis of total EGFR in H1975 cells after treatment with 200 nM of EGFRn or CTX with or without fusion to EndoTag1 or M6Pn for 48 h. f, Western blot analysis of total EGFR in wild-type or IGF2R-knockout HeLa cells after treatment with 200 nM EGFRn or EGFRn–IGF_EndoTags for 48 h. g, Quantitative proteomics analysis of protein abundance in H1975 cells. Data are mean of three biological replicates. Two-tailed unpaired t-test with Welch’s correction. EGFRmb, EGFR minibinder. h, Western blot analysis of PD-L1 in MDA-MB-231 cells after treatment with 200 nM ATZ or ATZ–pLYTACs for 48 h. i–m, Schema (i) of in vivo study. Five million A20 tumour cells were inoculated subcutaneously into BALB/c mice on day 0, then 5 mg kg⁻¹ of indicated reagent (n = 6) was administered intratumourally on days 10, 13 and 16 (j–m). j, Tumour growth curve over time. One-way ANOVA. k, Tumour mass measured at day 21. l, Overall survival of treated mice. P value by log-rank (Mantel–Cox) test. m, Body weight of treated mice. One-way ANOVA. In j–l, data are mean ± s.e.m. of n = 6 biological independent samples. NS, not significant. Source Data

Fig. 3. Clearance of soluble proteins by IGF2R pLYTACs.

a, Schema for the use of soluble pLYTACs with IGF_EndoTags. b, Cellular uptake of LHDB–AF647 via LHDA–IGF_EndoTags in Jurkat cells. Cells were incubated with 33 nM LHDB–AF647 with or without 1 μM LHDA–IGF_EndoTags for 24 h, washed twice with cold PBS and analysed by flow cytometry. c, Remaining supernatant LHDB–AF647 levels in Jurkat cells. Jurkat cells were incubated with 100 nM LHDB–AF647 with or without 500 nM LHDA–IGF_EndoTags. At timepoints 24 h and 48 h, the cells were pelleted down, and IgG in the supernatant was quantified using a Neo2 plate reader. IgG level was normalized to the IgG-alone control group. d, Cellular uptake of IgG–AF647 via protein G–IGF_EndoTags in K562 cells. Cells were incubated with 33 nM IgG–AF647 with or without 1 μM protein G–IGF_EndoTag3 for 24 h, washed twice with cold PBS and analysed by flow cytometry. The fold change in MFI was calculated by normalizing to the IgG–AF647-alone group. e, Remaining IgG–AF647 levels in the supernatant of Jurkat cells. Jurkat cells were incubated with 133 nM IgG–AF647 with or without 100 nM protein G–IGF_EndoTag3. At timepoints 24 h and 48 h, the cells were pelleted down, and IgG–AF647 in the supernatant was quantified using a Neo2 plate reader. The IgG–AF647 level was normalized to the IgG–AF647-alone control group at each timepoint. f, Confocal imaging of IgG–AF647 co-localization with lysosome in HeLa cells. g, Confocal imaging of IgG–AF647 co-localization with lysosome in HeLa (IGF2R-knockout) cells. Representative images of three replicated samples. f,g, Cells were incubated with 200 nM IgG–AF647 and 1 μM protein G–IGF_EndoTag3 for 24 h, washed and stained with LAMP2A antibody followed by AF488-labelled secondary antibody. Scale bars, 20 µm. Data in b–e are mean ± s.e.m. of n = 3 biologically independent samples.

Fig. 4. Logic-gated targeted degradation and locally secretable degraders.

a, Schematic illustration of AND gate logic for EGFR degradation in the presence of HER2. EGFR–Key, designed fusion protein composed of an EGFR-binding domain and a ‘key’ domain; HER2–LOCKR, designed fusion protein composed of an HER2-binding domain and a ‘LOCKR’ domain with a BCL2-recognizing peptide designed to bind to LOCKR domain at weak affinity. At close proximity, the key domain will bind to LOCKR and release the BCL2-recognizing peptide. b, Flow cytometry quantification of EGFR on the cell surface. K562 cells overexpressing EGFR only (K562-HER2⁻) or K562 cells overexpressing both EGFR and HER2 (K562-HER2⁺) were incubated with combinations of 100 nM of EGFRn–Key, HER2–LOCKR and BCL2–EndoTag2 for 24 h. c, Schematic of the use of EGFR–pLYTAC secretion to degrade EGFR in target cells. d, Flow cytometry quantification of cell surface EGFR in cells treated with cell supernatant or exogenous EGFRn–IGF_EndoTag1 for 24 h. For the secretion groups, IGF2R-knockout HeLa cells were transfected with viral vectors encoding EGFRn–IGF_EndoTag1 or LHDA–pLYTACs. Cell supernatants were collected and incubated with K562 cells overexpressing EGFR. Data in b,d are mean ± s.e.m. of n = 3 biologically independent samples.

Fig. 5. EndoTags enhance signalling.

a, Schematic illustrating the designed SNIPR system consisting of an extracellular LHDB protein that recognizes the LHDA ligand, a cleavable membrane proximal domain and an intracellular domain that releases transcription factor, which induces expression of blue fluorescent protein (BFP). Upon ligand binding of LHDB to LHDA SNIPR, the IGF_EndoTag triggers the endocytosis of the complex; signalling activation is quantified by BFP fluorescence intensity. b, Activation of an LHDA-responsive SNIPR driving a BFP reporter circuit in Jurkat T cells by a non-EndoTag ligand (LHDA-C2, a homodimer composed of two LHDA molecules in C2 rotational symmetry) is much weaker than by a similarly flexibly linked EndoTag-containing ligand (IGF_EndoTag1) (n = 3; mean ± s.e.m.). c, Dose–response curve of signal activation with LHDA fusion with IGF_EndoTags. b,c, Jurkat T cells expressing LHDB SNIPR were incubated with IGF_EndoTags at the titrated concentration. n = 3 replicates. Data are mean ± s.e.m. d, Relative activation of Jurkat T cells by IGF_EndoTag1 ligand in the presence of chemical inhibitors. Data are mean ± s.e.m. of measurements normalized to the activation of an inhibitor-free vehicle control sample; n = 3 biologically independent samples. e, Left, live confocal imaging of lysosome co-localization with EndoTag ligands and SNIPR receptors at 24 h. The lysosomes were stained with AF488-labelled LysoTracker. Right, images were acquired at 0.25 h, 3 h, 6 h and 24 h for the experiment in c and analysed for co-localization between the labelled EndoTag ligand and LysoTracker signal. Scale bars, 20 µm for e. n = 4 images per timepoint with at least 10 cells per image; data are mean ± s.e.m.

Extended Data Fig. 1. Binder design strategy and epitope selection.

a, Rifdock-based binder design pipeline. b, Selected target region for IGF2R D6, yellow region highlighted the selected residues for rifdock. c, Design model for D6mb in complex with IGF2R D6. d, Selected target region for IGF2R D11, yellow region highlighted the selected residues for rifdock. e, Design model for D11mb in complex with IGF2R D11. f, Selected target region for Sortilin, yellow region highlighted the selected residues for rifdock. g, Design model for Sort_EndoTag (green) in complex with Sortilin. h, Selected target region for ASGPR orthogonal binding sites, yellow region highlighted the selected residues for rifdock. i, Design model for ASmb1 in complex with ASGPR.

Extended Data Fig. 2. Binding affinity measurement for IGF2R and ASGPR minibinders.

a, BLI binding affinity measurement for D6mb against IGF2R D6. b, BLI binding affinity measurement for D11mb against IGF2R D11. c, BLI binding affinity measurement for EndoTag2 against IGF2R D6 (left) and D11 (right). d, BLI binding affinity measurement for EndoTag3 against IGF2R D6 (left) and D11 (right). e, BLI binding affinity measurement for EndoTag4 against IGF2R D6 (left) and D11 (right). f, BLI binding affinity measurement for ASmb1 against ASGPR. g, BLI binding affinity measurement for Sort_EndoTag against Sortilin. All affinity data was collected by Octet R8 and binding affinity is estimated by Octet ForteBio software package.

Extended Data Fig. 3. Computational design strategy to make rigid IGF_EndoTags (EndoTag3 and EndoTag4).

Starting from the structure of IGF-2 in complex with IGF2R, de novo minibinders were generated and screened against the IGF-2 binding sites at IGF2R domain 6 and domain 11, separately. Individual binders for each domain (D6mb and D11mb) were fused with flexible linkers or a rigid fusion interdomain connection. For flexible fusion, multiple linker lengths and fusion directions were sampled. For rigid fusion, the two major binding helices from D6mb and one major binding helix from D11mb were extracted as starting motifs. With protein Inpainting, geometries and fusion orders were sampled, and ranked based on Rosetta and alphafold2 metrics.

Extended Data Fig. 4. Cellular uptake evaluation IGF_EndoTags.

a, Cellular uptake comparison between homodimer and heterodimer fusion of D6mb and D11mb. b, Cellular uptake comparison of linker length of D11mb-D6mb fusion with various lengths of GS linkers in the middle. c, Cellular uptake of IGF_EndoTags. Jurkat cells were treated with biotinylated 100 nM IGF_EndoTags or IGF-2, and 33 nM Strapavidin-AF647 for 24 h. After washing 3 times, the cellular uptake was measured by flow cytometry. The data were normalized with the control group treated with 33 nM Strapavidin-AF647 alone. The data was collected as mean values ± SEM across n = 3 biologically independent samples. d, Fluorescence Microscopy imaging of IGF-2 co-localized with lysosomal. e, Fluorescence Microscopy imaging of EndoTag2 co-localized with lysosomal. f, Fluorescence Microscopy imaging of EndoTag4 co-localized with lysosomal. For a, and b, 200 nM biotinylated fusion proteins were incubated with 50 nM of Strapavidin-AF647 and incubated with Jurkat cells for 24 h. After wash 3 times with cold PBS, the cellular uptake was measured by flow cytometry. For d-f, HeLa cells were incubated with 100 nM of biotinylated IGF-2, EndoTag2 or EndoTag4 for various time length. After cells were washed twice and fixed, they were stained with anti-LAMP2A antibody followed by goat secondary anti-IgG Alexa Fluo 488 antibody and DAPI. Epifluorescence imaging was conducted on a Yokogawa CSU-X1 microscope. These images are representative of three independently replicated samples per time point.

Extended Data Fig. 5. Receptor degradation with EndoTags.

a, Levels of EGFR after treatment with 10 nM or 100 nM CTX-M6P or CTX-IGF_EndoTag1 in H1975 cells for 48 h. b, Levels of EGFR after treatment with 100 nM EGFRn or EGFRn-IGF_EndoTags in H1975 cells for 48 h. c, Levels of CTLA4 with 200 nM of CTLA4mb or CTLA4mb-IGF_EndoTag1 in Jurkat-CTLA4 cells after treatment for 3 h. d, Fold change in abundance of EGFR with treatment of EGFRn compared with control (untreated group). e, Fold change in abundance of EGFR with treatment of EGFRn-IGF_EndoTag1 compared with control (untreated group). f, Fold change in abundance of EGFR with treatment of CTX compared with control (untreated group). g, Fold change in abundance of EGFR with treatment of CTX-IGF_EndoTag1 compared with control (untreated group). h, Flow cytometry analysis of surface PD-L1 levels in MDA-MB-231 cells after treatment with 200 nM ATZ or ATZ-pLYTACs. MFI was normalized by the PD-L1 level of untreated groups. The data was collected as mean values ± SEM across n = 3 biologically independent samples. i, Functional EGF signaling assay. Hela WT cells were pre-treated with 100 nM of EGFRn, EGFRn-IGF_EndoTag2 or PBS control for 24 h, and then washed and stimulated with 100 nM of human EGF for 15 min followed by phosphorylation flow cytometry using anti-pERK AF-488. Data represents mean of biological triplicates and error bar indicates standard deviation. P values were determined by unpaired two-tailed t-test. For d-g, the proteomic data was collected in H1975 cells with the treatment 100 nM of corresponding reagents for 48 h and data collected is the replicated with sample size n = 2. P values were calculated by a two-tailed unpaired t-test with Welch’s correction.

Extended Data Fig. 6. Clearance of soluble proteins with EndoTags.

a, LHDB-AF647 cellular uptake ability comparison among flexible and rigid LHDA-IGF_EndoTags in K562 cells. b, LHDB-AF647 cellular uptake ability comparison among flexible and rigid LHDA-IGF_EndoTags in Jurkat cells. c, IgG-AF647 cellular uptake ability comparison between flexible and rigid designs. d, Quantitative clearance of IgG-AF647 in cell media comparison between flexible and rigid designs. e, IgG-AF647 cellular uptake ability comparison between pLYTACs and M6Pn. f, Quantitative clearance of IgG-AF647 in cell media comparison between pLYTACs and M6Pn. g, Quantitative clearance of IgG-AF647 in cell media with titrated proteinG-EndoTag3. h, Quantitative clearance of IgG-AF647 in cell media with titrated proteinG-EndoTag4. For a,b, cells were incubated with 100 nM LHDB-biotin + 33 nM Strapavidin-AF647 with/without 500 nM LHDB-pLYTACs for 48 h, washed twice with cold PBS and analyzed by flow cytometry. The fold change in MFI (mean fluorescence intensity) was calculated by normalizing the LHDB-AF647 alone group. For c,e, cells were incubated with 33 nM IgG-AF647 with/wihtout 1uM proteinG-IGF_EndoTags for 24 h, washed twice with cold PBS and analyzed by flow cytometry. The fold change in MFI (mean fluorescence intensity) was calculated by normalizing the IgG-AF647 alone group. For a,b,c,e, data are presented as mean values ± SEM with biologically replicates with n = 3. For d,f,g,h, cells were incubated with 33 nM IgG-AF647 with/without proteinG-IGF_EndoTags. At timepoints 24 h, 48 h, the cells were pelleted down, and supernatant IgG-AF647 levels were quantified by Neo2 plate reader. The percentage of IgG-AF647 level was normalized with the IgG-AF647 alone control group. For d,f,g,h, data are presented as mean values ± SEM with biologically replicates with n = 3. For a-c,e, p values were determined by unpaired two-tailed t-test.

Extended Data Fig. 7. Orthogonality of EndoTag.

a, Confocal imaging of biotinylated IGF-2 labeled with AF-555-Streptavidin (green). Hela cells were pre-incubated with 100 nM non-labelled IGF_EndoTags, PBS control for 24 h. After washing with PBS cells were incubated with AF-555 labeled Streptavidin together with biotinylated IGF-2 (the mix was preincubated for 10 min) for 4 h. Cells were washed and lysosomes were stained with AF488-labeled Lysotracker for 30 min. b, Averaged IGF-2 intensity in single cells based on a Lysotracker cell mask. Data represents mean and error bar indicates SEM (N = 6 images per condition with at least 10 cells per image). c, Binding of AF-647 labeled IGF_EndoTag2 on the cell surface of WT and GNPTAB KO UMRC2 cells on ice. d, Binding of AF647 labeled transferrin to HeLa cells with and without 100 nM of Tfr-EndoTag on ice. e, Internalization of transferrin-647 in HeLa cells with and without 100 nM TfR-EndoTag treatment. For c-e, data represents mean of 3 biological replicates and error bar indicates SEC. P values were determined by unpaired two-tailed t-test.

Extended Data Fig. 8. In vivo PD-L1 degradation by ATZ-pLYTAC.

Western Blot analysis of PD-L1 levels in tumor tissues colleted at sacrifice at day21 from different treatment groups of mice with β-actin as loading control.

Extended Data Fig. 9

a, Confocal imaging of lysosome co-localization of IgG-AF647 with lysosome in Hela WT cells with ProteinG-IGF_EndoTags treatment. b, Confocal imaging of lysosome co-localization of IgG-AF647 with lysosome in Hela IGF2R KO cells with ProteinG-IGF_EndoTags treatment. For a,b, the cells were incubated with 200 nM IgG-AF647 and 1uM of proteinG-IGF_EndoTag for 24 h, washed and stained with anti-LAMP2A antibody followed by AF488-labelled secondary antibody. The scale bar indicates 40 µm for a and 50 µm for b.