Structural organization of the retriever-CCC endosomal recycling complex

Abstract

The recycling of membrane proteins from endosomes to the cell surface is vital for cell signaling and survival. Retriever, a trimeric complex of vacuolar protein-sorting-associated protein (VPS)35L, VPS26C and VPS29, together with the CCC complex comprising coiled-coil domain-containing (CCDC)22, CCDC93 and copper metabolism domain-containing (COMMD) proteins, plays a crucial role in this process. The precise mechanisms underlying retriever assembly and its interaction with CCC have remained elusive. Here, we present a high-resolution structure of retriever in humans determined using cryogenic electron microscopy. The structure reveals a unique assembly mechanism, distinguishing it from its remotely related paralog retromer. By combining AlphaFold predictions and biochemical, cellular and proteomic analyses, we further elucidate the structural organization of the entire retriever-CCC complex across evolution and uncover how cancer-associated mutations in humans disrupt complex formation and impair membrane protein homeostasis. These findings provide a fundamental framework for understanding the biological and pathological implications associated with retriever-CCC-mediated endosomal recycling.

Extended Data Fig. 1.. Purification and cryo-EM structural determination of Retriever.

(A) Gel filtration chromatography of the purified Retriever complex. (B) Representative cryo-EM micrograph. (C) Representative 2D class averages. (D) Euler angle distribution plots for Retriever (upper) and the locally refined VPS29 with the NT “belt” peptide and the CT region of VPS35L (lower). (E) Local resolution map of Retriever (upper) and the locally refined VPS29 with the NT “belt” peptide and the CT region of VPS35L (lower). (F) Fourier Shell Correlation (FSC) plot for Retriever (upper) and the locally refined VPS29 with the NT “belt” peptide and the CT region of VPS35L (lower). (G) Schematic showing cryo-EM data processing steps for obtaining 3D reconstruction of Retriever. The three maps deposited to PDB/EMDB are labeled.

Extended Data Fig. 2.. Structural comparison between Retriever and Retromer.

(A) Surface representation of electrostatic potentials of Retriever vs. Retromer (PDB: 7U6F). (B) Superimpose of individual subunits from Retriever (colored) vs. Retromer (gray). (C) Intermolecular interface between VPS35L and VPS29 vs. VPS35 and VPS29, shown as surface representations of electrostatic potentials. (D) same as in (C), with VPS35L and VPS26C vs. VPS35 and VPS26A.

Extended Data Fig. 3.. Cellular and proteomic analysis of VPS35L mutants.

(A) Huh-7 hepatocellular carcinoma cells carrying the indicated mutations in VPS35L (EV, empty vector). Immunoprecipitation of VPS35L followed by western blot for the indicated proteins is shown. (B) Immunoprecipitation of VPS35L followed by competitive elution of native complexes using HA peptide, and separation of the complexes in blue native gels. After transfer, the complexes were immunoblotted with the indicated antibodies. (C) Heatmap representation of protein-protein interaction results using proteomics. VPS35L was immunoprecipitated from the indicated Huh-7 stable cell lines (in triplicate samples) and the results are expressed as fold compared to Huh-7 control cells (darker blue depicts greater fold difference). Statistical significance is indicated in color scale (yellow indicating p<0.05, and grey indicating p>0.05). (D) Immunoprecipitation of VPS35L carrying indicated point mutations expressed in HEK293T cells and immunoblotting for the indicated proteins.

Extended Data Fig. 4.. VPS35L localization and PM proteome effects.

(A) Immunofluorescence staining for VPS35L (green channel, using HA antibody), FAM21 (red channel), and nuclei (DAPI, blue channel) in the indicated stable Huh-7 cell lines. (B) Immunofluorescence staining for VPS35L (green channel, using HA antibody), FAM21 (red channel), and nuclei (DAPI, blue channel) in the indicated HeLa knockout cell lines transfected with HA-VPS35L. (C-D) Representative gating and acquisition parameters for Villin and CD14 staining by flow cytometry.

Extended Data Fig. 5.. AlphaFold Multimer prediction of CCDC22-CCDC93 binding to Retriever.

(A & D) Overlay of all 25 AlphaFold Multimer models of Retriever alone (A) or CCDC22-CCDC93-Retriever (D) with the cryo-EM model of Retriever. AFM models of Retriever are grey. (B & E) Representative AFM models colored using pLDDT scores. High scores indicate high reliability in local structure prediction. (C & F) PAE score matrix of the AFM model shown in (B & E). Low scores (deep color) indicate high reliability in the relative position in the 3D space. Boundaries of protein sequences and important structure regions are indicated.

Extended Data Fig. 6.. AlphaFold Multimer prediction of CCDC22-CCDC93 binding to DENND10.

(A) AlphaFold Multimer prediction of DENND10 binding to full-length (FL) CCDC22-CCDC93. (B) Representative AFM models colored using pLDDT scores. (C) PAE score matrix of the AFM model shown in (B). Boundaries of protein sequences and important structure regions are indicated. (D) Superimpose of the AFM model of DENND10 with the crystal structure of DENND1a (PDB: 6EKK). Rab35 binding surface of DENND1a and CCDC22-CCDC93 binding surface of DENN10 are colored to show the partial overlap of the two surfaces.

Extended Data Fig. 7.. AlphaFold Multimer prediction of CCDC22-CCDC93 binding to COMMD.

(A-C) Overlays of AlphaFold Multimer models and schematic showing CCDC22-CCDC93 binding to COMMD decamer ring for proteins from Human (A), Zebrafish (B), and Amoeba (Dictyostelium) (C). (D-F) Representative AFM models colored using pLDDT scores. (G-I) PAE score matrices of the AFM models shown in (D-F). Boundaries of protein sequences and important structure regions are indicated.

Fig. 1.. Cryo-EM structure of Retriever reveals a unique assembly mechanism.

(A) Cryo-EM map (EMD: 40886; PDB: 8SYO) and schematic of the Retriever complex. Dotted lines represent the putative flexible linker sequence in VPS35L not observed in the map. (B) Structural comparison between Retriever (top) and Retromer (bottom, PDB: 7U6F). Secondary structural elements of the remotely homologous proteins, including VPS35L vs. VPS35 and VPS26C vs. VPS26A, are labeled. The “belt” sequence unique to VPS35L is traced by yellow dotted lines.

Fig. 2.. The N-terminal “belt” sequence unique to VPS35L is key to Retriever assembly.

(A) Cryo-EM density of the “belt” sequence interacting with VPS35L and VPS29. (B) Alignment of the “belt” sequences from representative species from animal to amoeba and plants. Residues shown in (A) are marked with arrowheads. (C-D) Key interactions between the “belt” sequence (represented in cartoon, with carbon in green, oxygen in red, and nitrogen in blue) and its binding surface on VPS35L (C) and VPS29 (D). The binding surface is colored based on conservation score calculated by Consurf, with color to white gradients representing the most (ConSurf score = 9) to the least conserved residues (ConSurf score = 1). Contacting residues are shown as sticks. Yellow dashed lines indicate polar interactions. (E-F) Immunoprecipitation of VPS35L NT-deletion mutants expressed in HEK293T cells. Interactions with indicated components of Retriever and CCC were assessed by immunoblotting.

Fig. 3.. VPS35L bridges VPS26C and VPS29 through conserved surfaces.

(A-D) Interaction surface of VPS35L with VPS29 (A-B) and VPS26C (C-D). The binding surface is colored based on conservation score using the same scheme shown in Fig. 2. Contacting residues are shown as sticks. Yellow dashed lines indicate polar interactions. For clarity, the backbones of VPS29 and VPS26C in (B) and (D) are shown as loops. (E) Mutation rates (%) for VPS35L and VPS35 across multiple tumor types. (F) Overall structural model of Retriever showing the location of cancer-associated mutations on the surface of VPS35L. Residues mutated in this study are outlined with a black box. For clarity, VPS29 and VPS26 are shown as ribbons. (G-H) Immunoprecipitation of VPS35L (G) or VPS29 (H) carrying indicated point mutations expressed in HEK293T cells. Interactions with various components of Retriever and CCC were assessed by immunoblotting.

Fig. 4.. Disruption of Retriever assembly affects membrane protein homeostasis.

(A) Immunofluorescence staining for VPS35L (green channel, using HA antibody), LAMP1 (red channel), and nuclei (DAPI, blue channel) in the indicated stable Huh-7 cell lines. (B) Quantification of the correlation coefficient for VPS35L and LAMP1 localization for the images shown in (A). Each dot represents an individual cell. (C) Immunofluorescence staining for ITGB1 (green channel), FAM21 (red channel), and nuclei (DAPI, blue channel) in the indicated stable Huh-7 cell lines. (D) Quantification of the correlation coefficient for ITGB1 and FAM21 localization for the images shown in (C). Each dot represents an individual cell. (E) Surface biotinylation and protein isolation, followed by proteomic quantification was performed and protein abundance was compared against VPS35L WT in the indicated cell lines stable Huh-7 cell lines. Red indicates values for proteins with at least 50% reduction compared to VPS35L WT cells, blue represents values that were not significantly reduced, while N/A represents proteins that could not be quantified. (F) Phalloidin staining for F-Actin (green channel) and nuclei (DAPI, blue channel) in the indicated stable Huh-7 cell lines. (G) Quantification of the cortical actin staining in the images shown in (F). Each dot represents an individual cell. (H-I) Quantification of Villin (H) and CD14 (I) fluorescence staining intensity as determined by FACS, expressed as % compared to VPS35L WT cells.

Fig. 5.. Structural model of CCDC22-CCDC93 binding to Retriever.

(A) Overlay of AlphaFold Multimer models and schematic showing Retriever binding to CCDC22-CCDC93. For clarity, inconsistent models (5 out of 25 total models) are excluded. Unreliable structural regions showing inconsistency between models and high PAE scores are removed, including the peptide linker following the “belt” sequence in VPS35L (dotted green line). (B) Interaction surface between Retriever and CCDC22-CCDC93 colored by conservation score using the same scheme shown in Fig. 2. Key interactions are shown as sticks and polar interactions are represented with a dashed yellow line. Residues mutated in this study are outlined with a black box. (C) Coomassie blue-stained SDS PAGE gel showing indicated variants of MBP-CCDC22 NN-CH-VBD/MBP-CCDC93 VBD dimers (200 pmol) pulling down Retriever (60 pmol). (D-F) Immunoprecipitation of indicated mutants of CCDC22 (D), CCDC93 (E), and VPS35L (F) expressed in HEK293T cells and immunoblotting of indicated proteins. (G) Immunoprecipitation and immunoblotting of VPS35L from parental HeLa cells and a VPS29 knockout line derived from these cells.

Fig. 6.. Structural model of CCDC22-CCDC93 binding to DENND10.

(A) Overlay of all 25 AlphaFold Multimer models and schematic showing DENND10 binding to CCDC22-CCDC93. (B) Gel filtration of DENND10 and CCDC22-CCDC92 DBD, individually and in combination. Coomassie blue-stained SDS-PAGE gels of the indicated fractions are shown. The arrowhead indicates the peak fraction of the trimer. (C) Interaction surface between DENND10 and CCDC22-CCDC93 DBD colored by conservation score using the same scheme shown in Fig. 2. Key interactions are shown as sticks and polar interactions are represented with a dashed yellow line. Residues mutated in this study are outlined with a black box. (D-E) Coomassie blue-stained SDS PAGE gels showing MBP-tagged CCDC22-CCDC93 DBD (200 pmol) pulling down DENND10 (500 pmol). Mutations in corresponding constructs are indicated. (F) Immunoprecipitation of CCDC93 carrying indicated point mutants expressed in HEK293T cells and immunoblotting for the indicated proteins.

Fig. 7.. Structural model of CCDC22-CCDC93 binding to COMMD.

(A-C) Overlay of all 25 AlphaFold Multimer models and schematic showing COMMD decamer ring binding to CCDC22-CCDC93, with (A) highlighting the central ring of the COMM domain, (B) highlighting the globular domains on the two sides of the ring, and (C) highlighting the conformation of CCDC22 and CCDC93 CBDs. (D) Interaction surface between the COMMD ring (surface representation) with CCDC22-CCDC93 CBDs (cartoon). Key interactions are shown as sticks and polar interactions are represented with a dashed yellow line. Residues mutated in this study are outlined with a black box. (E) Immunoprecipitation of CCDC22 carrying indicated point mutations expressed in HEK293T cells and immunoblotting for the indicated proteins.

Fig. 8:. Overall model of the Retriever-CCC complex.

(A) Schematic showing the domain organization and the corresponding interaction partners of CCDC22 and CCDC93 derived from AlphaFold Multimer prediction. (B) Overall structural model and schematic of the Retriever-CCC complex derived from AlphaFold Multimer prediction of individual subcomplexes. The peptide linkers in CCDC22 and CCDC93 serving as distance constraints are shown as dashed lines.