Structure of the endosomal Commander complex linked to Ritscher-Schinzel syndrome

Abstract

The Commander complex is required for endosomal recycling of diverse transmembrane cargos and is mutated in Ritscher-Schinzel syndrome. It comprises two sub-assemblies: Retriever composed of VPS35L, VPS26C, and VPS29; and the CCC complex which contains twelve subunits: COMMD1-COMMD10 and the coiled-coil domain-containing (CCDC) proteins CCDC22 and CCDC93. Combining X-ray crystallography, electron cryomicroscopy, and in silico predictions, we have assembled a complete structural model of Commander. Retriever is distantly related to the endosomal Retromer complex but has unique features preventing the shared VPS29 subunit from interacting with Retromer-associated factors. The COMMD proteins form a distinctive hetero-decameric ring stabilized by extensive interactions with CCDC22 and CCDC93. These adopt a coiled-coil structure that connects the CCC and Retriever assemblies and recruits a 16th subunit, DENND10, to form the complete Commander complex. The structure allows mapping of disease-causing mutations and reveals the molecular features required for the function of this evolutionarily conserved trafficking machinery.

Keywords: AlphaFold; CCC complex; CCDC22; CCDC93; COMMD; Commander; DENND10; Endosome; Retriever; Retromer; Ritscher-Schinzel syndrome; VPS29.

Graphical abstract

Figure 1

Architecture of the human Retriever complex (A) Schematic of Retriever and CCC sub-complexes that form the Commander assembly. (B) Low resolution cryoEM envelope of human Retriever with docked AlphaFold2 model (Methods S1). Insets show details of: (i). VPS35L:VPS26C interface; (ii). VPS35L:VPS29 interaction; (iii). β-hairpin of VPS35L interacting with VPS29; (iv). intramolecular interaction of N terminus of VPS35L with its C terminus; (v). PL motifs in the N terminus of VPS35L interacting with the hydrophobic surface of VPS29. (C and D) GFP-nanotrap of GFP-VPS35L mutants targeting the interface with (C) VPS26C and (D) VPS29. (E) GFP-nanotrap of GFP-VPS29 mutants targeting the major interfaces within Retriever. (F) GFP-nanotrap of GFP-VPS35L mutants targeting the β-hairpin. (G) GFP-nanotrap of GFP-VPS35L mutants targeting the N-terminal sequence mediating intramolecular interactions with the VPS35L C terminus. All blots are representative of three independent experiments. Data S1 shows quantified and raw blots (n = 3). See also Figure S1 and Methods S1.

Figure 2

A unique structure in VPS35L regulates VPS29 interaction (A) VPS35L peptides were titrated into VPS29 and binding affinity measured by ITC. Top shows the raw data and bottom shows the integrated and normalized data fitted with a 1:1 binding model. VPS35L (16-38) had a slightly higher affinity (1.87μM ± 0.8 μM) than VPS35L (28-37) (6.8μM ± 1 μM), while the L27D/L35D mutant peptide showed no binding. K_d values and standard error of the mean (SEM) are calculated from n = 3. (B) A 1.35-Å crystal structure of VPS29 bound to VPS35L (16-38) confirms the binding of the core ³⁴PL³⁵ motif to VPS29 and extended interaction of adjacent residues predicted by AlphaFold2. (C) GFP-nanotrap of GFP-VPS35L mutants in the ²⁶PL²⁷ and ³⁴PL³⁵ sequences. Data S1 shows quantified and raw blots (n = 3). (D–F) Expression of VPS35L(R293E) in a VPS35L knock-out HeLa cells fails to: (D) rescue the localization of VPS35L or the CCC complex to Retromer-decorated endosomes as observed with wild-type VPS35L; (E) the expression and stability of VPS26C and the steady-state cell surface level of α5β1-integrin; (F) the trafficking of α5-integrin away from LAMP1-positive late endosomes/lysosomes. Data S1 shows quantified band intensities and raw blots. (D and F) Pearson’s coefficients were quantified from >30 cells per 3 independent experiments. Pearson’s coefficients for individual cells and means are presented by smaller and larger circles, respectively, colored according to the independent experiment. The means (n = 3) were compared using a two-tailed unpaired t test. Error bars represent the mean, S.D. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001. See also Figure S1.

Figure S1

Comparative architecture of Retriever and Retromer assembly and context specific role of VPS29 in accessory protein binding, related to Figures 1 and 2 (A) Comparison between Retriever and Retromer assemblies. (B) VPS35L PL motif binding to VPS29 mimics association of Retromer accessory proteins, TBC1D5 (5GTU) and ANKRD27 (6TL0), and the Legionella effector RidL (5WYH) to VPS29. (C) Recombinant Strep-tagged VPS26A-Retromer and Strep-tagged VPS26C-Retriever were incubated with recombinant his-tagged TBC1D5 and subjected to Strep-tactin affinity isolation. Coomassie staining and Western analysis reveals robust association with Retromer but limited association with Retriever. Representative of two independent experiments. (D, E, and H) HEK293T cells were transfected with GFP and (D) GFP-TBC1D5, (E) GFP-VPS35 and GFP-VPS35L, and (H) mCherry-RidL (1–200) or full length (FL) RidL and subjected to GFP- or mCherry-nanotrap. Representative of three independent experiments. (F) VPS35L KO cells do not have elevated lysosomal RAB7 levels. HeLa WT or HeLa KO cells were imaged by confocal microscopy. Scale bars represent 10 μm. Representative images from 3 independent experiments. (G) Quantification of Pearson’s coefficients between RAB7 and LAMP1 from (F). For each condition, 30 cells were quantified per 3 independent experiments (90 cells total). Pearson’s coefficients for individual cells are represented by transparent circles, colored according to the independent experiment. Error bars represent the mean, S.D. Mean represented by solid triangles, colored by replicate. Normality of data was checked prior to one-way ANOVA followed by Dunnett test for multiple comparisons. ^∗∗∗∗ = p < 0.0001, ns = not significant.

Figure 3

The COMMD proteins assemble into specific heteromeric complexes (A) Purification of COMMD sub-complexes. The ten human COMMD proteins were co-expressed in E. coli and purified via His-tags on different subunits (Methods S1) followed by gel filtration. Peptide mass spectrometry identified the subunits co-purified with each tagged protein (Methods S1) and reveals three distinct stable tetrameric complexes of COMMD1-6-4-8 (subcomplex A), COMMD2-3-4-8 (subcomplex B), and COMMD5-10-7-9 (subcomplex C). (B) 3.3-Å crystal structure of the tetrameric subcomplex C (COMMD5-10-7-9), primarily built around the three major binding interfaces shown in more detail below. (C) Key residues involved in the COMMD5-COMMD10 interface. (D) Key residues that form a β-sheet extension between COMMD5-COMMD10 and COMMD7-COMMD9 dimers. (E) Key COMMD10 residue Leu129 binds in a hydrophobic pocket to stabilize the tetramer. (F) The unique COMMD9 HN domain interface in which residues form stable and specific tetrameric interactions focusing on Trp157 of COMMD5. (G) Key interactions involving the COMMD9 linker between the HN and COMM domains centered around Ile118. (H) Similar to (F) showing the COMMD10 HN domain interactions with three subunits, centered on the COMMD7 Trp139 conserved sidechain.

Figure S2

Overall, 3D and local resolution estimations, related to Figure 4 (A) Gold standard FSC (Fourier Shell Correlation) plots for the CryoSPARC reconstruction. Resolution was estimated at FSC = 0.143. (B) Gold-standard (blue) and map vs. model (red) FSC plots. Resolution of gold-standard estimated at FSC = 0.143, model-vs-map estimated at FSC = 0.5. (C) Directional FSC plots and sphericity values for the CryoSPARC reconstruction. These were calculated using a 3D-FSC server (https://3dfsc.salk.edu/). (D) Gold standard (blue) and model-vs-map (red) FSC plots for the RELION4.0 reconstruction. Gold standard resolution was estimated at FSC = 0.143, model-vs-map resolution was estimated at FSC = 0.5. (E) Directional FSC plots and sphericity values for the RELION4.0 reconstruction. These were calculated using a 3D-FSC server (https://3dfsc.salk.edu/). (F) Local resolution estimates for the RELION4.0 reconstruction. The reconstruction was colored according to local resolution estimation in RELION.

Figure S3

AlphaFold2 modeling of CCC complex across evolution, related to Figure 4 (A) Alphafold2 colored by the confidence metric (pLDDT) of human COMMD1-10 and the N-terminal domains of CCDC22 and CCDC93 with PAE plots of the top 2 ranked models. (B) Same view as in Figure 4A showing the fit of the core CCC subunits to the cryoEM density. (C–E) Further modeling of the COMMD decamer was conducted using sequences from (C) Homo sapiens, (D) Danio rerio, and (E) Salpingoeca rosetta. Each model displayed highly connected structural correlations between subunits based on PAE plots and consistent decamer assembly.

Figure 4

CryoEM structure of the human CCC complex (A) CryoEM structure of the CCC complex revealing the COMMD proteins, the CH domain of CCDC93, and linker regions of CCDC22 and CCDC93. Linker domains of CCDC22 and CCDC93 visible in our cryoEM map form intricate interactions with the decameric COMMD structure, leading to a highly intertwined structure. The CH domain of CCDC22 and extended coiled-coil regions of the CCDC proteins are not visible in current cryoEM maps due to flexibility relative to the stable COMMD decamer. (B) Molecular surface highlighting the organization of the HN domains of COMMD1, 4, 2, 10, and 7 on one side of the COMM domain ring, and COMMD8, 3, 5, and 9 on the other side. Human COMMD6 lacks the HN domain. For clarity, CCDC22 and CCDC93 are omitted. (C) Schematic model of COMMD decamer and arrangement of the sub-complexes. (D) Interweaving of CCDC22 and CCDC93 within the COMMD ring. (E) Interactions stabilizing the CCDC93 CH domain contact with the central COMMD ring, via the HN domain of COMMD4. (F and G) The PxxR sequences in CCDC22 that form turn structures: (F) the ¹⁴⁵PHLR¹⁴⁸ motif binds the HN domain of COMMD5; and (G) the ¹⁹⁹ PVGR²⁰² motif binds the COMMD3 HN domain. See also Figures S2, S3, S4, and S5.

Figure S4

CCDC22 and CCDC93 linkers make extensive contacts with the central COMM domain ring and peripheral HN domains, related to Figure 4 (A) Details of the five interfaces between the COMMD heterodimers of the heterodecameric ring. The central schematic is as shown in Figure 4D to provide a reference for each interface. Structural panels show adjacent heterotetramers in the same orientation, placing the strictly conserved Trp sidechain of each subunit as the focal point. Many specific interactions between adjacent subunits determine the precise COMMD organization. (B and C) Interfaces between CCDC22 and the HN of COMMD3 and COMMD8, and (C) between CCDC93 and HN domains of COMMD2 and COMMD4.

Figure S5

Evolutionary origins of the CCC complex, related to Figure 4 (A) Maximum-likelihood phylogeny of COMMD1-10 proteins from 23 representative eukaryotic taxa inferred under the best-fitting Q.yeast+R5 substitution model. Each COMMD forms a strongly supported (>90% bootstrap) clan in the unrooted phylogeny, and each clan contains representatives from all major lineages (supergroups) of eukaryotes; this implies that all ten COMMD subunits were already present in the last eukaryotic common ancestor (that is, LECA appears 10 times in the tree). Based on the absence of COMMD homologues in Bacteria and Archaea, this protein family likely originated on the eukaryotic stem and proliferated via a series of gene duplications prior to the radiation of the modern eukaryotic groups. (B) Presence-absence patterns of COMMD and CCDC22/CCDC93 in a set of representative modern eukaryotes. The presence-absence pattern of COMMD genes in modern eukaryotes, taken together with the phylogeny in (A), indicates that these genes have been lost independently in different eukaryotic lineages. (C) Digitonin-solubilized COMMD knockout eHap cell lines expressing the indicated COMMD^FLAG construct were affinity enriched using Flag agarose beads followed by label free quantitative proteomics. The threshold of significant enrichment was determined to be 2-fold (log2 fold change = 1) based on the distribution of unenriched proteins. Black and colored dots indicate significantly enriched proteins. Red, COMMD subunits; Blue, CCDC subunits; Green, Retriever subunits.

Figure 5

DENND10 associates with the central coiled-coil domains of CCDC22 and CCDC93 (A) Structure of DENND10 complex with the dimeric CC1 and CC2 coiled-coil regions of CCDC22 and CCDC93 predicted by AlphaFold2. Model quality and predicted alignment errors are shown in Figure S6. (B) DENND10 was titrated into purified wild-type and mutant CC1-CC2 complexes (CCDC22(325–485) + CCDC93(310–488)) and binding was measured by ITC. Top shows the raw data and bottom shows the integrated and normalized data fitted with a 1:1 binding model. The binding affinities were as follows: WT, 34 nM ± 0.5 nM; CCDC22 (V360E), 47.9 nM ± 3.5 nM; and CCDC93 (M392R), 89.1 nM ± 10 nM. No binding was detected for CCDC93H406R or E410K. (C) Analytical size exclusion chromatography of DENND10 (magenta), CC1-CC2 complex (cyan), and DENND10 mixed with CCDC22-CCDC93 forming a stable complex (orange). (D and E) GFP-nanotrap of GFP-DENND10 (D) or CCDC22 and CCDC93 mutants. Data S1 shows quantified band intensities and raw blots (n = 3). See also Figure S6.

Figure S6

AlphaFold2 modeling of the DENND10-CCDC22-CCDC93 complex, related to Figure 5 (A) AlphaFold2 of human DENND10 and the CC1-CC2 coiled-coil domains of CCDC22 and CCDC93 colored by the confidence metric (pLDDT). The PAE plots of the top 2 ranked models are shown. (B) Same model as in (A) but with DENND10 shown in surface representation colored according to conservation with CONSURF. (C) Top panel shows analytical size exclusion chromatography of DENND10 (magenta), CC1-CC2 complex (cyan) and DENND10 mixed with the CCDC22-CCDC93 forming a stable complex (orange). Bottom panel shows Coomassie stained gel of the peak fractions. (D) Predicted DENND10 structure bound to CCDC22-CCDC93 in the same orientation alongside the crystal structure of DENND1B in complex with RAB35 (PDB ID: 3TW8)..

Figure 6

Assembly of the Commander holo-complex (A) Model of Commander complex combining cryoEM and crystal structures of the CCC and Retriever sub-assemblies and AlphaFold2 modeling of the coupling of CCC and Retriever via the C-terminal coiled-coil regions of CCDC22 and CCDC92 and the CH domain of CCDC22. The general approach is shown in Methods S1. (B) GFP-nanotrap of GFP-VPS35L wild type (WT) and mutants targeting the predicted interface with the CCC complex. (C) GFP-nanotrap of GFP-CCDC22 or CCDC93 mutants targeting the predicted interface with Retriever. Data S1 shows quantified band intensities and raw blots (n = 3). See also Figures S6 and S7.

Figure S7

Conserved and electrostatic surfaces of the Commander complex, related to Figure 6 (A) Overview of Commander as a ribbon diagram (left), with conserved surfaces mapped with CONSURF (middle) and with electrostatic surface potential calculated with ChimeraX.. (B) CCDC22-CCDC93-DENND10 interface where a conserved surface aligns with a region for binding the WASH complex subunit FAM21. (C) Conserved pocket in VPS35L-VPS26C interface. (D) CCDC93 CH domain showing highly conserved surface properties.

Figure 7

Structural and functional impacts of Commander mutations causing XLID and RSS (A) VPS35L and CCDC22 mutations associated with XLID and RSS mapped onto the Commander structure. (B) Volcano plots of enriched (red circles) or depleted (blue circles) interactors in seven-plex TMT-based proteomics comparing GFP-VPS35L wild-type and GFP-VPS35L mutants causative for RSS (n = 3).