Skip to main page content
U.S. flag
See More

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Jul;643(8070):252-261.
doi: 10.1038/s41586-025-09059-y. Epub 2025 May 28.

EndoMAP.v1 charts the structural landscape of human early endosome complexes

Miguel A Gonzalez-Lozano  1   2 , Ernst W Schmid  3 , Enya Miguel Whelan  1   2 , Yizhi Jiang  1   2   4 , Joao A Paulo  1 , Johannes C Walter  3   5 , J Wade Harper  6   7
Affiliations

EndoMAP.v1 charts the structural landscape of human early endosome complexes

Miguel A Gonzalez-Lozano et al. Nature. 2025 Jul.

Abstract

Early or sorting endosomes are dynamic organelles that play key roles in proteome control by triaging plasma membrane proteins for either recycling or degradation in the lysosome1,2. These events are coordinated by numerous transiently associated regulatory complexes and integral membrane components that contribute to organelle identity during endosome maturation3. Although a subset of the several hundred protein components and cargoes known to associate with endosomes have been studied at the biochemical and/or structural level, interaction partners and higher-order molecular assemblies for many endosomal components remain unknown. Here, we combine crosslinking and native gel mass spectrometry4-7 of purified early endosomes with AlphaFold8,9 and computational analysis to create a systematic human endosomal structural interactome. We present 229 structural models for endosomal protein pairs and additional higher-order assemblies supported by experimental crosslinks from their native subcellular context, suggesting structural mechanisms for previously reported regulatory processes. Using induced neurons, we validate two candidate complexes whose interactions are supported by crosslinks and structural predictions: TMEM230 as a subunit of ATP8 and ATP11 lipid flippases10 and TMEM9 and TMEM9B as subunits of the chloride-proton antiporters CLCN3, CLCN4 and CLCN5 (ref. 11). This resource and its accompanying structural network viewer provide an experimental framework for understanding organellar structural interactomes and large-scale validation of structural predictions.

PubMed Disclaimer

Conflict of interest statement

Competing interests: J.W.H. is a co-founder of Caraway Therapeutics (a subsidiary of Merck, Inc.) and is a scientific advisory board member for Lyterian Therapeutics. All other authors declare no competing interests.

Figures

Fig. 1
Fig. 1. EEA1+ endosomal proteome analysis through dual complexomics strategies.
a, EndoMAP.v1 workflow schematic depicting integration of XL–MS, BN–MS, scoring method and structural predictions to create an endosomal protein complex structural interaction landscape. b, Endosomal scoring method; known (blue) and candidate (black) endosomal proteins ranked on the basis of combined scoring method, with higher values indicating higher probability of a protein being endosomal. The inset shows receiver operating characteristic curves for each individual metric and its combination for annotating endosomal proteins. Partial area under the curve values at 10% false-positive percentage: combined score, 6.9%; PPIs, 6.1%; dataset count, 4.0%; abundance, 2.1%. c, Correlation heat map of BN–MS co-fractionation data showing unsupervised clustering of well-known endosomal complexes. Number of proteins included in each complex is indicated in brackets. d, Co-fractionation profiles of selected protein complexes from BN–MS. e, Summary of DSSO crosslinks identified in Endo-IP samples, including intraprotein and interprotein crosslinks involving high-confidence endosomal proteins. f, Pie chart showing the number of DSSO crosslinks within and between topological compartments based on Uniprot. g, Density plots showing the distribution of Cα–Cα distances (Å) for intraprotein and interprotein DSSO crosslinks for all structures available in the PDB for the entire XL–MS dataset. The vertical dashed line indicates the maximum distance allowed by the crosslinker. h,i, Identified DSSO crosslinks (red lines) mapped into the endolysosomal V-ATPase (h, PDB 6WM2) and the class II PI3P lipid kinase complex (i, PDB 7BL1). Panel a adapted from ref. , CC BY 4.0. Source data
Fig. 2
Fig. 2. Assembly of an integrated endosome protein complex structural landscape.
a, Core component of the network containing 1,722 nodes organized into 41 communities (indicated by numbers) and 3,489 edges. Significantly enriched protein complexes of selected communities are provided in the top left (see Supplementary Table 2 for full list of communities). Diamonds and circular nodes represent high-confidence endosomal and other proteins, respectively. Solid and dashed edges represent interactions identified by at least one crosslink or only co-fractionating, respectively. Red edges indicate interaction previously reported. b, Distribution of path distances between proteins within and between the same complex compared with proteins without complex annotation. c, Distribution of fraction of direct neighbours in the same complex for each protein compared with a randomized network control. d, Systematic AF-M and AlphaLink2 predictions of protein interactions identified by XL–MS and match with the crosslink distance constraints. e, Distribution of Cα–Cα distances (Å) for interprotein DSSO crosslinks reflecting AF-M predictions with SPOC > 0.33 (orange) and SPOC < 0.33 (red). Only residues with pLDDT > 70 were considered. f, Distribution of AF-M ipTM scores and average pLDDT for predictions with ipTM > 0.3. Numbers of interprotein crosslinks evaluated and exceeding the DSSO crosslinker distance constraints are indicated by point size and the colour, respectively. g, Percentage of pairwise AF-M predictions with more or fewer than 50% of crosslinks within the distance constraint (orange and red, respectively) relative to the SPOC and ipTM score. h, ipTM scores for AF-M compared with AlphaLink2 predictions. Colour gradient represents the score difference; higher in AlphaLink2 (red) or AF-M (blue). Source data
Fig. 3
Fig. 3. TMEM230 binds endosomal P4 lipid flippases ATP11B, ATP8A1 and ATP8A2 and interface variants disrupt interaction.
a, AF-M prediction for TMEM230–ATP11B–TMEM30A, in blue, cyan and magenta, respectively. TMEM230 Y29, R78 and C terminus (Ct; D120–D121), as purple space fill, and N terminus (Nt) are indicated. ipTM and SPOC scores are provided for the ATP11B–TMEM230 interaction. b, TMEM230–ATP11B–TMEM30A BN–MS profiling. c, Overlay of AF-M and AlphaLink2 predictions for TMEM230–ATP11B. AF-M: TMEM230 (dark blue), ATP11B (cyan), crosslink (red line and arrowhead). AlphaLink2: TMEM230 (light blue), ATP11B (teal), crosslink (wheat line and arrowhead). d, HA–TMEM230 and Flag–ATP11B co-precipitation after transfection (HEK293 cells). Anti-Flag immunoprecipitates or input samples were immunoblotted for the indicated proteins. e, Basic pocket in ATP11B predicted to interact with the acidic TMEM230 C terminus (yellow). Red spheres represent aspartic residues of TMEM230. f, Identification of TMEM230-interacting proteins in iNeurons. Volcano plot showing the proteomic analysis of anti-TMEM230 immunocomplexes from WT H9 compared with H9 TMEM230−/− iNeurons (n = 3 biologically independent replicates). g, Heat map showing the log2[fold changes] in the abundance of all significantly enriched proteins in TMEM230 IPs in H9 TMEM230−/− iNeurons with or without lentiviral expression of WT and variant HA–TMEM230. Asterisks indicate significantly enriched proteins (q value < 0.05, fold change > 1.5). pep., peptide; Triplemut., TMEM230(Y29C/R78L/X121W). h, Co-precipitation of HA–TMEM230 and HA–TMEM230(Y29C/R78L/X121W) with Flag–ATP11B and TMEM30A–V5 in transfected HEK293 cells, as examined using immunoblotting of anti-HA immunocomplexes. i, Schematic of experimental design for proteomic analysis of early endosomes (TMT multiplex set 2, plex 2) and PNS (TMT multiplex set 1, plex 1) in 21-day iNeurons derived from WT, TMEM230−/− and TMEM230X121W cells in biological triplicate (Supplementary Table 4). FAIMS, high-field asymmetric waveform ion mobility spectrometry. j, Violin plots (log2[fold change]) for the indicated cohorts of proteins of PNS from TMEM230X121W and TMEM230−/− (KO) iNeurons, relative to WT cells. Two-sided paired t-test; *P < 0.01; **P < 0.001; ***P < 0.0001 (n = 3 biologically independent replicates). For violin plots, the middle line corresponds to the median; the lower and upper lines correspond to the first and third quartiles, respectively. PM, plasma membrane. k, SynGO location and function enrichment analysis of proteins significantly regulated in Endo-IP from TMEM230X121W iNeurons (Supplementary Table 4). The indicated categories were significantly enriched (−log10[q value]). SV, synaptic vesicle. Panel i adapted from ref. , CC BY 4.0; illustration of MS machine from NIAID NIH BioArt Source. Source data
Fig. 4
Fig. 4. TMEM9 and TMEM9B are core subunits of endosomal CLCN3, CLCN4 and CLCN5 Cl−–H+ antiporters.
a, EndoMAP.v1 interactions for CLCN3, CLCN4, CLCN5, TMEM9 and TMEM9B. Diamonds and circular nodes represent endosomal and other proteins, respectively. Solid and dashed edges represent interactions identified by at least one crosslink or only co-fractionation, respectively. b, BN–MS profiling for CLCN3, CLCN4, CLCN5, TMEM9 and TMEM9B. c, AF-M predictions for CLCN3–TMEM9 pair and heterotetramer. The locations of DSSO crosslinks are indicated with the red line and arrowhead. d,e, Co-localization analysis of TMEM9–GFP and mCh–CLCN3 in SUM159 cells by live-cell imaging. Mander’s coefficients of GFP and mCh puncta are shown in e (n = 39 in 3 independent replicates, mean ± s.e.m.), with an example of a cell shown in d. f, Mander’s coefficient analysis of co-localization between TMEM9–GFP, mCh–CLCN3, anti-EEA1 and anti-LAMP1 in fixed SUM59 cells as determined by immunofluorescence. The number of fields of view across three biological replicates is indicated (mean ± s.e.m.) and P values from linear mixed-effects model analysis of variance. g, Example of TMEM9–GFP, mCh–CLCN3 and anti-EEA1 staining in a cell expressing high levels of CLCN3 (left panels), which promotes the formation of swollen endolysosomes. Traces of the white line in the bottom panel show the overlap of the three proteins in the limiting membrane of endosomes (right panel). h, Volcano plot showing the proteomic analysis of anti-HA IPs from TMEM9−/− iNeurons with or without lentiviral expression of TMEM9–HA (n = 4 biologically independent replicates). i, Schematic of experimental design for proteomic analysis of early endosomes and PNS in 21-day iNeurons derived from WT cells, TMEM9−/− cells and two different clones of TMEM9−/−TMEM9B−/− (DKO) cells in biological triplicate (Supplementary Table 5). j, Volcano plot showing the proteomic analysis of Endo-IPs from TMEM9−/−TMEM9B−/− (DKO clone 2) versus WT iNeurons (day 21; n = 3 biologically independent replicates). CTSF, cathepsin F. k, TMT reporter signal intensity for CLCN3, CLCN5, TMEM9 and TMEM9B in Endo-IPs from iNeurons with the indicated genotypes (n = 3 biologically independent replicates). DKO1, TMEM9−/−TMEM9B−/− (clone 1); DKO2, TMEM9−/−TMEM9B−/− (clone 2). Scale bars (d and g), 5µm. Panel i adapted from ref. , CC BY 4.0; illustration of MS machine from NIAID NIH BioArt Source. Source data
Fig. 5
Fig. 5. Towards a structural proteomic landscape for early endosomes.
a, Systematic AF-M structural predictions for three-way clique assemblies within EndoMAP.v1 and match with the crosslinker distance constraints. b, Pairwise AF-M prediction for VPS35–RAB7A (left) and tetramer prediction for retromer–RAB7A complex (right) and associated crosslinks from EndoMAP.v1. ipTM and SPOC scores for pairwise combination are shown. c, AF-M structural predictions and interprotein crosslinks within the BORC endolysosomal positioning complex. Pairwise AF-M predictions (left), three-way clique predictions (middle) and eight-protein predictions (right) are shown along with associated interprotein crosslinks. ipTM and SPOC scores for pairwise combinations are indicated. d, Pairwise AF-M predictions and associated crosslinks for a RUFY1–RUFY2 heterodimer (right) and for interaction of the RUFY2 N-terminal helical domain with ARL8B (left). e, Pairwise AF-M predictions and associated crosslinks for LAMTOR4 and LAMTOR5 (left), RRAGA and RRAGC (middle), and crosslinks mapped onto the ragulator structure (PDB 6U62) (right). DSSO crosslinks (red) and DHSO or DMTMM crosslinks (cyan) are indicated with lines and arrowheads.
Extended Data Fig. 1
Extended Data Fig. 1. Endosomal proteome scoring method and optimization of large-scale Endo-IP.
a, Overview of datasets used for endosomal scoring, including the number of proteins identified in each and across datasets, and the number of well-known endosomal proteins identified in the indicated studies. b, Multiple correspondence analysis (MCA) showing an overview of the relationship among datasets from a. Each node represents a dataset color-coded by isolation method and proportional size to the total number of proteins identified. c, Bar plot depicting the number of proteins identified across multiple datasets for several subcellular compartments. d, Line graph showing the percentage of proteins identified across all 16 datasets in a and their protein abundance in our Endo-IP experiments from HEK293 cells (Supplementary Table 1) represented as loess (Locally Estimated Scatterplot Smoothing) regression line and 95% confidence level interval band. f,g, Number of bait proteins, protein-protein interactions (PPIs), and PPIs per protein in Bioplex (panel f) and Open Cell (panel g) according to organelle assignment (n number of proteins in each category is indicated on top). For box plots, the middle line corresponds to the median, the lower and upper end of the box correspond respectively to the first and third quartiles, and the whiskers extend from the box to 1.5 times the inter-quartile range. h, Schematic of the purification steps from Endo-IP to endosomal pellet used for complexomics. i,j, Number of proteins identified (panel i) and abundance per compartment (panel j) in endosomal pellet compared to the input (PNS), supernatant, or NP40 eluate from the Endo-IP as depicted in panel h. k, Violin plot showing the fold-change enrichment of proteins from individual organelle compartments in endosomal pellets compared to input (PNS). l, Volcano plot showing fold-changes and FDR adjusted p-value for proteins in endosomal pellets compared to input (PNS) (n = 3 biologically independent replicates). DEqMS algorithm was used for statistical analysis with multiple testing correction as implemented in. h, Images modified from ref.  (Copyright (2024) National Academy of Sciences). Source data
Extended Data Fig. 2
Extended Data Fig. 2. Application of correlation profiling and cross-linking proteomics to endosomes purified by Endo-IP.
a, Co-fractionation profiles of selected protein complexes from BN-MS. b, Number of Bioplex interactions identified by BN-MS compared to co-fractionation PCProphet scores. c, Number of Bioplex interactions identified using PCProphet in either 2 or 3 replicates of the Endo-IP BN-MS compared to the maximum number of proteins per complex allowed in the analysis. d, Box plot depicting the protein MS signal intensity in Endo-IP compared to the number of DSSO cross-links identified for each protein (n number of proteins in each category is indicated on top). e, Box plot depicting the minimum protein MS signal intensity for PPIs identified by BN and DSSO cross-linking compared to all proteins identified in Endo-IP (n number of interactions in each category is indicated on top). f, Distribution of protein copy number (log10) for cross-linked proteins compared to the whole proteome. g-i, Box plots depicting the protein copy number (g), number of interactors in BioPlex (h), and molecular weight (i) compared to the number of interprotein DSSO cross-links identified for each protein (n number of proteins in each category is indicated on top). j, Venn diagram showing the number of protein pairs identified by yeast two hybrid (YTH), Bioplex, and cross-linking proteomics for the same set of proteins. k, Boxplot showing co-fractionation SECAT p-values for cross-linked proteins identified by different number of DSSO cross-links (n number of interactions in each category is indicated on top). l,m, Number (panel l) and rank (panel m) of cross-linked protein interactions that have been previously reported (or not) compared to their co-fractionation SECAT p-value (Supplementary Table 2). SECAT was used for statistical analysis,. n, Overview of protein interactions within EndoMAP.v1 including the method, organelle and previous reports. o, Venn diagram showing the overlap of endosomal interactions between EndoMAP.v1 and Bioplex for interactions in which both proteins are present in both datasets. For all box plot panels, the middle line corresponds to the median, the lower and upper end of the box correspond respectively to the first and third quartiles, and the whiskers extend from the box to 1.5 times the inter-quartile range. Source data
Extended Data Fig. 3
Extended Data Fig. 3. EndoMAP.v1 network characterization and application of AlphaFold-M across DSSO cross-linked protein pairs.
a, Degree distribution (number of edges per node) of the complete network. b, Power law log-log plot of the complete network showing the degree of a node (number of edges) and the probability. c, Distribution of the shortest path distances between all proteins in the complete interaction network. d, Distribution and number of PPIs within and between selected organelles (Supplementary Table 2). e, Criteria for network filtering to create an integrated endosomal network (EndoMAP.v1, see METHODS). f, Mapping of known protein complexes from CORUM onto the core components of the EndoMAP.v1 network (Supplementary Table 2). g,h, DisGeNET enrichment analysis of endosomal proteins as defined by our scoring method (panel g) and Gene Ontology (GO:0005768, panel h). Top 15 categories by highest gene ratio are depicted. Disorders related to the nervous system are indicated in bold. p-values by hypergeometric test were adjusted with Benjamini-Hochberg correction. i, Enrichment analysis of the endosomal proteome within several neurodegenerative diseases (LSD, Lysosomal Storage Disorders; ALS, Amyotrophic Lateral Sclerosis, PD, Parkinson’s disease; ASD, Autism Spectrum Disorders; DD/ID, epilepsy and severe neurodevelopmental disorder). j, Mapping of neurodegenerative disease related proteins onto the core component of EndoMAP.v1 network (see METHODS, Supplementary Table 2). k, Distribution of shortest path distances within various classes of neurodegenerative disease related proteins. Three different sources of disease genes were used to retrieve proteins related to PD (see METHODS). l, Distances between DSSO cross-linked lysines for AF-M predictions compared to structures in the PDB. Green and orange dots represent interprotein and intraprotein cross-links, respectively. Filled and empty dots represent predictions with SPOC > 0.33 or SPOC < 0.33, respectively. m, Distribution of Cα-Cα distances (Å) for intraprotein DSSO cross-linked lysines in all AF-M predictions compared to all lysines. n, Distribution of Cα-Cα distances (Å) for interprotein DSSO cross-linked lysines in all AF-M predictions compared to all lysines. o, Distribution of SPOC scores and average pLDDT for predictions with SPOC > 0. Number of interprotein DSSO cross-links evaluated and exceeding the cross-linker distance restrain are indicated by point size and the color, respectively. p, Box plot showing the distribution of SPOC scores relative to the number of DSSO cross-links identified for each interaction (n number of interactions in each category is indicated on top). The middle line corresponds to the median, the lower and upper end of the box correspond respectively to the first and third quartiles, and the whiskers extend from the box to 1.5 times the inter-quartile range. q,r, Distribution of Cα-Cα distances (Å) for intraprotein (q) and interprotein (r) DSSO cross-linked lysines in AF-M predictions involving endosomal proteins compared to all lysines. s, Distribution of Cα-Cα distances (Å) for interprotein DSSO cross-links reflecting predictions involving endosomal proteins with SPOC > 0.33 (orange) and SPOC < 0.33 (red). Source data
Extended Data Fig. 4
Extended Data Fig. 4. EndoMAP.v1 extension by AlphaLink2 and XL-MS using DHSO/DMTMM cross-linkers.
a, Overlap of DSSO cross-linking data analyzed using XlinkX at 5% FDR compared to Scout at 1%FDR. b, Number of protein interactions based on DSSO cross-links identified with XlinkX and Scout for known interactions and across the selection criteria used in EndoMAP.v1 (i.e. filtering for AF-M score, endosomal protein and cross-link distance). c, ipTM scores for AF-M compared to AlphaLink2 predictions. Color gradient represents the score difference; higher in AlphaLink2 (red) or AF-M (blue). d, Distances between DSSO cross-linked lysines for AF-M compared to AlphaLink2 predictions. Green and orange dots represent interprotein and intraprotein cross-links, respectively. e-i, Individual and overlay AF-M and AlphaLink2 predictions for several protein pairs (see Supplementary Text). DSSO and DHSO/DMTMM interprotein cross-links are indicated with red and cyan lines and arrowheads, respectively. j, Mapping DHSO/DMTMM cross-linking data to the proteins and interactions identified with DSSO. k, Pie chart showing the number of protein pairs identified with both DMTMM and DSSO (top) or DHSO and DSSO (bottom). l, Identified DSSO (red) and DHSO/DMTMM (cyan) cross-links mapped into the endolysosomal V-ATPase (PDB:6WM2). m,n, Distribution of Cα-Cα distances (Å) for intraprotein (m) and interprotein (n) cross-linked residues in AlphaLink2 predictions. o, Distribution of Cα-Cα distances (Å) for interprotein cross-links reflecting predictions with SPOC > 0.33. Source data
Extended Data Fig. 5
Extended Data Fig. 5. Interface variants disrupt interaction of TMEM230 with endosomal P4 lipid flippases ATP11B and ATP8A1/2.
a, Individual and overlay AF-M and AlphaLink2 predictions for TMEM230 and ATP11B. AF-M: TMEM230 (light blue), ATP11B (cyan), cross-link (red line and arrowhead). AlphaLink2: TMEM230 (dark blue), ATP11B (teal), cross-link (wheat line and arrowhead). b, Overlay of yeast DNF1-LEM3 structure (PDB:7DRX) in the EP2 conformation with AF-M prediction for ATP11B-TMEM30A-TMEM230. c, Co-precipitation of Flag-ATP11B and TMEM30A-V5 with HA-TMEM230. The indicated plasmids were transfected into HEK293 cells and α-HA immunoprecipitates or input samples were immunoblotted for the indicated proteins. Black dots indicate proteins expressed in each sample. d, Sequence validation of TMEM230-/- and TMEM230X121W clones in H9AAVS1-NGN2;Flag-EEA1 cells (H9-Flag-EEA1), showing the location of the sgRNA used (green) and base pairs deleted to create an out of frame mutation and point mutation, respectively. e, Immunoblot of total cell lysates from the indicated H9-Flag-EEA1 cell lines probed with α-TMEM230. The X121W mutation adds a six-residue extension (WHPPHS), which can be detected as a band with slightly higher molecular weight. Stain-free gel was used to indicate equal loading of extracts. f, Volcano plots (log2FC relative to TMEM230-/- cells) of TMEM230 immunoprecipitations in H9-TMEM230-/- iNeurons with or without lentiviral expression of WT and interface variant HA-TMEM230 proteins. g, Mass spectrometry (MS) TMT reporter signal for ATP11B and TMEM30A in the indicated TMEM230 variant immunoprecipitation from iNeurons. Dots indicate individual biological replicates (n = 2, except n = 3 for Control given the limitation of the maximum number of TMT channels). h, Immunoblots of total cell extracts from TMEM230-/- iNeurons transduced with lentiviruses expressing the indicated variants of HA-TMEM230 protein. Stain-free gel was used as loading control. i, AF-M prediction for a TMEM230-ATP8A1-TMEM30A complex (Y29, R78, and C-terminal D120-D121, purple space fill). The location of a cross-link between ATP8A1 and TMEM30A is indicated by the red line and arrowhead. ipTM = 0.74 for ATP8A1-TMEM230 prediction. j, Volcano plot for Endo-IP proteomic analysis from H9-Flag-EEA1 iNeurons (21 days) (n = 3 biologically independent replicates). Proteins annotated as endosomal (green), lysosomal (blue), or plasma membrane (PM, orange) are indicated. k, Immunofluorescence microscopy showing the colocalization of Flag-EEA1 (green) with RAB5 (magenta) in iNeurons from H9-Flag-EEA1 cells. l, Violin plot showing the fold-change enrichment (log2) of proteins from individual organelle compartments (color-coded as panel j) in Endo-IP samples from H9-Flag-EEA1 iNeurons (day 21). m, Immunoblots of Endo-IP or input samples (PNS) from H9-Flag-EEA1 iNeurons and untagged H9 control (21 days). Blots were probed with the indicated antibodies.
Extended Data Fig. 6
Extended Data Fig. 6. Proteomic profiling of postnuclear supernatant (PNS) and Endo-IP from TMEM230 mutant iNeurons.
a,b, Volcano plots of PNS proteomic analysis from TMEM230-/- (panel a) and TMEM230X121W (panel b) iNeurons compared to WT (day 21) (n = 3 biologically independent replicates). c, Violin plot showing the fold-change enrichment (log2) of proteins from individual organelle compartments in PNS from TMEM230-/- and TMEM230X121W iNeurons compared to WT (day 21). d, SynGO location enrichment analysis of proteins significantly regulated in PNS from TMEM230X121W iNeurons (Supplementary Table 4). The indicated categories were significantly enriched (−log10q-value). e,f, Volcano plots of Endo-IP proteomic analysis from TMEM230-/- (panel e) and TMEM230X121W (panel f) iNeurons compared to WT (day 21) (n = 3 biologically independent replicates). g, Heatmap showing the abundance fold-changes (log2) for all significantly regulated proteins in Endo-IPs from TMEM230-/- or TMEM230X121W iNeurons (21 day) compared to WT. Synaptic proteins annotated in SynGO (see METHODS) are indicated in bold. Asterisks indicate significantly regulated proteins (q-value < 0.05 and fold-change > 1.5). Abundance fold-changes in PNS are also indicated, except for proteins not detected (nd). h, Heatmap for the abundance fold-changes (log2FC) of selected proteins in PNS and Endo-IPs from TMEM230-/- or TMEM230X121W iNeurons (21 day) compared to WT. Asterisks indicate significantly regulated proteins (q-value < 0.05 and fold-change > 1.5) and nd for proteins not detected. i, Summary of pairwise AF-M predictions harboring candidate disease variants within 2 amino acids of the interface for endosomal and non-endosomal proteins. j, Candidate disease variants at the interaction interface of pairwise protein AF-M predictions. Predicted aligned error plots (left), predicted structures with ipTMs (center left, interprotein DSSO cross-links indicated by red lines) and close-up view of disease variant residues (yellow) at the interaction interface (center right, and right; dotted lines indicate predicted hydrogen bonds). Source data
Extended Data Fig. 7
Extended Data Fig. 7. TMEM9/9B are core subunits of endosomal CLCN3/4/5 Cl−-H+ antiporters.
a, Endosomal score and rank of TMEM9/9B and CLCN3/5/7, with higher values corresponding to endosomal proteins. b, Individual and overlay AF-M and AlphaLink2 predictions for CLCN3-TMEM9. AF-M: TMEM9 (dark blue), CLCN3 (cyan), cross-link (red bar and arrowhead). AlphaLink2: TMEM9 (light blue), CLCN3 (teal), cross-link (wheat bar and arrowhead). c,d AF-M predictions for CLCN3-TMEM9B and selected CLCN-TMEM9/9B heterotetramers. The locations of DSSO cross-links are indicated with the red line and arrowhead. The location of variants found in CLCN5 in Dent’s Disease retrieved from UniProt are shown in red (right, panel c). e, Overlay of the CLCN5-TMEM9 heterotetramer prediction with the CLCN7-OSTM1 heterotetramer structure (PDB: 7JM7). f, Example of TMEM9-GFP, mCh-CLCN3, and α-LAMP1 staining in a cell expressing high levels of CLCN3, which promotes the formation of swollen endolysosomes. Line traces show the overlap of the 3 proteins in the limiting membrane of endolysosomes (bottom right panel). g, Co-precipitation of CLCN3/5-Flag and TMEM9/9B-HA. The indicated plasmids were transfected into HEK293 cells and α-HA or α-Flag immunoprecipitations or input samples were immunoblotted with the indicated antibodies. Loading controls as stain-free gels are shown. Black dots indicate proteins expressed in each sample. h,i, Sequence validation of H9 TMEM9-/- cells, showing the location of the sgRNA used (green) to create frameshift mutations in TMEM9 (panel h) and subsequently, used the indicated sgRNA (green) to create frameshift mutations in TMEM9B (panel i). j, (Left) Volcano plot for Endo-IP proteomic analysis from H9-Flag-EEA1 iNeurons (n = 3 biologically independent replicates). (Right) Violin plot showing the fold-change enrichment (log2) of proteins from individual organelle compartments in Endo-IP from H9-Flag-EEA1 iNeurons. Proteins annotated as endosomal (green), lysosomal (blue), or plasma membrane (PM, orange) are indicated. Source data
Extended Data Fig. 8
Extended Data Fig. 8. Proteomic profiling of postnuclear supernatant (PNS) and Endo-IP from TMEM9-/- and TMEM9/9BDKO iNeurons.
a, Volcano plots for PNS (post-nuclear supernatants) proteomic analysis from TMEM9-/- and two clones of TMEM9/9BDKO iNeurons (day 21) compared to WT (n = 3 biologically independent replicates). b, Volcano plots for Endo-IP proteomic analysis from TMEM9-/- and one clone of TMEM9/9BDKO iNeurons compared to WT (n = 3 biologically independent replicates). c, Immunoblots of input and Endo-IP samples from the experiment outlined in Fig. 4i. Blots were probed with the indicated antibodies. d, Heatmap showing the abundance fold-changes (log2) for all significantly regulated proteins in Endo-IPs from TMEM9-/- and TMEM9/9BDKO iNeurons (day 21) compared to WT. Asterisks indicate significantly regulated proteins (q-value < 0.05 and fold-change > 1.5). Abundance fold-changes in PNS are also indicated, except for proteins not detected (nd).
Extended Data Fig. 9
Extended Data Fig. 9. 3-way Clique and higher order AF-M predictions reveal extensive SNARE interactions and assemblies.
a, Pairwise (top) and 3-way clique (bottom) AF-M predictions and associated DSSO cross-links for components of the Class II PI3K complex. Identified cross-links are also mapped onto the cryo-EM structure of the PI3K complex (PDB:7bl1) (lower right). b, Overlay of AF3 predictions of a VPS29-VPS35-VPS26A-RAB7AGTP complex and associated DSSO cross-links with a RAB7AGTP crystal structure (PDB:1T91). c, Summary of cross-link and AF-M predictions for SNARE components and their interactors. R-SNARE, Q-SNARE, known and candidate regulators and RAB proteins found with cross-links within EndoMAP.v1 are shown. Lines indicate one or more cross-links and are shown in distinct forms to facilitate visualization of connections. Colored dots indicate SPOC score for each AF-M pairwise prediction. d, Cross-links and pairwise AF-M predictions for “core” SNARE components VAMP3, STX7, STX8, and VTI1B. ipTM and SPOC scores are indicated for pairwise combinations. e, Examples of a subset of 3-way clique predictions and associated cross-links involving core SNARE components as well as NAPA. f, Core SNARE AF-M predictions and associated cross-links. The prediction resembles a post-vesical fusion-like conformation. g, AF-M predictions and associated cross-links for SNARE association with soluble fusions factors. h, Predicted interactions and cross-links for association of VPS16 with either STX8 or STX8 in the core SNARE complex. i, Core SNARE assembly predictions and associated cross-links with candidate interactors SCAMP1 and SCFD1. j, Summary of physical interactions involving SCAMP proteins in OpenCell and cross-links identified in our study. Intraprotein cross-links are not shown (see Supplementary Table 2). k, Summary of physical interactions involving PTTG1IP proteins in OpenCell and cross-links identified in our study. l, AF-M prediction for tetrameric SNARE complex composed of VTI1B, STX7, STX8, and VAMP7. m, Pentameric prediction for VTI1B, STX7, STX8, and VAMP7 together with PTTG1IP. Grey rectangles represent the transmembrane section of the complex. Left, cross-links not shown; Right, cross-links shown. n, Pentameric AF3 prediction for VTI1B, STX7, STX8, and VAMP8 together with PTTG1IP. DSSO and DHSO/DMTMM interprotein cross-links are indicated with red and cyan lines and arrowheads, respectively.
Extended Data Fig. 10
Extended Data Fig. 10. Endosomal Regulatory Proteins, Channels, Cargo, and Trafficking Complex AF-M Predictions.
a, Pairwise AF-M predictions and associated cross-links for two pairs of RABs with high scoring predictions. b,c, Pairwise AF-M predictions and associated cross-links for selected RABGEF complexes present in EndoMAP.v1 (panel b), and for a RAB11A-SH3BP5 complex (panel c) overlayed with a previously determined structure of the complex (PDB:6DJL). d, Pairwise AF-M prediction and associated cross-links for a RAB8A-SYTL4 (synaptotagmin-like) Snare complex. e-g, AF-M predictions and associated cross-links for selected channel/transporter assemblies in EndoMAP.v1. LRRC8 proteins (panel e) form hexamers and are components of volume regulated anion channels important for cell volume homeostasis. OSTM1-CLCN7 (panel f) is an endolysosomal voltage-gated channel mediating exchange of chloride against protons and is known to form a heterotetramer. CLCN7 was found cross-linked to RMC1 (panel g), a subunit of the CCZ1-MON1 GEF for RAB7 on endolysosomes. h, Pairwise AF-M predictions for cross-link containing AP1 components AP1G1 and either AP1S1 or AP1S2 (left) and tetramer AF-M prediction for AP1G1-AP1B1-AP1M1 and either AP1S1 or AP1S2 (upper panel). Pairwise and 3-way clique predictions for AP2 components AP2M1, AP2B1, or AP2A2 (lower panel). i, Pairwise predictions and associated cross-links for ESCRT and ubiquitin (Ub)-related modules within EndoMAP.v1. j, Pairwise predictions and associated cross-links for INSR and IGF1R. k, Pairwise AF-M predictions and associated cross-links for selected HOPS complex components (left) and a 3-way clique prediction (right) that maintains compatible cross-link distances. l, Pairwise predictions and associated cross-links for the FLOT1/2 complex that participates as a scaffolding protein within caveolar membranes, and the ITSN1-EPS15L complex that links endosomal membrane trafficking with actin assembly machinery. For all panels, DSSO and DHSO/DMTMM interprotein cross-links are indicated with red and cyan lines and arrowheads, respectively. Intraprotein cross-links are not shown (see Supplementary Table 2).
Extended Data Fig. 11
Extended Data Fig. 11. V-ATPase as an interaction hub.
a, DSSO cross-links identified between components of the V-ATPase (purple), the BORC complex (red), the LAMTOR complex (blue) and RAB proteins (green). b, Hypothetical model for association of MTOR-Ragulator complex (PDB:7UXH) with V-ATPaseADP (PDB:6WM2) based on two DSSO cross-links between LAMTOR2 and LAMTOR4 with ATP6V1C1 (red lines). Docking model was generated using HADDOCK (see METHODS). c, Pairwise AF-M prediction and associated cross-link for MEAK7 and ATP6V1B2 (left) is compared with the MEAK7-ATP6V1B2 sub-complex from PDB:7U4T (right). d, MEAK7-V1-ATPase (PDB:7U4T) together with overlay of AF-M prediction for MEAK7-ATP6V1B2. Cross-links between MEAK7 and either ATP6V1B2 or ATP6V1D are shown by red lines. e, MEAK7-ATP6V1B2 AF-M prediction modeled on PDB:7U4T and identified cross-links with ATP6V1D (red arrowhead) is shown on the left. MEAK7-ATP6V1D AF-M prediction and associated cross-link is shown on the right. f, Endosomal RABs cross-link with the ATPV0A1 subunit of the V0-ATPase. The structure of the V0-ATPase complex, with ATP6V0A1 shown in salmon, is presented on the far left. The AF-M predictions for 4 endosomal RABs and the detected cross-links are also shown. Intraprotein cross-links are not shown (see Supplementary Table 2). g, Screenshot from our EndoMAP.v1 website and AF-M prediction viewer at https://endomap.hms.harvard.edu/. Left panel shows the output of a search for TMEM230. Right panel shows the output of AF-M prediction for TMEM230-ATP11B interaction, together with predicted alignment error plots.
See this image and copyright information in PMC

Update of

References

    1. Cullen, P. J. & Steinberg, F. To degrade or not to degrade: mechanisms and significance of endocytic recycling. Nat. Rev. Mol. Cell Biol.19, 679–696 (2018). - PubMed
    1. Naslavsky, N. & Caplan, S. The enigmatic endosome - sorting the ins and outs of endocytic trafficking. J. Cell Sci.10.1242/jcs.216499 (2018). - PMC - PubMed
    1. Vietri, M., Radulovic, M. & Stenmark, H. The many functions of ESCRTs. Nat. Rev. Mol. Cell Biol.21, 25–42 (2020). - PubMed
    1. Liu, F., Rijkers, D. T., Post, H. & Heck, A. J. Proteome-wide profiling of protein assemblies by cross-linking mass spectrometry. Nat. Methods12, 1179–1184 (2015). - PubMed
    1. Yu, C. & Huang, L. New advances in cross-linking mass spectrometry toward structural systems biology. Curr. Opin. Chem. Biol.76, 102357 (2023). - PMC - PubMed

LinkOut - more resources