Ubiquitination regulates ER-phagy and remodelling of endoplasmic reticulum

Abstract

The endoplasmic reticulum (ER) undergoes continuous remodelling via a selective autophagy pathway, known as ER-phagy¹. ER-phagy receptors have a central role in this process², but the regulatory mechanism remains largely unknown. Here we report that ubiquitination of the ER-phagy receptor FAM134B within its reticulon homology domain (RHD) promotes receptor clustering and binding to lipidated LC3B, thereby stimulating ER-phagy. Molecular dynamics (MD) simulations showed how ubiquitination perturbs the RHD structure in model bilayers and enhances membrane curvature induction. Ubiquitin molecules on RHDs mediate interactions between neighbouring RHDs to form dense receptor clusters that facilitate the large-scale remodelling of lipid bilayers. Membrane remodelling was reconstituted in vitro with liposomes and ubiquitinated FAM134B. Using super-resolution microscopy, we discovered FAM134B nanoclusters and microclusters in cells. Quantitative image analysis revealed a ubiquitin-mediated increase in FAM134B oligomerization and cluster size. We found that the E3 ligase AMFR, within multimeric ER-phagy receptor clusters, catalyses FAM134B ubiquitination and regulates the dynamic flux of ER-phagy. Our results show that ubiquitination enhances RHD functions via receptor clustering, facilitates ER-phagy and controls ER remodelling in response to cellular demands.

Fig. 1. Ubiquitination of FAM134B RHD.

a, MS workflow for the analysis of GFP–FAM134B co-immunoprecipitated from cell lysates following treatment with Torin 1 (6 h, orange) and DMSO (basal, blue) or from mock-treated cells (GFP empty, light blue). HF, Qexactive HF (Ultra-High-Field Orbitrap) mass spectrometer; LFQ, label free quantification; 3xR 3 biological replicates. Panel a was partly generated using Servier Medical Art (Servier), licensed under a Creative Commons Attribution 3.0 unported license. b, Ubiquitination of FAM134B under control conditions and in response to 250 nM Torin 1 for 6 h. The diGly peptide intensities of FAM134B are normalized to the total intensities of modified and non-modified FAM134B peptides (data are mean ± s.d.; n = 3 independent experiments, two-way ANOVA, Bonferroni post-hoc test). c, Schematic organization of FAM134B. The RHD consists of two transmembrane segments (TM, green) separated by a linker and two conserved amphipathic helices (AH, yellow). The conserved lysine residues (blue) and ubiquitinated lysines (red) are highlighted. d, TUBE-2 pulldown assay showing increased time-dependent ubiquitination of FAM134B following Torin 1 treatment. WB, western blot. e, Densitometric quantification of the immunoblot signals for ubiquitinated HA–FAM134B normalized to total HA–FAM134B levels (d). Data are mean ± s.d.; n = 3 independent experiments; one-way ANOVA, Bonferroni post-hoc test. f, FAM134B RHD ubiquitination assay in cells showing a reduction of ubiquitination when the 17 conserved lysine residues are replaced with arginine (Myc–Ub immunoprecipitation (IP)). Lack of RHD ubiquitination reduces binding to LC3B-II and the abundance of oligomeric species (HA–FAM134B IP). IB, immunoblot; exp., experiment. g, Densitometric quantification of the immunoblot signals from f (lanes 1 and 2): LC3B-II bound to HA–FAM134B WT or HA–FAM134B 17KR, and the oligomers. Data are mean ± s.d.; n = 5 and n = 3 independent experiments, respectively; one-tailed unpaired Student’s t-test. h, Confocal fluorescence microscopy analysis of HA–FAM134B co-labelled with LC3B and Ub in cells treated with 250 nM Torin 1 for 2 h. Arrows indicate autophagosome (LC3B positive in blue) that colocalizes with HA FAM134B (green) and ubiquitin (red). Scale bar, 10 µm. i, Quantification of the fluorescence signal of HA–FAM134B–LC3B puncta that colocalizes or not with Ub from h (based on Pearson’s correlation coefficients). Data are mean ± s.d., n = 10 cells. Source data

Fig. 2. Role of Ub in the structure and function of RHD.

a, Equilibrated structures of ubiquitinated and non-ubiquitinated FAM134B RHD variants in MD simulations highlight the arrangement of Ub moieties (cyan and blue) with respect to the RHD (green) and the POPC bilayer (orange beads). b, Ubiquitination accelerates membrane curvature induction. K160–Ub (left) induces faster transitions of bicelles to closed vesicles than non-ubiquitinated RHDs (cumulative distribution function (CDF) of waiting times shown on the right; compare cyan, blue and purple versus green). c, In vitro liposome remodelling experiments using purified protein samples (GST–Ub, GST–RHD_90–264, GST–Ub–RHD_90–264–Ub) incubated with liposomes for 8 h at 25 °C. The top panel shows representative negative-stain transmission electron micrographs. Remodelled proteoliposomes were quantified by measuring their diameters (dotted lines) using ImageJ (version 1.51w). The bottom panel shows violin plots of liposome size distributions. Violin plots show the boxplots with median value (black dot), the interquartile range (black shaded region), the minimum and maximum values (1.5× the interquartile region) and mirrored probability density estimates on sides (coloured shaded region). GST–RHD mean = 64.40 nm; GST–Ub–RHD_90–264–Ub mean = 39.10 nm; GST–Ub mean = 127.47 nm. GST–RHD_90–264 n = 625; GST–Ub–RHD_90–264–Ub n = 961; GST–Ub n = 1,573; Kruskal–Wallis or Dunn’s post-hoc test. d, Ub promotes receptor clustering and membrane remodelling. Snapshots show the arrangement of nine (K160 + K264)–Ub–RHD molecules in model POPC bilayers (orange) at the end of the simulations. MD simulations were performed under four different conditions by altering bilayer asymmetry, ΔN = 0 and 300, and protein–protein interaction strength, α = 1 and 0.65, as quantified in the panels below. For the rows: time traces of the box width (L_X) during the four simulations (top); vertical displacement (z) of individual ubiquitinated proteins (centre-of-mass positions shown as green lines), with the highest and lowest points in the membrane shown as orange lines, and the intervening range in light grey (middle); the size of the largest protein cluster as a function of time for different simulation conditions (bottom). Source data

Fig. 3. Effect of RHD ubiquitination on the flux of ER-phagy and FAM134B cluster size.

a, The ER-phagy reporter system mCherry–GFP–FAM134B. b, U2OS TRex stable cell lines expressing mCherry–GFP–FAM134B WT or mCherry–GFP–FAM134B 17KR. Scale bar, 10 µm. c, ER-phagy flux was quantified as the ratio between mCherry⁺GFP^– and mCherry⁺GFP⁺ puncta. n = 4 independent experiments in which the total number of cells per condition for mCherry–GFP–FAM134B WT were: 482 (basal (DMSO)), 738 (BafA1), 673 (EBSS), 842 (Torin 1) and 667 (Torin 1 + BafA1). The number of cells per condition for mCherry–GFP–FAM134B 17KR: 440 (DMSO), 864 (BafA1), 723 (EBSS), 968 (Torin 1), 535 (Torin 1 + BafA1). Data are mean ± s.d.; one-way ANOVA, Bonferroni post-hoc test. d, DNA-PAINT super-resolution image of HA–FAM134B. Microscale (red) and nanoscale (yellow) clusters (red) are indicated. Scale bars, 10 µm (left panel) and 1 µm (right panel; magnified region from the left panel). e, Relative frequency distribution of HA–FAM134B WT and HA–FAM134B 17KR cluster areas (110 > n_cluster < 251) identified in U2OS cells under basal conditions (n_WT = 10 cells, n_17KR = 8 cells) or following Torin 1 treatment (n_WT = 11 cells, n_17KR = 14 cells). f, DNA-PAINT super-resolution imaging of HA–FAM134B WT and LC3B-II nanoscale clusters (yellow dashed circles). Dashed white line in top panel indicates cell outline. Scale bars, 10 µm (top panel) and 1 µm (bottom panel; magnified region from the top panel). g, Relative frequency distribution of the diameter of HA–FAM134B and HA–FAM134B 17KR nanoscale cluster (n_cells = 4, n_{WT clusters} = 1,278; n_{17KR clusters} = 1,255). Histograms were fitted with a log-normal distribution followed by a non-parametric one-tailed Mann–Whitney U-test. Cluster diameters were determined from the mode of the log-normal distribution using the mean and standard deviation (HA–FAM134B 17KR: µ = 88nm, σ = 24 nm; HA–FAM134B WT: µ = 123 nm, σ = 46 nm). h, Quantitative analysis of FAM134B copy numbers in nanoscale clusters. Scale bar, 1 µm. Relative frequency distribution of the inverse dark times (τ_D) of single-molecule binding time intervals recorded with DNA-PAINT (right). In the inset, the grey bars indicate calibration. Histograms were fitted with a log-normal distribution: FAM134B 17KR µ = 3.1 × 10⁻⁴ s⁻¹, σ = 3.5 × 10⁻⁴ s⁻¹; FAM134B WT µ = 9.4 × 10⁻⁴ s⁻¹, σ = 7.1 × 10⁻⁴ s⁻¹; calibration cluster µ = 1.9 × 10⁻⁴ s⁻¹, σ = 2.8 × 10⁻⁴ s⁻¹. Source data

Fig. 4. Ubiquitination site profiling and protein interactors of FAM134B-containing oligomers.

a, Schematic representation of the bimolecular complementation affinity purification assay of FAM134B dimers. The full-length FAM134B was fused to the C-terminal of the two non-fluorescent complementary fragments of the Venus fluorescent protein (V1–FAM134B and V2–FAM134B). b, Confocal fluorescence microscopy analysis of the bimolecular fluorescence complementation signal resulting from the interaction between V1–FAM134B and V2–FAM134B. Fixed cells expressing V1–FAM134B and V2–FAM134B were stained with anti-Ub(FK2) (red) or anti-LC3B (grey). Arrows indicate triple colocalization of the Venus signal (green), ubiquitin (red) and LC3B (grey). Scale bar, 10 µm. c, Histogram analysis of the fluorescence intensity distribution reveals that dimeric FAM134B colocalized into Ub⁺ and LC3B⁺ vesicles. d, Single-sided volcano plot of the quantitative label-free interactome of FAM134B homodimers depicting RHD-containing ER proteins (blue), autophagy-related proteins (green) and ubiquitination machinery (red). Data represent three independent experiments, one-sided unpaired Student’s t-test. e, Heatmap comparing the interaction of FAM134B WT homodimers versus FAM134 17KR homodimers with RHD-containing ER proteins, autophagy-related proteins and ubiquitination machinery. Interaction partners with log₂ enrichment > 2.0 and –log₁₀ P > 1.3 were plotted and compared (one-sided unpaired Student’s t-test). Source data

Fig. 5. Effect of AMFR on FAM134B RHD ubiquitination and ER-phagy.

a, HeLa cells were transfected with siNT or siAMFR, and were treated with 250 nM Torin 1 for the indicated time. Protein extracts were analysed by western blot for FAM134B, AMFR or vinculin. b, Densitometric quantification of the western blot signal of FAM134B in a (data are mean ± s.d.; n = 3 independent experiments; two-way ANOVA, Bonferroni post-hoc test). c, Ubiquitination of SFB-tagged FAM134B in AMFR-knockdown cells. The diGly peptide intensities were normalized to total intensities of modified and non-modified FAM134B peptides (data are mean ± s.d.; n = 3 independent experiments; two-way ANOVA, Bonferroni post-hoc test). d, U2OS TRex stable cell lines expressing mCherry–GFP–FAM134B WT were transfected with siNT, AMFR-targeting siRNA #1 or siRNA #2. Cells were treated as indicated. The flux of ER-phagy was quantified as the ratio between mCherry⁺GFP^– and mCherry⁺GFP⁺ puncta. The data are representative of three independent experiments in which the total number of cells per condition were: 699 siNT/DMSO, 683 siNT/BafA1, 695 siNT/Torin 1, 678 siNT/Torin 1 + BafA1, 627 siAMFR #1/DMSO, 668 siAMFR #1/BafA1, 713 siAMFR #1/Torin 1, 591 siAMFR #1/Torin 1 + BAfA1, 624 siAMFR #2/DMSO, 593 siAMFR #2/BafA1, 575 siAMFR #2/Torin 1 and 615 siAMFR #2/Torin 1 + BafA1. Data are mean ± s.d., one-way ANOVA, Bonferroni post-hoc test. e, Diameters of freeze-fractured liposomes incubated with either non-ubiquitinated or ubiquitinated GST–FAM134B. Data are mean ± s.e.m. normalized to the mean liposome diameter of GST control, representing three independent liposome preparations and experiments (n = 1,064 for the non-Ub sample, n = 793 for the Ub sample and n = 434 for the GST control; Kruskal–Wallis/Dunn’s post-test). f, Model in which the E3 ligase AMFR is recruited to ER-phagy receptor clusters to induce the ubiquitination of FAM134B. This event triggers changes in the conformation and composition of ER-phagy receptor clusters, enabling the clusters to grow in size, thus controlling ER remodelling and ER-phagy. mATG8s, mammalian ATG8 proteins. The schematic in panel f was created using BioRender (https://biorender.com). Source data

Extended Data Fig. 1. Constitutive and inducible FAM134B-RHD ubiquitination regulates ER-phagy.

a, TUBE-2 pulldown assay showing increased ubiquitination of endogenous FAM134B following Torin 1 treatment. Cells were treated with DMSO (control), 200 nM BafA1 for 6 h, 250 nM Torin 1 for 6 h, or a combination of 250 nM Torin 1 and 200 nM BafA1 for 6 h. Protein samples were analysed by SDS-PAGE and western blotting, as indicated. b, Densitometric quantification of ubiquitinated FAM134B (FAM134B-Ub) in panel a (Data are mean ± s.d.; n = 3 independent experiments, one-way ANOVA, Bonferroni post-hoc test). c, Cells were treated with DMSO (control), 200 nM BafA1 for 6 h or 10 µM MG132 for 6 h. Detergent-soluble extracts were analysed by western blot with antibodies against FAM134B and UbP4D1. The panels show representative immunoblots. d, Cycloheximide (50 µg/ml) chase for 0–6 h in HeLa cells without or with 200 nM BafA1. Detergent-soluble extracts were analysed by western blot with antibodies against FAM134B and vinculin. e, Densitometric quantification of FAM134B (normalised to vinculin) in panel d (Data are mean ± s.d.; n = 3 independent experiments, one-way ANOVA, Bonferroni post-hoc test). f, Cells were treated with DMSO (control) or 10 µM TAK243 for 1, 2 and 4 h before TUBE-2 pulldown assays. Endogenous ubiquitination of FAM134B was detected by western blot (n = 1 experiment). g, Cycloheximide (50 µg/ml) chase for 0–6 h in HeLa cells with or without 10 µM TAK243. Detergent-soluble extracts were analysed by western blot with antibodies against FAM134B, UbP4D1 and vinculin. h, Densitometric quantification of FAM134B (normalised to vinculin) in panel g (Data are mean ± s.d.; n = 3 independent experiments; One-way ANOVA, Tuckey’s post-hoc test). i, Ubiquitination assay in cells showing FAM134B-RHD ubiquitination despite the replacement of eight conserved lysine residues with arginine (myc-Ub-IP). LC3B-II binding to FAM134B and the formation of high-molecular-weight species (oligomers, SDS-resistant) of FAM134B were not affected by 8KR (Flag FAM134B-IP) (n = 1 experiment). j, Co-localisation of HA-FAM134B/LC3B⁺ puncta per cell in cells expressing FAM134B-WT or 8KR. Number of cells per condition for HA-FAM134B WT: 2039 (DMSO), 2366 (BafA1), 1515 (EBSS), 2280 (Torin1), 1987 (Torin1+BafA1). Number of cells per condition for HA-FAM134B-8KR: 2416 (DMSO), 2494 (BafA1), 1753 (EBSS), 2652 (Torin1), 2580 (Torin1+BafA1). Bars represent means ± s.d. (n = 6 independent experiments; One-way ANOVA, Bonferroni post-hoc test). k, Isolated membranes of cells expressing HA-FAM134B WT or 17KR were treated with 0.3 mM EGS and analysed by SDS-PAGE. Oligomers were visualised by western blot. (n = 1 experiment). l, U2OS cells stably expressing HA-FAM134B-WT under the control of a doxycycline-inducible promoter were treated with DMSO (control) or 200 nM BafA1 for 6 h. m, HeLa cells stably expressing HA-FAM134B-WT or the LIR mutant under the control of a doxycycline inducible promoter were treated with DMSO (control) or 200 nM BafA1 for 6 h. (n = 1 experiment). n, TUBE-2 pulldown assay showing the accumulation of endogenous ubiquitinated HA-FAM134B LIR-mutant. The panels show representative immunoblots. Source data

Extended Data Fig. 2. Ubiquitin-membrane and ubiquitin-RHD interactions.

a, Time series of the number of contacts between CG-beads of Ubiquitin and POPC bilayers in MD simulations of mono-ubiquitinated RHDs. b, Snapshots from MD simulations of K160-Ub (left) and K264-Ub (right) showing the structures of the Ub moieties (in cyan and blue), and their interactions with the membrane (orange beads). Modified lysines (red) show the relative orientations of the Ub-moieties with respect to the amphipathic helices (yellow) and the cytosolic portion of the RHD. c, Time series of the number of contacts between CG beads of RHD and covalently linked Ub moieties. d, Residue-wise contact maps between Ub and RHDs. The solid red lines indicate the position of the lysine residue on the RHD linked to Ub. e, Time series of the number of contacts between CG beads of Ubiquitin and POPC bilayers in MD simulations of bi-mono-ubiquitinated RHD, (K160+K264)-Ub. f, Snapshot from MD simulation of (K160 + K264)-Ub variant showing how bi-mono-ubiquitination of the RHD reduces the interactions between ubiquitin and POPC bilayer. The two Ub-moieties are involved in intra-molecular/cis interactions on top of the cytosolic face of the RHD. g,h Time series (g) and probability distribution (h) of the radii of gyration, R_g, sampled by ubiquitinated and non-ubiquitinated FAM134B-RHDs during the 10-μs MD simulations.

Extended Data Fig. 3. Ubiquitinated RHDs induce membrane curvature.

a–d, Time series of mean curvature (H) of bicelle containaing single FAM134B RHD. One hundred MD simulations for each system were initiated with different initial velocities to study curvature induction (positive/negative) and transition of the bicelles into closed vesicles (H = +0.14 nm⁻¹) at 300 K. e, Violin plots of estimated waiting times required for bicelle-to-vesicle transitions induced by WT and Ub-variants of FAM134B-RHD. n = 92 (FAM134B-RHD), 100 (K160-Ub), 99 (K264-Ub), 100 (K160+K264-Ub) runs f, Comparison of mean ± SEM curvature time traces (black line ± shaded region) for each system shows that Ub-RHDs induce bicelles to curve swiftly, resulting in faster kinetics. g, Table showing the kinetics of the in silico curvature induction process. The numbers n₊ and n₋ indicate bicelle transitions resulting in bilayer curvature away from and towards the upper/cytoplasmic leaflet. Observed waiting times (t) for vesicle formation were recorded as the time taken for the bilayer or bicelle discs to reach a curvature of |H| = 0.14 nm⁻¹. The vesicle formation rate k_sys = 1/(t′ + τ) was estimated from exponential fits of the cumulative distribution functions of Poisson-distributed waiting times (t′) and a constant lag time (τ). The acceleration in vesicle formation due to a protein inclusion was estimated as the ratio k_sys/k_WT. h–k, Time series of bicelle mean curvature (H) for 20 simulations initiated at 280 K for each system. Fewer bicelles with embedded K160-Ub and WT-RHD proteins (4/20 and 5/20) transitioned into closed vesicles at 280 K.

Extended Data Fig. 4. Curvature preference of ubiquitinated FAM134B-RHDs.

a,b, Selected snapshots from 20-μs MD simulations of single K160-Ub and K264-Ub variants on buckled POPC bilayers. The Ub-variants sample regions of high mean curvature and preferentially occupy the top of the buckle. c, Histograms of the mean curvature H(x, y) sampled by FAM134B RHD (green), K160-Ub (cyan) and K264-Ub (steel blue) in coarse-grained simulations (1-ns intervals for 20 μs) indicate a preference for highly curved regions of the buckle. For reference, the distribution of local mean curvature values on the empty buckled membrane (red) was estimated by random sampling of points in the xy plane, ignoring small curvature corrections. The time-averaged values of H(x, y) for each system are also provided. d, Confocal fluorescence microscopy analysis of cells transiently expressing HA-tagged FAM134B RHD_90–264 and the chimaera Ub-RHD_90–264-Ub. Fixed cells were stained with anti-HA (green) and anti-REEP5 (red). Scale bar = 10µm. e, Co-localisation analysis of RHD/REEP5+ puncta per cell in cells expressing FAM134B RHD_90-264 or Ub-RHD_90-264-Ub (bar plots of data are presented, statistical significance was determined using a non-parametric two-tailed Mann-Whitney U-test, n = 10 cells per condition) Bars represent means ± s.d. f, Ubiquitin moieties tethered to different RHDs trigger inter-molecular Ub-Ub interactions, which induce RHD clustering and dimerisation on top of the buckle (left) after ~18 μs. (right) Time series of inter-molecular Ub-Ub contacts across the two molecules. The inset shows the average Ub-Ub contact map stabilising the dimer structure (21–25 μs). g, Snapshots of membrane tubule with 10 Ub-RHDs (left) Snapshots of Ub-RHD cluster formation that locally deform membrane tubules from ideal cylindrical geometry. Coloured squares (red, blue and green) highlight the number of molecules (dimers/trimers) that form the clusters locally on different parts of the tubule. Ubiquitinated RHD clusters shape the tubule (right). Zoomed RHD clusters show Ub-Ub interactions (blue moieties). Source data

Extended Data Fig. 5. Ub-Ub interactions from MD simulations of ubiquitinated RHDs.

9 (K160+K264)-Ub-RHD molecules were simulated in a 3x3 square grid embedded in POPC bilayer under 4 different simulation conditions (ΔN = 0 and 300, and α = 1 and 0.65) to quantify the Ub-Ub interactions (schematic at the top). a-d, The two Ub moieties of (K160+K264)-Ub-RHD molecules are engaged in both cis/intra-molecular (left) and trans/inter-molecular interactions (right). Interactions under each simulation condition were quantified by computing an average Ub-Ub contact map, ${\mathrm{UU}}_{\mathrm{cnts}}=\left[{\sum }_{{i\in \mathit{Ub}}_1}{\sum }_{{j\in \mathit{Ub}}_2}\sigma \left(|r_{\mathit{ij}}|\right)\right]$, where the sums extend over residue COM positions of interacting Ub moieties and $\sigma \left(|r_{\mathit{ij}}|\right)$ is a smooth sigmoidal counting function to limit interactions below the cut-off (r_ij ≤ 10 Å). Interaction maps were averaged over only the bound states ($\sum {\mathrm{UU}}_{\mathrm{cnts}}\ge 5$) for all nine intra-molecular Ub-Ub pairs (left) and 144 inter-molecular Ub-Ub pairs (right), respectively. e, Box plots showing the distribution of lifetimes of trans/inter-molecular Ub-Ub interactions from each of the four simulation conditions. The lifetime for each trans-Ub-Ub bound state was estimated as a contiguous stretch (in ns) where (($\sum {\mathrm{UU}}_{\mathrm{cnts}}\ge 5$) for 10 ns). Lifetimes from a total of 216, 276, 143 and 304 trans-Ub-Ub bound states were sampled over the course of 5-µs runs for each of the four simulation conditions, respectively.

Extended Data Fig. 6. Map of Ub-Ub interactions.

a, (Top) Schematic showing the cis/intra-molecular Ub-Ub interactions and b, the trans/inter-molecular Ub-Ub interactions mapped onto the 3D structure of Ub. Strongly interacting (red), moderately interacting (white), and weakly interacting (blue) sites are coloured, and the cartoon backbone size is scaled accordingly. (Bottom) The same interactions are mapped along the sequence of Ub to highlight the various secondary structural elements and residues involved, revealing that hairpins β12 and β34, along with the C-terminal region of Ub, dominate the intra-molecular interactions of K160-Ub and K264-Ub. However, trans-Ub-Ub interactions are spread throughout the Ub sequence, indicating their non-specific nature.

Extended Data Fig. 7. Effect of RHD ubiquitination on the flux of ER-phagy and FAM134B cluster size.

a, Representative confocal images of HA-FAM134B WT and 17KR (red) co-stained with LC3B (green). b, Co-localisation of HA-FAM134B/LC3B⁺ puncta and the c, corresponding area (µm²) were lower for 17KR compared to WT. The data are representative of three independent experiments in which the total number of cells per condition were n_cells = 488 WT, 392 17KR. (Data are mean ± s.d.; two-tailed unpaired Student’s t-test). d, Representative negative-stain transmission electron microscopy (TEM) images of remodelled proteoliposomes (scale bars = 200 nm). Empty liposomes were incubated with purified GST empty, GST-FAM134B-WT or GST-FAM134B-17KR for 18 h at 25 °C. Images show examples of representative proteoliposomes. e, Violins shows the box-plots with median value, white dot, interquartile range (black shaded region), min and max values (1.5 x interquartile region) and mirrored probability density estimates on sides. (WTmean = 28.25; 17KR mean = 27.73; GSTmean = 128.78), GST empty (n = 167), GST-FAM134B-WT (n = 277) or GST-FAM134B-17KR (n = 297); Kruskal-Wallis/Dunn’s post-hoc test. f, Quantitative TEM analyses of diameters of freeze-fractured liposomes that were incubated with either GST control, GSTFAM134B WT or GSTFAM134B17KR. Data, mean ± SEM presented as normalized to the mean liposome diameter of GST control. Two independent liposome preparations and experiments. Kruskal-Wallis/Dunn’s post-test. (n = 310 for WT, n = 250 for 17KR and n = 208 for GST control). g, Schematic representation of the ER-phagy reporter system RFP-GFP-KDEL. h, Representative confocal images of U2OS TRex stable cell lines co-expressing RFP-GFP-KDEL with either HA-FAM134B WT or HA FAM134B 17KR. Cells were treated for 16 h with 1 µg/ml doxycycline to induce the expression of both proteins. Cells were fixed, permeabilised, and stained for HA and LC3B. Bar = 10 µm. i, ER-phagy flux was quantified as the ratio between RFP⁺/GFP^– and RFP⁺/GFP⁺ puncta, quantified using CQ1 software. Cells were treated with DMSO (control), 200 nM BafA1 for 6 h, EBSS for 6 h, 250 nM Torin 1 for 6 h or 250 nM Torin 1 plus 200 nM BafA1 for 6 h. Data are means ± s.d. of n = 6 independent experiments, in which the number of RFP-GFP-KDEL cells per condition are: 2366 (DMSO), 2202 (BafA1), 1460 (EBSS), 2228 (Torin1), 2378 (Torin1+BafA1). Number of RFP-GFP-KDEL/HA-FAM134B WT cells: 670 (DMSO), 773 (BafA1), 578 (EBSS), 986 (Torin1), 747 (Torin1+BAfA1), Number of HA-FAM134B 17KR cells: 1160 (DMSO), 1151 (BafA1), 1313 (EBSS), 1353 (Torin1), 1295 (Torin1+BafA1). One-way ANOVA, Bonferroni post-hoc test. j, RFP-GFP-KDEL, RFP-GFP-KDEL/HA-FAM134B WT and RFP-GFP-KDEL/HA-FAM134B 17KR cells were left untreated or treated with EBSS for 8 h. Detergent-soluble extracts were analysed by western blot using antibodies against RFP, HA, REEP5 and β-actin. (n = 1 experiment). k, The E1 inhibitor decreases the flux of ER-phagy in mCherry-GFP-FAM134B-WT cells induced with Torin 1 (Data are mean ± s.d.; one-way ANOVA, Bonferroni post-hoc test). ER-phagy flux was quantified as the ratio between mCherry⁺/GFP^– and mCherry⁺/GFP⁺ puncta, quantified using CQ1 software. Data are means ± s.d. of n = 5 independent experiments in which the number of cells per condition were: (DMSO) 837 basal, 1072 BafA1, 1038 Torin1, 966 BafA1+Torin1. Number of cells (10 µM TAK243): 729 basal, 1174 BafA1, 1060 Torin1, 1121 Torin1+BafA1. l, Two-colour DNA-PAINT super-resolution image of HA-FAM134B (magenta, R2-ATTO655) and the autophagosomal membrane marker LC3B-II (green, R1-ATTO655) (i). White box indicates the magnified region shown in (ii). (iii) Point localisations of HA-FAM134B from the magnified region shown in (ii) and corresponding Voronoi diagrams (blue polygons) with red line representing FAM134B cluster contour (iv). Clusters are identified based on previously determined thresholds (density factor, minimum number of localisations and minimal distance parameter). Scale bars = 10 µm (i) and 1 µm (ii–iv). m, Box plot of HA-FAM134B-WT and HA-FAM134B-17KR nanoscale and microscale cluster sizes. For nanoscale clusters, ubiquitination-deficient FAM134B significantly reduces the cluster diameter (35–202 nm, median = 85 nm) compared to its WT counterpart (33–286 nm, median = 114 nm). Nanoscale cluster (n_cells = 4, n_WTclusters = 1278; n_17KRclusters = 1255). For larger microscale clusters, significantly larger areas were detected for HA-FAM134B-WT (0.017–0.20 µm², median = 0.08 µm²) compared to the 17KR mutant (0.01–0.21 µm², median = 0.03 µm²). Torin 1 treatment further increased the HA-FAM134B cluster area with the effect being stronger for ubiquitinated HA-FAM134B-RHD (median_WT = 0.23 µm², median_17KR = 0.16 µm²). Quantitative analysis of nanoscale clusters was carried out using the DBSCAN algorithm and microscale clusters were identified using SR-tessellation. Box-plots of FAM134B wildtype (magenta, grey dots) and FAM134B 17 KR (green, grey squares) nanoscale cluster diameters (left panel) and microscale cluster areas (right panel, grey dots) showing median values (horizontal lines in boxes), the interquartile ranges (width of the boxes) and whiskers defining minimum and maximum values (excluding outliers). A non-parametric one-tailed Mann-Whitney U-test was applied to the data. n, HA-FAM134B forms nanoscale clusters within the ER network. Two-colour super-resolution image of HA-FAM134B-WT (magenta, R2-ATTO655) and ER-membrane marker REEP5 (green, R1-ATTO655), with (ii) showing the magnified region from box (i). Scale bars = 10 µm (i) and 1 µm (ii). Source data

Extended Data Fig. 8. Protein interactors of FAM134B/FAM134C-containing oligomers.

a, HEK 293T cells were transfected with a control plasmid (GFP), V1-FAM134B-WT, V2-FAM134B-WT, V1-FAM134C-WT or V2-FAM134C-WT. Cells were also co-transfected with V1-FAM134B-WT and V2-FAM134B-WT or V1-FAM134C-WT and V2-FAM134C-WT. The GFP trap was analysed by western blot. (n = 1 experiment). b, Ubiquitinated lysine residues identified by proteomics analysis in the RHD of immuno-isolated dimeric FAM134B: diGly peptides were significantly enriched (log₂ enrichment > 2.0 and –log₁₀ p value > 1.3, one-tailed unpaired Student’s test). n = 3 independent experiments. Schematic of the FAM134B-RHD showing the localisation of ubiquitinated lysine residues. c, Single-sided volcano plot of the quantitative label-free interactome of FAM134C homodimers and d, FAM134B/FAM134C heterodimers depicting identified RHD-containing ER proteins (blue), autophagy-related proteins (green), and components of ubiquitination machinery (red) (log₂ enrichment > 2.0 and –log₁₀ p value > 1.3). Data are means ± s.d. of n = 3 independent experiments. e, Heat map comparing the interaction of RHD-containing ER proteins, autophagy-related proteins and the ubiquitination machinery with WT FAM134B homodimers, FAM134C homodimers and FAM134B/FAM134C heterodimers (immuno-isolated using BiCAP). Interaction partners with log₂ enrichment > 2.0 and –log₁₀ p value > 1.3 were plotted. n = 3 independent experiments, one-tailed unpaired Student’s test. f, Venn diagram of interactors of FAM134B homodimers, FAM134C homodimers and FAM134B /FAM134C heterodimers. Numbers represent significantly enriched interaction partners (log₂ enrichment > 2.0 and –log₁₀ p > 1.3, one-tailed unpaired Student’s test). n = 3 independent experiments. g, Annotation enrichment analysis of the interactome of FAM134B and FAM134C heterodimers. Bars represent significantly enriched gene ontology biological process (GOBP), gene ontology cellular component (GOCC), gene ontology molecular function (GOMF), and domain enrichment (Pfam). h, Confocal fluorescence microscopy analysis of the BiFC signal produced by interactions between V1-FAM134B and V2-FAM134B. Fixed cells expressing V1-FAM134B and V2-FAM134B were stained for LC3B (red) and p62 (blue). Scale bar = 10 µm. i, Pearson’s correlation coefficients obtained from the co-localisation analysis of fluorescent signals representing FAM134B clusters and p62 or LC3B. Data are means ± s.d. of n = 10 cells per analysis. Source data

Extended Data Fig. 9. Analysis of the functional interaction between AMFR and FAM134B in cells.

a, Confocal fluorescence microscopy analysis of the BiFC signals following the interaction between V1-FAM134B and V2-FAM134B. Fixed cells expressing V1-FAM134B and V2-FAM134B were stained for AMFR (red) and LC3B (blue). Scale bar = 10 µm. b, The fluorescence intensity distribution reveals that FAM134B clusters (V1FAM134B+V2FAM134B) co-localise with endogenous AMFR and LC3B in punctate structures. c, Ubiquitination assay of HA-FAM134B in cells co-expressing WT-AMFR-Flag or the catalytically inactive AMFR-Flag (C356G H361A) variant. d, Densitometric quantification of western blot signals for ubiquitinated HA-FAM134B following co-expression with WT-AMFR-Flag or the catalytically inactive AMFR-Flag (C356G H361A) variant as presented in Fig. 9c. Data are means ± s.d. of n = 3 independent experiments, two-tailed unpaired Student’s t test. e, Schematic of the BiCAP method to study the functional interaction between FAM134B and AMFR. Full-length FAM134B was fused to the C-terminal of the non-fluorescent N-terminal (V1) fragment of Venus protein (V1-FAM134B), whereas full-length AMFR was fused to the N-terminal of the non-fluorescent C-terminal (V2) fragment (AMFR-V2). f, Confocal microscopy analysis of fixed cells co-expressing V1-FAM134BWT and AMFR-V2 WT or V1-FAM134BWT and AMFR-V2 C356G H361A stained for Ub and LC3B (magenta). Scale bar = 10 µm. g, Pearson’s correlation coefficients obtained from co-localisation of the fluorescent signals representing refolded Venus and Ub(FK2) in cells co-expressing V1-FAM134BWT and AMFR-V2 WT or V1-FAM134BWT and AMFR-V2 C356G H361A (Extended Data Fig. 9e), (Two-tailed non-parametric Mann-Whitney U-test was applied to the data) Data are means ± s.d. n_{V1FAM134B/V2AMFRWT} = 68 puncta and n_{V1FAM134B/V2AMFR C356G H361A} = 52 puncta. h, Single-sided volcano plot of the quantitative label-free interactome of the affinity-purified (using BiCAP) WT-AMFR-V2/V1-FAM134BWT complex. RHD-containing ER proteins (blue), autophagy-related proteins (green), and ubiquitination machinery (red) (log₂ enrichment > 2.0 and –log₁₀ value > 1.3. Data are means of n = 3 independent experiments; one-tailed unpaired Student’s test. i, Venn diagram of interactors of FAM134B homodimers and AMFR/FAM134B heterodimers. Numbers represent significantly enriched peptides (log₂ enrichment > 2.0 and –log₁₀ p value > 1.3. Data are means of n = 3 independent experiments, one-tailed unpaired Student’s test. j, Comparison of Torin 1-induced ubiquitination of FAM134B within complexes with either AMFR-WT (WT-AMFR-V2/V1-FAM134B) or the catalytically inactive AMFR mutant (C356G H361A AMFR-V2/V1-FAM134B). Graphs show the abundance of FAM134B peptides carrying a diGly modification expressed as intensities. Data are means ± s.d. of n = 3 independent experiments and the identified diGly peptides intensities were normalised to the total intensities of modified and non-modified peptides (two-tailed unpaired Student’s test). k, Heat map comparing the interaction of RHD-containing ER proteins, autophagy-related proteins and the ubiquitination machinery of WT-AMFR-V2/V1-FAM134B and AMFR C356G H361A-V2/V1-FAM134B complexes. Interaction partners with log₂ enrichment > 2.0 and –log₁₀ p value > 1.3 were plotted. Data are means of n = 3 independent experiments, one-tailed unpaired Student’s test. Source data

Extended Data Fig. 10. Analysis of the functional interaction between AMFR and FAM134B in cells and in vitro.

a, HeLa cells were transfected with control siRNA (siNT), siRNA#1 or siRNA#2 targeting AMFR (siAMFR) for 72 h. Detergent-soluble protein extracts were analysed by western blot using antibodies against FAM134B, AMFR or vinculin (loading control). b, Densitometric quantification of endogenous FAM134B (normalised to vinculin) in panel a (Data are means ± s.d. of n = 5 independent experiments, One-way ANOVA, Bonferroni post-hoc test). c, Confocal fluorescence microscopy of U2OS cells stably expressing HA-FAM134BWT transfected with either control siRNA (siNT) or siRNA#1 targeting AMFR (siAMFR) for 72 h, followed by incubation with 250 nM Torin 1 for 6 h. Cells were fixed and stained for HA-FAM134B and endogenous LC3B, respectively. d, Quantification of HA-FAM134B-WT/LC3B-II-containing puncta per cell of images in panel c. Scatter plot graphs represent means ± s.d. (n_siNT = 24 cells, n_siAMFR = 33 cells; two-tailed Mann-Whitney-U-test). e, HeLa cells were treated with 250 nM Torin 1 for the indicated time (h). Densitometric quantification of endogenous AMFR (normalised to vinculin), as presented in Fig. 5a. Data are means ± s.d. of n = 3 independent experiments; two-way ANOVA, Bonferroni post-hoc test. f, U2OS cells stably expressing HA-FAM134BWT or HA-FAM134B-17KR were incubated with 250 nM Torin 1 for 0, 2, 4, 6 and 8 h. Detergent-soluble extracts were analysed by western blot using antibodies against HA, AMFR and vinculin. g,h, Densitometric quantification of HA-FAM134B and AMFR (normalised to vinculin) from panel f. Data are means ± s.d. of n = 3 independent experiments, two-way ANOVA, Bonferroni post-hoc test. i, U2OS cells stably expressing HA-FAM134BWT or HA-FAM134B-LIR incubated with 250 nM Torin 1 for 0, 2, 4, 6 and 8 h. Detergent-soluble extracts were analysed by western blot using antibodies against HA, AMFR and vinculin. j, Densitometric quantification of AMFR (normalised to vinculin) in panel i. Data are means ± s.d. of n = 3 independent experiments, two-way ANOVA, Bonferroni post-hoc test. k, and l, Purification of AMFR from HEK 293T cells. m,n, Mass spectrometry (MS) analysis of the in vitro ubiquitination of full length GST-tagged FAM134B using recombinant AMFR. (Data are means ± s.d. of n = 3 independent experiments; two-tailed unpaired Student’s t-test). o–q, MS analysis of the in vitro ubiquitination of His-RHD_90–264-Strept-II and His-Ub-RHD_90–264-Ub-Strept-II using recombinant AMFR. Data are means ± s.d. of n = 3 independent experiments; two-tailed unpaired Student’s t-test. r, Immunodetection of native His-RHD_90–264-Strept-II following ubiquitination by AMFR. The ubiquitination reaction was analysed by western blot after blue native polyacrylamide gel electrophoresis (BN-PAGE) using antibodies against His₆ or UbP4D1. Control reaction (in the presence of AMFR, no ATP). Source data