Main

In a previous study2, we identified FAM134B loss-of-function mutations in patients with autosomal recessive hereditary sensory and autonomic neuropathy (HSAN)2. This disorder is characterized by the degeneration of sensory and autonomic neurons that leads to numbness and the inability to feel pain. These symptoms in turn cause severe injuries and tissue damage2. Our studies further showed that FAM134B is an ER-resident membrane-shaping protein that can bind LC3 proteins and mediate the engulfment of parts of ER sheets by autophagosomes and their subsequent lysosomal degradation1. The neurodegeneration that occurs in humans is also observed in Fam134b-deficient mice1. Together with the consequences observed following the disruption of the functional counterpart of FAM134B in yeast (Atg40)4, these results suggest that the role of ER-phagy in cell viability is evolutionarily conserved.

Within the past decade, several other ER-resident membrane-shaping proteins with central reticulon homology domains (RHDs) have been associated with similar neurodegenerative disorders, including ATL1, ATL3, REEP1, REEP2, SPAST, RTN2, ARL6IP1 and LNPK5,6,7. Mutations in ARL6IP1 cause SPG61, a neurodegenerative disorder characterized by progressive leg spasticity (hereditary spastic paraplegia (HSP)) in combination with loss of sensory and pain perception, thus overlapping with typical symptoms of HSAN2,3,8. The underlying mechanisms, however, remained largely elusive. Here we show that membrane-embedded clusters of ubiquitinated ARL6IP1 and FAM134B are required for effective ER remodelling and ER-phagy, defects of which result in severe neurodegeneration.

Degeneration of neurons in Arl6ip1 knockout mice

To resolve the pathophysiology of the ARL6IP1-related disorder, we studied fibroblasts obtained from a patient with SPG61. This patient harboured the homozygous carboxy-terminal frameshift mutation ARL6IP1 c.577–580delAAAC (NCBI Nucleotide database identifier NM_015161.3; K193Ffs variant) (Fig. 1a). Fibroblasts obtained from the patient’s father, who was unaffected and a heterozygous carrier, and from an unrelated healthy individual (as a control) were also analysed. In silico analysis indicated that this frameshift mutation is predicted to result in the replacement of the 11 amino acids of the C terminus by 36 alternative residues3. Suggesting nonsense-mediated decay in vivo, ARL6IP1 transcripts were absent in the patient’s cells, as assessed by real-time PCR with exon spanning primers annealing 5′ end of the deletion (Fig. 1b). Immunoblot analyses of lysates from cells transfected with a plasmid encoding the K193Ffs variant produced a band with a slightly higher molecular weight (Fig. 1c). No variant protein was detected in fibroblasts obtained from the patient’s father or from the patient when using an antibody directed against the cytoplasmic loop of ARL6IP, a result that is in agreement with nonsense-mediated decay (Fig. 1c). Because the ARL6IP1 c.577–580delAAAC variant represents a knockout (KO) allele, we generated Arl6ip1 KO mice to model the ARL6IP1-associated disorder (Fig. 1d). KO of the gene was confirmed by both quantitative PCR (qPCR) of RNA isolated from mouse embryonic fibroblasts (MEFs) (Fig. 1e) and immunoblot analyses of MEFs and tissue lysates (Fig. 1c,f). Compared with wild-type (WT) mice, KO animals did not gain appropriate body weight (Extended Data Fig. 1a). Consistent with the CNS-related phenotypes reported for some patients8,9, the weight of the brain in Arl6ip1 KO mice was significantly decreased (0.46 g in WT mice compared with 0.39 g in KO mice at 6 months of age; n = 3, unpaired Student’s t-test, P = 0.016). Moreover, cortical neuron (Extended Data Fig. 1b) and Purkinje cell (Extended Data Fig. 1c) counts were reduced. As a correlate of the HSP-related gait disorder, we measured the foot base angle of the hind paw at the moment when the toe was lifted10 and found that it decreased with age in KO mice (Fig. 1g). Moreover, the grip strength of the upper limbs was reduced in KO mice (Fig. 1h). As reported for deceased patients with HSP, we observed that some axons connecting cortical and spinal cord motor neurons were swollen and full of dysfunctional organelles (Extended Data Fig. 1d). Because some patients develop muscle hypotonia and weakness, we also quantified spinal cord motor neurons, which progressively decreased in KO mice (Extended Data Fig. 1e). In agreement, the electrophysiological analysis showed a severe reduction in compound muscle action potential amplitudes (Fig. 1i). Consistent with neurogenic muscle atrophy, we found reduced musculus gastrocnemius mass (Extended Data Fig. 1f), degenerating grouped skeletal muscle fibres (Extended Data Fig. 1g) as well as degenerating intramuscular nerve fibres (Fig. 1j) and motor end plates (Fig. 1k and Extended Data Fig. 1h) in KO animals. Immunoblot analyses of brain protein lysates showed that the abundance of some ER-resident proteins with RHDs was altered in KO mice (Extended Data Fig. 1i). Sensory fibres were also degenerated, as evidenced by the substantial loss of sensory amplitudes in electrophysiological analyses of peripheral nerves (Fig. 1l). This result correlates with sensory loss and loss of pain perception in patients3,8. The ultrastructural analysis of peripheral nerves also showed swollen axons full of dysfunctional organelles and tubulofilamentous material (Fig. 1m), and ladder-like expansions of transverse ER sheet structures (Fig. 1n). This result is similar to that reported previously for mice mutant for both ALT1 and REEP1 (ref. 11). An analysis of cell bodies of peripheral sensory neurons (dorsal root ganglia (DRG)) uncovered a substantial expansion of ER sheets in Arl6ip1 KO mice (Fig. 1o). Because of these observations and similar phenotypes in patients, we propose that ARL6IP1 plays a part in FAM134B-dependent ER-phagy.

Fig. 1: Neurodegeneration with ER sheet expansion in Arl6ip1 KO mice.
figure 1

a, Schematic of disease-associated ARL6IP1 variants. b, No ARL6IP1 transcripts were detected in fibroblasts obtained from a patient carrying a homozygous mutation of ARL6IP1K193Ffs (qPCR with three replicates, two-sided unpaired Student’s t-test, P = 0.0001). c, No variant protein was detected in cells from the patient (two experiments). d, Arl6ip1 KO strategy. Frt sites, white triangles; loxP sites, black triangles. Predicted 14 amino acid product for the KO allele. e, No Arl6ip1 transcripts were detected in KO MEFs (qPCR with three replicates, two-sided unpaired Student’s t-test, P = 0.0007). f, Absence of ARL6IP1 in KO tissue lysates (two experiments). g, Decreasing foot base angle in KO mice (n = 6 WT mice and n = 5 KO mice; two-sided unpaired Student’s t-test, 2 month-old WT versus KO, P = 0.0007; 8-month-old WT versus KO, P = 0.0016; KO 2 months versus 8 months, P = 0.038). h, Diminished forelimb grip strength in 2-month-old (P = 0.0004) and 8-month-old (P = 0.0001) KO mice (n = 6 WT and KO mice each; two-sided unpaired Student’s t-test). i, Decreased compound muscle action potentials (CMAPs) in KO mice (n = 6 WT mice and n = 7 KO mice; repeated-measures analysis of variance (ANOVA), F = 18.6, P = 0.0015). j, TEM of intact WT and degenerating KO intramuscular nerve fibre (indicated by the asterisk). k, TEM of intact WT and degenerating KO motor end plate. l, Decreased sensory amplitudes in KO mice (n = 6 WT mice and n = 7 KO mice; repeated-measures ANOVA, F = 6.08, P = 0.0314). m, TEM of transversely cut intact WT and degenerating KO sciatic nerve axon. n, TEM of longitudinally cut sciatic nerve axons with ladder-like transverse ER sheet expansions (arrowheads) in KO but not WT mice. For in, analyses were performed using 6-month-old mice. o, TEM images (left) and quantification (right) show ER sheet expansions in lumbar spinal ganglion neurons in 12-month-old KO mice. ER sheet areas are coloured and higher magnifications are indicated. n = 100 cells from n = 3 WT mice and n = 177 cells from n = 3 KO mice; two-sided unpaired Student’s t-test, P = 0.0006. Data shown as the mean ± s.e.m. Scale bars, 1 µm (j,m,n), 2 µm (k) or 2.5 µm (o).

Source Data

ARL6IP1 is part of ER-phagy complexes

On the basis of previous data12 and results of fluorescence protease protection assays, we concluded that ARL6IP1 is characterized by the presence of RHD-like structural elements. That is, two long hydrophobic regions (transmembrane helical hairpins TM1+2 and TM3+4) separated by an accessible linker segment with both the amino and C termini facing the cytoplasm (Fig. 2a). The predicted structural model (produced using AlphaFold) of ARL6IP1 contains two membrane-embedded helical hairpins (TM1+2 and TM3+4) with two amphipathic helices (Fig. 2b and Extended Data Fig. 2a–c). Purified ARL6IP1 bound to liposomes in vitro (Fig. 2c) and increased the proportion of smaller liposomes (Fig. 2d), similar to FAM134B1,13,14. Despite predicted putative LC3-interacting regions (LIRs) in the N terminus or the cytoplasmic loop between RHDs, and in contrast to FAM134B, ARL6IP1 did not bind LC3 proteins (Fig. 2e), which suggested that ARL6IP1 is not an ER-phagy receptor on its own. However, ARL6IP1 was detected as an interaction partner in both a yeast two-hybrid screen for FAM134B-binding proteins1 and in a proteomics analysis of FAM134B interactors (Fig. 2f). This result indicates that ARL6IP1 may be indirectly linked to the autophagy machinery through FAM134B. A sequence alignment analysis showed that both proteins are closely related and share all the signature membrane remodelling elements (Extended Data Fig. 2c). The co-precipitation of endogenous FAM134B and ARL6IP1 in MEFs or HEK293T cells (Fig. 2g) or endogenous FAM134B and expressed haemagglutinin (HA)-tagged ARL6IP1 (HA–ARL6IP1) in U2OS cells (Extended Data Fig. 2d) confirmed the interaction between both proteins. In agreement, ARL6IP1–Myc and FAM134B–HA co-localized to the same ER regions in MEFs (Fig. 2h). The co-localization results from MEFs was supported by data from proximity ligation assays using antibodies directed against endogenous FAM134B and ARL6IP1 (Fig. 2i). Using different deletion variants of ARL6IP1 and FAM134B, it became evident that their central parts containing the RHDs are required for the interaction between both proteins (Extended Data Fig. 2e–g). Notably, tagged FAM134A and FAM134C, the two homologues of FAM134B, also co-immunoprecipitated with ARL6IP1 in HEK293T cells (Extended Data Fig. 2e).

Fig. 2: FAM134B interacts with ARL6IP1.
figure 2

a, Left, cytosolic location of the ARL6IP1 C terminus. Middle, COS-7 cells transiently expressing RFP-tagged or GFP-tagged ER proteins subjected to fluorescence protease protection assays by sequential administration of digitonin to permeabilize plasma but not intracellular membranes and then trypsin. Right, C-terminal fluorescent tags of ARL6IP1, FAM134B or CD3 are destroyed by trypsin, whereas luminal RFP–KDEL is protected (three experiments with n = 29 (CD3–RFP), 16 (RFP–KDEL), 15 (FAM134B–GFP) and 22 (ARL6IP1–GFP) cells). b, Three-dimensional model of ARL6IP1 built using AlphaFold2 showing the relative organization of key structural elements (grey and yellow) and their relative orientation in a model bilayer (orange beads). c, Recombinant Trx–His–ARL6IP1 and untagged ARL6IP1 float with lipid membranes in sucrose density gradients to the liposome fraction 2 (immunoblot analysis, n = 2). d, Left, TEM images of freeze-fractured incubations of liposomes with Trx–His or recombinant Trx–His–ARL6IP1. Right, mean liposome diameters are decreased with Trx–His–ARL6IP1 (1 experiment with n = 1,817 (Trx–His; average 380 nm) and 1,824 (Trx–His–ARL6IP1; average 150 nm) vesicles analysed; two-sided Mann–Whitney U-test, P = 0.0001). e, HEK293T cells were transfected with the indicated constructs and immunoblotted (IB). Pull-down with anti-Myc coupled beads shows that LC3-II co-precipitates with FAM134B–Myc but not ARL6IP1–Myc (n = 1). f, The FAM134B interactome in U2OS cells includes ARL6IP1 (single-sided volcano plot). Notable hits with a log2(enrichment factor) > 1 and –log10(P value) > 1.3 are highlighted (one-sided paired Student’s t-test with three biological replicates). g, Co-precipitation of endogenous ARL6IP1 and FAM134B from MEFs and from HEK293T cells (n = 1). IP, immunoprecipitation. h, Overexpressed ARL6IP1–Myc and FAM134B–HA co-localize in MEFs. i, Proximity ligation assays suggest a proximity of less than 40 nm between endogenous ARL6IP1 and FAM134B. Specificity was confirmed by the absence of signals in the respective KO MEFs. Quantitative data are shown as the mean ± s.e.m. Scale bars, 200 nm (d) or 5 µm (h,i).

Source Data

We next used biomolecular complementation affinity purification (BiCAP)15 to study the interactions of ARL6IP1 and FAM134B in vivo (Fig. 3a,c). For this purpose, ARL6IP1 and FAM134B were linked to the V1 or V2 segment of the fluorescent Venus protein. The Venus signals for V1–ARL6IP1 and V2–ARL6IP1 homodimers and for V1–FAM134B and V2–ARL6IP1 heterodimers were distributed along the ER. By contrast, no Venus signal was observed for the non-canonical ER-phagy receptor CCPG1 (ref. 16) or for the ARL6IP1 variant lacking TM1 and TM2 (Extended Data Fig. 3a), which did not co-precipitate with FAM134B (Extended Data Fig. 2f). To characterize the functional relevance of this interaction, we immunoprecipitated homodimers and heterodimers and analysed interacting proteins by liquid chromatography and mass spectrometry (LC–MS)17 (Fig. 3b,d,e). Within all interacting proteins, 7% exclusively interacted with ARL6IP1 homodimers, 52.4% exclusively with FAM134B homodimers and approximately 40% with both (Extended Data Fig. 3b). Among the top ten gene ontology terms identified for interaction partners of ARL6IP1 homodimers were ER structural components (Extended Data Fig. 3c) such as FAM134B and FAM134C as well as the RHD proteins RTN1, RTN3 and RTN4 (Fig. 3b,e and Extended Data Fig. 3c,e). Notably, the main non-neuronal RTN4 isoform RTN4B was previously identified to interact with FAM134C and to have a role in autophagy18. Components of the autophagic vesicle formation machinery such as LC3B and GABARAPL2 were only identified as binding partners of ARL6IP1–FAM134B heterodimers but not ARL6IP1 homodimers (Fig. 3b,d,e and Extended Data Fig. 3c,d). Both ARL6IP1 and FAM134B homodimers interacted with components of the ubiquitination machinery (Fig. 3b,d,e). This included different E3 ligases, such as the ER-resident E3 ligase AMFR (also known as gp78) and HUWE1, and the deubiquitinating enzymes USP9X and USP24. Taken together, our data suggest that heterocomplexes of ARL6IP1 and FAM134B are part of ER-phagy receptor clusters.

Fig. 3: ARL6IP1 supports the formation of a multi-receptor complex with FAM134B and other ER-shaping proteins.
figure 3

a, ARl6IP1 N-terminally tagged with the non-fluorescent N-terminal (V1) and ARl6IP1 N-terminally tagged with the C-terminal (V2) fragment of the Venus protein exhibit fluorescence after interaction. b, Single-sided Volcano plot of the label-free interactome of ARL6IP1 homodimers revealed that FAM134B, FAM134C and other RHD-containing ER proteins (blue) and proteins of the ubiquitination machinery (green) are binding partners. Only notable hits with log2(P value) > 1 and –log10(P value) > 1.3 are labelled in colour (one-sided paired Student’s t-test with three biological replicates). c, FAM134B N-terminally tagged with the non-fluorescent N-terminal (V1) and ARL6IP1 N-terminally tagged with the C-terminal (V2) fragment of the Venus protein exhibit fluorescence after interaction. d, Single-sided volcano plot of the label-free interactome for V1-FAM134B–V2-ARL6IP1 heterodimers include autophagy-related proteins LC3B and GABARAPL2 (red) and FAM134B as known binding partners. Notable hits with log2(P value) > 1 and –log10(P value) > 1.3 are labelled in colour (one-sided paired Student’s t-test with three biological replicates). Red dots indicate autophagy-related proteins. e, Heatmap of log2 enrichment of V1-ARL6IP1–V2-ARL6IP1, V1-FAM134B–V2-ARL6IP1 or V1-FAM134B–V2-FAM134B over mock from notable hits indicated in b and d.

Ubiquitination promotes LC3B binding of FAM134B

As AMFR mediates the ubiquitination of FAM134B17, we proposed that ubiquitination may also participate in the regulation of ARL6IP1. The LC–MS analysis identified several ubiquitinated lysine residues within ARL6IP1 in both ARL6IP1 homodimers and ARL6IP1–FAM134B heterodimers (Fig. 4a and Extended Data Fig. 4a,c–e). Most of the residues were located close to the RHDs of ARL6IP1, with K96 being significantly ubiquitinated and K114 and K130 potentially ubiquitinated (Fig. 4a). FAM134B exhibited an even higher number of ubiquitinated lysine residues, namely K90, K160, K278, K374 and K485 and potentially K247 and K264, which are also located close to its RHDs17 (Fig. 4b and Extended Data Fig. 4b,f–l). Co-immunoprecipitation assays following the overexpression of HA–ARL6IP1 and Myc–ubiquitin in HEK293T cells confirmed that a significant amount of ARL6IP1 is ubiquitinated in the presence of FAM134B (Fig. 4c).

Fig. 4: LC3B binding to FAM134B–ARL6IP1 complexes depends on ARL6IP1 ubiquitination.
figure 4

a,b, Position and log2(fold changes) of ubiquitinated lysine residues identified by LC–MS in ARL6IP1 (cells expressing V1-ARL6IP1–V2-ARL6IP1 or V1-FAM134B–V2-ARL6IP1) (a) and FAM134B (cells expressing V1-FAM134B–V2-FAM134B or V1-FAM134B–V2-ARL6IP1) (b). Significant sites are indicated in red (–log10(P value) > 1.3, one-sided unpaired Student’s t-test). c, Anti-Myc IP from HEK293T cells co-expressing HA–ARL6IP1, GFP–FAM134B and Myc–ubiquitin (Ub) confirms ARL6IP1 and FAM134B ubiquitination (one experiment). d, Snapshots from coarse-grained molecular dynamics simulations showing the most populated conformations of non-ubiquitinated and ubiquitinated ARL6IP1 (ubiquitin purple) in 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC 16:0/18:1) bilayers (orange). e, Interaction with FAM134B but not endogenous REEP5 is promoted by ARL6IP1 ubiquitination (two experiments). f, LC3B co-precipitation with V1-FAM134B–V2-ARL6IP1-7KR is reduced compared with V1-FAM134B–V2-ARL6IP1 (three experiments; one-way ANOVA, F = 46; Holm–Sidak’s post-test, FAM134B–FAM134B versus FAM134B–ARL6IP1, P = 0.0014; FAM134B–FAM134B versus FAM134B–ARL6IP1-7KR, P = 0.0002; FAM134B–ARL6IP1 versus FAM134B–ARL6IP1-7KR, P = 0.0242). g, Pearson’s correlation analysis for ubiquitin-positive and LC3B-positive Venus puncta for V1-FAM134B–V2-ARL6IP1 and V1-FAM134B–V2-ARL6IP1-7KR in ARL6IP1 KO U2OS cells (1 experiment with n = 11 (ubiquitin) and 15 (LC3B) cells; two-sided unpaired Student’s t-test, P = 0.006 (ubiquitin) and P = 0.0001 (LC3B)). h, Left, ubiquitinated ARL6IP1 is detected after TUBE2 pull-down after co-expression of GFP–FAM134B and HA–ARL6IP1 with AMFR–Flag but not AMFR-RINGmut–Flag. Right, ARL6IP1–FAM134B interaction is promoted by AMFR but not AMFR-RINGmut–Flag (n = 1). i, Myc pull-down after co-expression of Myc–ubiquitin–HA–ARL6IP1 with AMFR–Flag or AMFR-RINGmut–Flag confirms ARL6IP1 ubiquitination with active AMFR (n = 1). j, Schematic (left) and quantification (right) of LC–MS of GFP pull-downs of co-expressed V1-ARL6IP1 and AMFR-V2 or AMFR-RINGmut-V2: AMFR ubiquitinates ARL6IP1 at K96 and K114 (one-sided unpaired Student’s t-test: V1-ARL6IP1–V2-ARL6IP1 versus V1-ARL6IP1–AMFR-V2 K96, P = 0.0075; K114, P = 0.006; V1-ARL6IP1–AMFR-V2 versus V1-ARL6IP1–AMFR-RINGmut-V2 K96, P = 0.038; K114, P = 0.044). k, In vitro ubiquitination of Trx–His–ARL6IP1 with AMFR decreases the mean liposome diameter in TEM of freeze-fractured liposomes by about 20%. Incubations without ATP served as control (2 experiments with n = 393, 346 and 223 vesicles analysed (left to right); two-sided Kruskal–Wallis test with Dunn’s post-hoc test: Trx–His–ARL6IP1 and ATP–Trx–His–ARL6IP1 + ATP, P = 0.0024; Trx–His–ARL6IP1 – ATP and control, P < 0.0001; Trx–His–ARL6IP1 + ATP and control, P < 0.0001). Quantitative data shown as the mean ± s.e.m.

Source Data

We next simulated the structural dynamics of non-ubiquitinated ARL6IP1 and the ubiquitinated ARL6IP1 (K96-Ub) embedded in phosphocholine bilayers using coarse-grained molecular dynamics simulations (up to 10 µs). Ubiquitination resulted in a more compact conformation in which the ubiquitin moiety interacted with the cytosolic loops (Fig. 4d). To assess the functional relevance of ARL6IP1 ubiquitination, we replaced all seven predicted ubiquitinated lysine residues with arginine residues (ARL6IP1-7KR). Transiently expressed HA–ARL6IP1-7KR co-precipitated with green fluorescent protein (GFP)-tagged FAM134B (GFP–FAM134B), albeit to a lesser extent (Fig. 4e). A reduced interaction was also evident from co-immunoprecipitation of endogenous FAM134B with HA-ARL6IP1-7KR (Extended Data Fig. 2f). HA-ARL6IP1-7KR also co-localized with the ER sheet marker CLIMP63 (ref. 19) (Extended Data Fig. 5a) and with FAM134B–Myc (Extended Data Fig. 5b). Moreover, the shaping properties of ARL6IP1-7KR tagged with a thioredoxin–histidine tag (Trx–His–ARL6IP1-7KR) did not differ from Trx–His–ARL6IP1 in liposome-binding assays (Extended Data Fig. 5c). ARL6IP1-7KR therefore preserves basic functions of ARL6IP1.

Next, we cloned ARL6IP1-7KR and FAM134B constructs with the complementing V1 and V2 segments of the Venus protein. Precipitation with anti-GFP-beads showed an efficient pull-down of V1-FAM134B–V2-ARL6IP1 heterodimers with LC3B. By contrast, binding to LC3B was significantly reduced for complexes with the ubiquitination-deficient ARL6IP1-7KR variant (Fig. 4f and Extended Data Fig. 5d). Co-expression of V1-ARL6IP1-7KR and V2-ARL6IP1 or V1-FAM134B and V2-ARL6IP1 in ARL6IP1 KO U2OS cells resulted in a regular distribution of the Venus signal in the ER (Extended Data Fig. 5e). Co-localization analysis using Pearson’s correlation showed that Venus puncta co-labelling with ubiquitin or LC3B in ARL6IP1 KO U2OS cells were reduced for ARL6IP1-7KR (Fig. 4g and Extended Data Fig. 5e).

Because the ER-resident E3 ligase AMFR was found in the interactome of V1-FAM134B–V2-ARL6IP1 complexes, we tested whether AMFR ubiquitinates ARL6IP1. We detected ARL6IP1 and FAM134B in the tandem ubiquitin binding entity (TUBE2) pull-down from cell lysates upon co-expression of GFP–FAM134B and HA–ARL6IP1 together with AMFR–Flag, but not with the catalytically inactive AMFR-C356G-H361A variant (AMFR RINGmut-Flag) (Fig. 4h, left). Moreover, the interaction between ARL6IP1 and FAM134B was promoted by AMFR (Fig. 4h, right). ARL6IP1 was detected in the Myc pull-down assay after co-expression of Myc–ubiquitin and ARL6IP1 with active AMFR, but not with the inactive AMFR RINGmut variant (Fig. 4i). When V1-ARL6IP1 was expressed with AMFR-V2, the ratio of peptides ubiquitinated at K96 or K114 in comparison to total peptides was approximately doubled compared with co-expression with AMFR RINGmut-V2 (Fig. 4j). We further detected ubiquitinated endogenous ARL6IP1 after induction of ER stress (Extended Data Fig. 5f), and we verified that ARL6IP1-K96 was ubiquitinated by AMFR in vitro (Extended Data Fig. 5g).

We considered that ubiquitination of ARL6IP1 might be important for membrane remodelling during ER-phagy. Accordingly, shaping assays showed that in vitro ubiquitination of ARL6IP1 by AMFR resulted in reduced mean liposome diameters (Fig. 4k). We therefore propose that the ubiquitination of RHDs of ARL6IP1 and FAM134B by AMFR is probably involved in ER remodelling during ER-phagy.

Impaired ER-phagy flux after ARL6IP1 disruption

To assess whether FAM134B-mediated ER-phagy is compromised in the absence of ARL6IP1, we overexpressed the mCherry–GFP–FAM134B reporter in WT and Arl6ip1 KO MEFs (Fig. 5a). Staining for LC3B enabled the quantification of autophagosomes and autolysosomes because GFP is quenched in the acidic lumen of autolysosomes. Puncta labelled for LC3B, mCherry and GFP (autophagosomes) and puncta positive for LC3B and mCherry but negative for GFP (autolysosomes) were decreased in KO MEFs and in patient fibroblasts (Fig. 5b,c and Extended Data Fig. 6a). A defect in ER-phagy was further confirmed following siRNA-mediated knockdown of ARL6IP1 in U2OS cells expressing the mCherry–GFP–FAM134B reporter after induction with doxycycline (Extended Data Fig. 6b,c). Similar results were also obtained after knockdown of ARL6IP1 or FAM134B in HeLa cells alone or in combination, when the defect was more severe (Extended Data Fig. 7a,b). The quantification of SEC62-positive autophagosomes and autolysosomes in either starved Arl6ip1 or Fam134b KO MEFs further confirmed a defect in ER-phagy. By contrast, SEC62-negative autophagosomes and autolysosomes as a readout for bulk autophagy were not affected (Extended Data Fig. 7c). Intact bulk autophagy was additionally verified using the mCherry–GFP–LC3 reporter after knockdown of ARL6IP1 in U2OS cells (Extended Data Fig. 7d,e). Mitophagy (Extended Data Fig. 7f) and pexophagy (Extended Data Fig. 7g) were not affected following the disruption of ARL6IP1. Notably, ER-phagy in Arl6ip1 KO MEFs was rescued by ARL6IP1 or ARL6IP1-7KR overexpression (Extended Data Fig. 8).

Fig. 5: FAM134B-mediated ER-phagy requires ARL6IP1 in mice and humans.
figure 5

a, Rational of the mCherry–GFP–FAM134B reporter. b,c, ARL6IP1 loss-of-function compromises ER-phagy. Arl6ip1 WT and KO MEFs (b) or fibroblasts from the patient and the healthy individual (control) (c) were transfected with mCherry–GFP–FAM134B and stained for LC3B. Quantifications of LC3B+mCherry+GFP+ puncta and LC3B+mCherry+GFP puncta per cell area suggest that the formation of autophagosomes (3 experiments with 10 cells per genotype each; two-sided Mann–Whitney U-test; MEFs, P = 0.0479; human cells, P = 0.0005) and autolysosomes (3 experiments with 10 cells per genotype each; two-sided Mann–Whitney U-test; MEFs, P = 0.0307; human cells, P = 0.0096) is impaired. d,e, ARL6IP1 loss-of-function enlarges ER sheets. Arl6ip1 WT and KO MEFs (d) or fibroblasts from the patient and healthy individual (e) were stained for the ER sheet protein CLIMP63 and the relative CLIMP63+ area per cell calculated (3 experiments with 15 cells per genotype; two-sided Mann–Whitney U-test; MEFs, P = 0.0001; human cells, P = 0.0001). f,g, TEM images showed increased numbers of small highly curved ER protrusions emanating from ER sheets in the absence of ARL6IP1 (ER sheets in light purple, ER-emerging spikes in pink) in MEFs (f) and human fibroblasts (g) (1 experiment with 55 cells per genotype; two-sided Mann–Whitney U-test; P = 0.0059 (f) and P = 0.0001 (g)). h,i, ARL6IP1 loss-of-function decreases cell viability in the presence of the ER stressors tunicamycin (Tunicam.) or thapsigargin (Thapsig.) with or without the proteasome inhibitor MG132 in MEFs (h) and human fibroblasts (i) (3 experiments with 2 replicates; two-sided Mann–Whitney U-test; MEFs, thapsigargin + MG132, P = 0.0022; tunicamycin, P = 0.041; tunicamycin + MG132, P = 0.0022; patient cells, thapsigargin, P = 0.0022; thapsigargin + MG132, P = 0.0087; tunicamycin, P = 0.0022; tunicamycin + MG132, P = 0.026). j, Cartoon showing ARL6IP1 (blue) ubiquitinated (red balls) by AMFR in a complex with FAM134B (red), which binds to LC3 (purple) during ER-phagy. Quantitative data are shown as the mean ± s.e.m. Individual experiments are indicated by differently coloured data points. Scale bars, 500 nm (f,g), 10 µm (d,e) or 20 µm (b,c). The models in a and j were created using BioRender (https://www.biorender.com).

Source Data

Because FAM134B is involved in the lysosomal clearance of ER subdomains containing ERAD-resistant misfolded proteins such as pro-collagen20, we also assessed whether the clearance of pro-collagen is affected following disruption of ARL6IP1. The ratio of cells with an intracellular accumulation of pro-collagen was substantially increased in both Fam134b and Arl6ip1 KO MEFs (Extended Data Fig. 9a). A defect in collagen exocytosis could be excluded because collagen concentrations in cell culture supernatants did not differ between genotypes (Extended Data Fig. 9b). Comparable results were obtained for fibroblasts from the patient and from the healthy individual (Extended Data Fig. 9c,d), which were confirmed by immunoblot analysis (Extended Data Fig. 9e,f). Co-staining with LAMP2 further indicated that the delivery of pro-collagen to lysosomes is impaired in Fam134b and Arl6ip1 KO MEFs (Extended Data Fig. 9g).

Collectively, these data suggest that FAM134B-mediated ER-phagy requires ARL6IP1 and that it is likely to be promoted by ARL6IP1 ubiquitination.

Abnormal ER sheets after ARL6IP1 disruption

In agreement with a severe defect of FAM134B-dependent ER-phagy flux, the relative area covered by CLIMP63-labelled ER sheets was increased in maximum intensity projections of confocal z-stacks of Arl6ip1 KO MEFs (Fig. 5d) and in patient fibroblasts (Fig. 5e). The area labelled for RTN4, a marker of ER tubules21, was decreased, which closely resembled findings for Fam134b KO MEFs (Extended Data Fig. 10a).

Transmission electron microscopy (TEM) analyses revealed an increased number of spike-like ER protrusions in Arl6ip1 KO MEFs (Fig. 5f). The numbers of spike-like ER protrusions was higher in fibroblasts from the patient compared with those from the healthy individual (Fig. 5g). Notably, such ER protrusions were also increased in fibroblasts from the patient’s father (heterozygous for K193Ffs) compared with the healthy individual, but significantly less compared with the patient (Extended Data Fig. 10b).

The compromised ER structure was accompanied by diminished cell viability in response to ER stressors such as thapsigargin or tunicamycin with or without the proteasome inhibitor MG132 in KO MEFs (Fig. 5h) and in fibroblasts from the patient (Fig. 5i). By contrast, the viability of fibroblasts from the patient’s father only differed marginally from control cells (Extended Data Fig. 10c). Notably, the alterations in ER structure were more pronounced in human fibroblasts compared with MEFs, whereas the sensitivity was less for human fibroblasts in our cell viability assays. This disparity may be explained by species-specific differences or differences between embryonic and adult tissues. Overall, we showed that the disruption of ARL6IP1 leads to a defect in ER remodelling and ER structure and decreases resistance to ER stress.

We conclude that the loss of ARL6IP1 crucially impairs ER-phagy and decreases cellular fitness. As an in vivo correlate of defective ER-phagy, ER sheets were expanded in DRG neurons in Arl6ip1 KO mice, which degenerated over time and led to sensory loss. In agreement, ARL6IP1 knockdown was reported to result in the expansion of ER sheets in HeLa cells22 and the fragmentation of ER tubules in Drosophila axons23. Here we showed that ARL6IP1 is required for FAM134B-mediated ER-phagy. However, it does not act as an ER-phagy receptor on its own (Fig. 5j) because it cannot bind LC3 proteins itself. Instead, ARL6IP1 forms multi-protein clusters with FAM134B and other ER-shaping proteins. Notably, the RHDs of FAM134B and ARL6IP1 were ubiquitinated within such ER-phagy competent multi-protein clusters. Moreover, the ubiquitination of ARL6IP1 increased its shaping properties and its binding to LC3 by FAM134B. Our observation that membrane protrusions emanating from ER sheets are increased suggests that ER membrane remodelling is incomplete in the absence of ARL6IP1. As a new mechanism, we propose that ER-phagy is controlled by the ubiquitin-dependent formation of heteromeric ER-phagy receptor complexes. Ubiquitination therefore acts as a higher level of regulation acting on top of RHD phosphorylation to promote FAM134B oligomerization and ER fragmentation24. As ER-phagy is also involved in the degradation of ER subdomains containing ERAD-resistant misfolded proteins, which has been shown for ATZ polymers25 and for endogenous pro-collagen20, compromised ER-phagy following the disruption of either FAM134B or ARL6IP1 coupled with age-associated attenuation of autophagy might contribute to an accumulation of misfolded or aggregated proteins within the ER. Subsequently, this leads to impaired proteostasis and progressive neurodegeneration. Because ARL6IP1 also binds to FAM134A and FAM134C, two broadly expressed close homologues of FAM134B2, which also serve as ER-phagy receptors26, this may explain why mutations in ARL6IP1 lead to more severe disease2,3,8,9,27. In light of the evolutionarily conserved function of FAM134B, the formation of heteromeric clusters of ubiquitinated membrane-shaping proteins to remodel the ER may represent a more general principle of cell homeostasis.

Methods

To study the function of ARL6IP1 and the consequences of its disruption, we investigated Arl6ip1 KO mice and different cell culture models, including fibroblasts from a patient carrying the homozygous mutation AR6IP1K193Ffs, from the patient’s father (unaffected and carrying the heterozygous allele) and from a healthy individual. Mouse experiments were performed on a C57BL/6 background after backcrossing for more than four generations. Mice were maintained in groups of up to 3 mice per cage at 21 ± 2 °C, air humidity of ≥45%, 15-fold air exchange, 14–10 h day–night cycle and maximum 500 lx. Mice had free access to standard mouse chow and water. Littermates of the same sex were randomly assigned to experimental groups. Experiments were conducted in a blinded manner with regard to cell, mouse and human genotypes. Figure legends include details of replicate experiments used to generate datasets. All animal experiments were approved by the Thüringer Landesamt für Lebensmittelsicherheit und Verbraucherschutz (registration numbers 02-055/14 and UKJ-17-006). Studies using human fibroblasts were approved by the local ethics committee.

Plasmids are presented in Supplementary Table 1. cDNAs were cloned into the pDONR223 vector using a BP Clonase Reaction kit (Invitrogen, 11789100) and further recombined into the Gateway destination vectors pcDNA5-FRT/TO-N-mCherry-EGFP, pcDNA3.1-N-HA, pHAGE-GFP, pcDNA3.1-N-Flag, pcDNA3.1-C-Flag, pcDNA3.1-N-SBP-Flag, pGEX6-GST, and the biomolecular complementation affinity purification system vectors pDEST-V1-ORF, pDEST-V2-ORF, pDEST-ORF-V1 and pDEST-ORF-V2 using a LR Clonase Reaction kit (Invitrogen). Plasmids encoding untagged ARL6IP1, Trx–His-tagged ARL6IP1 and Trx–His–ARL6IP1-7KR were generated by subcloning ARL6IP1 from pGEX6P1 into pPal7 and pET32a, respectively. ARL6IP1 in pGEX6P1 was cloned by PCR using pClneo-ARL6IP1 WT-GFP as template. The three sgRNA guides of ARL6IP1 were cloned into the pLentiCRISPR v2 vector.

Primers are presented in Supplementary Table 2, primary antibodies in Supplementary Table 3 and secondary antibodies are presented in Supplementary Table 4.

Generation of Arl6ip1 KO mice

The EUCOMM embryonic stem cell clone HEPD0752_7_A11 (Source Bioscience) was injected into C57BL/6J donor blastocysts. Next, 15–30-week-old F1 female offspring from C57BL/6J and CBA/J matings served as foster mice. Resulting chimeras were mated with C57BL/6J. For all experiments, littermates were used, which had been backcrossed for at least four generations. Genotyping was performed by PCR with three primers (Arl6ip1-forward: 5′-GTAATATTCTGAGCACTGCCT-3′, Arl6ip1-KO-reverse: 5′-TGCCATAATGACCTAATACTGTTGTG-3′, Arl6ip1-WT-reverse: 5′-CTAAGCACAGGCTATGAACC-3), which produced a WT band of 537 bp or the KO band of 350 bp.

Generation of ARL6IP1 CRISPR–Cas9 KO cell lines

The ARL6IP1 knockout U2OS cell line was generated using a lentiviral CRISPR–Cas9 system. sgRNAs are reported in Supplementary Table 5 (design at https://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design). The lentiviral plasmids were generated as previously reported28. The forward and reverse oligonucleotides were annealed and phosphorylated using T4 polynucleotide kinase (BioLabs). The oligonucleotides were ligated into the Cas9 vector pLenti-Puro-v2 and pLenti-Puro-EGFP using the BsmBI site. The lentiviral plasmids were co-transfected into HEK293T cells together with the packaging vectors pPAX2 and pDM2.G for lentivirus production. After 48 h, the medium containing lentiviral particles was collected, centrifuged to remove dead cells and stored at –80 °C. To generate the ARL6IP1 KO cell line, fresh U2OS or HeLa cells were infected with lentiviral particles with the three different sgRNAs for 48 h and then selected using 5 μg ml–1 of puromycin. The surviving cells were maintained in DMEM supplemented with 2 μg ml–1 puromycin. When using pLenti-Puro-EGFP as backbone of the sgRNA, cells were also FACS-sorted for GFP expression (SONY SH800S Cell Sorter, version 2.1.6). KO was verified by western blotting.

Motor performance

For the beam walk test, mice were placed on an elevated beam of 1 m in length and 4 cm in width, with the home cage at the end. After habituation on three consecutive days, the mouse was videotaped from behind during its movement on the beam. The foot base angle of the hind limb was measured at the moment when the toe was lifted.

For the grip strength analysis, mice were lifted at the tail base, brought to a trapeze-shaped handle connected with a force sensor (Grip Strength Meter, Ugo Basile). When the mouse spontaneously grabbed the handle, the mouse was gradually pulled away from the handle until it was released.

For the electrophysiological analysis of peripheral nerves, anaesthetized mice (100 mg kg–1 ketamine and 16 mg kg–1 xylazine) were placed on a heating pad. One pair of needle electrodes with a tip distance of 5 mm (WE30030.1H10, Science Products) was inserted near the base of the tail and a second pair 30 mm distal to the stimulation site close to the tip of the tail. For the analysis of motor fibres, the stimulus was applied through the proximal electrodes and the response recorded using the distal electrodes. Compound muscle action potentials and sensory nerve action potentials were evoked with increasing intensity (0–15 V, increment of 1 V, 50 µs duration, interstimulus interval of 20 s). Sum action potentials were filtered (high-pass filter 3 Hz, low-pass filter 1.3 kHz) and digitized with a sampling frequency of 10 kHz. Amplitudes were determined from peak to peak.

Histology, neuron count and TEM of mouse tissues

Mice were deeply anaesthetized and transcardially perfused with PBS (pH 7.4) followed by 4% paraformaldehyde (PFA) in PBS for 10 min. After dissection, tissues were post-fixed in 4% PFA in PBS for at least 1 h. Tissues were incubated in sucrose (10% sucrose for 4 h and in 30% sucrose overnight at 4 °C), frozen on dry ice and cut with a sliding microtome (Leica SM 2000R) in 30-µm-thick free-floating sections and stored in PBS supplemented with sodium azide at 4 °C until further use. For NeuN staining, free-floating sections were permeabilized with 0.25% Triton X-100, blocked in 5% normal goat serum (NGS) for 1 h and incubated with mouse anti-NeuN (Millipore) 1:500 at 4 °C overnight. After washing, sections were incubated with the appropriate secondary antibodies (Invitrogen) at 1:1,000 and Hoechst 33342 (Thermo Scientific). Sections were mounted with Fluoromount-G (Southern Biotech). Images were acquired using Cellobserver Z1 (Zeiss) with the tile-scan module and further analysed using ImageJ.

For paraffin embedding, the samples were dehydrated overnight in a series of ethanol and xylol baths (Leica TP20 Tissue Processor), embedded with paraffin (Leica HistoCore Arcadia) and cut into 5 µm sections with a microtome (ThermoScientific Microm HM 355S). For histological analyses, sections were either stained with haematoxylin and eosin (Sigma-Aldrich) or cresyl-violet (Sigma-Aldrich). Images were acquired using a Zeiss AxioLab A1 microscope and further analysed using ImageJ.

For TEM of tissue sections, animals were perfused with 2.5% glutaraldehyde in PBS unless indicated otherwise. For the analysis of DRGs, mice were perfused with 4% PFA and 2% glutaraldehyde in PBS. After dissection, tissues were post-fixed overnight. Tissues were contrasted with 1% osmium tetroxide, dehydrated and infiltrated with epoxy resin. Ultrathin sections were stained with uranyl acetate and lead citrate, mounted on copper grids and viewed with a Philips CM10 or Zeiss EM 900 digital (DRGs) transmission electron microscope.

TEM of cultured cells

MEFs and human fibroblasts were fixed by adding an equal volume of double strength fixative (4% PFA, 5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.4) to the culture medium and incubated for 20 min at room temperature. The fixative mixture was then replaced with one volume of single strength fixative (2% PFA and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.4) for another 2 h at room temperature. After 5 washes with 0.1 M sodium cacodylate buffer (pH 7.4), cells were processed for dehydration and embedding in Epon resin29. To preserve their original morphology, the monolayer culture of fibroblasts of the patient and of his father were embedded in their original position in their culture flasks. By contrast, WT and Arl6ip1 KO MEFs were scraped into 2% low-melting-point agarose before the dehydration process and the embedding in Epon resin, as previously described29. Subsequently, 70 nm ultrathin sections were cut using a Leica EM UC7 ultra microtome (Leica Microsystems) and stained with uranyl acetate and lead citrate29. Cell sections were analysed using an 80 kV transmission electron microscope CM100bio TEM (FEI).

Skeletal muscle fibre bundle staining

Muscles freshly dissected from 2-month-old mice were fixed in 2% PFA for 15 min and subsequently washed with PBS. Fibre bundles were prepared and used for further analyses. After overnight permeabilization with 0.2% Triton X-100 in PBS, samples were blocked with 5% NGS for 1 h followed by an incubation with α-bungarotoxin-Alexa 555 (Invitrogen) 1:500 and mouse anti-NF200 overnight at 4 °C. After washing with PBS, single myofibre bundles were incubated with the corresponding secondary antibodies (Invitrogen) in a dilution of 1:1,000 for 1 h at room temperature. Nuclei were stained with Hoechst 33258 (Invitrogen). Myofibres were washed with PBS and mounted using Fluoromount-G (Southern Biotech). Images were acquired using a Zeiss LSM 880 confocal microscope with Airyscan using the z-stack module. Z-projections with maximum intensities processed using ImageJ are shown.

Cell culture

HEK293T, U2OS and HeLa cells were obtained from the American Type Culture Collection. Their identities were authenticated by STR analysis. U2OS TRex cells were provided by S. Blacklow (Brigham and Women’s Hospital and Harvard Medical School), HeLa TRex were provided by S. Taylor (Manchester University). WT, Fam134b KO and Arl6ip1 KO MEFs were isolated from embryos and immortalized using SV40 large T antigen. All cell lines were regularly tested for mycoplasma contamination using a LookOut Mycoplasma PCR Detection kit (Sigma). Cells were maintained at 37 °C with 5% CO2 in DMEM medium (Gibco) supplemented with 10% FBS (Gibco) and 100 U ml–1 penicillin and streptomycin (Gibco).

Inducible cell lines were induced with 1 µg ml–1 doxycycline (Sigma-Aldrich).

Bafilomycin A1 (LC-Laboratories) was used at a concentration of 200 ng ml–1, Torin1 (LC-Laboratories) at 250 nM. EBSS medium was obtained from Gibco. For each treatment, cells were plated the day before to perform the experiments when cells had a confluence of 50–60%. For transient expression, DNA plasmids were transfected using GeneJuice (Merck-Millipore), Turbofect (Thermo Scientific) or Lipofectamine 2000 (Invitrogen).

U2OS TRex cell lines were used to generate stable cell lines using the Flp-IN TRex system (Invitrogen) or the lentiviral vector p-Lenti N-HA. For ARL6IP1 knockdown experiments, the respective cells were transfected with either 30 pmol siNT (non-targeting sequence, Qiagen) or with 30 pmol double-stranded ARL6IP1 siRNA (Integrated DNA Technologies; hs.RiARL6IP1.13.2) using Lipofectamine RNAiMAX transfection reagent (Invitrogen, 13778075). For ARL6IP1 and FAM134B double-knockdown experiments, HeLa cells were transfected with ARL6IP1 siRNA and FAM134B siRNA (siRNA RETREG1 18 J-016936-18-0002 and siRNA RETREG1 21 J-016936-21-0002, respectively). Experiments were performed 72 h after transfection.

Protein isolation from cells and tissue lysates

Cells were collected and lysed in RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% (v/v) NP-40, 1% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 1 mM EDTA and complete protease inhibitor (Roche)). Tissue lysates were prepared using a Ultra-Turrax T8 tissue homogenizer (IKA-WERKE) in RIPA buffer. After sonication, homogenates were spun down at 16,900g to remove nuclei and insoluble debris. The supernatant was stored at –80 °C until further use.

Western blotting

Proteins were denatured at 90 °C for 5 min in Laemmli buffer, resolved by SDS–PAGE and transferred to methanol-activated PVDF membranes (Amersham Hybond P 0.45 µm). Membranes were blocked for 1 h in 10% skim milk in TBS-T (Tris-buffered saline with Tween, 20 mM Tris, 150 mM NaCl, 0.1% Tween 20, pH 7.5) and incubated overnight at 4 °C with the specific primary antibody followed by 1 h incubation with the respective secondary antibody at room temperature. Detection was carried out using Clarity Western ECL substrate (Bio-Rad) and a LAS 4000 automated detection system (GE Healthcare). Bands were quantified using ImageJ.

Real-time qPCR

RNA was isolated by TRIzol–chloroform extraction. RNA was reverse-transcribed using a GoScript reverse transcription kit (Promega). qPCR was performed with a final amount of 20 ng of cDNA and EvaGreen Mix (Bio-Rad) with primer pairs for either mouse Gapdh (forward, GCTCATGACCACAGTCCAT; reverse, GTCATCATACTTGGCAGGTTT), mouse Arl6ip1 (forward, GCTCTAATAAATGGACCACTG; reverse, GCACAAATGTCACAATCAGGT), human GAPDH (forward, GAAGGCTGGGGCTCATTT; reverse, GGACTGTGGTCATGAGTC) or human ARL6IP1 (forward, GCTCCAATAAATGGACCACTGA; reverse, GGAAGTCACTATCAGGTAGGT) on a CFX96 Touch Real-Time PCR detection system (Bio-Rad).

Immunofluorescence and autophagic flux analysis

For immunostainings, cells were fixed with 4% PFA at room temperature or ice-cold methanol, washed with PBS and permeabilized with 0.25% Triton X-100 in PBS at room temperature for 10  min, or with 0.1% saponin in PBS at room temperature for 1 min followed by blocking with 5% NGS in PBS for 1  h at room temperature. Incubation with primary antibodies diluted in 5% NGS and 0.25% Triton X-100 in PBS was carried out overnight at 4 °C or at room temperature for 1 h. After three consecutive washes with PBS, secondary antibodies were incubated for 1 h at room temperature. After three washes, cells were incubated at room temperature with Hoechst 33258 (Invitrogen) for 10 min. After a final wash with PBS, the coverslips were mounted with Fluoromount-G solution (ThermoFisher, 00-4958-02). Images were acquired using a confocal microscope (Zeiss LSM 880 with Airyscan).

For the analysis of ER-phagy, cells were transiently transfected with the mCherry–GFP–FAM134B reporter construct. Cells were fixed with ice-cold methanol for 10 min 24 h after transfection. Then cells were permeabilized with 0.25% Triton X-100 in PBS and blocked with 4% NGS for 1 h and stained for LC3B. Images were taken with a LSM 880 and analysed using the ComDet (v.0.5.5) plugin for ImageJ (https://github.com/ekatrukha/ComDet; settings: particle size = 10 pixels, co-localization distance = 7, intensity threshold = 20). The cell border was selected and the cell area determined. Only signals within the cell border were analysed. The intensity threshold was set at 1,000 for all channels, except for human fibroblasts, for which the threshold for mCherry was set at 200. Pearson’s coefficients were calculated using the JaCOP plugin in ImageJ and normalized to the ER area.

To assess whether ARL6IP1 is involved in bulk autophagy, we knocked down ARL6IP1 in U2OS cells stably expressing the mCherry–GFP–LC3 reporter. Autophagy was triggered by 6 h of EBSS exposure or 6 h of 250 nM Torin1 exposure. Images were acquired with a high-content microscope–Yokogawa CQ1 confocal imaging cytometer. To assess whether ARL6IP1 is involved in mitophagy, ARL6IP1, CRISPR–Cas9 KO HeLa cells were transfected with the mitophagy reporter mCherry–GFP–FIS1. Autophagic flux was triggered with 40 μM CCCP for 4 h. Pexophagy was assessed after siRNA-mediated knockdown of ARL6IP1 in U2OS cells after induction of the doxycycline-inducible reporter mCherry–GFP–PMP34 (ref. 30) at baseline or after starvation with EBSS for 20 h. Images were acquired with a confocal microscope (Zeiss LSM 880 with Airyscan). The red and yellow puncta were manually counted.

Rescue experiment

Arl6ip1 WT and KO MEFs were seeded in 24-well-plates at 40,000 cells per well. After 24 h, cells were transfected with the mCherry–GFP–FAM134B plasmid in combination with either the ARL6IP–HA or the ARL6IP1-7KR–HA plasmid. After 48 h, cells were fixed with ice-cold methanol for 10 min, permeabilized with 0.25% Triton X-100 in PBS and blocked with 4% NGS for 1 h and stained for LC3B and HA and further processed as described above. Images were taken with a LSM 880 and analysed using the ComDet (v.0.5.5) plugin for ImageJ (https://github.com/ekatrukha/ComDet; settings: particle size = 10 pixels, co-localization distance = 10, intensity threshold = 200/200/15). Only signals within the cell border were analysed.

ER stress induction and cell viability count

MEFs or human fibroblasts were seeded in 6-well-plates and cultured to 70–80% confluency. Cells were washed with PBS and incubated with new medium with 1.5 µM thapsigargin (Sigma, T9033-5MG) or 5 µg ml–1 tunicamycin (Santa Cruz) without or in combination with 1 µM MG132 (Calbiochem, 474787-10MG) to inhibit the proteasome. After 24 h for MEFs and 48 h for human fibroblasts, the culture medium was removed, cells washed with PBS and trypsinized. All cells were pooled, centrifuged at 800 r.p.m. for 5 min (Heraeus Sepatech Megafuge 2.0R) and resuspended in fresh medium. Cell viability was measured by trypan blue exclusion with an automatic counting device (Bio-Rad TC20 automatic cell counter).

Interactome analysis and sample preparation for MS

For the LC–MS interactome analysis, HA–FAM134B expression was induced in U2OS cells with doxycycline. HEK293T cells were transiently transfected with the constructs V1-ARL6IP1, V2-ARL6IP1, V1-FAM134B and V2-FAM134B. For in vivo ubiquitination, HEK293T cells were transiently transfected with the plasmids V1-ARL6IP1, V2-ARL6IP1, AMFR-V2 and AMFR-V2-C356G-H361A. After 24 h, cells were lysed with 1% Triton X-100 IP buffer (50 mM Tris-HCL, 150 mM NaCl and 0.5 mM EDTA). Lysis buffer without detergents was added to protein lysates to dilute Triton X-100 to 0.3%. Then samples were incubated with HA-agarose beads (Sigma-Aldrich, A2095) or GFP-Trap beads (Chromotek, gta-20) overnight at 4 °C on a rotating platform. Protein-bound beads were washed three times with lysis buffer supplemented with 0.1% Triton X-100 and once with lysis buffer without detergents. HA IP samples were incubated with 40 µl denaturing buffer (2% sodium deoxycholate, 1 mM tris(2-carboxyethyl)phosphine, 4 mM chloroacetamide and 50 mM Tris-HCl pH 8.5) and heated at 95 °C for 10 min. Samples were mixed 1:1 with 500 ng LysC (Promega), incubated for 3 h at 37 °C and digested with 500 ng of trypsin in 50 mM Tris-HCl, pH 8.5, overnight at 37 °C. GFP IP samples were denatured with 25 µl denaturing buffer at 60 °C for 30 min. After cooling down, the samples were digested with 25 µl 50 mM Tris with 1 µl of trypsin (500 ng) at 37 °C overnight. Reactions were stopped by addition of 150 µl of isopropanol containing 1% trifluoroacetic acid (TFA). Peptides were cleaned up by loading them onto SDB-RPS stage tips (Sigma). After one wash with 1% TFA in isopropanol and one wash with 0.2% TFA in water, peptides were eluted using 80% acetonitrile and 1.25% ammonia. Eluted peptides were dried, tandem mass tagged labelled and processed for LC–MS measurements.

LC–MS analysis

Dried peptides were reconstituted in 2% acetonitrile, 0.1% TFA and analysed on a Q Exactive HF mass spectrometer coupled to an easy nLC 1200 (ThermoFisher Scientific) using a 35-cm-long, 75 µm inner diameter fused-silica column packed in-house with 1.9 µm C18 particles (Reprosil pur, Dr. Maisch) and kept at 50 °C using an integrated column oven (Sonation). Peptides were eluted using a nonlinear gradient from 4 to 28% acetonitrile over 45 min and directly sprayed into the mass spectrometer equipped with a nanoFlex ion source (ThermoFisher Scientific). Full scan MS spectra (300–1,650 m/z) were acquired in profile mode at a resolution of 60,000 at m/z 200, a maximum injection time of 20 ms and an automatic gain control target value of 3 × 106 charges. Up to 15 most intense peptides per full scan were isolated using a 1.4 Th window and fragmented using higher energy collisional dissociation (normalized collision energy of 28). MS/MS spectra were acquired in centroid mode with a resolution of 30,000, a maximum injection time of 45 ms and an automatic gain control target value of 1 × 105. Single charged ions, ions with a charge state above 4 and ions with unassigned charge states were not considered for fragmentation, and dynamic exclusion was set to 20 s to minimize the acquisition of fragment spectra of already acquired precursors.

Proteomics data processing

MS raw data were processed using MaxQuant (v.1.6.10.43) applying default parameters. Acquired spectra were searched against the human ‘one sequence per gene’ database (taxonomy identifier 9606) downloaded from UniProt (accessed 3 March 2020; 20,531 sequences) and a collection of 244 common contaminants (“contaminants.fasta” provided with MaxQuant) using the Andromeda search engine integrated in MaxQuant31,32. Identifications were filtered to obtain false discovery rates (FDRs) below 1% for both peptide spectrum matches (minimum length of 7 amino acids) and proteins using a target–decoy strategy33. Protein quantification and data normalization relied on the MaxLFQ algorithm implemented in MaxQuant34. The MS proteomics data have been deposited to the ProteomeXchange Consortium35 through the PRIDE partner repository36 with the dataset identifiers PXD032718, PXD032720 and PXD039184. All acquired raw files were processed using MaxQuant (v.1.6.10.43) and the implemented Andromeda search engine. For protein assignment, spectra were correlated with the UniProt human database (v.2019) including a list of common contaminants. Searches were performed with tryptic specifications and default settings for mass tolerances for MS and MS/MS spectra. Carbamidomethyl at cysteine residues, oxidations at methionine, acetylation at the N terminus were defined as a fixed modification. The minimal peptide length was set to 7 amino acids and the FDR for proteins and peptide-spectrum matches to 1%. The match-between-run feature with a time window of 1 min was used. For further analysis, Perseus software (v.1.6.6.0) was used and first filtered for contaminants and reverse entries as well as proteins that were only identified by a modified peptide.

Proximity ligation assays

Proximity ligation assays (PLAs) were performed using a Duolink in situ red starter kit mouse/rabbit (DUO92101, Sigma-Aldrich) according to the manufacturer’s instructions with rabbit anti-FAM134B (ref. 1) and rabbit anti-ARL6IP1 (PRS3305, Sigma-Aldrich) antibodies using the Minus (DUO92010) and the Plus probe (DUO92009). Rabbit anti-FAM134B-Plus was used in a 1:5 dilution and rabbit anti-ARL6IP1-Minus in a 1:10 dilution on WT, Arl6ip1 and Fam134b KO MEFs after PFA (4%) fixation and permeabilization with 0.25% (v/v) Triton X-100 in PBS.

Fluorescence protease protection assay

We followed a fluorescence protease protection assay protocol as previously described37. In brief, 75,000 COS-7 cells per well were seeded on 18 mm coverslips coated with 0.1 mg ml–1 PLL. The next day, cells were transfected with the respective constructs using Lipofectamine 2000. After two more days, cells were washed with pre-warmed intracellular buffer (50 mM HEPES, pH 7.5, 23  mM NaCl, 3 mM MgCl2, 100 nM CaCl2, 1 mM EGTA and freshly added 107 mM potassium glutamate, 1 mM ATP and 2  mM dithiothreitol) and transferred to a heated perfusion chamber filled with the same buffer. Live cell imaging for both GFP and RFP was initiated on a Zeiss Cell Observer Z1 with a frame every 20 s starting with a pre-permeabilization image followed by manual administration of 18 µM digitonin. After 120 s, the buffer was replaced by intracellular buffer containing 6 mM freshly added trypsin. Analysis was carried out using ImageJ by drawing the outline of the selected cell and measuring the mean fluorescence intensity of the surrounded area subtracted by the background intensity taken from a cell free spot of the same frame. For further analysis, the area under the curve was calculated between 160 and 720 s.

Purification of recombinant proteins

Trx–His fusion proteins were purified from Escherichia coli treated with 0.5 mM IPTG (overnight, 18 °C). Eluted proteins were concentrated using Amicon Ultra-4-10k centrifugal filter units (Millipore) and then either dialysed at 4 °C against HN buffer (20 mM HEPES/KOH pH 7.4, 150 mM NaCl and 2.5 mM dithiothreitol) (material used for Fig. 4 and Extended Data Fig. 5 analyses) or against liposome buffer (25 mM HEPES-KOH, pH 7.2, 25 mM KCl, 2.5 mM magnesium acetate and 100 mM potassium glutamate) (protein used for studies presented in Fig. 2).

Liposome preparation, liposome incubations, freeze-fracturing and TEM

Liposomes were prepared using Folch-fraction type I lipids (Sigma-Aldrich) according to previously described procedures10,38. Liposome co-floatation assays were performed as previously reported10. In brief, liposomes and purified recombinant protein were incubated for 15 min at 37 °C in 0.3 M sucrose in liposome buffer, mixed with 75% sucrose in liposome buffer, overlaid with 200  μl 35% sucrose and 200 μl liposome buffer and then centrifuged at 200,000g for 30 min at 28 °C. Six fractions were collected from top to bottom and analysed by SDS–PAGE and fluorescence-based western blotting using a LICOR Odyssey system (LICOR Bioscience).

For shaping assays presented in Fig. 2, 1 mg of liposomes was incubated with 5 μM protein in liposome buffer containing 0.3 M sucrose for 15 min at 37 °C. For shaping assays presented in Fig. 4 and Extended Data Fig. 5, 1 mg of liposomes was incubated at 37 °C with 2.5 μM protein in HN buffer containing 0.3 M sucrose for 15 min. Thereafter, 20 μg proteinase K was added to avoid liposomal aggregates. The reaction was performed for 40 min at 45 °C (ref. 39). Small aliquots of the liposome suspension were then used for freeze-fracturing. The grids with the samples were systematically explored using an EM 900 electron microscope (Zeiss) operated at 80 kV. Images were acquired with a Wide-angle Dual Speed 2K (Tröndle) CCD camera. The diameters of liposomes were determined using ImageJ.

In vitro ubiquitination assay of ARL6IP1 with recombinant AMFR and sample preparation for MS

For the ubiquitination assay with AMFR, 1 µM purified Trx-His–ARL6IP1, 10 µM ubiquitin (in-house), 10 mM ATP and 10 mM MgCl2 were incubated with 0.8 µM AMFR (provided by B. Schulman, Max Planck Institute of Biochemistry), 100 nM E1 UBA1 (in-house) and 0.8 µM E2 UBE2G2 (Biotechne) in 150 mM NaCl, 50 mM Tris-HCl, pH 7.5, at 37 °C for 2 h. The reaction mixture was analysed by SDS–PAGE and Coomassie staining or immunoblot analysis for GST (Cell Signaling Technology), His (Cell Signaling Technology) and ubiquitin (Cell Signaling Technology) or by mapping of ubiquitinated lysines by MS.

Ubiquitination assays in cells, co-immunoprecipitations and TUBE2 pull-down

The ubiquitination of ARL6IP1 was assessed in HEK293T cells transfected with Myc–ubiquitin, HA–ARL6IP1 and either AMFR–Flag or its catalytically inactive RING mutant. Cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 1% Triton X-100, 10 mM N-ethylmaleimide and protease inhibitors (Roche Diagnostics, 5892791001)). The lysates were then incubated on ice for 15 min and centrifuged at 12,000g at 4 °C for 30 min. Next, 40 µl of the supernatant was collected, mixed with Laemmli sample buffer, boiled for 5 min at 95 °C and stored at –20 °C as input control. Myc-tagged ubiquitinated proteins were immunoprecipitated from lysates cleared with Myc-Trap Agarose (Chromotek, yta-10). Beads were washed three times with lysis buffer, heated at 95 °C for 5 min, subjected to SDS–PAGE and analysed by immunoblotting to detect the HA-Tag. For other immunoprecipitation assays, cleared lysates were incubated with GFP-Trap (Chromotek, gta-20), HA-agarose beads (Sigma-Aldrich, A2095) or TUBE2 agarose beads (Life Sensors, UM402) and incubated at 4 °C overnight. The next day, tubes were centrifuged (800g, 4 °C, 5 min), the supernatants removed and the beads washed with ice-cold lysis buffer. Input and co-precipitated fractions were analysed by SDS–PAGE and immunoblotting. For co-immunoprecipitation of endogenous ARL6IP1 and FAM134B, a confluent 15 cm dish of Arl6ip1 KO or WT MEFs was collected. Lysates were cleared by centrifugation at 12,000g for 10 min and incubated with the ARL6IP1 primary antibody at 4 °C overnight. Protein A agarose beads (Roche, 11719408001) were added and incubated at 4 °C for 4 h. Beads were then washed three times with lysis buffer, re-suspended in Laemmli buffer and boiled. Supernatants were analysed by SDS–PAGE and immunoblotting.

Modelling and simulations of ARL6IP1

The atomic model of human ARL6IP1 was built using the AI-based AlphaFold (v.2) program40. Five models were constructed, and the top-ranked model was chosen as it had maximal overlap with predicted secondary structures and consensus transmembrane topology, a higher pLDDT score and a relatively lower predicted alignment error (AF confidence measure).

Coarse-grained (CG) molecular dynamics simulations were performed using the MARTINI model (v.2.2)41,42. CG models of ubiquitinated and non-ubiquitinated versions were built by using martinize.py43. DSSP assignments were used to generate backbone restraints to preserve local secondary structure44,45. In the ubiquitinated protein (ARL6IP1-K96-Ub), the iso-peptide bond between K96 and the terminal glycine (G76) of ubiquitin (Protein Data Bank identifier 1UBQ) was modelled by modifying the side chain lysine bead (SC2/+1) into a neutral backbone bead (BB/0) and restraining the distance between the terminal bead of ubiquitin and the lysine side chain to 0.35 nm with a force constant of k = 1,250 kJ (mol nm2)–1. CG protein models were embedded into POPC (16:0-18:1 PC) bilayers spanning the xy plane of a periodic simulation box (20 × 20 × 20 nm3) solvated with CG-water containing 150 mM NaCl using the insane.py script43. All systems were first energy minimized and then equilibrated using the Berendsen thermostat46 and barostat47 along with position restraints on protein backbone beads followed by production runs with a 20 fs time step for a total of 10 μs. The system temperature and pressure were maintained at 310 K and 1 atm with the velocity rescaling thermostat48 and the semi-isotropic Parrinello–Rahman barostat49, respectively. All simulations were performed using gromacs (v.2019.3)50,51.

Statistical analysis

All experiments were performed in at least three independent biological replicates unless indicated otherwise. Data are presented as the mean ± s.e.m. unless indicated otherwise. Data analyses were performed using GraphPad Prism 9. For statistical analysis, raw data were analysed for normal distribution with the Kolmogorov–Smirnov test or with graphical analysis using Q-Q-plot. If appropriate, we used one-way analysis of variance (with Bonferroni post-hoc test unless indicated otherwise), repeated-measures two-way analysis of variance, Kruskal–Wallis H-test, Student’s t-test (two-sided unless indicated otherwise) or Mann–Whitney U-test. P values less than 0.05 were considered significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.