Changes in the cell's protein-making structure occur with aging and rely on its recycling process

Many ER proteins are enriched in specific structural and functional subdomains^20,28,36. To study ER dynamics during ageing, we developed transgenic C. elegans strains with labels capable of distinguishing rough or tubular ER. SEC-61.B is an essential component of the ER translocon enriched in rough ER sheets²⁸. Genomic green fluorescent protein (GFP) fusion produced a native marker, GFP::SEC-61.B, with enrichment in rough ER and relative exclusion from established smooth ER subdomains, such as muscle sarcoplasmic reticulum (Extended Data Fig. 1a). Importantly, the use of native markers circumvents the loss of subdomain fidelity associated with overexpressed and/or heterologous markers^28,37 (Extended Data Fig. 1b). To visualize tubular ER subdomains, we inserted mKate2 at the carboxy terminus of the sole C. elegans reticulon RET-1. Among the most highly enriched smooth ER proteins²⁰, reticulons are conserved hairpin-domain proteins that stabilize membrane curvature in ER tubules and sheet edges²⁹. Super-resolution imaging of these endogenous GFP::SEC-61.B and RET-1::mKate2 proteins revealed canonical SEC-61.B-enriched rough ER sheets linked by a RET-1-enriched network of tubules (Fig. 1a). To further validate the labelling, we swapped fluorophores on RET-1 from mKate2 to GFP and combined this with an alternative translocon subunit³⁸, TRAP-1::mCherry. Super-resolution imaging of trap-1::mCherry;ret-1::GFP animals revealed sheet structures bearing distinct enrichment of RET-1 on tubules and sheet edges, with TRAP-1 concentrated on the sheet faces (Fig. 1b). When co-labelled, SEC-61.B and TRAP-1 exhibited strong co-localization (Extended Data Fig. 1c,d).

Interestingly, the relative abundance of SEC-61.B or RET-1 labelling in different tissues also reflected expected rough or smooth ER specialization (Fig. 1c). Enrichment of smooth ER tubules, observed as high levels of RET-1, was noted in neuronal projections, muscle sarcoplasmic reticulum and the smooth muscle-like spermathecae, whereas the hypodermis and intestine, which execute robust collagen and lipoprotein secretion, were highly enriched for SEC-61.B-labelled rough ER³⁶ (Fig. 1c). Together, these results provide support for native labelling of these ER proteins as a reliable platform for monitoring ER subdomain dynamics in vivo across tissues in C. elegans.

To investigate whether ER morphology is dynamic across lifespan, we focused first on the hypodermis, a thin and metabolically flexible cell type well-suited for visualizing fine ER structures in vivo. In young adults, SEC-61.B marked a dense network of cisternal structures extending throughout the cell, with relatively few resolvable RET-1-enriched tubules (Fig. 2a). We then aged animals to day 7 of adulthood, a stage coinciding with the onset of age-dependent functional decline. Consistent with previous descriptions of ageing and progeria models^39,40, the SEC-61.B-marked nuclear envelope developed deep invaginations and other structural distortions at this age (Fig. 2a). However, the peripheral ER network was also markedly remodelled (Fig. 2a–c), despite being relatively unexplored in ageing contexts. Chief among the changes was a striking decline in the total amount of ER but the morphology was also altered, with SEC-61.B localizing to small sporadic clusters connected by a largely tubular network (Fig. 2a–c).

To quantify these changes and avoid co-expression artifacts, we imaged single-tagged GFP::SEC-61.B and RET-1::GFP animals individually (Extended Data Fig. 2a–h). Both markers showed substantial declines in intensity (approximately 70%) and footprint, or area of the cell occupied by the ER (25–30%), alongside a twofold increase in the perimeter-area ratio (PAR; Extended Data Fig. 2c–h). These findings indicate reduced organelle volume and content with age combined with a shift towards a more tubular network. We confirmed these trends independently of fluorescence imaging using epitope tags and immunoblotting in whole-animal lysates (Extended Data Fig. 2i–l). Interestingly, these blots revealed a more pronounced decline in RET-1 levels across the organism than for SEC-61.B. This result probably reflects certain tissue-specific contributions to the overall protein levels in whole-animal lysates (compare RET-1 and SEC-61.B levels in the intestine, discussed in the next section) but the low levels of RET-1 still raised the question of how a largely tubular network would be supported by tubule-promoting proteins in aged animals. We thus tested the complementary ER tubulating factor in C. elegans, YOP-1 (ref. ⁴¹). In contrast to RET-1, the levels of YOP-1 remained largely stable (Extended Data Fig. 2m,n), indicating potential compensation between tubulating factors. Finally, we investigated whether luminal, widely distributed ER proteins also exhibit similar age-dynamics. The GRP78 (also known as BiP) orthologues HSP-3::mScarlet and HSP-4::mScarlet exhibited similar expression and localization trends with age (Extended Data Fig. 3a–d). On the other hand, the levels of the mitochondrial outer membrane marker TOMM-20::mCherry remained stable, demonstrating the specificity of this dynamic to the ER (Extended Data Fig. 3e,f). Collectively these results provide support for a major decline in ER mass while highlighting that individual ER-shaping factors are differentially impacted by ageing.

Next, we tracked the kinetics and ultrastructural impacts of ER remodelling. Daily imaging of GFP::SEC-61.B animals revealed a 36% decline in intensity over the first three days of adulthood, revealing ER remodelling to be an early stage transition (Fig. 2d,e). We also imaged the ER in ageing wild-type animals via transmission electron microscopy (TEM). Consistent with fluorescence imaging, young animals exhibited densely packed stacks of rough ER sheets that gave way to sparse tubular networks in the hypodermis with a reduced footprint and increased PAR (Fig. 2f–h and Extended Data Fig. 4a–d).

These age-dependent morphological changes may correspond to functional shifts from proteostasis of the rough ER sheets to those that are more tubule-associated, such as lipid metabolism². Notably, this potential transition would be consistent with widely reported declines in global protein synthesis with age^17,42. To explore this model, we mined recent proteomic datasets examining the age-dependent proteome of C. elegans across tissues⁴³. We identified ER-resident proteins associated with either proteostasis or lipid metabolism and compared their levels in young and aged animals (Fig. 2i–k and Supplementary Table 1). Although the proteostasis network of the ER undergoes broadscale decline, mirroring the loss of rough ER (Fig. 2i,j), proteins involved in lipid metabolism generally stayed consistent with roughly half increasing with age (Fig. 2i,k). We conclude that ER morphological shifts mirror a functional shift from proteostasis to lipid metabolism based on the declines in total ER abundance and shift in ER-structure function from tightly packed stacks of rough ER cisternae towards diffuse tubular networks.

We also aimed to determine which age-dependent changes in the ER network might be generalizable across cell types. Consistent with the secretory functions of the intestine, intestinal cells possess dense, rough ER networks³⁶ (Fig. 1c). Similar to the hypodermis, the intestine experienced dramatic loss of SEC-61.B at the protein level, reduction in ER footprint and an increase in PAR (Fig. 3a–d). Both the loss of ER mass and sheet-like structures was supported by TEM imaging (Extended Data Fig. 4e,f). Furthermore, we found that male worms exhibited similar age-dependent declines in SEC-61.B intensity, footprint and morphology in both the hypodermis (Fig. 3m–p) and intestine (Extended Data Fig. 5a–d), revealing that ER loss is independent of egg production⁴⁴. Finally, muscle cells (Fig. 3e–h) and neurons (Fig. 3i–l) also exhibited declines in SEC-61.B intensity, ER footprint and shifts in ER morphology, as measured in the muscle belly and soma, respectively. These results reveal that ER remodelling occurs across most major tissue types in C. elegans.

Although aspects of age-dependent ER dynamics seem to be generalizable across many cell types, we also observed tissue-specific changes. For example, RET-1::GFP labels the young intestinal ER but is virtually undetectable in aged intestine (Extended Data Fig. 5e,f), exhibiting a more pronounced decline than SEC-61.B in this tissue. This near-complete loss of RET-1 in the intestine probably explains why the RET-1 western blots measured from whole animals decline more sharply than SEC-61.B (Extended Data Fig. 2). In addition, muscle cells and neurons each harbour specialized smooth ER subdomains within anatomically distinct regions of the cell: the myofilament lattice and neurites, respectively^36,45. Providing support for a model where rough ER subdomains are preferentially lost during ageing, these exclusively smooth ER subdomains seem to be resistant to age-related ER declines, with an approximate 15% decline in RET-1::GFP intensity and 8% decline in sarcoplasmic-reticulum footprint (Extended Data Fig. 5g–i) compared with declines of approximately 55% and 22% in the ER of the muscle belly (Fig. 3e–h). Contrasting every other tissue, neurite ER exhibits notable increases in RET-1 levels during ageing (Extended Data Fig. 5j–l). We observed an approximately twofold increase in RET-1::wrmScarlet intensity with age in representative sensory (ALN and PLM) and motor (DA7) neurons, whereas the cytosolic GFP levels remained static, indicating that the increases in RET-1 are independent of broadscale neuronal changes (Extended Data Fig. 5k,l).

Collectively, these results highlight that age-dependent ER remodelling is a ubiquitous phenomenon in C. elegans but can take distinct forms depending on the tissue. Although a decline in ER volume is the most generalizable effect, notable exceptions such as neurites exist. Furthermore, multiple major metabolic tissues exhibit structural changes consistent with preferential turnover of rough ER and associated sheet-like morphologies during ageing.

We next set out to determine whether age-onset remodelling of ER is evolutionarily conserved by examining eukaryotes ranging from yeast to mammals. First, the Tabula Muris Senis project^46,47 recently performed RNA sequencing analysis of diverse mouse tissues and revealed tissue-specific and animal-wide gene expression trajectories at the messenger RNA level. When gene expression changes were averaged across tissues to identify universal ageing signatures, one of the most strongly downregulated clusters was enriched for genes functioning in ‘protein processing in the ER’ (KEGG mmu0414; Fig. 4a).

Furthermore, a parallel atlas of proteomic changes across several mouse tissues during ageing enabled us to determine whether these mRNA trajectories correspond with protein-level remodelling⁴⁸. To mine this dataset of samples taken from whole-tissue lysates, we extracted data from the same set of genes involved in ER protein processing identified by Schaum and colleagues⁴⁶, comparing young (6 months) and aged (30 months) animals. We found consistent downregulation of these genes at the protein level in multiple mouse tissues, most notably skin, liver and kidney (Fig. 4b). Unbiased gene ontology enrichment analysis of these complete proteomic datasets additionally revealed another signature of rough ER remodelling focused on the translocon and protein targeting to the ER, again revealing the strongest and most consistent downregulation in skin, liver and kidney (Fig. 4c). Given that mitochondrial dysfunction is an established hallmark of ageing, we next extracted genes annotated for localization to either the mitochondria or ER to compare the scope and magnitude of changes, focusing on the skin, which showed the most consistent downregulation of ER translocon factors. The mitochondrial and ER proteomes both exhibited substantial progressive alterations relative to the complete cellular proteome (P = 3.21 × 10⁻⁷ and 9.11 × 10⁻⁸, respectively) and the downregulation in each case was similar in scope and magnitude (47% ER proteins downregulated versus 45% of mitochondrial proteins with false detection rate (FDR) < 0.05; Fig. 4d,e). On examination of the categories of proteins that are either down- or upregulated in the ER, we again found notable similarities to C. elegans. The downregulated proteins were enriched for functions in secretion, protein metabolism and ER organization, whereas the smaller group of upregulated genes were enriched for roles in lipid metabolism (Fig. 4f,g). Although the proteomes of the liver and kidney revealed smaller-scale downregulation of the ER proteome, the downregulated proteins were similarly enriched for roles in secretion (Extended Data Fig. 5m,n). Intriguingly, ER proteins with roles in autophagy were consistently upregulated in these tissues (Extended Data Fig. 5m,n). Overall, these results uncover the downregulation of rough ER-associated protein synthesis and secretion as a widespread aspect of the ageing process in mammals.

Finally, we aimed to visualize ER network changes with age in other species. Early neuropathological studies previously suggested that outstanding characteristics of the ageing rodent brain include a decrease in ribosomes and progressive ‘disorganization’ of ER cisternae in neurons^49,50, leading us to investigate neurons in mouse brain sections by scanning electron microscopy. We examined cortical neurons of three- and 18-month-old mice, and confirmed statistically significant declines in ER volume with age, even at this relatively early stage of ageing (Fig. 4h,i). Surprisingly, we did not consistently identify stereotypical stacked cisternae even in young animals and thus, were unable to confirm previous observations of cisternal disorganization at this particular age (PAR unchanged, P = 0.71). Finally, we also examined the ER of chronologically aged yeast cells using fluorescence microscopy. We labelled distinct ER proteins known to localize broadly across ER subdomains, Sec61 and Elo3. Both exhibited early and striking declines in intensity and footprint (Fig. 4j–q), overall mirroring findings in C. elegans, although ultrastructural analysis of the uniquely organized yeast ER requires further study. Together, these results demonstrate that functional and structural ER remodelling during ageing is a conserved phenomenon, however, extrapolation across diverse mammalian types necessitates deeper dedicated analyses.

Next, we set out to identify mechanisms promoting ER remodelling during ageing. Given the pronounced loss of both ER protein and membrane mass during ageing, we reasoned that autophagic and/or lysosomal degradation processes may be involved. Current models suggest that multiple distinct pathways act in concert to deliver ER components to lysosomes: (1) macro-ER-phagy, which requires both canonical autophagy initiation factors like ULK1/Atg1/unc-51 and LC3/GABARAP/Atg8 lipidation; (2) micro-ER-phagy, independent of both ULK1 and Atg8; (3) RecovER-phagy, a micro-ER-phagy variant that utilizes Atg8 and (4) vesicular ER-to-lysosome pathways, which also generally depend on Atg8 lipidation^31,32. To help distinguish between these routes, we fed animals double-stranded RNA (dsRNA) targeting either Atg1/unc-51 or Atg8/lgg-1. Autophagy inhibition via depletion of Atg1/unc-51 and Atg8/lgg-1 had relatively small effects on ER size and shape in young adults, indicating that autophagy-mediated turnover plays a limited role in shaping the ER through development (Fig. 5a–d). However, knockdown of Atg1/unc-51 strongly suppressed age-onset changes in SEC-61.B protein levels, footprint and PAR (Fig. 5a–d), and this suppression of age-effects was virtually complete during Atg8/lgg-1 impairment (Fig. 5a–d). These trends generally remained consistent between ER markers (Extended Data Fig. 6a–d) and tissues (Extended Data Fig. 6e–l), although Atg8/lgg-1 RNA interference (RNAi) resulted in modest elevations in RET-1 levels in young animals (Extended Data Fig. 6a,b) and especially strong elevation of RET-1 levels in the aged intestine (Extended Data Fig. 6i,j). Notably, ULK1/unc-51 knockdown also caused the appearance of expanded sheet-like structures (Fig. 5a and Extended Data Fig. 6a), which were not apparent in the Atg8/lgg-1 animals. Together, these findings suggest that macro-ER-phagy, dependent on ULK1/UNC-51, is a primary route for ER-sheet turnover, whereas Atg8-dependent ULK1-independent (vesicular and RecovER-phagy) routes contribute to bulk ER degradation and help balance network morphology.

The ER is an important source of membrane for autophagosome formation as well as a potential target of autophagosomes via ER-phagy, and both of these roles could potentially result in reduced ER mass^31,51,52. To confirm the ER itself is targeted as cargo for ER-phagy, we employed TRAP-1::mCherry. Unlike GFP and mKate2 labels, mCherry is resistant to lysosomal acidity and degradation, enabling analysis of lysosomal targeting⁵³. In young adult animals, TRAP-1::mCherry and GFP::SEC-61.B labels reveal the same ER network organization (Extended Data Fig. 1c,d). However, TRAP-1::mCherry uniquely begins accumulating in distinct puncta by day 3 of adulthood (Fig. 5e). Consistent with the age-onset formation of these puncta, we observed low co-localization of TRAP-1::mCherry with lysosomal membrane marker, SCAV-3::GFP, in animals on the first day of adulthood and a dramatic increase of TRAP-1::mCherry in SCAV-3⁺ lysosomes by day 3 (Fig. 5e,g). Atg8/lgg-1 depletion prevented the formation of TRAP-1 puncta and suppressed lysosomal targeting (Fig. 5f, g). Furthermore, the role of autophagy in driving age-onset ER loss is evolutionarily conserved, as the decline of Sec61 in yeast correlated with re-localization to the vacuole via an Atg1-dependent mechanism (Fig. 5h,i). We confirmed the appearance of ER components inside lysosomes through both three-dimensional (3D) fluorescence imaging (Fig. 5j) and TEM (Fig. 5k) of C. elegans. Overall, these results are consistent with a model where activation of ER-phagy in adulthood promotes turnover and remodelling of the ER.

We next exploited C. elegans to investigate whether established paradigms of lifespan extension involve ER remodelling by ER-phagy. We selected a panel of mechanistically diverse interventions, that is, reduced insulin/insulin-like signalling (daf-2(e1370)), reduced mechanistic target of rapamycin (mTOR) signalling (raga-1(ok386)), germline removal (glp-1(e2141)) and inhibition of translation (dsRNA targeting ifg-1, ifg-1(RNAi)). Consistent with ER remodelling acting as an adaptive step in ageing, these life-extending interventions universally induced dramatic ER remodelling, even at the onset of adulthood (Fig. 6a–e). All interventions promoted substantial reductions in GFP::SEC-61.B intensity and footprint as well as increased PAR (Fig. 5f–h). Although these gross trends resemble changes during normal ageing, we also observed aspects unique to long-lived animals. In contrast to the evenly distributed rough ER sheets under normal conditions (Fig. 6a), cisternal ER networks were condensed into perinuclear regions with large peripheral regions devoid of translocon-labelled ER instead occupied by sparse tubular networks (Fig. 6b–e). This organization suggests enhanced functional compartmentalization of ER networks into distinct sheet- versus tubule-filled regions of the cell, which may correspond with remodelling of other organelle networks and the subcellular architecture more globally. We found that GFP::SEC-61.B intensity, footprint and PAR were restored to levels comparable to controls in animals fed Atg8/lgg-1 dsRNA to impair autophagy (Extended Data Fig. 7a–d), indicating that ER remodelling in long-lived animals involves ER-phagy. However, the regionalized clustering of ER sheets persisted during Atg8/lgg-1 inhibition (Extended Data Fig. 7a), indicating that this aspect of ER network remodelling in long-lived animals is independent of autophagy. Finally, ageing in these conditions resulted in additional declines in ER (Extended Data Fig. 7e–h), consistent with a model where life-extending interventions promote ER-phagy throughout life. Overall, these findings highlight ER network dynamics as an aspect of both life-extending interventions and the normal ageing process.

Mediators of selective ER-phagy are only beginning to be explored in C. elegans⁵⁴, so we next aimed to elucidate ER-centric processes and molecular mediators linked to turnover during ageing. We thus performed a candidate screen of 35 genes that are orthologues of established ER-phagy mediators or ER-resident proteins harbouring an LIR motif (Supplementary Table 2). As context for our screen, we employed mTOR/raga-1-pathway mutants due to their robust ER depletion even in young animals (Fig. 6c) and the utility of TOR impairment in previous genetic screens for ER-phagy receptors^34,55,56. Surprisingly, examination of RET-1::GFP in this background demonstrated that neither canonical ER-phagy receptor orthologues nor the COPII components sec-23, sec-24.1, sec-24.2 and sec-31 (ref. ⁵⁷) had any discernible effect on ER remodelling in the raga-1 mutants. However, RNAi of tmem-131 (tmem-131(RNAi)), a conserved ER transmembrane protein with a predicted LIR motif, resulted in amelioration of ER loss and an expansion of ER sheets (Fig. 7a). We thus investigated whether tmem-131 promotes turnover of ER during ageing, and found that tmem-131(RNAi) restored SEC-61.B levels and the ER footprint of aged animals to youthful levels (Fig. 7b–e). These results indicate that tmem-131 is required for ER clearance in ageing contexts.

The luminal N terminus of TMEM-131 bears a procollagen binding domain, and the cytoplasmic tail interacts with the transport protein particle (TRAPP) III complex via TRAPPC8/trpp-8 to promote COPII trafficking of collagen cargoes⁵⁸. However, bioinformatic predictions also indicate a cytosolic domain architecture reminiscent of established ER-phagy receptors^33,59, including intrinsically disordered regions^60,61 and a conserved LIR motif⁶² (residues 1517–1520) outside the TRAPPIII-binding region⁵⁸ (Fig. 7d,e). Furthermore, TMEM-131 was recently proposed as a candidate autophagy receptor in human cells based on proteomic profiling of autophagosomes⁶³. Considering procollagens are an established substrate for both basal and stress-induced ER-phagy^64,65, these features suggest TMEM-131 could link luminal ER collagen state to ER abundance (Fig. 7d).

We first tested whether collagen secretion defects could recapitulate the ER phenotypes of tmem-131. Although knockdown of TMEM-131’s binding partner, TRAPPC8/trpp-8, replicates the collagen export defects of tmem-131 (ref. ⁵⁸), trpp-8(RNAi) failed to rescue ER mass similarly to tmem-131 or Atg8/lgg-1 depletion (Fig. 7b,c and Extended Data Fig. 8a,b). In addition, COPII components play key roles in collagen trafficking, but their impairment did not produce a notable rescue of the ER (Extended Data Fig. 8c). Knockdown of the collagen metallopeptidase dpy-31 (ref. ⁶⁶) also contrasted sharply with tmem-131, producing a mild reduction in SEC-61.B levels in young animals and a small shift in ER PAR in aged animals (Extended Data Fig. 8d–g). These results collectively argue against a model where general collagen export defects are sufficient to explain the expanded ER mass in ageing tmem-131 animals.

The data instead illustrate a more specialized role for TMEM-131. Aligned with previous reports that procollagen stress is relieved by micro-ER-phagy and ER-to-lysosome-associated degradation-like ER-phagy^64,65, impairment of collagen export via either tmem-131(RNAi) or trpp-8(RNAi) elevated baseline TRAP-1 co-localization with lysosomes (Fig. 7f and Extended Data Fig. 8h). Consistent with this, immunoblots of GFP::LGG-1 revealed a pattern in tmem-131 animals consistent with elevated autophagic flux, including reductions in total and lipidated LGG-1 with a twofold increase in free proteolyzed GFP (Extended Data Fig. 8i–l). This suggests that bulk autophagy is activated for procollagen quality control in the absence of tmem-131. However, despite the elevated baseline, tmem-131 blunted the age-induced increase in ER trafficking to lysosomes to a similar extent as Atg8/lgg-1 depletion (Fig. 7f,g) and reduced overall lysosomal trafficking of ER relative to both aged control and trpp-8(RNAi) worms (Extended Data Fig. 8h). This distinction suggests that TMEM-131 plays a specific role in the age-onset macro-ER-phagy that depletes ER volume, which can be at least partly uncoupled from procollagen quality control.

Providing support for the potential for direct recruitment of autophagy machineries, pull-down assays using the TMEM-131 cytosolic domain revealed clear binding with the human Atg8/lgg-1 orthologue GABARAP (Extended Data Fig. 8l). Mutation of the bioinformatically identified LIR motif reduced human GABARAP binding by 25–30% (Extended Data Fig. 8m,n). Although single-LIR receptors generally exhibit stronger dependence, ER-phagy receptors containing multiple LIR motifs (for example, Rtnl3) can form multivalent interactions with the autophagy machinery and exhibit similar partial effects⁶⁷. Following this result, we indeed discovered at least one additional LIR motif in the cytosolic domain on manual examination (1559: FMNL). To investigate whether TMEM-131 behaves similarly to canonical ER-phagy receptors in vivo, we labelled and imaged endogenous TMEM-131 under ER-phagy-inducing starvation conditions. At baseline, TMEM-131 seemed to localize throughout the ER, forming many foci that did not correspond well with autophagosomes and probably relate to roles in collagen secretion (Extended Data Fig. 8o). Starvation resulted in a clear decline in TMEM-131 levels and footprint, reflecting a global loss of ER, yet TMEM-131 co-localization with autophagosomal puncta increased (Extended Data Fig. 8o–q). Atg8/LGG-1 puncta were consistently present at the boundaries of TMEM-131-labelled ER, including examples of fully co-localized puncta indicative of autophagosomal recruitment (Extended Data Fig. 8o, inset). Finally, consistent with TMEM-131 regulating selective turnover of the ER, tmem-131 depletion failed to promote an increase in mitochondrial volume during ageing (Extended Data Fig. 8r,s). Notably, age-associated mitochondrial fragmentation seemed to be accelerated by tmem-131(RNAi), which may suggest that early stage ER remodelling can promote downstream consequences for other organelle networks. Collectively, these data highlight the requirement of TMEM-131 for age-onset ER turnover. Structural similarities to ER-phagy receptors, evidence for Atg8 binding and recruitment to autophagosomes suggest that TMEM-131 may play direct roles in the targeting of ER by the autophagy machinery. Given that declines in global collagen and matrix secretion are a conserved facet of ageing⁶⁸, our data may point to a model where TMEM-131 promotes ER turnover when collagen clients are depleted, thus coordinating ER demand with supply.

Because we observed ER-phagy-based remodelling of the ER across many tissue types in the worm, we investigated whether TMEM-131’s role extends to other tissues. In contrast to the hypodermis, however, we observed relatively little effect of tmem-131(RNAi) on intestinal ER during ageing (Fig. 7h,i and Extended Data Fig. 9a,b). We therefore returned to the panel of candidate ER-phagy receptors that we initially screened in the hypodermis to test whether an alternate mediator may act in the intestine. We again recovered a single hit from our panel that in this case seemed to fully rescue the age-dependent loss of intestinal ER: the highly conserved, central player in UPR inositol-requiring enzyme (IRE)-1 (Fig. 7h,i and Extended Data Fig. 9a,b). Ire-1 had no impact on the age-dependent changes of the hypodermal ER (Extended Data Fig. 9c–f), just as tmem-131 had little impact on the intestine (Fig. 7h,i). Follow-up experiments revealed that knockdown of the IRE-1 target, the soluble transcription factor XBP-1, also rescues ER loss during ageing in the intestine (Extended Data Fig. 9g–j). Conversely, parallel branches of the UPR, atf-6 and PERK/pek-1, did not demonstrate roles in age-induced ER remodelling (Extended Data Fig. 10). Collectively, these results indicate that the IRE-1–XBP-1 axis promotes age-dependent ER turnover and remodelling, probably through downstream signalling rather than generalized ER stress or a direct role for IRE-1 as an ER-phagy receptor. Together, these findings indicate that the physiological mediators of ER-phagy are not universal but may instead align with the predominant functions of the ER in each tissue. Here TMEM-131 bridges collagen secretion and ER-phagy in the hypodermis, which is responsible for supporting the collagen-based cuticle, whereas ire-1–xbp-1 performs similarly in the intestine, a central hub of UPR signalling in C. elegans¹.

To investigate causal roles of ER-phagy in ageing and lifespan, we turned to yeast and C. elegans models. In yeast, previously validated ER-phagy-specific adaptors allow us to experimentally decouple ER-phagy from bulk macro-autophagy. Given that mTOR inhibition exhibited the strongest ER remodelling phenotype in C. elegans (Fig. 6) and robustly induces both ER-phagy and lifespan extension in yeast^34,56,69,70, we investigated whether ER-phagy mediators are important for longevity in rapamycin-treated cells. Alongside macro-autophagy-deficient atg5-deletion mutants (Δatg5) as positive controls, we utilized deletion mutants of the canonical receptor for ER-phagy of peripheral ER (Δatg40)⁵⁵ as well as deletion mutants of sec66 (Δsec66), which we selected based on its selective localization in ER membranes and its experimentally verified impairment of ER-phagy but not macro-autophagy³⁴. Strikingly, the Δsec66 mutants phenocopied the complete suppression of lifespan of macro-autophagy-deficient Δatg5 mutants and the Δatg40 mutants similarly exhibited a blunted rapamycin effect (Fig. 8a–e). We also tested the role of the tmem-131 and ire-1–xbp-1 ER-phagy-inducing pathways in long-lived C. elegans harbouring mutations in the mTOR activator raga-1. Given the tissue-specific effects of both tmem-131 and ire-1–xbp-1 on ER remodelling, we suspected that multitissue ER-phagy impairment via the combined loss of both pathways would exhibit the strongest impact on lifespan. Consistent with this prediction, we observed a statistically significant suppression of longevity when either tmem-131 or xbp-1 were impaired alone (Fig. 8f–h), with a much more complete suppression when impaired in combination (Fig. 8i). The combined loss of tmem-131 and xbp-1 resulted in a notably short-lived animal regardless of mTOR/raga-1 status. Besides emphasising the importance of ER-phagy in animal health, this dramatic result may also reflect secondary interactions between the two pathways⁵⁸. Collectively, our results indicate that ER-phagy plays a key role in mTOR-dependent longevity.

Worms were grown and maintained at 20 °C on 6-cm nematode growth media (NGM) plates seeded with Escherichia coli (OP50-1). Single E. coli colonies were cultured overnight in LB medium at 37 °C; each NGM plate was then seeded with 100 µl liquid culture was and cultured for two days at room temperature before use. For all experiments, worms were synchronized via timed egg lay. In experiments involving glp-1 mutants, glp-1 mutants and controls were moved to 25 °C 24 h after synchronization and incubated for 24 h to impair germline development. For experiments involving males, synchronized egg lays were conducted with mated worms and early larval stage L4 progeny were segregated by sex for subsequent experiments. For starvation experiments, young L4 animals were transferred to unseeded NGM plates containing carbenicillin (100 µg ml⁻¹) to prevent bacterial growth. A complete list of worm strains used in this study is in Supplementary Table 4.

Gene-specific RNAi feeding clones (HT115) were obtained from the Ahringer RNAi Library (Source Bioscience), sequence-verified and cultured as described earlier for the OP50-1 strain, with the exception of the addition of 100 µg ml⁻¹ carbenicillin (final concentration) to the LB medium and NGM plates. At least an hour before use, control and experimental lawns were spotted with 100 µl IPTG solution (0.1 M isopropyl β-D-1-thiogalactopyranoside (IPTG), 100 µg ml⁻¹ carbenicillin and 12.5 µg ml⁻¹ tetracycline) to induce dsRNA expression. For ifg-1 RNAi, feeding began at the L4 developmental stage. For the lgg-1 and unc-51 RNAi ageing experiments, the parental generation was also fed dsRNA from hatch for maximum penetrance. Otherwise, the worms were fed dsRNA from hatch.

Synchronized worms were raised as described earlier. For each condition, an agarose mounting pad—10% agarose in M9 buffer was prepared before imaging by briefly heating the agarose solution and flattening it between two slides (#1.5, VWR). Approximately 3 µl of 0.1 µm polybead microsphere suspension (Polysciences) was added to the pad, the worms were manually picked into the bead solution and a coverslip (#1.5, VWR) was added to immobilize the worms. For neuron-specific imaging, the worms were mounted instead on a 2% agarose pad in 6 µl of 10 mM levamisole (Fisher Scientific). Imaging was performed on an Eclipse Ti2 inverted microscope with a Yokogawa CSU-W1 spinning disc, a Plan Apo λ objective (×10/0.45 or ×100/1.45; Nikon) and a Prime 95B sCMOS camera (Teledyne Photometrics). Type B or LDF immersion oils (Cargille Labs) were used, always with consistent oil type across all replicates. Fluorophores were excited at 488 and 561 nm (dichroic mirror Di01-T405/488/568/647-13×15×0.5 (Semrock)), with respective ET525/36 m and ET605/52 m emission filters (Chroma). In addition, DIC images (89101m emission filter) were captured for each fluorescence image. For native TMEM-131::mKate2 imaging, 4× signal integration was used to compensate for low expression/intensity levels. Super-resolution images were obtained using an LSM980 microscope (Zeiss) with a ×63/1.40 Plan-Apochromat objective and Airyscan 2 detector. Acquisition settings, including exposure times, were identical across all conditions in each experiment. For consistency, focal planes were selected at nuclear midlines in the same cells across all conditions, specifically the anterior region of the hypodermis (hyp7), the first intestinal rings (int1), body-wall muscles and the ALN neuron. A plane was selected within the myofilament lattice to image the sarcoplasmic reticulum. For neural ER projections, imaging was acquired via Z-stack, with analysis performed on maximum intensity projections of the ALM, DA7 and PLM.

For imaging experiments, fluorescently labelled SEY6210 yeast cultures were cultured in synthetic complete dextrose medium at 30 °C in a gyratory shaker (220 r.p.m.) to mid-log phase (0 h). At each experimental time point, samples were removed for imaging and the culture was then maintained for subsequent time points. A chronological lifespan assay was performed as previously described⁸¹. Control and mutant yeast strains from the BY4741 gene-knockout library were cultured to mid-log phase in Longo’s Aging synthetic complete dextrose medium buffered at pH 6.0 before being subcultured into fresh medium with a starting optical density at 600 nm of 0.1 and treated with either 15 nM rapamycin (in 95% ethanol) or an equivalent volume of 95% ethanol. The cultures were then grown to post-replicative phase after incubation for three days at 30 °C in a gyratory shaker (220 r.p.m.). For viability measurements, the cell samples were diluted in sterile water (10:3 dilution), plated on YPD plates and incubated at 30 °C. Following a three-day incubation, the colony-forming units were counted, with subsequent colony-forming-unit counts normalized as a percentage of the day 3 colony-forming-unit count. For each condition, the AUC value from three independent biological replicates was calculated as the lifespan statistic for comparison⁶⁹. The AUC values were normalized within each genotype as the ratio between the rapamycin- and ethanol-treated groups. A complete list of yeast strains used in this study is provided in Supplementary Table 4.

Before imaging, yeast were concentrated via centrifugation at 3,500g, pipetted onto a slide and immobilized with a cover slip. Images were acquired with a DeltaVision Elite Imaging system (Olympus IX-71 inverted microscope, Olympus ×100 oil objective (1.4 numerical aperture) and DV Elite sCMOS camera; GE Healthcare). Images were obtained in red (Alexa Fluor 594; 475 nm excitation, 523 nm emission), green (fluorescein isothiocyanate; 575 nm excitation, 632 nm emission), blue (4′,6-diamidino-2-phenylindole; 390 excitation, 435 emission) and DIC channels.

Super-resolution Airyscan 2 images were deconvolved using the ZEN Blue v.3.5 software (Zeiss). NIS-Elements Advanced Research (NIS-AR, Nikon) was used to process images for all other worm experiments. Before analysis, images were subjected to automated two-dimensional or 3D deconvolution and subsequent denoising via the Denoise.ai module. Rolling-ball background removal was performed on each colour channel with identical settings for all images in the experimental replicate. For categorical analysis of ER morphology, images were blinded and the JavaScript command ‘Math.random();’ was used within the GA3 module of NIS-AR to select three random regions within the hypodermal region of interest. Subregion morphology was categorized using the Qualitative Annotations plugin in Fiji. For ER intensity and morphological quantitation, regions of interest were manually drawn around the cell of interest, specifically excluding nuclei and autofluorescent gut granules. GA3 was used for threshold-based segmentation of the ER. For experiments in which marked variability in ER fluorescence precluded the use of constant thresholds, thresholds were manually set on blinded images. In experiments involving ageing of lysosomal resistant mCherry markers, an upper threshold was also applied for ER segmentation to exclude ER contained in lysosomes. Images for yeast experiments were deconvolved using softWoRx and the integrated density of fluorescence intensity units were normalized to the cell surface area using Fiji. Yeast ER footprint and PAR were calculated from maximum intensity projections. To enable visualization of ER in aged samples where fluorescence intensity was substantially lower, minimal modifications to brightness, contrast and ɣ were applied on representative images after analysis and always uniformly across all conditions.

Worms were collected from bacterial lawns and washed 3× with M9 buffer. The worm pellets were resuspended in 0.15 M sucrose prepared in M9 buffer and loaded into a 200-µm-deep well of an A-type carrier (Leica). The assembly was covered with the flat side of B-type carrier (Leica) and vitrified using a high-pressure freezing machine (Leica EM ICE). The frozen specimens were stored in liquid nitrogen until further processing with a modified freeze-substitution process. Briefly, freeze-substitution was performed using a cocktail of 0.5% glutaraldehyde and 0.1% tannic acid in acetone. The samples were transferred into an automatic freeze-substitution machine (AFS2, Leica) pre-cooled to −140 °C. They were then warmed to −90 °C over 3 h, followed by incubation at −90 °C for 114 h. Next, the samples were washed (30 min each) 4× with acetone. The final acetone wash was replaced with 2% osmium tetroxide in acetone. Following this, the samples were warmed to −20 °C over 12 h and maintained at −20 °C for an additional 10 h, after which they were warmed to 0 °C over 4 h. The samples were then transferred into an ice-filled bath and washed (30 min each) 4× with acetone and infiltrated with Spurr’s resin (Electron Microscopy Sciences). During the final steps of resin infiltration, individual worms were transferred from the carriers into resin moulds with a fine needle and the samples were cured at 60 °C for 72 h. Thin sections (70 nm) were cut using a Leica UC7 ultramicrotome and a Diatome diamond knife, positioned on grids and stained with uranyl acetate and Reynold’s lead citrate. These sections were then imaged on a Tecani T12 TEM operating at 100 keV using an AMT CMOS camera. The Amira software (Thermo Fisher Scientific) was used for image processing and analysis. Briefly, the hypodermal ER and cytosolic area were manually segmented with the freehand masking tool and the ER footprint and PAR were calculated.

Male FVB/NJ mice obtained from the Jackson Laboratory were housed in cages of five mice and maintained on a 12-h light–dark cycle with controlled temperature (21 °C) and humidity. The mice had ad libitum access to chow food (Rodent Diet 20 5053, PicoLab) and water. All animal procedures were approved by the Institutional Animal Care and Use Committee of Vanderbilt University (M2000086-00/01).

Brain tissue for scanning electron microscopy imaging was prepared as previously described. Briefly, one male mouse from each age group (6 and 18 months) was killed and perfused with 37 °C Ringer’s solution (0.79% NaCl, 0.038% KCl, 0.02% MgCl₂·6H₂O, 0.018% Na₂HPO₄, 0.125% NaHCO₃, 0.03% CaCl₂·2H₂O, 0.2% dextrose and 1,000 U heparin) for 30 s, followed by a transcardiac perfusion with fresh ice-cold 2.5% glutaraldehyde and 2% paraformaldehyde in 0.15 M sodium cacodylate at 10 ml min⁻¹ for 10 min. The brains were dissected and sliced along the sagittal plane with a vibratome to prepare 150-µm-thick sections and post-fixed overnight at 4 °C. The brain slices were washed in cold 0.15 M sodium cacodylate buffer and then fixed in 2% osmium tetroxide and 1.5% potassium ferrocyanide in 0.15 M sodium cacodylate buffer for 1 h at room temperature (approximately 23 °C). The samples were washed several times in double distilled water (ddH₂0), placed in a 0.5% thiocarbohydrazide solution for 30 min and then washed 5× in ddH₂0. Next, the slices were placed in a 2% aqueous osmium tetroxide solution for 1 h, thoroughly washed in ddH₂0, transferred to a 2% aqueous uranyl acetate solution and incubated at 4 °C overnight. The next day, the slices were washed with ddH₂0, placed into Walton’s lead aspartate solution for 30 min and baked at 60 °C in a bench-top oven. The baked samples were thoroughly washed with ddH₂0, followed by a serial dehydration series on ice using ice-cold 70%, 90%, 100% and 100% ethanol, followed by dry acetone (10 min per step). The brain slices were subsequently washed in 1:3, 1:1 and 3:1 Durcupan ACM:acetone solutions (12 h in each solution). Finally, the slices were incubated in three washes (24 h each) of 100% Durcupan ACM before being baked for 48 h at 65 °C for solidification.

Slices containing layer 2 motor cortical neurons were imaged on a Zeiss Gemini class scanning electron microscope. Images were blinded and one neural soma from the central region of each image was selected for analysis. The ER and somatic cytosolic area were manually segmented in NIS-AR and the ER footprint was calculated.

To generate extrachromosomal array wbmEx222[eft-3p::GFP::sec-61.b::unc-54 UTR], a C. elegans codon-optimized GFP(S65C,Q80R) sequence was fused to the native sec-61.b sequence in an expression vector to generate plasmid pKB22. The plasmid was column-purified and microinjected as previously described.

To generate ret-1::GFP and ret-1::mKate2, codon-optimized GFP(S65C,Q80R) and mKate2 sequences were amplified with an N-terminal GGGGS linker from the plasmids pKB51(GFP) and pKB52(mKate2) using a common forward primer (Fwd, 5′-GTGCTCCAGTCGCCGCTGAAGAGAAGAAGGATCAAGGTGGCGGAGGTTCTGG-3′) and specific reverse primers (Rev-GFP, 5′-CAATGGCAAAGTGTGTTCTTCTTTCAATCGATTTATTTGTATAGTTCATCCATG-3′ or Rev-mKate2, 5′-CAATGGCAAAGTGTGTTCTTCTTTCAATCGATTTAACGGTGTCCGAGCTTGGATG-3′; IDT). The repair templates were column-purified and injected as previously described with a guide (Dharmacon) against the following target sequence: 5′-AATGGCAAAGTGTGTTCGGA-3′. For ret-1::wrmScarlet11, a single-stranded-oligodeoxynucleotide repair template including the wrmScarlet11 sequence (5′-GTGCTCCAGTCGCCGCTGAAGAGAAGAAGGATCAAGGTGGCGGAGGTTCTTACACCGTCGTCGAGCAATACGAGAAGTCCGTCGCCCGTCACTGCACCGGAGGAATGGATGAGTTATACAAGTAAATCGATTGAAAGAAGAACACACTTTGCCATTGTTTTTC-3′) was synthesized (IDT). The repair template was injected with the same guide used for the GFP and mKate2 fusions. To generate the tmem-131::mKate2 allele, 2×GGGGS::mKate2 was amplified with from pKB52 via subsequent rounds of PCR using the primers: F1, 5′-AATGCCACAGGACACCGACAATGAAAACGACGAGAAGAACAATGGTGGCGGAGGTTCTGG-3′; R1, 5′-ATATAATGGAGAATGTGTTGTAGAATGGAATTATTAACGGTGTCCGAGCTTGGATG-3′; F2, 5′-CAACACAACAACCATCAACTTCTCAAATGCCACAGGACACCGAC-3′ and R2, 5′-CATATAGATTTGGATAGCAGATAAGTAGATATATAATGGAGAATGTGTTGTAG-3′ (IDT). The purified repair template was injected with guides targeting the following target sequences: 5′-TTCTCAAATGCCACAAGATA-3′ and 5′-CAGATAAGTAGATATAATAT-3′ (Dharmacon). All fusion strains were sequence-verified and outcrossed with N2 worms at least four times.

For neuron-specific RET-1 fluorescence, we first generated extrachromosomal array bugEx10[rab-3p::wrmSct1-10::unc-54 UTR+rab-3p::GFP::unc-54 UTR] through the cloning and subsequent co-injection of pBK01 (62 ng µl⁻¹) and pBK02 (53 ng µl⁻¹) into N2 worms. To generate plasmid pBK01(rab-3p::GFP::unc-54 UTR), the atf-6 sequence was excised from plasmid pKB41 via BamHI and AgeI digest, and the backbone was re-ligated after Klenow treatment. To generate plasmid pBK02(rab-3p::wrmSct1-10::unc-54 UTR), the wrmSct1-10::unc-54 UTR sequence was amplified from plasmid pJG100 with the primers Fwd, 5′-CCCGGGATGGTATCGAAGGGAGAAGC-3′ and Rev, 5′-ACTAGTCTTCCACTGAGCCTCAAA-3′ (IDT). The PCR product was ligated into a vector containing the rab-3p sequence following XmaI and SpeI digestion. After micro-injection, GFP⁺ worms were crossed with ret-1::wrmScarlet11 worms to generate the strain BUZ112(ret-1::wrmScarlet11;bugEx10).

The TMEM-131 transmembrane segment was identified using Deep TMHMM⁸², after which the coding sequence for the cytosolic C terminus (residues 1191–1831, ‘TMEM-131_c’) was amplified from N2 worm complementary DNA using the primers F, 5′-TATCTGGAAGGTGATCGTGCTATTGC-3′ and R, 5′-TTAGTTATTCTTTTCATCATTCTCGTTGTCCGTATC-3′ (IDT). Homology to plasmid pMal-c5e was added via PCR with the primers F, 5′-TACCGCATATGTCCATGGGCTATCTGGAAGGTGATCGTG-3′ and R, 5′-CTGAGCCTTTCGTTTTATTTGATTAGTTATTCTTTTCATCATTCTCGTTGTCC-3′ (IDT). The PCR product was column-purified (Invitrogen) and inserted into linearized custom 6×HIS::MBP pMal-c5e vector (NotI-HF/HindIII-HF) via Gibson assembly (NEB) to generate plasmid pED09. To generate a LIR mutant TMEM peptide recoding FVHV as SSSS (TMEM ∆LIR), plasmid pED09 was amplified with the primers F, 5′-GAGCTCCTCCCCACCAACTTCTTTAAC-3′ and R, 5′-GAGTCATCAGTTGGAGCTGACGTC-3′ (IDT). The PCR product was isolated via gel purification and treated with KLD enzyme mix (NEB) to degrade any remaining template and circularize the product as plasmid pED10. Both plasmids were sequence-verified and individually transformed into Rosetta2 (DE3) competent cells for protein expression.

PCR was used to amplify mCherry (pFA6a-mCherry-KanMX6), mNeonGreen (pFA6a-mNeonGreen-NATMX) and mTagBFP2 (pFA6a-mTagBFP2-TRP1) to create DNA tagging cassettes as previously described⁸³. The cassettes were gel-purified and transformed into SEY6210 yeast cells as previously described⁸⁴. C-terminal integration of the fluorescent tag into the endogenous coding sequence was confirmed by PCR and fluorescence microscopy.

Synchronized populations of at least 400 worms were raised as described earlier on 10-cm NGM plates seeded with 500 µl overnight OP50-1 cultures. To assist with ageing a large population, L4 worms expressing 3×FLAG::SEC-61.B were transferred to lawns spotted with 250 µl 5-fluoro-2′-deoxyuridine (1 mg ml⁻¹ solution) 24 h before use. For RET-1::GFP blotting, germline tumours were observed with extremely high levels of RET-1, necessitating a germlineless glp-1 background. Ret-1::gfp;glp-1 worms were heat shocked as described earlier. The worms were collected from bacterial lawns and washed 3× with M9; the worm pellets were then flash-frozen with liquid nitrogen and stored at −80 °C. They were then lysed at 4 °C by sonication in RIPA buffer (50 mM Tris, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate and 0.1% SDS, pH 7.5) containing protease inhibitors (Roche Applied Science). The lysates were centrifuged at 14,000g and 4 °C for 15 min and the supernatants were collected. Protein concentrations were determined using a Pierce BCA assay (Thermo Scientific). An equal mass of each sample was diluted in Laemmli Sample Buffer (BioRad) with 2-mercaptoethanol, heated (95 °C for 15 min) and resolved via SDS–PAGE electrophoresis.

Proteins were transferred to activated polyvinylidene fluoride membranes over 2 h at 4 °C. Ponceau (Sigma Aldrich) staining, imaging and washing was conducted as per the manufacturer’s recommendations. Target proteins were probed through the following steps: an overnight 4 °C incubation in blocking solution (5% non-fat milk in TBST solution), incubation in primary antibody solution (5% milk in TBST + antibody—anti-FLAG (1:1,500) or anti-GFP (1:1,000; Santa Cruz)) for 2 h at room temperature, three TBST washes (10 min each) at room temperature and incubation with secondary antibody solution (5% milk in TBST + antibody (horseradish peroxidase (HRP)-conjugated, anti-mouse; 1:10,000)) for 1 h at room temperature. After another three TBST washes (10 min) at room temperature, the membranes were incubated with HRP substrate (Pierce ECL western blotting substrate, Thermo Scientific) as per the manufacturer’s recommendations. All wash and incubation steps were performed with gentle nutation. Ponceau and ECL imaging were performed using an Amersham Imager 600 (GE Life Sciences).

GFP::LGG-1 blots included the following modifications: room temperature transfer using a Trans-Blot Turbo system (BioRad), blocking for 1 h at room temperature, overnight incubation with anti-GFP (Roche) primary antibody at 4 °C, HRP-conjugated secondary antibody incubation for 1 h, developing with ProSignal Dura ECL (Genesee) and imaging with Invitrogen iBright500. Bands were manually selected and quantified using Fiji and HRP signal was normalized to α-tubulin for analysis.

We expressed 6×HIS::GST::GABARAP in a pGEX-6P-1 in BL21(DE3)pLysS cells (Invitrogen) for 16–20 h at 22 °C after induction with 0.4 mM IPTG. Both 6×HIS::MBP::TMEM-131_c and 6×HIS::MBP::TMEM-131(LIR Mutant)_c were expressed from Rossetta2 (DE3) cells in LB broth containing 100 µg ml⁻¹ ampicillin cultured to an optical density at 600 nm of approximately 0.5 and induced with 1 mM IPTG at 18 °C for about 16 h. Purification of TMEM-131 and GABARAP protein variants were performed identically. Briefly, the cells were resuspended in buffer (25 mM HEPES, 500 mM NaCl, 20% (wt/vol) glycerol, 1 mM dithiothreitol (DTT) and 0.4 M L-arginine, pH 7.4) and lysed by homogenization, spun at 18,000 RCF for 35 min. The lysates were filtered and passed over a HISTrap column (Cytiva). The resin was washed with buffer 1 (25 mM HEPES, 500 mM NaCl, 20% (wt/vol) glycerol and 1 mM DTT, pH 7.4). Protein was eluted in with buffer 1. The pooled fractions were loaded onto a HiPrep 26/60 Sepharcryl S300 column (Cytiva) equilibrated with buffer 1. After size exclusion, the final samples were collected and stored at −80 °C.

The entire pull-down protocol was performed on ice or at 4 °C. Amylose resin (New England Biolabs) was washed with buffer A (25 µM HEPES, 500 mM NaCl and 20% (wt/vol) glycerol, pH 7.5) and then incubated with 6×HIS::MBP::TMEM-131, 6×HIS::MBP::TMEM-131(LIR mutant) or MBP for 30 min. The slurry was washed thoroughly with buffer B (25 mM HEPES, 125 mM NaCl and 5% (wt/vol) glycerol, pH 7.5). Next, 6×HIS::GST::GABARAP was added to the solution at a final in-reaction concentration of 28 µM and incubated for 30 min at a final salt concentration of 159 mM NaCl. Each slurry was washed three times with 10×reaction volume of buffer B, completely removing the remaining supernatant after the last wash. Between washes, the beads were pelleted by centrifugation at 5,000g. The beads were incubated with 10 µl elution buffer (25 mM HEPES, 175 mM NaCl and 100 mM maltose) for 15 min. Small holes were poked in the bottom of each tube with 21 G needles. The supernatant was separated by centrifugation into a second 1.5-ml tube. Next, SDS sample buffer was added to the samples, which were then boiled. The positive control was loaded at a 2% positive loading control. The samples were run on a 10% bis-Tris gel (Thermo Scientific) at 150 V for 120 min. The gel was stained with SimplyBlue SafeStain Coomassie dye (Thermo Scientific) and imaged using a BioRad Universal Hood II. For quantitation, band-intensity-measurement images (tagged image file format, tiff) of scanned Coomassie gels were made with FIJI and GABARAP-TMEM pull-down bands were normalized to the input.

Data from published C. elegans proteomic datasets⁴³ were mined by identifying proteins annotated for ER localization (Wormbase WS292, GO_0005783) and WormCat2.0 annotated functions in either proteostasis (categories: chaperone, protein modification and stress response) or lipid metabolism (Supplementary Table 1), and relative abundances for each set of proteins were contrasted between day 6 and day 1. For mining of the Tabula Muris Senis RNA expression atlas, expression Z scores of genes annotated as ‘ER protein processing’ (KEGG mmu04141) from Schaum et al.⁴⁶ were extracted from public processed data repository (twc-stanford.shinyapps.io/maca↗), averaged across all tissues as in the original manuscript and replotted. For mammalian proteome mining, processed expression data from soluble whole-tissue-lysate samples⁴⁸ were extracted from public repository (https://aging-proteomics.info↗), using log₂-transformed fold changes relative to 6-month-old mice. Pantherdb (version 19.0) was used to extract genes annotated for mitochondria, ER and protein localization to ER gene ontology terms, and the Panther enrichment test⁸⁵ (Fisher’s Exact with FDR) was used to determine statistical enrichment of specified gene sets relative to a reference set of all genes. A cutoff of FDR < 0.05 was used for determining up- or downregulated proteins among all ER and mitochondrial proteomes. To identify enrichment of cellular processes within subsets of differentially regulated ER proteins, the Panther overrepresentation test was used and the top ten biological processes by FDR are listed. Imaging experiments were conducted in three independent repeats and data were pooled before analysis. Sample sizes were based on previous experiments with similar designs. For imaging experiments, randomization occurred via the picking of random worms to a slide and the experimenter was blinded to experimental conditions before either image acquisition or analysis, as feasible. Blinding was performed with the Fiji script blind-files, provided as part of the Lab-utility-plugins update site. Isolated comparisons between two independent groups were conducted via a two-tailed Student’s t-test or Welch’s t-test when noted otherwise. Western blot data were analysed via two-tailed paired ratio t-tests. Categorical morphologic analyses were conducted using a χ² test. For experiments requiring multiple comparisons, one- or two-way ANOVAs were conducted as appropriate with post-hoc two-tailed tests to determine intergroup differences. Data distribution was assumed to be normal but not formally tested. Prism (v.9 and v.10, GraphPad) was used for statistical analysis and plot generation. Graph midlines indicate the mean ± 95% CI unless otherwise specified. Statistical tests, P values and sample sizes are indicated in the figure legends or on the corresponding graphs.

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

Nature Cell Biology thanks Ana Paula Arruda, Ruoxi Wang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

is available for this paper at 10.1038/s41556-025-01860-1.

The online version contains supplementary material available at 10.1038/s41556-025-01860-1.

Data supporting this work are available through the supplementary information. Worm and yeast strains generated in this study will be available on request and/or deposited to the Caenorhabditis Genetics Center (CGC) repository. Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding author on reasonable request.