Lysosome lipid signalling from the periphery to neurons regulates longevity

Lysosomes are key cellular organelles that metabolize extra- and intracellular substrates. Alterations in lysosomal metabolism are implicated in ageing-associated metabolic and neurodegenerative diseases. However, how lysosomal metabolism actively coordinates the metabolic and nervous systems to regulate ageing remains unclear. Here we report a fat-to-neuron lipid signalling pathway induced by lysosomal metabolism and its longevity-promoting role in Caenorhabditis elegans. We discovered that induced lysosomal lipolysis in peripheral fat storage tissue upregulates the neuropeptide signalling pathway in the nervous system to promote longevity. This cell-non-autonomous regulation is mediated by a specific polyunsaturated fatty acid, dihomo-γ-linolenic acid, and LBP-3 lipid chaperone protein transported from the fat storage tissue to neurons. LBP-3 binds to dihomo-γ-linolenic acid, and acts through NHR-49 nuclear receptor and NLP-11 neuropeptide in neurons to extend lifespan. These results reveal lysosomes as a signalling hub to coordinate metabolism and ageing, and lysosomal signalling mediated inter-tissue communication in promoting longevity.

carboxypeptidase E, peptidylglycine α-amidating monooxygenase and neuroendocrine chaperone 7B2, respectively (Fig. 1d). Among them, egl-21, pgal-1 and pghm-1 are specifically expressed in neurons 13,14 . Their inductions suggest a cell-non-autonomous regulation of neuronal genes by peripheral lysosomal lipolysis. There are also neuropeptides transcriptionally upregulated in lipl-4 Tg, including 3 insulin-like peptides (ILPs), 12 FMRFamide-related peptides (FLPs) and 19 neuropeptide-like proteins (NLPs) (Fig. 1e and Supplementary Table 1). Next, we examined the role of the neuropeptide signalling pathway in longevity regulation using the loss-of-function mutant of egl -21, egl-21(lf). The EGL-21 enzyme is required for neuropeptide processing by removing basic residues from the C-terminus of cleaved peptides 15 , and the egl-21 gene is exclusively expressed in neurons as visualized by both single-molecule fluorescence in situ hybridization (smFISH) and its green fluorescent protein (GFP) transgenic reporter (Fig. 1f,g). We found that lipl-4 Tg cannot prolong the lifespan of egl-21(lf) (Extended Data Fig. 1b and Supplementary Table 4), suggesting that the induction of neuropeptide signalling contributes to lipl-4-induced longevity. Previous genomic RNA interference (RNAi) screens found that inactivation of egl-3, encoding the convertase upstream of EGL-21, extends lifespan, which is suppressed by inactivation of the daf-16/FOXO transcription factor 16 . Similarly, we found that egl-21(lf) has extended lifespan and this lifespan extension requires daf-16 (Extended Data Fig. 1c and Supplementary Table 5). Given that ILPs regulate lifespan 17,18 and DAF-16/FOXO is the key mediator of longevity caused by IIS reduction 19 , the requirement of DAF-16 suggests that the longevity effect conferred by egl-21(lf) is possibly due to reduced agonist ILP maturation and IIS. In contrast, the lifespan extension in lipl-4 Tg is not suppressed by daf-16 RNAi (Extended Data Fig.  1d and Supplementary Table 5), indicating a negligible role of ILPs for lipl-4-induced longevity. These results suggest that, in order to test whether the reduction of NLPs or FLPs in egl-21(lf) affects lipl-4-induced longevity, we should inactivate daf-16 to eliminate the contribution from ILP reduction. We found that, with daf-16 RNAi, egl-21(lf) fully abrogates the lifespan extension conferred by lipl-4 Tg ( Fig. 1h and Supplementary Table 5). Together, these results suggest that neuropeptide processing is required for intestinal lysosomal lipolysis to promote longevity, which is probably associated with NLPs and/or FLPs but not ILPs.

Neuronal NLP-11 neuropeptide promotes longevity
To identify specifically involved neuropeptides, we performed an RNAi-based screen to search for neuropeptide-encoding genes whose inactivation suppresses the lifespan extension in lipl-4 Tg (Supplementary Table 3). We discovered that RNAi inactivation of nlp-11 in a neuronal RNAi-sensitive background suppresses lipl-4 Tg longevity without affecting WT lifespan (Fig. 2a and Supplementary  Tables 3 and 5). We further generated a clustered regularly interspaced short palindromic repeats (CRISPR) deletion mutant for nlp-11, nlp-11(lf) (Extended Data Fig. 2a) and crossed it with lipl-4 Tg. We found that nlp-11(lf) reduces the lifespan extension in lipl-4 Tg but not WT lifespan ( Fig. 2b and Supplementary Table 4). nlp-11 is transcriptionally upregulated in lipl-4 Tg (Extended Data Fig. 2b), and overexpression of nlp-11 driven by its endogenous promoter sufficiently prolongs lifespan (Fig. 2c, Extended Data Fig. 2c and Supplementary Table 4). nlp-11 expresses in both neurons and the intestine (Extended Data Fig. 2d). To examine where nlp-11 functions to regulate longevity, we knocked down nlp-11 selectively in the intestine and found that this intestine-only inactivation does not affect the lifespan extension in lipl-4 Tg (Fig. 2d and Supplementary  Table 5). We also overexpressed nlp-11 in either neurons or the intestine using tissue-specific promoters and found that only neuron-specific overexpression of nlp-11 is sufficient to prolong lifespan (Fig. 2e, Together, these results demonstrate that neuronal nlp-11 is specifically responsible for the longevity effect conferred by intestinal lysosomal lipolysis.

Lysosome-derived PUFAs regulate neuropeptide and longevity
Lysosomal acid lipase catalyses FFA release from TAGs and/or CEs 8 . Through lipidomic profiling of FFAs, we found that the levels of polyunsaturated fatty acids (PUFAs) are increased in lipl-4 Tg (Fig. 3a). To test whether these PUFAs are derived from lysosomal lipolysis, we purified lysosomes and profiled different classes of lipids. We found that, compared with WT, the level of TAGs is reduced by approximately threefold in lysosomes purified from lipl-4 Tg (Extended Data Fig. 3a). Moreover, 186 out of 305 detected TAG species (61%) are decreased in lipl-4 Tg lysosomes, with 63% of them containing PUFAs (Fig. 3b). These results suggest that the induction of PUFAs is probably due to increased lysosomal lipolysis of TAGs.
To test the hypothesis that these PUFAs serve as cell-nonautonomous signals to regulate neuropeptides, we utilized loss-offunction mutants of fat-1 and fat-3 that encode ω-3 fatty acid desaturases and Δ6-desaturase, respectively required for PUFA biosynthesis 20 (Fig. 3c). With these desaturase mutants, the upregulation of neuropeptide genes (  Table 4). FAT-1 and FAT-3 function in the intestine and neurons to catalyse PUFA biosynthesis locally 21,22 . We selectively reduced intestinal PUFAs by knocking down fat-1 and fat-3 only in the intestine and found that intestine-only inactivation of either fat-1 or fat-3 fully abrogates the lifespan extension in lipl-4 Tg (Fig. 3e,f and Supplementary Table 5). Together, these results suggest that PUFAs derived from intestinal lysosomal lipolysis mediate both neuropeptide induction and longevity.
Peripheral lipid chaperone LBP-3 promotes longevity. FFAs have low aqueous solubility and must be bound to proteins in order to diffuse through the lipophobic environment. A family of proteins termed fatty acid binding proteins (FABPs) function as lipid chaperones, which reversibly bind FFAs and their derivatives to mediate their trafficking and signalling effects 23,24 . We tested whether specific FABPs facilitate the action of intestinal PUFAs on neurons, and focused on three FABPs, LBP-1, LBP-2 and LBP-3, that carry putative secretory signals. We found that RNAi inactivation of lbp-2 or lbp-3 but not lbp-1 specifically suppresses the induction of neuropeptide genes caused by lipl-4 Tg (Extended Data Fig. 4a,b). To confirm the RNAi knockdown results, we generated CRISPR deletion mutants of lbp-2 and lbp-3 (Extended Data Fig. 4c) and crossed them with lipl-4 Tg. We found that only lbp-3 but not lbp-2 deletion suppresses the induction of neuropeptide genes (Fig. 4a). Deletion of lbp-3 also suppresses lipl-4 Tg longevity without affecting WT lifespan ( Fig. 4b and Supplementary Table 4).
We also found that transgenic strains that constitutively express lbp-3 (lbp-3 Tg) live longer than WT worms (Fig. 4c, Extended Data Fig. 4d and Supplementary Table 4) and show upregulation of egl-3, egl-21 and nlp-11 (Fig. 4d). We further profiled transcriptome changes in lbp-3 Tg using RNA-seq (Supplementary Table 2 Table 4). These results support that specific neuropeptides act downstream of LIPL-4-LBP-3 signalling to regulate longevity. Next, we found that transgenic strains that selectively overexpress lbp-3 in the intestine exhibit the upregulation of neuropeptide genes (Fig. 4h) and lifespan extension (Fig. 4i, Extended Data Fig. 4g and Supplementary  Table 4). Together, these results support that the specific lipid chaperone LBP-3 mediates fat-to-neuron communication to regulate neuropeptides and longevity.

LBP-3 relies on secretion for its regulation.
To further understand the function of LBP-3 in this endocrine regulation, we examined whether LBP-3 can be secreted from the intestine. In C. elegans, coelomocytes are scavenger cells that take up secreted materials from the body cavity and serve as a monitor of secreted proteins 25 . We generated a transgenic strain expressing an intestine-specific polycistronic transcript encoding both LBP-3-red fluorescent protein (RFP) fusion and GFP, such that GFP indicates cells expressing lbp-3 and RFP directly labels LBP-3 protein. Without tagging with any proteins, GFP was detected ubiquitously within intestinal cells (Fig. 5a). LBP-3-RFP fusion, on the other hand, was detected within intestinal cells and also in coelomocytes (Fig. 5a), which shows LBP-3 secretion from the intestine into the body cavity. We also discovered that this secretion of LBP-3-RFP is elevated in lipl-4 Tg (Fig. 5b,c). Within intestinal cells, LBP-3 protein is detected in the cytosol and also at lysosomes that are marked by LMP-1 and stained with LysoTracker (Extended Data Fig. 5a). Moreover, we generated transgenic strains that overexpress LBP-3 without its secretory signal only in the intestine and found no neuropeptide gene induction (Fig. 5d) or lifespan extension (Fig. 5e, Extended Data Fig. 5b and Supplementary Table 4) in these strains. RFP fusion of this non-secretable LBP-3 was not detected in coelomocytes (Fig. 5f). Thus, LBP-3 protein requires secretion from the intestine to systemically regulate neuropeptides and longevity, which is triggered by LIPL-4-induced lysosomal lipolysis.

DGLA regulates LBP-3 secretion, neuropeptides and longevity.
To examine how specific PUFAs and LBP-3 coordinate with each other, we first examined the effect of PUFAs on LBP-3 secretion. We used fat-3 RNAi to reduce PUFA biosynthesis in peripheral tissues and found reduction of LBP-3-RFP secretion in lipl-4 Tg (Fig. 6a,b). Thus, PUFA induction by lipl-4 Tg promotes LBP-3 secretion from the intestine. It is known that the fat-3(lf) mutant lacks 20-carbon PUFAs, including -6 DGLA and arachidonic acid (AA) that are induced by lipl-4 Tg and -3 eicosapentaenoic acid (EPA) and eicosatetraenoic acid (ETA) with no induction (Fig. 3a). We then tested whether LBP-3 binds to DGLA, AA, EPA and ETA using a competitive fluorescence-based binding assay. In this assay, when bound to LBP-3, amphipathic 1-anilinonaphthalene-8-sulfonic acid (1, shows enhanced fluorescence that is quenched once outcompeted by FFAs. We found that DGLA (K d = 10.96 μM), AA (K d = 2.9 μM) and EPA (K d = 4.76 μM) but not ETA bind to LBP-3 (Fig. 6c). Next, we supplemented DGLA, AA or EPA to worms and measured neuropeptide gene expression. We found that, in fat-3(lf), DGLA supplementation is able to restore the upregulation of neuropeptide genes caused by lipl-4 Tg (Fig. 6d). Neither AA nor EPA supplementation shows such an ability (Fig. 6d). To examine tissue specificity of this restoration, we conducted qRT-PCR analysis using dissected intestine and found that DGLA supplementation causes no intestinal induction of egl-3 or egl-21 (Extended Data Fig. 6a). We also imaged egl-21 messenger RNA transcripts using smFISH and found an increase in neurons by DGLA supplementation (Extended Data Fig. 6b). Moreover, DGLA supplementation sufficiently restores the increased LBP-3-RFP secretion (Fig. 6e,f) and lifespan extension in lipl-4 Tg with fat-3 inactivation ( Fig. 6g and Supplementary Table 5). Together, these results suggest that the induction of DGLA by lysosomal lipolysis promotes secretion of the LBP-3 lipid chaperone from the intestine, and LBP-3-DGLA signals to neurons to regulate neuropeptides and longevity.
Next, we compared the predicted structures of LBP-2 and LBP-3 using AlphaFold2 and found changes in two cap-like α-helixes that are responsible for lipid binding (Fig. 7e). A sequence alignment between LBP-2 and LBP-3 reveals that ten amino acids are different in these regions (Fig. 7f). We thus designed a chimeric protein by replacing the two cap-like LBP-3 α-helixes from N38 to K60 with those present in LBP-2 ( Fig. 7f and Extended Data Fig. 7b). We generated    Tables 4 and 5 for full lifespan data. In d, error bars represent mean ± s.e.m., n = 5 (WT and lbp-3 Tg for egl-3 and nlp-11) and n = 3 (WT and lbp-3 Tg for egl-21) biologically independent samples, **P = 0.007 for egl-3, **P = 0.007 for nlp-11 and ***P = 0.0002 for egl-21 by two-way ANOVA with Holm-Sidak correction, ~2,000 worms per replicate. In h, error bars represent mean ± s.e.m., n = 4 biologically independent samples, *P = 0.0497, ****P < 0.0001 by two-way ANOVA with Holm-Sidak correction, ~2,000 worms per replicate. Source numerical data are available in source data.
transgenic lines expressing this chimeric protein selectively in the intestine and confirmed that the chimeric protein expresses normally (Extended Data Fig. 7c). We found that overexpression of the chimeric protein, like LBP-2, does not induce neuropeptide gene expression ( Fig. 7g and Extended Data Fig. 7d). No lifespan extension was detected in these transgenic strains either (Fig. 7h, Extended Data Fig. 7e and Supplementary Table 4). These results suggest that the lipid binding specificity of LBP-3 towards DGLA is necessary for its regulation of neuropeptides and longevity.
Neuronal transduction of peripheral lipid signals. To examine whether secreted LBP-3 is taken up by neurons, we generated a transgenic strain that specifically expresses GFP nanobody (GBP) in neurons together with polycistronic mKate, and then crossed it with a transgenic strain expressing GFP-fused LBP-3 only in the intestine. In this line, if secreted LBP-3-GFP proteins from the intestine are taken up by neurons, neuronal GBP will capture them, making GFP visible in neurons (Fig. 8a). In supporting LBP-3-GFP uptake by neurons, we detected GFP signals in mKate-positive neurons (Fig. 8b). As controls, we did not detect neuronal GFP signals in either the GBP or the LBP-3-GFP transgenic strain alone (Extended Data Fig. 8a,b). Moreover, we generated a transgenic strain that expresses GBP fused with the extracellular domain of SAX-7 (GBP-SAX-7) 26 in neurons, and then crossed it with the intestine-specific LBP-3-GFP transgenic strain. In this line, LBP-3-GFP signals were also detected in neurons (Extended Data Fig. 8c), supporting a close proximity between secreted LBP-3 and the neuronal surface. Together, these results reveal LBP-3 as an endocrine lipid chaperone, which is transported from the intestine to neurons to mediate lipl-4-induced longevity. Previously, we discovered LBP-8 as a cell-autonomous mediator of lipl-4-induced longevity. To investigate the interaction between LBP-3 and LBP-8, we examined the transcriptional levels of the neuropeptide genes in the long-lived lbp-8 transgenic strain (lbp-8 Tg) and found only negligible changes (<35%) (Fig. 8c). Both lbp-3 and lbp-8 are partially required for the lifespan extension caused by lipl-4 Tg (Fig. 4b) 1 , and lbp-3 Tg and lbp-8 Tg have an additive effect in prolonging lifespan (Fig. 8d). In the intestine, LBP-8 facilitates the lysosome-to-nucleus retrograde transport of lipid signals that activate nuclear receptors NHR-49 and NHR-80 to promote longevity 12 .  Table 4 for full lifespan data. Source numerical data are available in source data.
When examining the involvement of NHR-49 and/or NHR-80, we found that the loss-of-function mutation of nhr-49 fully suppresses the upregulation of neuropeptide genes in lipl-4 Tg and lbp-3 Tg (Fig. 8e,f), but the nhr-80 mutation has a negligible effect (<18% reduction; Fig. 8e). Importantly, in the nhr-49 mutant background, neuron-specific restoration of nhr-49 fully rescues the upregulation of neuropeptide genes (Fig. 8f) and the lifespan extension ( Fig. 8g and Supplementary Table 4) conferred by lbp-3 Tg. Furthermore, we confirmed the neuronal expression of nhr-49 using a transgenic strain expressing NHR-49-mKate2 fusion driven by its endogenous promoter (Fig. 8h). After crossing this line with the GFP reporter line of nlp-11, we found many overlaps between neurons with nlp-11 expression and NHR-49 localization (Fig. 8h). On the basis of CenGenApp analysis 27 , there are 57 overlapping neurons (Extended Data Fig. 8d and Supplementary Table 6). Thus, lysosomal lipid signals from the periphery act through neuronal NHR-49 to regulate neuropeptides and longevity (Fig. 8i).

Discussion
This study supports an emerging paradigm that lysosomes are the critical signalling hub for longevity regulation. We have identified two lipid chaperones mediating the signalling role of lysosomes, LBP-8 for lysosome-to-nucleus retrograde signalling and LBP-3 for fat-to-neuron endocrine signalling, which act in parallel to regulate longevity. Overexpression of non-secretable LBP-3 causes decreased transcription of neuropeptide genes, which is probably  c, After incubation with C. elegans liposome, fatty acids bound to LBP-2 or LBP-3 proteins were analysed with mass spectrometry. DGLA shows a high percentage of occupancy in LBP-3 but not LBP-2. d, In the C. elegans liposome, the percentage of EPA is more than ten times higher than that of DGLA or AA. e, LBP-2 (blue) and LBP-3 (grey) superposition structures predicted using AlphaFold2. f, An LBP-2 and LBP-3 protein alignment generated using t-coffee. Secondary structures are displayed above the alignment. The LBP-2 sequence utilized for the replacement in the LBP-3 chimeric protein is highlighted in purple. g,h, Intestine-specific overexpression of chimeric lbp-3(chim) does not affect the transcription of egl-3 and egl-21 (g) nor prolongs lifespan (h). In a and b, error bars represent mean ± s.e.m., n = 3 biologically independent samples; NS, P > 0.05 by two-way ANOVA with Holm-Sidak correction, ~2,000 worms per replicate. In c and d, n = 3 biologically independent samples. In g, error bars represent mean ± s.e.m., n = 4 biologically independent samples; NS, P > 0.05 by two-way ANOVA with Holm-Sidak correction, ~2,000 worms per replicate. In h, n = 3 biologically independent samples; NS, P > 0.05 by log-rank test, 98-120 worms per replicate. See Supplementary Table 4 for full lifespan data. Source numerical data are available in source data.
due to reduced DGLA secretion. Thus, secreted LBP-3 may regulate the basal expression of neuropeptides even without the induction of lysosomal lipolysis. Mammalian FABP4 secreted from adipocytes has been implicated in the hormonal control of metabolism 28 , and FABP5 at the blood-brain barrier contributes to the brain uptake of docosahexaenoic acid, a PUFA essential for cognitive function 29 . Therefore, FABP secretion may function as an evolutionarily conserved mechanism to facilitate lipid transportation from peripheral metabolic tissues to the central nervous system. Nuclear receptors are the best-known mediators of lipid signals in transcriptional responses, and several C. elegans nuclear receptors have been implicated in regulating longevity, including NHR-49, NHR-80, NHR-62 and DAF-12 (refs. 1,[30][31][32][33][34] ). In particular, PPARα has high binding affinity for FAs and plays a crucial role in metabolic tissues to regulate lipid catabolism 35 . PPARα also expresses at a high level in the nervous system; however, its neuronal function and regulation remain poorly understood. Our studies reveal that, in the nervous system, NHR-49 regulates neuroendocrine gene expression in response to peripheral lipid signals. Previous studies have also shown that neuronal NHR-49 mediates the longevity effect conferred by neuronal AMPK activation 36 . Thus, lipids may be crucial endocrine signals that couple peripheral metabolic status  Table 4 for full lifespan data. In e, error bars represent mean ± s.e.m., n = 4 biologically independent samples, ***P = 0.0004 and ****P < 0.0001 by two-way ANOVA with Holm-Sidak correction, ~2,000 worms per replicate. In f, error bars represent mean ± s.e.m., n = 4 biologically independent samples, *P = 0.019, **P = 0.009 and ***P = 0.0006 for WT versus lbp-3 Tg, P = 0.0003 for lbp-3 Tg versus lbp-3Tg;nhr-49(lf) and P = 0.0003 for lbp-3Tg;nhr-49(lf) versus lbp-3Tg;nhr-49(lf);neu-nhr-49, ****P < 0.0001 by two-way ANOVA with Holm-Sidak correction, ~2,000 worms per replicate. Source numerical data are available in source data.
with neuronal transcription via PPARα. We observed the internalization of secreted LBP-3 into neurons, and FABPs are known to cooperate with PPARs in regulating transcriptional responses 37 . We thus hypothesize that LBP-3-DGLA could directly act in concert with neuronal NHR-49 to regulate neuropeptide genes. However, we could not rule out the possibility that, within neurons, secondary lipid signals are derived from internalized DGLA and activate NHR-49 in the nucleus.
Our work highlights the crucial role of PUFAs in regulating lipl-4-induced longevity. The induction of PUFAs has also been linked with dietary-restriction-mediated longevity 38 . Previous studies have reported that monounsaturated fatty acid (MUFA) supplementation is sufficient to extend C. elegans lifespan 6 . One MUFA species, palmitoleic acid, is increased in lipl-4 Tg. However, through profiling the liposome bound to LBP-3, we did not detect an enrichment of palmitoleic acid. Although we could not rule out the possibility that the induction of palmitoleic acid contributes to lipl-4-induced longevity, it might not be involved in the LBP-3-mediated endocrine signalling mechanism.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41556-022-00926-8.

C. elegans strain maintenance. C. elegans strains obtained from Caenorhabditis
Genome Center or generated in this study are listed in Supplementary Table 7. Strains were maintained on standard nematode growth medium (NGM) agar plates seeded with corresponding bacteria at 20 °C. The full list of C. elegans strains is provided in Supplementary Table 7.

Molecular cloning and generating transgenics.
Tissue-specific lbp-3 or nlp-11 expression vectors were generated using Multisite Gateway System (Invitrogen) as previously described 39 .
For generating nlp-11 and lbp-3 transgenic lines driven by their endogenous promoters, the whole genomic region including the promoter, 5′ untranslated region (UTR), coding sequence and 3′ UTR were first PCR-amplified and then fused together with sl2-GFP::unc-54 3′UTR via fusion PCR. For nlp-11 and lbp-3, 2.2 kbp and 1.1 kbp of the upstream promoter region was used, respectively.
The Gibson Assembly Method (NEB) was used to generate the following vectors. To amplify egl-21 and nhr-49, 612 bp and 2.4 kbp of the upstream promoter region was used, respectively. lbp-3ns fused to both RFP and sl2::gfp, lbp-3 fused to GFP, SAX-7 fused to GBP and GBP alone were amplified using tissue-specific promoter vectors. The chimeric lbp-3 sequence was ordered using IDT, while the 3xHA sequence was PCR-amplified and ligated into the tissue-specific promoter vectors.
Transgenic strains were generated and integrated as previously described 1 and backcrossed to N2 at least five times.
Generating deletion mutants using CRISPR. All gene-specific mutations were generated using saturated single guide RNA (sgRNA) targeting throughout the nlp-11 locus. sgRNAs were identified using the http://crispr.mit.edu/ website. Possible sgRNAs were then screened for predicted efficacy using http://crispr.wustl.edu/. For nlp-11 deletion, we followed the protocol suggested by Dickinson et al. 40 and Ward et al. 41 , while for lbp-3 deletion, we identified candidates using the protocol suggested by Paix et al. 42 and Arribere et al. 43 . Genotyping PCR was performed using nlp-11 and lbp-3 spanning primers listed in Supplementary Table 8. Candidate worms with notable band shifting were saved and back-crossed at least four times with N2.
Lifespan assays. Lifespan assays were performed as previously described 1 . For integrated transgenic strains and newly isolated CRISPR deletion mutants, the strains were backcrossed at least five times before lifespan analysis (Supplementary  Tables 4 and 5).
For lifespan assays involving strains containing mutation of egl-21 or fat-3, 5′-fluorodeoxyuridine (FUDR) at a final concentration of 100 μM was added at L4 stage to prevent ageing-irrelevant lethality due to internal eggs' hatching. All the other lifespan assays did not use FUDR.
qRT-PCR. Total RNA was isolated as previously described 1 . Synthesis of complementary DNA was performed using the amfiRivert Platinum cDNA Synthesis Master Mix (GenDEPOT). Quantitative PCR was performed using Kapa SYBR fast PCR kit (Kapa Biosystems) in a Realplex 4 PCR machine (Eppendorf), and values were normalized to rpl-32 as an internal control. All data shown represent three to four biologically independent samples. Primers used in this study are listed in Supplementary Table 8.
Intestines were dissected from day 1 adult lipl-4 Tg with fat-3 RNAi supplemented by vehicle or DGLA for 12 h, and spun down at 20,000g for 2 min at 4 °C. Supernatant was removed, and 10 µl of worm lysis buffer (containing 1:100 diluted DNase; both from Ambion Power SYBR Green Cells-to-Ct kit) was added. Lysis reaction was incubated at room temperature for 5 min, stopped by adding 2 µl of Stop Solution and incubated at room temperature for 2 min following the manufacturer protocol. The synthesized cDNA was used undiluted for qRT-PCR. Fluorescent microscopy. Day 1 adult worms were mounted on 2% agarose pads containing 0.5% NaN 3 as anaesthetic on glass microscope slides. Fluorescent images were taken using confocal FV3000 (Olympus). Polygon selection tool was used to select coelomocytes' area to be quantified, and average pixel intensity was calculated with the 'analyze-measure' command. A similar strategy was used when performing smFISH quantification using the polygon selection on the head region of the worms. After subtracting the background intensity, all measurements were averaged to obtain mean and standard deviation. In each imaging session, around 10-20 animals were analysed.
Lysotracker staining. LysoTracker Red DND-99 (Molecular Probes) was diluted in ddH 2 O to 1,000 μM, and 6 μl was added to each 3.5 cm standard NGM plate (containing 3 ml of agar) seeded with OP50. The plates were kept in the dark for 24 h to allow the lysotracker solution to diffuse evenly throughout the plate. Approximately 10-20 worms were added to each plate at the L4 stage and kept in the dark for 1 day at 20 °C before confocal imaging.
FFA profiling. For each sample, 40,000 age-synchronized worms were grown on NGM plates seed with OP50 bacteria and collected as young adults. The worms were washed three times in M9 buffer and returned to empty NGM plates for 30 min for gut clearance. Following intestinal clearance, worms were washed twice more in M9 buffer, pelleted in a minimal volume of M9 and flash-frozen in liquid nitrogen. Proprietary recovery standards were added to each sample before extraction, and the samples were extracted using methanol, chloroform and water. The extracted samples in chloroform were dried and resuspended in methanol and isopropanol (50:50, vol/vol). The samples were analysed using a Vanquish UPLC and an LTQ-FT mass spectrometer with a linear ion-trap front end and a Fourier-transform ion cyclotron resonance back end (Thermo Fisher Scientific). The mobile phase A was 5 mM ammonium acetate with pH 5 and mobile phase B was 2-propanol and acetonitrile (20:80, vol/vol). The FFAs were first identified using lipidsearch 4.2.27 software (Thermo Fisher Scientific), and then validated using standards.
Lysosome-specific lipidomics. Briefly, we generated a transgenic strain expressing a C-terminal RFP-and HA-tagged lysosomal membrane protein LMP-1 driven by the ubiquitous promoter sur-5 and crossed with lipl-4 Tg worms. For each sample, 200,000 age-synchronized worms were grown on NGM plates seed with OP50 bacteria and collected as young adults. The worms were washed three times in M9 and one time in KPBS followed by Dounce homogenization in ice until half of the worms were broken in half. The lysate was spun at 1,000g for 3 min at 4 °C to remove debris. The supernatant was incubated with 160 µl of anti-HA magnetic beads (Thermo Fisher Scientific) for 6 min at 20 °C and transferred to a magnetic stand followed by KPBS cold wash. The lipids were extracted by adding solvent methanol:water:methyl tert-butyl ether (3:2.5:10, v/v/v) into the beads. The organic top layer was dried and reconstituted with 100 µl methanol:isopropanol (1:1, v/v). The samples were analysed using a Vanquish UPLC and a Lumos orbitrap mass spectrometer (Thermo Fisher Scientific). The mobile phase A was 5 mM ammonium formate with 0.1% formic acid in water and acetonitrile (50:50, vol/ vol) and mobile phase B consisted of 2-propanol, acetonitrile and water (88:10:2, vol/vol). A reverse-phase column Thermo Accucore Vanquish C18+ was used to separate the lipids, which were detected in both positive and negative ionization modes. Mass spectra were acquired in full-scan and data-dependent MS2 mode. For MS2 scanning, 20 dependent scans were acquired in each cycle. The MS2 resolution was 30k; HCD was used to fragment precursor ions with stepped collision energy 25, 30, 35; AGC target was 50,000. High-throughput analysis of lipidomic data was performed using Lipidsearch software (Thermo Fisher Scientific). Lipid quantification use precursor ion area and lipid were identified by matching product ion spectra to lipidsearch library. Both precursor and product ion mass tolerance were set at 5 ppm. M-score threshold was set 2.0. Both positive and negative data were aligned on the basis of retention time tolerance 0.1 min and mean value. The final data were filtered on the basis of preferred ion adduct for each lipid class.
Lipid feeding. Age-synchronized worms were grown on NGM plates seeded with OP50 bacteria to day 1 adulthood. AA, DGLA and EPA (Nu-Check Prep) were dissolved in DMSO and diluted into OP50 bacterial food to a final concentration of 1 mM. Then, 300 μL of each mixture was added to standard 6-cm NGM plates that were dried in a laminar flow hood under dark conditions. Worms were collected after 12 h of lipid feeding under dark conditions followed by RNA extraction and qRT-PCR.
Competitive fluorescence-based binding assay. Quantification of ligand binding was conducted via competition of the probe 1,8-ANS as previously described 1 .
LBP-2 and LBP-3 binding with C. elegans liposome. For each sample, 100,000 age-synchronized worms were grown on NGM plates seeded with OP50 bacteria and collected as young adults. The worms were washed three times in M9 buffer, one time in 1× PBS, pelleted in a minimal volume of 1× PBS and flash-frozen in liquid nitrogen. Lipids were extracted from C. elegans lysates using the Bligh and Dyer method 44 .
Fatty acids were extracted from LBP-2 and LBP-3 after binding with C. elegans lipid extracts. Fatty acid derivatives were generated as previously described 45 . Briefly, dried lipid extracts were incubated with 200 μl of oxalyl chloride (2 M in dichloromethane) at 65 °C for 5 min, then dried down with nitrogen gas. Then, 3-picolylamide fatty acid derivatives were formed through incubation with 3-picolylamine (1% in acetonitrile) at room temperature for 5 min and then dried down with nitrogen gas.
Mass spectrometry for LBP-2: 5 μl of the LBP-2-derived fatty acid sample resuspended in methanol was injected onto a ThermoScientific Accucore C18 (4.6 × 100 mm, 2.6 µm) column using the ExionLC AD UPLC system at a 0.8 ml min −1 flow rate, and a gradient solvent system containing 10 mM ammonium acetate, pH 7 in H 2 O (solvent A) and 10 mM ammonium acetate, pH 7 in 100% acetonitrile (solvent B). Samples were chromatographically resolved using a stepwise gradient starting at 40% solvent B for 3 min, 100% solvent B for 5 min and then 65% solvent B for 2 min. Derivatized fatty acids were detected using ABSciex QTrap5500 triple quadrupole mass spectrometer in positive ion mode. The following multiple reaction-monitoring transitions were used to detect the most abundant derivatized fatty acids. Derivatized fatty acids were quantified in Multiquant 3.0.2 software (AB Sciex) using a calibration curve with the following fatty acids: myristic acid, palmitic acid, oleic acid, linoleic acid, AA and docosahexaenoic acid.
Mass spectrometry for LBP-3: 20 μl of the LBP-3-derived fatty acid sample resuspended in ethanol was injected onto a Zorbax Eclipse Plus C18 (2.1 × 100 mm, 3.5 µm) at a 0.15 ml min −1 flow rate with column temperature of 45 °C, and a gradient solvent system containing solvent A and solvent B (both as above). Samples were chromatographically resolved using a stepwise gradient starting at 10% solvent B for 4 min, 40% solvent B for 3 min, 100% solvent B for 4 min, 65% solvent B for 1 min, 40% solvent B for 2 min and then 10% solvent B for 6 min. Derivatized fatty acids were detected using an Agilent 6495c triple quadrupole mass spectrometer coupled to an Agilent 1295 II UPLC in positive ion mode. The following multiple reaction-monitoring transitions were used to detect the most abundant derivatized fatty acids. Derivatized fatty acids were quantified in Mass Hunter Quantitative Analysis 10.1 software using a calibration curve with the following fatty acids: decanoic acid, undecanoic acid, dodecanoic acid, myristic acid, palmitic acid, oleic acid, linoleic acid, stearic acid, EPA, AA and DGLA.
Western blot. At least 300 worms per genotype were grown on seeded NGM plates before being collected and snap-frozen on dry ice. The samples were lysed in worm lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA and 0.1% NP-40) containing a protease and phosphatase inhibitor cocktail (cOmplete Protease Inhibitor Cocktail, cat. no. 11697498001; PhosSTOP, cat. no. 4906845001; both from Sigma) and homogenized with motorized pellet pestle. The lysates were then centrifuged, and the supernatants were used for protein quantification and western blotting analysis. Next, the proteins were separated with the NuPAGE system (Thermo Fisher, 4-12% Bis-Tris protein gel), and transferred to PVDF membrane (Thermo Fisher). The membranes were blocked with 5% BSA in TBST. The primary antibody against HA is anti-HA rabbit mAb (Cell Signaling, #C29F4, 1:1,000), which detects the HA tag and was tested for its specificity using WT and LBP-3::HA lysates. The anti-β-actin antibody is from Santa Cruz (sc-47778, 1:2,000). Protein detection was performed using chemiluminescent substrate (ECL Western Blotting Reagents, Sigma-Aldrich, GERPN2106) and images acquired using a gel imaging system (ImageQuant LAS 500, Thermo Fisher Scientific).
smFISH. Custom Stellaris FISH Probe was designed against egl-21 by utilizing the Stellaris RNA FISH Probe Designer (Biosearch Technologies, www.biosearchtech. com/stellarisdesigner). The samples were hybridized with the egl-21 Stellaris RNA FISH Probe set labelled with Quasar 670 (Biosearch Technologies), following the manufacturer's instruction. Briefly, around 300 worms were washed from small plates with M9, transferred to 1.5 ml Eppendorf tubes, resuspended in 1 ml fixation buffer (3.7% v/v formaldehyde in 1× PBS) and incubated at room temperature for 45 min with rotation. Fixed worms were washed twice with 1× PBS, resuspended in borate triton ß-mercaptoethanol solution buffer and kept at 4 °C overnight to permeabilize the cuticle. Worms were washed twice with Borate triton solution, twice with PBST and once in 1× PBS. Samples were resuspended in 1 ml wash buffer (10% v/v formamide in 1× Wash Buffer A, Biosearch Technologies, cat. no. SMF-WA1-60) at room temperature for 2-5 min, before being hybridized for 12 h at 37 °C in 100 µl hybridization buffer (10% v/v formamide in Hybridization Buffer, Biosearch Technologies, cat. no. SMF-HB1-10) containing 1 µl of 125 nM final concentration reconstituted probe. Samples were incubated in wash buffer at 37 °C for 30 min, then resuspended in 1 ml of DAPI nuclear staining (5 ng ml −1 DAPI in wash buffer). Worms were incubated in the dark at 37 °C for 30 min and resuspended in 1 ml of wash buffer at room temperature for 2-5 min. Worms were transferred using a Pasteur pipette into a small drop of Vectashield Mounting Medium (Vector Laboratories), before imaging.

RNA-seq preparation and analysis.
Total RNA from WT and lipl-4 Tg was extracted from around 3,000 worms in three different biological replicates using Trizol extraction combined with column purification (Qiagen). Total RNA from WT and lbp-3 Tg was extracted using phenol-chloroform method. Sequencing libraries were prepared using the TruSeq Stranded mRNA Sample Preparation kit (Illumina) following the manufacturer's instructions. Libraries were pooled together and sequenced using Illumina NextSeq 500 system. RNA-seq reads were aligned to the C. elegans reference genome using hisat2 with the default setting. HTSeq was used to count the read numbers mapped to each gene. DESeq2 was used to normalize the raw counts and identify differentially expressed genes (|fold change| ≥1.5; false discovery rate <0.05).
Protein sequence alignment. The alignment was generated by tcoffee and illustrated by ESPript.
Statistics and reproducibility. For all figure legends, asterisks indicate statistical significance as follows: NS, not significant (P > 0.05), *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Data were obtained by performing independently at least three biological replicates, unless specified in the figure legends. No statistical method was used to pre-determine the sample size. No data were excluded from the analyses. Two-tailed Student's t-test or one-way or two-way analysis of variance (ANOVA) with Holm-Sidak corrections was used as indicated in the corresponding figure legends. n indicates the number of biological replicates. For survival analysis, statistical analyses were performed with SPSS software (IBM) using Kaplan-Meier survival analysis and log-rank test. Details on samples size, number of biological replicates and statistics for each experiment are provided in Supplementary Tables 4 and 5. For FFA profiling, statistical analysis was performed using two-way ANOVA test with Holm-Sidak correction comparing lipl-4 Tg (n = 9 biological replicates) vs. WT (n=6 biological replicates), while for lipid profiling using lysosomal isolation, t-test was used to compare lipl-4 Tg (n=4 biological replicates) versus WT (n = 3 biological replicates). For RNA-seq, two-sided Wald test in R package DEseq2 was used. For qRT-PCR, t-test or one-way or two-way ANOVA with Holm-Sidak correction was used as indicated in the corresponding figure legends. Figures and graphs were constructed using GraphPad Prism 7 (GraphPad Software) and Illustrator (CC 2019; Adobe). The researchers involved in the study were not blinded during experiments or outcome assessment.  . 2). a) A schematic representation of the nlp-11(rax51) loss-of-function mutation. White boxes represent the nlp-11(rax51) deletion, black lines represent introns. The mutant lacks the three exons and the transcriptional start site. b) nlp-11 is transcriptionally up-regulated by lipl-4 Tg. c) Constitutive expression of nlp-11 driven by its endogenous promoter extends lifespan. d) The transgenic strains expressing nlp-11 under its endogenous promoter (nlp-11p::nlp-11::sl2-GFP) reveals the expression of nlp-11 in intestinal and neuronal cells. Scale bar 100µm. e, f) Neuron-specific overexpression of nlp-11 prolongs lifespan (e), but intestine-specific overexpression has no such effect (f). (b) Error bars represent mean ± s.e.m., n=3 biologically independent samples, **** p<0.0001 by two-tailed Student's t-test, ~2000 worms per replicate. (c, e, f) n=3 biologically independent samples, n.s. p>0.05 and *** p<0.001 by long-rank test, 60-120 worms per replicate. See Supplementary Table 4 for full lifespan data. Source numerical data are available in source data. Fig. 5 | LBP-3 secreted from the periphery regulates neuropeptide and longevity (related to Fig. 5). a) LBP-3 and the lysosomal membrane protein LMP-1 are visualized by their GFP and RFP fusions, respectively. LBP-3::GFP colocalizes with LMP-1::RFP and Lysotracker Red staining at lysosomes, and is also detected in the cytosol. Scale bar 10µm. b) Intestine-specific overexpression of lbp-3 lacking its secretory signal (lbp-3ns) fails to extend lifespan. n=3 biologically independent samples, n.s. p>0.05 by long -rank test, 72-100 worms per replicate. See Supplementary Table 4 for full lifespan data. Fig. 7 | DgLA binding specificity of LBP-3 mediates its effects (related to Fig. 7). a) After incubation with the C. elegans liposome, fatty acids bound to LBP-2 or LBP-3 proteins were analyzed with mass spectrometry. In addition to PUFAs shown in Fig. 5, SFAs and MUFAs also bind to LBP-2 and LBP-3 with different preferences. b) The LBP-3 chimeric protein structure predicted using AlphaFold2. The two helix regions from LBP-2 are highlighted in purple. c) Western-blot of WT and transgenic worms overexpressing LBP-3 chimeric proteins fused with 3XHA-tag. LBP-3::3xHA fusion proteins are detected in both chimeric lines. ß-actin is used as a control. d, e) Intestine-specific overexpression of the chimeric lbp-3(chim) does not affect the transcription of egl-3 or egl-21 (d) and fails to extend lifespan (e). (d) Error bars represent mean ± s.e.m., n=3 biologically independent samples, n.s. p>0.05 by two-way ANOVA with Holm-Sidak correction, ~300 worms per replicate. (e) n=3 biologically independent samples, n.s. p>0.05 by long-rank test, 72-100 worms per replicate. See Supplementary Table 4 for full lifespan data. Source numerical data and unprocessed blots are available in source data.