Articles https://doi.org/10.1038/s41556-019-0391-5

ER–lysosome contacts enable cholesterol sensing by mTORC1 and drive aberrant growth signalling in Niemann–Pick type C

Chun-Yan Lim1,2, Oliver B. Davis1,2, Hijai R. Shin1,2, Justin Zhang1,2, Charles A. Berdan1,3, Xuntian Jiang4, Jessica L. Counihan1,3, Daniel S. Ory4, Daniel K. Nomura 1,3 and Roberto Zoncu 1,2*

Cholesterol activates the master growth regulator, mTORC1 kinase, by promoting its recruitment to the surface of lysosomes by the Rag guanosine triphosphatases (GTPases). The mechanisms that regulate lysosomal cholesterol content to enable mTORC1 signalling are unknown. Here, we show that oxysterol binding protein (OSBP) and its anchors at the endoplasmic retic- ulum (ER), VAPA and VAPB, deliver cholesterol across ER–lysosome contacts to activate mTORC1. In cells lacking OSBP, but not other VAP-interacting cholesterol carriers, the recruitment of mTORC1 by the Rag GTPases is inhibited owing to impaired transport of cholesterol to lysosomes. By contrast, OSBP-mediated cholesterol trafficking drives constitutive mTORC1 activa- tion in a disease model caused by the loss of the lysosomal cholesterol transporter, Niemann–Pick C1 (NPC1). Chemical and genetic inactivation of OSBP suppresses aberrant mTORC1 signalling and restores autophagic function in cellular models of Niemann–Pick type C (NPC). Thus, ER–lysosome contacts are signalling hubs that enable cholesterol sensing by mTORC1, and targeting the sterol-transfer activity of these signalling hubs could be beneficial in patients with NPC.

he exchange of contents and signals between organelles is key of cholesterol within the lysosome, compromising its functional- to the execution of cellular programs for growth and homeo- ity and triggering NPC—a fatal metabolic and neurodegenerative stasis, and failure of this communication can cause disease. A disease14. LDL stimulates Rag- and SLC38A9-dependent activation T 7 form of organelle communication involves exchange of cholesterol of mTORC1 and, in cells lacking NPC1, mTORC1 is hyperactive and other lipids by specialized carriers that are located at physical and cannot be switched off by cholesterol depletion, although the contacts between the ER and other membranes1–3. Cholesterol was mechanistic basis for this constitutive activation remains unclear. recently identified as an essential activator for the master growth Following its NPC1-dependent export from the lysosome, cho- regulator, mTORC1 kinase. Cholesterol promotes mTORC1 recruit- lesterol can be detected in several acceptor compartments, including ment from the cytosol to the lysosomal membrane, where mTORC1 the ER, Golgi and plasma membrane, but whether these compart- triggers downstream programs for biomass production and sup- ments represent separate routes or stations in a common export pression of catabolism4–7. However, the mechanisms that deliver pathway is unclear1,3,15,16. Cholesterol can also be transferred back cholesterol to the lysosomal membrane to enable mTORC1 activa- from the ER to several acceptor organelles, including the lysosome, tion are unknown. More generally, whether and how interorganelle through specialized carriers that reside at membrane contacts1,3,15,16. contacts govern cell-wide programs for growth and quality control The points at which cholesterol is made available for mTORC1 acti- are not understood. vation along these routes are unclear7. Under low cholesterol, mTORC1 cannot interact with its lyso- An important class of cholesterol carriers are the oxysterol somal scaffold, the Rag GTPases, and remains inactive in the cyto- binding protein (OSBP)-related proteins (ORPs)1–3. ORPs contain, sol. By contrast, stimulating cells with cholesterol triggers rapid at their C terminus, large hydrophobic cavities that shield choles- Rag-GTPase-dependent translocation of mTORC1 to the lysosomal terol molecules from the polar cytosolic environment and can also surface and activation of its kinase function7. Experiments in cells accommodate phospholipids17,18. The founding member of this and reconstituted systems suggest that the Rag GTPases specifically family, OSBP, localizes at contacts between the ER and trans-Golgi, sense the cholesterol content of the lysosomal limiting membrane7. where it is thought to transfer ER-derived cholesterol to the Golgi in This cholesterol pool regulates the Rags, at least in part, through exchange for phosphatidylinositol 4-phosphate (PtdIns4P)19–21. SLC38A9, which is a multi-pass amino acid permease that is also OSBP was recently proposed to function at contacts between the required for the activation of mTORC1 by amino acids7–10. ER and endolysosomes22,23. In concert with its binding partners on The cellular origins of the cholesterol pool that activates the ER, VAPA and VAPB (VAPA/B), OSBP regulates the PtdIns4P mTORC1 are unclear. Exogenous cholesterol carried by low-density content of endolysosomes, which—in turn—affects their actin- lipoprotein (LDL) is trafficked to the lysosomal lumen, and from dependent motility22. Whether OSBP also controls cholesterol levels there it is exported to acceptor membranes by a mechanism that on the lysosomal limiting membrane, and how the transport activ- requires the putative cholesterol carrier, NPC1 (refs. 11–13). Genetic ity of OSBP across ER–lysosome contacts affects the activation of inactivation of NPC1 in humans leads to substantial accumulation mTORC1 are unknown.

1Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, CA, USA. 2The Paul F. Glenn Center for Aging Research, University of California, Berkeley, Berkeley, CA, USA. 3Department of Nutritional Sciences and Toxicology, University of California, Berkeley, Berkeley, CA, USA. 4Diabetic Cardiovascular Disease Center, Washington University School of Medicine, St. Louis, MO, USA. *e-mail: [email protected]

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

Here we find that, unique among several sterol carriers associ- Notably, knockdown of OSBP—but not ORP11, ORP8 or ORP5 ated with the lysosome, OSBP establishes a lysosomal cholesterol (which did not score as lysosome-associated)—largely reversed pool that is essential for Rag GTPase-dependent mTORC1 activa- D4H*–mCherry accumulation on the lysosomal surface of NPC1- tion. In cells lacking NPC1, unopposed ER-to-lysosome transport null human fibroblasts and HEK293T cells (Fig. 1f–h, Supplementary by OSBP drives aberrant build-up of cholesterol on the lysosomal Fig. 1g). By contrast, OSBP depletion did not correct luminal choles- limiting membrane. Genetic and chemical inhibition of OSBP terol accumulation of NPC1-null lysosomes, as shown by unchanged reverses cholesterol accumulation on the limiting membrane of cholesterol staining using filipin (Fig. 1g, Supplementary Fig. 1g). NPC1-null lysosomes, suppresses aberrant mTORC1 signalling and We confirmed this result using the natural product OSW-1, which restores defective autophagy, a major driver in the pathogenesis of inhibits OSBP (along with closely related OSBP2) with nanomolar NPC. Thus, ER–lysosome contacts emerge as signalling nodes that potency20,33. Treatment with OSW-1 caused the D4H*–mCherry sig- coordinately regulate mTORC1 signalling and autophagy, and their nal, but not the filipin signal, to disappear from lysosomes of NPC1- manipulation by OSBP inhibitors could be beneficial in NPC and null human fibroblasts, MEFs and HEK293T cells (Supplementary mTORC1-driven diseases. Fig. 2a–f). Thus, unique among the ORPs, OSBP seems to be responsible Results for the accumulation of cholesterol on the limiting membrane but OSBP mediates cholesterol build-up on the limiting membrane not within the lumen of NPC1-null lysosomes. of NPC1-null lysosomes. In cells lacking NPC1, mTORC1 is con- stitutively active and refractory to inhibition by cholesterol-deplet- OSBP localizes to ER–lysosome contacts and controls lysosomal ing agents7. This result seems paradoxical because, in the absence cholesterol levels in wild-type cells. OSBP was shown to mainly of NPC1, cholesterol should be trapped within the lysosomal inte- reside at ER–trans-Golgi-network contacts19,34,35. However, in rior and unable to reach its limiting membrane11–13 where mTORC1 HEK293A cells, we also detected stably expressed green fluorescent activation occurs7. A possible resolution of this paradox is that, protein (GFP)-tagged OSBP around LAMP2-positive lysosomes in NPC1-null lysosomes, not only the interior but also the limit- (Supplementary Fig. 3a). GFP–OSBP also co-localized with endog- ing membrane could be cholesterol-enriched, a possibility that is enous mTOR and with the p18 subunit of the Ragulator/LAMTOR consistent with the altered trafficking pattern of NPC1-null lyso- complex, which anchors the Rag GTPases and mTORC1 to the lyso- somes24–26. However, owing to limitations of current biochemical somal membrane5,36 (Supplementary Fig. 3b). fractionation and cholesterol-imaging approaches, it has not been Association of OSBP with LAMP2-, mTOR- or p18-positive possible to univocally determine the cholesterol status of the lim- vesicles was abolished by deletion of its pleckstrin homology (PH) iting membrane of NPC1-null lysosomes. Therefore, we used a domain, which anchors OSBP to the Golgi or endolysosomal mem- recently established cholesterol biosensor, mCherry-tagged D4H* brane by binding to PtdIns4P22,37–39 (Supplementary Fig. 3a,b). As (derived from the Clostridium Perfringens theta-toxin. *, improved OSBP transfers PtdIns4P from Golgi and endolysosomes towards version of the original D4H by introducing Y415A and A463W the ER, it is thought to weaken its own association with these organ- mutations, see Methods), which binds to membranes at which elles19,20. Consistent with this, a transport-defective OSBP mutant cholesterol exceeds 10% molar content27 (Supplementary Fig. 1a). (ΔELSK)40 was more strongly clustered on LAMP2-positive struc- Recombinant D4H*–mCherry was delivered to semi-permeabilized tures than wild-type OSBP (Supplementary Fig. 3a). Similarly, cells using a liquid-nitrogen pulse that breached the plasma mem- inhibiting OSBP with OSW-1 resulted in strong clustering of GFP– brane but left the lysosomal membrane intact (Fig. 1a), as shown by OSBP to lysosomes (Supplementary Fig. 3c). Staining HEK293A retention of LysoSensor staining (Supplementary Fig. 1b). cells with an antibody that recognizes endogenous OSBP showed In non-permeabilized HEK293T cells, D4H*–mCherry bound partial co-localization with LAMP2 that was enhanced by short to only the outer leaflet of the plasma membrane, consistent with its hairpin RNA (shRNA)-mediated depletion of VAPA/B, as previ- inability to cross membrane bilayers (Supplementary Fig. 1c, top). ously shown22 (Supplementary Fig. 3d). In semi-permeabilized cells with intact NPC1 function, includ- Next, we investigated whether OSBP regulates limiting-mem- ing human fibroblasts, mouse embryonic fibroblasts (MEFs) and brane cholesterol of not only NPC1-null but also wild-type lyso- HEK293T cells, D4H*–mCherry did not bind to the cytoplasmic somes. As D4H*–mCherry did not bind to lysosomes with intact leaflet of lysosomes or other endomembranes, indicating that NPC1 function, we performed lipidomic analysis of immunopu- their cholesterol content is below 10% (Fig. 1b, Supplementary rified lysosomes (Supplementary Fig. 4a). Inhibiting OSBP before Fig. 1c,d). By contrast, D4H*–mCherry showed strong lysosomal immunopurification caused a 30% reduction in cholesterol content, accumulation in fibroblasts from patients carrying a pathogenic suggesting that a significant fraction of the lysosomal cholesterol NPC1 mutation (NPC1mut; Fig. 1b), in Npc1−/− mouse-derived pool is ER-derived and transferred by OSBP (Fig. 2a). Whole-cell MEFs28,29 (Supplementary Fig. 1d), in HEK293T cells deleted for lipidomics showed that, although the total cholesterol content of NPC1 using CRISPR–Cas9 (ref. 7; Fig. 1c) and in control human OSBP-depleted cells was unchanged, a larger fraction was esteri- fibroblasts treated with the chemical NPC1 inhibitor, U18666A30 fied to oleic acid than in control cells20,38 (Supplementary Fig. 4b,c). (Supplementary Fig. 1e,f). Thus, not only the lumen but also the Moreover, filipin staining showed strong accumulation of choles- limiting membranes of NPC1-null lysosomes show considerable terol in the ER of cells treated with OSBP-targeting shRNA or with cholesterol accumulation. the OSBP inhibitor OSW-1 (Supplementary Fig. 4d,e). As LDL-derived cholesterol cannot escape from the lumen of To independently visualize OSBP-dependent sterol transport NPC1-null lysosomes11–13, we hypothesized that the peripheral to the lysosomal limiting membrane, we pulse–chased cells with a cholesterol build-up revealed by D4H*–mCherry might be trans- fluorescent cholesterol analogue, TopFluor-cholesterol. In DMSO- ferred across interorganelle contacts. To identify carriers that treated cells, TopFluor-cholesterol initially appeared at the plasma could mediate cholesterol transfer to lysosomes, we carried out a membrane, and began accumulating in lysosomes by 30 min proteomics-based analysis of immunopurified lysosomes7,31,32. This (Supplementary Fig. 5a,b). By contrast, OSW-1 treatment delayed analysis revealed several peptides from three ORP-family carriers: clearance of TopFluor-cholesterol from the plasma membrane and OSBP, ORP8 and ORP11 (ref. 2; (Fig. 1d, Supplementary Table 1). caused it to acquire an ER-like distribution, never building up in lyso- We confirmed lysosomal association of endogenous OSBP, ORP11 somes to a noticeable degree (Supplementary Fig. 5a,b). Although and ORP8 by immunoblotting of the immunopurified lysosomal off-target inhibition of other ORPs cannot be ruled out entirely, the samples (Fig. 1e). most likely explanation for this result is that OSBP mediates a key

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles

a b GST–D4H* d LAMP2 –mCherry Filipin Identified Unique peptide LAMP2 ORPs counts

Fix WT D4H* OSBP 8, 5, 8, 6, 12 Filipin NPC 1 ORP8 5, 1, 4, 4, 2 Liquid N2 Merge ORP11 5, 1, 4, 3, 2 LAMP2

mut D4H* Filipin NPC 1 Merge e GST–D4H* Input Lyso-IP(kDa) Wash 75 –mCherry HA c Flag–GFP– GST–D4H* Tmem192 –mCherry Filipin NPC1 250 Fix 20 Tmem192 p18

ResWT D4H* Raptor 150 100 Wash Filipin Filipin OSBP 100 Merge sgNPC 1 ORP11 Tmem192 ORP8 100 75 D4H* Pex5 Filipin PDI 50 sgNPC1 37 Merge VDAC

f g Flag–GFP– GST–D4H* h Tmem192 –mCherry Filipin Tmem192 1.5 1.5 Adjusted P = 0.0001 P = 1.16578 × 10–13 **** D4H* ****

1.0 –Dox Filipin 1.0 Merge 0.5 0.5 filipin signals Tmem192 filipin signals D4H*-positive D4H*-positive

0 sgNPC1 + D4H* 0 shOSBP-TetOn

+Dox shOSBP- Luc Filipin shRNA: ORP8 ORP5 TetOn: –Dox +Dox ORP11 OSBP Merge NPC1mut sgNPC1

Fig. 1 | Cholesterol accumulates at the limiting membrane of NPC1-deficient lysosomes in an OSBP-dependent manner. a, In situ labelling assay for detecting cholesterol at the lysosomal membrane. b, Cholesterol is deposited at the lysosomal membrane in human NPC1 patient-derived fibroblasts. WT mut Control (NPC1 ) and NPC1 (NPC1 ) fibroblasts were fixed, breached with a liquid N2 pulse, subjected to cholesterol labelling by GST–D4H*–mCherry and filipin, and stained for LAMP2. Scale bars, 10 µm. c, NPC1 deletion by CRISPR–Cas9 genome editing in cells results in cholesterol accumulation at the lysosomal membrane. NPC1-deleted HEK293T cells expressing Flag–GFP–Tmem192, either naive or reconstituted with Flag-tagged NPC1, were processed for cholesterol labelling. Scale bars, 10 µm. Insets: three examples per genotype. d, Proteomic analysis of affinity-purified lysosomes. Lysosomes were immunopurified by anti-Flag M2 beads (or anti-haemagglutinin (HA) magnetic beads) from HEK293T cells expressing LAMP1–mRFP–2×Flag or Tmem192–mRFP–3×HA and analysed by mass spectrometry. Unique peptide counts for identified ORPs are shown; n = 2 independent experiments, 5 biologically independent samples in total (Supplementary Table 1). mRFP, monomeric red fluorescent protein. e, Pull-down of lysosomes revealing the lysosomal association of ORPs. Lysosomes were purified by anti-HA beads and immunoblotted for the indicated proteins. IP, immunoprecipitation. f, Quantification of co-localization of D4H*–mCherry with filipin-labelled cholesterol deposits in NPC1-null cells depleted of ORPs. The box plots show the minimum, first quartile, median, third quartile and maximum; 10 fields of view per genotype; n represents the number of cells: shRNA targeting luciferase (shLuc) (n = 13), shOR8 (n = 12), shORP11 (n = 12), shOSBP (n = 11), shORP5 (n = 10); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 compared with NPC1mut-shLuc (Supplementary Fig. 1g). g,h, Concomitant depletion of OSBP in NPC1- null cells reduces lysosomal membrane cholesterol levels. Cells were depleted of OSBP using doxycycline-induced shRNA and processed for cholesterol labelling. Scale bars, 10 μm. Insets: three examples per genotype. For h, the box plots are as described in f; 10 fields of view per group; n represents the number of cells: −Dox (n = 64), +Dox (n = 72); statistical analysis was performed using two-tailed unpaired t-tests; ****P = 1.16578 × 10−13 compared with the −Dox group. Insets in b,c and g represent regions of the respective panels magnified by 3.2 times. Experiments in b and c were performed three times; experiments in e and g were performed twice. step in a pathway that moves cholesterol from the plasma membrane of canonical substrates S6-kinase 1 (S6K1) and 4E-binding protein 1 to the lysosomal limiting membrane through the ER. (4E-BP1). By contrast, knockdown of ORP8, ORP11 and ORP5 had a smaller or no effect on the kinase activity of mTORC1 (Fig. 2d). OSBP is essential for mTORC1 activation by cholesterol. As lyso- Inhibiting OSBP activity with OSW-1 also suppressed cho- somal membrane cholesterol drives mTORC1 activation7, we tested lesterol-induced mTORC1 lysosomal recruitment and activation the requirement for OSBP in mTORC1 lysosomal recruitment in a dose-dependent manner (Fig. 2e, Supplementary Fig. 6b,c). and kinase activation. In HEK293T cells, we found that OSBP was Even under complete medium conditions, OSBP ablation strongly required for cholesterol-induced translocation of mTOR to LAMP2- reduced lysosomal mTORC1 recruitment (Supplementary Fig. 6d,e) positive structures, whereas ORP8, ORP11 or ORP5 were dispens- and signalling (Supplementary Fig. 6f). By contrast, OSBP deple- able (Fig. 2b,c, Supplementary Fig. 6a). Consistent with the loss of tion affected acute mTORC1 activation by amino acids only lysosomal localization, depleting OSBP strongly reduced cholesterol- modestly, suggesting that OSBP is not directly involved in amino dependent mTORC1 activation, as shown by loss of phosphorylation acid sensing (Supplementary Fig. 6g).

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

a b shRNA: Luciferase OSBP 1.5 P = 1.61895 × 10–6 mTOR LAMP2 mTOR LAMP2 **** mTOR 1.0 LAMP2 –Chol Merge 0.5

Relative levels of levels Relative mTOR lysosomal cholesterol LAMP2 0 +Chol Merge DMSO OSW-1

c d e OSW-1 (nM) shRNA: Luc ORP8 ORP11 OSBP ORP5 1 10 20 50 MCD + + + + + + + + + + MCD + + + + + + –Chol +Chol Cholesterol – + – + – + – + – + 1.5 (kDa) Cholesterol – + + + + + (kDa) Adjusted P = 0.0001 75 75 Low Low **** p-T389-S6K1 75 p-T389-S6K1 75 High High 1.0 75 75 S6K1 S6K1 20 20 p-S65-4E-BP1 0.5 P-S65-4E-BP1 20 20 4E-BP1 mTOR-positive LAMP2 signals 4E-BP1 p-T308-Akt 50 0 ORP8 100 p-S473-Akt 50 ORP11 100 Luc Akt shRNA: ORP8 ORP5 50 ORP11 OSBP 100 100 OSBP OSBP ORP5 75 mTOR 250 50 Actin Raptor 150 50 Fold change: pS6K/S6K 0.15 1.00 0.11 0.19 0.10 0.28 0.09 0.09 0.10 0.64 Actin Fold change: p4E-BP1/4E-BP1 0.14 1.00 0.10 0.72 0.21 0.71 0.13 0.14 0.11 0.08

Fig. 2 | OSBP is required for cholesterol-dependent lysosomal recruitment and activation of mTORC1. a, OSBP inhibition by OSW-1 treatment reduces the lysosomal cholesterol content. HEK293T cells expressing Tmem192–mRFP–3×HA were treated with DMSO or 20 nM OSW-1 for 8 h. Lysosomes were purified and analysed by mass spectrometry. Data are mean ± s.d. of n = 4 biologically independent samples per treatment. Statistical analysis was performed using a two-tailed unpaired t-test; ****P = 1.61895 × 10−6 compared with DMSO-treated cells. Immunoblots are provided in Supplementary Fig. 4a. b, OSBP is specifically required for lysosomal recruitment of mTORC1 by cholesterol. Cells depleted of OSBP were subjected to cholesterol depletion and restimulation, where indicated, followed by immunofluorescence for endogenous mTOR and LAMP2. Representative images are shown. Scale bars, 10 μm. Insets in b represent regions of the respective panels magnified by 3.2 times. c, Quantification of co-localization of mTOR with LAMP2-positive lysosomes in cells expressing the indicated shRNAs. Data are mean ± s.d.; 10 fields of view per genotype or condition; n represents the number of cells: shLuc −cholesterol (Chol) (n = 182), shLuc +Chol (n = 181), shORP8 −Chol (n = 104), shORP8 +Chol (n = 91), shORP11 −Chol (n = 157), shORP11 +Chol (n = 156), shOSBP −Chol (n = 176), shOSBP +Chol (n = 182), shORP5 −Chol (n = 160), shORP5 +Chol (n = 166); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 compared with shLuc +Chol. Representative images are provided in Supplementary Fig. 6a. d, ORPs affect the activation of mTORC1 by cholesterol. Cells were sterol-depleted using methyl-β-cyclodextrin (MCD, 0.5% w/v) for 2 h and restimulated for 2 h using 50 µM cholesterol complexed with 0.1% MCD (MCD:cholesterol). Cell lysates were analysed for the levels of the indicated proteins and for the phosphorylation status of S6K1 (T389), 4E-BP1 (S65). Fold changes of protein phosphorylation are indicated. e, OSW-1 inhibits sterol-induced mTORC1 signalling in a dose-dependent manner. Cells were subjected to cholesterol starvation and restimulation in the presence of DMSO or OSW-1 at the indicated concentrations and immunoblotted for the indicated proteins. The experiments in b and d were performed three times; the experiments in a and e were performed twice. See unprocessed blots in Supplementary Fig. 9. Statistics source data are provided in Supplementary Table 2.

OSBP depletion had no effect on the lysosomal localization OSBP mediates ER-to-lysosome cholesterol transfer that enables of RagC and p18/LAMTOR1 (Fig. 3a, Supplementary Fig. 6h). mTORC1 activation. To determine whether and how the choles- OSBP loss therefore impairs the ability of the Ragulator–Rag terol-transport function of OSBP participates in mTORC1 regu- complex to recruit mTORC1, but not its integrity. Consistent with lation, we reconstituted OSBP-depleted cells with isoforms that this interpretation, stable expression of GTP-locked RagBQ99L lack key functions (Fig. 4a, Supplementary Fig. 7a). Disabling mutant, which bypasses mTORC1 inactivation by both amino OSBP attachment to the lysosome (by deleting or mutating the PH acid and cholesterol depletion4,6,7, largely rescued the effect of domain) or to the ER (by mutating the phenylalanines in an acidic OSBP depletion on lysosomal mTORC1 recruitment (Fig. 3b,c) tract (FFAT) motif that mediates the interaction of OSBP with and signalling (Fig. 3d). Similarly, mTORC1 signalling was resis- VAPA/B) or both (by expressing the OSBP-related domain (ORD) tant to OSBP depletion in cells lacking the NPRL3 component alone)38, resulted in failure to rescue mTORC1 lysosomal recruit- of the GATOR1 complex or the KPTN subunit of the KICSTOR ment (Fig. 4a–c, Supplementary Fig. 7a) and activation by choles- complex, which inactivates the Rag GTPases under low nutri- terol (Supplementary Fig. 7b,c). Disabling the OSBP transport cycle ents41–43 (Fig. 3e,f). by deleting or mutating critical residues in the ORD17,40 also abol- Taken together, these data indicate that OSBP enables mTORC1 ished cholesterol-dependent activation of mTORC1 (Supplementary activation by cholesterol through a pathway that converges on the Fig. 7b,c). Thus, mTORC1 activity requires both the membrane- Rag GTPases. anchoring and lipid-transporting functions of OSBP.

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles

a b shRNA: Luciferase OSBP 1.5 –Chol +Chol Adjusted P = 0.0001 mTOR LAMP2 mTOR LAMP2 **** mTOR 1.0 LAMP2 –Chol 0.5 Merge LAMP2 signals mTOR 0 Control mTOR-, p18- or RagC-positive mTOR-, Luc Luc Luc LAMP2 shRNA: OSBP OSBP OSBP +Chol mTOR p18 RagC Merge

c –Chol +Chol mTOR Adjusted P = 0.0001 **** LAMP2 1.5 Adjusted P = 0.0001 –Chol Adjusted P = 0.0001 **** Merge

**** Q99L 1.0 mTOR Rag B LAMP2

0.5 +Chol mTOR-positive LAMP2 signals Merge

0 shRNA: Luc Luc OSBP OSBP Control RagBQ99L

d e Control sgNPRL3 f Control sgKPTN

shRNA: Luc OSBP shRNA: Luc OSBP Luc OSBP shRNA: Luc OSBP Luc OSBP Q99L Q99L MCD + + + + + + + + MCD + + + + + + + +

Rescue: Control RagB Control RagB Cholesterol – + – + – + – + (kDa) Cholesterol – + – + – + – + (kDa) 75 75 MCD + + + + + + + + Low Low p-T389-S6K1 75 p-T389-S6K1 75 Cholesterol – + – + – + – + (kDa) High High 75 75 75 Low S6K1 S6K1 p-T389-S6K1 75 20 20 High p-S65-4E-BP1 p-S65-4E-BP1 20 20 S6K1 75 4E-BP1 4E-BP1 37 Q99L Flag–RagB NPRL3 50 NPRL3 50 100 100 100 OSBP OSBP OSBP 50 Actin Actin 50 Actin 50

Fold change: pS6K/S6K 0.21 1.00 1.48 1.47 0.20 0.36 1.14 0.92 Fold change: pS6K/S6K 0.13 1.00 0.13 0.15 1.09 1.15 0.52 0.79 Fold change: pS6K/S6K 0.14 1.00 0.15 0.23 1.01 1.03 0.41 0.52 Fold change: p4E-BP1/4E-BP1 0.40 1.00 0.20 0.24 1.32 2.01 1.09 0.68 Fold change: p4E-BP1/4E-BP1 0.47 1.00 0.36 0.33 1.48 1.52 0.82 1.11

Fig. 3 | OSBP regulates mTORC1 upstream of the Rag GTPases. a, OSBP depletion abolishes sterol-dependent mTORC1 recruitment to lysosomes without compromising the integrity of the mTORC1 scaffolding complex. Quantification of co-localization of mTOR, p18 and RagC with LAMP2-positive lysosomes. Data are mean ± s.d.; 6 fields of view per genotype or condition; n represents the number of cells: mTOR for shLuc −Chol (n = 126), shLuc +Chol (n = 91), shOSBP −Chol (n = 115), shOSBP +Chol (n = 106); p18 for shLuc −Chol (n = 105), shLuc +Chol (n = 101), shOSBP −Chol (n = 92), shOSBP +Chol (n = 89); RagC for shLuc −Chol (n = 88), shLuc +Chol (n = 90), shOSBP −Chol (n = 77), shOSBP +Chol (n = 79); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 compared with shLuc +Chol (Supplementary Fig. 6h). b, OSBP mediates mTORC1 lysosomal recruitment upstream of the Rag GTPases. Control HEK293T and cells expressing the constitutively active RagBQ99L were depleted of OSBP and evaluated for sterol-dependent lysosomal localization of mTORC1. Scale bars, 10 μm. Insets in b represent regions of the respective panels magnified by 3.2 times. c, Quantification of co-localization of mTOR with LAMP2-positive lysosomes. Data are mean ± s.d.;10 fields of view per genotype or condition; n represents the number of cells: control-shLuc −Chol (n = 185), control-shLuc +Chol (n = 146), control-shOSBP −Chol (n = 154), control-shOSBP +Chol (n = 164); RagBQ99L-shLuc −Chol (n = 142), RagBQ99L-shLuc +Chol (n = 148), RagBQ99L-shOSBP −Chol (n = 128), RagBQ99L-shOSBP +Chol (n = 143); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001. d, OSBP activates mTORC1 signalling through the Rag GTPases. Cells were starved of cholesterol, refed where indicated and analysed for mTORC1 signalling. e,f, Loss of GATOR1 or KICSTOR largely rescues the effects of OSBP depletion on mTORC1 signalling. NPRL3-deleted (e) or KPTN-deleted (f) HEK293T cells were analysed for sterol-induced mTORC1 signalling. Fold changes of protein phosphorylation are shown in d–f. Experiments in d–f were repeated independently twice. Unprocessed blots are provided in Supplementary Fig. 9. Statistics source data are provided in Supplementary Table 2.

As OSBP resides at both ER–lysosome and ER–Golgi contacts, ER and, most importantly, Golgi failed to do so (Fig. 4d). Rescue we rigorously tested whether OSBP regulates mTORC1 through of mTORC1 signalling by lysosome-targeted OSBP required its lipid transport from ER to lysosomes and not elsewhere. We recon- lipid-transport activity, as mTORC1 signalling was abolished by stituted OSBP-depleted cells with OSBP isoforms in which the mutations in the ORD that disrupt either cholesterol or PtdIns4P PH domain is replaced with organelle-specific targeting signals, binding (Fig. 4e). while retaining attachment to the ER by VAP (Supplementary The OSBP transport cycle involves exchange of cholesterol Fig. 7d). Only OSBP fused to a lysosomal targeting sequence (the with PtdIns4P20,38. Consistent with this, OSBP depletion caused 39 N-terminal amino acids of p18)5,44 rescued mTORC1 activation strong accumulation of PtdIns4P on lysosomes, as shown using by cholesterol, whereas OSBP isoforms targeted to mitochondria, a genetically encoded PtdIns4P probe, GFP–P4M37 (Fig. 5a).

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

a c –Chol +Chol 1.5 Adjusted P = 0.0001 **** 1 807 FFAT 1.0 PH ORD PtdIns4P ER Cholesterol/PtdIns4P 0.5 binding targeting transport mTOR-positive LAMP2 signals 0 – WT ΔN ΔPH ORD Rescue: PH-FFAT b shRNA: OSBP shRNA: OSBP Rescue: – WT d mTOR LAMP2 mTOR LAMP2 shRNA: OSBP mTOR Lyso- Golgi- Mito- ER- Rescue: – WT ΔPH ΔPH ΔPH ΔPH ΔPH LAMP2

–Chol MCD + + + + + + + + + + + + + + Merge Cholesterol – + – + – + – + – + – + – + (kDa) mTOR 75 Low LAMP2 p-T389-S6K1 75

+Chol High Merge 75 S6K1 shRNA: OSBP Flag–GFP– WT 100 Rescue: N PH OSBP Δ Δ Endogenous 75 mTOR LAMP2 mTOR LAMP2 Flag 100 mTOR 75 50 LAMP2 Actin –Chol Merge e OSBP mTOR shRNA: Lyso- LAMP2 ΔPH- Lyso- Lyso- Lyso- H522A/ PH- PH- +Chol Δ Δ Merge Rescue: – WT ΔPH ΔPH H523A ΔELSK K736A shRNA: OSBP MCD + + + + + + + + + + + + + + Rescue: PH-FFAT ORD Cholesterol – + – + – + – + – + – + – + (kDa) mTOR LAMP2 mTOR LAMP2 75 Low mTOR p-T389-S6K1 75 High LAMP2 75

–Chol S6K1 Merge Flag–GFP– WT 100 OSBP mTOR Endogenous 75 100 LAMP2 Flag 75 +Chol Merge Actin 50

Fig. 4 | Both the lipid-transporting and lysosome-targeting functions of OSBP are required for mTORC1 activation by cholesterol. a, Domain organization of OSBP and related functions. b, The membrane-tethering and lipid-transfer functions of OSBP are required for lysosomal recruitment of mTORC1. OSBP-depleted cells and cells that were re-engineered as indicated were subjected to cholesterol starvation and restimulation, followed by immunofluorescence for endogenous mTOR and LAMP2. Scale bars, 10 µm. Insets in b represent regions of the respective panels magnified by 3.2 times. c, Quantification of co-localization of mTOR with LAMP2-positive lysosomes in cells reconstituted with different truncations as indicated. Data are mean ± s.d.; 10 fields of view per genotype or condition; n represents the number of cells: shOSBP −Chol (n = 173), shOSBP +Chol (n = 170), rescued WT −Chol (n = 185), rescued WT +Chol (n = 154), rescued ΔN −Chol (n = 179), rescued ΔN +Chol (n = 131), rescued ΔPH −Chol (n = 169), rescued ΔPH +Chol (n = 168), rescued PH-FFAT −Chol (n = 188), rescued PH-FFAT +Chol (n = 179), rescued ORD −Chol (n = 185), rescued ORD +Chol (n = 168); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 compared with rescued WT +Chol. d, Linking OSBP to the lysosome but not to other organelles restores the activation of mTORC1 by cholesterol. Cells were depleted of OSBP and reconstituted with PH-domain-deleted isoforms that carry organelle-targeting sequences and were cholesterol starved and restimulated. Cell lysates were analysed for the indicated proteins and phosphorylated proteins. Mito, mitochondria. e, The lipid-transfer activity of lysosome-targeted OSBP is essential for mTORC1 activation. Cells were depleted of OSBP and reconstituted with the indicated lyso-OSBP constructs and subjected to cholesterol starvation and restimulation. Cell lysates were analysed for the indicated proteins and phosphorylated proteins. Experiments in d and e were repeated independently twice. Unprocessed blots are provided in Supplementary Fig. 9. Statistics source data are provided in Supplementary Table 2.

To rule out a role of lysosomal PtdIns4P build-up in mTORC1 to contribute to the mTORC1 inhibition triggered by inactivation inhibition, we fused the catalytic domain of Sac1 phosphatase, of OSBP. which hydrolyses PtdIns4P to phosphatidylinositol, to the lyso- Finally, consistent with OSBP functioning at ER–lysosome con- some-anchoring sequence of p18 (refs. 5,37,44; lyso-Sac1; Fig. 5b). tacts, shRNA-mediated depletion of VAPA/B suppressed lysosomal In OSBP-depleted cells, lyso-Sac1 completely reversed lysosomal mTORC1 localization (Fig. 6a,b) and signalling (Fig. 6c). Although build-up of PtdIns4P caused by OSBP inactivation, whereas a other FFAT-containing proteins, including STARD3 and ORP1L16,45, catalytically inactive lyso-Sac1 mutant (C389S) failed to do so also transfer cholesterol across ER–lysosome contacts, they were (Fig. 5c). However, neither the catalytically active nor the inac- completely dispensable for cholesterol-dependent mTORC1 sig- tive lyso-Sac1 restored mTORC1 activation in OSBP-depleted nalling (Fig. 6d,e). Moreover, depletion of STARD3 or ORP1L did cells (Fig. 5d). Thus, lysosomal PtdIns4P accumulation is unlikely not reverse accumulation of cholesterol at the lysosomal surface in

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles

a b Lyso-Flag–GFP– Lyso-Flag–GFP– Sac1cat-WT Sac1cat-CS GFP– LAMP1– P4M mRFP cat

P4M

LAMP1 GFP– Sac 1

Sac1 GFP –Dox cat PtdIns4P PI

Merge Flag

P4M LAMP2 shOSBP-TetOn LAMP1

+Dox Lysosome Merge

c Lyso-Flag–GFP– Lyso-Flag–GFP– Merge

Sac1cat-WT Sac1cat-CS shRNA: Luciferase OSBP Luciferase OSBP

d Cells Lyso-Flag–GFP– Lyso-Flag–GFP– Control expressing: Sac1cat-WT Sac1cat-CS

shRNA: Luc OSBP Luc OSBP Luc OSBP mCherry–P4M MCD + + + + + + + + + + + cat Cholesterol – + – – + – + – + – + (kDa) 75 Low p-T389-S6K1 75 High GFP–Sac 1 S6K1 75 100 OSBP

Flag 75 Merge Actin 50

Fig. 5 | Excess lysosomal PtdIns4P after OSBP inhibition does not cause mTORC1 inhibition. a, Depletion of OSBP results in accumulation of PtdIns4P at lysosomes. HEK293A cells stably expressing the PtdIns4P-specific probe, GFP–P4M, along with LAMP1–mRFP were depleted of OSBP by doxycycline- induced shRNA. Cells were plated on glass-bottom dishes, allowed to attach overnight and imaged live on a spinning-disk confocal microscope. Representative microscopy images are shown. Scale bars, 10 μm. Insets in a represent regions of the respective panels magnified by 3.2 times. b, Confirmation of lysosomal localization of the lysosome-targeted Sac1 catalytic-domain (Sac1cat) truncations in HEK293A cells. Schematic diagram showing that the Sac1 catalytic domain is targeted to the lysosome to hydrolyse its target PtdIns4P (left). PI, phosphatidylinositol. Right, representative microscopy images are shown. Scale bars, 10 μm. c, Ectopic expression of a lysosome-targeted Sac1 catalytic domain eliminates excess lysosomal PtdIns4P in OSBP-depleted cells. HEK293T cells co-expressing mCherry–P4M and either catalytically active (WT) or inactive (CS) forms of lyso-Sac1 were depleted of OSBP and subjected to live-cell imaging. Representative confocal microscopy images are shown. Scale bars, 10 μm. d, Lysosomal build-up of PtdIns4P caused by OSBP depletion is not responsible for mTORC1 inhibition. Control cells and cells expressing catalytically active and inactive lyso- Sac1 were depleted of OSBP, followed by cholesterol starvation and restimulation, lysis and immunoblotting for the indicated proteins and phosphorylated proteins. Unprocessed blots are provided in Supplementary Fig. 9. Experiments in a–d were repeated independently twice.

NPC1-null cells, as shown by unchanged D4H*–mCherry staining proximity ligation assay (PLA) for VAPA and LAMP2, which (Fig. 6f,g). Thus, OSBP and VAPs form a cholesterol-transport- yields a fluorescent signal when the two membrane markers come ing system at ER–lysosome contacts that is uniquely involved in within 10–30 nm of each other. VAPA–LAMP2 PLA confirmed that mTORC1 activation by cholesterol. ER–lysosome contacts are significantly increased in NPC1-null compared with wild-type cells (Fig. 7a). NPC1 regulates the number of ER–lysosome contacts and their To determine the mechanistic basis for enhanced ER-to- sterol-transporting activity. Our D4H*–mCherry results strongly lysosome cholesterol transport, we carried out OSBP–VAP co- suggest that ER-to-lysosome cholesterol transport is increased immunoprecipitation experiments and found that the OSBP–VAP in NPC1-null cells. Consistent with this, ER–lysosome contacts interaction is stronger in NPC1-null than in wild-type cells. appeared enlarged in NPC1-null cells. In control human fibroblasts Interestingly, in wild-type cells, the OSBP–VAP interaction is pro- co-expressing the lysosomal and ER markers GFP–Tmem192 and moted by cholesterol, whereas it is constitutively high in NPC1- mCherry–Sec61b, respectively, physical contacts between lyso- deleted cells (Fig. 7b). Cholesterol may therefore promote its own somes and the ER could be readily visualized23,46 (Supplementary transport to the lysosome by increasing OSBP–VAP complex Fig. 8a). In NPC1mut fibroblasts, the ER domains in direct contact formation, whereas NPC1 inhibits this process leading to OSBP- with lysosomes appeared significantly enlarged and were more per- dependent accumulation of cholesterol on the limiting membranes sistent than those in control cells (Supplementary Fig. 8a). Further of NPC1-null lysosomes. investigation using electron microscopy (EM) showed numerous ER tubules located close to the characteristically enlarged NPC1- OSBP inactivation corrects aberrant mTORC1 signalling and null lysosomes, recognizable by the accumulation of undigested defective autophagy in NPC1-null cells. We previously observed membranes and electron-dense material within their lumen14 that, in NPC1-null cells, mTORC1 signalling is increased and (Supplementary Fig. 8b). As membrane contacts cannot be uni- resistant to cholesterol-depleting agents7. On the basis of the vocally identified by EM, we bolstered these observations with a D4H*–mCherry data and the increased formation of OSBP–VAP

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

a b shRNA: Luciferase VAPA and VAPB –Chol +Chol 1.5 mTOR LAMP2 mTOR LAMP2 Adjusted P = 0.0001 mTOR **** 1.0 LAMP2 –Chol Merge 0.5 mTOR-positive LAMP2 signals mTOR 0

LAMP2 Luc

+Chol shRNA: VAPA/B Merge

c d e shRNA: shRNA: Luc STARD3 OSBP Control sgORP1L Luc VAPA/B VAPA VAPB OSBP MCD + + + + + + MCD + + + + MCD + + + + + + + + + + Cholesterol – + – + – + Cholesterol – + – + – + – + – + (kDa) Cholesterol – + – + (kDa) 75 (kDa) 75 75 Low p-T389-S6K1 Low p-T389-S6K1 75 75 p-T389-S6K1 75 High S6K1 High 75 75 S6K1 S6K1 20 ORP1L 100 p-S65-4E-BP1 20 VAPA 25 4E-BP1 100 50 VAPB 25 OSBP ORP1S 100 50 50 OSBP STARD3 Actin 50 50 Actin Actin

f GST-D4H* g LAMP2 –mCherry Filipin P = 0.8341 LAMP2 Adjusted P = 0.0001 **** D4H* 1.5 P = 0.9304 Filipin

Luciferase Merge 1.0

LAMP2 0.5 filipin signals

D4H* D4H*-positive - shRNA

mut Filipin STARD3 0 Merge

NPC 1 Luc OSBP ORP1L LAMP2 STARD3 D4H* NPC1mut- shRNA NPC1mut- sgRNA

OSBP Filipin Merge

LAMP2 D4H* - sgRNA mut

ORP1L Filipin Merge NPC 1

Fig. 6 | VAPA and VAPB, the ER anchors for OSBP, are essential for cholesterol-dependent mTORC1 activation. a, Depletion of VAPA and VAPB abolishes the lysosomal recruitment of mTORC1 by cholesterol. HEK293T cells were depleted of VAPA and VAPB using shRNA and evaluated for mTOR lysosomal localization. Scale bars, 10 μm. Insets in a represent regions of the respective panels magnified by 3.2 times. b, Quantification of co-localization between mTOR- and LAMP2-positive lysosomes in control cells and cells depleted of VAPA and VAPB. Data are mean ± s.d.; 10 fields of view per genotype or condition; n represents the number of cells: shLuc −Chol (n = 124), shLuc +Chol (n = 74), shVAPA and shVAPB −Chol (n = 178), shVAPA and shVAPB +Chol (n = 163); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****P = 0.0001 compared with shLuc +Chol. c, VAPA and VAPB are necessary for cholesterol-dependent mTORC1 activation. HEK293T cells were depleted of VAPA and VAPB, either individually or in combination, and evaluated for sterol-induced mTORC1 signalling using immunoblotting. d,e, STARD3 and ORP1L are not required for cholesterol- dependent mTORC1 signalling. Cells depleted of STARD3 or OSBP using shRNA (d) and ORP1L knockout cells (e) were assayed for sterol-induced mTORC1 activation by immunoblotting. f, Depletion of STARD3 or ORP1L has no effects on both peripheral and internal pools of lysosomal cholesterol in NPC1-deficient cells. NPC1-deficient fibroblasts obtained from patients with NPC, in which STARD3 or OSBP was depleted using shRNA or in which ORP1L was deleted using CRISPR–Cas9 genome editing, were subjected to cholesterol labelling by GST–D4H*–mCherry and filipin, and stained for endogenous LAMP2. Scale bar, 10 µm. Insets in f represent regions of the respective panels magnified by 3.2 times. g, Quantification of co-localization of D4H*– mCherry with filipin-labelled cholesterol deposits in NPC1-null cells that were silenced for STARD3, OSBP and ORP1L, respectively. The box plots show the minimum, first quartile, median, third quartile and maximum; 10 fields of view per genotype; n represents the number of cells: shLuc (n = 12), shSTARD3 (n = 10), shOSBP (n = 10), sgORP1L (n = 11); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 compared with NPC1mut-shLuc. Experiments in c–e were performed independently twice with similar results. The unprocessed blots are provided in Supplementary Fig. 9. Statistics source data are provided in Supplementary Table 2.

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles

a b Control sgNPC1

VAPA + LAMP2 VAPA LAMP1 F359A/F360A F359A/F360A Flag–GFP– Flag–GFP– Flag–GFP– Flag–GFP– Flag–GFP– Flag–GFP– Cells expressing OSBP-WT OSBP-WT P = 2.02047 × 10–19 Tmem192 OSBP Tmem192 OSBP WT 2.0 MCD – + + – – + + – **** Cholesterol – – + – – – + – (kDa) NPC 1 1.5 VAP A 25 VAP B 25 1.0 IP: Flag 100 mut Flag 0.5 PLA signals (AU)

NPC 1 50

0 VAP A 25 WT mut VAP B 25 c sgNPC1ResWT sgNPC1 NPC1 NPC1 100 Flag Input 50 shRNA: Luc OSBP Luc OSBP 75 p-T389-S6K1 MCD + + + + + + + + 75 – + – + – + – + S6K1 Cholesterol (kDa) 75 NPC1 250 Low 50 p-T389-S6K1 75 Actin High 75 S6K1 d shRNA: Luciferase OSBP Flag 250 mTOR LAMP2 mTOR LAMP2 NPC1 250 mTOR 100 OSBP 50 LAMP2 Actin –Chol Merge ResWT e –Chol +Chol P = 0.9976 mTOR

P = 0.9996 sgNPC1 1.5 Adjusted P = 0.0001 Adjusted P = 0.0001 LAMP2

**** **** +Chol Merge 1.0 mTOR

0.5 LAMP2 –Chol mTOR-positive LAMP2 signals Merge 0 Luc Luc mTOR shRNA: OSBP OSBP sgNPC 1 sgNPC1ResWT sgNPC1 LAMP2 +Chol Merge

Fig. 7 | OSBP inhibition abrogates aberrant mTORC1 signalling in NPC1-deficient cells. a, Increased ER–lysosome contacts in NPC1 fibroblasts (left). Cells were fixed, permeabilized, labelled with VAPA or LAMP2 antibodies, or in combination as indicated, and analysed using PLA. Scale bars, 10 µm. Right, quantification of PLA signals. The box plots show the minimum, first quartile, median, third quartile and maximum; 60 fields of view per group; n represents the number of cells: NPC1WT (n = 60) and NPC1mut (n = 60). Statistical analysis was performed using two-tailed unpaired t-tests; ****P = 2.02047 × 10−19 compared with NPC1WT. b, Cholesterol regulates OSBP–VAP interaction in an NPC1-dependent manner. Control and NPC1-null cells expressing the Flag–GFP–Tmem192, Flag–GFP–OSBPWT and Flag–GFP–OSBPF359A/F360A were starved and refed with cholesterol, lysed and processed for Flag immunoprecipitation. Both immunoprecipitation and input samples were analysed by SDS–polyacrylamide gel electrophoresis (SDS–PAGE) and immunoblotting for the indicated proteins. The experiments were repeated twice with similar results. c, Depletion of OSBP abolishes constitutive mTORC1 signalling in NPC1-null cells. NPC1-deleted HEK293T cells, either naive or reconstituted (sgNPC1ResWT) with stably expressed Flag-tagged NPC1, were depleted of OSBP and analysed for sterol-induced mTORC1 signalling by immunoblotting. The experiments were repeated twice. d,e, OSBP depletion abrogates constitutive mTORC1 lysosomal localization in NPC1-null cells. NPC1-deleted HEK293T cells, either naive or reconstituted with stably expressed Flag-tagged NPC1, were depleted of OSBP and evaluated for sterol-dependent mTORC1 lysosomal localization. Scale bars, 10 μm. Insets in d represent regions of the respective panels magnified by 3.2 times. For e, data are mean ± s.d.; 10 fields of view per genotype or condition; n represents the number of cells: sgNPC1ResWT-shLuc −Chol (n = 166), sgNPC1ResWT-shLuc +Chol (n = 110), sgNPC1ResWT-shOSBP −Chol (n = 181), sgNPC1ResWT-shOSBP +Chol (n = 132), sgNPC1-shLuc −Chol (n = 127), sgNPC1-shLuc +Chol (n = 142), sgNPC1-shOSBP −Chol (n = 147) and sgNPC1-shOSBP +Chol (n = 131); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001. The unprocessed blots are provided in Supplementary Fig. 9. Statistics source data are provided in Supplementary Table 2. complexes (Fig. 7a,b, Supplementary Fig. 8a,b), we hypothesized (Fig. 7c). Moreover, OSBP depletion suppressed constitutive lyso- that constitutive mTORC1 activation may be driven by OSBP- somal mTORC1 localization in NPC1-deleted cells (Fig. 7d,e). dependent accumulation of cholesterol on the limiting membrane A major downstream consequence of aberrant activation of of NPC1-null lysosomes. mTORC1 is suppression of autophagy. Indeed, as in other neuro- Consistent with this model, knockdown of OSBP resulted in degenerative disorders, autophagic flux is decreased in NPC and it the complete inactivation of mTORC1, irrespective of NPC1 status may drive the defective cellular quality control that contributes to

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

a Adjusted P = 0.0001 b Human **** fibroblasts NPC1WT NPC1mut 2.5 Adjusted P = 0.0001 Adjusted P = 0.0001 **** **** OSW-1 – + – + + – + – – 2.0 Torin1 – – – – + + – – – BafA1 – – – – – – + + – 1.5 Amino acid – – – – – – – – + starvation (kDa) 75 1.0 p62 GFPRFP l 15 ll Low 0.5 LC3B l 15 ll High 0 NPC1 250 Actin 50 BafA1 Torin1 BafA1 Torin1 DMSO OSW-1 DMSOOSW-1 Fold change: 0.13 1.00 0.31 1.03 0.98 0.98 1.06 0.99 1.03 LC3B-ll/actin Npc1+/+ Npc1–/–

c Human e P = 0.8715 fibroblasts NPC1WT NPC1mut P = 0.7858 1.5 Adjusted P = 0.0001 shRNA: Luc OSBP Luc OSBP **** 1 2 3 1 2 3 1 2 3 1 2 3 (kDa) 75 1.0 p62

l LC3B 15 ll 0.5 p62-positive

100 LAMP2 signals OSBP

NPC1 250 0 50 Actin Torin1 BafA1 BafA1 DMSO DMSO OSW-1 Fold change: 0.08 0.08 0.12 0.99 0.99 1.02 0.64 0.97 0.84 1.22 1.96 1.36 LC3B-ll/actin + OSW-1

NPC1WT NPC1mut d p62 LAMP2

p62

WT f LAMP2 PM Naive DMSO NPC 1 ER Merge

Growth Autophagy p62

LAMP2 ? VAP OSBP

DMSO Rag mTORC1 Merge Ragulator NPC1

p62 SLC38A9 LAMP2 OSW-1 Merge Lysosome

p62

mut Niemann–Pick type C LAMP2 ER Torin1 NPC 1 Merge Growth Autophagy

p62

VAP OSBP Rag LAMP2 mTORC1 Ragulator BafA1 NPC1 Merge

p62 SLC38A9

LAMP2 Lysosome Merge BafA1 + OSW-1

Fig. 8 | OSBP inhibition restores autophagic function in NPC1-deficient cells. a, OSW-1 increases autophagic flux in both Npc1+/+ and Npc1−/− MEFs expressing the GFP-LC3-RFP-LC3ΔG reporter. The normalized GFP/RFP fluorescence ratio is expressed as fold change over DMSO-treated Npc1+/+ cells. Data are mean ± s.d.; n represents number of biologically independent samples: Npc1+/+ DMSO (n = 7), Npc1+/+ OSW-1 (n = 7), Npc1+/+ BafA1 (n = 7), Npc1+/+ Torin1 (n = 7), Npc1−/− DMSO (n = 6), Npc1−/− OSW-1 (n = 6), Npc1−/− BafA1 (n = 6) and Npc1−/− Torin1 (n = 6); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 as indicated. b, OSW-1 treatment abolishes p62 accumulation in NPC1 fibroblasts. Cells were subjected to the indicted treatments and immunoblotted for p62 and LC3B. The experiments were repeated twice. c, OSBP depletion reduces p62 accumulation in NPC1-deficient cells obtained from patients with NPC. Cells were depleted of OSBP and analysed by immunoblotting; n = 3 biologically independent samples per genotype. Fold changes of LC3B-II to actin in b and c are indicated. The experiment was performed once. d,e, p62 clearance by OSW-1 in NPC1 fibroblasts. Cells were subjected to the indicated treatments for 8 h and stained for p62 and LAMP2. Scale bars, 10 μm. Insets in d represent regions of the respective panels magnified by 3.2 times. For e, data are mean ± s.d.; 10 fields of view per genotype or treatment; n represents the number of cells: NPC1WT DMSO (n = 14), NPC1mut DMSO (n = 10), NPC1mut OSW-1 (n = 11), NPC1mut Torin1 (n = 11), NPC1mut BafA1 (n = 11) and NPC1mut BafA1 + OSW-1 (n = 12); statistical analysis was performed using ANOVA with Dunnett’s multiple comparison test; ****adjusted P = 0.0001 compared with DMSO-treated NPC1mut cells. Statistics source data are provided in Supplementary Table 2. f, Across ER–lysosome contacts, OSBP–VAP generates a peripheral pool of lysosomal cholesterol that recruits and activates mTORC1 through the Rag GTPases. By contrast, NPC1 suppresses mTORC1 by limiting the levels of peripheral cholesterol. The steady-state levels of mTORC1 activation are therefore balanced by the opposite transport route of cholesterol. In cells lacking NPC1, cholesterol accumulates both in the lysosomal lumen and by unopposed ER-to-lysosome transport via OSBP on the lysosomal membrane. In turn, excess lysosomal-membrane cholesterol drives constitutively increased mTORC1 signalling and inhibits autophagy. PM, plasma membrane.

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles this disease29,47,48. Using a cleavable, dual-colour LC3 reporter49, we at the lysosomal surface and interacting with the mTORC1 scaf- verified that autophagic flux is reduced in Npc1−/− MEFs compared folding complex7. Subsequently, OSBP–VAP may return cholesterol with wild-type MEFs (Fig. 8a). Notably, OSBP inhibition using to the lysosomal surface across ER–lysosome contacts in a con- OSW-1 increased autophagic flux both in Npc1+/+ and Npc1−/− cells trolled manner. Consistent with this model, exogenously delivered with potency that was similar to that of the catalytic mTOR inhibi- TopFluor-cholesterol requires OSBP to exit the ER and reach the tor, Torin1 (Fig. 8a). lysosomal limiting membrane. NPC1-defective MEFs and fibroblasts obtained from patients Alternatively, NPC1 may promote the lateral segregation of lyso- showed pronounced accumulation of the autophagic adapter somal-membrane cholesterol into domains analogous to the ones p62 (also known as SQSTM1), indicating disrupted autopha- reported on the yeast vacuole, and which may not be accessible to gic degradation50. OSW-1 resolved p62 accumulation in a dose- the mTORC1 scaffolding complex59,60. Ultimately, mTORC1 activity dependent manner, decreasing its levels to those of control cells results from the balance between the sterol-transport activities of (Supplementary Fig. 8c,d), as did shRNA-mediated OSBP knock- OSBP–VAP and NPC1. Although we were unable to detect a direct down (Fig. 8c). OSW-1-induced clearance of p62 was prevented interaction between OSBP and NPC1 in co-immunoprecipitation by the vacuolar H+-ATPase inhibitor, bafilomycin A1 (BafA1; studies, we did find that NPC1 status affects the interaction between Fig. 8b, Supplementary Fig. 8e). Importantly, mTORC1 inhibi- OSBP and VAPA/B. NPC1 may therefore exert feedback regulation tion by Torin1 was as effective as OSW-1 in clearing p62, and their on OSBP, possibly through its cholesterol transport activity. Further combined treatment provided no additional benefit, suggesting studies are required to shed light on the mechanistic basis for this that both compounds act through a mechanism converging on regulation. More broadly, these data support the idea that ER–lyso- mTORC1 (Fig. 8b, Supplementary Fig. 8f). some contacts are bona fide signalling structures that can modulate Immunofluorescence for p62 in human patient fibroblasts the rate of cholesterol transport in response to physiological stim- showed considerable accumulation of intracellular aggregates in uli. Moreover, given that neither NPC1 nor OSBP are required for NPC1mut cells that partially overlapped with LAMP2-positive lyso- acute mTORC1 response to amino acids7 (Supplementary Fig. 6g), somes (Fig. 8d,e). Consistent with the immunoblotting results, ER–lysosome contacts are probably dedicated to the regulation of intracellular p62 aggregates were largely cleared by treatment with mTORC1 by cholesterol inputs. OSW-1 and Torin1, and this effect was reversed by treatment with By modulating the cholesterol content of the lysosomal limit- BafA1 (Fig. 8d,e). Together, these data demonstrate that inhibiting ing membrane, OSBP has a profound influence on the initiation aberrant OSBP-driven mTORC1 signalling in NPC cells is sufficient and execution of autophagy. This was especially evident in cellular to restore delivery of autophagic cargo to the lysosome and its sub- models of NPC, in which OSBP inhibition was sufficient to restore sequent degradation. autophagic flux, leading to clearance of the canonical autophagic substrate, p62. Previous studies in cells obtained from patients have Discussion uncovered defects in both autophagosome–lysosome fusion and in Here we identify OSBP as the key component of a transport cir- the subsequent breakdown of autophagic substrates29,47,48,61. These cuit that coordinately regulates cholesterol levels and mTORC1 defects can be explained, at least in part, by the observed accumula- activation at the lysosomal membrane (Fig. 8f). In concert with its tion of cholesterol on the lysosomal limiting membrane (Fig. 1b,c), ER membrane anchors, VAPA/B, OSBP delivers cholesterol in the which could lead to mistrafficking of lysosomes and compromise ER-to-lysosome direction in exchange for PtdIns4P22,38. Cholesterol their ability to fuse with other organelles24,62,63. In the context of deposited on the lysosome by OSBP may directly interact with NPC, OSBP inhibition restores autophagic function, probably by the mTORC1 scaffolding machinery that includes SLC38A9, reducing cholesterol build-up at the lysosomal membrane, which Ragulator and vacuolar H+-ATPase, leading to Rag GTPase activa- has the dual effect of enhancing autophagy initiation (by mTORC1 tion and mTORC1 recruitment5,7,10,31 (Fig. 8f). Notably, OSBP was inhibition) and, potentially, restoring lysosomal trafficking and unique among VAP-interacting proteins in directly controlling fusion with autophagosomes. mTORC1 activation. The partial decrease in mTORC1 signalling The availability of nanomolar OSBP inhibitors, such as OSW-133, caused by depleting ORP8 and ORP11 probably occurs by indi- suggests that chemical modulation of OSBP could be a promising rect mechanisms that are not connected to lysosomal cholesterol strategy to correct lysosomal function and to restore proteostasis trafficking16,45,51–53. in NPC, as well as in diseases driven by excess mTORC1 signalling. In yeast, intracellular transport of ergosterol has been impli- cated in TORC1 regulation by nitrogen, as well as its inactivation Online content under stress54,55. Through the transport action of OSBP–VAP, ER– Any methods, additional references, Nature Research reporting lysosome contacts in metazoan cells emerge as signalling nodes in summaries, source data, statements of code and data availability and which information about cholesterol availability is integrated and associated accession codes are available at https://doi.org/10.1038/ translated into mTORC1-driven growth programs. As the majority s41556-019-0391-5. of endolysosomes contact the ER over time scales of several min- utes, the regulatory action of these junctions on mTORC1 signalling Received: 25 February 2019; Accepted: 15 August 2019; is expected to be pervasive and to couple the pace of cell growth to Published: xx xx xxxx the levels of intracellular cholesterol2,3. Our results provide evidence of how NPC1 regulates lysosomal References cholesterol levels, a problem with important implications for meta- 1. Gatta, A. T. & Levine, T. P. Piecing together the patchwork of contact sites. bolic and neurodegeneration research14,56. After accepting luminal Trends Cell Biol. 27, 214–229 (2017). 2. Phillips, M. J. & Voeltz, G. K. Structure and function of ER membrane cholesterol from NPC2, NPC1 may facilitate its insertion into the contact sites with other organelles. Nat. Rev. Mol. Cell Biol. 17, 69–82 (2016). lysosomal limiting membrane for subsequent export, or directly 3. Telen, A. M. & Zoncu, R. Emerging roles for the lysosome in lipid hand it over to acceptors that remain unidentified11–13,57. How does metabolism. Trends Cell Biol. 27, 833–850 (2017). cholesterol handling by NPC1 lead to mTORC1 inhibition? Recent 4. Kim, E., Goraksha-Hicks, P., Li, L., Neufeld, T. P. & Guan, K. L. Regulation evidence indicates that LDL-derived cholesterol is rapidly removed of TORC1 by Rag GTPases in nutrient response. Nat. Cell Biol. 10, 935–945 (2008). from the lysosome and transported to the plasma membrane in an 5. Sancak, Y. et al. Ragulator-Rag complex targets mTORC1 to the lysosomal 58 NPC1-dependent manner . Rapid NPC1-dependent export may surface and is necessary for its activation by amino acids. Cell 141, prevent cholesterol that is liberated from LDL from accumulating 290–303 (2010).

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

6. Sancak, Y. et al. Te Rag GTPases bind raptor and mediate amino acid 39. Hammond, G. R., Machner, M. P. & Balla, T. A novel probe for signaling to mTORC1. Science 320, 1496–1501 (2008). phosphatidylinositol 4-phosphate reveals multiple pools beyond the Golgi. 7. Castellano, B. M. et al. Lysosomal cholesterol activates mTORC1 via an J. Cell Biol. 205, 113–126 (2014). SLC38A9-Niemann-Pick C1 signaling complex. Science 355, 1306–1311 (2017). 40. Perry, R. J. & Ridgway, N. D. Oxysterol-binding protein and 8. Jung, J., Genau, H. M. & Behrends, C. Amino acid-dependent mTORC1 vesicle-associated membrane protein-associated protein are required for regulation by the lysosomal membrane protein SLC38A9. Mol. Cell. Biol. 35, sterol-dependent activation of the ceramide transport protein. Mol. Biol. Cell 2479–2494 (2015). 17, 2604–2616 (2006). 9. Rebsamen, M. et al. SLC38A9 is a component of the lysosomal amino acid 41. Bar-Peled, L. et al. A tumor suppressor complex with GAP activity for the sensing machinery that controls mTORC1. Nature 519, 477–481 (2015). Rag GTPases that signal amino acid sufciency to mTORC1. Science 340, 10. Wang, S. et al. Lysosomal amino acid transporter SLC38A9 signals arginine 1100–1106 (2013). sufciency to mTORC1. Science 347, 188–194 (2015). 42. Panchaud, N., Peli-Gulli, M. P. & De Virgilio, C. Amino acid deprivation 11. Gong, X. et al. Structural insights into the Niemann-Pick C1 (NPC1)- inhibits TORC1 through a GTPase-activating protein complex for the Rag mediated cholesterol transfer and ebola infection. Cell 165, 1467–1478 (2016). family GTPase Gtr1. Sci. Signal. 6, ra42 (2013). 12. Kwon, H. J. et al. Structure of N-terminal domain of NPC1 reveals distinct 43. Wolfson, R. L. et al. KICSTOR recruits GATOR1 to the lysosome and is subdomains for binding and transfer of cholesterol. Cell 137, 1213–1224 (2009). necessary for nutrients to regulate mTORC1. Nature 543, 438–442 (2017). 13. Li, X., Saha, P., Li, J., Blobel, G. & Pfefer, S. R. Clues to the mechanism of 44. Menon, S. et al. Spatial control of the TSC complex integrates insulin and cholesterol transfer from the structure of NPC1 middle lumenal domain nutrient regulation of mTORC1 at the lysosome. Cell 156, 771–785 (2014). bound to NPC2. Proc. Natl Acad. Sci. USA 113, 10079–10084 (2016). 45. Zhao, K. & Ridgway, N. D. Oxysterol-binding protein-related protein 1L 14. Maxfeld, F. R. Role of endosomes and lysosomes in human disease. regulates cholesterol egress from the endo-lysosomal system. Cell Rep. 19, Cold Spring Harb. Perspect. Biol. 6, a016931 (2014). 1807–1818 (2017). 15. Radhakrishnan, A., Goldstein, J. L., McDonald, J. G. & Brown, M. S. 46. Hoyer, M. J. et al. A novel class of ER membrane proteins regulates Switch-like control of SREBP-2 transport triggered by small changes in ER er-associated endosome fssion. Cell 175, 254–265 (2018). cholesterol: a delicate balance. Cell Metab. 8, 512–521 (2008). 47. Ordonez, M. P. et al. Disruption and therapeutic rescue of autophagy in a 16. Wilhelm, L. P. et al. STARD3 mediates endoplasmic reticulum-to-endosome human neuronal model of Niemann Pick type C1. Hum. Mol. Genet. 21, cholesterol transport at membrane contact sites. EMBO J. 36, 1412–1433 (2017). 2651–2662 (2012). 17. Im, Y. J., Raychaudhuri, S., Prinz, W. A. & Hurley, J. H. Structural mechanism 48. Pacheco, C. D., Kunkel, R. & Lieberman, A. P. Autophagy in Niemann-Pick C for sterol sensing and transport by OSBP-related proteins. Nature 437, disease is dependent upon Beclin-1 and responsive to lipid trafcking defects. 154–158 (2005). Hum. Mol. Genet. 16, 1495–1503 (2007). 18. Tong, J., Yang, H., Yang, H., Eom, S. H. & Im, Y. J. Structure of Osh3 reveals 49. Kaizuka, T. et al. An autophagic fux probe that releases an internal control. a conserved mode of phosphoinositide binding in oxysterol-binding proteins. Mol. Cell 64, 835–849 (2016). Structure 21, 1203–1213 (2013). 50. Bjorkoy, G. et al. Monitoring autophagic degradation of p62/SQSTM1. 19. Mesmin, B., Antonny, B. & Drin, G. Insights into the mechanisms of sterol Methods Enzymol. 452, 181–197 (2009). transport between organelles. Cell. Mol. Life Sci. 70, 3405–3421 (2013). 51. Chung, J. et al. PI4P/phosphatidylserine countertransport at ORP5- and 20. Mesmin, B. et al. Sterol transfer, PI4P consumption, and control of membrane ORP8-mediated ER–plasma membrane contacts. Science 349, 428–432 (2015). lipid order by endogenous OSBP. EMBO J. 36, 3156–3174 (2017). 52. Du, X. et al. A role for oxysterol-binding protein-related protein 5 in 21. Goto, A., Charman, M. & Ridgway, N. D. Oxysterol-binding protein activation endosomal cholesterol trafcking. J. Cell Biol. 192, 121–135 (2011). at endoplasmic reticulum-Golgi contact sites reorganizes phosphatidylinositol 53. Du, X. et al. Oxysterol-binding protein-related protein 5 (ORP5) promotes 4-phosphate pools. J. Biol. Chem. 291, 1336–1347 (2016). cell proliferation by activation of mTORC1 signaling. J. Biol. Chem. 293, 22. Dong, R. et al. Endosome-ER contacts control actin nucleation and 3806–3818 (2018). retromer function through VAP-dependent regulation of PI4P. Cell 166, 54. Mousley, C. J. et al. A sterol-binding protein integrates endosomal lipid 408–423 (2016). metabolism with TOR signaling and nitrogen sensing. Cell 148, 702–715 (2012). 23. Rowland, A. A., Chitwood, P. J., Phillips, M. J. & Voeltz, G. K. ER contact sites 55. Murley, A. et al. Sterol transporters at membrane contact sites regulate defne the position and timing of endosome fssion. Cell 159, 1027–1041 (2014). TORC1 and TORC2 signaling. J. Cell Biol. 216, 2679–2689 (2017). 24. Rocha, N. et al. Cholesterol sensor ORP1L contacts the ER protein VAP to 56. Ballabio, A. & Gieselmann, V. Lysosomal disorders: from storage to cellular control Rab7-RILP-p150 Glued and late endosome positioning. J. Cell Biol. damage. Biochim. Biophys. Acta 1793, 684–696 (2009). 185, 1209–1225 (2009). 57. Infante, R. E. et al. NPC2 facilitates bidirectional transfer of cholesterol 25. Willett, R. et al. TFEB regulates lysosomal positioning by modulating between NPC1 and lipid bilayers, a step in cholesterol egress from lysosomes. TMEM55B expression and JIP4 recruitment to lysosomes. Nat. Commun. 8, Proc. Natl Acad. Sci. USA 105, 15287–15292 (2008). 1580 (2017). 58. Infante, R. E. & Radhakrishnan, A. Continuous transport of a small fraction 26. Lebrand, C. et al. Late endosome motility depends on lipids via the small of plasma membrane cholesterol to endoplasmic reticulum regulates total GTPase Rab7. EMBO J. 21, 1289–1300 (2002). cellular cholesterol. eLife 6, e25466 (2017). 27. Maekawa, M. & Fairn, G. D. Complementary probes reveal that 59. Toulmay, A. & Prinz, W. A. Direct imaging reveals stable, micrometer-scale phosphatidylserine is required for the proper transbilayer distribution of lipid domains that segregate proteins in live cells. J. Cell Biol. 202, 35–44 (2013). cholesterol. J. Cell. Sci. 128, 1422–1433 (2015). 60. Tsuji, T. et al. Niemann-Pick type C proteins promote microautophagy 28. Lofus, S. K. et al. Murine model of Niemann-Pick C disease: mutation in a by expanding raf-like membrane domains in the yeast vacuole. eLife 6, cholesterol homeostasis gene. Science 277, 232–235 (1997). e25960 (2017). 29. Sarkar, S. et al. Impaired autophagy in the lipid-storage disorder Niemann- 61. Maetzel, D. et al. Genetic and chemical correction of cholesterol accumulation Pick type C1 disease. Cell Rep. 5, 1302–1315 (2013). and impaired autophagy in hepatic and neural cells derived from Niemann- 30. Lu, F. et al. Identifcation of NPC1 as the target of U18666A, an inhibitor of Pick Type C patient-specifc iPS cells. Stem Cell Rep. 2, 866–880 (2014). lysosomal cholesterol export and Ebola infection. eLife 4, e12177 (2015). 62. Fraldi, A. et al. Lysosomal fusion and SNARE function are impaired by 31. Zoncu, R. et al. mTORC1 senses lysosomal amino acids through an inside-out cholesterol accumulation in lysosomal storage disorders. EMBO J. 29, mechanism that requires the vacuolar H+-ATPase. Science 334, 678–683 (2011). 3607–3620 (2010). 32. Abu-Remaileh, M. et al. Lysosomal metabolomics reveals V-ATPase- and 63. Settembre, C. et al. A block of autophagy in lysosomal storage disorders. mTOR-dependent regulation of amino acid efux from lysosomes. Science Hum. Mol. Genet. 17, 119–129 (2008). 358, 807–813 (2017). 33. Burgett, A. W. et al. Natural products reveal cancer cell dependence on oxysterol-binding proteins. Nat. Chem. Biol. 7, 639–647 (2011). Acknowledgements 34. Levine, T. P. & Munro, S. Te pleckstrin homology domain of oxysterol- We thank all of the members of the Zoncu laboratory for their insights; R. Perera and binding protein recognises a determinant specifc to Golgi membranes. M. Rape for reading the manuscript; M. D. Shair (Harvard University) for providing us Curr. Biol. 8, 729–739 (1998). with the OSW-1 compound; G. Fairn (University of Toronto) for the D4H cholesterol 35. Ridgway, N. D., Dawson, P. A., Ho, Y. K., Brown, M. S. & Goldstein, J. L. probe cDNA; G. Hammond (University of Pittsburgh) and S. Grinstein (Hospital for Translocation of oxysterol binding protein to Golgi apparatus triggered by Sick Children) for the P4M reporter and the scSac1 catalytic domain cDNAs; H. Arai ligand binding. J. Cell Biol. 116, 307–319 (1992). and N. Kono (University of Tokyo) for the anti-OSBP rabbit polyclonal antibody. 36. Su, M. Y. et al. Hybrid structure of the RagA/C-Ragulator mTORC1 C.-Y.L. is the recipient of an American Heart Association postdoctoral fellowship activation complex. Mol. Cell 68, 835–846 (2017). (18POST34070059). This work was supported by the NIH Director’s New Innovator 37. Levin, R. et al. Multiphasic dynamics of phosphatidylinositol 4-phosphate Award (1DP2CA195761-01), the Pew–Stewart Scholarship for Cancer Research, the during phagocytosis. Mol. Biol. Cell 28, 128–140 (2017). Damon Runyon-Rachleff Innovation Award, the Edward Mallinckrodt Jr. Foundation 38. Mesmin, B. et al. A four-step cycle driven by PI(4)P hydrolysis directs sterol/ Grant and the Packer Wentz Endowment (to R.Z.), a National Institutes of Health PI(4)P exchange by the ER-Golgi tether OSBP. Cell 155, 830–843 (2013). grant (R01 HL067773; to D.S.O).

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles

Author contributions Additional information C.-Y.L. and R.Z. conceived the study. C.-Y.L., O.B.D., H.R.S. and R.Z. designed Supplementary information is available for this paper at https://doi.org/10.1038/ experiments. C.-Y.L., H.R.S., O.B.D. and J.Z. performed experiments. X.J., C.A.B. and s41556-019-0391-5. J.L.C. performed mass spectrometry measurements. D.K.N. and D.S.O. provided advice Correspondence and requests for materials should be addressed to R.Z. on experimental design and data analysis. C.-Y.L. and R.Z. wrote the manuscript. All of the authors reviewed and edited the manuscript. Reprints and permissions information is available at www.nature.com/ reprints. Competing interests R.Z. is a co-founder, consultant and stockholder of Frontier Medicines Corporation. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in D.K.N. is a co-founder, stock-holder and scientific adviser for Artris Therapeutics and published maps and institutional affiliations. Frontier Medicines Corporation. © The Author(s), under exclusive licence to Springer Nature Limited 2019

Nature Cell Biology | www.nature.com/naturecellbiology Articles NATuRE CEll BIOlOGy

Methods with DMEM supplemented with 0.1% MCD and 0.5% LDS and 50 μM cholesterol 64 Antibodies and chemicals. Details of the antibodies used in this study are for 2 h. LDS was prepared as previously described . provided in Supplementary Table 3. LDL (J65039) was obtained from Alfa Aesar. Amino acids, lysozyme (62970), cholesterol (C3045), flipin (F9765), MCD Cell lysis and immunoprecipitation. Cells were rinsed with cold PBS and lysed (C4555), BafA1 (B1793) and U18666A (U3633) were obtained from Sigma- in lysis buffer (1% Triton X-100, 10 mM β-glycerol phosphate, 10 mM sodium Aldrich. TopFluor-cholesterol (810255P) was obtained from Avanti Polar Lipids. pyrophosphate, 4 mM EDTA, 40 mM HEPES, pH 7.4, and one EDTA-free Torin1 (4247) was obtained from Tocris Bioscience. Pierce protease inhibitor protease inhibitor tablet per 50 ml). Cell lysates were cleared by centrifugation in a tablets (A32965) were obtained from Termo Fisher Scientifc. Glutathione microcentrifuge at 17,000g for 10 min at 4 °C. Cell lysate samples were prepared by Sepharose 4B (17-0756-05) was obtained from Pierce. LysoSensor Green DND-189 addition of 5× sample buffer, heated at 70 °C for 10 min, resolved by either 10% or (L7535) and DMEM (11965) were obtained from Life Technologies. Amino acid- 12% SDS–PAGE gels, and analysed by immunoblotting. free RPMI (R9010-01) was obtained from US Biologicals. OSW-1 was a gif from For Flag immunoprecipitations, 25 μl of a well-mixed 50% slurry of anti- 1 M. D. Shair (Harvard University). Flag M2 Affinity Gel (Sigma, A2220) was added to each lysate (1 mg ml− ) and incubated at 4 °C in a shaker for 2 h. The immunoprecipitates were washed five Mammalian cell culture. Human HEK293A, HEK293T, HEK293T NPC1-null, times, twice with lysis buffer and three times with lysis buffer with 500 mM NaCl. HEK293T NPRL3-null and HEK293T KPTN-null cells, and Npc1+/+ and Npc1−/− Immunoprecipitated proteins were denatured by adding 50 μl of 2× urea sample MEFs were cultured in DMEM supplemented with 10% (v/v) inactivated fetal buffer and heating to 37 °C for 15 min, resolved by SDS–PAGE and analysed by bovine serum (Sigma-Aldrich), penicillin (100 U ml−1), streptomycin (100 μg ml−1) immunoblotting. and 2 mM l-glutamine. Human NPC1WT (GM23151) and NPC1mut fibroblasts (GM03123) were obtained from Coriell Cell Repositories in full compliance with Immunofluorescence. HEK293T or HEK293A cells were seeded on fibronectin- the Office for Human Research Protections, Department of Health and Human coated glass coverslips in 6-well plates (35 mm diameter per well) at 300,000– Services regulations. The statement of research intent and the signed MTA were 500,000 cells per well and allowed to attach overnight. The next day, cells were submitted and approved by the Coriell Institutional Review Board for use of patient subjected to sterol depletion and restimulation, where indicated, and fixed in 4% cells in this study. Human fibroblasts were grown in DMEM supplemented with paraformaldehyde (PFA) for 15 min at room temperature. The coverslips were

15% fetal bovine serum. All of the cell lines were maintained at 37 °C and 5% CO2. rinsed twice with PBS and cells were permeabilized with 0.1% (w/v) saponin in PBS The cells were free of mycoplasma contamination, and routinely checked using for 10 min. After rinsing twice with PBS, the slides were incubated with primary MycoAlert Mycoplasma Detection Kit (Lonza, LT07-318) and Hoechst staining. antibodies in 5% normal donkey serum (Jackson ImmunoResearch, 017-000-121) for 1 h at room temperature, rinsed four times with PBS and incubated with Plasmids, viruses and stable cell lines. The pLKO.1 lentiviral vector (The RNAi fluorophore-conjugated secondary antibodies derived from goat or donkey (Life Consortium, Broad institute) was used to express the shRNA against the proteins Technologies, diluted 1:400 in 5% normal donkey serum, PBS) for 45 min at room of interest: OSBP, TRCN0000155752; ORP8, TRCN0000147289; ORP11, temperature in the dark. After washing four times with PBS and one time with TRCN0000161454; ORP5, TRCN0000180915; VAPA, TRCN0000029130; VAPB, deionized water, the coverslips were mounted on glass slides using Fluoromount-G TRCN0000151599; STARD3, TRCN0000150515; Luciferase as non-targeting control, (SouthernBiotech, 0100-01) and imaged using a spinning-disk confocal system built TRCN0000072243. Where indicated, a doxycycline-inducible shRNA system using on a Nikon Eclipse Ti microscope with Andor Zyla-4.5 sCMOS camera. Tet-pLKO-puro (Addgene, 21915) was used to silence OSBP using doxycycline. For protein expression, cDNA encoding proteins was cloned into the pLJM1 Live-cell imaging of PtdIns4P distribution. Cells expressing the PtdIns4P lentiviral vector carrying a Flag tag or a Flag tag followed by a GFP marker fluorescent reporter (either GFP–P4M or mCherry–P4M) were seeded on 35 mm accordingly. OSBP cDNA was codon-optimized and its variants (R108L, F359A/ glass-bottom dishes (MatTeK, P35GC-1.5-14-C), allowed to attach overnight and F360A, H522A/H523A, ΔELSK (deletion of residues 430–433) and K736A) were imaged using a spinning-disk microscope. Before imaging, cells were transferred generated using Agilent QuikChange II site-directed mutagenesis. For organelle- to imaging buffer containing 125 mM NaCl, 5 mM KCl, 1.3 mM CaCl2, 1.2 mM targeted OSBP truncations, the coding sequence of p18 N-terminal 1–39 amino MgCl2, 25 mM HEPES and 5 mM d-glucose with pH adjusted to 7.4 with NaOH. acids was inserted into the N terminus of OSBP ΔPH cDNA to generate Lyso- ΔPH-OSBP; Golgi-ΔPH was generated by fusing the coding sequence of the Co-localization analysis. For quantification of cells with the phenotype of interest, Giantin C-terminal 3131–3259 amino acids to the C terminus of ΔPH; Mito-ΔPH 20–25 non-overlapping whole-field images were randomly acquired throughout was generated by inserting the coding sequence of amino acids 1–145 of TOM20 the slide of each sample. All of the RAW images were exported from the Andor into the N terminus of ΔPH; and ER-ΔPH was generated by inserting the coding iQ3 imaging software v.3.3.1 (Andor Technology, Oxford Instruments). Fiji v.1.51j sequence of the 860–887 amino acids of the ORP8 C-terminal transmembrane (ImageJ, NIH) was used to quantify the fluorescence intensity (https://fiji.sc). region 860–887 into the C terminus of ΔPH. To determine the co-localization of mTOR, OSBP or p62 with LAMP2-positive GFP–P4M and mCherry–P4M cDNAs were obtained from G. Hammond lysosomes, or co-labelling of D4H*-labelled vesicles with the intracellular filipin and cloned into pLJM1 vectors. The scSac1 catalytic domain cDNAs (WT and deposits, individual channels were segmented to identify bona fide vesicles using a C389S, catalytically inactive) were obtained from S. Grinstein, processed for binary mask through an intensity thresholding function. Co-localization between human codon optimization and fused to the C terminus of Lyso-Flag-GFP in the the respective channels (mTOR–LAMP2, OSBP–LAMP2, p62–LAMP2 or D4H*– pLJM1 background. For pGEX-KG-D4H*-mCherry, D4H–mCherry cDNA was filipin) was measured using the ‘AND’ function and was plotted as the fraction obtained from G. D. Fairn and inserted into the pGEX-KG vector. The construct of LAMP2-positive (or filipin-positive) vesicles that were also positive for mTOR was renamed to D4H*–mCherry as two mutations, Y415A and A463W, were (or OSBP, p62 or D4H*). Where indicated, co-localization data were normalized introduced to improve the sensitivity of cholesterol binding, in addition to the to the reference group for multiple comparisons. Unless otherwise stated, ten D434S mutation in the original D4H cDNA. The autophagic flux reporter pMRX- fields of view per condition were processed using thresholding for quantification. IP-GFP-LC3-RFP-LC3ΔG was obtained from Addgene (84572). The number of cells included for analysis for each condition and the number of All proteins were either acutely knocked down or stably expressed through independent experiments are indicated in the figure legends accordingly. lentiviral infection. Lentiviruses were produced by co-transfecting HEK293T cells with a lentiviral expression/shRNA cassette and packaging plasmids (pMD2.5G, Lysosome immunoprecipitation. Lysosomes carrying Tmem192–mRFP–3×HA Addgene, 12259; and psPAX2, Addgene, 12260) using the calcium phosphate were purified as described previously7,32. Confluent HEK293T cells stably transfection method. Viral supernatant was collected 24 h after transfection, expressing Tmem192–mRFP–3×HA in a 15 cm culture dish were scraped in centrifuged at 1,000g for 5 min and filtered with a 0.45 μm filter. The viruses were chilled PBS and collected by centrifugation at 300g for 10 min at 4 °C. Cells were further concentrated using Lenti-X Concentrator (Clontech, 631231) according resuspended, lysed using a 23G syringe in KPBS buffer (136 mM KCl, 10 mM to the manufacturer’s instructions. Target cells were plated in 6-well plates with KH2PO4 pH 7.25) and centrifuged at 700g for 10 min at 4 °C. The post-nuclear 8 μg ml−1 polybrene (Millipore, TR-1003-G) and incubated with virus-containing supernatant was then incubated with 50 μl of anti-HA magnetic beads that were medium for 24 h. After removal of viruses, cells were supplemented with fresh prewashed in KPBS buffer for 5 min at 4 °C. Lysosome-bound beads were then medium containing 2 μg ml−1 puromycin (Calbiochem, 540411). Protein expression washed three times in KPBS buffer and either resuspended in 2× urea sample was confirmed by immunoblotting and fluorescence microscopy. buffer for immunoblotting or extracted of lipids with methanol and ethyl acetate To generate ORP1L knockout cells, guide RNA, a sequence targeting human for lysosomal cholesterol measurement. OSBPL1A exon 2 (AATCACTTCATTCCTCGCCA), was cloned into the lentiCRISPRv2 vector (Addgene, 52961). HEK293T cells were infected with viral Lysosomal cholesterol measurement using mass spectrometry. The lysosomal supernatants generated from lentiCRISPRv2 and selected with puromycin for 2 d. samples were extracted from anti-HA magnetic beads with methanol and ethyl Clonal lines were subsequently performed by limiting dilution, and validated by acetate with d7-cholesterol (2 μg per sample), and further derivatized with nicotinic immunoblot analysis of ORP1L expression. acid to improve the mass-spectrometric detection sensitivity of cholesterol. Measurement of cholesterol was performed using a Shimadzu 10A HPLC system Cholesterol starvation and stimulation. HEK293T cells in culture dishes were and a Shimadzu SIL-20AC HT auto-sampler coupled to a Thermo Scientific rinsed with serum-free media and incubated in 0.5% MCD supplemented with TSQ Quantum Ultra triple quadrupole mass spectrometer. Data processing was 0.5% lipid-depleted serum (LDS) for 2 h and, where indicated, were stimulated7 performed using Xcalibur v.4.0 (Thermo Fisher Scientific). A quality control (QC)

Nature Cell Biology | www.nature.com/naturecellbiology NATuRE CEll BIOlOGy Articles sample was prepared by pooling the aliquots of the study samples and was used to Labelling of cholesterol with GST–D4H*–mCherry and filipin. In situ labelling monitor the instrument stability. The QC was injected six times in the beginning of cholesterol was performed by adopting a protocol as previously described16. to stabilize the instrument and was injected every four study samples to monitor Cells were plated on fibronectin-coated coverslips, treated as indicated, fixed with the instrument performance. The data were accepted if the coefficient variance of 4% PFA for 10 min at room temperature and permeabilized in a liquid nitrogen cholesterol in the QC sample was <15%. The data were reported as the peak area bath for 30 s. After blocking with 1% BSA in PBS for 1 h, the cells were incubated ratio of cholesterol to d7-cholesterol. with recombinant GST–D4H*–mCherry (1:100 dilution, 1% BSA in PBS) for 2 h. The samples were then rinsed twice with PBS, re-fixed with 4% PFA for 10 min, Whole-cell lipidome analysis. HEK293T cells cultured in 6-well plates were incubated with filipin solution (0.5 mg ml−1 in PBS) for 2 h, immunostained with depleted of OSBP by doxycycline-induced shRNA, washed twice with PBS, LAMP2 or LAMP1, where indicated, and washed four times with PBS. Coverslips collected by scraping and collected by bench-top centrifugation at 1,000g at were mounted on glass slides using Fluoromount-G and imaged on a spinning-disk 4 °C, and cell pellets were flash-frozen and stored at −80 °C until metabolome confocal microscope. extraction. Lipid metabolites were extracted in 4 ml of a 2:1:1 mixture of chloroform:methanol:Tris buffer with inclusion of internal standards C12:0 TopFluor–cholesterol trafficking. HEK293T cells stably expressing Tmem dodecylglycerol (10 nmol) and pentadecanoic acid (10 nmol). Organic and aqueous 192–mRFP–3×HA were incubated with DMEM containing 1.5% LDS, treated layers were separated by centrifugation at 1,000g for 5 min, and the organic layer with DMSO or 20 nM OSW-1 for 2 h. A total of 20 nmol ml−1 TopFluor-cholesterol was collected. The aqueous layer was acidified (for metabolites such as LPA) by complexed with 0.1% MCD in DMEM containing 1.5% LDS was loaded to the cells adding 0.1% formic acid, followed by the addition of 2 ml of chloroform. The for 10 min on ice. After rinsing with PBS, cells were incubated at 37 °C, chased at mixture was vortexed and the organic layers were combined, dried down under the indicated times, fixed with 4% PFA and imaged using a spinning-disk liquid nitrogen and dissolved in 120 μl of chloroform, of which 10 μl was analysed confocal microscope. by both single-reaction monitoring (SRM)-based liquid chromatography with tandem mass spectrometry (LC–MS/MS) or untargeted LC–MS. LC separation PLA for in situ detection of ER–lysosome contacts. Human NPC1WT and was achieved using a Luna reverse-phase C5 column (50 mm × 4.6 mm with 5 μm NPC1mut fibroblasts were seeded on fibronectin-coated coverslips, fixed with diameter particles, Phenomenex). Mobile phase A was composed of a 95:5 ratio of 4% PFA and subjected to PLA according to manufacturer’s protocol (Sigma- water:methanol, and mobile phase B consisted of 2-propanol, methanol, and water Aldrich, Duolink In Situ Detection Reagents FarRed, DUO92013). In brief, after at a ratio of 60:35:5. Solvent modifiers 0.1% formic acid with 5 mM ammonium permeabilization with 0.1% saponin, the cells were blocked, incubated with anti- formate and 0.1% ammonium hydroxide were used to assist ion formation as well VAPA and anti-LAMP2 antibodies—either alone or in combination, as indicated— as to improve the LC resolution in both positive and negative ionization modes, for 1 h at room temperature, hybridized with PLA probes, ligated, amplified and, respectively. The flow rate for each run started at 0.1 ml min−1 for 5 min, to alleviate finally, stained with Hoechst. The number of PLA puncta (Fig. 7a), indicative of a backpressure associated with injection of chloroform. The gradient started at 0% close apposition between VAPA (ER) and LAMP2 (lysosome), were quantified as B and increased linearly to 100% B over the course of 45 min with a flow rate of discussed in the section of co-localization analysis. 0.4 ml min−1, followed by an isocratic gradient of 100% B for 17 min at 0.5 ml min−1 before equilibrating for 8 min at 0% B with a flow rate of 0.5 ml min−1. Autophagic flux reporter assay. Npc1+/+ and Npc1−/− MEFs expressing the GFP- Mass spectrometry analysis was performed with an electrospray ionization LC3-RFP-LC3ΔG reporter49 were generated by retrovirus-mediated transduction. (ESI) source using an Agilent 6430 QQQ LC–MS/MS. The capillary voltage was set After puromycin selection, cells were seeded in fibronectin-coated 96-well plates at to 3.0 kV, and the fragmentor voltage was set to 100 V. The drying gas temperature 25,000 cells per well and grown overnight. The cells were then treated with 20 nM was 350 °C, the drying gas flow rate was 10 l min−1 and the nebulizer pressure was OSW-1, 100 nM BafA1 or 250 nM Torin1 for 24 h and fixed with 4% PFA in PBS 35 p.s.i. Representative metabolites were quantified by SRM of the transition from for 15 min. The fluorescence intensity of GFP and RFP per well was measured precursor to product ions at associated collision energies. Data were normalized using a microplate reader (SpectraMax i3, Molecular Devices) with an excitation/ to the internal standards, and also external standard curves of metabolite classes emission of 488 nm/509 nm and 584 nm/607 nm, respectively. against the internal standards. The relative levels of all lipids measured were normalized to their respective controls. For absolute quantification of cellular Statistics and reproducibility. Data were expressed as mean ± s.d. where indicated cholesterol levels, as shown in Supplementary Fig. 4b, serial dilution of free in the figure legends. Statistical analyses were performed using unpaired two-tailed cholesterol, cholesteryl oleate and cholesteryl ester, ranging from 1 mM to 1 nM, was Student’s t-tests or one-way ANOVA for group comparisons, where indicated, prepared, processed by 2:1 chloroform:methanol extraction along with appropriate using GraphPad Prism v.7.0a (GraphPad Software). For box plots, the upper internal standards and analysed using LC–MS. The ratios of the integrations for and lower edges of the box indicate the first and third quartiles (25th and 75th cholesterol/C12MAGE, cholesteryl oleate/C12MAGE and choleseteryl ester/ percentiles) of the data, the middle line indicates the median and the whiskers C12MAGE were determined and plotted in inverse. A polynomial trend line was extend to the minimum and maximum. The levels of statistical significance are fit to the data and the resulting equation was used to calculate the amounts of the indicated by asterisks, and the exact P values are indicated in each figure legend lipids at 1 million cells per cell pellet. along with the statistical tests. Experiments in Fig. 8c and Supplementary Figs. 4c and 6h were performed once. All of the other experiments reported here were Purification of recombinant GST–D4H*–mCherry. GST-fusion D4H*–mCherry performed at least two to three times independently and all attempts at replication was expressed in BL21 Escherichia coli and induced for protein production with were successful with similar results. 0.4 mM IPTG for 20 h at 18 °C when bacterial growth reached an optical density at

600 nm (OD600) of ~0.6. Bacteria pellets were suspended in buffer A (20 mM Tris-Cl Reporting Summary. Further information on research design is available in the pH 8.0, 0.1 M NaCl, 1 mM dithiothreitol (DTT)) containing a protease inhibitor Nature Research Reporting Summary linked to this article. cocktail tablet (Pierce) and incubated with 0.35 mg ml−1 lysozyme on ice for 30 min. The bacterial lysate was sonicated on ice using an ultrasonic processor Q700 Data availability (QSONICA Sonicators) with 12 cycles of a 10 s pulse followed by a 10 s break at Mass spectrometry data that support the findings of this study are provided 50% amplitude output for 4 min. The lysate was treated with 0.5% Triton X-100 and in Supplementary Table 1 and have been deposited in ProteomeXchange with rocked at 4 °C for 15 min. The lysate was cleared by centrifugation at 17,000g at 4 °C accession number PXD014733. Statistics source data for Figs. 1–4 and 6–8 for 30 min. The pre-equilibrated Glutathione Sepharose 4B beads (GE Healthcare) and Supplementary Figs. 1–6 are provided in Supplementary Table 2. All data in buffer A with 0.1% Triton X-100 were incubated with clear supernatant for 2 h supporting the findings of this study are available from the corresponding author at 4 °C. GST-fusion D4H*–mCherry proteins were eluted from beads with 25 mM on reasonable request. l-glutathione solution in 50 mM Tris-Cl pH 8.8 and 200 mM NaCl. The fractions of the eluate were pooled and filtered sequentially using 3 kDa and 30 kDa centrifugal filter units (Amicon Ultra, Millipore) to remove glutathione and concentrate References the protein in 50 mM Tris-Cl pH 8.0, 5 mM DTT. The protein solution was then 64. Goldstein, J. L., Basu, S. K. & Brown, M. S. Receptor-mediated endocytosis supplemented with 20% sucrose and frozen in liquid nitrogen. The qualities of the of low-density lipoprotein in cultured cells. Methods Enzymol. 98, recombinant protein of different batches were validated by Coomassie blue staining. 241–260 (1983).

Nature Cell Biology | www.nature.com/naturecellbiology