SNX19 restricts endolysosome motility through contacts with the endoplasmic reticulum

The ability of endolysosomal organelles to move within the cytoplasm is essential for the performance of their functions. Long-range movement involves coupling of the endolysosomes to motor proteins that carry them along microtubule tracks. This movement is influenced by interactions with other organelles, but the mechanisms involved are incompletely understood. Herein we show that the sorting nexin SNX19 tethers endolysosomes to the endoplasmic reticulum (ER), decreasing their motility and contributing to their concentration in the perinuclear area of the cell. Tethering depends on two N-terminal transmembrane domains that anchor SNX19 to the ER, and a PX domain that binds to phosphatidylinositol 3-phosphate on the endolysosomal membrane. Two other domains named PXA and PXC negatively regulate the interaction of SNX19 with endolysosomes. These studies thus identify a mechanism for controlling the motility and positioning of endolysosomes that involves tethering to the ER by a sorting nexin.

ndosomes and lysosomes (henceforth jointly referred to as "endolysosomes" or "ELs") are membrane-bound organelles that primarily mediate uptake, degradation and recycling of both exogenous materials delivered by endocytosis and endogenous materials delivered by biosynthetic transport or autophagy 1,2 . In addition, ELs participate in many other cellular processes, including the regulation of cell metabolism, proliferation, adhesion and migration, immunity, exocytosis, and plasma membrane repair 3,4 . These diverse functions of ELs are highly dependent on their ability to move throughout the cytoplasm 5 . EL motility is mediated by interactions with motor proteins such as kinesins and dynein-dynactin, which drive movement along microtubules in the plus-end (i.e., anterograde) and minus-end (i.e., retrograde) directions, respectively 5,6 . This movement is influenced by interactions of ELs with other organelles, mainly the endoplasmic reticulum (ER) 5,6 . Interactions with ELs occur at specific domains of the organelles known as membrane contact sites (MCS) 7 . One type of ER-EL interaction that influences EL motility involves the ER-anchored ubiquitin ligase RNF26, which recruits and ubiquitinates the cytosolic sequestosome 1 protein SQSTM1, enabling binding to various EL adaptors having ubiquitin-binding domains such as EPS15, TAX1BP1, and TOLLIP 8 . Another type is represented by the ER-anchored protein protrudin (also known as ZFYVE27), which binds ELs through coincident detection of the small GTPase RAB7 and the phosphoinositide phosphatidylinositol 3-phosphate [PI(3)P] on the EL surface 9 . Finally, a third type involves the ER-anchored protein VAP (VAP-A and VAP-B paralogs), which interacts with the endolysosomal cholesterol-sensing protein ORP1L 10 . ER-EL tethering mediated by each of these three systems is reversed by the action of the deubiquitinating enzyme USP15 8 , the adaptor protein FYCO1 9 , and high cholesterol 10 , respectively, freeing the ELs for interaction with microtubule motors. The balance between interactions with microtubule motors and ER-tethering factors determines the overall movement and distribution of ELs within the cytoplasm.
Studies in yeast identified Mdm1 as another protein that bridges the nuclear envelope (a domain of the ER) with the vacuole (the functional equivalent of the metazoan lysosome) at MCS known as nuclear-vacuole junctions (NVJ) 11 . Further studies showed that yeast Mdm1 also associates with lipid droplets (LDs), clustering LDs next to NVJ and thus contributing to LD expansion under conditions of increased fatty acid availability 12 . Drosophila melanogaster has a single Mdm1 ortholog named snazarus or SNZ, which also regulates LD homeostasis, although in this case at MCS between the ER and the plasma membrane 13 . Mammals have 4 Mdm1/SNZ homologs named SNX13, SNX14, SNX19, and SNX25 11,14 (Fig. 1a). All of these proteins belong to a subset of the larger family of sorting nexin (SNX) proteins, which are defined by the presence of a phox-homology (PX) domain 14 (Fig. 1a). In addition to a PX domain, this subset comprises two predicted N-terminal transmembrane (TM) domains, a PXassociated domain (PXA), and a C-Nexin domain (PXC) (Fig. 1a) 11,14 . The best characterized of the mammalian Mdm1/ SNZ orthologs is SNX14 15,16 . Like the yeast and Drosophila orthologs, SNX14 localizes to ER-LD contact sites and promotes fatty acid-mediated LD growth 16 . Whereas the role of Mdm1, SNZ, and SNX14 in lipid homeostasis is now well established, to date there is no evidence that SNX14 or the other mammalian orthologs mediate tethering of the ER to lysosomes or other endolysosomal organelles, much less that they regulate EL positioning.
Herein we report that SNX19, but not SNX14, promotes ER-EL contacts and that these contacts restrict EL motility. Contacts are dependent on the two N-terminal TM domains that anchor SNX19 to the ER membrane and the PX domain that binds PI(3)P on the EL membrane. Furthermore, we show that the PXA and PXC domains inhibit the ability of SNX19 to mediate ER-EL tethering, indicating that they participate in an auto-regulatory mechanism. SNX19-mediated tethering to the ER decreases EL motility and contributes to the maintenance of a perinuclear population of ELs. Finally, we show that SNX19 can be additionally recruited to ER-LD contact sites in the presence of excess fatty acids. SNX19 thus has the properties of a protein capable of mediating both ER-EL and ER-LD tethering.

Results
SNX19 localizes to endoplasmic reticulum-endolysosome contact sites. SNX13, SNX14, SNX19, and SNX25 constitute a subfamily of mammalian sorting nexins that contain two predicted N-terminal TM domains, followed by PXA, PX, and PXC domains (Fig. 1a) 11,14 . In addition, SNX13, SNX14, and SNX25 have a regulator of G-protein signaling (RGS) domain that is absent in SNX19 (Fig. 1a) 11,14 . The best-characterized member of this family is SNX14, which localizes to ER-LD contact sites and regulates LD growth upon fatty acid addition 15,16 . In contrast, the localization and function of SNX19 are poorly understood. Besides the lack of an RGS domain, SNX19 differs from SNX14 in that the SNX19 PX domain binds PI(3)P whereas the SNX14 PX domain does not bind any phosphoinositide 17,18 . The fact that PI (3)P is enriched in ELs 19 prompted us to investigate if SNX19 could mediate ER-EL contacts.
Although SNX19 is ubiquitously expressed (http://www. proteinatlas.org), we were unable to determine the intracellular localization of the endogenous protein by immunofluorescence microscopy of various human cell lines using commercially available antibodies, probably due to the low abundance of this protein (0.59 ppm; for comparison, LAMP1 abundance is 115 ppm; https://pax-db.org/). To overcome this limitation, we transfected a plasmid encoding SNX19 tagged with a Cterminal green fluorescent protein (GFP) (SNX19-GFP) into human osteosarcoma U-2 OS cells, which are large, flat, and thus highly amenable to light microscopy analyses. Confocal microscopy of fixed cells showed localization of SNX19-GFP to a network that was identified as the ER by co-staining for the ER marker calnexin (Fig. 1b). With the exception of very high expressors, transfected and untransfected cells did not show appreciable differences in the overall structure and distribution of the ER network. Deletion of the two predicted TM domains resulted in localization of the truncated protein (SNX19 ΔTM -GFP) to the cytosol (Fig. 1c), demonstrating that SNX19 is anchored to the ER via the TM domains.
A similar pattern of localization to the ER was observed in live U-2 OS cells expressing Halo-tagged SNX19 (Fig. 1d, left panels). However, closer inspection of these live cells revealed the presence of higher intensity SNX19-Halo puncta along with the ER (Fig. 1d, left panels, arrowheads). To determine whether these puncta represented contacts with other organelles, we coexpressed SNX19-Halo with various fluorescent organelle markers and assessed their co-localization by confocal microscopy. We observed that SNX19-Halo puncta were closely apposed to a subset of organelles containing the EL marker LAMP1-RFP 20 (Fig. 1d). Additional experiments showed apposition of transgenic SNX19-GFP puncta with endogenous LAMP1 and early endosomal marker EEA1 ( Supplementary Fig. 1a). In contrast, we did not observe a significant association of SNX19-Halo with the LD marker GFP-ADRP 21 or the mitochondrial marker DsRed-Mito 22 under normal conditions of culture (Fig. 1d). Time-lapse imaging of live cells co-expressing SNX19-GFP and LAMP1-RFP showed that both signals remained together over time, irrespective of whether they appeared static (most cases) or motile (fewer cases) ( Fig. 1e and Supplementary Movie 1), indicating that these puncta represent bona fide ER-EL associations. Quantification of puncta containing both SNX19-GFP and LAMP1-RFP and that co-localized for at least 10 s in live cells (a rather strict requirement) revealed an average of 8 contacts per cell, with 25% of cells exhibiting between 10 and 35 contacts per cell, at any given time (Fig. 1f). Further live-cell imaging showed that SNX19-GFP foci associated with LAMP1-RFP were continuous with a fainter SNX19-GFP signal distributed throughout the ER, identified by labeling with Halo-tagged SEC61B (Fig. 1g).
Overexpressed LAMP1 has been shown to localize not only to lysosomes but also to early and late endosomes 23 Fig. 1b), which were additionally transfected with plasmids encoding SNX19-GFP and LAMP1-BFP, and incubated live for 16 h with the fluid-phase endocytic marker Alexa Fluor 647 (AF647)-Dextran followed by a 2 h chase in regular medium ( Supplementary Fig. 1c). Quantification of the co-localization with these markers showed that the majority of SNX19-GFP contacts occurred with organelles having all three markers (32%), or combinations of LAMP1-BFP plus RAB5A-Halo (27%), or LAMP1-BFP plus AF647-Dextran (15%), which we took to represent late endosomes, early endosomes and lysosomes, respectively ( Fig. 1h and Supplementary Fig. 1d).
From these experiments, we concluded that: (1) SNX19 mediates contacts of the ER mainly with early/late endosomes, and, to a lesser extent, terminal lysosomes, and (2) LAMP1 is a valid pan-EL marker for identifying contacts with SNX19.
SNX19 contacts endolysosomes via interaction of its PX domain with PI(3)P. To investigate the mechanism by which SNX19 contacts ELs, we considered the possibility that a specific SNX19 domain could recognize EL membrane lipids. While little is known about the PXA and PXC domains, the PX domain of SNX19 has been shown to bind PI(3)P 17,18 . Indeed, assays using PIP Strips confirmed specific binding of the recombinant GSTtagged PX domain of SNX19 (GST-PX; Fig. 2a) to PI(3)P, and to a lesser extent, phosphatidylserine (PS) (Fig. 2b). Moreover, mutation of a PX domain residue predicted to be involved in PI (3)P binding 17 (R582 to Q) abrogated binding to PI(3)P on the PIP Strips (Fig. 2a, b; GST-PX R582Q ).
To examine a possible role of PI(3)P recognition in the association of SNX19 with ELs, we examined the distribution of full-length SNX19-GFP with mutation of R582 to Q. We observed that this mutant construct still localized to the ER network but showed reduced association with LAMP1-RFP-positive puncta in U-2 OS cells (Fig. 2c, d). In addition, we tested the effects of drugs that decrease or increase intracellular PI(3)P levels. To deplete PI (3)P, we treated cells with SAR405, an inhibitor of VPS34, the lipid kinase that converts PI to PI(3)P on endosomes (Fig. 2e) 24 . The effectiveness of this inhibitor in our cell system was demonstrated by its ability to dissociate the PI(3)P probe 2xFYVE-GFP from ELs ( Supplementary Fig. 2). We found that SAR405 treatment decreased the number of SNX19-GFP puncta that co-localized with LAMP1-RFP (Fig. 2f, g). Conversely, to increase PI(3)P levels we used apilimod, an inhibitor of PIKFYVE, the lipid kinase that converts PI(3)P to PI(3,5)P 2 25 (Fig. 2e). Treatment with apilimod caused EL swelling (Fig. 2f), as previously reported 25 . Importantly, this treatment resulted in a dramatic increase in the number of SNX19-GFP puncta that were associated with the swollen LAMP1-RFP-positive ELs (Fig. 2f, g).
Taken together, these results indicated that SNX19-EL contacts are mediated by binding of the SNX19 PX domain to PI(3)P on the EL membrane.
Deletion of the SNX19 PXA or PXC domains increases association with endolysosomes. Since little is known about the PXA and PXC domains, we investigated their role in the interaction of SNX19 with ELs. To this end, we generated SNX19 deletion mutants lacking either the PXA domain (SNX19 ΔPXA -GFP) or a C-terminal segment containing the PXC domain (SNX19 1-659 -GFP) (Fig. 3a). Confocal microscopy showed that both SNX19 ΔPXA -GFP and SNX19 1-659 -GFP had a significantly larger number of contacts with LAMP1-RFP-positive ELs relative to full-length SNX19-GFP (Fig. 3b, c). Foci containing full-length or mutant SNX19-GFP together with LAMP1-RFP were always associated with the ER network, as visualized by co-labeling with a BFP-tagged KDEL ER-retrieval signal (Fig. 3d) or Halo-tagged VAP-A tethering protein ( Supplementary  Fig. 3a-c). The latter two ER proteins, however, were not visibly concentrated at SNX19-GFP-LAMP1-RFP contact sites ( Fig. 3d and Supplementary Fig. 3a-c). These observations indicated that the PXA and PXC domains inhibit the interaction of SNX19 with ELs.
The inhibitory effect of these domains likely explains why a SNX19-GFP mutant lacking the two transmembrane domains localizes to the cytosol (Fig. 1c), despite having a PX domain intrinsically capable of binding PI(3)P on the EL membrane (Fig. 2). Indeed, deletion of the PXA domain in addition to the transmembrane domains resulted in a SNX19-GFP construct that localized to a subset of LAMP1-RFP-positive vesicles (Supplementary Fig. 3d). We termed SNX19 mutants lacking the PXA or PXC domains "hypertethers" because of their enhanced ability to tether ELs to the ER.
The identification of the SNX19 hypertethers greatly facilitated the ultrastructural visualization of SNX19-EL contacts by correlative light and electron microscopy (CLEM). In these experiments, cells expressing SNX19 ΔPXA -GFP or SNX19 1-659 -GFP together with LAMP1-RFP and Mito-BFP (mitochondrial marker) were allowed to endocytose BSA-gold for 24 h, fixed and imaged by high-resolution structured illumination microscopy (SIM) (Fig. 4a, c), after which selected contacts were analyzed by transmission EM (Fig. 4b, d and Supplementary Fig. 4a, b). Labeling mitochondria with Mito-BFP served two important purposes: (1) it facilitated the manual alignment of fluorescence and EM images, as these organelles are large and easily identifiable landmarks, and (2) they were used as an internal specificity control for SNX19 contacts, which occurred with LAMP1-RFP but not Mito-BFP labeled organelles. Using this approach, we could observe ER cisternae (colored green) that were tightly apposed to organelles with the morphology of ELs at sites of enrichment of both hypertethers (yellow arrowheads) (Fig. 4b, d and Supplementary Fig. 4a, b). Three-D reconstruction from serial sections revealed the extensive nature of ER-EL contacts at those sites ( Supplementary Fig. 4c). CLEM also showed that the LAMP1-RFP-positive EL compartment enriched in hypertethers most often had the morphology of a multivesicular body (MVB), as intraluminal vesicles could be readily observed in their interior (Fig. 4b, d and Supplementary Fig. 4a). The CLEM image in Fig. 4b additionally shows that a neighboring LAMP1-RFP-positive structure lacking SNX19 ΔPXA -GFP and containing internalized BSA-gold (black arrowhead) has the morphology of a lysosome. These observations supported our earlier finding that SNX19 contacts mostly endosomes, and to a lesser extent terminal lysosomes ( Fig. 1h and Supplementary  Fig. 1d). LDs, identified as small, electron-dense (i.e., osmiophilic) bodies were often found in the proximity of SNX19-decorated ELs, but the LDs themselves did not have associated SNX19 under the conditions of these experiments ( Fig. 4b and Supplementary Fig. 4a, white arrowheads).
Loss of SNX19 decreases both contacts of endolysosomes with the ER and perinuclear distribution of endolysosomes. Next, we examined the contribution of SNX19 to the overall extent of ER-EL contacts within the cell. To this end, we knocked out the gene encoding SNX19 in U-2 OS cells using CRISPR-Cas9 ( Supplementary Fig. 5a-c) and analyzed the effect of the KO on ER-EL contacts using two methods. In the first method, we allowed cells to endocytose BSA-gold for 24 h to label ELs, and then examined serial sections of plastic-embedded cells by transmission EM for the presence of ER-EL contacts defined by apposition of the two organelle membranes at ≤30 nm 26 (Fig. 5a). Quantification of the EL perimeter length engaged in contacts with ER elements relative to the total EL perimeter length revealed a statistically significant~2-fold decrease in the extent of ER-EL contacts in SNX19-KO relative to WT cells (Fig. 5b).
The second method was a proximity ligation assay (PLA) 27 using primary antibodies to endogenous LAMP1 and calnexin, followed by PLA probes. Immunofluorescence microscopy showed fewer fluorescent puncta in SNX19-KO relative to WT cells (Fig. 5c, d), indicative of a reduction in the overall number of ER-EL contacts.
Having established that SNX19 contributes to overall ER-EL contacts, we next examined whether these contacts influence the cytoplasmic distribution of ELs. Immunostaining for various organelle markers revealed that the LAMP1-positive EL population in SNX19-KO cells was more dispersed than in wild-type (WT) cells (Fig. 5e). Quantification of this dispersal using a semi-automated shell analysis of 60-90 cells per condition showed that the fraction of peripheral ELs rose from a median of 0.15 in WT cells to 0.42 in SNX19-KO cells (Fig. 5f). Re-expression of SNX19-GFP in SNX19-KO cells partially reversed the EL dispersal phenotype (Fig. 5e, f), confirming that this phenotype was due to the loss of SNX19. In contrast to the distribution of ELs, the distribution of the ER, the Golgi complex, and mitochondria immunolabeled with antibodies to calnexin, GM130, and Tomm20, respectively, was not affected by SNX19 KO (Supplementary Fig. 5d-f). These experiments thus demonstrated that SNX19-mediated ER-EL contacts contribute to the maintenance of the perinuclear population of ELs.
SNX19 restricts endolysosome motility. The EL dispersal phenotype of SNX19-KO cells suggested that SNX19-mediated ER-EL contacts constrain ELs to the perinuclear region. To determine whether this phenomenon was due to an effect on EL motility, we performed live imaging of ELs in U-2 OS cells at~10 fps, followed by analysis using automated particle detection and tracking algorithms 28,29 . Briefly, we applied a feature detection algorithm to identify LAMP1-RFP objects and then fit multiple Gaussians to approximate the center of the objects. Next, we reconstructed individual trajectories for the entire video (i.e., over a 10-20 s period) (Fig. 6a) and applied a moment scaling spectrum (MSS) analysis to determine the motion type of individual trajectories based on the trend of the object's diffusion over time 28 . This approach resulted in a mean of~200 trajectories extracted per cell and a total of >5000 trajectories per condition analyzed. The majority of the EL trajectories fit within three motion-type categories: confined (blue), free (green), or directed (red), with a much smaller portion of the trajectories determined to be unclassified (magenta) (Fig. 6a, b). Confined trajectories represent ELs that are either static (e.g., confined-immobile) or constrained by barriers to their movement (e.g., confinedmobile). Free ELs move sporadically with frequent changes in direction, probably resulting from intermittent association with microtubule motors (Fig. 6a, b), while directed ELs move directionally, consistent with movement driven by a strong association with kinesin or dynein motors along microtubules (Fig. 6a-c). Free and directed ELs are overall more motile than confined ELs, and thus, for simplicity, we combined free and directed motion types into a single category when quantifying EL motility.
This analysis showed that SNX19 KO resulted in an increased proportion of free plus directed ELs and a decreased proportion of confined ELs (Fig. 6d), indicative of increased EL motility. In contrast, cells transfected with plasmids encoding SNX19-GFP or the hypertethers SNX19 ΔPXA -GFP or SNX19 1-659 -GFP exhibited a decreased proportion of free plus directed ELs and an increased proportion of confined ELs relative to GFP-transfected cells ( Fig. 6e, f). Even in the same field of observation (Fig. 6g), ELs having associated SNX19-GFP (indicated by arrows) were less motile than neighboring ELs devoid of SNX19-GFP (no arrows) (Fig. 6g, h).
Further analysis of EL movement in the peripheral cytoplasm of WT U-2 OS cells (only region where individual ER tubules can be resolved) showed that SNX19-GFP-positive, mostly static LAMP1-RFP puncta were associated with BFP-KDEL-labeled ER tubules 100% of the time, whereas SNX19-GFP-negative, motile LAMP1-RFP puncta were associated with BFP-KDEL-labeled ER tubules~80% of the time ( Supplementary Fig. 6a, b). These data indicated that the binding of SNX19-GFP not only decreases the motility but also enhances the alignment of ELs with ER tubules. It's important to point out that the alignment of 80% motile ELs with ER does not necessarily reflect physical tethering between these organelles, but could result from the fact that both ELs and ER are independently attached to the same microtubules in the peripheral cytoplasm 30,31 .
We also tested if apilimod treatment, a condition that enhances SNX19-EL association (Fig. 2f, g), affects EL motility. Because PIKFYVE inhibition also causes EL swelling, an effect that alone may lower EL motility due to friction-based constraints, we treated cells acutely (30 min) with apilimod. This duration of treatment was sufficient to cause an increase in SNX19-EL tethering without a gross volumetric gain of ELs (Fig. 6i). Under these conditions, the acute apilimod treatment strongly reduced EL motility (Fig. 6j).
From these experiments, we concluded that ER-EL contacts mediated by SNX19 restrict EL motility.
SNX14 does not localize to ER-EL contact sites. The SNX19 homolog SNX14 was previously shown to localize to ER-LD contacts upon treatment of cells with fatty acids 16 . To determine if SNX14 could also localize to ER-EL contacts, we compared the localization of SNX19-GFP and SNX14-mNeonGreen relative to LAMP1-RFP in U-2 OS cells in the absence of added fatty acids. We observed that, like SNX19-GFP, SNX14-mNeonGreen localized to the reticular ER (Fig. 7a). However, whereas SNX19-GFP also localized to puncta containing LAMP1-RFP, SNX14-mNeonGreen did not (Fig. 7a, b). Moreover, treatment with apilimod greatly increased the number of SNX19-GFP-EL contacts, but did not induce the formation of SNX14-mNeonGreen-EL contacts (Fig. 7b, c).
Conversely, we tested whether SNX19 could be recruited to LDs upon treatment with fatty acids. We observed that treatment of cells with oleic acid (OA) for 2 h did cause SNX19-GFP to associate with LDs labeled with GFP-ADRP (Fig. 7d, e) and monodansylpentane (MDH) (Supplementary Fig. 7), as was previously reported for SNX14 16 . Despite this association, 3-color time-lapse imaging of live cells revealed that SNX19-Halo-EL contacts were maintained upon treatment with OA and that SNX19-Halo could independently contact ELs (labeled with LAMP1-RFP) and LDs (labeled with GFP-ADRP) under this condition (Fig. 7f, g).
Taken together, these experiments indicated that SNX19 can contact both ELs and LDs, whereas SNX14 only contacts LDs.

Discussion
Our studies have thus identified SNX19 as a player in the mechanisms that regulate EL motility and positioning. SNX19 is anchored to the ER by its two N-terminal TM domains and contacts ELs by the interaction of its PX domain with the endolysosomal lipid PI(3)P. These interactions result in the concentration of SNX19 at sites of ER-EL contact. Since the ER is generally less motile than ELs, the net result is a reduction in  Characteristics of SNX19-mediated ER-EL contacts. Immunofluorescence microscopy and CLEM showed that ER-anchored SNX19 interacts with a range of EL compartments, from early endosomes to lysosomes, with the greatest frequency seen for organelles with characteristics of late endosomes/MVBs (Figs. 1h, 4b, d and Supplementary Figs. 1, 4a). This distribution closely matches that of PI(3)P in the endolysosomal system 19 , consistent with the specificity of the SNX19 PX domain for this phosphoinositide ( Fig. 2b) 17,18 . These properties are reminiscent of those of protrudin, another protein that is anchored to the ER by three Nterminal hydrophobic domains and that interacts with ELs via a Cterminal PI(3)P-binding FYVE domain 9 . Since the levels of PI(3)P are subject to regulation by nutrients and other stimuli [32][33][34] , the PI (3)P-binding specificity of both SNX19 and protrudin provides a means to control the formation of ER-EL contact sites. These similarities notwithstanding, SNX19 and protrudin differ in other respects. First, whereas SNX19 KO causes dispersal of ELs toward the cell periphery (Fig. 5e, f), protrudin knockdown (KD) causes clustering of ELs in the perinuclear region 9 . These opposite effects might be explained by the ability of protrudin to interact with the small GTPase RAB7 and to transfer the microtubule motor kinesin-1 to the RAB7-effector protein FYCO1 on ELs for transport of ELs toward the cell periphery 9 . Thus, rather than immobilizing ELs in the perinuclear region, protrudin mobilizes them toward the cell periphery. We considered the possibility that SNX19 could exert its effects on EL motility and positioning through interactions with microtubule motors. However, co-immunoprecipitation and pulldown experi1ments failed to detect interactions of SNX19 with kinesin-1, kinesin-3, and dynein ( Supplementary Fig. 6c-e). Furthermore, protrudin has an FFAT motif (i.e., two phenylalanines within an acidic tract) that mediates binding to VAP, resulting in a concentration of protrudin at a specific type of ER-EL contact site 35 . In contrast, SNX19 lacks FFAT or similar motifs, and coexpression of VAP-A-Halo with SNX19-GFP or SNX19-GFP hypertethers does not result in accumulation of VAP-A-Halo at SNX19-GFP ER-EL contact sites ( Supplementary Fig. 3a-c and Supplementary Movie 2). This property also distinguishes SNX19 from other proteins that localize to ER-EL contact sites by virtue of VAP-FFAT-motif interactions, such as the sterol-binding proteins ORP1L 10 , STARD3, and STARD3NL 36 , and the phospholipidtransport protein VPS13C 37 . SNX2 is another sorting nexin that localizes to ER-EL contact sites via interactions of its PX domain with PI(3)P and an FFAT motif with VAP 38 . In this case, however, the interactions function to modulate PI(4)P-and WASHdependent fission of recycling endosomal tubules 38 . Thus, the SNX19-mediated ER-EL contacts identified here differ from those described before, as they are devoid of VAP. Further studies will be needed to determine whether SNX19 interacts with any small GTPases or other components of ER-EL contact sites, and how these interactions are regulated.
Regulatory role of the SNX19 PXA and PXC domains. The finding that deletions of SNX19 regions comprising the PXA or PXC domains enhance ER-EL association (Fig. 3) suggests that these regions play a regulatory role. This regulation could involve an intramolecular autoinhibitory interaction, whereby the PXAand PXC-containing regions prevent PI(3)P binding by the PX domain from the same molecule. Alternatively, the PXA-and PXC-containing regions could inhibit the PX domain of another SNX19 molecule in trans. Furthermore, the PXA-and PXCcontaining regions could interact with other proteins that, in turn, inhibit ER-EL tethering. Although little is known about the molecular functions of these domains, recent work has shown that the PXA domain of S. cerevisiae Mdm1 binds LDs, fatty acids, and the fatty acyl-CoA ligase Faa1 12 . In addition, the PXC domain of human SNX14 binds LDs via an amphipathic helix 16 . The SNX19 PXC domain is also predicted to contain at least two amphipathic helices 39 . Therefore, another possibility is that the PXA and PXC domains of SNX19 regulate the PX domain by binding to other lipids rather than proteins. These lipids could be on LDs, raising the possibility that SNX19 becomes activated for EL tethering in the proximity of LDs. Alternatively, such lipids could be on the EL membranes themselves. In this regard, it is worth noting that the S. pombe protein Pxa1 (not to be confused with the S. cerevisiae peroxisomal fatty acyl-CoA transporter Pxa1), whose single defining feature is an N-terminal PXA domain, is involved in protein sorting to the vacuole 40  For each section that captured a gold-containing EL, the perimeter of the EL membrane was measured. The same was done for the portion of the EL membrane that was closely apposed to an ER membrane (i.e., contact length). All EL perimeters and contact lengths across the sections that captured a given EL were summed up, the total contact lengths were divided by the total EL perimeter lengths and multiplied by 100 to compute the percentage of the EL perimeter that contacts the ER. Contacts were quantified in multiple sections from n = 8 cells per condition across two independent experiments. Graph shows a line at the mean and SD. Unpaired t-test was performed, ** two-tailed p-value = 0.009. c WT or SNX19-KO U-2 OS cells were fixed, immunostained with rabbit anti-LAMP1 and mouse anti-calnexin antibodies. A proximity ligation assay (PLA) was then performed using anti-mouse and anti-rabbit probes, and detected in the 647-nm channel. The PLA signal is shown in white in two representative confocal micrographs. d Quantification of PLA puncta from c. A threshold was applied to eliminate background signal and the total number of puncta per cell was computed in three experiments, each consisting of 60-80 cells per condition. Graph shows a line at the mean and SD. Unpaired t-test was performed, *** two-tailed p-value < 0.001. e WT or SNX19-KO U-2 OS cells were fixed, immunostained with anti-LAMP1 antibody, and imaged by confocal microscopy. Bottom row shows a SNX19-KO cell that was transfected with a SNX19-GFP construct. f Quantification of peripheral ELs from experiments such as that shown in e. Cell outlines were traced and the outline was iteratively shrunken by 2 μm to generate 5 "shells" per cell. The scheme on top depicts combinations of the 3 outermost shells in gray, and 2 innermost shells in white. LAMP1 signal in the gray region (peripheral ELs) was quantified and divided by LAMP1 signal in combined gray and white regions (total ELs). Box and whisker plot represents data from 3 independent experiments where 7-20 cells were analyzed per condition in each experiment. The box extends from 25th to 75th percentiles, the horizontal line represents the median, and the whiskers represent the minimum and maximum values. Statistical significance was calculated with the Mann-Whitney test; *** two-tailed p-value < 0.001, ** two-tailed p-value = 0.0081. Source data are provided as a Source Data file. Scale bars: 500 nm in a, 250 nm insets and 10 μm in c and e.  (Fig. 7a, b). Not even when PI(3)P levels are raised by treatment with apilimod is SNX14 able to interact with ELs (Fig. 7b, c). These functional differences likely stem from the inability of the SNX14 PX domain to bind PI(3)P because of a single amino-acid substitution in the PI(3)P-binding site 17,18 . Likewise, the PX domain of Drosophila SNZ is incapable of binding PI(3)P, probably explaining why this protein does not localize to ER-EL contacts but to ER-PM contacts 13 . The PX domain of SNX25 also does not bind PI(3)P, instead binding PS, PI(3,4)P 2 , PI(4,5)P 2 , PI(3,5)P 2 and PI(3,4,5)P 3 17,18 . The SNX13 PX domain shares with the SNX19 PX domain the property of binding PI(3)P 17,18 , and indeed a SNX13-GFP construct also colocalizes with LAMP1-RFP at ER-associated foci ( Supplementary  Fig. 6f), making SNX13 another candidate for mediating ER-EL contacts. These observations highlight the functional diversification of this SNX subfamily in mammals. An activity that SNX19 and SNX14 do share is their association with LDs upon treatment of cells with OA (Fig. 7d-f and  refs. 15,16 ). For SNX14, this activity enables its function in lipid metabolism at the ER-LD interface 16 . SNX19 could therefore be involved in the same process. Nevertheless, SNX19 does not seem able to functionally compensate for SNX14 loss, as KO of SNX14 alone in U-2 OS cells causes defects in LD morphology 16 . Further assessment of our SNX19-KO cells will be needed to determine whether SNX19 plays a role in lipid homeostasis and to test if SNX14 and SNX19 have overlapping functions at ER-LD contacts.
Other functions of SNX19. SNX19 has been shown to interact with islet antigen-2 (IA-2), a protein tyrosine phosphatase and autoantigen in type I diabetes 41,42 that is expressed in brain neuroendocrine cells, pituitary gland, and pancreas. In pancreatic β cells, SNX19 silencing results in fewer dense-core vesicles (DCVs) and decreased insulin content and secretion 43 . Given that both IA-2 and PI(3)P are present on DCVs 44,45 , SNX19 may be recruited to DCVs by coincident detection of these two factors to control DCV abundance. Unlike IA-2, however, SNX19 expression is not restricted to regulated secretory cells, consistent with a more general function in controlling EL motility. SNX19 has been additionally implicated in the partitioning of the dopamine D1 receptor (D1R) to lipid rafts 46 and in chondrogenic differentiation 47 . It is unclear, however, if these functions relate to the ability of SNX19 to tether ELs to the ER.
Disease connections. Whereas biallelic mutations in the SNX14 gene are known to cause autosomal recessive spinocerebellar ataxia 20 (SCAR20), a disorder characterized by cerebellar atrophy, ataxia, and severe intellectual disability 48,49 , mutations in the SNX19 gene have not yet been shown to cause a Mendelian disorder [50][51][52][53][54][55] . Genetic and epigenetic analyses, however, have linked variations in SNX19 expression to proliferation and apoptosis in pancreatic β cells 56 , coronary heart disease 57 , and cancer 58,59 . In addition, extensive analyses have revealed an association of higher levels 51,52 or rare transcript variants of SNX19 53 with schizophrenia risk. Among these variants, a prominent risk-associated transcript is predicted to encode a protein lacking the C-terminal PXC domain, a mutation similar to our SNX19 1-659 hypertether. The expression of higher levels of normal SNX19 or normal levels of PXC-deleted SNX19 would be expected to enhance EL tethering to the ER, decreasing the motility of ELs. This defect, particularly in relevant neuronal types, could contribute to the pathogenesis of this form of schizophrenia. Since SNX19 can also associate with LDs, defective handling of lipids at the ER-LD interface could also underlie the development of this disorder.
Concluding remarks. In general, sorting nexin proteins have been shown to play various roles in intracellular cargo sorting and cell signaling 14 . Studies on Mdm1, SNZ and SNX14 expanded these roles to the formation of ER-organelle contacts for the regulation of lipid metabolism 11,13,16 . Our studies on SNX19 now demonstrate another function for a sorting nexin in the regulation of EL motility and positioning by tethering to the ER.

Methods
All primers used in this study are listed in Supplementary Table 1.
Plasmid constructs. The following plasmid constructs were generated in this study: SNX19-GFP: Human SNX19 cDNA was PCR-amplified using primers AS191 and AS318 and inserted into EcoRI-digested pEGFP-N1 using Gibson Assembly (E2611L, New England Biolabs, Ipswich, MA, USA).
SNX19-Halo: The Halo tag cDNA was PCR-amplified using primers AS306 and AS307 and swapped into SNX19-GFP by digesting out the GFP sequence with AgeI and NotI. SNX13-GFP: Human SNX13 cDNA was PCR-amplified using primers AS187 and AS188 and inserted into EcoRI-digested pEGFP-N1 using Gibson Assembly.
GST-PX R582Q : Generated by PCR-amplification of the PX domain of SNX19 R582Q -GFP, using the primers AS293 and AS294. The fragment was inserted into BamHI/XhoI-digested pGEX-6P-3 bacterial expression vector by Gibson Assembly.
VAP-A-Halo: Rat VAP-A cDNA was PCR-amplified using the primers AS380 and AS381 and swapped into SNX19-Halo vector by digesting out SNX19 with HindIII and AgeI.
All cDNAs generated in this study were sequence-verified prior to use.  The sources of plasmids encoding other proteins were as follows: LAMP1-RFP was a gift from Walther Mothes (Addgene plasmid #1817), GFP-ADRP was from GeneCopoeia (Germantown, MD, USA), DsRed-Mito was from Clontech (Mountain View, CA, USA), Halo-SEC61B was a kind gift from the Lippincott-Schwartz laboratory 60 , 2xFYVE-GFP was a gift from Sergio Grinstein 61 , Mito-BFP and BFP-KDEL were gifts from Gia Voeltz (Addgene plasmid #49151 and 49150, respectively), GFP-SKIP was previously made in our lab 62 , GFP-RILP was previously described 63 .
Cells The 100 μL transfection mixes were allowed to sit for an additional 20 min at room temperature. During this period, the medium on cells was replaced with 500 μL Opti-MEM, and cells were returned to the incubator. After 20 min, transfection mixes were added to the cells dropwise, and cells were returned to the incubator. The medium was changed to 500 μL complete DMEM after 1 h. Transfections for CLEM imaging were conducted the same way as above but scaled to 25 mm coverslips in 6-well dishes. DNA amounts and Lipofectamine reagent were used at 4× the amounts indicated above, each pre-mixed in 250 μL Opti-MEM and combined to produce 500 μL transfection mixes which were added dropwise to cells in 1 mL Opti-MEM after 20 min. Following a 1 h incubation, the medium was changed to 2 mL complete DMEM.
Generation of CRISPR-Cas9 knock-in and knock-out cells. The RAB5A-Halo knock-in (KI) HeLa cell line was generated by CRISPR-Cas9 and homologydirected repair (HDR). This was done using transient co-transfection of pX458 plasmid, a gift from Feng Zhang (Addgene plasmid #48138) bearing the guide 5′-GCCAATATTCTTCTTTCTGG, targeting a PAM sequence 60 bp upstream the start codon of RAB5A in exon 2, and a donor vector containing the Halo tag coding sequence and left and right homology arms (each~800 bp), to the RAB5A genomic region flanking the sgRNA site for targeted insertion of the Halo Tag at the N terminus of RAB5A. A single point mutation was introduced in the donor vector to destroy the PAM sequence and allow proper insertion of the tag in the genome while active Cas9 was transiently expressed. Single-cell sorting of GFP-positive cells was performed. After two weeks, the same cells were incubated with Janelia Fluor 549 dye 64 (see "Inhibitors and other reagents" section for Halo Tag labeling), rinsed twice with fresh medium, lifted with trypsin, and immediately sorted for GFPnegative/Janelia Fluor 549-positive signal to only recover the cells with the insertion of the Halo Tag. Homo/heterozygotic monoclonal cell lines expressing endogenous RAB5A-Halo were screened by PCR using the primers Rab5 Halo KI F and Rab5 Halo KI R. A clone that was edited on one allele was kept for this study (Supplementary Fig. 1b).
To inactivate the SNX19 gene in U-2 OS cells, the CRISPR-Cas9 system was used with two sgRNA sequences. Initially, 6 different sgRNAs targeting the first exon of SNX19 near the start codon were designed using the online tool Benchling (https://www.benchling.com). The guides were cloned into pX458 plasmid and cotransfected into U-2 OS cells in different combinations. Three days later, cells were harvested and genomic DNA was isolated. PCR was conducted using the primers AS407 and AS408 to yield a band of 1.1 kb in WT cells and a short band (100-600 bp shorter depending on the sgRNA combination) in CRISPR-edited cells. The PCR products were run on a 1% agarose gel and those sgRNA combinations that together produced the most prominent short band were used as the final guides for generating KO cells. The 2 final guides (sgRNA1 5′-CCGTTCC AGGAAACTCCAGC and sgRNA2 5′-GCTGAAGTCACCTATACGCG) were cotransfected into U-2 OS cells and 24 h later cells were single-cell sorted into 96-well plates on GFP-positive signal and allowed to grow in complete DMEM for 2-3 weeks. When colonies were visible, cells were transferred to larger wells until a confluent well of a 6-well dish could be harvested and PCR-screened as described above. Those colonies that had undergone non-homologous end joining, and showed only the short PCR band, were kept as "KO" and their PCR products were sequence-verified ( Supplementary Fig. 5a, b). Inhibitors and other reagents. To inhibit VPS34, cells were treated with 1 μM SAR405 (Cat# 2716, Axon Medchem, Groningen, The Netherlands) for 2 h prior to imaging. PIKFYVE was inhibited with 200 nM apilimod (Cat# sc-480051, Santa Cruz Biotechnology, Dallas, TX, USA) for 2 h. As a control, cells were treated for 2 h with an equivalent volume of DMSO, the solvent that both inhibitors were reconstituted in. Where indicated, acute treatments were conducted using 1 μM apilimod for 30 min.
To visualize Halo-tagged proteins, cells expressing Halo Tag constructs or the HeLa RAB5A-Halo-KI cells were incubated in a medium containing 500 nM Janelia Fluor 549 or 646 cell-permeable dyes (Cat# 6147 and 6148, Tocris, Minneapolis, MN, USA) for 30 min in a 37°C, 5% CO 2 incubator. The cells were washed once with complete DMEM and imaged in fresh pre-warmed complete DMEM.
To label bona fide lysosomes in HeLa RAB5A-Halo-KI cells, cells were allowed to endocytose Alexa Fluor 647 (AF647)-conjugated Dextran (Cat# D22914, ThermoFisher Scientific, Waltham, MA, USA) at 10 μg/mL overnight, followed by a minimum of 2 h chase in Dextran-free complete DMEM the following day prior to imaging.
To induce LD growth, cells were treated with oleic acid (Cat# O1008, Sigma Aldrich, St. Louis, MO, USA) for 2 h at a 200 μM final concentration. Briefly, oleic acid was dissolved in complete DMEM at 60°C, vortexed, cooled to 37°C, and added to cells. As a control, some cells were treated with the equivalent volume of methanol, the solvent that the oleic acid was reconstituted in. In some experiments, LDs were visualized by staining with MDH (AUTOdot Autophagy Visualization Dye, Cat# SM1000a, Abcepta, San Diego, CA, USA). This was done by washing live cells twice with warm Opti-MEM and adding 1:1,000 MDH in Opti-MEM, for 15 min, in a cell culture incubator. The cells were washed once with Opti-MEM and cell culture medium was added back to the cells. The cells were imaged live and LDs were visualized in the 405 nm channel.
Immunostaining. All immunostaining was conducted at room temperature. Cells were fixed in 4% paraformaldehyde in PBS for 20 min followed by 3 washes in PBS. The cells were then permeabilized in PBS plus 0.1% v/v Triton X-100 for 5 min followed by an additional 2 washes with PBS. Cells were blocked in 0.5% w/v BSA in PBS for 30 min, then incubated for 1 h with primary antibodies in 0.5% w/v BSA-PBS. Cells were washed 3 times with 0.5% w/v BSA-PBS followed by incubation with fluorophore-conjugated secondary antibodies for 40 min, protected from light. Cells were then washed 3 times with 0.5% w/v BSA-PBS followed by 3 washes with PBS alone. Finally, coverslips were mounted onto glass slides with Fluoromount-G containing DAPI (Cat# 00-4959, Invitrogen, Carlsbad, CA, USA) and stored in the dark until imaged. For long-term storage, slides were kept at 4°C.
Confocal microscopy. Fixed-and live-cell imaging were conducted on a laserscanning confocal LSM880 microscope (Zeiss, Oberkochen, Germany) with a PlanApo ×63 1.4 NA objective and the Definite Focus system. Fixed and mounted cells were imaged at room temperature with no CO 2 while live cells were imaged on a 37°C heated stage in a 5% CO 2 chamber.
Live-cell imaging requiring higher acquisition speeds than that obtainable with laser-scanning confocal microscopy was conducted on a spinning-disk confocal Eclipse Ti Microscope (Nikon, Tokyo, Japan), fitted with a PlanApo VC ×60 objective with 1.4 NA and an Evolve 512 EMCCD camera (Photometrics, Tucson, NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24709-1 ARTICLE NATURE COMMUNICATIONS | (2021) 12:4552 | https://doi.org/10.1038/s41467-021-24709-1 | www.nature.com/naturecommunications AZ, USA). Cells were kept in focus using the Perfect Focus system and maintained on a 37°C heated stage in a 5% CO 2 humidified chamber. To achieve the high frame rates (>10 fps) required for EL motility tracking, cells were imaged by excitation with red and green lasers and, after emitted light was passed through a beam-splitter (525 nm/605 nm), simultaneous detection using a dual-camera setup.
PIP strip immunoblotting. To analyze phospholipid-binding specificity, the purified GST-PX domain of SNX19, GST-PX R582Q domain mutant, or GST alone were incubated with PIP Strips (Cat# P-6001, Echelon Biosciences, Salt Lake City, UT, USA) according to the supplier's protocol. Briefly, the PIP Strips were blocked overnight at 4°C in a blocking solution composed of 3% w/v BSA in PBS containing 0.1% v/v Tween 20 (PBS-T) under gentle agitation. The following day, the PIP Strips were incubated with purified GST or GST-PX at 0.5 μg/mL in blocking solution and gently agitated at room temperature for 1 h. The strips were washed 3 times with PBS-T, 5 min per wash, and incubated for an additional hour at room temperature with anti-GST HRP-conjugated antibody (1:2,000) in blocking solution. The strips were washed 3 times with PBS-T, 5 min per wash, and the signal was detected by chemiluminescence using SuperSignal West Femto Maximum Sensitivity Substrate (Cat# 34096, ThermoFisher Scientific).
Correlative light and electron microscopy (CLEM). U-2 OS cells were seeded onto 25-mm round, photoetched gridded coverslips (Cat# 72265-50, Electron Microscopy Sciences, Hatfield, PA, USA) and transfected the following day. To facilitate the identification of endolysosomal organelles in the EM images, the cells were allowed to endocytose 6-nm BSA-gold particles (Cat# 25483, Electron Microscopy Sciences, Hatfield, PA, USA) for 24 h. The following day, cells were fixed for CLEM in 2% paraformaldehyde, 0.1% glutaraldehyde in PBS at room temperature for 20 min. Cells were washed twice with PBS and placed in a 25-mm round coverglass chamber filled with PBS for imaging. Regions around cells of interest were mapped with a transmitted and fluorescent light on a Zeiss LSM880 microscope (Plan-Apochromat ×63 1.4 NA objective lens) using a tile and stitch module, to capture grid labels as well as cell positions and fluorescence patterns. SIM imaging of fixed cells was then performed on a Zeiss Elyra PS.1 microscope with a Plan-Apochromat ×63 1.4 NA objective lens at room temperature. Three orientation angles of the excitation grid with five phases each were acquired for each z plane. Images were then reconstructed with the SIM module in Zeiss ZEN Black software version 14.0.18.201, using the automatic setting.
After imaging, the cells were placed into a fixative solution composed of 2.5% glutaraldehyde, 2% formaldehyde, 2 mM CaCl 2 in 0.1 M sodium cacodylate, pH 7.4, for 10 min at room temperature followed by 45 min on ice. The coverslips were washed 4 times for 5 min each, post-fixed with 2% OsO 4 for 2 h in the same buffer at 4°C, extensively washed with water, stained with 2% uranyl acetate in water, dehydrated through a series of increasing ethanol concentrations (30%, 50%, 70%, 90%, 3 changes of 100%) and embedded in EMBed 812 epoxy resin (Cat# 13940, Electron Microscopy Sciences, Hatfield, PA, USA). After resin polymerization, the coverslip was removed with hydrofluoric acid. Cells previously imaged by light microscopy were identified by their position on the grid. A 1 × 1-mm area containing the cell(s) of interest was cut out using a jeweler saw, mounted on an aluminum holder, and trimmed to 300 × 300 μm. Serial 50-70 nm sections were cut parallel to the plane of the coverslip and mounted on formvar/carbon-coated slot (0.5 × 2 mm) EM grids. Sections were stained with 2% uranyl acetate in 50% ethanol and imaged in a FEI Tecnai 20 transmission electron microscope operated at 120 kV. Images were recorded on an AMT XR81 wide-field CCD camera. Images were analyzed in Fiji/ImageJ (https://imagej.nih.gov/ij/docs/guide/146.html), and 3D reconstructions were generated in Amira software version 6.5.0 (Thermo Scientific, Waltham, MA, USA).
Measurement of ER-EL contacts by electron microscopy. WT or SNX19-KO U-2 OS cells were plated on coverslips and incubated with 6-nm BSA-gold particles (Cat# 25483, Electron Microscopy Sciences, Hatfield, PA, USA) for 24 h. The next day, cells were washed twice in PBS, fixed in 4% paraformaldehyde in PBS at room temperature for 20 min and washed again three times in PBS. Cells were further fixed and processed for EM as described in the "CLEM" section. Eight cells were randomly selected in each condition, and a perinuclear and a peripheral region of interest was imaged for each cell, across 20-30 z-sections (~50 nm sections) from the base of the cell, resulting in~50 images per cell. Each image set was analyzed to first identify ELs by the presence of gold particles. Then, for all the sections that captured a given EL, the perimeter of the EL membrane was traced in Fiji/ImageJ and the length was scored in each section. Within the same sections, the lengths of all ER contacts with the same EL membrane (if present) were scored; if no ER contacts were observed, the length of ER contact was zero. The EL perimeter lengths across all sections in which a given EL is present, were summed up. The same was done for the ER contact lengths for that EL. The ER contact length was then divided by the EL perimeter length and multiplied by 100 to get a percentage of the total EL perimeter that contacts the ER. Typically, between 5 and 10 ELs were quantified per cell in this way; thus, we report the per-cell average of these values (Fig. 5b).
Proximity ligation assay (PLA). WT and SNX19-KO U-2 OS cells were plated on coverslips and, the following day, PLA was conducted using a mouse primary antibody to the endogenous ER protein calnexin and a rabbit primary antibody to the endogenous EL protein LAMP1 (antibody information under "Antibodies" section), followed by mouse and rabbit PLA probes from the DuoLink In Situ Red Starter Kit Mouse/Rabbit (Cat# DUO92101, Sigma-Aldrich, St. Louis, MO, USA) and detection with DuoLink In Situ Detection Reagents FarRed (Cat# DUO92103, Sigma-Aldrich). The PLA was conducted as per the manufacturer's instructions and coverslips were mounted with Fluoromount-G mounting medium containing DAPI (Cat# 00-4959-52, ThermoFisher, Waltham, MA, USA). Cells were imaged by confocal microscopy in the far red (647 nm) and DAPI (405 nm) channels.
A threshold was applied in the far red channel in Fiji/ImageJ to all acquired images to eliminate background, and all images were converted to binary images. The PLA signal was selected and the Watershed feature was applied to separate any signal dots that were apparently touching. Dots that were at least 5 pixels in size were counted and the total number of dots per cell (individual cells identified by DAPI nuclear stain) was measured across 3 experiments for each of the two conditions.
Measurement of endolysosomal peripheral dispersal. To quantify the peripheral distribution of LAMP1 structures, cells with a relatively round shape were chosen for the analysis, as narrow, elongated cells could not be quantified accurately by this method. These requirements were pre-determined and applied to all conditions. Cell outlines were traced in Fiji/ImageJ by briefly boosting the LAMP1 signal such that cytosolic fluorescence could be visualized. The LAMP1 signal was returned to original settings, a threshold was applied to eliminate the background and the total cellular LAMP1 signal was measured. The cell outline was iteratively shrunken by 2 μm to produce a total of 5 shells, shell 1 being the perinuclear region of the cell and shell 5 being the outermost edge of the cell. The LAMP1 signal within each shell was measured. Since WT cells contained the largest fraction of their lysosomes within shells 1 and 2 (represented by the white region in cell scheme in Fig. 5f), we computed the peripheral/total LAMP1 signal as the fraction of LAMP1 signal present in outer shells 3-5 (represented by gray region in the scheme in Fig. 5f) relative to total LAMP1 signal.
Endolysosome motility tracking. WT or SNX19-KO U-2 OS cells expressing LAMP1-RFP with or without GFP control, SNX19-GFP, SNX19 ΔPXA -GFP or SNX19 1-659 -GFP, were recorded for 10-30 s as described in the "Confocal microscopy" section. To determine motion types of individual ELs, recorded LAMP1-RFP data were subjected to Moment Scaling Spectrum (MSS) analysis as previously described 28 (software is publicly available at https://www. utsouthwestern.edu/labs/jaqaman/software/). Endolysosomes exhibited three major motion types: confined, free, or directed. Free and directed tracks were combined when reporting EL motility as they both represent ELs that are overall more motile than confined.
ER-EL dynamic associations in live-cell imaging experiments. U-2 OS cells cotransfected with plasmids encoding SNX19-GFP, LAMP1-RFP, and the ER marker BFP-KDEL were imaged live by confocal fluorescence microscopy at a frame rate of 2.4 s. Fiji/ImageJ was used to apply a threshold in each channel to the time-lapse recordings. SNX19-positive LAMP1-labeled organelles outside of the perinuclear area (where the ER signal cannot be resolved) were followed for 10 consecutive