Snx14 proximity labeling reveals a role in saturated fatty acid metabolism and ER homeostasis defective in SCAR20 disease

Significance

SCAR20 disease is an autosomal-recessive spinocerebellar ataxia primarily affecting children, and results from loss-of-function mutations in the SNX14 gene. Snx14 is an endoplasmic reticulum (ER)-localized protein that localizes to ER–lipid droplet (LD) contacts and promotes LD biogenesis following exogenous fatty acid (FA) treatment. Why Snx14 loss causes SCAR20 is unclear. Here, we demonstrate that, following exposure to saturated FAs, Snx14-deficient cells have defective ER homeostasis and altered lipid saturation profiles. We reveal a functional interaction between Snx14 and FA desaturase SCD1. Lipidomics shows Snx14-deficient cells contain elevated saturated lipids, closely mirroring SCD1-defective cells. Furthermore, SCD1 overexpression can rescue Snx14 loss. We propose that Snx14 maintains cellular lipid homeostasis, the loss of which underlies the cellular basis for SCAR20 disease.

Abstract

Fatty acids (FAs) are central cellular metabolites that contribute to lipid synthesis, and can be stored or harvested for metabolic energy. Dysregulation in FA processing and storage causes toxic FA accumulation or altered membrane compositions and contributes to metabolic and neurological disorders. Saturated lipids are particularly detrimental to cells, but how lipid saturation levels are maintained remains poorly understood. Here, we identify the cerebellar ataxia spinocerebellar ataxia, autosomal recessive 20 (SCAR20)-associated protein Snx14, an endoplasmic reticulum (ER)–lipid droplet (LD) tethering protein, as a factor required to maintain the lipid saturation balance of cell membranes. We show that following saturated FA (SFA) treatment, the ER integrity of SNX14^KO cells is compromised, and both SNX14^KO cells and SCAR20 disease patient-derived cells are hypersensitive to SFA-mediated lipotoxic cell death. Using APEX2-based proximity labeling, we reveal the protein composition of Snx14-associated ER–LD contacts and define a functional interaction between Snx14 and Δ-9 FA desaturase SCD1. Lipidomic profiling reveals that SNX14^KO cells increase membrane lipid saturation following exposure to palmitate, phenocopying cells with perturbed SCD1 activity. In line with this, SNX14^KO cells manifest delayed FA processing and lipotoxicity, which can be rescued by SCD1 overexpression. Altogether, these mechanistic insights reveal a role for Snx14 in FA and ER homeostasis, defects in which may underlie the neuropathology of SCAR20.

Cells regularly internalize exogenous fatty acids (FAs) and must remodel their metabolic pathways to process and properly store FA loads. As it is a central cellular currency that can be stored, incorporated into membrane lipids, or harvested for energy, cells must balance FA uptake, processing, and oxidation to maintain homeostasis. Defects in any of these elevates intracellular free FAs (FFAs), which can act as detergents and damage organelles. Excessive membrane lipid saturation can also alter organelle function and contribute to cellular pathology, known as lipotoxicity (1, 2). Failure to properly maintain lipid compositions and storage contributes to many metabolic disorders (3), including type 2 diabetes (4), obesity (5), cardiac failure (6, 7), and various neurological diseases (8).

Properties of FAs such as their degree of saturation and chain length are key determinants of their fate within the cell (9). High concentrations of saturated FAs (SFAs) in particular are highly toxic, as their incorporation into organelles affects membrane fluidity and can trigger lipotoxicity and cell death (10–13). To prevent this, cells desaturate SFAs into monounsaturated FAs (MUFAs) before they are subsequently incorporated into membrane glycerophospholipids or stored as triglycerides (TGs) in lipid droplets (LDs). LD production provides a lipid reservoir to sequester otherwise toxic FAs, providing a metabolic buffer to maintain lipid homeostasis (14, 15).

As LDs are created by and emerge from the ER network, interorganelle communication between the ER and LDs is vital for LD biogenesis (16). Consequently, numerous proteins that contribute to LD biogenesis, such as seipin (17, 18) and the diacylglyceride acyltransferase (DGAT) (19), are implicated in ER–LD cross-talk. Previously, we identified Snx14, a sorting nexin (SNX) protein linked to the cerebellar ataxia disease spinocerebellar ataxia, autosomal recessive 20 (SCAR20) (20–22), as a novel factor that promotes FA-stimulated LD growth at ER–LD contacts (23, 24). Snx14 is an ER-anchored integral membrane protein. During periods of elevated FA flux, Snx14 is recruited to ER–LD contact sites, where it promotes the incorporation of FAs into TG as LDs grow (23). In line with this, SNX14^KO cells exhibit defective LD morphology following oleate addition, implying Snx14 is required for proper FA storage in LDs. Related studies of Snx14 homologs in yeast and Drosophila indicate a conserved role for Snx14-family proteins in FA homeostasis and LD biogenesis (25, 26).

Despite these insights, why humans with Snx14 loss-of-function mutations develop the cerebellar ataxia disease SCAR20 remains enigmatic. Given the proposed role of Snx14 in lipid metabolism, and that numerous neurological pathologies arise through defects in ER lipid homeostasis (27–29), here we investigated whether Snx14 loss alters the ability of cells to maintain lipid homeostasis in response to FA influx. Our findings indicate that Snx14-deficient cells are hypersensitive to SFA exposure and manifest defects in ER morphology and ER-associated lipid metabolism.

Results

SNX14^KO Cells Are Hypersensitive to SFA-Associated Lipotoxicity.

Previously, we showed that, following oleate addition, Snx14 enriches at ER–LD contacts to promote LD growth, and Snx14 loss disrupts LD homeostasis (23). To further dissect Snx14 function in maintaining lipid homeostasis, we interrogated how SNX14^KO cells respond to exposure to various SFAs and unsaturated FAs. We exposed wildtype (WT) or SNX14^KO U2OS cells to titrations of specific SFAs or MUFAs for 48 h and monitored cell viability using an established crystal violet assay (30, 31). Exposure to MUFAs including palmitoleate (16:1) and oleate (18:1) did not perturb cell viability of either WT or SNX14^KO cells even at 1,000-μM concentrations (Fig. 1 A and A′). In contrast, treatment of WT cells with increasing concentrations of SFAs such as palmitate (16:0) or stearate (18:0) resulted in decreased cell survival, as previously reported (13, 32) (Fig. 1 B and B′).

Fig. 1.

Snx14-deficient cells are hypersensitive to SFAs. (A) Percent surviving cells denoted as cell viability of WT and SNX14KO cells following treatment with palmitoleate (0, 250, 500, 750, 1,000 μM, FA|16:1) for 2 d. (A′) Treatment of WT and SNX14^KO cells with concentrations of oleate (FA|18:1) for 2 d. Assay was repeated thrice in triplicate. Values are mean ± SEM. (B) Cell viability of WT and SNX14^KO cells showing SNX14^KO cells as hypersensitive following concentrations of palmitate (0, 250, 500, 750, 1,000 μM, FA|16:0) for 2 d. (B′) Exposure to concentrations of stearate (FA|18:0) for 2 d in WT and SNX14^KO cells. The assay was repeated thrice in triplicate. Values represent mean ± SEM. Significance test between WT and SNX14^KO (n = 3); ***P < 0.0001,**P < 0.001,*P < 0.01, multiple t test by Holm–Sidak method with alpha = 0.05. (C) Cell viability of fibroblasts derived from two WT (SNX14⁺/SNX14⁺) subjects and one SCAR20 patient with homozygous recessive mutations in SNX14 showing hypersensitivity of SCAR20 cells following 2-d exposure to 0, 1,000, and 2,000 μM palmitate. Assay repeated thrice in triplicate. Values represent mean ± SEM (n = 3; *P < 0.01 relative to WT-1; ##P < 0.001 and ###P < 0.0001 relative to WT-2, multiple t test by Holm–Sidak method with alpha = 0.05). (D) Cell viability of WT and SNX14^KO cells following exposure to palmitate concentration (0, 250, 500, 750, 1,000 μM) and treatment with 10 μM etomoxir for 2 d. Assay repeated thrice in triplicate. Values represent mean ± SEM. (E) Cell viability of WT and SNX14^KO cells following addition of increasing palmitate concentration (0, 250, 500, 750, 1,000 μM) in the presence of 50 μM myriocin for 2 d. Assay repeated thrice in triplicate. Values represent mean ± SEM.

Intriguingly, SNX14^KO U2OS cells were hypersensitive to SFA-induced death compared to WT cells (Fig. 1 B and B′). The concentration of palmitate at which ∼50% of WT cells survive is ∼1,000 μM, but only ∼500 μM for SNX14^KO (Fig. 1B). Similarly, exposure to ∼600 μM stearate resulted in ∼50% cell viability for WT cells, but this only required ∼300 μM for SNX14^KO (Fig. 1B′). Consistent with this, SCAR20 patient-derived fibroblasts (20, 24), which are homozygous for loss-of-function Snx14 mutations, exhibited significantly reduced cell viability compared to control fibroblasts following palmitate addition (Fig. 1C). SCAR20 patient cells exhibited ∼50% viability at ∼1,000 μM palmitate exposure, whereas more than 50% WT cells were viable even at 2,000 μM palmitate treatment. Collectively, these observations indicate that Snx14-deficient cells are hypersensitive to SFA exposure.

Once internalized, FFAs are esterified and shunted into several distinct metabolic fates, including their incorporation into ceramides (33), glycerophospholipids (34), or neutral lipids (35). They can also be harvested by catabolic breakdown in oxidative organelles like mitochondria (2). To begin to dissect why SNX14^KO cells were hypersensitive to SFAs, we conducted a systemic analysis of each FA-associated pathway. Since intracellular ceramide accumulation can itself be toxic (33), we tested whether pharmacologically lowering ceramide synthesis could rescue SNX14^KO SFA-associated toxicity. We treated cells with myriocin, which inhibits the SPT complex that incorporates palmitate into newly synthesized ceramides (10). Myriocin treatment did not rescue SNX14^KO SFA hypersensitivity, suggesting that elevated ceramides do not contribute to SNX14^KO palmitate-induced cell death (Fig. 1D). Next, we examined whether mitochondrial FA oxidation was required for SNX14^KO hypersensitivity by treating cells with etomoxir that inhibits mitochondrial FA uptake (36, 37). This did not suppress palmitate-induced lipotoxicity in SNX14^KO cells, indicating perturbed mitochondrial FA oxidation is likely not causative of Snx14-associated lipotoxicity (Fig. 1E).

SFA-Induced Lipotoxicity in SNX14^KO Cells Is Associated with Defects in ER Morphology.

A major FA destination is their incorporation into membrane glycerophospholipids via de novo lipid synthesis at the ER. Excessive SFA incorporation into diacyl-glycerophospholipids can impact ER membrane fluidity and drive cellular stress (12, 38). To determine whether SFA exposure affected ER homeostasis in SNX14^KO cells, we examined ER morphology of WT and SNX14^KO cells exposed to palmitate via fluorescence confocal microscopy. Whereas the immunofluorescently stained ER network of WT cells was reticulated and regular following overnight exposure to 500 μM palmitate, SNX14^KO cells displayed perturbed ER morphology. The ER of SNX14^KO cells appeared fragmented and contained drastic bulges within the tubular network following palmitate treatment (Fig. 2A, red arrows). A subpopulation of SNX14^KO cells even exhibited solubilized ER lumen marker in the cytoplasm, suggesting drastic defects in ER integrity. We quantified these ER morphological alterations into distinct classes (Fig. 2 A and B, white arrows indicating ER morphology of each class), revealing ∼70% of SNX14^KO cells exhibited irregular or fragmented ER structure following palmitate exposure, compared to only ∼22% of WT (Fig. 2B). These perturbations became more prominent when examining ER ultrastructure with transmission electron microscopy (TEM). Here, WT and SNX14^KO cells were either left untreated or cultured in media containing 500 μM of palmitate for 6 h or 16 h. Even with palmitate, WT cells exhibited normal tubular ER networks and did not manifest any significant change in the ER morphology (Fig. 2C). SNX14^KO cells not exposed to palmitate also exhibited normal ER morphology. However, the ER of SNX14^KO cells exposed to 6 h palmitate appeared swollen and dilated, forming sheet-like extensions within the thin-section plane (Fig. 2C, red arrows). This ER dilation was more pronounced following 16 h treatment (Fig. 2C, red arrows).

Fig. 2.

Palmitate-induced hypersensitivity in SNX14^KO is associated with defective ER morphology. (A) IF labeling of the ER with α-HSP90B1 (ER marker) antibody before and after overnight palmitate treatment in WT and SNX14^KO cells. (Scale bar, 10 μm.) Red arrows indicate the drastic bulges in the ER. White arrows indicate each class of ER morphology where A is regular ER, B is partially fragmented ER, C is fully fragmented ER, and D shows soluble ER marker. (B) Percentage of palmitate-treated WT and SNX14^KO cells quantified and grouped based on whether the ER morphology is regular (A), partially fragmented (B), fully fragmented (C), or completely soluble (D). A total of ∼100 cells were quantified from 3 experiments. Values represent mean ± SEM. (C) TEM micrographs of WT and SNX14^KO cells with no palmitate treatment and with palmitate treatment for 6 h and overnight to visualize ER ultrastructure. (Scale bars: 1 μm; Inset, 0.5 μm.) Red arrows indicate ER dilation in palmitate-treated SNX14^KO cells.

Since Snx14 is implicated in LD biogenesis, we also examined LD morphology in WT and SNX14^KO cells exposed to palmitate. Whereas WT cells generated small LDs in response to palmitate, SNX14^KO cells exhibited significantly fewer LDs that stained poorly with the LD dye monodansylpentane (MDH), suggesting defective palmitate processing and LD incorporation (SI Appendix, Fig. S1 A and B). Collectively, these observations suggest that SNX14^KO cells manifest altered ER architecture and LD homeostasis following prolonged exposure to palmitate.

Changes in ER lipid homeostasis can induce the unfolded protein response (UPR) as well as caspase-dependent apoptotic cell death (13). To understand whether such responses were associated with palmitate-induced cell death in Snx14-deficient cells, we monitored them in SNX14^KO cells. As expected, palmitate treatment elevated s-Xbp1 levels compared to no treatment in both WT and SNX14^KO cells, but there was not a significant difference in s-Xbp1 levels between the two cell lines (SI Appendix, Fig. S1C). In line with this, addition of the IRE1 inhibitor 4μ8c or PERK inhibitor GSK2606414, both of which suppress branches of UPR signaling, did not rescue palmitate-induced cell death in SNX14^KO cells, suggesting altered or hyperactive UPR signaling was not causative of SNX14^KO cell death following palmitate exposure (SI Appendix, Fig. S1D). To dissect whether SFAs induced an apoptotic response in SNX14^KO cells, we also treated cells with the caspase-6/8 inhibitor SCP0094. This did not rescue palmitate-induced cell death, indicating SNX14^KO cells were not manifesting hyperactive caspase-6/8–dependent apoptosis (SI Appendix, Fig. S1D).

APEX-Based Proteomics Reveals Snx14 Is in Proximity to Proteins Involved in SFA Metabolism.

Given that Snx14 was required for maintaining ER morphology and LD biogenesis following palmitate exposure, we next investigated what proteins Snx14 interacted with that may promote lipid homeostasis. Previously, we utilized an ascorbate peroxidase APEX2-based proximity technology to examine the localization of APEX2-tagged Snx14 at ER–LD contact sites using TEM (23). APEX2 tagging also enables the local interactome of a protein of interest to be interrogated. The addition of biotin-phenol and hydrogen peroxide to APEX2-expressing cells induces the local biotinylation of proteins within ∼20 nm of the APEX2 tag. These biotinylated proteins can be subsequently affinity-purified via streptavidin beads and identified via mass spectrometry (MS) (39) (Fig. 3A). As ER–LD contacts are lipogenic ER subdomains with known roles in FA processing, we hypothesized that Snx14’s enrichment at ER–LD junctions represented an opportunity to use APEX2 technology to reveal proteins that contribute to ER lipid metabolism in conjunction with Snx14.

Fig. 3.

APEX2-based proteomics reveals the Snx14-associated ER–LD proteome. (A) Schematic diagram of Snx14-EGFP-APEX2. With oleate treatment, APEX2-tagged Snx14 enriches at ER–LD contacts, and labeling reaction will biotinylate Snx14 interactors at the contacts. (B) Co-IF staining of cells stably expressing Snx14-EGFP-APEX2 with anti-EGFP (green) antibody, streptavidin-conjugated Alexa 647 fluorophore (biotinylated proteins; red) antibody, and LDs stained with MDH (blue) followed by confocal imaging revealed colocalization of Snx14 and biotinylated proteins surrounding LDs. (Scale bar, 10 μm.) (C) Biotinylated proteins pulled down by streptavidin-conjugated beads from HEK293 and U2OS cells expressing Snx14-EGFP-APEX2 or no APEX2 (negative control) were Coomassie-stained. Western blot of the same pulled-down lysates with streptavidin-HRP antibody revealed biotinylation of several proteins in Snx14-EGFP-APEX2 relative to the negative control. (D) Heat map of ∼60 proteins highly enriched in Snx14-APEX2 proteomics and cross-referenced with other LD proteomics studies (40–44) (black box represents detected, and white box represents no detection). Heat map shows enrichment of the proteins in Snx14-APEX2 proteomics (red columns) as well as the deenrichment in the cyto-APEX2 and ER-APEX2 proteomics (blue columns; n.d., not detected). (E) Venn diagram of multistage analysis of the MS data, which identified protein abundance in Snx14-EGFP-APEX2, no-APEX2, and EGFP-APEX2-Sec61. A total of 2,376 proteins are enriched more than twofold in Snx14-EGFP-APEX2 (1, purple circle). Among them, 674, which are also enriched in APEX2-tagged Sec61, are excluded (2, black circle). 305 of group 1 are ER-associated according to Gene Ontology analysis (3, red circle). 310 of group 1 comprise genes whose CRISPR/Cas9 deletion causes palmitate sensitivity in (46) (4, yellow circle). Overlap of sets 1, 3, and 4 consists of 29 ER-associated proteins, which are enriched for Snx14 and important for palmitate metabolism. Set belongs to the Snx14 interactome and is compared with its fly homolog Snz interactome. This final comparison indicated SCD1 as one of the top candidates. (F) Cell viability of WT and SNX14^KO cells treated with 4 μM SCDi with increasing palmitate concentration (0, 250, 500 μM) for 2 d. Assay repeated thrice in triplicates. Values represent mean ± SEM (n = 3; *P < 0.1, **P < 0.001, ***P < 0.0001, multiple t test by Holm–Sidak method with alpha = 0.05).

We generated U2OS cells stably expressing Snx14-EGFP-APEX2 and HEK293 cells transiently expressing Snx14-EGFP-APEX2 and exposed them to oleate to induce Snx14 recruitment to ER–LD contacts. We then treated cells with biotin-phenol for 30 min and hydrogen peroxide for 1 min. As expected, coimmunofluorescence staining of U2OS cells for Streptavidin-Alexa647 and EGFP revealed their colocalization, suggesting proteins in close proximity to Snx14-EGFP-APEX2 were biotinylated (Fig. 3B). Biotinylated proteins from both U2OS and HEK293 cells were then affinity-purified with streptavidin beads. Gel electrophoresis followed by Coomassie staining and anti-streptavidin/HRP Western blotting revealed many biotinylated proteins in Snx14-EGFP-APEX2–expressing cell lysates, but not in controls lacking APEX2 (Fig. 3C). The bead-enriched biotinylated proteins from both Snx14-EGFP-APEX2 and the control lacking APEX2 were then identified by tandem MS/MS proteomic analysis.

Previous proteomics studies have isolated LDs and characterized LD-associated proteins (40–44), and several of these were also detected in our APEX2-based approach (Fig. 3D). However, to confirm that these peptide hits were specific to the Snx14-EGFP-APEX2 interactome, we also conducted proteomics on biotinylated proteins from cells expressing a soluble EGFP-APEX2 (cyto-APEX) as well as cells expressing APEX2-tagged Sec61β, a general ER marker (ER-APEX). High-abundance peptides in the Snx14-EGFP-APEX2 proteomics that were correspondingly low in the cyto-APEX and ER-APEX were thus considered high-confidence hits (Fig. 3D and SI Appendix, Table S1). Notably, this list included well-characterized LD surface proteins Perilipin 3 (PLIN3), Perilipin 2, and PNPLA2. In fact, PLIN3 was one of the most enriched proteins from both cell line samples (>65-fold; SI Appendix, Fig. S3A), consistent with it being a highly abundant LD protein that coats newly synthesized LDs. Proteins recently highlighted as localizing to ER–LD contact sites were also identified, including VPS13C and VPS13A (45). We also detected several Rab proteins such as Rab7a, Rab10, and Rab11b, which had previously been detected on LDs using an LD-targeted APEX approach (44). Additionally, we detected enzymes involved in ER-associated lipid synthesis, including lyso-phospholipid acyltransferases LPCAT1, LPCAT3, and LPGAT, as well as enzymes driving sterol metabolism like the sterol O-acyltransferase SOAT1, the squalene synthase FDFT1, and the squalene epoxidase SQLE. Last, we noted several enzymes involved in FA metabolism and desaturation, including ACSL4, ELOVL1, ELOVL5, FASN, SCD1, FADS1, and FADS2.

To identify Snx14 functional interactors that may play a role in FA metabolism that was defective in SNX14^KO cells, we implemented a multistage analysis approach (Fig. 3E). First, we selected proteins that were greater than 2.0-fold enriched in Snx14-EGFP-APEX2 samples over negative controls lacking APEX2 (Fig. 3E, circle 1, and SI Appendix, Fig. S2A). Next, we focused on proteins annotated to localize to the ER network but also specific to the Snx14 interactome, since the ER was the primary organelle that manifested morphological alterations in SNX14^KO cells following palmitate exposure. We did this by eliminating proteins detected at similar levels in both the ER-APEX and Snx14-EGFP-APEX2 proteomics (Fig. 3E, circle 2, and SI Appendix, Fig. S2 A and B). We then applied Gene Ontology enrichment analysis on the annotated ER-associated proteins specific to Snx14 (Fig. 3E, circle 3, and SI Appendix, Fig. S2C), revealing that ∼250 of these proteins are associated with cellular metabolism, and ∼100 with lipid metabolism (SI Appendix, Fig. S2D).

With this smaller candidate list, we compared them to a recently published genome-wide CRISPR/Cas9 screen identifying proteins whose loss sensitized cells to palmitate-induced lipotoxicity (46). Remarkably, this unbiased screening identified Snx14 in the top 6% of proteins that were protective against palmitate (gene score, −2.28). This study also identified 310 proteins detected in our Snx14 interactome that exhibited a negative score greater than −1.8 for palmitate sensitivity (46) (SI Appendix, Table S2, SI Appendix, Fig. S2A, orange dots, and SI Appendix, Fig. S2E, circle 4). These palmitate-sensitive proteins included FADS1, FADS2, SCD1, CEPT1, and HSD17B12 (SI Appendix, Fig. S2F).

As a final stage of analysis, we compared our list to our recently published interactome of Snx14 Drosophila melanogaster ortholog Snz, which also functions in lipid homeostasis in fruit fly adipocytes (26). Snz proteomics identified the enzyme DESAT1, the major fly Δ-9 FA desaturase, as a key Snz functional interactor that was required for Snz-driven TG synthesis in Drosophila (26). DESAT1 is the ortholog of human Δ-9 FA desaturase SCD1, which catalyzes the conversion of SFAs to MUFAs prior to their incorporation into glycerophospholipids or neutral lipids (47). Indeed, several lines of evidence converged on SCD1 and its desaturation activity as being functionally linked to Snx14: 1) both proteins are ER-associated integral membrane proteins, and SCD1 was highly enriched in Snx14-EGFP-APEX2 proteomics in both U2OS and HEK293 cell lines (Fig. 3D and SI Appendix, Fig. S2A); 2) SCD1 genetic or pharmacological perturbation hypersensitizes cells to palmitate-induced cell death, similar to Snx14 loss (13, 48, 49) (Fig. 3F); and 3) in genome-wide screening, both Snx14 and SCD1 scored similarly in impact to palmitate sensitivity (SCD1, −2.42; Snx14, −2.28) (46). Note that, here, we are operationally defining a potential functional interaction between SCD1 and Snx14 based on their phenotypical similarities; this is distinct from any potential physical interaction. Based on this analysis, we chose to investigate the functional interplay between Snx14 and SCD1-associated FA desaturation in ER lipid homeostasis.

SNX14^KO Cell FA Elevations Are Similar to Cells with Reduced SCD1 Activity.

To begin dissecting the functional interplay between Snx14 and FA metabolism, we conducted whole-cell lipidomic FA profiling. We pulsed WT and SNX14^KO cells for 2 h with 500 μM palmitate, extracted polar lipids, and conducted gas chromatography-MS analysis to profile polar lipid fatty acyl chain length and saturation. Indeed, SNX14^KO cells exhibited a significant increase in total polar lipid-derived FAs (SI Appendix, Fig. S3A), as well as a significant increase in the abundance of 16-carbon length saturated fatty acyl chains (FA|16:0) in polar lipids relative to WT cells (Fig. 4A). To determine whether this increase mimics alterations in polar lipids when SCD1 function is perturbed, we also treated WT cells with SCD1 inhibitor (SCD1i). Indeed, WT cells exposed to SCD1i exhibited similar increases in 16:0 saturated fatty acyl chains in their polar lipid fraction, mirroring elevations observed in SNX14^KO cells (Fig. 4A). Furthermore, SCD1i-treated SNX14^KO cells exhibited no additional 16:0 acyl chain increases in the polar lipid profile, implying SCD1i treatment and Snx14 loss may perturb the same pathway.

Fig. 4.

SNX14^KO cell lipidomic profiling is similar to cells with reduced SCD1 activity. (A) Abundance of FAs (16:0) derived from polar lipids of WT and SNX14^KO cells relative to untreated WT under the following conditions: no treatment, palmitate treatment, and treatment with palmitate and SCD1i. Values represent mean ± SEM (n = 3, *P < 0.01, multiple t test by Holm–Sidak method with alpha = 0.05). (B) Heat map indicating the relative change in abundance of individual lipid species of U2OS WT and SNX14^KO cells before and after palmitate treatment relative to untreated WT cells. (C) Heat map indicating the relative change in abundance of 60 different TG species of WT and SNX14^KO cells relative to untreated WT. These cells were either untreated or treated with palmitate alone or in the presence of SCD1i. The label “N” indicates the number of double bonds in those group of TG species. The vertical serial number represents a different TG species where “1” is TG|50:0|(NL-16:0) and “11” is TG|48:1|(NL-16:1), and they exhibit the most changes in TG in the treated SNX14^KO cells. (D) Abundance of TG (with zero or one unsaturation) relative to untreated WT as analyzed in WT and SNX14^KO cells quantified C. Prior to lipid extraction and lipidomics, these cells were either untreated or treated with palmitate alone or in the presence of SCD1i. Values represent mean ± SEM (n = 3; *P < 0.01, multiple t test by Holm–Sidak method with alpha = 0.05). (E) Heat map indicating the relative change in abundance of nine different LPC species of WT and SNX14^KO cells relative to untreated WT. These cells were either untreated or treated with palmitate alone or in the presence of SCD1i.

TG, PC, and Lysolipids Exhibit Significant Elevations in Saturated Acyl Chains in SNX14^KO Cells.

To dissect how specific lipid classes are altered by Snx14 loss, we conducted global lipidomics via quantitative liquid chromatography (LC)-MS analysis. WT and SNX14^KO cells were either left untreated or exposed to 2 h of palmitate, harvested, and analyzed. For comparative analysis, we again treated some cells with SCD1i. Importantly, this LC-MS analysis revealed fatty acyl properties such as chain length and saturation types of major lipid classes, including TG, diacylglyceride (DAG), phosphatidylethanolamine, phosphatidylcholine (PC), phosphatidic acid, phosphatidylserine (PS), and lysophospholipids. Analysis revealed a significant (∼40%) increase in the levels of DAG, lysophosphatidylethanolamine (LPE), and PC in SNX14^KO cells relative to WT cells following palmitate treatment (Fig. 4B). However, relative abundances of most lipid classes between these two samples with or without palmitate treatment were not drastically altered.

Next, we examined the change in acyl-chain saturation within each lipid class. In general, there was a greater proportion of saturated fatty acyl chains within several lipid classes. The most significant change in saturated fatty acyl chains in palmitate-treated SNX14^KO cells relative to WT was in TG and lyso-PC (LPC). Examining ∼60 different TG species revealed the most significant increase in TG species containing two or more saturated acyl chains. These TG species contained either zero or one unsaturation among all three acyl chains (denoted as an N score of 0 or 1; Fig. 4C). For example, TG|48:1(NL-16:1), which contains one monounsaturated (16:1) acyl chain and two fully saturated acyl chains, was significantly elevated in SNX14^KO cells following palmitate treatment (Fig. 4C, species 11). Notably, TG 50:0 (NL-16:0) and 50:0 (NL-18:0), both of which contained three saturated acyl chains, were also elevated in SNX14^KO cells compared to WT following palmitate addition and SCD1 inhibition, suggesting SNX14^KO cells incorporated SFAs into TG more than WT cells (Fig. 4C, species 1 and 2). For more global analysis, we pooled the abundances of all TG species comprising only zero or one total fatty acyl unsaturation. Indeed, TG pooling revealed that TG containing zero or one unsaturation was significantly increased in palmitate-treated SNX14^KO cells and closely mirrored levels of WT cells treated with SCD1i (Fig. 4D). Collectively, this suggests that the TG lipid profile of palmitate-treated SNX14^KO cells exhibits increased acyl-chain saturation and is similar to cells with decreased SCD1 function.

Among polar lipid species, the abundance of LPC species containing the saturated fatty acyl chain 18:0 was significantly increased in palmitate-treated SNX14^KO cells relative to WT (Fig. 4E). In fact, levels of 18:0-containing LPC in SNX14^KO cells treated with palmitate closely mirrored WT cells treated with palmitate and SCD1i, again indicating that SNX14^KO cells closely resembled cells with inhibited SCD1 function by lipid profiling (Fig. 4E). In line with this, there was a decrease in LPC species containing 18:1 or 18:2 acyl chains following palmitate treatment. LPE species containing 18:0 fatty acyl chains were also significantly increased in SNX14^KO cells even without palmitate treatment (SI Appendix, Fig. S3B). Profiling of PC, one of the most abundant glycerophospholipids, also revealed an increase in PC species with 18:0 fatty acyl chain in SNX14^KO cells, and that was more pronounced with palmitate treatment followed by SCD1 inhibition (SI Appendix, Fig. S3D). PS lipid profiles were similarly altered in SNX14^KO, with increases in PS species containing 18:0 saturated fatty acyl groups and a corresponding decrease in PS with 18:1 or 18:2 acyl chains (SI Appendix, Fig. S3C).

Collectively, LC-MS lipidomics suggests: 1) SNX14^KO cells exhibit elevated TG species containing two or more saturated acyl chains, 2) SNX14^KO cells have increased levels of LPC and LPE lysolipids containing saturated fatty acyl chains, 3) SNX14^KO cells exhibit more PC and PS with an 18:0 saturated acyl chain and less with 18:1 or 18:2, and, 4) for TG and LPC, the lipid profiles of SNX14^KO cells exposed to palmitate closely mirror cells with inhibited SCD1 function.

SCD1 Activity Can Rescue SNX14^KO Palmitate-Induced Lipotoxicity.

Given the increased lipid-saturation profiles of SNX14^KO cells, we next examined SCD1 protein levels in the context of Snx14 loss. SCD1 protein abundance was similar in WT and SNX14^KO cells under ambient conditions, but became elevated in both cell lines when exposed to 500 μM palmitate. Notably, SCD1 protein levels were significantly more elevated in SNX14^KO cells exposed to palmitate. Snx14-Flag–overexpressing cells returned SCD1 protein levels to WT levels, implying that SNX14^KO cells may further elevate SCD1 protein expression to compensate for Snx14 loss (Fig. 5 A and B).

Fig. 5.

SCD1 activity can rescue SNX14^KO palmitate-induced lipotoxicity. (A) Western blot of SCD1 and Hsp90B1 (ER marker) before and after overnight palmitate treatment in WT, SNX14^KO, and Snx14Flag-overexpressed (O/E) cells. (B) Ratio of intensity of SCD1 protein over Hsp90B1 from A, plotted as fold change relative to untreated WT, (ratio is set as 1). Values represent mean ± SEM. Significance test between before and after palmitate treatment (n = 3): #P < 0.1, ##P < 0.001, multiple t test by Holm–Sidak method with alpha = 0.05. Significance test between WT, SNX14^KO, and Snx14FlagO/E (n = 3): *P < 0.01, multiple t test by Holm–Sidak method with alpha = 0.05. (C) Cell viability of WT and SNX14^KO cells showing sensitivity of both WT and SNX14^KO cells rescued with SCD1 overexpression following addition of palmitate (0, 500, 750, 1,000 μM) for 2d. Assay repeated thrice in triplicate. Values represent mean ± SEM. Significance test WT and WT+SCD1 (n = 3): **P < 0.001, ***P < 0.0001, multiple t test by Holm–Sidak method with alpha = 0.05. Significance test with SNX14^KO and SNX14^KO+SCD1 (n = 3): ##P < 0.001, ##P < 0.0001, multiple t test by Holm–Sidak method with alpha = 0.05. (D) TLC of neutral lipids and FFAs performed in WT and SNX14^KO cells with 500 μM palmitate for 0, 2, 4, 8, or 16h. Quantification of relative fold change in FFA (normalized to cell pellet weight) with respect to untreated WT from this TLC. Values represent mean ± SEM (n = 3, *P < 0.01, multiple t test by Holm–Sidak method with alpha = 0.05). (E) TLC of neutral lipids and FFAs performed in WT and SNX14^KO cells before and after SCD1 overexpression following exposure to 500 μM palmitate for 0, 4, or 16 h. Quantification of relative fold change in FFA (normalized to cell pellet weight) with respect to untreated WT from this TLC. Values represent mean ± SEM (n = 3; *P < 0.01, **P < 0.001, multiple t test by Holm–Sidak method with alpha = 0.05). (F) IF labeling of ER with α-HSP90B1 (ER marker) antibody in SNX14^KO cells before and after overexpression of SCD1 following overnight palmitate treatment. (Scale bar, 10 μm.) (G) Percentage of palmitate-treated WT and SNX14^KO cells quantified and grouped based on whether the ER morphology is regular (A), partially fragmented (B), fully fragmented (C), or completely soluble (D). A total of ∼100 cells were quantified from 3 experiments. Values represent mean ± SEM.

Since SNX14^KO cells exhibited slightly elevated SCD1 levels, we queried whether ectopic SCD1 overexpression could reduce SNX14^KO-associated palmitate hypersensitivity. Strikingly, SCD1 overexpression rescued SNX14^KO cell viability, and SNX14^KO cells now responded similarly to WT cells exposed to dose-dependent palmitate treatment (Fig. 5C). SNX14^KO cells were also rescued by exposure to mixtures of palmitate and oleate, the MUFA and product of SCD1 enzymatic activity (SI Appendix, Fig. S4A). Since SNX14^KO cells also displayed defective LD morphology with palmitate, we tested whether SCD1 overexpression could restore LD levels. Indeed SNX14^KO cells overexpressing SCD1 manifested more LDs following palmitate exposure (SI Appendix, Fig. S2 A and B). Since LDs are lipid reservoirs, we also determined whether this SCD1-mediated rescue required the incorporation of FAs into TG for LD storage. We exposed SNX14^KO cells to DGAT1/2 inhibitors (DGATi) in the presence of palmitate and monitored cell viability. Surprisingly, SCD1 overexpression rescued SNX14^KO cells even with DGATi, suggesting this SCD1-mediated rescue functions upstream of TG synthesis (SI Appendix, Fig. S4B).

Since SNX14^KO and SCD1i-exposed cells displayed some lipidomics similarities, and SCD1 overexpression mitigated palmitate-induced SNX14^KO cell death, we hypothesized that SNX14^KO cells had defects processing SFAs. To test this, we exposed WT and SNX14^KO cells to 500 μM palmitate for 0, 2, 4, 8, and 16 h, extracted whole cell lipids, and conducted thin layer chromatography (TLC) to monitor changes in FFAs and neutral lipids. As expected, both WT and SNX14^KO cells exhibited elevated FFA levels following 2 and 4 h palmitate exposure, but SNX14^KO exhibited significantly elevated FFAs compared to WT (Fig. 5D and SI Appendix, Fig. S4C). SCD1 overexpression reversed this FFA elevation at 4 h in SNX14^KO cells, implying the elevated FFA pool was composed of SFAs that could be processed by SCD1 (Fig. 5E and SI Appendix, Fig. S4D). As a key control, we monitored uptake of radiolabeled ¹⁴C-palmitate in WT and SNX14^KO cells to test whether elevated FFA accumulation in SNX14^KO cells was due to increased FA uptake. We confirmed that FA uptake was not altered by Snx14 loss (SI Appendix, Fig. S4E). This indicates that SNX14^KO cells accumulate FFAs following palmitate uptake, consistent with a defect in palmitate processing, and these effects can be reversed by SCD1 overexpression.

To understand whether the ER fragmentation observed in SNX14^KO cells is associated with defects in palmitate processing, we examined ER morphology in SCD1-overexpressed SNX14^KO cells following palmitate exposure. Remarkably, ectopic expression of SCD1 in palmitate-treated SNX14^KO cells rescued ER morphology (Fig. 5 F and G). Collectively, this suggests that SCD1 overexpression can rescue SNX14^KO elevated lipid saturation, LD morphology, and ER fragmentation.

Snx14 Functionally Interacts with SCD1 in the ER Network.

Since Snx14 and SCD1 appeared to have phenotypical similarities, and SCD1 overexpression could rescue aspects of Snx14 loss, we next examined whether the two proteins could coimmunoprecipitate (co-IP), which would suggest their close proximity to each other within the ER network. We generated U2OS cell lines stably expressing either 3×Flag-tagged Snx14 or GFP, conducted anti-Flag immunoprecipitation, and Western-blotted the immunoprecipitated (IP) lysates for endogenous SCD1. SCD1 was detected in the Snx14-Flag sample but not GFP-Flag (Fig. 6A). Intriguingly, Snx14-Flag co-IPed endogenous SCD1 both with and without palmitate addition (Fig. 6A). To determine whether this interaction was specific, we Western-blotted for PLIN3, which was highly enriched in the Snx14 APEX2 proteomics (Fig. 3D and SI Appendix, Fig. S2A). Notably, PLIN3 was not detected in the Snx14 co-IP lysate (SI Appendix, Fig. S5A).

Fig. 6.

Snx14 interacts with SCD1 in the ER network. (A) Western blotting with anti-Flag and anti-SCD1 antibody of 2% input lysate from GFP-Flag– and Snx14-Flag–expressing cells with and without palmitate treatment reveals relative expression of GFP-Flag, Snx14-Flag, and SCD1. Co-IP of SCD1 with Flag-tagged constructs reveals presence of SCD1 in Snx14-Flag– and not GFP-Flag–enriched beads when Western-blotted with anti-Flag and anti-SCD1 antibody. (B) Schematic diagram of Snx14 fragments C-terminally tagged with either 3× Flag or EGFP. Snx14^FL depicts the full-length human Snx14. Snx14^N is the N-terminal fragment that spans from the beginning and includes TM, PXA, and RGS domains. Snx14^PXCN is the C-terminal half including the PX domain and C-Nexin domains. (C) Lanes represent 2% input and IP from GFP-Flag, all 3× Flag-tagged Snx14 constructs (Snx14^FL, Snx14^N, Snx14^PXCN), and SCD1i-treated Snx14^FL-3×Flag–expressing U2OS cells. Western blotting with anti-Flag and anti-SCD1 antibody reveals relative expression of all of the Flag-tagged constructs and SCD1 in all these samples. (D) IF staining of U2OS cells expressing Snx14^FL, Snx14^N, and Snx14^PXCN with anti-EGFP (green) or anti-SCD1 (red) antibody and imaged with a confocal microscope. LDs were stained with MDH (blue). The cells were either untreated or treated with oleate. (Scale bar, 10 μm.)

Next, we dissected what regions of Snx14 were sufficient to co-IP SCD1. We used cell lines stably expressing the Flag-tagged N-terminal half of Snx14 encoding the transmembrane (TM), PXA, and RGS domains (Snx14^N) or a C-terminal half with the PX and C-Nexin domains (Snx14^PXCN; Fig. 6B). Co-IP experiments revealed that Snx14^N, but not Snx14^PXCN, was sufficient to pull down SCD1 (Fig. 6C). To test whether this Snx14:SCD1 co-IP required SCD1 desaturase activity, we conducted the co-IP with full-length Snx14 (Snx14^FL) in the presence of SCD1i (50). Inhibited SCD1 still co-IPed with Snx14^FL, indicating this interaction did not require SCD1 enzymatic activity (Fig. 6C).

Given Snx14 could co-IP endogenous SCD1, we tested whether this interaction occurred in intact cells. We exploited the previous observation that Snx14 enriches at ER–LD contacts following oleate addition (23). Given this, we queried whether Snx14 overexpression was sufficient to drive the accumulation of SCD1 at ER–LD interfaces, since it normally resides throughout the ER network. As expected, immunofluorescence (IF) labeling of endogenous SCD1 in non–oleate-treated cells transfected with Snx14-EGFP revealed SCD1 colocalization with Snx14 throughout the ER network (Fig. 6D and SI Appendix, Fig. S5B). Following oleate addition, Snx14-EGFP accumulated at ER–LD contacts and colocalized with SCD1 foci, which also enriched at these sites (Fig. 6D, red arrows, and SI Appendix, Fig. S5B). In contrast, expression of Snx14^PXCN-EGFP, which accumulates around LDs (23), failed to colocalize with SCD1 foci, consistent with Snx14^PXCN being insufficient to co-IP SCD1 (Fig. 6D and SI Appendix, Fig. S5B). Snx14^N-EGFP localized throughout the ER network following oleate, since it lacks the previously identified LD-targeting region in the CN domain (23), and this colocalized with SCD1 throughout the ER network but did not promote SCD1 foci at ER–LD contacts, consistent with co-IP results (Fig. 6D and SI Appendix, Fig. S5B). Collectively, this suggests that Snx14 co-IPs SCD1 from cell lysates and can interact with SCD1 in cells. It should be noted that the ER–LD SCD1 accumulations we observe require Snx14 overexpression; we do not detect SCD1 enrichment at ER–LD contacts under more ambient conditions.

Snx14 Loss Does Not Impact SCD1 Enzymatic Activity, but Snx14 Requires Its FA-Binding PXA Domain for Function.

To further dissect how Snx14 and SCD1 contribute to SFA metabolism at the ER, we examined whether Snx14 regulates SCD1 enzymatic activity. We directly assayed SCD1 activity in vitro via a well-established radiolabel-based desaturase assay (10). We pulsed microsomal fractions isolated from WT or SNX14^KO cells with (9,10-³H)-stearoyl-coenzyme A, an SCD1 substrate. SCD1 activity releases free ³H when stearoyl-CoA is monounsaturated, which can be directly detected by scintillation counting. As a positive control, we treated samples with SCD1i and detected a significant decrease in free ³H, indicating reduced SCD1 activity (SI Appendix, Fig. S6A)-. However, there was no significant change in relative SCD1 activity in SNX14^KO cells, suggesting Snx14 is not required for SCD1 activity.

Since Snx14 yeast ortholog Mdm1 had previously been shown to directly bind to FAs in vitro via its PXA domain (25), we next interrogated whether Snx14 requires its PXA domain for function. We generated cell lines stably expressing Snx14 lacking its PXA domain (Snx14^FLΔPXA; Fig. 7A) and queried whether this construct could co-IP with SCD1 as well as process the accumulated FFAs in SNX14^KO cells following palmitate exposure. Intriguingly, Snx14^FLΔPXA could co-IP with SCD1 (Fig. 7B). However, in contrast to SNX14^KO cells expressing full-length Snx14 (Snx14^FL), which have normal FFA levels, SNX14^KO-expressing Snx14^FLΔPXA cells exhibited elevated FFA levels similar to SNX14^KO cells. This indicates that the Snx14 PXA domain was required for FFA processing (Fig. 7C). Importantly, here also we confirmed that FA uptake was not altered by expressing Snx14^FL or Snx14^FLΔPXA in SNX14^KO cells, indicating that the elevated FFA accumulation in SNX14^KO cells expressing Snx14^FLΔPXA was not due to increased FA uptake (SI Appendix, Fig. S4E).

Fig. 7.

Snx14 loss does not impact SCD1 enzymatic activity, but Snx14 requires FA-binding PXA domain for function. (A) Schematic diagram of Snx14 fragments C-terminally tagged with 3× Flag. Snx14FL depicts the full-length human Snx14. Snx14FLΔPXA and Snx14FLΔTM are the full-length Snx14 excluding the PXA domain and TM domain, respectively. (B) Lanes represent 1% input and IP from GFP-Flag, 3× Flag-tagged Snx14 construct (Snx14FL, Snx14FLΔPXA, Snx14FLΔTM) expressing U2OS cells. Western blotting with anti-Flag and anti-SCD1 antibody reveals relative expression of all of the Flag-tagged constructs and SCD1 in all these samples. Snx14FLΔPXA could co-IP SCD1 similar to Snx14FL, whereas Snx14FLΔTM could not pull down SCD1. (C) Quantification of fold change in FFA (normalized to cell pellet weight) relative to untreated WT from TLC of whole-cell neutral lipids extracted from WT, SNX14^KO, and SNX14^KO cells expressing either EV, Snx14FL, or Snx14FLΔPXA, which are either untreated or treated with palmitate for 4 h. Values represent mean ± SEM (n = 3; *P < 0.01, multiple t test by Holm–Sidak method with alpha = 0.05). (D) Cell death (as a percentage) after exposure to 500 μM of palmitate in WT and SNX14^KO and on readdition of empty vector (EV), Snx14FL, Snx14N, Snx14PXCN, Snx14FLΔTM, or Snx14FLΔPXA to SNX14^KO. Values represent mean ± SEM. Significance test compared with WT (n = 3): *P < 0.01, multiple t test analysis by Holm–Sidak method with alpha = 0.05.

Next, we assayed whether the PXA domain was necessary to rescue SNX14^KO cell viability following palmitate exposure. Indeed, Snx14^FLΔPXA failed to rescue SNX14^KO cell viability, indicating the PXA was required for Snx14-mediated protection from palmitate (Fig. 7D). However, SNX14^KO cells expressing Snx14^N were protected from palmitate, suggesting this minimal Snx14 fragment, which contains an ER-anchored PXA domain, is sufficient for palmitate protection. However, an Snx14 construct which lacked the N-terminal TM domain but encoded all other domains (Snx14^FLΔTM) could not rescue the SNX14^KO cell viability, indicating that TM-mediated ER association is also required for Snx14 function (Fig. 7 A and D). Collectively, this suggests that Snx14 loss does not impact SCD1 enzymatic activity in vitro, but implies that the ER-associated Snx14 PXA domain may interact with FAs at or near the ER network in a manner that promotes SCD1-mediated FA processing.

Discussion

FAs are essential cellular components that act as energy substrates, membrane components, and key signaling molecules. These functions are dependent on the chemical features of FAs such as their chain length and saturation degree, which influence membrane fluidity and impact organelle structure and function (10, 46). Lipid homeostasis depends on FA processing and channeling to specific organelle destinations. When cells experience elevated intracellular FA levels, they respond by increasing FA processing and storage. Excess FAs are incorporated into neutral lipids and stored in LDs, which protect cells from lipotoxicity (51). Much of the machinery to achieve this resides at the ER, and proper ER–LD cross-talk by proteins such as seipin and the FATP1–DGAT2 complex is essential to maintain lipid homeostasis (17–19). Our earlier work revealed a role for Snx14 in ER–LD cross-talk, the loss of which contributes to the cerebellar ataxia disease SCAR20 for unknown reasons (23). Here, we report the proteomic composition of Snx14-associated ER–LD contacts and provide mechanistic insights into the function of Snx14 in FA metabolism. We find that Snx14 loss impacts the ability of cells to maintain proper lipid saturation profiles. In line with this, following SFA exposure, SNX14^KO cells display defects in ER morphology, FFA accumulation, and elevated SFA incorporation into membrane glycerophospholipids, lysophospholipids, and TG, which ultimately impacts cell homeostasis.

Utilizing APEX2-based proteomics analysis, we reveal the protein composition of Snx14-positive ER–LD contacts formed during oleate treatment. Notably, these contacts contain proteins associated with the LD surface such as PLIN2, PLIN3, and PNPLAP2. We also detect numerous proteins involved in FA processing (ACSL4, SCD1) and lipid/sterol biosynthesis (LPCAT1, LPCAT3, SOAT1), indicating that ER–LD contacts act as lipogenic ER subdomains for several lipid species, as well as hotspots for FA processing. These proteomics reveal many proteins previously identified through a similar APEX-based proteomics study that targeted APEX to the LD surface, providing consistency in the method, as well as underscoring the tight functional association between LDs and the ER network (44). Among these proteins, we focused on investigating the functional interplay between Snx14 and SCD1-associated SFA metabolism, which provided important insights in understanding the hypersensitivity of SNX14^KO cells to palmitate exposure. Although we cannot conclude that Snx14 and SCD1 directly interact, their phenotypical similarities and functional interplay provide important insights that may help in the therapeutic treatment of SCAR20.

Through lipidomic analysis and biochemistry, we show that Snx14 loss phenotypically mimics aspects of the enzymatic inhibition of Δ-9-FA desaturase SCD1, which catalyzes the oxidization of SFAs to MUFAs (Fig. 8) (47). In line with this, SCD1 overexpression can rescue palmitate sensitivity in Snx14-deficient cells. We also find that SCD1 and Snx14 can co-IP from cell lysates, and that Snx14 overexpression can promote SCD1 enrichment at ER–LD contacts under distinct metabolic conditions. The data presented are consistent with a model where Snx14 may promote FA processing to maintain ER homeostasis. One possibility is that Snx14 acts as an organizational scaffold for enzymes including SCD1 within the ER network, enabling FAs to be processed. In agreement with this, Snx14 requires its FA-binding PXA domain to function, implying that Snx14 could bind FA substrates and present them to SCD1 during periods of elevated FA influx. However, it should be noted that SNX14^KO cells do not display altered SCD1 enzymatic activity, indicating that Snx14 is not a direct enzymatic regulator of SCD1 per se, nor required for its activity. We also cannot rule out the possibility that Snx14 may interact with or present FAs to other enzymes in the ER network, which could explain other alterations in the lipidomics profile of SNX14^KO cells.

Fig. 8.

Working model for Snx14 in maintaining ER lipid homeostasis when treated with excessive SFAs.

Snx14 is highly conserved in evolution, and related studies of Snx14 orthologs provide mechanistic insights into Snx14’s role in lipid homeostasis. Drosophila ortholog Snz functionally interacts with SCD1 ortholog DESAT1 in the adipocyte cell periphery, a subcellular region rich in FA processing (26). Similarly, yeast ortholog Mdm1 functions in FA activation and LD biogenesis near the yeast vacuole (25, 52). As such, Mdm1-deficient yeast are sensitive to high dosages of exogenous FAs. Studies of both Snz and Mdm1 also reveal these proteins localize to specific subregions of the ER network, and thereby help to demarcate ER subdomains through their interorganelle tethering capabilities. Thus, an emerging possibility is that Snx14-family proteins act as organizational scaffolds within the ER network and recruit enzymes into ER subdomains to more efficiently process lipids.

Collectively, our observations provide a framework for understanding how Snx14 loss contributes to the cerebellar ataxia disease SCAR20. A growing number of neurological diseases are attributed to loss of proteins that function in ER-localized lipid metabolism or in the maintenance of ER architecture (27–29). Our data are consistent with a model where Snx14 loss perturbs the ability of cells to maintain FA metabolism and membrane lipid composition, ultimately elevating SFA levels and saturated fatty acyl chain incorporation in membrane lipids. Such alterations will ultimately effect membrane fluidity and organelle function and contribute to cellular lipotoxicity and the progressive death of cells like neurons, a pervasive symptom of SCAR20 disease (20, 22). Additionally, although we focused our analysis here on the interplay between Snx14 and SFA metabolism, the APEX2-based proteomics and lipidomics analysis will provide insights into understanding proteins and lipogenic reactions at ER–LD contacts.

Materials and Methods

See SI Appendix for Materials and Methods. There is discussion of cell biology and biochemical assays. See also for lipidomics details.

Data Availability.

All study data are included in the article and SI Appendix.