Targeting defective sphingosine kinase 1 in Niemann – Pick type C disease with an activator mitigates cholesterol accumulation

Niemann – Pick type C (NPC) disease is a lysosomal storage disorder arising from mutations in the cholesterol-trafficking protein NPC1 (95%) or NPC2 (5%). These mutations result in accumulation of low-density lipoprotein-derived cholesterol in late endosomes/lysosomes, disruption of endocytic trafficking, and stalled autophagic flux. Additionally, NPC disease results in sphingolipid accumulation, yet it is unique among the sphin-golipidoses because of the absence of mutations in the enzymes responsible for sphingolipid degradation. In this work, we examined the cause for sphingosine and sphingolipid

Niemann-Pick type C (NPC) disease is a lysosomal storage disorder arising from mutations in the cholesterol-trafficking protein NPC1 (95%) or NPC2 (5%). These mutations result in accumulation of low-density lipoprotein-derived cholesterol in late endosomes/lysosomes, disruption of endocytic trafficking, and stalled autophagic flux. Additionally, NPC disease results in sphingolipid accumulation, yet it is unique among the sphingolipidoses because of the absence of mutations in the enzymes responsible for sphingolipid degradation. In this work, we examined the cause for sphingosine and sphingolipid accumulation in multiple cellular models of NPC disease and observed that the activity of sphingosine kinase 1 (SphK1), one of the two isoenzymes that phosphorylate sphingoid bases, was markedly reduced in both NPC1 mutant and NPC1 knockout cells. Conversely, SphK1 inhibition with the isotype-specific inhibitor SK1-I in WT cells induced accumulation of cholesterol and reduced cholesterol esterification. Of note, a novel SphK1 activator (SK1-A) that we have characterized decreased sphingoid base and complex sphingolipid accumulation and ameliorated autophagic defects in both NPC1 mutant and NPC1 knockout cells. Remarkably, in these cells, SK1-A also reduced cholesterol accumulation and increased cholesterol ester formation. Our results indicate that a SphK1 activator rescues aberrant cholesterol and sphingolipid storage and trafficking in NPC1 mutant cells. These observations highlight a previously unknown link between SphK1 activity, NPC1, and cholesterol trafficking and metabolism.
NPC disease is caused by genetic mutations in the cholesterol transport proteins NPC1 and NPC2, with the vast majority of cases attributed to NPC1 mutations (95%) (1)(2)(3)(4). Whereas NPC pathology often first presents systemically with hepatospleno-megaly, morbidity and mortality are almost entirely attributed to the development of a complex progressive neurodegeneration (5). It has been suggested that NPC proteins act sequentially in late endosomes/lysosomes (LE/L), where the soluble NPC2 protein binds unesterified cholesterol and transfers it to the luminal sterol-binding site of membrane-associated NPC1, which then facilitates its transport out of the LE/L compartment (6,7). Aberrant functions of NPC proteins lead to excessive accumulation of unesterified cholesterol in the liver and spleen and also accumulation of sphingolipids, particularly in the brain (8). It was suggested that cholesterol accumulation can cause a secondary reduction in the activity of acidic sphingomyelinase that may explain accumulation of sphingomyelin (9). However, it is still not clear why sphingosine, the final breakdown product of all sphingolipids in the lysosome, also accumulates LE/L in NPC disease (10)(11)(12)(13)(14)(15). Intriguingly, sphingosine is the first detectable lipid that accumulates upon pharmacological inhibition of NPC1 (10) and is the only lipid that can induce an NPC phenotype when added exogenously at the levels observed in NPC cell models (11). Sphingolipid catabolism depends on egress of sphingosine from lysosomes and phosphorylation to sphingosine-1-phosphate (S1P), and then it can be irreversibly degraded by S1P lyase (16) or dephosphorylated to sphingosine that is reutilized for synthesis of ceramide and more complex sphingolipids in the ER (17). SphK1 is a cytosolic enzyme, and multiple growth factors, including vascular endothelial growth factor (VEGF), activate and translocate it to the plasma membrane, leading to increased S1P formation (18). This S1P in turn is exported from cells by transporters and signals by binding to S1PR1-5 to regulate many physiologic responses (19). It was previously suggested that loss of Purkinje neurons in NPC is caused by defective SphK activity because of reduced levels of VEGF and implicated SphK1 in the neuropathogenesis of NPC (20). Inhibition or deletion of SphK1 has also been shown to impair endocytic trafficking and accumulation of sphingosine in the LE/L, similar to the cellular phenotype of NPC (21,22).

This article contains supporting information.
Here, we showed that accumulation of sphingosine in several types of NPC1 mutant cells correlated with decreased SphK1 activity. Treating NPC1 mutant cells with a novel SphK1 activator that we identified reduced sphingosine accumulation, ameliorated autophagic defects, and restored the impaired cholesterol trafficking. Conversely, inhibition of SphK1 in WT cells with an isozyme-specific inhibitor increased accumulation of cholesterol. Our results highlight a link between lysosomal cholesterol accumulation and SphK1 activity and suggest that a SphK1 activator could be an attractive therapeutic option in NPC1 disease.

NPC1 MUTANT fibroblasts have defects in sphingoid base phosphorylation
In agreement with previous studies demonstrating that in addition to cholesterol, sphingosine also accumulates in NPC disease (10,11,15), we found that sphingosine levels were elevated in NPC1 mutant fibroblasts derived from well-characterized patients with a wide disease spectrum (NPC25 and NPC26) compared with control fibroblasts from a healthy individual (Fig. 1A). In addition, dihydrosphingosine, an intermediate in de novo sphingolipid biosynthesis, also accumulated in these mutant fibroblasts. However, despite accumulation of these sphingoid bases, S1P and dihydro-S1P levels were reduced (Fig. 1A). As this could result from either decreased sphingoid base phosphorylation by sphingosine kinases (SphKs) or increased degradation by S1P lyase, we next determined their enzymatic activities. Sphingosine kinase activity was assessed by short incubation of cells with C17-sphingosine, and levels of C17-S1P were quantified by liquid chromatography, electrospray ionization-tandem MS (LC-ESI-MS/MS).
Formation of C17-S1P was greatly reduced in the mutant fibroblasts compared with WT cells (Fig. 1B). Furthermore, an in vitro isoform-specific assay demonstrated that SphK1 enzymatic activity was significantly reduced in the mutant fibroblasts (Fig. 1C). In contrast, S1P lyase activity was not significantly different in the NPC1 mutant cells compared with WT ( Fig. 1D). Taken together, these results suggest that impaired SphK1 activity leads to increased levels of sphingoid bases.

Identification of a novel SphK1 isoform-specific activator
A previous study suggested that pathogenesis in NPC neurons resulted from defective SphK1 activity due to impaired VEGF levels and that correction of this activity by VEGF can reduce NPC pathological changes (20), suggesting that enhancing SphK1 activity could be a potential therapeutic intervention for this disorder. Therefore, it was important to identify a small molecule capable of increasing SphK1 activity. To this end, we used a high-throughput fluorescence assay to screen a chemical library with NBD-sphingosine as substrate that was used previously to identify a new inhibitor of SphK1 (23). In this screen, we discovered several compounds that increased SphK1 activity (data not shown; see Fig. 2A). However, only one compound, 2-hydroxy-3-{[4-(2-methoxyethoxy)benzoyl]oxy}-N-[2-(4-methoxyphenoxy)ethyl]propan-1-aminium, hereafter referred to as SK1-A (Fig. 2B), consistently stimulated SphK1 activity. In contrast, compound 3 in further assays with recombinant SphK1 did not increase its enzymatic activity. A more indepth kinetic analysis of the effects of SK1-A resulted in an estimated half-maximal activation constant (AC 50 ) of 4.5 mM for SphK1 (Fig. 2C), with a complete absence of any effects on SphK2 activity (Fig. 2D). Furthermore, with sphingosine as substrate, maximum activation of SphK1 was observed at a concentration as low as 0.5 mM (Fig. 2E). Moreover, SK1-A enhanced endogenous SphK activity in WT but not in SphK1 knockout mouse embryonic fibroblasts (Fig. S1). Thus, we next determined whether SK1-A could stimulate SphK1 activity in NPC1 mutant cells.

SK1-A increases sphingosine phosphorylation and reduces autophagic defects in NPC1 mutant fibroblasts
Next, cells were incubated with SK1-A, and cellular SphK activity was determined. As expected, C17-sphingosine accumulated in vehicle-treated NPC mutant fibroblasts, whereas C17-S1P formation was reduced compared with WT fibroblasts (Fig. 3A). Treatment of NPC mutant fibroblasts with SK1-A corrected this phosphorylation defect, reduced the high level of C17-sphingosine, and increased C17-S1P to a similar level observed in WT fibroblasts (Fig. 3A). Nevertheless, SK1-A treatment did not affect SphK1 protein expression in NPC1 mutant fibroblasts (data not shown).
Autophagy, an essential process that ensures that misfolded or dysfunctional proteins are degraded by lysosomes, is known to be impaired in NPC and contributes to disease progression (24,25). Because it was suggested that sphingosine accumulation due to impaired SphK activity causes defective endocytic and autophagic fluxes in NPC mutant cells (20), and we and others have shown that SphK1 is rapidly recruited to vesicles during endocytosis and autophagy to regulate these processes (21,22,26), it was of interest to determine the effects of SK1-A on autophagy. Consistent with previous results (24), in NPC1 mutant fibroblasts we found increased processing of the cytosolic soluble MAP1LC3/LC3 (microtubule-associated protein 1 light chain 3) form (LC3-I) to its lipidated membrane-bound form (LC3-II), a marker of increased formation of autophagosomes or amphisomes. This increased LC3-II level was reduced by treatment with SK1-A (Fig. 3, B and C). Likewise, accumulation of p62, an adapter protein that recruits polyubiquitinated and aggregated proteins to autophagosomes by its interaction with LC3-II (27), was significantly higher in NPC1 mutant fibroblasts compared with WT cells (Fig. 3, B and C), as was reported previously (24,28). The elevated levels of p62 in NPC1 mutant fibroblasts were also decreased by SK1-A (Fig. 3, B and C). Consistent with the accumulation of late endosomes in NPC1 mutant cells due impaired recruitment of components of the SNARE machinery to late endosomes (24,29), Rab7 and other late endosome markers, such as cation-independent mannose 6-phosphate receptor (M6PR), were all increased compared with WT, and treatment with SK1-A reduced their levels ( Fig. 3, B and C). Thus, increased phosphorylation of sphingosine and reduced accumulation of sphingosine correlates with improved autophagic and endocytic defects in NPC1 mutant cells.

Unexpected effects of SphK1 activation on cellular cholesterol trafficking
The accumulation of unesterified cholesterol inside the LE/L compartment is the hallmark cellular defect in NPC disease (1, 2, 5-7). Because autophagic defects can increase accumulation of cholesterol (24,30) and up-regulation of autophagy is beneficial in several neurodegenerative diseases (31), we sought to examine the effect of SK1-A on cholesterol accumulation in NPC1 mutant fibroblasts. Remarkably, treatment with 2 mM SK1-A significantly decreased the levels of unesterified (free) cholesterol in NPC25 mutant fibroblasts (Fig. 4A). The reduction in unesterified cholesterol could result from increased trafficking from the LE/L compartment or a result of reduction in uptake. Thus, cholesterol ester formation was examined, as this requires delivery of cholesterol to acyl-CoA:cholesterol acyltransferase at the ER (32-34). SK1-A treatment dramatically increased cholesterol ester formation in NPC1 mutant fibroblasts (Fig. 4B), suggesting that activation of SphK1 by SK1-A increased delivery of cholesterol to the ER. To confirm these results, cells were stained with filipin, a fluorescent polyene antibiotic that specifically binds unesterified cholesterol (35), to visualize cholesterol levels inside the cells. Vehicle-treated NPC25 and NPC26 fibroblasts showed higher cholesterol levels compared with WT2 cells, yet after SK1-A treatment, this accumulation was greatly reduced (Fig. 4, C and D). These results suggest that activation of SphK1 can reduce the pathogenic accumulation of cholesterol in these cells.

Inhibition of SphK1 increases cholesterol levels in NPC1 WT cells
We have previously shown that inhibition of SphK1 with the isoform-specific SphK1 inhibitor SK1-I induces a stall in the late endosomal compartment, producing dilated Rab7a-positive vesicles containing sphingosine (22). As restoring SphK1 activity reduced cholesterol in NPC1 mutant fibroblasts (Fig. 4), we sought to determine whether inhibition of SphK1 activity would have an inverse effect on cellular cholesterol in cells expressing WT NPC1 protein. Treatment of multiple cell types, including WT2 human fibroblasts, WT mouse embryonic fibroblasts (MEFs), and WT (JP17) CHO cells with SK1-I for 2 h significantly increased levels of free cholesterol with a decrease in cholesterol esters (Fig. 5, A and B). Similarly, increased filipin staining was observed after treatment of these cells with SK1-I (Fig. 5, C and D), supporting the notion that inhibition of SphK1 and sphingosine elevation reduced trafficking of cholesterol from lysosomes to the ER.

Activation of SphK1 by SK1-A reduces accumulation of sphingoid bases and complex sphingolipids in NPC1 null cells
NPC is a complex disease in which many mutations have been identified, and affected individuals display a broad range of disease severity and age of onset of symptoms, ranging from infant to adult (5). Therefore, we sought to examine the effects of SK1-A in the most severe type of NPC disease where NPC1 is completely absent.
SphK1 activity in isogenic CHO cells with homozygous deletion of the NPC1 gene (NPC1 KO ) was greatly reduced (Fig. 6A), similar to the defect observed in NPC1 MUTANT fibroblasts (Fig.  1C). Likewise, no major effects on S1P lyase activity were detected in NPC1 KO cells (Fig. 6A). Similar to NPC1 MUTANT fibroblasts, CHO NPC1 KO cells accumulated sphingoid bases sphingosine and dihydrosphingosine (Fig. 6B). As expected, rescuing the expression of NPC1 in knockin cells that stably express FLAG-tagged human NPC1 in an endogenous NPC1 knockout background (FLAG-NPC1 KI ) reversed accumulation of the sphingoid bases (Fig. 6B). Moreover, the reduced conversion of C17-sphingosine to C17-S1P (Fig. 6C) in NPC1 KO cells compared with WT CHO cells also supports the notion that this is the result of reduced SphK1 activity. Furthermore, treatment of NPC1 KO cells with the SphK1 activator SK1-A, which increased conversion of C17-sphingosine to C17-S1P (Fig. 6C), also reduced levels of sphingosine and dihydrosphingosine to almost the same levels observed in WT CHO cells (Fig. 6D).
Similarly, SK1-A significantly reduced accumulation of complex sphingolipids, including ceramides, monohexosylceramides, and sphingomyelins in NPC1 KO cells (Fig. 7, A-C). Interestingly, an examination of the individual species of these complex sphingolipids, defined by the number of carbons present in their acyl chains, indicates that with the exception of C16:0 ceramide, all other acyl chain species of the complex sphingolipids were reduced by SK1-A treatment (Fig. 7, A-C).

SK1-A corrects autophagic and cholesterol-trafficking defects in NPC1 null cells
Similar to NPC1 mutant fibroblasts (Fig. 3), NPC1 KO cells also have autophagic and endocytic defects, as shown by their  increased levels of LC3-I and LC3-II as well as Rab7 and M6PR compared with WT cells (Fig. 8, A and B). Activation of SphK1 in these cells with SK1-A ameliorates these defects (Fig. 8, A  and B), as assessed by a reduction in LC3-II, Rab7, and M6PR levels. Previous work suggested that autophagic flux is impaired in NPC mutant cells because of a loss in vesicular fusion between autophagosomes and late endosomes, inhibiting amphisome formation and cargo degradation (36,37). Autophagic flux can be directly monitored using expression of a tandem monomeric RFP-GFP-tagged LC3 reporter, which emits both green and red fluorescence when localized to autophagosomes yet only emits red fluorescence when localized to autolysosomes because of fluorescence quenching of the GFP protein in the acidic environment of lysosomes (38). Consistent with previous reports (36,37,39), the majority of accumulated vesicles in NPC1 KO CHO cells expressing RFP-GFP-tagged LC3 show increased yellow (red 1 green) fluorescence, indicating a block in autophagosome and lysosome fusion compared with WT cells (Fig. 8, C and D). Treatment with SK1-A resulted in the correction of impaired autophagosome maturation in NPC1 KO cells (Fig. 8, C and D), suggesting that SK1-A can restore autophagic flux. Importantly, treatment of NPC1 KO CHO cells with SK1-A decreased cellular unesterified (free) cholesterol levels (Fig. 9A) and increased cholesterol esters (Fig.  9B). In contrast, treatment with 100 nM S1P, a ligand for the S1PRs, had no effect on cholesterol accumulation in NPC1 KO CHO cells (Fig. S2). Furthermore, visualization of unesterified cholesterol by filipin staining confirmed that SK1-A treatment of NPC1 KO CHO cells decreased free cholesterol (Fig. 9, C and D). Taken together, these data indicate that activation of SphK1 with SK1-A restores cholesterol-trafficking defects in diverse NPC1 aberrant cells.

Discussion
Although the causative relationship between NPC mutations and cholesterol accumulation has been extensively studied, surprisingly little is known about the mechanism behind sphingosine and complex sphingolipid accumulation. Here, we report that the enzymatic activity of SphK1, one of the kinases that phosphorylate sphingosine to S1P, is reduced in diverse NPC1 mutant fibroblasts and in NPC1-deleted cells. As S1P cleavage by S1P lyase is the only exit point for complex sphingolipid degradation, we suggest that this decreased phosphorylation of sphingoid bases accounts for accumulation of sphingolipids in NPC. Consistent with our findings, a previous study showed that SphK activity is reduced in NPC mouse Purkinje neurons because of decreased levels of VEGF (20). Moreover, treatment with VEGF, which activates SphK1 by binding to VEGFR2, decreased sphingosine accumulation as well as improved neuronal survival and rota-rod scores in NPC1 knockout mice (20). Taken together, these results suggest that reversing defective SphK1 activity in NPC might be beneficial. In this study, we identified SK1-A as a potent and specific small-molecule activator of SphK1. SK1-A enhanced SphK activity and decreased accumulation of sphingoid bases in NPC1 mutant human fibroblasts and in NPC1 knockout cells. Previous studies have shown that autophagic flux is impaired in NPC1 disease (40)(41)(42) because of defective amphisome formation caused by failure of the SNARE machinery (24). Furthermore, inhibition of autophagy caused accumulation of cholesterol (24,30), suggesting that defective autophagic flux could be a causative factor in NPC1 disease. Methyl-b-cyclodextrin, which reduces lysosomal cholesterol accumulation in NPC disease restores impaired autophagy flux in NPC1-deficient cells through activation of AMP-activated protein kinase (43). We showed that activation of SphK1 with SK1-A suppressed defective autophagic flux and endocytic defects in NPC1 mutant and knockout cells. These results support recent findings that SphK1 regulates autophagic flux (44,45) and cooperates with autophagy to maintain endocytic membrane trafficking (21).
Increasing autophagy has been suggested as a potential strategy for several neurodegenerative disorders (31). Unfortunately, although efforts to enhance autophagy using mTOR-dependent (rapamycin) and mTOR-independent (lithium) induction corrected the autophagic defects in NPC1 mutant cells and increased cell viability, they did not correct the hallmark cholesterol-trafficking defects (24). Remarkably, however, we found that treatment with SK1-A not only reduced autophagic defects but also reversed cholesterol accumulation in NPC1 mutant and deleted cells.
It has been elegantly established that low-density lipoprotein-derived free cholesterol is transferred by the soluble NPC2 protein in the lysosome to the membrane-bound NPC1 for subsequent egress from this compartment (7,46,47). Perhaps the most surprising discovery of our work is that SphK1 activation alone is sufficient to reduce lysosomal accumulation of free cholesterol and restore production of cholesterol esters even in NPC1 null cells. Thus, cholesterol egress can be restored in an NPC1-independent manner. There are several potential mechanisms that might explain NPC1-independent SphK1 in NPC1 disease transport of cholesterol. We recently showed that expansion of membrane contact sites by overexpression of ORP1L-D-ORD mutant, which is known to act as a constitutively active tether between late endosomes and the ER without transporting cholesterol (48,49), rescued the lysosomal cholesterol accumulation in NPC1-deficient cells (50). It is possible that when membrane contact sites are expanded, another cholesterol transporter, such as STARD3 (51), ORP1L (48,49,(52)(53)(54)(55), or Gramd1b (50), could compensate for loss of NPC1. It is also possible that cholesterol could diffuse along a concentration gradient across the expanded membrane contact sites without the need for a cholesterol transporter (56). It is tempting to speculate that conversion of sphingosine to S1P could regulate formation of membrane contact sites between late endosomes/ lysosomes and the ER; however, other possibilities cannot be excluded. For example, it has been suggested that the ABC transporter A1 (ABCA1) has lysosomal cholesterol transport activity that requires functional NPC2 but not NPC1 (57). Furthermore, intracellular cholesterol trafficking can be enhanced by direct interaction of NPC2 with the phospholipid lysobisphosphatidic acid in the absence of NPC1 (58). It is still not known whether activation of SphK1 can increase ABCA1 transporter activity or levels of lysobisphosphatidic acid. Moreover, it was recently shown that reduction of the elevated sphingomyelin in NPC1 null cells decreased the cholesterol accumulation (59). Therefore, the dramatic decrease in sphingomyelin because of activation of SphK1 by SK1-A could be responsible for restoration of cholesterol trafficking via Rab9-dependent vesicular trafficking of cholesterol in NPC1-deficient cells (59). In addition, we cannot exclude the possibility that SK1-A also has effects on cholesterol metabolism and/or trafficking independent of its effects on SphK1.
Together these results demonstrate that impaired SphK1 activity in NPC cells contributes to accumulation of sphingosine and storage of complex sphingolipids and highlight a novel link between SphK1 activity, NPC1, and cholesterol trafficking and metabolism. Furthermore, the SphK1 activator SK1-A that we have characterized might lead to development of a new

Experimental procedures
Cell culture NPC1 MUTANT dermal fibroblasts (NPC25 and NPC26) and corresponding WT cells (WTA and WT2) were originally derived from NPC1 patients evaluated as part of an NICHD, National Institutes of Health, Institutional Review Boardapproved protocol and obtained with consent. The following mutations were present: NPC25 (fs-exon 20, c.2979dupA, and p.N701K); NP-26 (c.3742_3745delCTCA fs-exon24 and p. R1059Q). WT (JP-17 parental), NPC1 KO , and NPC1 KO CHO cells stably expressing a FLAG-tagged NPC1 protein (referred to as NPC1 KI ) described previously (60) were a generous gift from Dr. Judith White (University of Virginia). Mouse embryonic fibroblasts (MEFs) were purchased from ATCC. Human fibroblasts were cultured in MEM (Thermo Fisher Scientific, Waltham, MA, USA), CHO cells were cultured in MEM-a (Thermo Fisher Scientific), and MEFs were cultured in Dulbecco's modified Eagle's medium (Thermo Fisher Scientific), each supplemented with 10% fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA, USA) containing 100 units/ml penicillin and 100 mg/ml streptomycin (Thermo Fisher Scientific) as described (34).

Sphingolipid analyses
Cells were washed three times with ice-cold PBS, lipids were extracted, and sphingolipids were quantified by LC-ESI-MS/ MS (5500 QTrap; AB Sciex, Framingham, MA, USA) (61) and normalized to total protein.

Determination of SphK activity in cells
Cells were seeded on 6-well plates 24 h prior to a 30-min pulse with 1 mM C17-sphingosine (Avanti Polar Lipids). Cells were then washed three times with ice-cold PBS and then harvested directly by adding 1 ml of ice-cold methanol to each well. Sphingolipids were extracted, and C17-sphingosine and C17-S1P were quantified by LC-ESI-MS/MS and normalized to total lipid phosphates. SphK activity in mouse embryonic fibroblasts was determined with NBD-sphingosine as described previously (22).

Identification of the SphK1 activator SK1-A
Potential specific SphK1 inhibitors and activators were evaluated by high-throughput library screening fluorescence assays with recombinant SphK1 and SphK2 as described previously (23,62). Briefly, compounds were dissolved in DMSO and initially screened by measuring effects on rates of production of NBD-S1P from NBD-sphingosine (Avanti) at a concentration of 650 mM. Candidates showing inhibition or stimulation of SphK1 activity in this assay were further characterized with NBD-sphingosine at lower concentrations to obtain IC 50 values. SphK1 isoform specificity was determined using previously reported isoform-specific reaction conditions (63): SphK1 reaction buffer contained 30 mM Tris-HCl (pH 7.4), 0.5% Triton X-100 (MilliporeSigma), 150 mM NaCl, and 10% glycerol; SphK2 buffer contained 30 mM Tris-HCl (pH 7.4), 200 mM KCl, and 10% glycerol. Assays were carried out in 384-well plates (Greiner Bio-One, Frickenhausen, Germany) in the presence of recombinant SphKs (15 nM). Reactions were initiated by adding an ATP-Mg mixture (1 mM ATP, 2 mM MgCl 2 , 40 mM Tris-HCl, pH 7.4) and followed in a TECAN Infinite M1000 fluorescence plate reader (Männedorf, Switzerland) at 37°C. Excitation and emission wavelengths were 550 and 584 nm, respectively. All data were analyzed using Prism (GraphPad, La Jolla, CA, USA).

Measurements of cholesterol and cholesterol esters
Cholesterol and cholesterol esters were quantified using the Amplex Red cholesterol assay kit as described previously (34). Briefly, cells in 6-well plates were treated as indicated in the figure legends, washed three times with ice-cold PBS, and lysed. To determine cholesterol esters, lysates were incubated without or with 0.2 units/ml cholesterol esterase. Lysates were subsequently incubated with 2 units/ml each of horseradish peroxidase and cholesterol oxidase and 300 mM Amplex Red reagent for 30 min at 37°C in a 100 mM potassium phosphate buffer (pH 7.4) containing 50 mM NaCl, 5 mM cholic acid, and 0.1% Triton X-100. Fluorescence was measured with a TECAN Infinite M1000 fluorescent plate reader (TECAN, Männedorf, Switzerland). Levels of unesterified cholesterol and cholesterol esters were calculated from standard curves and normalized to total protein measured with the Bio-Rad protein assay kit.

Filipin staining
Cells were seeded on 22-mm coverslips in 6-well plates. After 24 h, cells were incubated with test compounds for 2 h. Cells were then washed and fixed in 4% paraformaldehyde for 15 min and washed. After incubating with 1.5 mg/ml glycine to quench the paraformaldehyde, cells were washed in PBS and stained with Filipin III (0.1 mg/ml; MiliporeSigma) for 2 h. Coverslips were then washed and mounted onto glass slides using Prolong Gold anti-fade reagent (Thermo Fisher Scientific) and imaged on an Axioimager A1 fluorescent microscope using Zen software (Zeiss, Jena, Germany). Images were quantified using ImageJ (National Institutes of Health) as described previously (34).

RFP-GFP-LC3 autophagic flux measurements
CHO cells were seeded on 96-well glass bottom plates (Cellvis, Mountain View, CA, USA) and infected with 80 viral particles (RFP-GFP-LC3; Life Technologies, Inc.) per cell for 48 h to express LC3 fused to a tandem green fluorescence-and red fluorescence-tagged protein. Autolysosomes are then displayed as red-only structures, as their acidic compartment causes the loss of the green fluorescent signal. After treatment without or with SK1-A, GFP-, and RFP-positive cells were visualized at 3100 in a Zeiss Cell Observer Spinning Disc (Zeiss, Germany) confocal microscope equipped with a growth chamber maintained at 37°C and 5% CO 2 . RFP and GFP fluorescence was collected with identical settings across wells, and GFP/RFP ratios were quantified using ImageJ (45).

Statistics
Data were analyzed by Student's t test when comparing two groups or by ANOVA with Fisher's LSD, Dunnett's, or Tukey's post hoc analysis as appropriate using Prism. Results are expressed as means 6 S.D. p ≤ 0.05 was considered statistically significant. Experiments were repeated three times with similar results.

Data availability
All data are contained within the article.