Sphingolipid distribution at mitochondria-associated membranes (MAMs) upon induction of apoptosis[S]

Journal of Lipid Research Volume 61, 202

It has been shown that ceramide induces cell death specifically when generated in mitochondria (9). Recently, a direct implication of sphingolipids in MOMP has been further highlighted by a study showing the cooperation of microsomal sphingolipids (namely S1P and its metabolite hexadecenal) with pro-apoptotic members of the BCL-2 family, respectively BAK and BAX, to promote cytochrome c release (10). These data suggest that nonmitochondrial membranes are necessary to provide an appropriate sphingolipid environment for the executive phase of apoptosis. Indeed, ceramide has been shown to be present in the so-called mitochondria-associated membranes (MAMs), an elusive structure that connects the ER to mitochondria (11,12). Primitively, MAM functions have been assigned to lipid transfer from the ER to mitochondria and to Ca 2+ exchange between the two organelles (13). The protein composition of MAMs appears to be tissue-specific with a corpus of proteins mainly involved in mitochondrial functions (14). Recently, new roles have been attributed to MAMs, including the control of autophagy/mitophagy, mitochondrial fusion/fission, and apoptosis (13). This structure has also been implicated in the basic mechanisms of neurodegenerative diseases and diabetes (15)(16)(17). However, the precise molecular mechanisms and actors implicated in MAM function in such an essential program as apoptosis remain poorly characterized. Because many MAM functions appear to be similar to those implicating sphingolipids (18), we analyzed the composition of the main subspecies of sphingolipids and related enzymes in vivo and in vitro under resting and apoptotic conditions. MS allows analysis and quantification with reproductive accuracy of a large number of lipid species and subspecies in a very small amount of biological samples of diverse origins (cell, cell medium, tissue, blood, etc.). In this work, we used these properties to attempt to define the distribution of a large number of sphingolipids and their different subspecies between ER, MAM, and mitochondria by LC-MS/MS.

Subcellular fractionation
Livers from C57BL6 mice (about 10 g) or U251 cells from 40 petri dishes (representing about 16 × 10 7 cells) were homogenized and subcellular fractionation prepared by differential centrifugation as previously described (19). Briefly, after homogenization, unbroken cells and nuclei were removed, and then an intermediate pellet was isolated from a fraction later centrifuged for ER and cytosol isolation and finally separated by ultracentrifugation (20,000 g for 30 min and then 100,000 g for 1 h). The crude mitochondrial fraction (MF) was obtained by centrifugation of the previous resuspended pellet at 10,000 g for 10 min. Crude MF was then layered on top of 30% Percoll medium and centrifuged at 95,000 g for 30 min. After this centrifugation, a dense band of pure mitochondria (PM) was observed at the bottom of the tube and a diffused white band of MAM at the top of the tube. These fractions were collected with a Pasteur pipette, diluted, and then pelleted respectively at 6,300 g for 10 min and 100,000 g for 1 h. The different fractions were diluted in different volumes as described in supplemental Fig. S1.

Microscopy
Transmission electron microscopy was done on cells fixed with 4% glutaraldehyde in PBS (pH 7.4) followed by a postfixation with 2% OsO 4 . After dehydration in a graded series of ethanol, adherent cells were embedded in epoxy resin, and thin sections (60 to 70 nm) were cut on a Reichert Ultracut E microtome and stained with uranyl acetate and lead citrate for observation at 80 KV under a JEM-1010 transmission electron microscope (JEOL).
For microscopic analyses, cells were grown on gelatin-coated coverslips. The cells were incubated with MitoTracker Red (Life Technologies) for 30 min at 37°C, washed two times with PBS, and then fixed in 4% paraformaldehyde for 30 min. The cells were washed with PBS and then mounted with Prolong antifade (Life Technologies) polymerizing solution and observed under a microscope with a 63× objective (Zeiss with apotome).

FACS analysis
Cells treated or not treated with STS were harvested after 2, 4, 6, or 8 h, permeabilized in PBS/0.002% digitonin, and fixed for 15 min with 2% paraformaldehyde. Cytosolic cytochrome c is lost during mild digitonin-induced membrane permeabilization, apoptotic cells thus appear negative for cytochrome c staining. Staining for cytochrome c (Clone 6H2.B4; BD Biosciences, San Jose, CA) and active Bax (Clone 6A7; BD Biosciences) was performed overnight at 4°C. The percent of positive cells was determined compared with isotype. For cell viability assessment, cells were harvested after STS treatment and the number of dead PI positive cells was determined after 5 min incubation with propidium iodide (Sigma).

Lipid purification
Targeted lipid purification was carried out using NH 2 SPE cartridges (100 mg). Total extract was deposited on cartridges preconditioned with 2 ml of hexane. Different classes of lipids were separated following the method described in (20). The four recovered fractions (F2, F4, F5, and F6) were dried down under nitrogen and removed in optimized volume of UPLC eluent A.

UPLC-ESI-MS/MS analysis
The sphingolipids were quantified by LC-ESI-MS/MS. The purified sphingolipids were analyzed on an Acquity H-Class UPLC system (Waters Corporation, Milford, MA) combined with a Waters Xevo TQD triple quadrupole mass spectrometer. All analyses were performed using ESI in the positive ion mode (ESI+) with multiple reaction monitoring. All sphingolipid classes were analyzed on a Waters C18 BEH column (2.1 × 50 mm) with 1.8 m particle size equipped with a 0.5 M prefilter. The column heater was set at 43°C for all compounds except ceramide-1-phosphate and sphingosine-1-phosphate where column temperature was set at 60°C. The autosampler temperature was maintained at 10°C. The mobile phases consisted of methanol (eluent A) or water (eluent B) containing 1% formic acid and 5 mM of ammonium formate. Purified sphingolipid fractions (sample and calibration) were eluted according to the elution step program.
Measurement and data analysis were collected by Mass Lynx software version 4.1 (Waters, Manchester, UK). Integration and quantification were performed using the TargetLinks™ software (Waters).
ASMase reaction (21) was initiated by the addition of 1 g protein sample in a volume of 100 l containing 73 pmol substrate in 12.5 mM sucrose buffer (pH 5.0) with 1% Triton X-100 and 0.1 mM of ZnCl 2 or 5 mM of EDTA activation buffer. EDTA and ZnCl 2 buffers were used respectively for lysosomal ASMase (L-ASMase) and total ASMase activity quantification.
NSMase activity was magnesium dependent (22). The reaction was initiated by 1 g of protein sample in a volume of 100 l containing 10 pmol of substrate in 200 mM of HEPES buffer (pH 7.4) with 20 mM of MgCl 2 .
The reaction was processed at 37°C over 24 h and stopped by adding chloroform:methanol (1:1, v/v). Samples were centrifuged at 1,000 g for 5 min. The organic phase was harvested, evaporated under N 2 , and resuspended in chloroform/ methanol (2:1, v/v).
Five microliters were spotted on silica gel 60 thin-layer chromatography plates (Merck-Millipore; 105626). Ceramide and uncleaved SM were separated using a mixture of chloroform:methanol (95:5, v/v) as a solvent and were quantified on a ChemiDoc MP imaging system (Bio-Rad) with the ImageLab software (Bio-Rad). Enzymatic activity is presented as the hydrolysis rate of SM (picomoles) per time (hours) and per protein concentration (micrograms).

Statistical analysis and animal studies
Two-way ANOVA and, as posttest, Tukey's multiple comparison test were used in this study. Research involving animals were performed in agreement with the French National Ethics Committee. Investigations with animals were approved by the local institutional review committee of the animal house facilities.

ER, MAM, and mitochondria sphingolipid composition in mouse liver
We first analyzed the mitochondrial and ER proteins present in the different submembrane fractions from mouse liver by immunoblots (see the Materials and Methods). As illustrated in Fig. 1A, we found a repartition of these different proteins from ER to MAM to MF similar to that described in previous studies: CR (23) and the longchain fatty acid-CoA ligase 4 (FACL-4) (24) were found in ER and MAM, while VDAC1 and cytochrome c were predominantly found in mitochondria. Similar but not identical results were found in mouse brain (supplemental Fig. S2).
Next, we analyzed and quantified the sphingolipid content and, in particular, precursors and metabolites of ceramides in the different liver fractions by LC-ESI-MS/ MS. Ceramides and SM were present mostly in MAM and the ER (Fig. 1B). On the other hand, sphinganine was essentially present in MAM and PM fractions, while sphingosine was found mostly in PM (Fig. 1C). Statistical data for lipid measurements are shown in supplemental Fig. S3.
Glycosylated forms of ceramides are biologically active and versatile derivatives of ceramides. These glucosyl-and lactosyl-ceramides are thought be synthesized in the late ER compartments. although there is some evidence that they are present in MAM (11). We confirmed the latter observation and found that, in mouse liver, glucosyl-ceramide is essentially present in MAM, while lactosyl-ceramide is located both in PM and MAM (Fig. 1D).
Analysis by the chain lengths of the sphingolipids shows some differences between the different fractions: MAM and PM fractions were enriched in C18 compared with ER, while ER and PM fractions were enriched in C24 and C24:1 compared with MAM (Table 1). However, major and minor chain lengths in the different fractions appear to be roughly constant and no chain length is specific from any organelle (except from C18 ceramide, which is the only ceramide found in cytosol) ( Table 1).
Our data support previous observations on the presence of sphingolipids in MAM. However, sphingolipids present in MAM are quantitatively and qualitatively different from those found in ER and mitochondria.

Sphingolipid composition of ER, MAM, and U251 in U251 cells: effect of the induction of apoptosis
Sphingolipids have been implicated in apoptosis, but little is known about their intracellular distribution between ER, MAM, and mitochondria during the induction of this cell death program (8). Induction of apoptosis in liver is a multimodal and complex phenomenon that involves different processes. In order to obtain a more homogeneous population, we investigated the sphingolipid composition in resting and apoptotic human glioma U251 cells. This cell line has been extensively studied as a model for apoptosis (25). As illustrated in Fig. 2A, transmission electron microscopy studies reveal that U251 cells exhibited ER/ mitochondria contacts that resemble classical MAM morphology. Resting U251 cells were subjected to subcellular fractionation as described in the Materials and Methods (19). We first analyzed by immunoblots using various markers of mitochondria and ER proteins specific for the different fractions. As illustrated in Fig. 2B, we found a classical   repartition of these different proteins from ER to MAM and MF: CR and RTN3 (14) were found in ER and MAM, while VDAC, TOM20, and cytochrome c were predominantly found in mitochondria; GM130 was slightly detected in the ER fraction, suggesting a mild contamination with Golgi membranes. We examined the sphingolipid composition of the different fractions and found that, at variance with mouse liver, ceramide and SM were partitioned between PM, MAM, and ER, and sphingosines were localized in MAM and PM while sphinganine was essentially mitochondrial (Fig. 2C). Of note, the ratio of glycosylated ceramides in the different subcellular compartments was inverted compared with mouse liver (Fig. 2C).
Next, we monitored the kinetics of the induction of apoptosis in STS-treated cells by analyzing the disruption of the mitochondrial network. As shown in Fig. 3A, the mitochondrial network in STS-treated cells was maintained at 2 h but started to be disrupted at 4 h. The main cellular events leading to apoptotic cell death in STS-treated U251 cells can be defined by first a change in the conformation of the pro-apoptotic protein Bax [monitored by the monoclonal 6A7 (26)] and the release of cytochrome c (released cytochrome c is lost during washing steps in the permeabilized cells; apoptotic cells are thus cytochrome c negative) as a demonstration of the apoptotic MOMP (Fig. 3B, left and middle graphs). The final step of cell death can be quantified by the induction of nonselective plasma membrane permeability visualized by propidium iodide staining (Fig. 3B, right graph). Based on the different results, we define a chronology of apoptotic cell death in U251 cells with early apoptotic events taking place before 2 h of STS treatment. Subcellular fractionation indicates that during the early mitochondrial-dependent apoptotic events, there are no major changes in protein distribution between ER/ MAM/PM (Fig. 3C).
Next, we examined the sphingolipid contents 2 and 4 h after the induction of apoptosis. No change in the total cellular concentrations of glucosyl-and lactosyl-ceramides is observed under these conditions until 4 h (Fig. 4A). However, a significant decrease in sphingosine and SM concentrations is observed in U251 cells after 2 h of STS treatment, while total cellular sphinganine content decreases only at 4 h. An increase in ceramide is observed at 4 h (Fig. 4A), which corresponds to a significant rise in C16 ceramide (Fig. 4B), a form shown to be associated with apoptosis (27). Thus, a modification in the expression of sphingolipids, in particular ceramide C16, closely accompanies the first steps of mitochondrial apoptosis, as previously shown for UV-induced apoptosis (28).

Subcellular distribution of sphingolipids and related enzymes in apoptotic U251 cells
To further substantiate this assertion, we analyzed the distribution of sphingolipids in the different organelles of U251 cells at 2 and 4 h after the induction of apoptosis by STS (i.e., maximum Bax activation analyzed by exposure of the 6A7 epitope and the onset of cytochrome c release from mitochondria, see Fig. 3). We observe an important increase in ceramide content in PM 4 h after STS treatment (Fig. 5A). In contrast, the content in SM is specifically Fig. 3. Induction of apoptosis by STS-treatment in U251 cells. A: Impact of STS-induced apoptosis on mitochondrial network. Polarized mitochondria are stained by Mitotracker (red) and nucleus with Dapi (blue) (scale bar represents 20 µm). B: Cells were treated by STS and stained for 6A7 active Bax (left) or cytochrome c (cyt c) (middle) as described in the Materials and Methods, or unfixed cells were stained with propidium iodide (PI) (right) and analyzed by flow cytometry (*P < 0.05, ***P < 0.0001 vs. T0). C: Impact of STS-induced apoptosis on the subcellular distribution of the fraction markers (CR, TOM20, and cytochrome c). decreased in MAM (Fig. 5B). The sphingosine contents are not affected by STS treatment (Fig. 5C). Quite surprisingly, we observe a similar increase in sphinganine in mitochondria (Fig. 5D), although the global cellular concentration is decreased (Fig. 4A). On the other hand, no change in the distribution of sphingosine and glucosyl-and lactosylceramides is observed (Fig. 5 E, F). Our results show that the distribution of sphingolipids in the MAM and mitochondria is regulated differently from total cellular contents.  Because of the putative importance of the modulation of SM and ceramide in MAM/PM during apoptosis, we analyzed the distribution of ASMase activity [both L-ASMase and secreted ASMase (S-ASMase)] in the different organelles in U251 cells 4 h after the induction of apoptosis by STS. We observe a repeated increase in ASMase activity in MAM but not in PM (Fig. 6A-C, supplemental Fig. S4A). The increase in ASMase activity is statistically significant in each individual experiment but not significant when all results are compiled (supplemental Fig. S4A) due to interexperiment variability. ASMase activity is separated into L-ASMase and S-ASMase; both L-and S-ASMase activities are increased during the early induction of apoptosis in the MAM fraction (Fig. 6B, C). As for total ASMase activity, the increase in L-and S-ASMase activity in the MAM fraction is statistically significant in each experiment but not when all data are analyzed due to important inter-experiment variation (supplemental Fig. S4A). Of note, NSMase activity is not affected during the time of treatment, in any compartment ( Fig. 6D and supplemental Fig. S4A). Given the suggested association of lysosomes to the mitochondrial membranes in mitochondrial fusion sites (29), we next questioned the presence of lysosomal membranes associated to mitochondria and MAM. However, while the presence of the lysosomal protein LAMP2 is detected in all the subcellular fractions isolated (Fig. 6E), the ASMase activity is not altered in the MF and PM fraction. Of note, the microscopic observation of U251 MF reveals the regular proximity between mitochondria, ER, and lysosomes (supplemental Fig. S4B). The amount of lysosomes copurified with the MF is not increased by STS (Fig. 6F, supplemental Fig. S4C), suggesting that the increase in ASMase activity measured in MAM is not due to the recruitment of lysosomes. Of note, confocal imaging reveals that ASMase is sometimes found in the proximity to mitochondria (supplemental Fig. S5), even if no significant increase in the colocalization between ASMase and the mitochondrial marker TOM20 is observed in apoptotic cells.

DISCUSSION
Apoptosis can be initiated in many different ways, but it is generally admitted that most signals are converted into a common mechanism at the mitochondrial level, which in turn triggers the execution phase of the cell death program. Sphingolipids play a major role in the regulation of cell death and trigger different types of apoptosis depending on its intracellular localization from plasma membrane to mitochondrial outer membrane. Several studies have underlined the importance of lipids in the BCL-2 family functions (8,10,30,31). Elusive structures called MAMs play a major role in the coupling between ER apoptotic signals and mitochondria mainly by supporting the ER-mitochondria transport and synthesis of lipids and calcium exchange between these two organelles (18,32). Compartmentalization of sphingolipid-metabolizing enzymes into different subcellular locations has been extensively described (1). De novo ceramide synthesis takes place in the ER where serine and palmitoyl-CoA are condensed into sphinganine (2). Ceramide is synthesized from sphinganine by ceramide synthases located at various membranes (ER, nuclear membrane, mitochondria, MAM) (33). Sphingosine can be recycled via the salvage pathway to regenerate ceramide. Ceramide generation can also occur via SM hydrolysis through the activity of sphingomyelinase. Based on the specific intracellular localization of sphingolipid enzymes and their products, sphingolipid functions have been largely restricted to the site of generation but may also be involved in specific functions in other subcellular compartments. For instance, little is known about the MAM-specific roles and distribution, if any, in sphingolipid metabolism. In this work, we used LC-MS/MS to define the distribution of a large number of sphingolipids and their different subspecies between ER, MAM, and mitochondria in human glioma U251 cells at the basal level and during the apoptotic process. We specifically analyzed the sphingolipids specific to the three pathways capable of generating ceramide: sphinganine for the de novo pathway, SM for the breakdown pathway, and sphingosine for the salvage pathway (2). The addition of sugar to ceramide to form complex sphingolipids was analyzed by measuring lactosyland glucosyl-ceramides. No significant difference was measured in the composition of the different species regarding the chain lengths during apoptosis induction. The glycosylated forms of ceramide were not altered during STS treatment (Figs. 4A; 5E, F) either at the cellular level or in the fractions. At the cellular level, ceramide (especially C16 ceramide) increased after 4 h concomitant to a decrease in sphinganine (Fig. 4A, B). These events were preceded by a decrease in the cell SM and sphingosine levels (Fig. 4A). These results suggest that the increase in ceramide induced by apoptosis induction entails several processes: first, SM hydrolysis and activation of the salvage pathway (sphingosine recycling), and then de novo synthesis of ceramide from sphinganine by ceramide synthases. At the organelle level (Fig. 5), the ceramide increase is mainly measured in mitochondria, which is consistent with the apoptotic function of ceramide described in the literature (9). We measured no concurrent SM decrease in mitochondria; however, we measured it in MAM (Fig. 5B), suggesting that SM hydrolysis occurs in MAM rather than in mitochondria, followed by ceramide transfer to mitochondria. This is consistent with the sphingomyelinase activity increase we measured in MAM but not in mitochondria ( Fig. 6A-D, supplemental Fig. S4A). With reference to de novo synthesis, Fig. 5D shows that sphinganine concentration is mostly raised in mitochondria. Serine palmitoyltransferase is known as an ER-located enzyme (33). Sphinganine might thus be produced in the ER and transit to the mitochondria through the MAM to then be modified by mitochondrial ceramide synthases. Sphingosine decrease measured in total cells (Fig. 4A) is not recovered at the organelle level (Fig. 5C). This suggests that the recycling of sphingosine to ceramide happens in a compartment that we did not purify in our isolation protocol. Of note, our protocol is not intended for the purification of plasma membrane, nucleus, or late endosomes/lysosomes. Sphingosine-to-ceramide conversion is reported in the lysosomal fraction, which could account for the results we observe. Thus, a quantitative and qualitative analysis of sphingolipids in the different subcellular compartments showed a different composition in the different organelles. We measured a specific activation of ASMase (L-and S-forms) in MAM during apoptosis, which could explain the decrease of SM in MAM and the concomitant increase of ceramide measured in mitochondria (Fig. 5A). ASMase activity was not typically described in MAM. We observed the copurification of lysosomes with MAM and also with other subcellular fractions (Fig. 6E). Of note, lysosome-mitochondria contacts were observed at mitochondrial fission sites (29), as were ER-mitochondria contacts. It is conceivable that tripartite ER-mitochondria-lysosomes contacts exist in cells, underlying the measurement of ASMase activity in some but not all mitochondria. The fact that we measured the sphingosine decrease in the total cells but not in the fractions isolated nevertheless suggests that not all lysosomes connect to mitochondria/MAM; specific enzymatic activities might be segregated in lysosomes depending on their localization in the cell and their interaction with other organelles. It seems that lysosomes were not massively recruited to the MF during apoptosis (Fig. 6F, supplemental Fig. S4C); the gain in ASMase activity measured in MAM may thus be linked to an activation induced by oxygen species produced in stressed mitochondria (34).
Based on our results, we can provide a subcellular timeline on ceramide and related metabolites during apoptosis. Ceramide production and especially C16-ceramide increase at the early stages of apoptosis through a possible conversion from sphingosine and SM first, then through de novo synthesis from sphinganine, all of these events happening upstream from the mitochondrial permeabilization and caspase activation (Fig. 3), while lactosyl-and glucosylceramide are not affected. Sphinganine has been shown to be involved itself in apoptosis through the disruption of the mitochondrial respiratory chain and inhibition of kinase (35,36). The regulation of contacts between organelles, including MAM, might thus be a key determinant in the regulation of sphingolipid synthesis and thereby participate to the early stage of apoptosis.