Regulation of mitochondrial ceramide distribution by members of the BCL-2 family.

Apoptosis is an intricately regulated cellular process that proceeds through different cell type- and signal-dependent pathways. In the mitochondrial apoptotic program, mitochondrial outer membrane permeabilization by BCL-2 proteins leads to the release of apoptogenic factors, caspase activation, and cell death. In addition to protein components of the mitochondrial apoptotic machinery, an interesting role for lipids and lipid metabolism in BCL-2 family-regulated apoptosis is also emerging. We used a comparative lipidomics approach to uncover alterations in lipid profile in the absence of the proapoptotic proteins BAX and BAK in mouse embryonic fibroblasts (MEFs). We detected over 1,000 ions in these experiments and found changes in an ion with an m/z of 534.49. Structural elucidation of this ion through tandem mass spectrometry revealed that this molecule is a ceramide with a 16-carbon N-acyl chain and sphingadiene backbone (d18:2/16:0 ceramide). Targeted LC/MS analysis revealed elevated levels of additional sphingadiene-containing ceramides (d18:2-Cers) in BAX, BAK-double knockout MEFs. Elevated d18:2-Cers are also found in immortalized baby mouse kidney epithelial cells lacking BAX and BAK. These results support the existence of a distinct biochemical pathway for regulating ceramides with different backbone structures and suggest that sphingadiene-containing ceramides may have functions that are distinct from the more common sphingosine-containing species.

7 cells) pellets were shaken for 30 s in glass vials (with PTFElined caps) with 2 ml:1 ml:1 ml chloroform-methanol-citric acid buffer (100 mM trisodium citrate, 1 M NaCl [pH 3.6]). The vials were vortexed for 15 s and centrifuged at 2,200 g at 4°C for 6 min. The organic (bottom) layer containing lipids was carefully taken up with a glass Pasteur pipet, transferred to a new glass vial, and concentrated under nitrogen. Lipids were reconstituted in chloroform and 1/4-1/8 of sample used for LC/MS.

Lipidomic analysis of WT, DKO, and single-knockout MEFs
Lipidomic analysis was performed with an Agilent 1200 Series HPLC online with an Agilent 6220 ESI-TOF (Agilent Technologies). Data were acquired in positive and negative ionization modes. For negative mode, a Gemini (Phenomenex) or Inspire (Dikma Technologies) C18 column (5 µm, 4.6 mm × 50 mm) was used with a guard column (C18, 2 µm frit, 2 mm × 20 mm). Solvent A was 95:5 water-methanol with 0.1% ammonium hydroxide, and solvent B was 60:35:5 isopropanol-methanol-water with 0.1% ammonium hydroxide. For positive mode, a Luna (Phenomenex) C5 or Bio-Bond (Dikma Technologies) C4 column (5 µm, 4.6 mm × 50 mm) was used with a guard column (C4, 2 µm frit, 2 mm × 20 mm). Solvent A was 95:5 water-methanol with 0.1% formic acid and 5 mM ammonium formate, and solvent B was 60:35:5 isopropanol-methanol-water with 0.1% formic acid and 5 mM ammonium formate. Identical gradient was used for both modes. The gradient was held at 0% B between 0 and 5 min, changed to 20% B at 5.1 min, increased linearly from 20% B to 100% B between 5.1 min and 45 min, held at 100% B between 45.1 min and 53 min, and returned to 0% B at 53.1 min and held at 0% B between 53.1 min and 60 min to allow column reequilibration. Flow rate was maintained at 0.1 ml/min between 0 and 5 min to counter the pressure increase caused by chloroform injection. The fl ow rates were 0.4 ml/min between 5.1 min and 45 min and 0.5 ml/min between 45.1 min and 60 min. Injection volume was 10-30 µl. Capillary, fragmentor, and skimmer voltages were 3.5 kV, 100 V, and 60 V, respectively. Drying gas temperature was 350°C, drying gas fl ow rate was 10 l/min, and nebulizer pressure was 45 psi. Data were collected in both profi le and centroid modes using a mass range of 100-1,500 Da.
Data were analyzed via targeted and untargeted approaches. Targeted analysis of known lipids was performed with manual integration in MassHunter Qualitative Analysis (Agilent Technologies) using a symmetric m/z expansion of either 50 or 100 ppm ( m/z expansions of 20, 50, and 100 ppm were compared and there were no signifi cant differences in peak integration). For untargeted analysis, raw data were converted to .mzXML format using trapper ( 27 ) and analyzed by XCMS ( 28 ) operated in R with the following parameters: family = "s", plottype = "m", bw = 10, and metlin = 0.15; retention correction (retcor) was iterated at least three times to maximize peak alignment across samples. XCMS output fi les were fi ltered by statistical signifi cance ( P р 0.05), fold change ( у 3), and reproducibility across four independent data sets, and the remaining ions were further verifi ed by manual integration in Qualitative Analysis. Database searches were performed in LIPID MAPS ( 29 ) and METLIN ( 30 ). A neutral mass of 535.50 was used to search the LIPID MAPS Structural Database ( 29 ) using the text/ontology-based search option with a mass expansion of ±0.01 and the METLIN ( 30 ) database using the Simple search option with a mass tolerance of ±20 ppm.
For construction of volcano plot ( see Fig. 2B ), ions between 0 and 5 min and 50 and 60 min were removed because these ions fall within the aqueous portions of the LC method and are unlikely to be lipids; the remaining ions between 5 and 50 min were examined in qualitative analysis. Isotopes were manually removed, and the resulting parent ions were graphed by their to the inner mitochondrial membrane, and cardiolipin peroxidation during apoptosis is a crucial step in releasing cyt c ( 18,19 ). Previous studies also demonstrate that sphingolipids cooperate with BAX and BAK in promoting membrane permeabilization (20)(21)(22)(23). The interplay between lipids and apoptosis led us to speculate that BAX and BAK might regulate mitochondrial lipids.
We reasoned that an LC/MS analysis of the BAX and BAK-regulated lipidome could provide an unbiased view into potential lipid metabolic pathways regulated by BAX and BAK. Our lipidomic analysis of WT and DKO mouse embryonic fi broblasts (MEFs) identifi ed distinct differences in mitochondrial lipid composition between DKO and WT cells, revealing a previously unknown function for BAX and BAK in cellular sphingolipid metabolism.

Tissue culture and harvest
MEFs were maintained at 37°C and 5% CO 2 in DMEM (11965, Life Technologies) with 10% FBS (HyClone or Seradigm), penicillin, streptomycin, nonessential amino acids, and L -glutamine (fi nal concentration, 6 mM). Immortalized baby mouse kidney epithelial cells (iBMKs) were maintained in the same medium with 5% FBS, penicillin, streptomycin, and L -glutamine (6 mM). All cells were passaged (1:4-1:6) at least three times and harvested at confl uency (48-52 h after last passage) by scraping into cold PBS on ice. Cell pellets for lipid extraction were used immediately for extraction or frozen at Ϫ 80°C until further use. Mitochondria isolation was performed immediately after cell harvesting without freezing.

Ceramide quantifi cation, coinjection, and tandem MS
A d 31 -d18:1/16:0 ceramide standard was used for quantifi cation of endogenous C16-C24 ceramides on the TOF instrument using general profi ling conditions. The standard was added to chloroform before lipid extraction to account for metabolite loss during extraction. Tandem MS in positive mode was performed with an Agilent 1200 Series HPLC online with an Agilent 6410 Triple Quad MS (Agilent Technologies) in Product Ion mode. Precursor ions were m/z 538.5 (d18:1/16:0 ceramide) and m/z 536.5 (d18:2/16:0 and d18:1/16:1 ceramides). Time fi lter width was 0.07 min. Fragmentor voltage, collision energy (CE), skimmer voltage, and ⌬ EMV were 132, 26, 15, and 400 V, respectively. Capillary voltage was 4.0 kV, skimmer voltage was 15 V, drying gas temperature was 350°C, drying gas fl ow rate was 8 l/min, and nebulizer pressure was 35 psi for all Triple Quad measurements. Injection volume was 30 µl .
Tandem MS in negative mode was performed with an Agilent 6500 Series Q-TOF LC/MS system. Precursor ions were m/z 534.4892 (d18:2/16:0 and d18:1/16:1 ceramides) and m/z 536.5048 (d18:1/16:0 ceramide). Injection volume was 15 µl. VCap, nozzle, skimmer1, and fragmentor voltages were 4,500, 2,000, 0, and 365 V, respectively. CE was 26 V. Octopole RF peak voltage was 750 V. Isolation width was set to Medium. Gas and sheath gas temperatures were 250°C; gas and sheath gas fl ow rates were 13 l/min and 12 l/min, respectively; and nebulizer pressure was 55 psi. Positiveand negative-mode tandem MS was recorded for mitochondrial and whole cell lipids from WT and DKO MEFs to verify the presence of the same ceramide species in all sample fractions.
The Triple Quad instrument operated in multiple reaction monitoring (MRM) mode was used for coinjection. MEF whole cell lipid extract was used as sample, synthetic ⌬ 4, 11-d18:2/16:0 ceramide was used as standard, and a combination of sample and standard used for coinjection. All samples were dissolved in chloroform such that the fi nal volume was 10 µl. The fragmentor and CE were the same as Product Ion. The transition in negative mode was m/z 534.5 → m/z 280.3, ⌬ EMV was 400 V, and MS1 resolution was set to Wide and MS2 resolution to Unit. LC parameters were identical to comparative lipidomics. d 12 -d18:2/16:0 ceramide was detected by dynamic MRM using the transition m/z 548.6 → 274.3 in positive mode. Retention time was 42.55 min, and the retention window was 10 min. All other parameters were identical to coinjection.

Comparative metabolite profi ling reveals elevated levels of unsaturated C16 ceramide in DKO mitochondria
Lipids were extracted ( 25,26 ) from WT and DKO MEF mitochondria ( 24 ) and subjected to LC/MS to identify lipidomic changes in the absence of BAX and BAK ( Fig.  2A ). Major lipid classes (free fatty acids, phospholipids, sphingolipids, acylglycerols, acylcarnitines, ether lipids, was added. The reaction was allowed to proceed overnight where it warmed from Ϫ 78°C to room temperature (rt) and the color dissipated until white. Saturated aqueous sodium bicarbonate was added, and the aqueous phase was extracted three times with diethyl ether (10 ml). The combined organic phases were washed with brine, dried over anhydrous sodium sulfate, concentrated, and purifi ed by silica gel chromatography (15:85 ethyl acetatehexanes, R f = 0.3) to yield 3 (36 mg, 32% yield) as a colorless oil.  ( Fig. 1A , 3 ).  ( Fig. 1A , 4 ).

Cellular production of d 12 -d18:2/16:0 ceramide
For each sample, 20 µl 83 mM d 14 -palmitoleic acid in DMSO was mixed with 10.5 ml MEF media, and 10 ml of the media introduced to DKO MEFs in a 10 cm tissue culture dish. 20 µl DMSO was used as negative control. Treated cells were incubated at 37°C for 2 h or 21 h before harvest by scraping into cold PBS on ice and lipid extraction. (i.e., sphingadiene-containing ceramide) or the N-acyl chain. We refer to ceramides with two double bonds as "diunsaturated ceramides" and ceramides with one double bond as "monounsaturated ceramides."

Diunsaturated ceramides are upregulated in DKO MEF mitochondria
To account for less abundant diunsaturated ceramides not identifi ed during untargeted XCMS analysis, monounsaturated and diunsaturated C16-C24 ceramide levels were measured by isotope-dilution MS ( 26,38 ). A d 31 -d18:1/16:0 ceramide standard was added to biological samples during lipid extraction, allowing ratiometric quantifi cation of endogenous ceramides. For each N-acyl chain, both the monounsaturated and diunsaturated species were considered.
Of all detectable ceramides ( Fig. 3A ), ceramides with C16 acyl chains occurred at the highest level ( ‫ف‬ 300-1,400 pmol/mg protein). The next highest level belonged to ceramides, with C24 acyl chains ( ‫ف‬ 150-600 pmol/mg protein) ( Fig. 3B ). A decrease in cellular concentration was observed for C18 and C20 ceramides ( Fig. 3B ), corresponding to an ‫ف‬ 100-fold decrease relative to C16 ceramides ( Fig. 3B ). Comparison of WT and DKO MEFs revealed a unique change in the ceramide profile. Levels of most monounsaturated ceramides (i.e., d18:1/16:0) were unaltered between WT and DKO MEF mitochondria. By contrast, a signifi cant increase ( ‫ف‬ 2-4.5-fold) in most diunsaturated ceramides was observed in the absence of BAX and BAK ( Fig. 3B ). This increase led to an ‫ف‬ 2-fold rise in the total diunsaturated-to-monounsaturated mitochondrial ceramide ratio ( Fig. 3C, D ). cardiolipins, cholesterol, and cholesterol esters) from four independent data sets were analyzed to ensure data quality and reproducibility. Levels of most lipids were unaltered between WT and DKO MEFs (supplementary Table 2), demonstrating that the majority of lipid pathways were unaffected by the absence of BAX and BAK.
Data analysis was also performed using the XCMS software package ( 28 ), which aligns, quantifi es, and statistically ranks changes in detected ions between WT and DKO groups. XCMS uncovered a metabolite with a detected m/z of 534.4886 in negative ionization mode that was consistently elevated in DKO mitochondria relative to WT ( Fig.  2B ). Our fi nding indicates that the loss of BAX and BAK can infl uence cellular metabolism, and the observation of one change in a background of mostly unaltered metabolites demonstrates specifi city in the metabolic pathway regulated by BAX and BAK.

BAX and BAK regulate d18:2-Cers in two different cell types
We next used a targeted MRM detection method to measure all detectable d18:2-and d18:1-ceramides (supplementary Table 1) in MEF and iBMKs. The absence of BAX and BAK raised the levels of C16-C24 d18:2-Cers relative to monounsaturated ceramides in both cell lines ( Fig. 6A , B ). Thus, BAX and BAK regulate d18:2-Cers in at least two distinct cell lines. BCL-2 regulation of these ceramides is also highly specific; d18:2-Cers were upregulated, whereas other diunsaturated ceramides (e.g., d18:1/16:1) were unchanged ( Fig. 6C ). Therefore, the location of the additional double bond in the backbone sphingosine is essential for ceramide regulation by BAX and BAK.

DISCUSSION
Several BCL-2 proteins affect cellular metabolism, including glucose and fatty acid metabolism (46)(47)(48), calcium homeostasis ( 49 ), and insulin secretion ( 50 ). Similar to these fi ndings, our data demonstrate a novel metabolic function for BAX and BAK in the regulation of endogenous ceramides. Specifi cally, we fi nd that BAX and BAK preferentially regulate diunsaturated, sphingadiene-containing ceramides (d18:2-Cers). The presence of the second double bond in the sphingoid We also monitored the levels of additional sphingolipids. No differences in additional sphingolipids or in palmitic or palmitoleic acids were observed between WT and DKO MEF mitochondria ( Fig. 4A , B ; supplementary Table  2), indicating that changes in ceramides are not the direct consequence of an alteration in the fatty acid composition or in other sphingolipids. Most ceramides were also unchanged between WT and BAX or BAK single-knockout (SKO) MEFs ( Fig. 4C, D ), suggesting that elevation of these diunsaturated ceramides required both BAX and BAK.
Identical fragmentation patterns were observed between the DKO-elevated ceramide and the synthetic d18:2/16:0 ceramide standard ( Fig. 5C ). For the initial synthesis, we installed the cis double bond in the d18:2/16:0 ceramide at C11-C12 ( Fig. 1A ). A double bond at this position would arise from potential incorporation of palmitoleyl ( ⌬ 9-cis -16:1)-CoA during de novo ceramide biosynthesis ( 34,42 ) (also see DISCUSSION). In a coelution study, this   backbone (i.e., d18:2-Cers) is essential for BAX and BAK regulation. Diunsaturated ceramides with the second double bond in the N-acyl chain were not regulated by BAX and BAK. Ceramide levels were also unaltered in BAX or BAK SKO MEFs, suggesting that BAX and BAK compensate for each other in this metabolic function. Whereas mitochondrial morphogenesis depends on both BAX and BAK ( 51 ), only BAK is required for longchain d18:1-Cers release in response to a range of apoptotic agents ( 21,22 ). Our observation that d18:2-Cers, but not d18:1-Cers, are regulated by both BAX and BAK supports the existence of subpopulations of structurally distinct ceramides that could be selectively modulated by different BCL-2 proteins. To our knowledge, this is the fi rst example of ceramide regulation based on backbone structure, and this result suggests that the backbone specifi city in ceramide biosynthetic enzymes requires exploration. This metabolic function of BAX and BAK is likely connected to their apoptotic function because of the proapoptotic activity of ceramides ( 15 ). Ceramides are bioactive lipids involved in multiple cellular processes, including apoptosis, proliferation, and autophagy ( 17 ). Cellular ceramide levels rise in response to a range of apoptotic signals ( 22 ), and ceramide macrodomains in the outer mitochondrial membrane act as sites of BAX activation during irradiation-induced apoptosis ( 52 ). Although numerous studies have demonstrated a proapoptotic role for ceramides, the apoptotic functions of ceramides also depend on lipid structure. Dihydroceramide inhibits ceramide channel formation in isolated mitochondria ( 53 ), and long-chain and very long-chain ceramides have been shown to exhibit opposite effects toward cancer cell survival ( 54 ). Our discovery of preferential regulation of d18:2-Cers by BAX and BAK suggests that there could also exist distinct cellular ceramide pools with different infl uences over cellular viability, which we will investigate in future studies.
Sphingadiene-containing sphingolipids are not new. Sphingadienes with a trans double bond at C4 and a cis double bond at C14 are a component of human serum sphingomyelins ( 43,44 ). Sphingadienes with 14 or 16 carbons and conjugated cis double bonds at C4, 6 are present in Drosophila ( 55 ), and cis -C4, 8-sphingadienes are found in plants ( 56 ). C16 sphingoid backbones are also present in human plasma due to the activity of an additional serine palmitoyltransferase isoform (SPTLC3) that favors myristoyl (C14:0)-CoA ( 35,45 ). Although there could be multiple biochemical pathways or dietary sources for generating mammalian sphingadiene-containing ceramides ( Fig. 7A ), we have examined the incorporation of unsaturated palmitoleic acid (d 14 -C16:1 fatty acid [FA]) into the sphingoid backbone. d 14 -16:1 FA-treated cells produced labeled ceramides to suggest that unsaturated FA incorporation could be one of the routes toward d18:2-Cers production ( Fig. 7B ). Additional questions, such as the known preference of serine palmitoyltransferases for palmitoyl-CoA over cis -unsaturated fatty acyl-CoA substrates ( 57,58 ) and how differently saturated cellular FA pools could be directed toward biosynthesis of sphingosineor sphingadiene-containing ceramides, will need to be further addressed.
The discovery of the DKO-elevated d18:2-Cers relied on tandem MS. Most research into the role of ceramides in apoptosis focuses on the levels and functions of d18:1-Cers ( 15 ). Our work indicates that it would be of interest to revisit these studies using tandem MS and targeted LC/ MS approaches to evaluate whether d18:2-Cers were overlooked. The authors thank Prof. Eileen White for the generous gift of WT and DKO iBMK cells; Dr. Sunia A. Trauger at the Harvard FAS Small Molecule Mass Spectrometry core for acquisition of the negative mode ceramide tandem MS spectra; Dr. Qian Chu and Dr. Christopher Vickers for manuscript reviews; and members of the Saghatelian and Walensky research groups for general technical guidance and discussions.