Receptor-interacting Ser/Thr kinase 1 (RIPK1) and myosin IIA–dependent ceramidosomes form membrane pores that mediate blebbing and necroptosis

Formation of membrane pores/channels regulates various cellular processes, such as necroptosis or stem cell niche signaling. However, the roles of membrane lipids in the formation of pores and their biological functions are largely unknown. Here, using the cellular stress model evoked by the sphingolipid analog drug FTY720, we show that formation of ceramide-enriched membrane pores, referred to here as ceramidosomes, is initiated by a receptor-interacting Ser/Thr kinase 1 (RIPK1)–ceramide complex transported to the plasma membrane by nonmuscle myosin IIA–dependent trafficking in human lung cancer cells. Molecular modeling/simulation coupled with site-directed mutagenesis revealed that Asp147 or Asn169 of RIPK1 are key for ceramide binding and that Arg258 or Leu293 residues are involved in the myosin IIA interaction, leading to ceramidosome formation and necroptosis. Moreover, generation of ceramidosomes independently of any external drug/stress stimuli was also detected in the plasma membrane of germ line stem cells in ovaries during the early stages of oogenesis in Drosophila melanogaster. Inhibition of ceramidosome formation via myosin IIA silencing limited germ line stem cell signaling and abrogated oogenesis. In conclusion, our findings indicate that the RIPK1–ceramide complex forms large membrane pores we named ceramidosomes. They further suggest that, in addition to their roles in stress-mediated necroptosis, these ceramide-enriched pores also regulate membrane integrity and signaling and might also play a role in D. melanogaster ovary development.

Formation of membrane pores/channels regulates various cellular processes, such as necroptosis or stem cell niche signaling. However, the roles of membrane lipids in the formation of pores and their biological functions are largely unknown. Here, using the cellular stress model evoked by the sphingolipid analog drug FTY720, we show that formation of ceramide-enriched membrane pores, referred to here as ceramidosomes, is initiated by a receptor-interacting Ser/Thr kinase 1 (RIPK1)-ceramide complex transported to the plasma membrane by nonmuscle myosin IIA-dependent trafficking in human lung cancer cells. Molecular modeling/simulation coupled with site-directed mutagenesis revealed that Asp 147 or Asn 169 of RIPK1 are key for ceramide binding and that Arg 258 or Leu 293 residues are involved in the myosin IIA interaction, leading to ceramidosome formation and necroptosis. Moreover, generation of ceramidosomes independently of any external drug/stress stimuli was also detected in the plasma membrane of germ line stem cells in ovaries during the early stages of oogenesis in Drosophila melanogaster. Inhibition of ceramidosome formation via myosin IIA silencing limited germ line stem cell signaling and abrogated oogenesis. In conclusion, our findings indicate that the RIPK1-ceramide complex forms large membrane pores we named ceramidosomes. They further suggest that, in addition to their roles in stressmediated necroptosis, these ceramide-enriched pores also regulate membrane integrity and signaling and might also play a role in D. melanogaster ovary development.
Formation of nonselective pores within the plasma membrane in response to cellular stress results in disruption of the integrity of the lipid bilayer, increasing osmotic pressure and leading to necrosis (1,2). One of the pore-forming processes is regulated by the receptor-interacting protein 3 (RIPK3)mixed-lineage kinase domain-like (MLKL) complex (3,4), resulting in regulated necrosis, necroptosis, which also involves activation of RIPK1 (5-7). Although protein-dependent membrane pore formation has been demonstrated to play various biological roles in the plasma membrane (8), mitochondrial membrane (9), and nuclear membrane (10), the functions of lipid-protein complexes in membrane pore formation and their biological functions have been largely unknown. FTY720 (fingolimod or Gilenya, Novartis) is a sphingosine analog drug (11,12) that is approved for the treatment of refractory multiple sclerosis (13)(14)(15). FTY720 is mainly phosphorylated by sphingosine kinase 2 (SPHK2) (16), and phosphorylated FTY720 (P-FTY720) exhibits a functional antagonist action against sphingosine 1-phosphate receptor 1 (S1PR1) (17)(18)(19). P-FTY720 binding leads to internalization and degradation of S1PR1, inhibiting its G protein-coupled receptor signaling, required for the egress of lymphocytes from lymphoid organs to the bloodstream, resulting in immune suppression (20,21). In addition to its immune-suppressive functions, FTY720 also exhibits anti-cancer functions in various models, such as reducing colitis-mediated colon cancer in SphK2-deficient mice (22), inhibiting leukemic stem cell expansion (23), and attenuating xenograft-derived lung cancer growth in mouse models (24). In lung cancer, our previous data revealed that FTY720, without the need for phosphorylation, induces necroptosis via activation of tumor suppressor protein phosphatase 2A (PP2A) and RIPK1 (25). We found that, although FTY720-induced necroptosis depended on the kinase domain of RIPK1, it was RIPK3/MLKL-independent, suggesting a novel form of necroptosis (25) that here is referred to as type II necroptosis. However, the mechanisms of type II necroptosis induction, RIPK1-dependent and RIPK3/MLKL-independent, by FTY720 remain unknown.
Interestingly, RIPK1 deficiency in hematopoietic cells results in cell death, leading to bone marrow failure, suggesting a role of RIPK1 in the survival of hematopoietic stem cells and progenitor cells (38). However, the precise roles of RIPK1 and/or necroptosis in stem cell development and/or signaling remain largely unknown. In Drosophila melanogaster, ovary development (oogenesis) begins with the formation of cysts of interconnected germ cells, produced by germ line stem cells, enveloped by somatic follicle cells (39). Oogenesis in insect cells requires a complex mechanism for differentiation of germ line and somatic cells (40). The Drosophila ovary contains three types of stem cell populations, germ line stem cells (GSCs), somatic stem cells, and escort stem cells, which are contained in a well-defined anatomical structure called the germarium (41)(42)(43). Thus, insect ovaries present an ideal model to study the roles of various mechanisms in the regulation of stem cell development and signaling.
Given the importance of ceramide in inducing cell death and necroptosis, here we sought to investigate whether ceramide generation and/or trafficking plays any roles in FTY720-mediated necroptosis and/or stem cell development/signaling. Our data suggest that FTY720-mediated necroptosis depends on the formation of nonselective membrane pores, regulated by ceramide-RIPK1 complexes, that are transported to the plasma membrane by nonmuscle myosin IIA (NMIIA) through endocytic vesicular transport. These data also reveal a novel endogenous role for the formation of ceramide-RIPK1-NMIIA-enriched membrane pores in germ line stem cells in D. melanogaster during oogenesis without any external stress stimuli or exposure. These data suggest that. in addition to their roles in stress-mediated necroptosis, ceramide-enriched pores might play important bio-logical roles, at least during D. melanogaster ovary development, regulating membrane integrity and signaling.

FTY720 induces the formation of ceramide-enriched membrane pores, ceramidosomes, at the plasma membrane
To determine whether FTY720-mediated necroptosis is associated with plasma membrane alterations, we carried out live-cell imaging, transmission EM (TEM), and scanning EM (SE) studies using GFP-expressing A549 cells in the absence/ presence of FTY720 at various time points (0 -240 min). Data showed that FTY720 induced membrane blebbing, followed by plasma membrane rupture and necroptosis in a time-dependent manner (Fig. 1, A and B, and Fig. S1A). Plasma membrane blebbing and rupture in response to FTY720 (10 M) was also observed in H1341 lung cancer cells (Fig. S1B). Interestingly, FTY720-mediated blebbing and plasma membrane rupture were largely abrogated when the cells were pretreated with blebbistatin (44), a known inhibitor of NMIIA-mediated blebbing, in A549 cells, as detected using live-cell imaging, TEM, and SE ( Fig. 1, C-G). Moreover, alterations of plasma membrane integrity and induction of necroptotic cell death in response to FTY720 were also evident when the YOPRO-1 (45) and propidium iodine (PI) dyes were detected in H1341 cells using flow cytometry and by trypan blue exclusion assays (Fig.  S1, C-E). Thus, these data suggest that FTY720-mediated necroptosis is linked to NMIIA-mediated blebbing, resulting in membrane rupture and cell death.
Because FTY720 induced necroptosis was independent of RIPK3 and MLKL, as shown previously (25), and FTY720 is known to modulate ceramide synthesis, we investigated whether FTY720-mediated necroptosis is linked to ceramide generation, a bioactive lipid that is known to alter membrane integrity and induce blebbing (46). Although there were no detectable changes in the overall abundance of ceramide in response to FTY720, as measured by LC-MS/MS (47) in A549 cell pellets (Fig. S1F), FTY720 (20 M) exposure resulted in the formation of ceramide platforms/macrodomains, which seemed to contain holes/pores in the center, on the plasma membrane in A549 cells, as detected by immunofluorescence using two distinct anti-ceramide antibodies (48,49) (Fig. 2, A  and B). There was no accumulation of sphingosine 1-phosphate (S1P), as detected by anti-S1P antibody (50), or cholesterol, as detected by anti-cholesterol antibody (51), on the plasma membrane with/without FTY720, used as a negative controls (Fig.  S2, A and B). Moreover, NBD-FTY720 (20 M) showed no drug accumulation on the plasma membrane of A549 cells (Fig.  S2C), suggesting that FTY720 itself does not accumulate in the ceramide-enriched membrane macrodomains, which appear to be around 1-2 m in diameter. In addition, ceramide accumulation was not detectable in the Golgi, ER, early endosomes, caveolin-enriched lipid microdomains, or exosomes in response to FTY720 in these cells (Fig. S2D), indicating the possibility of these being novel structures. To confirm the presence of ceramide in FTY720-mediated membrane macrodomains, we depleted ceramides in H1341 cells by overexpressing acid ceramidase (AC) (

Ceramidosome-dependent necroptosis
acid (Fig. S2E). Ectopic expression of AC using adenoviral GFP tagged acid ceramidase (Ad-GFP-AC) reduced total ceramide levels ( Fig. S2F) and abrogated the formation of ceramide-enriched membrane macrodomains (Fig. 2, C and D). Recently, ceramiderich membrane areas in the plasma membrane were shown to form pores (54). This, together with our novel data with ceramide labeling, suggests that FTY720 exposure induces the formation of membrane pores, which are enriched in endogenously generated ceramides, altering plasma membrane integrity and inducing necrotic cell death. These ceramide-enriched membrane pores are here referred to as ceramidosomes. Moreover, our data also showed that exposure of A549 cells to other sphingolipids, such as myriocin (as FTY720 is a myriocin analogue), sphingosine, S1P, or lysophosphatidic acid did not induce ceramidosomes, whereas exposure of cells to FTY720 and its nonphosphorylatable analogue OSU-2S (55) largely mediated ceramidosome formation (Fig. 2E). In addition, a FTY720 mimetic selected from the ChemBridge library in silico (compound 5) was able to induce ceramidosomes, suggesting that ceramidosome formation is not dependent on selective FTY720 exposure (Fig. 2E).
The phosphorylated form of FTY720 (P-FTY720) did not induce ceramidosomes (Fig. 2E). We confirmed this by knocking down sphingosine kinase 2 (SPHK2 or SK2), a key enzyme that phosphorylates FTY720 to P-FTY720 (Fig. S2G). FTY720 and P-FTY720 levels were measured in A549 cells using LC-MS/MS (Fig. S2H), and data showed a significant decrease of P-FTY720 with SK2 knockdown. Knocking down SK2 had no effect on ceramidosome formation compared with controls, scrambled shRNA-transfected, and vehicle-treated cells (Fig.  S2, I and J), confirming that this was an FTY720-and not P-FTY720 -dependent mechanism. Interestingly, in H1341 cells, induction of ceramidosomes was detected with 0.1-1 M FTY720 exposure for 60 min, whereas A549 cells required a higher dose of FTY720 (Fig. 2, F-H), indicating different sensitivities of these cells to FTY720. Although ceramidosome formation in response to FTY720 was evident in multiple cancer lines, including H1650 (non-small-cell lung cancer), H1341 (small cell lung cancer), and 22A (head and neck squamous cell carcinoma) (Fig. 2I), they were not evident in noncancerous skin keratinocytes, normal human lung fibroblasts, and NIH3T3 mouse embryonic fibroblasts (Fig. S2K). Taken

Ceramidosome-dependent necroptosis
together, these data suggest that FTY720 and FTY720 mimetics, but not P-FTY720, mediate the formation of ceramidosomes in various cancer cells, but not in noncancerous keratinocytes or fibroblasts, in a dose-and time-dependent manner preceding blebbing, plasma membrane rupture, and necroptosis.
Moreover, to confirm formation of membrane pores by FTY720-mediated ceramidosomes, we measured inside-out transfer of BODIPY in A549 cells by immunofluorescence. The BODIPY signal was captured within ceramidosomes (detected by co-localization of BODIPY with ceramide) in response to FTY720 at 1.5-3 h (Fig. 3A). This resulted in a reduction of the cytoplasmic BODIPY signal between 2.5-4 h of FTY720 exposure (Fig. 3B), suggesting that ceramidosomes mediate the formation of membrane pores that induces inside-out transfer of BODIPY in response to FTY720. Interestingly, FTY720-mediated ceramidosomes had no effect on the internalization of BODIPY in these cells (Fig. 3C). In addition, ceramidosomeinduced membrane pores (2.5 h), leading to plasma membrane rupture (4 h), were also detectable by TEM using gold labeling of ceramide signals in response to FTY720 (Fig. 3D) compared with controls; TEM was performed using anti-mouse IgG antibody in A549 cells (Fig. 3E). These data suggest that ceramide molecules are detectable within membrane pores and ruptured membranes (Fig. 3D, center and right panels, arrows) in response to FTY720 treatment at 2.5 and 4 h, respectively. As a control, anti-mouse IgG antibody was used in gold labeling and TEM studies with/without FTY720. Although ruptured membranes were detectable in response to FTY720 at 4 h, there was no detectable gold labeling around these pores in these controls (Fig. 3E, right panel). Thus, these data support that FTY720mediated ceramidosomes mediate the formation of membrane pores, leading to alterations of plasma membrane integrity.

C16-ceramide is the main ceramide species contained in ceramidosomes during necroptosis
Ceramide can be generated by hydrolysis of sphingomyelin (SM) by sphingomyelinases (SMases) or synthesized de novo by the functions of the rate-limiting enzymes serine palmitoyl transferase (SPT) and CerS1-6 in response to cellular stress  (56,57). To investigate the source of the ceramide enriched in membrane pores in response to FTY720, we generated A549/ H1341 cells stably expressing shRNAs against SMase (nSM1-3), SPT (SPT1-3), and CerS1, CerS2, CerS4, CerS5, and CerS6 using lentiviral particles. CerS3 was omitted because of its selective expression in the skin and testes (58). shRNA-dependent knockdown of nSMases or SPT enzymes had no effect on ceramidosome formation with/without FTY720 (Fig. S3, A-D). Interestingly, knockdown of ceramide synthase 5 or 6, which generate C16-ceramide, but not CerS1, CerS2, or CerS4, abrogated ceramidosomes (Fig. 4, A and B) and protected cells from necroptosis in response to FTY720 (Fig. 4C). Reduction of C16ceramide by silencing CerS5 and/or CerS6 was confirmed by lipidomics analysis compared with Scr-shRNA-transfected controls (Fig. S3, E and F). Thus, these data suggest that ceramidosomes contain C16-ceramide generated by CerS5 and/or CerS6 in response to FTY720.

Ceramidosomes mediate RIPK1-dependent necroptosis
Because FTY720-mediated necroptosis was shown to be dependent on RIPK1 but not RIPK3 or MLKL, we then assessed whether RIPK1 played any roles in the formation of ceramidosomes in response to FTY720. To determine whether RIPK1 is localized within ceramide-enriched membrane pores, we first measured the co-localization of RIPK1 and ceramide by immunofluorescence using anti-RIPK1 and anti-ceramide antibodies. FTY720 exposure induced the co-localization of RIPK1 and ceramide compared with vehicle-treated controls within the plasma membrane (Fig. 5A). This co-localization between RIPK1 and ceramide was also detectable by proximity ligation assay (PLA) using anti RIPK1 and anti-ceramide antibodies (Fig. 5, B and C) and immunogold TEM using gold-labeled anti-RIPK1 (labeled with 10 nM gold particles) and anti-ceramide (labeled with 1.4 nM gold particles) antibodies in response to FTY720 compared with vehicle-treated controls (Fig. 5D). Thus, these data suggest that RIPK1 is co-localized with ceramide at the plasma membrane within ceramidosomes in response to FTY720-mediated stress.
Then, to determine whether RIPK1 was necessary for ceramidosome formation, we measured the effects of RIPK1 knockdown on ceramidosomes in response to FTY720. shRNA-mediated knockdown of RIPK1 (Fig. S4A) prevented ceramidosome formation in response to FTY720 (Fig. 5, E and F) and protected cells from necroptosis compared with Scr-shRNA-transfected and vehicle-treated cells (Fig. 5G). In con-   A, co-localization of RIPK1 and ceramide was measured by confocal microscopy. A549 cells were treated with 20 M FTY720 for 2 h and stained with anti-ceramide antibody and anti-RIPK1 antibody and detected by immunofluorescence. Co-localization of RIPK1 and ceramide is evident on the plasma membrane in response to FTY720 treatment (the bottom right inset shows a zoom-out of the framed area). B and C, this interaction was detected by PLA in A549 cells treated with FTY720 (20 M) or vehicle (DMSO). This is quantified in C. Data represent means Ϯ S.D. from five independent experiments, analyzed by paired Student's t test (n ϭ 5; *, p ϭ 0.006). D, RIPK1-ceramide co-localization was further detected by immunogold transmission EM. H1341 cells were treated with FTY720 or vehicle (DMSO). The micrograph shows increased RIPK1 (10 nM gold) association with ceramide (1.4 nM gold). E and F, to study the effects of RIPK1 silencing on ceramidosome formation, A549 cell were stably transfected with lentiviral shRNA against RIPK1. These cells were treated with 20 M FTY720 or vehicle (Veh) for 2 h and stained with anti-ceramide antibody and Rhodamine phalloidin (F-actin). shRIPK1 had significantly less ceramide-enriched membrane pores formed, as quantified in F. Data represent means Ϯ S.D. from three independent experiments, analyzed by paired Student's t test (n ϭ 3; *, p ϭ 0.01). G, an LDH assay was used to study the effect of RIPK1 knockdown on FTY720-induced necroptosis. shRIPK1-A549 cells were less sensitive to FTY720 treatment compared with shScr cells. Data represent means Ϯ S.D. from three independent experiments, analyzed by paired Student's t test (n ϭ 3; *, p ϭ 0.048; **, p ϭ 0.008; ns, not significant). Scale bars ϭ 10 m.

Ceramidosome-dependent necroptosis
trast, knocking down RIPK3 using shRNA had no effect on ceramidosomes formation (Fig. S4B-D) or cell death (Fig. S4E). Given that RIPK3 is known to activate MLKL, which induces membrane pores for the execution of necroptosis, we also measured the effects of MLKL inhibition on ceramide-enriched membrane pore formation using the pharmacological inhibitor necrosulfonamide (NSA). NSA-mediated MLKL inhibition had no effect on FTY720-mediated ceramidosome formation or necroptosis (Fig. S4, F-H). As an additional control, we also explored whether there were other known pore-forming proteins involved in the execution of FTY720-mediated necroptosis. To achieve this, we measured the effects of purinergic receptor protein (P2X7), which is known to mediate membrane pores in response to cytotoxic stress, such as increased ATP accumulation (59), on FTY720-mediated ceramidosome formation and necroptosis. The data showed that pharmacological inhibition of P2X7 using the inhibitor AZ10606120 (60) had no effect on induction of ceramidosomes or necroptosis in the absence/presence of FTY720 (Fig. S4, I-K). These data are consistent with our previous studies, indicating that although RIPK3 and MLKL were dispensable, RIPK1 is necessary for FTY720-induced ceramidosome formation and induction of type II necroptosis, which is a noncanonical necroptosis mediated by RIPK1 independently of RIPK3/MLKL signaling.

RIPK1 directly binds ceramide to form ceramidosomes
To investigate whether the association between ceramide and RIPK1 was due to their direct interaction, we first measured ceramides by lipidomics after immunoprecipitation of RIPK1 or RIPK3 using anti-RIPK1 or anti-RIPK3 antibodies and lipid extraction on agarose beads used for pulling down possible RIPK1-ceramide or RIPK3-ceramide complexes, as we described previously (36,61) (Fig. 6, A and B). These data revealed that there was an increase in C16-ceramide along with dihydro (dh)-C16-ceramide in FTY720-treated cell extracts bound to RIPK1 compared with controls (vehicle-treated or anti-RIPK3 antibody immunoprecipitations) (Fig. 6B). It should be noted that, although ceramidosomes are composed of mainly C16-ceramide, RIPK1-ceramide complexes might also include other endogenous ceramides, such as C14-ceramide, C18-ceramide, C20-ceramide, or C24-ceramide (and, to a lesser extent, sphingosine) compared with controls (Fig. 6B). In reciprocal studies, binding of ceramide with RIPK1 was measured in vitro using commercially available biotinylated C6-ceramide and recombinant human RIPK1, which were isolated by avidin-conjugated agarose beads. The amount of RIPK1 bound to biotin-ceramide was measured by ELISA using anti-RIPK1 antibody. It was found that ceramide bound to RIPK1 with a K d of 1.9 nM (Fig. 6C), and the stoichiometry of RIPK1-ceramide binding was found to be 1:2.4, calculated as we described previously (61). In addition, increased formation of heterodimers of RIPK1 upon FTY720 exposure was also detected using nondenaturing Western blot analysis (Fig. 6D). Of note, exposure to FTY720 did not alter monomeric RIPK1 expression (Fig. S5A). These biochemical data were then used for in silico modeling of the RIPK1-ceramide complex containing the RIPK1 dimer associated with four molecules of C16-ceramide (Fig. 6E) by molecular docking and simulations using the already existing X-ray crystal structure of RIPK1 (62) (PDB codes 4NEU and 5HX6). The top-ranked model of the RIPK1-ceramide complex identified potential residues of RIPK1 that were important for ceramide binding, including Asp 147 and Asn 169 . Then, to confirm and validate the model for RIPK1-ceramide binding, we performed site-directed mutagenesis to generate D147A, N148A, N169A, or N186A conversions and measured their effects on co-localization with ceramide at the plasma membrane (Fig. 6, F and G). WT-RIPK1-GFP was co-localized with ceramide at the plasma membrane, whereas the association between ceramide and these mutants was reduced (40 -70%) compared with WT-RIPK1 in response to FTY720 (Fig. 6, F and  G). The loss of ceramide binding of N169A-RIPK1, which was selected based on high level of expression, was confirmed using biotinylated C6-ceramide and partially purified N169A-RIPK1-GFP, as described above, which showed no binding ( Fig. 6H and Fig. S5A). These data were also consistent with the effects of altered ceramide-RIPK1 association because of N169A mutations of RIPK1, which protected H1341 cells from necroptosis in response to FTY720 (Fig. 6, I and J). Similarly, reduced ceramide interaction in response to N147A-RIPK1 mutation (Fig.  6, F and G) also prevented these cells from FTY720-mediated necroptosis compared with WT-RIPK1 (Fig. 6K). Thus, these data suggest that RIPK1-ceramide binding plays a key role in the generation of membrane pores (ceramidosomes) and induction of necroptosis after FTY720-mediated cellular stress. These data also suggest that, at least under in vitro conditions, RIPK1 binding to ceramide is not dependent on the fatty acyl chain length of ceramide.

The ceramide-RIPK1 complex is transported to the plasma membrane by NMIIA
To establish whether the ceramide-RIPK1 complex is transported to the plasma membrane via vesicular trafficking, we treated A549 cells with FTY720 at 37°C or at 4°C and measured ceramide-RIPK1-enriched membrane pores at the plasma membrane. The data demonstrated that, although ceramidosomes were formed in cells incubated at 37°C in response to FTY720, incubation of cells at 4°C largely prevented the formation of these structures at the plasma membrane. Interestingly, when cells that were initially incubated at 4°C were later transferred to 37°C, formation of ceramidosomes was restored (Fig. S6, A and B), indicating the involvement of temperature-sensitive trafficking, which is usually dependent on vesicular transport. This was confirmed by pretreating A549 cells with a known vesicular trafficking inhibitor, Brefeldin A (10 g/ml) (63), for 24 h before FTY720 exposure, which prevented ceramide-RIPK1 complex transport at the plasma membrane (Fig. S6, C and D). Interestingly, vesicular membrane trafficking of the ceramide-RIPK1 complex was also dependent on an intact actin cytoskeleton because inhibiting actin polymerization using Latrunculin A (64) attenuated ceramidosome formation (Fig. S6, E and F).
Because blebbistatin, which is an inhibitor of NMIIA, also protected ceramide-enriched pores (Fig. 1, C and D), and NMIIA is known to transport cargo proteins to the plasma membrane through the actin cytoskeleton (65), we then

Ceramidosome-dependent necroptosis
hypothesized that NMIIA, a recently identified tumor suppressor protein (66,67), might be key for trafficking of the ceramide-RIPK1 complex to the plasma membrane to induce the formation of ceramidosomes in response to FTY720. To test this hypothesis, we first measured the co-localization of NMIIA with ceramide and RIPK1 using immunofluorescence. FTY720 exposure induced co-localization of NMIIA with ceramide and RIPK1 (Fig. 7, A and B, and Fig. S6G) compared with vehicletreated controls. The co-localization of NMIIA and RIPK1 within ceramidosomes at the plasma membrane in response to FTY720 was also consistent with TEM studies performed using gold-labeled anti-NMIIA (labeled with 1.4 nM gold particles) and anti-RIPK1 (labeled with 10 nM gold particles) antibodies (Fig. 7C). Interestingly, PLA analysis showed that, although NMIIA associated highly with RIPK1, there was no detectable interaction between NMIIA and ceramide (Fig. 7, D-G). These data suggest that FTY720-mediated formation of ceramidosomes depends on the ceramide-RIPK1 complex, which is then transported to the plasma membrane by NMIIA via RIPK1-NMIIA association.
We then determined whether NMIIA knockdown inhibits the formation of ceramidosomes in response to FTY720. In fact, shRNA-mediated knockdown of NMIIA, a 75% reduction in NMIIA protein abundance compared with Scr-shRNAtransfected controls (Fig. S6I), reduced the formation of ceramidosomes (80%) in response to FTY720 compared with Scr-shRNA-transfected and vehicle-treated controls (Fig. 7, H and  I). Down-regulation of NMIIA using shRNA also protected A549 cells from FTY720-mediated necroptosis compared with controls (Fig. 7J). A 3D image of the ceramide-RIPK1-NMIIA complex was then generated using the immunofluorescence images shown in Fig. S6G using ImarisCell software (Fig. S6I).

Ceramidosome-dependent necroptosis
The involvement of NMIIA in the formation of ceramideenriched membrane pores was further confirmed using blebbistatin, which attenuated the formation of ceramidosomes (Fig. S6J). However, pharmacologic inhibitors of kinesin spindle protein (ARRY520) (68), kinesin (K858) (69), and Dynein (ciliobrevin) (70), respectively, had no effect on the formation of ceramide-RIPK1-enriched pores after FTY720-mediated stress (Fig. S6J). Thus, these data suggest that NMIIA is key for transport of the ceramide-RIPK1 complex to the plasma membrane and that the ceramide-RIPK1-NMIIA complex integrates ceramidosomes to form a membrane macrodomain and subsequent membrane pores, leading to FTY720-induced necroptosis.

NMIIA interacts with RIPK1 for trafficking of the RIPK1-ceramide complex to the plasma membrane
Because our studies suggest that trafficking of the ceramide-RIPK1 complex to the plasma membrane is facilitated by NMIIA via its interaction with RIPK1, we then generated a model of NMIIA-RIPK1, containing the RIPK1-ceramide complex, using molecular modeling and simulations (Fig. 8A), as described above (Fig. 6E). This model (the top-ranked model among various others) suggested a few possible residues of RIPK1 that might be involved in NMIIA interaction, including Leu 112 , Cys 256 , Arg 258 , Leu 293 , and Glu 295 (Fig. 8A, right panel). We then confirmed this model by measuring the effects of WT-RIPK1-GFP compared with various mutants of RIPK1 with L112A, C256A, R258A, L293A, or E295A conversions with the GFP tag on NMIIA interaction using Western blotting (Fig.  S7A) and the formation of ceramidosomes on the plasma membrane using immunofluorescence (Fig. 8, B and C). The data showed that alterations of RIPK1-NMIIA interaction because of R258A, L293A, or E295A mutations on RIPK1 limited the formation of ceramidosomes in response to FTY720 (Fig. 8, B and C). We picked mutant L293A for live-cell imaging, and the mutation also protected cells from FTY720-mediated necroptosis compared with cells expressing WT-RIPK1-GFP (Fig. 8, D  and E). To visualize the ceramide-RIPK1-NMIIA complex (ceramidosome) that forms large pores on the plasma membrane in response to FTY720, we isolated ceramidosomes from the plasma membrane by using a method for the isolation of giant plasma membrane vesicles (GPMVs) (71). After isolation, GPMVs were fixed and labeled using gold-labeled anti-ceramide, anti-RIPK1, or anti-NMIIA antibodies for visualization Figure 7. The ceramide-RIPK1 complex is transported to the plasma membrane by NMIIA. A, co-localization of NMIIA with ceramide in ceramide-enriched membrane pores was detected by immunofluorescence using anti-ceramide and anti-NMIIA antibodies (the bottom right inset shows a zoom-out of the framed area). B, co-localization of RIPK1 and NMIIA in ceramide-enriched membrane pores was detected by immunofluorescence using anti-ceramide and anti-RIPK1 antibodies. Scale bars ϭ 10 m. C, immunogold TEM showing co-localization of NMIIA (1.4 nm) and RIPK1 (10 nm) in H1341 cells treated with FTY720. D, PLA assay in A549 cells detected increased RIPK1 and NMIIA association in response to FTY720. The PLA assay is quantified in E. Data represent means Ϯ S.D. from three independent experiments, analyzed by paired Student's t test (n ϭ 5; **, p ϭ 0.004). F and G, the PLA assay did not show ceramide-NMIIA interaction in response to FTY720. The detected PLA signal is quantified in G. Data represent means Ϯ S.D. from three independent experiments, analyzed by paired student t test (n ϭ 5; ns, not significant). H and I, A549 cells with stable NMIIA knockdown (shMYH9) were stained for ceramide-enriched membrane pores using anti-ceramide and anti-RIPK1 antibodies. Scale bars ϭ 10 m. This is quantified in I. Data represent means Ϯ S.D. from three independent experiments, analyzed by paired Student's t test (n ϭ 3; *, p ϭ 0.01; **, p ϭ 0.003). J, LDH assay in stable shMYH9 cells treated with increasing doses of FTY720 show significant protection from FTY720-induced cell death. Data represent means Ϯ S.D. from three independent experiments, analyzed by paired Student's t test (n ϭ 3; *, p ϭ 0.004; **, p ϭ 0.01; ***, p Յ 0.0008).

Ceramidosome-dependent necroptosis
by TEM (Fig. 8F). The TEM images of these GPMVs demonstrate that, although ceramide surrounds the inner channel, RIPK1 and NMIIA appear to have a general distribution within these vesicles (Fig. 8F), consistent with the molecular modeling (Fig. 8A) and the 3D image (Fig. S6I). Overall, these data suggest that RIPK1-NMIIA interaction, involving the Arg 258 , Leu 293 , and Glu 295 residues of RIPK1, plays a key role in the transport of the ceramide-RIPK1 complex to the plasma membrane for the formation of ceramidosomes and large membrane pores, involved in the execution of FTY720-mediated necroptosis.

Ceramidosomes are formed de novo in D. melanogaster ovaries during development
Because ceramide-RIPK1-NMIIA-mediated ceramidosomes, which form large membrane pores, appear to be highly organized and well-structured, we investigated whether they are also formed in various cell types in vivo without FTY720 exposure or any exogenous stress stimuli. To determine this, we elected to use D. melanogaster ovaries because of the presence of various stages of development and cell types (72)(73)(74). We isolated ovary tissues from D. melanogaster and performed immunofluorescence labeling using an anti-ceramide antibody. The confocal microscopy images confirmed the presence of these ceramide macrodomains in a few (two/three) distinct cells (Fig. 8G). Because there are only two/three GSCs present in ovaries, co-localization of ceramide macrodomains and GSCs was detected by immunofluorescence using anti-ceramide and anti-1B1/hts (a stem cell marker) (75) antibodies in the ovary tissues. These data revealed a distinct co-localization of ceramide macrodomains within the GSCs in the early stages of ovary development (Fig. 8G).
To confirm the co-localization detected between 1B1 and ceramide, we also analyzed the effects of the inhibition of ceramide synthesis using fumonisin B1 (FB1), a ceramide synthase inhibitor, on the formation of ceramidosomes and ovary development. The data showed a large decrease in ceramide and 1B1 co-localization H and I, isolated ovaries from D. melanogaster exposed to FB1/vehicle were stained for ceramide and 1B1. There was decreased co-localization between ceramide and 1B1 in flies exposed to FB1. The rightmost panels show zoomed-out pictures of the framed areas. Quantification of 1B1 and ceramide co-localization is shown (I). J, to investigate the effect of genetic loss of zipper (NMIIA), ovarioles isolated from zipper RNAi flies were stained for ceramide and a stem cell marker (1B1). As shown, zipper RNAi flies had stunted ovary development (whole ovary from zipper, bottom panel) compared with the germarium (from the control, top panel).

Ceramidosome-dependent necroptosis
without affecting oogenesis (Fig. 8, H and I). These data suggest that, without complete elimination of ceramidosome formation, ovary development might not be affected.
To further study whether ceramide-enriched macrodomains play any functional roles in ovary development, we obtained zipper (an NMIIA homolog) (76) RNAi flies and drove RNAi using the Nos RNAi driver for ovary-selective zipper knockdown. We then determined the effects of zipper knockdown on the formation of ceramide pores/macrodomains and ovary development. Remarkably, siRNA-mediated knockdown of zipper resulted in defective ovary development in insects, leading to a completely sterile phenotype (Fig. 8J and Fig. S7B), consistent with attenuation of ceramidosome formation in the ovaries. Taken together, our FB1 and zipper studies indicate a possible role of ceramidosomes in GSC signaling and oogenesis.
Overall, these data suggest that ceramide-enriched macrodomains containing ceramide and the NMIIA homologue zipper are selectively formed de novo without any external drug treatment or stress stimuli in GSCs found in the early developmental stages of ovaries, playing a key role in ovary development, at least in D. melanogaster.

Discussion
In this study, our data demonstrate that the ceramide-RIPK1-NMIIA complex forms large membrane pores that are referred to here as ceramidosomes. The formation of ceramidosomes in response to FTY720-mediated stress resulted in plasma membrane blebbing, leading to compromised membrane integrity and necroptosis in cancer cells. FTY720/ceramidosome-mediated necroptosis was RIPK1-dependent; however, unlike the canonical mechanism, RIPK3 and MLKL were dispensable in this process. Moreover, our data suggest that ceramidosomes can also form selectively without any external stress stimuli or treatment, at least in the early stages of oogenesis, in germ line stem cells in D. melanogaster. Partial inhibition of ceramide synthesis or knockdown of NMIIA (zipper) limited ceramidosome formation, and NMIIA knockdown resulted in alterations of ovary development, suggesting important roles for ceramidosomes in GSC signaling during development.
It is known that ceramide induces necroptosis in response to tumor necrosis factor ␣-mediated stress in L929 cells, which is protected by acid ceramidase-dependent hydrolysis of ceramide (77). Previous data also suggest that ceramide-mediated necroptosis plays a key role in hyperosmosis-induced cell death in erythrocytes (suicidal death of erythrocytes, eryptosis), which is associated with membrane blebbing and necrosis (78). However, the mechanisms by which ceramide induces membrane blebbing and necroptosis have been unknown. A recent study showed that ceramide nanoliposomes induced MLKLdependent necroptosis in ovarian cancer cells (1). Our data suggest that FTY720-mediated necroptosis depends on the formation of novel ceramide-enriched membrane pores (ceramidosomes) that are associated with formation of the ceramide-RIPK1-NMIIA complex. This ceramidosome-dependent necroptosis appears to be distinct from canonical necroptosis, as RIPK3 and MLKL were dispensable for this process in response to the FTY720-induced stress response.
As shown by our data, formation of ceramidosomes in response to FTY720 was also accompanied by smoothening of the plasma membrane (loss of membrane ruffles and microvilli that were visible by SE in vehicle-treated cells). Interestingly, this phenomenon was previously evident in cells depleted of plasma membrane cholesterol, which led to massive endocytosis associated with recruitment of sphingosine kinase 1 (SphK1) (79). Whether cholesterol metabolism plays a role in the formation of these necroptotic ceramidosomes remains to be investigated.
Previous research has suggested that ceramide forms lipid channels in vitro and that these ceramide channels exist at the mitochondrial membrane and are disassembled by anti-apoptotic Bcl-2 family proteins (80,81). However, whether ceramide channels in mitochondria play any role in the regulation of cell death remains unknown. To this end, our study demonstrates that FTY720-mediated ceramidosomes are formed with the involvement of ceramide-RIPK1 association, possibly in the ER, and are then trafficked to the plasma membrane through actin filaments by NMIIA. We hypothesize that ceramides associated with RIPK1 play a key role in the insertion of the ceramide-RIPK1-NMIIA complex into the plasma membrane through the inner leaflet, leading to the formation of ceramidosome-mediated pores. Sustained ceramidosome-induced membrane pores then alter membrane integrity, leading to necroptosis in response to stress stimuli such as FTY720. The formation of ceramide pores within the RIPK1 dimer, as our molecular modeling/simulation studies predicted, however, needs to be further investigated and validated.
It has been shown that NMIIA is involved in the trafficking of protein cargo to various biological membranes with distinct functions, such as controlling exocytosis or positioning of organelles in cells (82,83). Our data are also consistent with the tumor suppression function of NMIIA (84); it has been reported that genetic deletion results in squamous cell carcinoma development in mice through inactivation of p53 signaling. Similarly, here we report the prevention of ceramidosomemediated necroptosis by shRNA-mediated NMIIA knockdown in response to FTY720 in lung cancer cells. However, whether ceramidosomes form in other cell types without any external stress stimuli was unknown.
To address this, we used D. melanogaster as a model organism, specifically ovaries, as they contain multiple types of cells at various stages of development. The data suggested that ceramidosomes form in germ line stem cells and that inhibition of ceramidosome formation using siRNA-mediated knockdown of zipper (an NMIIA homologue) inhibited ovary development, possibly by alteration of germ line stem cell signaling, without affecting their survival. However, the mechanisms by which ceramidosomes regulate ovary development are unknown and need to be determined. It is known that there are nanotubes in the D. melanogaster testes that mediate niche stem cell signaling (85). It remains unknown whether ceramidosomes play similar roles in ovaries. It is possible that ceramidosomes might regulate germ line stem cell niche organization by enhancing epidermal growth factor receptor (EGFR) signaling (86,87). Interestingly, ceramide macrodomains appeared to also contain central holes/channels when co-localized with 1B1 in ova-

Ceramidosome-dependent necroptosis
ries, a known component of fusomes, which are cell-to-cell communication channels (88), controlling GSC proliferation and differentiation signals in ovaries. Thus, it is possible that ceramidosomes might regulate the trafficking of hydrophobic signaling molecules, such as sterols and sterol-derived molecules (89), from neighboring cells to GSCs to regulate stem cell niche organization. Interestingly, our data suggest that FTY720-induced ceramidosomes might induce membrane pores, which enhance inside-out transfer of BODIPY, without affecting internalization of this molecule. These data suggest that ceramidosomes might involve the inside-out transfer of various hydrophobic molecules, including various lipids, which might then act as signaling molecules for neighboring cells. Alternatively, these data also suggest that ceramidosomes might play a role in detoxification of cells when endogenous stress-related lipid molecules reach high/toxic concentrations by inducing their externalization/clearance. These possible roles of ceramidosome-induced membrane pores, however, need to be further investigated. This study presents a novel lipid-protein complex involving the ceramide-RIPK1 complex, which is trafficked to the membrane to form pores; ceramidosomes, inducing membrane rupture and necroptosis in response to FTY720 in lung cancer cells; or in germline stem cell signaling de novo, regulating ovary development.

Immunofluorescence labeling
Cultured cells were incubated with the respective treatment, washed twice with 1ϫ PBS, fixed with prewarmed 4% paraformaldehyde (PFA) for 15 min at room temperature, permeabilized for 10 min with 0.01% Triton X-100, and blocked for 20 min with 1% BSA before incubation with primary antibody in blocking solution overnight at 4°C. The following day, the cells were incubated with secondary antibody (Alexa Fluor 488 anti-mouse) and Rhodamine phalloidin. The coverslips were mounted on slides with a drop of Prolong anti-fade reagent (Invitrogen) and sealed with clear nail polish. Labeling patterns were observed using an Olympus FV10i laser-scanning confocal microscope. Images were analyzed with Fluoview Fv10i software. Ceramidosome quantification was performed using Fiji (ImageJ).

Proximity ligation assay
Proximity ligation assays were performed using the Duolink In Situ Red Kit (Sigma) according to the manufacturer's instructions and then analyzed as described previously (61).

Co-immunoprecipitation and Western blotting
For immunoprecipitation, cells were grown to 70% confluency and treated with FTY720 or vehicle. Cells were rinsed with Ceramidosome-dependent necroptosis cold 1ϫ PBS, scraped off, and lysed in Pierce immunoprecipitation lysis buffer (nondenaturing) with protease inhibitor mixture and phosphatase inhibitor. Immunoprecipitation was done with RIPK1, RIPK3, or myosin IIA antibody. Lysates were precleared with 40 l of protein A/G beads and control IgG and incubated in a rotary shaker for 1 h at 4°C. Samples were centrifuged at 2,000 ϫ g for 5 min. The supernatant was collected and incubated with the respective antibody at 4°C for 1 h before adding 40 l of agarose beads and rotating overnight. The following day, samples were centrifuged at 2,000 ϫ g for 5 min; the beads were washed with cold 1ϫ PBS four times, followed by addition of 2ϫ loading dye. The samples were boiled using a heating block before Western blotting. For assays, cells were homogenized in radioimmune precipitation assay buffer (25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, and EDTA (20 l of 0.5 M stock)), including protease inhibitor mixture, using a 26.5-gauge syringe five times and incubated for 20 min on ice. The samples were centrifuged at 12,000 ϫ g for 15 min at 4°C. The supernatant was then collected, and proteins were quantified using the Bradford method. 4ϫ loading dye was added to samples and boiled using a heating block at 95°C for 5 min before loading. SDS gradient gels, 4 -20%, were used to run the samples using the Bio Rad Criterion apparatus, followed by semidry transfer onto a polyvinylidene difluoride membrane. Blocking was done with 5% milk in 0.1% PBS. Primary antibodies were used at 1:2,000 dilution overnight at 4°C. Proteins were analyzed by Western blotting using the following antibodies: rabbit RIPK1 (Cell Signaling Technology), ␤-actin (Sigma), anti-GFP (Santa Cruz Biotechnology), and nonmuscle myosin IIA (Sigma).

Quantitative RT-PCR
RNA was extracted from cell pellets using the RNeasy kit (Qiagen) according to the manufacturer's instructions. Complementary DNA was generated using equal amounts of RNA from each sample and the iScript complementary DNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. Reactions were carried out using SsoFast probes mixture (Bio-Rad) and TaqMan primer probes (Thermo Fisher Scientific) in a Bio-Rad CFX96 RealTime system as described by the manufacturer.

Membrane permeability/apoptosis kit with YO-PRO-1 and PI
Cells were cultured to 70 -80% confluency and then treated with 20 M FTY720 or vehicle. The cells were collected by scraping off the dishes, rinsed with cold 1ϫ PBS, and resuspended in 1 ml of cold PBS; dyes were then added at 1 l/ml concentration. The samples were kept on ice for 20 -30 min before analysis by flow cytometry on an LSR Fortessa/X-20. For live-cell imaging, cells were treated with 20 M FTY720 or vehicle and incubated with the dyes (1 l/ml). Pictures were taken every 30 min for a period of 3 h. For the RIPK1 mutants, membrane permeability was analyzed by propidium iodide only in GFP-positive cells.

Trypan blue exclusion assay
Cells were seeded in 6-well plates and allowed to adhere for ϳ20 h. Treatment was with 20 M FTY720 in DMSO or the corresponding amount of vehicle control. Medium containing dead cells was pelleted with trypsinized cells, resuspended in 1ϫ PBS, and then counted in a hemocytometer after addition of trypan blue dye (Sigma-Aldrich) at a 1:10 dilution.

Lipidomics analysis
Cells were collected by scraping off from the dishes and washed twice with cold PBS. Further preparation of samples and measurement of endogenous ceramides by LC-MS/MS followed a protocol described previously (47). Briefly, samples were supplemented with internal standards, and 2 ml of isopropyl alcohol:water:ethyl acetate (30:10:60 (v:v:v)) was added to the extracts. Samples were subjected to two rounds of vortexing and sonication, followed by 10 min of centrifugation at 4,000 rpm. The supernatant or top layer was used as lipid extract and subjected to LC-MS/MS for analysis of ceramide species. Lipid extraction and analyses were performed by the MUSC Lipidomics Shared Resources. P i was used for normalization.

Isolation of ceramide-enriched membrane pores
We followed a protocol published previously for isolating giant plasma membrane vesicles (71). In summary, cells were grown in 100-mm dishes to 80% confluency. These cells were treated with vehicle/FTY720. The cells were rinsed twice with GPMV buffer (10 mM HEPES, 150 mM NaCl, and 2 mM CaCl (pH 7.4)) and incubated with vesiculation buffer (25 mM PFA/2 mM DTT in GPMV buffer) for 1 h in a 37°C incubator. GPMVs were collected and processed for EM.

Ceramide docking to RIPK1
Using the software MOE from CCG (90), a complete contiguous model of the kinase domain from RIPK1 was created based on X-ray-derived coordinates (PDB codes 4NEU and 5HX6 (91, 92)). The protein was protonated at T ϭ 310 K (pH Ceramidosome-dependent necroptosis 7.3), salt at 200 mM, and using the generalized Born/volume integral (GB/VI) electrostatics. The final structure was energy minimized using the AMBER12 force field. C16-ceramide was docked to the RIPK1 monomer using the entire surface as a target. Initial placement calculated 100 poses using triangle matching with London dG scoring; the top 50 poses were then refined using force field and affinity dG scoring, as we described previously (36,61).

Molecular interaction analysis
The heterodimer structures from PDB code 3L82 and the best heterodimer output from ClusPro for Fbxo4:Fxr1 were interrogated for intermolecular contacts. Protein contact thresholds were 4.5 for hydrophobic interactions, 4.2 for ionic bonds, 2.5 for disulfide bonds with a sequence separation of 4, and a network separation of 0. Molecular images were prepared using MOE, as we described previously (36,61).

Molecular modeling of ceramidosomes forming membrane pores
The theoretical model was created in several steps, as we described previously (36,61). First the top C16-ceramide-RIPK1 results from the in silico probe were duplicated to make a symmetric RIPK1 homodimer containing two molecules of ceramide. Two molecules of ceramides were added per molecule of RIPK1 and placed along the active site as a stack to create a homodimer of RIPK1 with four total ceramides. Finally, the C-terminal domain (CBD) of NMIIA was attached to each molecule of RIPK1. This was done using a prediction derived from a CluPro protein-protein interaction simulation. Part of the recognition domain of CBD from myosin molecules is defined and uses a conserved amphipathic helix (Glu 1547 -Lys 1555 in the case of NMIIA). An NMIIA (also known as MYH9) CBD domain homology model was created using the CBD (also known as a globular tail domain) from MyoVa (PDB code 4KP3). Using the structural file for RIPK1 and the CBD of NMIIA, bimolecular docking was performed. Using the docking server ClusPro (93), the RIPK1 structure (as described above) was uploaded as the ligand, and the NMIIA CBD structure was uploaded as the receptor. The binding residues were not selected, nor were any residues selected to block docking, and default settings were used to allow maximum freedom of docking poses. The best pose and four of the top 10 poses indicated an interaction with the Glu 1547 -Lys 1555 helix and indicated that it is the favored interaction site. Using protein superposition between the CluPro results and the aforementioned RIPK1 dimer, the CBD was attached to each molecule of RIPK1.

Quantitative detection of ceramide-RIPK1 binding
H1341 cells were transfected with GFP-tagged RIPK1 WT, RIPK1 N169A, or empty vectors. Cells were lysed by freezing and thawing in nondenaturing buffer (50 mM Tris (pH7.4), 150 mM NaCl, 1 mM EDTA, 0.5% NP-40, and 1:500 protease inhibitor mixture). Lysates (2 mg of protein) were precleared with 40 l of agarose beads. The precleared lysates were incubated with anti-GFP magnetic beads at 4°C for 18 h with gentle agitation. The beads were then washed with 1ϫ PBS and incubated with unlabeled ceramide or biotin ceramide in 200 l of binding buffer. RIPK1-biotin-ceramide complexes were quantified using competitive ELISA with streptavidin-coated plates. The results were analyzed using GraphPad Prism 7 software, as we described previously (36,61).

Ultrastructural analysis using TEM
A549 cells were collected by scraping off tissue culture plates, rinsed with 1ϫ PBS, and fixed in 2% glutaraldehyde in 1ϫ PBS. After post-fixation with 2% (v/v) osmium tetroxide, specimens were embedded in Epon 812, and sections were cut orthogonally to the cell monolayer with a diamond knife. These sections were visualized with a JEOL1010 transmission electron microscope.

Immunogold analysis using transmission EM
H1341 cells were treated with FTY720/vehicle and collected by scraping the cells off the plate. The cells were rinsed with 1ϫ PBS and fixed for 15 min with 4% paraformaldehyde. The cells were rinsed twice with 1ϫ PBS and permeabilized with 0.01% saponin. They were rinsed again twice with 1ϫ PBS and blocked for 20 min with 1%BSA (in 1ϫ PBS) before incubation with primary antibody (anti-ceramide or anti-mouse IgG, 1:100) at 4°C for 18 h. They were then rinsed twice with 1% BSA before incubation with secondary antibody (gold-conjugated donkey anti-mouse/rabbit) at 1:200 dilution for 1 h at 37°C. Cells were rinsed twice with 1% BSA, post-fixed with 1% glutaraldehyde (in 1ϫ PBS) for 10 min, rinsed with distilled H 2 O, and enhanced with silver staining before Epon embedding and imaging.

SE
After treatment, samples were fixed with primary fixative (2% glutaraldehyde and 2% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.35)). Secondary fixation was achieved with 1% osmium tetroxide in 0.1 M cacodylate buffer. After both fixation steps, samples were washed three times with double-distilled H 2 O and then dehydrated with increasing ethanol concentrations (20% to 100%). Samples were submerged in 100% ethanol, and critical point drying was achieved with Tousimis Autosamdri-931 (Rockville, MD). Before imaging, 3-6 nM iridium was applied to samples, and images were acquired with a Magellan 400 field emission scanning electron microscope (FEI, Hillsboro, OR) at the Notre Dame Integrated Imaging Facility.

Germ line knockdown of zipper by shRNA
Fly stocks were obtained from the Bloomington Drosophila Stock Center and maintained at 25°C. Zipper was knocked down in the germ line by driving zipper shRNA using Gal4nos.NGT. Adult ovaries in F1 progeny from this cross were rudimentary.

Drosophila ovary immunostaining and microscopy
Immunostaining of ovaries was performed as described earlier (74). ␣-Ceramide was used at 1:100 (48). ␣-1B1 was obtained from the Developmental Studies Hybridoma Bank and used at 1:50 dilution. Confocal images were taken using a Zeiss LSM 880 NLO microscope and Plan-Apochromat ϫ63/ 1.40 oil differential interference contrast objective. Images were analyzed using Fiji (ImageJ).

Ceramidosome-dependent necroptosis Immunogold labeling of Drosophila ovaries
Isolated ovaries from D. melanogaster flies were stained with gold-labeled ceramide as described previously (94). Briefly, paraformaldehyde tissues were sliced into ultrathin sections using a microtome. The sections were mounted on carbon-coated copper grids. The grids were then rinsed, blocked, and incubated with ceramide antibody and then gold-labeled antibody before contrast staining with uranyl acetate and imaging.

FB1 exposure of D. melanogaster
10 -15 adult male and female OregonR flies were exposed to food containing 100 M FB1. Flies not exposed to FB1 were used as a control. F1 generation flies from exposed and nonexposed parents were used for immunostaining adult ovaries using anti-1B1 and anti-ceramide antibodies.

Statistical analysis
Data were reported as mean Ϯ S.D. Statistical analysis was performed by analysis of variance or Student's t test using Prism/GraphPad software version 7. p Ͻ 0.05 was considered statistically significant, as described previously (36,61).