Direct binding to ceramide activates protein kinase Czeta before the formation of a pro-apoptotic complex with PAR-4 in differentiating stem cells.

We have reported that ceramide mediates binding of atypical protein kinase C (PKC) zeta to its inhibitor protein, PAR-4 (prostate apoptosis response-4), thereby inducing apoptosis in differentiating embryonic stem cells. Using a novel method of lipid vesicle-mediated affinity chromatography, we showed here that endogenous ceramide binds directly to the PKCzeta.PAR-4 complex. Ceramide and its analogs activated PKCzeta prior to binding to PAR-4, as determined by increased levels of phosphorylated PKCzeta and glycogen synthase kinase-3beta and emergence of a PAR-4-to-phosphorylated PKCzeta fluorescence resonance energy transfer signal that co-localizes with ceramide. Elevated expression and activation of PKCzeta increased cell survival, whereas expression of PAR-4 promoted apoptosis. This suggests that PKCzeta counteracts apoptosis, unless its ceramide-induced activation is compromised by binding to PAR-4. A luciferase reporter assay showed that ceramide analogs activate nuclear factor (NF)-kappaB unless PAR-4-dependent inhibition of PKCzeta suppresses NF-kappaB activation. Taken together, our results show that direct physical association with ceramide and PAR-4 regulates the activity of PKCzeta. They also indicate that this interaction regulates the activity of glycogen synthase kinase-3beta and NF-kappaB.

In previous studies, we have shown that the simultaneous elevation of the sphingolipid ceramide and the atypical PKC/ 1 inhibitor protein PAR-4 is critical for the induction of apoptosis in differentiating mouse embryonic stem (ES) cells (1)(2)(3). We demonstrated for the first time that during cell division, PAR-4 and the neuroprogenitor (NP) marker protein nestin are asymmetrically distributed to the two daughter cells (1). Only the PAR-4(Ϫ)/ nestin(ϩ) daughter cell survives, whereas its PAR-4(ϩ)/nestin(Ϫ) sibling undergoes ceramide-induced apoptosis. In other studies, we have found that PAR-4 sensitizes differentiating mouse as well as human embryoid body-derived cells (EBCs) toward apoptosis induction by ceramide and ceramide analogs (2). We also reported that the novel ceramide analog S18 (N-oleoyl serinol) eliminates residual pluripotent Oct-4(ϩ)/PAR-4(ϩ) stem cells from EBCs, enriches for NPs, and prevents teratoma formation from stem cell-derived neural transplants in mouse brain (2). It remained to be elucidated, however, which mechanism drives ceramide-induced or ceramide analog-induced apoptosis in the PAR-4(ϩ) cells.
We and others have suggested that ceramide activates atypical PKC but also promotes inhibition of atypical PKC by PAR-4 (1)(2)(3)(4)(5)(6)(7)(8)(9). To resolve this apparent paradox, we have analyzed the ceramide-induced or ceramide analog-induced formation of a protein complex between PKC and PAR-4 and determined the effect of ceramide analogs on the activation of NF-B and the degree of apoptosis. Ceramide and ceramide analogs have been shown to enhance phosphorylation of PKC in vitro and in vivo, however, without demonstrating direct and specific physical interaction of ceramide with atypical PKC (4 -8). Filter or solid phase binding assays are commonly used to test the affinity of a protein to its lipid ligand. The binding of PKC to ceramide, however, has not been reported yet using this or a similar type of binding assay. Most recently, it has been shown that solid phase/overlay binding assays fail to show the specific interaction of proteins with ceramide (10). We developed a novel assay based on mixed ceramide/phospholipid vesicles to determine binding of atypical PKC to ceramide. Our results show that PAR-4 binds to PKC that is associated with ceramide and has first been activated due to ceramide-mediated elevation of enzyme phosphorylation. Thus, NF-B-dependent cell survival is enhanced or suppressed by ceramide, which is dependent on the absence or presence of PAR-4, respectively. Polar Lipids Inc. (Alabaster, AL). Polyclonal anti-PAR-4 rabbit IgG (catalog number sc-1807), polyclonal anti-SUMO-1 rabbit IgG (catalog number sc-9060), polyclonal anti-PKC rabbit IgG (catalog number sc-216), polyclonal anti-PKC␦ rabbit IgG (catalog number sc-937), and monoclonal anti-PAR-4 mouse IgG (catalog number sc-1666) were from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-cleaved caspase 3 rabbit IgG (catalog number 9661S), polyclonal anti-phosphorylated (phospho-Thr 403/410 ) PKC/ rabbit IgG, and polyclonal antiphosphorylated (phospho-Ser 9 ) GSK-3␤ rabbit IgG were purchased from Cell Signaling (Beverly, MA). Monoclonal anti-ceramide mouse IgM (clone 15B4) was from Alexis (San Diego, CA). Donkey anti-mouse, anti-rabbit, and anti-goat IgG Cy2, Cy3, and Cy5 conjugates; Cy2conjugated donkey anti-mouse IgM (-chain specific); Cy3-conjugated goat anti-mouse IgG (Fc␥-fragment specific); goat anti-mouse IgG horseradish peroxidase conjugate; normal rabbit IgGs; and normal donkey serum were purchased from Jackson ImmunoResearch (West Grove, PA). Myristoylated PKC pseudosubstrate inhibitor peptide, biotinylated PKC substrate peptide, and NF-B translocation inhibitor peptide SN50 were from Calbiochem. NBD-C6-ceramide/bovine serum albumin conjugate, ER tracker Blue-White DPX, Mito-Tracker® Red CM-H 2 XPhos, Alexa Fluor 594-conjugated wheat germ agglutinin, Vybrant CM-diI, Zenon direct labeling kit for rabbit IgG, and Alexa 350and 488-conjugated Annexin V were obtained from Molecular Probes (Eugene, OR). The magnetic activated cell sorting (MACS) kit including Annexin V-conjugated magnetic beads was from Miltenyi Biotec (Auburn, CA). Streptavidin-coated membranes (SAM 2 ® Biotin Capture Membrane), pNF-B-luc reporter plasmid, and the dual luciferase reporter assay were from Promega (Madison, WI). [␥-32 P]ATP (specific activity, 6000 Ci/mmol) was obtained from ICN Biomedicals (Irvine, CA). The pan-caspase FLICA (fluorochome inhibitor of caspases) assay kit was from Chemicon (Temecula, CA). The Lipofectamine 2000 transfection reagent was obtained from Invitrogen. All reagents were of analytical grade or higher.
Preparation of Lipid Vesicles and PKC Binding Assays-Lipid vesicles were obtained from dried mixtures of equimolar amounts of phosphatidylserine (PS) (105 g) and C16-ceramide (85 g) following modified procedures for large liposome preparation (10 -17). The lipid mixtures were resuspended and sonicated for 1 h in 100 l of buffer A consisting of 50 mM Tris/HCl (pH 8.1) and 150 mM NaCl (10 -17). After adding 300 l of buffer A supplemented with 0.1 mM MnCl 2 , the samples were centrifuged at 12,000 ϫ g for 20 min at 4°C. The pellet (large lipid vesicles) was resuspended in 100 l of buffer A and incubated with 1 nmol of Vybrant CM-diI for 1 h at 37°C to visualize the vesicle fraction after MACS separation. A detergent-free cell lysate was prepared by sonication of differentiating ES cells or EBCs in 300 l of hypotonic buffer (10 mM Tris/HCl (pH 7.0) with protease and phosphatase inhibitors) followed by removal of membranous debris by centrifugation. The cleared lysate was added to the lipid vesicle suspension, and the mixture was incubated for 2 h at 4°C. The reaction mixture was supplemented with 20 l of 20ϫ Annexin V binding buffer and 50 l of a solution containing magnetic beads conjugated to Annexin V followed by incubation for 1 h at 4°C. MACS was performed according to the manufacturer's (Miltenyi) protocol. The presence and quantity of lipid vesicles were determined by monitoring the Vybrant CM-diI fluorescence in the flow-through and elution fractions using a microplate fluorescence reader. The linear correlation between the amount of vesicular lipid and fluorescence intensity of vesicle-bound Vybrant CM-diI was verified by quantitative high-performance thin-layer chromatography of the lipid mixture applied to Annexin V-based MACS. The specificity of the binding reaction of PKC or other proteins to the Cer/PS vesicles was verified by an antibody competition assay using 1 g of anti-PKC rabbit polyclonal antibody to incubate the stem cell-derived lysate for 1 h at 4°C prior to incubation with the lipid vesicles. The protein binding to Cer/PS vesicles in the MACS eluate was analyzed by SDS-PAGE and immunoblotting.
ES Cell Differentiation, Immunocytochemistry, and FRET Analysis-In vitro neural differentiation of mouse ES cells (ES-J1) followed a serum deprivation protocol as previously published (1). Apoptosis was induced by the addition of 80 M of the novel ceramide analog S18, 30 M C2-ceramide, or 2 M C16-ceramide. C16-ceramide was dissolved in ethanol/2% dodecane before addition to the medium in a 1:1000 dilution. Controls were obtained with diluted vehicle only. For immunocytochemistry and FRET analysis, cells were fixed with 4% p-formaldehyde/phosphate-buffered saline, followed by permeabilization with 0.2% Triton X-100 for 5 min at room temperature. Nonspecific binding sites were saturated by incubation with 3% ovalbumin and 2% donkey serum in phosphate-buffered saline at 37°C for 60 min. Cells were stained with primary and fluorescence-labeled secondary antibodies at a concentration of 5 or 10 g/ml, respectively. In case of simultaneous labeling with anti-PAR-4 mouse monoclonal IgG and anti-ceramide mouse monoclonal IgM, we used isotype-specific secondary antibodies (Fc␥-or -chain specific, respectively) for the immunodetection reaction. In control experiments, we ruled out any potential cross-reactivity of the isotype-specific secondary antibody with non-matching primary antibodies (Supplemental Fig. S1). The ceramide specificity of the 15B4 antibody was verified using a pre-adsorption assay with ceramide or sphingomyelin following a protocol as previously reported (18) and shown in Supplemental Fig. S2. Epifluorescence microscopy was accomplished using a Zeiss Axioplan Deconvolution microscope or a Zeiss Axiophot microscope equipped with a Spot digital camera. Confocal fluorescence microscopy and FRET were performed using a Zeiss LSM confocal laser scanning microscope equipped with a two-photon argon laser at 488 nm (Cy2, green fluorescent protein) and 543 nm (Cy3, HcRFP) or 633 nm (Cy5), respectively. Acceptor (Cy5) bleaching was achieved by repetitive scanning at 633 nm until the Cy5 signal was minimized. Data were collected from images at a resolution of 1024 ϫ 1024 pixels, and the fluorescence increase of the Cy3 signal was calculated by densitometry based on pixel-to-pixel counting within selected areas of the images.
PKC Activity Assay-EBCs (5 ϫ 10 6 cells) were treated for 20 min with 500 l of lysis buffer containing 1% Brij96V in 50 mM Tris/HCl (pH 7.5), 5 mM MgCl 2 , 100 g/ml bovine serum albumin, and protease and phosphatase inhibitor mixtures at 4°C. The cell lysate was centrifuged for 15 min at 12,500 ϫ g. For each activity assay, 100 l of the supernatant were supplemented with 8 g of PS pre-dissolved in dimethyl sulfoxide and 40 M biotinylated PKC substrate peptide (final concentration) pre-dissolved in de-ionized water. The reaction was started by the addition of 50 M of non-labeled ATP (final concentration) and 10 Ci of [␥-32 P]ATP and then incubated at room temperature. At various time periods of incubation (10,20,40, and 60 min), aliquots of 20 l were taken, and the reaction was stopped by the addition of 12 l of 7.5 M guanidinium-HCl. For binding of radiolabeled peptide on streptavidin-coated membranes (SAM 2 ®), 15 l of the reaction mixture were spotted on the membrane, and unbound radiolabeled phosphate was removed according to the instructions of the manufacturer (Promega). The radiolabeled product peptide bound to the streptavidincoated membranes was quantified by scintillation counting.
NF-B Activity Assay-Differentiating EBCs at the NP2 (24 h after replating of dissociated EBs) or NP3 stage (48 h after replating of dissociated EBs, for details see Ref. 1) were transfected with 1 g of pNF-B-luc, a multimerized B-luciferase reporter gene plasmid, and 0.2 g of pRL-CMV (Renilla luciferase) internal control plasmid to normalize on the efficacy of transfection using the Lipofectamine 2000 procedure. Cells were co-transfected with pcDNA3.1-PAR-4 or pcDNA3.1 empty vectors as controls. Twenty-four hours after transfection, cells were treated for 15-24 h with different concentrations of S18 and/or PKC pseudosubstrate inhibitor (PZI). Cell extracts were prepared, and the levels of luciferase activity were determined using the dual luciferase reporter assay system in accordance with the manufacturer's (Promega) instructions.

RESULTS
Ceramide-containing Lipid Vesicles Bind to PKC-It has been shown that PKC can be activated in vitro by supplementation of the enzyme assay with ceramide (4 -8, 19, 20). However, solid phase or vesicle binding assays demonstrating sustained and specific binding of ceramide to PKC have not been reported yet. Recently published results suggested that solid phase/overlay binding assays are not suitable to test the affin-ity of proteins to ceramide (10). We attempted to test binding of PKC to ceramide by using a lipid vesicle binding assay. The preparation of pure ceramide (C16-ceramide) vesicles, however, failed as verified by microscopic inspection, density gradient centrifugation, and filter exclusion assays (data not shown). Therefore, we developed a novel technique termed lipid vesicle-mediated affinity chromatography (LIMAC) to isolate ceramide-binding proteins. Based on procedures used for phosphatidylserine liposome aggregation assays, we prepared lipid vesicles from a variety of phospholipid and ceramide mixtures (10 -17). Lipid vesicles obtained from PS, phosphatidylcholine (PC), and/or ceramide (Cer) were labeled with the fluorescent membrane marker dye Vybrant CM-diI. The labeled vesicles were then subjected to Annexin V MACS to isolate the PScontaining lipid vesicles due to binding of PS to Annexin Vconjugated magnetic beads. vesicles were composed of the two lipids (Fig. 1B).
To test binding of PKC to ceramide, lipid vesicles were incubated with recombinant human PKC prior to Annexin V-based MACS. Protein bound to the vesicles in the elution fraction was analyzed by SDS-PAGE and immunoblotting. Fig.  1C shows that PKC was only co-purified with Cer/PS vesicles (lanes 2 and 6), but not with PS/PC vesicles (lane 1), indicating that binding of PKC was specific for ceramide.
The absence of PKC in the elution fraction of PS/PC vesicles suggested that PS by itself did not sustain binding of the lipid vesicles to the enzyme, although PS has been reported to activate PKC in vitro (4,(21)(22)(23). In contrast to numerous studies showing the specific association of PS with classical PKC␣ (24 -26), binding data for the physical association of PS to PKC are not available yet. Absence of binding of PKC to PS/PC vesicles may have been due to a low affinity of the enzyme to PS. Alternatively, the combination of lipids in the vesicles may have favored binding of PKC to Cer/PS rather than PS/PC vesicles. The binding of PKC to Cer/PS vesicles was suppressed by incubation of PKC with anti-PKC antibody (Fig. 1C, lane 8) prior to the addition of the lipid vesicles. This result indicated that binding of the antibody to PKC blocked access of ceramide to its binding site on the enzyme. The antibody itself did not bind to the vesicles as shown by the absence of immunostaining in the elution fraction of Cer/PS vesicles that were incubated with anti-PKC IgG alone (Fig.  1C, lane 7). This antibody competition assay clearly demonstrated that PKC bound specifically to the Cer/PS vesicles, ruling out nonspecific hydrophobic interaction. To visualize co-localization of ceramide and PKC, Cer/PS vesicles were labeled with NBD-C6-ceramide prior to the addition of PKC and immunofluorescence staining of the vesicle-bound enzyme. Fig. 1D shows that NBD-C6 ceramide was co-localized with Alexa 546-linked anti-PKC antibody, indicating that PKC specifically bound to ceramide in Cer/PS vesicles (arrows). Our results also suggested that ceramide-containing lipid vesicles could be used to isolate ceramide-binding proteins.
We tested whether the LIMAC procedure could be used to bind endogenous PKC from a detergent-free EBC lysate that was incubated with Cer/PS vesicles. Fig. 2 shows that the Cer/PS vesicles bound to PKC (lane 3) and PAR-4 (lane 6). Binding of PKC or PAR-4 to Cer/PS vesicles was not complete, as indicated by protein detected in the flow-through fractions (Fig. 2, lane 2). Incomplete binding is commonly observed with binding assays and indicates a binding equilibrium depending on the concentration of lipid vesicles and protein in the incubation reaction. Alternatively, only a portion of PKC may have been accessible to binding to the Cer/PS vesicles, whereas the non-binding portion was partitioned to the flow-through fraction. Immunostaining for PAR-4 showed enrichment of an additional protein with a molecular mass of about 60 kDa (Fig. 2,  lanes 6 and 7). This protein also stained for SUMO-1 (Fig. 2,  lane 9) indicating that the PKC-containing Cer/PS vesicles bound specifically to PAR-4 that was modified by sumoylation. This result was consistent with the observation that a portion of non-modified PAR-4 was recovered in the flow-through fraction (Fig. 2, lane 5) and suggested that modification of PAR-4 by sumoylation increased its affinity to the lipid vesicles or vesicle-bound PKC. PAR-4 and its sumoylated form were not co-purified in the Cer/PS eluate when binding of PKC to the vesicles was suppressed by prior incubation with anti-PKC antibody (Fig. 2, lanes 8 and 10). This suggests that PAR-4 did not directly bind to the Cer/PS vesicles but co-purified in a protein complex with ceramide-associated PKC. Non-related proteins such as actin were only found in the flow-through fraction of the MACS column, demonstrating the specificity of the PKC association to Cer/PS vesicles (Fig. 2, lanes 11 and  14). PKC␦, another kinase proposed to interact with ceramide, was eluted in trace amounts, indicating that this enzyme was only weakly associated with Cer/PS vesicles (Fig. 2, lanes 13  and 16). Hence, our results suggested that ceramide-containing lipid vesicles bound specifically to PKC in a cell lysate and that ceramide-associated PKC formed a complex with PAR-4.
PKC Bound to Intracellular Ceramide Is Phosphorylated and Forms a Complex with PAR-4 -To determine intracellular binding of PKC to endogenous ceramide, we transfected EBCs with vectors encoding PKC-EGFP and tested the co-distribution/association with ceramide using a ceramide-specific mouse monoclonal antibody. Although this antibody or a similar antibody has shown a broader specificity toward sphingolipids in solid phase/overlay assays, it was found to react specifically with intracellular ceramide when used with fixed cells (1,18,(27)(28)(29)(30)(31)(32)(33)(34). Our results and those of other groups indicated that solid phase/overlay binding assays are not well suited to test the affinity of proteins to ceramide. The reason for low or absent binding to ceramide may be a different conformation of ceramide incorporated into a lipid membrane as compared with that coated on a solid surface. It is likely that this also applies to ceramide-specific antibodies, which results in a different binding specificity depending on whether ceramide is presented by a lipid membrane or coated on a solid surface. To show ceramide-specific binding, the anti-ceramide antibody was pre-adsorbed to ceramide (18) or sphingomyelin prior to immunocytochemistry. Supplemental Fig. S2, A shows that only pre-adsorption with ceramide (but not sphingomyelin) reduced the antibody-derived fluorescence signal. These results suggested that the anti-ceramide antibody was able to distinguish ceramide from sphingomyelin when used for immunocytochemistry. Our results did not exclude, however, that the antibody may have also bound to dihydroceramide, the precursor of ceramide. Co-staining with markers for the ER and Golgi indicated that the perinuclear region showing elevation of endogenous ceramide was the ER but not the Golgi (Supplemental Fig. S2, B and C). The distribution of the antibody-derived fluorescence signal was consistent with the subcellular site for de novo ceramide biosynthesis (ER), whereas sphingomyelin synthesized in the Golgi was not recognized.
We transfected cells with a vector encoding PKC-EGFP to determine the dependence of PKC phosphorylation on the expression level of the enzyme and the distribution of ceram- ide. Fig. 3A shows that the immunofluorescence signal for pPKC/ was strongest in cells that ectopically expressed PKC-EGFP and/or showed elevated levels of endogenous ceramide. The strong signal for PKC-EGFP indicated the specificity of the antibody reaction for pPKC, although the antibody used was raised against the phosphorylated epitope of both PKC and PKC. In cells with elevated ceramide, pPKC/ was mainly distributed to the nuclear envelope and the perinuclear region, consistent with the subcellular sites that showed the strongest immunostaining for ceramide.
Using co-immunoprecipitation assays, we have reported that the ceramide analog S18 induces the formation of a protein complex between PKC and PAR-4 (2). To test this complex formation in living cells, we transfected EBCs with the cDNAs encoding PKC-EGFP and PAR-4-HcRFP prior to incubation with S18. In cells co-expressing PKC-EGFP and PAR-4-HcRFP, time-lapse fluorescence microscopy showed that incubation with 80 M S18 induced a perinuclear complex of the two fusion proteins about 30 -60 min prior to apoptosis as indicated by intensive Hoechst staining (Fig. 3B). In control experiments, we verified that this subcellular distribution was not observed when the fluorescent proteins were expressed by themselves (Supplemental Fig. S3). We tested whether the endogenously expressed proteins were also co-distributed in response to elevation of endogenous ceramide. Fig. 4 shows that ceramide was co-distributed with PKC and PAR-4 in cells with increased ceramide levels. This co-localization was mainly found in apoptotic cells, as indicated by staining for Annexin V, and was consistent with our previous observation that simultaneous elevation of ceramide and PAR-4 induced apoptosis in neural stem cells (1,2). These results showed that the subcellular distribution of ectopically expressed PKC-EGFP and PAR-4-HcRFP in response to ceramide or ceramide analogs was equivalent to that of the endogenous proteins.
The observation that both ceramide-associated endogenous PKC and PKC-EGFP were phosphorylated (Fig. 3A) suggested that PKC was also phosphorylated when associated with PAR-4. To determine the physical association of endogenous ceramide, pPKC/, and PAR-4, we used confocal scanning laser immunofluorescence microscopy and FRET analysis, a technique commonly used to test the formation of a complex between proteins and/or other binding partners. Fig. 5 shows strong FRET signals from ceramide (Cy2) to PAR-4 (Cy3) and from PAR-4 (Cy3) to pPKC/ (Cy5). The two FRET signals co-localized in the perinuclear region and were more intense when cells underwent apoptosis as indicated by the appearance of condensed nuclei. The number of condensed nuclei showing perinuclear staining of a ceramide⅐pPKC⅐PAR-4 complex was increased by 2-fold when EBCs were incubated with 2 M C16-ceramide or 80 M S18. Consistent with our previous observation (Fig. 3A), the co-localization of a strong pPKC/ signal with PKC-EGFP suggested that the FRET signal resulted from a complex of ceramide and PAR-4 with pPKC. The specificity and authenticity of the FRET Note that the intensity of the pPKC/ signal was dependent on the expression level of PKC and the concentration and subcellular distribution of ceramide. B, EBCs (NP2 stage) were co-transfected with cDNAs encoding PKC-EGFP and PAR-4-HcRFP. At 24 h post-transfection, cells were incubated with 80 M S18. The fluorescence signals were recorded by time-lapse fluorescence microscopy at 37°C. Note that PKC-EGFP and PAR-HcRFP were co-distributed to a perinuclear region prior to staining of condensed (apoptotic) nuclei with Hoechst dye.

FIG. 4. Perinuclear co-distribution of endogenous ceramide, PKC, and PAR-4 in apoptotic cells.
Immunocytochemistry was performed on EBCs (NP2 stage) using antibodies against ceramide (Cy2, isotype-specific), PAR-4 (Cy3, isotype-specific), and PKC (Cy5, all signals are visualized in different pseudocolors). Apoptotic cells were identified by staining with Alexa 350-conjugated Annexin V. Note that PKC and PAR-4 were co-localized in apoptotic cells that also showed a strong signal for ceramide. The perinuclear co-distribution of ceramide, PKC, and PAR-4 appears white due to overlay of three pseudocolors. signal were verified by donor de-quenching due to acceptor bleaching of a Cy5-labeled secondary antibody against the anti-pPKC/ antibody (Supplemental Fig. S4). Acceptor bleaching reduced quenching of the Cy3 signals from the secondary antibodies detecting ceramide and PAR-4, indicating that ceramide was associated with a pPKC⅐PAR-4 protein complex. In summary, our results suggested that endogenous ceramide as well as S18 enhanced phosphorylation of PKC and that PAR-4 formed a protein complex with the ceramide-associated and phosphorylated enzyme.
Ceramide-induced Activation of PKC Is Compromised by Binding to PAR-4 -In our previous study, we have reported that antisense-knockdown of PAR-4 resulted in the reduction of apoptosis that was inducible with endogenous ceramide or exogenously added ceramide analogs (1). Here, we quantified the degree of ceramide analog-induced apoptosis in dependence on the level of PKC activity and the presence of PAR-4. We expressed the fluorescent fusion proteins of either PKC (PKC-EGFP) or PAR-4 (PAR-4-HcRFP) and determined the effect of the expression level of the two proteins on ceramide analog-inducible apoptosis. Using Annexin V-based MACS, we quantified the distribution of the ectopically expressed proteins by cell counting of PKC-EGFP(ϩ) or PAR-4-HcRFP(ϩ) cells in the two fractions of non-apoptotic (Annexin V(Ϫ)) or apoptotic (Annexin V(ϩ)) cells. We found that 83 Ϯ 9% of the PKC-EGFP(ϩ) cells were non-apoptotic, whereas 86 Ϯ 10% of the PAR-4-HcRFP(ϩ) cells were apoptotic. This result indicated that elevated expression of PKC counteracted ceramide-inducible apoptosis, whereas enhanced expression of PAR-4 promoted apoptosis. PKC-dependent reduction of apoptosis was consistent with our observation that ectopically expressed PKC-EGFP showed enhanced phosphorylation and thus activation of the enzyme (Fig. 3A). Cells that overexpressed PAR-4, however, were undergoing apoptosis in the presence of ceramide analogs.
The cells obtained from the two MACS fractions were lysed, and the protein was analyzed by SDS-PAGE and immunoblotting for the expression of endogenous PKC or PKC-EGFP and endogenous PAR-4 or PAR-4-HcRFP. Fig. 6A shows that PKC-EGFP was mainly found in the fraction of non-apoptotic Annexin V(Ϫ) cells, whereas PAR-4-HcRFP was mainly detected in the fraction of apoptotic Annexin V(ϩ) cells. The observation that PKC-EGFP was only found in the Annexin V(Ϫ) fraction suggested that overexpression of PKC protected against apoptosis, which is consistent with the results from cell counting. The distribution of endogenous PAR-4 was similar to that of PAR-4-HcRFP, whereas endogenous PKC was distributed to fractions of both apoptotic and non-apoptotic cells. This distribution suggested that ceramide analog-dependent activation of PKC did not induce apoptosis unless PAR-4 was expressed as well. Hence, cells expressing PKC but no or only low levels of PAR-4 were non-apoptotic in response to ceramide analogs. However, cells expressing similar levels of both PKC and PAR-4 were apoptotic. This result was consistent with our previous studies showing that ceramide analogs induce apoptosis only in PAR-4-expressing cells (1,2).
Using radioactive labeling and immunoprecipitation assays, we have previously shown that the serinol analogs of ceramide induced intracellular phosphorylation of PKC, even at a concentration that resulted in the induction of apoptosis (4). Here, we tested how the degree of PKC phosphorylation affected the activity of the enzyme in the presence of PAR-4. Fig. 6B shows that incubation of EBCs with S18 resulted in the elevation of PKC phosphorylation, consistent with enhanced phosphorylation of GSK-3␤, an intracellular substrate of PKC. In vitro assays of PKC activity showed that incubation of EBCs with S18 at a low concentration (40 M) resulted in a 2-fold increase in the activation of PKC (Fig. 6C). At this concentration in the medium, S18 was enriched to ϳ70% of the concentration of endogenous ceramide found in EBCs (data not shown). At a higher concentration of S18 in the medium (80 M), the activation of PKC was ablated, concurrent with an increased level of apoptosis (Fig. 6, C and D). Although ectopic expression of PAR-4 by itself was not sufficient to significantly inhibit PKC, its activity was dramatically reduced when cells were transfected with PAR-4 and at the same time incubated with S18 (Fig. 6C). Induction of apoptosis by incubation of EBCs with a cell-permeable pseudosubstrate inhibitor of PKC (PZI) showed that inhibition of PKC was sufficient to induce apoptosis (Fig. 6, C and D). PZI has been shown to specifically inhibit PKC or PKC in various cell systems (35)(36)(37)(38)(39). Hence, we cannot exclude that some of the results obtained with PZI may be due to inhibition of PKC. In summary, our data suggested that binding of ceramide or ceramide analogs induced phosphorylation and activation of atypical PKC. In cells with elevated expression of PAR-4, binding of PAR-4 to activated atypical PKC resulted in the inhibition of the enzyme and the induction of apoptosis.
Ceramide-activated PKC Regulates NF-B in Dependence on the Level of PAR-4 Expression-It has been suggested that ceramide-mediated complex formation between atypical PKC and PAR-4 reduces the activity of NF-B and that this results in the induction of apoptosis (1, 3, 4, 7, 40 -45). We have FIG. 5. Ceramide/S18-induced activation of PKC prior to perinuclear association with PAR-4. Immunocytochemistry was performed using antibodies against ceramide, pPKC/, and PAR-4, and a potential association was detected by FRET analysis (Cy2 (ceramide) to Cy3 (PAR-4) and Cy3 (PAR-4) to Cy5 (pPKC/)) using confocal laser scanning microscopy. The secondary antibodies were isotype-specific and did not cross-react with non-matching primary antibodies (see controls in Supplemental Fig. S1). Note that the FRET signals were co-localized in the nuclear envelope and a perinuclear region in cells with condensed (apoptotic) nuclei (arrows). Also note that FRET signals were absent or were not co-localized in cells with regularly shaped (non-apoptotic) nuclei.
previously shown that in differentiating stem cells, the induction of apoptosis by endogenous ceramide and exogenously added ceramide analogs is concentration-dependent and re-quires the expression of PAR-4 (1, 2). To test the effect of ceramide analogs on the activity of NF-B, we co-transfected EBCs at different NP stages with a vector encoding the NF-B FIG. 6. Regulation of PKC activity and induction of apoptosis by ceramide/S18 and PAR-4. A, EBCs (NP2 stage) were transfected with cDNAs encoding PKC-EGFP or PAR-4-HcRFP. Twenty-four hours post-transfection, cells were incubated overnight with 80 M S18. Cells were trypsinized, labeled with Annexin V-conjugated magnetic beads, and fractionated using MACS. The picture shows immunoblots performed with solubilized protein from the non-apoptotic Annexin V(Ϫ) and apoptotic Annexin V(ϩ) cell fractions. Note that endogenous PKC was distributed to the non-apoptotic as well as apoptotic cell fraction, whereas endogenous PAR-4 was only found in apoptotic cells, indicating that PKC protected against S18-inducible apoptosis unless PAR-4 was also expressed. Caspase 3* denotes activated (cleaved) caspase 3. B, EBCs (NP2 stage) were incubated with or without 50 M S18 overnight, and PKC/ phosphorylated at Thr 410 (pPKC/) or GSK-3␤ phosphorylated at Ser 9 (pGSK-3␤) was immunostained on Western blots obtained with EBC-derived protein. C, protein was solubilized from EBCs and incubated with [␥-32 P]ATP and a biotinylated PKC substrate peptide (50 M). The radioactively labeled product peptide was isolated using streptavidin-coated membranes, and the transferred phosphate (dpm/mg cell protein/min) was quantified by scintillation counting. White bars represent means from four independent incubation reactions with S.E. indicated as error bars. D, EBCs were dissociated by trypsinization, and EBs were replated and cultivated for 24 h (NP2 stage) prior to the addition of various apoptosis inducers (40 M S18, 80 M S18, 30 M C2-ceramide, 2 M C16-ceramide, and 30 M PZI). Apoptosis was quantified after 15 h using labeling of apoptotic cells with FLICA assays and cell counting. White bars represent the means (percentage of apoptotic cells within the total population of EBCs) from four independent incubation reactions with S.E. indicated as error bars. enhancer element (B4)-driven luciferase and a vector containing the cDNA of PAR-4 prior to incubation with S18. Fig. 7 shows that the effect of S18 on the NF-B activity was dependent on the S18 concentration and the differentiation (NP) stage.
At the NP2 stage, S18 incubation at low concentration (40 M) activated NF-B-driven luciferase expression to a degree that was consistent with that of ceramide-induced activation of PKC (Figs. 6 and 7). At a higher concentration (80 M) of S18, however, the NF-B activity was reduced, in particular in cells with elevated expression of PAR-4.
At the NP3 stage, we found NF-B activation even at a high concentration (120 M) of S18 (Fig. 7B). This result was consistent with that of our previous study showing that at a later stage of neural differentiation, the degree of S18-inducible apoptosis dropped concomitantly with the reduced level of PAR-4 expression (2). Reduction of the NF-B activity was only seen at a very high concentration of 250 M S18 or on co-transfection with PAR-4. Low NF-B activity concurrent with elevated apoptosis was also observed after incubation of NPs with the PKC pseudosubstrate inhibitor (PZI) or a combination of 40 M S18 and PZI (Fig. 7, A and B). Inhibition of NF-B with the translocation inhibitor peptide SN50 (50 M) elevated apoptosis by FIG. 7. Regulation of NF-B activity by ceramide analogs and PAR-4. A and B, EBCs were co-transfected with the NF-B reporter vector, an internal control vector (Renilla luciferase), and either the PAR-4 vector or an empty vector at the NP2 or NP3 stage. The transfected cells were incubated overnight with S18 at various concentrations or with 30 M PZI. Note that at the NP3 stage, S18 activated NF-B at a concentration that suppressed NF-B activation at the NP2 stage. Also note that S18-induced activation of NF-B was always suppressed by elevated expression of PAR-4. C, model for ceramide/PAR-4-induced apoptosis in stem cells. Cytosolic PKC is in its non-activated form due to binding of the pseudosubstrate domain to the catalytic domain (Step 1). Binding to ceramide triggers a conformational change and activation of PKC by phosphorylation at Thr 410 (Step 2). Activation of PKC stimulates downstream targets, in particular the NF-B signaling pathway for cell survival. Phosphorylated PKC binds to PAR-4, which results in inhibition of the enzyme and down-regulation of the NF-B signaling pathway (Step 3). 50%, indicating that suppression of NF-B activation was sufficient to induce apoptosis in NPs. Thus, our results suggested that the effect of the ceramide analog on the activity of NF-B and the induction of apoptosis was dependent on the expression level of PAR-4. In cells expressing a low level of PAR-4, S18 activated NF-B, whereas in cells expressing a high level of PAR-4, S18-mediated NF-B activation was suppressed, and apoptosis was induced. DISCUSSION We have reported that simultaneous elevation of the sphingolipid ceramide and the atypical PKC inhibitor protein PAR-4 induces apoptosis in differentiating ES cells (1)(2)(3). Based on this observation, we hypothesized that ceramide may sensitize stem cells toward PAR-4-dependent inhibition of PKC. Previous studies have shown that ceramide activates PKC in vitro and in vivo, but without providing experimental evidence of sustained binding (4 -8). It is well known that activation of enzymes can be due to weak or nonspecific interaction of the effector with the enzyme or even a complex between the substrate and the effector. Solid phase/overlay or filter binding assays, however, failed to demonstrate a specific physical association of ceramide to known ceramide-activated enzymes such as protein phosphatase 2a (10). We show here the specific affinity of PKC to ceramide using a novel binding assay based on MACS sorting of lipid vesicles that was developed in our laboratory. The PKC-dependent induction of apoptosis was tested by incubation with natural ceramide (C16ceramide) and the novel ceramide analog S18 that was synthesized in our laboratory and used to determine the significance of ceramide for the induction of apoptosis in stem cells (1)(2)(3)(4). We defined the mechanism by which ceramide may induce PAR-4-mediated inhibition of PKC and how this inhibition promotes apoptosis. Our results suggest that direct binding of ceramide triggers phosphorylation of PKC and its physical association with PAR-4, which results in down-regulation of NF-B-dependent cell survival and induction of apoptosis. Fig. 7C depicts a model that incorporates the effect of ceramide on PKC and its subsequent effect on binding of PAR-4. It should be noted that this model is likely to apply to other atypical PKC species as well, in particular PKC that has also been found to bind to PAR-4 (41,46,47). In fact, some of the results obtained in our study using the antibody against pPKC/ or the pseudosubstrate inhibitor of PKC (PZI) may involve both atypical PKCs. Cytosolic PKC or PKC has only low activity due to auto-inhibition by binding of the pseudosubstrate domain to the catalytic domain of the enzyme (Fig. 7C, Step 1) (48,49). It has been suggested that binding of ceramide or structurally related lipids to the C1B domain may trigger a conformational change of the enzyme inducing an activated state (Fig. 7C, Step 2) (50, 51). This activation state is indicated by the ceramide analog-induced phosphorylation of the Thr 410 epitope in the activation loop domain of PKC (48,52) and, subsequently, elevated phosphorylation of GSK-3␤, an intracellular substrate of PKC (53)(54)(55)(56). The role of the PKC-dependent phosphorylation of GSK-3␤ for the regulation of stem cell apoptosis remains to be elucidated. The physical association of ceramide with PKC is consistent with our observation that phosphorylated PKC co-distributes with a ceramide-rich compartment. This model is also supported by the observation that PKC binds to ceramide-containing vesicles that show a similar lipid composition as the perinuclear mitochondria-associated membrane subcompartment of the ER (57,58). At the cytosolic face of this compartment, PAR-4 may bind to ceramide-associated and activated atypical PKC, thereby inhibiting the activity of the enzyme (Fig. 7C, Step 3). A previous model has suggested that ceramide either activates PKC or promotes binding to PAR-4 (7). It remained to be elucidated, however, where and how ceramide and PAR-4 interact with PKC. It also remained to be shown that there is a direct physical interaction of ceramide with PKC. Alternatively, ceramide may activate another kinase that phosphorylates PKC, or ceramide may first bind to PAR-4, which then inhibits PKC. Our model thus extends on the previously suggested mechanism in that ceramide activates PKC prior to binding of PAR-4. The ceramide binding assay suggests that a portion of PAR-4 is modified by sumoylation, whose significance for binding of PAR-4 to PKC will be investigated in future studies. Consistent with our model, ceramide or ceramide analogs activate NF-B dependent gene expression, unless PAR-4 or another inhibitor of PKC (e.g. PKC pseudosubstrate) is present. Most inhibitory is the combination of ceramide analogs and PAR-4, which supports our model that pre-activation of PKC by ceramide or ceramide analogs sensitizes the enzyme to inhibition by PAR-4. Inhibition of PKC is concurrent with enhanced apoptosis, indicating that atypical PKC-dependent NF-B activation is critical for cell survival. We suggest that ceramide may have beneficial as well as detrimental effects, depending on the expression level and distribution of PAR-4. In cells with a low level of PAR-4, direct binding to ceramide activates PKC/ and supports cell survival and/or neural differentiation. In cells with a high level of PAR-4, ceramide induces complex formation of activated PKC/ with PAR-4, which results in the inhibition of the enzyme and down-regulation of NF-B-dependent cell survival. In future studies, we will determine by which mechanism ceramide association elevates phosphorylation of atypical PKC and how this regulates the activity of its intracellular substrate(s).