Ceramide kinase regulates phospholipase C and phosphatidylinositol 4, 5, bisphosphate in phototransduction

Abstract

Phosphoinositide-specific phospholipase C (PLC) is a central effector for many biological responses regulated by G-protein–coupled receptors including Drosophila phototransduction where light sensitive channels are activated downstream of NORPA, a PLCβ homolog. Here we show that the sphingolipid biosynthetic enzyme, ceramide kinase, is a novel regulator of PLC signaling and photoreceptor homeostasis. A mutation in ceramide kinase specifically leads to proteolysis of NORPA, consequent loss of PLC activity, and failure in light signal transduction. The mutant photoreceptors also undergo activity-dependent degeneration. Furthermore, we show that a significant increase in ceramide, resulting from lack of ceramide kinase, perturbs the membrane microenvironment of phosphatidylinositol 4, 5, bisphosphate (PIP₂), altering its distribution. Fluorescence image correlation spectroscopic studies on model membranes suggest that an increase in ceramide decreases clustering of PIP₂ and its partitioning into ordered membrane domains. Thus ceramide kinase–mediated maintenance of ceramide level is important for the local regulation of PIP₂ and PLC during phototransduction.

Signal transduction via G-protein–coupled receptors (GPCRs) is vital for many cellular processes including vision, olfaction, taste, and neurotransmission. Extensive studies on proteins constituting this family and their interactions reveal complex signaling networks regulated at multiple levels (1). Lipids play equally important roles in GPCR signaling, as most of the signal transduction machinery is membrane associated. How lipids regulate GPCR signaling is being addressed in recent years (2). Our current knowledge of how lipids influence each other to form membrane microenvironments and how this modulates proteins during signal transduction in a multicellular organism is limited. In this study, we address this issue in the context of Drosophila phototransduction, a prototypic G-protein–coupled phosphoinositide cascade, by genetically modulating the sphingolipid ceramide.

Analyses of Drosophila phototransduction have led to the identification, characterization, and regulation of many signaling components (3). Phototransduction begins with the absorption of light by rhodopsin, followed by the activation of a G protein (Gα_q). Gα_q activates the critical effector NORPA, a phospholipase C (PLC) that catalyzes the hydrolysis of phosphatidylinositol 4, 5-bisphosphate (PIP₂) into two important second messengers, diacylglycerol and inositol 1, 4, 5-trisphosphate. Activation of PLC leads to gating of two transduction channels, transient receptor potential (TRP) and TRP-like. Although many of the proteins involved in phototransduction have been well characterized, we are only beginning to understand how lipids and enzymes involved in lipid metabolism regulate this cascade (4–7). Sphingolipids are integral components of all eukaryotic cell membranes and also act as second messengers in diverse signaling pathways (8). The sphingolipid biosynthetic pathway is an evolutionarily conserved route that generates and interconverts various sphingolipids such as ceramide, sphingosine, ceramide 1–phosphate and sphingosine 1–phosphate (9). We showed earlier that modulating this biosynthetic pathway by targeted overexpression of Drosophila neutral ceramidase (CDase), an enzyme that converts ceramide to sphingosine, rescues retinal degeneration in an arrestin mutant, and facilitates membrane turnover in a rhodopsin null mutant by modulating the endocytic machinery (10–12). Although these studies established that ceramide metabolism is important for survival of photoreceptors, they did not evaluate its role in signaling events during phototransduction.

Ceramide kinase (CERK), a recently cloned lipid kinase, phosphorylates ceramide to ceramide 1–phosphate (C-1-P), thereby decreasing ceramide levels (13, 14). Here we show that Drosophila ceramide kinase (DCERK) regulates PLC activity, function, and the local organization of PIP₂ in GPCR signaling by controlling the ceramide level. Genetic, biochemical, and electrophysiological analyses of DCERK deficient flies reveal a severe down regulation of NORPA and failure in phototransduction. Increased ceramide levels in the mutant also alter the level and membrane microenvironment of PIP₂ that correlates with a failure of NORPA to localize to the membranes. Using fluorescence image correlation spectroscopy in supported bilayers, we show that ceramide perturbs both the protein-dependent and -independent compartmentalization of PIP₂, thus providing a biophysical basis for the effect of ceramide on PIP₂. These findings show that sphingolipids and phospholipids cooperate in vivo to establish a suitable membrane microenvironment for signaling mediated by PLC.

Results

Identification and Characterization of Drosophila CERK (DCERK).

A BLAST search of the Drosophila melanogaster genome identified CG16708 as the Drosophila homolog of the CERK gene and was named DCERK. DCERK is on the right arm of the third chromosome at 82F11–83A1. It encodes a protein of 687 aa and is 35% identical to human CERK (Supporting Information (SI) Fig. S1). Western analyses using monoclonal antibodies raised against DCERK protein showed that it is expressed during all developmental stages (Fig. S2A). Membrane association analyses showed that DCERK is an integral membrane protein that is released from membranes by detergent treatment (Fig. S2B). Hydrophobicity analysis predicts two transmembrane helices (residues 258–282 and 515–533) in the protein. Immunofluorescence analyses of Schneider cells and third-instar larval wing discs using DCERK antibodies showed that the protein predominantly localized to the plasma membrane (Figs. 1A, S2 C and D). Like the mammalian enzyme, DCERK has a diacylglycerol kinase domain and a Ca²⁺/calmodulin binding domain. The N terminus of mammalian CERK contains a pleckstrin homology (PH) domain that binds PIP₂ and targets CERK to the membrane. Although primary sequence comparisons of DCERK do not reveal a PH domain except for a conserved tryptophan residue, secondary structure predictions indicate that DCERK has a noncanonical PH domain. We confirmed that DCERK could phosphorylate ceramide by establishing stable cells expressing tagged DCERK or control vector and measuring enzyme activity in membrane fractions by a chemiluminescence assay that measures ATP depletion (15). DCERK used ceramide but not sphingosine or diacylglycerol as a substrate (Fig. S3 A and B). Sphingolipid enriched fractions were prepared from stable cells expressing DCERK and ceramide and C-1-P levels were measured by ultraperformance liquid chromatography in conjunction with mass spectrometry (UPLC-MS/MS). The analyses confirmed that ceramide kinase overexpression increased C-1-P and decreased ceramide level (Fig. S3C).

Fig. 1.

dcerk¹ Is a severe hypomorph showing light-dependent photoreceptor degeneration. (A) Immunostaining of Schneider cells with antibodies to DCERK and N-cadherin, a plasma membrane marker. DCERK shows plasma membrane staining. (B) Western blot analysis of one, three, and five head extracts (1 h, 3 h, and 5 h) from w¹¹¹⁸ and dcerk¹ with DCERK antibody (3H7) reveals less than 2% DCERK protein in the mutant lanes. The blot is probed with an antibody to inositol polyphosphate 1–phosphatase (IPP) as a loading control. (C) RT-PCR analysis shows significant reduction in DCERK transcript in dcerk¹. Reactions with actin serve as a control. Transmission electron micrographs (TEM) showing photoreceptors of 5-day-old (D) w¹¹¹⁸ (E) dcerk¹ with extensive degeneration, (F) DCERK transgene in dcerk¹ showing degeneration is rescued, and (G) dark-raised dcerk¹ showing no significant degeneration. Bars, 2 μm.

dcerk¹ Mutant Undergoes Light-Dependent Photoreceptor Degeneration.

To understand how CERK regulates ceramide metabolism in vivo and, in particular, to analyze whether it regulates photoreceptor structure and function, we generated mutants in DCERK. Two P element lines (CG16708^BG01100 and P{EPgy2}EY07850) inserted in the DCERK gene were identified in Fly Base. Because both lines expressed significant amounts of DCERK protein, we excised the P element CG16708^BG01100 (inserted in the first intron) using standard techniques. The genetic scheme used to mobilize it is described in Fig. S4A. A total of 640 independent excision lines were established of which 57 were lethal. The lethal lines were analyzed by transgenic rescue using a wild-type copy of DCERK, whereas the viable lines were analyzed by western analysis for decrease or lack of DCERK protein. This led to the isolation of dcerk¹ mutant, and western analysis of dcerk¹ fly head extracts showed less than 2% intact protein compared with control level (Fig. 1B). Reverse transcription–polymerase chain reaction (RT-PCR) analysis of dcerk¹ showed a significant reduction in the DCERK transcript level (Fig. 1C). PCR analysis suggested that ≈6 kb of the original P element left behind affected the DCERK transcript level (Fig. S4B). The above data show that dcerk¹ is a severe hypomorphic mutant of DCERK. dcerk¹ Flies are semilethal, and show fertility defects. All of the phenotypes were rescued by the introduction of a DCERK transgene into dcerk¹.

Because our earlier published results indicated that ceramide is an important regulator of photoreceptor viability, we analyzed photoreceptors of dcerk¹ for possible defects. As seen in Fig. 1E, dcerk¹ showed severe photoreceptor degeneration, cells appeared vacuolated with almost no intact rhabdomeres. A time course depicting progressive degeneration of the photoreceptors is shown in Fig. S4C. The degenerative phenotype was rescued by introducing a transgene expressing DCERK (Fig. 1F). Degeneration of dcerk¹ depended on light activation of phototransduction, as photoreceptors of dark raised flies did not show morphological signs of degeneration (Fig. 1G).

dcerk¹ Mutant Does Not Respond to Light, and Its PLC Level and Activity Are Severely Downregulated.

To test whether photoreceptor function was also defective in dcerk¹, we performed electroretinogram recordings (ERGs) from control, mutant, and rescued flies raised in the dark. Surprisingly, despite their intact morphology, mutants did not respond to the light stimulus (Fig. 2A). However dcerk¹ rescued with a DCERK transgene showed normal ERGs (Fig. 2 A and B). The lack of light response in dcerk¹ retina suggested that CERK regulates a crucial step in phototransduction. Mutations in NORPA, the critical effector of phototransduction, led to defective light-induced electrical responses, with null mutants being blind (16–18). Because dcerk¹ were blind, we checked the steady-state level of NORPA by immunoblotting using 1-day-old dcerk¹ maintained in the dark. We observed that NORPA protein was absent in dcerk¹ and that the expression of a DCERK transgene in the mutant rescued the NORPA level (Fig. 2C). In overexposed blots, a small amount of NORPA can be detected in 1-day-old flies (Fig. S5C). Real-time PCR analysis revealed that the amount of NORPA transcript was not different between w¹¹¹⁸ and dcerk¹, suggesting that DCERK primarily regulates NORPA posttranscriptionally (Fig. S5D). Although w¹¹¹⁸ retinal extracts showed high PLC activity, mutant extracts lacked this activity, and it was restored in the rescued flies (Fig. 2D). Our results thus far show that the lack of DCERK downregulates NORPA and that this results in loss of its activity and function in photoreceptor cells.

Fig. 2.

Severe downregulation of NORPA expression in dcerk¹ flies renders them unresponsive to light. (A) ERG recordings from dark-raised w¹¹¹⁸, dcerk¹, and rescued flies. The dcerk¹ do not respond to light, whereas rescued flies exhibit robust response. Cn bw flies were also used as control, as the rescued flies were in cn bw background. (B) Quantitative analysis of ERG recordings shows that the amplitude in the rescued flies is comparable to that in control flies. (C) Western blot of head extracts probed with NORPA antibody shows that its expression is severely reduced in dcerk¹ and rescued in dcerk¹ expressing DCERK transgene. (D) Retinal extracts of dcerk¹ do not show PLC activity, whereas activity of rescued flies is comparable to that of w¹¹¹⁸. Error bars denote standard deviation (n = 3).

Localization and Levels of Other Phototransduction Components Are Not Affected in dcerk¹ Mutant.

In Drosophila photoreceptors, INAD, a scaffolding protein, organizes signaling components including NORPA into a supramolecular complex to maximize the efficiency of phototransduction (19, 20). We tested whether DCERK specifically influenced NORPA or if it could affect the concentration or localization of other signaling components by western and immunofluorescence analyses. As seen in Fig. S6 A and B, no significant difference was observed between control and dcerk¹ for any of the proteins tested in 1-day-old flies. Immunostaining with a NORPA antibody revealed that its level was too low to be detected (Fig. S6B). By day 14, dark-raised dcerk¹ showed loss of TRP (Fig. S6C). Fig. S6D shows immunostaining of eyes with antibodies to DCERK and Rh1 in w¹¹¹⁸ and dcerk¹. In these sections, CERK is present in rhabdomeres and other membranous compartments of photoreceptors, overlying cone, and lens cells, whereas Rh1 shows specific rhabdomere staining of R1–R6 cells. That CERK can localize to rhabdomeres is also seen in thin sections of photoreceptors co-stained with antibodies to DCERK and INAD (Fig. S7A).

DCERK Regulates NORPA and Photoreceptor Function by Modulating the Ceramide Level.

To test whether the substrate, ceramide, or the product C-1-P was responsible for regulating NORPA, we measured their levels in lipid extracts from control and dcerk¹ by UPLC-MS/MS. The mutant flies showed a 250–300% increase in ceramide levels in all of the species measured and a modest decrease (40%) in C-1-P levels when compared with w¹¹¹⁸ (Fig. 3A). The levels of other sphingolipids such as ceramide phosphoethanolamine and glucosyl ceramide did not change significantly (<30%) between w¹¹¹⁸ and dcerk¹ extracts. It is likely that other pathways or proteins could contribute to the synthesis of C-1-P, as it is decreased by only 40% in the mutant. To determine whether the significant increase in ceramide level in dcerk¹ was crucial for downregulating NORPA, we overexpressed CDase in dcerk¹. CDase overexpression decreases the ceramide level by catalyzing the hydrolysis of ceramide to sphingosine and a fatty acid (10). Targeted overexpression of CDase in dcerk¹ photoreceptors rescued NORPA, suggesting that an increase in ceramide contributed to a decrease in NORPA level (Fig. 3 B and C). Lipid extracts from dcerk flies expressing CDase indeed showed a decrease in ceramide in all species measured except C-20 (Fig. S7B). ERGs carried out on dcerk¹ expressing CDase showed that overexpression of CDase rescues the lack of light response observed in dcerk¹ (Fig. 3D). CDase expression also rescued the retinal degeneration in dcerk¹ (Fig. 3E).

Fig. 3.

dcerk mutant accumulates ceramide, and reducing the ceramide level restores NORPA expression and function. (A) Total lipids extracted from dcerk¹ and w¹¹¹⁸ flies are subjected to UPLC-MS/MS. Individual ceramide and ceramide 1–phosphate species (with tetradecasphinganine backbone to which 18–, 20–, 22–, or 24–carbon chain fatty acids are attached) are measured and values are calculated for dcerk¹ relative to w¹¹¹⁸. (B) Expression of ceramidase (CDase) transgene rescues NORPA expression in dcerk¹. (C) Quantification of NORPA shows that the level is comparable in the CDase dcerk¹ rescued flies compared with the control. (D) ERGs showing CDase transgene rescues light response in dcerk¹. (E) CDase overexpression rescues retinal degeneration in dcerk¹.

Ceramide Levels Regulate the Localization and Level of PIP₂.

Because PLC acts on PIP₂, its localization and function is modulated by the lipid bilayer (21). We evaluated whether an increase in ceramide could alter global membrane properties such as bilayer fluidity. We measured fluorescence polarization as a measure of membrane fluidity, and these experiments showed that membrane fluidity did not change significantly in dcerk¹ (22, Fig. S7C). As bulk membrane properties like fluidity were not affected, we decided to test whether the membrane microenvironment was perturbed in dcerk¹ due to an increase in ceramide concentration. Because conventional procedures that involve ³Hinositol labeling followed by HPLC separation of radiolabeled phosphoinositides could not be adapted to flies, we measured the PIP₂ level in control and dcerk¹ by standardizing a coupled nonradioactive TLC separation of phosphoinositides followed by an enzyme-linked immunosorbent assay (ELISA) using a monoclonal antibody that recognizes PIP₂ with high specificity (23). dcerk¹ showed a 60% increase in the PIP₂ level compared with the control (Fig. 4A). NORPA null mutants lacking the major PLC used as a control showed an accumulation of PIP₂. We then analyzed whether increased ceramide could affect the distribution of PIP₂ using the PIP₂ antibody in thin sections of control and dcerk¹ mutant photoreceptor cells. As seen in Fig. 4B (labeled PIP₂), PIP₂ appeared to be enriched in doughnut-shaped clusters and spots at the rhabdomeres and their base. We then confirmed that these clusters recruit PLC by expressing PLCδ1PH-GFP in photoreceptors and staining thin sections of photoreceptors with a GFP antibody. In this widely used tool to visualize the distribution of PIP₂, the PH domain of PLCδ1, which specifically binds to PIP₂ at the plasma membrane, is fused to GFP (24). As seen in Fig. 4B (labeled GFP), in addition to rhabdomere staining, clusters were visible at the membrane. The plasma membrane was stained with cadherin in these sections (labeled Cadherin). Fig. 4B (labeled PIP₂+GFP) shows the overlay of PIP₂ and GFP, suggesting that PLCδ can co-localize to PIP₂ enriched clusters at the membrane. To test whether clustering is unique to PIP₂, we also expressed the PH domain of general receptor for phosphoinositides-1 (GRP1) fused to GFP which binds specifically to phosphatidylinositol-3,4,5 -P₃(PIP₃). As seen in Fig. 4C, punctate dots of PIP₃ staining can be seen at the rhabdomeres and their base but big clusters seen with PIP₂ are absent here. To analyze the distribution of PIP₂ in dcerk¹, thin sections of photoreceptors were stained with PIP₂ antibody. Unlike w¹¹¹⁸, PIP₂ was not clustered but fragmented in dcerk¹ and staining was diffuse (Fig. 4D and Fig. S7D). The diffuse staining in the mutant is better appreciated in experiments using quantum dot conjugated secondary antibody (Fig. S8A). norpA null mutant photoreceptors stained with PIP₂ antibody showed PIP₂ clusters (Fig. 4D and Fig. S8A). The ceramide level in norpA mutant would be unaltered and the microenvironment likely to be unperturbed. Morphometric analyses showed there were six to eight PIP₂ clusters in w¹¹¹⁸ and norpA null mutant ommatidium, whereas there were two to three in dcerk¹ (Fig. S7E). Our results suggest an increase in the ceramide level disrupts the organization of PIP₂-enriched areas at the plasma membrane.

Fig. 4.

Increased ceramide level alters PIP₂ membrane microenvironment in dcerk¹. (A) The PIP₂ level is elevated in dcerk¹ compared with w¹¹¹⁸. (B) Immunostaining of photoreceptors expressing PLCδPH GFP in w¹¹¹⁸ cn bw background with PIP₂, GFP and cadherin antibodies. PIP₂ clustering can be seen at the membranes, and the overlay shows that PLCδPH GFP localizes to these clusters. A cartoon of the staining shows PIP₂ in green, GFP in red, and overlay in yellow. Purple rhabdomeres are from overlay of cadherin in blue and GFP in red. (C) Immunostaining of photoreceptors expressing GRP1PH GFP in w¹¹¹⁸ cn bw background with GFP and cadherin antibodies. PIP₃ staining, unlike PIP₂, is not in large clusters. Last panel is a magnified view of one ommatidium; arrows show small dots of PIP₃ in the rhabdomere vicinity. (D) PIP₂ is disorganized in dcerk¹. Photoreceptors of w¹¹¹⁸, dcerk¹ and norp A null are immunostained with PIP₂ antibody. The w¹¹¹⁸ and norp A null photoreceptors show PIP₂ staining in clusters, which are less in dcerk¹.

NORPA That Is Not Membrane Associated Is Not Functional and Is Targeted for Degradation.

As the interaction of PLC with its substrate PIP₂ requires membrane association, and as PIP₂ organization is affected in dcerk¹, we examined the membrane association of NORPA in dcerk¹. We tested whether reducing ubiquitin-mediated proteosomal degradation in vivo would allow us to detect NORPA in dcerk¹. We used a dominant temperature-sensitive (DTS) mutation, DTS5, that affects the β6 proteosomal subunit (25). If NORPA is targeted for degradation in dcerk¹, then expressing DTS5 in dcerk¹ photoreceptors should result in restoration of NORPA. As seen in Fig. 5A, this was indeed the case; quantitative analysis of NORPA levels is shown in Fig. S8B. We fractionated head extracts from control and dcerk¹ expressing the DTS5 subunit into pellet and supernatant fractions and looked for NORPA by western analysis. These studies showed that in control almost all of the NORPA fractionates with the pellet underscoring its membrane localization, whereas most of the NORPA in dcerk expressing DTS5 was in the soluble fraction and hence cytosolic. To test whether this NORPA was functional, ERGs were carried out on dcerk¹ expressing DTS5. Whereas DTS5 flies showed a light response, DTS5 flies in dcerk¹ showed no response (Fig. 5C). Photoreceptor degeneration in dcerk¹ could also not be rescued in these flies (Fig. 5D). Taken together, a likely explanation for our results is that NORPA that is not membrane associated because of the disorganization of PIP₂ is unstable and is targeted for degradation.

Fig. 5.

NORPA is unable to bind efficiently to membranes in dcerk¹ and is degraded. (A) NORPA expression is detected in western analysis of dcerk¹ expressing DTS5. Lanes 1: w¹¹¹⁸; lane 2: dcerk¹; lane 3: UAS-DTS5;dcerk¹; lane 4: GMR-Gal4;dcerk¹; and lane 5: GMR-Gal4/UAS-DTS5;dcerk¹. (B) NORPA is unable to bind to membranes in dcerk¹. Fractionation experiments on head extracts from dcerk¹ expressing DTS5 subunit in photoreceptors (1) and control DTS5 subunit (2) show that a significant fraction of NORPA is soluble in dcerk¹ expressing DTS5, whereas it is in the pellet fraction in the control extract. (C) Amplitude values from ERGs show that DTS5 flies in dcerk¹ show no light response, indicating NORPA is not functional. (D) Ultrastructure of photoreceptors expressing DTS5 in dcerk¹ still degenerate.

Ceramide Affects PIP₂ Clustering and Its Partitioning into Liquid Ordered Membrane Domains.

Cellular membranes are not homogeneous but compartmentalized into liquid ordered (L_o), liquid disordered (L_d), and gel (L_β) phases. In the L_d phase, lipids diffuse freely, while in the L_β phase they are immobile and the L_o phase is an intermediate stage. The existence of PIP₂ compartmentalization in vivo has been rationalized by two hypotheses: (i) PIP₂ molecules can cluster independent of proteins or other lipids, for example, through hydrogen bonding between their head groups; or (ii) PIP₂ can be laterally sequestered by certain proteins such as MARCKS, GAP43 into cholesterol-rich membrane domains (26). We explored each of these scenarios in model membranes with defined components to understand how ceramide accumulation could influence PIP₂. To address protein-independent clustering, we prepared L_d supported bilayers made of dioleoylphosphatidylcholine (DOPC), 3% molar brain PIP₂, and a trace amount of fluorescent PIP₂ (BodTMRPIP₂) for visualization. Under these conditions, BodTMRPIP₂ forms clusters that can be visualized using fluorescence laser scanning microscopy (Fig. 6). Fluorescence correlation spectroscopy was used to show that membrane fluidity is not affected by the presence of the clusters (Fig. S8C). It was recently shown that in free-standing membranes, PIP₂ forms similar clusters that depend on the presence of divalent cations such as Ca²⁺ (27). Next, we evaluated the effects of incorporating ceramides in these preparations on the clustering and distribution of PIP₂, and image correlation spectroscopic (ICS) analysis was performed to quantify the degree of PIP₂ clustering (Fig. 6G). The spatial segregation of BodTMRPIP₂ depends on the presence of brain PIP₂ in the lipid mixture, as simple DOPC samples show no significant heterogeneities (Fig. 6A). The incorporation of either brain ceramide or short-chain C6 ceramide significantly decreases cluster formation (Fig. 6 E and F). C-1-P does not prevent PIP₂ clustering, whereas cholesterol seems to promote it (Fig. 6 C and D). From these experiments, it appears that the presence of ceramide has a direct effect on the clustering of PIP₂ in model membranes. A possible explanation for this phenomenon lies in the peculiar biophysical properties of the ceramide molecule. Ceramide has a small polar head (like cholesterol), and it is both donor and acceptor for hydrogen bonding (28). The former feature might explain why ceramide and cholesterol interact with PIP₂: the energetically unfavorable contact between water and their hydrophobic moieties are avoided by the large PIP₂ polar head (“umbrella effect”) (29). Nevertheless, the distinctive structural properties of ceramide (e.g., the amidic group that is absent in cholesterol and phospholipids) could influence the hydrogen bond network of surrounding molecules. Ceramide might act as a local chaotrope that perturbs the PIP₂–water–hydrogen bond network, thus allowing the electrostatic repulsion between PIP₂ molecules to disperse the clusters (30).

Fig. 6.

Ceramide affects protein-independent PIP₂ clustering and GAP-43P–induced PIP₂ partition in ordered domains. Fluorescence images of BodTMRPIP₂ (0.005% mol) in supported bilayers containing (A) DOPC, (B) DOPC + 3% BrPIP₂, (C) B + 5% Cer-1-P, (D) B + 5% cholesterol, (E) B + 5% C6 ceramide, and (F) B + 5% brain ceramide at room temperature. The presence of ceramide affects clustering of PIP₂. (G) Degree of clustering of PIP₂ for all samples described above. This quantity is defined as inverse of the cluster density, as calculated via ICS on an average of eight 23 × 23-μm² images for each sample. *E and F are statistically distinguishable from B with P < 0.01. **D is statistically distinguishable from B with P < 0.1. (H) Fluorescence image of a supported bilayer showing L_o domains in L_d phase. Red color refers to BodTMRPIP₂ and green to BodChol (0.01% mol). The L_o domains appear dark in both channels. (I) When the bilayer is treated with palmitoylated GAP-43P, the L_o domains are fragmented and show a significant degree of fluorescence in the red channel. (J and K) Analogous imaging of bilayers with ceramide-rich domains in L_d phase, without and with GAP-43P respectively. Ceramide-rich domains are dark in both channels. (L) Partition of BodTMRPIP₂ in L_o or ceramide-rich domains. This is calculated as the ratio between the average red signal in L_o (or ceramide) domains—dark in the green channel and the signal in the L_d phase—bright in the green channel. *I is statistically distinguishable from H with P < 0.01 and **K is statistically indistinguishable from sample J with P = 0.3. (Scale bars, 2 μm.)

Protein/lipid dependent PIP₂ clusters (scenario 2): A second line of thought regarding PIP₂ compartmentalization is that certain proteins, for example, the diacylated protein GAP-43, can bind PIP₂ and partition into small raft domains, thus offering a molecular basis for PIP₂ segregation (31). In this context, ceramide might play an important role inducing structural alterations in the lateral organization of the membrane bilayer. Nonphosphorylated long-chain ceramides can form specific membrane microdomains that can exist in gel phase (L_β) within raft domains of L_o phase (32). Such ceramide-rich domains are characterized by very tight lipid packing. To test whether such lateral segregation properties of ceramide within membrane bilayers could explain the loss of PIP₂ aggregation seen in vivo in CERK mutants, we examined the effect of increasing ceramide on PIP₂ organization in supported bilayer showing raft-like phase separation. BodTMRPIP₂ was added to the lipid mixture in trace amounts to study the organization of PIP₂, whereas a green fluorescent lipid analog (BodChol) was used as a marker for the L_d phase. In these preparations, PIP₂ localizes to the L_d phase of the membranes (Fig. 6H) because of the multiple unsaturation in the sn-2 hydrocarbon chain that would make PIP₂ partitioning into tightly packed domains energetically unfavorable. In this set of experiments, we monitored PIP₂ partition into ordered domains, either raft-like or ceramide-rich, as this could be a more relevant mechanism for PIP₂ segregation in vivo, and we did not focus on the Ca²⁺ dependent PIP₂ clustering only partially observable in the L_d phase because of washing with 15 mM ethylenediaminetetraacetic acid (EDTA). Following a published protocol (31), we verified that the addition of doubly palmitoylated peptide GAP-43P (i.e., the N terminus plus the basic effector domains of GAP-43) relocates a portion of the PIP₂ to the L_o (raft-like) phase of the membrane (Fig. 6I). Substitution of ≈50% of sphingomyelin with ceramide results in bilayers containing ceramide-rich domains in a matrix of L_d phase. Here again, PIP₂ is localized to the L_d phase (Fig. 6J). Interestingly, the addition of the acylated GAP-43P showed no significant change in the dimension of the ordered domains (Fig. 6K). Furthermore, very little of the PIP₂ relocates to these ceramide rich domains. Figure 6L shows quantification of the fluorescent intensity originating from PIP₂ in the ordered domains compared with that from PIP₂ in the disordered phase. These experiments suggest where PIP₂ clustering in vivo is determined by the partition of this lipid into small protein/lipid domains, the accumulation of ceramide, and the formation of highly ordered, ceramide-rich domains could effectively decrease PIP₂ segregation. The conclusions drawn from fluorescence experiments described above correlate well with the observed effects of increased ceramide in DCERK mutants on PIP₂ organization in photoreceptor cells.

Discussion

In this study, we show that DCERK regulates the ceramide level to maintain PLC and membrane organization of PIP₂ during phosphoinositide-mediated GPCR signaling in Drosophila. Our results are summarized in a model depicted in Fig. S9. Morphometric analysis of PIP₂ clusters suggests that there are still some clusters in dcerk¹; thus, the loss of NORPA in dcerk mutant may not be due only to the loss of PIP2 clustering. Although the effect of ceramide on PIP₂ is central, increased ceramide could affect other membrane properties that can downregulate NORPA or other proteins. Also, INAD is required for localization of NORPA and TRP is required for localization of INAD (20). Because TRP protein is affected over time in dcerk¹, additional interactions mediated by INAD and TRP could also contribute to NORPA stability. Photoreceptor degeneration seen in cerk mutants is also not simply caused by loss of NORPA protein, as it is more severe and sets in earlier than in norpA null mutants. There could be other effects of ceramide contributing to degeneration.

Recent clinical studies have identified mutations in the human ceramide kinase like (CERKL) gene in patients with autosomal recessive retinitis pigmentosa (33). No CERKL homolog has been identified in the Drosophila genome. Interestingly, DCERK shares 31% identity with human CERKL (Fig. S1), and possibly DCERK could perform some CERKL functions also. It would be worthwhile to test whether CERK regulates PLCβ4, the closest homolog of NORPA among mammalian PLCs, thereby participating in the mammalian visual process. Recent analyses of cerk^−/− mice revealed that CERK could function in cerebellar Purkinje cells (which are also enriched in PLCβ4) and in neutrophil homeostasis (34, 35). Because the mechanism by which mammalian CERK regulates these processes is not known, it would be interesting to test whether these functions are also mediated through PLC. Although our current understanding limits a direct co-relationship between mammalian CERK and PLCs, it is likely that similar ceramide-regulated microenvironments could operate in other phosphoinositide-dependent signaling such as insulin signaling in adipocytes.

In summary, our data show that modulation of the ceramide level by CERK regulates PIP₂ and PLCβ function in Drosophila. Because PIP₂ and PLC are fundamental components of GPCR signaling, uncovering their regulation by ceramide through CERK should lead to a better understanding of lipid regulation in signaling.

Methods

Genetic Screen and Isolation of dcerk Mutants.

The genetic scheme and method to generate jump-out lines from DCERK P-element are outlined in SI Materials and Methods. Five hypomorphic mutants including dcerk¹ were obtained from the screen.

Methods for electron microscopy and immunohistochemistry, assay for ceramide kinase activity, estimation of sphingolipids by mass spectrometry and PIP₂ by ELISA, and model membrane experiments are described in detail in SI Materials and Methods.