Protein Phosphatase 2Cϵ Is an Endoplasmic Reticulum Integral Membrane Protein That Dephosphorylates the Ceramide Transport Protein CERT to Enhance Its Association with Organelle Membranes*

Protein phosphatase 2Cϵ (PP2Cϵ), a mammalian PP2C family member, is expressed in various tissues and is implicated in the negative regulation of stress-activated protein kinase pathways. We show that PP2Cϵ is an endoplasmic reticulum (ER) transmembrane protein with a transmembrane domain at the amino terminus and the catalytic domain facing the cytoplasm. Yeast two-hybrid screening of a human brain library using PP2Cϵ as bait resulted in the isolation of a cDNA that encoded vesicle-associated membrane protein-associated protein A (VAPA). VAPA is an ER resident integral membrane protein involved in recruiting lipid-binding proteins such as the ceramide transport protein CERT to the ER membrane. Expression of PP2Cϵ resulted in dephosphorylation of CERT in a VAPA expression-dependent manner, which was accompanied by redistribution of CERT from the cytoplasm to the Golgi apparatus. The expression of PP2Cϵ also enhanced the association between CERT and VAPA. In addition, knockdown of PP2Cϵ expression by short interference RNA attenuated the interaction between CERT and VAPA and the sphingomyelin synthesis. These results suggest that CERT is a physiological substrate of PP2Cϵ and that dephosphorylation of CERT by PP2Cϵ may play an important role in the regulation of ceramide trafficking from the ER to the Golgi apparatus.

transmembrane protein with homologs widely distributed from yeast to human (1)(2)(3). Recently, evidence has accumulated that in mammalian cells VAPA participates in the regulation of inter-organelle transport of membrane lipids by recruiting lipid transfer proteins to the ER membrane. VAPA associates with a short, conserved peptide sequence termed the "two phenylalanines in an acidic tract" (FFAT) motif that is found in several lipid transfer proteins including ceramide transport protein CERT, oxysterol-binding protein, Opi1 protein, and PITP/Nir/ rdgB families (4 -11). VAPA is composed of two conserved domains, an amino-terminal immunoglobulin-like ␤ sheet responsible for FFAT motif binding and a carboxyl-terminal transmembrane domain (8). In addition to its role in recruiting FFAT motif-targeted proteins to ER membranes, VAPA has been proposed to function in vesicle trafficking (1,(12)(13)(14), in the organization of the microtubule network (10,15), and in the replication of hepatitis C virus RNA (16,17).
In mammalian cells ceramide is synthesized in the ER and transported to the Golgi apparatus where it is converted to sphingomyelin (SM). The ceramide transport protein CERT plays a key role in the ER-to-Golgi trafficking of ceramide (18 -20). CERT consists of several distinct domains including a Steroidogenic acute regulator-related lipid transfer (START) domain capable of specifically extracting ceramide from membrane, a pleckstrin homology (PH) domain that serves to target the Golgi apparatus by recognizing phosphatidylinositol 4-monphophatate, and a FFAT motif, which interacts with VAPA. In addition to these functional domains, CERT possesses a regulatory sequence referred to as a serine repeat (SR) motif between the PH domain and FFAT motif (21). The Ser/ Thr residues of the SR motif are phosphorylated in vivo. These phosphorylation sites match the typical consensus motif for protein kinase CK1 in which Ser-132 would serve as priming phosphorylation site for sequential phosphorylation of downstream Ser/Thr residues within the motif. Enhanced phosphorylation of these sites results in down-regulation of CERT activity; under such conditions ceramide transport from the ER to the Golgi apparatus diminishes as a result of repression of both the ceramide transfer activity of the START domain and the phosphatidylinositol 4-monophosphate binding activity of the PH domain (21). A loss of SM and cholesterol from the plasma membrane induces dephosphorylation of the SR motif, resulting in activation of CERT. However, the protein phosphatase(s) responsible for dephosphorylation of CERT has not yet been identified.
PP2C⑀ was originally identified by us as a negative regulator of stress-activated protein kinase signaling pathways (28,29). Ectopic expression of PP2C⑀ in mammalian cells represses the activity of transforming growth factor ␤-activated kinase 1 and apoptosis-regulating kinase 1, two mitogen-activated protein kinase kinase kinases. PP2C⑀ keeps these kinases in an inactive state in quiescent cells by associating with and dephosphorylating them. PP2C⑀ associates with transforming growth factor ␤-activated kinase 1 and apoptosis-regulating kinase 1 in quiescent cells, but the association was transiently suppressed in response to treatment of the cells with interleukin-1 and H 2 O 2 , respectively, which activate these respective kinases. On the basis of these results we proposed that PP2C⑀ regulates transforming growth factor ␤-activated kinase 1 and apoptosis-regulating kinase 1 pathways by a common regulatory mechanism (28,29).
PP2C⑀ has a unique hydrophobic region composed of 60 amino acids at the amino terminus whose function is not yet known. During the course of studies to elucidate the functional role of the amino-terminal region of PP2C⑀, we noticed that PP2C⑀ is an ER resident integral membrane protein and identified VAPA as a binding partner of PP2C⑀ on ER. Furthermore, we obtained evidence suggesting that PP2C⑀ regulates CERT function through dephosphorylation of its SR motif.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes and other modifying enzymes used for DNA manipulation were obtained from New England Biolabs (Beverly, MA). Lipofectamine 2000 was purchased from Invitrogen. Glutathione Sepharose-4B, protein G-agarose beads, polyvinylidene difluoride membrane, ECL plus kit and L-[U-14 C]serine were obtained from GE Healthcare. Horseradish peroxidase-labeled secondary antibody was obtained from Cell Signaling (Beverly, MA). Anti-hemagglutinin (HA), anti-GST, anti-GS28, and anti-protein disulfide isomerase (PDI) antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-GM130 antibody was purchased from BD Biosciences. Alexa 488-conjugated anti-rabbit IgG and Alexa 568-conjugated anti-mouse IgG were obtained from Molecular Probes (Eugene, OR). Complete protease mixture was obtained from Roche Applied Science. Methyl-␤-cyclodex-trin was obtained from Sigma-Aldrich. All other reagents were purchased from Wako Pure Chemicals (Osaka, Japan).
Production of Anti-PP2C⑀ Antiserum-Polyclonal antibodies recognizing PP2C⑀ were raised in rabbits against a mixture of the peptides QHLQDYEKDKENSVLC and CPLSHDHK-PYQLKERKR (corresponding to residues 149 -163 and 225-241 of mouse PP2C⑀, respectively). Antibodies were affinitypurified on NSH-Sepharose covalently coupled to these peptides.
Yeast Two-hybrid Screening-A cDNA encoding a dominant negative mutant of mouse PP2C⑀(D302A) was cloned into the pGBK-T7 to produce "bait" vector. This construct was used to screen a human brain cDNA library in the pACT2 vector. The dominant negative mutant was used because it was expected to associate with its substrate more stably than the wild type in the cells (28). In a screen of 1 ϫ 10 6 library clones, 5 independent clones encoding human VAPA were isolated.
Plasmid Constructs-Human VAPA and human CERT cDNAs were obtained by reverse transcriptase-PCR from total RNA of HEK293 cells. Plasmids expressing GST-VAPA and HA-CERT were prepared by inserting these cDNAs into pCX-GST-MS and pcDNA-HA-MS vectors, respectively. Plasmids expressing PP2C⑀ and PP2C⑀-FLAG were constructed by inserting the PP2C⑀ cDNA into pcDNA-3 and pcDNA-MS-FLAG vectors, respectively. Plasmids expressing ⌬N-PP2C⑀ and FLAG-⌬N-PP2C⑀ were constructed by inserting the PP2C⑀ cDNA encoding amino acids 58 -360 into pcDNA-3 and pcDNA-FLAG-MS vectors, respectively. Site-directed mutagenesis was carried out to generate deletion mutants of PP2C⑀ and VAPA using the directed PCR method. To produce GFP fusion proteins, pEGFP-N3-M vector was used in which an inherent initiating ATG codon was mutated to GTG to prevent non-fused GFP protein from being produced.
Cell Culture and Transfection-HEK293 cells and HeLa cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% (v/v) fetal bovine serum. At ϳ90% confluency, the cells were transfected using Lipofectamine 2000. The total amount of DNA used for transfection was 1.6 g per well of a 12-well plate. After transfection, the cells were cultured for 24 -48 h before harvest. Cells transfected with the indicated expression plasmids were washed twice with phosphate-buffered saline (PBS) and lysed with ice-cold lysis buffer containing 50 mM Tris-HCl, pH 7.5, 1% (v/v) Triton X-100, 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1 mM sodium orthovanadate, 50 mM NaF, 10 mM ␤-glycerophosphate, 5 mM sodium pyrophosphate, and a Complete protease inhibitor mixture. Cell lysate containing 200 g of protein was incubated for 0.5 h with glutathione-Sepharose beads (5 l). After washing the beads 5 times with lysis buffer, the bound protein was subjected to 10% (w/v) SDS-PAGE and then transferred onto polyvinylidene difluoride membranes. The membranes were incubated with primary antibody for 1 h at 25°C followed by incubation with horseradish peroxidase-conjugated secondary antibody for 0.5 h at 25°C and developed by chemiluminescence using ECL plus kit. Dilution of antibody was 1:500 for anti-GST antibody, 1:1000 for anti-PDI, anti-HA, and anti-PP2C⑀ antibodies, and 1:2000 for horseradish peroxidase-conjugated secondary antibody. The intensity of the band was quantified by imaging analyzer (FLA-7000, Fuji Film, Tokyo, Japan).
Subcellular Fractionation of Mouse Brain-Mouse brain (1.2 g) was homogenized in 3 volumes of solution A (0.32 M sucrose, 1 mM NaHCO 3 , 1 mM MgCl 2 , and 0.5 mM CaCl 2 ), and the homogenate was centrifuged at 1400 ϫ g for 10 min to eliminate nuclei and cell debris. The supernatant was centrifuged at 5000 ϫ g for 10 min. The precipitates containing mitochondria were resuspended in solution A. The supernatant was centrifuged at 100,000 ϫ g for 90 min to produce cytosol and microsomal fraction. The microsomal fraction was homogenized in solution A, and membrane proteins were extracted by the addition of 2% (v/v) Triton X-100.
Isolation of Microsomal Membrane from Cultured Cells-The membrane fraction of cultured cells was prepared as described by Nohturfft et al. (30). Briefly, cells were scraped into ice-cold PBS and resuspended in buffer B (10 mM HEPES-NaOH, pH 7.5, 0.25 M sorbitol, 10 mM potassium acetate, 1.5 mM magnesium acetate , and Complete protease inhibitor mixture). The suspension was passed through a 23-gauge needle 25 times and centrifuged at 1000 ϫ g for 5 min. The resulting supernatant was centrifuged at 16,000 ϫ g for 20 min. The pellet was resuspended in buffer C (50 mM HEPES, pH 7.5, 0.25 M sorbitol, 70 mM potassium acetate, 2.5 mM magnesium acetate, 5 mM EGTA, and protease inhibitor) and centrifuged again at 16,000 ϫ g for 3 min. The resulting precipitates were resuspended in Buffer C to obtain the microsomal fraction.
Fractionation by Triton X-114-Triton X-114 was added to the microsomal fraction at 4°C to a final concentration of 2% (v/v) to solubilize the membrane proteins. The Triton X-114treated sample was layered over a cushion of 0.25 M sucrose containing 0.06% (v/v) Triton X-114, incubated at 30°C for 5 min, and centrifuged for 3 min at 300 ϫ g. The supernatant and detergent layers were subjected to SDS-PAGE.
Proteinase K Protection Assay-The microsomal fraction was incubated for 60 min at 30°C with 0.1 mg/ml proteinase K with or without the addition of Triton X-100 to a final concentration of 1% (v/v). Proteolysis was stopped by adding 4 mM phenylmethylsulfonyl fluoride, and the samples were analyzed by SDS-PAGE.
Sucrose Density Gradient Centrifugation-HEK293 cells (4 ϫ 10 7 ) were homogenized with a buffer containing 0.1 M KH 2 PO 4 , pH 6.8, 5 mM MgCl 2 , 5 mM dithiothreitol, and 0.5 M sucrose, passed through a 23-gauge needle 25 times, and centrifuged at 1000 ϫ g for 10 min at 4°C. The postnuclear supernatant was fractionated over sucrose gradients as described for bovine brain (31). In brief, the postnuclear supernatant (1.8 ml) was overlaid on top of 0.86 M sucrose (1 ml) and 1.25 M sucrose (1 ml) step gradient. The gradient was centrifuged in a Hitachi RPS-56T rotor at 100,000 ϫ g for 90 min at 4°C. The 0.86/1.25 M interface was collected, adjusted to 1.6 M sucrose, and placed in the bottom of the tube followed by overlaying with 1.25 M (0.7 ml), 1.0 M (0.7 ml), 0.86 M (0.7 ml) and 0.5 M (0.7 ml) sucrose solutions. The gradient was centrifuged in a Hitachi RPS-56T rotor at 100,000 ϫ g for 2.5 h at 4°C. Fractions were collected from the bottom of the tube and analyzed by SDS-PAGE.
Indirect Immunofluorescence Microscopy-Cells seeded on coverslips coated with poly-D-lysine were fixed with 3% (w/v) paraformaldehyde in PBS for 10 min and then permeabilized with PBS containing 0.05% (v/v) Triton X-100 for 30 min at 25°C. The coverslips were blocked by incubation with 3% (w/v) bovine serum albumin in PBS for 30 min followed by 1 h of incubation with primary antibody (dilutions of antibody were 1:1000 for anti-PDI and anti-GM130 antibodies and 1:100 for anti-HA antibody). Cells were then incubated with Alexa 488conjugated anti-rabbit IgG (1:2000) and Alexa 568-conjugated anti-mouse IgG (1:2000) for 1 h. The immunostained samples were mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, CA) and visualized by a confocal microscopy (Carl Zeiss LSM5, Oberkochen, Germany). All experiments were performed at least three times with the representative results presented.
Knockdown of PP2C⑀ mRNA by Short Interference RNAs (siRNAs)-siRNAs against human PP2C⑀ were synthesized by Invitrogen. Nucleotide sequence of the sense strands of ⑀-3 and ⑀-4 are UAGCAGAGCAAUCAAACACGUUGUG and UGAUCAUGAGACAAAGGAAUAGCGU, respectively. The oligonucleotide (50 pmol) and plasmid (1.6 g) were incubated with Lipofectamine 2000 separately and mixed immediately before transfection to HeLa cells. Twenty-four hours after the transfection, total RNA was isolated, and the amount of PP2C⑀ mRNA was determined by real time PCR system (Takara TP-800, Tokyo, Japan).
Metabolic Labeling of Lipid with [ 14 C]Serine-The metabolic labeling of HeLa cells with radioactive serine was performed as described previously (11). HeLa cells transfected with siRNAs were incubated with 10 mM methyl-␤-cyclodextrin for 30 min and then incubated in 1.5 ml of serum free medium containing 37 kBq of L-[U-14 C]serine at 37°C for 2 h. Labeled lipids extracted from the cells were separated on TLC plate (methyl acetate, n-propanol, chloroform, methanol, 0.25% CaCl 2 ϭ 25/25/25/10/9) and quantified by FLA-7000 imaging analyzer.

RESULTS
PP2C⑀ Is Localized in the ER-Unlike PP1, PP2A, and PP2B, whose subcellular localization and activities are controlled by separate regulatory subunits, the members of PP2C family are monomeric proteins and, therefore, are thought to contain distinct domains that influence the subcellular localization or function. In this context we noticed that PP2C⑀ has an aminoterminal non-catalytic region composed of about 60 amino acids in addition to the carboxyl-terminal catalytic domain composed of about 300 residues. Because the amino-terminal region was hydrophobic, we speculated that this region might function as a membrane association domain. To test this possibility we transfected HEK293 cells with a plasmid for either full-length PP2C⑀ or PP2C⑀ lacking the amino-terminal 57 amino acids (⌬N-PP2C⑀) and separated the postnuclear supernatant of the cell extracts into cytosolic and membranous fractions. Western blot analysis using anti-PP2C⑀ antiserum revealed that the full-length PP2C⑀ localized to the membrane fraction, whereas ⌬N-PP2C⑀ was recovered in the cytosolic fraction (Fig. 1a). These results suggest that PP2C⑀ is a membrane-associated protein and that the amino-terminal hydrophobic region is required for its membrane association.
To determine the subcellular localization of endogenous PP2C⑀, we separated the postnuclear supernatant of mouse brain extracts into cytosolic, mitochondrial, and microsomal fractions. Western blot analysis demonstrated that endogenous PP2C⑀ was mainly localized in the microsomal fraction ( Fig.  1b). We then carried out a sucrose density gradient centrifugation of the homogenates prepared from HEK293 cells expressing PP2C⑀. As shown in Fig. 1c, PP2C⑀ was co-fractionated with PDI, an ER resident protein, but not with GS28, a Golgi marker protein. Furthermore, PP2C⑀-GFP fusion proteins expressed in HEK293 cells exhibited a diffuse reticular pattern that overlapped with staining of PDI when observed by confocal microscopy (Fig. 1d). These results suggest that PP2C⑀ is an ER resident protein.
PP2C⑀ Is a Type 1 Integral Membrane Protein of the ER-We next determined whether PP2C⑀ was an ER transmembrane protein, a peripheral membrane protein that associates with transmembrane proteins or a luminal protein that resides in the lumen of ER as a soluble protein. To this end, the microsomal fraction prepared from HEK293 cells expressing PP2C⑀ was treated with a high concentration of salt, alkali, or Triton X-100. Western blot analysis indicated that, whereas PP2C⑀ remained associated with the membrane fraction under a high salt concentration or alkaline conditions (Fig. 2a, lanes 2 and 3), it was solubilized upon treatment with Triton X-100 (Fig. 2a,  lane 4), suggesting that PP2C⑀ is either a transmembrane or luminal protein. To discriminate between these two possibilities, the microsomal fraction prepared from HEK293 cells expressing PP2C⑀ was treated with Triton X-114, and the solubilized proteins were subjected to phase separation at 30°C. PP2C⑀ was recovered in the detergent phase, which contains transmembrane proteins, but not in the hydrophilic phase, which is expected to contain luminal proteins (Fig. 2b). Thus, PP2C⑀ is likely to be a transmembrane (TM) protein.
To determine whether the carboxyl-terminal catalytic domain of PP2C⑀ projects into the cytosol or the ER lumen, we treated the microsomal fraction prepared from HEK293 cells expressing PP2C⑀ with proteinase K with or without Triton X-100. Our expectation was that, if the carboxyl terminus of PP2C⑀ faces the cytosol like SREBP1, it would be digested by proteinase K even in the absence of Triton X-100. However, PP2C⑀ would be digested by proteinase K only in the presence of Triton X-100 if its carboxyl-terminal region projects into ER lumen like PDI because it is protected from digestion by proteinase K. We found that PP2C⑀ was readily digested by proteinase K irrespective of the presence of Triton X-100, as was observed with SREBP1. This finding suggests that PP2C⑀ exists as a transmembrane protein whose carboxyl-terminal catalytic domain faces the cytoplasm (Fig. 2c).
To further confirm the conclusion described above, we carried out the protease protection assay using the microsomal fraction prepared from the cells expressing GST-PP2C⑀ in which GST tag was fused to the amino terminus of PP2C⑀.

JOURNAL OF BIOLOGICAL CHEMISTRY 6587
Although the majority of GST-PP2C⑀ expressed in the cells was recovered in the cytosolic fraction (data not shown), a significant amount of GST-PP2C⑀ was localized to the microsomal fraction (Fig. 2d, lanes 1 and 4). When immunoblotting was carried out using the anti-PP2C⑀ antibody after the protease digestion of the microsomal fraction, the bands of GST-PP2C⑀ were found to disappear upon the protease treatment (Fig. 2d,  lanes 5 and 6). In contrast, a faster migrating band whose molecular mass was 27 kDa appeared in accordance with the disappearance of the full-length GST-PP2C⑀ when stained with anti-GST antibody, indicating that GST portion (ϳ26.5 kDa) was protected from digestion by protease (Fig. 2d, lanes 2 and  3). These results suggest that the GST portion of the expressed GST-PP2C⑀ was localized in the microsomal lumen, whereas the carboxyl-terminal portion of the GST-PP2C⑀ existed outside of the microsome, supporting the conclusion that the amino-terminal region of PP2C⑀ functions as the TM domain. We were interested in identifying the TM domain of PP2C⑀. To this end we expressed a series of amino-terminal deletion mutants of PP2C⑀ (Fig. 3a, ⌬1-⌬7) in HEK293 cells and examined their subcellular localization. Of these mutants, ⌬1, ⌬5, ⌬6, and ⌬7 were localized only to the membrane fraction (Fig.  3b, lanes 2, 6, 7, and 8). Two mutants (⌬2 and ⌬4) distributed equally to membranous and cytosolic fractions (Fig. 3b, lanes 3  and 5). Most of mutant ⌬3, which lacked amino acids between 26 and 45, exhibited cytosolic localization (Fig. 3b, lane 4). Because the region between amino acids 26 and 45 is predicted to form an ␣-helix (data not shown) and is rich in hydrophobic amino acids, it may function as a TM domain (Fig. 3a). To directly show that the amino-terminal region of PP2C⑀ is involved in the membrane localization, we constructed three GFP fusion proteins, ⌵1-, N2-, and N3-GFP, which were composed of the amino-terminal regions of PP2C⑀ with three different lengths fused to GFP (Fig. 3a). These proteins were expressed in HEK293 cells, and their subcellular localization was determined. All fusion proteins were localized to the membrane fraction (Fig. 3c), indicating that amino-terminal region containing amino acids between 26 and 45 is involved in the membrane localization. The amino acid sequence surrounding the TM domain of mouse PP2C⑀ is conserved in orthologs in insect and sea urchin genomes, suggesting that membrane localization of PP2C⑀ is widely conserved (Fig. 3d).
PP2C⑀ Interacts with VAPA in Vivo-To identify the substrate(s) or the protein(s) responsible for regulation of PP2C⑀ activity, we carried out a yeast two-hybrid screen for interaction partners. The coding region of full-length PP2C⑀(D302A) (a mutant of PP2C⑀ deficient in phosphatase activity) was fused to the GAL4 DNA binding domain and used as bait. In a screen of 1 ϫ 10 6 human brain library clones, VAPA was identified as an interacting partner of PP2C⑀. VAPA interacted with PP2C⑀ but not with GAL4 DNA binding domain, PP2C␣, PP2C␤, PP2C␦, or PP2C in the yeast two-hybrid system, suggesting that the interaction between VAPA and PP2C⑀ is specific (data not shown). To examine whether this interaction also occurs in mammalian cells, a GST fusion protein of VAPA was expressed with PP2C⑀-FLAG in HEK293 cells, and the lysates from the cells were subjected to a GST pulldown assay. As shown in Fig. 4a, VAPA in fact interacted with PP2C⑀ but not with PP2C␤, suggesting that the interaction between PP2C⑀ and VAPA in cells is specific. ⌬N-PP2C⑀ did not interact with VAPA, indicating that the amino-terminal non-catalytic region of PP2C⑀ is required for this interaction (Fig. 4a, lane 3 in the upper panel).
We next attempted to identify the region of PP2C⑀ that is required for the interaction with VAPA. GFP fusion proteins containing full-length PP2C⑀ (FL-GFP) or various lengths of the amino-terminal region of PP2C⑀ (N1-GFP, N2-GFP, and N3-GFP in Fig. 3a) were expressed in HEK293 cells, and interaction with GST-VAPA was examined. Of these GFP fusion proteins, only FL-GFP and N3-GFP (the fusions bearing residues 1-68 of PP2C⑀) interacted with GST-VAPA (Fig. 4b, lanes  4 and 5 in the upper panel). These results suggested that the hydrophilic region adjacent to the TM domain is also required for the interaction of PP2C⑀ with VAPA (see Fig. 3a).  The green box shows GFP protein that is fused to the amino-terminal region of the PP2C⑀ deletion mutants. b, deletion mutants of PP2C⑀ were transiently expressed in HEK293 cells, and subcellular localization was determined as in Fig. 1a. c, GFP proteins fused to the various length of the amino-terminal region of PP2C⑀ (N1-, N2-, and N3-GFP) and full-length PP2C⑀ (FL-GFP) were expressed in HEK293 cells, and subcellular localization was determined. C and M indicate the cytosolic fraction and membrane fraction, respectively. d, amino acid sequence alignment (red, acidic; blue, basic; green, hydrophobic; yellow, polar) of the transmembrane domain of PP2C⑀ of mouse, insect (Drosophila melanogaster), and sea urchin (Strongylocentrotus purpuratus).
To identify the region of VAPA that interacts with PP2C⑀, we constructed GST fusion proteins containing either a VAPA mutant that lacks its TM domain (GST-VAPA-1) or the TM domain of VAPA without the rest of the protein (GST-VAPA-2) (Fig. 4c). GST-VAPA-2 but not GST-VAPA-1 was able to interact with PP2C⑀, suggesting that the TM domain is responsible for the interaction with PP2C⑀ (Fig. 4d).
PP2C⑀ Dephosphorylates CERT in a VAPA-dependent Manner-Recently, we have reported that the ceramide transport protein CERT interacts with VAPA in the ER (11). This interaction is required for CERT to extract ceramide from the ER before transportation of ceramide to the Golgi apparatus. We also showed that the SR motif in CERT is phosphorylated at multiple Ser/Thr residues and that hyperphosphorylation of these sites results in down-regulation of ceramide trafficking function of CERT (21). We, therefore, were interested in determining whether PP2C⑀ associated with VAPA regulates CERT function through dephosphorylation. We first tested the possibility that PP2C⑀ dephosphorylates CERT in vivo. HA-CERT was expressed in HEK293 cells with or without GST-VAPA in the presence or absence of co-expressed PP2C⑀-FLAG, PP2C⑀(D302〈)-FLAG which is a phosphatase activity-deficient mutant of PP2C⑀ or FLAG-PP2C␤. As judged by a shift in SDS-PAGE mobility, PP2C⑀-FLAG, but not PP2C⑀(D302A)-FLAG, caused dephosphorylation of HA-CERT when GST-VAPA was co-expressed in the cells (Fig. 5, lanes 3, 4, 7, and 8 in  the upper panel). In contrast, HA-CERT remained phosphorylated when it was co-expressed only with PP2C⑀-FLAG (Fig. 5,   lanes 5 and 6 in the upper panel). FLAG-PP2C␤ did not dephosphorylate HA-CERT even in the presence of GST-VAPA (Fig.  5, lanes 9 and 10 in the upper panel). These results suggest that PP2C⑀ specifically dephosphorylates CERT in a manner depending on VAPA expression.
Expression of PP2C⑀ Induces Redistribution of CERT to the Golgi Apparatus-We have previously shown that the treatment of cells with sphingomyelinase resulted in redistribution of CERT from the cytoplasm to the Golgi apparatus in accordance with its dephosphorylation (21). We, therefore, examined the effect of expression of PP2C⑀ on subcellular localization of  CERT. HA-CERT and GST-VAPA were transiently expressed in HeLa cells, and the intracellular distribution of HA-CERT was analyzed by indirect immunostaining with an anti-HA antibody and an antibody against GM130, a Golgi marker protein.
Although HA-CERT was distributed throughout the cytoplasm when expressed only with GST-VAPA (Fig. 6a), co-expression of PP2C⑀ greatly redistributed HA-CERT to the Golgi apparatus (Fig. 6d). CERT(S132A), which was previously shown to act as a mimic of the dephosphorylated form of CERT, was localized to the Golgi region even in the absence of PP2C⑀, as we demonstrated previously (Fig. 6g) (21). These results suggest that PP2C⑀ enhances phosphatidylinositol 4-monophosphate binding activity of CERT through its dephosphorylation. PP2C⑀ Enhanced the Interaction between CERT and VAPA-We next examined whether dephosphorylation of CERT by PP2C⑀ affected the interaction between CERT and VAPA. When HEK293 cells were transfected with expression plasmids of GST-VAPA and HA-CERT, no interaction between these proteins was observed (Fig. 7a, lanes 1 and 2 in the top panel). In contrast, expression of PP2C⑀-FLAG in combination with GST-VAPA and HA-CERT caused an increase in the interaction between GST-VAPA and HA-CERT, in accordance with dephosphorylation of CERT (Fig. 7a, lanes 3 and 4 in the top and middle panels). Neither interaction between CERT and VAPA nor dephosphorylation of CERT was observed when PP2C⑀(D302A) (Fig. 7a, lanes 5 and 6 in the top and middle panels) or PP2C␤ (Fig. 7b, lanes 7 and 8 in the top and middle panels) were co-expressed. These results suggest that PP2C⑀ enhances the interaction between CERT and VAPA through dephosphorylation of CERT. Interestingly, although expression of ⌬N-PP2C⑀ caused a mobility shift of CERT protein to a similar extent as full-length PP2C⑀ (Fig. 7b, lanes 5 and 6 in the

HA-CERT GM130 DIC
WT WT S132A S132A WT WT To examine whether the enhanced interaction between CERT and VAPA by PP2C⑀ is mediated through dephosphorylation of the SR motif, we conducted binding assays with the S132A mutant of CERT. As shown in Fig. 7d, the interaction between CERT(S132A) and VAPA was observed even in the absence of PP2C⑀. Co-expression of PP2C⑀ did not further increase CERT(S132A)-VAPA interaction (Fig. 7d), whereas the interaction between wild-type CERT and VAPA was dependent on PP2C⑀ (Fig. 7c). These results suggest that PP2C⑀ regulates the interaction between CERT and VAPA through dephosphorylation of the SR motif in CERT.
Knockdown of Expression of PP2C⑀ Attenuates the Interaction between VAPA and CERT-We were interested in determining whether endogenous PP2C⑀ is indeed responsible for dephosphorylation of CERT. To this end we used RNA interference to knock down the expression of PP2C⑀ in vivo. Although the interaction of CERT and VAPA is low in HEK293 cells unless PP2C⑀ is co-expressed, a significant interaction between these proteins is observed in HeLa cells even in the absence of coexpression of PP2C⑀. We, therefore, used HeLa cells in knockdown experiments. Transfection of HeLa cells with two distinct siRNAs directed against PP2C⑀ mRNA (⑀-3 and ⑀-4) resulted in a 70 and 75% reduction in PP2C⑀ expression 24 h after transfection, respectively (Fig. 8a). At the same time point, interaction between HA-CERT and GST-VAPA was decreased by 35 and 47% in the cells transfected with ⑀-3 and ⑀-4, respectively (Fig. 8b, lane 3-6).
We finally examined whether knockdown of PP2C⑀ affects SM synthesis in intact cells. We have previously shown that the treatment of cells with cholesterol adsorbent methyl-␤-cyclodextrin stimulated dephosphorylation of CERT (21). We, therefore, employed this experimental condition for the knockdown study. The amount of radioactivity incorporated into SM during 2 h of labeling in HeLa cells were reduced by 20 and 40% by transfection with two siRNAs, ⑀-3 and ⑀-4, respectively, without affecting the rate of ceramide synthesis (Fig. 8c). These results support the conclusion that PP2C⑀ is involved in regulation of ceramide trafficking in the cells.

DISCUSSION
PP2C⑀ Is a Transmembrane Protein-In the present study we provide evidence that PP2C⑀ is an ER resident type I integral membrane protein, with a carboxyl-terminal catalytic domain facing the cytoplasm and a short amino-terminal hydrophobic domain localizing inside the membrane. Genomic analysis revealed that the amino acid sequence surrounding the transmembrane domain of mammalian PP2C⑀ is conserved in distantly related organisms, such as insects and sea urchins (Fig.  3d), supporting the notion that membrane localization is important for the in vivo function of PP2C⑀.
We previously proposed that PP2C⑀ was translated beginning at the fourth Met codon of the open reading frame to produce a 303-amino acid polypeptide based on the results of the experiments using in vitro translation system (28). To examine whether PP2C⑀ was translated from fourth Met codon in vivo, we examined the size of the endogenous protein.
Endogenous PP2C⑀ in mouse brain co-migrated with recombinant protein translated from the first Met codon, which consists of 360 amino acids (result not shown). Furthermore, the fact that amino acid sequence surrounding the transmembrane domain of mouse PP2C⑀ is conserved in distantly related organisms supports the idea that the amino-terminal non-catalytic region is translated in these animals. In contrast, the sequence surrounding fourth Met codon of mouse PP2C⑀ is not conserved among other organisms (data not shown). Collectively, these lines of evidence suggest that endogenous PP2C⑀ is actually longer than was previously thought, although we cannot rule out the possibility that the shorter polypeptide, which is composed of 303 amino acids, also exists in vivo. Physiological Significance of Dephosphorylation of CERT by PP2C⑀-In this study we provide evidence that PP2C⑀ dephosphorylates CERT in vivo. Our conclusion that CERT is a physiological substrate of PP2C⑀ is based on the following observations. First, PP2C⑀ dephosphorylated CERT in a manner depending on VAPA expression, whereas PP2C␤ did not dephosphorylate CERT in vivo, even in the presence of co-expressed VAPA. Second, dephosphorylation of CERT by PP2C⑀ caused redistribution of CERT from the cytoplasm to the Golgi apparatus. Third, dephosphorylation of CERT by PP2C⑀, but not PP2C␤, also caused the enhanced interaction between CERT and VAPA. Fourth, knockdown of PP2C⑀ mRNA by siRNA reduced the interaction between CERT and VAPA. These lines of evidence together with the observation that PP2C⑀ specifically associated with VAPA both in the yeast twohybrid system and in mammalian cells strongly suggest that VAPA is a physiological binding partner of PP2C⑀ in the ER. Considering the fact that dephosphorylation of CERT by PP2C⑀ occurred in a VAPA-dependent manner and that PP2C⑀ and CERT were found to bind to different regions of VAPA, we propose that VAPA may act as a scaffold protein that interacts with PP2C⑀ and CERT simultaneously and stimulates dephosphorylation of CERT by PP2C⑀.
Previously, we showed that CERT is phosphorylated at multiple sites in the SR motif and that the phosphorylation of these sites results in down-regulation of ceramide trafficking activity of CERT through inhibition of the START and PH domains (21). Our finding that PP2C⑀ dephosphorylates CERT and that dephosphorylation enhanced the interaction of CERT with both ER and Golgi membranes suggest that PP2C⑀ functions as an activator of ceramide transport (Fig. 9). This conclusion is supported by the fact that knockdown of PP2C⑀ expression reduced the level of ceramide-trafficking SM synthesis. In this context we previously suggested that dephosphorylation of the SR motif was required for the association of CERT with phosphatidylinositol 4-monphophatate in the Golgi apparatus but not with VAPA in the ER. Our conclusion was based on the observation that CERT(S132A) and CERT(10E) in which all the Ser/Thr residues of the SR motif were replaced by Glu to mimic the phosphorylated state had similar affinity for VAPA as wildtype CERT (21). Although the reason for the discrepancy between our previous and present studies is not known, the difference in experimental conditions for the pulldown assay may account for it. Although the cells were treated with chemical cross-linker before the pulldown assay in the previous study, no cross-linking reagent was used in this study. A pulldown assay without cross-linking pretreatment may detect only strong and stable interaction between CERT and VAPA. Alternatively, the conformational change induced by replacement of Ser/Thr residues with Glu only partially mimicked that induced by phosphorylation of these sites. This partial conformational change may not have been enough to inhibit the association between the mutant CERT and VAPA, although it was sufficient to suppress the interaction between the mutant CERT and Golgi. Further studies are required to answer this question.
One unexpected observation was that the expression of ⌬N-PP2C⑀ caused a mobility shift of CERT protein to the same extent as wild-type PP2C⑀, although in contrast to the expression of wild-type PP2C⑀, ⌬N-PP2C⑀ expression had only a modest stimulatory effect on the association of CERT with VAPA (Fig. 7). These results suggest that the phosphorylation site(s) in SR motif of CERT responsible for the regulation of interaction with VAPA differs from that responsible for the mobility shift on SDS-PAGE and that localization to ER membrane is required for PP2C⑀ to dephosphorylate the former site(s) precisely. Alternatively, association with PP2C⑀ may be required for VAPA to keep an appropriate conformation for binding to the dephosphorylated CERT.
CERT-mediated ceramide trafficking is thought to occur at the narrow cytoplasmic gaps termed the ER-Golgi membrane contact site. Two models have been proposed to depict this process (19,32). In the "neck-swinging model," CERT might simultaneously associate with both ER and Golgi, and the ceramide is transferred by the START domain by its oscillating motion. In contrast, the "short-distance shuttle model" proposes that CERT might quickly shuttle between the two organelle membranes. Our observation that dephosphorylation of CERT by PP2C⑀ enhances its affinity for both the ER membrane and the Golgi apparatus would support the neck-swinging model (Fig. 9).
It has been reported that VAPA interacts with hepatitis C virus nonstructural protein 5A (NS5A) and that the interaction of NS5A with VAPA is required for efficient replication of the hepatitis C virus RNA genome (17). Interestingly, the ability of NS5A to bind VAPA is enhanced by dephosphorylation of NS5A, whereas hyperphosphorylation of NS5A disrupts viral RNA replication (17,33). Whether PP2C⑀ and VAPA also regulate the hepatitis C virus life cycle in human liver through dephosphorylation of NS5A remains to be elucidated. Acknowledgment-We thank to Yuko Nagaura for technical assistance. FIGURE 9. A model of PP2C⑀ function in the regulation of ceramide transfer at a membrane contact site (MCS). CERT normally exists in the cytosol in the hyperphosphorylated form, in which the function of PH and START domains is fully repressed, and the binding activity of FFAT motif is partially inhibited (a). CERT weakly interacts with VAPA, and the Ser/Thr residues within the SR motif are dephosphorylated by PP2C⑀ on the ER membrane (b). Upon dephosphorylation, CERT forms active conformation in which PH, FFAT, and START domains are fully activated (c). PI4P, phosphatidylinositol 4-monophosphate.