PITPs [PI (phosphatidylinositol) transfer proteins] bind and transfer PI between intracellular membranes and participate in many cellular processes including signalling, lipid metabolism and membrane traffic. The largely uncharacterized PITP RdgBβ (PITPNC1; retinal degeneration type B β), contains a long C-terminal disordered region following its defining N-terminal PITP domain. In the present study we report that the C-terminus contains two tandem phosphorylated binding sites (Ser274 and Ser299) for 14-3-3. The C-terminus also contains PEST sequences which are shielded by 14-3-3 binding. Like many proteins containing PEST sequences, the levels of RdgBβ are regulated by proteolysis. RdgBβ is degraded with a half-life of 4 h following ubiquitination via the proteasome. A mutant RdgBβ which is unable to bind 14-3-3 is degraded even faster with a half-life of 2 h. In vitro, RdgBβ is 100-fold less active than PITPα for PI transfer, and RdgBβ proteins (wild-type and a mutant that cannot bind 14-3-3) expressed in COS-7 cells or endogenous proteins from heart cytosol do not exhibit transfer activity. When cells are treated with PMA, the PITP domain of RdgBβ interacts with the integral membrane protein ATRAP (angiotensin II type I receptor-associated protein; also known as AGTRAP) causing membrane recruitment. We suggest that RdgBβ executes its function following recruitment to membranes via its PITP domain and the C-terminal end of the protein could regulate entry to the hydrophobic cavity.
Phosphoinositides play diverse roles in cells and participate in many aspects including signalling, cytoskeletal regulation, ion channel regulation and membrane traffic. PITPs [PI (phosphatidylinositol) transfer proteins] have emerged as key regulators that interface the spatial distribution of PI coupled with its conversion into seven phosphorylated derivatives [1,2]. PITPs are a family of lipid-binding proteins that transfer PI between membrane compartments. Originally identified as soluble proteins of approximately 35 kDa , the family of PITP-related proteins has subsequently grown to five members subdivided into the class I PITPs, α and β (35 kDa), and the Class II RdgB (retinal degradation type B) proteins based on the sequence of the PITP domain [4–6]. The defining feature of a member of the PITP family is the presence of an N-terminal PITP domain. The class I PITPs PITPα and PITPβ are best characterized; PITPα is enriched in neurons and is required for axonal outgrowth [7,8], whereas PITPβ functions in retrograde transport from the Golgi to the endoplasmic reticulum by COPI (coatamer protein I)-containing vesicles . The founding member of the class II proteins is the Drosophila melanogaster (Dm)-RdgBα . The RdgB proteins are named retinal degeneration type B, after the phenotype observed upon disruption of this gene in flies: impairment of visual transduction coupled with retinal degeneration . There are two RdgBα isoforms in humans, RdgBαI (also known as Nir2/PITPNM1) and RdgBαII (also known as Nir3/PITPNM2). RdgBα proteins are multi-domain proteins that, in addition to the PITP domain, possess a FFAT motif that anchors the protein to the endoplasmic reticulum by binding to VAP [VAMP (vesicle-associated membrane protein)-associated protein], a DDHD metal-binding domain of 180 residues and a LSN2 domain of 130 amino acids, both of unknown function. Individual members of the PITP family are likely to be involved in specific functions, as distinctive phenotypes are observed when their genes are ablated in model organisms, including mice. In vibrator mice, the 80% reduction in PITPα levels leads to neurodegeneration followed by juvenile death, whereas ablation of the PITPα gene (gene symbol PITPNA) results in a more severe phenotype where death occurs within days after birth [12,13]. In contrast, ablation of the gene for PITPβ (gene symbol PITPNB) or for RdgBαI (gene symbol PITPNM1) results in embryonic lethality [14,15].
Comparatively little is known about the smaller soluble RdgBβ protein. Two splice variants of RdgBβ have been described; a long splice variant, referred to throughout this paper as RdgBβ (-sp1) (332 amino acids, 38 kDa) translated from an mRNA comprising exons 1–8 and 10, and the short splice variant -sp2 (268 amino acids, 32 kDa) translated from an mRNA comprising exons 1–9 [16,17]. Expression of RdgBβ has been examined at the mRNA level and transcripts are enriched in the heart, muscle, kidney, liver, peripheral blood leucocytes, brain and testes [16,17]. However, the presence of the endogenous protein has mainly been described from high-throughput proteomic screens [18–22].
The PITP domain of human RdgBβ is 41% identical with PITPα and retains all of the key residues that are essential for PI binding . The class I PITPs are thought to interact with and dock on to cell membranes via residues following the G-helix at the C-terminus of the protein [24,25]. Interestingly, the main source of variation between the PITPs occurs at this C-terminal region. The C-terminus of PITPα interacts with DCC (deleted in colorectal cancer), the receptor for netrin-1, whereas the C-terminus of PITPβ determines Golgi localization [7,26]. We therefore set out to examine whether the C-terminus of RdgBβ could interact with other proteins. It had been reported previously that RdgBβ could interact with ATRAP (angiotensin II type I receptor-associated protein, also known as AGTRAP), in a genome-wide yeast two-hybrid screen for interacting partners [27,28].
In the present study we identify that RdgBβ binds directly to 14-3-3 proteins and to ATRAP. 14-3-3 binds to two phosphorylated serine residues in the C-terminal region, whereas ATRAP binds to the PITP domain of RdgBβ, but only following treatment of cells with PMA. We also demonstrate that RdgBβ is subject to rapid turnover in cells, is ubiquitinated and degraded at the proteasome. Comparison of the in vitro PI transfer activity between recombinant PITPα and RdgBβ proteins show that the concentration required for RdgBβ proteins to show activity is 100-fold greater than that for PITPα. Thus examination of the PI transfer activity of endogenous or overexpressed RdgBβ [WT (wild-type) and mutants that cannot bind 14-3-3] from cellular extracts demonstrates no activity. We propose that RdgBβ has to be recruited to membranes by ATRAP via its PITP domain to enable lipid exchange to take place at the adjacent membrane at membrane contact sites.
The following antibodies were obtained from Santa Cruz Biotechnology: pAb (polyclonal antibody) pan 14-3-3 (K-19) (sc-629), pAb anti-FLAG [OctA-Probe (D-8)] (sc-807) and mAb (monoclonal antibody) anti-HA (haemagglutinin) [HA-Probe (F-7)] (sc-7392). The mAb anti-FLAG (DDK) was obtained from OriGene Technologies (TA50011) and the pAb anti-ubiquitin (U5379) was from Sigma–Aldrich. An anti-PITPNC1 pAb was obtained from Orbigen (PAB-02250) and the anti-ATRAP mAb (ab57555) was obtained from Abcam. Antibodies against ARF1 (ADP-ribosylation factor 1) (pAb 678), PITPβ (mAb 1C1) and PITPα (pAb 674) were made in-house and have been described previously [29–31].
Mammalian and bacterial expression plasmids
pEFPLink2-FLAG-RdgBβ (human) was a gift from Professor J. Hsuan (Department of Medicine, University College London, London, U.K.) , and was used as a template to clone RdgBβ into the pRSET-C-His expression vector (Invitrogen) for bacterial expression using the XhoI/EcoRI restriction sites, and into the pIRES2 EGFP (enhanced green fluorescent protein) bicistronic vector (BD Bioscience) for mammalian expression of the untagged protein. The RdgBβ-sp2 IMAGE clone cDNA (IMAGE: 4299595) (human) was obtained from the MRC (Medical Research Council) Geneservice (Cambridge, U.K.) and was cloned into the pRSET-C-His vector for bacterial expression using the XhoI/EcoRI restriction sites. The human sequences of both RdgBβ and RdgBβ-sp2 were amplified by PCR from the pRSET-C vectors with the addition of an N-terminal FLAG tag, introduced from the forward primer, and cloned into the pcDNA3.1(−) vector for mammalian expression. RdgBβ-sp2 was also cloned into pcDNA3.1(−) without the addition of a protein tag, using the XhoI/HindIII restriction sites. Point mutations were generated using the QuikChange® site-directed mutagenesis kit (Stratagene): pRSET-C-RdgBβ-sp1 1-263 (mutation of Leu264 to a stop codon, L264_STOP), pcDNA3.1-FLAG-RdgBβ S274A, S299A and SS274/299AA and pcDNA3.1-FLAG-RdgBβ 1-251 (I252_STOP). All constructs were sequenced for verification. pcDNA3-HA-ATRAP (mouse) was a gift from Dr M. Lopez-Ilasaca (Brigham and Women's Hospital, Harvard Medical School, Boston, MA, U.S.A.) .
Purification of recombinant proteins expressed in Escherichia coli
His-tagged RdgBβ, RdgBβ-sp2, truncated RdgBβ (residues 1–263), PITPα and PITPβ proteins were expressed in the E. coli strain BL21(DE3)pLysS and purified using HIS-Select Nickel Affinity gel as described previously . The recombinant proteins were desalted into Pipes buffer [20 mM Pipes, 137 mM NaCl and 3 mM KCl (pH 6.8)] and analysed by SDS/PAGE for purity. The protein concentration was adjusted accordingly and the proteins were stored at −80 °C.
Production and validation of RdgBβ-specific polyclonal antisera [Ab:101 (antibody 101)]
Antibodies were raised in two rabbits using two internal peptides [TKYEDNKGSNDTIFD (residues 111–125) and ACDETIPERYYK (residues 140–151)] common to both rodent and human RdgBβ splice variants, sp1 and sp2. Peptide synthesis and immunization was performed by Eurogentech. For validation, three different siRNAs (small interfering RNAs) (1 plus 5, and 6 alone) were used in two independent combinations exactly as described, except that a single round of knockdown was carried out . The siRNA sequences against human and monkey RdgBβ were oligonucleotide number 1, 5′-GGAUUUGGAGCCUAAUUAATT-3′; oligonucleotide number 5, 5′-CACCGUAGACGAGUACAATT-3′; and oligonucleotide number 6, 5′-GAGCGCUACUACAAAGAAUTT-3′, and were obtained from Qiagen.
Trypsin digest of RdgBβ proteins
RdgBβ or RdgBβ-sp2 recombinant protein (100 μg in 500 μl) was incubated with 5 μl of trypsin-EDTA solution (Sigma–Aldrich) at 37 °C. Aliquots (50 μl) were removed directly into SDS/PAGE sample buffer at each time point. For the Coomassie-Blue-stained gel, 2 μg of protein was loaded per lane; for the Western blot analysis, 30 ng of protein was loaded per lane. Ab:101 was used to detect the trypsin-digested proteins. Transfer activity was measured using 200 μg of protein/ml as described below.
Culture and electroporation of COS-7 cells, and fractionation of cytosol by size-exclusion chromatography
COS-7 cells, cultured in DMEM (Dulbecco's modified Eagle's medium) supplemented with 10% heat-inactivated FBS (fetal bovine serum), 4 mM L-glutamine, 0.5 i.u./ml penicillin and 50 μg/ml streptomycin, were electroporated with the appropriate plasmid as described previously . At 48 h later, the cells were trypsinized, washed and re-suspended in 240 μl of SET buffer [0.25 M sucrose, 1 mM EDTA and 10 mM Tris/HCl (pH 7.4)] with protease inhibitor cocktail (Sigma–Aldrich). To harvest the cytosol, the cells were sonicated 3×15 s followed by ultracentrifugation at 43000 rev./min for 60 min at 4 °C (rotor type TLA55, Beckman Optima benchtop ULTRA). The cytosol was decanted, re-centrifuged at 15000 g for 10 min and 200 μl (16 mg of protein) was immediately loaded on to a Superose 12 10/300 column with a bed volume of 24 ml (GE Healthcare). The column was calibrated using a kit containing proteins of molecular mass 200, 67, 43, 25 and 13.7 kDa, and their elution profile is indicated in Figure 5(A). For the recombinant proteins, 200 μg (diluted in SET buffer to 200 μl) was loaded on to the column. The column was eluted with Pipes buffer and 0.5 ml fractions were collected. A 50 μl aliquot of each fraction for the COS-7 cytosol, or 3 μl of each fraction for the recombinant protein was used for the Western blot analysis using Ab:101. Animals were killed by Schedule 1 procedure (according to the appropriate Home Office and Institutional guidelines) and, once the heart was removed, a syringe containing PBS was used to retrogradely perfuse the heart to remove the blood. To obtain rat heart cytosol, the perfused heart was chopped finely in 600 μl of SET buffer and homogenized with a dounce homogenizer. After 10 strokes, the homogenate was transferred into a 15 ml tube, sonicated 3×15s and cytosol was obtained as described above.
Preparation of membranes and cytosol
COS-7 cells were electroporated with the appropriate plasmids. After 24 h, the cells were treated with PMA (100 nM) as indicated. At 16 h later, the cells were trypsinized, washed and re-suspended in 300 μl of PBS in the presence of a protease inhibitor cocktail (Sigma). To prepare membranes and cytosol, the cells were sonicated on ice (50 μm, 3×15 s) and centrifuged for 10 min (1800 g, 4 °C) to pellet the nuclei and unbroken cells. The lysate was centrifuged for 60 min at 43000 rev./min at 4 °C (rotor type TLA55, Beckman Optima benchtop ULTRA) to pellet the membranes, and the supernatant consisting of the cytosolic fraction was retained for further analysis. To remove contaminating cytosolic proteins, the membranes were resuspended in 1 ml of SET buffer and centrifuged at 43000 rev./min at 4 °C for 60 min (rotor type TLA55, Beckman Optima benchtop ULTRA). The membranes were resuspended in 100 μl of SET buffer or in PBS. Protein concentrations were determined for both the membrane and cytosolic fractions. The distribution of proteins between membranes and the cytosol was analysed through separation of the proteins by SDS/PAGE followed by Western blot analysis.
COS-7 cells were seeded in 10 cm cell culture dishes in fresh medium at (1–2)×105 cells/ml. Transfection was carried out 6 h after seeding using FuGENE™ HD reagent (Roche Diagnostics) following the manufacturer's instructions (DNA/FuGENE™ HD ratio, 2:6). Cells were harvested on ice 48 h after seeding in 1 ml of lysis buffer [50 mM Tris/HCl (pH 7.4), 150 mM NaCl, 10 mM MgCl2 and 1% Triton X-100] with protease and phosphatase I and II inhibitors (Sigma–Aldrich, P8340, P2850 and P5726 respectively). The cells were sonicated and lysates were obtained after centrifugation (15000 g for 15 min at 4 °C). The BCA (bicinchoninic acid) assay was used to determine the protein concentration of the lysate. For immunoprecipitation, 300 μg of protein was incubated with 40 μl of equilibrated anti-FLAG M2 affinity gel (Sigma–Aldrich, A2220) on a rotating wheel for 2 h at 4 °C. The tubes were centrifuged at 15000 g for 2 min and the supernatant was discarded. The beads were washed three times with lysis buffer; the final wash was carried out in a fresh Eppendorf tube. After the final centrifugation, the supernatant was aspirated and replaced with 20 μl of 2×SDS/PAGE sample buffer. The proteins were separated by SDS/PAGE and then transferred on to nitrocellulose membrane for Western blot analysis using the appropriate primary antibodies. For the protein kinase inhibitor treatments, the culture medium was replaced with fresh medium containing the appropriate inhibitor at the concentrations indicated 16 h prior to cell harvest: H-89 [PKA (protein kinase A) inhibitor], LY294002 [PI3K (phosphoinositide 3-kinase) inhibitor], BIM-I (bisindolylmaleimide-I, also known as GF 109203X) [PKC (protein kinase C) inhibitor] (all from Calbiochem, 371963, 440206 and 203290 respectively), and PMA (Sigma–Aldrich, P8139). For treatment with the proteasome inhibitor MG-132 (Z-Leu-Leu-Leu-al) (Sigma, C2211), cells were treated with 20 μM MG-132 for 4 h.
14-3-3 far-Western blot assay for direct binding
The far-Western blot assay was carried out exactly as described previously . In brief, FLAG immunoprecipitates and recombinant RdgBβ proteins were subjected to SDS/PAGE. The proteins were transferred on to nitrocellulose membrane which was then blocked for 1 h in 5% BSA/TBS-T [TBS (Tris-buffered saline, 20 mM Tris/HCl, pH 7.4, and 150 mM NaCl) containing 0.02 % Tween 20]. The membrane was washed and incubated with in vitro DIG (digoxigenin)-labelled BMH1 and BMH2 Saccharomyces cerevisiae 14-3-3 isoforms in 5% BSA/TBS-T for 2 h at room temperature (22 °C). The membrane was washed again and then finally incubated with horseradish peroxidase-conjugated anti-DIG antibody for 45 min before detection of the bound 14-3-3 protein with ECL (enhanced chemiluminescence; GE Healthcare). 14-3-3 far-Western blot reagents were a gift from Professor C. MacKintosh (MRC Protein Phosphorylation Unit, College of Life Sciences, University of Dundee, Dundee, Scotland, U.K.).
CHX (cyclohexamide) treatment
COS-7 cells were electroporated with pcDNA3.1-FLAG-RdgBβ WT or S299A mutant plasmids and divided into 16×3.5 cm cell culture dishes (2 ml per dish). At 48 h after transfection, the medium was removed and replaced with fresh medium containing 100 μg/ml CHX. After 0, 2, 4 or 6 h, the medium was removed and the cells washed twice in PBS. The cells were harvested in 300 μl of RIPA buffer [150 mM NaCl, 1% (v/v) Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS and 50 mM Tris/HCl (pH 7.5)] and incubated on ice for 30 min. The tubes were centrifuged at 15000 g for 15 min at 4 °C. The supernatant was transferred to a fresh tube. The protein concentration of the lysate was determined using the BCA assay. Protein (50 μg) was boiled with SDS/PAGE sample buffer and loaded into each lane of the gel. Proteins were transferred on to nitrocellulose membranes and probed with the anti-FLAG mAb DDK. Membranes were stripped and re-probed for endogenous PITPβ.
Assay for PI transfer activity
PI transfer activity was assayed by measuring the transfer of [3H]PI from radiolabelled rat liver microsomes to unlabelled synthetic liposomes [PC (phosphatidylcholine)/PI molar ration of 98:2] by recombinant proteins or fractions (100 μl), as described previously . Transfer activity was calculated as a percentage of the total radioactivity present in the assay after subtraction of the number of counts transferred in the absence of a PITP source. For the recombinant proteins, transfer activity was monitored in duplicate samples; for fractions obtained after size-exclusion chromatography, individual fractions were analysed singly. All data presented are representative of at least three independent experiments and, for the recombinant proteins, at least two protein preparations.
Localization of RdgBβ and ATRAP by immunofluorescence
COS-7 cells were co-transfected with FLAG-tagged RdgBβ and HA-tagged ATRAP using FuGENE™ HD and were cultured on glass coverslips for 48 h. Cells were treated with 100 nM PMA for 4 h. The cells were washed twice with PBS and were fixed with 4% (w/v) paraformaldehyde and subsequently permeabilized with digitonin (40 μg/ml) for 10 min on ice and washed with ice-cold PBS. The cells were incubated with primary antibodies [mAb anti-HA for ATRAP and pAb OctA (anti-FLAG) for RdgBβ] as indicated followed by fluorescent conjugated secondary antibodies [Alexa Fluor® 488 (for ATRAP) or Alexa Fluor® 546 (for RdgBβ) (Molecular Probes)]. Fluorescence was recorded by excitation at 488 nm or 546 nm with a light source (Excite 120) using an Olympus IX80 microscope fitted with a ×100 oil-immersion objective. Images were acquired with a charge-coupled device camera (ORCA) cooled to −35 °C and controlled with the CellF software (Olympus).
Characterization of a pAb (Ab:101) against RdgBβ
A pAb made against two internal peptides of RdgBβ detects both splice variants including a C-terminal-truncated protein of the long splice variant (residues 1–263). The antibody does not cross-react with PITPα or PITPβ, or to the RdgBα PITP domain (Figure 1A, left-hand panel). When COS-7 cell lysates were probed with the antibody, no band corresponding to either splice variant of RdgBβ was seen (Figure 1A, middle panel). However, the antibody detected RdgBβ when overexpressed (indicated by an arrow). The antibody also recognized an endogenous protein present as a doublet band at 35 kDa; this protein was not susceptible to RNAi (RNA interference), indicating that it is a cross-reactive band. We also examined the detection level by titration of recombinant RdgBβ proteins and show that the antibody could just detect ∼5 ng and above (Figure 1A, right-hand panel).
The C-terminus of RdgBβ is disordered and contains putative 14-3-3-binding sites
Bioinformatic analysis shows that the RdgBβ C-terminus is a low complexity region susceptible to proteolytic cleavage
Analysis of the sequence of RdgBβ indicates that it has an N-terminal PITP domain (Pfam family IP_trans), which shares 41% identity with PITPα. PITPα, RdgBβ and RdgBβ-sp2 differ mainly at their C-termini (Figure 1B). The N-terminal PITP domain of PITPα and PITPβ is a compact structure containing a lipid-binding cavity. It consists of eight β-strands which form a large concave sheet, flanked by two long α-helices [24,25]. This cavity is closed by a ‘lid’ composed of a C-terminal α-helix (G-helix) and an 11-amino-acid extension in PITPα and PITPβ (Figure 1B). Movement of the ‘lid’ is required for lipid exchange . The putative G-helix in RdgBβ is indicated by a dotted line (Figure 1B). Although the RdgBβ-sp2 C-terminal extension is similar in length to that of PITPα, the RdgBβ C-terminus is extended by ∼60 amino acid residues (Figure 1B).
To understand more about the C-terminus of RdgBβ, the human sequence (UniProt accession number Q9UKF7) was uploaded to the DISOPRED2 Disorder Prediction Server . This identified a region of disorder beginning at residue 252 in RdgBβ and continuing to the extreme C-terminus of the protein (amino acid 332). To confirm this experimentally, the recombinant RdgBβ protein was incubated with trypsin alongside RdgBβ-sp2 as a control, and aliquots of the recombinant protein were removed at specified time points for analysis by Coomassie-Blue-stained SDS/PAGE and by Western blot (Figure 1C). RdgBβ-sp2 has a much shorter C-terminus than RdgBβ (16 residues compared with 80 residues following the compact domain predicted by DISOPRED2). Consistent with the absence of trypsin cleavage sites in the C-terminus of RdgBβ-sp2 after residue 252 (Figure 1B), RdgBβ-sp2 is resistant to trypsin cleavage (Figure 1C). This, together with earlier observations of trypsin digestion of PITPα where the compact PITP domain was found to be resistant to proteolytic cleavage [37,38], indicates that the stable product (∼37 kDa) reached by 10 min in the trypsin digestions shown in Figure 1(C) is the compact PITP domain. In support of this, the trypsinized protein exhibits PI transfer activity comparable with the untreated protein in the in vitro assay (Figure 1C).
Further bioinformatic analysis of the C-terminus of RdgBβ indicates the presence of two PEST sequences (Figure 1B, denoted by *). PEST sequence analysis was performed using the pestfind algorithm on the Emboss server (http://emboss.bioinformatics.nl/). The PEST sequence motif was first identified in 1986 as a proteolytic signal region [39,40]. Such regions are rich in proline (P), glutamic acid (E), serine (S) and threonine (T) residues and are most common in rapidly degraded proteins.
RdgBβ contains a 14-3-3 protein-binding site at its C-terminus
Large-scale phosphoproteomic studies have reported that the C-terminus of RdgBβ is phosphorylated in vivo at Ser274 and Ser299 (Figure 1B). Phosphoproteomic analysis of the developing mouse brain [fore-brain and mid-brain from E16.5 (E is embryonic day)]  and HeLa cells in G1 phase of the mitotic cycle identified Ser274 as being phosphorylated . Ser299 was found to be phosphorylated in human embryonic stem cells , following receptor tyrosine kinase stimulation of cancer cell lines (H1703, H3255 and MKN45 cells) by EGF (epidermal growth factor), PDGF (platelet-derived growth factor) and c-Met , after stimulation of signalling by the oncogenic mutant of Flt3 , as well as in Jurkat cells treated with the phosphatase inhibitors calyculin and pervanadate (observed on more than 25 separate occasions) (PhosphoSitePlus, Cell Signalling Technology) .
A survey of the sequence surrounding pSer299 (p indicates the phosphorylated form) using the cell signalling interaction prediction program Scansite at high stringency  and the PhosphoMotif Finder tool at the Human Protein Reference Database , indicate that this region is a consensus-binding site for 14-3-3 proteins. Two modes of 14-3-3 binding have previously been identified [44,45]; the region surrounding Ser299 conforms to the mode I site RSX(pS)XP (where pS is the phosphorylated serine residue), with a lysine residue in place of the first serine residue in RdgBβ (Figure 1B). Interestingly this is very similar to the region surrounding Ser274, which conforms to the sequence RSXXX(pS)XP. 14-3-3 proteins form dimers, with each monomer providing one phospho-serine-binding site. In this way 14-3-3 proteins have been shown to bind two tandem sites on a single target protein resulting in strong ‘bidentate’ binding [44–46]. Ser274 and Ser299 could form two such 14-3-3-binding sites on RdgBβ. RdgBβ-sp2 lacks these residues and so is not predicted to bind 14-3-3 proteins.
To examine whether RdgBβ is present as a complex with 14-3-3 proteins in vivo, COS-7 cells were transfected with FLAG-tagged RdgBβ. RdgBβ was immunoprecipitated via its FLAG tag and Western blotted using a pan 14-3-3 antibody (Figure 2A, top right-hand panel). In addition, the immunoprecipitate was probed with 14-3-3 by far-Western blotting in which recombinant 14-3-3 proteins with a DIG tag were incubated with the membrane, and 14-3-3 binding was revealed using anti-DIG Fab fragments conjugated with horseradish peroxidase (Figure 2A, middle right-hand panel). Both membranes were stripped and re-probed with an anti-FLAG antibody to confirm the presence of FLAG–RdgBβ. Only one is shown (Figure 2A, bottom right-hand panel). These experiments show that 14-3-3 is present in the FLAG–RdgBβ immunoprecipitate, and that binding to 14-3-3 is a direct interaction.
RdgBβ binds directly to endogenous 14-3-3 proteins
We hypothesized that the recombinant RdgBβ (purified after expression in E. coli) would not bind 14-3-3 proteins as it lacks the phosphorylation required for 14-3-3-binding. FLAG–RdgBβ immunoprecipitate was used as a positive control; recombinant RdgBβ, RdgBβ-sp2 and truncated RdgBβ (residues 1–263) were analysed alongside and were all subjected to 14-3-3 far-Western blotting (Figure 2B). The recombinant 14-3-3 only binds to the COS-7 cell-expressed RdgBβ which has had the opportunity to be phosphorylated, not to recombinant RdgBβ lacking the phosphorylation, nor to RdgBβ-sp2 or RdgBβ (1–263) which lack the consensus 14-3-3-binding sites.
To confirm the residues responsible for binding to 14-3-3, serine-to-alanine point mutations of Ser274 and Ser299 were constructed (S274A, S299A and SS274/299AA double mutant) and expressed in COS-7 cells. Immunoprecipitation of WT and mutant RdgBβ revealed that mutation of either Ser274 or Ser299 abolished binding to 14-3-3 (Figure 3A). We conclude that 14-3-3 needs to bind both pSer274 and pSer299 for stable association.
Phosphorylation of both Ser274 and Ser299 of RdgBβ is required for 14-3-3 binding
The consensus sites for 14-3-3 binding shows high similarity to the consensus sequences for phosphorylation by the basophilic kinases PKA, PKB (protein kinase B)/Akt and PKC . To establish which kinase may be responsible for either of the serine phosphorylations, we treated COS-7 cells expressing FLAG–RdgBβ with inhibitors of these kinases (H89 for PKA inhibition, LY294002 for inhibition of PKB indirectly through PI3K, and BIM-I for inhibition of PKC) for 16 h. Only BIM-I caused a reduction in 14-3-3 binding (Figure 3B), suggesting that PKC could be responsible for one or both of the serine phosphorylations, pSer274 and pSer299.
A reduction of RdgBβ expression was observed in the lysates of BIM-I-treated cells (Figure 3B), suggesting that 14-3-3 binding could protect RdgBβ from degradation. The C-terminal region of RdgBβ is unstructured and contains two PEST sequences (Figure 1B), suggesting that the protein could be subject to rapid turnover. Moreover, binding to 14-3-3 would shield the PEST sequence and may provide protection. We therefore examined the rate of WT RdgBβ and S299A mutant degradation after treatment with the protein synthesis inhibitor CHX (Figure 4A). WT RdgBβ was reduced by half within 4 h, whereas the mutant protein that cannot bind 14-3-3 was reduced by half within 2 h. At the end of the 6 h period, very little of the phospho-defective RdgBβ mutant was present. We also compared the turnover rate of another soluble PITP, PITPβ, which was not seen to degrade in the 6 h experiment time.
RdgBβ undergoes rapid turnover in COS-7 cells
A common mechanism by which intracellular proteins are degraded is via a ubiquitin-dependent mechanism involving the 26S proteasome. Treatment of COS-7 cells expressing WT FLAG–RdgBβ with the proteasome inhibitor MG-132 revealed that RdgBβ is ubiquitinated (Figure 4B). Phosphorylation of substrate proteins may promote recruitment of the E3 ubiquitin ligase, which binds to the substrate and attaches ubiquitin to it. Phosphorylation is typically required on serine and threonine residues in a PEST sequence to activate the signal and promote degradation of the substrate protein . We therefore examined whether phosphorylation of Ser274 or Ser299 on RdgBβ was required for ubiquitination (Figure 4C). FLAG–RdgBβ and phospho-defective mutant immunoprecipitates were probed with an anti-ubiquitin antibody, which showed that RdgBβ-sp1 was ubiquitinated, regardless of its phosphorylation state.
Assessment of PI transfer activity of RdgBβ when complexed with 14-3-3
To examine the consequences of 14-3-3 binding to RdgBβ on lipid-transfer activity, WT and mutants of RdgBβ that could not bind 14-3-3 were expressed in COS-7 cells. COS-7 cells express negligible levels of endogenous RdgBβ (Figure 1A), but do express endogenous PITPα and PITPβ  and therefore it was necessary to separate these PITPs from RdgBβ. Cytosol prepared from cells overexpressing WT FLAG–RdgBβ was fractionated by size-exclusion chromatography, examined for the PITP proteins by Western blot analysis, and individual fractions were assessed for PI transfer activity. Endogenous PITPα and PITPβ co-elute in fractions 23–25 and are well-separated from RdgBβ-containing fractions (17–19) (Figure 5A). RdgBβ-containing fractions also contain 14-3-3 (Figure 5A). Although the fractions (23–25) enriched in PITPα/PITPβ show robust PI transfer activity , the RdgBβ-containing fractions (17–19) exhibit none (Figure 5A). Next we considered the possibility that the complex of RdgBβ with 14-3-3 is unable to transfer because the C-terminus is immobilized due to 14-3-3 binding. We therefore expressed the mutant S299A and S274A RdgBβ proteins which are unable to bind to 14-3-3. After separation of the proteins by gel filtration, the fractions were blotted for PITPα and PITPβ, as well as for the mutant RdgBβ proteins. PITPα/PITPβ eluted in fractions 23–25; these fractions exhibited PI transfer activity, whereas fractions (17–19) containing the mutant proteins lacked transfer activity. As a control, bacterially expressed recombinant RdgBβ (200 μg) was similarly separated by gel filtration; recombinant RdgBβ elutes in fractions 17–21 and is associated with PI transfer activity (Figure 5A).
PI transfer activity of COS-7 cell-derived RdgBβ-sp1 and recombinant RdgBβ proteins
Overexpressed RdgBβ proteins, whether bound to 14-3-3 or not, are devoid of lipid-transfer activity, although the recombinant protein is active even after gel filtration. One possibility is that RdgBβ proteins are less active compared with PITPα (and PITPβ), and that concentrations achieved by overexpression are insufficient. Comparison of the concentration-dependence of transfer activity between PITPα and RdgBβ confirmed that this was indeed the case (Figure 5B). Unlike PITPα, which shows transfer activity at even 200 ng/ml, RdgBβ just begins to show activity when its concentration exceeds 10 μg/ml. Significant activity is only observed when RdgBβ concentrations rise above 50 μg/ml. In separate experiments, we calculated that ∼300 ng/ml of RdgBβ was present in the RdgBβ-enriched fractions, a level insufficient to exhibit lipid transfer. It is possible that the in vitro lipid-transfer assay used in the present study was not optimal for RdgBβ, and therefore permeabilized HL60 cells were tested as the donor (rather than microsomes) . We also varied the composition of the acceptor liposomes. None of these modifications improved the efficiency of lipid transfer by RdgBβ.
Endogenous RdgBβ is present in rat heart
To investigate the function of RdgBβ in vivo, we needed to identify cell lines and tissues that express endogenous RdgBβ proteins. Using Ab:101, as well as a commercial antibody called PITPNC1 (Orbigen, PAB-02250) which was raised against amino acids 67–266 of RdgBβ, we screened lysates (50–100 μg) prepared from several cell lines and rat tissues by Western blot analysis. The cell lines examined included COS-7, HEK (human embryonic kidney)-293, HKC-8 cells (kidney proximal tubule cell line), Jurkat (a T-cell line), HL60 cells (a promyelocytic leukaemic cell line), HeLa cells, Hep-G2 (a liver cell line), PC12 cells (a chromaffin cell line) and RBL-2H3 mast cells. Rat tissues examined included liver, brain, heart, kidney and muscle. The Western blot signal from the lysates was ambiguous as both antibodies recognized several weak bands at the appropriate size. RdgBβ transcripts are highly enriched in the heart  and we therefore fractionated rat heart cytosol by size-exclusion chromatography and Western blot fractions with both anti-RdgBβ antibodies, as well as for 14-3-3 (Figure 6). As a positive control, we used fractions enriched in (overexpressed) untagged RdgBβ prepared from COS-7 cell cytosol. Fractions 17–21 show immunoreactivity with both antibodies for a protein of the right size. Moreover, recombinant proteins and COS-7 cell-expressed proteins also eluted in the same fractions (see Figure 5). We suggest that rat heart does contain RdgBβ, and its elution on gel filtration partially overlaps with 14-3-3 proteins, suggesting that some of the RdgBβ proteins are likely to be in a complex with 14-3-3. The fractions were also analysed for PI transfer and a peak of activity was observed that could be accounted for by the presence of endogenous PITPα and PITPβ (Figure 6). No PI transfer activity was associated with RdgBβ proteins.
RdgBβ is present in rat heart cytosol
RdgBβ associates with ATRAP
The results described above regarding the absence of in vitro transfer activity for RdgBβ reinforces our recent speculation that PITP proteins function by being recruited to specific membrane compartments, potentially by interaction with integral membrane proteins . The best example is that of PITPα which is recruited by DCC, the receptor for netrin-1 . It has been reported that RdgBβ interacts with an integral membrane protein, ATRAP, in a high-throughput yeast two-hybrid screen, where systematic mapping of protein–protein interactions of 8100 human proteins was undertaken [27,28]. To validate this interaction in mammalian cells, FLAG-tagged RdgBβ was co-expressed with HA-tagged ATRAP; however, no interaction between the two proteins was observed. Since ATRAP is protective following chronic stimulation with angiotensin II in vivo, we reasoned that the RdgBβ association would be regulated by downstream signalling from the angiotensin II receptor. Angiotensin II is a G-protein-coupled receptor and stimulates phospholipase C. Events downstream to phospholipase C is the activation of PKC, and to mimic chronic angiotensin II signalling we examined complex formation between RdgBβ and ATRAP with PMA, an activator of PKC. FLAG–RdgBβ immunoprecipitates contained ATRAP only following treatment of the cells with PMA (Figure 7A). BIM-I inhibited the interaction, suggesting that events downstream of PKC activation were required (Figure 7B). A 4–16 h PMA treatment was required for the interaction; conditions that would have down-regulated some isoforms of PKC , but not PKD (protein kinase D, PKCμ) . In addition, HA immunoprecipitates, in which ATRAP was immunoprecipitated via its HA tag, contained FLAG–RdgBβ (results not shown).
RdgBβ interacts with ATRAP following treatment with PMA
To examine whether the binding of RdgBβ to ATRAP was affected by 14-3-3 binding at the C-terminus, we engineered a C-terminal-truncated RdgBβ mutant (1–251), that lacks the 14-3-3-binding site. In addition, we used the S299A mutant that does not bind 14-3-3 (Figure 3A). The constructs were expressed in COS-7 cells and the cells were incubated with PMA for 16 h. It was noted that PMA treatment caused increased expression of RdgBβ (Figure 8). Immunoprecipitation of the RdgBβ constructs revealed that binding to ATRAP was independent of 14-3-3 binding. Thus we can conclude that ATRAP binds to the PITP domain, whereas 14-3-3 binds to the unstructured C-terminus. This also implies that RdgBβ-sp2 should also interact with ATRAP (Figure 1). (Expression of RdgBβ-sp2 in COS-7 cells is extremely low and therefore the interaction was inconclusive). The interaction between ATRAP and the PITP domain of RdgBβ is specific, as neither PITPα nor PITPβ was found to interact with ATRAP.
ATRAP interacts with the PITP domain of RdgBβ
Recruitment of RdgBβ to membranes upon PMA treatment
ATRAP is an integral membrane protein , whereas RdgBβ is a soluble protein. In previous studies, it was reported that RdgBβ localized throughout the cytoplasm, but was absent from the nucleus in COS-7 and in PC12 cells [16,17], whereas ATRAP was found to localize mainly at the endoplasmic reticulum and the Golgi in HEK cells . We examined the co-localization of ATRAP and RdgBβ after co-expression in COS-7 cells. RdgBβ localized to the cytosol showing a uniform staining throughout, but was excluded from the nucleus, whereas ATRAP showed perinuclear staining. Upon treatment with PMA, increased expression of RdgBβ was observed, and co-localization of RdgBβ with ATRAP at the perinuclear region was also observed.
Since treatment with PMA does cause an interaction between the two proteins, we surmised that RdgBβ would get recruited to membranes after PMA treatment and this could be monitored biochemically. COS-7 cells were co-transfected with RdgBβ and ATRAP and treated with PMA. Membranes and cytosol were prepared and blotted for RdgBβ and ATRAP. ATRAP was present in the membranes, whereas RdgBβ was mainly cytosolic. Upon PMA treatment, RdgBβ was found to translocate to membranes (Figure 9A). In the presence of PMA, RdgBβ levels were much higher in cells, and this could be due to an inhibition of its degradation (Figure 8A). Therefore we calculated the percentage distribution of RdgBβ between membranes and cytosol for the untreated and PMA-treated cells. Membranes prepared from PMA-treated cells had more RdgBβ (after accounting for increased expression), with a corresponding decrease in the cytosolic fractions.
Translocation of RdgBβ to membranes upon PMA treatment
ATRAP is highly enriched in the heart and kidney  and we confirmed that COS-7 cells, a kidney fibroblast-like cell line, do contain endogenous ATRAP (Figure 9B). Translocation of RdgBβ to membranes by PMA treatment was also observed in COS-7 cells. Finally, we examined whether a mutant RdgBβ that did not bind 14-3-3 would also translocate to membranes. Because of the increased expression of RdgBβ by PMA, the results are expressed as the ratio of PMA-treated membranes to control membranes. There was an 8-fold increase in the membranes compared with a 2-fold increase in the cytosol for WT RdgBβ. For the mutant that does not bind 14-3-3, there was a 14-fold increase, suggesting that 14-3-3 binding hinders membrane translocation. These changes were specific for RdgBβ, as changes in the distribution of 14-3-3, PITPα or PITPβ was not observed (Figure 9B).
Proteins belonging to the PITP family have a defining PITP domain at their N-terminus, but differ markedly in their C-terminal residues. Of the three soluble PITPs, splice variant 1 of RdgBβ (referred throughout as RdgBβ) has the most extensive C-terminal region at 80 amino acids long. This region contains two residues, Ser274 and Ser299, which are phosphorylated in vivo [18–20]. These residues lie in consensus sequences for 14-3-3-binding , and in the present study we provide compelling evidence that RdgBβ directly binds to 14-3-3 proteins via its C-terminus and that phosphorylation of Ser274 and Ser299 is required to mediate this interaction. 14-3-3 proteins are cup-shaped dimeric proteins and each subunit of the dimer is able to bind one discrete phospho-serine- (or phospho-threonine)-containing ligand. In RdgBβ there are two 14-3-3-binding sites, which are separated by 25 amino acids. A minimal distance of ten residues is sufficient for 14-3-3 proteins to bind to a single polypeptide , suggesting that the 14-3-3 dimer engages a single molecule of RdgBβ. Mutation of either serine residue to alanine is sufficient to inhibit 14-3-3 binding, indicating that both residues are essential for optimal binding. Owing to the dimeric nature of 14-3-3 proteins, a phosphopeptide with two binding motifs can bind 14-3-3 proteins with a 30-fold greater affinity than a phosphopeptide containing a single motif . This suggests that the high affinity of 14-3-3 binding to RdgBβ only occurs when both residues are phosphorylated.
In addition to the presence of 14-3-3-binding sites on the C-terminus of RdgBβ, this region also contains two PEST sequences which, in many other proteins, act as signals for degradation . Binding to 14-3-3 proteins would mask the PEST sequence and therefore prevent the degradation. Nonetheless, RdgBβ is rapidly degraded in resting cells with a half-life of 4 h, suggesting that binding of 14-3-3 proteins to RdgBβ is highly dynamic. 14-3-3 binding does afford some protection since a mutant of RdgBβ that is unable to bind 14-3-3 is degraded even faster; its half-life is reduced to 2 h. The recognition of the PEST sequence and the mechanism of subsequent degradation of the substrate protein has been shown in many studies to be mediated by ubiquitin and requires a properly functioning proteasome . The PEST sequence by its very nature contains a high proportion of serine and threonine residues. In many studies phosphorylation of serine and threonine residues in the PEST sequence is a pre-requisite for degradation of the protein [52,53]. However in the present study, mutation of the phosphorylated serine residues to alanine had no effect on ubiquitination of RdgBβ, indicating that phosphorylation is not required for activation of the PEST sequence and protein degradation in this case. In addition, bioinformatic analysis shows that the C-terminus of RdgBβ is largely disordered and consequently is susceptible to proteolytic cleavage by trypsin. Such disordered sequences have also been shown to be necessary for the degradation of proteins at the proteasome, effectively giving the proteasome a point from which to unravel the target protein to allow access of proteases . Transplantation of a disordered region from one protein to another can increase the rate of degradation of a protein, although the disordered region itself is not sufficient for proteasomal degradation: ubiquitination is also required . These results suggest that there exists an E3 ubiquitin ligase or ligases that are responsible for targeting RdgBβ-sp1 for degradation and there are >600 E3 ligases which are sub-divided by the presence of RING, HECT or U-box domains that could be responsible .
RdgBβ is a member of the PITP family and is able to bind and transfer PI in vitro . The entrance to the hydrophobic cavity in PITPα is controlled by the C-terminus (see Figure 1). Thus our prediction was that 14-3-3 binding to the C-terminus of RdgBβ is likely to immobilize the C-terminus and restrict lipid exchange. To test this we expressed WT RdgBβ and mutants that could not bind 14-3-3 in COS-7 cells and separated these proteins from endogenous PITPs. No transfer activity was detected with either construct. RdgBβ is highly expressed in the heart, but we were unable to detect any in vitro transfer activity associated with the endogenous protein. The lack of activity is attributed to insufficient protein and this was confirmed by examination of the concentration-dependence for transfer activity. Although RdgBβ proteins are active in vitro, the concentration required is 50–100 μg/ml. In comparison, PITPα requires 300–500 ng/ml, a difference of 100-fold. Thus we would suggest that RdgBβ proteins would need to be concentrated at the membrane by interaction with membrane proteins for lipid exchange. Despite the inability to transfer PI in the in vitro assay, RdgBβ binds PI in its hydrophobic cavity (Figure 5C).
We have identified that RdgBβ interacts with ATRAP, an 18 kDa transmembrane protein that contains three hydrophobic domains and an extended hydrophilic cytoplasmic C-terminal tail (Figure 10). This interaction between RdgBβ and ATRAP is observed after treatment with PMA. Binding of RdgBβ to ATRAP occurs via the PITP domain, a site distinct from the 14-3-3-binding site on RdgBβ (Figure 10). We suggest that ATRAP may prefer to bind RdgBβ that is not bound to 14-3-3, conditions that may allow interaction of the C-terminus with an adjacent membrane initiating the process of lipid exchange. This conclusion is supported by the observation that the RdgBβ Ser274 mutant translocates to the membrane at a higher level (Figure 9B) than WT RdgBβ, and the cytosolic pool is depleted to a greater extent when results are calculated as the distribution between membrane and cytosol, as in Figure 9(A). ATRAP was originally identified in a yeast two-hybrid screen as an interacting protein for the angiotensin II type I receptor. Residues 110–120 of ATRAP were found to interact with the C-terminus of the receptor [32,51]. Interaction between ATRAP and the receptor is observed in the basal state and is enhanced upon stimulation with angiotensin II. It is generally thought that ATRAP increases internalization of the angiotensin II type I receptor leading to a suppression of angiotensin II-mediated signalling, including phospholipase C activation [51,56,57]. In the heart, angiotensin II stimulation promotes cardiac hypertrophy, concomitant with a decrease in cardiac ATRAP expression, and transgenic mice overexpressing ATRAP are protected from cardiac hypertrophy provoked by chronic angiotensin II infusion . The highest expression of ATRAP is found in the kidney, where it is localized in the proximal tubule, particularly the brush border. Blood pressure and plasma volume are raised in mice lacking ATRAP. Thus ATRAP has a protective effect on the deleterious effects of angiotensin II. Since both ATRAP and RdgBβ are highly expressed in the heart and the kidney, it would suggest that RdgBβ could modulate the function of ATRAP. Angiotensin II stimulates phospholipase C via Gq and this results in the downstream activation of PKC. The effects of angiotensin II on cardiac hypertrophy require a chronic stimulation with angiotensin II, and mice are infused with angiotensin II for 14 days to induce cardiac hypertrophy . Thus ATRAP interactions with RdgBβ most probably occur following chronic stimulation with angiotensin II. We suggest that PMA mimics this chronic stimulation by angiotensin II. In addition to interacting with the angiotensin II type I receptor, ATRAP can bind a number of other proteins which are independent of angiotensin II signalling. This includes CAML (calcium-modulating cyclophilin ligand)  and RACK1 (receptor for activated C kinase-1) . Thus the consequences of the RdgBβ interaction with ATRAP may go beyond angiotensin II signal transduction. Since RdgBβ possesses PI transfer activity, binding of RdgBβ to ATRAP at the membrane could promote PI exchange at the opposing membrane (Figure 10, see inset).
Molecular interactions between RdgBβ, 14-3-3 proteins and ATRAP leading to lipid exchange
The results of the present study indicate that RdgBβ-sp1 is an unusual protein whose expression and activity is tightly regulated in cells. The C-terminus of RdgBβ-sp1 is highly disordered and contains PEST sequences that were originally identified as degradation signals owing to their abundance in rapidly degraded proteins [39,40]. Proteins with unstructured regions predominantly have signalling or regulatory roles and are often reused in multiple pathways to produce different physiological outcomes . The C-terminus of RdgBβ has all the properties found in unstructured proteins where proteolytic degradation contributes to controlling its abundance. We suggest that RdgBβ plays a central role in angiotensin II-mediated cell signalling both in the heart and the kidney, two organs where RdgBβ proteins are enriched together with ATRAP.
ADP-ribosylation factor 1
angiotensin II type I receptor-associated protein
deleted in colorectal cancer
human embryonic kidney
PI transfer protein
protein kinase A
protein kinase B
protein kinase C
retinal degeneration type B
small interfering RNA
Tris-buffered saline containing Tween 20
Kathryn Garner and Shamshad Cockcroft conceived the study, performed the bioinformatic analysis and wrote the paper. Kathryn Garner, Michelle Li and Natalie Ugwuanya performed the experimental work and participated in discussion and data analysis.
We thank Professor Carol MacKintosh for suggesting and providing the reagents for the 14-3-3 far-Western blots, and Professor Justin Hsuan and Dr Marco Lopez-Ilasaca for providing the plasmids for human RdgBβ and for ATRAP respectively. We would like to thank Clive Morgan, Alison Skippen, Roman Holic and Nicolas Carvou who have contributed to the early parts of the project.
This work was funded by the British Heart Foundation [grant number FS/08/044/25498]