Golgi-Associated Protein Kinase C-ε Is Delivered to Phagocytic Cups: Role of Phosphatidylinositol 4-Phosphate

Protein kinase C-ε (PKC-ε) regulates IgG-dependent phagocytosis by mediating the membrane fusion necessary for pseudopod extension (1). Wild type, but not PKC-ε^−/−, macrophages spread on IgG surfaces, adding membrane as measured by increased membrane capacitance; re-expression of PKC-ε in null cells restores spreading, confirming a role for PKC-ε in membrane fusion (2). This is consistent with PKC-ε involvement in exocytosis, including release of dense granules from chromaffin cells (3), secretion by lacrimal cells (4), and insulin release (5). How PKC-ε mediates membrane fusion is an open question.

PKC-ε is a member of the Ca²⁺-independent, diacylglycerol-activated subset of PKCs and is unique in that it contains an actin-binding motif. This implicates it in cytoskeletal functions; it is associated with actin filaments and keratin and is involved in integrin-dependent cell motility and spreading (6). Interestingly, its overexpression in NIH3T3 fibroblasts elicits a transformed phenotype (7) and its downregulation inhibits metastasis (8). As metastatic cells are quite motile, PKC-ε is likely involved in cell motility in cancer models. Indeed, PKC-ε is upregulated in prostate (8) and breast (9) cancers and is considered an oncogene (10). However, its mechanism of action has yet to be determined.

Structurally, all PKCs consist of regulatory and catalytic domains connected by a hinge. Catalytic domains are homologous, containing the ATP binding pocket as well as the substrate- binding site. The regulatory domains vary and the isoforms are categorized by their lipid and calcium-binding properties (11). All regulatory regions contain a pseudosubstrate sequence that, when bound in the active site, maintains the enzyme in an inactive state. Interaction of the regulatory domain with lipids and receptor for activated C kinases causes conformational changes that release the pseudosubstrate leading to enzyme activation. Although studies with chimeric PKCs suggested that the regulatory domain targets the catalytic domain to its site of action (12, 13), numerous exceptions have been reported. For example, PKC-βII V5 domain (in the catalytic region) interacts with RACK1 for PKC targeting in cardiac myocytes (14), the isolated regulatory domain of PKC-ε supports neurite extension (15), and catalytically inactive PKC-α activates phospholipase D (16). Thus, intact PKC can signal via regulatory domain–targeting of the catalytic region, the regulatory domain alone can promote cellular responses, and catalytically inactive PKC can mediate signaling. The reports detailed in this study establish that the pseudosubstrate of PKC-ε binds phosphatidylinositol 4-phosphate (PI4P), tethering it to the trans-Golgi network (TGN). This TGN-associated pool of PKC-ε is necessary for its concentration at the phagocytic cup and Fc [γ] receptor (FcγR)–dependent membrane fusion. To our knowledge, this is the first demonstration that a PKC pseudosubstrate–lipid interaction directs PKC concentration and membrane fusion in response to receptor ligation.

ACK lysis buffer: 0.15 M NH₄Cl, 1 mM KHCO₃, and 0.1 mM EDTA (pH 7.4). HBSS: HBSS⁺⁺: HBSS containing 4 mM sodium bicarbonate, 10 mM HEPES, and 1.5 mM each CaCl₂ and MgCl₂. Bone marrow media: DMEM containing 10% FBS, 20% l-cell conditioned media, and 0.03% NaHCO₃, and 50 μg/ml gentamicin. PIK93 was from Echelon Biosciences (Salt Lake City, UT). Wortmannin (Wt) and LY29004 (LY) were from Cayman Chemical (Ann Arbor, MI).

PKC-ε^−/+ heterozygotes on the C57/Bl6 background were purchased from The Jackson Laboratory (stock# 004189; Bar Harbor, ME) and bred in the Albany Medical Center Animal Resources Facility. Heterozygotes were crossed to produce the wild type and PKC-ε^−/− mice. All animal procedures were approved by the Albany Medical Center Institutional Animal Care and Use Committee. Cells obtained from one mouse represent one independent experiment.

PKC-ε^+/+ and PKC-ε^−/− mice were euthanized and their femurs removed. The bone marrow was removed and the RBC lysed with ACK lysis buffer. Bone marrow stem cells were differentiated by incubation in phenol-red free (high glucose) DMEM containing 10% FBS, 20% l-cell conditioned media, and sodium bicarbonate. Bone marrow–derived macrophages (BMDM) were used 7–14 d after harvesting.

The RAW cell subclone, LacR/FMLPR.2 (RAW cells) (17) was used. Cells were maintained in (high glucose) DMEM, nonessential amino acids, sodium pyruvate, GlutaMAX (all from Life Technologies), and 10% newborn calf serum (HyClone).

Oligonucleotides were designed to encode the pseudosubstrate domain of PKC-ε (aa 147–165) or PKC-δ (aa 136–154) flanked by BamHI and EcoRI sites. Complementary oligonucleotides were annealed in 10 mM Tris (pH 8), 50 mM NaCl, 1 mM EDTA at 50 μM, phosphorylated, and subcloned into pGEX-2T (GE Healthcare) linearized with BamHI and EcoRI. Constructs were transformed into Rosetta 2 cells (EMD Millipore). Isopropyl-β-d-thiogalactopyranoside (1 mM) was added to logarithmically growing bacteria for 2 h. Protein was extracted with BPERII (Thermo Fisher) supplemented with 10 mM MgCl₂, protease inhibitors, 1% lysozyme, and DNaseI. GST fusion proteins were purified from clarified bacterial lysates using glutathione-agarose beads (Sigma-Aldrich) and glutathione elution according to the manufacturer’s directions.

Biotinylated PolyPIPosomes were purchased from Echelon Bioscience and the assay done according to the manufacturer’s instructions. Briefly, incubations containing 1 μl of liposome and 5 nMol GST-peptide in 500 μl TBST were rotated (room temperature) for 60 min. Liposomes were immobilized on Streptavidin magnetic beads (Dynabeads; Life Technologies) in a magnetic field, washed twice with TBST, and subjected to Western blot analysis for GST. An aliquot of the input was run in parallel to confirm equivalent addition of GST protein.

Acid-washed 2 μm borosilicate beads (Duke Standards; Thermo Scientific) were sequentially coated with poly-l-lysine, dimethylpimelimidate, 12N HCl, and Alexa 568-conjugated (IgG-free) BSA. Free active groups were blocked by a 1 h incubation in 0.5 M Tris, (pH 8) before opsonizing with rabbit anti-BSA IgG.

Briefly, 15 mm glass coverslips were rinsed with ethanol and dH₂O and dried with compressed air to prepare for coating. The coverslips were aldehyde silane-coated by sequential incubation with an oxygen plasma cleaner (5 min) and vaporized silane [11-(Triethoxysilyl) undecanal, 70°C, 1 h, sealed in curing oven]. To covalently link proteins to the silane-coated coverslips, the coverslips were washed with PBS and incubated with 25 μg/ml normal rabbit IgG (60 min, room temperature). After protein attachment, the remaining active sites were blocked 0.5 M Tris (pH 8) (4°C, overnight). Before use, the coverslips were washed twice in HBSS⁺⁺.

Synchronized phagocytosis was performed as detailed previously detailed (2). Briefly, cells were pretreated (37°C, 30 min) with inhibitors (100 nM Wt, 50 μM LY, 0.1 or 10 μM PIK93, DMSO served as the no-treatment control) and cooled on ice (15 min). IgG-opsonized beads were added (four beads per cell) and allowed to bind (15 min, 4°C), then transferred to a 37°C water bath. Cells were fixed after 5 min (4% PFA/PBS) and imaged by spinning disk confocal microscopy (Olympus IX81-DSU) with a 100×/1.4 N.A. oil objective; Hamamatsu electron-multiplying charge-coupled device camera, driven by MetaMorph software (Molecular Devices, Sunnyvale, CA). Image analysis for concentration at cups and nascent phagosomes in RAW cells was performed using ImageJ/FIJI and the Localization Index calculated as detailed (2). Briefly, the target associated membrane was selected using the three-pixel brush and the intensity normalized to an equivalent area of noninvolved membrane.

For live imaging, the ice steps were eliminated. Cells were pretreated with inhibitors, brought into focus on the stage of the spinning disk, and IgG-opsonized beads were added. Images were taken every 5 s for 10 min. The rate of bead internalization was calculated as the number of frames from the first indentation of the plasma membrane through the first frame showing completely enclosed particles × 5 (seconds between frames) (2).

The GFP plasmid was purchased from Clontech (Mountain View, CA). GFP-P4M and GFP-hSac1-K2A (K2A) were generous gifts from T. Balla (18), and P. Mayinger (19), respectively. Construction of the εK437W and pBM-GW-eGFP IRES-Puro vector and retroviral plasmid for PKC-ε-GFP have been published (2).

Delivery of plasmids for expression of exogenous proteins was done by transfection for cell lines and viral transduction for primary macrophages.

RAW cells were transfected with Mirus TransIT-neural at a 3:1 (Mirus:DNA) ratio. A detailed methodology has been published (2, 20).

The details of retroviral pseudovirion production and BMDM transduction have been published (2). Briefly, Phoenix ecotropic packaging cells (National Gene Vector Biorepository, Indianapolis, IN) were transfected with the viral plasmids and the supernatants harvested 72 h postinfection. Viral transduction of bone marrow cells was done by spinfection 24 and 48 h after harvesting. Puromycin selection began ∼96 h after harvesting and cells differentiated. BMDM were used 8–14 d after isolation; ∼90% of the selected cells expressed the transduced protein.

Whole cell patch-clamping was done essentially as described (2). Briefly, BMDM were removed from plates and treated in suspension (HBSS⁺⁺) with inhibitors (30 min, room temperature), allowed to settle onto IgG-coated coverslips (20 min), then placed at 37°C to promote cell spreading. At 15 min, the coverslips were transferred to a holder and bathed in a balanced salt solution [10 mM HEPES, (pH 7.4), containing 145 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 2 mM MgCl₂, and 5 mM glucose] for capacitance measurements. Cells were maintained at a 0 mV holding potential during experiments. Data are reported as mean capacitance ± SEM. Statistical significance was determined by one-way ANOVA with Bonferroni correction.

Live imaging of PKC-ε-GFP during IgG-mediated phagocytosis in RAW 264.7 cells and primary macrophages (MØ) reveal two temporally distinct waves of concentration, early at the phagocytic cup and later, as a flash on nascent phagosomes [(2, 20) and Supplemental Video 1]. The (diacylglycerol binding) C1B and pseudosubstrate (εPS) domains of PKC-ε are necessary and sufficient for concentration at both sites (2). C1B binds DAG, released by phospholipase C-γ1 (20); the docking partner for εPS is unknown.

The pseudosubstrate of PKC-ε contains 12 basic amino acids, including two triplets (Fig. 1). Such polybasic clusters are a signature of phosphoinositide phosphate (PIP) binding proteins (21), leading us to test the ability of εPS to bind PIPs. Using GST-conjugated peptides and PolyPIPosomes (liposomes that mimic the composition and curvature of cellular membranes), we identified PI4P and PI(3,5)P₂ as binding partners for εPS (Fig. 1). The homologous region of PKC-δ, which neither contains polybasic triplets nor concentrates during phagocytosis (1), does not bind PIPs.

Given that εPS binds PI4P and PI(3,5)P₂, we asked if LY and/or Wt block PKC-ε concentration at cups. This pair of inhibitors was chosen as both inhibit PI3 kinase but Wt also blocks the type III PI4 kinases. Whereas Wt blocks PKC-ε concentration at cups and nascent phagosomes, LY inhibits PKC-ε concentration only at nascent phagosomes (Fig. 2A, 2B). Additionally, Wt, but not LY, significantly slows the rate of target internalization (Fig. 2C), consistent with a role for PI4P in PKC-ε–dependent phagocytosis. To determine if Wt depletes PI4P under the conditions that inhibit phagocytosis, we expressed the PI4P reporter P4M (18) in RAW cells and quantified the effect of Wt and LY on P4M intensity. Wt, but not LY, significantly reduces perinuclear PI4P levels (Fig. 3A) consistent with the hypothesis that a PI4P–PKC-ε interaction is necessary for PKC-ε concentration at cups and regulation of phagocytosis. When the same experiment was done using PKC-ε-GFP, we found a modest, but not statistically significant, decrease in Golgi-associated PKC-ε (Fig. 3B). This apparent disparity may reflect the fact that Wt prevents PI4P formation, which would have little effect on that pool of PI4P already bound to PKC-ε.

Given the body of literature concluding the PI3K is necessary for phagocytosis (22), our results suggesting that the Wt effect at cups is PI3K independent was somewhat surprising. The disparity likely arises from the differences in experimental design. Fig. 2B reports the rate of bead internalization, extracted from live imaging; published studies quantify the number of particles internalized after 60 min. Although LY does not decrease the initial rate of phagocytosis, it does inhibit phagocytosis at 60 min (data not shown), in agreement with the literature. Notably, our data parallel the findings that the exocytosis phase of phagocytosis is PI3K independent (23). Thus, PI3P acts later in phagocytosis, likely at the level of membrane/receptor recycling.

Given that PKC-ε translocation is necessary for membrane fusion for IgG-mediated phagocytosis (2), if PI4P is required for PKC-ε translocation, then inhibiting PI4 kinase (PI4K) should block FcγR-dependent membrane fusion. Thus, we tested the effect of PIK93, an inhibitor of the Wt-sensitive PI4Ks (24) on macrophage spreading on IgG surfaces. Using whole cell patch clamping, we determined that cells treated with Wt or PIK93 have significantly lower capacitance than solvent-treated controls (Fig. 4A). The fact that the capacitance of WT cells treated with either Wt or PIK93 is equivalent to that of PKC-ε^−/− BMDM supports the hypothesis that PI4P is required for PKC-ε–mediated membrane fusion during phagocytosis.

Patch clamping used 10 μM PIK93 as it inhibits both Wt-sensitive PI4KIIIα and IIIβ. As PIK93 has an isoform selectivity, we used 0.1 and 10 μM to test the dose dependency: 0.1 μM preferentially blocks PI4K IIIβ with 10 μM needed to substantively inhibit PI4KIIIα (24). Both concentrations reduce PI4P Golgi concentration in RAW cells (Fig. 3A). When tested for their effects in primary BMDM, both concentrations significantly reduce PKC-ε concentration at cups (Fig. 4B, 4C) and slow IgG-mediated phagocytosis (Fig. 4D).

Live imaging of BMDM expressing PKC-ε-GFP reveals a decrease in perinuclear PKC-ε intensity during active phagocytosis (Supplemental Video 1). PKC-ε concentration at the Golgi was dramatically reduced by 0.1 μM PIK93 (Fig. 4B, Supplemental Fig. 1, Supplemental Video 2), consistent with its selective inhibition of PI4KIIIβ, the isoform producing PI4P in the Golgi (25). The hypothesis that PKC-ε associates with PI4P in the TGN is supported by several lines of evidence: ∼20% of PKC-ε in macrophages is membrane associated (26), PKC-ε concentrates in the perinuclear region in macrophages [(2), Supplemental Video 1 and arrowheads, Fig. 4A], and PI4P is enriched in the TGN (25). If PKC-ε binds PI4P in the Golgi, then PI4P and perinuclear PKC-ε should colocalize. This is the case, as evidenced by a Pearson correlation of 0.86 (40 cells from three independent experiments) in macrophages coexpressing the PI4P reporter, P4M, and PKC-ε (Fig. 5A); colocalization of PKC-ε with the Golgi marker GM130 (Supplemental Fig. 2) with a Pearson coefficient of 0.64 confirms that perinuclear PKC-ε is indeed Golgi associated. Notably, the perinuclear intensity of PKC-ε is significantly lower in cells coexpressing P4M (Fig. 3B), which would be expected if P4M and PKC-ε compete for Golgi PI4P. Second, if PKC-ε is tethered to the Golgi via PI4P, then selective removal of Golgi-associated PI4P should reduce perinuclear PKC-ε. This was tested by coexpressing the Golgi-directed PI4 phosphatase, K2A (27) with either P4M or PKC-ε. K2A expression substantively reduces both perinuclear PI4P (Fig. 3A) and PKC-ε (representative image, Fig. 5B, quantitation Fig. 3B), paralleling the effects of 0.1 μM PIK93 on Golgi-associated PI4KIIIβ (24) and perinuclear PKC-ε in RAW cells (Fig. 3) and BMDM (note lack of perinuclear PKC-ε in Supplemental Video 2 compared with Supplemental Video 1, arrowheads in Fig. 4B, Supplemental Fig. 1). Additionally, if the pool of PKC-ε that concentrates at cups originates at the Golgi, we would predict that expression of K2A would prevent or reduce that concentration. Indeed, coexpression of K2A and PKC-ε revealed a statistically significant decrease in PKC-ε at cups in cells expressing K2A compared with those expressing PKC-ε alone (Fig. 5C). Finally, if Golgi-associated PKC-ε traffics to the cup on microtubules, nocodazole should prevent the addition of membrane during cell spreading on IgG surfaces. Indeed, nocodazole inhibits membrane addition as determined by whole cell patch clamping (Fig. 6). Together, these data support the conclusion that PKC-ε binds PI4P on Golgi and this interaction is necessary for the translocation of PKC-ε to the phagocytic cup and the membrane fusion necessary for efficient phagocytosis.

PKC-ε concentrates in two waves during phagocytosis, at the forming phagosome (the cup) and after phagosome closure, appearing as a flash [(2, 20) and Supplemental Video 1]. The C1B and εPS domains were shown to be necessary and sufficient for both waves of concentration (2). C1B binds diacylglycerol but the binding partner for εPS was unknown. We narrowed the required εPS sequence to 19 aa [diagrammed in Fig. 1, identified in (2)] and noted the predominance of basic amino acids, a signature of PIP binding (21). Although the polybasic nature of εPS might suggest electrostatic binding, this is clearly not the case as εPS binds PI4P but not the isomers PI3P and PI5P, compounds with the same net charge. Similarly, there is a specificity of PI(3,5)P₂ over isomers PI(3,4)P₂ and PI(4,5)P₂ and no apparent interaction with the most negatively charged PIP, PI(3,4,5)P₃. Precedence for such selectivity comes from the PH domain literature. Indeed, this selectivity has been exploited in generating a toolbox of lipid reporters (28), including the P4M used in this study.

Based on the liposome binding assays, and the acknowledged role for PI3K in IgG-mediated phagocytosis (22), we hypothesized that the εPS-PI(3,5)P₂ interaction was responsible for PKC-ε concentration. Thus, the differential effects of Wt and LY on PKC-ε concentration at the cup were surprising and focused our attention on the ability of Wt to inhibit the Type III PI4 kinases. PI4P modulators, combined with the P4M lipid reporter provided us with the tools with which to modulate PI4P (PIK93, K2A), monitor its presence (P4M), and reproduce the Wt findings. Notably, all treatments that decreased Golgi-associated PI4P reduced perinuclear PKC-ε, its concentration at cups, and slowed phagocytosis. Conversely, LY did not substantively reduce perinuclear PKC-ε (Fig. 2) nor did it reduce cup-associated PKC-ε or alter phagocytic rate, providing a control lipid inhibitor that did not alter PI4P. Our conclusion, based on the totality of the results, is that PKC-ε associates with PI4P in the TGN and it is this pool that concentrates at the cup. One argument against this model is that brefeldin A, a drug that disrupts the Golgi, does not alter IgG-mediated phagocytosis (29). When tested in our system, brefeldin A had no effect on the PKC-ε perinuclear localization (data not shown), consistent with reports that it neither disrupts TGN trafficking (30) nor inhibits IgG-mediated phagocytosis (29). This provides additional evidence that PKC-ε is localized at the TGN. Although not directly tested, several pieces of evidence argue that PI4KIIIβ is producing the PI4P for PKC-ε tethering. These include: localization of PI4KIIIβ to the trans Golgi (31) and loss of Golgi-associated PI4P or disruption of phagocytosis at a concentration of PIK93 that targets PI4KIIIβ (24) (Figs. 2, 3).

The simplest model that incorporates our data and the literature is that the PKC-ε that concentrates at the cup originates at the Golgi and travels on vesicles in a PI4P-dependent manner; PI4KIIIβ facilitates the Golgi to plasma membrane trafficking (25, 31). PKC-ε–positive vesicles would fuse into the cup, providing a membrane for pseudopod extension, explaining the loss of membrane fusion in response to PI4K inhibition (Fig. 4A). As Golgi-to-plasma membrane vesicles traffic on microtubules, one test of the model would be to determine the effect of nocodazole on membrane fusion. Whole cell patch clamping revealed that nocodazole prevents FcγR-mediated membrane fusion, with nocodazole-treated wild type cells having significantly lower capacitance than their untreated counterparts, equivalent to that of PKC-ε^−/− BMDM (Fig. 6). These data support a vesicular trafficking model. That re-expression of full length, but not catalytically inactive, PKC-ε in PKC-ε^−/− cells restored membrane fusion bridges these studies to our previous findings that catalytically inactive PKC-ε translocates, but does not support phagocytosis (2). Thus, the regulatory domain directs vesicle targeting, with catalytic activity needed for membrane fusion.

In conclusion, although these studies focused on PKC-ε at the cup, it is of interest to consider that treatments that block PKC-ε association with the Golgi, block its concentration at the cup, and retard phagocytosis have little effect on PKC-ε concentration at nascent phagosomes (Fig. 2, Supplemental Fig. 3). This suggests there may be two pools of PKC-ε: one at the Golgi regulated by εPS-PI4P and one potentially regulated by εPS-PI(3,5)P₂ binding (Fig. 1). The differential regulation of these two putative pools is intriguing and under investigation.

We thank Dr. Lynn Cassimeris for insightful analysis of our real-time imaging. Additionally, we thank Dr. Joseph Mazurkiewicz and the Albany Medical College Imaging Core Facility for use of the confocal microscopes, Kate Tubbesing for help with Imaris imaging, Dr. Elaine Raines and laboratory for guidance on viral constructs and production of pseudovirions, Tiffany Wood-Cammock for generation of preliminary data, and Dr. Daniel J. Loegering and Dr. James Drake for helpful discussions and reading the manuscript. Finally, many thanks to Deborah Moran for work on preparing the manuscript and the Johnathan R. Vasiliou Foundation for generous support of this work.

The authors have no financial conflicts of interest.