The Carboxyl-terminal Domain of Atypical Protein Kinase Cζ Binds to Ceramide and Regulates Junction Formation in Epithelial Cells^,

Abstract

Atypical protein kinase Cs (PKCs) (aPKCζ and λ/ι) have emerged as important binding partners for ceramide, a membrane-resident cell signaling lipid that is involved in the regulation of apoptosis as well as cell polarity. Using ceramide overlay assays with proteolytic fragments of PKCζ and vesicle binding assays with ectopically expressed protein, we show that a protein fragment comprising the carboxyl-terminal 20-kDa sequence of PKCζ (C20ζ, amino acids 405–592) bound to C16:0 ceramide. This sequence is not identical to the C1 domain (amino acids 131–180), which has been suggested to serve as a potential ceramide binding domain. Using immunocytochemistry, we found that a C20ζ protein fragment ectopically expressed in two epithelial cell types (neural progenitors and Madin-Darby canine kidney cells) co-distributed with ceramide. Stable expression of C20ζ-EGFP in Madin-Darby canine kidney cells disrupted the formation of adherens and tight junctions and impaired the epithelium integrity by reducing transepithelial electrical resistance. Disruption of cell adhesion and loss of transepithelial electrical resistance was prevented by incubation with C16:0 ceramide. Our results show, for the first time, that there is a novel ceramide binding domain (C20ζ) in the carboxyl terminus of aPKC. Our results also show that the interaction of ceramide with this binding domain is essential for cell-to-cell contacts in epithelia. Therefore, ceramide interaction with the C20ζ binding domain is a potential mechanism by which ceramide and aPKC regulate the formation of junctional complexes in epithelial cells.

Epithelial cells play essential roles in multicellular organisms by forming physiological and mechanical barriers and controlling tissue architecture, because they acquire apicobasal and cell-to-cell (planar) polarity (1, 2). Adherens junctions (AJs)2 and tight junctions (TJs) are major structures responsible for cell-to-cell adhesion in epithelial cells (3). The regulation of junction formation requires endocytosis, redistribution, and recycling of junctional proteins, such as E-cadherin (4), and ZO-1. Many factors, including EGF, EGFR, Src kinase, Rho family GTPases Cdc42 and Rac1, and atypical PKC (aPKC), have been found to regulate junction formation (5–9). In Madin-Darby canine kidney (MDCK) cells, Cdc42 modulates AJs by regulating E-cadherin ubiquitination and degradation (9), whereas aPKC directly localized at TJs is required for the asymmetric differentiation of the premature junction complex during epithelial cell polarization (1, 10).

The protein kinase C (PKC) family comprises serine/threonine kinases, which consist of a carboxyl-terminal catalytic domain and an amino-terminal regulatory domain (Fig. 1A). The regulatory domain includes an inhibitory pseudosubstrate domain and allosteric sites for activation by phosphatidylserine and, depending on the isoform, calcium (C2 domain) and/or diacylglycerol (C1 domain). aPKC is a subfamily of PKC, which consists of the isoforms ζ and λ/ι. The aPKC isoforms contain only half of the C1 domain, and hence, their activity is not affected by calcium or diacylglycerol/phorbol esters (see Fig. 1A and Refs. 11–13).

FIGURE 1.

Binding of ceramide to the COOH terminus of PKCζ. A, primary structure of aPKC, the caspase 3 proteolytic fragment ζCasp II, and the NH₂-terminal deletion mutant C20ζ-EGFP. B, 2 μg of recombinant His-tagged PKCζ was proteolytically digested by 20 ng of recombinant caspase 3. Proteolysis by caspase 3 occurred first after amino acid 239 (4-h incubation) and then after amino acid 459 (10-h incubation, ζCasp II). C, binding to ceramide spotted on nitrocellulose (overlay assay). FL PKCζ and the COOH-terminal proteolytic fragment ζCasp II bound to C16 ceramide. D, C16 ceramide vesicle binding assay (LIMAC). Ectopically expressed C20ζ-EGFP prepared from a cell lysate was bound to ceramide vesicles; EGFP was not. Protein was detected using anti-aPKC and anti-GFP antibodies. Lanes 1–3, loading input for ceramide vesicles; lanes 4–6, eluate of vesicle binding columns (output). Lanes 7 (input) and 8 (output) show that PKCζ-EGFP did not bind to vesicles prepared with sphingomyelin (SM) instead of ceramide. E, subcellular fractionation of cells expressing FL PKCζ-EGFP or C20ζ-EGFP.

Apart from its function in apoptosis (13–15) and cell growth (16), aPKC has been found to play a pivotal role in cell polarity, both in neuroepithelial cells (17–20) or other epithelial cell types (1, 10). Consistently, the gene knock-out of aPKC shows loss of cell junction formation and detachment of neural progenitor cells from the neuroepithelium (8, 21–23). We and others have found that the sphingolipid ceramide activates aPKC, recruits it to structured microdomains, and regulates cell polarity and motility (24–28). Using lipid vesicle-mediated affinity chromatography (LIMAC) assays, we showed for the first time that ceramide directly binds to aPKC (25). Yet which domain of aPKC binds to ceramide is not known.

Using lipid overlay and LIMAC assays, we show here that a COOH-terminal 20-kDa domain of PKCζ (C20ζ) binds to ceramide. Similar to its full-length counterpart, the C20ζ protein fragment resides in cellular membranes, where it co-distributes with ceramide in both C17.2 (neural progenitor) and MDCK cells. To study the function of this ceramide binding domain, we established a stably transfected MDCK cell line expressing C20ζ-EGFP. In these cells, the protein level of E-cadherin is reduced, and the cellular distribution of E-cadherin, ZO-1, and β-catenin is disrupted when compared with EGFP-transfected cell lines. Further, transepithelial electrical resistance (TER) assays show that the C20ζ-EGFP cell line has reduced impedance when compared with the control cell line expressing EGFP. This finding suggests that the C20ζ protein fragment is a dominant negative mutant of PKCζ. The effects of this dominant negative mutant can be, at least partially, rescued by incubation with C16:0 ceramide, suggesting that ceramide regulates aPKC and aPKC-dependent cell junction formation by interaction with the COOH-terminal domain.

EXPERIMENTAL PROCEDURES

Materials—Dulbecco's modified Eagle's medium was purchased from Cellgro (Herndon, VA). Purified recombinant human PKCζ, anti-PKCζ rabbit antiserum, and goat anti-rabbit IgG horseradish peroxidase conjugate were obtained from Sigma. N-Palmitoyl-d-erythro-sphingosine (C16:0 ceramide) was purchased from Avanti Polar Lipids Inc. (Alabaster, AL). polyclonal anti-PKCζ rabbit IgG (catalogue number sc-216) was from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Polyclonal anti-phosphorylated (phospho-Thr^403/410) PKCζ/λ rabbit IgG and polyclonal anti-phosphorylated (phospho-Ser⁹) GSK-3β rabbit IgG were purchased from Cell Signaling (Beverly, MA). Polyclonal anti-ceramide mouse IgM MAS00020 was from Glycobiotech (Kuekels, Germany). Polyclonal anti-ceramide rabbit IgG was generated in our laboratory, as previously described (29). Donkey anti-mouse, anti-rabbit, and anti-goat IgG Cy2, Cy3, and Cy5 conjugates; Cy2-conjugated donkey anti-mouse IgM (μ-chain-specific); Cy3-conjugated goat anti-mouse IgG (Fc γ fragment-specific); and goat anti-mouse IgG horseradish peroxidase conjugate were purchased from Jackson ImmunoResearch (West Grove, PA). Myristoylated PKCζ pseudosubstrate inhibitor peptide and His-tagged recombinant PKCζ were from Calbiochem. The magnetic activated cell sorting kit, including Annexin V-conjugated magnetic beads, was from Miltenyi Biotec (Auburn, CA). The Lipofectamine 2000 transfection reagent was obtained from Invitrogen. All reagents were of analytical grade or higher.

Construction of PKCζ Mutant Plasmids and Transfection of Cells—For the construction of C20ζ-EGFP, cDNA was amplified from full-length PKCζ using primers 5′-aactcgagatgtatatcgcccccgaaatcctg-3′ (KpnI site) and 5′-atctagactcacggactcctcagcagacag-3′ (XbaI site). The amplification product was ligated into pEGFP-N1 vector between the restriction enzyme sites KpnI and XbaI. The hemagglutinin-tagged PKCζ-C20 was constructed by inserting a PCR product into pcDNA3.1/myc-his (B), for which the primers used were 5′-aaggtaccatgggctacccatacgatgttccagattacgcttatatcgcccccgaaatcctg-3′ (KpnI) and 5′-aatctagactcacggactcctcagcagacag-3′ (XbaI site). Transfections were performed using the Lipofectamine 2000 procedure following the manufacturer's (Invitrogen) protocol.

Cell Culture—MDCK II and C17.2 cells were generous gifts from Drs. Quansheng Du (Medical College of Georgia, Augusta, GA) and Evan Snyder (Burnham Institute for Medical Research, La Jolla, CA), respectively. Cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum, penicillin, and streptomycin in a 5% CO₂ atmosphere at 37 °C. To generate cell lines stably expressing EGFP or C20ζ-EGFP, MDCK cells were transfected with either EGFP (Promega) or C20ζ-EGFP cDNA and selected in Dulbecco's modified Eagle's medium with 1.5 mg of Geneticin/ml of medium (Invitrogen) for 2 weeks. Single colonies were isolated, subjected to fluorescence-activated cell sorting for the expression of EGFP, and maintained in Dulbecco's modified Eagle's medium with 0.8 mg/ml Geneticin.

Partial Proteolysis of PKCζ and Lipid Binding Assays—Recombinant His-tagged PKCζ (2 μg) was proteolytically digested in 50 mm Tris/HCl buffer, pH 7.4, 1 mm dithiothreitol, 5 mm EDTA by adding 20 ng of recombinant caspase 3 (EMD), followed by incubation at 37 °C. The reaction was stopped after 4 or 10 h by adding 0.5 mm of the caspase 3 inhibitor benzyloxycarbonyl-VAD-fluoromethyl ketone (final concentration). The NH₂ terminus was removed by binding to His Select zinc chelate chromatography on 96-well plates (Sigma). The COOH-terminal fragments were recovered from the supernatant and used for proteomics analysis, nitrocellulose membrane lipid overlay, and LIMAC assays as described previously (25). Proteomics analysis confirmed that the COOH-terminal fragment obtained after 4 h of caspase 3 digestion started with amino acid 240, whereas that obtained after 10 h of digestion started with amino acid 460 (fragment similar to C20ζ).

Lipid overlay binding assays were performed as described in our previous publication (29). Briefly, C16:0 ceramide was spotted onto enzyme-linked immunosorbent assay plates or nitrocellulose membranes, and nonspecific binding sites were saturated by incubation with 1% bovine serum albumin or 5% dry milk in PBS, respectively. COOH-terminal fragments equivalent to 500 ng of cleaved PKCζ were added and incubated overnight at 4 °C. Plates or membranes were washed with PBS and then further incubated with anti-PKCζ antibody in 1% bovine serum albumin/PBS or 5% dry milk/PBS. Bound full-length PKCζ or COOH-terminal fragments were detected using anti-rabbit IgG secondary antibody conjugated with horseradish peroxidase, followed by peroxidase substrate or the ECL reaction.

The LIMAC reaction was performed with EGFP, PKCζ-EGFP, and the C20ζ-EGFP protein fragment following procedures as described previously (25). In brief, ceramide vesicles were obtained by sonication and centrifugation of a mixture of equimolar amounts of phosphatidylserine and C16:0 ceramide. The resuspended pellet (large lipid vesicles) was incubated with detergent-free lysates obtained from MDCK cells stably expressing EGFP (control), full-length PKCζ-EGFP, or the C20ζ-EGFP protein fragment. The protein binding to Cer/phosphatidylserine vesicles in the magnetic activated cell sorting eluate was analyzed by SDS-PAGE and immunoblotting.

Immunocytochemistry Analysis—Cells were fixed with 4% p-formaldehyde/PBS, followed by permeabilization with 0.1% Triton X-100 for 5 min at room temperature. Nonspecific binding sites were saturated by incubation with 3% ovalbumin and 2% donkey serum in phosphate-buffered saline at 37 °C for 60 min. Cells were stained with primary and fluorescence-labeled secondary antibodies at a concentration of 5 or 10 μg/ml, respectively. Epifluorescence microscopy was accomplished using a Zeiss Axioplan Deconvolution microscope or a Zeiss Axiophot microscope equipped with a Spot digital camera. Confocal fluorescence microscopy was performed using a Zeiss LSM confocal laser-scanning microscope equipped with a two-photon argon laser at 488 nm (Cy2, green fluorescent protein) and 543 nm (Cy3) or 633 nm (Cy5), respectively.

Cell Impedance Assay—Measurements of TER were performed using an electrical cell-substrate impedance sensing system (ECIS) (Applied BioPhysics Inc., Troy, NY) (30). Briefly, equal numbers (50,000) of MDCK cells were plated on poly-l-lysine-coated gold electrode arrays (8W10E) and grown to confluence in the growth medium. TER was monitored for 30 min to establish a steady state at 1200–1400 ohms, and the electrical resistance across the monolayer was recorded over a period of 50 h. Values from each microelectrode were pooled at discrete time points and plotted versus time as the mean ± S.E.

RESULTS

The COOH-terminal 20-kDa Sequence of PKCζ (C20ζ) Comprises a Ceramide Binding Domain—Our previous studies showed that ceramide and ceramide analogs bind and activate aPKC (25, 31, 32). However, it has not been shown which protein domain interacts with ceramide. The identification of specific ceramide binding domains on aPKC is fundamental to its function in the regulation of cell polarity by ceramide. To determine the protein sequence that interacts with ceramide, we tested ceramide binding of defined fragments derived from aPKC. It has been shown in previous studies that limited proteolysis with caspase 3 generated defined NH₂- and COOH-terminal fragments of PKCζ (33). To localize the ceramide binding site, we used these proteolytic fragments for binding assays. Proteolysis of recombinant, His-tagged human PKCζ at two specific cleavage sites (amino acid 239 or 459) resulted in two fragments, a COOH-terminal 45-kDa (amino acids 240–592) and a 15-kDa fragment (amino acids 460–592; Fig. 1, A and B). The identity of the two fragments was confirmed by proteomics analysis and by immunoblotting, using an antibody that was raised against the COOH-terminal sequence of PKCζ (Fig. 1B and supplemental Table 1). The two fragments lacked the C1B domain that has been suggested to interact with ceramide (34).

We determined binding of full-length PKCζ and the 15-kDa fragment to ceramide using lipid overlay assays, which were performed by first spotting C16:0 ceramide on enzyme-linked immunosorbent assay plates and then detecting bound PKCζ by immunostaining. Full-length PKCζ showed a saturation kinetics indicating nonallosteric binding with a K_D of 25 nm (supplemental Fig. 2). Although binding of the 15-kDa fragment was detected, the limited intensity of the detection method (antibody against full-length PKCζ) did not allow for a kinetic analysis. Therefore, we used another method, overlay assays with C16:0 spotted on nitrocellulose membranes to detect binding of the 15-kDa fragment. To our surprise, the 15-kDa fragment bound to ceramide as tightly as full-length PKCζ despite lacking the C1B domain (Fig. 1C). To further confirm that the COOH-terminal domain of PKCζ binds to ceramide, the cDNAs of full-length PKCζ and a COOH-terminal fragment (C20ζ, amino acids 405–592) containing the sequence of the proteolytic 15-kDa fragment (Fig. 1A) were generated and expressed as EGFP fusion constructs in MDCK cells. The ectopically expressed full-length (FL) PKCζ-EGFP and C20ζ-EGFP were used for LIMAC (24, 25). This assay was based on the isolation of proteins specifically binding to ceramide-containing lipid vesicles. It was previously used in our laboratory to detect, for the first time, the association of full-length PKCζ with ceramide (25). Fig. 1D shows that endogenous aPKC, FL PKCζ-EGFP, and C20ζ-EGFP bound to the ceramide vesicles but not EGFP itself or actin as controls. FL PKCζ-EGFP did not bind to sphingomyelin or any other phospholipids used for the vesicle assay, as shown previously (25). Taken together, the lipid overlay and LIMAC assays confirmed that a COOH-terminal domain of PKCζ (C20ζ) bound directly to ceramide.

The COOH-terminal Protein Fragment C20ζ Co-distributes with Ceramide in MDCK Cells and Neural Progenitors—We and others have previously shown that aPKC associates with ceramide in various intracellular membranes and the plasma membrane (24, 26, 28). To determine if the COOH-terminal domain C20ζ is also translocated to cellular membranes and co-distributes with ceramide, we first performed a subcellular fractionation assay with MDCK cells expressing FL PKCζ-EGFP or C20ζ-EGFP. Fig. 1E shows that all of the expression products were predominantly distributed to the membrane fraction, indicating that the COOH-terminal domain of PKCζ contains a sequence that anchors the protein (fragment) to the membrane.

To define the subcellular distribution of this COOH-terminal domain, we performed immunocytochemistry with a neural progenitor cell line (C17.2 cells) and MDCK cells transiently expressing C20ζ-EGFP or hemagglutinin-tagged C20ζ. In these two epithelial cell types, the C20ζ protein fragment was co-distributed with ceramide at the cell membrane and intracellular membrane vesicles (Fig. 2, A and B, and supplemental Fig. 4). Transient expression of C20ζ-EGFP altered the distribution of ceramide to a more focal localization at the membranes, which indicates a close interaction between the COOH-terminal protein fragment and ceramide (Fig. 2B). EGFP expressed as a control was not distributed to the membrane and remained in the cytosol, or it was translocated to the nucleus (not shown). There was minimal co-distribution of C20ζ with sphingomyelin at the membrane, indicating that binding was specific for ceramide (supplemental Fig. 4). The translocation of C20ζ to cellular membranes and its co-distribution with ceramide, regardless of being conjugated to EGFP or hemagglutinin, was consistent with the results of the in vitro binding assays and suggested that the COOH-terminal domain mediates association of PKCζ with ceramide in living cells as well.

FIGURE 2.

C20ζ-EGFP co-distributed with ceramide in C17.2 and MDCK II cells. A, C17.2 cells were transiently transfected with C20ζ-EGFP. 48 h post-transfection, cells were fixed and stained for ceramide (red). Scale bars (two left panels), 5 and 1 μm, respectively. B, MDCK II cells were transiently transfected with C20ζ-EGFP. 48 h post-transfection, cells were fixed and stained for ceramide (red). Note that ceramide co-distributed with C20ζ-EGFP at the cell membrane and in the perinuclear region. Scale bar, 5 μm.

The C20ζ Protein Fragment Acts as a Dominant Negative Mutant of PKCζ and Disrupts the Formation of Cell-to-cell Junctions in MDCK Cells—To determine the biological function of ceramide binding to the COOH-terminal domain of PKCζ, we generated MDCK cells stably expressing C20ζ-EGFP and cells expressing EGFP or FL PKCζ-GFP as controls. Fig. 3A shows that, using both anti-PKCζ and GFP antibodies, the cell lines expressed FL PKCζ or C20ζ as EGFP-conjugated proteins. We hypothesized that the catalytically inactive C20ζ-EGFP protein fragment acts as dominant negative mutant competing with endogenous PKCζ for binding to ceramide. Since inhibition of PKCζ has been shown to reduce proliferation and induce apoptosis, we first tested the effect of C20ζ-EGFP on cell growth and survival in MDCK cells. Apoptosis and proliferation assays showed that stable expression of C20ζ-EGFP did not affect cell growth or death in MDCK cells (data not shown).

FIGURE 3.

C20ζ-EGFP down-regulated expression of E-cadherin, which was rescued by C16 ceramide. A, characterization of MDCK II cell lines stably expressing C20ζ-EGFP. Top, anti-PKCζ antibody; bottom, anti-GFP antibody used for immunostaining; EGFP and full-length PKCζ-EGFP (FL-EGFP)-expressing cells were also established. B, equal number of EGFP- and C20ζ-EGFP-expressing cells were grown to confluence and collected for immunoblot analysis detecting phospho- and total PKCζ (pPKCζ and tPKCζ), phospho- and total GSK-3β (pGSK-3β and tGSK-3β), E-cadherin (E-Cad), β-catenin (β-cat), and ZO-1. β-Actin was used as loading control. Note that E-cadherin was down-regulated in C20ζ cell lines. C, equal numbers of EGFP- and C20ζ-EGFP-expressing cells were grown to confluence and treated with a myristoylated pseudosubstrate inhibitor of PKCζ (PZI, 50 μm; EGFP-expressing cells) or 0.8 μm C16 ceramide (C20ζ-EGFP-expressing cells). Cells were collected, and immunoblot analysis was performed for E-cadherin. α-Tubulin (α-tub) was used as loading control. D, quantitation of C normalized to loading control.

We then determined the degree of phosphorylation of PKCζ at Thr⁴¹⁰, a site known to be phosphorylated by phosphatidylinositol-dependent kinase 1, which primes PKCζ for its further activation (20, 25). We also tested phosphorylation of GSK-3β, a downstream target of PKCζ. Fig. 3B shows that expression of C20ζ-EGFP reduced the level of both phosphorylated PKCζ and GSK-3β. However, the total protein levels of these two proteins were also reduced, suggesting that the expression of C20ζ-EGFP did not specifically affect phosphorylation of endogenous PKCζ or GSK-3β. The effect on the protein level of PKCζ and GSK-3β was specific, as shown by comparison with the unchanged level of β-actin used as a protein standard in the immunoblotting reaction.

Since aPKC is also essential for AJs or TJs between adjacent cells, we determined whether the expression of C20ζ-EGFP would interfere with the formation of junctional protein complexes. Immunoblot analysis showed that the level of the junctional proteins β-catenin and ZO-1 did not change in C20ζ-EGFP-expressing cells when compared with EGFP controls (Fig. 3B, right). In contrast, the level of E-cadherin was reduced to less than 50% (Fig. 3, B and C) of that found in EGFP controls. These data were intriguing, since PKCζ has been shown to be critical for both AJ and TJ formation (1, 8, 35). These data also provided strong evidence that C20ζ-EGFP acted as dominant negative mutant of PKCζ.

We determined whether inhibition of endogenous PKCζ would mimic the effect of C20ζ-EGFP. Fig. 3C shows that incubation of control cells with the myristoylated pseudosubstrate inhibitor of PKCζ (PZI) reduced the protein level of E-cadherin to that of C20ζ-EGFP-expressing cells, which was consistent with data from the literature showing that knock-out of aPKC affects the formation of AJs (8). Our result confirmed that activation of PKCζ was required to maintain the adherens junction in MDCK cells (10) and that dominant negative C20ζ-EGFP compromised the downstream effect of active PKCζ. To evaluate if competition with endogenous PKCζ for binding to ceramide mediated the effect of C20ζ-EGFP, we supplemented the medium with 0.8 μm C16:0 ceramide. In previous studies, it has been shown that ceramide is rapidly taken up from the medium and enriched in various cell types (26). Fig. 3, C and D, shows that incubation of C20ζ-EGFP-expressing cells with ceramide partially restored the level of E-cadherin to that of EGFP-expressing control cells. This result suggested that additional ceramide could overcome the competition between endogenous PKCζ and C20ζ-EGFP. Other sphingolipids, such as sphingosine-1-phosphate (S1P), could not overcome the dominant negative effect of C20ζ-GFP, indicating that the rescue effect was specific for ceramide (not shown).

Further, immunohistochemistry was performed to determine the effect of C20ζ-EGFP on the subcellular distribution of E-cadherin and ZO-1. Consistent with the results from the immunoblots (Fig. 3, B and C), we found that the immunofluorescence signal of E-cadherin was dramatically diminished. The signal intensity for E-cadherin was inversely correlated with the expression level (intensity of green fluorescence) of C20ζ-EGFP (Fig. 4A). To our surprise, the distribution of ZO-1 and β-catenin to adjacent cell membranes was disturbed, although the overall expression of the proteins remained unchanged, as shown by the immunoblot signals (Fig. 3B, right). The distribution of ZO-1 and E-cadherin was partially restored by supplementation of the medium with C16:0 ceramide (Fig. 4, A and B), consistent with its rescue effect on the protein level of E-cadherin (Fig. 3, C and D). Taken together, these results suggested that 1) C20ζ-EGFP is a dominant negative mutant of PKCζ that disrupted the formation of junctional protein complexes, and 2) competition with endogenous PKCζ for binding to ceramide mediated this effect. Our results implied that the interaction of ceramide with PKCζ is critical for the formation of junctional protein complexes and the establishment of cell-to-cell contacts.

FIGURE 4.

C20ζ-EGFP interrupted cell-cell junction formation. A–C, an equal number of EGPP- and C20ζ-GFP-expressing cells were seeded on coverslips and grown to confluence. Some of the wells were treated with 0.8 μm C16 ceramide for 24 h. Cells were fixed and stained for E-cadherin (A), ZO-1 (B), and β-catenin (C). Note that the membrane distribution of E-cadherin, ZO-1, andβ-catenin was disrupted by C20ζ-EGFP, whereas C16 ceramide rescued it. Scale bar, 10 μm.

Expression of C20ζ-EGFP Disrupts Epithelial Integrity—The effect of C20ζ-EGFP on the formation of junctional protein complexes suggested that cell-to-cell contacts and epithelial integrity would be disturbed by expression of this protein fragment. One of the characteristics of a functional epithelium is the development of TER, a quantitative indicator of epithelial integrity (36). We used an ECIS to study the effect of C20ζ-EGFP on TER. The ECIS measures the electrical impedance (resistance) between the top and bottom of a cell layer, which is directly correlated to the “tightness” of cell-to-cell contacts. Fig. 5 shows that in control MDCK cells expressing EGFP, TER increased and reached a plateau within 24 h, followed by a gradual decline, most likely because of the exhaustion of nutrients. C20ζ-EGFP-expressing cells showed maximum impedance at a similar time point (20–22 h of incubation) but at a 2.5-fold smaller value of resistance (Fig. 5). Most interestingly, C16 ceramide partially rescued the impedance in C20ζ-EGFP-expressing cells. This result was consistent with the aberrant subcellular distribution/reduced protein level of junctional proteins (Figs. 3, B and D, and 4) in C20ζ-EGFP-expressing cells and suggested that the interaction of ceramide with aPKC is critical for epithelial integrity.

FIGURE 5.

C20ζ-EGFP disrupted epithelial cell integrity. Equal numbers of EGPP- and C20ζ-EGFP-expressing cells (50,000) were seeded on poly-l-lysine-coated gold electrode arrays (8W10E) and grown to confluence in the growth medium. 0.8 μm C16 ceramide was added 1 h before measurement. Measurements of TER were performed using an ECIS instrument. TER was first monitored for 30 min to establish a steady state at 1200–1400 ohms, and then the electrical resistance across the monolayer was recorded over a period of 50 h. Values from each microelectrode were pooled at discrete time points, and the normalized impedance (normalized to initial impedance) was plotted versus time as the mean ± S.E. of the mean. Data shown were normalized to initial impedance. This experiment was repeated three times. ^*, p < 0.05; ^**, p < 0.01.

DISCUSSION

aPKC (ζ and λ/ι) has emerged as one of the key regulatory protein kinases in embryonic development and pathological conditions, such as cancer (37–40). aPKC is involved in multiple cellular processes, one of which is cell polarity (1, 20, 35), in which it is critical for establishment of apicobasal polarity (1, 10), cell migration (19, 20, 41, 42), motility (26), and AJ and TJ formation (8, 10, 43). aPKC forms a protein complex with PAR-3, PAR-6, and Cdc42, which regulates PAR-1, GSK-3β, and other downstream effectors that are involved in cell polarity (1, 20, 42, 44). Regarding its upstream factors, aPKC has been reported to be activated by Cdc42 (19) or Dishevelled during astrocyte migration or neuronal axon formation, respectively (44). We and others have found that the sphingolipid ceramide recruits and activates aPKC (24–28). However, how aPKC associates with ceramide is not known.

Ceramide is a cell signaling lipid that has gained attention during the last 2 decades (45, 46). In addition to the traditional view of being a mediator of cell-stress responses, including apoptosis (25) and cell senescence (46), we recently found that ceramide is an essential regulator of cell polarity (24, 26). Several candidate proteins have been shown to be activated by ceramide in vitro and in cells. These include protein phosphatase 1 (46, 47), protein phosphatase 2A (47), kinase suppressor of Ras (48), cathepsin D (49), and aPKC (25, 27). Activation of PKCζ by ceramide has been implicated in the regulation of apoptosis, membrane potential, and cell polarity (24, 26, 27, 49).

Using a lipid overlay assay, we show here that a proteolytically derived, COOH-terminal fragment of PKCζ binds to ceramide. The association of the COOH terminus with ceramide was confirmed by binding to ceramide-containing lipid vesicles using an ectopically expressed, COOH-terminal 20-kDa fragment of PKCζ (C20ζ). Consistently, subcellular fractionation studies and immunocytochemistry studies showed that C20ζ is distributed to cellular membranes and co-distributes with ceramide-rich areas of the cell membrane and intracellular vesicles.

We then determined if binding of the COOH terminus to ceramide is significant for the biological function of PKCζ. C20ζ, expressed as fusion protein with EGFP, disrupts epithelial integrity by down-regulating the protein level of E-cadherin and dislocating other junctional proteins from the cell membrane, suggesting that C20ζ-EGFP is a dominant negative mutant of PKCζ. We hypothesized that the dominant negative effect results from a competition between ectopically expressed C20ζ-EGFP and endogenous PKCζ for binding to ceramide. Hence, ectopic expression of the catalytically inactive C20ζ-EGFP would prevent activation of endogenous PKCζ by ceramide and impair biological processes regulated by this activation. Consistent with this hypothesis, direct inhibition of endogenous PKCζ with PZI reduces the level of E-cadherin to that of cells expressing C20ζ-EGFP. Further, incubation of C20ζ-EGFP-expressing cells with C16 ceramide normalizes the protein level of E-cadherin and restores, at least in part, epithelial integrity as quantified by immunoblotting and ECIS analysis, respectively. This finding is consistent with intracellular enrichment of C16 ceramide providing additional binding sites for endogenous PKCζ and, therefore, reactivation of PKCζ-dependent cell signaling pathways for polarity. Therefore, it is likely that binding of ceramide to the COOH terminus of aPKC is a common pathway to activate biological processes critical for cell polarity and motility.

The C20ζ protein fragment encompasses only half of the catalytic domain and the rest of the COOH-terminal amino acid sequence, and it does not contain the inhibitory pseudosubstrate and the C1b domain. The lack of the C1b domain is of particular significance, since it has been previously speculated that this domain may bind to ceramide (34). Although our data do not exclude the possibility of an additional ceramide binding site, the dominant negative effect of the C20ζ protein fragment counteracted by ceramide suggests that ceramide binding to the COOH terminus of PKCζ is critical for cell polarity and epithelial integrity. The amino acid sequence comprising the C20ζ fragment is the most conserved region when comparing PKCζ, PKCλ, and PKCι within one species (supplemental Fig. 1) and PKCζ between different species (supplemental Fig. 2). Therefore, it is likely that all of the aPKC isoforms are regulated by binding of their COOH termini to ceramide.

This is the first report showing that the COOH terminus of aPKC can bind to ceramide and that this binding is of biological significance for cell polarity and cell-to-cell junctions. In future studies, we will further define this ceramide binding site and determine its affinity for distinct ceramide species.