QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

Vol. 19, Issue 3, 912-928, March 2008

This Article Abstract Full Text (PDF) All Versions of this Article: E07-06-0596v1 19/3/912    most recent Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Google Scholar Articles by Langlois, S. Articles by Laird, D. W. PubMed PubMed Citation Articles by Langlois, S. Articles by Laird, D. W.

Caveolin-1 and -2 Interact with Connexin43 and Regulate Gap Junctional Intercellular Communication in Keratinocytes

Stéphanie Langlois^*, Kyle N. Cowan^*,, Qing Shao^*, Bryce J. Cowan, and Dale W. Laird^*,

Departments of *Anatomy and Cell Biology, Physiology and Pharmacology, and Surgery, University of Western Ontario, London, ON N6A 5C1, Canada; and Department of Dermatology and Skin Science, University of British Columbia, Vancouver, BC V5Z 4E8, Canada

Submitted June 22, 2007; Revised December 10, 2007; Accepted December 19, 2007
Monitoring Editor: Robert Parton

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Connexin43 (Cx43) has been reported to interact with caveolin (Cav)-1, but the role of this association and whether other members of the caveolin family bind Cx43 had yet to be established. In this study, we show that Cx43 coimmunoprecipitates and colocalizes with Cav-1 and Cav-2 in rat epidermal keratinocytes. The colocalization of Cx43 with Cav-1 was confirmed in keratinocytes from human epidermis in vivo. Our mutation and Far Western analyses revealed that the C-terminal tail of Cx43 is required for its association with Cavs and that the Cx43/Cav-1 interaction is direct. Our results indicate that newly synthesized Cx43 interacts with Cavs in the Golgi apparatus and that the Cx43/Cavs complex also exists at the plasma membrane in lipid rafts. Using overexpression and small interfering RNA approaches, we demonstrated that caveolins regulate gap junctional intercellular communication (GJIC) and that the presence of Cx43 in lipid raft domains may contribute to the mechanism modulating GJIC. Our results suggest that the Cx43/Cavs association occurs during exocytic transport, and they clearly indicate that caveolin regulates GJIC.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Gap junctions are specialized intercellular membrane channels that allow direct transfer of molecules of <1000 Dalton to selectively pass from one cell to another (Sohl and Willecke, 2004; Laird, 2005). Gap junctional intercellular communication (GJIC) plays a critical role in the coordination of development, tissue function, and cell homeostasis (Segretain and Falk, 2004; Wei et al., 2004). Gap junction channels are composed of two hemichannels, termed connexons, each provided by one of the two contacting cells and each connexon is composed of six transmembrane proteins, called connexins (Cxs) (Laird, 2006). The large gene family encoding Cx comprises 21 members in human (Sohl and Willecke, 2004; Wei et al., 2004). Cx43 is the most abundant gap junction protein in a wide spectrum of tissues, including the epidermis (Kretz et al., 2004; Laird, 2006). Cx43 is most likely assembled into connexons in the trans-Golgi network and then traffics along microtubules to the plasma membrane where they diffuse laterally and aggregate into gap junction plaques (Saez et al., 2003; Martin and Evans, 2004; Segretain and Falk, 2004; Laird, 2006; Lauf et al. 2002), although one report suggests that Cx43 containing vesicles may be directly targeted to adherens junctions adjacent to gap junctions (Shaw et al., 2007). The function of existing gap junction channels can be controlled by their opening and closing, whereas the regulation of their delivery, assembly, and removal constitutes additional mechanisms to control GJIC (Saez et al., 2003; Segretain and Falk, 2004; Laird, 2006). The extent of intercellular coupling is thus finely regulated by the control of connexin channels present at the plasma membrane, their functional state, and their permeability. All of these properties are determined by the constituent proteins of the connexon and their interaction with other protein partners and lipids (Cascio, 2005).

Various gap junction proteins are targeted to cholesterol-sphingolipid–rich membrane microdomains, called lipid rafts (Schubert et al., 2002; Lin et al., 2003, 2004; Barth et al., 2005; Laing et al., 2005; Locke et al., 2005), with several of these Cxs, including Cx43, having been shown to interact with caveolin (Cav)-1 (Lin et al., 2003, 2004; Schubert et al., 2002). Caveolins are cholesterol-binding integral membrane proteins that play a crucial role in the formation of caveolae, which are specialized invaginated raft domains (Parton et al., 2006). It has been recently identified that the Golgi complex is the site where newly assembled caveolar domains show up first, which are then directly transported and fuse to the plasma membrane (Tagawa et al., 2005). The formation and exit of caveolae from the Golgi complex is associated with caveolin oligomerization, acquisition of detergent insolubility, and association with cholesterol to form a mature caveolae-like exocytic structure (Parton et al., 2006; Parton and Simons, 2007). Although it is still unknown whether these proteins are transported to the plasma membrane in caveolar carriers or via other exocytic pathways in a caveolin-dependent manner, the efficient plasma membrane delivery of the angiotensin II type 1 receptor, the insulin receptor, and transient receptor potential channel protein (TRPC1) has been reported to depend on Cav-1 (Brazer et al., 2003; Cohen et al., 2003; Wyse et al., 2003; Parton and Simons, 2007). In addition, caveolins also function as scaffolding proteins capable of binding lipids and recruiting numerous signaling molecules into caveolae, as well as regulating their activity (Cohen et al., 2004). Various channels have been shown to localize to and be functionally regulated by lipid rafts/caveolae (Lockwich et al., 2000; Martens et al., 2001; Yarbrough et al., 2002; Barbuti et al., 2004; Brainard et al., 2005; Wang et al., 2005; Balijepalli et al., 2006). Cav-1 has been shown to modulate the activity of several channels (Trouet et al., 1999, 2001; Toselli et al., 2005), and more recently, it has been reported that Cav-1 regulates the function of the TRCP1 and the large conductance Ca(²⁺)-activated K⁺ channels by a direct physical interaction (Wang et al., 2005; Kwiatek et al., 2006; Remillard and Yuan, 2006).

Although Cx43 has been reported to target to lipid rafts and interact with Cav-1 (Schubert et al., 2002), and it has been suggested that rafts might be involved in Cx trafficking (Locke et al., 2005), the role of the Cx43/Cav-1 association, whether this interaction is direct or not, and whether other members of the caveolin family bind Cx43 had yet to be established. To investigate the role of caveolins in regulating Cx43 function, we chose to use keratinocytes, which endogenously express Cx43, Cav-1, and Cav-2. Cx43 and Cav-1 have both been associated with keratinocyte differentiation and transformation (Fitzgerald et al., 1994; Sando et al., 2003; Maass et al., 2004), and skin carcinogenesis (Kamibayashi et al., 1995; Capozza et al., 2003). In this study, we mainly used a rat epidermal keratinocyte (REK) cell line that is phenotypically similar to basal keratinocytes in that they retain the ability to differentiate into organotypic epidermis (Maher et al., 2005; Langlois et al., 2007). In addition, Cx43 is found at the plasma membrane within gap junction plaques and within intracellular compartments in keratinocytes, thus allowing for the investigation of a possible role of caveolins in constitutive Cx43 trafficking and/or internalization. Using REKs, we have found that, in addition to Cav-1, Cx43 coimmunoprecipitates and colocalizes with Cav-2. The Cx43/Cav-1 colocalization was also observed in vivo in keratinocytes from human epidermis. Our results suggest that newly synthesized Cx43 and Cavs associate in the Golgi apparatus and most likely traffic together to the plasma membrane in lipid rafts. Mutation or deletion of Cx43 C-terminal tail (CT) inhibits its association with Cav-1 and -2, indicating that this domain is required for the Cx43/Cavs interaction. Our Far Western analysis confirmed the importance of the CT domain of Cx43 for its association with Cavs and indicates that Cx43CT can bind directly to Cav-1, suggesting that the Cx43/Cav-2 interaction might occur via Cav-1. Disruption of lipid rafts by using methyl-β-cyclodextrin (MβC) and diminution of Cav-1 and -2 expression by using small interfering RNA (siRNA), both reduced GJIC. These data were further confirmed by overexpressing Cav-1 in 293T cells, which express Cx43 but are deficient in Cavs, resulting in a significant enhancement of GJIC. Treatment with MβC and isolation of lipid rafts suggest that the presence of Cx43 in these specialized microdomains contributes to the mechanism regulating GJIC.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Cell Culture
REKs and human embryonic kidney 293T were cultured in DMEM supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 100 µg/ml streptomycin, and 2 mM glutamine. 293T cells were transfected using Metafectene transfection reagent (Biontex, Martinsried, Germany), as suggested by the manufacturer.

Engineering of the Cx43 DNA Constructs and Retroviral Infections
Generation of constructs containing cDNA for human Cx43-green fluorescent protein (GFP), G21R-GFP, G138R-GFP, G60S-GFP, and fs260-GFP within the adaptor protein (AP)2 replication-defective retroviral vector (Galipeau et al., 1999) were described previously (Flenniken et al., 2005; Gong et al., 2006). In the cDNA encoding for 244-GFP Cx43 mutant, GFP was fused in frame to truncated Cx43 at amino acid 243 with the addition of seven amino acid linker sequence (GATCCACCGGTCGCCACC), and the DNA sequence was verified. Stable REK cell lines expressing Cx43-GFP, G21R-GFP, G138R-GFP, G60S-GFP, fs260-GFP, or 244-GFP cDNAs were generated by retroviral infection as described previously (Mao et al., 2000; Qin et al., 2002).

Engineering of the Cav-1 Small Interfering RNA Construct and Retroviral Infections
The coding region of rat Cav-1 gene was used for target gene sequence template. The siRNA sequence from Cav-1, ACCGCTTGCTGTCTACCATCTT (rat Cav-1 gene, GenBank accession no. BC078744, from 333 to 354), was selected using software from GenScript (Scotch Plains, NJ). GenScript synthesized the small DNA insert encoding a short hairpin targeting Cav-1 gene and cloned it into a pRNA-H1.1/Retro vector, which contains the human H1 promoter and the selection marker hygromycin. For simplicity, we refer to RNA interference (RNAi) reagents as siRNA. The Cav-1–targeted siRNA construct, together with an unrelated control siRNA vector, were used to make infectious viral supernatant as we described previously (Shao et al., 2005). REKs were infected, cultured in selection medium containing 50 µg/ml hygromycin, and antibiotic-resistant cells were passed at least three times before further experimentation (Shao et al., 2005).

Human Tissue Samples
Human facial skin samples were obtained after informed consent from patients attending the Vancouver General Hospital Skin Care Center (University of British Columbia, Vancouver, BC, Canada). During the reconstructive phase, after surgical removal of a cutaneous tumor, samples were taken from discarded normal tissue remote to the primary tumor site. Tissue collection was performed by Dr. Bryce J. Cowan in accordance with the ethical principles set forth in the Declaration of Helsinki and as instituted at the University of British Columbia. Samples were fixed in 10% neutral-buffered Formalin and subsequently embedded in paraffin.

Western Blotting
Wild-type, transfected, or infected cells were washed three times with cold phosphate-buffered saline (PBS) and scraped into 25 mM 2-(N-morpholino) ethanesulfonic acid, pH 6.5, and 0.15 M NaCl (MBS)-buffered saline containing 1 mM NaF, 1 mM Na₃VO₄, and protease inhibitor. The collected cells were sonicated and centrifuged to remove the debris from the lysates. Protein concentrations were determined using bicinchoninic acid assay (Pierce Biotechnology, Rockford IL). Proteins were resolved by SDS-polyacrylamide gel electrophoresis (PAGE), transferred to nitrocellulose membranes, and incubated in blocking buffer (LI-COR, Lincoln, NE) for 1 h, and then they were incubated with primary antibodies (β-actin, AC-15; Sigma-Aldrich. St. Louis, MO), early endosome marker EEA1 (ab2900; Abcam, Cambridge, MA), Cav-1 (clone polyclonal antibody [pAb]; BD Biosciences, San Jose, CA), Cav-2 (monoclonal antibody [mAb] clone 65, BD Transduction Laboratories, San Jose, CA), Cav-3 (clone 26; BD Biosciences), Cx43 (C6219; Sigma-Aldrich), or a mouse anti-Cx43NT (Goldberg et al., 2002) overnight at 4°C. After three washes with Tris-buffered saline with Tween 20, infrared fluorescent-labeled secondary antibodies (IRDye 800 anti-rabbit or anti-mouse, Rockland Immunochemicals, Gilbertsville, PA; or Alexa-680 anti-rabbit or anti-mouse, Invitrogen, Carlsbad, CA) were incubated at room temperature for 1 h, and immunoblots were processed and quantified using the Odyssey infrared-imaging system (LI-COR).

Coimmunoprecipitation
Identical amounts of proteins from each lysate were incubated in immunoprecipitation (IP) lysis buffer (150 mM NaCl, 10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 0.5% NP-40, and 1% Triton X-100) containing 1 mM NaF and 1 mM Na₃VO₄ overnight at 4°C in the presence of 2 µg/ml specific anti-Cav-1 (mAb clone 2297) or anti-Cav-2 (mAb clone 65) antibodies (BD Biosciences). The immune complexes were collected by incubating the mixtures with 30 µl (50% suspension) of protein A-Sepharose beads. Nonspecifically bound proteins were removed by washing the beads three times in 1 ml of IP lysis buffer, and bound material was solubilized in 30 µl of twofold concentrated Laemmli sample buffer, boiled for 5 min, resolved by SDS-PAGE, and transferred to nitrocellulose membranes. Blots were probed with specific antibodies to detect Cx43, Cav-1, and Cav-2.

Immunofluorescence
To detect the different pools of caveolins, cells grown on glass coverslips were fixed with either 3.7% formaldehyde or 80% methanol:20% acetone (Bush et al., 2006). 293T cells were fixed with 3.7% formaldehyde. Cells were permeabilized for 10 min in PBS with 0.2% bovine serum albumin (BSA) and 0.1% Triton X-100, and double-labeled for Cx43 and caveolins. Cells were labeled with either a polyclonal (1:200; Sigma-Aldrich) or a monoclonal anti-Cx43 antibody (clone P4G9 from the Fred Hutchinson Cancer Research Center Antibody Development Group, Seattle, WA). A fluorescein isothiocyanate (FITC)-conjugated anti-rabbit or anti-mouse IgG (1:200; Jackson ImmunoResearch Laboratories, West Grove, PA) was used as the secondary antibody for connexin labeling. For immunolocalization of caveolins, cells were labeled with a polyclonal anti-Cav-1 (1:50–1:100; pAb) or a monoclonal anti-Cav-2 antibody (1:50–1:100; clone 65). Texas Red-conjugated anti-rabbit or anti-mouse immunoglobulin G (IgG) (1:200; Jackson ImmunoResearch Laboratories) was used as the secondary antibody. For colabeling of Golgi markers, wild-type (WT) REKs and REKs overexpressing Cx43-GFP were labeled with either trans-Golgi network38 antibody (TGN38) (1:200; BD Biosciences Transduction Laboratories, Lexington, KY), or GPP130 (1:200; Covance, Berkeley, CA). FITC- and Texas Red-conjugated anti-rabbit or anti-mouse IgG were used as the secondary antibody. Cell cultures were then stained with Hoechst 33342 (1:1000; Invitrogen).

Human skin sections (5 µm in thickness) were deparaffinized in xylene, rehydrated in graded alcohols, and washed in PBS. Antigen retrieval was performed using Vector Antigen Unmasking Solution (Vector Laboratories) according to the manufacturer's protocol. To block nonspecific antibody binding, sections were incubated for 1 h at room temperature in PBS containing 3% BSA and 0.1% Triton X-100. Primary antibodies (Cx43 [P4G9], Cav-1 [1:100, pAb; BD Transduction Laboratories) were next applied to sections for 90 min at 37°C and diluted, where applicable, in PBS with 1% BSA, 0.1% Tween 20, and 0.01% SDS. Appropriate secondary antibodies diluted 1:100 in PBS with 1% BSA were then applied, conjugated to either Texas Red (Jackson ImmunoResearch Laboratories) or Alexa Fluor 488 (Invitrogen), and incubated for 1 h at room temperature. Hoechst 33342 (1:1000) was used as a nuclear stain. Sections were mounted with Vectashield mounting medium (Vector Laboratories, Burlingame, CA). In all cases, labeling was visualized with an LSM 510 META inverted confocal microscope (63x oil objective; Carl Zeiss, Jena, Germany).

Drug Treatments: Brefeldin A (BFA), MβC, and Cycloheximide (CHX)
Cells were treated for 6 h with 5 µg/ml BFA, for 1 h with 10 mM MβC, and for 6 h with 10 µg/ml CHX. BFA, MβC, and CHX were purchased from Sigma-Aldrich. After treatment, cells were fixed for immunofluorescence microscopy, lysed for immunoprecipitation and Western blotting, or used for Triton solubility and dye transfer assays.

Triton Solubility
Cells were washed three times with cold PBS. Triton solubility was assessed as described previously (Schubert et al., 2002). Briefly, 600 µl of cold MBS containing 1% Triton X-100 plus 1 mM NaF, 1 mM Na₃VO₄, and proteases inhibitors was added to the cells. After a 30-min incubation without agitation on ice, the soluble fraction was collected. Six hundred microliters of 1% SDS was added to the 60-mm-diameter plate to dissolve the remaining Triton X-100–insoluble material. Equal volumes (25 µl) of the Triton X-100 soluble and insoluble fractions were separated by SDS-PAGE and subjected to immunoblotting as described above. EEA1 was used to verify the quality of the isolated fractions (Estall et al., 2004).

Isolation of Lipid Rafts
Lipid rafts were isolated as described previously (Ostrom and Insel, 2006). Briefly, three 100-mm-diameter plates were washed three times with cold PBS, scraped into 750 µl of MBS containing 1% Triton X-100, and passed through a tight-fitting Dounce homogenizer (20 strokes). The sample was mixed with an equal volume of 90% sucrose (prepared in MBS lacking Triton X-100), transferred to a centrifuge tube, and overlaid with 4 ml of 35% sucrose, and overlaid with 4 ml of 5% sucrose (prepared in MBS lacking Triton X-100). The samples were centrifuged at 200,000 x g for 18 h in a Beckman SW41Ti rotor. Twelve fractions of 1 ml were collected starting at the top of the gradient, and aliquots of each fraction were subjected to SDS-PAGE and immunoblotting.

Cell Surface Biotinylation and Coimmunoprecipitations
Coimmunoprecipitations of biotinylated proteins were performed on the basis of a previously described protocol (Ray et al., 1999). During the biotinylation procedure, all reagents and cultures were kept on ice. Cultures were washed three times in PBS, and then they were incubated in the same solution with EZ-link NHS-LC-biotin (0.5 mg/ml; Pierce Biotechnology) at 4°C for 20 min. Cells were washed once with PBS containing 100 mM glycine, and then they were incubated in glycine buffer for 15 min. Cells were washed again with glycine buffer and lysed either in IP buffer or in SDS lysis buffer (1% Triton X-100 and 0.1% SDS in PBS). Lysates were rocked for 1 h, and supernatants were either incubated overnight with 40 µl of neutravidin-agarose beads (Pierce Biotechnology) or first subjected to IP by using Cav-1 or Cav-2 antibodies. After IP, the beads were washed twice with coIP buffer and PBS, and then they were dried by aspiration. Immunoprecipitated proteins were collected in 2% SDS at 55°C for 5 min, diluted into IP buffer 5:1, and incubated with 40 µl of neutravidin-agarose beads overnight. Neutravidin beads were washed in IP buffer and PBS as described above, resuspended in Laemmli buffer, boiled for 5 min, resolved by SDS-PAGE, and transferred to nitrocellulose membranes. Blots were probed with specific antibodies to detect Cx43, Cav-1, and Cav-2.

Far Western Overlay Assay
Glutathione transferase (GST)-Cx43CT transformed into DH5 Escherichia coli was kindly provided by Dr. Lampe (Fred Hutchinson Cancer Research Center), and the GST fusion protein was expressed and purified using the Bulk GST purification module following the manufacturer's protocol (GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom). At the end of the purification steps, GST-Cx43CT was incubated with thrombin to obtain purified Cx43CT (amino acids 236–382).

Cav-1 was immunoprecipitated from three different sets of lysates from control or Cav-1–transfected 293T cells, immunoprecipitates were resolved by SDS-PAGE, and then they were transferred to nitrocellulose membrane. The membrane was blocked in 2% nonfat dry milk in PBS and incubated with Cx43CT (2 µg/ml in 2% dry nonfat milk/PBS) overnight at 4°C. The membrane was washed three times with PBS containing 0.05% Tween 20, and then it was incubated with anti-Cx43 antibody (1:500; Sigma-Aldrich), which recognizes the C-terminal domain of Cx43, for 2 h at room temperature. The membrane was washed three times, incubated with anti-rabbit antibody (Alexa 680 [red], 1:5000) for 1 h, and the immunoblot was processed using the Odyssey infrared-imaging system (LI-COR). To confirm that the band detected using Cx43CT corresponded to Cav-1, the same blot was probed with anti-Cav-1 antibody, washed, and incubated with anti-mouse antibody (IRDye 800 [green], 1:5000) as described above.

Dye Transfer
Confluent control or MβC-treated REKs or REKs infected with control or Cav-1–targeted siRNA constructs were used for microinjection. 293T cells transiently transfected with a control empty vector or with cDNA encoding Cav-1 monometric red fluorescent protein (mRFP) were also used. Random cells selected within control or MβC-treated WT and Cav-1 knockdown REKs were pressure microinjected. In the case of 293T cells, one cell expressing mRFP-tagged Cav-1 within a cluster of cells was pressure microinjected. In all cases, cells were microinjected with 1% Lucifer yellow in double-distilled H₂O (Invitrogen) until the cell was brightly fluorescent (<5 s) by using an Eppendorf FemtoJet automated pressure microinjector. After 1 min, the percentage of microinjected cells that transferred Lucifer yellow to at least one contacting cell was determined using a DM IRE2 inverted epifluorescent microscope (Leica Microsystems, Wetzlar, Germany). Digital images were collected with a charge-coupled device camera (Hamamatsu Photonics, Tokyo, Japan) by using OpenLab software (distributed by Quorum Technologies, Guelph, ON, Canada). At least 38 microinjections were performed for each experimental condition.

Statistics
All statistical data were analyzed using a one-way analysis of variance followed by a Tukey's test for significance, except for the data on the effect of MβC treatment on dye transfer in REKs and of Cav-1 overexpression on dye transfer in 293T cells, and the percentage of Cx43 present in raft and nonraft fractions in these cells. Because these data contained only two groups, they have been analyzed using two-tailed Student's t test.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Cx43 Coimmunoprecipitates with Cav-1 and Cav-2 in REKs
In addition to Cav-1, the caveolin gene family in mammals includes Cav-2 and -3. Cav-1 is expressed ubiquitously, although at various levels in different tissues. Cav-2 is coexpressed with Cav-1 in many cell types and tissues, whereas Cav-3 is predominantly expressed in muscle cells (Williams and Lisanti, 2004). As shown in Figure 1A, WT REKs and REKs that express Cx43-GFP or the empty vector (V) express Cav-1 and -2, but not Cav-3. To examine the association of Cx43 with Cav-1 and its possible interaction with Cav-2, coimmunoprecipitation experiments were carried out. As reported previously (Schubert et al., 2002), the various phosphorylated species of endogenous Cx43 (P₀ and P), and Cx43-GFP, were found in Cav-1 immunoprecipitates (Figure 1B). When cell lysates were immunoprecipitated with antibodies directed against Cav-2, Cx43-GFP and Cx43 (P₀ and P) were also coimmunoprecipitated (Figure 1C). In keeping with the literature (Scherer et al., 1997; Scheiffele et al., 1998), Cav-1 was found associated with Cav-2 (Figure 1C). These results suggest that Cx43 interacts with Cav-1 and -2 as part of a multiprotein complex in REKs.

View larger version (27K): [in this window] [in a new window]   Figure 1. Cx43 coimmunoprecipitates with Cav-1 and Cav-2 in REKs. (A) WT REKs were infected with a vector encoding Cx43-GFP, or a control empty vector (V). Western blot showed that REKs express endogenous Cx43, Cav-1, and Cav-2, but not Cav-3 (positive control: muscle). (B) Both Cx43 (P0 and P forms) and Cx43-GFP were found in Cav-1 immunoprecipitates. (C) Cx43 (P0 and P forms) and Cx43-GFP, and Cav-1, coimmunoprecipitated with Cav-2. IP, immunoprecipitation; IB, immunoblotting.

Cx43 Mainly Colocalizes with Cav-1 and Cav-2 in Intracellular Compartments of REKs
The next step was to identify in which cellular location the Cx43/Cavs interaction occurs. It is known that Cav-1 is the major structural component of caveolae and that it is also found in the Golgi complex of many cell types, whereas Cav-2 is often colocalized with Cav-1 (Williams and Lisanti, 2004). Given that no single staining condition is capable of detecting all the Cav-1 in a cell simultaneously (Bush et al., 2006), two different fixation methods were used. As shown in Figure 2, the localization of Cx43 was not dependent on the fixation method as in both cases (formaldehyde [A] or methanol:acetone [B] fixation) Cx43-GFP was found at the plasma membrane (Figure 2, A and B, top, green, arrowheads), and in intracellular compartments in a Golgi-like pattern (Figure 2, A and B, top, green, arrows). In formaldehyde-fixed cells, Cav-1 (Figure 2A, red, top, arrow) and Cav-2 (Figure 2A, red, middle, arrows) were mainly detected in intracellular compartments in a Golgi-like pattern. Overlay images revealed that Cx43-GFP is colocalized with Cav-1 and -2 (Figure 2A, yellow, top and middle, arrows) in intracellular compartments, most likely in the Golgi complex. As reported previously in Caco-2 cells (Breuza et al., 2002), colocalization of Cav-1 (Figure 2A, green, bottom) with Cav-2 (Figure 2A, red, bottom) was observed in a Golgi-like pattern (arrow) in WT REKs. Fixation of cells with methanol:acetone predominantly allowed the detection of Cav-1 (Figure 2B, red, top) and -2 (Figure 2B, red, middle) at the cell surface (Pol et al., 2005; Bush et al., 2006). Overlay images suggest that the majority of Cx43-GFP was not colocalized with Cav-1 and -2 at the plasma membrane and vice versa (Figure 2B, top and middle), but that few Cx43-GFP puncta were colocalized with Cavs (yellow, arrowheads) at the plasma membrane. However, in keeping with the literature (Scherer et al., 1996), the majority of Cav-1 (Figure 2B, red, bottom) and Cav-2 (Figure 2B, green, bottom) was colocalized at the cell surface (yellow, arrowheads). These data suggest that the Cx43/Cavs interaction mainly occurs in intracellular compartments, most likely in the Golgi complex, in REKs.

View larger version (78K): [in this window] [in a new window]   Figure 2. Cx43 mainly colocalizes with Cav-1 and -2 in intracellular compartments of REKs. (A) REKs or REKs overexpressing Cx43-GFP were fixed with formaldehyde and labeled with Cav-1 and -2. Confocal imaging of REKs overexpressing Cx43-GFP (A, green, top and middle) revealed Cx43 in gap junctions plaques at the cell surface (arrowheads) and in intracellular localization in a Golgi-like pattern (arrows). Cav-1 (red, top, arrow) and -2 (red, middle, arrows) were mainly detected in intracellular compartments in a Golgi-like pattern. Overlay images suggest that Cx43-GFP is colocalized (yellow, top and middle, arrows) with Cav-1 and -2 in the Golgi. Colocalization (yellow) of Cav-1 (green, bottom) with Cav-2 (red, bottom) was observed in intracellular compartments of REKs (arrow). (B) REKs or REKs overexpressing Cx43-GFP were fixed with methanol/acetone and labeled with Cav-1 and -2. Cx43-GFP (B, green, top and middle) was found in intracellular compartments (arrow) and at the plasma membrane (arrowheads), whereas Cav-1 (B, red, top) and -2 (B, red, middle) were mainly detected at the plasma membrane. Overlay images revealed that the majority of Cx43-GFP was not colocalized with Cav-1 and -2 at the plasma membrane. However, few Cx43-GFP plaques were colocalized with Cavs (yellow, arrowheads) at the plasma membrane. Colocalization (yellow, arrowheads) of Cav-1 (red, bottom) with Cav-2 (green, bottom) was observed at the plasma membrane of REKs. Blue, nuclei. Bars, 10 µm.

View larger version (50K): [in this window] [in a new window]   Figure 3. Cx43 colocalizes with Cav-1 in keratinocytes from human skin epidermis. Confocal imaging of human epidermis indicates that Cx43 (green, top) was detected in the vital layers (between dashed line and dotted line) and absent in the cornified layer (between dashed lines). Cav-1 staining (red, middle) was pronounced in basal and spinosum layers (between dashed line and dotted line), and it was absent in the cornified layer (between dashed lines). Overlay images suggest that Cx43 is colocalized with Cav-1 in the vital layers of human epidermis (yellow, arrowheads, bottom). Blue, nuclei. Bar, 10 µm.

View larger version (104K): [in this window] [in a new window]   Figure 4. Cx43 colocalizes with Cavs in the Golgi apparatus. (A) REKs overexpressing Cx43-GFP were labeled with TGN38 (red, left), and REKs were colabeled for Cav-1 (red, middle) and TGN38 (green) or with Cav-2 (red, right) and GPP130 (green). Overlay images show that Cx43-GFP, Cav-1, and Cav-2 were colocalized (yellow) with the resident Golgi proteins. REKs or REKs overexpressing Cx43-GFP were treated with BFA for 6 h before their fixation with formaldehyde and labeling with Cav-1 and -2. (B) Treatment with BFA resulted in a loss of most Cx43-GFP (green, top and middle) at gap junction plaques (arrowheads), compared with untreated cells (green, inset, top). Golgi localization of Cx43-GFP (green, top and middle), Cav-1 and -2 (red, top and middle) was lost in BFA-treated cells. Colocalization of Cav-1 (green, bottom) with -2 (red, bottom) in the Golgi was lost in REKs. Blue, nuclei. Bar, 10 µm. (C) Lysates from untreated and BFA-treated REKs were subjected to Western blot. BFA treatment dramatically reduced the P species of Cx43. β-Actin was used as a loading control. Treatment with BFA reduced the amount of Cx43 found in Cav-1 (D) and Cav-2 (E) immunoprecipitates, without affecting the amount of Cav-1 coimmunoprecipitating with Cav-2 (E). IP, immunoprecipitation; IB, immunoblotting.

View larger version (98K): [in this window] [in a new window]   Figure 5. Newly synthesized Cx43 and Cavs colocalize in the Golgi apparatus. REKs overexpressing Cx43-GFP were treated with CHX for 6 h before their fixation with formaldehyde and labeling with Cav-1 (A) and -2 (B). Treatment with CHX resulted in a loss of Cx43-GFP in gap junction plaques (A and B, green, arrowheads). It also resulted in a loss of the intracellular pools of Cx43-GFP (A and B, green, arrows), Cav-1 (A, red, arrows), and Cav-2 (B, red, arrows) in the Golgi apparatus. Blue, nuclei. Bars, 10 µm.

View larger version (26K): [in this window] [in a new window]   Figure 6. A population of Cx43 interacts with Cavs at the plasma membrane in lipid rafts. WT REKs or REKs infected with Cx43-GFP, or the control empty vector (V), were treated with 10 mM MβC for 1 h. Untreated (A) and treated cells (B) were lysed in 1% Triton X-100 to obtain insoluble (I) and soluble (S) fractions, which were analyzed by Western blotting. EEA1 was used as a fractionation control. Note that Cx43, Cav-1, and Cav-2 were mainly found in the Triton-insoluble fractions (I), but that they were translocated to Triton-soluble fractions (S) after MβC treatment. (C) WT REKs were subjected to sucrose density gradient centrifugation. Twelve fractions were collected and analyzed by Western blot. Note that the fractions 4–6 (raft) are enriched in the P forms of Cx43, whereas the fractions 8–12 (nonraft) are enriched in the P0 form. (D) Fractions 4–6 (R) and fractions 8–12 (NR) were pooled and equal amounts of proteins were subjected to Western blotting. REKs were biotinylated (+), or subjected to mock biotinylation (–), and then subjected to coimmunoprecipitations by using Cav-1 and Cav-2 antibodies, as presented in the scheme (E). SDS extracts were collected, and an aliquot (2%) was analyzed by Western blot (F and G, lanes 1 and 2); the remainder of the lysates underwent neutravidin precipitation before SDS-PAGE (F and G, lanes 3 and 4). Cx43, Cav-1, and Cav-2 were recovered when REKs were treated with biotin (F and G, lane 4), but not when biotin was omitted (F and G, lane 3). The Cx43/Cav-1 and Cx43/Cav-2 complexes were recovered by anti-Cav-1 (F, lanes 5 and 6) and anti-Cav-2 IP (G, lanes 5 and 6), eluted from the antibodies, and aliquots (10%) were analyzed directly. The remainder of the immunoprecipitates underwent neutravidin precipitations followed by SDS-PAGE (F and G, lanes 7 and 8). Biotinylated Cx43 was coimmunoprecipitated with Cav-1 (F, top, lane 8) and Cav-2 (G, top, lane 8), and biotinylated Cav-1 was also present in Cav-2 immunoprecipitates (G, middle, lane 8). Lysates from untreated and MβC-treated WT REKs were subjected to Western blot (H) and IPs (I and J). Treatment with MβC reduced the amount of Cx43 found in Cav-1 (I) and Cav-2 (J) immunoprecipitates. IP, immunoprecipitation; IB, immunoblotting.

Because our colocalization data suggest that only a small population of Cx43 and Cavs are associated at the cell surface, we wanted to confirm the presence of the Cx43/Cavs complex at the plasma membrane; thus, we performed biotinylation/coimmunoprecipitation experiments (Figure 6E). Extracellular proteins were labeled with biotin on ice to limit intracellular transport to, and internalization from, the cell surface. After biotin treatment, the various phosphorylated species of Cx43 (P and P₀) were recovered with neutravidin, and Cav-1 and Cav-2 (Figure 6, F and G, lane 4), indicating their presence at the plasma membrane. No Cx43 or Cavs were recovered when biotin was omitted as a control (Figure 6, F and G, lane 3). In parallel, the Cx43/Cavs complexes were recovered by Cav-1 and Cav-2 immunoprecipitations to determine whether any of the Cx43 bound to Cavs had been biotinylated. When the Cx43 bound to Cav-1 (Figure 6F, lanes 5 and 6), and to Cav-2 (Figure 6G, lanes 5 and 6), was eluted and reprecipitated with neutravidin, Cx43 was recovered from biotin-treated REKs (Figure 6, F and G, top, lane 8), but not from the untreated cells (Figure 6, F and G, top, lane 7). In addition, when the proteins bound to Cav-2 have been eluted and reprecipitated with neutravidin, Cav-1 was also recovered (Fig, 6G, middle, lane 8) confirming the association of Cav-1 with Cav-2 at the plasma membrane. Although the amounts were low, biotinylated Cav-1 and Cav-2 were recovered in their respective immunoprecipitates (Figure 6, F and G, bottom, lane 8). To further assess whether the Cx43/Cavs association detected at the plasma membrane occurs in lipid rafts, REKs were treated with MβC, and coimmunoprecipitation experiments were performed. Although there were variations in the extent of dissociation, overall the treatment with MβC resulted in the reduction of the association of Cx43 with Cav-1 (Figure 6I) and Cav-2 (Figure 6J) without significantly altering the expression levels of these proteins (Figure 6H). However, treatment with MβC did not affect the extent of the association between Cav-1 and -2 (Figure 6J). Together, these results indicate a population of Cx43 is targeted to lipid rafts and interacts with Cavs in these specialized microdomains at the plasma membrane.

The C-terminal Tail of Cx43 Is Required for Its Association with Cavs
It has been reported previously that Cx43 interacts with the caveolin-scaffolding and the C-terminal domains of Cav-1 (Schubert et al., 2002). However, the domain(s) of Cx43 involved in this association and in its newly described interaction with Cav-2 remained to be determined. To identify which part of Cx43 molecule is involved in its interaction with Cavs, we stably overexpressed five GFP-tagged Cx43 mutants (G21R, G138R, G60S, fs260, and 244) in REKs, and WT Cx43-GFP (Figure 7A). These mutants were specifically chosen because they each represent mutations or truncations in distinct Cx43 domains (Figure 7A). Although some of these Cx43 mutant motifs were unlikely to be topologically available to interact directly with Cavs embedded in the plasma membrane, they were still assessed for their ability to associate with Cavs as the interaction could be indirect and occurring through protein complexes that could possibly involve the transmembrane and/or extracellular domains of the Cx43 molecule. The G21R, G138R, and fs260 mutants are linked to oculodentodigital dysplasia (ODDD), which is an autosomal dominant disorder caused by mutations in the gap junction 1 gene (GJA1) encoding Cx43 (Paznekas et al., 2003; van Steensel et al., 2005). Interestingly, the G21R substitution is close to a consensus sequence present in Cx43 (residues 25–32) for putative binding to a Cav-1 scaffolding domain (XXXXX, where is an aromatic amino acid [Phe, Tyr, or Trp]; (Lin et al., 2003). The G60S substitution occurs in a highly conserved Cx43 amino acid within the first extracellular loop, and it is the same mutation found in a mouse model of ODDD (Flenniken et al., 2005). The fs260 mutant is the result of a dinucleotide deletion (780–781del) in the GJA1 gene resulting in a frame-shift, yielding 46 aberrant amino acids after residue 259 and a shortened protein at residue 305 (van Steensel et al., 2005). In a previous report, it was shown that two truncated Cx43 mutants at residues 257 and 374 could still interact with Cav-1 (Schubert et al., 2002); thus, in our study, we included a substantially more truncated mutant (244) to assess whether the C-terminal tail was required for binding to Cavs.

View larger version (48K): [in this window] [in a new window]   Figure 7. The C-terminal tail of Cx43 is required for its association with Cavs. (A) A predicted topology of GFP-tagged Cx43 depicting four transmembrane domains (M1–M4), two extracellular loops (EL-1 and EL-2), a cytoplasmic loop (CL), and an N terminus and a C terminus (CT) are shown. Similar to WT Cx43, the C-terminal domain of all the Cx43 mutants used in this study were fused to GFP. The G21R, G138R, and G60S are missense mutations present in the first transmembrane domain, the cytoplasmic loop, and the first extracellular loop, respectively. The fs260 mutant represents a frameshift where residues 260–305 are aberrant amino acid residues and in which the C-terminal domain is truncated. The 244 mutant is truncated after residue 243 in its C terminus. (B) REKs overexpressing Cx43-GFP or GFP-tagged Cx43 mutants were subjected to Western blot. β-Actin was used as a loading control. (C) Triton solubility of GFP-tagged Cx43 constructs in REKs overexpressing GFP-tagged WT Cx43 and mutants. Insoluble (I) and soluble (S) fractions were analyzed by Western blotting. EEA1 was used as a fractionation control. Amount of Cx43-GFP and GFP-tagged Cx43 mutants in each fraction was quantified, and data are expressed as percentage of Cx43 in Triton-soluble and -insoluble fractions (n = 3, *p < 0.05). Cx43-GFP and the G21R, G138R, and G60S mutants coimmunoprecipitated with Cav-1 (D) and Cav-2 (E), but not the 244 and fs260 mutants. An antibody directed against the N-terminal domain (NT1) of Cx43 was used to detect all the Cx43 constructs. If present in the Cavs immunoprecipitates, the 244 mutant would have been detected as a band just below the IgG heavy chain band, whereas the fs260 mutant would have been detected as a slightly higher band than the IgG band. IP, immunoprecipitation; IB, immunoblotting.

The C-terminal Tail of Cx43 Can Bind Directly to Cav-1
Cx43 has been shown to coimmunoprecipitate with exogenous Cav-1 in transfected 293T cells (Schubert et al., 2002), which do not express any Cavs endogenously (Schlegel and Lisanti, 2000), suggesting that Cx43 can interact with Cav-1 independently of Cav-2. Because Cav-2 is known to form oligomers with Cav-1 (Scheiffele et al., 1998), its interaction with Cx43 might be indirect and occurring through Cav-1. To investigate whether Cx43 can directly bind to Cav-1, Cav-1 immunoprecipitates from independent sets of Cav-1–transfected and untransfected 293T cells were incubated with a soluble carboxy-terminal fragment of Cx43 designated (Cx43CT) in a Far Western blotting approach. In Cav-1 immunoprecipitates only, Cx43CT bound to a band doublet at 21–25 kDa (Figure 8, top, red). The same blot was then reprobed with anti-Cav-1 antibodies (middle, green), which strongly recognized a band at 21 kDa and weakly recognized a second slightly higher molecular weight protein. These bands most likely correspond to the and β isoforms of Cav-1. Taking advantage of the Odyssey infrared-imaging system, the overlay images revealed that the lower molecular weight band detected by Cx43CT corresponded to Cav-1 (bottom, yellow) suggesting that Cx43 directly binds to Cav-1.

View larger version (50K): [in this window] [in a new window]   Figure 8. Cx43 can bind directly to Cav-1. Cav-1 was immunoprecipitated from two sets of Cav-1-transfected and untransfected 293T cells, and the immunoprecipitates were resolved by SDS-PAGE and transferred to nitrocellulose. The blot was probed with Cx43CT (Far Western [FW]), which bound to an 21- to 25-kDa doublet only in immunoprecipitates from Cav-1–expressing cells (red, top). Bound Cx43CT was detected using antibodies against the C-terminal domain of Cx43. The blot was reprobed with anti-Cav-1 antibody (green, middle) and overlay images revealed that the Cx43CT bound to the same band detected with anti-Cav-1 antibodies (yellow, bottom).

View larger version (37K): [in this window] [in a new window]   Figure 9. RNAi knockdown of Cav-1 and -2 expression reduces the pool of Cx43 present in lipid rafts and GJIC in REKs. WT REKs were infected with a retrovirus encoding a control (Ctl) RNAi, or an RNAi targeting Cav-1. (A) Western blot show that Cav-1 and Cav-2 expression was reduced by 80 and 40%, respectively, whereas Cx43 level remained unchanged in cells treated with Cav-1 siRNA. β-actin was used as a loading control. (B) Triton solubility of Cx43, Cav-1, and Cav-2. EEA1 was used as a fractionation control. S, soluble fractions; I, insoluble fractions. (C) Labeling of endogenous Cx43 by using an anti-Cx43 antibody obtained from Sigma (green) revealed its location within intracellular compartments (arrows) and at the plasma membrane (arrowheads) in both Ctl and Cavs-knockdown REKs. (D) Labeling of endogenous Cx43 by using the P4G9 anti-Cx43 antibody (green) mainly detected Cx43 at the plasma membrane (arrowheads) in both Ctl and Cavs-knockdown REKs. (E) Ctl and (F) Cavs-knockdown REKs were subjected to sucrose density gradient centrifugation. Twelve fractions were collected and analyzed by Western blot. (G) Fractions 4–6 (R) and fractions 8–12 (NR) were pooled, and equal amounts of proteins were submitted to Western blotting. (H) Percentage of Cx43 present in raft (fractions 4–6) and in nonraft fractions (8–12) were quantified when equal amounts of proteins were analyzed by Western blotting (n = 4, *p < 0.05 compared with the equivalent fractions in control cells). (I) Untreated and MβC-treated WT REKs were microinjected with Lucifer yellow. Data were expressed as percentage of microinjected cells that passed dye (n represents the number of microinjected cells from three independent experiments, *p < 0.05). (J) WT, Ctl REKs, or REKs that were infected with Cav-1 siRNA were microinjected with Lucifer yellow. Data are expressed as percentage of microinjected cells that passed dye (n represents the number of microinjected cells from three independent experiments, *p < 0.001 compared with WT and Ctl REKs).

Overexpression of Cav-1 Induces the Translocation of Cx43 into Lipid Rafts and GJIC in 293T Cells
To further examine the relationship between Cavs and Cx43, we overexpressed Cav-1 in 293T cells, which express Cx43 (Figure 10A) but have no detectable levels of any known caveolin (Schlegel and Lisanti, 2000). Consistent with our results in REKs, overexpression of Cav-1 in 293T cells did not alter Cx43 expression and/or overall phosphorylation level (Figure 10A). Cav-2 was not detected by Western blotting in either WT or Cav-1–overexpressing 293T cells (data not shown). As reported previously (Schubert et al., 2002), Cx43 and Cav-1 were predominantly Triton-insoluble in these cells (Figure 10B). Confirming our results obtained from Cavs-knockdown REKs, the overexpression of Cav-1 did not alter the Triton solubility of Cx43 (Figure 10B). Accordingly, Cx43 reached the plasma membrane and could form gap junction plaques in control 293T cells (Figure 10C, green, top, arrowheads), suggesting that these processes do not strictly depend on Cav-1. In addition, no obvious difference in Cx43 localization was observed between control and Cav-1–overexpressing 293T cells by using the P4G9 antibody, which mainly recognizes Cx43 at the plasma membrane (Figure 10C). Although the majority of exogenous Cav-1 was not colocalized with Cx43, overlay images suggest that a small population of Cx43 colocalizes with Cav-1 at the cell surface (Figure 10C, bottom, arrowheads). To confirm that Cav-1 plays a role in targeting Cx43 into lipid raft domains, raft and nonraft membrane fractions were isolated from control (Figure 10D) and Cav-1–transfected 293T cells (Figure 10E), and the presence of Cx43 and Cav-1 were evaluated by Western blotting. The presence of Cx43 in R and in NR fractions was also assessed in pooled fractions (4–6 [R] and 8–12 [NR]) loaded at equal amount of proteins (Figure 10F). Cx43 was detected in lipid raft fractions in control 293T cells (Figure 10F), but the proportion present in these microdomains was slightly increased in cells overexpressing Cav-1 (Figure 10G, 67.0 ± 6.6% compared with 56.5 ± 7.0% in control cells; p < 0.05, n = 3). To further confirm the results obtained using REKs, we then tested whether Cav-1 overexpression could stimulate GJIC. We found that although 293T cells express low levels of endogenous Cx43, these cells were poorly coupled, because only 10% of microinjected cells passed Lucifer yellow dye (Figure 10, H and I). Interestingly, 70% of microinjected cells overexpressing Cav-1-mRFP passed dye (Figure 10, H and I), confirming that Cav-1 regulates GJIC.

View larger version (29K): [in this window] [in a new window]   Figure 10. Overexpression of Cav-1 in 293T cells induces the translocation of Cx43 into lipid rafts and increases GJIC. 293T cells were transfected with Cav-1 or the empty vector (V), subjected to Western blotting, and probed for Cx43 and Cav-1 (A). β-Actin was used as a loading control. (B) Triton solubility of Cx43 and Cav-1. EEA1 was used as a fractionation control. S, soluble fraction; I, insoluble fractions. (C) 293T cells were transfected with Cav-1 or control empty vector (V) and labeled for Cx43 (P4G9 anti-Cx43 antibody, green) and Cav-1 (red). Confocal imaging of control 293T cells (top) revealed Cx43 at the cell surface (arrowheads) in the absence of Cav-1. No obvious difference was observed in Cx43 localization at the plasma membrane (arrowheads) in Cav-1–expressing 293T cells, compared with control cells (V). Overlay images suggest that a population of Cx43 at the plasma membrane colocalizes with Cav-1 (yellow, bottom, arrowheads). (D) Control empty vector (V) and (E) Cav-1–transfected 293T cells were subjected to sucrose density gradient centrifugation. Twelve fractions were collected and analyzed by Western blot. (F) Fractions 4–6 (R) and fractions 8–12 (NR) were pooled, and equal amounts of proteins were submitted to Western blotting. (G) Percentage of Cx43 present in raft (fractions 4–6) and in nonraft fractions (8–12) were quantified when equal amounts of proteins were analyzed by Western blotting (n = 3, *p < 0.05 compared with the equivalent fractions in cells transfected with the empty vector). (H) 293T cells were transfected with Cav-1-mRFP or control empty vector (V) and microinjected with Lucifer yellow. (I) Data are expressed as percentage of microinjected transfected cells that passed dye (n represents the number of microinjected cells from three independent experiments, *p < 0.05). Blue, nuclei. Bars, 10 µm.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The Golgi complex has been identified as the site where newly assembled caveolae domains occur first before their rapid vesicular delivery to the plasma membrane where they insert as a stable caveolar domain (Tagawa et al., 2005). It is thought that Cav-1 and -2 hetero-oligomers represent the assembly units that drive caveolae formation. In the trans-Golgi network of epithelial cells, Cav-1/-2 hetero-oligomers have been reported to be sorted into basolateral vesicles and formed caveolae, whereas caveolae were not normally observed on the apical surface when only Cav-1 was found (Scheiffele et al., 1998). The formation of a Cav-1/-2 hetero-oligomeric complex has been shown to be required for Cav-2 transport from the Golgi to the plasma membrane in caveolae (Parolini et al., 1999). Our confocal data indicate that the majority of Cav-1 and -2 are colocalized in the Golgi apparatus and at the plasma membrane, suggesting that they mainly exist as hetero-oligomers in REKs. In addition, Cx43 and Cav-1 were both found in Cav-2 immunoprecipitates, suggesting that Cx43 interacts with these Cav-1/-2 hetero-oligomers. Our biotinylation experiments further indicate that a portion of the complex formed of Cx43 and Cav-1/-2 hetero-oligomers is present at the cell surface. These Cav complexes seemed to be stable because BFA and MβC treatment did not alter the association between Cav-1 and -2, but it led to a significant dissociation of Cx43 from Cavs. Taking into account previous studies on caveolae biogenesis (Scheiffele et al., 1998; Parolini et al., 1999; Tagawa et al., 2005; Parton et al., 2006), our results suggest that Cx43 is associated with Cavs in newly assembled caveolae, which exit the Golgi and most likely traffic together to be delivered in lipid rafts/caveolae at the plasma membrane. Cavs would thus be involved in the exocytic transport of Cx43, rather than in its internalization from the plasma membrane.

Caveolae assembled in the Golgi do not exchange caveolin molecules with each other or with any other pool once they reach the cell surface. In resting cells, caveolae at the cell surface are mainly immobile because they are tightly associated with the actin cytoskeleton and microfilaments (Tagawa et al., 2005). Our confocal data indicate that the majority of Cx43 and Cavs are not colocalized at the cell surface, suggesting that the Cx43/Cavs complex dissociate after its delivery in lipid rafts to the plasma membrane. Accordingly, it has been suggested that nonjunctional Cx26 and Cx32 are present in lipid rafts and that rafts might be involved in trafficking of plasma membrane connexin channels to gap junctions (Locke et al., 2005). Similarly, a model has been proposed in which connexons are delivered to the plasma membrane and reside transiently in lipid raft domains, requiring the tight junction protein zona occludens (ZO)-1 for recruitment into gap junctional plaques. Because ZO-1 is an actin binding protein, interactions with the actin cytoskeleton may also be involved in the recruitment of connexins to gap junctions (Laing et al., 2005). In keeping with our work, little Cx43 would thus reside in lipid rafts under steady-state conditions.

Our mutational analysis of Cx43 indicates that its C-terminal tail is involved in its interaction with Cavs. Indeed, the 244 mutant, which is almost completely truncated of its intracellular C-terminal domain, did not coimmunoprecipitate with Cavs. Similar to WT Cx43, this mutant was found in a Golgi-like intracellular compartment and within gap junction plaques (data not shown). These findings, combined with our confocal data of Cx43 localization in control 293T cells, indicate that Cavs are not essential for Cx43 transport to the cell surface. The trafficking of the GFP-tagged 244 mutant is also independent of the presence of wild-type Cx43, because it can reach the plasma membrane when expressed in Cx-negative HeLa cells (data not shown). Because the 244 mutant behaved similarly to WT Cx43 regarding its cellular localization and Triton solubility, its lack of interaction with Cavs was not due to a defect in its trafficking or lack of targeting to detergent-insoluble domains. Confirming the results obtained using the