Impaired Trafficking of Connexins in Androgen-independent Human Prostate Cancer Cell Lines and Its Mitigation by α-Catenin*

Gap junctions, composed of connexins, provide a pathway of direct intercellular communication for the diffusion of small molecules between cells. Evidence suggests that connexins act as tumor suppressors. We showed previously that expression of connexin-43 and connexin-32 in an indolent prostate cancer cell line, LNCaP, resulted in gap junction formation and growth inhibition. To elucidate the role of connexins in the progression of prostate cancer from a hormone-dependent to -independent state, we introduced connexin-43 and connexin-32 into an invasive, androgen-independent cell line, PC-3. Expression of these proteins in PC-3 cells resulted in intracellular accumulation. Western blot analysis revealed a lack of Triton-insoluble, plaque-assembled connexins. In contrast to LNCaP cells, connexins could not be cell surface-biotinylated and did not reside in the cell surface derived endocytic vesicles, in PC-3 cells, suggesting impaired trafficking to the cell surface. Intracellular accumulation of connexins was observed in several androgen-independent prostate cancer cell lines. Transient expression of α-catenin facilitated the trafficking of both connexins to the cell surface and induced gap junction assembly. Our results suggest that impaired trafficking, and not the inability to form gap junctions, is the major cause of communication deficiency in human prostate cancer cell lines.

Gap junctions, composed of connexins, provide a pathway of direct intercellular communication for the diffusion of small molecules between cells. Evidence suggests that connexins act as tumor suppressors. We showed previously that expression of connexin-43 and connexin-32 in an indolent prostate cancer cell line, LNCaP, resulted in gap junction formation and growth inhibition. To elucidate the role of connexins in the progression of prostate cancer from a hormone-dependent to -independent state, we introduced connexin-43 and connexin-32 into an invasive, androgen-independent cell line, PC-3. Expression of these proteins in PC-3 cells resulted in intracellular accumulation. Western blot analysis revealed a lack of Triton-insoluble, plaque-assembled connexins. In contrast to LNCaP cells, connexins could not be cell surface-biotinylated and did not reside in the cell surface derived endocytic vesicles, in PC-3 cells, suggesting impaired trafficking to the cell surface. Intracellular accumulation of connexins was observed in several androgen-independent prostate cancer cell lines. Transient expression of ␣-catenin facilitated the trafficking of both connexins to the cell surface and induced gap junction assembly. Our results suggest that impaired trafficking, and not the inability to form gap junctions, is the major cause of communication deficiency in human prostate cancer cell lines.
Cell-cell and cell-matrix adhesion is involved not only in maintaining the structural integrity of cells in tissues but also in governing a wide array of cell behavior (1)(2)(3). Cell-cell and cell-matrix adhesion molecules frequently cluster at specific contact areas to form cell structures, such as adherens junctions, tight junctions, desmosomes, and focal adhesion plaques (1)(2)(3)(4)(5). Loss of these junctions has profound consequences on cellular growth, differentiation, and apoptosis during neoplastic development in several tumor model systems (1). Recent studies have shown that expression of cell-cell and cell-matrix adhesion molecules is decreased in prostate cancer (PCA) 1 cell lines and that impairment or loss in the expression of these molecules is associated with the malignant potential of prostate epithelial cells (6 -12). Direct support for the role of cell adhesion molecules in controlling the invasive behavior of PCA cells has come from studies showing that forced expression of ␣-catenin, an E-cadherin-associated protein, and C-CAM (7)(8)(9)(10) in human PCA cell lines mitigates their malignant phenotype. These studies suggest that direct cell contact-dependent interactions among epithelial cells in prostate tumors are likely to play an important role in PCA progression.
In addition to cell-cell and cell-matrix adhesion junctions, epithelial cells also form a highly specialized class of cell junctions called gap junctions, which are membrane appositions that are traversed by clusters of channels through which molecules up to 1 kDa can directly pass between adjoining cells (13). The channels are bicellular structures formed by the members of a family of about 20 related but distinct proteins named connexins (Cxs). Connexins first assemble into hexamers to form connexons that align and join with connexons in adjacent cells to form cell-cell channels, which get clustered to form gap junctions (14 -16). In addition to its well documented role in the maintenance of tissue homeostasis and synchronization of cellular behavior, it has been proposed that altered gap junctional communication and/or impaired expression of Cxs may be one of the genetic or epigenetic changes involved in the initiation and progression of neoplasia (13,(17)(18)(19). This notion has been well supported by several independent studies (20 -24) showing that forced expression of Cx genes in several Cx-deficient tumor cell lines attenuates their malignant phenotype. A recent study (25) showed that transgenic mice deficient in Cx32, a Cx abundantly expressed in liver, developed a higher incidence of age-related liver tumors and were more susceptible to the tumor-promoting effect of liver-specific chemical carcinogens.
Although a number of tumor suppressor genes and oncogenes have been implicated in the development of PCA, no consistent genetic or epigenetic changes are known to be associated with its initiation and progression. What is clear, however, is that the incidence of PCA increases with age and is characterized by the progression from an indolent, slow-growing, and hormone (androgen)-dependent state to an invasive, hormone-independent state (10,11,26). Thus, identification of cellular and molecular events that play formative roles in driving the expansion and clonal selection of incipient PCA cells from an androgen-dependent state to an androgen-independent state is essential for understanding PCA progression and designing strategies for its intervention (11,26). Our previous studies showed that, compared with normal prostate epithelial cells, gap junctional communication in PCA cell lines was either absent or reduced (27) and that forced expression of Cx32 and Cx43, the two Cxs expressed by the well differentiated epithelial cells of the prostate, into an indolent, androgen-dependent and Cx-deficient human PCA cell line, LNCaP, inhibited growth, retarded tumorigenicity, and induced differentiation (20). These studies also showed that Cxs were localized at cell-cell contact areas in epithelial cells of well differentiated prostate tumors, and they began to accumulate intracellularly as the tumors progressed to more invasive and undifferentiated stages with an eventual loss of expression in advanced stages (20).
Prostate epithelial cells from the most invasive forms of human androgen-independent prostate carcinomas show frequent impairment and/or deletion of cadherins and their associated proteins, such as ␣-, ␤-, and ␥-catenins (10,11,26). Because bi-directional signaling between cell adhesion molecules and Cxs may be important in initiating the formation of gap junctions, we investigated if forced expression of Cx43 and Cx32 into an invasive, androgen-independent PCA cell line PC-3, with deficient cadherin-mediated adhesion due to the deletion of the ␣-catenin gene (7), would abrogate its malignant phenotype in a manner similar to that of LNCaP cells, which have functional cadherin-mediated adhesion. Our findings showed that, in contrast to androgen-dependent PCA cell lines, expression of Cx43 and Cx32 in PC-3, and several other androgen-independent cell lines, resulted in the intracellular accumulation of Cxs due to defective trafficking and that transient expression of ␣-catenin, a cadherin-associated protein, triggered trafficking and assembly of Cxs into gap junctions.

EXPERIMENTAL PROCEDURES
Materials-Cell culture media were obtained from Invitrogen. Defined fetal bovine and dialyzed fetal calf sera were from HyClone Laboratories (Logan, UT). Tissue culture plasticware was from Nalge Nunc International (Rochester, NY). Seakem GTG-agarose was from FMC BioProducts (Rockland, ME). TRIzol reagent, geneticin (G418), RNA molecular weight markers, and FuGENE 6 transfection reagent were from Invitrogen. Yeast tRNA, poly(A), poly(C), and herring sperm DNA were from Roche Molecular Biochemicals. Fluorochrome-conjugated secondary antibodies were from Jackson ImmunoResearch (West Grove, PA). Ultrapure formamide was from Clontech. Lucifer Yellow (LY, lithium salt), rhodamine-and Alexa 594-conjugated dextrans (M r 10,000, lysine-fixable), and Alexa 488-and Alexa 594-conjugated mouse and rabbit secondary antibodies were from Molecular Probes (Eugene, OR). The super-signal chemiluminescent substrate was from Pierce (Rockford, IL). Enhanced chemiluminescent kit (ECL plus) was from Amersham Biosciences. GeneScreen Plus nylon membranes and [ 32 P]dCTP were from PerkinElmer Life Sciences. Trans 35 S-label was from ICN Biomedical (Irvine, CA). Restriction enzymes and pre-stained protein molecular weight markers were from New England Biolabs (Beverly, MA). BCA reagent for protein determination was from Pierce.
Cell Culture-Human PCA cell lines PC-3 (ATCC CRL 1740) and LNCaP (ATCC CRL 1740) were grown in RPMI containing 7.5% defined fetal bovine serum in an atmosphere of 5% CO 2 , 95% air. Stock cultures were maintained in 12 ml of RPMI in 75-cm 2 flasks and sub-cultured weekly at 1.5 ϫ 10 5 cells/flask with a medium change at 3-or 4-day intervals as described previously (27). LNCaP clones expressing Cx43 and Cx32 stably were isolated and grown in culture medium supplemented with 200 g/ml G418 (active) as described (20). The growth characteristics and the hormonal dependence of these cell lines/clones have been described previously (20,27). The retroviral packaging cell lines PA317 (ATCC CRL 9078) and PG13 (ATCC CRL 10686) were grown in RPMI containing 10% defined fetal bovine serum as described previously (20,28).
Cells were immunostained after fixing with methanol/acetone, paraformaldehyde, and histochoice (depending on the antibody) as described previously (20,27,29). Briefly, 5 ϫ 10 4 cells were seeded in 6-well clusters containing glass coverslips and allowed to grow to confluence. They were washed 3 times with PBS, fixed for 10 min, and immunostained at room temperature with various antibodies. Secondary antibodies (rabbit or mouse) conjugated with fluorescein, CY2, CY3, Texas Red, Alexa 488, and Alexa 594 were used as appropriate. Images of immunostained cells were acquired with Leica DMRIE microscope (Leica Microsystems, Wetzler, Germany) equipped with Hamamatsu ORCA-ER CCD camera (Hamamatsu City, Japan). For colocalization studies, serial z sections (0.5 m) were collected and analyzed using image processing software (Openlab 3.01; Improvision, Inc., Lexington, MA).
Retroviral Vectors and Plasmids-Plasmids containing cDNAs for various Cxs were obtained from several sources as described previously (20,(27)(28)(29). Retroviral vector LXSN (33) was a generous gift of Dr. Dusty A. Miller (Fred Hutchinson Cancer Center, Seattle, WA). Retroviral vectors containing rat Cx43 and Cx32 in sense orientation were constructed as described (20) and designated LXSNCx43S and LXSNCx32S, respectively. The retroviral 5Ј long terminal repeat and SV40 virus promoter drive the expression of Cx cDNAs and neomycin phosphotransferase, respectively (33). Plasmids pECFP-N1, pEGFP-N1, pEGFP-N3, and PEYFP-N1 were purchased from Clontech (Palo Alto, CA). Chimeras of these fluorescent proteins fused to the carboxyl termini of Cx43 and Cx32 were constructed according to the standard molecular biology methods. The details of these constructs and their function and assembly into functional gap junctions will be described elsewhere. 2 Plasmid pcDNA3-␣-catenin (chicken) was constructed by cloning a chicken 3.5-kb HindIII to XbaI fragment, encompassing the coding range of chicken ␣-catenin cDNA from pUC21 vector, into the HindIII and XbaI site of pcDNA3.
Retrovirus Production and Infection of Cells-Control and recombinant retroviruses harboring Cx cDNAs were produced in an amphotropic packaging cell line PA317 and, to improve the efficiency of retroviral mediated gene transfer into human cells, in gibbon ape leukemia virus envelope-based packaging cell line PG13 as described previously (20). The titer of recombinant retroviruses produced from the most stable and the best producing PA317 and PG13 clones were 2 ϫ 10 6 and 4 ϫ 10 5 G418-resistant colony-forming units/ml, respectively, as measured in rat Morris hepatoma cells (20,28). PC-3 cells were infected with equivalent titer (4 ϫ 10 5 colony-forming units/ml) of recombinant retroviruses LXSN (Neo control), LXSNCx32S and LXSNCx43S, and selected in G418 (400 g/ml, active) for 2-3 weeks as described previously (20,28). Glass cylinders were used to isolate individual G418-resistant clones, which were expanded, frozen, and maintained in G418 (200 g/ml).
Isolation of RNA and Northern Blot Analysis-Total RNA was extracted from two 10-cm dishes of confluent cells using Trizol reagent as described previously (29). Ten to 20 g of RNA was analyzed on 1% agarose/formaldehyde gels, transferred to nylon filters, pre-hybridized, and hybridized with 32 P-labeled DNA probes for various Cxs or glyceraldehyde-3-phosphate dehydrogenase, washed in 0.1ϫ SSC at 65°C for 1-2 h, and the membranes exposed to Fuji-RX x-ray film for 1-24 h. The labeled DNA probes were prepared using gel-purified fragments (100 ng) and a random priming kit (Roche Molecular Biochemicals). The probes were labeled to a specific activity of 10 8 -10 9 cpm/g DNA and used at 10 6 cpm/ml hybridization buffer.
Western Blot Analysis-Triton X-100 solubility/insolubility of Cxs was assayed essentially as described by Musil and Goodenough (34,35). Preparation of cell lysates and Western blot analysis of Cxs was as described previously (20,27) except for the following modifications. After centrifugation at 50,000 ϫ g for 50 min on a tabletop Beckman ultracentrifuge (model TL-100) to separate Triton X-100-insoluble and -soluble fractions, the Triton X-100-insoluble pellet was dissolved in 500 l of solubilization buffer (70 mM Tris/HCl, pH 6.8, 8 M urea, 2.5% SDS, and 0.1 M dithiothreitol). Total Triton X-100-soluble and -insoluble fractions were mixed with 4ϫ Laemmli buffer to a final concentration of 1ϫ and boiled at 100°C for 5 min (Cx43) or incubated at room temperature for 60 min (Cx32) before separating by SDS-PAGE.
Detergent (Triton X-100) Extraction of Connexin-43 and Connexin-32 in Situ-Cells were seeded in 6-well clusters containing glass coverslips as described above (see "Antibodies and Immunostaining"), allowed to grow to confluence, and were washed once with PBS at room temperature. Half of the coverslips were extracted with 2 ml of 1% Triton X-100 (weight/weight) in solution B (30 mM HEPES; 140 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 3 mM glucose) containing a mixture of protease inhibitors for 30 min at 4°C with occasional gentle shaking. Control cells were treated identically except for the omission of 1% Triton X-100. Cells were immunostained with the antibodies against Cx43 and Cx32 as described above (see "Antibodies and Immunostaining").
Metabolic Labeling and Cell Surface Biotinylation-PC3 and LNCaP cells (2.5 ϫ 10 5 ) were seeded on 6-cm dishes and grown to 80 -90% confluence. Cells were incubated for 30 min at 37°C in methionine-and cysteine-free DMEM (Invitrogen) containing 2 mM L-glutamine and 5% dialyzed fetal calf serum (pulse medium), and labeled with 0.15 mCi/ml Tran 35 S-label (ICN Biomedicals, Irvine, CA) for 30 min at 37°C (2.5 ml per dish). Cells were chased in normal cell culture medium supplemented with 0.5 mM methionine and 0.5 mM cysteine (chase medium) at 37°C. Arrival of newly synthesized Cxs at the cell surface at various time intervals was assayed by incubating cells in freshly prepared EZ-Link TM Sulfo-NHS-SS-Biotin reagent (Pierce) in PBS at 0.5 mg/ml for 30 min at 4°C and quenched with 15 mM glycine. Lysis of monolayers, immunoprecipitation of Cxs, and the recovery of biotinylated Cxs were done essentially as described by VanSlyke and Musil (36) with the following modification. After cell lysis, 1/5th of the total immunoprecipitate was used for detecting the total and 4/5th for detecting the biotinylated fraction of Cxs at various time intervals. The samples were boiled for 5 min in 1ϫ SDS-PAGE sample buffer and resolved by SDS-PAGE. Quantitation was done by PhosphorImaging (STORM 840, Amersham Biosciences) using ImageQuant software.
Dextran Uptake and Endosome Labeling-PC-3 and LNCaP cells were seeded on glass coverslips as described above (see "Antibodies and Immunostaining") and grown to 60 -70% confluence. Endocytosis of dextrans was achieved by incubating cells with 10 mg/ml of Alexa Fluor 594-Dextran (M r 10,000, lysine-fixable, Molecular Probes) in DMEM at 37°C for 30 min. Cells were then rinsed briefly with PBS and incubated with DMEM without dextrans for 30 min at 37°C, rinsed again three times with PBS before fixing (with 2% paraformaldehyde), and immunostained for connexins as described (see "Antibodies and Immunostaining").
Transient Transfection of PC-3 Cells-FuGENE 6 transfection reagent was used to transfect cells with various plasmids mentioned above according to the manufacturer's instructions. Briefly, cells (5 ϫ 10 4 ) were seeded on glass coverslips in 6-well culture plates (for immunostaining) or 100-mm dishes (for Western blotting) containing 2 and 12 ml of complete medium, respectively, and incubated at 37°C. After 16 h, cells were transfected with various plasmids using 3 l of Fu-GENE 6 reagent: 1 g of DNA complex in 100 l of serum-free medium. The transfection complex was preincubated for 15 min at room temperature before transfection. The medium was replaced by fresh medium 5 h after transfection, and cells were fixed for immunostaining or lysed for Western blotting (see above) after 24 -48 h.
Communication Assays-Gap junctional communication was assayed either by microinjecting fluorescent tracer Lucifer Yellow (443 Da, 5% aqueous solution) as described previously (28,38) or by scrapeloading method (39). Briefly, LY was microinjected into test cells by Eppendorf InjectMan and FemtoJet microinjection systems (models 5271 and 5242, Brinkmann Instruments) mounted on a Leica DMIRE2 microscope. The microinjected cells were viewed with the aid of Sony 3 CCD color video camera (Sony Corp., Japan) and the number of fluorescent cells (excluding the injected one) scored ϳ10 min after injection served as an index of junctional transfer. For scrape loading, cell culture medium from freshly confluent 6-cm dishes was removed and replaced with 2.5 ml of medium containing rhodamine-conjugated fluorescent dextrans (10 kDa, 1 mg/ml; fixable) and LY (0.05%).

Overexpression of Connexin-43 and Connexin-32 in PC-3
Cells-We chose PC-3 cells for these studies because they are well characterized, highly invasive, and androgen-independent cells (10,11,26). Moreover, our previous study (27) showed that they communicated poorly, formed few gap junctions, expressed a low level of Cx43 mRNA and protein, and expressed no other Cxs. Connexin-43 and Cx32 were expressed in PC-3 cells by infecting with a control (LXSN) and Cx-harboring (LXSN43S and LXSN32S) recombinant retroviruses. Expression of Cxs was confirmed in several randomly isolated G-418resistant clones by Northern and Western blot analysis, and the data from one representative clone are shown in Fig. 1. Northern blot analysis of total RNA isolated from Cx43-and Cx32-expressing clones showed abundant expression of retrovirally transcribed 4.4-kb Cx43 (Fig. 1A, labeled R-Cx43; lane PC-43-1) and Cx32 (Fig. 1B, labeled R-Cx32; lane PC-32-1) mRNAs that were not expressed by the parental (lane PC-WT) and control clones (lanes PC-NEO-4 and PC-NEO-1). In addition, parental PC-3 cells and all retrovirally transduced clones expressed a low level of endogenous 3-kb human Cx43 mRNA (Fig. 1A, labeled E-Cx43), which was detected only upon overexposure of blots (data not shown). Western blot analysis (Fig.  1, C and D) showed that parental PC-3 cells (lane labeled PC-WT) and PC-3 clone isolated after infecting with LXSN (lane labeled PC-NEO) expressed neither Cx32 nor Cx43, whereas clones isolated after infecting with LXSN43S and LXSN32S expressed abundant Cx43 (Fig. 1C, lane labeled PC-43-1) and Cx32 (Fig. 1D, lane labeled PC-32-1). Taken together, these data show that retroviral transduction of Cx32 and Cx43 in PC-3 cells results in the abundant expression of both Cxs at the mRNA and protein level.
Intracellular Accumulation of Connexin-43 and Connexin-32 in PC-3 Cells-To determine whether Cx43-and Cx32-expressing PC-3 clones formed gap junctions, we immunostained cells from several clones with Cx-specific antibodies. Fig. 2 shows the typical immunostaining pattern for Cx43 and Cx32 in one such clone. The results showed that a major portion of both Cxs remained localized in the intracellular compartments (Fig. 2,  1st and 3rd rows), and punctate dots characteristic of gap junctional plaques were rarely observed at cell-cell contact areas (Fig. 2, white arrows in the middle and right panel of top row). Moreover, in many cells intense intracellular immunostaining was observed not only in the perinuclear areas but also throughout the cytoplasm (Fig. 2, yellow arrows, middle panels, 1st and 3rd rows) and near the cell surface membrane (not shown). In contrast to PC-3 cells, both Cxs were assembled into gap junctions in indolent, Cx-expressing LNCaP clones, and very little intracellular accumulation was observed (Fig. 2, 2nd and 4th rows, junctions indicated by arrows). Because similar results were obtained with all other independently isolated PC-3 clones, we chose only 1 clone for each Cx subtype for further study.
Detergent Insolubility of Intracellular Connexin-32 and Connexin-43 in PC-3 Cells-A widely accepted biochemical attribute of Cxs upon assembly into gap junctional plaques is their insolubility in Triton X-100 (34 -36). To corroborate the immunocytochemical data, and to rule out the possibility that intracellular accumulation was due to the formation of gap junctions in the intracellular stores, we extracted Cx32-and Cx43-expressing PC-3 cells in situ with 1% Triton X-100 for 30 min (see "Experimental Procedures") before immunostaining with Cxspecific antibodies. We also analyzed the Triton X-100 solubility of Cxs by Western blot analysis. Only a small fraction of total Cx43 (Fig. 3A) and Cx32 (Fig. 3B) was converted into a Triton X-100-insoluble form in invasive PC-3 cells, whereas a major fraction of both Cxs was Triton X-100-insoluble in LN-CaP cells and in RL-CL9 cells (29) which form abundant gap junctions composed of Cx43 (compare lanes labeled T, S, and I under PC-3, LNCaP, and RL-CL9). In PC-3 cells, nearly all intracellular Cx32-and Cx43-specific immunostaining was lost upon in situ extraction (Fig. 3C, compare Control and Extracted in 1st and 3rd rows), whereas in LNCaP cells there was no effect (compare Control and Extracted in 2nd and 4th rows).
Intracellular accumulation of Cx43 and Cx32 in PC-3 cells was not an artifact of overexpression, because androgen-dependent Cx-expressing LNCaP clones seemed to express more Cxs compared with PC-3 clones (see Fig. 3) even though an equal amount of total protein was analyzed by Western blot analysis (see also "Discussion"). The results shown in Fig. 3 suggest that in contrast to LNCaP cells, both Cx32 and Cx43 accumulate intracellularly in PC-3 cells and that intracellularly accumulated Cxs were not assembled into gap junctions ectopically based on the assumption that Triton X-100 solubility of gap junctions formed intracellularly is not significantly different from those formed at the cell surface. Moreover, these data also agree with our previous in vivo studies, which showed intracellular accumulation of Cx43 and Cx32 and/or loss of formation of gap junctions in epithelial cells of aggressive prostate carcinomas (20).
Communication in Connexin-expressing PC-3 Clones-To examine whether formation of only a few immunocytochemically detectable gap junctions was sufficient to promote gap junctional communication in Cx-expressing PC-3 clones, we studied the junctional transfer of 443-Da fluorescent tracer, LY, by microinjection and scrape loading. Fig. 4 shows representative photographs of junctional transfer of LY in control and Cxexpressing PC-3 cells. There was no significant difference in the junctional transfer of LY in the Cx-expressing PC-3 cells compared with the control cells (see Fig. 4 legend for details). The data obtained with the microinjection were independently corroborated by the scrape-loading method (39), which measures the communication capacity of several hundred cells simultaneously (Fig. 4 legend). Similar data were obtained with three other control and Cx43-and Cx32-expressing PC-3 clones (data not shown). Taken together, the data suggest that, in contrast to Cx-expressing LNCaP clones and normal RL-CL9 cells (20), reintroduction of Cx43 and Cx32 into PC-3 cells does not significantly enhance communication.
Cell Surface Biotinylation, Dextran Uptake, and Trafficking of Connexins-Besides intense perinuclear staining, we also observed punctate immunostaining scattered throughout the cytoplasm and near the cell surface (see Fig. 2, cells with yellow arrows). These observations prompted us to investigate whether intracellular accumulation of Cxs was due to their inability to traffic from intracellular stores to the cell surface or due to internalization and recycling back into the cytoplasmic stores after arrival at the cell surface because of lack of cell-cell contacts conducive for the formation of gap junctions. To test this notion, Cx-expressing PC-3 and LNCaP cells were cell surface-biotinylated and immunoprecipitated after metabolic labeling with Tran 35 S-label for detecting the total and biotinylated fraction of connexins at various time intervals (see "Experimental Procedures"). Fig. 5 shows that Cx43 and Cx32 were readily biotinylated in LNCaP cells in three independent experiments, whereas no biotinylated fraction of connexins could be detected in PC-3 cells (see figure legends for the efficiency of biotinylation of connexins). The failure to detect a significant amount of the biotinylated form of Cx32 and Cx43 in PC-3 cells was not due to inefficient cell surface biotinylation of proteins as judged visually by immunofluorescence microscopy using Alexa fluor-conjugated streptavidin (data not shown). These data suggest that trafficking of Cxs from the cytoplasm to the cell surface is impaired in PC-3 cells.
Because of the inefficient biotinylation of membrane proteins at 4°C, particularly of Cxs after their assembly into gap junctions (36), and to dispel the possibility that intracellular accumulation was caused by endocytosis of Cxs after their cell surface arrival and not by defective trafficking, we performed the following experiment. Dextrans were allowed to accumulate in the endocytic vesicles in PC-3 and LNCaP cells at 37°C for 30 min (shorter time intervals were not investigated), and colocalization of dextrans with Cxs was studied by fluorescence microscopy as described under "Experimental Procedures." The results of Fig. 6 show that in PC-3 cells, Cxs did not colocalize with the cell surface-derived endosomes as there was a clear demarcation between endocytosed dextrans and Cx immunostaining. On the other hand in LNCaP cells, appreciable cytoplasmic immunostaining for Cxs was colocalized with the endocytosed dextrans (Fig. 6, higher magnification), indicating normal Cx trafficking to cell surface followed by their endocytosis. These data suggest that in PC-3 cells, intracellular accumulation of Cxs is caused by impaired trafficking of Cxs to the cell surface and not by endocytosis. (37,40) have shown degradation of Cxs by both proteasomal and lysosomal pathways. Therefore, we next investigated whether intracellular accumulation of Cx43 and Cx32 was due to their resistance to degradation via these pathways. Fig. 7 shows that treatment with leupeptin, an inhibitor of lysosomal function, and ALLN, an inhibitor of the proteasomal pathway (40,41), further increased intracellular accumulation of Cx32 and Cx43 as judged by immunocytochemical (Fig. 7A) and by Western blot analyses (Fig. 7, B and C). Similar data were obtained with lactacystin (not shown), which is a more specific inhibitor of proteasomal pathway (41). Although we did not measure the half-life of intracellular Cx32 and Cx43 in PC-3 cells by pulse-chase analysis, our data suggest that intracellular Cxs are constantly degraded by lysosomal and proteasomal pathways regardless of whether or not they traffic to the cell surface.

Trafficking and Assembly of Connexins into Gap Junctions
Induced by ␣-Catenin-The Triton X-100-solubility and poor cell surface biotinylation of Cxs in PC-3 cells suggested that their trafficking to the cell surface was impaired. Moreover, the data in Fig. 7 further ruled out the possibility that intracellular accumulation of Cxs was caused by their aggregation into plaques resistant to degradation via proteasomal and lysosomal pathways. Because cadherin-mediated adhesion is nonfunctional in PC-3 cells (42), due to a deletion of the ␣-catenin gene (7), we asked if restoring cell-cell adhesion, which is conducive to the formation of gap junctions, would trigger the trafficking of Cxs from intracellular stores to the cell surface. Chicken ␣-catenin, which shows 90% homology to human ␣-catenin (43), was introduced transiently into Cx-expressing PC-3 clones using Lipofectin. In addition, yellow fluorescent protein chimeras of Cx43 and Cx32 were introduced into wild type PC-3 cells together with chicken ␣-catenin. Because of the interplay between junctional complexes, we examined expression of E-and N-cadherin after transient expression of ␣-catenin. Fig. 8 shows that ␣-catenin expression not only triggered the trafficking of Cxs from the intracellular stores to the cell surface but also induced their assembly into gap junctions as judged by immunocytochemical analysis (Fig. 8A) and by Triton X-100 insolubility assay of Cx32 (Fig. 8B) and Cx43 (Fig.  8C). We do not as yet know whether gap junctions formed after transient transfection of ␣-catenin into PC-3 cells are functional or nonfunctional. Our data also show that PC-3 cells expressed N-cadherin (and not E-cadherin), and its expression level did not seem to change after transient expression of ␣-catenin (data not shown). Furthermore, we found that transient expression of ␣-catenin also increased the expression of ZO-1 at the areas of cell-cell contact (data not shown).

FIG. 2. Immunolocalization of Cx32
and Cx43 in invasive PC-3 and indolent LNCaP prostate cancer cell lines. Cells were immunostained with polyclonal anti-Cx43 and monoclonal anti-Cx32 antibodies as described under "Experimental Procedures." Note most immunostaining in invasive PC-3 cells that express Cx43 (first row) and Cx32 (third row) is intracellular with only a few punctate dots characteristic of gap junctional plaques seen at the cell-cell contact areas (first row, arrows). In indolent LN-CaP cells that express Cx43 (second row) and Cx32 (fourth row), no intracellular immunostaining is observed, and most punctate dots are localized at the cell-cell contact areas (second and fourth rows, arrows). Note also that in some PC-3 cells (yellow arrows) scattered punctate dots are observed throughout the cytoplasm. No intracellular immunostaining was observed in non-permeabilized cells or when primary antibodies were omitted.

Intracellular Accumulation of Connexins Is a Common Feature of Androgen-independent Prostate Cancer Cell
Lines-We next investigated whether impaired trafficking of Cxs from intracellular stores to the cell surface was observed in other androgen-independent human PCA cell lines. Chimeras of Cx32 and Cx43 fused to yellow fluorescent proteins were introduced transiently into several androgen-independent PCA cell lines using the androgen-dependent LNCaP cells as a positive control ( Fig. 9 and Table I). The site of localization of these chimeras as well as formation of gap junctions were examined at 24 and 48 h post-transfection. Fig. 9 shows that transient transfection of Cx32-YFP and Cx43-YFP into ALVA-31, an androgen-independent PCA cell line, resulted in intracellular localization, whereas transfection into LNCaP cells resulted in localization of Cxs at cellcell contact areas indicative of the formation of gap junctions (indicated by arrows in bottom row). The data from several such experiments are summarized in Table I (see legends) and corroborate the data of Fig. 9.

DISCUSSION
The study reported here was motivated by our previous findings. First, we had observed that epithelial cells in well differentiated prostate tumors assembled Cxs into gap junctions whereas those in invasive and poorly differentiated prostate tumors did not and, instead, contained Cxs that were localized intracellularly (20). Second, we had shown that the forced expression of Cx32 and Cx43, the two Cxs expressed by the well differentiated epithelial cells of the normal prostate (44,45), into an indolent, androgen-dependent, but Cx-deficient PCA cell line, LNCaP, inhibited growth, induced differentiation, and retarded tumorigenicity (20). The main findings of our present study show that, in contrast to indolent LNCaP cells, forced expression of Cx43 and Cx32 into an invasive cell line PC-3, and several other androgen-independent PCA cell lines, resulted in intracellular accumulation. The accumulation of Cxs was probably not caused by the formation of ectopic gap junctions in the intracellular compartments, their aggregation into plaques resistant to degradation via proteasomal and lysosomal pathways, and their endocytosis upon arrival at the cell surface but by impaired trafficking to the cell surface. Most significantly, transient expression of ␣-catenin, a cadherinassociated protein that links cadherins to the cytoskeleton elements (46), induced trafficking of Cx32 and Cx43 from intracellular stores to the cell surface.
In contrast to LNCaP cells, we failed to detect significant amounts of cell surface as well as endocytosed Cx32 and Cx43 in PC-3 cells, suggesting that intracellular accumulation was caused by impaired or inefficient trafficking and not by endocytosis and internalization. Moreover, the following evidence substantiates our notion that intracellular accumulation was not caused by an artifact of overexpression of Cx32 and Cx43 and by mutations in the Cxs that might impede trafficking (49 -52) but rather by the pathological state of the PC-3 cells themselves. First, transient transfection of ␣-catenin abrogated impaired trafficking of Cx32 and Cx43 and induced their assembly into gap junctions (Fig. 8). Second, neither Cx32 nor Cx43 accumulated significantly in LNCaP cells, which expressed more Cxs than PC-3 cells when an equal amount of total protein was analyzed by Western blot analysis (Fig. 3). Third, intracellular Cx32 and Cx43 continued to be degraded via proteasomal and lysosomal pathways, indicating that they had not aggregated into degradation-resistant plaques.

FIG. 3. Triton X-100 insolubility of Cx32 and Cx43 in invasive PC-3 and indolent LNCaP prostate cancer cell lines. Triton X-100-soluble and -insoluble extracts from PC-3 and LNCaP cells were analyzed by Western blot analysis as described under "Experimental
Procedures." Note that only a small fraction of total Cx43 (A) and Cx32 (B) was not soluble in Triton X-100 in invasive PC-3 cells, whereas a major fraction of total Cx43 and Cx32 was Triton X-100-insoluble in LNCaP cells and in RL-CL9 cells, which are derived from rat liver and form abundant junctions composed of Cx43 only (29). C, Triton X-100 insolubility of Cx32 and Cx43 in PC-3 and LNCaP cells in situ. PC-3 and LNCaP cells were extracted in situ with 1% Triton X-100 and immunostained with anti-Cx43 and anti-Cx32 antibodies. Note that in Cx43-and Cx32-expressing PC-3 cells, nearly all intracellular Cxs are lost upon in situ extraction compared with Cx-expressing LNCaP cells which form abundant gap junctions. T, total; S, Triton X-100-soluble fraction; and I, Triton X-100-insoluble fraction.
The intrinsic and extrinsic determinants crucial for regulating gating, degradation, trafficking, and assembly of Cxs into gap junctions are poorly understood (19,40,(53)(54)(55). Impaired trafficking of wild type Cx43 and Cx32, leading to their intracellular accumulation, in PC-3 cells and its abrogation by ␣-catenin raises several intriguing questions about the molecular mechanisms involved in the trafficking of Cxs and their assembly into gap junctions. It has been proposed that bidirectional signaling between cell adhesion molecules and Cxs may be important in initiating the formation of gap junctions (34,56). Consistent with this notion, several studies have shown that restoration of cadherin-based cell-cell adhesion induces the assembly of Cxs into gap junctions (56), and conversely, its abolition impedes that assembly (57,58). Because previous studies (7) showed that transient expression of ␣-catenin in PC-3 cells triggered the recruitment of E-cadherin from the cytoplasm to the cell surface and restored cell-cell adhesion, we reasoned that intracellular accumulation of Cx32 and Cx43 might be caused by deficient E-cadherin-mediated cell-cell adhesion. Therefore, we expressed ␣-catenin in PC-3 cells. We chose to express ␣-catenin transiently because previous attempts to express it stably in PC-3 cells had failed. 3 Although our data are consistent with the possibility that transient expression of ␣-catenin triggered trafficking of both Cx32 and Cx43 and induced gap junction formation, the mechanism involved is likely to be complex, and restoration of E-cadherin based cell-cell adhesion alone may not be the cause.
First, in agreement with other studies, we found that both the parental and Cx-expressing PC-3 cells expressed N-cadherin and not E-cadherin at cell-cell contacts, suggesting that N-cadherin mediates cell-cell adhesion in these cells (7,59,60). Moreover, N-cadherin levels did not change significantly upon transient expression of ␣-catenin as assessed immunocytochemically and by Western blot analysis (data not shown). The reasons for the discrepancy between our data and those of others regarding E-cadherin expression in PC-3 cells are not understood. Second, we also found that ␣-catenin expression not only triggered the trafficking and assembly of Cxs into gap junctions but also recruited ZO-1, a tight junction and adherens junction-associated protein (72), to the cell surface. Third, ZO-1, Cx32, and Cx43 were not only co-localized with ␣-catenin at the areas of cell-cell contact but were also in the cytoplasm (data not shown). The significance of this finding is not understood at present and remains to be explored in the future (61,62). Fourth, the trafficking of Cx32 and Cx43 from intracellular stores to the cell surface as well as formation of gap junctions could be induced significantly upon increasing intracellular cAMP levels, 4 suggesting that alternative pathways exist, in- were seeded on glass coverslips and allowed to grow to 60% confluence. They were then allowed to uptake dextrans by incubating with the culture medium containing Alexa fluor 594-conjugated dextrans for 30 min. After rinsing once with PBS, cells were incubated with the culture medium for 30 min, and the medium was removed. Cells were rinsed with PBS three times, fixed with 2% paraformaldehyde, and immunostained for Cxs as described under "Experimental Procedures." Note a clear demarcation between endocytosed dextrans and Cx43 immunostaining in PC-3 cells and detectable colocalization of Cx43 with the dextrans in LNCaP cells. The cells in right-most panels represent higher magnification of those marked by white arrows in the third panels. Similar data were obtained with Cx32-expressing PC-3 and LNCaP cells. been shown to interact with vinculin, ZO-1, ␣-actinin, and actin in addition to binding to E-or N-cadherin and has also been proposed to be one of the key regulators of the structural integrity of several junctional complexes (63)(64)(65). Previous studies implicating cadherin-mediated cell-cell adhesion in facilitating the assembly of Cxs into gap junctions utilized cell lines in which Cxs trafficked normally, did not accumulate intracellularly, and in which only the capacity to assemble Cxs into gap junctions was defective (56,68,69). In addition, the role of adhesion in facilitating the formation of gap junctions is further complicated by studies showing inhibition of gap junction formation upon restoration of cell-cell adhesion mediated by N-cadherin (70) or vice versa (57,58). Our data clearly show that N-cadherin is expressed at the regions of cell-cell contact (data not shown). Previous studies have shown that ␣-catenin controls the strength of cell-cell adhesion through its interaction with ␣-actinin and the cytoskeleton (see Refs. 63 and 72 for discussion). Because the adhesive force of the extracellular domain of N-cadherin is sufficient for increasing cell-cell adhesion and there was no difference in the strength of cell-cell adhesion between control and ␣-catenin transfected cells as assessed by cell aggregation assay, 5 it is possible that ␣-catenin induces the trafficking of Cxs and their assembly into gap junctions by modulating the state of cell-cell adhesion.
Recent findings (1,5,(63)(64)(65)(66)(67) have shown that cell-cell or cell-matrix adhesions are a versatile and complex array of interactions, modulations, and signaling events rather than just adhesion and that there is extensive cross-talk between various junctional complexes formed as a result of these adhesions. Therefore, it is possible that ␣-catenin may induce trafficking of Cxs and their assembly into gap junctions by activating cellular signaling pathways in addition to modulating cell-cell adhesion (72). For example, conditional ablation of ␣-catenin in keratinocytes has been shown to increase proliferation by activating mitogen-activated protein kinase, independent of effects on cell-cell adhesion (73). Several signal transduction pathways have been shown to regulate the formation and dissolution of gap junctions (13)(14)(15)(16)(17)(18)(19). In this regard we note that mitogen-activated protein kinase has been shown to be constitutively activated not only in PC-3 cells but also in several other androgen-independent PCA cell lines in which ␣-catenin has been found to be deleted (74). Impaired trafficking of Cxs has been observed previously (50 -52) in a number of other diseases such as X-linked Charcot-Marie-Tooth disease, sensorineural hearing loss, erythrokeratoderma variabilis, visceroatrial heterotaxy, and cataract, but the impairment in most cases has been causally linked to the mutations in the Cxs themselves and not to the pathological state of the cells. In several studies intracellular accumulation of Cxs was attributed to an artifact of overexpression and/or internalization and ectopic formation of gap junctions (47,48). Therefore, the role of cell-cell adhesion in enhancing the formation of gap junctions indirectly via its effect on trafficking or via activation of multiple signal transduction pathways may have been overlooked. Our data implicate impaired trafficking of Cxs as an additional cause of communication deficiency in tumor cells and support the notion that regulating transport of Cxs by physiological effectors may be a mechanism to control the assembly of gap junctions as suggested by others (71).
The incidence of PCA increases with age and is characterized by progression from an indolent, slow-growing androgen-dependent state to an invasive, androgen-independent state (75)(76)(77). Because intracellular accumulation of Cx43 and Cx32 was observed only in epithelial cells from invasive prostate carcinomas (20) and in androgen-independent cell lines (present study), but not in an indolent and androgen-dependent cell line 5 P. P. Mehta and R. Govindarajan, unpublished observations. LNCaP (20), it is tempting to speculate that the pathways governing the trafficking of Cxs and their subsequent assembly into gap junctions become altered during the progression of PCA from an androgen-dependent to -independent state. Our Cx43-and Cx32-expressing androgen-independent PC-3 cells, in which ␣-catenin is deleted, and androgen-dependent LNCaP cells, in which the cadherin-based adhesion system has remained intact (20), offer two in vitro systems that mimic the behavior of Cxs in vivo and should prove useful in elucidating the molecular mechanisms by which ␣-catenin controls the FIG. 8. Transient transfection of ␣-catenin induces trafficking of connexins and formation of gap junctions. A, parental PC-3 cells were grown on glass coverslips and transiently transfected with pcDNA3-␣-catenin and pCx43-YFP and pCx32-CFP. After 48 h, cells were fixed in 2% paraformaldehyde, permeabilized, and immunostained with antibody against ␣-catenin (red). The colocalization of Cxs (green) with ␣-catenin was studied after acquiring z section images and deconvolution. Note the formation of puncta (gap junctions) at the cell-cell contact areas in cells expressing ␣-catenin (arrows). The last row is a higher magnification of the third row. Similar data were obtained with connexin-32-and connexin-43-expressing PC-3 clones. Nuclei, stained with DAPI, are blue. B and C, connexin-32-and connexin-43-expressing PC-3 clones were seeded in 10-cm dishes and allowed to grow to 50% confluence for 3 days. Cells were transiently transfected with pcDNA3-␣-catenin, and the assembly of gap junctions was studied by Triton X-100 insolubility 48 h after transfection. Control, untransfected cells; Transfected, cells transfected with pcDNA3-␣-catenin. Note the appearance of Triton X-100-insoluble bands in cells transfected with pcDNA3-␣-catenin. T, total; S, Triton X-100-soluble fraction; and I, Triton X-100-insoluble fraction; TRF, transfected; CNT, controls.
FIG. 9. Impaired trafficking of connexin-43 and connexin-32 in other human prostate cancer cell lines. Several androgen-independent human PCA cell lines (see Table I) were grown on glass coverslips and transiently transfected with pCx43-YFP and pCx32-CFP. Androgen-dependent LNCaP cells were used as a positive control. Cells were fixed in 2% paraformaldehyde 48 h post-transfection, and localization of connexin-43 and connexin-32 in the intracellular stores and at the cell-cell contact areas was observed under fluorescent microscope. Note intracellular localization of Cx43 and Cx32 chimeras fused to fluorescent proteins in an androgen-independent PCA cell line (arrows) and formation of gap junctions at the cell-cell contact areas (arrows, bottom row) in androgen-dependent LNCaP cells. The nuclei (blue) were stained with DAPI.
trafficking of Cx43 and Cx32 and their assembly into gap junctions.