Connexins are gap junction proteins that form aqueous channels to interconnect adjacent cells. Rat osteoblasts express connexin43 (Cx43), which forms functional gap junctions at the cell surface. We have found that ROS 17/2.8 osteosarcoma cells, UMR 106-01 osteosarcoma cells, and primary rat calvarial osteoblastic cells also express another gap junction protein, Cx46. Cx46 is a major component of plasma membrane gap junctions in lens. In contrast, Cx46 expressed by osteoblastic cells was predominantly localized to an intracellular perinuclear compartment, which appeared to be an aspect of the TGN as determined by immunofluorescence colocalization. Hela cells transfected with rat Cx46 cDNA (Hela/Cx46) assembled Cx46 into functional gap junction channels at the cell surface. Both rat lens and Hela/Cx46 cells expressed 53-kD (nonphosphorylated) and 68-kD (phosphorylated) forms of Cx46; however, only the 53-kD form was produced by osteoblasts. To examine connexin assembly, monomers were resolved from oligomers by sucrose gradient velocity sedimentation analysis of 1% Triton X-100–solubilized extracts. While Cx43 was assembled into multimeric complexes, ROS cells contained only the monomer form of Cx46. In contrast, Cx46 expressed by rat lens and Hela/Cx46 cells was assembled into multimers. These studies suggest that assembly and cell surface expression of two closely related connexins were differentially regulated in the same cell. Furthermore, oligomerization may be required for connexin transport from the TGN to the cell surface.
Gap junction channels mediate intercellular communication by allowing the direct transfer of ions and small aqueous molecules between neighboring cells. Gap junction channel proteins, or connexins, have been identified with sizes in the range of 26–56 kD (for reviews see 6, 7, 17, 21, 32). Channels composed of different connexins have distinct properties, and most tissues express more than one connexin. There is increasing evidence that multiple connexins can interact to form heteromeric gap junction channels (25, 52).
Regulation at the level of gene expression is clearly one important way that gap junction composition can be regulated (5, 10, 12, 14, 31, 50, 58). Connexin transport and assembly into gap junction channels are other possible points where cells can regulate the formation of gap junction channels (17, 38). A number of studies have shown that newly synthesized connexins are transported through the normal secretory apparatus (18, 39, 40, 46). However, in contrast with most multimeric membrane protein complexes that are formed in the ER (3, 13, 28), the identity of the site of connexin oligomerization remains controversial (32). Both ER (33) and Golgi apparatus (41) have been suggested as sites for connexin assembly into gap junction channels or hemichannels.
In previous studies we have examined the expression of connexins in human and rat osteoblastic cells (11, 30, 53). All of the cells examined express Cx43 (α1) in junctional plaques at the cell surface. In addition, some osteoblasts also produce Cx45 (α6), which also shows plasma membrane localization (53). In this paper we have characterized a third endogenous connexin expressed by rat osteoblastic cells, Cx46 (α3).
Cx46 expression has typically been associated with plasma membrane gap junction channels in lens (14, 22, 25, 29, 43, 49, 57). We found that Cx46 was expressed by primary rat osteoblastic cells and two osteosarcoma cell lines, ROS-17/2.8 (ROS) and UMR 106-01 (UMR) cells. In contrast with Cx43 and Cx45, Cx46 was largely retained in an intracellular perinuclear compartment. Little, if any, Cx46 accumulated on the cell surface, as determined by immunofluorescence microscopy. Instead, Cx46 was retained in a trans-aspect of the Golgi apparatus, probably the TGN. In contrast, we found that Hela cells stably transfected with rat Cx46 (Hela/Cx46) formed functional gap junction channels by Cx46 at the plasma membrane.
In addition to the differential transport of Cx43 and Cx46 by ROS cells, these cells also showed differences in connexin assembly. While Cx43 was assembled into hexameric gap junction hemichannels by ROS cells, the intracellular pool of Cx46 remained as unassembled monomers. Thus, transport and assembly of Cx43 and Cx46 are differentially regulated by osteoblastic cells. These data are discussed in the context of potential roles for connexin oligomerization in the regulation of transport.
Materials And Methods
ROS 17/2.8 and UMR 106-01 cells were cultured in MEM (No. 11095-056; GIBCO BRL, Gaithersberg, MD) containing 10% heat-inactivated bovine calf serum (BCS)1 (Hyclone, Logan, UT), 2 mM glutamine, 1 mM sodium pyruvate, 1% nonessential amino acids (GIBCO BRL), 5 U/ml penicillin, and 5 μg/ml streptomycin (MEM + BCS). Primary osteoblastic cells were isolated from fetal rat calvaria as previously described (51) and cultured in MEM + BCS for no more than six passages. Hela cells were kindly provided by Jeanette Pingle and Matthew Thomas (Washington University, St. Louis, MO).
Total cellular RNA was isolated using guanidium isothiocyanate and probed for rat Cx46 and human γ-actin as previously described (30).
A number of immunological reagents were generous gifts from the following researchers. Tissue-culture supernatant from murine hybridoma cells producing anti-Cx50 (MP70; α8) was from Drs. J. Kistler (University of Auckland, New Zealand) and D. Goodenough (Harvard University, Boston, MA) (14, 15). Tissue-culture supernatant containing monoclonal murine anti–rat TGN38 was from Dr. G. Banting (University of Bristol, UK) (24). Production of rabbit polyclonal antiserum that recognizes Cx43 was previously described (4).
Rabbit anti–Cx46 IgG was produced using a 6His-tagged Cx46 fusion protein. A cDNA consisting of the COOH-terminal tail sequence of rat Cx46 was produced by PCR amplification using a full-length rat Cx46 cDNA as a template (sense primer: 5′-CGAGA GCGAC ATATG CTAGA GATTT ACCAC CT-3′/antisense primer: 5′-AGCGA GCGGA TCCTT TCTAC CTGTT GATTT GA-3′). The resulting PCR product was amplified and inserted into the pET-15b vector (Novagen, Madison, WI) to produce a bacterial expression vector containing cDNA encoding for a fusion protein containing the Cx46 COOH terminus, a thrombin cleavage site, and six histidine residues (Cx46-his). Bacterial expression of Cx46-his was induced by treatment with isopropylthio-β-d-galactoside, cell extracts were prepared, and Cx46-his was affinity purified using a nickel/His-Bind chelation resin column (Novagen). This product was used to induce polyclonal antiserum in rabbits by established protocols.
Polyclonal serum that recognized Cx46 was initially identified by immunoblot analysis of rat lens total protein preparations that showed reactive bands with the expected Mr at 53 and 68 kD, consistent with previous reports (26, 29, 57). Preincubation of anti-Cx46 antiserum with antigen eliminated binding to both bands in samples prepared from lens and ROS cells (see Fig. 3, lanes 2 and 4). The perinuclear staining pattern obtained for osteoblasts labeled with anti-Cx46 (e.g., see Fig. 2) was eliminated by preincubation of the antiserum with Cx46-his protein (not shown). Also, cells that do not express Cx46 did not show any labeling by immunofluorescence or immunoblotting (see Fig. 4).
Cells were initially cultured on glass coverslips for 1–3 d before examination. The cells were washed twice in PBS, and then fixed and permeabilized for 2 min at room temperature using methanol/acetone (1:1). The cells were washed once with PBS containing 0.5% Triton X-100 and then twice with PBS containing 0.5% Triton X-100 and 2% heat-inactivated normal goat serum (PBS/TX/HIGS). The cells were then incubated with primary antiserum diluted to optimal titer in PBS/TX/HIGS for 45–60 min at room temperature. After washing three times with PBS/TX/HIGS, the cells were further incubated with fluorescent anti-IgG diluted in PBS/TX/ HIGS for 45–60 min. The cells were then washed and viewed by fluorescence microscopy using appropriate optics. For Cx46 labeling, anti-Cx46 antiserum was diluted 1:1,000 into PBS/TX/HIGS, and the secondary antiserum was rhodamine-conjugated goat anti–rabbit IgG (Boehringer Mannheim Biochemicals, Indianapolis, IN) diluted 1:500 in PBS/TX/HIGS. For double labeling experiments, Cx43 was labeled using monoclonal antiserum from Zymed Laboratories (South San Francisco, CA), and staining with fluorescent lectins was substituted for labeling with antisera to visualize some organelles. In some instances, images were obtained with a BioRad MRC-1000 confocal fluorescence microscopy system (Hercules, CA); all other images were obtained by epifluorescence microscopy using a Zeiss Axioscope (Thornwood, NY) and GIPSSPC image processing system (Georgia Instruments, Roswell, GA).
Protein Preparation and Immunoblotting
Total cell protein samples were prepared as previously described (30) and resolved by SDS-PAGE using standard methods and 10% polyacrylamide gels. The proteins were then electrophoretically transferred to polyvinylidene difluoride (PVDF) membranes (transfer buffer: 50 mM Tris, 380 mM glycine, 0.025% (wt/vol) SDS, 20% MeOH), blocked with blotto (40 mM Tris, 5% (wt/vol) Carnation powdered milk, 0.1% (vol/vol) Tween20) for 1 h at room temperature, and then blocked overnight with antiserum diluted into blotto. The membranes were then washed, and specific bands were detected using HRP-conjugated goat anti–rabbit IgG (Tago, Burlingame, CA) and enhanced chemiluminescence (ECL; Amersham Intl., Little Chalfont, UK). Cx50 immunoblots were detected using peroxidaseconjugated goat anti–mouse IgG + IgM (Boehringer Mannheim Biochemicals). After a series of exposures, films showing unsaturated bands were digitized and quantitated using an image analysis system (Georgia Instruments).
Sucrose Gradient Velocity Centrifugation
This procedure was adapted from that of Musil and Goodenough (40). Cells cultured for 1–3 d in 100-mm tissue-culture dishes were washed twice with PBS, and then scraped into PBS at 4°C and centrifuged at 500 g for 5 min. The cell pellet was resuspended in 2.75 ml incubation buffer (0.14 M NaCl, 5.3 mM KCl, 0.35 mM Na2PO4, 0.35 mM KH2PO4, 0.8 mM MgCl2, 2.7 mM Hepes, pH 7.4) containing protease (10 mM N-ethylmaleimide, 1 mM phenylmethylsulfonylchloride, 2 μg/ml leupeptin, 1 μg/ml pepstatin) and phosphatase inhibitors (1 mM NaVO4, 10 mM NaF) at 4°C and homogenized with a ball-bearing cell homogenizer (1). The preparation was then centrifuged at 500 g for 5 min to obtain a postnuclear supernatant and brought to 1% Triton X-100. After a 30-min incubation at 4°C, the sample was centrifuged at 100,000 g for 30 min to remove Triton-insoluble material, and then layered onto a 3.6-ml 5–20% sucrose gradient in incubation buffer + 0.1% Triton X-100 on top of a 0.2-ml 25% sucrose cushion. Samples from intact lens were treated in a similar manner, except that they were minced with a razor blade and incubated in 1% Triton X-100 for 12 h. The gradient was centrifuged using an SW55 rotor in a model L7-55 ultracentrifuge (Beckman Instruments, Inc., Palo Alto, CA) for 16 h at 148,000 g. After centrifugation, samples were collected from the bottom of the tube and diluted 1:1 with denaturing SDS sample buffer. Connexins were detected by immunoblotting as described above.
ROS cells were incubated in methionine-deficient MEM containing 10% dialyzed BCS for 1 h at 37°C, and then labeled for 5 h at 37°C with 300 μCi per dish 35S-Trans label (ICN Pharmaceuticals, Costa Mesa, CA). The cells were then harvested, and Triton X-100–soluble and –insoluble fractions were prepared as described above. The insoluble pool was resuspended, both fractions were diluted with PBS containing 0.25% BSA and 0.2% gelatin, and SDS was added to 0.5% final concentration. The samples were precleared with protein A–Sepharose and preimmune serum for 90 min at room temperature, and then incubated overnight with specific antiserum and protein A–Sepharose that was preblocked with BSA and gelatin. The beads were washed, resuspended in reducing SDS-PAGE sample buffer, and heated to 68°C to release proteins associated with the beads.
Rat Cx46 cDNA was excised with EcoRI and inserted into the expression vector pSFFVneo (8). Cx46/pSFFVneo was transfected into Hela cells by calcium phosphate precipitation. The cells were incubated overnight at 37°C, and then shocked by treatment with 15% glycerol for 2 min. The cells were then washed and plated in selective medium containing 1 mg/ml G-418. Clones of G-418–resistant cells expressing Cx46 were identified by immunofluorescence and immunoblotting.
Cx46 Is Expressed by Osteoblasts
By Northern blot analysis, both ROS and UMR cells were found to express Cx46 mRNA (Fig. 1). The intracellular distribution of Cx43 and Cx46 was examined by indirect immunofluorescence (Fig. 2). Consistent with our previous work (30, 53), ROS and primary rat calvarial cells showed high levels of Cx43 expression localized to areas where the cells were in close contact, corresponding to gap junctions. While UMR cells express relatively low levels of Cx43 as compared with ROS cells (53), we were able to locate areas where Cx43 was immunolocalized to the cell surface (Fig. 2 c). In contrast, all three cell types showed expression of Cx46. In each case, Cx46 was predominantly localized to an intracellular compartment in the perinuclear region of the cell.
Lens cells typically show the highest levels of Cx46 expression (14, 22, 25, 29, 43, 49, 57). When total lens protein extracts were examined by Western blotting, two forms of Cx46 were found with Mr of 68 and 53 kD (Fig. 3, lanes 1 and 5). This is consistent with previous reports, where the 68-kD band corresponds to a phosphorylated form of Cx46 (26, 29) and may also have other posttranslational modifications (29). In contrast, immunoblots of total proteins isolated from ROS, UMR, and primary osteoblastic cells derived from rat calvaria showed that these cells expressed only the 53-kD form of Cx46. This finding suggests that Cx46 expressed by osteoblastic cells is likely to be nonphosphorylated, although low levels of Cx46 phosphorylation cannot be ruled out. Osteoblasts do not show a general deficiency in connexin phosphorylation, since multiple forms of Cx43 corresponding to phosphorylated species were observed in immunoblots of total protein samples from ROS cells (Fig. 3, lane 9) (38, 39). Also, ROS cells showed qualitatively similar amounts of Cx43 and Cx46 expression, as determined by immunoblotting (Fig. 3, lanes 9 and 10).
Elfgang et al. (16) found that Hela cells were highly deficient in connexin expression. We confirmed by Western blotting and immunofluorescence that wild-type Hela cells did not express Cx46 (Fig. 4), Cx43, or Cx50 (not shown). Thus, Hela cells were stably transfected with rat Cx46 cDNA (Hela/Cx46) to examine the targeting of Cx46 in nonosteoblastic cells.
In contrast with osteoblasts, Cx46 was present at the cell surface of Hela/Cx46 cells, as well as in intracellular compartments (Fig. 4,c). While these cells were not extensively coupled, Cx46 expressed by Hela/Cx46 cells formed functional gap junction channels, as determined by intercellular transfer of microinjected Lucifer yellow (Fig. 5). Texas red ovalbumin did not transfer between Hela/Cx46 cells. This indicates that retention of Cx46 in an intracellular compartment depends upon the cell type examined. Furthermore, since Cx46 was readily detected at the surface of Hela/Cx46 cells, it is likely that accumulation of appreciable amounts of Cx46 on the osteoblast cell surface would be visible by immunofluorescence microscopy. However, this does not rule out the possibility that Cx46 is transported to the cell surface and rapidly cleared by reinternalization.
The expression of Cx46 by transfected Hela cells was also examined by immunoblotting (Fig. 4,a). Hela/Cx46 cells showed both the 68- and 53-kD forms of Cx46 that are found associated with lens. This contrasts with results obtained with osteoblastic cells, where only the lower molecular mass form was present (Fig. 3). Thus, the ability to produce the 68-kD form of Cx46 correlated with the ability to form gap junctions containing Cx46. Whether one process is a prerequisite for the other is not known at present.
Cx46 Is Retained in the TGN
The identity of the intracellular compartment containing Cx46 was determined with a series of fluorescence colocalization experiments in ROS cells. To label endocytic compartments, cells were incubated at 37°C for 1 h with medium containing the fluorescent fluid phase marker, Lucifer yellow. ROS cells labeled in this manner were fixed and immunostained for Cx46, and then examined by fluorescence microscopy. In these double-labeled cells, fluorescently labeled endocytic compartments and intracellular vesicles containing Cx46 showed distinct intracellular distributions (Fig. 6, a and b).
ER in fixed and permeabilized cells was preferentially labeled using a fluorescent lectin, fluorescein-conjugated conconavalin A (FITC-ConA), which recognizes the high mannose core oligosaccharide component of glycoproteins. Areas of the ER that are prominently labeled with FITC-ConA, such as the nuclear envelope and lattices in the periphery of the cell, showed little labeling for Cx46 (Fig. 6, c and d).
The Cx46-containing compartment in ROS cells colocalized with the Golgi apparatus by double-labeling immunofluorescence using a marker for the medial-Golgi stack (anti-GCI) (9). As shown in Fig. 7, a and b, both the antiGCI and Cx46 showed prominent perinuclear localization. However, these two proteins were in distinct compartments. This was revealed by treatment of the cells with brefeldin A (BFA), which preferentially disassembles cis and medial aspects of the Golgi apparatus (54). When ROS cells were preincubated for 5 min with BFA and then fixed and examined by immunofluorescence, the GCI marker redistributed to the periphery of the cell, while the Cx46 staining pattern was relatively intact (Fig. 7, c and d). This suggests that Cx46 was located in a secretory compartment beyond the medial Golgi, either the trans-Golgi cistern or the TGN.
To further examine this possibility, we performed colocalization experiments using a monoclonal anti–rat TGN38 antibody (24). As shown in Fig. 7, e and f, both Cx46 and TGN38 colocalize to the perinuclear region of the cell. The distribution of these two proteins was not affected by shortterm treatment with BFA (not shown). However, in cells incubated for 30 min in the presence of BFA, both Cx46 and TGN38 condensed into the same concentrated spot in the nuclear region (Fig. 7, g and h). This is similar to the pattern of TGN38 labeling in normal rat kidney (NRK) cells treated for 30–60 min with BFA (34, 48). Taken together, these observations suggest that the intracellular pool of Cx46 likely corresponds to an aspect of the TGN.
We attempted to stimulate the transport of Cx46 to the plasma membrane by plating osteoblastic cells under a variety of culture conditions. Under the conditions examined, little, if any, Cx46 was found to be localized to the cell surface. We also incubated ROS cells for short periods (2–12 h) before immunofluorescence labeling with a number of agents known to affect bone cell metabolism (47), such as β-glycerophosphate (2.3–5.8 mM) + ascorbate (0.1–0.25 mg/ml), dexamethasone (10–250 μM), PMA (0.1 μM) ± bromo-A23187 (0.1 μg/ml), parathyroid hormone (0.125–1.0 μM), β-estradiol (0.5–50 μg/ml), 1,25-dihydroxy-vitamin D3 (5–25 ng/ml), and dibutyryl-cAMP (0.1 mM), as well as a variety of plating densities and mechanical stress. None of these treatments altered the intracellular distribution of Cx46 in ROS cells.
Osteoblasts Retain Cx46 as a Monomer
To examine the assembly of Cx43 and Cx46, we used techniques previously described by Musil and Goodenough (40) using 1% Triton X-100 at 4°C, conditions that enable the solubilization of intact Cx43 hexamers in other cell types. Cx43 and Cx46 were solubilized to different extents (Fig. 8). Note that Cx43 already assembled into gap junctional plaques is resistant to solubilization under these conditions (39). As determined by immunoprecipitation analysis, ∼50% of the Cx43 in ROS cells was soluble in 1% Triton X-100 at 4°C, while virtually all of the Cx46 was solubilized under the same conditions.
Detergent-extracted connexin preparations were resolved into monomers and multimers by sucrose gradient velocity sedimentation (Fig. 9). Cx43 from ROS cells resolved into two fractions. The major fraction centered at 8% sucrose corresponds to Cx43 monomers, while the peak at 15% sucrose corresponds to multimers. Multimeric Cx43 consisted of 21.4 ± 2.8% (n = 3) of the total Triton X-100–soluble Cx43. The preponderance of Cx43 monomers over hexamers was expected, since the Triton-soluble fraction of Cx43 is likely to contain mostly newly synthesized Cx43 (39, 40). However, some monomeric Cx43 may be due to spontaneous disassembly of hexameric Cx43 during the course of the experiment.
Treatment of NRK cells with BFA inhibits the oligomerization of Cx43, suggesting that Cx43 is assembled into oligomers in a late Golgi apparatus compartment (40). If this is the case, then treatment of ROS cells with monensin, which inhibits cis to medial transport through the Golgi apparatus, should also inhibit Cx43 oligomerization. As shown by immunofluorescence microscopy in Fig. 10, monensin-treated ROS cells showed Cx43 and Cx46 retained in a similar perinuclear region of the cell. This is consistent with results obtained with monensin-treated cardiac myocytes (45). When oligomerization of Cx43 was examined in monensin-treated cells, only 8% (n = 2) of the Triton-soluble Cx43 isolated from monensin-treated cells was assembled into multimers, as compared with control cells that had two- to threefold more Cx43 in the hemichannel form (Fig. 11). Also, since we could detect this difference in the amount of Cx43 assembled into hemichannels, it is likely that nonspecific aggregation of Cx43 did not occur during the solubilization and sucrose gradient fractionation procedure.
In contrast with Cx43, Cx46 solubilized from ROS cells migrated as a single peak centered at 8% sucrose (Fig. 9 b). Since nearly all of the Cx46 was soluble in 1% Triton X-100, this suggests that all of the Cx46 was being retained by ROS cells in a monomeric form.
One possible concern in using Triton X-100–solubilized preparations is that Cx46-containing oligomers may not be stable under these conditions. To examine this possibility, we prepared Triton X-100–soluble extracts from both Hela/Cx46 cells and whole rat lens. While Hela/Cx46 cells were solubilized using the same conditions used for ROS cells, extensive incubation with 1% Triton X-100 at 4°C was required to solubilize detectable amounts of material from rat lens. Both the 53- and 68-kD forms of Cx46 were solubilized under these conditions (Figs. 12 and 13).
When Triton X-100–soluble extracts from lens were examined by sucrose gradient fractionation, the majority of the Cx46 was found to be in oligomeric forms (Fig. 12,a). We also examined this gradient for Cx50, another abundant lens connexin (14, 15). The profile for Cx50 was very comparable to that observed for Cx46 in lens (Fig. 12,b). Similar results were obtained by sucrose gradient analysis of Cx46 solubilized from Hela/Cx46 cells, although the peak at 14–18% was significantly reduced (Fig. 13,a). The sucrose gradient profiles of Cx46 isolated from lens and Hela/Cx46 cells are consistent with the notion that multimers containing Cx46 were largely stable in the presence of Triton X-100. This is further supported by Fig. 13 b, where Cx46 extracted from Hela/Cx46 cells and then dissociated by treatment with 0.2% SDS migrated on the gradient as a single peak at 8–9% sucrose.
In this paper we have examined the expression of Cx46 by osteoblastic, lens, and transfected Hela cells. Osteoblastic cells retained Cx46 as a monomer in an intracellular compartment that is likely to be an aspect of the TGN. In contrast, Cx46 was assembled into oligomers by lens and Hela/Cx46 cells and transported to the cell surface. This raises the possibility that connexin assembly and transport are interlinked and regulated processes.
Connexins are assembled into a hemichannel in an intracellular compartment before transport to the plasma membrane (33, 40). This hemichannel then pairs with another hemichannel on a neighboring cell to form a complete gap junction channel. There is evidence supporting both ER and Golgi apparatus as sites for connexin oligomerization. Kumar et al. (33) have found by EM that BHK cells transfected with Cx32 (β1) form intracellular gap junctional structures in the ER as well as the plasma membrane. This indicates that Cx32 assembly into hemichannels also occurred at the ER and is consistent with “classical” models for transmembrane protein assembly (3, 13, 28). An alternative model has been proposed by Musil and Goodenough (40). Using sucrose gradient fractionation to analyze connexin assembly, they found that Cx43 is assembled in a late Golgi compartment of NRK and CHO cells, most likely the TGN. When cells were treated with BFA, oligomerization was inhibited. Similar results were obtained using temperature-sensitive mutant cell lines deficient in ER to Golgi transport.
We used sucrose gradient fractionation to examine the oligomerization state of Cx43 and Cx46 present in Triton X-100 extracts. Treatment of ROS cells with monensin, which inhibited cis- to medial-Golgi transport, also reduced Cx43 hemichannel formation by >62% (Fig. 10). While this evidence is not conclusive, it is consistent with a trans-Golgi compartment as the site of Cx43 oligomerization in this system. Since Cx43 assembly was not completely inhibited, it is possible that Cx43 hemichannels may also be assembled in the ER. Alternatively, some of the hemichannels that we detected in the presence of monensin may be due to Cx43 disassembled from gap junctions that is en route to degradation.
We found that Cx46 oligomers were stable in Triton X-100 (Figs. 12 and 13). Sucrose gradients have also been used to analyze connexin channels isolated from ovine lens (25). Analogous to chick, ovine, and bovine lens (25, 29), it is likely that most hemichannels isolated from rat lens contained Cx50 in addition to Cx46. Preliminary analysis of ROS and UMR cells indicates that these cells do not express Cx50 (not shown). An absolute requirement for Cx50 is not likely for Cx46 incorporation into gap junction channels, since we found that Hela/Cx46 cells were able to form functional gap junctions in the absence of Cx50, albeit at relatively low levels. Formation of gap junctions purely from Cx46 has also been described in other systems (59, 60).
Cx46 was retained in a monomer, rather than a hemichannel form, by ROS cells. Based on colocalization experiments shown above, it seems likely that osteoblasts transported Cx46 to compartments where connexin oligomerization occurs. This is consistent with a requirement for connexin oligomerization as a prerequisite for transport from the TGN and/or retention at the cell surface. The notion of quality control at the level of the Golgi apparatus is supported by the observation that some mutant viral glycoproteins exit the ER without being properly assembled, but are retained in the Golgi apparatus (20, 42).
Immunofluorescence was used to determine whether Cx46 was present at the plasma membrane of osteoblastic cells. This technique does not completely rule out the possibility that Cx46 transport to the cell surface occurs, but suggests that it is rapidly reinternalized after delivery to the plasma membrane. TGN38 (48) and furin (37, 56) are examples of proteins that continuously recycle at low levels between the TGN and plasma membrane, yet are not readily detectable at the cell surface by conventional indirect immunofluorescence. Furthermore, the communication-deficient cell lines L929 and S180, which do not show accumulation of Cx43 at the plasma membrane by immunofluorescence (41), contain a pool of Cx43 that is accessible by surface biotinylation (39). Interestingly, S180 and L929 cells show low levels of Cx43 phosphorylation (41), analogous to our observations with Cx46 in osteoblasts. However, these situations are not completely comparable, since both S180 and L929 cells assemble Cx43 into hemichannels (40). Cx43 hemichannels at the cell surface have also been observed in other systems (35).
It is also possible that Cx46 is transported by osteoblasts along a regulated secretory pathway that is distinct from the transport pathway for Cx43. Rat uterine myometrial cells provide a precedent for the stimulated transport of connexins to form gap junctions (2, 10, 19, 23, 44). Whether the intracellular compartment containing Cx46 is a regulated secretory compartment is not known at present. However, storage vesicles that accumulate other transmembrane proteins subject to stimulated transport to the plasma membrane, such as GLUT4, accumulate in the perinuclear region in other cell types (27, 36).
Since Cx43 and Cx46 had distinct intracellular distributions in osteoblastic cells, this indicates that these two connexins were not randomly intermixing in these cells. It is not known whether Cx43 and Cx46 can form heteromeric channels in other systems, and there is no a priori reason that these two α-connexins cannot intermix. In fact, Cx43 has been observed to colocalize near lens gap junctional plaques that also contain Cx46 (14). Also, Cx43 and Cx46 expressed by Xenopus oocytes can form functional heterotypic channels (55, 59).
Since Cx43 and Cx46 have different fates in osteoblastic cells, our data support the notion that connexin transport is a specific process and not due to the bulk flow of connexins along a “default” secretory pathway. Also, oligomerization seems to be interconnected with connexin targeting and transport to the cell surface. Further work will be required to find structural determinants that either target connexins to specific intracellular locations or control connexin transport at the level of gap junction channel assembly.
We are grateful to Dr. S. Kornfeld and the anonymous reviewers for their very helpful comments. We also thank A. Robertson for technical assistance with the RNA blot and Dr. R. Civitelli for providing calvarial cells.
1. Abbreviations used in this paper: BCS, bovine calf serum; BFA, brefeldin A; NRK, normal rat kidney; PVDF, polyvinylidene difluoride.
This research was supported in part by National Institutes of Health grants GM45815 and DK46686.
Please address all correspondence to Michael Koval, Department of Medicine, Washington University School of Medicine, Campus Box 8051, 660 S. Euclid Avenue, St. Louis, MO 63110. Tel.: (314) 362-6606. Fax: (314) 3629230. e-mail: firstname.lastname@example.org
As of July 1, 1997, M. Koval's address is Institute for Environmental Medicine and Department of Physiology, University of Pennsylvania School of Medicine, 1 John Morgan Building, 3620 Hamilton Walk, Philadelphia, PA 19104.