|Home | About | Journals | Submit | Contact Us | Français|
Gap junctions are intercellular channels that connect the cytoplasms of adjacent cells. For gap junctions to properly control organ formation and electrical synchronization in the heart and the brain, connexin-based hemichannels must be correctly targeted to cell-cell borders. While it is generally accepted that gap junctions form via lateral diffusion of hemichannels following microtubule-mediated delivery to the plasma membrane, we provide evidence for direct targeting of hemichannels to cell-cell junctions through a pathway that is dependent on microtubules; through the adherens-junction proteins N-cadherin and β-catenin; through the microtubule plus-end-tracking protein (+TIP) EB1; and through its interacting protein p150(Glued). Based on live cell microscopy that includes fluorescence recovery after photobleaching (FRAP), total internal reflection fluorescence (TIRF), deconvolution, and siRNA knockdown, we propose that preferential tethering of microtubule plus ends at the adherens junction promotes delivery of connexin hemichannels directly to the cell-cell border. These findings support an unanticipated mechanism for protein delivery to points of cell-cell contact.
Gap junctions, which are formed by the serial coupling of hemichannels of adjacent cells, allow direct sharing of ions and small cytoplasmic molecules. Each hemichannel is a hexamer of connexin, and the most common isoform is Connexin43 (Cx43). Gap junctions reside at cell-cell borders, where their density is of critical importance. In the heart, for example, gap junctions are concentrated at the intercalated disc (ID) that joins the ends of cardiomyocytes, and this is where they ensure propagation of action potentials (Gros and Jongsma, 1996; Shaw and Rudy, 1997). Altered Cx43 gap junction distribution following cardiac ischemia contributes to malignant ischemic arrhythmias (Kaprielian et al., 1998; Peters et al., 1997; Shaw and Rudy, 1997). Preventing or reversing this process offers a strategy to repair damaged heart (Abraham et al., 2005; Reinecke et al., 2004). Understanding the molecular mechanism of gap junction localization at the cell-cell border, therefore, is important not only for addressing the basic cell biological question of gap junction formation but also for developing treatments for life-threatening diseases.
A model for gap junction formation has emerged from biochemical and cell biological studies over the past decade. The half-life of Cx43 is between 1 and 3 hr (Beardslee et al., 1998; Laird et al., 1991), which indicates that gap junction trafficking is a dynamic process. Gap junction hemichannels that are packaged in vesicles emerge from the Golgi, reach the cortical membrane via microtubules, which have insertion sites all over the cell surface (Jordan et al., 1999; Lauf et al., 2002), and form clusters of gap junctions known as plaques at cell-cell borders. FRAP studies show that connexin hemichannels can float freely within the cortical membrane (Lauf et al., 2002). Over a time course of tens of minutes or hours, newer hemichannels are evident first at the plaque perimeter and later on throughout the plaque (Gaietta et al., 2002; Lauf et al., 2002), which leads to this commonly held view of gap junction plaque formation: after their microtubule-mediated delivery, hemichannels diffuse laterally within the membrane to cell-cell border regions. In this model, hemichannels that coalesce at the periphery of the plaques then move inwards, and those reaching the center of the plaque become internalized for degradation as annular gap junctions (Laird, 2005; Segretain and Falk, 2004).
The current model is the simplest scenario that could account for the experiments reported thus far, but it does not explain how plaques occur at specific locations of the cell membrane. While junctional proteins such as ZO-1 may retain connexins (Giepmans and Moolenaar, 1998), few other connexin-binding proteins have been identified. In multiple tissues, gap junctions colocalize with adherens junctions (AJs) formed by cadherins. At the cardiac ID, the AJs are assembled prior to the establishment of a Cx43 gap junction plaque (Angst et al., 1997). Notably, mislocalized Cx43 plaques in ischemic myocardium are associated with similarly mislocalized AJs (Matsushita et al., 1999). Indeed, transfection of E-cadherin into gap junction-incompetent cells allows the transfected cells to form functional gap junctions (Mege et al., 1988). Moreover, N-cadherin knockout mice cannot form gap junctions (Luo and Radice, 2003), and conditional knockout of N-cadherin in the heart causes poor expression and mislocalization of gap junctions, thus leading to arrhythmogenic death (Li et al., 2005).
Since Cx43-laden microtubules must interact with the plasma membrane to deliver the hemichannels and because cadherin-mediated signaling is implicated in microtubule regulation (Jamora and Fuchs, 2002), we wondered whether plaque formation involves microtubule interaction with the AJ. We looked into the possible involvement of the microtubule plus-end-tracking proteins (+TIPs) because they control microtubule dynamics and could be captured by membrane-associated proteins (Akhmanova and Hoogenraad, 2005; Mimori-Kiyosue et al., 2005). Among the +TIPs, the dimeric EB1 associates directly with the plus ends of microtubules and has dual binding sites for a number of proteins, including p150(Glued) (Berrueta et al., 1999), which is a component of the dynein/dynactin complex. The dynein/dynactin complex, in turn, can tether microtubules at AJs (Chausovsky et al., 2000; Ligon et al., 2001).
In this study, we analyzed plaques formed by fluorescently tagged Cx43 in HeLa cells that do not express endogenous Cx43. Within a few minutes following the photobleaching of a plaque and its surrounding area in our FRAP studies, there was rapid replenishment of fluorescent Cx43 within the plaque but not in the bleached surrounding area. The recovery was sensitive to reagents that alter microtubule dynamics. Time lapse imaging of live cells expressing fluorescently tagged EB1 and Cx43 further showed that microtubule plus ends approached the plaque more frequently than they approached the rest of the cortical membrane and that they remained at the plaque for longer periods of time. Moreover, gap junction plaque formation was disrupted by siRNA knockdown of EB1, of its binding partner p150(Glued), and of the AJ protein β-catenin. It was further disrupted by the application of peptides that compromise homophilic cadherin interactions at the AJ. In addition, TIRF microscopy showed enhanced EB1 interaction and Cx43 vesicle delivery to the cell surface that rested on coverslips coated with cadherin extracellular domains. Thus, mammalian cells use +TIPs for targeted delivery of Cx43 to form gap junction plaques that are spatially coincident with AJs; this mechanism could restrict intercellular communication to cells of the same tissue type.
In isolated HeLa cells transiently transfected with a construct containing Cx43 tagged with yellow fluorescent protein (Cx43-YFP), Cx43-YFP was concentrated in the perinuclear Golgi. The protein then translocates to the cell-cell border when two cells came into contact. Imaging isolated cells with low levels of Cx43-YFP expression with long exposure times indicated that Cx43-YFP is localized to microtubules (Figure 1A). To avoid artifacts due to fixation, we repeated the imaging but did so in pairs of live HeLa cells with shorter exposure time at multiple Z planes, and we used deconvolution software to resolve the distribution of GFP-α-tubulin and Cx43-RFP (Figure 1B), which showed a strong correlation between cortical microtubule ends and Cx43 plaques. Within sparsely populated plaques, it is possible to resolve individual microtubules and their orientations and locations relative to Cx43 distribution, revealing direct association of microtubule tips with the plaque regions with the highest Cx43 intensity (Figures 1C and 1D).
Next, in FRAP studies we used a confocal microscope to follow Cx43-YFP dynamics in live HeLa cells with or without treatments that alter microtubule dynamics. Entire plaques and their surrounding areas were subjected to fluorescence photobleaching, and the recovery of fluorescence was monitored every 20 s for 5 min, revealing rapid repopulation of the plaque in its original geometry. Plaque reappearance was inhibited by 100 μM nocodazole, which depolymerizes microtubules, and 10 μM Taxol, which stabilizes microtubules but prevents growth at the plus end (Morrison et al., 1998; Figure 2).
After bleaching in the horizontal XY plane, we recorded, depending on plaque geometries, in the horizontal plane for horizontally aligned plaques (data not shown) and in the vertical plane for vertically aligned plaques (Figure 2A), and we found no significant differences in recovery kinetics. Whereas Cx43-YFP in the cell-cell border as well as the lower level of Cx43-YFP in surrounding regions were readily detectable, pooled data from both regions revealed rapid plaque repopulation, while nonplaque regions acquired much lower levels of fluorescence (Figure 2B); this suggests direct delivery of connexin to the plaque.
The efficiency of plaque formation was quantified by allowing HeLa cells transfected with Cx43-YFP to form plaques over 18–24 hr prior to the nocodazole (100 μM, 30 min) or Taxol (10 μM, 2 hr) treatment and then the effect of these reagents on isolated pairs of cells expressing Cx43-YFP was assessed (Figure 3A) by measuring the percent of the cell-cell border with cortical connexin plaques (Figure 3B). Drug treatment caused a marked reduction in the concentration of Cx43 at the cell-cell border.
To test whether microtubule plus ends preferentially interact with the cortical membrane in the vicinity of connexin plaques, we cotransfected HeLa cells with Cx43-RFP and fluorescently tagged EB1, which is a marker for rapidly growing microtubule plus ends (Figure 4A), then compared time series of EB1 dynamics at Cx43 plaques, at regions of cell-cell border without Cx43 plaques, or at cell edges not in contact with other cells (Figure 4B). EB1-capped microtubules grew toward the connexin plaque three times more frequently than those approaching membrane not in contact with other cells, and each microtubule lasted on average 3.5 times longer when it was within 2 μm of a plaque (Figure 4C; Movie S1). The significantly greater microtubule plus end orientation (frequency of cortical approaches) and stabilization (duration of appearance near the cortex) at plaques (Figure 4C) are a likely underestimate because we excluded nonplaque regions without EB1. Greater EB1 orientation and stabilization at plaques may account for the rapid repopulation of plaques after photobleaching (Figure 2).
Since imaging with widefield epifluorescence involves collapsing the depth of a three-dimensional structure into a plane, we used TIRF microscopy that limits the imaging depth to within 50–100 nm of the coverslip and examined cells that had been transfected with EB1-GFP and plated on coverslips coated with the extracellular domain of N-cadherin, thus allowing homophilic N-cadherin-N-cadherin contacts to form between the cell and the coverslip. Control coverslips were created with IgG. There was a much greater density of EB1 comets in the immediate vicinity of the plasma membrane and a longer EB1-cell membrane interaction in cells plated on coverslips coated with cadherin extracellular domains (Figures 4D and S1), which is consistent with the greater abundance of EB1 comets found near the cell-cell border (Figures 4A–4C). These results indicate that cadherin-based AJs may capture and anchor EB1-tipped microtubules to the plasma membrane.
For direct visualization of vesicles being targeted for fusion with the cell membrane, we again used TIRF microscopy and monitored Cx43-YFP in cells plated on coverslips coated with either the extracellular domain of N-cadherin (Figures 5A and 5B) or IgG (control; Figures 5C and 5D).
With TIRF microscopy to capture Cx43-YFP signals that originated only from membrane and submembrane regions, we analyzed individual Cx43 vesicle movements in the central square of the image (Figures 5A and 5C). Starting from the appearance of a vesicle in the last minute of a 5 min sequence of images (marked as time = 0 s), we manually traced the appearances and paths of five vesicles, as indicated by five differently colored lines with colored arrows marking their final positions (Figures 5B and 5D). At the end of the traced path, the vesicles typically became brighter and immobile in cells that were in contact with cadherin extracellular domains (Figure 5B; see also kymograph in Figure S2 and Movie S2). We interpreted these terminal events as being vesicle fusion with the plasma membrane, which is consistent with previous observations that most gap junctions are immobile within the plasma membrane (Jordan et al., 1999).
The vesicles of cells contacting the extracellular domain of N-cadherin were notable for a short prefusion path length (indicating directed delivery) and a high incidence of fusion (Figures 5A, 5B, and 5E). In contrast, in cells plated on coverslips coated with IgG and no N-cadherin (Figures 5C, 5D, and 5E) vesicles exhibited few fusion events and longer path lengths (the vesicles that disappeared by the end of the minute are marked by an “X” at the end of their respective paths; see also the kymograph in Figure S2 and Movie S3). In addition to performing visual analysis and measuring path length for 158 manually tracked vesicles (Figure 5E), we quantified fusion events that were automatically computed for 35 cells (Figure 5F), which revealed that the cells plated on coverslips with N-cadherin domains have six times as many fusion events as the control cells have.
In our TIRF imaging, there were rare occasions where microtubule orientation made it possible to observe a dynamic sequence in which several connexin vesicles apparently moved progressively toward the membrane just above the cadherin extracellular domains along a microtubule and fused with the membrane, thus causing a plaque at the membrane to grow in size (Movie S4). Subsequent immunofluorescence labeling of microtubules revealed their presence along the tracks for Cx43 vesicle movements. Consistent with membrane distribution of Cx43 being a function of directed delivery, the membrane patch with lowest overall Cx43 density corresponds to the regions with lowest microtubule density (middle panel, Movie S4). In addition, the computed vesicle velocity along the track, 1.8 ± 0.2 μm/s (n = 3 vesicles), is within the range of kinesin-based transport along microtubules (Hirokawa et al., 1998), which is consistent with the notion of microtubule-based Cx43 delivery directly into a growing gap junction plaque.
Next, we asked whether Cx43 plaque formation requires specific proteins that interact with the microtubule plus ends. Because Taxol can dissociate EB1 from microtubule plus ends (Morrison et al., 1998) and because EB1 appears to be involved in microtubule tethering (Figure 4), we tested the involvement of EB1 in plaque formation by exposing HeLa cells to EB1 siRNA for 24 hr before transfecting them with Cx43-YFP. Knockdown of EB1 protein (Figures S3 and S4) caused a substantial reduction of connexin plaque formation at the cell-cell border (Figures 6C and S5) and at the cell membrane that was in contact with cadherin extracellular domain as revealed by TIRF imaging (Figures S6 and S7), but it did not prevent AJ formation in HeLa cells or hepatocytes (Figures S8 and S9). Taken together with our finding of a concentration of EB1 comets near the plaque (Figures 4C, 4D, S1, and S5; Movie S1), the requirement of endogenous EB1 for plaque formation suggests that EB1 at the microtubule plus ends may interact with proteins at the cell-cell border and enhance targeted delivery of Cx43. Lending further support to the importance of EB1 in plaque formation, EB1 coimmunoprecipitated with Cx43-YFP in transfected HeLa cells (Figure S10).
The cadherin-based AJ is a highly efficient “glue” that binds cells together and may also regulate the microtubule cytoskeleton (Chausovsky et al., 2000; Waterman-Storer et al., 2000). Moreover, β-catenin, which is the cytoplasmic binding partner of cadherin, binds dynein (Ligon et al., 2001). The dynein/dynactin complex is a potential cortical anchor of microtubules (Fuchs and Karakesisoglou, 2001) since p150(Glued) in the dynactin complex has a binding domain for EB1 (Askham et al., 2002; Hayashi et al., 2005). Therefore, we tested for the involvement of p150(Glued) and β-catenin in forming gap junction plaques.
Compared to HeLa cells expressing Cx43-YFP, with p150(Glued) in a cytoplasmic pool and at the cell-cell border near the plaques formed by Cx43-YFP, siRNA knockdown of p150(Glued) greatly decreased Cx43-YFP accumulation at the cortical membrane of the cell-cell border (Figures 6A, bottom row, and and6C).6C). Reducing p150(Glued) expression also decreased EB1 dwell time at the cell border (Movie S5) but did not prevent AJ formation (Figure S11), which is consistent with the role of p150(Glued)-containing dynein/dynactin in anchoring microtubules at the cell-cell border (Dujardin and Vallee, 2002; Koonce and Samso, 2004). Dynein/dynactin may thus help Cx43-containing vesicles on microtubules reach the cortex and offload connexin to the plasma membrane at the cell-cell border.
If p150(Glued) connects the microtubule plus end to the AJ, β-catenin likely constitutes part of the cortical anchor, given that both β-catenin and the p150(Glued)-containing dynactin complex bind dynein. Indeed, immunocytochemistry of HeLa cells transfected with Cx43-YFP, with or without coexpression of β-catenin siRNA (Figure 6B), revealed that not only was Cx43-YFP colocalized with β-catenin, but siRNA knockdown of β-catenin also significantly reduced plaque formation (Figures 6B and 6C).
We then examined the distribution of endogenous p150(Glued), β-catenin, and Cx43 in adult ventricular cardiomyocytes and found that they colocalize at the IDs. Colocalization of cardiac Cx43 and β-catenin (Figure 6D) is consistent with the observation that the ID is rich in both gap junctions and AJs. The colocalization of p150(Glued) and Cx43, both in isolated cardiomyocytes and at intact cell-cell borders of isolated pairs (Figure 6E), indicates that p150(Glued) is poised in native heart cells to connect cadherins with microtubule plus ends for Cx43 delivery.
We next asked whether plaque formation also depends on cadherin. The extracellular domain of cadherin engages in homophilic interactions with cadherin from adjoining cells while the cytoplasmc domain of cadherin binds β-catenin. In order to test if functional N-cadherin on the cell membrane contributes to the formation of gap junction plaques, we applied a peptide that blocks homophilic intercellular N-cadherin interactions (Frenzel and Johnson, 1996) to HeLa cells already transfected with Cx43-YFP and adherent on coverslips. This treatment disrupted most cell-cell contacts, although isolated cell pairs could still be found (Figure 7A). Cells were fixed 24 hr after exposure to the peptides, and the cortical connexin at the cell-cell border was quantified for isolated cell pairs (Figure 7B), revealing that N-cadherin-blocking peptides caused a major reduction of connexin deposition at the cell-cell border. Another way to disrupt cadherin-mediated AJ formation is to induce actin depolymerization (Ko et al., 2001). We found that actin depolymerization agents also reduced Cx43 plaque formation (data not shown). Taken together, these observations suggest that homophilic cadherin interactions transmit the signal of cell-cell contact to induce targeted Cx43 delivery.
Although microtubules are uniformly laden with connexin (Figure 1A), most microtubules do not reach the cellular cortex. However, regions containing gap junction plaques always have at least one associated microtubule (Figure 1B) and appear to interact with EB1-capped microtubules more frequently and for a greater duration (Figure 4C) than those regions that do not contain gap junction plaques. These observations raise the possibility of targeted, microtubule-mediated delivery of connexin.
If microtubules deliver Cx43 directly to the plaques, fluorescent Cx43 would repopulate bleached plaques with a distribution that mimics the prebleach plaque morphology rather than appearing first at the plaque perimeter. Indeed, our FRAP studies revealed a fast repopulation (recovery half-time of 2.8 min) that restored the original plaque morphology (Figure 2). Moreover, nocodazole and Taxol worked immediately in the postbleach period to retard recovery of the plaque, which indicates that the fast repopulation of a plaque involves microtubule-mediated transport from the cell interior.
How might our findings be reconciled with experiments that led to the lateral diffusion model of gap junction plaque formation? Previous studies have documented fusion between Cx43-containing vesicles and the plasma membrane of isolated cells as well as the presence and mobility of tagged Cx43 on the nonjunctional surface of HeLa cells (Jordan et al., 1999; Lauf et al., 2002). However, gap junction delivery to the plasma membrane depends not only on the potential for Cx43 hemichannels to arrive at non-contact regions (Jordan et al., 1999; Lauf et al., 2002) but also on the relative frequency of microtubule approaches to the cell-cell border versus other regions of the plasma membrane (Figure 4). Taken together with the colocalization of Cx43 plaques with β-catenin (Figures 6B and 6D), studies to date are consistent with microtubule-based transport that delivers connexin to the AJ to a greater extent than it does to the rest of the cell membrane.
Plaque repopulation has also been investigated using both FRAP and successive labeling of tetracysteine-tagged Cx43 with different fluorophores (Gaietta et al., 2002; Lauf et al., 2002). In the FRAP studies, 30 min to several hours after bleaching a portion of the plaque, YFP-tagged Cx43 was observed to accumulate at plaque edges (Lauf et al., 2002). The time resolution and the low fluorescence sensitivity necessitated by the imaging of the remaining unbleached plaque in this study would have precluded the detection of rapid plaque repopulation within a few minutes (Figure 2). In the tetracysteine-labeling studies (Gaietta et al., 2002), newer Cx43 appears at the perimeter of the plaque when examined 4–8 hr after the initial Cx43 labeling, though there appear to be low levels of newer Cx43 within the central “old” plaques. Given that gap junctions are internalized at the center of a plaque within several hours (Laird, 2005), it seems likely that the size of the plaque is maintained by the formation of new plaques as membranes at the plaque perimeter come into close apposition. The bright rim of fluorescent Cx43 observed in the earlier studies could, under this scenario, represent new plaques formed entirely by freshly delivered Cx43.
We are not aware of a report that directly shows tagged Cx43 from nonborder regions of the cell membrane reaching an existing plaque, though this type of lateral diffusion probably accounts for the residual recovery that occurred when photobleaching followed nocodazole treatment in this and previous studies (Lauf et al., 2002; Thomas et al., 2005). In our FRAP studies, we bleached the entire plaque and observed repopulation at 20 s intervals, which revealed a rapid, microtubule-dependent repopulation of the original plaque (Figure 2). Using peptides that prevent homophilic cadherin interactions without interfering with cadherin placement on the cell surface, we further showed that homophilic cadherin-cadherin interaction is important for gap junction formation (Figures 7A and 7B). These findings are consistent with previous reports of the calcium dependence of cell-cell communication (Davidson et al., 1984) that has been attributed to an adhesion event and the cadherin family of proteins (Musil and Goodenough, 1990; Takeichi, 1990). In agreement with reports that gap junction plaque formation is preceded by cell-cell contact and homophilic cadherin-cadherin interaction (Angst et al., 1997; Hertig et al., 1996; Kostin et al., 1999), our results support a model for targeted Cx43 delivery that involves more frequent and longer-lasting microtubule attachment to the AJ (Figure 4). In particular, our TIRF microscopy data (Figures 5 and S6; Movie S4) indicate that AJs are preferential Cx43 delivery sites that depend on EB1. This mechanism of plaque localization at AJs places cell-cell communication channels precisely at the locations of cell-cell contact, though it does not preclude microtubule-mediated transport of connexin to nonborder regions of the plasma membrane generally or lateral diffusion of connexin within the cell membrane. The ability of Cx43 to bind microtubules (Giepmans et al., 2001) could also contribute to microtubule interactions at plaques, perhaps after their capture via cadherin-associated proteins and +TIPs.
Our FRAP studies revealed that nocodazole and Taxol both cause disruption of the fast repopulation of Cx43 plaques (Figure 2). While nocodazole depolymerizes microtubules, Taxol stabilizes them but disrupts their interactions with EB1 (Nakata and Hirokawa, 2003), which normally associates with rapidly growing microtubule plus ends (Schuyler and Pellman, 2001). We found that EB1-capped microtubules interact more frequently and for a longer period of time at plasma membrane-containing plaque than at regions of plasma membrane that did not contain plaque (Figures 4C and 4D), that EB1 coimmunoprecipitates with Cx43-YFP expressed in HeLa cells (Figure S10), and that EB1 knockdown decreases plaque formation at the cell-cell border (Figures 6C, S5, and S6), thereby implicating EB1 as a major player in gap junction formation.
How might the plus end-binding protein EB1 facilitate connexin delivery? The C terminus of EB1 binds the p150(Glued) subunit of the dynein/dynactin complex (Askham et al., 2002; Hayashi et al., 2005). Moreover, dynein/dynactin localizes with AJs at cell-cell contact points through direct binding with β-catenin (Ligon et al., 2001). In this manner, dynein/dynactin may serve as an anchor for microtubules at the AJ. We found that knockdown of p150(Glued) or β-catenin disrupts the formation of gap junction plaques (Figure 6) and that p150(Glued) knockdown impairs capture of EB1 at cortical regions of cell-cell contact (Figure 4C; Movie S5). While it is possible that targeted connexin delivery involves other cytoskeletal elements, +TIPs, and cortical proteins, this study identifies some key components of functional importance: the microtubule plus end-binding protein EB1, the EB1-binding protein p150(Glued) as part of the dynein/dynactin complex that could tether microtubules to the AJ, and β-catenin as the cytoplasmic enforcer of homophilic cadherin-cadherin interaction (Figures 7C–7E). Most likely, actin also acts as an important initial sensor of cell-cell interaction and guides the localization of AJs (Drees et al., 2005) with assistance from Rho-GTPases (Noren et al., 2001). It will be interesting to determine how other proteins associated with the cytoskeleton or cell-cell junctions may contribute to targeted delivery of gap junction proteins.
The proposed model for connexin trafficking suggests a mechanism for membrane protein localization at specific regions of the plasma membrane, which is a pathway of particular interest in developmental biology. Based on preferential adhesion between cells expressing the same cadherin type(s) or similar cadherin levels, cadherin-cadherin interactions affect cell sorting during tissue development (Wheelock and Johnson, 2003). Our findings suggest that connexin reaches the AJ via direct, microtubule-mediated delivery so that gap junctions may form preferentially between cell types that express the same type of cadherin. Mutations in Cx43 are associated with deafness, cataracts, germ cell development defects, oculodentodigital dysplasia, and cardiac outflow abnormalities (Wei et al., 2004). It would be interesting to explore the potential involvement of microtubule-mediated Cx43 delivery to AJs in these developmental processes and pathological conditions.
Cancer cells tend to lose gap junction communication that is normally maintained between nonmalignant cells and, once they invade other tissues, generally form gap junctions only with other cancer cells. Just as loss of E-cadherin or upregulation of N-cadherin can lead to tumor invasiveness (Mareel and Leroy, 2003), both gain of and loss of connexin expression are associated with cancers. Cx43 expression is downregulated in a variety of cancers, including prostate, lung, bladder, and cervical cancers, as well as in gliomas and melanomas (Mesnil, 2002). Like Cx43, Cx26 targeting in HeLa cells was also compromised by siRNA knockdown of EB1 (Figure S12). Reduced gap junction coupling may help cells complete malignant transformation by limiting the spread of Ca2+-mediated apoptosis (Krutovskikh, 2002), while overexpression of connexins may lead to increased invasiveness of human glioblastomas (Oliveira et al., 2005). At present, there are few data linking cadherin to connexins in malignant cells. Testing for a correlation between targeting of connexins and cadherin-based AJs may be a first step toward developing new strategies to limit tumor growth and metastasis.
Cardiac excitation is an iterative process whereby inward current depolarizes the plasma membrane of a cardiomyocyte and initiates an action potential, which, in turn, drives a depolarizing current that spreads to adjacent cardiomyocytes through gap junctions. The flow of current for proper cardiac action potential propagation requires that electrical coupling through gap junctions be both robust and oriented along the long axis of cardiac fibers. Decreased or disorganized gap junction coupling due to myocardial ischemia leads to ventricular arrhythmias of sudden cardiac death and contributes to the pathogenesis of congestive heart failure. In addition, disorganization of gap junctions in ischemic myocardium is preceded by disorganized AJs (Matsushita et al., 1999). Our model for +TIPs-mediated gap junction trafficking suggests that AJ disruption contributes directly to the rearrangement of gap junction distribution. Furthermore, we provide a mechanism for gap junction localization that will likely apply to the subcellular targeting of other membrane proteins.
Details of standard technique, antibodies used, sequences, and nonessential imaging are in Supplemental Experimental Procedures.
HeLa and rat hepatic epithelial cells (WB-F344, which express endogenous Cx43) were cultured in standard mammalian cell conditions. FuGene 6 (Roche) was used for all cDNA tranfections. For immunocytochemistry, cells were cultured on 12 mm coverslips (Warner Instrument Corp.) that were precoated with bovine fibronectin (50 μg/ml). Taxol (10 μM) and nocodazole (100 μM; Sigma-Aldrich) were applied for 2 hr prior to fixation unless otherwise specified. N-cadherin peptide N-Ac-INPISGQ-NH2 (Williams et al., 2000; 10 mM; Anaspec, San Jose, CA) was applied shortly after transfection. Cells were fixed with methanol (−20°C) for 5 min.
For siRNA knockdown, cells were transfected with siRNA (all RNA from Dharmacon, Lafayette, CO) at a final concentration of 100 nM using Lipofectamine RNAiMAX Reagent (Invitrogen, Carlsbad, CA). For cells transfected with both cDNA and siRNA, siRNA transfection occurred 24 hr prior to cDNA transfection.
Cx43-YFP signal in isolated cell pairs was collected for a fixed exposure time. Cell borders were identified by double staining with N-cadherin, and the percentage of Cx43-YFP plaque was calculated by dividing the number of pixels with intensity greater than a preset intensity threshold by the number of pixels defining the border. Statistical significance was tested by one-tailed Student’s t test.
HeLa cells were transfected with Cx43-YFP, harvested, lysed, and exposed to agarose beads that had been pretreated with anti-EGFP antibody. Material bound to washed beads was eluted, boiled, separated, and probed with anti-EB1 antibody.
FRAP studies were performed on HeLa cells transfected with Cx43-YFP using a Zeiss LSM confocal microscope with a 63× objective. Sampling rate was 20 s per time point, and bleach time was 40 s at high laser intensity.
Widefield imaging was performed on HeLa cells that had been cotransfected with Cx43-mRFP and either GFP-tubulin (Clontech) or GFP-EB1. A Nikon TE2000-U inverted fluorescence microscope was used with a Photometrics Coolsnap HQ CCD camera. Stacks for deconvolution were obtained with 100 nm spacing, and deconvolution was performed using Autoquant software. Statistical significance was evaluated by one-tailed Student’s t test.
Uncoated glass-bottom chambers (Number 1.0, MatTek Corporation) were precoated with poly-L-lysine (PLL; 1 mg/ml) and allowed to dry overnight in a sterile hood. The amino groups on the PLL were then activated for 90 min with 2.5% gluteraldehyde in 50 mM Na-phosphate buffer (pH 6.8), washed several times, and exposed to Cy5-tagged anti-human Fc secondary antibody (diluted 1:10 in 100 mM sodium carbonate, pH 9.4; Jackson ImmunoResearch Laboratories). +Cadherin coverslips were then generated via the application of a fusion protein between human Fc and N-cadherin (diluted 1:10 in PBS; R&D Systems). −Cadherin coverslips were generated by exposure to buffer once the secondary antibody was applied. HeLa cells were plated on the coverslips and transfected with Cx43-YFP.
Imaging was carried out using a Nikon TE-2000E inverted microscope with a 60× 1.45 NA TIRF objective equipped for 1.5× Optivar magnification and through-the-objective TIRF illumination using a 488 nm argon laser. TIRF imaging of Cx43-YFP in the presence of drugs and knockdown reagents as well as for GFP EB1 and of Cx43-RFP was carried out with a 100× 1.49 NA TIRF objective using either a 488 nm, 514 nm, or 543 nm argon laser. Cells were maintained in Hank’s BSS + 5% Fetal Bovine Serum (UCSF Cell Culture Facility) at 37°C. Time lapse sequences were acquired at a continuous rate of two frames per second with 200 ms of exposure per frame. Cx43-RFP imaging required 400 ms of exposure per frame with capture at the same frame rate.
Automated quantification of vesicle fusion events was achieved by first performing a temporal average of the last minute of images acquired during a 5 min acquisition. This technique enriched for fixed over mobile vesicles. A minimum intensity threshold of 300 AU/pixel was applied to remove background signal (same threshold for each cell), and vesicles were counted by selecting for particles between 400 nm and 1000 nm in diameter. All TIRF image processing was performed with ImageJ software (NIH).
Rat ventricular myocytes were isolated from adult Sprague Dawley rats (200 to 300 g; Charles River) after dissociation with collagenase type 2 (Worthington, Lakewood, NJ) with previously described methods (Hu et al., 2003). Attachment to coverslips, fixation, and immunostaining followed the same protocols as those used for HeLa cells (above).
We are grateful to Dr. Dale Laird (Ontario) for the Cx43-YFP and the Cx26-YFP constructs, Dr. James Trosko (Michigan State) for the WB-F344 hepatic epithelial cell line, Dr. Bert Vogelstein (Johns Hopkins) for the GFP-EB1 construct, Dr. Xiang Qian (Jan Lab) for technical assistance with cardiomyocytes, and other members of the Jan Lab for comments on this manuscript. This work was supported by National Institutes of Health Grants NIMH 065334 (L.Y.J.), NIDA 10711 and 10154 (M.v.Z.), NHLBI 075449 (R.M.S.), and American Heart Association Grant 0475022N (R.M.S.). Y.-N.J. and L.Y.J. are Howard Hughes Medical Institute Investigators.
Supplemental Data include 13 figures, five movies, Supplemental Experimental Procedures, and Supplemental References and can be found with this article online at http://www.cell.com/cgi/content/full/128/3/547/DC1/.