Jump to
Abstract
The intercellular geometry of connexin43 (Cx43) gap junctional coupling is key to coordinated spread of electrical activation through the ventricle of the mammalian heart. A progressive redistribution of electrical and mechanical junctions into intercalated discs occurs during postnatal development. Breakdown of disc-localized pattern in the adult heart, to recapitulate immature distributions, is thought to be key to the genesis of conduction disturbance and arrhythmia. Recently, ZO-1 (a PDZ-MAGUK protein), has been suggested to have a role in generating coupling geometries between myocytes. We therefore investigated the codistribution of ZO-1 with Cx43 and N-cadherin in the adult rat ventricle using quantitative immunoconfocal and immunoelectron microscopy. These analyses indicated that, whereas ZO-1 and Cx43 codistribute within discs, only low to moderate point-by-point colocalization of Cx43 and ZO-1 is found within these domains compared with the relatively high level of colocalization between N-cadherin and ZO-1. By contrast, levels of association between Cx43 and ZO-1 increased rapidly and significantly (P<0.001) after partial or complete enzymatic dissociation of myocytes from intact ventricle—a treatment known to induce gap junction endocytosis. Coimmunoprecipitation using Cx43- and ZO-1-specific antibodies confirmed that significantly (P<0.03) increased ZO-1 is precipitated relative to Cx43 in freshly dissociated myocytes as compared with intact ventricle. On immunoblots, decreases in Cx43 relative mobility, consistent with increased phosphorylation, were observed following myocyte dissociation. The increased ZO-1-Cx43 association that occurs after remodeling of myocyte intercellular contacts indicates the possibility of unanticipated roles for ZO-1 in gap junction turnover during cardiac development and disease processes.
The gap junction (GJ) is an aggregate of channels that functions in intercellular communication by directly linking the cytoplasm of adjacent cells.1 These channels provide a pathway for metabolic coupling and homeostatic exchange of molecules, as well as for propagation of electrical activation in excitable tissues such as the heart.2–5⇓⇓⇓ Each of the apposed membranes in a GJ is composed of a honeycomb-like array of transmembrane channels termed connexons. Connexons from adjacent cells combine to form an intercellular channel that permits particles of less than 1000 Daltons to pass freely. Connexons, in turn, are composed of 6 protein subunits known as connexins, of which multiple subtypes exist in various animal species, cells, and tissue types. In the atrial and ventricular myocardia of adult and developing mammalian heart, connexin43 (Cx43)-containing GJs contribute to low-resistance pathways for propagation of electrical activation.1–7⇓⇓⇓⇓⇓⇓ Atrial myocardium furthermore expresses Cx40 and Cx45, connexins also found in the pacemaker and conduction systems.3–5,7⇓⇓⇓
In the mammalian neonate, GJs and Ca2+-dependent adhesion junctions (adherens junctions and desmosomes) are distributed uniformly across the sarcolemma of ventricular myocytes.8–10⇓⇓ During postnatal development, the distribution of these junctions undergo a progressive remodeling, leading to the development of a highly organized structure at myocyte termini known as the intercalated disc. There is a lag in the rate at which GJs accrue at the disc compared with that of mechanical junctions.10 The eventual preferential colocalization of GJs at intercalated discs is a significant component of the normal maturation of electromechanical function and stability of cardiac activation.11,12⇓ The understanding of the mechanisms regulating differentiation of cell-cell contacts at intercalated discs has pertinence to human cardiac disease.2,4,13⇓⇓ Breakdown of disc localization of GJs in ventricular myocardium13 and, as recently reported by Chien and coworkers in conduction tissues,14 to recapitulate distributions found at immature developmental stages may be a key factor in the genesis of conduction disturbance and arrhythmia.
The mechanisms involved in cumulative targeting of Cx43-GJs to the intercalated disc are uncharacterized. At present, 2 hypotheses have been proposed to explain how GJs become preferentially localized to this region of specialized sarcolemma. In the first, it has been independently suggested by 2 groups that there may be different rates of turnover of GJs in different domains of sarcolemma.10,15⇓ In particular, it has been proposed that GJs associated with concentrations of adhesion junctions at the disc are preferentially maintained over GJs located in myocyte lateral domains. In the second hypothesis, Cx43-containing GJs are actively maintained at the intercalated disc via direct interactions with cytoskeletal-related anchoring or adapter proteins and other disc components. This latter hypothesis has received support from identification of zonula occludens-1 (ZO-1), a PDZ domain-containing protein localized at the intercalated disc,16–18⇓⇓ as a candidate molecule mediating anchorage of Cx43 to the cytoskeleton. Originally discovered in association with the tight junction plaque, ZO-1 is one member of the membrane-associated guanylate kinase family, which are known to function in protein targeting, signal transduction, and determination of cell polarity.19 ZO-1 has been linked to the function of tight junctions,19 adherens junctions,16 and more recently, GJs.17,18,20⇓⇓ Evidence supporting protein-protein interactions between Cx43 and ZO-1 have come from landmark work involving coimmunoprecipitation17,18⇓ and the yeast 2-hybrid system.20 Together this work has shown that the terminal 5 amino acids (DDLEI) of Cx43 bind the second PDZ domain of ZO-1 and that this interaction may be regulated by src-tyrosine phosphorylation of Cx43.18
In the following study, we have used immunoconfocal, immunoelectron microscopy, and coimmunoprecipitation to examine the relationship between Cx43 and ZO-1 in the ventricular myocardium of the adult rat heart. Our data indicate that association between ZO-1 and Cx43 is surprisingly limited to low to moderate levels in intact ventricular myocardium. Furthermore, we find that disruption of intercellular contacts between myocytes, a treatment inducing gap junction endocytosis,21 results in a dynamic change in the pattern of association between ZO-1 and Cx43. Our data suggests previously unanticipated roles for ZO-1 in the turnover of Cx43 during, or after, gap junctional endocytosis.
Materials and Methods
Myocyte Isolation
All reagents were purchased from Sigma Chemical Company, unless otherwise indicated. In this and subsequent procedures, 90-day-old (adult) Holtzman Sprague-Dawley rats were used. Rats were purchased from Harlan Biosciences (Indianapolis, Ind). Experiments were conducted in accordance with the Guide for the Use of Experimental Animals at the Medical University of South Carolina and conformed to the Guide for the Care and Use of Laboratory Animals published by the NIH (Publication No. 85-23, revised 1985). The myocyte isolation procedure used was based on the Langendorff perfusion method developed by Isenberg and Klockner22 with minor modifications (see expanded Materials and Methods section in the online data supplement available at http://www.circresaha.org).
Immunoconfocal and Immunoelectron Microscopy
Immunolabelings of ZO-1 (rabbit polyclonal, Zymed Laboratories, San Francisco, Calif), Cx43 (mouse monoclonal, Chemicon International, Temecula, Calif), and N-cadherin (mouse monoclonal) were carried out and imaged on a BioRad MRC 1024 laser scanning confocal microscope (LSCM) using standard protocols outlined previously10,23⇓ (see data supplement). Double immunolabelings were done using FITC and CY-5 secondary fluors to eliminate “bleed over” between detection channels. For immunoelectron microscopy, ventricular samples were prepared as previously described,24 with modifications as outlined (see data supplement), labeled using a rabbit polyclonal antibody directed against ZO-1 (Zymed), and viewed on a JEOL 200CX EM.
Western Blotting
Immunoblotting was carried out using standard methodologies as outlined previously23 with slight modifications (see data supplement). Primary antibodies used were against Cx43 (Sigma rabbit polyclonal anti-Cx43) and ZO-1 (Zymed rabbit anti-ZO-1). NIH Image was used for densitometric analyses of gels scanned on a ChemImager gel imager (Alpha Innotech Corp).
Immunoprecipitation-Immunoblotting
The immunoprecipitation method used was based on a protocol by Lukas et al25 with modifications (see data supplement). Immunoprecipitations were performed using 2 anti-Cx43 antibodies against independent Cx43 epitopes (Sigma rabbit polyclonal anti-Cx43 No. C6219, Zymed polyclonal rabbit anti-Cx43 No. 71-0700). To confirm reproducibility, immunoprecipitations were repeated at least 3 times. To test specificity of the reaction, further control immunoprecipitations were carried out in the presence of a peptide inhibitor (Zymed polyclonal rabbit anti-Cx43 peptide). Immunoblots were performed to detect Cx43 and ZO-1 as described.
Colocalization and Statistical Analyses
FITC and CY-5 channels were thresholded to precisely match immunolabeling patterns as previously described,10,23⇓ superimposed, and ZO-1 particles overlapping Cx43 particles were counted. The area of thresholded Cx43 particles and area of myocyte profile were also measured using NIH Image software. A pixel-by-pixel colocalization analysis, using proprietary BioRad software, was also used to assess levels of ZO-1 colocalization with Cx43. One-way ANOVAs were carried out on the data using Minitab statistical software. A more detailed description of the colocalization and statistical analyses is given in the expanded Materials and Methods section that can be found in the online data supplement available at http://www.circresaha.org.
Results
Cx43 and ZO-1 Show Limited Coimmunolocalization Within Intercalated Discs
Consistent with previous findings, our immunoconfocal surveys of rat ventricle confirmed that ZO-1 was localized in endothelial tissues, interstitial cells, and at zones of cell-to-cell contact between myocytes, including intercalated discs.17,18,26⇓⇓ To examine association between ZO-1 and Cx43 or N-cadherin at intercalated discs in detail, we next performed double immunolabelings followed by optical sectioning at high resolution on the confocal microscope.
Initially, analyses were confined to sections of ventricular myocytes arrayed longitudinally (Figure 1). In such perspectives, discs are typically resolved as linear profiles. These studies confirmed that Cx43, N-cadherin, and ZO-1 were all localized within intercalated discs. However, detailed inspection and analysis of immunofluorescence colocalization indicated that the proportional degree of point-by-point overlap between Cx43 and ZO-1 immunolabeling was not as high as that of N-cadherin and ZO-1 (compare Figures 1a and 1b with 1c and 1d).
Figure 1. a and b, Cx43 (red-CY-5 secondary fluor) and ZO-1 (green-FITC secondary fluor); c and d, N-cadherin (red-CY-5) and ZO-1 (green-FITC) double immunolabelings in intact ventricular myocardium. Overlap between the fluors is shown in yellow. Details shown in b and d correspond to the boxed regions in a and c. Scale bar=25 μm.
The difference in relative levels of colocalization was more apparent if discs were optically sectioned and reconstructed en face (Figure 2 and 3⇓). In such views, components of the disc may be observed as part of an extended 2-dimensional surface.27 Although points of overlap occurred between immunolabeled ZO-1 and Cx43, for the most part, the 2 immunolocalized proteins appeared to have relatively independent patterns of distribution (Figures 2a through 2c). A view of the signal channel images for Cx43 (Figures 2d through 2f) and ZO-1 (Figures 2g through 2i) in the same single optical sections confirms that the signal overlap between these 2 proteins is low to moderate. In contrast, ZO-1 and N-cadherin colocalized at relatively much higher spatial densities in these en face disc perspectives (Figures 3a through 3c). A much higher degree of correspondence between the distribution of these proteins is apparent in the single optical sections labeled for N-cadherin (Figures 3d through 3f) and ZO-1 (Figures 3g through 3i).
Figure 2. a through c, 3-D reconstruction of a ventricular intercalated disc in en face orientation immunolabeled for Cx43 in red and ZO-1 in green (a), with single channel perspectives of Cx43 and ZO-1 in b and c, respectively. *Cx43 and ZO-1 colocalizations. d through f and g through i, Single optical sections of Cx43 and ZO-1 labelings, respectively, through the disc. Arrows point to Cx43 immunopositive structures that show low levels of colocalization with ZO-1. Scale bar=5 μm.
Figure 3. a through c, 3-D reconstruction of an intercalated disc immunolabeled for N-cadherin in red and ZO-1 in green (a), with single channel perspectives shown of N-cadherin and ZO-1 in b and c, respectively. *Areas in which both N-cadherin and ZO-1 signal are correspondingly low. d through f and g through i, Single optical sections of N-cadherin and ZO-1 labelings, respectively, through the disc. Arrows indicate areas of significant colocalization. Scale bar=5 μm.
Because the results outlined appeared to differ from previous reports, we examined the association between GJs and ZO-1 in intact adult ventricular myocardium using immunogold electron microscopy (EM). Ultrastructurally defined GJs and cell adhesion junctions were readily identified in ultrathin sections of adult ventricle labeled with antibodies against ZO-1. The undulating profiles of adherens junctions were always extensively labeled with gold particles (Figure 4a), including those directly adjacent GJs (Figure 4b). Similarly, vascular endothelial cells also showed prominent accumulations of gold particles at presumed tight junctions (data not shown). By contrast, though wholly consistent with the immunoconfocal analyses, all 22 ultrastructurally defined GJs observed in this survey showed no specific immunogold labeling (Figure 4b).
Figure 4. a and b, Immunogold electron microscopy of fascia adherens (labeled fa with white arrows in a and b) and gap junctions (labeled gj with black arrow in b) in ventricular myocardium. Scale bar=500 nm.
Colocalization of Cx43 and ZO-1 Increases After Dissociation of Ventricular Myocytes
Based on our immunoconfocal and immuno-EM data, we concluded that there was evidence for only limited colocalization between ZO-1 and Cx43 at intercalated discs in vivo. Nonetheless, the data for interaction between ZO-1 and Cx43 from the earlier coimmunoprecipitation and yeast hybrid studies were compelling.17,18,20⇓⇓ We speculated that if significant interaction occurred, it may be transient, and thus not easily resolved in intact myocardium. One well-characterized dynamic change in GJ distribution occurs following disruption of intercellular contacts. During isolation of adult cardiomyocytes, the entire population of GJs is internalized, undergoing redistribution from the sarcolemma to cytoplasmic vesicles known as annular GJs and cell surface located gap junctional vesicles.21 We therefore undertook immunoconfocal analyses of freshly isolated ventricular myocytes. In such cells, striking increases in levels of colocalization between immunolabeled Cx43 and ZO-1 were found (Figures 5a through 5c) relative to those observed at GJs in intact myocardium (Figures 1a and b). It should be noted that although association increased, a proportion of ZO-1-immunolabeled particles was not associated with Cx43 and vice versa. This increase in ZO-1-Cx43 colocalization is rapid, being detectable within 5 minutes of initiation of collagenase dissociation (Online Figure 1 available at http://www.circresaha.org). Interestingly, this figure also reveals immunolocalization of ZO-1 in myocyte nuclei in partially dissociated cells.
Figure 5. a through f, Freshly isolated ventricular myocytes immunolabeled for ZO-1 (FITC-green) and Cx43 or N-cadherin (CY-5-red). a and b show ZO-1 and Cx43 immunolabeling, respectively. c, Color merge of a and b. d and e, N-cadherin and ZO-1 immunolabeling, respectively. f, Color merge of d and e. The inset is of the boxed region in f. Internalized ZO-1 particles are indicated with arrows. Scale bars=50 μm.
An increase in association after cell isolation was not observed between ZO-1 and N-cadherin (Figures 5d through 5f). Indeed, in freshly isolated ventricular myocytes, ZO-1 and N-cadherin colocalization appeared to decrease. Furthermore, whereas N-cadherin was maintained exclusively at the remnants of discs, clear dispersion of particulate ZO-1 to the cytoplasm was observed in isolated cells (Figure 5f, arrows inset). Presumably, a fraction of such ZO-1 particles were associated with internalized GJs, such as those seen in Figure 5c.
Next, we undertook quantitation of the colocalization between Cx43 and ZO-1 in the intact and dissociated ventricular preparations (Figure 6). This analysis was done using 2 approaches. In the first approach, we analyzed codistribution of Cx43 particles with ZO-1 in single optical sections of cells or tissues. Our presumption was that such particles corresponded to individual sarcolemma-localized or internalized (annular) GJs. Figure 6a (particles) shows that the degree of colocalization between Cx43 particles and ZO-1 particles increases significantly (P<0.001) following myocyte isolation. In the second approach (pixels, Figure 6a), we examined the pattern of colocalization on a pixel-by-pixel basis. The level of pixel-by-pixel colocalization increased over 3-fold (P<0.001) following dispersion—a similar increase to that observed in the particle colocalization analysis. Finally, we undertook a 0- to 60-minute time course examining colocalization between Cx43 and ZO-1 following isolation of myocytes (Figure 6b). Colocalization percentages of ZO-1 with Cx43 approximately doubled in the first minute after myocyte isolation and climbed to a peak of 60% at 30 minutes before falling back to 40% after an hour of culture of isolated myocytes. Each change in colocalization percentage over this time course was significantly different (P<0.01) from the time point preceding it. In contrast to the changing pattern of ZO-1-Cx43 association observed following myocyte dissociation, there was no significant variation in the area of Cx43 per cell over the time course examined (Figure 6b).
Figure 6. a, Cx43-ZO-1 colocalization analyses carried out pixel-by-pixel (PIXELS), or by Cx43 particle (PARTICLES), on optical sections of intact ventricular myocardium (intact) and isolated ventricular myocytes (isolated). Increases in ZO-1 colocalization with Cx43 measured within 60 minutes of dissociation, as determined by either quantitative method, were significant (P<0.001). b. On the left axis (squares), colocalization levels of ZO-1 with Cx43 (as determined by the pixel-by-pixel method on confocal optical sections) in intact ventricular myocardium (ie, the 0 time point) and isolated ventricular myocytes cultured for 1, 30, or 60 minutes after dissociation are plotted. **Difference between a given time point and the intact ventricular myocardial preparation that is significant at P<0.01. ##Difference between the 30- and 60-minute time points is significant at P<0.01 from the preceding time points. On the right x-axis (triangles), a corresponding time course is illustrated of Cx43 signal area (μm2) per cell area after myocyte isolation.
Coimmunoprecipitation Indicates an Increased Cx43 and ZO-1 Interaction and Potential Increases in Cx43 Phosphorylation After Myocyte Dissociation
Using anti-Cx43 antibodies, it was found that both Cx43 and ZO-1 could be coimmunoprecipitated, under nondenaturing conditions, from intact myocardium and freshly isolated myocytes (Figure 7a). Immunoprecipitations were performed using 2 anti-Cx43 antibodies against independent Cx43 epitopes. Western blots of ZO-1 and Cx43 immunoprecipitated by either of the anti-Cx43 antibodies used revealed that relatively more ZO-1 was consistently coimmunoprecipitated from isolated myocytes than intact myocardium. Densitometric analyses indicated that the relative level of ZO-1 (as a ratio of Cx43 levels in the sample) brought down by anti-Cx43 antibodies increased significantly (P<0.03) between intact and isolated preparations (Figure 7b). Addition of peptide to which anti-Cx43 antibody was raised abolished immunoprecipitation of both Cx43 and ZO-1, indicating that the immunoprecipitation of ZO-1 was due to the specific interaction between Cx43 in the nondenatured preparations and the anti-Cx43 antibody (Figure 7a). Taken together, these data are consistent with the confocal colocalization analyses and strongly support that a rise in direct interaction between Cx43 and ZO-1 occurs after disruption of intercellular contact between myocytes.
Figure 7. a, Western blots of Cx43 and ZO-1 in anti-Cx43 immunoprecipitates from intact myocardium (i. int) and isolated myocytes (ii. iso). Preincubation with peptide to which the Cx43 antibody was raised (Cx43 IP+PI) inhibited immunoprecipitation of Cx43 and ZO-1 from both intact (iii. int) and isolated (iv. iso) samples. b, Ratios of immunoprecipitated Cx43 to ZO-1 in intact or isolated cells. The difference between ratios is significant at P<0.03. c, Anti-Cx43 and anti-GAPDH Western blots of intact myocardium (int) and myocytes isolated for 10, 30, and 60 minutes.
The induction of endocytosis of Cx43-containing GJs in cultured cells has been reported to be associated with an increase in the phosphorylated isoforms of Cx43.28 We therefore undertook Western blotting with emphasis on increasing separation in the 40- to 50-kDa range of protein relative mobility. These blots revealed that after dispersion of cardiac myocytes, there was a shift toward increased levels of higher molecular mass isoforms of Cx43 at 10, 30, and 60 minutes, as compared with Cx43 isoforms in intact myocardium (Figure 7c). The change in relative mobility of Cx43 has been correlated with changes in phosphorylation status of the protein. Thus, the decrease in relative mobility of Cx43 in PAGE following myocyte isolation likely corresponds to an increase in phosphorylated Cx43 isoforms. The majority of Cx43 protein detectable after myocyte isolation is consistent with a more phosphorylated form of Cx43. A comparison between the lanes of the Western blot shown in Figure 7c indicates that there is no decrease in Cx43 protein in the post-isolation time course relative to immunoblotted GAPDH levels.
Discussion
Here, using high-resolution analyses, we show that ZO-1 and Cx43 have low to moderate point-by-point colocalization within intercalated discs in normal ventricular myocardium. This is in contrast to the precise and extended pattern of colocalization observed between ZO-1 and N-cadherin, a component of fascia adherens. Relative levels of association between Cx43 and ZO-1 increase conspicuously after dissociation of living ventricular myocytes from intact heart. Previous workers have characterized that this treatment, which results in the break-up of electrical and mechanical intercellular couplings, induces the formation of cytoplasmic gap junctional vesicles known as annular junctions.21 That this increase in ZO-1 immunolocalization at internalized GJs is accompanied by increases in direct interaction between Cx43 and ZO-1 at the protein-protein level is supported by coimmunoprecipitation analyses. Taken together, our findings are consistent with the growing evidence29,30⇓ that previously hypothesized functions for MAGUK proteins (eg, ZO-1, dlg, PSD-95 and PSD-93) in active localization of proteins to membrane subdomains and in ion channel aggregation may need reevaluation, or at least, expansion to consider other possible modalities. Indeed, although an involvement of ZO-1 in stabilization and function of GJ channel clusters at intercalated discs remains possible, our data are also consistent with ZO-1 partnering with Cx43 subsequent to the disruption of functional communicative contact between myocytes. A similar conclusion has recently been reported in 42GPA9 Sertoli cells treated with lindane.31
Previous reports have suggested that ZO-1 may be an important factor in the intercalated disc-targeting and function of Cx43-containing cardiac GJs.17,18,20⇓⇓ These conclusions were based largely on studies of cultured cells, and included coimmunoprecipitation, yeast 2-hybrid assays, transfection experiments, and low-resolution single immunolabelings of Cx43 and ZO-1. To date, high-resolution immunoconfocal or immuno-EM analyses of the pattern of association within intact myocardial preparations have not been undertaken. Together, with data from coimmunoprecipitation experiments, our detailed spatial analyses indicate that low to moderate levels of Cx43-ZO-1 association occurs in normal adult myocardium. The results shown here nonetheless are consistent with those of previous reports in that significant contact between these 2 proteins can occur in cells in vivo, albeit under specific circumstances; namely, ZO-1-Cx43 interactions appear to be particularly increased at former GJs, internalized within living ventricular myocytes.
Previous studies of isolated myocytes suggest that all gap junctional membrane is immediately internalized (ie, no gap junctional membrane remains exposed to the myocyte exterior) directly following myocyte dissociation.21 The majority (>70% in rabbit and guinea pig) of these junctions in freshly dissociated cells are annular gap junctions. This internalization occurs mainly at or just subjacent to sites at which former cell contact occurred, including near the remnants of intercalated discs.21 Over time courses, annular gap junctions did not undergo significant reduction in number (ie, did not break down) nor did they traffic more deeply into the myocyte interior from initial subsarcolemmal locations. There was no apparent evidence for repopulation of gap junctions to the cell surface. As such, all Cx43 immunolabeling shown in Figure 5 is internalized and probably mostly located within annular gap junctions.
In guinea pig, though 77% of junctions in isolated myocytes at time zero are annular gap junctions, the percentage rises to 91.5% after a few hours.21 This has been interpreted as evidence that in guinea pig, gap junction endocytosis continues following dissociation. The rat undergoes a similar progressive time course of gap junction endocytosis following dissociation of myocytes.32 It is thus possible that a progressive time course of gap junction endocytosis after cellular dissociation explains, or at correlates with, the cumulative time course of Cx43-ZO-1 colocalization that we show in Figure 6b.
Although a distinction should be drawn between processes occurring in vivo and those resulting from enforced cellular dissociation, annular GJs have been shown in a wide range of normal and diseased tissues. Intact myocardial tissues in which internalized GJ vesicles have been located include mammalian myocardium during early postnatal life33 and adult hearts affected by myocardial pathologies.34 Nonmyocardial tissues in which annular GJs have been observed include ovarian granulosa cells, bone, iris epithelium, oral epithelium, hepatocytes, and embryonal carcinoma cells.35 More recently, Laird et al36 used GFP-tagged Cx43 to provide direct evidence that annular GJ formation is a normal mechanism for removal and turnover of membrane channel aggregates. On an ultrastructural scale, this study reveals that the process is swift and dramatic. Within minutes, GJs are torn as extended bi-membranous structures from the plasma membrane and internalized intact within a neighboring cell. On the basis of these in vivo and in vitro studies, it is probable that rapid generation of cytoplasmic GJ vesicles is “part and parcel” of the remodeling of cell-cell contacts that occurs during normal developmental and disease processes. Finally, consistent with the low levels of ZO-1-Cx43 association we report in undisrupted ventricular muscle, it is noteworthy that annular GJs are rare in nondiseased myocardial tissues from adult mammals.33,34⇓
One reason that we initially focused on internalized GJs as a potential target for increased Cx43-ZO-1 interaction is that earlier workers had reported that elements of the actin cytoskeleton were closely associated with annular junctions.37,38⇓ As the C-terminal of ZO-1 is thought to interact with the actin cytoskeleton,16,17⇓ we reasoned that ZO-1 may also be localized to annular GJs. A logical extension then is to surmise that ZO-1 may be involved in the generation of these cytoplasmic membrane structures per se. Actin is known to directly participate in endocytosis processes,39 including specific involvement in GJ endocytosis.38 ZO-1, via its ability to simultaneously interact with Cx43 and cytoskeleton, may confer specificity to actin-based contractile processes. The tearing out and internalization of structures as large and potentially noncompliant as GJ channel clusters in toto presumably requires powerful, site-directed mechanisms of force generation.
A putative role for ZO-1 in GJ endocytosis re-raises the prospect for ZO-1 involvement in generation of myocyte coupling patterns in development and disease, although not necessarily just as a multivalent adapter, stabilizing the components of electrical and mechanical junctions within intercalated discs. We have formerly hypothesized that discs at myocyte termini may differentiate by mechanisms that include more rapid turnover of Cx43-containing GJs at lateral domains of ventricular myocytes.10 Interestingly, annular GJs are significantly more common in ventricular muscle during the postnatal period in which intercalated disc differentiation occurs.33 Whether this internalized GJ population in postnatal myocardium has associated ZO-1 remains to be explored. Other explanations of the increased association of Cx43 and ZO-1 after cell dissociation also deserve consideration. These alternates would include roles for ZO-1 in targeting of annular GJs to the lysosome and/or recycling of Cx43 from annular junctions back to functional GJs in the plasma membrane.
Phosphorylation of Cx43 represents an important mechanism for the regulation of GJ gating, assembly, trafficking, and degradation.40,41⇓ Cx43 in normal intact myocardium can be resolved into 2 distinct phosphorylated species on Western blotting, although as many as 5 species have been reported.42 We found that an increase in potential Cx43 phosphoisoforms, as discriminated by electrophoretic mobility shift, consistently accompanies myocyte isolation and the induction of GJ endocytosis. A similar decrease of the relative mobility of Cx43 on immunoblots has been previously reported following induction of GJ endocytosis in studies of cultured cell lines.28 Further evidence supporting our observation comes from reports that antibodies specific for the nonphosphorylated isoform of Cx43 immunolocalize at internalized annular GJs at only relatively low levels.42 In this respect, it is also important to note that no overall alteration in Cx43 levels occurs over the time course following myocyte dissociation, as assessed by both immunoconfocal quantitation and immunoblotting. The changes in Cx43-ZO-1 association and Cx43 relative mobility observed in this study are therefore unlikely to be the result of selective degradation of gap junctions not containing phosphorylated Cx43 and not associated with ZO-1.
In conclusion, this study establishes a number of observations and raises some equally important questions. Based on the results, it seems that direct interactions between ZO-1 and Cx43 are dynamic and capable of undergoing rapid change after disruption of functional contact between myocytes. This change occurs over a specific time frame in living cells, with levels of association quantifiably rising and falling within 60 minutes of myocyte dissociation. Finally, the increased Cx43-ZO-1 interaction we observe is correlated with an equally rapid change in the relative mobility (and possibly phosphorylation status) of Cx43. Whether phosphorylation of Cx43 is a prerequisite for increased ZO-1-Cx43 interaction remains to be determined. Of other remaining questions, perhaps the most important is whether ZO-1 directly participates in GJ endocytosis. Future experiments targeting Cx43-ZO-1 interaction using dominant negative approaches might enable discrimination of whether this interaction is necessary during or subsequent to the genesis of annular GJs. Resolution of this question may provide important new avenues for understanding the mechanistic basis of gap junctional remodeling in cardiac development and disease.
Acknowledgments
This work was supported by grants from the NHLBI (HL56728) and National Science Foundation (9734406). The excellent technical assistance of Jeff Davis and Jane Jourdan is acknowledged with gratitude. We thank Dr Cheng Gang for his suggestions and discussion.
Footnotes
Original received July 3, 2001; resubmission received October 24, 2001; revised resubmission received December 10, 2001; accepted December 19, 2001.
References
This Issue
Jump to
Article Tools
Share this Article
- Increased Association of ZO-1 With Connexin43 During Remodeling of Cardiac Gap JunctionsPermalink:
