INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

MYOCARDIUM is not a true electrical syncytium. Rather, current must be transferred from one discrete cell to another to electrically activate the heart. This process occurs at gap junctions, which are specialized sarcolemmal organelles composed of arrays of membrane-spanning channels that connect the cytoplasmic compartments of adjacent cells and permit intercellular passage of current (1, 20). Working ventricular myocytes are extensively interconnected by numerous gap junctions that are among the largest in any mammalian tissues (11, 20). It is likely, therefore, that the presence of numerous large gap junctions is critical to normal ventricular conduction.

Many observations suggest that the number, size, and spatial distribution of gap junctions play an important role in determining the conduction properties of different cardiac tissues. For example, ventricular myocytes are extensively interconnected to neighboring myocytes in both end-to-end and side-to-side orientations (17, 27), whereas atrial myocytes of the crista terminalis are interconnected mainly end to end, consistent with the much greater degree of anisotropy of conduction in the crista compared with the ventricle (33). The total length of gap junction profiles per unit myocyte area is ~25 times less in the sinus node than in ventricular myocytes (26). Reduced gap junction profile length and gap junction protein expression levels in diseased ventricular myocardium have been implicated in the pathogenesis of slow conduction and in unidirectional conduction block leading to reentrant arrhythmogenesis (17, 21, 22). Conversely, increased gap junction profile length and protein expression has been correlated with increased conduction velocity in cultured cells exposed to cAMP (4). These observations and others suggest that the number and size of gap junctions are important determinants of conduction properties in different parts of the heart under both physiological and pathophysiological conditions. Little is known, however, about how specific patterns of intercellular connections are established or how the number and size of gap junctions is determined in different tissues of the heart.

We recently characterized myocardial conduction in mice created by Reaume et al. (24) that are heterozygous for a null mutation in the gene encoding connexin43 (Cx43), the predominant ventricular gap junction channel protein. These mice (Cx43 +/) express ~50% of the wild-type level of Cx43 in their ventricles and have ~40% slower ventricular epicardial conduction (10, 35). Whole cell action potentials of ventricular myocytes from Cx43 +/ and wild-type (Cx43 +/+) hearts are indistinguishable (10) and there appears to be no marked differences in ventricular wall structure or fiber orientation (10). These observations suggest that slow ventricular conduction in Cx43 +/ mice is related directly to diminished Cx43 in gap junctions (i.e., fewer intercellular channels resulting in increased resistance to current transfer) rather than differences in other determinants of conduction velocity such as active membrane properties or the anatomy of the conduction pathway. However, there have been no reported studies of cardiac gap junction structure in Cx43 +/ mice.

In the present study, we analyzed Cx43 +/ mice to gain insight into the relationship between the amount of a connexin being expressed by a cardiac tissue and the structure of gap junctions in that tissue. We asked whether the ~50% reduction in Cx43 protein content in ventricular tissue would change the number and/or size of gap junctions interconnecting ventricular myocytes in Cx43 +/ mice. With the use of quantitative immunohistochemistry and electron microscopy, we found that the mean size of an individual gap junction is the same in Cx43 +/ and +/+ ventricular myocardium, but the number of gap junctions is significantly reduced in Cx43 +/ tissue. These results suggest that under conditions in which expression of gap junction channel proteins is diminished, ventricular myocytes remain interconnected by large gap junctions, which could be an adaptive response.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Tissue acquisition and processing for immunofluorescence and electron microscopy. The hearts of five adult Cx43 +/ mice and five adult Cx43 +/+ controls (18.5 ± 7 wk of age for both groups) were rapidly excised, rinsed in PBS, and fixed in 10% neutral-buffered Formalin in preparation for immunofluorescence microscopy. Each Formalin-fixed heart was cut into transverse slices ~1-mm thick, dehydrated in a series of alcohols, embedded in paraffin, sectioned at a thickness of 6 µm, and mounted on gelatin-coated slides. The hearts of an additional five adult Cx43 +/ mice and five adult Cx43 +/+ controls were rapidly excised, rinsed in PBS, and dissected with the aid of a stereo microscope to remove the major left ventricular papillary muscles of the mitral valve. The papillary muscles were fixed for 24 h in 2% glutaraldehyde in 0.1 mmol/l cacodylate buffer (pH 7.4) in preparation for electron microscopy. The fixed papillary muscles were postfixed in 1% OsO₄, dehydrated in ethanol, and embedded in Spurr's low-viscosity epoxy resin. After the resin had been polymerized, ultrathin (50-100 nm) sections were prepared in a plane parallel to the long axis of the papillary muscle to ensure that the ventricular myocytes of the papillary muscles were being cut in the longitudinal plane.

Immunohistochemistry. Formalin-fixed, paraffin-embedded sections of a mouse left ventricle were immunostained using immunochemical reagents and methods that have been validated in a previous study (15). Primary antibodies included a commercially available polyclonal rabbit anti-Cx43 antiserum (Zymed), which we have used previously to characterize expression of Cx43 in the mammalian heart (4, 12, 15). We also used a commercially available polyclonal rabbit antiserum directed against a synthetic peptide corresponding to the COOH-terminal amino acids of chicken N-cadherin (anti-pan-cadherin, Sigma). This immunochemical reagent reacts with members of the cadherin family that form adherens type junctions but does not recognize the desmosomal cadherins (8). The pan-cadherin polyclonal antibody specifically immunostains fascia adherens junctions but not desmosomes in intercalated disks in chicken, mouse, rat, guinea pig, rabbit, and human cardiac muscle (8, 14, 23). To eliminate interpreparative variability in the immunofluorescence data, sections from all hearts (5 of each genotype) were immunostained in a single batch using identical aliquots of diluted primary and secondary antibodies and other reagents.

Quantitative analysis of immunoreactive signal by confocal microscopy. Sections from each heart stained with anti-Cx43 or anti-pan-cadherin antibodies were examined by fluorescence microscopy at a magnification of ×400 using a ×40 oil immersion lens with numerical aperture of 1.0, lateral resolution of 0.23 µm, and depth resolution of 1.06 µm. Use of this lens in combination with a 50-µm pin hole resulted in a full-width, half-maximal focal plane with a thickness of ~1.0 µm. Areas selected for analysis consisted of well-preserved, compact bundles of left ventricular myocytes cut in a plane parallel to the long axis of the cells. Each individual test area encompassed ~26, 400 µm² and included profiles of ~30 to 50 cardiac myocytes. Five test areas were analyzed for each antibody in each of five Cx43 +/ hearts and five Cx43 +/+ hearts for a total of 25 test areas in each group. Each test area was scanned by the laser at the approximate middepth of the section. Each test area was digitized into a 1,024 × 1,024 matrix (~1.05 × 10⁶ pixels/test area) (Fig. 1). In general, the level of background fluorescence was low, and the immunoreactive material recognized by both anti-Cx43 and anti-pan-cadherin antibodies was concentrated in discrete spots at points of intercellular apposition. The low background and highly localized, intense signal facilitated identification of a signal intensity threshold exceeded only by the specific discrete spots of signal at appositional membranes. Accordingly, an arbitrary threshold was identified for each set of sections stained with a given primary antibody that clearly distinguished the high-intensity signal concentrated in discrete spots from all other areas in the test field (Fig. 1). Any structures with a signal intensity that exceeded this threshold were assumed to represent gap junctions or fascia adherens components of intercalated disks in sections stained with anti-Cx43 or anti-pan-cadherin antibodies, respectively. A much lower threshold value was also identified below which only areas of the slide not covered by cells were defined. Thus the proportion of total tissue area occupied by gap junctions or fascia adherens junctions was defined as the number of pixels with signal intensity exceeding the high threshold divided by the total number of pixels exceeding the low threshold. The percentage of total test area occupied by tissue was defined as the total number of pixels exceeding the low threshold divided by the total number of pixels in the image space (1,024 × 1,024 pixel array).

View larger version (70K): [in this window] [in a new window]   Fig. 1.   Quantitative analysis of connexin 43 (Cx43) expression and gap junction number and size by confocal microscopy and digital image processing. A: representative confocal image (1-µm-thick optical cut) from a Cx43 +/+ heart stained with anti-Cx43 antibody. Specific immunoreactive signal appears as white structures against the darker (lower intensity) background of myocytes cut in a longitudinal plane of section. B: same image as in A following subtraction of all pixels that did not achieve the signal threshold that distinguished "gap junctions" (clusters of pixels exhibiting intense immunoreactive signal) from other structures. C: histogram showing number of pixels at each signal intensity level (0-255 gray scale) in A. All tissue sections from Cx43 +/+ and +/ hearts were immunostained in a single batch to eliminate interpreparative variability (see text). A signal intensity threshold was arbitrarily identified that best distinguished clusters of intense signal from all other structures. This threshold value (150 on the 0-255 gray scale) was used in analysis of all confocal images from all Cx43 +/+ and +/ hearts. Total Cx43 expression was defined as aggregate number of pixels that achieved this threshold as a proportion of total number of pixels representing tissue area. In A total Cx43 expression was 0.75%, a value close to the mean expression level observed in Cx43 +/+ sections (0.79%). D: high-resolution image of a specific cluster of Cx43 immunoreactive signal identified with arrows in A and B. Full range of signal intensities of individual pixels can be seen at this magnification. E: same image as D following subtraction of all pixels that did not achieve the signal threshold that distinguished "gap junctions" from other structures. Image processing software identified 8 individual gap junctions ranging in size from 9 to 103 pixels. Because the minimum gap junction size was 5 contiguous pixels, the cluster of 3 pixels included in this image was not scored as a gap junction. F: histogram showing number of individual pixel clusters (i.e., number of "gap junctions") scored at each cluster sized expressed either in terms of the number of pixels per cluster or the cluster area in the image shown in A. Minimum cluster size is 5 pixels. The largest gap junction contained 271 pixels. Distribution of gap junction sizes is similar to the distribution of gap junction profile lengths reported in a previous quantitative ultrastructural study of canine left ventricular myocardium.

Ultrastructural morphometry of ventricular myocardial gap junctions and intercalated disks. Left ventricular papillary muscles were analyzed by electron microscopy because they are composed of myocytes arranged in bundles parallel to the long axis of the muscle, thus greatly improving the reproducibility of sectioning the samples in the true longitudinal axis. We assumed that the quantitative relationships between the number and size of gap junctions and intercalated disks in the papillary muscles of Cx43 +/ and +/+ mice were the same as in the larger areas of the left ventricular free walls analyzed by confocal microscopy. Ultrathin sections of left ventricular papillary muscles cut in a plane parallel to the long axis of the myocytes were analyzed according to methods used in previous studies (11, 17, 26). Areas were selected for analysis at low magnification (below the resolution of gap junctions) solely on the basis of technical considerations as follows. Test areas were composed of compact groups of myocytes cut in true longitudinal section and free of technical artifacts such as tears in the section. Myocytes adjacent to collagenous septa or perivascular regions of blood vessels larger than capillaries were excluded. Three test areas were analyzed for each of the five Cx43 +/ and +/+ hearts for a total of 15 test areas for each group. Each test area was photographed at a final print magnification of ×5,000. At this magnification, a typical test area included ~750 µm² of tissue area and contained portions of 3 to 5 myocytes. The amount of total section area occupied by myocytes and the total sarcolemmal length (both junctional and nonjunctional) were measured in each ×5,000 test area by computer-aided planimetry. In addition, every gap junction and intercalated disk profile identified within every test area was rephotographed at a final print magnification of ×30,000. The total length of gap junction and intercalated disk profiles and the lengths of individual gap junction profiles were measured in each intercellular junction region. Calculated morphometric parameters included the amount of sarcolemma and intercalated disk profile length per test area, the amount of intercalated disk profile length per 100-µm sarcolemmal length, the number of individual gap junction profiles per 100-µm sarcolemma or intercalated disk length, and the mean length of individual gap junction profiles.

Statistical analyses. All data are expressed as means ± SD. The statistical significance of differences between means for confocal immunofluorescence data were determined using a nested ANOVA. The analysis was carried out using the NESTED procedure in SAS (34). ANOVA was used to determine whether differences in gap junction number or size between Cx43 +/+ and Cx43 +/ samples varied as a function of the minimum "gap junction" size (number of contiguous high-signal intensity pixels). A nested ANOVA was used to test for significant differences between means in the ultrastructural morphometry data. Nested analyses of variance were used to account for the fact that multiple samples taken from the same heart are not fully independent of one another. In the nested ANOVA performed herein, the overall variance of the dependent variable was partitioned as the sum of the variances due to differences between mouse genotypes, between hearts of the same genotype, and between the five or three samples selected from each heart. Through a comparison of each component of variance with the appropriate error term, the nested ANOVA determined the significance of both differences between hearts and differences between genotypes (Cx43 +/+ and Cx43 +/ mice). In this study, our only interest was in evaluating differences between the two mouse genotypes, and we report only the P values that were associated with these tests. It would not have been appropriate to use a simple unpaired t-test to compare the two groups of mice because the correlation between samples from the same heart violates the independent assumption of a t-test. P < 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Confocal immunofluorescence analysis of gap junctions and fascia adherens junctions in Cx43 +/ and +/+ mice. As shown in representative images in Fig. 2, high-intensity specific immunoreactive signal for Cx43 or N-cadherin was clearly identified at points of intercellular apposition. Little or no signal was observed intracellularly. It was obvious from inspection of sections stained with anti-Cx43 antibodies that the amount of signal was reduced in the left ventricles of Cx43 +/ mice compared with Cx43 +/+ mice, whereas it appeared that the amount of signal was roughly equal in ventricular sections from each group stained with anti-pan-cadherin antibodies. Quantitative analysis of digitized confocal images confirmed these conclusions. As shown in Table 1, the proportion of total tissue area occupied by high-intensity immunoreactive Cx43 signal was reduced significantly in sections of Cx43 +/ ventricular myocardium (0.43 ± 0.13%) compared with Cx43 +/+ tissue (0.79 ± 0.24%, P = 0.02). The magnitude of this reduction, 46%, is close to the previously reported 50% reduction in total tissue content of Cx43 measured by immunoblotting of Cx43 +/ ventricular homogenates (10, 35).

View larger version (66K): [in this window] [in a new window]   Fig. 2.   Representative laser scanning confocal images of anti-Cx43 and anti-cadherin stained left ventricular myocardium from Cx43 +/+ and Cx43 +/ mice. Thin line of apparent signal surrounding some cells in sections stained with anti-cadherin antibody is nonspecific and was of much lower intensity than discrete patches of specific signal. This nonspecific signal did not achieve the signal intensity threshold to be included in confocal analysis of number and size of fascia adherence junctions. Bar = 25 µm.

View this table: [in this window] [in a new window]   Table 1.   Quantitative confocal immunofluorescence analysis of Cx43 and N-cadherin signal in Cx43 +/+ and Cx43 +/ left ventricle

Electron microscopic morphometry of gap junctions and intercalated disks in Cx43 +/ and +/+ left ventricular papillary muscles. Immunofluorescence microscopy allows only indirect visualization of gap junctions or fascia adherens junctions and cannot resolve individual junctions with certainty. Thus quantitative ultrastuctural analysis was performed to confirm the results of confocal immunofluorescence microscopy. Although the amount of tissue that can be analyzed practically by electron microscopy is far less than with confocal microscopy, the advantage of the ultrastructural approach is that definitively identified gap junctions and intercalated disks can be measured with precision. Figure 3 illustrates representative individual images that reflect the general findings of the quantitative measurements, which are reported in Table 3. All sampling parameters were the same in Cx43 +/ versus +/+ samples. Specifically, the proportion of total tissue area occupied by cardiac myocytes in papillary muscle samples was 92 ± 5% and 90 ± 7% (P = 0.40) in Cx43 +/ and +/+ samples, respectively. This value is comparable to previously reported measurements in canine myocardium (17). Similarly, an equivalent amount of total myocyte surface membrane (junctional and nonjunctional sarcolemma) was analyzed in the two groups (145 ± 32 vs. 163 ± 36 µm of cell surface membrane length per test area in Cx43 +/ vs. +/+ samples, respectively, P = 0.13). Furthermore, the total length of ultrastructurally identified intercalated disk profiles (which encompassed all intercalated disk components including fascia adherens junctions, desmosomes, gap junctions, and undifferentiated membrane regions) was similar in Cx43 +/ and +/+ test areas (24 ± 9 vs. 26 ± 12 µm, respectively, P = 0.64) and, when expressed as a proportion of total sarcolemmal length, intercalated disk length was nearly identical in both groups (17 ± 8 vs. 16 ± 7% in Cx43 +/ and +/+ samples, respectively, P = 0.83). The total amount of gap junction length per test area or per 100-µm sarcolemmal length was, however, reduced by ~35% in Cx43 +/ versus +/+ samples (Table 3). This reduction in aggregate gap junction length was due entirely to a reduction in the number of individual gap junctions. The mean number of gap junctions/100-µm intercalated disk length, the total number of gap junctions per test area, and the total number of gap junctions in all test areas were reduced by ~35-40% in Cx43 +/ versus +/+ samples (6.6 ± 4.8 vs. 11.2 ± 6.9 gap junctions/100-µm intercalated disk length, P < 0.05; 1.7 ± 1.4 vs. 2.7 ± 1.5 gap junctions per test area; and 26 vs. 40 total gap junctions in all test areas, respectively). In contrast, the mean size of an individual gap junction profile was identical in Cx43 +/ and +/+ ventricular papillary muscle myocytes (0.37 µm for both groups, P = 0.98). The data presented in Table 3 should be interpreted with the recognition that because four different interdependent variables were tested, multiple comparisons could have resulted in P values that were significant by chance. In addition, the P value of 0.03 for total number of gap junctions per 100-µm intercalated disk length is of borderline significance. Therefore, although the ultrastructural data suggest a significant reduction in the number of gap junctions (but not mean size) in Cx43 +/ hearts, the statistical significance of the difference is equivocal when considered in isolation. However, when these data are interpreted in the context of the immunofluorescence data, we conclude that the results obtained by electron microscopy confirm the independent findings obtained by confocal analysis, demonstrating a reduction in the number rather than the size of gap junctions in Cx43 +/ hearts.

View larger version (125K): [in this window] [in a new window]   Fig. 3.   Representative transmission electron micrographs of intercellular junctions in left ventricular papillary muscle from Cx43 +/+ and Cx43 +/ mice. Arrows bracket profiles of individual gap junctions. Bar = 0.5 µm.

View this table: [in this window] [in a new window]   Table 3.   Ultrastructural morphometry of intercellular junctions in left ventricular papillary muscle of Cx43 +/+ and Cx43 +/ mice

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Since the development of antibodies that recognize cardiac connexins (2), numerous immunohistochemical studies have defined gap junction distributions in functionally disparate conductive tissues of the heart (3, 5, 9, 16-19, 21, 22, 27, 29). Cardiac tissues that conduct impulses rapidly, such as working atrial and ventricular myocardium and the His-Purkinje fibers of the ventricular conduction system, have numerous, large gap junctions (5, 9, 18). In contrast, tissues that conduct slowly, such as the sinus and atrioventricular nodes, contain myocytes that are interconnected by considerably fewer and smaller junctions (5, 18, 19, 26). Side-to-side intercellular connections are far more extensive in ventricular myocardium than in the crista terminalis in which conduction velocity is much more anisotropic (27). Changes in patterns of electrical coupling have also been linked to pathophysiologically relevant changes in conduction. For example, rearrangements in the spatial distribution of connections between myocytes in atrial and ventricular myocardium have been described in association with aging or ischemic injury (17, 21, 22, 30). Similar structural changes have been observed in the hearts of patients with diseases commonly associated with atrial or ventricular arrhythmias (6, 7, 22, 29). All of these observations suggest that the number, size, and three-dimensional distribution of gap junctions are important determinants of the distinct conduction properties of different tissues of the heart. However, virtually nothing is known about the establishment or regulation of specific patterns of intercellular junctions.

One of the limitations in efforts to define relationships between gap junction structure and impulse propagation is that experimental interventions that alter connexin expression or gap junction distribution may affect multiple determinants of conduction. The advantage of the Cx43 +/ mouse is that altered ventricular conduction appears to be due solely to diminished expression of Cx43, and thus it provides an opportunity to directly analyze gap junction structure-function relationships. Although junctional conductance values recorded directly from Cx43 +/ cell pairs have not been reported, current evidence suggests that changes in connexin expression levels that occur without accompanying changes in active membrane properties or tissue structure do have functional significance and cause a change in conduction velocity (4, 10, 35). Working ventricular myocytes express another gap junction protein, connexin45 (Cx45), but recent evidence indicates that they express Cx45 in much lower quantities than Cx43 (3). The other cardiac gap junction protein, connexin40, is not expressed in working ventricular myocytes (5, 9, 24, 27, 35). The contribution of Cx45 to gap junction size in ventricular myocyte is unknown. These two proteins colocalize in the same junctions (12) and appear to be interspersed rather than segregated into subregions within an individual junction. Thus the effects of the presence of Cx45 on the size of a gap junction may be apparent by both immunofluorescence (with anti-Cx43 antibodies) and electron microscopy. However, because the amount of Cx45 appears to be much less than Cx43 in ventricular myocytes (3), and Cx45 content is the same in Cx43 +/ and +/+ ventricles (10), we have assumed that the contribution of Cx45 to overall gap junction size is small and is not different in Cx43 +/ and +/+ ventricular myocytes.

The results of the present study indicate that diminished expression of Cx43 in Cx43 +/ mice is associated with a significant reduction in the number of gap junctions that interconnect ventricular myocytes but no decrease in the average size of these gap junctions. The size and number of fascia adherens junctions and the total amount of intercalated disk membrane length are unchanged, suggesting that the extent and distribution of sites of mechanical coupling are not altered by the diminished electrical coupling in Cx43 +/ mice. These conclusions are based on quantitative digital image processing of Cx43 immunoreactive signal and confirmed by direct measurements of gap junctions and intercalated disk profiles by electron microscopy. Despite the scatter in the ultrastructural data, the results are concordant with and validate the immunofluorescence data.

One interpretation of these data is that when faced with a reduced connexin level, it is functionally more advantageous for ventricular myocytes to maintain normal gap junction size rather than number. Results of recent experimental studied by Rohr et al. (25) and theoretical studies by Shaw and Rudy (28) have demonstrated the importance of gap junctional coupling in action potential propagation and arrhythmogenesis, especially in the setting of slow conduction due to diminished coupling. Analysis of the effects on conduction of specific patterns of remodeling of intercellular junctions, including one in which gap junction size is maintained but gap junction number is reduced, must await studies in models that take into consideration the actual distribution of gap junctions in cells and tissues. Spach and Dolber (31) and Spach and Heidlage (32) have developed two-dimensional models that incorporate the geometrical arrangement of gap junctions on a cellular scale. Simulations of conduction in models of this type under conditions in which gap junction number or size are varied could provide insights into the functional consequences of the specific pattern of remodeling observed in ventricular myocytes of Cx43 +/ mice.

One way that neighboring myocytes might maintain large gap junction channel arrays is to create localized regions of the sarcolemma that are conducive to formation of stable gap junctions. Because the packing of connexins in the lipid bilayer is so concentrated, regions of the cell membrane containing gap junctions are stiff and liable to disrupt the membrane when subjected to shear stress (20). This probably accounts for the intimate juxtapositioning of gap junctions and mechanical junctions in cardiac myocyte membranes. Fascia adherens junctions and desmosomes may mechanically stabilize certain regions of the membranes of interconnected cells and, thereby, create a local environment of low shear stress that favors formation and maintenance of large channel arrays. Because ventricular myocytes of Cx43 +/ mice do not differ from their normal counterparts in terms of the number and size of fascia adherens junctions or the amount of intercalated disk membrane, the diminished connexin pool in Cx43 +/ myocytes may assemble into normally large channel arrays at these sites.

Of particular relevance to the question of the spatial relationship between gap junctions and mechanical junctions is a study of human myocardium by Peters et al. (23) who observed that the gap junctions and fascia adherens junctions of neonatal and infantile ventricular myocytes share only loose spatial proximity and are distributed as tiny, punctate spots of Cx43 and N-cadherin immunoreactive signal over the entire surface of the cell. With increasing age, intercalated disks enlarge, fascia adherens junctions accumulate near the ends of the myocytes, and gap junctions and fascia adherens junctions become more closely juxtaposed until the adult pattern is achieved by ~6 years of age. The mechanisms underlying this maturation process are unknown, but these observations are consistent with the hypothesis that the size and location of ventricular myocyte gap junctions are determined by the extent to which mechanical junctions stabilize membrane regions that favor gap junction formation.

Changes in gap junction size have been described in chronic ischemic heart disease. In an ultrastructural study of canine left ventricular myocytes bordering healed infarcts, we observed a marked, highly selective reduction in the largest gap junctions (17). Recently, Kaprielian et al. (13) described downregulation of Cx43 immunoreactive signal in hibernating myocardium in patients with chronic left ventricular dysfunction. They observed a loss of the larger gap junctions normally seen at the ends of cells, suggesting that remodeling of ventricular membrane structure in the setting of chronic ischemic heart disease causes selective loss of large gap junctions. Future studies in which defined mechanical loads are placed on multicellular preparations of myocytes and concomitant changes in mechanical and electrical junctions are characterized may help identify mechanisms regulating the number and size of intercellular junctions in the heart.