Gap junction distribution in adult mammalian myocardium revealed by an anti-peptide antibody and laser scanning confocal microscopy

Gap junctions are aggregates of intercellular channels that in the heart facilitate the transmission of electrical current between myocytes (for a review, see Page and Manjunath, 1986). By means of these channels, low-resistance pathways are formed, linking all the myocytes within the myocardium (Weidmann, 1952, 1966, 1970; Barr et al. 1965; Sommer and Johnson, 1979; De Mello, 1982; Page and Manjunath, 1986). The electrical impulses borne rapidly and repeatedly along these pathways stimulate the contraction of individual myocytes in synchrony, which in turn provides the motive force for the pumping action of the heart.

In the ventricle of the mammalian heart, the principal sites of cell-to-cell plasma membrane interaction occur at the intercalated disk. Models for the organisation and distribution of gap junctions within the intercalated disk have been derived almost exclusively from electron-microscopic studies (McNutt, 1970; Page and McCallister, 1973; Forbes and Sperelakis, 1985; Severs, 1985, 1990). From longitudinal thin sections of myocytes, the intercalated disk is commonly envisaged as a series of membrane steps following the ends and sides of the myofibril terminations of neighbouring cells. Gap-junctional membrane is located predominantly in the lateral-facing sides of the disk steps, although small gap junctions also occur in the transverse regions abutting the myofibril ends, in association with fascia adherens junctions (Hoyt et al. 1989). The fasciae adherentes (also known as intermediate junctions or sheet desmosomes) link the myofibrils to the plasma membrane and fasten adjacent myocytes together.

Such an idealised picture of intercalated disk structure embodies features that are consistently observed, yet important details of disk architecture and gap junction organisation have remained unsettled (see, for example, Hoyt et al. 1989; Severs, 1989b). One reason for this is that electron microscopy normally permits the sampling of only small portions of the disk and hence is restricted in its ability to provide comprehensive overviews of structural detail over the intercalated disk as a whole. Sectional views of disks can often reveal intricacies of geometry not encompassed in current models, but the significance of these is often difficult to evaluate. However, such features of geometry may point to the existence of patterns of organisation, other than the idealised one, that have not been identified. The advent of specific antibody probes against cardiac gap-junctional protein has recently permitted the imaging of heart gap junctions by light microscopy (Beyer et al. 1989; Yancey et al. 1989; Luke et al. 1989; Laird and Revel, 1990). This has opened up the possibility of studying gap junction distribution and intercalated disk organisation over relatively large areas of cardiac tissue compared to the spatially limited (though higher magnification) views afforded by ultrastructural observation. We have recently raised and characterised three site-specific anti-peptide antibodies to a 43K (K=10³M_r) rat cardiac gap-junctional protein (Harfot et al. 1990; Gourdie et al. 1990a). Here we apply one of these antibody probes, in combination with laser scanning confocal microscopy, to compare and quantify gap junction organisation in relation to intercalated disk structure in a range of mammalian species, including man. The advantage of this approach over standard immuno-light microscopy is that laser scanning confocal microscopy permits optical sectioning through substantial depths of tissue (White et al. 1987; Shotton, 1989), enabling the reconstruction in three-dimensions of the disposition of the entire population of immunostained gap junctions within individual intercalated disks.

Peptides matching portions of the amino acid sequence of a 43K rat cardiac gap junction protein predicted from the nucleotide sequence of a cDNA clone (Beyer et al. 1987) were synthesised according to the method of Houghten (1985). These peptides were kindly provided by Dr N. B. Gilula (Research Institute of the Scripps Clinic, La Jolla, California) and were made to match residues 101-112, 131-142 and 237-248, segments of the 382-amino acid molecule deduced to be exposed on the cytoplasmic surface of the membrane (Beyer et al. 1987; Gourdie et al. 1990a). The peptides were coupled to keyhole limpet haemocyanin using glutaraldehyde, as described in detail elsewhere (Harfst et al. 1990). Antisera were raised in Sandy half-lop rabbits (two per antigen) using an initial injection of 0·4 mg of peptide linked to haemocyanin, mixed with an equal volume of Freund’s complete a<adjuvant (total volume, 2 cm³). The injections were subcutaneous, made at a series of sites along the animals’ backs. Each animal was boosted with injections of 0·2 mg of peptide linked to haemocyanin in Freund’s incomplete adjuvant at intervals that were determined by screening results. The animals were bled from ear arteries and veins after administration of 0·5 mg Hypnorm, and serum was prepared, aliquoted and frozen at -80 °C. The first bleed was taken before injection of the antigen to provide pre-immune serum, the second was at 2 weeks after the initial injections, and thereafter the animals were bled weekly (or occasionally at slightly longer intervals, e.g. immediately after boosts). Screening of sera for reactivity was carried out by dot blot assay (modified from Hawkes et al. 1982), and specificity of the antisera was confirmed by dot blots using the individual peptides, and in Western blots of isolated cardiac plasma membrane and gap junction preparations (Harfst et al. 1990). Western blotting experiments showed that unpurified antisera taken from periods of peak antibody production (as determined from dot blot screening) detected the same protein bands as did affinity-purified antibodies (Harfst et al. 1990). For immunohistochemical studies, unpurified antiserum to the peptide matching residues 131-142, designated ‘HJ’, was found to be highly effective, and so this antiserum was used throughout the present study.

Immunolocalisation was carried out on sections of wax-embedded or frozen myocardium from the following species: rat (Sprague-Dawley, ∼250 g), guinea pig (DH, ∼400g), golden hamster (∼90g), mouse (Balb/c —20g), rabbit (New Zealand White, —2·5 kg), and man. All laboratory animals were maintained on standard chow diets and tap water, available ad libitum. Human material came from samples of left and right ventricle free wall (mid-region) of the explanted hearts of patients undergoing heart transplantation. Two patients (one 60-year-old male and one 53-year-old female) were suffering from severe ischaemic heart disease; one patient (a 36-year-old male) had dilated cardiomyopathy. In all, left ventricular function was significantly impaired but right ventricular function was near normal.

For wax-embedding, samples of human heart tissue and whole intact hearts from the other five species were fixed for 2-6 h in Zamboni’s fixative (2% paraformaldehyde, 0·2% picric acid in 0·1 M phosphate-buffered saline, pH 7·0; Toshimori et al. 1987). The excised tissue samples from human hearts were from defined anatomical locations to enable reliable comparison. After fixation, the specimens were washed in tap water, dehydrated in ethanol, placed in chloroform and embedded in wax following standard histological procedure. Whole hearts were oriented to permit longitudinal sectioning in a mid-plane through the whole organ so as to reveal the four cardiac chambers with their surrounding myocardium (left and right ventricles and atria). These four-chambered views were taken at a sectional plane that revealed the crest of the interventricular septum and the subjacent bundle of His. Sections of — 10 μm thickness were cut, dewaxed in xylene, rehydrated in a graded ethanol series and rinsed in tap water. Incubation in buffered trypsin (0·1 % trypsin (Sigma), 0·1% CaCl₂, 20mM Trizma base, pH7·4) then followed for 10 min at room temperature. This step is a standard procedure for unmasking epitopes that have been altered by aldehyde fixation (Harlow and Lane, 1988). The sections were washed with tap water and then blocked using 0·1M L-lysine in phosphate-buffered saline containing 0·1% Triton X-100 (as a wetting agent). Incubation with the primary antiserum, anti-HJ, was then carried out at a 1:100 dilution in the phosphate-buffered saline overnight at room temperature. After washing with phosphate buffer, secondary antibody treatment with swine antirabbit conjugated to fluorescein isothiocyanate (Dako; 1:20 dilution) was given for 1 h in the dark at room temperature. After the final washing in phosphate-buffered saline, the slides were mounted using Citifluor mounting medium (a medium that reduces fading and autofluorescence). The following controls were routinely done in each staining run; (1) omission of primary antiserum, (2) use of preimmune serum in place of primary antiserum.

For cryosectioning, 25 μm thick sections were prepared from excised blocks of unfixed frozen tissue from anatomically defined locations. The sections were fixed by immersion in methanol at —2O°C for 3min and then treated with 0·1M L-lysine and taken through the same series of immunolabelling steps as those described for dewaxed sections. Treatment with primary antiserum was for lh using a 1:10 dilution. Controls using preimmune serum, and omission of the primary antiserum were carried out in parallel.

Immunolabelled sections were examined using a Bio-Rad Lasersharp MRC-500 laser scanning confocal microscope. Gain and contrast levels were set according to procedures standardised to ensure that the image collected demonstrated a full range of grey level values from black (0 pixel intensity level) to peak white (255 pixel intensity level). All sections were initially surveyed at low magnification so that overall patterns of distribution of immunofluorescence could be compared qualitatively in cells oriented at different angles with respect to the section plane. Detailed comparisons were carried out at both high and low magnification. Confocal sections were either viewed individually, or series of optical sections (taken at 1-μm focus steps in the optical axis (z-axis) of the microscope) were digitally combined into z-axis projections or z-axis red/green anaglyphs using the laser scanning confocal microscope’s SOM software ‘maximum projection’. The SOM software’s “Thru-View’ sub-menu option was used to generate projections around the y-axis of optical sections stacks, producing images of rotated intercalated disks.

A detailed analysis of immunoβtained gap junction size and numerical density per disk was carried out on wax-sections from the six mammalian species. Individual fluorescent spots were assumed to correspond to discrete junctions and the longest axis dimension of the spot was taken as a measurement of gap junction size (for details on validity and limitations of this approach, see Discussion).

In the laboratory animal species, images of intercalated disks were recorded from immunoβtained sections of whole hearts prepared as described above. This was done in a defined zone - the central region of the left ventricular free wall, midway between the epicardial and endocardial surfaces. Immunostained gap junction size was also analysed in a corresponding location within rat left atrial free wall. In the human material, a detailed qualitative comparison of histologically unaffected or minimally altered right and left ventricular myocardium was first made, avoiding visible abnormalities (e.g. healed infarcts in the ischaemic hearts). Histopathological assessment was done on haematoxylin and eosin-stained wax sections. The principal criteria for abnormality were disarray of myocyte orientation, reduction of myofibrillar content, vacuolation of cytoplasm, attenuated cell shape and nuclear enlargement. Myocardial areas showing such changes, and those lying alongside were excluded from the study. On the basis of clinical and histopathological assessment, the histologically normal right ventricular myocardium (which was least affected by disease) was selected for data collection.

Intercalated disks with their full complement of gap junctions were reconstructed for quantitative analysis as follows. An intercalated disk, in en-face or near en-face aspect and contained entirely within the thickness of the tissue section, was selected using standard light-microscope epifluorescence. The microscope was then swapped to confocal mode and the lower z-limit (as indicated by the disappearance of punctate immunolabelling) of the disk was determined by moving focus up and down manually. The focus motor was then engaged and the upper limit of punctate immunolabelling was determined by quickly moving up through the disk in 1-μm steps. Once the upper limit was located, the whole immunolabelled disk was optically sectioned in 1-μm steps. Optical sectioning was performed using a 60 × objective (Nikon, PlanApo, 1·4 NA), Kalman averaging and a zoom 4 setting on the computer. During recording, the confocal aperture was set at a position between 30 % and 35 % of its full adjustable range. This was done in order to optimise the efficiency of light collection whilst maintaining sub-micrometre depth of field within optical sections. Disk recording proceeded immediately after immunostaining.

The design of the analysis was as follows. From each of the six species, hearts of three individuals were used. For each rat, guinea pig, mouse, hamster and human heart, 15 intercalated disks were recorded within the defined ventricular zone, giving a total of 45 disks per species. The subsets of 15 disks were subdivided into three groups (five disks each) according to disk size. The three size classes of disks were: (1) 0-49 μm² area, (2) 50-100 μm² and (3) greater than 100 μm². For the rabbits, disks were recorded only within size classes (1) and (2), as after extensive survey a total of only four intercalated disks larger than 100 μm² were found. For the recording of fluorescent spots the disk was divided into two zones, peripheral and central (details below). In summary, then, the study design was: 6 species ×3 individuals×3 disk sizes×5 optically sectioned disks×2 disk zones. A total of 255 disks, containing 13 898 fluorescent spots were used in the analysis of variance. For rat atrium, the same four-chamber heart sections were used. In sections from each of the three rats, five randomly selected fields were optically sectioned in five 1-μm steps within the defined atrial tissue zone (i.e. 3 rats×5 optically sectioned fields). A total of 1447 fluorescent spots was measured in the atrium.

Image analysis and image processing were performed on z-axis ‘maximum pixel’ intensity projections of optically sectioned disks and rat left atrial free wall fields using the PC-IMAGE (Paul Smith, Foster Findlay Associates Ltd) image analysis package. Initially a 3×3 median convolving filter was passed over the whole image to remove background noise. The outline of each intercalated disk, as marked by peripherally located immunostained gap junctions, was traced using a mouse. A second region was traced within each disk that excluded these peripherally located junctions. Immunostained gap junctions within this region are referred to as centrally located. Immunostained junctions external to this region, but within the mouse-traced outer border of the intercalated disk, are referred to as peripheral junctions. The number and size of immunostained gap junctions within each disk region were measured as follows. A pixel intensity threshold was adjusted such that the bright, fluorescent-labelled gap junctions were demarcated by an overlying colour binary image. The correspondence between the particles in the two images was checked by toggling back and forth between the full (256 grey levels) grey scale image and the binary image. Once correspondence was achieved, the labelled disks were inspected at high zoom to ensure that clearly separate juxtaposed junctions were not contiguous. If such artefactual merging occurred between proximate junctions, the binary image was edited to separate them. Pixels occurring singly in the binary image were rejected as noise. At the magnification used for data collection and display on the monitor, a single pixel equated to 60 nm; hence, the smallest immunostained junctions measured by the technique were 90-120 nm in size.

Calibrated automatic measurement of intercalated disk area and the number and long axis length of individually labelled junctions in the colour binary image was then performed and the data were stored to disk. In some rare cases, elongate junctions demonstrated marked curvature. For the purpose of long axis measurement, these junctions were ‘straightened’ by using the PC-IMAGE ‘Binary Edit’ option. The areas of intercalated disks, as measured by the areas circumscribed by the mouse-traced outlines, should be regarded as comparative indices of the disk areas, rather than absolute values, in view of the highly undulating topology of the disk demonstrated ultrastructurally. The numerical density of fluorescent spots was calculated as the number of spots within a ventricular intercalated disk divided by the total visible area of the disk traced; hence these comparative data are likely to overestimate the true numerical density of gap junctions.

Analyses of variance (ANOVA) of immunostained gap junction length and density were performed using Minitab computer software (Minitab Corp.). ANOVAs of the lengths were performed on natural logarithmically (logβ) transformed data. This was done as the length values demonstrated marked deviation from normal distributions (see Figs 12 and 13, below). Following transformation, a Minitab “Nscore’ test of the logo-transformed data revealed it to be normally distributed P<0·01). Variation in length and numerical density between individuals within species was not significant and hence ANOVAs were performed on data pooled within a species. The means given in Table 1 are the natural antilogarithms of the mean values produced by the ANOVAs of logo-transformed data. Immunostained gap junction numerical density within intercalated disks was normally distributed P0·01).

Standard phase-contrast microscopy was earned out on an Olbrich 4000 prototype research microscope, equipped with Nikon optics and a customised stand. An important feature of the instrument is the lighting source at the base, which allows variation of fighting conditions. This facility was exploited to combine epifluorescent imaging with oblique back-lighting of immunostained tissue sections (a method developed by Dr Gary Greenberg, USC School of Dentistry, Los Angeles) to give a comparison of the results obtained by laser scanning confocal microscopy with optimised conventional light microscopy. Oblique lighting gives a three-dimensional impression by differential shading of the structures in the section. Superimposition of immunofluorescently labelled gap junctions upon the cellular detail revealed by oblique back-lighting of myocardial tissue is achieved photographically by double-exposure.

Conventional thin-section electron microscopy was carried out according to routine procedures (Severs et al. 1985).

In ventricular tissue from mammalian myocardium, gap junctions are characteristically located at the intercalated disks adjoining neighbouring myocytes (McNutt, 1970; Page and McCallister, 1973; Forbes and Sperelakis, 1985; Severs, 1985, 1990). To explain the interpretation of the immunostaining patterns obtained, the basic features of intercalated disk and gap junction organisation, as seen by standard light microscopy and thin-section electron microscopy, are illustrated in Figs 1 and 2. In longitudinally sectioned cells viewed by phase-contrast light microscopy, the intercalated disks are seen as white lines traversing myocytes at right angles to, or diagonally across, the long axes of the cells. By thin-section electron microscopy, the characteristic stepped structure of the intercalated disk is resolved, with gap junctions located predominantly on the membrane segments lying parallel to the cell’s long axis, perpendicular to the conspicuous darkly stained fasciae adherentes that link the myofibril ends of apposing cells.

In longitudinally sectioned rat ventricular myocytes examined by laser scanning confocal microscopy after immunostaining with anti-HJ antiserum, sharply defined punctate lines of fluorescence crossing the short axes of the myocytes are observed (Fig. 3). The general pattern of immunostaining in survey view (20 × objective, NA 0·5) matches precisely the configuration of intercalated disks seen by the standard techniques of phase-contrast light microscopy and thin-section electron microscopy (Figs 1 and 2). Immunostained sections examined by conventional epifluoresence light microscopy show immunostaining patterns corresponding to those observed by laser scanning confocal microscopy but with the intercalated disks delineated as blurred rather than punctate lines (Fig. 9, below). Controls in which the anti-HJ antiserum treatment was omitted or substituted with pre-immune serum showed no fluorescent labelling.

Immunostained sections of rat ventricular tissue examined by laser scanning confocal microscopy show no evidence for the presence of gap junctions external to recognisable intercalated disks. This is in contrast to the pattern of localisation observed in longitudinally sectioned rat atrial myocytes (Fig. 4), where extensive punctate staining along the lateral surfaces of adjacent myocytes is common. Some atrial myocytes demonstrate punctate staining along the entire length of their side-byside contacts. Chains of spots do also occur in intercalated disk-like structures, but these are smaller and less well defined than those of ventricular tissue.

Fig. 5 compares, at higher magnification, the patterns of gap junction immunostaining in longitudinally oriented ventricular myocytes in rat, guinea pig, rabbit, mouse, hamster and man, as viewed by laser scanning confocal microscopy. The general features of gap junction staining from these mammals is similar, in that it is consistently confined within clearly delineated intercalated disks. At the higher magnification afforded by the 60 × objective (NA 1.4), the ability to resolve the individual spots within the disks is markedly improved. The presence of a faint background fluorescence enables the positions of the cells to be identified so that the transverse widths of the disks can be seen in relation to the cell bodies. Staggered lines of punctate fluorescence often extend across the full widths of abutting cells; shorter lines are situated at bifurcations. In human, the linear arrays of spots tend to merge into narrow bands of fluorescence that are often terminated by distinct large spots at either end. Rabbit intercalated disks are noticeably less uniform in organisation than those of the other species, with some punctate fluorescence observed in zones of lateral interaction between cells, usually in association with recognisable disks. The rabbit also shows a higher general level of background fluorescence, attributed to non-specific binding of the secondary (anti-rabbit) antibodies used in the staining procedure.

Further insights into gap junction organisation at the disk are obtained by viewing myocytes cut in a variety of section planes and by exploiting the ability of the confocal microscope to provide projections of optical section series. An optical section series taken in six 1-μm steps through obliquely oriented hamster ventricular myocytes is illustrated in Fig. 6, and a z-axis projection of the series, in which the optical sections are projected into a composite image spanning the full 6 μm depth of field, is shown in Fig. 7. Viewing the immunostaining in this section plane permits sharper visualisation of individual gap junctions than can be achieved in longitudinally sectioned cells, and disks of various shapes, sizes and orientations are clearly delineated. The ability of the confocal microscope to sample information principally in the plane of focus is well illustrated in this series. Though the punctate fluorescence from most individual gap junctions can be traced through consecutive pairs of sections, the overall pattern of punctate staining progressively changes through the series as new immunostained junctions come into view, and those formerly viewed disappear. An example is situated just above the centre of the field in Fig. 6, where the serial views combine to reveal a ring of triangularly distributed fluorescent spots, outlining a disk in near face-on aspect, when viewed in Fig. 7. The densest accumulation of spots occurs at the edge of the disk. These peripherally distributed spots are larger than the less frequent, albeit well-stained spots seen within the disk.

Below the ‘triangular’ disk, interconnected myocytes are observed in oblique orientation. Here, the myocytes are organised in linear series. The main boundary between cells at their ends is marked by dense ‘collar-like’ distributions of fluorescent spots around the lateral edges of disks. Some fluorescent spots are observed within the interior zone ofdisks, though the frequency of these varies. However, despite the foreshortening effect of the viewing angle on obliquely oriented disks, a more prominent accumulation of large fluorescent spots appears to be present at the disk periphery.

That this pattern of immunostaining represents a consistent feature of gap junction organisation at the disk is clearly demonstrated in projections of disks viewed directly face-on at higher magnification (Fig. 8). In all species examined, the periphery of the disk is typically outlined by large fluorescent spots, often elongated in shape, which enclose numerous smaller spots within the central region. Though this feature of the disk, newly revealed by laser scanning confocal microscopy, is common to all mammalian species examined, interspecies variations in its prominence are nevertheless apparent. The pattern is most marked in man and rat, where large elongate peripherally located immunostained gap junctions often show a pronounced alignment with the outer border of the disk (Fig. 8). In the other species there is lesser tendency for such alignment, and relatively fewer large junctions mark the disk periphery.

Perception of three-dimensional organisation is further enhanced by viewing red/green anaglyphs of immunostained intercalated disks (Fig. 10). The example in Fig. 10 has been assembled from a series of six 1-um optical sections from rat ventricle. An elliptical, obliquely oriented intercalated disk appearing as a ‘bowl-shaped’ structure is observed between two adjacent myocytes. Large spots occur around the ‘lip’ of the ‘bowl; smaller spots, below the ‘lip’, give the impression that the intercalated disk has a concave, interior surface.

The ‘Thru view’ facility of the SOM software permits rotation of disks, allowing face-on views to be correlated directly with the edge-on perspectives normally viewed in longitudinally sectioned cells. A group of projections of the human intercalated disk in Fig. 8, taken at six l5º-angle increments around the y-axis of a stack of 10 optical sections, is shown in Fig. 11. The rotations suggest that the structure is composed of two shallow bowl-shaped disks skewed slightly with respect to one another in threedimensional space. In Fig. HE, where the lower smaller disk is seen oriented edge-on, with its interior obscured, the larger disk is tilted slightly, so that its centrally located junctions are still visible. As the disks are rotated from face-on to edge-on aspects, the characteristic peripheral ring arrangement becomes progressively less obvious. At the same time, the apparent shape of many of the peripheral junctions alters. Junctions that appear elongate or strip-like, with their long axes aligned around the circumference of the face-on disk, progressively broaden out as the disk is rotated to edge-on view, eventually appearing as irregular or rounded domains. This suggests that such junctions are viewed edge-on in face-on disks, and face-on in edge-on disks, i.e. they lie on lateral-facing zones of the cells. When viewed in this way, it becomes clear that immunostained junctions are seldom markedly elongate, and where they have a noticeably longer axis, this shows no preferential orientation.

Table 1 summarises data on numerical density of immunostained junctions (number of junctions μ m^-2 ‘disk’) in ventricular myocardium of the six species. Overall, the absolute number of immunostained junctions per intercalated disk increases with size of the disk, but the numerical density decreases as disk size increases. This trend is significant within all species P<0·05), except guinea pig (Table 1). When disks of all sizes are pooled there is no significant difference P>0·05) in the overall numerical density of immunostained gap junctions between the species.

Histograms illustrating the frequency distributions of fluorescent spot longest axis for the six species are shown in Figs 12 and 13. In all the myocardia examined, the distributions of this index of junction size are asymmetric, being skewed towards the lower end of the range, with the majority of immunostained junctions (>97%) being less than 2 μm long. The small fraction of immunostained

junctions above this size extended up to a maximum of 7·9 μm (mouse ventricle), but only 13 out of the 15 345 immunostained junctions measured in all species were greater than 4 μm in length.

The frequency distribution in the rat atrium shows a flatter curve and lower spread of values than that of rat ventricle (Fig. 13), with the longest immunostained junctions measuring 2·9 μm and 5·4 μm, respectively. There is a highly significant difference (F= 131·0, P<0·01) in size between atrial and ventricular immunostained junctions, those in the atrium being on average 33% longer than those in ventricle (mean spot lengths for atrium and ventricle, 0·56 μm and 0·42 μm, respectively).

Table 2 summarises the results of ANOVA of immunostained gap junction size in the six species. Though overall there was significant variation between species in the mean size of immunostained junctions; means in five of the six species were similar. The cause of the apparent variation between species was that the rabbit had significantly smaller junctions P<0·01) than the other five species. Small differences in spot size between disks of different size were found within species but, though some of these differences reached statistical significance, no consistent trend between these two variables was found across the species examined.

The major result revealed by ANOVA is a pronounced, highly significant difference P<0·01) in size between immunostained junctions located in the central regions of disks compared with those at the periphery in all species. The ANOVA F levels determined for this source of variation were consistently an order of magnitude greater than all other sources of junction length variation examined in this study (Table 2).

In the present study, we have exploited a specific cardiac gap junction antiserum in combination with laser scanning confocal microscopy to extend our understanding of the distribution and arrangement of gap junctions between different types of cardiac muscle cell.

The antiserum used, referred to as anti-HJ, is one of three polyclonal antisera that we have raised to synthetic peptides whose amino acid sequences match different putative cytoplasmically exposed portions of the connexin43 molecule, a cardiac gap junction protein identified in molecular cloning studies (Beyer et al. 1987). Our demonstration that in longitudinally sectioned ventricular myocytes, anti-HJ gives a localised punctate staining pattern that mirrors precisely the positions of intercalated disks seen by standard light microscopy and electron microscopy, suggests very strongly that specific gap junction labelling is being achieved. When anti-HJ is omitted from the staining procedure, or replaced with preimmune serum, no immunofluorescence is observed, demonstrating that the staining observed in the experimental samples is due specifically to the binding of anti-HJ. Direct evidence that gap junctions and not some other component of the intercalated disk are indeed the specific target for anti-HJ binding comes from a detailed characterisation of the antisera (Harfst et al. 1990). In brief, each of the antisera binds specifically: (1) to the peptide to which it was raised; (2) to the proteins of isolated preparations of enriched cardiac gap junctions separated by SDS-polyacrylamide gel electrophoresis; and (3) to ultrastructurally recognisable cardiac gap junctions, as determined by dot blotting, Western blotting and immunogold cytochemistry, respectively (Harfst et al. 1990).

The different patterns of immunostained gap junction localisation observed between atrial and ventricular tissues can be explained by reference to the characteristic morphological features of these two myocyte types. Atrial myocytes are slender (5-10 μm width) compared with their counterparts in the ventricle (25-35 μm width) and they are less uniformly oriented, forming ramifying rather than parallel muscle fibres (McNutt and Fawcett, 1969; Simpson et al. 1973; Severs, 1989a). Although classical step-like intercalated disks are seen by electron microscopy of atrial tissue, they occupy correspondingly shorter widths across the cell. Often, the differentiation between intercalated disk and general (non-disk) sarcolemma is less clear-cut than in the ventricle, and consequently lateral-facing segments of disk membrane may stretch for long distances along the cells’ sides (Forbes and Sperelakis, 1985; Severs, 1985, 1990). This occurs not only at the cell tips, but also in association with small, often poorly distinguished, transverse disk segments located at various points along the cell body. The net result is a ‘spreading’ of the gap junctions over the sides of the cell, rather than their being confined solely to sharply defined regions in the manner of the ventricular myocyte. The distribution of immunostained atrial gap junctions that we report here is thus consistent with the existing observations and data but, without the benefit of laser scanning confocal microscopy it has not previously been possible to appreciate the full extent of such side-to-side coupling.

Our comparison of gap junction localisation in the ventricular myocardium of different species demonstrates that the general plan of gap junction distribution and intercalated disk architecture, and the epitopes of the component gap-junctional protein detected, are common to and conserved within a range of mammalian species, including man. This basic similarity encompasses overall immunostaining pattern, numerical density and size distribution of immunostained junctions, and differentiation of a border of larger junctions at the disk periphery. Within this plan, distinct species differences are nevertheless discernible. Most notably, in the rabbit a small proportion of immunostained gap junctions was detectable at apparent regions of lateral contact between myocytes. These contacts were far less numerous and extensive than those observed in the rat atrium, but show some similarity to the junction distribution observed in the ventricular myocardium of immature rats (Gourdie et al. 19906). The apparent lateral staining in the rabbit may represent a population of gap junctions that is quite separate from the usual form of intercalated disk or, bearing in mind the smaller overall apparent disk sizes measured in the rabbit, they may simply reflect the presence of disks that have a more extended organisation, with less clearly defined limits than those found in other species.

Although the results obtained in human paralleled those from the other species, any detailed conclusions as to their similarity have to be qualified in view of the source of the material examined. The human specimens came from areas of ventricle that appeared histopathologically unaffected in three patients whose hearts were removed at surgery for two different disease conditions (ischaemic heart disease and dilated cardiomyopathy). The samples were not age-matched with those from the other species, and right rather than left ventricle was analysed (the right ventricle being less affected by disease). These variables may well influence some of the data obtained (e.g. hypertrophy is a common response in cardiac disease, and this might be expected to lead to an increase in size of the intercalated disk). Clearly, the use of such material is far from ideal in attempting to establish the characteristics of gap junction organisation in the normal human heart, but this was the only practicable approach available. It seems reasonable to conclude, however, that features present in the human material examined that are also common to all the other species, notably the peripheral ring arrangement, and the data on size and numerical density of immunostained junctions, are in all likelihood features of the normal human heart too.

A major finding of the present study is our demonstration that the intercalated disk in mammalian ventricular myocardium is characterised by a conspicuous peripheral ring of large gap junctions. In larger disks, the rings of brightly stained junctions demarcate the boundaries of the principal ends of the main cell body, and the diameter of the ring thus often corresponds to that of the cell itself. Smaller, and less abundant gap junctions are typically located in the interior zone of the disk, though some large bright spots may also be present in this zone. The gross morphology of the large disks, as revealed by laser scanning confocal microscopy, is more accurately likened to a stack of two bowls than to the classically envisaged irregular step-like structure. The organisation of immunostained gap junctions in smaller disks, located at the interacting ends of stub-like branches of the main cell body, resembles that of the larger disks, though numerical density of immunostained junctions is inversely related to disk size.

The perspective of gap junction organisation at the intercalated disk thus revealed by confocal microscopy differs substantially in emphasis from the existing concepts and models derived from standard light- and electron-microscopical observations. Thin-section and freeze-fracture electron microscopy (e.g. see Page and McCallister, 1973; Forbes and Sperelakis, 1985; Hoyt et al. 1989; Severs, 1985, 1990) provide multiple highly detailed

views of small portions sampled within disks, but in the absence of complete overviews of entire individual disks, a recurrent feature at a more gross level of organisation is easily overlooked, particularly in structures (like the disk) that are of large size and have a complex, variable threedimensional topology. Visualisation of these new features in the present study stems from the ability of the laser scanning confocal microscope to detect the detailed threedimensional distribution of fluorescent labels through substantial depths of tissue (White et al. 1987; Shotton, 1989). Conventional immunohistochemical techniques have also been used previously to visualise gap junctions at the light-microscopical level (Beyer et al. 1989; Yancey et al. 1989; Luke et al. 1989) but as is also apparent from the present study, this approach has inadequate resolution and is unable to provide clear views of immunostained junctions free from out-of-focus blurr over sufficiently large depths of focus. Moreover, the thinner tissue sections normally required (usually <10 μm) for standard light microscopy as compared with confocal microscopy lessen the probability of including entire intercalated disks within the width of the tissue slice examined.

In addition to the integrated overviews of gap junction organisation and intercalated disk structure made possible by confocal microscopy, the images of en-face oriented disks integrated from optical section series are well-suited for morphometric analysis. An alternative morphometric approach, involving silver-enhanced gold immunostaining and standard light microscopy of longitudinally oriented canine ventricular myocytes, has recently been proposed (Luke et al. 1989). However, this method has the drawback that when intercalated disks are viewed edge-on (the predominant orientation of disks in longitudinally sectioned myocytes), larger gap junctions at the disk periphery will appear superimposed upon one another, and obscure the numerous smaller junctions present in the disk central zone. Owing to the ability of the confocal microscope to reduce out-of-focus blur from fluorescently labelled structures (White et al. 1987; Shotton, 1989), sharper visualisation of discrete immunolabelled gap junctions is achieved in confocal images, and so measurements are correspondingly more precise than those obtained from standard light microscopy. Moreover, the laser scanning confocal microscope enables more comprehensive data to be collected and superior sampling as the total population of immunolabelled gap junctions within complete intercalated disks can be analysed.

How far are these theoretical advantages of laser scanning confocal microscopy translated, in practice, to accurate data on junction number and size? Given that the measured sizes of immunostained junctions are often less than the 1-μm optical section steps used in this study, it might be argued that counts made on intercalated disks reconstructed from those optical sections will fail to reveal all the junctions present in the disk. In practice, however, this is not the case. Although the confocal microscope reduces out-of-focus blur from above and below the focal plane, a limited spread of collected light about focus nevertheless occurs (Inoue, 1989). This is demonstrated practically in Fig. 6 where punctate fluorescence from the same individual gap junctions can be seen in register from one 1-μm stepped optical section to the next. Although estimation of the depth of field of the laser scanning confocal microscope is not straightforward, given the objectives used and the setting of the confocal aperture in this study, it can be predicted from a relationship plotted by Wells et al. (1989) that a small overlap in the point spread of light intensity along the z-axis would be expected between sequential 1-μm spaced optical sections. Other potential sources of error are: that dumbell-shaped junctions may be analysed as two separate spots, or that two closely adjacent junctions may appear as a single spot, though carefull procedures to deal with such eventualities were devised for the data collection procedure.

The measurements of immunostained gap junction longest axis correspond well with the dimensions of cardiac gap junctions reported in previous electron-microscopical studies (e.g. see Shibata and Yamamoto, 1979; Gros et al. 1981; see, however, Hoyt et al. 1989). In particular, our data on rat ventricle match closely those published by Shibata and Yamamoto (1979) from freezefracture replicas. These authors found that slightly greater than 50% of junctions were less than 0·1 μm² in area, equivalent to 0·36 μm in diameter (assuming circularity of smaller maculae). Similarly, in the present work, 55 % of immunostained gap junctions measured in the rat ventricle had long axes of 0·4 μm or less. The mean length of immunostained gap junctions in rat atrium was slightly larger than that for rat ventricle as a whole, but was comparable to that of the peripherally located ventricular junctions. The difference between the overall mean sizes for the two tissues may reflect constraints on the dimensions of junctions located within the central regions of ventricular intercalated disks, imposed by the presence of the fascia adherens junctions that occupy a major portion of these regions. In all species, only a small fraction of ventricular gap junctions (3%) were longer than 2 μm, though in the mouse a single junction of 7·9 μm was recorded. Thus, we found no evidence for a major population of very large ribbon-shaped gap junctions oriented with their long axes (5·1 μm mean length) transverse to the long axis of the cell, as proposed in the model described by Hoyt et al. (1989). By rotating disks through 90°, immunostained junctions can often be seen to have an irregular rather than a perfectly rounded shape, but such junctions are seldom markedly elongated, and where a longer axis is apparent, this shows no favoured orientation.

Could the immunodetection of a prominent ring of large gap junctions at the periphery of the ventricular intercalated disk be an artifact of the staining technique? If access of the labelling reagents to the cytoplasmically located epitopes of the gap-junctional protein were more limited in the interior regions of the disks than in the peripheral zones, then a differential pattern of the type we describe might be produced artifactually. However, our preparations show clear and consistent staining at all levels in the tissue slice, and where large gap junctions are detected in the central regions of disks, they fluoresce as brightly as those at the periphery. These two observations suggest that the trypsin treatment used in the procedure, aided by inclusion of detergent as a wetting agent in the lysine blocking step, are adequate in ensuring even penetration of the immunostaining reagents both throughout the thickness and within the individual cells of the entire tissue slice.

It thus appears justified, in the light of the present findings, to reconsider some aspects of gap junction organisation and intercalated disk morphology deduced from electron microscopy. Longitudinal steps of membrane bearing gap junctions in the interior zone of the intercalated disk may, on the basis of our observations, either be less frequent or less extensive than previously supposed, or their gap junctions may be substantially smaller and less common than those of the outermost longitudinalfacing segment circumscribing the cell periphery. Some of the large gap junctions described by Hoyt et al. (1989) situated in a lip of membrane overlapping the cell periphery could correspond to the peripheral ring of gap junctions we describe, though these authors did not distinguish peripheral from interiorly located longitudinal segments of disk membrane as the preferential site for large gap junctions. The few large bright spots that we detect in the central zone of the disk most likely represent large gap junctions situated in segments of longitudinalfacing membrane located within the central zone. Many of the smaller spots seen in the central zone probably represent small gap junctions that are known to occur in the transversely oriented interdigitating regions, i.e. in the membrane zones containing the fasciae adherentes (Hoyt et al. 1989; Severs, 1990).

What might be the functional significance of the gap junction organisation we describe? One possibility is that a peripheral ring of large gap junctions represents the most efficient arrangement for rapid longitudinal spread of electrical impulses between myocytes. If two rod-shaped ventricular myocytes are envisaged, interconnected end-to-end by an intercalated disk, then the large peripheral gap junctions can be imagined as lining the inside surface of an exterior rim of membrane created by the abutment of the two cells. The principal gap junction ring thus lies directly in the path of the depolarising current moving along the surface of one cell to its neighbour, thus increasing the rapidity of electrical current propagation in the direction of the long axes of the myocytes. The role of the smaller gap junctions in the disk interior is less certain; whether these junctions differ in function from their counterparts at the disk periphery is an intriguing possibility for future investigation.

We are very grateful to Dr N. B. Gilula (Research Institute of the Scripps Clinic, La Jolla, California) for supplying the peptides used for raising antibodies. The substantial efforts of Dr Liz Harfst in raising and initially characterising the antipeptide antisera are gratefully acknowledged. We thank Mr Stephen Rothery for his photographic expertise and assistance, and Dr Sian Harding for organising the samples of human tissue. This work was supported by grants from the Medical Research Council and the British Heart Foundation (grant no. 86/39). Dr C. R. Green is a Royal Society University Research Fellow.