The spatial distribution and relative abundance of gap ... - CiteSeerX

Journal of Cell Science 105, 985-991 (1993) Printed in Great Britain © The Company of Biologists Limited 1993

985

The spatial distribution and relative abundance of gap-junctional connexin40 and connexin43 correlate to functional properties of components of the cardiac atrioventricular conduction system Robert G. Gourdie1,*, Nicholas J. Severs2, Colin R. Green1, Stephen Rothery2, Patricia Germroth3 and Robert P. Thompson3 1Department 2Department 3Department

of Anatomy and Developmental Biology, University College London, Gower Street, London WC1E 6BT, UK of Cardiac Medicine, National Heart and Lung Institute, Dovehouse Street, London SW3 6LY, UK of Cell Biology and Anatomy, Medical University of South Carolina, 171 Ashley Avenue, Charleston, South Carolina 29425, USA *Author for correspondence

SUMMARY Electrical coupling between heart muscle cells is mediated by specialised regions of sarcolemmal interaction termed gap junctions. In previous work, we have demonstrated that connexin42, a recently identified gapjunctional protein, is present in the specialised conduction tissues of the avian heart. In the present study, the spatial distribution of the mammalian homologue of this protein, connexin40, was examined using immunofluorescence, confocal scanning laser microscopy and quantitative digital image analysis in order to determine whether a parallel distribution occurs in rat. Connexin40 was detected by immunofluorescence in all main components of the atrioventricular conduction system including the atrioventricular node, atrioventricular bundle, and Purkinje fibres. Quantitation revealed that levels of connexin40 immunofluorescence increased along the axis of atrioventricular conduction, rising over

10-fold between atrioventricular node and atrioventricular bundle and a further 10-fold between atrioventricular bundle and Purkinje fibres. Connexin40 and connexin43, the principal gap-junctional protein of the mammalian heart, were co-localised within atrioventricular nodal tissues and Purkinje fibres. By applying a novel photobleach/double-labelling protocol, it was demonstrated that connexin40 and connexin43 are colocalised in precisely the same Purkinje fibre myocytes. A model, integrating data on the spatial distribution and relative abundance of connexin40 and connexin43 in the heart, proposes how myocyte-type-specific patterns of connexin isform expression account for the electrical continuity of cardiac atrioventricular conduction.

INTRODUCTION

tricular bundle and is finally distributed via the Purkinje system to synchronise the contraction of the ‘working’ ventricular myocardium. Individual channels (connexons) of the gap junction are composed of transmembrane proteins (connexins), encoded by a multigene family (Willecke et al., 1991; Kumar and Gilula, 1992). Transcripts for at least seven different connexin isoforms have been reported in vertebrate hearts (Beyer et al., 1987, 1988; Willecke et al., 1991; Kanter et al., 1992) and recent evidence from avian hearts suggests that different connexin isoforms form channels with distinct electrophysiological properties (Veenstra et al., 1992). In mammalian heart, the principal connexin, connexin43 (Beyer et al., 1987, 1989), is localised to gap junctions between ‘working’ atrial and ventricular myocytes (Beyer et al., 1989; Gourdie et al., 1991; Yancey et al., 1992; Fromaget et al., 1992; Dolber et al., 1992). The presence of

During the heart beat, the atria and ventricles contract in rapid succession. The coordination of this sequence results from the orderly propagation of action potential through the cardiac atrioventricular conduction system. At the cellular level, continuity of atrioventricular electrical conduction is mediated by gap junctions. These junctions are composed of aggregates of intercellular membrane channels that act as conduits for the diffusion of ions and hence the flow of electrical current between myocytes (Barr et al., 1965; Severs, 1990; Page, 1992). The conduction and propagation of action potential occurs through a well established sequence of specialised cardiac tissues. From initiation at the sinuatrial node, action potential spreads through the atria and is slowed as it coalesces at the atrioventricular node. The action potential is then accelerated along the atrioven-

Key words: heart, conduction tissue, gap junction, connexins

986

R. G. Gourdie and others

low levels of connexin43 have also been reported in the specialised muscle tissues of the atrial nodes (Pressler et al., 1991; Anumonwo et al., 1992; Gourdie et al., 1992; ten Velde et al., 1992), but not within the atrioventricular bundle (Gourdie et al., 1990a; van Kempen et al., 1991; Gourdie et al., 1992). This lack of connexin43 is puzzling, as electron microscopic studies show that gap junctions are abundant in the atrioventricular bundle of a variety of mammalian species (Kawamura and James, 1971; Viragh and Challice, 1973; Arluk and Rhodin, 1974; Marino, 1979). Kanter and co-workers (1992) recently cloned and sequenced cDNAs for two novel connexins from dog heart; connexin40 and connexin45. Using antisera specific to connexin40, connexin43 or connexin45, these authors demonstrated that in isolated canine cardiomyocytes, all of these connexins were localised to ultrastructurally defined gap junctions. One possible explanation for the paradoxical lack of connexin43 identified in certain well-coupled myocardial tissues might be the expression of another connexin. Evidence supporting this hypothesis is our finding that connexin42, thought to be an avian homologue of canine connexin40 (Haefliger et al., 1992; Kanter et al., 1992), is specifically localised to gap junctions in avian conducting tissues, including cells of the atrioventricular bundle (Gourdie et al., 1993). In the present study we set out to determine whether a parallel distribution of connexin40 occurs in mammalian myocardium and, if so, how the differential expression of connexin isoforms allows for electrical coupling between conduction tissues and ‘working’ myocardium. The results reveal that electrical continuity along the cardiac atrioventricular conduction system involves overlapping distributions of connexin40 and connexin43. MATERIALS AND METHODS Immunofluorescence labelling of connexins Preparation of cardiac tissues from 40- and 90-day-old SpragueDawley rats, immunolabelling for connexins using rabbit polyclonal anti-connexin antibodies (i. e. HJ (anti-connexin43), 1:500 dilution in phosphate buffered saline (PBS), and anti-connexin40, 1:200 dilution in PBS) and controls were carried out as described previously (Gourdie et al., 1991, 1992). Prior to immunolabelling of atrioventricular nodal tissue with HJ antisera, a trypsin concentration of 0.01% was used for protease unmasking of antigenic sites. Localisation of primary rabbit antibodies was detected by indirect immunofluorescence using biotinylated anti-rabbit antibodies (Amersham, UK, 1:250 in PBS, 1 h, 20˚C) and streptavidin conjugated to fluorescein or Texas Red (Amersham, UK, 1:250 in PBS, 1 h, 20˚C). The characterisation and specificity of antisera against connexin43 (Harfst et al., 1990; Gourdie et al., 1990b, 1991, 1992) and connexin40 (Kanter et al., 1992) has been described in previous studies.

Photobleach double-labelling In a photobleach double-labelling procedure, a second anti-connexin43 (mCx43, Zymed, California) antibody was used. This antibody is a mouse monoclonal antibody raised against a synthetic peptide immunogen matching residues 250-272 of rat connexin43. This peptide sequence is located within the C-terminal domain of connexin43 and has little homology with amino acid sequences within other known connexins. Prior to conducting

experiments, serial histological sections were stained with either this monoclonal antibody or HJ, the rabbit polyclonal anti-connexin43 antibody. The two antibodies demonstrated identical patterns of immunolocalisation in atrioventricular nodal, atrioventricular bundle, Purkinje fibres and ‘working’ myocardial tissues. These patterns were consistent with earlier reports of the spatial distribution of connexin43 in myocardial tissues from adult rat (Gourdie et al., 1991, 1992; van Kempen et al., 1991), indicating the specificity of the mouse monoclonal antibody (mCx43) for connexin43. The protocol used in the double-labelling experiment was as follows. A standard immunofluorescence labelling run for connexin40 was carried out. Immunopositive Purkinje fibres in the left ventricular wall were identified and imaged by confocal optical sectioning. A small rectangle of tissue, including the imaged Purkinje fibre, was then photobleached using the scanning laser of the confocal microscope (25 milliwatt argon ion laser, high power, 0 neutral density, 40 min). Subsequent optical sectioning of the photobleached region using a ×60, 1.4 NA objective lens indicated no detectable fluorescence within previously labelled Purkinje fibres. After photobleaching, the tissue was re-fixed in 2% paraformaldehyde for 5 min and washed in two changes of PBS for 30 min. The subsequent immunolabelling protocol for mCx43 (1:20 dilution in PBS) was the same as that for HJ (anticonnexin43), except that immunofluorescence localisation of mCx43 was performed using secondary anti-mouse Igs conjugated to fluorescein (DAKO, UK, 1:20 in dilution PBS, 1 h, 20˚C). In control incubations, no re-labelling of photobleached connexin40positive cells was observed following treatment of the tissue with anti-mouse Igs conjugated to fluorescein (1:20, 1 h, 20˚C) that had not been previously incubated with the mouse monoclonal against connexin43.

Quantitation of immunofluorescence A quantitative study was done on histological sections from three adult rat hearts (90 days). Connexin40 (anti-connexin40) and connexin43 (HJ) immunofluorescence was measured on single optical sections (zoom 2) using PC-IMAGE software as described previously (Gourdie et al., 1991). Initially, a median convolving filter was passed over the whole image to remove background noise. A pixel intensity threshold was then done such that the bright, fluorescence-labelled gap junctions were demarcated by an overlying colour binary image. The correspondence between the two images was checked by toggling back and forth between the grey and binary images. Once correspondence was judged to be satisfactory, calibrated automatic measurement of the total area of immunolabelled fluorescence was recorded from the binary image. Atrioventricular node, distal atrioventricular bundle, Purkinje fibres of the left ventricular wall and ‘working’ myocardium of the interventricular septum were sampled three times at different locations on histological sections from each of the three hearts. Within the atrioventricular node, three samples were taken at anterior, posterior and mid-nodal regions. Quantitation of connexin40 and connexin43 immunofluorescence in Purkinje fibres was carried out on serial histological sections similar to those shown in Fig. 2a and b (below), i.e. where distinctively shaped Purkinje fibres could be identified on sequential serial sections. Estimates of numbers of nuclei were made by restaining connexin40 and connexin43 immunolabelled sections with propidium iodide (Ockleford et al., 1981) and doing counts of nuclei at microscopic fields similar to those sampled for connexin quantitation. The numerical data are given as the overall mean area µm2 (± standard error) of immunofluorescence per nucleus (index of gap junction per cell).

Connexins in heart conduction tissues Confocal microscopy All imaging was done using a Bio-Rad MRC-500 confocal microscope (Bio-Rad Microscience Ltd, Hemel Hempsted, UK). Those images indicated as projections were taken using a ×60, 1.4 NA objective, at zoom levels of 1, 2 or 4, and projected by maximum pixels.

RESULTS Connexin40 and connexin43 co-localise in atrioventricular node In Fig. 1a, high levels of punctate connexin43 immunofluorescence are observed in atrial and ventricular ‘working’ myocardium adjacent to atrioventricular node. The photobleached area (arrowed) in Fig. 1a indicates a region of nodal tissue that has been optically sectioned in 24 × 0.3 µm z-focus steps at high magnification ( ×60, 1.4 NA objective, zoom 4). Digital projection of connexin43 immunolabelling in this 7.2 µm deep volume reveals small puncta distributed around atrioventricular nodal cells (Fig. 1b). In Fig. 1c a similar distribution of connexin40 immunolabelling is observed in nodal tissue. This projected volume was imaged at a coincident location on a histological section serial to that shown in Fig. 1b. Co-distribution of connexin40 and connexin43 occurred at uniformly low levels in the mid and peripheral atrioventricular node. A total of 224 junction-like punctae can be counted in Fig. 1b and c, and 96% of these fluorescent structures have longest chord diameters below 0.5 µm. This size range is consistent with ultrastructural morphometry of gap junctions in mammalian nodal tissues (Marino, 1979, Masson-Pevet et al., 1979; Sugi and Hirakow, 1986). Connexin40, but not connexin43, localises in the atrioventricular bundle In Fig. 1d a survey view of the interventricular septum and atrioventricular bundle is shown. Connexin40 immunolabelling is observed between atrioventricular bundle myocytes, but not between the myocytes of the ‘working’ ventricular myocardium subjacent to the conduction tissues (Fig. 1e). Conversely, in a histological section serial to that shown in Fig. 1e, connexin43 is specifically immunolocalised to ‘working’ ventricular myocardium (Fig. 1f), no connexin43 immunolocalisation is observed in atrioventricular bundle tissue. Connexin40 and connexin43 co-localise in Purkinje fibres Subendocardial strands of Purkinje fibres lining the sides of the interventricular septum and walls of the ventricles demonstrated high levels of connexin40 immunolabelling (Fig. 2a). In serial histological sections, these strands could be traced along the sides of the interventricular septum and were in continuity with the bundle branches. As is the case for connexin42 localisation between avian cardiac Purkinje fibre cells (Gourdie et al., 1993), high levels of punctate connexin40 immunoreactivity were concentrated at intercalated disk-like structures in rat Purkinje fibres (Fig. 2a). In contrast to the distal atrioventricular bundle, Purkinje fibres co-labelled for both connexin40 and connexin43. Fig.

987

2a and b are serial histological sections through a distinctive subendocardial strand of cells positive for both connexin40 and connexin43. In Fig. 2b a delicate protrusion of ‘working’ myocardium forms a narrow zone of contact with the strand of Purkinje fibre. This contact zone shows high levels of punctate connexin43 immunofluorescence and identifies a site of potential coupling between conduction and ‘working’ myocardium. To establish whether the two connexins are co-localised precisely in identical Purkinje fibres, a novel protocol was developed permitting re-labelling of the same histological section. Fig. 3a and b shows a Purkinje fibre (arrowed) labelled by indirect immunofluorescence using rabbit polyclonal connexin40 antibodies. Following photobleaching by scanning laser (Fig. 3c), and re-labelling using a mouse monoclonal antibody against connexin43, the same Purkinje fibre and adjacent ‘working’ myocardium demonstrate connexin43 immunolabelling (Fig. 3d). Quantitation of connexin40 and connexin43 in atrioventricular conduction tissues The area of connexin40 and connexin43 immunofluorescence per nucleus was quantified in atrioventricular nodal, atrioventricular bundle, Purkinje fibres and ‘working’ ventricular myocardium. The data are summarised in Fig. 4, along with a diagrammatic representation of the spatial distribution of the two connexins in the cardiac atrioventricular conduction system. Quantification indicates that ‘cellnormalised’ levels of connexin40 show a graded distribution along the atrioventricular conduction system, increasing approximately 10-fold between the atrioventricular node and atrioventricular bundle and a further 10-fold between the atrioventricular bundle and Purkinje fibres. The data are consistent with the relative abundance of morphologically identified gap junctions deduced from electron microscopy (Marino, 1979; Forbes and Sperelakis, 1985). DISCUSSION Here we show that connexin40 is present in gap junctionlike structures between myocytes comprising the major components of the cardiac atrioventricular conduction system. These components include the atrioventricular node, atrioventricular bundle and Purkinje fibres. Quantitation of gap-junctional connexin40 reveals that it is not uniformly distributed along the atrioventricular conduction axis, but increases in abundance 100-fold between the atrioventricular node and Purkinje fibres. Connexin43 is colocalised with connexin40 in atrioventricular nodal tissues and Purkinje fibres, but not in the distal atrioventricular bundle. This pattern of co-expression for connexin40 and connexin43 may have an important functional role in the intercellular propagation of action potential through the heart, as it enables electrical connection between specialised conduction cells and ‘working’ myocytes. Based on the data given in Fig. 4, the following sequence for conduction of action potential through the atrioventricular conduction system may be surmised. Atrial myocytes express connexin43 (Gourdie et al., 1991), and hence are able to couple electrically with atrioventricular nodal cells

988

R. G. Gourdie and others

Fig. 1. Connexin40 and connexin43 immunolabelling of atrioventricular node (avn) and atrioventricular bundle (avb) imaged by confocal scanning laser microscopy. (a) Survey of a section through the atrioventricular node (90-day rat) after labelling for connexin43. High levels of punctate immunolabelling occur in atrium (a) and ventricle (v) adjacent to the node. Note the photobleached area (arrows) within the atrioventricular node corresponding to the area in Fig. 1b. (b) Projection (zoom 4, 24 × 0.3 µm z-focus steps) of connexin43 immunolabelling between atrioventricular nodal cells. (c) Connexin40 immunolabelling in atrioventricular node; projection (zoom 4, 24 × 0.3 µm z-steps). This area was imaged at a corresponding location on the histological section serial to that shown in Fig. 1b. (d) Survey of a section through distal atrioventricular bundle (40-day rat) after immunolabelling for connexin40. The area optically sectioned and projected in (e) is arrowed. (e) Projection (zoom 1, 5 × 1 µm z-steps) showing specific connexin40 immunolocalisation to distal atrioventricular bundle tissue; note absence of label in subjacent ‘working’ ventricular myocardium (v). (f) Connexin43 is specifically immunolocalised to ‘working’ myocardium, but not atrioventricular bundle myocytes (avb); projection (zoom 1, 5 × 1 µm zsteps). Bars: (a) 30 µm; (b,c) 2 µm; (d) 200 µm; (e,f) 10 µm.

Connexins in heart conduction tissues

989

Fig. 3. A photobleach/double-labelling experiment showing that rabbit polyclonal connexin40 antibodies and mouse monoclonal connexin43 antibodies co-localise in the same Purkinje fibre. (a) Survey of a histological section through the left ventricular wall (90-day rat) following immunolabelling for connexin40. A group of connexin40-positive cells (indirectly labelled by biotinylated rabbit antibodies and streptavidin conjugated to Texas Red) is arrowed. (b) A projection (zoom 2, 15 × 1 µm z-steps) of connexin40 immunolabelling at the arrowed location (c). The same region (corners arrowed), following photobleaching using the confocal scanning laser (25 milliwatt argon ion laser, high power, 0 neutral density, 40 min). (d) Projection (zoom 2, 15 × 1 µm z-steps) at the photobleached region of tissue following re-immunolabelling with a mouse monoclonal antibody against connexin43 (mCx43). Connexin43 immunolabelling is localised in the same cells that were immunopositive for connexin40 (cf Fig. 3b). Bars: (a,c) 10 µm; (b,d) 5 µm.

Fig. 2. Connexin40 and connexin43 (HJ) immunolabelling of ‘working’ ventricular myocardium and Purkinje fibres (90-day rat). (a) A subendocardial strand of Purkinje fibre demonstrates high levels of punctate connexin40 immunolabelling localised in clusters representing intercalated disk-like structures. (b) The same strand on a serial histological section shows a correlated distribution of immunolabelling for connexin43 in the Purkinje fibre. High levels of punctate connexin43 immunofluorescence are observed in obliquely oriented disks between ‘working’ ventricular myocytes. The arrow indicates a delicate zone of contact between the Purkinje fibre and ‘working’ myocardium. Fig. 2a,b were projected from 15 × 1 µm optical sections taken at zoom 1. Bars, 10 µm.

Fig. 4. Summary of the spatial distributions and quantification of connexin40 (red dots) and connexin43 (green dots) in the atrioventricular conduction system. The numerical data are given as the overall mean area (± s.e.m.) in µm2 of immunofluorescence per nucleus (index of immunolabelled gap junction per cell) in each of the four myocardial tissues sampled.

990

R. G. Gourdie and others

expressing both connexin40 and connexin43. Within the node, propagation of action potential is slowed as intercellular diffusion of ions, and hence flow of electric current, is restricted by exceptionally small, sparse gap junctions. From the atrioventricular node, depolarising membrane potential accelerates as it is propagated down the atrioventricular bundle. This compartment of cells is coupled by gap junctions containing connexin40 and is insulated from adjacent ‘working’ myocardium expressing connexin43. Finally, the action potential is distributed into the Purkinje fibre system. Purkinje cells co-express connexin40 and connexin43, and are able to link the conduction system directly to ‘working’ myocardium, enabling coordinated depolarisation and contraction of the ventricles. Cell-type specificity of gap junction communication is implicit in this model. Similar specificity was first observed in cell culture, but has since been reported in vivo in systems as diverse as molluscan embryos (Van den Biggelaar and Serras, 1988; Serras et al., 1989) and mammalian skin (Kam and Hodgins, 1989; see also Risek et al., 1992). Evidence from studies of Xenopus oocyte pairs injected with mRNAs for different connexin isoforms suggests that where heteromolecular channels form, only particular combinations give functional junctions (Swenson et al., 1989; Barrio et al., 1991; reviewed by Kumar and Gilula, 1992). Interestingly, functional hybrid channels have asymmetric or rectifying conduction characteristics, a property that could be envisaged as useful in an electrically excitable tissue such as the myocardium. However, whilst different connexin isoforms have been immunolocalised within the same gap junctional aggregate of channels (Traub et al., 1989), the existence of heteromolecular channels has yet to be demonstrated in vivo. The spatial distributions and relative abundance of connexin40 and connexin43 correlate with functional properties of components of the atrioventricular conduction system and suggest a pattern of intercellular electrical couplings that ensures continuity in propagation of action potential through the heart. Important substructural divisions and heterogeneity of cell phenotype occur within components of the conduction system such as the atrioventricular node (Viragh and Challice, 1973; Anderson et al., 1974). Future work may therefore concentrate usefully on an even finer dissection of connexin isoform distribution within atrioventricular conducting tissues. For example, there is evidence for significant expression of at least three other connexin isoforms in mammalian heart (Beyer et al., 1988; Willecke et al., 1991; Kanter et al., 1992), for which no data on spatial distribution exist. Such information, together with the knowledge that different cardiac connexins form channels with distinct electrophysiological properties (Veenstra et al., 1992), may reveal further insight into the mechanisms governing the electrophysiological behaviour of the normal and abnormal heart. This work was supported by grants from the British Heart Foundation (grant no. 90/84), a travel fellowship from the Royal Society to Dr R. P. Thompson and US Public Health Service/National Institutes of Health (grant no. NIH/HL37704). We are grateful to Dr Eric Beyer (Washington University School of Medicine, USA) for connexin40 antiserum and Prof. R. H. Anderson (National

Heart and Lung Institute, UK) for his discussion. The editorial assistance of Dr Antonia Brizzolara is acknowledged with gratitude. Dr Rob Gourdie is a British Heart Foundation Research Fellow.

REFERENCES Anderson, R. H., Janse, M. J., van Capelle, F. J. L., Billett, J., Becker, A. E. and Durrer, D. (1992). A combined morphological and electrophysiological study of the atrioventricular node of the rabbit heart. Circ. Res. 35, 909-922. Anumonwo, J. M. B., Wang, H-Z., Trabka-Janik, E., Dunham, B., Veenstra, R. D. and Jalife, J. (1992). Gap junctional channels in adult mammalian sinus nodal cells. Circ. Res. 71, 229-239. Arluk, D. J. and Rhodin, A. G. (1974). The ultrastructure of calf heart conducting fibers with special reference to nexuses and their distribution. J. Ultrastruct. Res. 49, 11-23. Barr, L., Dewey, M. M. and Berger, W. (1965). Propagation of action potentials and the structure of the nexus in cardiac muscle. J. Gen. Physiol. 48, 797-823. Barrio, L. C., Suchyna, T., Bargiello, T., Xu, L. X., Roginski, R. S., Bennet, M. V. L. and Nicholson, B. J. (1991). Gap junctions formed by connexins 26 and 32 alone and in combination are differently affected by applied voltage. Proc. Nat. Acad. Sci. USA 88, 1883-1887. Beyer, E. C., Goodenough, D. A. and Paul, D. L. (1988). The connexins: a family of related gap junction proteins. In Gap Junctions (ed. E. L. Hertzberg and R. Johnson), pp. 167-175. New York: Alan R. Liss Inc. Beyer, E. C., Kistler, J., Paul, D. L. and Goodenough, D. A. (1989). Antisera directed against connexin43 peptides react with a 43-kd protein localised to gap junctions in myocardium and other tissues. J. Cell Biol. 108, 595-605. Beyer, E. C., Paul, D. L. and Goodenough, D. A. (1987). Connexin43: a protein from rat heart homologous to a gap junction protein from liver. J. Cell Biol. 105, 2621-2629. Dolber, P. C., Beyer, E. C., Junker, J. L. and Spach, M. S. (1992). Distribution of gap junctions in dog and rat ventricle studied with a double-label technique. J. Mol. Cell Cardiol. 24, 1443-1457. Forbes, M. S. and Sperelakis, N. (1985). Intercalated discs of mammalian heart: a review of structure and function. Tissue & Cell 17, 605-648. Fromaget, C., El Aoumari, A. and Gros, D. (1992). Distribution pattern of connexin 43, a gap junctional protein, during the differentiation of mouse heart myocytes. Differentiation 51, 9-20. Gourdie, R. G., Green, C. R., Severs, N. J., Anderson, R. H. and Thompson, R. P. (1993). Evidence for a distinct gap-junctional phenotype in ventricular conduction tissues of the developing and mature avian heart. Circ. Res. 72, 278-289. Gourdie, R. G., Green, C. R., Severs, N. J. and Thompson, R. P. (1990a). Three-dimensional reconstruction of gap junction arrangement in developing and adult rat hearts. Trans. R. Microsc. Soc. 1, 417-420. Gourdie, R. G., Green, C. R., Severs, N. J. and Thompson, R. P. (1992). Immunolabelling patterns of gap junction connexins in the developing and mature rat heart. Anat. Embryol. 185, 363-378. Gourdie R. G., Harfst E., Severs N. J. and Green, C. R. (1990b). Cardiac gap junctions in rat ventricle: Localization using site-directed antibodies and laser scanning confocal microscopy. Cardioscience 1, 75-82. Gourdie, R. G., Severs, N. J. and Green, C. R. J. (1991). Gap junction distribution in adult mammalian myocardium revealed by an antipeptide antibody and laser scanning confocal microscopy. J. Cell Sci. 99, 41-55. Haefliger, J-A., Bruzzone, R., Jenkins, N. A., Gilbert, D. J., Copeland, N. G. and Paul, D. A. (1992). Four novel members of the connexin family of gap junction proteins. J. Biol. Chem. 267, 2057-2064, Harfst, E., Severs, N. J. and Green, C. R. (1990). Cardiac myocyte gap junctions: Evidence for a major connexon protein with an apparent relative molecular mass of 70 000. J. Cell Sci. 96, 591-604. Kam, E. and Hodgins, M. B. (1989). Communication compartments in hair follicles and their implication in differential control. Development 114, 389-393. Kanter, H. L., Saffitz, J. E. and Beyer, E. C. (1992). Cardiac myocytes express multiple gap junctional proteins. Circ. Res. 70, 438-444. Kawamura, K. and James, T. N. (1971). Comparative ultrastructure of cellular junctions in working myocardium and the conduction system under normal and pathologic conditions. J. Mol. Cell Cardiol. 3, 31-60.

Connexins in heart conduction tissues Kumar, N. M. and Gilula, N. B. (1992). Molecular biology and genetics of gap junction channels. Semin. Cell Biol. 3, 3-16. Marino, T. A. (1979). The atrioventricular node and bundle in the ferret heart: a light and quantitative electron microscopic study. Amer. J. Anat. 154, 365-392. Masson-Pevet, M., Bleeker, W. K. and Gros, D. (1979). The plasma membrane of leading pacemaker cells in the rabbit sinus node: A qualitative and quantitative ultrastructural analysis. Circ. Res. 45, 621629. Ockleford, C. D., Bae-Li, H., Wakely, J., Badely, R. A., Whyte, A. and Page-Faulk, W. (1981). Propidium iodide as a nuclear marker in immunofluorescence. I. Use with tissue and cytoskeleton studies. J. Immunol. Meth. 43, 261-267. Page, E. (1992). Cardiac gap junctions. In The Heart and Cardiovascular System, 2nd edn (ed. H. A. Fozzard, E. Haber, R. B. Jennings, A. M. Katz and H. E. Morgan), pp. 1003-1047. New York: Raven Press. Pressler, M. L., Brodhecker, C. A., Huang, X-D., Zipes, D. P. and Sandusky, G. E. (1991). Differential expression of gap junctional proteins in specialized and working myocardium. Circulation, Supplement II, 84, 1290. Risek, B., Klier, G. and Gilula, N. B. (1992). Multiple gap junction genes are utilised during rat skin and hair development. Development 116, 639651. Serras, F., Damen, P., Dictas, W. J. A., Notenboom, R. G. E and Van den Biggelaar, J. A. M. (1989). Communication compartments in the ectoderm of the embryo of Patella vulgata. Roux’s Arch. Dev. Biol. 198, 191-200. Severs, N. J. (1990). The cardiac gap junction and intercalated disk. Int. J. Cardiol. 26, 137-173. Sugi, Y. and Hirakow, R. (1986). Freeze-fracture studies of the sinuatrial and atrioventricular nodes of the caprine heart, with special reference to the nexus. Cell Tiss. Res. 245, 273-279. Swenson, K. I., Jordan, J. R., Beyer, E. C. and Paul, D. L. (1989).

991

Formation of gap junction pairs by expression of connexins in Xenopus oocyte pairs. Cell 57, 145-155. ten Velde, I., De Maziere, A., de Jonge, D., Gros, D. and Jongsma, H. J. (1992). Gap junction distribution in and around sinoatrial node studied by immunohistochemistry. Histochem. J. 24, 575. Traub, O., Look, J., Dermietzel, R., Brummer, F., Hulser, D. and Willecke, K. (1989). Comparative characterisation of the 21 kDa and 26 kDa gap junction proteins in murine liver and cultured hepatocytes. J. Cell Biol. 108, 1039-1051. Van den Biggelaar, J. A. M. and Serras, F. (1988). Determinative decisions and dye coupling changes in the molluscan embryo. In Gap Junctions. Modern Cell Biology, vol. 7 (ed. E. Hertzberg and R. G. Johnson), pp. 483-493. New York: Alan R. Liss Inc. van Kempen, M. J. A., Fromaget, C., Gros, D., Moorman, A. F. M. and Lamers, W. H. (1991). Spatial distribution of connexin 43, the major cardiac gap junction protein, in the adult and developing rat heart. Circ. Res. 68, 1638-1651. Veenstra, R. D., Wang, H-Z., Westphale, E. M. and Beyer, E. C. (1992). Multiple connexins confer distinct regulatory and conductance properties of gap junctions in developing heart. Circ. Res. 71, 1277-1283. Viragh, Sz. and Challice, C. E. (1973). The impulse generation and conduction system of the heart. In Ultrastructure of the Mammalian Heart (ed. C. E. Challice, Sz. Viragh), pp. 43-89. New York: Academic Press. Willecke, K., Hennemann, H., Dahl, E., Jungbluth, S. and Heynkes, R. (1991). The diversity of connexin genes encoding gap-junctional proteins. Eur J. Cell Biol. 56, 1-7. Yancey, S. B., Biswal, S. and Revel, J-P. (1992). Spatial and temporal patterns of distribution of the gap junction protein connexin43 during mouse gastrulation and development. Development 114, 203-212. (Received 12 March 1993 - Accepted 5 May 1993)

Note added in proof Bruzzone et al., (Mol. Biol. Cell (1993) 4, 7-20) have recently reported that Xenopus oocytes solely expressing connexin40 do not form functional channels when coupled to oocytes expressing only connexin43. This specificity of connexin interaction would support our proposal that cellular co-expression of connexin40 and connexin43 at the atrioventricular node and at Purkinje fibres is required to ensure electrical continuity along the axis of atrioventricular conduction in the mammalian heart.