Gap junctions, homeostasis, and injury


  • Antonio De Maio,

    Corresponding author
    1. Division of Pediatric Surgery and Department of Physiology, Johns Hopkins University School of Medicine, Baltimore, Maryland
    • Johns Hopkins University School of Medicine, 720 South Rutland Avenue, Ross 746, Baltimore, MD 21205.
    Search for more papers by this author
  • Virginia L. Vega,

    1. Division of Pediatric Surgery and Department of Physiology, Johns Hopkins University School of Medicine, Baltimore, Maryland
    Search for more papers by this author
  • Jorge E. Contreras

    1. Division of Pediatric Surgery and Department of Physiology, Johns Hopkins University School of Medicine, Baltimore, Maryland
    Search for more papers by this author


Gap junctions (Gj) play an important role in the communication between cells of many tissues. They are composed of channels that permit the passage of ions and low molecular weight metabolites between adjacent cells, without exposure to the extracellular environment. These pathways are formed by the interaction between two hemichannels on the surface of opposing cells. These hemichannels are formed by the association of six identical subunits, named connexins (Cx), which are integral membrane proteins. Cell coupling via Gj is dependent on the specific pattern of Cx gene expression. This pattern of gene expression is altered during several pathological conditions resulting in changes of cell coupling. The regulation of Cx gene expression is affected at different levels from transcription to post translational processes during injury. In addition, Gj cellular communication is regulated by gating mechanisms. The alteration of Gj communication during injury could be rationalized by two opposite theories. One hypothesis proposes that the alteration of Gj communication attenuates the spread of toxic metabolites from the injured area to healthy organ regions. The alternative proposition is that a reduction of cellular communication reduces the loss of important cellular metabolisms, such as ATP and glucose. © 2002 Wiley-Liss, Inc.

Body homeostasis is maintained by the coordinated communication between cells within the same organ or different organ systems. Cross-talk between cells is accomplished by paracrine or autocrine mechanisms. Thus, molecules that are released into the extracellular environment, such as neurotransmitters, hormones, and growth factors, interact with receptors, triggering specific responses. These responses could be at the level of ionic currents, metabolic activation, and changes of gene expression, which are dependent on the effectors and cell type. An alternative mechanism of cellular communication is the direct transfer of molecules within cells without reaching the extracellular environment. This intercellular communication is mediated by gap junctions (Gj). Gj are low resistance pathways that allow the passage of ions (Na+, K+, Ca+2, H+, etc.), and low molecular weight metabolites (cycle nucleotides, inositol 1,4,5-trisphosphate, etc.) between neighboring cells (Bennett et al., 1991; Bruzzone et al., 1996; Goodenough et al., 1996). Thus, a signal can pass from one cell to the next traveling long distances (measured in cell layers) without being exposed to the extracellular milieu. Gj are involved in the regulation of many organ processes, such as the coordinated contraction of heart (Page and Manjunath, 1986) and hepatic glucose metabolism (Hooper and Subak-Sharpe, 1981).

Gj were first described by electron microscopy as a close separation (2–4 nm space) between neighboring cells. Additional studies demonstrated that they were formed by channel conglomerates on the surface of opposite cells that allow the passage of small cytosolic components (Bennett et al., 1991; Bruzzone et al., 1996). These channels are formed by two hemichannels, coined connexons, on the plasma membrane of adjacent cells (Fig. 1). The three-dimensional structure of the gap junction channel has been elucidated by electron crystallography (Unger et al., 1999). Each connexon is formed by the assembly of six identical protein subunits, termed connexins (Cxs). Cxs are integral membrane proteins containing four transmembrane, two extracellular, and two intracellular domains. The extracellular regions are responsible for the interaction between connexon on adjacent cells. The intracellular domains, which contain both the carboxy and amino terminal, are targets of post translational modifications that are important in the regulation of channel activity. These channels are grouped in very extensive areas of the membrane forming plaque-like clusters as visualized by atomic force microscopy (Fig. 2) (Hoh et al., 1983).

Figure 1.

Diagram of gap junction channels.

Figure 2.

Gap junction plaques. Atomic force microscope image of the extracellular surface of an isolated rat hepatic gap junction in phosphate buffered saline. Image size is 170 nm × 170 nm, courtesy of Jan Hoh.

Cxs have been classified according to their molecular weight and they have been named accordingly; i.e., Cx43, Cx37, Cx26 (Beyer et al., 1988; Bruzzone et al., 1996; Goodenough et al., 1996). Alternatively, they have been grouped based on their molecular homology; i.e., α1, β1, β2 (Kumar and Gilula, 1996). Expression of Cx genes is tissue specific. Several Cxs are present within the same organ or cell type. For example, Cx43 (α1) is the most abundant Cx in the heart as compared to Cx40 (α5) and Cx45 (α6), which are expressed at lower levels in this organ. In addition to myocytes, Cx43 is present in epithelial and endothelial cells of different tissues (Kumar and Gilula, 1986). In liver, Cx32 (β1) and Cx26 (β2) are present within the same hepatocyte. In rodents, their distribution in the hepatic acinus is different. Cx32, the most prominent, is homogeneously distributed throughout the liver, whereas Cx26 is more abundant within the periportal area with respect to the pericentral region (Paul, 1986; Gumucio, 1989). The deletion of Cx genes by homologous recombination has provided new information about the role of gap junctional communication in several organ systems. In addition, the identification of human mutations in several Cx genes that functionally affect cell coupling has contributed to a better understanding of gap junction biology. A brief description of major phenotypes observed in Cx knock-out mice and human mutations is presented in Table 1. The more extensively studied mutation of Cx genes is on human Cx32, which results in the development of Charcot-Marie-Tooth (X-type) diseases, a neuropathy condition (Nicholson et al., 1999; White and Paul, 1999).

Table 1. Phenotypes of Cx-deficient mice or human mutations
ConnexinKnock-out and mouse mutationsHuman mutations
  1. (1), (Gabriel et al., 1998); (2), (Kelsell et al., 2001a,b); (3), (Willecke et al., 1999); (4), (Nelles et al., 1996); (5), (Temme et al., 1997); (6), (Nicholson et al., 1999); (7), (Deans et al., 2001); (8), (Hormuzdi et al., 2001); (9), (Guldenagel et al., 2001); (10), (Belluardo et al., 1999); (11), (Goodenough et al., 1999); (12), (Simon and Goodenough, 1998); (13), (Kirchhoff et al., 1998); (14), (Reasume et al., 1995); (15), (Lecanda et al., 2000); (16), (Kurutovskikh and Yamasaki, 2000); (17), (Kruger et al., 2000); (18), (Kumai et al., 2000); (19), (Gong et al., 1997); (20), (White et al., 1998).

Cx26Death at the embryonic day 11 due to decreased transplacental transport of glucose analogous (1)Hearing loss
  Disorders of epidermal keratinization (2)
Cx30 Dominant hearing loss, Clouston's hydroticectodermal dysplasia, skin diseases (palmoplantar keratoderma), hair loss, nail defects, and often mental deficiency (2)
Cx30.3 Dominant: skin diseases (Erthrokeratoderma variabilis) (2)
Cx31Transient placental dysmorphogenesis (2)Recessive: nonsyndromic moderate-profound sensorineural hearing loss, dominant nonsyndromic high frequency hearing loss, skin diseases (Erthrokeratoderma variabilis) and neuropathies (2)
Cx32Normal appearance and behavior. Peripheral premature neuropathy. Lower hepatic glucose mobilization in response to electrical stimulation. High incidence of spontaneous or chemically-induced liver tumors (3, 4, 5)Type X of Charcot-Marie-Tooth characterized by myelin disruption and axonal degeneration of peripheral nerves (6)
Cx36Weaker and greatly restricts spatially synchronous activity of inhibitory networks in neocortex. Deficient in early visual transmission (7, 8, 9)Linked to a juvenile form of myoclonic epilepsy and an inherited abnormality in sensory responses associated with predisposition to schizophrenia (10)
Cx37Infertility of females (11) 
Cx40Major alterations of heart function 
 Development of arrhythmias dependent on the genetic background (12, 13) 
Cx43Death at the perinatal period. Malformation of the conus region overlying the pulmonary outflow tract. Failure at the pulmonary gas exchange. Heart embryonic alteration. Delayed ossification and osteoblast dysfunction (14, 15)Visceroatrail heterotaxia (16)
Cx45Death around embryonic day 10 due to heart failure. Impairment of epithelial-mesenchymal transformation of the cardiac endothelium 
 Development of abnormal vasculature, capillary formation in the labyrinthine part placenta, and arrest of arterial growth. Failure to develop a smooth muscle layer surrounding the major arteries of the embryo proper (17, 18) 
Cx46Development of senile-cataracts after 3 weeks of age due to aberrant proteolysis of crystalline proteins. Incidence of microphtalmia (19)Autosomal dominant congenital zonular pulverulent cataracts on human. (2)
Cx50Development of cataracts during the first week after birth (20)Congenital autosomal dominant zonular pulverulent cataracts (2)


The transfer of ions and small molecules through Gj is mediated by passive diffusion. Gj are permeable to organic ions and molecules of a size less than 1 or 1.2 kDa with a maximal diameter of approximately 1.5 nm. In addition, transfer selectivity is modulated by the charge of the passing molecule or ion (Bennett et al., 1991). Studies in vitro have shown that selectivity and permeability of the Gj depend on the type of Cx forming the channel (Elfgang et al., 1995; Veenestra et al., 1995; Veenestra, 1996; Bevans et al., 1998; Niessen et al., 2000). Tracers of different molecular weights and charges, such as Lucifer yellow (457 Da, − 2), propidium iodide (668 Da, + 2), and ethidium bromide (394 Da, + 1), have been used to evaluate the selectivity and permeability of the gap junction composed of different Cxs in vivo. For example, HeLa cells transfected with Cx32 gene have been observed to transfer Lucifer yellow, but not propidium iodide or ethidium bromide. The same cells transfected with Cx26 genes allow the passage of all these dyes (Elfgang et al., 1995). In heterologous systems obtained after cell transfection or reconstitution into liposomes, Gj composed of different Cxs displayed different selectivity and permeability to second messengers, such as inositol 1,4,5-trisphosphate, and cyclic nucleotides (Bevans et al., 1998; Niessen et al., 2000). Moreover, hemichannels made of two different Cxs (heteromeric) showed different conductances than the respective homomeric channels (Koval et al., 1995; Berthoud et al., 2000). Thus, differences in the selectivity for second messengers or conductance of Gj may play an important role in organ function. These observations may explain the large variety of Cx genes and their differential tissue distribution. The passage of ions and molecules across Gj is also regulated by gating mechanisms, which are modulated by voltage differences, reduction in the cytosolic pH, and increases in intracellular calcium concentrations (Spray et al., 1985; Bennett et al., 1991). In addition, phosphorylation of the intracellular domains alters the gating activity of most Cxs (Bennett et al., 1991; Bruzzone et al., 1996; Goodenough et al., 1996). Finally, changes in the cellular content of Cxs, which depends on gene expression, also affect the communication between cells (Bennett et al., 1991).


The genomic organization of Cx genes is well conserved. Cxs are encoded by single gene copies, which are localized on several chromosomes. Presently, 17 different Cx genes have been cloned from mice. These genes have been re-named Gj (Table 2). Each Cx gene contains two exons separated by a large intron. The first exon contains a small portion of the 5′ untranslated region of the message and the second exon includes the rest of the untranslated regions as well as the open reading frame. In contrast with the similarities at the level of genomic organization, the expression of Cx genes is more complex, which is regulated at different levels. In addition, Cx gene expression is tissue specific. The first level of Cx gene regulation is at the level of transcription. Promoters of Cx26 (Jin et al., 1998; Tu et al., 2001), Cx31 (Gabriel et al., 2001), Cx32 (Bai et al., 1993, 1995; Neuhaus et al., 1995, 1996; Piechocki et al., 2000), Cx37 (Seul and Beyer, 2000), Cx40 (Bierhuizen et al., 2000), and Cx43 (De Leon et al., 1994; Yu et al., 1994; Chen et al., 1995; Fernandez-Cobo et al., 1999, 2001), have been characterized. In most cases, the basal promoter activity is located within 300 bp upstream of the transcriptional initiation site. An Sp-1 binding site has been identified in several Cx genes, which appears to be a common important element in the basal transcriptional activity of several Cx genes, such as Cx26 (Jin et al., 1998), Cx32 (Bai et al., 1993; Piechocki et al., 2000), Cx40 (Bierhuizen et al., 2000), and Cx43 (Echetebu et al., 1999; Fernandez-Cobo et al., 2001). However, the presence of this element cannot explain the tissue specific expression of different Cx genes. Consequently, additional elements within the promoter region should be responsible for the exclusive regulation of Cx gene expression in different cell types. A promoter activity within the intron of some Cx genes has also been reported. Thus, two different Cx32 transcripts of different sizes and cell expression pattern have been detected (Bai et al., 1995; Neuhaus et al., 1995, 1996; Duga et al., 1999). A longer Cx32 transcript containing both exons, which is the expressed product of a promoter activity within the 5′ flanking region of the first exon, has been detected in liver (Neuhaus et al., 1995, 1996; Sohl et al., 2001). A shorter transcript, which lacks the first exon, is the result of a promoter activity within the intron. This shorter transcript has been detected in cell of the nervous system, such as Schwann cells (Neuhaus et al., 1995, 1996; Sohl et al., 1996). The transcriptional activity of Cx31 was also found to be modulated by parts of the intron (Gabriel et al., 2001).

Table 2. Chromosomal localization of Cx genes
Protein nameGene nameMouse chromosomal localization (× CM)Human chromosomal localization
  1. Information used to prepare this Table can be found on

Cx26Gjb214 (21)13 (q11–q12.1)
Cx29Gje1-pending5 (syntenic) 
Cx30Gjb614 (syntenic)13 (q12–q12)
Cx30.3Gjb44 (57.5)1 (p35.1–p35.1)
Cx31Gjb34 (syntenic)1 (p35.1–p35.1)
Cx31.1Gjb54 (57.5)1 (p35.1–p35.1)
Cx32Gjb1X (41)X (q13.1–q13.1)
Cx33Gja6X (68) 
Cx36Gja92 (F3)15(q14)
Cx37Gja44 (57.6)1 (p35.1–p35.1)
Cx40Gja53 (45.6)1 (p36–q12)
Cx43Gja110 (29)6 (q21–q23.2)
Cx45Gja711 (syntenic) 
Cx46Gja314 (21)13 (q11–q12.1)
Cx47Gja12-pending11 (34) 
Cx50Gja83 (52.7)1 (pter–qter)
Cx57Gja104 (16.1) 

The expression of Cx43 is modulated by different factors, which seem to act at the transcriptional level. Basic fibroblast growth factor (FGF) has been shown to stimulate Cx43 expression in cardiac fibroblasts (Doble and Kardami, 1995). Parathyroid hormone and thyroid hormone receptors have also been reported to regulate transcription of Cx43 gene (Stock and Sies, 2000; Mitchell et al., 2001). Sex-steroids have also been implicated in the increase of transcription of Cx43 in myometrium during parturition (Garfield et al., 1977; Dahl and Berger, 1978). Estrogen was reported to upregulate Gj in the myometrium, whereas progesterone had the opposite effect (Dahl and Berger, 1978; Burghardt et al., 1984). Analysis of the rat Cx43 promoter has revealed a putative progesterone/glucocorticoid recognition element that has been postulated to function as a negative regulatory element since the uterus lacks Gj under progesterone control (Yu et al., 1994). Sex steroids have also been shown to differentially modulate the expression of Cx26 and Cx43 in rat endometrium (Grummer et al., 1999). The expression of Cx43 genes is also modified by inflammatory factors, such as bacterial lipopolysaccharide (LPS) (Fernandez-Cobo et al., 1998), tumor necrosis factor-α (TNF-α) (Fernandez-Cobo et al., 1999), and interleukin 1β (Fernandez-Cobo and De Maio, unpublished communications). The precise mechanism of such regulation is not known. An AP-1 site present in the promoter of Cx43 is a good candidate for cytokine regulation. A transient increase in the expression of c-Fos and c-Jun has been correlated with an increase of Cx43 expression in primary myometrial cells treated with phorbol ester (Geimonen et al., 1996). In addition, c-Jun has been reported to contribute substantially to Cx43 transcription activity in the myometrium during labor (Echetebu et al., 1999). The importance of the AP-1 site on Cx43 transcription was demonstrated by co-transfection of NRK cells with a Cx43 promoter coupled to human growth hormone (as reporter gene), c-fos and c-jun genes. Considerable Cx43 promoter activity was observed in the presence of c-fos and c-jun, which was abolished by deletion of the AP-1 site on the promoter (Fig. 3). In spite of these observations, the activation of AP-1 and consequently increase of Cx43 transcription by inflammatory mediators remains to be demonstrated.

Figure 3.

Cx43 promoter activity after transfection in NRK cells. NRK cells (50% confluent monolayers) were co-transfected with pOPRSVI-CAT and combinations of c-fos and c-jun, and different Cx43 promoter constructs. pCx43-224, containing 224 bp upstream of the transcription initiation site; pCx43-224ΔAP, promoter with a deletion of the AP-1 site; pCx43-TATA, TATA box of Cx43 promoter. After 48 h of the transfection, the extracellular medium was collected and cell were lysed. The level hGH in the extracellular medium was measured using a commercial radioimmunoassay and the levels of CAT was analyzed by an ELISA assay in the cell lysate (Fernandez-Cobo et al., 1998). Protein levels were also determined in the cell lysate. The Cx43 promoter activity was expressed as the ratio between the specific activity of hGH and CAT. Notice the high Cx43 promoter activity after co-transfection with c-fos and c-jun in comparison with the absence of these genes. This activity was diminished if the AP-1 site was deleted.

Stability of Cx mRNAs has also been found to contribute to gene expression regulation. The best-documented case is for Cx32 in the liver. Transcription of this gene is relatively low under normal conditions as detected by nuclear run off analysis. On the contrary, Cx32 mRNA is very abundant in this organ suggesting that the message is very stable (Gingalewski et al., 1996b). The stability of Cx32 mRNA is decreased during inflammatory conditions and injury (Gingalewski et al., 1996b; Gingalewski and De Maio, 1997). This change of Cx32 mRNA stability resulted in a reduction of translation and consequently in the disappearance of the protein from the hepatocyte surface (Gingalewski et al., 1996b; Gingalewski and De Maio, 1997). In fact, hepatocytes showed reduced coupling during inflammation (De Maio et al., 2000). Although the mechanism for this accelerated degradation of Cx32 mRNA is not known, the size of the poly(A) tail seems to be a marker of the decay process. Moreover, the length of Cx32 mRNA poly(A) tail is an apparent marker for the message age. Newly transcribed Cx32 mRNAs have long poly(A) tails, whereas older messages display short tails (Theodorakis and De Maio, 1999). Cx32 expression could also be regulated at the level of translation. An internal ribosome entry site has been identified in the 5′ untranslated region of Cx32 mRNA. This Cx32 internal ribosome entry site is important for the selective expression of this gene in cells of the nervous system, such as Schwann cells (Hudder and Werner, 2000). An internal ribosome entry site has also been identified in Cx43 mRNA, which has been implicated in translational control (Schiavi et al., 1999). Moreover, stability of Cx43 mRNA is modulated by the intron of this gene (Gabriel et al., 2001).


The biosynthesis of Gj is a complex process. Cxs, like other membrane proteins, seem to be inserted cotranslationally into the endoplasmic reticulum (ER) membrane. The folding of these polypeptides that span the membrane several times is convoluted and probably requires the assistance of molecular chaperones, such as heat shock proteins (De Maio, 1999) and other ER chaperones (Ellgaard and Helenius, 2001). Proper transmembrane topology of Cx43 and Cx32 has been achieved using an in vitro translation system supplemented with microsomes (Falk and Gilula, 1998). The site of connexon assembly is controversial. The ER (Kumar et al., 1995; George et al., 1999) and/or trans-Golgi network (Musil and Goodenough, 1993; Koval et al., 1997) have been postulated as possible places for oligomerization. It has been proposed that the site of connexon assembly depends on the type of Cx involved (Yeager and Nicholson, 1996). Finally, the successfully assembled connexons are further transported to the cell surface, where they form large plaque-like structures (Bruzzone et al., 1996), which are exempted of other membrane polypeptides (Fig. 2). These compact gap junction regions provide a dense area of coupling producing considerable electric currents or metabolic gradients necessary to modify cellular function. The mechanism involved in the targeting of Cxs to a particular cell surface domain is still unclear.

An additional feature of almost all Cxs is their apparent rapid turnover rate (Saffitz et al., 2000), which is also dependent on the particular Cx. Thus, a short half-life (1–2 h) has been reported for Cx43 (Laird et al., 1991; Chen et al., 1995; Darrow et al., 1995; Beardslee et al., 1998) and Cx45 (Darrow et al., 1995), intermediate (3–12 h) for Cx32 (Fallon and Goodenough, 1981; Traub et al., 1983; Kren et al., 1993; De Maio et al., 2000) and Cx26 (Traub et al., 1987), and long (over 24 h) for Cx46 (Jiang et al., 1993). This rapid turnover rate adds a new level of regulation for gap junctional cellular communication. On the cell surface, connexons are apparently dynamic structures, which are removed with high frequency (Laird, 1996). Studies using a Cx43-green fluorescent fusion protein have shown that many gap junction plaques were relatively immobile and others moved laterally on the cell surface. Moreover, Cx43 was transported in highly mobile intermediates of two different sizes, which may represent distinct traffic pathways (Jordan et al., 1999). Electron microscopy evidence suggests that whole Gj are internalized, forming a vesicle within several channels, which have been named annular gap junction (Laird, 1996). These structures are abundant after cellular dissociation, ischemia and anoxia (Traub et al., 1983; Luke and Saffitz, 1991). Several studies revealed different pathways for the degradations of Gj including lysosomes and proteasomes (Laing and Beyer, 1995; Yeager and Nicholson, 1996; Musil et al., 2000). Following hepatic ischemia reperfusion, Cx32 disappears from the cell surface, and it was visualized in intracellular vesicles that contain lysosomal protein markers (Gingalewski and De Maio, unpublished communications). Moreover, lysosomal inhibitors have been shown to block the degradation of Cx43 (Laing et al., 1997). Additionally, ubiquitination and proteasome degradation of Cx43 has been reported (Laing and Beyer, 1995; Laing et al., 1997). Proteolysis of Cx43 could also be augmented by heat stress and temperature inducible factors, such as Hsp70, which may protect Cx43 from degradation (Laing et al., 1998). The proteasomal pathway seems involved in degradation of unfolded Cxs within the ER (VanSlyke et al., 2000). It is possible that Cxs are synthesized in excess because the number of subunits that can be assembled within a connexon is limited. This excess of Cxs may be degraded before reaching the cell surface. In summary, these observations suggest that Cxs may be degraded by alternative mechanisms. It could be that unfolded or misfolded Cxs as well as those synthesized in excess are degraded by the proteasomes system, whereas those forming connexons on the cell surface are disposed within lysosomes.

Hemichannels that are not coupled to another connexon have been detected on the cell surface, particularly outside the plaque region (Musil and Goodenough, 1991). These “unapposed” hemichannels may play an alternative role to gap junction communication. Under normal physiological conditions, these “unapposed” hemichannels are probably closed to prevent metabolic stress and death caused by the collapse of ionic gradients, loss of small metabolites, and influx of Ca+2 (Bennett et al., 1991; Contreras et al., 2002). Recent evidence suggests that these “unapposed” hemichannels are involved in cellular responses like the release of cytosolic components, such as NAD+ and ATP (Cotrina et al., 1998; Bruzzone et al., 2001), regulation of cell volume (Quist et al., 2000), and retina function (Kamermans et al., 2001).


The role of gap junctional cellular communication during injury is a relatively new turn in this field. This review focuses on work directed to understand the involvement of gap junction communication in the response to injury in three major organs: liver, heart, and brain, which have been the most extensively studied. The information that is presented does not pretend to be an exhaustive revision of the current status, but rather a mechanistic discussion.


Gap junction cellular communication is important in the propagation of signals that regulate hepatic metabolism (Hooper and Subak-Sharpe, 1981). Parenchymal cells (hepatocytes) express two different Cxs (Cx32 and Cx26). In rodents, Cx32 is homogeneously expressed in hepatocytes, whereas the level of Cx26 varies between hepatocytes within the portal (high expression) and pericentral (low expression) regions (Paul, 1986; Gumucio, 1989). The gradient of Cx26 expression is associated with differences in oxygen concentrations and metabolic activity within the hepatic acinus. Homotypic Cx32 and Cx26 Gj as well as heterotypic Cx26/Cx32 channels have been detected in hepatocytes (Valiunas et al., 1999). Nonparenchymal cells, such as sinusoidal endothelial cells, biliary epithelial cells, Kupffer cells, and stella (Ito) cells, express Cx43 abundantly (Neveu et al., 1995; De Feijter et al., 1996). The expression of different Cxs within hepatocytes may have an impact on the permeability of secondary messenger molecules, such as inositol 1,4,5-triphosphate and calcium ions (Niessen et al., 2000). The response to hormones, such as adrenaline and vasopressin, acting on the inositol phosphate pathway has been shown to propagate a calcium wave from the periportal to the perivenous area of the liver lobule. In addition, calcium ions propagate better between hepatocytes, which express mainly Cx32 (Nathanson and Burgstahler, 1992). Mice deficient in Cx32 have reduced glucose mobilization from hepatic glycogen stores after the electric stimulation of sympathetic nerves (Nelles et al., 1996). Additionally, the glycogenolytic hormone norepinephrine was found to induce a lower response, measured as glucose mobilized from glycogen stores, in Cx32 deficient mice as compared to wild type (Stumpel et al., 1998).

Expression of Cx genes within the liver is also affected during different types of injury. Cx32 expression was reduced after hepatic regional ischemia/reperfusion. In this case, blood flow through the left and median liver lobes were reversibly blocked (ischemic liver), while maintaining circulation to the rest of the organ (nonischemic). In the ischemic area, there was a simultaneous reduction in Cx32 mRNA and the encoding polypeptide from the plasma membrane within 1 h of reperfusion, with no effects during the ischemia period. On the contrary, Cx32 mRNA was reduced after 4 h of reperfusion in the nonischemic liver, which was followed by the disappearance of cell surface Cx32 within 24 h of the insult (Gingalewski and De Maio, 1997). The decrease in Cx32 expression within the ischemic liver is probably the result of two independent processes, accelerated degradation of Cx32 mRNA, and disappearance of the polypeptide from the cell surface. The latter may be due to the local effect of oxygen radicals that are produced during the reperfusion process. The effect of oxygen radicals on the downregulation of Cx32 has been suggested in the case of hepatic chemical injury (Saez et al., 1987). Additionally, protein synthesis is blocked in the ischemic area during ischemia/reperfusion (Gingalewski et al., 1996a). In the nonischemic area, the disappearance of Cx32 is probably due to the inflammatory process induced by this insult as proposed by other proinflammatory stimuli (see below).

Hepatic cholestasis (bile duct ligation) also resulted in a decrease in the expression of Cx32 (Traub et al., 1983; Fallon et al., 1995) and Cx26 (Fallon et al., 1995). Cx26 was shown to return to basal levels more rapidly than Cx32 after this injury (Fallon et al., 1995). In addition, vasopressin-induced calcium waves were significantly impaired after 1 day of cholestasis as compared with controls, probably due to reduced hepatic gap junction communication secondary to the decrease of Cx levels (Fallon et al., 1995). Other chronic liver diseases, such as viral induced-hepatitis and cirrhosis, have been shown to result in a significant decrease of Cx32 levels (Yamaoka et al., 2000). A reduction of Cx32 hepatic expression was similarly reduced after administration of several toxic chemicals that produced an irreversible liver damage, such as carbon tetrachloride (Saez et al., 1987; Miyashita et al., 1991), dimethylnitrosamine (Miyashita et al., 1991), and 1,1,1-trichloro-2,2-bis(p-chlorophenyl)ethane 1,1,1-trichloro-2,2-bis(p-chlorophenyl)ethane (Tateno et al., 1994). The effect of these chemicals may be due to the generation of free radicals and oxidative stress (Saez et al., 1987). In support of this hypothesis, liver perfusion with H2O2 resulted in a rapid degradation of Cx32 without changes in Cx32 mRNA levels (Fukuda et al., 2000). In addition, a direct correlation between antioxidant pretreatment and preservation of gap junction communication has been made in a model of chemical-induced hepatic injury (Kojima et al., 1996). Administration of cadmium, which results in an acute hepatic injury characterized by apoptosis and necrosis (Habeebu et al., 1998), also induced a decrease of Cx26 and Cx32 levels in a dose-dependent manner (Jeong et al., 2000). A reduction of hepatic Cxs was observed in regenerating liver after partial hepatectomy (Traub et al., 1983; Kren et al., 1993). This effect could be due to the inflammatory process triggered by tissue trauma during partial hepatectomy. Alternatively, Cx32 expression could be reduced as part of the proliferative process, which may not require the presence of Gj. Indeed, Cx32-deficient mice showed an elevated rate of hepatocyte proliferation in vivo as well as ex vivo conditions (Neveu et al., 1995). Interestingly, the frequency of spontaneous- and chemically-induced liver tumors was increased in the Cx32 deficient mice as compared with the wild type rodent (Temme et al., 1997). A reduction of Cx32 levels has also been observed during ex vivo culture of primary hepatocytes (Saez et al., 1989), which may be related to a differentiation process. Finally, Cx32 expression is reduced during malignant transformation (Yamasaki, 1991).

The inflammatory process induced by administration of LPS resulted in a reduction of the intrahepatic cellular communication. This decrease in cellular communication was associated with the disappearance of Cx32 and Cx26 from the hepatocyte plasma membrane (De Maio et al., 2000). This downregulation of Cx32 and Cx26 is probably the result of a rapid degradation of these polypeptides combined with a reduction in the synthesis of these polypeptides. In addition, Cx32 mRNA levels were reduced in response to LPS (Gingalewski et al., 1996b). Thus, it has been shown that Cx gene expression during inflammatory conditions is regulated at the level of mRNA stability (Gingalewski et al., 1996b; Theodorakis and De Maio, 1999). On the other hand, Cx26 mRNA levels were unaffected or even transiently increased, upon LPS stimulation without significant increase in the polypeptide level (De Maio et al., 2000; Temme et al., 2000). Studies on immortalized mouse hepatocytes have corroborated the effect of inflammatory molecules on the expression of Cx32 (Temme et al., 1998). Cx32 mRNA levels were also reduced during sepsis induced by cecal ligation and puncture (Gingalewski and De Maio, unpublished communications). Thus, hepatic Cx32 is very sensitive to inflammation independent of the stimuli. Interestingly, an increase of Cx43 expression in the liver was observed simultaneously with the decrease of hepatic Cx32 after cecal ligation and puncture, which could also be observed after administration of LPS (Gingalewski and De Maio, unpublished communications). It is possible that this increase of Cx43 expression occurs in nonparenchymal cells, such as endothelial and Kupffer cells. In fact, an increase of basal Cx43 mRNA levels has been observed in primary culture of endothelial cells incubated with LPS. In other organs, such as kidney and lung, an increase of Cx43 expression has been reported during inflammation (Fernandez-Cobo et al., 1998). The hepatic inflammatory process as assessed by expression of acute phase genes was not altered in Cx32 knock-out mice (Temme et al., 1998).


Gj play an essential role in normal heart function, mediating the spread of electrical impulses that stimulate the synchronized contraction of the cardiac chambers. In addition, Gj contribute to the coordination of activities between cells of the heart wall. Thus, the propagation of cardiac action potential depends on Gj (Page, 1992). Cx43 is expressed at high levels in the heart, whereas Cx40 and Cx45 are less abundant (Jalife et al., 1999). Cx37 is only expressed in cardiac endothelium (Vozzi et al., 1999). During heart failure, a depressed conduction of cardiac action potential has been associated with a decrease in Cx43 expression (Chen and Jones, 2000). In advanced stages of heart disease, reduction in Cx expression and intercellular coupling in conjunction with gap junction redistribution have been observed. These changes have been strongly implicated in the pathogenesis of lethal ventricular arrhythmias (Kanno and Saffitz, 2001). Downregulation of Cx43 and Cx45, measured as mRNA and protein levels, were observed in end-stage heart failure. On the contrary, a significant increase of Cx40 without changes of Cx37 levels were detected in this pathological condition (Dupont et al., 2001). A significant increase in the levels of phosphorylated Cx43, which is associated with a decrease in the propagation of calcium waves, has been observed during cardiac dysfunction (Toyofuku et al., 1999). A decrease in cardiac gap junction cellular communication has also been observed in post-ischemic myocardium (Kieval et al., 1992). Morphological studies have shown that the density of Gj in the ventricular myocardium was reduced in the heart after ischemia or chronic hypertrophy as compared to regions distant from the injured area (Peters et al., 1993). Changes in gap junction distribution in the infarct border zones seem to contribute to reentrant arrhythmias (Luke and Saffitz, 1991). A decline of electrical uncoupling during acute myocardial ischemia has been correlated with dephosphorylation and cellular redistribution of Cx43 as well as with an increase of intracellular Ca2+ and H+ concentrations and accumulation of amphipathic lipid metabolites. In addition, a reduction of myocardial coupling after ischemia has been associated with Cx43 desphosphorylation within Gj and the translocation of this protein from the cell surface to intracellular pools (Beardslee et al., 2000; Lerner et al., 2000).

Cx43 mRNA levels were dramatically and rapidly reduced in rat heart after injection of LPS. This decrease of Cx43 mRNA levels was apparently due to reduction in the transcription of this gene. Moreover, administration of actinomycin D resulted in a rapid decrease of Cx43 mRNA levels suggesting that this message has a short half-life (Fernandez-Cobo et al., 1999). The half-life of Cx43 (polypeptide) in cardiomyocytes is also short, in the order of 1–2 h (Saffitz et al., 2000). Thus, a high rate of Cx43 transcription and translation must be required to maintain normal levels of this protein within the heart. Heart Cx43 mRNA levels were also reduced following hepatic ischemia/reperfusion (Fernandez-Cobo et al., 1999). This observation suggests that the factor modulating the changes in Cx43 expression within the heart during injury is a circulating inflammatory mediator. Indeed, transfection of Cx43 promoter coupled to a reporter gene into a rat myoblast cell-line (H9c2) showed regulation of transcriptional activity by proinflammatory mediators, particularly TNF-α (Fernandez-Cobo et al., 1999). Circulating levels of TNF-α has been observed after many forms of cardiac injury, and has been implicated in some cases of heart failure (Odeh, 1993; Mann, 1996; Meldrum, 1998). TNF-α has been detected in the right atrial tissues of patients with severe heart failure (Sasayama et al., 1996). Alterations in myocardial contractility have been associated with the presence of TNF-α (Odeh, 1993; Mann, 1996; Meldrum, 1998). Direct administration of TNF-α to animals resulted in hypotension, metabolic acidosis, hemoconcentration, and death resembling the symptoms of septic shock (Tracey et al., 1986; Odeh, 1993). Indeed, several investigations have reported a depression of myocardial performance during septic and endotoxic shock (Abel, 1989). Thus, cardiac dysfunction during inflammatory conditions may be due to the effect of TNF-α on the expression of Cx43.

Cx37, Cx40, and Cx43 have been identified in endothelial and smooth muscle cells (SMCs) (Cai et al., 2001). In the vascular system, gap junction cellular communication is involved in the modulation of the vasomotor tone and in the maintenance of the homeostasis (Dora, 2001). Gap junction coupling has been implicated in a number of developmental and pathogenic processes of endothelial cells, such as angiogenesis, migration, and reparation after an injury (Christ et al., 1996). For example, a different pattern of Cx expression in the regenerating endothelium has been shown, which is characterized by an increase in Cx40 expression (Yeh et al., 200l). Differential expression of Cxs in both endothelial and SMCs were involved in the development of arteriosclerosis (Blackburn et al., 1995). In addition, proper Cx expression was required for coordinated migration during repair of an endothelial cell wound (Kwak et al., 2000).


Gj allow the electrical coupling between neurons and may regulate and/or synchronize neuron network activity (Bennett, 1997). Studies supporting this hypothesis show that Gj between neurons synchronize the inhibitory activity within neocortex (Beierlein et al., 2000; Tamas et al., 2000). Cx36 is the only one that has been unequivocally identified at the ultrastructural level within plaques of coupled neurons in adult rat brain and spinal cord (Rash et al., 2000, 2001). Although Cx32 has also been detected in neurons at the mRNA and protein level (Shiosaka et al., 1989; Nadarajah et al., 1997; Venance et al., 2000; Oguro et al., 2001), it has not been localized within plaques yet. Cx36 knock-out mice showed a weaker and greatly restricted spatially synchronous activity of inhibitory networks in the neocortex suggesting an essential role for this Cx (Deans et al., 2001; Hormuzdi et al., 2001). Coupling between neurons and glia cells has been characterized at the electrophysiological levels (Alvarez-Maubecin et al., 2000). However, the detection of any Cxs between these two cells has been elusive.

Gap junction communications have been associated with abnormal hypersynchrony of adjacent neuronal spike firing, which seems to be important in epilepsic seizures (Perez-Velazquez and Carlen, 2000). Moreover, gap junction blockers or activators alter the spontaneous synchronized events within CA1 area of the hippocampus (Perez-Velazquez et al., 1994). Conversely, downregulation of Cx36 expression in rat hippocampus was observed in an experimental epilepsy model induced by administration of kainite, which apparently results in a reduction of coupling between neurons (Sohl et al., 2000). Human Cx36 gene has been localized on a locus linked to juvenile form of myoclonic epilepsy and an inherited predisposition to schizophrenia (Belluardo et al., 1999). Selective increase of Cx32 and Cx36 expression without alterations of Cx43 levels was observed in hippocampus CA1 zone after global brain ischemia in mice. Additionally, this increase of Cx32 and Cx36 expression within CA1 zone occurred specifically in inhibitory interneurons, which form an extended network within the hippocampus. In contrast, no change in the expression of any Cx was detected in CA3 and dentate gyro zones of the hippocampus after an identical injury (Oguro et al., 2001). The increase of Cx32 and Cx36 expression in CA1 interneurons after ischemia has been associated with tolerance to subsequent ischemic episodes (Oguro et al., 2001). Moreover, Cx32 knock-out mice showed an enhanced sensitivity to cerebral ischemia (Oguro et al., 2001). These observations suggest an important role of Gj in neuron protection from global brain ischemia.

Astrocytes are essential to maintain extracellular homeostasis in the brain, contributing to neuron survival and function. For example, elevated concentrations of K+ and glutamate produced during normal brain activity, which are toxic for neurons, are cleared and diluted via Gj between astrocytes (Chen and Nicholson, 2000; Hansson et al., 2000). In addition, astroglia Gj mediate the propagation of Ca2+ waves, which seem important for the regulation of neuronal activity (Nedergaard, 1994). In the brain, astrocytes express Cx43 and low levels of Cx30, Cx40, and Cx45 (Nagy et al., 1997, 1999; Dermietzel et al., 2000). Among these Cxs, only Cx43 has been found to form functional Gj in astrocytes (Dermietzel et al., 1989; Giaume et al., 1991). Indeed, astrocytes derived from Cx43 deficient mice did not form functional Gj (Naus et al., 1997). Neurotransmitters, growth factors, and cytokines regulate Cx expression and gap junction permeability in astrocytes, which have important implications in the physiology and pathology of the brain. Astroglia gap junction communication in the striatal zone of the brain was reduced by endothelin and noradrenaline (Giaume et al., 1991; Venance et al., 1998). Gap junction communication between cortical astrocytes was reduced by nitric oxide and arachidonic acid (Bolanos and Medina, 1996; Rorig and Sutor, 1996; Martínez and Sáez, 1999). Addition of interleukin 1β resulted in a decrease of Cx43 levels and gap junction communication in human fetal astrocytes (Duffy et al., 2000). Administration of epidermal growth factor was also observed to downregulate Cx43 expression in rat cortical astrocytes (Ueki et al., 2001). Members of the FGF family, which are expressed in brain, trigger a selective reduction of Cx43 expression and gap junction communication in cultured astrocytes derived from different brain areas. Thus, FGF-9 resulted in downregulation of Cx43 expression and gap junction communication in cortical, striatal, and mesencephalic; FGF-2 in cortical and striatal, but not in mesencephalic; and FGF-5 in mesencephalic, but not in astrocytes from other areas (Reuss et al., 2000). Addition of tissue growth factor-β increased coupling of astrocytes and modifies Cx43 phosphorylation pattern, whereas it has the opposite effect in glioma cells (Robe et al., 2000). Thus, growth factors and inflammatory mediators have a wide range of effects on Gj between astrocytes.

Neuronal activity also increased astrocyte coupling without changing Cx43 expression (Rouach et al., 2000). Ageing has an impact on the size and number of gap junction plaques in astrocytes, without significant changes in functional coupling (Cotrina et al., 2001). Gj between astrocytes derived from different brain zones has different tracer dye permeability and Ca2+wage propagation (Blomstrand et al., 1999). The generation of arachidonic acid by-product during hypoxia/reoxygenation was associated with a reduction in gap junction communication in cultured astrocytes (Martínez and Sáez, 1999). Hypoxia as well as addition of iodoacetic acid (a metabolic inhibitior) resulted in a decrease of astrocyte coupling in brain slices (Cotrina et al., 1998). A reduction in gap junction communication associated with desphosphorylation of Cx43 within astrocytes was also observed after hypoxia (Cotrina et al., 1998) or in the presence of metabolic inhibitors (Contreras et al., 2002). In spite of these dramatic effects, gap junction communication was never completely blocked, which may still allow the passage of toxic metabolites between astrocytes (Cotrina et al., 1998; Contreras et al., 2002). Indeed, Gj appeared to augment the spatial extent of injury in a glioma cell line during metabolic inhibition and oxidative stress (Lin et al., 1998).

The infusion of gap junction blockers limited the infarct size after focal or global brain ischemia in rodents (Rawanduzy et al., 1997; Saito et al., 1997; Rami et al., 2001). However, heterozygous Cx43 null mice showed increased infarct zone damage after global ischemia as compared with normal mice (Siushansian et al., 2001). Blockers of Gj also increased neuronal vulnerability to oxidative stress in co-cultures of astrocytes and neurons (Blanc et al., 1998). Cx43 hemichannels on the astrocyte cell surface, which were not associated to another connexon, were open as detected by dye uptake during metabolic inhibition by the addition of iodoacetic acid and antimycin A. These hemichannels have been proposed to be involved in cell death (Contreras et al., 2002). Astrocytes derived from human epileptic foci showed a significantly high cell coupling as compared with tissue from nonepileptic region (Lee et al., 1995). In addition, elevated Cx43mRNA levels were detected in seizure areas of human brains. Conversely, normal levels of Cx43 mRNA and protein were observed in brain of patients with complex partial seizure disorder (Elisevich et al., 1997). In a rodent model of epilepsy induced by kainic acid, which results in neuronal death and gliolisis, a depletion of Cx43 in the zone of astrogliosis was observed (Vukelic et al., 1991). Another study using the same animal model showed a slight reduction in Cx43 and Cx30 expression in the hippocampus (Sohl et al., 2000).

Oligodendrocytes are involved in myelination of neuronal axon in the central nervous system and in the formation of Ranvier nodes, which accelerates the nerve impulse propagation by the axon. Oligodendrocytes express Cx26, Cx32, and Cx45, although their distribution in the brain is not totally clear (Dermietzel et al., 1989; Kunzelmann et al., 1997; Li et al., 1997; Nagy et al., 2001). Studies on Cx32-deficient mice suggest that Cx32 does not seem to be primarily involved in the generation of axon myelin sheaths in the central nervous system and neuron function. These observations are in contrast with the role of Cx32 in peripheral nervous system, in which mutation of Cx32, observed in Charcot-Marie-Tooth (X-type) diseases, were involved with the demyelination of nervous fibers and consequent neuronal dysfunction (Scherer et al., 1998; Nicholson et al., 1999; Willecke et al., 1999). Recently, an axon myelination defect was observed in the neocortex of Cx32 deficient mice, which may be associated with oligodendrocyte dysfunction (Sutor et al., 2000). In addition, an inhibitory synaptic transmission defect in the neocortex was associated with the lack of Cx32 (Sutor et al., 2000). A reduction in conduction velocity of central nerves and myelination has been observed in Charcot-Marie-Tooth (X-type) disease patients (Nicholson et al., 1999). However, it is not clear whether this dysfunction is related to oligodendrocytes, neurons, or both of them. Microglias are resident macrophages of the brain and they are associated to immunological and inflammatory responses in this organ. These cells are activated during diverse pathologies, such as head injury, ischemia, and neurodegenerative diseases (Kreutzberg, 1996). Cx43 became detectable in microglias that were recruited in brain stab wounds. In addition, microglia in culture conditions formed functional Gj after incubation with interferon-γ and TNF-α or LPS (Eugenín et al., 2001). Thus, the Gj between macrophages could be involved in the synchronization of inflammatory response (Saez et al., 2000).


The preceding observations clearly indicate that gap junction cellular communication is altered during injury. The question that emerges is the importance of these changes in cellular communication after stress. An idea is that a decrease in cellular communication is necessary to isolate the affected or damage cell(s) from the rest of the healthy tissue to avoid the spread of harmful or toxic signals. Pertinent to this hypothesis is the dissemination of apoptotic signals that could trigger this deadly pathway in distant cells. The alternative hypothesis is that a reduction of gap junction communication decreases the loss of important metabolites, such as glucose or ATP, for the cell. However, this decrease of metabolite sharing may be detrimental to the organ. A final possibility is that the reduction of gap junction cellular communication is related to an overall change of gene expression after injury. Although, there is timid evidence supporting this hypothesis, a reduction in expression of constitutive genes is observed during stress simultaneously with the increase in the expression of “protective genes,” such as heat shock and acute phase proteins. This phenomenon, which has been coined “the adaptive response to stress,” is based on the assumption that the cellular capacity for gene expression is limited (Wang et al., 1995). In this order of ideas, genes that are downregulated after stress correspond to nonvital proteins in the short term because otherwise the cell commits “suicide”. It is possible that Gj are not required for short-term cell survival and can be spared during early “emergency” conditions. However, the long-term absence of Gj is lethal for the organism. Regardless of these speculations, alteration of gap junction communication is an event that has been well associated with injury and it requires future and serious investigation. The alteration of gap junction communication may be important target in the treatment of many aliments that affect human beings.


We thank Jan Hoh for providing the electron force microscopy figure, Cynthia Gingalewski and Mariana Fernandez-Cobo for allowing us to present their unpublished observations, William B. Fulton for critically reading the manuscript, and Ana De Maio for her comments.