Gap Junctions

Abstract

Gap junctions are aggregates of intercellular channels that permit direct cell–cell transfer of ions and small molecules. Initially described as low-resistance ion pathways joining excitable cells (nerve and muscle), gap junctions are found joining virtually all cells in solid tissues. Their long evolutionary history has permitted adaptation of gap-junctional intercellular communication to a variety of functions, with multiple regulatory mechanisms. Gap-junctional channels are composed of hexamers of medium-sized families of integral proteins: connexins in chordates and innexins in precordates. The functions of gap junctions have been explored by studying mutations in flies, worms, and humans, and targeted gene disruption in mice. These studies have revealed a wide diversity of function in tissue and organ biology.

Gap junctions that allow material to move between cells are composed of connexins. The choice of connexin affects the permeability and other features of these junctions.

Gap junctions are clusters of intercellular channels that allow direct diffusion of ions and small molecules between adjacent cells. The intercellular channels are formed by head-to-head docking of hexameric assemblies (connexons) of tetraspan integral membrane proteins, the connexins (Cx) (Goodenough et al. 1996). These channels cluster into polymorphic maculae or plaques containing a few to thousands of units (Fig. 1). The close membrane apposition required to allow the docking between connexons sterically excludes most other membrane proteins, leaving a narrow ∼2 nm extracellular “gap” for which the junction is named (Fig. 2). Gap junctions in prechordates are composed of innexins (Phelan et al. 1998; Phelan 2005). In chordates, connexins arose by convergent evolution (Alexopoulos et al. 2004), to expand by gene duplication (Cruciani and Mikalsen 2007) into a 21-member gene family. Three innexin-related proteins, called pannexins, have persisted in vertebrates, although it is not clear if they form intercellular channels (Panchin et al. 2000; Bruzzone et al. 2003). 7Å-resolution electron crystallographic structures of intercellular channels composed of either a carboxy-terminal truncation of Cx43 (Unger et al. 1999; Yeager and Harris 2007) or an M34A mutant of Cx26 (Oshima et al. 2007) are available. The overall pore morphologies are similar with the exception of a “plug” in the Cx26 channel pore. The density of this plug is substantively decreased by deletion of amino acids 2–7, suggesting that the amino-terminus contributes to this structure (Oshima et al. 2008). A 3.5-Å X-ray crystallographic structure has visualized the amino-terminus of Cx26 folded into the mouth of the channel without forming a plug, thought to be an image of the open channel conformation (Maeda et al. 2009). The amino-terminus has been physiologically implicated in voltage-gating of the Cx26 and Cx32 channels (Purnick et al. 2000; Oh et al. 2004), lending support to a role for the amino-terminus as a gating structure. However, Cx43 also shows voltage-gating, and its lack of any structure resembling a plug remains unresolved. A comparison of a 1985 intercellular channel structure (Makowski 1985) with the 2009 3.5Å structure (Maeda et al. 2009) summarizes a quarter-century of X-ray progress (Fig. 3).

Figure 1.

A diagram showing the multiple levels of gap junction structure. Individual connexins assemble intracellularly into hexamers, called connexons, which then traffic to the cell surface. There, they dock with connexons in an adjacent cell, assembling an axial channel spanning two plasma membranes and a narrow extracellular “gap.”

Figure 2.

Electron microscopy of gap junctions joining adjacent hepatocytes in the mouse. The gap junction (GJ) is seen as an area of close plasma membrane apposition, clearly distinct from the tight junction (TJ) joining these cells. (Inset A) A high magnification view of the gap junction revealing the 2–3 nm “gap” (white arrows) separating the plasma membranes. (Inset B) A freeze-fracture replica of a gap junction showing the characteristic particles on the protoplasmic (P) fracture face and pits on the ectoplasmic (E) fracture face. The particles and pits show considerable disorder in their packing with an average 9-nm center-to-center spacing.

Figure 3.

A comparison of axial sections through gap-junction structures deduced from X-ray diffraction. The 1985 data (Makowski 1985) were acquired from gap junctions isolated biochemically from mouse liver containing mixtures of Cx32 and Cx26. The intercellular channel (CHANNEL) is blocked at the two cytoplasmic surfaces by electron density at the channel mouths along the sixfold symmetry axis. The 2009 data (Maeda et al. 2009), acquired from three-dimensional crystals of recombinant Cx26, resolve this density at the channel opening as the amino-termini of the connexin proteins, the 2009 model possibly showing an open channel structure.

Most cells express multiple connexins. These may co-oligomerize into the same (homomeric) or mixed (heteromeric) connexons, although only certain combinations are permitted (Falk et al. 1997; Segretain and Falk 2004). A connexon may dock with an identical connexon to form a homotypic intercellular channel or with a connexon containing different connexins to form a heterotypic channel (Dedek et al. 2006). Although only some assembly combinations are permitted (White et al. 1994), the number of possible different intercellular channels formed by this 21-member family is astonishingly large. This diversity has significance because intercellular channels composed of different connexins have different physiological properties, including single-channel conductances and multiple conductance states (Takens-Kwak and Jongsma 1992), as well as permeabilities to experimental tracers (Elfgang et al. 1995) and to biologically relevant permeants (Gaunt and Subak-Sharpe 1979; Veenstra et al. 1995; Bevans et al. 1998; Gong and Nicholson 2001; Goldberg et al. 2002; Ayad et al. 2006; Harris 2007).

Opening of extrajunctional connexons in the plasma membrane, described as “hemichannel” activity, can be experimentally induced in a variety of cell types. Because first observations of hemichannel activity were in an oocyte expression system (Paul et al. 1991) and dissociated retinal horizontal cells (DeVries and Schwartz 1992), the possible functions of hemichannels composed of connexins and pannexins has enjoyed vigorous investigation (Goodenough and Paul 2003; Bennett et al. 2003; Locovei et al. 2006; Evans et al. 2006; Srinivas et al. 2007; Schenk et al. 2008; Thompson and MacVicar 2008; Anselmi et al. 2008; Goodenough and Paul 2003). Hemichannels have been implicated in various forms of paracrine signaling, for example in providing a pathway for extracellular release of ATP (Cotrina et al. 1998; Kang et al. 2008), glutamate (Ye et al. 2003), NAD⁺ (Bruzzone et al. 2000), and prostaglandins (Jiang and Cherian 2003).

GAP JUNCTIONAL INTERCELLULAR CHANNELS ARE DYNAMICALLY REGULATED

Communication via intercellular channels is regulated at multiple levels. The most rapid timescales involve changing the unitary conductance of single channels or altering their probability of opening. Slower regulation is achieved by altering the number of channels present in the membrane by changing rates of synthesis and assembly, posttranslational modification and/or protein degradation. The mechanisms of regulation can overlap between these different time frames, for example, phosphorylation is involved both in changing single channel conductance and in protein trafficking to the cell surface and degradation. The different timescales will be considered in turn.

Rapid Regulation

On the shortest time scale, it is known that gap-junction channels are gated by voltage and can display multiple voltage-dependent conductance states (Turin and Warner 1977; Spray et al. 1979; Neyton and Trautmann 1985; Chen and DeHaan 1992; Bukauskas and Weingart 1993). Voltage-gating is a common property of connexins, although they show substantive differences in their sensitivities. Voltage-gating could explain the rectifying neuronal synapses observed in crayfish (Furshpan and Potter 1959), Drosophila (Allen et al. 2006), and hatchetfish (Auerbach and Bennett 1969; Hall et al. 1985), in which action potentials are permitted to pass orthodromically but not antidromically. This behavior requires a structural asymmetry that could be most simply modeled by a heterotypic intercellular channel in which one connexon showed fast voltage-dependent closure whereas the other did not. Indeed, rectification was observed in heterotypic junctions formed between connexins expressed in paired Xenopus oocytes (Dahl et al. 1987), but the time scale was too slow to completely explain rectifying synapses (Swenson et al. 1989). In addition to rapid closure of a channel in response to postsynaptic depolarization, rectification at an electrical synapse could also be achieved by opening channels in response to presynaptic depolarization. However, it requires at least 9.5 ms to reopen a closed Cx40 channel in this manner, which is also too slow to account for synaptic rectification (Bukauskas et al. 1995). Although rectifying synapses require near-instantaneous rectification of current, somewhat slower voltage inactivation may be functional in other contexts. For example, Cx45/Cx43 heterotypic junctions may rectify fast enough to influence dendro–dendritic interactions in the central nervous system or may modulate re-entry circuits in myocardium (Bukauskas et al. 2002a).

Fast rectification has been shown using Cx32/Cx26 heterotypic channels (Oh et al. 1999). However, neither connexin displays particularly fast homotypic voltage-dependent gating and thus the rectification observed cannot be predicted from the properties of the individual channels. A model is that the asymmetry of the heterotypic channel results in a separation of fixed positive and negative charges across the two junctional membranes and that rectification of ionic currents occurs within the channel rather than resulting from voltage-induced connexin conformational changes. Regardless, Cx26 and Cx32 are not typically found in excitable cells and are unlikely to participate in rectifying synapses.

Recently, Phelan et al. 2008 have explored innexin composition and physiology of rectifying synapses in the Drosophila giant fiber system. These rectifying synapses were shown to be composed of heterotypic channels formed from two different products of the shaking-B innexin gene: Shaking-B (neural+16) and Shaking-B (lethal). The former innexin is expressed in the presynaptic neuron and the latter in postsynaptic cell. Although technical limitations did not permit direct electrophysiological measurements in vivo, the two innexins were expressed in the Xenopus paired-oocyte system that allowed the characterization of both homotypic and heterotypic innexin interactions. Homotypic intercellular channels composed of Shaking-B (lethal) were highly voltage-dependent compared with those composed of Shaking-B (neural+16). However, in neither case did homotypic channels display rectification. In contrast, Shaking-B (neural+16) and Shaking-B (lethal) assembled heterotypic junctions that rectified. Importantly, channel closure was complete within 5 ms of the application of a transjunctional voltage, and displayed the appropriate gating polarity seen in vivo. However, crayfish junctions in vivo show channel gating within 0.8–1 ms (Furshpan and Potter 1959; Giaume et al. 1987) fivefold faster than the values measured using innexin channels in paired oocytes. Because it is not known how fast channels rectify in the fly, it is not yet possible to conclude that innexin composition explains the entire phenomenon. Regardless, this study provides the first molecular in vivo model to explain part of this 40-year-old conundrum.

Other than rectification, voltage gating of gap-junction channels may not be an important mode of channel regulation in vivo (Harris 2002). However, experimental manipulation of transjunctional voltage reveals a range of conductance states that are likely stabilized by other forms of channel regulation. For example, phosphorylation may function to favor one conductance state more than another, and hence be of great importance in terms of channel selectivity. Although phosphorylation is observed in most members of the connexin family (Lampe and Lau 2000; Lampe and Lau 2004; Laird 2005), most studies have focused on Cx43, which contains 21 serine and two tyrosine residues that are targets of phosphorylation by protein kinase A (PKA), protein kinase C (PKC), p34(cdc2)/cyclin B kinase, casein kinase 1, mitogen-activated protein kinase (MAPK), and pp60 (src) kinase (review (Solan and Lampe 2005)). Phosphorylation of Cx43 changes the shape of the current voltage relationship (Moreno et al. 1994). In particular, phosphorylation of serine368 (Lampe et al. 2000) by PKC results in a ∼50% reduction in unitary conductance. The change in conductance state likely reflects significant changes in channel permeation. For example, driving intercellular channels into subconductance states with transjunctional voltage has been shown to produce a change in charge selectivity (Bukauskas et al. 2002b) or a block of intercellular cAMP and dye-transfer (Qu and Dahl 2002) with little effect on macroscopic electrical coupling. Phosphorylation effects on permeation have also been noted with Cx43 hemichannels where dephosphorylation was correlated with increased channel permeability in liposome reconstitution studies (Kim et al. 1999).

Cx45 has been shown to change its open probability in response to activation of cAMP-dependent protein kinases (van Veen et al. 2000). Activation of pp60^v-src is correlated with tyrosine phosphorylation of Cx43 and concomitant channel inactivation (Swenson et al. 1990; Lampe et al. 2000; Lampe and Lau 2000; Lin et al. 2001; for a review see Pahujaa et al. 2007), although recent studies suggest this regulation may be complex, as src activation also led to phosphorylation of MAPK and PKC sites in Cx43 (Solan and Lampe 2008).

Although many studies have shown changes in channel conductance with phosphorylation in cell culture, there are also in vivo studies documenting this role. For example, during reinitiation of meiosis by luteinizing in developing mouse ovarian follicles, Cx43 is multiply serine phosphorylated via MAPK (Norris et al. 2008), resulting in closure of gap junctional channels between mural granulosa cells, and internalization of gap junctions (Gilula et al. 1978). Another example in which connexin phosphorylation has a clear physiological relevance is in light–dark adaptation, which is globally regulated in the retina by the extrasynaptic release of dopamine (Puopolo et al. 2001). Dopamine acts on most if not all retinal neurons to adjust the gain of neural networks so that sensitivity to contrast can be maintained as the intensity of background illumination changes. In the outer retina, dopamine release rapidly and reversibly leads to a decrease in junctional coupling between horizontal cells (Lasater and Dowling 1985; DeVries and Schwartz 1989; Xin and Bloomfield 1999), which among other actions decreases the size of their receptive field (i.e., restricts the response of a given horizontal cell to a smaller number of photoreceptors), with the overall effect being an improvement in contrast sensitivity. In the inner retina, dopamine has similar effects on junctional coupling between amacrine cells, particularly the AII amacrine, which expresses Cx36 and is a critical part of the rod photoreceptor signaling pathway. D1 dopamine receptor activation in mouse AII amacrine cells leads to a PKA-mediated phosphorylation of Cx36 correlating with a decrease of dye coupling in vivo (Urschel et al. 2006). In the teleost retina, it was shown using phospho-specific antibodies that the natural stimulus of dark-adaptation dramatically increased the levels of Cx35 (the teleost ortholog of Cx36) phosphorylation (Kothmann et al. 2007). Furthermore, these phosphorylation events occurred at sites shown to regulate Cx35 channel gating using in vitro expression studies (O’Brien et al. 2004).

Slow Regulation

A slower temporal level of regulation involves connexin biosynthesis and junctional plaque assembly and turnover (Segretain and Falk 2004). Connexins can show a remarkably rapid turnover rate for a membrane protein. For example, the in vivo half-life of Cx32 in gap junctional plaques from rodent hepatocytes is less than 5 hours (Fallon and Goodenough 1981) and turnover of Cx43 in tissue culture cells is even faster (Musil and Goodenough 1991; Laird et al. 1991). Gap junctions have been shown to turn over by addition of subunits at the edges and removal of subunits from the center of plaques (Gaietta et al. 2002; Lauf et al. 2002). Accretion of connexons at the edges of pre-existing plaques could require nothing more than lateral diffusion in the plasma membrane, but it is not at all clear how the selective removal of connexins/connexons/intercellular channels from the center of a plaque might be orchestrated. Gap junctions are also removed from the cell surface by gross internalization of the entire plaque, leaving large double-membrane vesicles in the cytoplasm (Albertini and Anderson 1975; Larsen et al. 1979; Jordan et al. 2001). Studies with cultured cells suggest that internalization is a clathrin-mediated process (Piehl et al. 2007; Nickel et al. 2008). The relationship between the removal of connexins from the center of pre-existing junctional plaques and the clathrin-dependent endocytosis of whole junctional plaques remains unclear.

Gap junction assembly is associated with multiple phosphorylation steps (Musil and Goodenough 1991). Cx43 is phosphorylated soon after synthesis, and trafficking of the protein through the Golgi to the plasma membrane is accompanied by phosphorylation of specific residues, suggesting a requirement for these modifications in protein transport (Solan and Lampe 2007). Consistent with this notion, chemical or temperature blockade of trafficking in the ER or Golgi results in incomplete Cx43 phosphorylation (Musil and Goodenough 1993). Phosphorylation is also used by different connexins to both block and enhance degradation (Laird et al. 1995). For example, it has been shown that phosphorylation protects Cx32 from calpain digestion (Elvira et al. 1993), while serine phosphorylation of Cx45.6, the chick lens counterpart of Cx50, stimulates protein turnover (Yin et al. 2008). Cx43 can be degraded by both the proteosomal and lysosomal pathways, although no ubiquitin ligase has been shown to specifically associate with a connexin (Laing and Beyer 1995; Berthoud et al. 2004). Proteosome inhibitors block connexin degradation and up-regulate both gap junction assembly and intercellular dye transfer, demonstrating control of gap-junctional intercellular communication (GJIC) via the degradation pathway (Musil et al. 2000). Cx43 dephosphorylation has been associated with disassembly of gap junctions in cells treated with the gap junction blocking agent 18 β-glycyrrhetinic acid (Guan et al. 1996).

Assembly is also affected by interaction with connexin binding partners. A Cx43-interacting protein, CIP85 can induce the turnover of Cx43 through the lysosomal pathway (Lan et al. 2005). Another important interactor is ZO-1, which colocalizes with Cx43 in myocardium and links this connexin to α-spectrin in HEK293 cells in culture (Toyofuku et al. 1998). Cx43 binds to the second PDZ domain of ZO-1 (Giepmans and Moolenaar 1998). Mutations in Cx43 that alter the consensus PDZ binding domain do not inhibit the formation of gap junctions or the activity of intercellular channels. However, there is a dramatic deregulation of plaque size and abnormally large gap junctions are observed (Falk 2000; Hunter et al. 2005). The size expansion results from increased accretion of cytoplasmic pools of Cx43 connexons to the edges of existing junctional plaques and not from de novo synthesis or inhibited degradation. ZO-1 is preferentially associated with the periphery of gap junctional plaques in cells expressing Cx43, suggesting that ZO-1 is a negative regulator of accretion. It has been proposed that accretion is suppressed by a ZO-1 mediated association with filamentous actin (Hunter and Gourdie 2008). In addition, Cx43 may directly associate with tubulin (Giepmans et al. 2001), possibly explaining the observed transport of Cx43 along microtubule tracks (Lauf et al. 2002) that in turn may influence the rate or location of plaque assembly.

The myriad forms of regulation of gap junction function seem surprisingly diverse in comparison to other membrane channels. As reviewed in the following section, the multiple cellular, tissue, and organ functions that have adapted gap-junctional communication as part of their mechanisms have developed a diverse set of regulatory strategies to provide the spatial and temporal controls required in different contexts. Indeed, in some cases, the evolution of multiple connexin genes may have occurred in part because of requirements for unique mechanisms of regulation. In other cases, for example with Cx43, which is used by many different cell types in specialized contexts, multiple regulatory mechanisms are needed to provide specialized control. It is clear from this diversity that the regulation of gap junctional intercellular communication must be experimentally determined on a case-by-case basis as different mechanisms have evolved to subserve this function in different cellular contexts.

UNIVERSAL FUNCTIONS OF GAP JUNCTIONS

The ability of adjacent cells to share ions through low-resistance pathways is fundamental to the function of electrically excitable cells, such as neurons, heart, and smooth muscle. Indeed, gap junctions (electrical synapses) were first discovered in myocardium and nerve because of their properties of electrical transmission between adjacent cells (Weidmann 1952; Furshpan and Potter 1957). In these contexts, connecting cells with gap junctions provides both increased speed in synaptic transmission and the ability to synchronize groups of cells for coordinated electrical and mechanical output.

In addition to electrically excitable cells, virtually all cells in solid tissues are joined by gap junctions. A core function of GJIC is to share metabolic demands across groups of cells and thereby buffer spatial gradients of nutrients or signaling molecules. For example, targeted deletion of Cx32 in mice has been shown to result in a loss of responsiveness to sympathetic stimulation, resulting in an impaired mobilization of glucose from glycogen stores. Postganglionic sympathetic axons terminate at the edges of the liver lobules and thus can only directly stimulate a fraction of the hepatocytes. Presumably, the remainder of the lobule is stimulated indirectly by diffusion of second messengers through gap junctions (Stümpel et al. 1998). Gap junctions may also function as suppressors of somatic cell mutations so that loss of a critical metabolic enzyme or ion channel in one cell might be compensated by its neighbors. For example, Lesch-Nyhan syndrome results from impaired activity of hypoxanthine phosphoribosyltransferase (HGPRTase), a key enzyme in the nucleotide salvage pathway. Impaired HGPRTase results in an elevated concentration of phosphoribosyl pyrophosphate, a marked increase in the rate of purine biosynthesis, and an overproduction of urate. Mutant fibroblasts from patients with Lesch-Nyhan syndrome can be metabolically rescued in cell culture by gap junction formation with normal cells (Cox et al. 1970), a process termed metabolic cooperation (Subak-Sharpe et al. 1969). Furthermore, metabolic cooperation likely accounts for the lack of symptoms in heterozygous female Lesch-Nyhan carriers. As HGPRTase is located on the X chromosome, random X-inactivation results in a mosaic of mutant and normal cells. Thus, individuals are asymptomatic because of metabolic rescue of mutant cells by adjacent nonmutant cells.

SPECIALIZED FUNCTIONS REVEALED BY CONNEXIN MUTATIONS

Human Mutations

Given the long phylogenetic history of gap junctions in metazoans (Fraser and Bode 1981; Potenza et al. 2002; Starich et al. 2003; Nogi and Levin 2005), it is not surprising that this method of cell–cell communication has been adapted to subserve a wide variety of physiological functions in different cell types. Many cell- and tissue-specific functions of GJIC have been brought to light by human mutations and targeted connexin gene deletion in mice (for reviews see Simon and Goodenough 1998; White and Paul 1999; Gerido and White 2004; Dobrowolski and Willecke 2008). In humans, mutations in Cx32 underlie X-linked Charcot-Marie-Tooth syndrome, a common peripheral demyelination neuropathy (Bergoffen et al. 1993), and mutations in Cx47 result in a central demyelinating condition called Pelizaeus-Merzbacher-Like-Disease (Uhlenberg et al. 2004). More than half of all profound hereditary deafness results from mutations in Cx26, which are often syndromic and involve skin disorders (Kelsell et al. 1997; Denoyelle et al. 1997). Similarly, although usually less severe, disorders of the skin and the auditory system accompany mutations in Cx31 and Cx30 (Common et al. 2002; Abrams et al. 2006; Yang et al. 2007; Apps et al. 2007; Yum et al. 2007). Familial cataracts are commonly associated with mutations in either Cx46 or Cx50, whose expression is largely restricted to the ocular lens (Gong et al. 2007; Richard 2005; van Steensel 2004; Vreeburg et al. 2007; Mese et al. 2007). Finally, mutations in Cx43 give rise to oculodentodigital dysplasia, a pleomorphic, syndromic condition affecting a large number of cell types (Paznekas et al. 2003).

Targeted Mutations in Mice

In mice, targeted mutations of connexins have uncovered a wide variety of gap-junction functions in various organs. In many of these cases, a given connexin occupies a particular niche, supplying an essential function that is not compensated by another connexin. For example, Cx26 deletion is embryonic lethal because of a disruption of glucose transport between syncytiotrophoblast I and II in the labyrinth layer of the placenta, which are coupled by gap junctions (Gabriel et al. 1998). In contrast, the human placenta contains only one giant syncytiotrophoblast and so is not vulnerable to Cx26 mutations. Cx45 deletions are also embryonic lethal (Kruger et al. 2000; Willecke et al. 2002), in this case likely the result of myocardial arrhythmia shortly after the heart begins to beat (Nishii et al. 2003). Cx37 knockouts are female sterile from a failure of ovarian follicle development at the antral stage. Presumably, loss of communication between oocyte and cumulus cells leads to premature resumption of meiosis and luteinization (Simon et al. 1997). The loss of Cx40, prevalent in the His-Purkinje system, results in cardiac arrhythmias resembling right-bundle-branch block in humans (Simon et al. 1998; Kirchhoff et al. 1998).

Unique roles played by some connexins have been shown by knockin experiments. The Cx43 coding sequence was replaced in three separate mouse lines with Cx32, Cx40, or Cx26 coding regions. All three animal lines showed new functional defects unique to each connexin, revealing that the three connexins were not able to substitute for Cx43 in all contexts (Plum et al. 2000; Winterhager et al. 2007). Although none of the lines displayed the pulmonary outflow defects seen in the Cx43KO mouse (Reaume et al. 1995), a knockin of Cx31 into the Cx43 locus did show the defect (Zheng-Fischhofer et al. 2006). Thus, connexins may have both unique and redundant functions.

SURPRISING AND PUZZLING RESULTS FROM CONNEXIN MUTATIONS

Other functions that emerge from connexin deletions may result from the loss of a complex interplay of multiple connexin-family members in an incompletely defined network, producing unexpected and unexplained outcomes. Some of these examples are explored here in more detail.

Gap Junctions in the Vascular System

Arterioles are composed of a longitudinal layer of endothelial cells facing the blood, which is separated by a basal lamina from a layer of circular smooth muscle cells that control lumen diameter. There is a surprising complexity of connexin expression in the arteriolar layers. Smooth muscle cells express mainly Cx43 (Gabriels and Paul 1998) and endothelial cells mainly Cx40 (Little et al. 1995; van Kempen and Jongsma 1999), although both cell types express both connexins. Cx32 expression has been reported in endothelial cells (Okamoto et al. 2009). Smooth muscle cells uniquely express Cx45 (Kruger et al. 2000), whereas only the endothelium contains Cx37 (Gabriels and Paul 1998; van Kempen and Jongsma 1999). In addition, there can be significant regional variations in the relative abundance of these connexins in the vessel wall. As an example, endothelial Cx43 is dramatically up-regulated at the expense of the other connexins in areas that experience shear stresses such as vessel branch points (Gabriels and Paul 1998). Not only are gap junctions formed within arteriolar layers, but junctions are also formed between smooth muscle and endothelial cells. The connexin content of the myoendothelial junctions is not yet clear, although in vitro studies suggest that the endothelial side contains largely if not exclusively Cx40 (Isakson and Duling 2005).

Gap junctions have been strongly implicated in the conducted spread of vasodilation. Local endothelial stimulation initiates a rapidly propagated, bidirectional wave of relaxation along the vessel axis (Welsh and Segal 1998; Figueroa et al. 2003; de Wit et al. 2006). An intact endothelium is required for conducted vasodilation, which does not decay with distance and so must contain a self-regenerative component. The propagation of vasomotor activity is significantly depressed in Cx40 KO but not Cx37 KO animals (Figueroa et al. 2003; de Wit et al. 2000). While it was initially surprising that the loss of Cx37, which is co-expressed in endothelial cells, had no effect on propagation, this could be explained by the fact that loss of Cx40 causes a dramatic (>20-fold) reduction in the levels of endothelial Cx37, while loss of Cx37 results in only a mild (∼fourfold) reduction in the levels of Cx40 (Simon and McWhorter 2003).

A simple model for the role of gap junctions in propagation is that endothelial stimulation results in a change in membrane potential that is passively conducted along the endothelial layer through gap junctions, critically those containing Cx40. However, this model does not explain self-propagation. Even more problematic, knockin of Cx45 into the Cx40 locus does not rescue the Cx40 KO phenotype, suggesting that ionic spread of membrane potential changes through endothelial–endothelial gap junctions is not a critical factor (Wolfle et al. 2007). On the other hand, studies using connexin-mimetic peptides to selectively inhibit junctional communication in rabbit iliac arteries suggest that although Cx40 is required for endothelium-dependent smooth muscle hyperpolarization, Cx43 is required for spread of that hyperpolarization within the smooth muscle layer (Chaytor et al. 2005). Taken together, these observations suggest another model in which propagation requires both myoendothelial gap junctions as well as gap junctions joining smooth muscle cells. In the first phase, endothelial stimulation leads to release of an endothelium-derived hyperpolarizing factor (EDHF), causing hyperpolarization of immediately adjacent smooth muscle. It has been suggested that EDHF signaling requires myoendothelial junctions (Griffith 2007), which are permeable to inositol trisphosphate and Ca²⁺ (Isakson et al. 2007). A second phase might involve electrotonic spread of hyperpolarization within the smooth muscle layer through gap junctions composed of Cx43. The extent of this spread would be modest as electrical coupling in this layer is relatively weak. In the third phase, smooth muscle must restimulate endothelial cells distal to the site of initial stimulus, regenerating additional rounds of EDHF release. Relaxation of smooth muscle accompanies release of a second factor, endothelium-derived relaxation factor (likely nitric oxide), which can move from endothelium to smooth muscle in the absence of gap junctions. This model is consistent with the loss of conducted vasodilation in the Cx40 KO, but not Cx37 KO, and predicts a Cx40 KO phenocopy in a smooth muscle-specific Cx43 KO, which has not yet been evaluated.

In addition to vasomotor responses, connexin knockouts can dramatically impact systemic blood pressure. Conditional disruption of Cx43 in vascular endothelial cells results in hypotension and bradycardia (Liao et al. 2001), accompanied by elevated plasma levels of nitric oxide because of increased activity of endothelial nitric oxide synthase. These phenotypes are currently without explanation and are not seen in another model of vascular deletion of Cx43 (Theis et al. 2001). In contrast to the hypotension accompanying vascular loss of Cx43, constitutive deletion of Cx40 results in hypertension (de Wit et al. 2006). In this case, disregulation of angiotensin levels may be responsible. In these animals, renin-producing cells are anatomically displaced during development (Kurtz et al. 2007) and are also less responsive to feedback inhibition by plasma angiotensin, leading to increased plasma levels of renin (Wagner et al. 2007). Why the loss of Cx40 results in this cellular localization defect is not known. Interestingly, although knockin of Cx45 into the Cx40 locus is unable to rescue propagation of the vasomotor activity (Wolfle et al. 2007), it abrogates the hyperreninemia, partially attenuating the systemic hypertension and restoring angiotensin-suppression of renin release (Schweda et al. 2008). Parenthetically, Cx45 deletion from smooth muscle in the juxtaglomerular apparatus later in development also results in increased renin secretion and significant blood pressure elevation (Hanner et al. 2008; Yao et al. 2008).

The double knockout (dKO) of Cx37 and Cx40 displays an additional phenotype not seen in either individual knockout. dKO animals die perinatally with dramatic vascular abnormalities. By E18.5, numerous hemorrhages are visible through the skin and internally in the testes, lungs, and intestines. Vasculogenesis is aberrant in the testis and in the connective tissues of the small bowel, but seemingly unaffected in other organs (Simon and McWhorter 2002; Simon and McWhorter 2003). It is not known if these new pathologies result from a combination of the individual regulation and selectivities of the individual connexins, or if this is because of unique properties exhibited by heteromeric or heterotypic intercellular channels.

Gap Junctions in the Ocular Lens

During development, the optic vesicle induces the overlying ectoderm to invaginate and pinch off a hollow sphere of cells, the lens vesicle. The posterior cells of the vesicle then elongate anteriorly as lens fibers, which contact the anterior cells occluding the vesicle lumen. The lens thus becomes a solid cyst of cells, with an anterior epithelium and posterior fibers. The organ eventually loses an enveloping basket of blood vessels, becoming totally avascular and therefore dependent on the aqueous humor for all metabolic needs. The lens continues to grow in volume throughout the life of the organism by appositional growth, differentiating new lens fibers from a stem cell population at the equatorial surface. The older fibers do not turn over, remaining in the lens interior. To achieve a high refractive index and transparency, the differentiating fibers synthesize high concentrations of soluble proteins, the crystallins, and then undergo a limited apoptosis, destroying their nuclei and all light-scattering organelles. Thus, the lens fibers are metabolically dependent on the anterior epithelial cells that retain their organelles. The lens fibers are joined to each other and to the epithelial cells by large numbers of gap junctions (Goodenough 1992). The asymmetric location of the Na⁺K⁺ATPase in the epithelium results in a translenticular potential and a DC current flow (Candia et al. 1970), modeled as the circulatory system of the lens (Rae 1979; Mathias 1985; Mathias and Rae 1989). As the high concentration of the crystallins requires a tight control of ionic balance to remain in solution, the ionic syncytium created by the gap junctions is essential for lens transparency.

Cx43, 46, and 50 are expressed in the lens. Cx43 and 50 are found abundantly in the lens epithelium (Beyer et al. 1987; Jiang et al. 1995; Martinez-Wittinghan et al. 2003). Cx46 and 50 are found joining the lens fibers where they colocalize to the same junctional plaques (Paul et al. 1991) and have been shown to co-oligomerize into the same connexons and intercellular channels (Konig and Zampighi 1995; Jiang and Goodenough 1996). Indeed, immunofluorescence studies have shown colocalization of Cx46 and 50 in all junctional plaques joining the fibers. Given this anatomical overlap, it is surprising that targeted deletion of Cx46 and 50 result in distinctly different phenotypes (Gong et al. 1997; White et al. 1998). First, both cause cataracts but with differences in timing of onset and in morphology. Second, deletion of Cx50, but not Cx46, results in a slower postnatal growth rate with concomitant decrease in lens size and microphthalmia (White et al. 1998). Interestingly, the normal growth rate is uniquely dependent on Cx50 because replacing the coding region of Cx50 with that of Cx46 (Cx50^46/46) does not fully rescue the lens mitotic rate (White 2002; Sellitto et al. 2004). The identity of the Cx50-dependent signal controlling mitosis is not known (White et al. 2007). The Cx46/Cx50 double knockout shows a phenotype more severe but predictable as the sum of the two individual connexin deletions (Xia et al. 2006).

Cx50^46/46 animals are completely free of cataracts (White 2002), suggesting that this pathology could be prevented by simply restoring adequate numbers of junctional channels. Thus, it is surprising that mice heterozygous for Cx46 and Cx50 at the Cx50 locus (Cx50^+/46) develop a cataract (Martinez-Wittinghan et al. 2003). Furthermore, this cataract is morphologically different from those in either Cx46KO or Cx50KO lenses. Although the latter two are primarily nuclear, the Cx50^+/46 cataract is largely subepithelial. Additional crosses show that the Cx50^+/46 cataract is insensitive to dosage of Cx46 at the Cx46 locus, proving that this unexpected phenotype is the result of changes in connexin stoichiometry in the epithelium, where Cx46 is not normally detected. Importantly, the phenotype only occurs when Cx50 and Cx46 are coexpressed in the epithelium, because no cataract is observed in the homozygous (Cx50^46/46) knockin (White 2002). In addition to the cataract, Cx50^+/46 lenses display impaired dye transfer both within the epithelial plane and between epithelium and underlying fibers (Martinez-Wittinghan et al. 2003). Why mixing of Cx46 and Cx50 in the epithelium should depress dye transfer and cause a novel cataract is completely without explanation because those connexins functionally interact in heterotypic and heteromeric configurations both in vivo and in expression systems (White et al. 1994; Jiang and Goodenough 1996; Hopperstad et al. 2000).

Demonstration of mechanisms underlying the specificity of connexin intercellular channels in these contexts is still missing. It was shown that fiber–fiber conductance was lower in the Cx50^46/46 knockin than WT (Martinez-Wittinghan et al. 2004), thus the knockin approach may provide equal numbers of channels but does not provide equal levels of coupling. Regardless, the relationship between coupling level and differential mitotic rates remains obscure. We favor the notion that differential permeability of intercellular channels may play a more important role, as connexin-dependent differences in small molecule permeability have been observed in several studies (Harris 2007). For example, Cx43 channel permeability to cAMP is approximately three times higher than Cx26 and approximately five times higher than Cx40 (Kanaporis et al. 2008), providing a conceptual framework for the observed differences in knockin phenotypes (Harris 2008).

Gap Junctions in Myelin and the Central Nervous System

Mutations in Cx32 associated with the X-linked form of Charcot-Marie-Tooth syndrome result in a peripheral neuropathy associated with myelin failure in Schwann cells. Cx32 forms “reflexive” gap junctions that the Schwann cell makes with itself at the paranodal membranes and incisures of Schmidt-Lantermann. This anatomy suggests that the reflexive junctions in myelin are essential for communication between perinuclear and adaxonal Schwann cell cytoplasm. Measurements of the rate of diffusion between these two cytoplasmic compartments in individual Schwann cells support this notion (Balice-Gordon et al. 1998). However, there is no significant difference between diffusion rates in WT and Cx32 KO animals. To explain this discrepancy, it was hypothesized that Cx29, which is equally abundant although with a somewhat different intracellular distribution, might substitute for the loss of Cx32. However, Cx29 does not accumulate in gap junctional plaques in vivo in oligodendrocytes or Schwann cells (Altevogt et al. 2002; Nagy et al. 2003; Altevogt and Paul 2004) or form function gap junctions when expressed in tissue culture cells (Altevogt et al. 2002). On the other hand, the Cx29 KO does show a myelin defect but one that is restricted to cell bodies of the spiral ganglion neurons in the organ of Corti (Tang et al. 2006).

An additional surprising role for connexins has been shown in the developing neocortex (Elias et al. 2007). Cx26 and Cx43 protein expression was substantively knocked down by electroporation of shRNAs into E16 embryonic cortex. Connexin knockdown resulted in the stalling of migration of neurons along radial glia in the intermediate zone and a loss of cells arriving in the lower and upper cortical plates. Further experiments showed that normal migration was dependent on neuronal rather than glial expression of connexins (Elias et al. 2007). Connexin knockdown neurons showed normal timing of exit from mitosis and no detectable changes in apoptosis, which is unexpected because changes in cell–cell communication and hemichannel involvement in Ca²⁺ waves have been correlated with stages of the mitotic cycle (Bittman et al. 2007). Surprisingly, a channel-dead mutant (Beahm et al. 2006) rescued the migration defect, whereas mutations that resulted in both the loss of connexon pairing (but not hemichannel activity) and the loss of interaction with cytoplasmic partners (C-terminal truncations) were unable to rescue (Elias et al. 2007). These data led to the conclusion that the adhesive properties of connexins, rather than channel activity, were required for correct neuronal migration. In this context, it is of interest that Cx43 hemichannels can confer adhesivity between HeLa and C6 glioma cells in culture (Cotrina et al. 2008).

In summary, connexins and innexins are universally used to promote intercellular interactions between cells in solid tissues and circulating elements of the blood (Wong et al. 2006). They show multiple levels of regulation from instantaneous to hours. Genetic studies have shown that gap junctions are involved in a wide variety of functions in homeostasis, regulation, regeneration, and development. Given that a complex spectrum of small molecules within a cell can potentially diffuse through gap-junctional channels into neighbors, the identification of the relevant small molecules subserving each function has been difficult. Connexons, the hexameric precursor to the gap-junction channel, can function as a hemichannel in nonjunctional membranes promoting paracrine signaling. Even without channel function, the adhesivity of connexons can provide critical migratory cues. Unraveling the multiple functions of connexins and innexins and the contributions to these functions controlled by channel selectivity and regulation, is fundamental to understanding many aspects of collective cellular behavior.