This summary is a proposed synthesis of available information for the non-specialist. It does not incorporate all the published data, is inconsistent with some, and reflects the biases of the author. Connexin proteins have a common transmembrane topology, with four alpha-helical transmembrane domains, two extracellular loops, a cytoplasmic loop, and cytoplasmic N- and C-terminal domains. The sequences are most conserved in the transmembrane and extracellular domains, yet many of the key functional differences between connexins are determined by amino-acid differences in these largely conserved domains. Each extracellular loop contains three cysteines with invariant spacing (save one isoform) that are required for channel function. The junctional channel is composed of two end-to-end hemichannels, each of which is a hexamer of connexin subunits. Hemichannels formed by some connexin isoforms can function as well-behaved, single-membrane-spanning channels in plasma membrane. In junctional channels, the cysteines in the extracellular loops form intra-monomer disulfide bonds between the two loops, not intermonomer or inter-hemichannel bonds. The end-to-end homophilic binding between hemichannels is via non-covalent interactions. Mutagenesis studies suggest that the docking region contains beta structures, and may resemble to some degree the beta-barrel structure of porin channels. The two hemichannels that compose a junctional channel are rotationally staggered by approximately 30 degrees relative to each other so that the alpha-helices of each connexin monomer are axially aligned with the alpha-helices of two adjacent monomers in the apposed hemichannel. At present there is a published 3D map with 7.5 A resolution in the plane of the membrane, based on electron cryomicroscopy of 2D crystals of junctional channels formed by C-terminal truncated Cx43. The correspondence between the imaged transmembrane alpha-helices and the known transmembrane amino-acid sequences is a matter of debate. Each of the approximately 20 connexin isoforms produces channels with distinct unitary conductances, molecular permeabilities, and electrical and chemical gating sensitivities. The channels can be heteromeric, and subfamilies among connexins largely determine heteromeric specificity, similar to the specificities within the voltage-dependent potassium channel superfamily. The second extracellular loop contains the primary determinants of the specificity of hemichannel-hemichannel docking (analogous to the tetramerization domain of potassium channels). The 7.5 A map shows that each monomer exposes only two transmembrane alpha-helices to the pore lumen. However the conductance state of the imaged structure and the effects of the C-terminal truncation are unknown, so it is possible that other transmembrane domains contribute to the lumen in other functional states of the channel. In the transmembrane region, SCAM and mutagenesis data suggest that parts of the first three transmembrane alpha-helices are exposed to the lumen. Some of these data are contradictory, but may reflect conformational or isoform differences. There is reason to think that the first part of the N-terminal cytoplasmic domain can line the pore in some conformations. In the extracellular part of junctional channels, the N-terminal portion of the first extracellular loop is exposed to the lumen. The unitary conductances through connexin channels vary over an order of magnitude, from 15 pS to over 300 pS. There is a range of charge selectivities among atomic ions, from slightly anion selective to highly cation selective, which does not correlate with unitary conductance. There appear to be substantial ion-ion interactions within the pore, making the GHK model of assessing selectivities of limited value. Pores formed by different connexins have a range of limiting diameters as assessed by uncharged and charged probes, which also does not correlate with unitary conductance (i.e. some have high conductance but have a narrow limiting diameter, and vice versa). Channels formed by different connexins have different permeabilities to various cytoplasmic molecules. Where it has been assessed, the selectivity among cytoplasmic molecules is substantial and does not correlate in an obvious manner with the size selectivity data derived from fluorescent tracer studies, suggesting there are chemical specificities within the pore that enhance or reduce permeability to specific cytoplasmic molecules, functionally analogous to the ability of some porins to facilitate transport of specific substrates. For example, heteromeric channels with different stoichiometries or arrangements of isoforms can distinguish among second messengers. The differences in permeability to cytoplasmic molecules have biological consequences; in most cases one connexin cannot fully substitute for another. Voltage and chemical gating mechanisms largely operate within each hemichannel, though there is evidence for inter-hemichannel allosteric effects as well. There are at least two distinct gating mechanisms. One (Vj-gating) is a voltage-driven mechanism that governs rapid transitions between conducting states. Its voltage sensor involves charges in the first several positions of the cytoplasmic N-terminal domain and possibly in the N-terminal part of the first extracellular loop, which may both be exposed to the lumen of the pore in some states. The polarity of Vj-gating sensitivity is connexin-specific, closing with depolarization for some connexins and with hyperpolarization for others. The polarity can be reversed by point mutations at the second position. The lower conductance states induced by Vj-gating correspond to physical restrictions of the pore, and thus restricted or eliminated molecular permeation. Since the channels are not fully closed by Vj-gating, it can be seen as a way to eliminate molecular signaling while leaving electrical signaling operational. A second, independent gating mechanism mediates slow transitions (approximately 10-30 ms) into and out of non-conducting state(s). These transitions can occur in response to voltage ('loop gating'), chemical factors such as pH and lipophiles ('chemical gating'), and the docking of two hemichannels (sometimes called the 'docking gate'). These slow transitions may reflect a common structural change induced by these several effectors (electrical, chemical and homodimerization). Alternatively, they could reflect distinct gating processes responding to one or more of these effectors, that are indistinguishable at the single-channel level and have yet to be resolved mechanistically. The slow or loop gate closes with hyperpolarization. As a result, where Vj-gating closes with depolarization, individual hemichannels can close in response to both polarities of voltage (but only to a subconductance state for the Vj-gating polarity). Because of this, it is difficult to assign a macroscopic voltage sensitivity, or its modification due to mutagenesis, chemical modification or heteromeric interactions, to one or the other of these very distinct voltage-sensitive processes. This distinction can be made reliably only at the single-channel level. The Vj-gating voltage sensor and the loop-gating voltage sensor appear to be independent structures, since the Vj-gating voltage sensitivity can modified without effect on loop gating. For some connexins, certain modifications of the C-terminal domain seem to interfere with the operation of the Vj-gate while leaving loop gating unaffected. In some connexins, but not all, the chemical sensitivity to pH can involve interactions between regions of the C-terminal domain and cytoplasmic loop. Whether these regions exert their effects directly by physically blocking the pore, or by allosteric mechanisms (which may be more consistent with the relatively long time-course of closure) is not clear. For several connexins, truncation of the C-terminal domain eliminates the pH sensitivity, and co-expressing the domain with the truncated connexin restores the pH sensitivity. This has a functional resemblance to the particle-receptor mechanism for N-type inactivation of Shaker channels. What is being protonated is not clear, and may involve cytoplasmic factors, such as endogenous aminosulfonates. For other connexins, the action of pH does not involve the C-terminal domain and seems due to direct protonation of connexin. PKC phosphorylation of serine(s) in the C-terminal domain can affect the substate occupancy of at least one connexin. Phosphorylation of series in the C-terminal domain by MAP kinase appears to facilitate an interaction between it and an unknown receptor domain to eliminate coupling. This process has yet to be studied at the single-channel level. It also has a functional analogy to the particle-receptor model of channel inactivation. Both MAP kinase phosphorylation-induced and pH-induced inhibition can be mediated in truncated connexins by the corresponding free peptide. However, the relation between these two mechanisms are unexplored, as are specific mechanisms of direct endogenous regulation of connexin channel activity. (ABSTRACT TRUNCATED)