While a number of different gap junction proteins have now been identified, hepatic gap junctions are unique in being the first demonstrated case where two homologous, but distinct, proteins (28,000 and 21,000 Mr) are found within a single gap junctional plaque (Nicholson, B. J., R. Dermietzel, D. Teplow, O. Traub, K. Willecke, and J.-P. Revel. 1987. Nature [Lond.]. 329:732-734). The cDNA for the major 28,000-Mr component has been cloned (Paul, D. L. 1986. J. Cell Biol. 103:123-134) (Kumar, N. M., and N. B. Gilula. 1986. J. Cell Biol. 103:767-776) and, based on its deduced formula weight of 32,007, has been designated connexin 32 (or Cx32 as used here). We now report the selection and characterization of clones for the second 21,000-Mr protein using an oligonucleotide derived from the amino-terminal protein sequence. Together the cDNAs represent 2.4 kb of the single 2.5-kb message detected in Northern blots. An open reading frame of 678 bp coding for a protein with a calculated molecular mass of 26,453 D was identified. Overall sequence homology with Cx32 and Cx43 (64 and 51% amino acid identities, respectively) and a similar predicted tertiary structure confirm that this protein forms part of the connexin family and is consequently referred to as Cx26. Consistent with observations on Cx43 (Beyer, E. C., D. L. Paul, and D. A. Goodenough. 1987. J. Cell Biol. 105:2621-2629) the most marked divergence between Cx26 and other members of the family lies in the sequence of the cytoplasmic domains. The Cx26 gene is present as a single copy per haploid genome in rat and, based on Southern blots, appears to contain at least one intron outside the open reading frame. Northern blots indicate that Cx32 and Cx26 are typically coexpressed, messages for both having been identified in liver, kidney, intestine, lung, spleen, stomach, testes, and brain, but not heart and adult skeletal muscle. This raises the interesting prospect of having differential modes of regulating intercellular channels within a given tissue and, at least in the case of liver, a given cell.

Connexins, like true cell adhesion molecules, have extracellular domains that provide strong and specific homophilic, and in some cases, heterophilic interactions between cells. Though the structure of the binding domains of adhesion proteins have been determined, the extracellular domains of connexins, consisting of two loops of ∼34–37 amino acids each, are not easily studied in isolation from the rest of the molecule. As an alternative, we used a novel application of site-directed mutagenesis in which four of the six conserved cysteines in the extracellular loops of connexin 32 were moved individually and in all possible pairwise and some quadruple combinations. This mapping allowed us to deduce that all disulfides form between the two loops of a single connexin, with the first cysteine in one loop connected to the third of the other. Furthermore, the periodicity of movements that produced functional channels indicated that these loops are likely to form antiparallel β sheets. A possible model that could explain how these domains from apposed connexins interact to form a complete channel is discussed.