Gap junctions are one of the most common forms of intercellular communication. They are composed of membrane proteins that form a channel permeable for ions and small molecules connecting the cytoplasm of adjacent cells. Although gap junctions provide similar functions in all multicellular organisms, vertebrates and invertebrates are believed to use unrelated proteins for this purpose [1–3]. The family of gap junction molecules called connexins is well-characterized in vertebrates, but no homologs of these proteins have been found in invertebrates [1–5]. In turn, only gap junction molecules with no sequence homology to connexins have been identified so far in insects and nematodes [3–7]. It was suggested that these are specific invertebrate gap junction proteins, and they were thus named innexins (invertebrate analog of connexins)[3]. Here, we demonstrate the presence of innexin homologs in different taxonomic groups, including vertebrates. Using PCR with degenerate primers, we cloned sequences homologous to innexins from mollusc central nervous system and flatworm whole-animal cDNA (Figure 1). This finding is important because it refutes the hypothesis that innexin proteins could represent a specific feature of recently postulated Ecdysozoa clade (‘moulting animals’, including among others arthropods and nematodes but not molluscs and flatworms)[3, 8]. Moreover, a database search using BLAST [9] for homology matches to the new mollusc and flatworm sequences revealed similarity to two human proteins: MRS1, function unknown, predicted from cDNA sequence submitted by GB Bolger and MR Steele (GenBank accession number AF093239) and a novel protein similar to MRS1 recently predicted from chromosome 22 DNA sequence (hPanx2 in Figure 1, accession number AL022328). A PSI-BLAST search [9] unambiguously detected the same two proteins even when seeded by one of the original innexin sequences, the Unc-7 gap junction protein from Caenorhabditis elegans: with an E-value inclusion threshold of 0.01, the two human homologs were detected with expectation (E) values of 10–5 at the first iteration. In reciprocal searches initiated by the human homologs, the C. elegans innexins were detected with E values of 10–9 in the second iteration. It can be argued that the presence of four (compositionally biased) transmembrane domains is a possible source of error while searching for homologous sequences. Theoretically, seeding BLAST searches with transmembrane region containing sequences may result in retrieval of similar membrane proteins that are, nevertheless, not homologous. However, in the case of innexins and related vertebrate sequences, because of the presence of a relatively well-conserved region containing two conserved cysteine residues just carboxy-terminal to the first transmembrane sequence, there is sufficient similarity outside the transmembrane regions to indicate homology: a PSI-BLAST search seeded by the hPanx2 sequence of the first putative extracellular loop flanked by only four amino acids from adjacent transmembrane regions with the E-value inclusion threshold of 0.05 revealed similarity to Unc-7 with E values of 10–12 in the second iteration.