The COL7A1 gene, which encodes type VII collagen, has been implicated as a candidate gene for dominantly and recessively inherited forms of dystrophic epidermolysis bullosa. In this study, hamster and human cDNA clones, which encode the previously uncharacterized carboxyl-terminal portion of type VII collagen, have been isolated and characterized. The previously uncharacterized carboxyl-terminal NC-2 non-collagenous domain is shown to be comprised of 161 amino acids in humans, 170 amino acids in hamster and to contain 8 conserved cysteines in each species. The 6 most carboxyl-terminal cysteines are contained in a conserved motif similar to domains found in Kunitz protease inhibitors, and most closely resembling a similar motif found in the carboxyl-terminal globular domain of the alpha 3 chain of type VI collagen. Also contained in the highly acidic NC-2 domain are a number of potential sites for phosphorylation by casein kinases. Human genomic clones containing 24 exons of COL7A1 were isolated and characterized. The NC-2 domain is encoded by 7 of these exons, which include a junctional exon encoding the end of the collagenous region and the beginning of the NC-2 domain and a final exon encoding the end of the NC-2 domain and 333 bp of 3' untranslated sequences. Comparison of hamster and human sequences shows the region surrounding the junction of the collagenous and NC-2 domains to be particularly conserved. This region is likely to contain residues involved in the proteolytic removal of the NC-2 domain and cysteines involved in formation of the disulfide linkages which stabilize type VII collagen dimers.