Abstract. Keratin 5 and keratin 14 have been touted as the hallmarks of the basal keratin networks of all stratified squamous epithelia. Absence of K14 gives rise to epidermolysis bullosa simplex, a human blistering skin disorder involving cytolysis in the basal layer of epidermis. To address the puzzling question of why this disease is primarily manifested in skin rather than other stratified squamous epithelia, we ablated the K14 gene in mice and examined various tissues expressing this gene. We show that a key factor is the presence of another keratin, K15, which was hitherto unappreciated as a basal cell component. We show that the levels of K15 relative to K14 vary dramatically among stratified squamous epithelial tissues, and with neonatal development. In the absence of K14, K15 makes a bona fide, but ultrastructurally distinct, keratin filament network with K5. In the epidermis of neonatal mutant mice, K15 levels are low and do not compensate for the loss of K14. In contrast, the esophagus is unaffected in the neonatal mutant mice, but does appear to be fragile in the adult. Parallel to this phenomenon is that esophageal K14 is expressed at extremely low levels in the neonate, but rises in postnatal development. Finally, despite previous conclusions that the formation of suprabasal keratin filaments might depend upon K5/K14, we find that a wide variety of suprabasal networks composed of different keratins can form in the absence of K14 in the basal layer.

Abstract. In epidermal cells, keratin intermediate filaments connect with desmosomes to form extensive cadherin-mediated cytoskeletal architectures. Desmoplakin (DPI), a desmosomal component lacking a transmembrane domain, has been implicated in this interaction, although most studies have been conducted with cells that contain few or no desmosomes, and efforts to demonstrate direct interactions between desmoplaldn and intermediate filaments have not been successful. In this report, we explore the biochemical nature of the connections between keratin filaments and desmosomes in epidermal keratinocytes. We show that the carboxy terminal "tail" of DPI associates directly with the amino terminal "head" of type II epidermal keratins, including K1, K2, K5, and K6. We have engineered and purified recombinant K5 head and DPI tail, and we demonstrate direct interaction in vitro by solution-binding assays and by ligand blot assays. This marked association is not seen with simple epithelial type II keratins, vimentin, or with type I keratins, providing a possible explanation for the greater stability of the epidermal keratin filament architecture over that of other cell types. We have identified an 18-amino acid residue stretch in the K5 head that is conserved only among type II epidermal keratins and that appears to play some role in DPI tail binding. This finding might have important implications for understanding a recent point mutation found within this binding site in a family with a blistering skin disorder.

Epidermal nevus of the epidermolytic hyperkeratotic type is a mosaic genetic disorder of suprabasal keratin. The correlation of mutations in the K10 gene with lesional skin and the correlation of the normal gene with normal skin provide evidence that genetic mosaicism can cause clinical mosaicism.