Human keratinocytes were grown on a dermal equivalent (or lattice) at the liquid-air interface in an attempt to reconstitute a functional epidermis in vitro. Although the multilayered epithelium thus obtained is well differentiated, as shown by the presence of keratohyaline granules and horny layer, several differences from its in vivo counterpart were also observed: In the reconstructed epidermis, basal keratinocytes do not have the cuboidal shape found in vivo; they synthesize bullous pemphigoid antigen and laminin, but the distribution of these antigens is not linear as in vivo; they contain the plasma-membrane antigens restricted to the basal layer in vivo (VM1, BC1), but these antigens are not polarized; lack of polarization is also evidenced by the distribution of actin. Differentiation markers appear but with a topography slightly different from that of epidermis in vivo; the 67-kD keratin does not appear in the first suprabasal layer as in vivo but above; involucrin, which appears in the granular layers in vivo appears as soon as the cells leave the basal layer. psi 3 antigen and fibronectin found in vivo only in hyperproliferative epidermis (wound healing, psoriasis) are detected. Hyperproliferation would also explain the unexpected straining of basal cells by KL1 monoclonal antibody. Because of the potential clinical or pharmacologic use of artificial epidermis, the question of whether the epidermis obtained in vitro can be considered as "normal" is discussed.