Major histocompatibility complex (MHC) class I genes typically encode polymorphic peptide-binding chains which are ubiquitously expressed and mediate the recognition of intracellular antigens by cytotoxic T cells. They constitute diverse gene families in different species and include the numerous so-called nonclassical genes in the mouse H-2 complex, of which some have been adapted to variously modified functions. We have identified a distinct family of five related sequences in the human MHC which are distantly homologous to class I chains. These MIC genes (MHC class I chain-related genes) evolved in parallel with the human class I genes and with those of most if not all mammalian orders. The MICA gene in this family is located near HLA-B and is by far the most divergent mammalian MHC class I gene known. It is further distinguished by its unusual exon-intron organization and preferential expression in fibroblasts and epithelial cells. However, the presence of diagnostic residues in the MICA amino acid sequence translated from cDNA suggests that the putative MICA chain folds similarly to typical class I chains and may have the capacity to bind peptide or other short ligands. These results define a second lineage of evolutionarily conserved MHC class I genes. This implies that MICA and possibly other members in this family have been selected for specialized functions that are either ancient or derived from those of typical MHC class I genes, in analogy to some of the nonclassical mouse H-2 genes.