The v-maf oncogene, originally identified as the transforming gene of the avian retrovirus AS42, encodes a protein with a basic-leucine zipper (bZIP) structure which is a typical motif for protein dimerization and DNA binding. The v-Maf protein forms homodimers and sometimes heterodimers with some bZIP proteins and works as a transcription factor. Recent studies of tissue-specific expression of cellular Maf family proteins suggest that Maf-related proteins play important roles in early development and cell differentiation. In our previous studies, two maf-related cDNA clones, maf-1 and maf-2, have been isolated from a rat liver cDNA library which facilitates the investigation of physiological roles of Mafs as well as their target genes in mammals. We determine here the binding DNA sequences of Maf-1 and Maf-2, respectively. Maf-1 recognizes a number of sequences containing a short consensus sequence, -GCTGAC-, half of the Maf recognition element (MARE) which has been previously identified for V-Maf. On the other hand, binding sequences of Maf-2 are limited and Maf-2 binds to the MARE preferentially. It is shown that dimer-forming specificities of Maf-1 and Maf-2 to Jun or Fos family proteins are variable. Maf-1 efficiently forms heterodimers with Fos family proteins. Compared to the case of Maf-2, though the DNA-binding specificity of the heterodimer depends on the kind of counterpart binding to Maf-1, a wide variety of target sequences are available for Maf-1. In contrast, Maf-2 shows restricted heterodimer-forming specificity and DNA-binding specificity, so that the target genes for Maf-2 are considered to be much more limited.