Carcinoembryonic antigen (CEA) is a widely used tumor marker, especially in the surveillance of colonic cancer patients. Although CEA is also present in some normal tissues, it is apparently expressed at higher levels in tumorous tissues than in corresponding normal tissues. As a first step toward analyzing the regulation of expression of CEA at the transcriptional level, we have isolated and characterized a cosmid clone (cosCEA1), which contains the entire coding region of the CEA gene. A close correlation exists between the exon and deduced immunoglobulin-like domain borders. We have determined a cluster of transcriptional starts for CEA and the closely related nonspecific cross-reacting antigen (NCA) gene and have sequenced their putative promoters. Regions of sequence homology are found as far as approximately 500 nucleotides upstream from the translational starts of these genes, but farther upstream they diverge completely. In both cases we were unable to find classic TATA or CAAT boxes at their expected positions. To characterize the CEA and NCA promoters, we carried out transient transfection assays with promoter-indicator gene constructs in the CEA-producing adenocarcinoma cell line SW403, as well as in nonproducing HeLa cells. A CEA gene promoter construct, containing approximately 400 nucleotides upstream from the translational start, showed nine times higher activity in the SW403 than in the HeLa cell line. This indicates that cis-acting sequences which convey cell type-specific expression of the CEA gene are contained within this region.