The monoclonal antibody-defined, tumor-associated antigen GA733 was purified from the SW948 human colorectal carcinoma cell line and its partial amino acid sequence was determined. By using a synthetic oligonucleotide probe, two recombinants were isolated from a total human genomic library. We prove the existence of a family of GA733 genes. One of the genomic isolates is demonstrated to be an intronless gene, which is transcribed in pancreatic carcinoma cell lines and in placenta. The GA733 proteins were observed to contain sequences homologous to a repeat unit occurring 10 times in thyroglobulin and once in the HLA-DR-associated invariant chain. A more evolutionarily distant relationship was found with the alpha chain of the interleukin 2 growth factor receptor.