Recent studies have suggested that carcinoembryonic antigen (CEA)-promoter sequences are active only in CEA-positive cells, filing in the criteria for tumor specific targeting of suicide genes. However, the present study on gene therapy of colon cancer and cell-specificity of CEA promoter, provide evidence that CEA-positive and CEA-negative cells transfected with E. coli cytosine deaminase (CD) gene under the control of CEA promotor sequence are sensitive to enzyme/pro-drug therapy with 5-fluorocytosine (5-FC). Individual clones derived from the CEA-negative cell lines: melanoma Hs294T and glioblastoma T98G after transfection with CD differed profoundly in their sensitivity to 5-FC. The IC50 values for several clones of the CEA-negative cells were almost the same as for CEA-positive colon cancer cells. Such 5-FC-sensitive clones derived from the population of CEA-negative cells, present even in small number, because of the very effective bystender effect of this enzyme/pro-drug system can cause severe problems during therapy by efficiently killing surrounding normal cells. Safety is the major issue in gene therapy. Our data suggest that the safety of gene-directed enzyme pro-drug therapy (GDEPT) with CEA promoter driven expression of therapeutic genes is not so obvious as it has originally been claimed.