During the last decade, an increasing number of papers have described the use of various genera of bacteria, including E. coli and S. typhimurium, in the treatment of cancer. This is primarily due to the facts that not only are these bacteria capable of accumulating in the tumor mass, but they can also be engineered to deliver specific therapeutic proteins directly to the tumor site. However, a major obstacle exists in that bacteria because the plasmid carrying the therapeutic gene is not needed for bacterial survival, these plasmids are often lost from the bacteria. Here, we report the development of a balanced-lethal host-vector system based on deletion of the glmS gene in E. coli and S. typhimurium. This system takes advantage of the phenotype of the GlmS(-) mutant, which undergoes lysis in animal systems that lack the nutrients required for proliferation of the mutant bacteria, D-glucosamine (GlcN) or N-acetyl-D-glucosamine (GlcNAc), components necessary for peptidoglycan synthesis. We demonstrate that plasmids carrying a glmS gene (GlmS(+)p) complemented the phenotype of the GlmS(-) mutant, and that GlmS(+) p was maintained faithfully both in vitro and in an animal system in the absence of selection pressure. This was further verified by bioluminescent signals from GlmS (+)pLux carried in bacteria that accumulated in grafted tumor tissue in a mouse model. The signal was up to several hundred-fold stronger than that from the control plasmid, pLux, due to faithful maintenance of the plasmid. We believe this system will allow to package a therapeutic gene onto an expression plasmid for bacterial delivery to the tumor site without subsequent loss of plasmid expression as well as to quantify bioluminescent bacteria using in vivo imaging by providing a direct correlation between photon flux and bacterial number.