Efficient methods for tumor targeting are eagerly awaited and must satisfy several challenges: molecular specificity, transport through physiologic barriers, and capacity to withstand extracellular or intracellular degradation and inactivation by the immune system. Through interaction with its hosts, the intestinal pathogen-produced Shiga toxin has evolved molecular properties that are of interest in this context. Its nontoxic B-subunit binds to the cellular toxin receptor, glycosphingolipid Gb 3 , which is highly expressed on human cancers and has recently been reported to be involved in the formation of metastasis in colorectal cancers. Its function as a target for cancer therapy has already been addressed in xenograft experiments. We here show that after oral or i.v. injections in mice, the B-subunit targets spontaneous digestive Gb 3 -expressing adenocarcinomas. The nontumoral mucosa is devoid of labeling, with the exception of rare enteroendocrine and CD11b-positive cells. As opposed to other delivery tools that are often degraded or recycled on cancer cells, the B-subunit stably associates with these cells due to its trafficking via the retrograde transport route. This can be exploited for the in vivo delivery of contrast agents to tumors, as exemplified using fibered confocal fluorescence endoscopy and positron emission tomography (PET) imaging. In conclusion, the data presented in this manuscript lay the groundwork for a novel delivery technology that, in addition to its use for molecular imaging applications such as noninvasive PET, could also be exploited for targeted tumor therapies.