CD47 is a transmembrane protein found on the surface of many cells in the body.Its expression is often up-regulated in many cancer cells.The receptor of CD47 was identified as SIRPa, which is expressed on phagocytic cells.Engagement of SIRPa by CD47 serves as "do not eat me" signal, thus inhibiting the phagocytic activity of macrophages.Cancer cells often hijack this pathway by boosting CD47 expression on their surface, in order to prevent innate cell-mediated phagocytosis.Mols. blocking the CD47-SIRPa interaction will unleash such inhibition and promote tumor destruction.Accordingly, anti-CD47 antibodies offer new hope to successful cancer treatment.Despite people are enthusiastic about CD47 as a potential target, the side effects associated with CD47 blockade have emerged as a major concern.For example, treatment with anti-CD47 antibody greatly reduced the number of circulating red blood cells and platelets that also express CD47.Thus, the class of CD47 antibodies that stimulate tumor cell killing while sparing normal cells in vivo is desirable for the cancer patients.We are interested in understanding whether humanized mouse models could aid the evaluation of a variety of monoclonal CD47 antibodies we generated.We first screened them using the CDX platform established at Biocytogen.Indeed, most antibodies were able to clear human tumor cells in B-NDG mice.Then we screened the toxicity of these antibodies using humanized knocked-in mice.Body weight and blood were collected and analyzed.Most antibodies were able to clear human tumor cells in B-NDG mice.However, when we screened the toxicity of these antibodies using humanized knockedin mice, many antibodies caused rapid death of animals upon the first administration.Interestingly, we found two clonal antibodies bearing the least toxicity since the treated animals exhibited minimal weight loss on Day 2 and then recovered.We concluded that humanized mice could expedite the development of safer CD47 antibodies that can be advanced to human clin. trials.