Patient-derived cancer xenografts (PDX) are widely used to identify and evaluate novel therapeutic targets, and to test therapeutic approaches in preclinical mouse avatar trials. Despite their widespread use, potential caveats of PDX models remain considerably underappreciated. Here, we demonstrate that EBV-associated B-lymphoproliferations frequently develop following xenotransplantation of human colorectal and pancreatic carcinomas in highly immunodeficient NOD.Cg-Prkdc Il2rg /SzJ (NSG) mice (18/47 and 4/37 mice, respectively), and in derived cell cultures in vitro. Strikingly, even PDX with carcinoma histology can host scarce EBV-infected B-lymphocytes that can fully overgrow carcinoma cells during serial passaging in vitro and in vivo. As serial xenografting is crucial to expand primary tumor tissue for biobanks and cohorts for preclinical mouse avatar trials, the emerging dominance of B-lymphoproliferations in serial PDX represents a serious confounding factor in these models. Consequently, repeated phenotypic assessments of serial PDX are mandatory at each expansion step to verify "bona fide" carcinoma xenografts.

Combining high-coverage whole-genome sequencing with functional analyses, Giessler et al. demonstrate that tumor initiation and long-term tumor formation in human colorectal cancer are driven by multiple genomic subclones and that the functional heterogeneity of colorectal cancer tumor clone–initiating cells is not based on genomic architecture.