Patient-Derived Xenograft (PDX) Models

In patient-derived xenograft (PDX) models, tumor tissue from cancer patients is implanted into immunocompromised mice, allowing the human tumor to grow in vivo. Fragments from primary or metastatic tumors are sectioned into small pieces and transplanted while preserving their original tissue architecture and cellular composition. Depending on the research purpose, these tumor fragments can be implanted subcutaneously, orthotopically (into the organ of origin), or heterotopically. Mouse strains that lack mature T cells, B cells, or NK cells, such as athymic nude, SCID, and NOD/SCID mice, are used to prevent immune-mediated rejection of the human tumor graft.

Among the benefits of PDX models, they more meaningfully preserve tumor properties when compared to cell line-derived xenografts (CDX). They maintain the 3D architecture, genomic landscape, and molecular heterogeneity of the original tumor, including the spatial organization and the mix of different cell populations. These models retain patient-specific features across different cancer stages, molecular subtypes, and treatment histories, making them particularly useful for translational oncology. Because of this high degree of clinical relevance, PDX models have become a reliable tool for a number of applications, such as the preclinical evaluation of novel therapeutics, validation of drug combinations, identification of treatment-responsive subpopulations, and investigation of mechanisms of drug resistance. They are especially valuable in the development of targeted therapies and immunotherapies, where preserving the native tumor microenvironment is critical for predicting patient outcomes. Humanized PDX models, which incorporate elements of the human immune system, are particularly valuable in immuno-oncology and for the development of immunotherapies.

PDX models can be generated de novo, but this process is often time-intensive and may not provide results quickly enough to meet certain project timelines. To overcome this limitation, several major repositories and biobanks provide ready access to well-characterized PDX models, representing diverse cancer types and genetic backgrounds. Large publicly accessible repositories include the NCI Patient-Derived Models Repository, the UHN Princess Margaret Living Biobank, and The Center for Patient Derived Models at the Dana-Farber Cancer Institute. Commercial organizations also maintain extensive PDX model libraries, such as Charles River Laboratories, The Jackson Laboratory, and Crown Bioscience. PDX model biobanks often provide searchable databases that allow selection based on key model attributes, including patient demographics, tumor type and molecular subtype, treatment history, anatomic origin, and host strain background. These collections can serve as a useful resource to accelerate preclinical evaluation of novel anticancer agents and combination therapies.

PDX models have some limitations as well. Over serial passages, these models can diverge from the original patient tumor. Within two to five passages, the human tumor stroma can be heavily replaced by murine extracellular matrix (ECM) and stromal cells, particularly cancer-associated fibroblasts (CAFs). This replacement disrupts human cytokine signaling and limits immune cell activation in the tumor microenvironment. In glioblastoma PDX models, in particular, later passages often show accelerated tumor growth and increased malignancy. Genomic fidelity also diminishes, with DNA copy number profiles diverging from the primary tumor. Over time, intratumoral heterogeneity and tumor differentiation also diminish, likely due to clonal selection and microenvironmental instability. Additionally, some PDX models develop spontaneous lymphomas from Epstein–Barr virus (EBV)–transformed B cells, which can confound drug sensitivity analyses.