Patient-derived xenograft models

A patient-derived xenograft (PDX) is created by implanting human tumor tissue into an immunocompromised animal host.^[S1] The design preserves a patient-derived tumor specimen in an in vivo system, but implantation site, host strain, passage, and replacement of human stroma can all change what the model represents.^[S3][S4][S5] This article explains how to specify those variables and how to interpret the evidence without treating a mouse response as a patient outcome.

What a PDX adds, and what it changes

Unlike a cell line-derived xenograft, a PDX begins with implanted patient tumor tissue and can retain features of that specimen that are absent from a long-established cell line.^[S1] Retention is not absolute. Host immune deficiency affects engraftment and which immune questions can be asked; implantation site changes the local environment; passaging can select mouse-adapted genomic changes; and human stromal components are replaced by mouse stroma.^[S3][S4][S5]

Evidence commonly used to justify PDX studies

Two frequently cited publications support different uses of PDX. A 2012 review describes the model and its experimental variables, while a 2015 study tests a large PDX panel as an in vivo compound-screening system.^[S1][S2] Neither removes the need to evaluate model drift, stroma, and host biology.

Model rationale and experimental variables

• {'items': ['Paper: Patient-derived tumour xenografts as models for oncology drug development.^[S1]', 'Publication: Nature Reviews Clinical Oncology, 2012.^[S1]']}

The review describes transplantation of patient-derived tumors into immunocompromised rodents and evaluates PDX as a platform for oncology drug development.^[S1] It is useful for framing heterogeneity, host selection, implantation, and passage. Its favorable assessment concerns the model class; it does not establish that any particular PDX preserves the donor tumor or anticipates a clinical response.

Large-panel in vivo screening

• {'items': ['Paper: High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response.^[S2]', 'Publication: Nature Medicine, 2015.^[S2]']}

Gao and colleagues assembled approximately 1,000 PDX models carrying diverse driver mutations and used them for in vivo compound screening.^[S2] The work provides supportive evidence that a sufficiently characterized panel can reproduce some drug-response and biomarker patterns relevant to trial hypotheses. It does not show that one PDX predicts one patient, and it does not establish universal clinical predictivity.

Host systems and shared resources

Two practical extensions require separate decisions: whether the study needs a human immune compartment, and whether an existing public model can answer the question before a new PDX is established. Neither decision changes the basic requirement to document host, passage, tumor identity, and assay performance.

Humanized mouse PDX designs

• {'items': ['Design question: does the mechanism require a human immune compartment that a conventional immunocompromised host cannot provide?^[S6]', 'Verified review: Humanized Mouse Models for the Preclinical Assessment of Cancer Immunotherapy, 2018.^[S6]']}

Conventional PDX hosts are immunocompromised, which limits experiments that depend on human immune responses. A review describes humanized mouse systems as one way to study cancer immunotherapy in a host containing human immune components.^[S6] The humanization method, donor matching, reconstitution measurements, graft-versus-host window, and cell populations present must be specified; “humanized PDX” is not a single standardized model.

Host choice changes the experiment. Nude, SCID, NOD-SCID, and NSG backgrounds differ in degree and type of immune deficiency and in engraftment behavior.^[S5] A proposal for immunotherapy work should therefore justify the host against the immune mechanism and document which human and mouse immune populations remain. Do not infer patient safety, immune-related adverse events, or biomarker performance from host nomenclature.

Public PDX resources

• {'items': ['NCI Patient-Derived Models Repository (PDMR): a repository that includes PDXs and other patient-derived model types.^[S7]', 'PDX Development and Trial Centers Research Network (PDXNet): an NCI network for research using PDX datasets.^[S8]', 'EurOPDX Consortium: a European PDX consortium with shared activities and consensus work.^[S9]']}

The NCI PDMR includes patient-derived xenografts, patient-derived cell cultures, cancer-associated fibroblasts, and patient-derived organoids.^[S7] PDXNet provides a network context for PDX datasets and studies.^[S8] These resources can be checked before commissioning new model establishment, but inclusion in a repository does not itself validate a model for a specific mechanism or endpoint.

EurOPDX is a separate European consortium whose public site documents workshops and consensus activity around PDX models.^[S9] When using any shared resource, request current passage, host history, molecular identity, consent and use restrictions, and raw characterization data rather than relying on the resource name.

Design variables that change interpretation

A model name and tumor type are not enough to compare proposals. The variables below determine which biology remains in the system and which conclusions the resulting data can support.^[S3][S4][S5]

A defensible commissioning strategy

Start with the decision the PDX experiment must support. Select models using donor and molecular criteria fixed in advance, justify host and implantation site, cap passage, confirm identity and target status before randomization, and prespecify exclusions and analysis. If a faster in vitro screen is used to narrow candidates, document that screen as a separate system and use the PDX to answer a defined in vivo question. Agreement between systems strengthens an internal evidence package; it does not establish clinical validity.

How to interpret a result

A PDX result is evidence that an intervention changed a measured endpoint in a specified patient-derived tumor, host, site, and passage. Panel-scale work can test response patterns across molecularly diverse models,^[S2] while genomic drift and stromal replacement constrain how far those patterns can be generalized.^[S3][S4] Report both the response and the model history, and do not translate tumor control in mice into an expected patient response.