Patient-Derived Xenograft (PDX) Models: Passage Drift, Host Strain, and Welfare Endpoints

Patient-derived xenograft (PDX) models implant a fragment of a patient’s tumor directly into an immunodeficient mouse rather than expanding it first as a cultured cell line. That distinction is what makes PDX oncology research scientifically attractive — the tumor’s original architecture, heterogeneity, and stromal composition survive the transplant far better than they do through years of monolayer culture — and it is also what makes PDX protocols harder to review than a standard xenograft study. An IACUC evaluating a PDX proposal is weighing three variables at once: whether the study design itself will still be measuring the tumor it started with by the time data comes out (passage drift), whether the immunodeficient host strain chosen for engraftment creates a welfare and biosecurity burden the facility can actually meet, and whether the protocol defines, in advance, what “the animal has reached its endpoint” means for a tumor-bearing mouse. This guide works through all three from the compliance side, not as a bench protocol.

What Distinguishes a PDX Model From a Cell-Line Xenograft

In a conventional xenograft, an established, often decades-old human cancer cell line is expanded in vitro and injected into a mouse. In a PDX model, primary tumor tissue — a surgical or biopsy specimen — is implanted directly into the host, most commonly subcutaneously or orthotopically, without an intervening culture step. Because the tissue is never selected for growth in plastic, PDX tumors retain more of the original tumor’s clonal heterogeneity, stromal and vascular architecture, and drug-response phenotype, which is why PDX panels and “avatar” models have become a standard tool for preclinical drug-response modeling and biomarker studies feeding into oncology trial design (see this site’s guide to oncology clinical trial endpoints for how that translational pipeline connects to RECIST-based trial readouts). The tradeoff is that PDX work is more animal-intensive to establish and propagate than a cell-line xenograft, which is exactly why the two issues below — drift and host strain — carry real weight in protocol review.

Passage Drift: Why Serial Propagation Threatens the Model’s Own Validity

A PDX line is expanded by re-implanting tumor fragments from one mouse into the next generation of hosts — each such transfer is a “passage.” Passaging is necessary because a single patient biopsy does not yield enough tissue to run a full study; the model has to be grown out first. It is also the model’s central vulnerability. Ben-David and colleagues (Ben-David et al., Nature Genetics, 2017) showed that PDX tumors undergo measurable, mouse-specific genomic evolution across serial passages: copy-number profiles diverge progressively from the founder patient tumor as passaging in the murine host selects for clones that propagate well in that environment, and the tumor’s original human stroma is largely replaced by mouse-derived stroma within the first few passages, changing the tumor microenvironment the drug-response readout is actually measuring. Put plainly, a late-passage PDX tumor is not a fixed proxy for the patient’s cancer — it is a moving target that has been reshaped by the act of propagating it.

For protocol review, this is a study-validity issue as much as an animal-welfare one, and the two are linked: an IACUC‘s charge under the Animal Welfare Act and PHS Policy includes confirming that a study is designed well enough to justify the animals it uses — a reduction principle, not just a welfare one. A protocol that proposes PDX work should specify a maximum passage number for the line in use, whether passage number is held constant across treatment and control arms (an uncontrolled passage-number mismatch between arms confounds drift with treatment effect), and whether the line is drawn from a cryopreserved, low-passage biobank stock rather than continuously re-passaged forward indefinitely. None of that is exotic bench detail — it is exactly the kind of design-quality question a reviewing committee is positioned to ask, and a protocol that is silent on passage number has left a real source of variability undocumented.

Host Strain Selection: Engraftment Efficiency Trades Against Welfare Burden

PDX engraftment depends on suppressing the host’s ability to reject transplanted human tissue, and the mouse strains used for this form a rough ladder of increasing immunodeficiency: athymic nude mice (T-cell deficient only), SCID mice (T- and B-cell deficient, though “leaky” residual immune function can develop with age), NOD-SCID mice (adds impaired NK-cell, macrophage, and complement function on top of the SCID defect), and NSG/NRG-class mice (IL2 receptor gamma-chain null, the most severely immunodeficient commonly used strains). Engraftment and tumor-take rates generally improve moving down that ladder, which is why low-cellularity biopsies, hematologic malignancies, and difficult-to-engraft solid tumors are often modeled in NSG-class hosts specifically.

The same deficiency that makes these strains better hosts for human tissue removes their capacity to control infection that an immunocompetent mouse would clear without incident — and NOD-SCID/NSG-class strains lose innate-immune components (NK cells, complement, macrophage function) beyond the adaptive-immune loss that defines nude and basic SCID mice, so susceptibility does not scale only with “immunodeficient vs. not,” it scales with where on that ladder a given protocol sits. That has direct facility consequences: specific-pathogen-free sourcing, individually ventilated caging, autoclaved feed/water/bedding, restricted personnel access, dedicated PPE, and active sentinel health monitoring stop being generic good practice and become the minimum needed to keep the study’s own animals alive to endpoint — see this site’s guides on animal biosafety level requirements and mouse husbandry standards for what that housing plan needs to cover in detail. A protocol should justify strain selection against this welfare and biosecurity cost, not only against engraftment efficiency — choosing the most permissive strain available by default, when a less severely immunodeficient strain would engraft the tumor type in question adequately, is a refinement question the committee should be able to ask and the investigator should be able to answer.

Welfare Endpoints an IACUC Protocol Must Define for Tumor-Bearing Animals

Because PDX hosts already carry elevated baseline risk before a tumor is added, humane endpoint criteria for tumor-bearing animals need to be defined across more than one axis, and a protocol should specify all of them before approval rather than leaving judgment calls to be made ad hoc during monitoring:

• Tumor burden. A maximum single or aggregate tumor size — commonly expressed as a maximum diameter or volume, or as a percentage of body weight — set by the institution’s animal care committee in consultation with the attending veterinarian; the specific figure is protocol- and institution-specific rather than one universal external number, which is exactly why the protocol document is where it has to be written down. The method and frequency of measurement (caliper formula, imaging) should be specified alongside the number itself.
• Ulceration and necrosis. Skin ulceration overlying a subcutaneous tumor is its own defined endpoint, independent of tumor size — a small tumor that has ulcerated can require intervention before a larger, intact one does. The protocol should distinguish impending from actual ulceration and specify what action (veterinary intervention vs. euthanasia) each grade triggers.
• Body condition scoring. The field-standard complement to weight-based monitoring is a body condition score — the widely used 5-point scale from Ullman-Culléré and Foltz (Laboratory Animal Science, 1999) grades emaciation by palpating spine and pelvic-bone prominence. This matters specifically in tumor-bearing animals because tumor mass can offset weight loss from underlying cachexia on a body-weight-only readout, masking a genuine decline; a protocol that monitors percent body-weight loss alone, without a body-condition component, has a real blind spot.

Monitoring frequency has to scale with the tumor’s expected growth kinetics for that model — daily observation as an animal approaches its projected endpoint is standard practice, not an optional intensification — and the protocol should name a defined veterinary-consultation trigger plus who is authorized to make the “endpoint met” determination, which is a training and personnel-qualification question covered in this site’s guide to IACUC training requirements. Consistent with expectations under the Guide for the Care and Use of Laboratory Animals, all of this needs to leave a documented record trail, not live only in an observer’s memory.

The 3Rs in PDX Study Design

Replacement is genuinely limited for PDX work — no accepted in vitro system reproduces heterogeneous human-tumor engraftment in a way that substitutes for the in vivo model, and a protocol should say so plainly rather than gesture at replacement it cannot deliver. Reduction and refinement both have real purchase, though. Reduction: maintaining a cryopreserved, low-passage PDX biobank means a line does not have to be re-derived (and re-established through animals) each time it is needed, and non-invasive tumor measurement — caliper or bioluminescent imaging — allows one cohort to be followed longitudinally instead of requiring serial-sacrifice cohorts at fixed timepoints. Refinement: well-defined, earlier humane endpoints substitute for a mortality or maximal-tumor-burden endpoint, and standard analgesia around the implantation procedure itself is part of the same calculus. A protocol’s 3Rs narrative for a PDX study should engage honestly with which of the three actually applies rather than treating all three as equally available.

Blinding and Randomization Under ARRIVE 2.0

Tumor volume measurement and body-condition/ulceration grading both involve real observer judgment, which is precisely the kind of measurement ARRIVE 2.0‘s experimental-design items are aimed at. Animals should be allocated to treatment and control arms by a defined randomization method — commonly stratified by baseline tumor volume so arms start balanced rather than confounding baseline tumor size with treatment assignment — and, where feasible, the person measuring tumors or scoring endpoints should be blinded to group assignment. See this site’s ARRIVE guidelines walkthrough for the full reporting checklist this maps to. A protocol or accompanying SOP should say explicitly who measures, whether that person is blinded, and how allocation is concealed — leaving this implicit is a common gap reviewers should be checking for specifically in tumor-model protocols, where the outcome measure is unusually observer-dependent.

What the IACUC Protocol Must Specify Before Approval

Pulling the threads above together, a PDX protocol should not be approvable until it states, in writing: the scientific justification for a PDX rather than cell-line model and for the specific host strain chosen; the maximum passage number in use and whether it is balanced across arms; the full set of humane endpoint criteria — tumor burden, ulceration/necrosis, and body condition score together, not any one alone; monitoring frequency and the documentation method; who is qualified to make an endpoint determination and what training that requires; the veterinary-consultation trigger; the housing and biosecurity plan matched to the host strain’s actual immunodeficiency level; and the randomization and blinding plan for tumor measurement. None of this is unusual by the standards of animal-research oversight generally — it is the same protocol-completeness logic the IACUC already applies elsewhere, made concrete for the specific ways a tumor-bearing, immunodeficient host raises the stakes on each item.

Frequently Asked Questions

What is passage drift in a PDX model, and does a protocol need to disclose it?

Passage drift is the genetic and phenotypic change a PDX tumor accumulates as it is serially re-implanted across mouse generations — published work (Ben-David et al., 2017) documents both copy-number divergence from the founder tumor and replacement of human stroma with murine stroma across passages. A protocol should state the maximum passage number used and whether it is held constant across study arms, since this is a design-validity variable a reviewing committee can reasonably expect to see addressed.

Why do PDX studies often use more severely immunodeficient mice than other xenograft work?

Primary tumor tissue, especially low-cellularity biopsies and some hematologic malignancies, engrafts poorly in less immunodeficient hosts. NOD-SCID and NSG/NRG-class strains, which additionally lack NK-cell, complement, and macrophage function beyond basic T/B-cell deficiency, engraft a wider range of tumor types — at the cost of substantially greater infection susceptibility that the housing and biosecurity plan has to account for.

Is a tumor size limit alone a sufficient humane endpoint for a PDX study?

No. Tumor burden, ulceration/necrosis, and body condition scoring are treated as distinct, complementary criteria because each can trigger independently — a small ulcerated tumor or a cachectic animal with tumor mass masking weight loss can both require intervention before an aggregate tumor-size threshold is reached.

Does the 3Rs framework meaningfully apply to PDX research, given that it requires a live host?

Replacement is genuinely constrained, since no accepted in vitro system reproduces heterogeneous PDX engraftment. Reduction (biobanking low-passage lines, longitudinal non-invasive imaging instead of serial-sacrifice cohorts) and refinement (earlier defined endpoints, procedural analgesia) both apply in practice and should be addressed specifically rather than treated as boilerplate.

Who sets the tumor burden limit for a specific PDX protocol?

The institution’s animal care and use committee, in consultation with the attending veterinarian, sets the specific number for a given protocol — there is no single universal external figure; it is written into the protocol itself and reviewed as part of approval.

How does ARRIVE 2.0 change tumor measurement practice in these studies?

It pushes protocols to specify a randomization method for treatment-arm allocation (commonly stratified by baseline tumor volume) and, where feasible, blinding of the person measuring tumors or scoring endpoints — both of which reduce observer-driven bias in what is otherwise a fairly subjective measurement.