Patient-derived xenografts: Practical considerations for preclinical drug testing

Patient-derived xenografts (PDXs) are increasingly being used to test new therapies or repurpose existing therapies as researchers and clinicians optimize precision oncology treatments.¹ This has been further accelerated with the increasing availability of new immunodeficient mice that have improved our ability to generate a wider variety of PDXs, including for challenging leukemia subtypes such as favorable risk acute myeloid leukemia (AML). Inspired by the conversation with Prof Richard Lock who features in a HemaSphere podcast reflecting on over 20 years of experience in preclinical testing,² this article reflects some of the practical considerations for establishing a PDX bank and their use in evaluating new therapies.

Immunodeficient mice provide the opportunity to engraft human leukemia cells and generate a PDX model. These are the models of systemic disease that infiltrate the bone marrow, spleen, and liver and disseminate throughout the peripheral blood. They are attractive models because they retain the cellular and molecular characteristics of the original disease with leukemia burden monitored through peripheral blood sampling or via bioluminescence.

To establish a PDX, patient cells are injected into the tail vein or intrafemorally of immunodeficient mice (Figure 1). This first round of engraftment or primagrafts usually has the slowest kinetics of engraftment time depending on the quality and source of the patient sample. Once leukemia develops in these primagrafts, the cells can be harvested from highly engrafted mice (e.g., human CD45+ve cells > 80% in the peripheral blood) and serially reinjected into secondary and tertiary recipients after which the kinetics of engraftment stabilises and is usually consistent across multiple transplants.

Importantly, when establishing new PDX samples, it is recommended that cells harvested from primagraft, and secondary transplant cells are protected and stored over the long term with only cells from tertiary transplants used in downstream experiments. This will ensure the longevity of the PDX bank and provide an important reference for quality assurance regarding clonal and genetic heterogeneity.

Alongside the technical establishment of the PDX, ensuring excellent record keeping (e.g., the time taken to reach 1% human CD45 cells in the peripheral blood) and adhering to the published minimum information standards for PDX models is important for the field in terms of reproducibility and sharing of PDX resources.³ This includes metadata on the original patient sample and sequencing methodology for the molecular characterization of the PDX (Figure 1). This characterization of the PDX is essential for downstream preclinical drug testing when individual samples are chosen based on the expression or a biomarker or the presence of a genetic mutation.

It is equally important that PDX samples used in downstream experiments are routinely checked (i.e., using single-nucleotide polymorphism arrays or whole genome sequencing) and referenced back to the primagraft and secondary transplants similar to the routine short tandem repeat profiling of cell lines. This characterization will quickly identify any mislabeled samples and ensure the long-term integrity of the PDX bank.

The most common immunodeficient mouse used in contemporary PDX generation is the NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mouse that is also referred to as NOD-scid IL2Rgamma^null, NOD-scid IL2Rg^null, or NOD scid gamma. These mice have no functional T cells and B cells and no natural killer cells, which allows the efficient engraftment of human cells.⁴ This strain has been useful for engraftment of lymphoid malignancies but more challenging for myeloid malignancies. Improved engraftment rates of AML samples have been achieved using NSG-SGM3 (NSG humanized with SCF, GM-CSF, and IL-3) including cases where cells were from intermediate-risk AML patients. Interestingly, this study also found that sex was a variable for engraftment with male patients agnostic to sex of recipient mice but female patients generated significantly higher engraftment into female.⁵

A new commercial immunodeficient mouse strain that can assist with AML PDX engraftment is the MISTRG (M-CSFh/h IL-3/GM-CSFh/h hSIRPh/h TPOh/h Rag2^−/− Il2rg^−/−) mouse that has human IL-3, GM-CSF, TPO, and M-CSF genes knocked-in to replace their murine counterparts, thereby expressing physiologically relevant levels of these cytokines. Importantly, these mice support the engraftment and maintenance of leukemia-initiating cell (LIC) cells, which have implications in testing new therapies designed to target this cell population.⁶ However, these mice can be difficult to obtain; therefore, another alternative is the NBSGW strain generated by crossing NSG mice with C57BL/6J-KitW-41J/J (C57BL/6.KitW41) mice is also permissive to AML engraftment even in the absence of preconditioning with ionizing radiation.⁷

One of the major challenges in preclinical testing is when to start treatment and when to assess event-free survival (EFS) when death is not considered an ethical endpoint. Furthermore, researchers are rarely blinded to the treatment mice receive, making it essential that an unbiased assessment of endpoint is applied consistently across different experiments and researchers. As explained by Professor Richard Lock, within the preclinical testing consortium, drug treatments begin when mice reach 1% human CD45 in the peripheral blood. The mice are then monitored once per week via peripheral blood draw for the duration of the study period, and objective response measures (ORM) are used to assess drug efficacy (Table 1). These measures were established in part due to the immense success rate of standard-of-care chemotherapy with new compounds that only result in progressive or stable disease irrespective of a statistical difference in EFS unlikely to be used clinically.

However, when researchers are establishing proof of principle efficacy using tool compounds, these ORM might not be appropriate and statistical differences in EFS sufficient to assess efficacy. Nevertheless, researchers should still record and report on the kinetics of leukemia burden during and after treatment and have a predetermined event cut-off to calculate EFS (e.g., 25% human CD45 in the peripheral blood) that will help produce robust data for assessing drug efficacy.

The assessment of targeted therapies is conventionally tested in a PDX model with each arm of the treatment regimen having approximately six mice. Treatment can then be compared to vehicle control and differences in EFS determined and charted using Kaplan–Meier survival curves (Figure 2). An alternative to this conventional method is the SMT format, which is recommended for assessing the efficacy of a drug across a large set of PDXs.⁸ The SMT allows response to be measured across diverse genetic subtypes of leukemia and can identify associated biomarkers. Another advantage of the SMT is the reduction of mouse numbers with a recent study assessing the CD123 antibody–drug conjugate pivekimab sunirine requiring a total of 78 mice to assess efficacy across 39 PDX models.⁹ This is in contrast to a conventional study assessing a second-generation proteasome inhibitor ixazomib in eight different T-ALL PDX samples that used 128 mice.¹⁰ However, SMTs can be logistically challenging to run and requires PDX samples with consistent leukemia growth kinetics (Figure 2).

PDX models continue to play a central role in understanding both leukemia biology and assessing novel agents. However, to quote the British statistician George Box, “All models are wrong, but some are useful” holds true for PDX models and their limitations should be recognized. Immunodeficient mice are, for example, very sensitive to anthracyclines due to the scid mutation or conversely can tolerate doses that are clinically unachievable. Similarly, establishing PDX models is only possible due to the lack of an adaptive immune system in host mice; therefore, interactions and impacts of a treatment on normal B- and T cells can be difficult to determine. Therefore, appreciating both the advantages and the limitations of the PDX model alongside standardized reporting in assessing efficacy will improve the development and translation of new therapies into the clinic.

Charles E. de Bock conceptualized and wrote the article.

The author declares no conflicts of interest.

No funding was received for this publication.