Posted by Anaid & Sam, July 2018

Repositive's Cancer Models Community: How to identify a good PDX model

This blog is a continuation of our previous post linked to the Cancer Models Community initiative and the Cancer Models Platform.

The initiative is aimed at promoting scientific discussion between a wide range of researchers from the biotechnology and pharmaceutical industries, as well as from contract research organisations (CROs).

In this new post we share what we have learned from experts whose job it is to identify, use and develop good Patient-Derived Xenograft (PDX) models. As we discussed in our previous blog, PDX models have emerged as important tools for pre-clinical and translational research. However, while PDX models have several advantages over standard Cell Line-Derived Xenografts (CDX), identifying the most suitable PDX models is not easy.

Here we share what both biopharma and biotech specialists regard as key characteristics for identifying a good model, the predictive value of PDX models, and new methods to overcome some of the challenges of working with them. The points highlighted here are notes from discussions during our Cancer Models Networking event at the AACR annual meeting in Chicago in mid-April.

We will return to the east coast of the USA in mid-July; we are attending the Hanson Wade Tumor Models Conference in Boston and hosting a networking event in Cambridge, MA. The event will focus on ‘Model selection for immuno-oncology and combination therapies’. If you have any questions or would like to arrange a meeting, please get in touch at cancer-models@repositive.io.

What makes a good PDX model, and what are the key characteristics for identifying one

A hot topic for debate is deciding on a common set of data that would allow a researcher to identify whether to use a model for a study or not. Although there is a consensus that the more information the better, the specifics of the data can vary from project to project. Biopharma and biotech specialists tend to agree on three sets of characteristics: molecular information, tumour growth dynamics, and clear annotation.

Molecular information: A key stage during pre-clinical research is identifying a subpopulation of patients that will potentially respond to the new compound. In this respect, molecular profiling can be important in providing information on mutations and chromosomal aberrations, such as duplication, deletion and translocation, many of which identify tumour suppressors or oncogenic drivers and potentially predict drugs likely to be efficacious in particular patient subgroups (Gao et al., 2015).

Tumour growth dynamics: It is known that tumours are inherently unstable, constantly evolving in response to selective pressures while in the patient (McGranahan & Swanton, 2017). Consequently, one of the challenges of using PDX models is that, over time, the models have a tendency to become less like the genotype and phenotype of the original patient tumour (Ben-David et al., 2017). Growth rate has been highlighted as a good initial indicator of whether a model has diverged and there is a general agreement that sudden changes in the growth rate can be indicative of a change in the model, and it should therefore be viewed with caution and characterised in more detail. Further investigation for the presence of key oncogenic mutations can be used to monitor the integrity of the PDX model.

Annotation: There is general agreement that the molecular profile of a model should be monitored over time, and the data carefully annotated. Specialists suggest P0, P2 and P10 are logical sampling points during passaging. Annotation over time should be clear, describing the total number of passages, the corresponding passage at which the molecular characterisation was carried out, and how the model has been managed prior to it being used. A list of key metadata fields that were agreed upon by the research community was recently published in the form of the PDX minimum information standard (PDX-MI) by Meehan et al. (2017).

One good model vs. a large and diverse panel of PDX models

Specialists tend to agree that identifying a single, good PDX model, which represents a single patient, has limited predictive value in drug discovery programmes. However, a comprehensive set of models that reflects variability and oncogenic mutations is an extremely powerful pre-clinical in vivo tool for compound screening and patient subgroup stratification (e.g. 1 × 1 × 1 experimental design or PDX clinical trial or PCT, Gao et al., 2015). This approach relies on having a large and diverse set of PDX models to make statistical predictions and identify associations between a genotype and drug response. Moreover, it may also establish mechanisms of resistance (Gao et al., 2015). Ultimately, this has the potential to inform patient selection in the clinic, increasing the efficiency and efficacy of clinical trials and ultimately speeding up the development of new therapies.

On the other hand, a single PDX model can function as a patient’s ‘avatar’ for exploring personalised cancer treatment. In this approach, the avatar models are mouse substitutes of specific human patients for drug testing or running in parallel with clinical studies (Hidalgo et al., 2014; Williams, 2018). This is technically challenging to implement, particularly for patients with very aggressive tumours, as the engraftment success can be low and the growth period long or unpredictable.

Modification of PDX models

One of the challenges of PDX models is their lack of sustained growth when cultured ex vivo. While PDX cells can be grown in 2D and 3D cultures, most PDX cell lines cease proliferating and undergo senescence after repeated passaging (Siolas & Hannon, 2013). Recently, scientists have overcome this limitation by reprogramming cells. A method called Conditional Reprogramming (CR) was developed in a collaboration between specialists at AstraZeneca and Propagenix and published in Molecular Cancer in December 2017 (Borodovsky et al., 2017).

The CR method can be used to achieve sustained expansion of human normal and tumour epithelial cells in 3D systems. Borodovsky and colleagues showed that PDX cells can be reprogrammed by co-culturing PDX epithelial cells with murine fibroblast feeders in the presence of Rho kinase inhibitor. This step allows cells to re-differentiate into their terminal lineage. Scientists have been able to show that reprogrammed PDX tissues retained the complexity of PDX tissue rather than the more homogeneous aggregation of cells that is typically seen in CDX models. This suggests that the 3D cell cultures derived from PDX are fundamentally more complex than cells from CDX.

Using this and similar methods, specialists are able to generate and expand stable PDX cell lines for ex vivo applications. Biopharma are positive about this new approach. Compared to 2D, 3D systems allow further genetic manipulation, retain the molecular characteristics of the original tumours, and offer the opportunity to model the tumour microenvironment in an ex vivo system that can stimulate dimensional aspects of the tumour.

Conclusion and outlook

Experts agree that there is no perfect model. All have challenges associated with them. While some specialists would say that:

“A good model is the one that is available at the time I need it” (Anonymous oncology researcher).

There is a general consensus that access to a large catalogue of models through a single portal is a valuable resource for the community.

It is also important to recognise that the parameters and characteristics highlighted here are a key element of targeted drug discovery programmes. For immuno-oncology programmes, the parameters and characteristics for model selection would be different. We will come back to this point in a future blog post, which will focus on immuno-oncology. This will also be a discussion point at our next Cancer Models Community event in Cambridge, MA on 18th July.

If you have any questions or would like further information, please contact cancer-models@repositive.io.

This blog post was co-authored by Sam Shelton and Anaid Diaz.


Ben-David U, et al., 2017. Patient-derived xenografts undergo mouse-specific tumor evolution. Nat Genet. 49(11):1567-1575. doi: 10.1038/ng.3967.

Borodovsky A, et al., 2017. Generation of stable PDX derived cell lines using conditional reprogramming. Mol Cancer. 6;16(1):177. doi: 10.1186/s12943-017-0745-1.

Gao H, et al., 2015. High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response. Nat Med.; 21(11):1318-25. doi: 10.1038/nm.3954.

Hidalgo M, et al., 2014. Patient-derived xenograft models: an emerging platform for translational cancer research. Cancer Discov.; 4(9):998-1013. doi: 10.1158/2159-8290.CD-14-0001.

McGranahan N, Swanton C, 2017. Clonal heterogeneity and tumor evolution: Past, present, and the future. Cell. 9;168(4):613-628. doi: 10.1016/j.cell.2017.01.018.

Meehan TF, et al., 2017. PDX-MI: Minimal information for patient-derived tumor xenograft models. Cancer Res. 77(21):e62-e66. doi: 10.1158/0008-5472.CAN-17-0582.

Siolas D, Hannon GJ, 2013. Patient-derived tumor xenografts: transforming clinical samples into mouse models. Cancer Res. 73(17):5315-9. doi: 10.1158/0008-5472.CAN-13-1069.

Williams JA, 2018. Using PDX for preclinical cancer drug discovery: The evolving field. J Clin Med. 7(3): 41. doi: 10.3390/jcm7030041.

Posted by
Sam Shelton

Sam Shelton

Business Development Associate
See all Sam's posts
Anaid Diaz

Anaid Diaz

Business Development Associate
See all Anaid's posts