Over 50 years ago, scientists took a primary colonic adenocarcinoma from a 74-year-old patient, diced it up, and injected it under the skin of immune deficient mice. The tumors grew over several weeks and were then transplanted to another set of mice. The researchers, Rygaard and Povlsen, showed that the tumors retained many characteristics of the original human one. However, getting the tumors to grow in mice wasn’t necessarily an easy task. Of the four tumors they injected, two breast carcinomas and two colon carcinomas, only one took.
Although the term “PDX” or “patient derived xenograft” emerged relatively recently, the concepts and questions have been around for decades. Some papers even described the possibility of “avatars” although that term wouldn’t appear until the 1980s. These original PDX models were promising, but also expensive and slow-going. So it’s no surprise that when cancer cell lines derived from tumors began to gain traction and were able to generate tumors in xenograft models, that this faster and cheaper version became more popular. Unfortunately, scientists realized that drugs that fought cancer in these mouse models hardly ever worked in human patients, in large part because the cell-line xenografts eventually lost key features of the original tumor. These severe limitations drove the resurgence of the PDX model in the early 2000s.
These animal models still look a lot like their predecessors. “Cancer tissue is implanted into immune-compromised mice and kept ‘alive’ by passaging it from one animal to the other,” explains Julia Schueler, research director, Charles River. “Those cancer tissues represent a highly realistic model of the human disease.” Adds Davy Ouyang, vice president, scientific research & innovation at Crown Bioscience,“They recapitulate the key characteristics of the parental patient tumors. This includes features such as genetic heterogeneity and cellular and histopathological architecture of the original tumor. Further analysis of tumor cells from PDX models also indicates that genomic and gene expression profiles are maintained between the patient tumor and derived PDX model.” Ouyang says that this type of model is the most translational preclinical model for cancer drug development efficacy screening. “They are highly useful for predicting drug efficacy and for understanding tumor characteristics, and therefore are able to provide data which is more predictive of patient therapeutic outcomes. PDX models have also proven useful for ‘co-clinical trials’ where in vivo studies using PDX models and clinical trials are run in parallel. In these studies, PDX models act as avatars to mimic patient disease and are used to examine therapeutic treatment options.”
PDX models come in many varieties. Carol Bult, principal investigator in the MGI consortium at The Jackson Laboratory (JAX), notes that host animals can include zebrafish, fruit flies, and of course, mice. When it comes to mice, models vary depending on strain (or type of immunocompromised mouse) and patient-derived tumor, explains Terina Martinez, field application specialist at Taconic. “Often, the host mouse strain is selected for maximal compatibility with the tumor.”
Megan MacBride, director of product marketing at Taconic, explains that Nude mice, which lack T cells, and severe combined immunodeficient (SCID) mice, which lack T and B cells, are widely used models. SCID mice on a nonobsese diabetic (NOD) inbred strain background are also popular because the NOD strain has reduced innate immunity and also carries a Sirpα allele, which is permissive for engraftment of human hematopoietic lineages, and SCID/nonobese diabetic (NOD) mice are some of the most widely used animal models.
Image: A CIEA NOG mouse from Taconic.
“Beyond spontaneous mutations, genetically engineered models (GEM) are also popular. Rag2 knockouts fail to generate mature T and B cells. Combined with an interleukin-2 receptor gamma (IL2rg) mutation, Rag2/Il2rg double knockout mice lack T, B and natural killer (NK) cells,” says MacBride.
There are now also triple immunodeficient strains like NOG and NSG mice. Schueler says that since NOG and NSG strains have been made available to the scientific community, “PDX of hematological malignancies could be established to a similar extent as their solid counterparts.” According to MacBride, both are equivalent super immunodeficient models; both are on the NOD inbred strain background that carries the SCID mutation and deletion of the Il2rg gene. Both MacBride and Schueler point out that although the animals are functional knockouts, the design is somewhat different. The NSG mice have a full null phenotype whereas the NOG mice carry a version of the gene with a disrupted cytoplasmic region that can produce a truncated protein. “Although this protein can bind cytokines, it cannot transduce signals, and is thus a functional knockout,” says MacBride. That said, both models lack T, B, and NK cells, and have dysfunctional macrophages and dendritic cells, and a defective complement system.
There is the question of where to implant the tumor material. Injecting it under the skin of mice (subcutaneously) usually translates to faster, easier, and non-invasive readouts. But will a breast carcinoma behave the same way under the skin as it would if transplanted, orthotopically to mouse mammary tissue? “The tumor microenvironment has a proven impact on tumor biology,” says Schueler. There is complicated bidirectional communication between the cancer and surrounding non-malignant cells, which can influence growth, invasion and metastasis, and mechanism of resistance. “The advantage of orthotopic implantation is a higher resemblance to the human disease, but comes with the downside of a more laborious read-out, usually some sort of in-life imaging.”
Human tumors in the context of a human immune system
PDX mice without an immune system allow scientists to observe how the tumor grows, essentially unimpeded. And what types of drugs might successfully attack it. In this regard, the PDX mice are a critical resource. But to get a more complete picture of a human tumor in the context of a human immune system, a more advanced mouse model is required. Bult says, “Mice can be ‘humanized’ in many ways.” This includes genetically modeling the animal to express a human gene instead of a murine one or injecting them with hematopoietic stem cells. “In an oncological context, these are mice injected with human immune cells,” explains Schueler. “This enables the testing of immune-modulating compounds that help to fight cancer.” And Martinez points out that humanizing can mean relatively simple or complex. “They can range from a simple knock-in of a single human gene to multifaceted super immunodeficient GEM mouse models engrafted with human immune system or other tissues/organs.”
According to Schueler, there are two paths to humanized mice in oncology that depend on the source of the human immune cells. Mice can be injected with CD34+ hematopoietic stem cells derived from umbilical cord or with adult peripheral blood monocytes (PBMCs). “There are pros and cons to both approaches. In a nutshell, the stem cell approach is more time and cost intensive but, specifically with the advent of second-generation NSG/NOG mice supporting the differentiation process of the stem cells, has to be seen as the most realistic model system in this scientific area.” Ouyang says, “Humanized CD34+ mice are suitable for long-term in vivo studies as they can develop multiple lineages of human immune cell populations, and do not develop [Graft vs. Host Disease] GvHD. PBMC-humanized mice are faster to generate but quickly develop GvHD and succumb to the disease within weeks of human cell transplant.”
Although the NSG and NOG platforms are genetically equivalent, MacBride says that the case is different with the next-generation models. And both JAX and Taconic have different portfolios with clear functional differences; these are outlined for some of the more prominent models on the Taconic site.
Just because humanized mice are available doesn’t mean they are required for PDX models, says Ouyang. “However, for testing immunotherapies, an animal model with an intact human immune system is required, which is where humanized models are best leveraged.”
All experts were quick to point out that no model is without faults. Martinez explains that when it comes to PDX models there are limitations when it comes to “modeling stromal cells and complex aspects of the tumor microenvironment. Additionally, due to the rapid rate of mutation in some cancers, genetic drift can occur in patient-derived tumors, which can constrain replicability and add noise to the experimental system.” Ouyang notes that not all tumors will successfully engraft, so a PDX model is not always feasible. And it can be both expensive and labor-intensive to generate these models. The process can also be slow-going, sometimes taking six months or more to establish.
More PDX models than ever before
There are several PDX repositories around the world, including the EurOPDX consortium, which was launched in 2013 and consists of 18 not-for-profit cancer centers and universities in Europe and the U.S., and has established more than 1,500 PDX models. Schueler says that Charles River offers over 500 fully characterized, proprietary PDX models. “There is an ongoing effort to broaden the PDX portfolio through international collaborations with major hospitals and universities. The database allows clients to search features of interest to facilitate tumor model selection." Taconic offers many models commonly used for PDX hosts, although they do not provide PDX tumors, while JAX has PDX models in house and offers additional services. And CrownBio has over 3,300 established PDX models, including models of specific disease pathways such as RET, ALK, EGFR, MET, IDH, RSPO, and HER2.
Image: CrownBio provides a large commercial collection of PDX, with over 3,300 models covering a range of indications and including unique models of specific disease pathways.
PDX models are surging ahead. But Bult says that it’s wrong to think that one size fits all. “It is probably more appropriate to think of developing a model platform that is comprised of many components than to think that any single model system will be sufficient to address the complexity of biology. It is also important to realize that models are constantly being improved.”
Host strains for PDX mice depend on the type of tumor, immune deficiency required, and potential for human immune system expression. Below are popular mouse varieties.
Nude: Caused by a mutation in the FoxN1 gene; missing thymus and repressed immune system; deficient in T and normal B lymphocytes, enhanced natural killer (NK) cell activity. Ideal for use as a first-line model, these mice are hairless (making tumor detection easier). The NMRI nude outbred strain is widely used, although it has a relatively low take-rate for human breast tumors compared to other nude or immunodeficient mice. MacBride notes this strain is not ideal for PDX development.
SCID: Spontaneous genetic mutation in the Prkdc gene causes absence of mature B and T lymphocytes. Some mice can develop partial immune reactivity; this “leakiness” can be modulated by background strain.
NOD/SCID: SCID mutation transferred onto a non-obese diabetic (NOD) background. The gold standard for many years, these mice have impaired T and B cell lymphocyte development and are deficient in NK cell function. NK cells play a key role in rejecting human tissues. Some limitations remain, including leakiness of SCID mice and limited lifespan.
NOG: IL-2 receptor-γ chain (a common receptor for several cytokines) knockout on NOD/SCID background. Mutation in Il2rg produces truncated protein, but is functional knockout. Mice lack T cells, B cells, and NK cells, have reduced dendritic cell function, reduced macrophage function, lack complement activity, and do not display leakiness of T and B cells with age. Unlike NOD/SCID, very low incidence of lymphoma. Show better engraftment of human cells and tissues. CD34+ hematopoietic stem cells (HSC) engraftment results in the huNOG mouse with multiple stable cell lineages. The peripheral blood mononuclear cell (PBMC) engrafted huPBMC-NOG model is also available. Several humanized NOG versions with human immune system engraftments, as well as next-generation models with human immune system transgenes are currently available through Taconic.
NSG: Genetically similar to NOG except has complete null allele at IL-2 receptor-γchain on NOD/SCID background, lacks mature T, B, and NK cells, lacks complement activity, has defective macrophages and dendritic cells. HSC engraftment results in the huNSG mouse. JAX carries CD34+ and PBMC engrafted NSG models.
Note: GEMs with impaired immune systems are also available from many companies. This list is not exhaustive. Microbiome differences exist between vendors which may impact experiments.