The Use of Animal Models for DIPG

Diffuse intrinsic pontine glioma (DIPG) is a rare tumor that arises in the pons of children and is currently incurable. The only active therapeutic agent is radiation, which unfortunately provides only temporary relief. Clinical trials for the past 30 years evaluating novel agents have failed to identify additional active and effective agents against this tumor. In recent years, molecular genetic technologies have been quite successful in identifying new molecular targets in numerous cancers which has led to targeted drug development, and an increasing number of promising targeted agents. The challenge is how to determine which novel agents or combination of agents should move into clinical trials for DIPG. As this is a rare tumor, it is impossible to test every new agent and combination of these agents in these patients. There are not enough patients to accomplish this feat. An additional complexity to new drug development is that DIPG tumors are heterogeneous and may be comprised of multiple different subtypes, where each subtype may respond differently to specific drugs. 

Why Do We Need Animal Models?

One idea that scientists thought might be helpful is to develop animal models that would allow for screening of novel agents and combinations. The results from these animal models could be predictive of anti-tumor activity in children with DIPGs, leading to the discovery of the most promising drugs for human DIPG trials. At this point in time, this ideal has yet to be fully realized. There is currently one main obstacle that must be overcome so that predictive animal models can be developed. Scientists need a better understanding of the genomic alterations that drive the growth of human DIPGs so as to guide the development of accurate animal models to potentially treat it. Unfortunately, scientists do not have a good understanding of the drivers of these tumors, although there are several research groups who are currently working on this, and it is believed that answers to these questions are within reach. 

In spite of the limited understanding of the biology of human DIPGs, there are several animal models that have been described, although it is not clearly known if they are predictive. In the next few paragraphs, I will review the various DIPG models that currently exist, and the advantages and disadvantages of each.

DIPG Animal Models Available

Most of the current animal models for DIPGs are rat allograft models and only recently a genetically engineered DIPG mouse model was developed as well.

The three main types of animal models for glioma are:

  1. Allograft: Chemically induced tumors in rats that mimic DIPG.
  2. Xenograft: Human tumor from patients, transplanted into mice or rats.
  3. Genetically Engineered Mouse Model/GEMM: Mice given new genes (transgenics) that cause tumors appearing similar to DIPG, and sharing some molecular signatures of the human tumor.

Allograft: There are several rat glioma cell lines available, which were generated by injecting rats with repeated dosing of chemotherapy until the rats developed gliomas, which were then cultured and propogated. There are currently several rat glioma cell lines available. Three of these cells lines (C6, 9L, and T9 gliomas) were induced by repeated injections of methylnitrosurea (MNU) to adult rats. Two other cell lines (RG2 and F98 gliomas) were chemically induced by administering ethylnitrosurea (ENU) to pregnant rats. In this case, the progeny developed brain tumors that subsequently were propagated in vitro and cloned.

Both MNU and ENU are alkylating agents, which mean that they damage DNA by adding alkyl groups to it. Most of the above mentioned rat glioma cell lines are being used to generate brainstem gliomas by direct injection into the brainstem of either rats of the same strain or immunodeficient rats. Depending on the number of cells injected, the cell line used, and the age of the rats at the time of injection, the rats go on to develop brainstem tumors one to several weeks later. A head-to-head comparison between 3 week-old and 10 week-old rats injected with the same cell line and the same number of cells into the brainstem demonstrated that the microenvironment of young rats allows for the formation of diffuse pontine tumors, while the microenvironment of older rats allows for the development of focal brainstem tumors. Interestingly, there have not been many trials testing systemic chemotherapy using such models. This is most likely due to the belief that systemic chemotherapy for the most part does not get into the brain tumor due to the blood-brain barrier (BBB). Therefore, most of these models have been mainly used to test for CED (convection enhanced delivery) of chemotherapy such as carboplatin.

Xenograft: Recently, a rat xenograft model was developed whereby adult human glioma cell lines were implanted into 6-week old immunodeficient rats. Prior to implantation, some of the cell lines were maintained in media with serum (these are usually grown as cells adherent to plastic); some of the cell lines were maintained as subcutaneous xenografts (grown under the skin of immunodeficient rats); and one cell line—the GS2 cell line was maintained as neurospheres (these are spherical colonies in suspension in media). 

It is well documented that gliomas which are cultured in epidermal growth factor (EGF) and basic fibroblast growth factor (bFGF), and are grown as neurospheres, have genomic signatures that most resemble the signatures of naturally developing (in vivo) gliomas. While an advantage of such a model is that the tumor cells are human, the disadvantage is that adult glioma cell lines have different genetic alterations than DIPGs. In addition, the in vitro culturing step likely alters the biology of the cells even in neurosphere conditions. Lastly, such models remove the role of the immune system in DIPG tumorigenesis.

GEMM: Genetically engineered mouse models for brain tumors have been developed since 2000 and are a more recent addition to the animal modeling toolbox. The advantage of such models is that the genetic alterations which initiate and drive tumor formation are known and recapitulate the genetic alterations present in the respective human tumors. Therefore, the genetic alterations of the respective human tumors should be determined so as to guide the development of the genetically engineered mouse model. Such models are helpful in determining if a particular genetic alteration can drive tumor formation. Not all genetic alterations are equally meaningful and scientists divide genetic alterations into “drivers” and “passengers” to imply that only certain genetic alterations can drive tumor growth (so-called driver mutations) while the role of other genetic alterations is less clear (so-called passenger mutations). 

There are numerous technologies that can be used to generate these mice. One brain tumor model uses conditional knockout mice where mice that have lost one copy of p53, PTEN, and NF-1 develop brain tumors through loss of heterozygosity. A second system is the RCAS-tv-a system, which allows for oncogene delivery by injection of virus producing cells into areas of interest such as the brainstem. Normal mice do not express the receptor (called tv-a) and so are not susceptible to infection by RCAS vectors. However, two transgenic mice were developed that express the tv-a receptor in progenitor/stem cell of the brain compartment: nestin tv-a mouse and GFAP tv-a mouse. (A transgenic mouse is a mouse which integrates an additional piece of DNA in its germline called a transgene, and so every cell in the mouse acquires this extra piece of DNA.)

Recently a genetically engineered mouse model for brainstem gliomas was developed which recapitulates the genetic alterations of a subset of the human disease. It was recently observed that PDGFRa is amplified in 20% to 30% of DIPGs which means that the receptor for PDGF ligand is expressed in high levels in a subset of DIPGs. The derived mouse model uses the retroviral delivery system described above, whereby a virus is used to deliver oncogenes to specific areas in the mouse brain. Tumors are generated by the over-expression of PDGF-b in nestin positive cells which line the floor of the 4th ventricle at postnatal day 1 to 3. Nestin positive cells are cells that express nestin, an intermediate filament that is expressed in progenitor cells in the brain. Overexpression of PDGF-b results in the formation of low-grade brainstem gliomas. The addition of Ink4a-ARF null genetic alteration (which has been described as a common alteration in human DIPGs and normally restrains cell division), together with PDGF-b overexpression, results in the formation of high-grade brainstem gliomas, or DIPGs, with high incidence by six weeks of age.

Advantages of genetically engineered mouse models are that a) the genetic alterations are clearly defined; b) the tumor forms in the right microenvironment (the brainstem); and c) the tumor forms at the right time period (pediatric). It is generated completely in vivo and develops de novo in the mouse. As tumors are a complex cellular setting, it is important that this environment is as close to reality as possible. Another advantage of the genetically engineered mouse model is that it can be used to determine the cells-of-origin for a particular tumor. The cells-of-origin for the recently developed DIPG mouse model were derived from cells lining the floor of the 4th ventricle and aqueduct. However it does not tell us with any certainty that the cells-of-origin for human DIPGs are similar cells. One disadvantage of this animal model is that it is probable that this genetically engineered DIPG model may be oversimplified, as human DIPGs likely contain more genetic alterations than simply PDGF-b overexpression and Ink4a-ARF loss. It remains to be determined whether therapeutic agents with antitumor activity in this animal model will also be active in children with DIPGs.

There is no perfect animal model. Ultimately there is a need for a predictive model of activity in the clinic. An added complexity is that the human tumors are heterogeneous and so it is likely that we will need to classify them into several groups based on their genetic alterations. Each subtype will then require specific therapy and an associated specific DIPG animal model. Of note, adult gliomas have recently been subdivided into three groups based on the genetic alterations of the tumors.

One advantage of rat brainstem glioma models is that the rat brainstem is larger than the mouse brainstem. As a result, it may be easier to conduct CED preclinical studies. The disadvantage of rat allograft models are that in most cases the genetic alterations of the tumors are not clearly defined and may change over time. Most of the cell lines are maintained on plastic dishes that are quite artificial. In addition, as mentioned, human tumors are complex with several cell types interacting including tumor cells and various stromal cells such as blood vessel cells, support cells, and immune cells. Therefore it is ideal for the animal to develop the tumor within a normal immune system and with all of the support cells being present from tumor onset.

Preclinical Testing

The question that often arises, is how much preclinical evidence is needed before a decision is made to move a new agent into the clinic for DIPGs? There are several levels of preclinical evidence and I personally believe that novel agents should move to the clinic after full preclinical testing has been done. This means that a novel agent has been tested in cell lines; in DIPG xenograft rat models where the xenograft originated from a DIPG tumor that has been propagated in vivo or, second best, neurospheres; and has also been tested in genetically engineered DIPG subsets. Depending on the therapeutic agent, some agents cannot be tested in cell lines at all and can only be tested in vivo or in neurospheres. An example is the sonic hedgehog pathway inhibitors, which cannot be tested in cell lines grown on plastic dishes, as the pathway is not functional in such conditions.

Once an agent shows strong promise in preclinical testing, a decision is then made to test it in patients with DIPGs. The first clinical study is a phase I study which is used to determine the correct dose to use in the clinic, as well as assess for toxicities of the drug. Even if a drug has already been tested in adults, it will still need to be tested in a phase I trial for children as children at times metabolize drugs differently than adults. Usually phase I studies are open for children with diverse cancers but it does not mean that a particular drug will not be active in DIPGs. Once the safety and dosing are established in a phase I study, then a phase II trial is designed to determine if the drug or drug combination is active against DIPGs. Phase II studies are usually done on a selected tumor subtype. Phase III studies are large studies that are used to confirm promising results from phase II studies.


At this point in time, most clinicians do not believe that animal models can help guide which therapeutic agents or combination of agents will be active in the clinic. The burden of proof lies with the animal-modeling field to continue to improve the animal models so that eventually they will be predictive of activity in the clinic. Similar to what has been done in adult gliomas, in the near future DIPGs will also be grouped into genomic subgroups, and genetically engineered animal models will be developed for each subtype. It is my hope that these genetically engineered DIPG mouse models will be predictive of activity in the clinic, but it remains to be seen if this will indeed be the case.