17th June 2021Cancer Models Forum
What do the next generation of humanised models need to look like?
Our Team recently attended the PREDiCT Innovation conference on Advancing Precise & Translational Preclinical Models for Evaluating Immunotherapies and we wanted to share some of our key takeaways from the "Panel Discussion: What Do the Next Gen Humanized Models Need to Look Like?". The panel included a mix of speakers from genOway, CrownBio, Pandion Therapeutics and IGM Biosciences.
Below we have put together a bullet point summary for each of the questions to provide an easy-to-read overview of the panel answers.
Q: What type of new immune-oncology models are biopharma interested in seeing?
A (Pandion Therapeutics):
- Interested in seeing preclinical models which possess both the selected key immune target, as well as other relevant components of the immune system context.
- Especially important if drug target is expressed by more than one cell type in the immune system (e.g. PD-1), as this can help to predict potential adverse effects.
A (IGM Biosciences):
- Interested in models that enable investigation of different T cell populations (memory T cells, central memory T cells, effector T cells, etc.) and their dynamics in response to antibody treatment.
- Also, interested in models that enable investigation of cytokine use for enhancing NK cell activity.
- Growing trend among biopharma towards looking at modulators of other immune cell types besides T cells – work is being done with vendors to develop these types of 2nd gen mice.
Q: What would be the best models to understand cytokine release mechanisms and adverse effects in preclinical trials?
- There isn’t a best model per se, because a model will always be deficient in some area so it is important to consider the MOA of a therapeutic and select preclinical models accordingly.
- For a T cell therapeutic, the magnitude of the cytokine response and type of cytokine produced is partly dependent on what elements around the T cell are being triggered by the antibody/T cell engager/etc.
- Important to keep in mind that any therapeutic, as long as it is targeting the immune system, will have a certain level of toxicity as a result of engagement with a given cell population and associated downstream effects.
- Having a model in which the immune system is as close as possible to the human situation (e.g. BRGSF-HIS mouse model) helps to provide a better assessment of potential toxicity.
Q: With CAR-T cell therapies being a commonly investigated indication, how could BRGSF-HIS mice be used here and is there any experience in looking at cytokine release syndrome (CRS) toxicity in CAR-T cell therapies?
- In terms of CAR-T cell therapies, BRGSF-HIS mouse models have been used in a context where there was no human immune system i.e. recipient mice with multiple melanoma cells were used to show that the CAR-T was functional.
- Model was also used to show that CAR-T cells could be depleted on demand due to the CD20 epitope present in their cell membranes, and protection against multiple melanoma was no longer observed here.
- The BRGSF-HIS mouse model has never been used in a context where the human immune system was implanted and developed, and then CAR-T was used for treatment. The main concern here is whether other reactions would occur between the CAR-Ts and the human immune system developed in an animal.
- Bottom line is that mouse biology is different from human biology so there is no ideal model that can answer all the questions with regards to toxicity.
- With CRS, we don’t focus so much on finding mice with a complete immune system - we do however want to make sure the drug target is present and supported. This allows us to check whether there is dysregulation of cytokine within a target cell upon drug administration.
Q: What does the next generation of humanised models look like? Will we ever have affordable, off-the-shelf models?
- The price of a model goes down with the use of the model. The price structure of a model depends on the price of model generation and breeding costs.
- While a lot of model generation can be done using CRISPR-Cas9, the breeding itself makes up the most expensive part of the process – there is potential to reduce costs here as the more animals are bred, the lower the cost per animal, however in order to do this, there has to be some visibility from biopharmas on the need for those models.
- This could save biopharma researchers up to 60% on the price per unit by planning experiments in advance and sharing future study plans with CROs.
- In terms of next generation models, there is a trend towards a desire for more and more complex models. Expectations for models are becoming more demanding, which means that models are becoming more expensive as we move away from generic models.
- One model type still missing in the context of syngeneic models is models expressing human Fcy receptors with a human-like expression pattern. We are currently working on this, however making these models is super complicated because you have to swap around 200kbs.
Q: What is the main message you would like to give the audience?
A (Pandion Therapeutics):
- Don’t be afraid of using humanised mouse models, but also don’t be blind to their limitations.
- Take different approaches to cover blind spots and weaknesses of different models and really understand what you’re actually doing.
A (IGM Biosciences):
- Although humanised mouse models have some limitations, they can really help you with your studies and are particularly useful for biomarker assessments.
A (Crown Bio):
- Humanised models really take research to the next level, especially when looking at toxicity (no need for surrogates as with syngeneic models).
- There is a plethora of humanised models out there, but it is important to refine your research question and thus model criteria in order to select the best models for your research.
- When picking a model, it is important to be nosy about how the model was generated, what was replaced, what was deleted, etc.
- For instance, there are many different types of PD-1 knock-in mice and the precise model selection can really impact the experimental setting and quality of data generated.