The importance of patient-specific models: an interview with Stephanie Seidlits

Written by RegMedNet

In this interview, Stephanie Seidlits, Assistant Professor, University of California, Los Angeles, discusses her research into patient-specific models for studying glioblastoma.

Jump to questions:

Please introduce yourself and your institution

I’m an assistant professor at the University of California, Los Angeles (UCLA; USA). I’m in the department of bioengineering trying to engineer neural tissue and, although I am fairly new here as an assistant professor, starting in 2014, I have been working on similar projects since my undergraduate days.

Last year, you published a paper on integrating the glioblastoma microenvironment into engineered experimental models in Future Science OA. Can you explain how this research came about?

Glioblastoma is a kind of brain cancer and I hadn’t actually worked in brain cancer models before becoming a faculty member at UCLA. Much of what I had been doing before was developing biomimetic cultures that we could use to study repair and stem cell response after brain and spinal cord injury. In talking to people at UCLA that were working in glioblastoma I realized there’s a lot of crossover, not only the cells we’re looking at but also the brain microenvironment, which seems to be very important in glioblastoma as well.

Glioblastoma is a very devastating disease; it’s the most common kind of cancer that originates in the brain but it’s essentially untreatable. Most people will succumb to the disease around 15 months from diagnosis so it seemed like a really important problem to get a handle on.

Back to top of page

Why is it important to consider the microenvironment of the native tissue when developing engineered experimental models?

One problem we are facing is that when you take cells out of a tumor, for example in a biopsy to study them in a lab and start to understand that cancer and develop treatments for it, often the way you culture these cells is put them in a dish surrounded by some sort of liquid medium. They grow into a ball of cells which respond very well when we treat them with drugs but that doesn’t translate clinically.

One of the first studies we did was taking cells isolated from the same patient and put them into a mouse model. We established tumors from these cells in the mouse’s brain and in a pocket under the skin on their back, and treated them with different drugs. We found that in tumors we established in the brain, it might take, for example, 20 days for the tumor to gain resistance to a drug and the mouse would die very quickly after that. However, in tumors established under the skin, it would take over 200 days for the cells to stop responding to the drugs. The cells are from the same patient and the tumor is genetically identical but they’re behaving very differently depending on what’s around it.

When we look at tissue in the brain, we see very different biochemical compositions compared to almost any other tissue in your body. There are different sugars and proteins that are around and activating the cells, and we also see differences in the physical microenvironments. The brain is also different mechanically and this physical signaling to the cells which also seems to be important for their behavior.

Back to top of page

Were the models described in your paper patient-specific? Are patient-specific experimental models necessary for current therapy development?

In the Future Science OA paper, we argue that we do need patient specific models and there are a couple of reasons for this. First, many of the cell lines that people have used over the years to study glioblastoma don’t seem to reflect how patient tumors act and the longer you keep the cells the more different they become from what they were originally. That’s a problem for developing drugs.

What we propose to do is use cells from individual patients and utilize them in models. There’s a lot of heterogeneity in tumor properties between patients, but even cells from a single patient can have very different responses. Trying to capture all of this heterogeneity seems to be really important for understanding tumor function. When we take lines that are isolated from different patients and put them into these models, we expect to see what mechanisms seem to be common across many different patients. This will help us find other drugs that might work in a single patient but not across many patients.

At the same time we can start to diagnose a specific patient. In contrast to animal models which would take months to establish, these tissue engineered models can provide answers much quicker and on a time scale that might be clinically actionable. We can test the patient ourselves and use this information to determine a treatment strategy.

Back to top of page

How close are we to existing engineered models replacing in vivo models? What more is required to get there?

There’s still a lot of work required to validate these models and get clinically translatable information out of these models. In a recent paper, we’ve shown equivalence of these in vitro models to an animal model, as far as mimicking the kinetics of drug resistance. However, even these animal models are not a perfect prediction of the human body. Even if we get close to those animals models, we still have to figure out how close they are to clinical models.

On the other hand, it’s unlikely we’ll ever completely replace the in vivo models because of the level of complexity. For example, there are things that might affect the cancer or the treatment that we can’t easily recapitulate in a culture, such as an intact blood brain barrier, or a functioning immune system with multiple glands, the endocrine system and the spleen that work together.

Our goal is to use these stem cell models to help us decouple all of the complex factors so we can really start to nail down the specific factors in the disease that are actionable as treatment. This will allow us to look at it in a more systemic way. We need an integrated approached, where we verify if our findings are still true in vivo once we add back the complexity. We’ll keep refining these models by going back and forth. 

Back to top of page

You are currently working on regenerative strategies for spinal cord injury. How does development of a microenvironment that is delivering therapeutics vary from one that is simply supportive?

In some ways, all microenvironments are very biologically active. If you add any sort of microscale structure or you’re changing the thickness of an interface, you affect the cells biologically. When we want to deliver a therapeutic, especially in the nervous system, it becomes very important that these materials are injectable. For example, in the case of spinal cord injuries, you don’t want to take out tissue to be able to implant a therapeutic. It needs to be minimally invasive and injectable where it will interface with existing tissue, and there’s a lot of different ways you can. You could have local delivery of a therapeutic where it just diffuses away and it’s less controlled or the cells could start breaking up the material that was delivering the drug so that the release becomes timed.

Back to top of page

What are your predictions for the development of engineering models in the future?

A big question in a lot of fields right now is what is the minimal level of complexity required to understand tissue function? You can always keep adding complexity as the body is incredibly complex but you don’t necessarily need it to answer specific questions. If I’m trying to deliver a drug that I want to cross the blood brain barrier I might need to have an intact artificial blood brain barrier in my model but if I’m looking at the interaction between cells and the vasculature around them, I might not need that feature. Your models become more specific to the questions you’re trying to answer.

As far as clinical applications, what we trying to do now is transfer these models. We can print out a high throughput array of the different cultures and we think this will be really valuable because we can take patient cells, produce 100 different cultures and treat them with 100 different drugs. At the moment in glioblastoma, we are seeing many different subtypes among patients but none of these have translated to robust improvements clinically when the specific genetic defect is drugged. If there was a way to see how the cells would respond, which hopefully would be very quick, we could treat a patient based on that functional response.

Back to top of page

Further reading: