This month, we sat down with Dr. Brinton Seashore-Ludlow. Dr. Seashore-Ludlow has a PhD in organic chemistry and synthetic methods from The Royal Institute of Technology (KTH) in Stockholm and did a postdoc at the Broad Institute of Harvard and MIT. Currently, she is a Senior Researcher at Karolinska Institutet and the Chemical Biology Consortium of Sweden focusing on precision cancer medicine.
Introduce yourself, your background, journey and also expertise as it pertains to the 3D culture space.
Dr. Seashore-Ludlow: My name is Brinton. I started out as a chemist. I started as a PhD candidate at Caltech and then I moved to KTH in 2007. I have a PhD in organic chemistry and synthetic methods from The Royal Institute of Technology (KTH) in Stockholm.
During my PhD, I realized that I wanted to focus on the intersection of chemistry and biology, so I decided to do a postdoc at Broad Institute of Harvard and MIT with Stuart Schreiber, who is the father of chemical biology.
I looked at how chemistry can be used to probe biological questions, and I worked on a large-scale drug profiling project. That got me to the idea of working in the chemical biology area.
I have been in that area ever since, working to tailor treatments to individual patients towards the precision medicine area.
3D Cell Models for Functional Precision Medicine & Drug Development
So where does that put you now in terms of current focus with your personal research?
Dr. Seashore-Ludlow: I’m working mostly in the functional precision medicine area. A lot of what I do revolves around high-content screening or high-content characterization of patient-derived models.
We use a lot of primary cells and formats that are compatible with multiple readouts, like larger drug assays or whole plates, for example, which is different than I think a lot of people working in the 3D space.
At what point did 3D models pop up on your radar? What research questions do you think they will help you answer?
Dr. Seashore-Ludlow: We’ve been getting patient cells from ovarian cancer patients, and these can come with many different cell types. It’s the cancer cells themselves, but then you get stromal cells, you get blood cells, you get lots of different cell types.
And we really wanted to be able to assay these quickly to see which drugs they responded to. But we found that when we tried to do this in a 2D setting, that over just the course of a three-day assay, we would see extreme growth of the fibroblast, but not so much growth of the cancer cells.
We thought that this was really biasing our measurements. There are a few companies working in the functional precision medicine area, like Exscientia. Then there are some groups that have shown that if you focus on cancer cell growth that you have a better predictive value, or better recapitulation of the in vivo context. So we thought, OK, we’ll try this in 3D.
And there we seem just to maintain growth of the cancer cells and minimize the growth of these other unwanted peripheral tumor microenvironments cells. So, that enabled us to streamline our assays.
Outside of the functional precision medicine stuff in in oncology, you’re leading a Vinnova-backed project focused on scaling the use of 3D liver spheroids in preclinical drug development. Where are you in that project, and what are the main project goals?
Dr. Seashore-Ludlow: The question behind that project was, ‘how do we take patient models that we know we can develop and better use them for drug discovery?’.
Because we know there’s a lot of heterogeneity between patients, and we know that understanding how effective a compound is across multiple patients is an aspect that we’re missing prior to clinical studies. This idea stems from the functional precision medicine world.
And we’re trying to meet the drug discovery world.
One of the key drivers for why 3D and patient-derived models aren’t used is that they’re often deemed to be more heterogeneous and more expensive. The expense is driven by needing extremely expensive cultured media.
It’s also thought that you won’t have enough material to support the life cycle of a product that’s run in drug discovery over multiple years. What would really help address both the challenges with cost and material availability is to miniaturize and get as much information out as possible.
That’s been the driving factor for the project.
It would really help my own research as well, because if I can perform the same drug tests in 100 cells, I can test many more drugs than if I have to test in 3000 cells.
Around that time, AstraZeneca was looking internally at what they should do to improve their own research. And Steve Rees came with this idea that he wanted to do 1000 measurements in 100 cells over 100 patients, which brought together all our expertise towards this project.
Where is the project currently?
Dr. Seashore-Ludlow: We’ve spent a lot of time looking at how to miniaturize. At first, we weren’t sure we were going to be focusing on 3D. But it really became clear that if we needed to work in precompetitive space, we needed to focus on certain types of toxicology or safety aspects.
And it became clear that iPSC or hepatic models in 3D better recapitulate the in vivo biology they do in 2D. 2D hepatocytes lose a lot of their features that we think of as hepatocytes very quickly, within one to two days.
That’s where we are now. Working on miniaturizing these 3D models.
We’ve set up a pipeline for qualifying these models. So, a range of different:
- Assays
- Compound treatments
- Measurements
that we think are required to say that a model is a good model.
Once that’s done, the plan is to conduct a screen on the miniaturized models.
Challenges With Scaling 3D Culture in Drug Development
Depending on who you speak to in pharma, some people are very pro 3D models and others push back. What are the remaining questions that are causing some people to still be to doubt the value of 3D culture?
Dr. Seashore-Ludlow: In general, there have not been many high-throughput assays adapted to the 3D setting.
When you add primary cultures on top of the 3D setting, you’re really making your life difficult. Meaning that you could potentially run into data that’s so heterogeneous that you can’t compare it between different runs.
Showing that you’re going to get the same readouts over multiple testing times is necessary to prove that your method is valuable.
There’s been a push in recent years to add high-content imaging to 3D methods. This has been challenging mostly because you get light diffraction, and with larger spheroids there’s not much you can really do to avoid that while maintaining high throughput.
It’s a drawback of the method. And there as well is where miniaturization can help. To be able to use high-content screening and look all the way through a spheroid.
I think it’s also taken time until we had microscopes that were compatible with this type of readout and well plates that were compatible in 3D settings.
We need to have all the pieces in place to validate and show that these 3D methods are better.
There is also a disconnect between academic standards and industry. I usually work with something that pharma would call low-throughput screening.
But as an academic, I call it high-throughput screening. The numbers, they are just totally different.
For me, it’s OK to store 30 gigabytes of data per patient. But for their data sets, especially image-based, data storage becomes a huge challenge. I can understand why they might have some questions about whether it’s necessary.
A final perspective in that area is that there are still a lot of questions about if we should do target-based screening or should we do phenotypic-based screening. I think the industry is still up in the air about that.
There are a lot of reviews from around 2018 looking at what has been the most successful. I think one consensus is that new entities generally come from phenotypic-based screens. You’re more likely to find a new mechanism of action or a new type of chemistry that way.
What are the practical differences between a target-based screen and a phenotypic-based screen
Dr. Seashore-Ludlow: There are many levels to that.
For target-based, people usually mean a purely protein-based assay. A purified protein of some sort that you’re looking at either physically binding to, or maybe you’re looking for inhibition of, for example, a kinase reaction.
People have made this a little bit more fancy by over expressing a target within a cell and then looking at expression or localization of a specific protein within a cell or using reporter assays.
For example, if you have a target in mind and when you treat with a certain compound you see localization of a protein from the membrane into the cytoplasm.
Then you can do phenotypic screening that is unrelated to any targets. You’re just looking for compounds that are active and change your phenotype. In cancer, for example, you often look for cell death. In some areas you’re looking for differentiation, or de-differentiation. You change the morphology or the state of the cells.
This can be done through any mechanism. You just screen and look for something that reverts that. This comes with challenges as well, and there I think we haven’t been ready to categorize those mechanisms the same way in 3D that are ready to do in 2D.
On the other hand, if I understand it correctly, that’s also where the 3D models have the most potential, right?
Dr. Seashore-Ludlow: Yes, exactly. In 3D, you can maintain the cell-cell context that you can’t do in the 2D setting, and you can have multiple cell types in an organoid that can form a physiologically relevant architecture.
What other specific reasons are there for using a 3D model instead of 2D in a drug discovery setting?
Dr. Seashore-Ludlow: One other aspect to consider is drug diffusion, or drug uptake. This is very different in 2D and 3D. In 3D, especially in the liver, some channels within the structure change how drugs are taken up. So, I think there are aspects of that are better studied in 3D.
When I talk to companies or academic groups who are looking to develop these models, 99% are all opting towards making very complex, heterogeneous models.
I don’t see many going for the smallest and simplest models that you are going for in the Vinnova project. Why is that?
Dr. Seashore-Ludlow: The ultra miniaturizing is driven by both the cost of the assay and the amount of information you can get from the material. In the functional precision medicine area, where material comes from cancer patients, you want to get as much information as possible from the limited amount you have.
What I see with the smaller models is that they’re often accessible to many of the typical assay types that we’re using. So, we can do imaging and typical readouts with a plate reader that are not always available to the larger models.
With the larger models, we can’t do imaging through the depths to get an understanding of the single cell architecture unless you go to things like light sheet microscopy, which is low throughput.
There are some intersections between the miniaturized models we’re working on. If they do recapitulate an organ, then we could combine them with multiple organ types to assess toxicity together. I think this is an interesting question. How do we combine these miniaturized models together?
I think this would answer some questions that I think that we’re currently not answering today.
I think a lot of the larger organoids that people are making have major drawbacks for screening purposes. Especially, if you want to go into functional precision medicine. A lot of them take time to establish.
You’re talking about three to five months to establish a line that you can culture continuously and use. And during that time, you might lose a lot of your primary cell phenotype that you had originally. There is also phenotypic drift that happens in the cultures.
I just heard a really interesting talk where somebody did a large-scale screen in organoids. They found what they thought were really interesting hits. But, every time they tried to validate those hits in the screen, after several passages, they didn’t get any response at the end.
The reason was because it was a Wnt inhibitor, and during culturing in the Matrigel that you use for organoids, they were driving Wnt resistance. So, they had a large-scale screen showing a dependence on signaling from the Matrigel.
Drug Development Requires Different 3D models for Each Stage
You mentioned some of the challenges with working with these larger, more complex models. Of course, throughput is one thing. No one’s really cracked the code of vascularization, and also imaging through them is difficult.
If some of these major challenges are solved, how do you see the interplay between them and the miniaturized models? Is there room for both?
Dr. Seashore-Ludlow: As you get closer to clinical translation, sometimes I think the larger ones will work there if they have more cell types. Because it’s hard to have representative number of cell types in a model that’s 40-100 cells. So, if you need that specific mixture of cell types in a miniaturized model, I think you’re going to end up going to something larger for statistical reasons.
And there I think those ones are similar to organs-on-a-chip, which are also interesting and a good complement to what we’re working on in the consortium because they answer very different questions.
But you will never be able to test them with many drugs. You can use them for 1-10 drugs, but you’re never going to use them for 1000 drugs. So, I think it depends on your question and where you are.
In order to find a chemistry or a drug that you’re interested in, you need to do some kind of screening. Once you have that, and you have a good understanding of exactly what you’re after, then I think you can go for these larger models where people have also previously gone for mouse models.
What are some of the common questions that you see people trying to answer with an organ-on-a-chip versus the miniaturized model?
Dr. Seashore-Ludlow: My idea is that you would be looking more at systemic toxicity models or looking to actually replace a mouse model, which is a very good idea because we know that mouse models are not necessarily predictive of a human response. So having a way to assess systemic toxicity without testing in a human is really valuable. And we’re not doing that in miniaturized models.
Probably it can be done, but we’re not there yet.
One thing that you do very well is put together different constellations of people for, for example, grant applications. It’s great because there’s still a lot of work being done in silos.
What kind of advice do you have to people on how to incentivize different stakeholders to come together and execute a project like the Vinnova-backed Enabling Near Patient Drug Development project?
Dr. Seashore-Ludlow: A lot of times we’re solving similar problems in academia and in industry. We just have different approaches to the same general problems. And oftentimes, we talk about them in different ways.
In pharma, for example, you have a line manager and different types of leadership structures that, for me, wasn’t clear before I actually started this collaboration with AstraZeneca. It’s been very different to learn how to communicate. We also have different respect for timelines.
In academia, we tend to be late with everything. We tend not to be organized, which is the opposite in pharma. The vision is often there and the processes to get to those visions are often in place. It’s that you need to break those down to have a new project, which is complicated because they take time to put in place.
Then I think it’s just about talking with different people. Most of the time, it’s easy to work with people, but you have to find that common communication style.
The Best Events and Resources for 3D Cell Culture
How do you stay up to date on recent developments in the field of 3D cell culture, whether that’s research groups, publications, events, newsletters, etc.?
Dr. Seashore-Ludlow: I skim Twitter for new publications. There are some events that I like to go to in the cancer fields. I think AACR is one where you can get inspired by models. People have made really cool and advanced models there, not so much for screening. It’s an interesting place to go to think about what I could do if I made certain models screenable.
In terms of screening, I think the ELRIG’s Advances in Cell-Based Assays is really fun and often has a lot of 3D areas in it. And same with SLAS’s Building Biology in 3D, which I think is also nice.
Being on the infrastructure side of CBCS, we get into a lot of projects, which gives me the opportunity to understand what people are doing. Many people are moving towards 3D patient-derived models in their research.
CBCS is part of something called EU Open Screen, which is a European Research Infrastructure Consortium (ERIC) that brings together screening sites across Europe.
There we also have projects that come in, and you can really see what is cooking across Europe.
For example, there is one called Impulse where we’re looking at how to improve the use of advanced models within our own ERIC. We’ve been doing a mapping of what is available out there. How do we do a better job? How do we offer this to potential clients?
Want to connect with Dr. Seashore-Ludlow? You can find her here:
LinkedIn: Brinton Seashore-Ludlow
