Professor Michael Shuler, Professor of Engineering, Cornell University
Date: 16th July 2021
Michael Shuler is Professor of Engineering, Cornell University, President and CEO of Hesperos Inc. Nomthandazo Ziba is completing a master's degree in pharmacology at Coventry University. This transcript has been edited for clarity and brevity.
Photograph of Professor Michael Shuler, credit: Shuler. Born in 1947 in Jollet, Illinois, Shuler did an undergraduate degree in chemical engineering at Notre Dame and then a doctorate in the area at the University of Minnesota. In 1974 he joined Cornell University where he started as assistant professor in biomedical engineering and became the Director of the School of Chemical Engineering. He later became the Founding Chair for Biomedical Engineering.
Thank you very much for joining me. To start with I would like to ask how you came to be involved in the organ on a chip technology?
I've been involved for a long time. When I first started, which was about 1989, I had the idea of using human cell cultures and animal cell cultures in drug development, but in a way that was integrated like the body based on what's called a physiologically based pharmacokinetic model. We did that for about 10 years using a macroscopic system. In 1998, I met Greg Baxter who was based at Cornell's nanofabrication facility and we decided to build microscale systems, linking pseudo organs together, as they would be in the body, using the physiologically based pharmacokinetic model as the basis for that. Such models were developed in the 1960s. A physiologically based pharmacokinetic model is a computer model, in which each compartment represents a different organ or tissue type and allows you to look at the interactions. The way that people used such models was to estimate drug distribution. For example, I can predict what happens for different doses and routes of administration. So we were interested in using this system as a way of improving drug development and discovery and to try to understand what might happen in the human body.
Just to go back a couple steps. Only about 10% of the drugs that enter human clinical trials come out as approved products. I talked to major pharmaceutical firms and something like two, five, six percent of animal models predict what's going to happen in humans. Most animal models therefore do not predict what's going to happen in humans. So we were motivated to try to find a human based model, which could predict what would happen in humans. Also, we looked to build animal based models, which allows us to ask some questions about mechanisms for drug metabolism we couldn’t ask otherwise. So that was the whole basic idea.
And in 1998 Greg and I realised that it was fairly expensive to do this with large cell cultures and flasks and stuff like that and we needed to make it much more efficient in terms of space and cost. By going to the microscale we could get by with cell cultures which are much smaller. It's still large in some sense, it's not a single cell, it can be 10,000 cells, or 100,000 cells, depending on the organ size. When we build the systems we build them based on the pharmacokinetic model, and in this case the liver is much larger than the pancreas etc, so the relative sizes of these organs has to be right too. In this case the number of cells in the liver model would be more than the pancreas model etc. We have a 2001 publication on a prototype and then we have three papers from 2004, for a workable system. We used this model, which was based on the toxicity of naphthalene and determined why it causes different toxicity in mice and rats. They respond differently (Viravaidya, Sin, Shuler). With the experimental model plus the modelling we could identify the reactive metabolite, (naphthoquinone) which was important because it was not originally known, and also, why rats and mice respond differently to naphthalene.
As time went on there was more and more time and interest in the field for using it in humans. It still makes some sense to do some of this in animals, particularly to test the mechanism of drug action and to test that what we're doing is actually real. Because we can test things on animals we can’t test in humans it is important to verify that the approach works in a animal model. But it also gives you some insights into mechanisms, even if you're just using animals. So we do some animal studies and try to see if the model that predicts animal responses is accurate then see if we can make a model that can predict how it works for humans for which we now have plentiful data. Most of our work right now is associated with humans. As I mentioned, the animals are not a very good predictor of what's going to happen in humans.
What we wanted to do was to reduce or ultimately replace the use of animals. Where we are right now is attempting to reduce reliance on animal models. Because we're now using human models, we can make better predictions about what's going to happen in human clinical trials. The big payoff for society and for drug companies right now is that you lose a lot of money in drug development. About half the drugs now fail at phase three and that's very expensive. So if, before you even start clinical trials, you have a better idea which drugs are going to be effective in humans and what dosage is effective in humans, and what the side effects might be and how you might mitigate them, then you can make the development process more effective. If instead of one in ten drugs currently come out as a useful product, could become instead one in five or one in three, that would lead to huge savings in terms of drug development costs. That would be a huge benefit for the consumer, because a lot of the cost of the drug is linked to the clinical trials. Also, all the drugs that go into clinical trials don't come out as products. I think the big benefit will be for societies to get more useful drugs. A lot of times in the preclinical trials, you look at different forms of the drug, and if all you're doing is just looking at animal trials, and a single human cell type in a dish, it’s not very authentic approach to predicting human response.
It may turn out that you're not using the right selection criteria. The real selection criteria needs to be what drug is going to be effective at a certain dose and not have any adverse side effects. Or at least if we know there might be side effects, we can set up clinical trials to mitigate those. In other words if you get drug A and give it with drug B it might mitigate potential side effects. Also, it can take into account that often drugs work well in 95% of the population but in the other 5%, because of their genetic background, may not work so well. So you can also use this system to screen for that issue and say this drug is suitable for people of a certain genetic background, but you really shouldn't give it to people that have another specific genetic background. It is the ultimate personalised medicine. I can put cells from your body on a chip and treat it with candidate drugs. I might say, for your case, this would be the best drug, and for somebody else's case, a different drug. That’s not what we are doing now but it could be done in the longer term.
So we started working on the microscale system in 1998 and in 2004 we published three papers, and the work has evolved quite a bit since then. When we first started, our studies were relatively short term, six hour runs, and they were based on tissue cultures. They were not 3D cultures, and not organised tissue cultures. A lot of the evolution has been to use this same framework, but use improved models of different organs in different organ compartments. I don't like to use the words 2D and 3D so much. I prefer to use the words organised and disorganised cultures.
Why, is that?
Because, let's say if I'm looking at your heart and I want to look at measuring conduction velocity in your heart, I can do that with a 2D system. I treat the surface, so I force the cells to attach just along a pattern with chemical cues for cell attachment and then I can measure conduction velocity. But if they're just in a big disorganised monolayer of cells, you can't measure electrical conduction velocity. The current just goes everywhere. But if you can align them, you can measure the time from when you create a stimulus to a point of action, of how long it takes for the electric current to go from one end of the pattern them to the other and observe if the drug modifies that. Additionally, with cells aligned you can do force measurement analysis for cell contraction measuring deflection of a cantilever with cells. For other tissues or organs, it is the metabolic function of cells which maybe important and that may require an organized 3 D system. But what we're really interested in is the function of cells which is where having a 3D system maybe important. As an example consider a cell model for the liver as the principle organ for drug metabolism. A simple model with a monolayer of hepatocytes may be satisfactory for some purposes, but a 3D model with multiple layers incorporating multiple cell types (Kupfer, Stellate, and endothelial cells in addition to hepatocytes) provides a more complete model. With a simple model I may capture certain behaviour, but there's other behaviours, which I won’t capture. So if we make the tissue models more authentic, one can expect the results to become better. There's always a price to doing a more complex system.
NomthandazoWhen it comes to pricing, is it going to be affordable for the normal person or even people in other communities and undeveloped communities?
I'm involved with a small company called Hesperos which in the last six years has been developing this system as a commercial technology. There's a trade-off between how authentic you want to make the system and the cost. So if I want really authentic tissues in 10 different compartments that gets to be very expensive. If I have a more targeted question, maybe where I'll only need three or four organs, I'm not ignoring other organs, I'm just saying I'm not worried about the metabolism or anything there. In the systems we're building right now, we can do it at a relatively cheap cost. We don't always use the most sophisticated cultures we could but we usually have a target that we're interested in. So for example, we may have a drug that elicits changes in neuromuscular junctions (or NMJ) which is important in a lot of diseases. The NMJ is a fairly sophisticated model. The rest of the tissue models may be less sophisticated. For example, we have a focus for example on cardiac cells. We have a very good liver-cardiac model. A lot of drugs fail because of cardiac toxicity. We have a three component model with liver, a target tissue that we're trying to treat, and cardiac tissue to tell us about side effects. Now there's certain things we’re obviously going to miss, but we're going to get a lot more good information than with only a single organ. If it's a side effect we are interested in, having the cardiac tissue in there is important. We also have many other tissue models to include. So depending on what I'm looking for, I can have other kinds of tissues in there. For a lot of drug companies, having a human based model, with even three or four cell types in it, will tell them a lot.
The other thing we also have to worry about is how the drug is administered. If it was given to you as an IV shot that's really easy. But let's say it goes through the skin, or the lung, then we have to build models of the skin and lung to measure the rate of absorption. Also consider the gastrointestinal (GI) tract. We have fairly good models for the GI tract. We have good models for the blood brain barrier, which is important because the brain is privileged immunologically. Unless you need to treat a brain tumour and stuff like that, the blood brain barrier can be a problem. If the drug is not supposed to be in the brain it can cause problems. The blood brain barrier is a very, very tight tissue, it doesn't allow many things to pass through it. So when we build a model of a blood brain barrier, or model of the GI tract, or the kidney there are barrier tissues inside those systems too. You want to make the system as complicated as necessary to answer the question you're interested in, but not a lot more complicated. I can make a system with lots of different tissue types represented but it's probably not economically feasible to run every drug through that.
We and other people are working on ways to make the system become more cost effective. That means improvements in terms of automating the system and in terms of the cell culture, which is the hardest part to do and most expensive. We have to worry about what we circulate as a blood surrogate because we want to mimic absorption/desorption.
In testing drugs one of the things we really worry about is nonspecific absorption. If I put a drug into a very small volume, the surface area is large, and the drug absorbs on the surface, you don't really see what you would see in the body. So you have to have ways to take that into account.I work with James Hickman (chief scientist at Hesperos and co-founder). He is trained as a surface chemist. We vary the surface chemistry to minimise unwanted absorption of a particular compound. But we still need to take this absorption into account so we usually run the system with that compound, try to determine what the amount of absorption is, and then actually determine what should remain in solution. We can measure the solution with HPLC for drug metabolites, and biomarkers to see if our predictions are right. We do have a fairly large number of drug companies that work with us on different compounds, different diseases.
NomthandazoI saw that you've worked with Roche before, especially with the cancer drugs, is that still ongoing?
It's not ongoing but we still talk with Roche. We have worked also with AstraZeneca and Sanofi. There's one drug that is in phase 2 clinical trials with Sanofi that used data from our efforts with a small company that they purchased to obtain regulatory approval. Most of our active grants right now to the company are actually with medium to smaller pharmaceutical firms who really don't have the capacity to try to build this into their own pipeline. Sometimes a small company using our technology is bought by a bigger firm. There's one drug that is probably in the Roche pipeline that is related to work that we've done with a very small company that they purchased.
The small companies have maybe the biggest need for us because they're not set up to do super large clinical trials. They have to partner with somebody and they want to make the best choice of product to get it to the market. Many people are working on trying to get to the point where regulatory agencies will ultimately accept the data for our “Body-on-a-Chip” systems in place of animal data. We're at a point now where regulators will accept this data, as part of an IND, at least for efficacy, but to fully replace animals for both efficacy and safety, will take them a long time.
What are the key challenges in taking this technology forward and where do you think the solutions might lie?
The key challenges right now are the cell cultures. The focus right now is on induced pluripotent stem cells (iPSC). They are relatively stable and more uniform than primary cultures would be and we can propagate them for a long period of time. So you can do a set of experiments over a long period of time and be relatively certain that the biological characteristics are fairly similar.
In terms of frontiers, we have done some work with immune cells. We have had a paper on that published in 2020 (Sasserath, Rumsey, McAleer) that focused on circulating immune cells throughout the body. That's where a lot of people are going to be interested because the immune response is one place where there's often a lot of problems, particularly with biologics and immuno-suppressant drugs. If we can build a useful human immune system model that will give us a lot of advantages.
One challenge is getting the cost down. That's where we're trying to be more efficient in the devices we make in terms of trying to automate as much as we can, both in terms of the cell culture and setup of the system and also the readouts from the system. The more we can automate, the cheaper the system will be to operate. The other challenges are adding capabilities. The immune system is one of the capabilities a lot of people would like to see added. We can add certain components without the full immune system, but still enough to get some sense of what's happening. An additional area of great interest are Neurological systems. We have a lot of experience with such models. I think it's important because it is hard to replicate neurological systems in many animal models. So those are probably a couple areas that I think are important.
And how about the legislation? What do you think can be changed to make the technology go into the market a bit quicker?
This is a very heavily regulated industry, as it should be. Most regulatory agencies are now interested in this technology. They're interested in working with companies and groups on it. Their duty is to make sure drugs are absolutely safe and efficacious for humans so there's a lot of things that need to be done now. Where we'll see a lot of the breakthrough in terms of regulatory agencies is probably with rare diseases initially because there are not very many good treatment alternatives. How to use the technology for individuals, for personalised medicine is also something they're interested in moving forward. Most of the regulatory agencies have set up research groups to explore the technology. And as they become more familiar with the technology, they'll have a better idea of what it can do, and also what its disadvantages are.
I suspect that as time goes along, data from these systems will become supplemental to animal data and then ultimately it will replace animal data. It's going to take a long time before that happens. I don't know if we need any laws really changed. I think it's really that the regulatory agencies have a mission which requires them to be very conservative and to be very sure of things. It's really up to us to make sure that we provide them with the data that they can be comfortable with in determining that what we're predicting will be useful and real.
What we're focused on right now is learning which drugs you take from preclinical trials into clinical trials and developing ways the technology is going to impact treatment further down the line. Getting from the preclinical to clinical is the biggest hurdle, and also where I think where the potential biggest advantages are. Because anytime you put a drug into clinical trials, which is a loser, it means that something that may have worked didn’t get a chance to go in. And I have talked principally about pharmaceuticals because the medical industry has been relatively receptive and interested in this kind of technology. This technology also applies to other regulated compounds like chemicals in cosmetics and also food ingredients.
What has been your favourite aspect of this research?
For me it's the fact that an idea we originally came up with is now being embraced by a much larger community. When we first had that idea, most people said it wouldn't work, etc, etc. We are past that stage now and a lot of people are interested in it. People are trying to make it work in different ways. Our focus is more on the body on the chip integration of different organs. Other people are working specifically on single organ models, and there you build a much more complex biological model. When you're building a body on a chip where you have different components, you're really more interested in the interaction between the components so you may have some less complex individual organ models, although the system allows you to build the model as complex as you want.
NomthandazoCould you please share with me a defining moment for you both in the lab and from the perspective of business.
I think actually, our first studies with naphthalene to demonstrate that the system could work. We did a good job with the model showing we take the data from here and make predictions about what's going to happen, animals, etc. And through a series of papers, we put it all together, that was probably the most important part. We've had a lot of other advances that I'm proud of. But the first time I put the system together and made it work was I think the most critical.
How did the name 'organ on-a-chip' come about?
The name ‘organ-on-a-chip’ is a term I did not come up with. The field developed it. I really came up with the idea of a ‘body on-a-chip’ which also became ‘animal-on-chip’ and ‘human-on-a-chip’. But it was really the body-on-a-chip that was our initial model. Now, people are talking about organ-on-a-chip, which is a single organ or in some cases multiple organs on a chip. The body-on-a-chip is a subcategory of that in the sense that organs on a chip are not always physiologically based. Our body-on-a-chip is physiologically based. That's why we rely on an physiologically based pharmacokinetic model. The term organ-on-a-chip, I think, started maybe about 12 years ago. Back in the 1990’s we called our system a cell culture analogue because actually, when we went to get things published with reviewers, they didn't like the idea of body-on-a-chip. So cell culture analogue was what we originally started with and then that evolved into the body-on-a-chip which seemed to be a reasonable term. There was an article about what we were doing in Newsweek, in 2005 which used the term and we took it from there. But organs on a chip recently has been a compromise umbrella term between different groups. And micro physiological systems is also the other term that's used. And that was really developed by NIH. I think NIH’s goal was to have a term which was not in favour of one group over another and was an inclusive term.
Is there anything else you'd like to share, which maybe we couldn't have covered?
It's been interesting to see the field evolve from an idea. Just a while after we got started some other groups did too. There was a relatively small number of papers until 2012 but that number has really blossomed into a very large number. Now there are several thousands of papers a year rather than five or ten papers a year. The other thing is that there are a lot of companies out there trying to do this now. And I think each of the companies have a different focus, which is good. I think that's actually fairly healthy. What the real breakthrough will be is when a drug is approved, based in part on data from a body on a chip. And we do actually have one case where we have some work that was used by a company as part of their application for an IND and isnow in clinical trials for treating a rare disease. I think, once there is regulatory acceptance of the system as being useful, that would be a key positive step. Now that regulators are actually very interested in technology, and they see the potential benefits. We have groups trying to explore the technology, see what works, and doesn't. Well, once we have several drugs that have been approved by regulatory agencies based on this technology, then I feel that will go mainstream at that point. And that will be probably the ultimate validation of the technology.
I think it's interesting that you feel there's more room for you, with the smaller companies as compared to the bigger pharmaceutical companies.
There's two things about that. One is that smaller companies don't have a large infrastructure to do drug testing. So if they can have a human based model it is an advantage to them. Larger companies typically have a group looking at this technology. They are not ignoring it, but they are very well set up to do testing on a large scale with animals and small companies aren't. Small companies are often looking for an alternative and a way to differentiate what they're doing. Often they don't have huge, deep financial pockets and resources so if they can get things done more quickly, which this technology promises to do, potentially that would be a big help for them. So a lot of the medium size and smaller companies have more incentive to try to develop this technology, whereas a large company has a very large infrastructure already in place that does the development in a way which they know works. They're still interested in making it better, but then they're going to do a lot of it in house, whereas I think the smaller companies are more interested in partnering with developers of the technology because they don't have the resources to develop it in house.
Thank you very much. I really appreciate this opportunity. And you are taking time out of your busy schedule to just try and come and give me all this information.
Sasserath, T, Rumsey, J, McAleer, CW, et al (2 June 2020) 'Differential Monocyte Actuation in a Three-Organ Functional Innate Immune System-on-a-Chip', Advanced Science.
Viravaidya, K, Sin, A, Shuler, ML (2004)'Development of a microscale cell culture analog to probe naphthalene toxicity', Comparative Study, Biotechnology Progress, 20/1, 316-23.Back