Stem Cell Research and Precision Medicine Today and Tomorrow: A Conversation with Lee L. Rubin, Ph.D.
Aiden Bowers
Dr. Lee L. Rubin, Ph.D. is a Harvard professor in the Department of Stem Cell and Regenerative Biology.
His work is primarily translational neuroscience research: formerly, his lab was focused on using induced pluripotent stem (iPS) cells to model diseases and discover therapeutics, including Parkinson’s disease, Amyotrophic Lateral Sclerosis (ALS), and Spinal Muscular Atrophy (SMA). Today, his lab is focused on the study of age and age-related factors, like GDF11, which might play a role in helping to improve symptoms of neurodegenerative disease in patients (Ozek, 2018).
iPS cells have revolutionized in vitro research, and are critical models in stem cell research. They can be genetically engineered and differentiated into most somatic cell types, and are most commonly used to model human development and diseases in vitro, screen drug candidates, and create cell therapies (Cerneckis, 2024).
This interview has been edited for length and clarity.
AB: What questions does your current research aim to answer?
LR: I’m interested in translational research — in other words, moving research towards the clinic as far as we can. I work mostly on diseases of the nervous system or of the neuromuscular system, and I think both are characterized by their complexity. Just for the sake of this discussion, we talk about two kinds of complexity: how people with the same disease can present as having very different disease, and the other is genetic complexity. Even within a single disease, Alzheimer’s disease, whatever, there are many different ways of getting Alzheimer's disease: you can have a mutation that gives you that disease early in your life, relatively speaking — 40s — if you have Down’s syndrome you will be predisposed to Alzheimer’s disease in your 50s, and if you have the most common form, aging driven, then it’s more like 60s or 70s. So you end up in exactly the same place, or a similar place, but the way you get there is quite different. It’s not obvious to me, anyhow, whether all that matters is where you end up vs. how you get there. Personalized medicine suggests, like is typically done in cancer, that the path is as important as where you end up.
I am interested in knowing whether or not that applies to the nervous system as well. Historically here a lot of our research was based on using iPS cells, that gives you a method by which you can objectively answer the questions of “if you have a drug, will it work on everybody with the disease?”
Theoretically, you can make iPS cells from lots of people and test the drug on their neurons and ask whether it works on these people only or these people, so that’s one thing that we’ve done. More recently, however, we have been doing something which will sound exactly the opposite of what I’ve just told you, and is exactly the opposite of what I’ve just told you, which has to do with aging.
Aging is a common driver of many diseases, and the thing that’s unique about aging is not that it’s associated with so many diseases, but, at least for a mouse brain, a lot of what goes on with aging is nearly completely reversible. If you think about it from a therapeutic point of view: whenever you hear about let’s say a drug or an antibody for Alzheimer’s disease, what people are talking about is slowing the rate of progression. Nobody is talking about giving you back a function that you already lost. But in aging, in a mouse, you can actually go back the other way.
Now, it’s two questions: in a mouse, where there are so many ways of doing this, what is the best way from a translational point of view, that is to say, the thing that you imagine would really be applicable to people, and the biggest question is would any of this really work in people? So my lab has shifted its balance from mostly iPS cells to mostly this aging stuff.
AB: What is your personal path to this research?
LR: I have sort of an unusual background here, which is to say I have worked in academia and in industry, and I came here from a biotech company. So when I say I am interested in translational stuff, I really am.
When I started, I was really a pure, basic scientist. The world has changed a lot since then, and I’ve changed a lot since then. What I learned by being in industry, which I didn’t expect really, is two things. One, I really enjoyed thinking that my research could actually have some impact on somebody’s life in a positive way.
The other is, in industry, what’s kind of nice — there are a lot of things that are not as nice as in academia — is the feeling that you’re really working together as a team. Everybody accepts if you’re working in that company you're committing to working together with other people to do something that’s really important from a health perspective. I like that feeling and I like the dynamic it creates with the people that you’re working with. I’ve tried to bring both things back here.
The reason I came here was the company where I was working did not want me to take this very new iPS kind of approach — it was too experimental in those days. You never hear that in academia, right? I literally came here to keep working on translational stuff. I say that to people and it sounds sort of crazy, because normally industry would be the place you would do that, but not if you want to do something that’s really not been validated as a standard, turn-the-crank system — if you’re really establishing a new approach, it’s really better in academia.
AB: As far as your research goes, why is the iPS cell model so pivotal? Why not a different model?
LR: I would say there are two reasons.
When you start talking about the bulk of human disease, which are genome diseases, there are good genes, bad genes, genes that have small effects, big effects — there’s just no way you can do that in anything other than a human cell. That’s one big reason: to capture human genetics I feel you have to use a human cell.
Then the question is, if you’re a neurobiologist, like I am, that cell is going to be a neuron. When I first started out — how did you get neurons? — you dissected chick embryos or mouse embryos and if you wanted to get motor neurons, which is what I worked on for years, you would get 3,000 per embryo. If you want to do a drug screen, like I’ve done in my lab, you need three billion. There’s no way of getting that number without using a pluripotent cell somewhere. When you put those two things together, the pluripotent cell captures the genetics, you can make unlimited numbers of neurons from them.
Then you have human disease genetics in a boatload full of cells that you can use for basically any purpose you want. For someone with my interests, it's not that the other systems can't tell you anything, but to me it's clearly most representative to people. I would rather make the system adaptable to what I want to do than work with a system that's not really modeling the disease process that I’m interested in.
AB: What do you see as the challenges of this model?
LR: The big challenge that everybody acknowledges is, if you work on late onset disease, which I told you I'm interested in, the neurons you make from iPS cells are fetal. They’re the human genome in a cell, the cell can be a motor neuron or whatever cell you’re interested in, but it’s the fetal version of that cell, so you’ve now made a compromise. You’ve gotten some of the stuff you want, but you’ve sacrificed other stuff. Again, how do you adapt to making those compromises? It’s not the ideal situation but in my mind it’s at least in the direction of the ideal situation.
Let’s say for a fly version of what I said to do in human cells, I feel not that it’s useless, but it involves, to me, more compromises for what I want to do.
AB: What are your visions for the future of the field of precision medicine in a neurological capacity?
LR: Let’s just say that there’s a possibility that you’re not going to identify one drug to treat these very complex diseases.
In the end, maybe you have ten drugs that you think might cover the population of people that have schizophrenia, just to give a very complex disease. iPS cells can be used in different ways — and this is a little bit more futuristic — right now, you can try and use it to discover drugs, if you make iPS cells from a bunch of people with schizophrenia. That’s one thing. A second thing is that, you could, in theory, say which schizophrenic patients will that drug work on, this genetic composition of people or this one? Third, and most importantly, you can do that before you test a drug in the clinic.
Can you achieve both at the same time, lowering the cost of drugs but nonetheless having the funding to do research?
Roughly 10% of drugs that go into a phase one study are approved, and for neuro it’s 3%, so 97% failure when it goes into the clinic — that influences the cost of drugs because the 3% cover the cost of the 97% that fail, so of course it’s going to jack up the price. For the company to stay in business, the successes have to pay for the failures, no way around that that I know of. If you could increase the efficiency, it seems at least possible that that would reduce drug costs. So I think if all this iPS stuff were to align, it would be possible to develop really innovative drugs and charge less for them.
It shouldn’t be one or the other, because both goals are totally important.
AB: What horizons do you see for the future applications of stem cell research, both for your own research and the method as a whole?
LR: I’d love to get another one or two drugs approved during my time here, and I’m not sure that’s impossible, so that’s one thing I’m excited about. The other thing that I’m working on with several colleagues is bringing more automation, industry stuff, to the way that we are making and using iPS cells. That’s a big push now between me and Paola Arlotta, who is on the organoid end, and Laurence Daheron, who runs the Harvard iPS core. I’m hoping that we can really increase our capacity to study many patients' cells at one time.
It’s in the planning stage. There’s been a lot of improvement in robotics. In fact, just today we had another zoom call with an automation group that’s been advising us on how to do this. There are things that can be automated, like, for instance when you feed cells you have to turn the dish to suck out the medium — there's automation for that. It’s kind of remarkable. The other thing that's very important is AI. We’d like to be able to image cells well-by-well. iPS cells are so variable, even the same line from time to time, different patients, how do you adapt to that? In theory, if you image each well and establish criteria even for silly things like when do you feed the cells, when do you differentiate them, I think this can be automated as long as it's equipped with some sort of machine learning algorithm.
I think that that’s possible now, and I’m kind of excited to see what that would look like.
About the Author
Aiden J. Bowers (’26) is currently a junior at Harvard College studying Neurobiology.
References
His work is primarily translational neuroscience research: formerly, his lab was focused on using induced pluripotent stem (iPS) cells to model diseases and discover therapeutics, including Parkinson’s disease, Amyotrophic Lateral Sclerosis (ALS), and Spinal Muscular Atrophy (SMA). Today, his lab is focused on the study of age and age-related factors, like GDF11, which might play a role in helping to improve symptoms of neurodegenerative disease in patients (Ozek, 2018).
iPS cells have revolutionized in vitro research, and are critical models in stem cell research. They can be genetically engineered and differentiated into most somatic cell types, and are most commonly used to model human development and diseases in vitro, screen drug candidates, and create cell therapies (Cerneckis, 2024).
This interview has been edited for length and clarity.
AB: What questions does your current research aim to answer?
LR: I’m interested in translational research — in other words, moving research towards the clinic as far as we can. I work mostly on diseases of the nervous system or of the neuromuscular system, and I think both are characterized by their complexity. Just for the sake of this discussion, we talk about two kinds of complexity: how people with the same disease can present as having very different disease, and the other is genetic complexity. Even within a single disease, Alzheimer’s disease, whatever, there are many different ways of getting Alzheimer's disease: you can have a mutation that gives you that disease early in your life, relatively speaking — 40s — if you have Down’s syndrome you will be predisposed to Alzheimer’s disease in your 50s, and if you have the most common form, aging driven, then it’s more like 60s or 70s. So you end up in exactly the same place, or a similar place, but the way you get there is quite different. It’s not obvious to me, anyhow, whether all that matters is where you end up vs. how you get there. Personalized medicine suggests, like is typically done in cancer, that the path is as important as where you end up.
I am interested in knowing whether or not that applies to the nervous system as well. Historically here a lot of our research was based on using iPS cells, that gives you a method by which you can objectively answer the questions of “if you have a drug, will it work on everybody with the disease?”
Theoretically, you can make iPS cells from lots of people and test the drug on their neurons and ask whether it works on these people only or these people, so that’s one thing that we’ve done. More recently, however, we have been doing something which will sound exactly the opposite of what I’ve just told you, and is exactly the opposite of what I’ve just told you, which has to do with aging.
Aging is a common driver of many diseases, and the thing that’s unique about aging is not that it’s associated with so many diseases, but, at least for a mouse brain, a lot of what goes on with aging is nearly completely reversible. If you think about it from a therapeutic point of view: whenever you hear about let’s say a drug or an antibody for Alzheimer’s disease, what people are talking about is slowing the rate of progression. Nobody is talking about giving you back a function that you already lost. But in aging, in a mouse, you can actually go back the other way.
Now, it’s two questions: in a mouse, where there are so many ways of doing this, what is the best way from a translational point of view, that is to say, the thing that you imagine would really be applicable to people, and the biggest question is would any of this really work in people? So my lab has shifted its balance from mostly iPS cells to mostly this aging stuff.
AB: What is your personal path to this research?
LR: I have sort of an unusual background here, which is to say I have worked in academia and in industry, and I came here from a biotech company. So when I say I am interested in translational stuff, I really am.
When I started, I was really a pure, basic scientist. The world has changed a lot since then, and I’ve changed a lot since then. What I learned by being in industry, which I didn’t expect really, is two things. One, I really enjoyed thinking that my research could actually have some impact on somebody’s life in a positive way.
The other is, in industry, what’s kind of nice — there are a lot of things that are not as nice as in academia — is the feeling that you’re really working together as a team. Everybody accepts if you’re working in that company you're committing to working together with other people to do something that’s really important from a health perspective. I like that feeling and I like the dynamic it creates with the people that you’re working with. I’ve tried to bring both things back here.
The reason I came here was the company where I was working did not want me to take this very new iPS kind of approach — it was too experimental in those days. You never hear that in academia, right? I literally came here to keep working on translational stuff. I say that to people and it sounds sort of crazy, because normally industry would be the place you would do that, but not if you want to do something that’s really not been validated as a standard, turn-the-crank system — if you’re really establishing a new approach, it’s really better in academia.
AB: As far as your research goes, why is the iPS cell model so pivotal? Why not a different model?
LR: I would say there are two reasons.
When you start talking about the bulk of human disease, which are genome diseases, there are good genes, bad genes, genes that have small effects, big effects — there’s just no way you can do that in anything other than a human cell. That’s one big reason: to capture human genetics I feel you have to use a human cell.
Then the question is, if you’re a neurobiologist, like I am, that cell is going to be a neuron. When I first started out — how did you get neurons? — you dissected chick embryos or mouse embryos and if you wanted to get motor neurons, which is what I worked on for years, you would get 3,000 per embryo. If you want to do a drug screen, like I’ve done in my lab, you need three billion. There’s no way of getting that number without using a pluripotent cell somewhere. When you put those two things together, the pluripotent cell captures the genetics, you can make unlimited numbers of neurons from them.
Then you have human disease genetics in a boatload full of cells that you can use for basically any purpose you want. For someone with my interests, it's not that the other systems can't tell you anything, but to me it's clearly most representative to people. I would rather make the system adaptable to what I want to do than work with a system that's not really modeling the disease process that I’m interested in.
AB: What do you see as the challenges of this model?
LR: The big challenge that everybody acknowledges is, if you work on late onset disease, which I told you I'm interested in, the neurons you make from iPS cells are fetal. They’re the human genome in a cell, the cell can be a motor neuron or whatever cell you’re interested in, but it’s the fetal version of that cell, so you’ve now made a compromise. You’ve gotten some of the stuff you want, but you’ve sacrificed other stuff. Again, how do you adapt to making those compromises? It’s not the ideal situation but in my mind it’s at least in the direction of the ideal situation.
Let’s say for a fly version of what I said to do in human cells, I feel not that it’s useless, but it involves, to me, more compromises for what I want to do.
AB: What are your visions for the future of the field of precision medicine in a neurological capacity?
LR: Let’s just say that there’s a possibility that you’re not going to identify one drug to treat these very complex diseases.
In the end, maybe you have ten drugs that you think might cover the population of people that have schizophrenia, just to give a very complex disease. iPS cells can be used in different ways — and this is a little bit more futuristic — right now, you can try and use it to discover drugs, if you make iPS cells from a bunch of people with schizophrenia. That’s one thing. A second thing is that, you could, in theory, say which schizophrenic patients will that drug work on, this genetic composition of people or this one? Third, and most importantly, you can do that before you test a drug in the clinic.
Can you achieve both at the same time, lowering the cost of drugs but nonetheless having the funding to do research?
Roughly 10% of drugs that go into a phase one study are approved, and for neuro it’s 3%, so 97% failure when it goes into the clinic — that influences the cost of drugs because the 3% cover the cost of the 97% that fail, so of course it’s going to jack up the price. For the company to stay in business, the successes have to pay for the failures, no way around that that I know of. If you could increase the efficiency, it seems at least possible that that would reduce drug costs. So I think if all this iPS stuff were to align, it would be possible to develop really innovative drugs and charge less for them.
It shouldn’t be one or the other, because both goals are totally important.
AB: What horizons do you see for the future applications of stem cell research, both for your own research and the method as a whole?
LR: I’d love to get another one or two drugs approved during my time here, and I’m not sure that’s impossible, so that’s one thing I’m excited about. The other thing that I’m working on with several colleagues is bringing more automation, industry stuff, to the way that we are making and using iPS cells. That’s a big push now between me and Paola Arlotta, who is on the organoid end, and Laurence Daheron, who runs the Harvard iPS core. I’m hoping that we can really increase our capacity to study many patients' cells at one time.
It’s in the planning stage. There’s been a lot of improvement in robotics. In fact, just today we had another zoom call with an automation group that’s been advising us on how to do this. There are things that can be automated, like, for instance when you feed cells you have to turn the dish to suck out the medium — there's automation for that. It’s kind of remarkable. The other thing that's very important is AI. We’d like to be able to image cells well-by-well. iPS cells are so variable, even the same line from time to time, different patients, how do you adapt to that? In theory, if you image each well and establish criteria even for silly things like when do you feed the cells, when do you differentiate them, I think this can be automated as long as it's equipped with some sort of machine learning algorithm.
I think that that’s possible now, and I’m kind of excited to see what that would look like.
About the Author
Aiden J. Bowers (’26) is currently a junior at Harvard College studying Neurobiology.
References
- Cerneckis, J., Cai, H., & Shi, Y. (2024). Induced pluripotent stem cells (iPSCs): Molecular mechanisms of induction and applications. Nature News. https://www.nature.com/articles/s41392-024-01809-0
- Ozek, C., Krolewski, R. C., Buchanan, S. M., & Rubin, L. L. (2018). Growth Differentiation Factor 11 treatment leads to neuronal and vascular improvements in the hippocampus of aged mice. Scientific reports, 8(1), 17293. https://doi.org/10.1038/s41598-018-35716-6