The Blog

Dr. Simon Melov on Single Cell Genomics

Dr. Simon Melov is a professor at the Buck Institute for Research on Aging. His lab is working on identifying the molecular hallmarks of aging, specifically in the context of cellular senescence. Recently, the lab published a paper in which it announced a discovery of a completely new senolytic compound, 25HC. This work was groundbreaking in several respects, including the implementation of single cell sequencing techniques to discover senolytic targets and the use of the novel senolysis marker developed by Dr. Judy Campisi’s group. We talked to Dr. Melov about this study, the phenomenon of cellular senescence, and the state of affairs in the longevity field in general.

Your group has discovered an entirely new senolytic using cutting-edge methods such as single cell sequencing. Could you tell us more about that?

I think it’s worth emphasizing a couple of key points about senescence without getting too much into the nitty-gritty. We’ve always had difficulty identifying causes and drivers of aging because aging is asynchronous in tissues.

It occurs in one cell at a time; all cells are not in lockstep, and this differentiates the process of aging from development, in the sense that development has a suite of coordinated changes which happen simultaneously. The technology for investigating changes in aging biology has always relied on averaging all cells within tissues.

You might take a sample of tissue, then you grind it up, and you’re, of course, homogenizing all the cells together so that you can get an average signal. There might be a difference with age compared to young tissues. In the end, you might infer something, but you would be incorrect because that average change reflects the sum of all those individual changes.

There may be more important things happening in subsets of cells, which are being obscured by this kind of homogenization. We’ve always had this problem, and the only way of addressing it in the past has been through microscopy, where you’re literally looking at cells in tissue sections. This has been incredibly slow and not particularly impactful, with the exception perhaps of descriptive changes where you can look at the incidence of fibrosis and things like that.

However, over the last decade or so, a slow-rolling revolution has been happening in the methodology for investigating changes at the single cell level. That technology is increasingly being applied to aging biology, and I’d make the argument that single cell genomics, which encompasses multiple different techniques, is perhaps the most impactful technology to hit aging in decades – specifically because aging happens one cell at a time. We need to understand that heterogeneity within a tissue before we can begin to develop rational therapeutics that will impact the functional decline, which is present in almost every tissue of our bodies as we age. Now that we have this tremendous technology, which is yielding enormous volumes of data, we can apply it, and I think it will pay big dividends and insights into aging biology.

We’re just at the “beginning of the beginning” of understanding aging biology. We’d been stymied over the last three decades or so from getting true understanding at the single cell level because of the technological limitations in looking at single cell changes. We’ve had some successes, but they haven’t been dramatic. There’s a lot of hyperbole in the aging field about dramatic breakthroughs. We’re always on the cusp of discovering this or discovering that, the phrase “In five years, we will have this,” but sadly, those five years are always in the future.

I’m very hopeful that single cell technology is going to make those kinds of statements redundant, because we actually will have breakthroughs, and our paper is the beginning of that kind of rational exploitation of a technology to gain fundamental insights into one aspect of aging, which is senescence.

Senescence is not the only driver of aging. I hesitate to say ‘minor’ because it depends on the tissue and on the organism, but it certainly contributes to functional decline. The magnitude of that contribution is still under active investigation. We don’t yet know how much of the functional decline associated with aging is caused by senescence itself. It’s a powerful biology, certainly at the single cell level. When you have a senescent cell, it’s secreting inflammatory factors, which can alter the tissue around it.

This is something we’re trying to enumerate in a large-scale mapping effort, which crosses multiple institutions across the US called the “SenNet”, which is based on single cell technology. We’re applying it heavily at the Buck, focusing on three tissues: muscle, the ovary, and the breast. Other institutions are focusing on different tissues, and that’s just to identify senescent cells in aging tissue, because we don’t yet have a good understanding of the burden of senescent cells in multiple different tissues.

To summarize, our paper is using the tool of single cell biology to discover potential targets which mediate senescence and then exploit those by uncovering new drugs or tool compounds that can attack those targets in senescent cells and hopefully confer beneficial effects.

We’re still in the early days, but we applied this technology to identify a target, which appeared to be important to senescent cells. It’s a protein called CRYAB, and we identified a molecule which seems to interact with that target, resulting in the death of senescent cells. This is the senolytic approach, which is somewhat similar to chemotherapy and cancer (killing the bad cells).

We have looked at this in many different cell types, human and mouse, and we found that the molecule 25-hydroxycholesterol seems to kill senescent cells in multiple different tissue types and species. We’re pretty excited about this as a general approach to drug development: using the technology of single cell profiling to uncover new targets and leverage that in the rational development of therapeutics for aging.

It looks like a black box approach – just look at the genes that are upregulated by senescence and find relevant inhibitors – which makes me think: do we even need to know how things work mechanistically?

We don’t, but certainly if you talk to drug development folks, they always like to have a mechanism. Editors in journals also like to have a mechanism. I’m much more agnostic about this in the sense that if you get a beneficial effect, you can just try and double down on that. I think mechanism is a very subjective terminology, because one person’s mechanism is another person’s hypothesis. For instance, it’s very difficult to unambiguously say you have the mechanism of action on COVID because there’s always something upstream, but the “mechanistic meme” has a lot of impact in different sectors of the community. They like to say that we understand the mechanism, but I think it more fair to say you understand just some of the mechanism. I think that’s true of most drugs.

Do you think that things like single cell techniques and AI are going to make this black box approach widely accepted and probably the primary tool at our disposal?

We’ve been doing single cell biology in the context of aging for more than a decade in my lab, and I’m convinced that the technology of single cell genomics, in particular, is perhaps the best application of that technology for a particular type of biology, which is aging.

You can argue that the return on investment for applying single cell genomics to aging paradigms is going to be enormous, much more impactful than in many other fields. Single cell techniques are also used in cancer, and that has some analogies to the situation with aging, and the methodology is used routinely in developmental work, and in many types of diseases. But, in terms of the bang for the buck, using single cell biology in the context of aging, we are going to make major discoveries which would just not be possible with any other tool.

You can combine it with AI because AI loves big data. Anyone who tells you that they’ve applied AI to n=5 or n=6 – they don’t know what they’re doing. You need zillions of data points for AI to work really well, and that’s exemplified by some of the best use cases of AI, which is probably in advertising, with billions of patterns of repetition in terms of clicks.

A more recent example and quite relevant one is the use of AI in Tesla, in autonomous driving, where they’ve logged millions of miles to try and work out how the car should turn in appropriate spots. That’s a comparatively simple problem, but it is enormously complicated when you try and break it down into individual components.

When you’re talking about genomics and the range of changes which can happen in aging tissue, that also is extremely complicated. For example, we don’t yet have a good sense of the variance across hundreds or even thousands of individuals with regards to gene expression changes, and it’s something which I think can be done in the near future, if not now.

It is expensive, that’s the downside of this. Not only expensive from the laboratory consumable side of things, but also computationally. When you start applying single cell workflows to hundreds of samples, it gets computationally heavy very quickly.

Do you think we might still hit the wall with senolytics somewhere in the future, maybe because we are going to discover serious adverse effects or that the doses that can be safely administered to humans are just not effective enough?

It’s extremely early to say. We’ve yet to see a successful application of aging biology in terms of a rationally designed therapeutic hitting the marketplace. That hasn’t happened yet. There was great optimism around resTORbio sometime ago. Their clinical trials failed for a variety of reasons, maybe not just because the paradigm was wrong. Maybe the design was wrong, that’s often a confounding factor in the success or failure of clinical trials. It’s not necessarily that the central hypothesis is incorrect. Or maybe they just didn’t hit on the right indication. The same thing happened, of course, with UNITY, their failure on their first initiative. They are starting to get some encouraging results in their Phase I data of the eye, which was released just a few weeks ago.

If there is a success story going to come out of the basic biology of aging, it may be senescent cells, the TOR pathway, maybe epigenetic changes being reversed – all of these things are going to take time.

Let’s go back to your research. Do you think your results, particularly on the restoration of muscle mass, can be replicated in humans?

I like to think that the mouse is a miniature human, but I have colleagues who tell me the mouse is a terrible model. It depends on who you talk to. There are many success stories using mice for developing therapeutics, where what you find in the mouse does hold in human beings, but there are also many failures. It depends on the context and I’m not sure we can make a strong prediction one way or the other.

We like to think that the core processes which we’re studying in the mouse will hold over to humans, otherwise we wouldn’t be doing it. We’d be just sitting in our armchairs, talking about what we should do instead of actually doing stuff. I think the mouse is a good model for aging. No model is perfect, all models have downsides. Something I heard recently, “all models are wrong, it’s just a question of how wrong.” So, we hope to use the least wrong model in moving the field forward. You really want to hope that what you’re doing has some relevance to human biology, but ultimately, we don’t know until there is a successful translation of what you find at the mouse model level into a human therapy.

I’m hopeful, because of the dramatic success of speeding up of clinical trials with COVID, which would have been unheard of five years ago, that it is now possible to speed up the clinical trial process if there’s sufficient incentive. You can make the argument that that is the case with aging; there’s an incentive to speed it up just because of the whole ‘silver tsunami’ argument.

I think that’s already happening, but there’s a disconnect between the research, the epidemiology, and the political apparatus, which is weird, because the political apparatus is right in the middle of that tsunami. You would think that the basic biology of aging would be funded at least at the level of Alzheimer’s disease, but it is not. That’s because there’s an effective lobby group for Alzheimer’s disease, but there is not an effective lobby group for the functional decline associated with aging. I don’t understand why those dots are not being connected.

Such groups are beginning to emerge, for instance, an Alliance for Longevity Initiatives, a lobbying group here in the US. Were you personally involved in any conversations with politicians, decision makers about that?

I’m not involved with that at all, so I don’t really have an opinion on it beyond seeing a lack of funding. There has been a steady small increase in funding for the biology of aging, but when you think about it, basically all major medical problems stem from aging, except childhood infectious diseases.

There’s remarkably little funding, considering that basic fact, and it doesn’t make sense. Researchers in this area have bemoaned this situation for many years because to those of us who studied aging, it’s obvious where the results are going to come from in terms of overall improvement of human health – beyond things like basic nutrition, vitamins, and so on, which is a given, and even that’s not particularly well funded.

I know that in the States there’s still an appallingly high level of just basic poverty, which could be solved overnight if Congress got its act together. So, sometimes things are obvious, and it just doesn’t happen for whatever reason politically. I don’t know why aging falls into this category. For some reason, Alzheimer’s disease and cancer resonate with the politicians.

I’m not sure why it’s so hard for some people to understand that aging is a huge problem. I don’t know if it’s solvable, but we at least should try to solve it with all we have.

As someone who’s been in the field for quite a while, we’ve been talking about this for decades. It’s not new. We’ve had politicians at the Buck, we’ve talked to them about this, and there’s nodding of heads and an apparent understanding of the problem and then nothing ever happens. I’m not sure why, one of the reasons might be this inability to accept the fact that aging is a plastic biology. It’s malleable and we can muck around with it in the lab. It’s trivial for us to manipulate lifespan in simple organisms now. Still hard in mice, but in simple organisms, like drosophila and C. elegans, it’s really easy.

When I got started in the field in the early nineties, it was really hard. There were only a couple of labs in the world who were able to do it. My old boss Tom Johnson was one of the first to do it, discovering genes which extended the nematodes’ lifespan. That was unheard of at the time. Today, it’s something that a graduate student can do in a summer, it’s really trivial.

That’s different from what happened with the late politician Arlen Specter. He got cancer, I can’t recall which, and then he made it his mission to increase the budget of the NIH (National Institutes of Health) – as a function of him being ill – and he was dramatically successful in doing that.

We haven’t had the same sort of thing happening with Alzheimer’s; it’s been more of a sustained lobbying effort by multiple groups. As you said, there are groups trying to do this with aging, but it’s hard maybe in part because aging is so poorly defined. What does it mean to be old apart from the passage of years? There’s a lot of debate over it.

On the other hand, it’s something that every person is familiar with or will become familiar with.

But it’s also like accepting the weather. The weather just happens, and you don’t try to do anything about it. You might want to predict where it’s going (which in our case would be equivalent to, say, putting dollars into nursing homes), but you’re not going to try and fundamentally alter the weather. That is, unless it’s about climate change.

That’s an encouraging example. We had seen the same denial with climate change for decades, but then it changed. 

That’s true, but there are still large chunks of the population who don’t accept it. They just don’t believe science basically, that’s the bottom line.

Going back to your research again, what exactly is your target protein CRYAB? Like many other targets of senolytics, it seems to be protecting cells from apoptosis, right?

Yes, that’s correct. It’s a small heat shock protein. Interestingly enough, it has been reported previously to aggregate with age in human skeletal muscle. There is a report from more than a decade ago where they looked at the level of aggregated CRYAB represented by its insoluble form, which accumulates with age in human skeletal muscle quite dramatically.

We’re following up on that in multiple ways at the moment. We’re trying to understand if this target plays into the whole proteostasis argument around aging, which says that there’s a failure to maintain conformation with age and this somehow confers a benefit to senescent cells.

We’re investigating that at this point. We’re having a particular emphasis on skeletal muscle. It’s difficult to talk about apoptosis in skeletal muscle fibers, because they are long syncytial cells. This is a weird tissue to work on in many ways because there’s no single nucleus. They’ve got thousands of nuclei per fiber. They also have support cells and things like that, and maybe that’s where the action is. We are drilling into that at this point.

We know for certain that when you make cells senescent, multiple different cell types, everything from cardiac support cells through to liver cells, they will respond to the molecule (25HC), which we discovered, which somewhat bizarrely is also an anti-viral molecule discovered in the course of the COVID-19 pandemic. It kills senescent cells preferentially. We don’t really understand the interaction, but we’re working on the mechanism of that at the moment.

This brings me to the senolysis marker that was mentioned in your paper. How does it work; what exactly does it measure?

This marker was discovered by Judy Campisi and Chris Wiley early last year. They reported it in the context of a chemotherapeutic model, and we reported it now in the context of aging and treating aged animals with senolytics. It’s produced only in senescent cells, and when the senescent cell is lysed, this molecule is released into the plasma and the urine where it can be detected.

You get an increase in this marker when you’re killing senescent cells – increased signal is good, decreased signal is bad. Interestingly, we found an endogenous signal present in animals not treated with the senolytic, which might represent continuous turnover of senescent cells in aged animals. We were able to show that in old animals compared to young animals, there is a marked increase in the presence of this molecule. Then it goes up further when you treat them with a senolytic.

This marker is very useful because it’s really hard to enumerate senescence or senolysis in vivo. A lot of people are interested in that topic, and this is one way of doing it, which seems to work pretty well.

Since lifestyle changes are probably among the most potent anti-aging interventions that we have now…

The only anti-aging interventions we have now.

Do we have any indication that those lifestyle changes can help specifically with the accumulation of senescent cells or the rate of senolysis?

That’s a great question and one we’re actively interested in investigating. It’s likely going to be a major component in another initiative, which we have at the Buck in partnership with the Astera program.

This is this gigantic project to do single cell mapping on thousands of mice treated with different interventions that are reported to extend lifespan. I suspect some of these will not replicate when we test them. We will be looking at exercise as one of the modalities of improving function across multiple different domains of aging in mice in conjunction with senolytics like 25HC, and then we’re going to be able to answer that question.

It’s not something we have data on yet, but it’s certainly something there’s a lot of interest in, given the profound effects of exercise on slowing the functional decline associated with aging. You can’t stop aging, but exercise and a good diet appear to be the only way we have of slowing things down. You can actually make the philosophical argument that all we’re really doing is optimizing the system instead of fundamentally affecting aging.

It might be more appropriate to say that when you’re not exercising and you’re overeating, you’re just screwing the system up and you’re speeding things up. You’re making the system worse. So, when you exercise and you eat well, all you’re doing is putting the system back to what where it should be.

The question is whether many of the prospective anti-aging interventions work the same way, by simply optimizing the system?

We don’t know the answer to that yet. That’s something we are hoping to answer in the course of the Astera program over the next 5 to 10 years. It’s the largest project that Buck has ever launched. The whole thing is funded by Jed McCaleb, a crypto billionaire. I think it’s going to be tremendously useful to the community at large in the long run.

Do you have any plans or thoughts about human trials of 25HC? 

I have thoughts and plans, but I don’t know if they are realistic at this point. 25HC is being touted as a potential therapeutic for COVID. I don’t know whether that conversation is meaningful, but there’s been almost half a dozen papers published on 25HC and COVID in the last year alone. I think it might move forward in that route, not necessarily as a result of our work.

Still, there’s an intriguing possibility that the mechanism of action of endogenously produced 25HC is to kill senescent cells. In fact, it may be that the cytokine storm, which is very common in severely affected COVID patients, is linked to a high level of senescent cells. I haven’t seen anything on that, it’s pure speculation on my part, but if that’s true, then administration of 25HC might attenuate that storm by killing senescent cells.

Human trials are extremely hard to pull off. Do you think that the system might be broken, that things should be easier, less restrictive?

Yes, it’s just tremendously hard. It should be hard on one hand, I don’t want to say that it should be easy to do clinical trials: the possibility of bad things happening is high when you stick unknown substances into human beings. But there are things we can do.

For instance, I have a collaborator at Boulder (the University of Colorado), an investigator called Doug Seals. He’s really good at small-scale clinical trials, and he does them on an NIH budget, so it’s not true that you need billions of dollars to do a decent clinical trial. His trials are modest but well-designed; he’s assessing functional changes in the cardiovascular system, and he’s been very successful at running these little trials which involve just a couple of dozen people instead of hundreds.

I think it is possible to be creative within the constraints of the system without necessarily having to commit huge sums of money to large scale trials.

What are your thoughts on the situation in the aging field today? Are you more optimistic or more pessimistic? 

I’m optimistic, and it comes back to the point I made initially about technology. I do think the technology is here now to get fundamental insights at the level of a single cell, which we’ve lacked over the last three or four decades.

We can also make the readout at the level of the single cell for thousands, to hundreds of thousands, to maybe millions of cells. We’re going to have an instrument at the Buck fairly soon in conjunction with Astera which allows us to do spatial profiling at the level of the single cell.

It’s true single cell spatial profiling, and it yields enormous volumes of data because you get the positions of the transcripts within the cells, and you’re able to localize them to specific organelles, to quantitate them. It’s going to give tremendous insights into what’s happening at any given instant especially in conjunction with putative therapeutics.

Because we’re so good at measuring function in animal models (we can measure nearly any functional readout you care to name in human beings that we have a good proxy for in mice), we’ve got to be able to link function with therapeutics at the single cell level. That’s going to break down barriers like we’ve never seen before.