#062 Dr. Steve Horvath on epigenetic aging to predict healthspan: the DNA PhenoAge and GrimAge clocks

Methylation Support Discussion

This measurement may be more fundamental, more interesting than many other markers, because, as we'll discuss in a moment, it's predictive value. In this episode, we're going to talk about two generations of epigenetic clocks.

"This measurement may be more fundamental, more interesting than many other markers, because, as we'll discuss in a moment, it's predictive value. In this episode, we're going to talk about two generations of epigenetic clocks."

Epigenetic Aging: Clock

This is the last episode of 2020, and while 2020 has been a tough year, this episode is fantastic. Today's guest is Dr.

"Insurance companies do this every day when they look at actuarial tables in combination with other factors to give you a price on your life insurance plan."

Epigenetic Aging: Clock

As a biomarker, telomeres have the shortcoming that they are highly dependent on cell types, since telomer length is governed partly by artifacts of cellular division. They also behave differently between tissues and species.

"As a biomarker, telomeres have the shortcoming that they are highly dependent on cell types, since telomer length is governed partly by artifacts of cellular division. They also behave differently between tissues and species."

Methylation Support: How To

The reason why this is not yet a viable strategy is because people who get a transplant often get so-called graft versus host disease. It's a dangerous procedure, but in theory, it could work.

"A grim age is a pretty good predictor of time to coronary heart disease. Surprisingly, these clocks even predict time to cancer. These are no doubt exciting times for the aging field. I want to tell you about two really profound experiments that tell us very interesting things about some of the drivers of epigenetic age. One is an accidental human experiment happening inadvertently in the clinic every single day. The other, an animal experiment notable especially for its ingenuity. Scientists have long known that paracrine signaling, the molecular and cellular niche created from cell-to-cell interactions, is extremely crucial in maintaining tissue health and integrity, and even, as it turns out, youth maintenance. In other words, keeping our tissues young. But what happens if you can bring an older animal into a younger animal's network of cellular signals on a daily, minute-to-minute basis? That's the question. In animals, the experimental surgical union of their vascular systems, known as parabiosis, leads to the transference of blood-borne factors from the younger animal into the older animal. Some evidence has shown that this can lead to a youthful phenotype and rejuvenating effect in certain tissues. While these experiments have been the excitement of the aging field for a while, despite their somewhat clinical impracticality, the results of Dr. Horvath's lab, he says, have been somewhat mixed. However, humans have been doing a somewhat modified version of this experiment too, but inadvertently. Surprisingly, at least insofar as epigenetic clocks are concerned, the results here are even more exciting than animal research. Let's say you take a 50-year-old and you give this person a bone marrow transplant from a 20-year-old. And so after the transplant, the blood in the recipient reconstitutes itself. The person has now new blood. And the question is, what's the age of the blood? Is that blood, does the blood have the age of the 20-year-old donor or the age of the 50-year-old recipient? There are now several scientific papers that really give an unequivocal answer, which is the reconstituted blood in the recipient has the age of the donor. And that effect persists for decades. Bone marrow transplants are different from parabiosis. For example, in leukemia, it is sometimes necessary to irradiate, destroy, and ultimately replace the entire hematopoietic system via a bone marrow transplant. This ultimately results in the recipient having cells that no longer track with their own epigenetic clock. Rather, they track with the donors. That means if you're unlucky enough to have leukemia, but lucky enough to have a younger donor, you might ultimately be able to enjoy younger T-cells, macrophages, white blood cells, and natural killer cells. A tissue-specific epigenetic age reversal effect that lasts, as you heard a moment ago, for decades. Unfortunately, cancer treatment as a whole does accelerate epigenetic age. So that's on some level a very exciting finding because it kind of hints to an idea that you could possibly rejuvenate people through transplantation. The reason why this is not yet a viable strategy is because people who get a transplant often get so-called graft versus host disease. So disease. So there are all sorts of complications. It's a dangerous procedure, but in theory, it could work. There are a few experimental ideas for how to interfere with the epigenetic clock or otherwise rejuvenate the methylome. One is using drugs to play with inhibiting the enzymes that directly interact with DNA methylation, known as DNA methyltransferases. These drugs already exist, and in mice, with disorders of the epigenetic machinery, these drugs are able to reverse some effects of those conditions. The other avenue, which is a bit more promising in my opinion, is manipulating the expression of genes that encode for transcription factors already known to reverse epigenetic age. Right now, there's an idea in the aging field to rejuvenate people by leveraging this fundamental insight that you call, it's called reprogramming. You can take an old cell, you administer certain factors. Called Yamanaka factors, these factors can revert a differentiated somatic cell back in epigenetic age to an embryonic or near embryonic state. Even more interesting, scientists know that by genetically engineering a switch, one that is turned on by a chemical, they have been able to add these Yamanaka factors to animals to reverse even certain aspects of tissue aging by pulsing these factors. But don't switch them on for too long, says Dr. Horvath, because too much of the Yamanaka factors may revert cells too far back, ultimately inducing malignancy. One of the biggest challenges in repeating this experiment in humans, however, is that genetic engineering is not without risk."

Epigenetic Aging: Clock

A grim age is a pretty good predictor of time to coronary heart disease. Surprisingly, these clocks even predict time to cancer.

"Maybe it is as simple as a vitamin D supplement, but maybe you need something much more radical, like, for example, a modification of the Yamanaka cocktail, you know. Right. Or it could be plasma transfusions or so. You know, we as a field need to experiment, you know, with what kind of interventions work. It could be hormones, by the way. Right. You know, it could be a hormone intervention. Age-associated methylation are commonly found near genes involved in development. If you had asked an aging researcher five years ago whether developmental processes matter in aging, they would have said no. Many people think of aging as noise or wear and tear, you know. But these epigenetic clocks have really linked development to tissue dysfunction in a direct manner. An epigenetic clock is a continuous readout that links prenatal tissues directly to very old samples. Dr. Horvath issues a caution about using the epigenetic clock as a clinical biomarker. When I predict, for example, that you will develop heart disease in 15 years, you know, there would be a big error bar associated with it, plus minus six years or so. Okay, so that is our long-winded intro. For a shorter intro, head to my YouTube channel where you will find a three-minute primer on epigenetic clocks. You couldn't have a better pair to this interview. Find that by searching FoundMyFitness YouTube or by going to foundmyfitness.com forward slash YouTube. It's just three minutes long. Let's get on to the actual podcast with Dr. Steve Horvath, where we talk all things epigenetic aging clocks. Hello, everyone. I'm sitting here with Dr. Steve Horvath, who is a professor of genetics and biostatistics at UCLA. Probably one of Steve's most well-known contributions to biology is the development of what's known as the Horvath aging clock. And since then, he's gone on to develop even more accurate aging clocks, which I'm so excited to talk about. I've talked about your work multiple times on multiple podcasts to multiple scientists. So thank you for spending time to talk with me today. Thank you for your interest. Thank you for visiting me. Well, Steve, maybe we could kind of start at the beginning with the, what is this Horvath aging clock that you had developed? Like, can you explain what it is? Well, the Horvath aging clock is what I sometimes call the so-called pan-tissue epigenetic clock. And so it is the most accurate molecular measure of age. It applies to all cells in the bodies, certainly all cells that have DNA, all tissues, all organs. It measures age in prenatal samples, in children, all the way to super centenarians, people who are over 110 years old. And it measures age. So if somebody provides, for example, DNA from blood or DNA from neurons or DNA extracted from saliva or urine, I can very accurately estimate their age. Their chronological age, like how old they are in years. Exactly, exactly. And that's already a deep question. So the clock does measure chronological age. However, it's of course not a perfect measure of chronologic age. There's always an error. For example, if I analyze the blood from a 50-year-old, the epigenetic clock may say, well, this person is actually 55 or 45. And so there's a small error. And this error is actually biologically meaningful. It's not just noise, but rather it is in part related to what people call biological age. That's super interesting there, that basically the error was actually related to biological aging because that was the next thing I was going to say was, you know, people age at different rates. Like even, you know, obviously chronologically they could be the same age. Yes. But if you look at a variety of biomarkers, in fact, there was a paper published a few years ago in PNAS that looked at, like, 18 different biomarkers. Yes. They looked at glycated hemoglobin, so HbA1c, Vo2max, triglycerides, telomere length, immunosenescence, a lot of the, you know, typical biomarkers that are clinically used to, health status. And it was basically found that people aged at very different rates based on those biomarkers. So some people biologically appeared much younger than their chronological age, and some people appeared much older than their chronological age. Exactly. So how now you've developed a different clock that can actually predict? That's right. Well, let me comment first on the term biologic age. It's a very intuitive term. Most people have a vague understanding of it. It somewhat relates to morbidity risk or mortality risk and also aging. But strictly speaking, it's not well defined because different researchers have different ways of measuring biological age. Some people use clinical markers that you mentioned, various markers of glucose levels or lipid levels and so on. Now, my take to measuring biologic aging is based on a chemical modification of the DNA molecule. It's DNA methylation, you know. And I mention it because depending on how you measure biologic age, you might get different answers. So a person might look bad in terms of glucose levels and you would say, well, they age faster than they should. However, it could turn out that according to methylation, they are actually in pretty good shape. So yeah. Have you seen that before where you can see people have, for example, like other clinical biomarkers that are unhealthy, like higher fasting blood glucose or maybe elevated triglycerides, elevated C-reactive protein, a marker of inflammation? Do you find that those typically correlate well with the epigenetic age? I wouldn't say it correlates well. It correlates, you know, so people who have higher levels of inflammation and what you mentioned, the epigenetic clock goes a little bit faster. But the word is there's a weak relationship, you know, because it is quite possible that somebody looks, turns out to be in good shape according to epigenetic aging rates. And the number one example I want to mention in this context are actually people of Hispanic ancestry. Unfortunately, Hispanics often have higher risk for diabetes, the higher metabolic syndrome. And however, according to the epigenetic clock, they actually age more slowly, you know. And so there's this really, this disconnect. And this is actually an interesting disconnect because there's something known as the Hispanic mortality paradox, you know. Hispanics, as I mentioned, have often a disadvantageous risk profile according to clinical biomarkers. But it turns out, on average, they live much longer than expected. They actually live longer lives than people of European ancestry."

Methylation Support Discussion

Then if you followed them for 60 years, you know, you would find this consistency that people who are slow agers at age 20, they also slow agers at age 60 or 80.

"Yes. So that's one line of evidence, but there's another. So people have these longitudinal epidemiological studies, and they may have collected a blood sample from a person when they were, let's say, 40 years old."

Epigenetic Aging: Clock

Then if you followed them for 60 years, you know, you would find this consistency that people who are slow agers at age 20, they also slow agers at age 60 or 80.

"Yes. So that's one line of evidence, but there's another. So people have these longitudinal epidemiological studies, and they may have collected a blood sample from a person when they were, let's say, 40 years old. And then 15 years later, they get a second blood sample. you know. And so you can then ask the question whether a person who was aging quickly at the first blood draw, did they still age quickly at the second blood draw, you know. And the answer is yes. And conversely, you observe the same for people who age more slowly. And that, in my opinion, this could already be observed when you study, let's say, a person at a young age, age 20, draw their blood. Then if you followed them for 60 years, you know, you would find this consistency that people who are slow agers at age 20, they also slow agers at age 60 or 80. So that does seem to imply the genetic... I mean, unless someone dramatically changed their lifestyle. How stable are these changes in methylation, these methylation patterns that are so-called, you know, the aging clock? How stable are they over a person's lifetime? I mean, they're... I mean, they are remarkably stable. So when we compare to any other genomic measurement, I mean, they would be far more stable than anything I'm aware of. They're far more stable than gene expression, proteomics, metabolomics measurements. All the omics are less stable. And that's really the biological reason why these epigenetic clocks are the most accurate measures of aging. It's just that methylation is so stable. And it's stable not just in vivo, but also when people collect DNA. So we've collected DNA, and then we didn't store the blood tubes properly, so they melted. And we extracted DNA, and the measurements were perfect. you know. So, even on a technical level, they're very stable. That's interesting because I've done experiments intentionally, though, almost the same, where we were collecting blood samples from participants in a trial and I was measuring DNA damage as biomarked by gamma H2AX. And I wanted to know how long I could have blood at room temperature before I started to get artifactual DNA damage happening. And so I did time course and found that after two hours of blood being at room temperature, there's just tons of DNA damage. I see. That, that's scary. That's right. That wouldn't be the case with DNA methylation. I know because I hired a phlebotomist here in LA to visit families to collect blood. And this person didn't have an air conditioner in his car, and it was the hottest day in LA ever. So all the blood tube melted. That was my experiment. And then I just couldn't send this phlebotomist out to go back to the families to collect blood, you know, I felt sorry for the family. So that's why we did this experiment, you know. Yeah. And all I can tell you, we got beautiful data. And so I see that over and over. And people sometimes ask me, what if you have a forensic sample, you know, so let's say a blood stain, you know, that was, let's say, in a room for a couple of weeks or even, let's say, a bone sample collected, you know, a couple of months later. All of these data, in my opinion, would still lead to very good methylation measurements for the purpose of measuring aging. Yeah, because it's so stable. It's just so stable. Do you find that because these methylation patterns obviously are changing with age, you were able to predict first chronological age with pretty 96% or so accuracy. And so, you know, obviously, while they're stable, at the same time, they are changing. Yes. But do they change, like, is there a pattern of change? Like, do they change, you know, every few years, all of a sudden, things rearrange? Or is there, like, a pattern you can see where things start to change every, like, block of time? Yeah. Do you know what I'm saying? Yeah, I do. It's a good question, you know. So our epigenetic... Let me start slowly and say the epigenetic clocks typically track several hundred locations in the genome. For example, the pan-tissue clock is based on 353 locations in the genome. And a question is whether each locus changes, each location gains methylation in a continuous fashion, you know. And probably not, you know. So I think what happens is some locations gain methylation, others lose methylation, and it's a bit random. But on aggregate, once you average hundreds of sites, you know, you kind of average out the noise, the variability, and that gives rise then to this very accurate age estimate. Okay. I definitely want to jump into some questions on mechanism and stuff too, but before I kind of go into the woods, predicting the biological age, your DNA methylation, pheno age, was, if I recall from reading your papers, which I read recently, was able to predict all-cause mortality, disease-specific mortality, like cardiovascular disease-related mortality. That's right. I mean, that was pretty interesting to me. Just to set the stage, so we have some epigenetic clocks whose purpose is really to measure chronologic age, period. But then we have other epigenetic clocks that are really defined to be lifespan predictors, or they are meant to predict time to death, or time to major onset of a disease, or what people call health span, how long are you healthy. And as you mentioned, we have actually two biomarkers. One is called DNA methylation pheno-age, but also another one that was named after the Grim Reaper, called DNA methylation Grim-age. Now, these biomarkers were developed really for that purpose of predicting health span and lifespan, as opposed to measuring aging. And the reason why we have these different tools is that we, of course, plan to use them in human clinical trials of anti-aging interventions. And in that context, you want to see whether an intervention actually resets an epigenetic clock. And that resetting then has a benefit in terms of delaying risk for various diseases. And you mentioned heart disease. A grim age is a pretty good predictor of time to coronary heart disease. Surprisingly, these clocks even predict time to cancer. And that is surprising because it's a measurement based on blood, you know. And so you could ask, why would a measurement in blood be predictive of the onset of various types of cancers in other solid tissues, you know. You can predict it before other clinical diagnostics in some cases? Yeah. I mean, let me start out by saying I'm not sure whether this biomarker is clinically useful, okay, because I'm very scared people think they can now measure their blood and I will predict you will get cancer in 10 years. It's not at that level. However, if you have, for example, a study of 1,000 women and somebody collected their blood in the 1990s, and so for each woman you have follow-up information whether she developed breast cancer or when she developed breast cancer. And if you then analyze the data, you will find that biomarkers such as Grimmage and other biomarkers actually do predict onset to cancer in a statistical fashion. The p-value would be quite significant, you know. But as I said, I wouldn't claim that this is right now ready for prime time in a clinical setting for finding high-risk individuals, because the effect sizes are too small. When I predict, for example, that you will develop heart disease in 15 years, there would be a big error bar associated with it, plus, minus six years or so. So for the individual, it might not be useful. That's a significant amount of time for a person. Yes, exactly. What about disease states? You mentioned cancer. People with cancer or Alzheimer's disease or Parkinson's disease, have you measured the epigenetic age of these individuals, and does it look like it's accelerated aging? Yes. So we looked at blood samples from Parkinson's cases and controls, and there's no question there's an age acceleration effect in blood. It's minor. It's one or two years, but it is there. Alzheimer's disease, we looked at prefrontal cortex samples from the religious order study. And again, we found age acceleration in the prefrontal cortex. When it comes to blood samples from Alzheimer's disease, I think there might be a signal, but if there is a signal, it's very weak. So what other disease did you mention? Cancer. I'm wondering, you know, cancer is a beast. I Thank you. But if there is a signal, it's very weak. What other disease did you mention? Cancer. I'm wondering, cancer is a beast. I mean, there's so many different types. Yeah, cancer is complicated. So the exciting insight is that, yes, blood methylation data indicate that blood samples collected before the person developed cancer show a slight epigenetic age acceleration. So that supports the view that faster epigenetic aging is predictive of future onset of cancer. And that finding has been validated by many groups. Problem is, this association weak. You need really a couple of thousand people to observe it. But the question is, what about if we, for example, measured tumor tissue? Well, tumor tissue, the signal is huge. So if I, for example, when I analyzed breast, malignant breast tissue samples from women with so-called luminal breast cancer, the epigenetic age acceleration is off the chart. So their breast tissue is much older than expected. But it's complicated. Have you compared it to their blood? Like, is tumor tissue the same? No, no, it would be different. I mean, I would say that, I mean, yeah, let's say in breast tissue, we find 10, 15 years age acceleration in malignant tissue, but in blood, I mean, that effect would be much smaller, if at all. That's very, it's very interesting because you mentioned the Alzheimer's disease blood also. Very weak. Very weakening signal. But if you mentioned the, if you measured actual brain tissue, which is where this, you know, it's a neurological disorder, you find a signal. The interesting thing is you said Parkinson's disease, you do measure the signal in the blood. Yes, that's true. It'd be really interesting to know if the immune system is playing a role in Parkinson's disease. Yeah, I can tell you the following. So we did this study of Parkinson's people. Why? Because I was interested in epigenetic aging. However, my software also produces estimates of blood cell counts. And so it turned out that the blood cell counts, in particular neutrophils, were really highly elevated in Parkinson's disease. Huge effect, you know. And so in certain ways, this was completely surprising to me. But this finding has now been validated over and over. So yes, PD cases have highly elevated neutrophil counts. That's very interesting. Yes. So yes, immune system plays a role. We don't know the causal direction, you know. Is it first an immune dysregulation that gives rise to PD, or is it the other way around? Yeah. There's been some interesting links to gut, the origination of the gut and the Parkinson's, and of course the immune system's involved in the gut and all that. So the reason I asked about the cancer tissue, though, was because, you know, another biomarker of aging telomere length, the longer the telomeres are, it's associated with, you know, basically better biological aging in a way because your telomeres get shorter with age. Right. So it's assumed that longer telomeres means younger, right? Yes. But some cancers find a way to like reactiv enzymes involved in building telomeres, like telomerase, for example. And so some cancers have longer telomeres. Absolutely. So if you just were looking at that one biomarker, you'd look at that tissue sample and think, oh, this is... Yeah. I mean, let me make a few comments about telomere length. And so, as you said, by now we know that there is a U-shape behavior. You don't want telomeres that are too short and you don't want to have telomeres that are too long, you know. And so, that's the first statement. The second statement is telomere length per se is actually not a good biomarker for predicting onset of various diseases. Most diseases don't have a strong relationship with telomere length. In particular, when it comes to predicting lifespan, you know, telomere length is actually a shockingly weak predictor of lifespan. For many years, people wrote articles where they claimed there is no relationship to lifespan. By now, the field has moved on to say, well, if you have very large data, you do see a relationship to lifespan. But if we compared telomere length versus an epigenetic clock such as grim age when it comes to predicting lifespan, time to cancer, time to coronary heart disease, I mean, there would be no comparison, you know. So in this sense, telomere length plays a very important role in certain disorders, you know, but it's just not a broad biomarker for aging. Right. How does the epigenetic clock, whether we're talking about the pheno age or DNA grim age, relate to other biomarkers of aging? So does it usually correlate? Like if you have, you know, so there's immunosenescence, which is associated with aging. Yes. relate to other biomarkers of aging. So does it usually correlate? Like if you have, you know, so there's immunosenescence, which is associated with aging, DNA damage, inflammation, there's telomere length. Does it correlate typically, like in the same direction? Yeah, it would. So talking about grim age or pheno age, they would correlate in a consistent fashion. You know, they, so they would have a weak correlation with telomere length to give you a number correlation 0.1. So it's actually a weak correlation, but yes, if you have a thousand people, you pick it up. In general, telomere biology is really a different hallmark of aging compared to epigenetic changes. They measure different aspects of aging, but yeah, there's consistency. So, he learned biology has been...I interviewed Dr. Alyssa Eppel on the podcast. She works closely with Dr. Elizabeth Blackburn. Yes. She's at UCSF, and she has published some studies showing that stress plays a big role in...big role. It plays a pretty good role in telomere biology, so you can find that different types of stress can actually affect telomere length. So lifestyle factors that affect the epigenetic clocks. So for example, diet, exercise, smoking, or even education. So how do those lifestyle factors in general affect epigenetic aging? Yeah, everything your grandmother ever told you about living a healthy lifestyle is kind of corroborated by our epigenetic clocks. So for example, people who eat vegetables, people who exercise, also actually educational level, to some extent even alcohol consumption, you know, show a beneficial effect. Now, the problem is these effects are weak. You know, again, you need a couple of thousand people, then you pick it up. In terms of statistical significance, there's no debate. These are, yeah, clearly these associations are there. But for the individual, the question is, what if I follow the perfect lifestyle? Do I make a big dent on epigenetic aging? And the answer is, unfortunately, not really. I mean, I'm as much of a health nut as many other people in Southern California, you know, so I'm trying to have a healthy lifestyle. But so, yes, you want to avoid diabetes and all of that, you know, and certainly you don't want to smoke. But the truth is a lifestyle intervention will never have a profound impact on aging at a population level, you know. Because what I would like to do is I would like to increase health span by 10, 15 years, you know, and in my opinion, lifestyle interventions won't get us there, you know. In healthy people, okay. So let's say you have a friend who smokes and is obese, then yes, tell your friend to adopt a healthier lifestyle because for this person, it will have a huge effect. But let's say you take somebody like me who is reasonably slender, doesn't smoke. And now you tell me, what about if you become a vegetarian, you know, or what if you double the amount of exercise you do? Will you have a strong effect on my lifespan? And the answer is no, not really. According to the epigenetic clock. So your, the physical activity was only like a week. Yeah, physical activity, yeah, exactly, unfortunately, a week. So I want to say correlation 0.08, for people who know what that means, that's a very weak correlation. In blood, though. Right. So because the question is maybe if we studied heart tissue or muscle, maybe we would find a much more pronounced effect, you know. in blood, we didn't see it. Well, that sort of brings the question about tissue types too as well. Yeah. Example is the effect of obesity on epigenetic aging. Turns out obese people age faster in blood. However, the strongest effect can be found in liver tissue. So obesity greatly accelerates the epigenetic age of liver tissue. And so I think a lot of stress factors really have an organ-specific effect. Conversely, anti-aging interventions also have an organ-specific effect. So, for example, when we evaluated the effect of postmenopausal hormone therapy in women, we found no beneficial effect in blood. However, interestingly, the buccal epithelial cells, so the cells inside of your mouth, they actually revealed that women who took hormone therapy were aging more slowly in these cells. Oh, interesting. Yes. Many cell types are also epithelial cells. I mean, many of your organs have epithelial cells. Yes, yes. Blood cells are a little different. I know the finding made sense because blood cells don't have as many estrogen receptors as buccal epithelial cells. Oh, really? So, yes."

Alzheimer Prevention Discussion

A grim age is a pretty good predictor of time to coronary heart disease. Surprisingly, these clocks even predict time to cancer.

"Yes. So that's one line of evidence, but there's another. So people have these longitudinal epidemiological studies, and they may have collected a blood sample from a person when they were, let's say, 40 years old. And then 15 years later, they get a second blood sample. you know. And so you can then ask the question whether a person who was aging quickly at the first blood draw, did they still age quickly at the second blood draw, you know. And the answer is yes. And conversely, you observe the same for people who age more slowly. And that, in my opinion, this could already be observed when you study, let's say, a person at a young age, age 20, draw their blood. Then if you followed them for 60 years, you know, you would find this consistency that people who are slow agers at age 20, they also slow agers at age 60 or 80. So that does seem to imply the genetic... I mean, unless someone dramatically changed their lifestyle. How stable are these changes in methylation, these methylation patterns that are so-called, you know, the aging clock? How stable are they over a person's lifetime? I mean, they're... I mean, they are remarkably stable. So when we compare to any other genomic measurement, I mean, they would be far more stable than anything I'm aware of. They're far more stable than gene expression, proteomics, metabolomics measurements. All the omics are less stable. And that's really the biological reason why these epigenetic clocks are the most accurate measures of aging. It's just that methylation is so stable. And it's stable not just in vivo, but also when people collect DNA. So we've collected DNA, and then we didn't store the blood tubes properly, so they melted. And we extracted DNA, and the measurements were perfect. you know. So, even on a technical level, they're very stable. That's interesting because I've done experiments intentionally, though, almost the same, where we were collecting blood samples from participants in a trial and I was measuring DNA damage as biomarked by gamma H2AX. And I wanted to know how long I could have blood at room temperature before I started to get artifactual DNA damage happening. And so I did time course and found that after two hours of blood being at room temperature, there's just tons of DNA damage. I see. That, that's scary. That's right. That wouldn't be the case with DNA methylation. I know because I hired a phlebotomist here in LA to visit families to collect blood. And this person didn't have an air conditioner in his car, and it was the hottest day in LA ever. So all the blood tube melted. That was my experiment. And then I just couldn't send this phlebotomist out to go back to the families to collect blood, you know, I felt sorry for the family. So that's why we did this experiment, you know. Yeah. And all I can tell you, we got beautiful data. And so I see that over and over. And people sometimes ask me, what if you have a forensic sample, you know, so let's say a blood stain, you know, that was, let's say, in a room for a couple of weeks or even, let's say, a bone sample collected, you know, a couple of months later. All of these data, in my opinion, would still lead to very good methylation measurements for the purpose of measuring aging. Yeah, because it's so stable. It's just so stable. Do you find that because these methylation patterns obviously are changing with age, you were able to predict first chronological age with pretty 96% or so accuracy. And so, you know, obviously, while they're stable, at the same time, they are changing. Yes. But do they change, like, is there a pattern of change? Like, do they change, you know, every few years, all of a sudden, things rearrange? Or is there, like, a pattern you can see where things start to change every, like, block of time? Yeah. Do you know what I'm saying? Yeah, I do. It's a good question, you know. So our epigenetic... Let me start slowly and say the epigenetic clocks typically track several hundred locations in the genome. For example, the pan-tissue clock is based on 353 locations in the genome. And a question is whether each locus changes, each location gains methylation in a continuous fashion, you know. And probably not, you know. So I think what happens is some locations gain methylation, others lose methylation, and it's a bit random. But on aggregate, once you average hundreds of sites, you know, you kind of average out the noise, the variability, and that gives rise then to this very accurate age estimate. Okay. I definitely want to jump into some questions on mechanism and stuff too, but before I kind of go into the woods, predicting the biological age, your DNA methylation, pheno age, was, if I recall from reading your papers, which I read recently, was able to predict all-cause mortality, disease-specific mortality, like cardiovascular disease-related mortality. That's right. I mean, that was pretty interesting to me. Just to set the stage, so we have some epigenetic clocks whose purpose is really to measure chronologic age, period. But then we have other epigenetic clocks that are really defined to be lifespan predictors, or they are meant to predict time to death, or time to major onset of a disease, or what people call health span, how long are you healthy. And as you mentioned, we have actually two biomarkers. One is called DNA methylation pheno-age, but also another one that was named after the Grim Reaper, called DNA methylation Grim-age. Now, these biomarkers were developed really for that purpose of predicting health span and lifespan, as opposed to measuring aging. And the reason why we have these different tools is that we, of course, plan to use them in human clinical trials of anti-aging interventions. And in that context, you want to see whether an intervention actually resets an epigenetic clock. And that resetting then has a benefit in terms of delaying risk for various diseases. And you mentioned heart disease. A grim age is a pretty good predictor of time to coronary heart disease. Surprisingly, these clocks even predict time to cancer. And that is surprising because it's a measurement based on blood, you know. And so you could ask, why would a measurement in blood be predictive of the onset of various types of cancers in other solid tissues, you know. You can predict it before other clinical diagnostics in some cases? Yeah. I mean, let me start out by saying I'm not sure whether this biomarker is clinically useful, okay, because I'm very scared people think they can now measure their blood and I will predict you will get cancer in 10 years. It's not at that level. However, if you have, for example, a study of 1,000 women and somebody collected their blood in the 1990s, and so for each woman you have follow-up information whether she developed breast cancer or when she developed breast cancer. And if you then analyze the data, you will find that biomarkers such as Grimmage and other biomarkers actually do predict onset to cancer in a statistical fashion. The p-value would be quite significant, you know. But as I said, I wouldn't claim that this is right now ready for prime time in a clinical setting for finding high-risk individuals, because the effect sizes are too small. When I predict, for example, that you will develop heart disease in 15 years, there would be a big error bar associated with it, plus, minus six years or so. So for the individual, it might not be useful. That's a significant amount of time for a person. Yes, exactly. What about disease states? You mentioned cancer. People with cancer or Alzheimer's disease or Parkinson's disease, have you measured the epigenetic age of these individuals, and does it look like it's accelerated aging? Yes. So we looked at blood samples from Parkinson's cases and controls, and there's no question there's an age acceleration effect in blood. It's minor. It's one or two years, but it is there. Alzheimer's disease, we looked at prefrontal cortex samples from the religious order study. And again, we found age acceleration in the prefrontal cortex. When it comes to blood samples from Alzheimer's disease, I think there might be a signal, but if there is a signal, it's very weak. So what other disease did you mention? Cancer. I'm wondering, you know, cancer is a beast. I Thank you. But if there is a signal, it's very weak. What other disease did you mention? Cancer. I'm wondering, cancer is a beast. I mean, there's so many different types. Yeah, cancer is complicated. So the exciting insight is that, yes, blood methylation data indicate that blood samples collected before the person developed cancer show a slight epigenetic age acceleration. So that supports the view that faster epigenetic aging is predictive of future onset of cancer. And that finding has been validated by many groups. Problem is, this association weak. You need really a couple of thousand people to observe it. But the question is, what about if we, for example, measured tumor tissue? Well, tumor tissue, the signal is huge. So if I, for example, when I analyzed breast, malignant breast tissue samples from women with so-called luminal breast cancer, the epigenetic age acceleration is off the chart. So their breast tissue is much older than expected. But it's complicated. Have you compared it to their blood? Like, is tumor tissue the same? No, no, it would be different. I mean, I would say that, I mean, yeah, let's say in breast tissue, we find 10, 15 years age acceleration in malignant tissue, but in blood, I mean, that effect would be much smaller, if at all. That's very, it's very interesting because you mentioned the Alzheimer's disease blood also. Very weak. Very weakening signal. But if you mentioned the, if you measured actual brain tissue, which is where this, you know, it's a neurological disorder, you find a signal. The interesting thing is you said Parkinson's disease, you do measure the signal in the blood. Yes, that's true. It'd be really interesting to know if the immune system is playing a role in Parkinson's disease. Yeah, I can tell you the following. So we did this study of Parkinson's people. Why? Because I was interested in epigenetic aging. However, my software also produces estimates of blood cell counts. And so it turned out that the blood cell counts, in particular neutrophils, were really highly elevated in Parkinson's disease. Huge effect, you know. And so in certain ways, this was completely surprising to me. But this finding has now been validated over and over. So yes, PD cases have highly elevated neutrophil counts. That's very interesting. Yes. So yes, immune system plays a role. We don't know the causal direction, you know. Is it first an immune dysregulation that gives rise to PD, or is it the other way around? Yeah. There's been some interesting links to gut, the origination of the gut and the Parkinson's, and of course the immune system's involved in the gut and all that. So the reason I asked about the cancer tissue, though, was because, you know, another biomarker of aging telomere length, the longer the telomeres are, it's associated with, you know, basically better biological aging in a way because your telomeres get shorter with age. Right. So it's assumed that longer telomeres means younger, right? Yes. But some cancers find a way to like reactiv enzymes involved in building telomeres, like telomerase, for example. And so some cancers have longer telomeres. Absolutely. So if you just were looking at that one biomarker, you'd look at that tissue sample and think, oh, this is... Yeah. I mean, let me make a few comments about telomere length. And so, as you said, by now we know that there is a U-shape behavior. You don't want telomeres that are too short and you don't want to have telomeres that are too long, you know. And so, that's the first statement. The second statement is telomere length per se is actually not a good biomarker for predicting onset of various diseases. Most diseases don't have a strong relationship with telomere length. In particular, when it comes to predicting lifespan, you know, telomere length is actually a shockingly weak predictor of lifespan. For many years, people wrote articles where they claimed there is no relationship to lifespan. By now, the field has moved on to say, well, if you have very large data, you do see a relationship to lifespan. But if we compared telomere length versus an epigenetic clock such as grim age when it comes to predicting lifespan, time to cancer, time to coronary heart disease, I mean, there would be no comparison, you know. So in this sense, telomere length plays a very important role in certain disorders, you know, but it's just not a broad biomarker for aging. Right. How does the epigenetic clock, whether we're talking about the pheno age or DNA grim age, relate to other biomarkers of aging? So does it usually correlate? Like if you have, you know, so there's immunosenescence, which is associated with aging. Yes. relate to other biomarkers of aging. So does it usually correlate? Like if you have, you know, so there's immunosenescence, which is associated with aging, DNA damage, inflammation, there's telomere length. Does it correlate typically, like in the same direction? Yeah, it would. So talking about grim age or pheno age, they would correlate in a consistent fashion. You know, they, so they would have a weak correlation with telomere length to give you a number correlation 0.1. So it's actually a weak correlation, but yes, if you have a thousand people, you pick it up. In general, telomere biology is really a different hallmark of aging compared to epigenetic changes. They measure different aspects of aging, but yeah, there's consistency. So, he learned biology has been...I interviewed Dr. Alyssa Eppel on the podcast. She works closely with Dr. Elizabeth Blackburn. Yes. She's at UCSF, and she has published some studies showing that stress plays a big role in...big role. It plays a pretty good role in telomere biology, so you can find that different types of stress can actually affect telomere length. So lifestyle factors that affect the epigenetic clocks. So for example, diet, exercise, smoking, or even education. So how do those lifestyle factors in general affect epigenetic aging? Yeah, everything your grandmother ever told you about living a healthy lifestyle is kind of corroborated by our epigenetic clocks. So for example, people who eat vegetables, people who exercise, also actually educational level, to some extent even alcohol consumption, you know, show a beneficial effect. Now, the problem is these effects are weak. You know, again, you need a couple of thousand people, then you pick it up. In terms of statistical significance, there's no debate. These are, yeah, clearly these associations are there. But for the individual, the question is, what if I follow the perfect lifestyle? Do I make a big dent on epigenetic aging? And the answer is, unfortunately, not really. I mean, I'm as much of a health nut as many other people in Southern California, you know, so I'm trying to have a healthy lifestyle. But so, yes, you want to avoid diabetes and all of that, you know, and certainly you don't want to smoke. But the truth is a lifestyle intervention will never have a profound impact on aging at a population level, you know. Because what I would like to do is I would like to increase health span by 10, 15 years, you know, and in my opinion, lifestyle interventions won't get us there, you know. In healthy people, okay. So let's say you have a friend who smokes and is obese, then yes, tell your friend to adopt a healthier lifestyle because for this person, it will have a huge effect. But let's say you take somebody like me who is reasonably slender, doesn't smoke. And now you tell me, what about if you become a vegetarian, you know, or what if you double the amount of exercise you do? Will you have a strong effect on my lifespan? And the answer is no, not really. According to the epigenetic clock. So your, the physical activity was only like a week. Yeah, physical activity, yeah, exactly, unfortunately, a week. So I want to say correlation 0.08, for people who know what that means, that's a very weak correlation. In blood, though. Right. So because the question is maybe if we studied heart tissue or muscle, maybe we would find a much more pronounced effect, you know. in blood, we didn't see it. Well, that sort of brings the question about tissue types too as well. Yeah. Example is the effect of obesity on epigenetic aging. Turns out obese people age faster in blood. However, the strongest effect can be found in liver tissue. So obesity greatly accelerates the epigenetic age of liver tissue. And so I think a lot of stress factors really have an organ-specific effect. Conversely, anti-aging interventions also have an organ-specific effect. So, for example, when we evaluated the effect of postmenopausal hormone therapy in women, we found no beneficial effect in blood. However, interestingly, the buccal epithelial cells, so the cells inside of your mouth, they actually revealed that women who took hormone therapy were aging more slowly in these cells. Oh, interesting. Yes. Many cell types are also epithelial cells. I mean, many of your organs have epithelial cells. Yes, yes. Blood cells are a little different. I know the finding made sense because blood cells don't have as many estrogen receptors as buccal epithelial cells. Oh, really? So, yes."

Young Plasma Exchange Discussion

I had read a study, I think it was one of your really good reviews that you published, where you talked about bone marrow transplants.

"And we found that mice that were a young mouse that was connected to an old mouse actually aged faster according to an epigenetic clock in mice. So that part confirmed these parabiosis experiments. So in other words, you can age a young mouse. But that's not what people are interested in. They're interested in the opposite. You take an old mouse and you connect it to a young mouse and then you study the brain of the old mouse and you want to see that the brain is rejuvenated. And for that scenario, actually our results were disappointing. We didn't see a rejuvenation effect. And so now we're trying to get additional data because our first study was underpowered, but one of these months we will have a definitive answer. Yeah. Some of these animal studies are really good for trying to understand mechanism. Yes. And all of this data suggests you've got a clock that can predict chronological age, you've got a clock that can look at your biological age and also predict time to death, lifespan, the grim age. And I mean, something clearly is changing these methylation patterns. So the question is, what is that? Is there a chronic signal that's doing it or is it just completely under genetic control? And maybe they're related, right? So what are your thoughts on the aging process? Yeah, I mean, well, when it comes to these epigenetic clocks, this is the number one weakness of these clocks, that we don't completely understand the molecular mechanism. And coming back to telomere length, that's a great advantage of telomere biology. We really understand very well what regulates telomere length, you know. But yeah, with the epigenetic clocks, this is a very active area of research. Top biologists and labs are working on that very question, you know. And there are many theories. Some people think stem cell biology plays an important role, and that's probably true for many tissues, you know. In certain ways, it could measure aspects of stem cells, for example, how often a stem cell divided. The problem with that interpretation is that the epigenetic clocks work beautifully in neurons, you know, which really don't rejuvenate over the lifespan, you know. And yes, another group thinks that epigenetic clocks might relate to circadian rhythm. So there have been some theories. Now, I believe that also processes that play a role in development must be playing a role here. And the reason is because my original pan-tissue epigenetic clocks works actually beautifully in prenatal brain samples. It works beautifully in various in vitro studies of so-called three-dimensional brains, you know, or in retina samples. So really, it captures aging of gestational age during development, you know. And during development, there's really no noise. This is a highly coordinated process. And so, yes, these processes also must play a role. Yeah, certainly. One thing I had thought about, and I want to get back to development because of the stem cell thing, but one thing that came to my mind with this chronic signal is one particular gene that is methylated during early age but then becomes demethylated as a person ages is this P16-Inc4A gene. It plays a role in cell cycle progression, meaning it basically stops the cell from going on to the next cycle. Yes. So obviously you don't want it to be active during development or early age because your cells are growing. Yes. But it also, so when it becomes active, it stops stem cells. Like hematopoietic stem cells, they aren't growing. So there's like this kind of role in aging where it's sort of basically stopping the stem cell from growing is going to have a negative impact on aging, although it can be positive for cancer because...well, positive for the person that has cancer because it can stop a cancer cell from growing, right? Yes, yes. So the question is, there's a group of demethylases that can take off the methyl group that become active and take it off of this gene. And they're activated by inflammation. So what I'm wondering is, is anyone looking at, obviously, these methyl groups are changing. And so the enzymes that are pulling off methyl groups and the dem demethylases, the enzymes that are putting them on there, methyltransferases, they must be doing something. Yeah, yes. And people are looking at it. What's changing those enzymes? Like, is there a signal there? Or is it a gene? Is it genetic control? Or what is it, you know, that can change? Yeah, you make very good points. So if you want to understand the epigenetic clock, clearly you start with so-called DNA methyltransferases or these TET enzymes. Why? Because they, on the one hand, add methyl groups or remove methyl groups. So that's a low-hanging fruit. And just recently, we and others, several groups, actually found no doubt, when you interfere with these enzymes, you affect epigenetic age. There are very exciting findings where people studied certain developmental disorders where mutations deactivated DNA methyltransferase or mutation rendered it overactive, you know. And sure enough, all of these mutations in humans, you know, affect epigenetic age. And so at that level, we know it has an effect as expected. And the effect is pronounced. It could add five or ten years to a person or the opposite, you know. And does that correlate with the lifespan of disease? See, we don't know. Exactly. That's the question. So some of these children have a developmental disorder and there's various syndromes, you know. And so one is so-called Sotos syndrome. And anyways, and so we see strong deviations in blood in both directions, plus five years, minus five years. So at that level, it's all confirmed. Plus, we right now have, we do mouse crosses, you know, where we knock out these DNA methyl transferases and just to very carefully study it in a very controlled setting. But I can already tell you, you'll find strong effects. Now, but the question you really ask is, well, on this upstream of those, you know, what regulates the clockwork, you know, which, what tells the DNA methyl transferase, go to this location, you know, and deposit and work your magic, you know. Right. And that's where we don't have an understanding yet. And that's why I mentioned the Jumanji demethylases, because that is just one, it's one group of demethylasesases that particularly plays a role in taking off the methyl groups for the Inc. 4A locus. And I remember trying to figure out inflammation plays a role in aging. And so I had seen studies showing, sure enough, TNF-alpha, these things play a role in activating those demethylases. And. And so it's like, well, is that something that's just over time, chronic, that there's a threshold or what? I don't know. Absolutely. I don't know. I've never seen the experiment done with that particular... Yeah. No, I mean, I've seen results from other groups that look at that issue, chronic inflammation, or even looking at these transposons. So there's now some very exciting results that show that some transposons become active in older tissues. And so I've seen some preliminary data where people said this was associated with increased epigenetic aging. And also our finding that HIV is very much associated with accelerated epigenetic aging also points again to this idea of a viral component, you know."

Time Restricted Eating: Fasting

Otherwise, I think my BMI would be 17 right now. There's an acquaintance of mine who runs a pretty popular aging blog.

"It's just happening. I'm trying to do this intermittent fasting, but I don't have much self-discipline."

Vitamin D: Supplementation

People are doing this stuff. It's just happening.

"But yeah, we will now analyze mice from the lab from joe takahashi who did various interventions and probably in a couple of weeks we will have some answers you know whether these strategies have a..."

Epigenetic Aging: Clock

They're interested in the opposite.

"But yeah, we will now analyze mice from the lab from joe takahashi who did various interventions and probably in a couple of weeks we will have some answers you know whether these strategies have a..."

Epigenetic Aging: Clock

Whereas these biomarkers such as an epigenetic clock is hopefully much closer to an innate aging process, you know. We don't want to, it's not my goal to reduce people's stress levels.

"Whereas these biomarkers such as an epigenetic clock is hopefully much closer to an innate aging process, you know."

Sleep Optimization: Rem

He runs the Human Sleep Center at UC Berkeley. I always hate these studies because I don't sleep well.

"And he just talked about all the, you know, basically various diseases and, you know, all-cause mortality and how everything just goes up when sleep quality goes down."

Cellular Senescence Discussion

He runs the Human Sleep Center at UC Berkeley. I always hate these studies because I don't sleep well.

"And so I certainly feel like... I always hate these studies because I don't sleep well. I like the studies where they study the so-called super sleepers, you know, who sleep only five hours a night and still are perfectly healthy."