ON APRIL 15, 2003, NEWSPAPERS, television programs, and websites around the world carried the story: the mapping of the human genome was complete.
There was just one pesky problem: it really wasn’t. There were, in fact, huge gaps in the sequence.
This wasn’t a case of the mainstream news media blowing things out of proportion. Highly respected scientific journals such as Science and Nature told pretty much the same story. It also wasn’t a case of scientists overstating their work. The truth is simply that, at the time, most researchers involved in the thirteen-year, $1 billion project agreed that we’d come as close as we possibly could—given the technology of the time—to identifying each of the 3 billion base pairs in our DNA.
The parts of the genome that were missing, generally overlapping sections of repetitive nucleotides, were just not considered important. These were areas of the code of life that were once derided as “junk DNA” and that are now a little better respected but still generally disregarded as “noncoding.” From the perspective of many of the best minds in science at the time, those regions were little more than the ghosts of genomes past, mostly remnants of dead hitchhiking viruses that had integrated into the genome hundreds of thousands of years ago. The stuff that makes us who we are, it was thought, had largely been identified, and we had what we needed to propel forward our understanding of what makes us human.
Yet by some estimates, that genetic dark matter accounts for as much as 69 percent of the total genome,1 and even within the regions generally regarded as “coding,” some scientists believe, up to 10 percent has yet to be decoded, including regions that impact aging.2
In the relatively short time that has come and gone since 2003, we have come to find out that within the famous double helix, there were sequences that were not just unmapped but essential to our lives. Indeed, many thousands of sequences had gone undetected because the original algorithms to detect genes were written to disregard any gene less than 300 base pairs long. In fact, genes can be as short as 21 base pairs, and today we’re discovering hundreds of them all over the genome.
These genes tell our cells to create specific proteins, and these proteins are the building blocks of the processes and traits that constitute human biology and lived experiences. And as we get closer to identifying a complete sequence of our DNA, we’ve come closer to having a “map” of the genes that control so much of our existence.
Even once we have a complete code, though, there’s something we still won’t be able to find.
We won’t be able to find an aging gene.
We have found genes that impact the symptoms of aging. We’ve found longevity genes that control the body’s defenses against aging and thus offer a path to slowing aging through natural, pharmaceutical, and technological interventions. But unlike the oncogenes that were discovered in the 1970s and that have given us a good target for going to battle against cancer, we haven’t identified a singular gene that causes aging. And we won’t.
Because our genes did not evolve to cause aging.
My journey toward formulating the Information Theory of Aging was a long one. And in no small part, it can be traced to the work of a scientist who toiled without fame but whose work helped set the stage for a lot of the longevity research being done around the world today.
His name was Robert Mortimer, and if there was one adjective that seemed to come up more than any other about him after he passed away, it was “kind.”
“Visionary” was another. “Brilliant,” “inquisitive,” and “hardworking,” too. But I’ve long been inspired by the example Mortimer set for his fellow scientists. Mortimer, who died in 2007, had played a tremendously important role in elevating Saccharomyces cerevisiae from a seemingly lowly, single-celled yeast with a sweet tooth (its name means “sugar-loving”) to its rightful place as one of the world’s most important research organisms.
Mortimer collected thousands of mutant yeast strains in his lab, many of which had been developed right there at the University of California, Berkeley. He could have paid for his research, and then some, by charging the thousands of scientists he supplied through the university’s Yeast Genetic Stock Center. But anyone, from impecunious undergraduates to tenured professors at the world’s best-funded research institutions, could browse the center’s catalog, request any strain, and have it promptly delivered for the cost of postage.3
And because he made it so easy and so inexpensive, yeast research bloomed.
When Mortimer began working on S. cerevisiae alongside fellow biologist John Johnston4 in the 1950s, hardly anyone was interested in yeast. To most, it didn’t seem we could learn much about our complex selves by studying a tiny fungus. It was a struggle to convince the scientific community that yeast could be useful for something more than baking bread, brewing beer, and vinting wine.
What Mortimer and Johnston recognized, and what many others began to realize in the years to come, was that those tiny yeast cells are not so different from ourselves. For their size, their genetic and biochemical makeup is extraordinarily complex, making them an exceptionally good model for understanding the biological processes that sustain life and control lifespans in large complex organisms such as ourselves. If you are skeptical that a yeast cell can tell us anything about cancer, Alzheimer’s disease, rare diseases, or aging, consider that there have been five Nobel Prizes in Physiology or Medicine awarded for genetic studies in yeast, including the 2009 prize for discovering how cells counteract telomere shortening, one of the hallmarks of aging.5
The work Mortimer and Johnston did—and, in particular, a seminal paper in 1959 that demonstrated that mother and daughter yeast cells can have vastly different lifespans—would set the stage for a world-shattering change in the way we view the limits of life. And by the time of Mortimer’s death in 2007, there were some 10,000 researchers studying yeast around the globe.
Yes, humans are separated from yeast by a billion years of evolution, but we still have a lot in common. S. cerevisiae shares some 70 percent of our genes. And what it does with those genes isn’t so different from what we do with them. Like a whole lot of humans, yeast cells are almost always trying to do one of two things: either they’re trying to eat, or they’re trying to reproduce. They’re hungry or they’re horny. As they age, much like humans, they slow down and grow larger, rounder, and less fertile. But whereas humans go through this process over the course of many decades, yeast cells experience it in a week. That makes them a pretty good place to start in the quest to understand aging.
Indeed, the potential for a humble yeast to tell us so much about ourselves—and do so quite quickly relative to other research organisms—was a big part of the reason I decided to begin my career by studying S. cerevisiae. They also smell like fresh bread.
I met Mortimer in Vienna in 1992, when I was in my early 20s and attending the International Yeast Conference—yes, there is such a thing—with my two PhD supervisors, Professor Ian Dawes, a rule-avoiding Australian from the University of New South Wales,6 and Professor Richard Dickinson, a rule-abiding Briton from the University of Cardiff, Wales.
Mortimer was in Vienna to discuss a momentous scientific endeavor: the sequencing of the yeast genome. I was there to be inspired. And I was.7 If I’d harbored any doubts about my decision to dedicate the opening years of my scientific career to a single-celled fungus, they all went away when I came face to face with people who were building great knowledge in a field that had hardly existed a few decades before.
It was shortly after that conference that one of the world’s top scientists in the yeast field, Leonard Guarente of the Massachusetts Institute of Technology, came to Sydney on holiday to visit Ian Dawes. Guarente and I ended up at a dinner together, and I made sure I was sitting opposite him.
I was then a graduate student using yeast to understand an inherited condition called maple syrup urine disease. As you might imagine from its name, the disease is not something most polite people discuss over dinner. Guarente, though, engaged me in a scientific discussion with a curiosity and enthusiasm that was nothing short of enchanting. The conversation soon turned to his latest project—he had begun studying aging in yeast the past few months—work that had its roots in the workable genetic map that Mortimer had completed in the mid-1970s.
That was it. I had a passion for understanding aging, and I knew something about wrangling a yeast cell with a microscope and micromanipulator. Those were essential skills needed to figure out why yeast age. That night, Guarente and I agreed on one thing: if we couldn’t solve the problem of aging in yeast, we had no chance in humans.
I didn’t just want to work with him. I had to work with him.
Dawes wrote him to tell him that I was keen to join his lab and I was “skilled at the bench.”
“It would be a pleasure to work with David,” he replied a few weeks later, the same way he probably did to so many other enthusiastic applicants. “But he’s got to come with his own funding.” Later I learned he had been excited only because he’d thought I was the other student he’d met at dinner.
I had a foot in the door, but my chances were slim. At the time, foreigners weren’t considered for prestigious postdoctoral awards in the United States, but I insisted I be interviewed and paid for a flight to Boston myself. I was interviewed by a giant in the stem cell field, Douglas Melton, for a Helen Hay Whitney Foundation Fellowship, which has been providing research support to postdoctoral biomedical students since 1947. After waiting in line outside his office with the other four candidates, I had my chance. This was my moment. I don’t remember being nervous. I figured I probably wouldn’t get the award anyway. So I went for it.
I told Melton about my lifelong quest to understand aging and find “life-giving genes,” then sketched out on his whiteboard how the genes work and what I’d be doing for the next three years if I got the money. To show my gratitude, I gave him a bottle of red wine that I’d brought from Australia.
Afterward, two things became clear. One, don’t bring wine to an interview because it can be seen as a bribe. And two, Melton must have liked what I said and how I said it, because I flew home, got the fellowship, and then got onto a plane back to Boston. It was, without a doubt, the most life-changing meeting of my life.8
At the time of my arrival, in 1995, I had expected to build our understanding of aging by studying Werner syndrome, a terrible disease that occurs in less than 1 in 100,000 live births, with symptoms that include a loss of body strength, wrinkles, gray hair, hair loss, cataracts, osteoporosis, heart problems, and many other telltale signs of aging—not among folks in their 70s and 80s but rather among people in their 30s and 40s. Life expectancy for someone with Werner is 46 years.
Within two weeks of my arrival in the United States, though, a research team at the University of Washington, headed by the wise and supportive grandfather of aging research, George Martin, announced that they had found the gene that, when mutated, causes Werner syndrome.9 It was deflating at the time to have been “scooped,” but the discovery allowed me to take a bigger first step toward my ultimate objective. Indeed, it became the key to formulating the Information Theory of Aging.
Now that the Werner gene, known as WRN, had been identified in humans, the next step was to test if the similar gene in yeast had the same function. If so, we could use yeast to more rapidly determine the cause of Werner syndrome and perhaps help us better understand aging in general. I marched into Guarente’s office to tell him I was now studying Werner’s syndrome in yeast and that’s how we would solve aging.
In yeast, the equivalent of the WRN gene is Slow Growth Suppressor 1, or SGS1. The gene was already suspected to code for a type of enzyme called a DNA helicase that untangles tangled strands of DNA before they break. Helicases are especially important in repetitive DNA sequences that are inherently prone to tangling and breaking. Functionality of proteins, such as the ones coded for by the Werner gene, is therefore vital, since more than half of our genome is, in fact, repetitive.
Through a gene-swapping process in which cells are tricked into picking up extra pieces of DNA, we swapped out the functional SGS1 gene with a mutant version. In effect, we were testing to see if it was possible to give the yeast Werner syndrome.
After the swap, the yeast cells’ lifespan was cut in half. Ordinarily, this would not have been news. Many events unrelated to aging—such as being eaten by a mite, drying out on a grape, or being placed in an oven—can and do shorten the lifespan of yeast cells. And here we’d messed with their DNA, which could have short-circuited the cells in a thousand different ways to cause early death.
But those cells weren’t just dying. They were dying after a precipitous decline in health and function. As the SGS1 mutants became older, they slowed down in their cell cycle. They grew larger. Both male and female “mating-type” genes (descendants of gene A) were switched on at the same time, so they were sterile and couldn’t mate. These were all known hallmarks of aging in yeast. And it was happening more quickly in the mutants we’d made. It certainly looked like a yeast version of Werner’s.
Using specialized stains, we colored the DNA blue and used red for the nucleolus, which sits inside the nucleus of all eukaryotic cells. That made it easier to see under the microscope what was happening at a cellular level.
And what was happening was fascinating.
The nucleolus is a part of the nucleus in which ribosomal DNA, or rDNA, resides. rDNA is copied into ribosomal RNA, which is used by ribosome enzymes to stitch amino acids together to make every new protein.
In the aged SGS1 cells, the nucleolus looked as if it had exploded. Instead of a single red crescent swimming in a blue ocean, the nucleolus was scattered into half a dozen small islands. It was tragic and beautiful. The picture, which would later appear in the August 1997 issue of the prestigious journal Science, still hangs in my office.
What happened next was both enchanting and illuminating. In response to the damage, like rats to the call of the Pied Piper, the protein called Sir2—the first known sirtuin, which is encoded by the gene SIR210 and descended from gene B—had moved away from the mating genes that control fertility and into the nucleolus.
That was a beautiful sight to me, but it was a problem for the yeast. Sir2 has an important job: it is an epigenetic factor, an enzyme that sits on genes, bundles up the DNA, and keeps them silent. At the molecular level, Sir2 achieves this via its enzymatic activity, making sure that chemicals called acetyls don’t accumulate on the histones and loosen the DNA packaging.
When sirtuins left the mating genes—the ones descended from gene A that controlled fertility and reproduction—the mutant cells turned on both male and female genes, causing them to lose their sexual identity, just as in normal old cells, but much earlier.
I didn’t understand at first why the nucleolus was exploding, let alone why the sirtuins were moving toward it as the cells grew older. I agonized over the question for weeks.
And then one night, after a long day in the lab, I woke up with an idea.
It came in the space between sleep-deprived delirium and deep dreaming. The wisps of a concept. A few words jumbled together. A muddled picture of something. That was enough, though, to jolt me awake and pull me from my bed.
I grabbed my notebook and went to the kitchen. There, hunched over the table in the early morning hours of October 28, 1996, I began to write.
I wrote for about an hour, jotting down ideas, drawing pictures, sketching out graphs, formulating new equations.11 Scientific observations that had previously made no sense to me were falling perfectly into a larger picture. Broken DNA causes genome instability, I wrote, which distracts the Sir2 protein, which changes the epigenome, causing the cells to lose their identity and become sterile while they fixed the damage. Those were the analog scratches on the digital DVDs. Epigenetic changes cause aging.
There was, I imagined, a singular process that controlled them all. Not a countless number of separate cellular changes or diseases. Not even a set of hallmarks that could be addressed one at a time. There was something bigger—and more singular—than any of that.
This was the foundation for understanding the survival circuit and its role in aging.
The next day I showed Guarente my notes. I was excited; it felt like the biggest idea I’d ever had. But I was nervous, too; afraid he would find a hole in my logic and tear it apart. Instead, he looked over my notebook quietly, asked a few questions, and sent me on my way with six words.
“I like it,” he said. “Go prove it.”
To understand the Information Theory of Aging, we need to pay another visit to the epigenome, the part of the cell that the sirtuins help control.
Up close, the epigenome is more complex and wonderful than anything we humans have invented. It consists of strands of DNA wrapped around spooling proteins called histones, which are bound up into bigger loops called chromatin, which are bound up into even bigger loops called chromosomes.
Sirtuins instruct the histone spooling proteins to bind up DNA tightly, while they leave other regions to flail around. In this way, some genes stay silent, while others can be accessed by DNA-binding transcription factors that turn genes on.12 Accessible genes are said to be in “euchromatin,” while silent genes are in “heterochromatin.” By removing chemical tags on histones, sirtuins help prevent transcription factors from binding to genes, converting euchromatin into heterochromatin.
Every one of our cells has the same DNA, of course, so what differentiates a nerve cell from a skin cell is the epigenome, the collective term for the control systems and cellular structures that tell the cell which genes should be turned on and which should remain off. And this, far more than our genes, is what actually controls much of our lives.
One of the best ways to visualize this is to think of our genome as a grand piano.13 Each gene is a key. Each key produces a note. And from instrument to instrument, depending on the maker, the materials, and the circumstances of manufacturing, each will sound a bit different, even if played the exact same way. These are our genes. We have about 20,000 of them, give or take a few thousand.14
Each key can also be played pianissimo (soft) or forte (with force). The notes can be tenuto (held) or allegretto (played quickly). For master pianists, there are hundreds of ways to play each individual key and endless ways to play the keys together, in chords and combinations that create music we know as jazz, ragtime, rock, reggae, waltzes, whatever.
The pianist that makes this happen is the epigenome. Through a process of revealing our DNA or bundling it up in tight protein packages, and by marking genes with chemical tags called methyls and acetyls composed of carbon, oxygen, and hydrogen, the epigenome uses our genome to make the music of our lives.
Yes, sometimes the size, shape, and condition of a piano dictate what a pianist can do with it. It’s tough to play a concerto on an eighteen-key toy piano, and it’s mighty hard to make beautiful music on an instrument that hasn’t been tuned in fifty years. Likewise, the genome certainly dictates what the epigenome can do. A caterpillar can’t become a human being, but it can become a butterfly by virtue of changes in epigenetic expression that occur during metamorphosis, even though its genome never changes. Similarly, the child of two parents from a long line of people with black hair and brown eyes isn’t likely to develop blond hair and blue eyes, but twin agouti mice in the lab can turn out brown or golden, depending on how much the Agouti gene is turned on during gestation by environmental influences on the epigenome, such as folic acid, vitamin B12, genistein from soy, or the toxin bisphenol A.15
Similarly, among monozygotic human twins, epigenetic forces can drive two people with the same genome in vastly different directions. It can even cause them to age differently. You can see this clearly in side-by-side photographs of the faces of smoking and nonsmoking twins; their DNA is still largely the same, but the smokers have bigger bags under their eyes, deeper jowls below their chins, and more wrinkles around their eyes and mouths. They are not older, but they’ve clearly aged faster. Studies of identical twins place the genetic influences on longevity at between 10 and 25 percent which, by any estimation, is surprisingly low.16
Our DNA is not our destiny.
Now imagine you’re in a concert hall. A virtuoso pianist is seated at a gorgeously polished Steinway grand. The concerto begins. The music is beautiful, breathtaking. Everything is perfect.
But then, a few minutes into the piece, the pianist misses a key. The first time it happens, it’s almost unnoticeable—an extra D, perhaps, in a chord that doesn’t need that note. Embedded in so many perfectly played notes, hidden among an otherwise flawless chord in an otherwise perfect melody, it’s nothing to worry about. But then, a few minutes later, it happens again. And then, with increasing frequency, again and again and again.
It’s important to remember that there is nothing wrong with the piano. And the pianist is playing most of the notes prescribed by the composer. She’s just also playing some extra notes. Initially, this is just annoying. Over time it becomes unsettling. Eventually it ruins the concerto. Indeed, we’d assume that there was something wrong with the pianist. Someone might even rush onto the stage to make sure she is all right.
Epigenetic noise causes the same kind of chaos. It is driven in large part by highly disruptive insults to the cell, such as broken DNA, as it was in the original survival circuit of M. superstes and in the old yeast cells that lost their fertility. And this, according to the Information Theory of Aging, is why we age. It’s why our hair grays. It’s why our skin wrinkles. It’s why our joints begin to ache. Moreover, it’s why each one of the hallmarks of aging occurs, from stem cell exhaustion and cellular senescence to mitochondrial dysfunction and rapid telomere shortening.
This is, I acknowledge, a bold theory. And the strength of a theory is based on how well it predicts the results of rigorous experiments, often millions of them, the number of phenomena it can explain, and its simplicity. The theory was simple, and it explained a lot. As good scientists, what we had left to do was to try our best to disprove it and see how long it survived.
To get started, Guarente and I had to get our eyes on some yeast DNA.
We used a technique called a Southern blot, a method of separating DNA based on its size and conformation and lighting it up with a radioactive DNA probe. In the first experiment, we noticed something spectacular. Normally, the rDNA of a yeast cell that is made visible by a Southern blot is tightly packed, like a new spool of rope, with a few barely visible wispy loops of supercoiled DNA. But the rDNA of the yeast cells we’d created in our lab—the Werner mutants that seemed to be aging rapidly—were madly unpacking, like a vacuum-sealed bag of yarn that had been ripped open.
LESSONS FROM YEAST CELLS ABOUT WHY WE AGE. In young yeast cells, male and female “mating-type information” (gene A) is kept in the “off” position by the Sir2 enzyme, the first sirtuin (encoded by a descendant of gene B). The highly repetitive ribosomal DNA (rDNA) is unstable, and toxic DNA circles form; these recombine and eventually accumulate to toxic levels in old cells, killing them. In response to DNA circles and the perceived genome instability, Sir2 moves away from silent mating-type genes to help stabilize the genome. Both male and female genes turn on, causing infertility, the main hallmark of yeast aging.
The rDNA was in a state of chaos. The genome, it seemed, was fragmenting. DNA was recombining and amplifying, showing up on the Southern blot as dark spots and wispy circles, depending on how coiled up and twisted they were. We called those loops extrachromosomal ribosomal DNA circles, or ERCs, and they were accumulating as the mutants aged.
If we had indeed induced aging, then we would see this same pattern emerge in yeast cells that had aged normally.
We don’t count the age of a single yeast cell with birthday candles. They simply don’t last that long. Instead, aging in yeast is measured by the number of times a mother cell divides to produce daughter cells. In most cases, a yeast cell gets to about 25 divisions before it dies. That, however, makes obtaining old yeast cells an exceptionally challenging task. Because by the time an average yeast cell expires, it is surrounded by 225, or 33 million, of its descendants.
It took a week of work, a lot of sleepless nights, and a whole lot of caffeinated beverages to collect enough regular old cells. The next day, when I developed the film to visualize the rDNA, what I saw astounded me.17
Just like the mutants, the normal old yeast cells were packed with ERCs.
That was a “Eureka!” moment. Not proof—a good scientist never has proof of anything—but the first substantial confirmation of a theory, the foundation upon which I and others would build more discoveries in the years to come.
The first testable prediction was if we put an ERC into very young yeast cells—and we devised a genetic trick to do that—the ERCs would multiply and distract the sirtuins, and the yeast cells would age prematurely, go sterile, and die young—and they did. We published that work in December 1997 in the scientific journal Cell, and the news broke around the world: “Scientists figured out a cause of aging.”
It was there and then that Matt Kaeberlein, a PhD student at the time, arrived at the lab. His first experiment was to insert an extra copy of SIR2 into the genome of yeast cells to see if it could stabilize the yeast genome and delay aging. When the extra SIR2 was added, ERCs were prevented, and he saw a 30 percent increase in the yeast cells’ lifespan, as we’d been hoping. Our hypothesis seemed to be standing up to scrutiny: the fundamental, upstream cause of sterility and aging in yeast was the inherent instability of the genome.
What emerged from those initial results in yeast, and another decade of pondering and probing mammalian cells, was a completely new way to understand aging, an information theory that would reconcile seemingly disparate factors of aging into one universal model of life and death. It looked like this:
Youth → broken DNA → genome instability → disruption of DNA packaging and gene regulation (the epigenome) → loss of cell identity → cellular senescence → disease → death.
The implications were profound: if we could intervene in any of these steps, we might help people live longer.
But what if we could intervene in all of them? Could we stop aging?
Theories must be tested and tested and tested some more—not just by one scientist but by many. And to that end, I was fortunate to have been put onto a research team that included some of the most brilliant and insightful scientists in the world.
There was Lenny Guarente, our indefatigable mentor. There was also Brian Kennedy, who started the yeast-aging project in Lenny’s lab and has since played a tremendously important role in understanding premature aging diseases and the impact of genes and molecules that increase health and longevity in model organisms. There were Monica Gotta and Susan Gasser at the University of Geneva, who are now some of the most influential researchers in the field of gene regulation; Shin-ichiro Imai, now a professor at Washington University, who discovered that sirtuins are NAD-utilizing enzymes and now does research into how the body controls sirtuins; Kevin Mills, who ran a lab in Maine, then became a cofounder of and chief scientific officer at Cyteir Therapeutics, which develops novel ways to fight cancer and autoimmune diseases; Nicanor Austriaco, who started the project with Brian, now teacher of biology and theology at Providence College, a great combo; Tod Smeal, chief scientific officer of cancer biology at the global pharmaceutical company Eli Lilly; David Lombard, who is now a researcher in the field of aging at the University of Michigan; Matt Kaeberlein, a professor at the University of Washington, who is testing molecules on dog longevity; David McNabb, whose lab at the University of Arkansas has made key and lifesaving discoveries about fungal pathogens; Bradley Johnson, an expert on human aging and cancer at the University of Pennsylvania; and Mala Murthy, a prominent neuroscientist now at Princeton.
Again and again I have been greatly privileged in the matter of those who work around me. And that was never truer than it was in Guarente’s lab at MIT. It was a dream team, and I often felt humbled by the people with whom I was surrounded.
When I began my career in this field, I dreamt of publishing just one study in a top-tier journal. During those years, our group was publishing one every few months.
We demonstrated that the redistribution of Sir2 to the nucleolus is a response to numerous DNA breakages, which happen as a result of ERCs multiplying and inserting back into the genome or joining together to form superlarge ERCs. When Sir2 moves to combat DNA instability, it causes sterility in old, bloated yeast cells. That was the first step of the survival circuit, though at the time we had no idea that it was as ancient and as essential to our very existence as it turned out to be.
We told the world that we could give yeast a Werner-like syndrome, causing exploded nucleoli.18 We described the ways in which mutants of SGS1, those we’d plagued with the yeast equivalent to the Werner syndrome mutation, accumulated ERCs more rapidly, leading to premature aging and a shortened lifespan.19 Crucially, by demonstrating that if you add an ERC to young cells they age prematurely, we had crucial evidence that ERCs don’t just happen during aging, they cause it. And by artificially breaking the DNA in the cell and watching the cellular response, we showed why sirtuins move—to help with DNA repair.20 That turned out to be the second step of the survival circuit.21 The DNA damage that gave rise to the ERCs was distracting Sir2 from the mating-type genes, causing them to become sterile, a hallmark of yeast aging.
It was epigenomic noise in its purest form.
It took another twenty years to learn if those findings in yeast were relevant to organisms more complex than yeast. We mammals have seven sirtuin genes that have evolved a variety of functions beyond what simple SIR2 can do. Three of them, SIRT1, SIRT6, and SIRT7, are critical to the control of the epigenome and DNA repair. The others, SIRT3, SIRT4, and SIRT5, reside in mitochondria, where they control energy metabolism, while SIRT2 buzzes around the cytoplasm, where it controls cell division and healthy egg production.
There had been many clues along the way. Brown University’s Stephen Helfand showed that adding extra copies of the dSir2 gene to fruit flies suppresses epigenetic noise and extends their lifespan. We found that SIRT1 in mammals moves from silent genes to help repair broken DNA in mouse and human cells.22 But the true extent to which the survival circuit is conserved between yeast and humans wasn’t fully known until 2017, when Eva Bober’s team at the Max Planck Institute for Heart and Lung Research in Bad Nauheim, Germany, reported that sirtuins stabilize human rDNA.23 Then, in 2018, Katrin Chua at Stanford University found that, by stabilizing human rDNA, sirtuins prevent cellular senescence—essentially the same antiaging function as we had found for sirtuins in yeast twenty years earlier.24
That was an astonishing revelation: over a billion years of separation between yeast and us, and, in essence, the circuit hadn’t changed.
By the time those findings appeared, though, it was clear to me that epigenomic noise was a likely catalyst of human aging. Two decades of research had already been leading us in that direction.25
In 1999, I moved from MIT across the river to Harvard Medical School, where I set up a new lab on aging. There I was hoping to answer a new question that had increasingly been occupying my thoughts.
I had noticed that yeast cells fed with lower amounts of sugar were not just living longer, but their rDNA was exceptionally compact—significantly delaying the inevitable ERC accumulation, catastrophic numbers of DNA breaks, nucleolar explosion, sterility, and death.
Why was that happening?
Our DNA is constantly under attack. On average, each of our forty-six chromosomes is broken in some way every time a cell copies its DNA, amounting to more than 2 trillion breaks in our bodies per day. And that’s just the breaks that occur during replication. Others are caused by natural radiation, chemicals in our environment, and the X-rays and CT scans that we’re subjected to.
If we didn’t have a way to repair our DNA, we wouldn’t last long. That’s why, way back in primordium, the ancestors of every living thing on this planet today evolved to sense DNA damage, slow cellular growth, and divert energy to DNA repair until it was fixed—what I call the survival circuit.
Since the yeast work, evidence that yeast aren’t so different from us has continued to accumulate. In 2003, Michael McBurney from the University of Ottawa in Canada discovered that mouse embryos manipulated to be unable to produce one of the seven sirtuin enzymes, SIRT1, couldn’t last past the fourteenth day of development—about two-thirds of the way into a mouse’s gestation period.26 Among the reasons, the team reported in the journal Cancer Cell, was an impaired ability to respond to and repair DNA damage.27 In 2006, Frederick Alt, Katrin Chua, and Raul Mostovslavsky at Harvard showed that mice engineered to lack SIRT6 underwent the typical signs of aging faster along with shortened lifespans.28 When the scientists knocked out a cell’s ability to create this vital protein, the cell lost its ability to repair double-strand DNA breaks, just as we had showed in yeast back in 1999.
If you are skeptical, and you should be, you might assume these SIRT mutant mice could just be sick and, therefore, short lived. But adding in more copies of the sirtuin genes SIRT1 and SIRT6 does just the opposite: it increases the health and extends the lifespan of mice, just as adding extra copies of the yeast SIR2 gene does in yeast.29 Credit for these discoveries goes to two of my previous colleagues, Shin-ichiro Imai, my former drinking buddy at the Guarente lab, and Haim Cohen, my first postdoc at Harvard.
In yeast, we had shown that DNA breaks cause sirtuins to relocalize away from silent mating-type genes, causing old cells to become sterile. That was a simple system, and we’d figured it out in a few years.
But is the survival circuit causing aging in mammals? What parts of the system survived the billion years, and which are yeast specific? Those questions are on the cutting edge of human knowledge right now, but the answers are beginning to reveal themselves.
What I’m suggesting is that the SIR2 gene in yeast and the SIRT genes in mammals are all descendants of gene B, the original gene silencer in M. superstes.30 Its original job was to silence a gene that controlled reproduction.
In mammals, the sirtuins have since taken on a variety of new roles, not just as controllers of fertility (which they still are). They remove acetyls from hundreds of proteins in the cell: histones, yes, but also proteins that control cell division, cell survival, DNA repair, inflammation, glucose metabolism, mitochondria, and many other functions.
I’ve come to think of sirtuins as the directors of a multifaceted disaster response corps, sending out a variety of specialized emergency teams to address DNA stability, DNA repair, cell survivability, metabolism, and cell-to-cell communication. In a way, this is like the command center for the thousands of utility workers who descended upon Louisiana and Mississippi in the wake of Hurricane Katrina in 2005. Most of the workers weren’t from the Gulf Coast, but they came, did their level best to fix what was broken, and then went home. Some were working in the storm-ravaged communities for a few days and others for a few weeks before returning to their normal lives. And for most, it wasn’t the first or last time they had done something like that; anytime there’s a mass disaster that impacts utilities, they swoop in to help.
When they’re home, those folks take care of the typical business of being at home: paying bills, mowing lawns, coaching baseball, whatever. But when they’re away, helping keep places like the Gulf Coast from descending into anarchy—a condition that would have had disastrous results for the rest of the nation—a lot of those things have to be put on hold.
When sirtuins shift from their typical priorities to engage in DNA repair, their epigenetic function at home ends for a bit. Then, when the damage is fixed and they head back to home base, they get back to doing what they usually do: controlling genes and making sure the cell retains its identity and optimal function.
But what happens when there’s one emergency after another to tend to? Hurricane after hurricane? Earthquake after earthquake? The repair crews are away from home a lot. The work they normally do piles up. The bills come due, then overdue, and then the folks from collections start calling. The grass grows too long, and soon the president of the neighborhood association is sending nastygrams. The baseball team goes coachless, and the team devolves into the Bad News Bears. And most of all, one of the most important things they do while at home—reproducing—doesn’t get done. This form of hormesis, the original survival circuit, works fine to keep organisms alive in the short term. But unlike longevity molecules that simply mimic hormesis by tweaking sirtuins, mTOR, or AMPK, sending out the troops on fake emergencies, these real emergencies create life-threatening damage.
What could cause so many emergencies? DNA damage. And what causes that? Well, over time, life does. Malign chemicals. Radiation. Even normal DNA copying. These are the things that we’ve come to believe are the causes of aging, but there is a subtle but vital shift we have to make in that manner of thinking. It’s not so much that the sirtuins are overwhelmed, though they probably are when you are sunburned or get an X-ray; what’s happening every day is that the sirtuins and their coworkers that control the epigenome don’t always find their way back to their original gene stations after they are called away. It’s as if a few emergency workers who came to address the damage done in the Gulf Coast by Katrina had lost their home address. Then disaster strikes again and again, and they must redeploy.
Wherever epigenetic factors leave the genome to address damage, genes that should be off, switch on and vice versa. Wherever they stop on the genome, they do the same, altering the epigenome in ways that were never intended when we were born.
Cells lose their identity and malfunction. Chaos ensues. The chaos materializes as aging. This is the epigenetic noise that is at the heart of our unified theory.
How does the SIR2 gene actually turn off genes? SIR2 codes for a specialized protein called a histone deacetylase, or HDAC, that enzymatically cleaves the acetyl chemical tags from histones, which, as you’ll recall, causes the DNA to bundle up, preventing it from being transcribed into RNA.
When the Sir2 enzyme is sitting on the mating-type genes, they remain silent and the cell continues to mate and reproduce. But when a DNA break occurs, Sir2 is recruited to the break to remove the acetyl tags from the histones at the DNA break. This bundles up the histones to prevent the frayed DNA from being chewed back and to help recruit other repair proteins. Once the DNA repair is complete, most of the Sir2 protein goes back to the mating-type genes to silence them and restore fertility. That is, unless there is another emergency, such as the massive genome instability that occurs when ERCs accumulate in the nucleoli of old yeast cells.
For the survival circuit to work and for it to cause aging, Sir2 and other epigenetic regulators must occur in “limiting amounts.” In other words, the cell doesn’t make enough Sir2 protein to simultaneously silence the mating-type genes and repair broken DNA; it has to shuttle Sir2 between the various places on an “as-needed” basis. This is why adding an extra copy of the SIR2 gene extends lifespan and delays infertility: cells have enough Sir2 to repair DNA breaks and enough Sir2 to silence the mating-type genes.31
Over the past billion years, presumably millions of yeast cells have spontaneously mutated to make more Sir2, but they died out because they had no advantage over other yeast cells. Living for 28 divisions was no advantage over those that lived for 24 and, because Sir2 uses up energy, having more of the protein may have even been a disadvantage. In the lab, however, we don’t notice any disadvantage because the yeast are given more sugar than they could possibly ever eat. By adding extra copies of the SIR2 gene, we gave the yeast cells what evolution failed to provide.
If the information theory is correct—that aging is caused by overworked epigenetic signalers responding to cellular insult and damage—it doesn’t so much matter where the damage occurs. What matters is that it is being damaged and that sirtuins are rushing all over the place to address that damage, leaving their typical responsibilities and sometimes returning to other places along the genome where they are silencing genes that aren’t supposed to be silenced. This is the cellular equivalent of distracting the cellular pianist.
To prove that, we needed to break some mouse DNA.
It’s not hard to intentionally break DNA. You can do it with mechanical shearing. You can do it with chemotherapy. You can do it with X-rays.
But we needed to do it with precision—in a way that wouldn’t create mutations or impact regions that affect any cellular function. In essence, we needed to attack the wastelands of the genome. To do that, we got our hands on a gene similar to Cas9, the CRISPR gene-editing tool from bacteria that cuts DNA at precise places.
The enzyme we chose for our experiments comes from a goopy yellow slime mold called Physarum polycephalum, which literally means “many-headed slime.” Most scientists believe that this gene, called I-PpoI, is a parasite that serves only to copy itself. When it cuts the slime mold genome, another copy of I-PpoI is inserted. It is the epitome of a selfish gene.
That’s in a slime mold, its native habitat. But when I-PpoI finds itself in a mouse cell, it doesn’t have all the slime mold machinery to copy itself. So it floats around and cuts DNA at just a few places in the mouse genome, and there is no copying process. Instead, the cell has no problem pasting the DNA strands back together, leaving no mutations, which is exactly what we were looking for to engage the survival circuit and distract the sirtuins. DNA-editing genes such as Cas9 and I-PpoI are nature’s gifts to science.
To create a mouse to test the information theory, we inserted I-PpoI into a circular DNA molecule called a plasmid, along with all the regulatory DNA elements needed to control the gene, and then inserted that DNA into the genome of a mouse embryonic stem cell line we were culturing in plastic dishes in the lab. We then injected the genetically modified stem cells into a 90-cell mouse embryo called a blastocyst, implanted it into a female mouse’s uterus, and waited about twenty days for a baby mouse to show up.
This all sounds complicated, but it’s not. After some training, a college student can do it. It’s such a commodity these days, you can even order a mouse out of a catalog or pay a company to make you one to your specifications.
The baby mice were born perfectly normal, as expected, since the cutting enzyme was switched off at that stage. We called them affectionately “ICE mice,” ICE standing for “Inducible Changes to the Epigenome.” The “inducible” part of the acronym is vital—because there’s nothing different about these mice until we feed them a low dose of tamoxifen. This is an estrogen blocker that is normally used to treat human cancers, but in this case, we’d engineered the mouse so that tamoxifen would turn on the I-PpoI gene. The enzyme would go to work, cutting the genome and slightly overwhelming the survival circuit, without killing any cells. And since tamoxifen has a half-life of only a couple of days, removing it from the mice’s food would turn off the cutting.
The mice might have died. They might have grown tumors. Or they might have been perfectly fine, no worse off than if they’d received a dental X-ray. Nobody had ever done this before in a mouse, so we didn’t know. But if our hypothesis about epigenetic instability and aging was correct, the tamoxifen would work like the potion that Fred and George Weasley used to age themselves in Harry Potter and the Goblet of Fire.
And it worked. Like wizardry, it did.
During the treatment, the mice were fine, oblivious to the DNA cutting and sirtuin distraction. But a few months later, I got a call from a postdoc who was taking care of our lab’s animals while I was on a trip to my lab in Australia.
“One of the mice is really sick,” she said. “I think we need to put it down.”
I asked her to text me a photo of the mouse she was talking about. When the photo came over my phone, I couldn’t help but laugh.
“That’s not a sick mouse,” I replied. “That’s an old mouse.”
“David,” she said, “I think you’re mistaken. It says here that it’s the sister of these other mice in the cage, and they’re perfectly normal.”
Her confusion was understandable. At 16 months old, a regular lab mouse still has a thick coat of fur, a sturdy tail, a muscular figure, perky ears, and clear eyes. A tamoxifen-triggered ICE mouse at the same age has thinning, graying hair, a bent spine, paper-thin ears, and cloudy eyes.
Remember, we’d done nothing to change the genome. We’d simply broken the mice’s DNA in places where there aren’t any genes and forced the cell to paste, or “ligate,” them back together. Just to make sure, later we broke the DNA in other places, too, with the same results. Those breaks had induced a sirtuin response. When those fixers went to work, their absence from their normal duties and presence on other parts of the genome altered the ways in which lots of genes were being expressed at the wrong time.
THE MAKING OF THE ICE MOUSE TO TEST IF THE CAUSE OF AGING MIGHT BE INFORMATION LOSS. A gene from a slime mold that encodes an enzyme that cuts DNA at a specific place was inserted into a stem cell and injected into an embryo to generate the ICE mouse. Turning on the slime mold gene cut the DNA and distracted the sirtuins, causing the mouse to undergo aging.
Those findings were aligned to discoveries being made by Trey Ideker and Kang Zhang, at UC San Diego, and Steve Horvath, at UCLA. Steve’s name stuck, and today he’s the namesake of the Horvath Clock—an accurate way of estimating someone’s biological age by measuring thousands of epigenetic marks on the DNA, called methylation. We tend to think of aging as something that begins happening to us at midlife, because that’s when we start to see significant changes to our bodies. But Horvath’s clock begins ticking the moment we are born. Mice have an epigenetic clock, too. Were the ICE mice older than their siblings? Yes, they were—about 50 percent older.
We’d found life’s master clock winder.
In another manner of thinking, we’d scratched up the DVD of life about 50 percent faster than it normally gets scratched. The digital code that is, and was, the basic blueprint for our mice was the same as it had always been. But the analog machine built to read that code was able to pick up only bits and pieces of the data.
Here’s the vital takeaway: we could age mice without affecting any of the most commonly assumed causes of aging. We hadn’t made their cells mutate. We hadn’t touched their telomeres. We hadn’t messed with their mitochondria. We hadn’t directly exhausted their stem cells. Yet the ICE mice were suffering from a loss of body mass, mitochondria, and muscle strength and an increase in cataracts, arthritis, dementia, bone loss, and frailty.
All of the symptoms of aging—the conditions that push mice, like humans, farther toward the precipice of death—were being caused not by mutation but by the epigenetic changes that come as a result of DNA damage signals.
We hadn’t given the mice all of those ailments. We had given them aging.
And if you can give something, you can take it away.
Like the gnarled hands of giant zombies breaking free of the rocky soil, the ancient bristlecone pine trees of California’s White Mountains strike haunting silhouettes against the dewy morning sun.
The oldest of these trees have been here since before the pyramids of Egypt were built, before the construction of Stonehenge, and before the last of the woolly mammoths left our world. They have shared this planet with Moses, Jesus, Muhammad, and the first Buddha. Standing some two miles above sea level, adding fractions of a millimeter of growth to their twisted trunks each year, defying lightning storms and periodic droughts, they are the epitome of perseverance.
It’s easy to stand in wonder of these great and ancient things. It’s easy to be swept away by their might and majesty. It’s easy to simply stare at them in awe. But there’s another way to view these antediluvian patriarchs—a harder way, but a way in which we should seek to view every living thing on this planet: as our teachers.
Bristlecones are, after all, our eukaryotic cousins. About half of their genes are close relatives of ours.
Yet they do not age.
Oh, they add years to their lives—thousands upon thousands of them, marked by the nearly microscopic rings hidden in their dense heartwood, which also record in their size, shape, and chemical composition climate events long past, as when the eruption of Krakatoa sent a cloud of ash around the globe in 1883, leaving a fuzzy ring of growth in 1884 and 1885, barely a centimeter from the outer ring of bark that marks our current time.32
Yet even over the course of many thousands of years, their cells do not appear to have undergone any decline in function. Scientists call this “negligible senescence.” Indeed, when a team from the Institute of Forest Genetics went looking for signs of cellular aging—studying bristlecones from 23 to 4,713 years old—they came up empty-handed. Between young and old trees, their 2001 study found, there were no meaningful differences in the chemical transportation systems, in the rate of shoot growth, in the quality of the pollen they produced, in the size of their seeds, or in the way those seeds germinated.33
The researchers also looked for deleterious mutations—the sorts of which many scientists at the time expected to be a primary cause of aging. They found none.34 I expect that if they were to look for epigenetic changes, they would similarly come up empty-handed.
Bristlecones are outliers in the biological world, but they are not unique in their defiance of aging. The freshwater polyp known as Hydra vulgaris has also evolved to defy senescence. Under the right conditions, these tiny cnidarians have demonstrated a remarkable refusal to age. In the wild they might live for only a few months, subject to the powers of predation, disease, and desiccation. But in labs around the world they have been kept alive for upward of 40 years—with no signs that they’ll stop there—and indicators of health don’t differ significantly between the very young and the very old.
A couple of species of jellyfish can completely regenerate from adult body parts, earning them the nickname “immortal jellies.” Only the elegant moon jelly Aurelia aurita from the US West Coast and the centimeter-long Turritopsis dohrnii from the Mediterranean are currently known to regenerate, but I’m guessing the majority of jellies do. We just need to look. If you separate one of these amazing animals into single cells, the cells jostle around until they form clumps that then assemble back into a complete organism, like the T-1000 cyborg in Terminator 2, most likely resetting their aging clock.
Of course, we humans don’t want to be mashed into single cells to be immortal. What use is reassembling or spawning if you have no recollection of your present life? We may as well be reincarnated.
What matters is what these biological equivalents of F. Scott Fitzgerald’s backward-aging Benjamin Button teach us: that cellular age can be fully reset, something I’m convinced we will be able to do one day without losing our wisdom, our memories, or our souls.
Though it’s not immortal, the Greenland shark Somniosus microcephalus is still an impressive animal and far more closely related to us. About the size of a great white, it does not even reach sexual maturity until it is 150 years old. Researchers believe the Arctic Ocean could be home to Greenland sharks that were born before Columbus got lost in the New World. Radiocarbon dating estimated that one very large individual may have lived more than 510 years, at least up until it was caught by scientists so they could measure its age. Whether this shark’s cells undergo aging is an open scientific question; very few biologists had so much as looked at S. microcephalus until the past few years. At the very least, this longest-living vertebrate undergoes the process of aging very, very slowly.
Evolutionarily speaking, all of these life-forms are closer to us than yeast, and just think of what we’ve learned about human aging from that tiny fungus. But it is certainly forgivable to consider the distances between pine trees, hydrozoans, cartilaginous fish, and mammals like ourselves on the enormous tree of life and say, “No, these things are just too different.”
What, then, of another mammal? A warm-blooded, milk-producing, live-birth-giving cousin?
Back in 2007, aboriginal hunters in Alaska caught a bowhead whale that, when butchered, was found to have the head of an old harpoon embedded in its blubber. The weapon, historians would later determine, had been manufactured in the late 1800s, and they estimated the whale’s age at about 130. That discovery sparked a new scientific interest in Balaena mysticetus, and later research, employing an age-determining method that measures the levels of aspartic acid in the lens of a whale’s eye, estimated that one bowhead was 211 years old when it was killed by native whalers.
That bowheads have been selected for exceptional longevity among mammals should perhaps not be surprising. They have few predators and can afford to build a long-lived body and breed slowly. Most likely they maintain their survival program on high alert, repairing cells while maintaining a stable epigenome, thereby making sure the symphony of the cells plays on for centuries.
Can these long-lived species teach us how to live healthier and for longer?
In terms of their looks and habitats, pine trees, jellyfish, and whales are certainly very different from humans. But in other ways, we’re very similar. Consider the bowheads. Like us, they are complex, social, communicative, and conscious mammals. We share 12,787 known genes, including some interesting variants in a gene known as FOXO3. Also known as DAF-16, this gene was first identified as a longevity gene in roundworms by University of California at San Francisco researcher Cynthia Kenyon. She found it to be essential for defects in the insulin hormone pathway to double worm lifespan. Playing an integral role in the survival circuit, DAF-16 encodes a small transcription factor protein that latches onto the DNA sequence TTGTTTAC and works with sirtuins to increase cellular survival when times are tough.35
In mammals, there are four DAF-16 genes, called FOXO1, FOXO3, FOXO4, and FOXO6. If you suspect that we scientists sometimes intentionally complicate matters, you’d be right, but not in this case. Genes in the same “gene family” have ended up with different names because they were named before DNA sequences were easily deciphered. It’s similar to the not uncommon situation in which people have their genome analyzed and learn they have a sibling on the other side of town.36 DAF-16 is an acronym for dauer larvae formation. In German, “dauer” means “long lasting,” and this is actually relevant to this story. Turns out, worms become dauer when they are starved or crowded, hunkering down until times improve. Mutations that activate DAF-16 extend lifespan by turning on the worm defense program even when times are good.
I first encountered FOXO/DAF-16 in yeast, where it is known as MSN2, which stands for “multicopy suppressor of SNF1 (AMPK) epigenetic regulator.” Like DAF-16, MSN2’s job in yeast is to turn on genes that push cells away from cell death and toward stress resistance.37 We discovered that when calories are restricted MSN2 extends yeast lifespan by turning up genes that recycle NAD, thereby giving the sirtuins a boost.38
Hidden within the sometimes byzantine way scientists talk about science are several repeating themes: low energy sensors (SNF1/AMPK), transcription factors (MSN2/DAF-16/FOXO), NAD and sirtuins, stress resistance, and longevity. This is no coincidence—these are all key parts of the ancient survival circuit.
But what about FOXO genes in humans? Certain variants called FOXO3 have been found in human communities in which people are known to enjoy both longer lifespans and healthspans, such as the people of China’s Red River Basin.39 These FOXO3 variants likely turn on the body’s defenses against diseases and aging, not just when times are tough but throughout life. If you’ve had your genome analyzed, you can check if you have any of the known variations of FOXO3 that are associated with a long life.40 For example, having a C instead of a T variant at position rs2764264 is associated with longer life. Two of our children, Alex and Natalie, inherited two Cs at this position, one from Sandra and one from me, so all other genes being equal, and as long as they don’t live terribly negative lifestyles, they should have greater odds of reaching age 95 than I do, with my one C and one T, and substantially greater than someone with two Ts.
It’s worth pausing to consider how remarkable it is that we find essentially the same longevity genes in every organism on the planet: trees, yeast, worms, whales, and humans. All living creatures come from the same place in primordium that we do. When we look through a microscope, we’re all made of the same stuff. We all share the survival circuit, a protective cellular network that helps us when times are tough. This same network is our downfall. Severe types of damage, such as broken strands of DNA, cannot be avoided. They overwork the survival circuit and change cellular identity. We’re all subject to epigenetic noise that should, under the Information Theory of Aging, cause aging.
Yet different organisms age at very different rates. And sometimes, it appears, they do not age at all. What allows a whale to keep the survival circuit on without disrupting the epigenetic symphony? If the piano player’s skills are lost, how is it possible for a jellyfish to restore her ability?
These are the questions that have been guiding my thoughts as I have considered where our research is headed. What might seem like fanciful ideas, or concepts straight out of science fiction, are firmly rooted in research. Moreover, they are supported by the knowledge that some of our close relatives have figured out a workaround to aging.
And if they can, we can, too.
Before most people could even fathom the idea of mapping our genome, before we had the technology to map a cell’s entire epigenome and understand how it bundles DNA to turn genes on and off, the developmental biologist Conrad Waddington was already thinking deeper.
In 1957, the professor of genetics, from the University of Edinburgh, was trying to understand how an early embryo could possibly be transformed from a collection of undifferentiated cells—each one exactly like the next and with the exact same DNA—to the thousands of different cell types in the human body. Perhaps not coincidentally, Waddington’s ponderings came in the dawning years of the digital revolution, at the same time that Grace Hopper, the mother of computer programming, was laying the foundation for the first widely used computer language, COBOL. In essence, what Waddington was seeking to ascertain was how cells, all running on the same code, could possibly produce different programs.
There had to be something more than genetics at play: a program that controlled the code.
Waddington conceived of an “epigenetic landscape,” a three-dimensional relief map that represents the dynamic world in which our genes exist. More than half a century later, Waddington’s landscape remains a useful metaphor to understand why we age.
On the Waddington map, an embryonic stem cell is represented by a marble at the top of a mountain peak. During embryonic development, the marble rolls down the hill and comes to rest in one of hundreds of different valleys, each representing a different possible cell type in the body. This is called “differentiation.” The epigenome guides the marbles, but it also acts as gravity after the cells come to rest, ensuring that they don’t move back up the slope or hop over into another valley.
The final resting place is known as the cell’s “fate.” We used to think this was a one-way street, an irreversible path. But in biology there is no such thing as fate. In the last decade, we’ve learned that the marbles in the Waddington landscape aren’t fixed; they have a terrible tendency to move around over time.
At the molecular level, what’s really going on as the marble rolls down the slope is that different genes are being switched on and off, guided by transcription factors, sirtuins and other enzymes such as DNA methyltransferases (DNMTs) and histone methyltransferases (HMTs), which mark the DNA and its packing proteins with chemical tags that instruct the cell and its descendants to behave in a certain way.
What’s not generally appreciated, even in scientific circles, is how important the stability of this information is for our long-term health. You see, epigenetics was long the purview of scientists who study the very beginnings of life, not folks like me who are studying the other end of things.
Once a marble has settled in Waddington’s landscape, it tends to stay there. If all goes well with fertilization, the embryo develops into a fetus, then a baby, then a toddler, then a teenager, then an adult. Things tend to go well in our youth. But the clock is ticking.
Every time there’s a radical adjustment to the epigenome, say, after DNA damage from the sun or an X-ray, the marbles are jostled—envision a small earthquake that ever so slightly changes the map. Over time, with repeated earthquakes and erosion of the mountains, the marbles are moved up the sides of the slope, toward a new valley. A cell’s identity changes. A skin cell starts behaving differently, turning on genes that were shut off in the womb and were meant to stay off. Now it is 90 percent a skin cell and 10 percent other cell types, all mixed up, with properties of neurons and kidney cells. The cell becomes inept at the things skin cells must do, such as making hair, keeping the skin supple, and healing when injured.
In my lab we say the cell has ex-differentiated.
Each cell is succumbing to epigenetic noise. The tissue made up of thousands of cells is becoming a melange, a medley, a miscellaneous set of cells.
As you’ll recall, the epigenome is inherently unstable because it is analog information—based on an infinite number of possible values—and thus it’s difficult to prevent the accumulation of noise and nearly impossible to duplicate without some information loss. The earthquakes are a fact of life. The landscape is always changing.
If the epigenome had evolved to be digital rather than analog, the valley walls would be the equivalent of 100 miles high and vertical, and gravity would be superstrong, so the marbles could never jump over into a new valley. Cells would never lose their identity. If we were built this way, we could be healthy for thousands of years, perhaps longer.
But we are not built this way. Evolution shapes both genomes and epigenomes only enough to ensure sufficient survival to ensure replacement—and perhaps, if we are lucky, just a little bit more—but not immortality. So our valley walls are only slightly sloped, and gravity isn’t that strong. A whale that lives two hundred years has probably evolved steeper valley walls and its cells maintain their identity for twice as long as ours do. Yet even whales don’t live forever.
I believe the blame lies with M. superstes and the survival circuit. The repeated shuffling of sirtuins and other epigenetic factors away from genes to sites of broken DNA, then back again, while helpful in the short term, is ultimately what causes us to age. Over time, the wrong genes come on at the wrong time and in the wrong place.
As we saw in the ICE mice, when you disrupt the epigenome by forcing it to deal with DNA breaks, you introduce noise, leading to an erosion of the epigenetic landscape. The mice’s bodies turned into chimeras of misguided, malfunctioning cells.
THE CHANGING LANDSCAPE OF OUR LIVES. The Waddington landscape is a metaphor for how cells find their identity. Embryonic cells, often depicted as marbles, roll downhill and land in the right valley that dictates their identity. As we age, threats to survival, such as broken DNA, activate the survival circuit and rejigger the epigenome in small ways. Over time, cells progressively move towards adjacent valleys and lose their original identity, eventually transforming into zombielike senescent cells in old tissues.
That’s aging. This loss of information is what leads each of us into a world of heart disease, cancer, pain, frailty, and death.
If the loss of analog information is the singular reason why we age, is there anything we can do about it? Can we stabilize the marbles, keeping the valley walls high and the gravity strong?
Yes. I can say with confidence that there is.
Regular exercise “is a commitment,” says Benjamin Levine, a professor at the University of Texas. “But I tell people to think of exercise as part of personal hygiene, like brushing their teeth. It should be something we do as a matter of course to keep ourselves healthy.”41
I’m sure he’s right. Most people would exercise a lot more if going to the gym were as easy as brushing their teeth.
Perhaps one day it will be. Experiments in my lab indicate it is possible.
“David, we’ve got a problem,” a postdoctoral researcher named Michael Bonkowski told me one morning in the fall of 2017 when I arrived at the lab.
That’s seldom a good way to start the day.
“Okay,” I said, taking a deep breath and preparing for the worst. “What is it?”
“The mice,” Bonkowski said. “They won’t stop running.”
The mice he was talking about were 20 months old. That’s roughly the equivalent of a 65-year-old human. We had been feeding them a molecule intended to boost the levels of NAD, which we believed would increase the activity of sirtuins. If the mice were developing a running addiction, that would be a very good sign.
“But how can that be a problem?” I said. “That’s great news!”
“Well,” he said, “it would be if not for the fact that they’ve broken our treadmill.”
As it turned out, the treadmill tracking program had been set up to record a mouse running for only up to three kilometers. Once the old mice got to that point, the treadmill shut down. “We’re going to have to start the experiment again,” Bonkowski said.
It took a few moments for that to sink in.
A thousand meters is a good, long run for a mouse. Two thousand meters—five times around a standard running track—would be a substantial run for a young mouse.
But there’s a reason why the program was set to three kilometers. Mice simply don’t run that far. Yet these elderly mice were running ultramarathons.
Why? One of our key findings, in a study we published in 2018,42 was that when treated with an NAD-boosting molecule that activated the SIRT1 enzyme, the elderly mice’s endothelial cells, which line the blood vessels, were pushing their way into areas of the muscle that weren’t getting very much blood flow. New tiny blood vessels, capillaries, were formed, supplying badly needed oxygen, removing lactic acid and toxic metabolites from muscles, and reversing one of the most significant causes of frailty in mice and in humans. That was how these old mice suddenly became such mighty marathoners.
Because the sirtuins had been activated, the mice’s epigenomes were becoming more stable. The valley walls were growing higher. Gravity was growing stronger. And Waddington’s marbles were being pushed back to where they belonged. The lining of the capillaries was responding as if the mice were exercised. It was an exercise mimetic, the first of its kind, and a sure sign that some aspects of age reversal are possible.
We still don’t know everything about why this happens. We don’t know what sorts of molecules will work best for activating sirtuins or in what doses. Hundreds of different NAD precursors have been synthesized, and there are clinical trials in progress to answer that question and more.
But that doesn’t mean we need to wait to take advantage of all that we’ve learned about engaging the epigenetic survival circuit and living longer and healthier lives. We don’t need to wait to take advantage of the Information Theory of Aging.
There are steps we can take right now to live much longer and much healthier lives. There are things we can do to slow, stop, and even reverse aspects of aging.
But before we talk about what steps we might take to combat aging, before I can explain the science-backed interventions that have the greatest promise for fundamentally changing the way we think about growing old, before we even begin to talk about the treatments and therapies that will be game changers for our species, we need to answer one very important question:
Should we?