If you ask someone to tell you the cause of aging, their answer will say a great deal about their journey into the aging field. The most common answer, “isn’t it telomere shortening?,” means someone first heard of aging theories around 2009 when the Nobel Prize for the discovery of telomere function was awarded. Focusing on aging-ending strategies or longevity escape velocity is a clear sign of having watched Aubrey de Grey’s TED talk. More recently, mentions of NAD or sirtuins suggest (like me!) their first appetite for the aging field came from reading David Sinclair’s Lifespan. Fresh longevity enthusiasts these days mostly get excited about bioelectricity after hearing Michael Levin on a podcast or two.

This is, of course, not ideal; it would be much better to gain an understanding of aging through surveying the ideas that have permeated the field for decades and choosing the one that feels most compelling. But, for all my years of searching, I have yet to find an accessible entry point into this fascinating world1. Why is this such a loss?

In the words of psychologist Kurt Lewin, “There is nothing as practical as a good theory.” It is difficult to intervene in the aging process without a direction, and it is tough to find a direction without a map. Hence, this article will draw one.

Two key axes emerge when looking at aging theories.

1. Does it primarily focus on evolution as an explanatory force for aging, or entropy?

2. Is it centered around one biological mechanism, or many?

If you take sixteen aging theories2 and map them on these axes, you get a plot that looks like the following:

Below, I’ll break down each of the quadrants, what they mean, and my best good-faith interpretation of what each listed theory claims3.

Table of Contents

1. Introduction

2. The Theories

   1. Entropy Drives One Molecular Cause

      1. Free Radical Theory

      2. Mutations

      3. Telomere Attrition

      4. Bioelectricity

      5. Information Theory

   2. Evolution Drives One Molecular Cause

      1. Hyperfunction Theory

      2. Pathogen Control

   3. Entropy Drives Many Causes

      1. Wear-and-Tear

      2. SENS

      3. Hallmarks of Aging

      4. Fedichev’s Physics-Based Aging

      5. Saturating Removal

   4. Evolution Drives Many Causes

      1. Medawar’s Mutation Accumulation

      2. Antagonistic Pleiotropy

      3. Software Design Flaw

      4. Disposable Soma

3. Conclusion

The Theories

Entropy Drives One Molecular Cause

This quadrant is a popular place for aging theories to go because it is fairly straightforward to understand. Aging is primarily caused by [biological phenomena of interest]; if one fixes [biological phenomena of interest] then that should slow the aging process significantly. The simplicity of these arguments, to their great credit, allows for relative ease in falsifying their core premise. It is comparatively much easier to experimentally test the relative importance of telomeres in aging compared to debating the applicability of, say, physics-based models. Consequently, I’ll spend some time writing on the various efforts scientists have made over the decades to prove out these theories.

Free Radical

The free radical theory of aging comes from the early Atomic Age, when scientists noticed the effects of radiation exposure look surprisingly similar (in many ways) to aging. The predominant thinking on aging at the time was influenced by fruit fly scientists, who observed that metabolic rate was inversely related to lifespan. Denham Harman combined these two ideas by hypothesizing that normal metabolism creates free radicals that lead to aging.

For a long period of time, this theory saw a great deal of popularity, especially when SOD2 and catalase were discovered as enzymes evolved specifically to deal with free radical damage. This set the field up nicely - if the free radical theory of aging were true, then more SOD2/catalase/antioxidants should lead to significant extension of lifespan.

Unfortunately, this did not pan out. SOD2 overexpression does not extend lifespan in mice, halving SOD2 does not decrease lifespan in mice, and taking antioxidant supplements leads to increased death in people. Damage from mitochondrial production of hydrogen peroxide does not even reach the nucleus. These findings, when paired with discoveries about the importance of free radicals in normal biology, have led to the aging field to look elsewhere.

In more recent years, the theory has seen a revival in the form of an increased emphasis on the mitochondria. As mentioned earlier, overexpressing antioxidant enzymes does not increase lifespan - but moving them to the mitochondria does. Mitochondrial function is implicated in the development of sarcopenia and other age-related diseases in humans. If mitochondria are important to aging, then the massive amounts of damaging free radicals they produce could be an obvious culprit.

Mutations

There is a rare disease, Werner’s syndrome, wherein individuals with a mutation in a DNA repair protein (WRN) look remarkably like they are aging about 25% faster than normal.

Despite the striking similarity, critics will note there are many ways in which Werner’s syndrome disease progression is quite different from aging. And yet there are 5-10 other genetic diseases with an uncanny resemblance to accelerated aging…. and they all come from having inherited mutations in DNA repair genes. This is quite unlikely to happen by chance, and leads to the immediate hypothesis that DNA repair is related to aging progression. The question is how, and one of the first possible answers was very direct - it’s the direct DNA mutations themselves that are causing the aging process. And there is a very strong relationship between mutational load in aging, both within- and across- species! See below:

But correlation is not causation, and immediately the same test arises here as it did with the free radical theory of aging - do increased mutations always lead to faster aging, and can you decrease mutations to slow aging? The jury is still out on the latter point, but a 2021 paper poured cold water on the former by finding a population of individuals that has an order of magnitude higher mutational burden with no detectable increase in aging rate.

The field has since shifted more towards focusing on cellular population dynamics - maybe mutations impacting individual cells are not the problem, but rather how they skew a stem cell niche. The bone marrow, for example, dramatically shrinks in terms of cellular diversity over time. The importance of this shrinking is not yet fully known.

Telomere Attrition

Telomeres are pieces of DNA located at the ends of chromosomes. Because of the way that DNA replication occurs, telomeres become slightly shorter each time a cell divides. After enough cellular divisions, the full telomeric region is gone and the cell no longer divides. The late Leonard Hayflick, who discovered that cells cannot divide forever (without making the connection to telomeres at the time!), immediately noted in the same manuscript the relevance for human aging:

The core idea is elegant. If telomere shortening leads to cells ceasing to grow, then this is a clear way through which aging could manifest in people. Unfortunately, the connection between telomeres and aging became shakier over time. Telomere length turns out to not correlate with lifespan in mouse strains, and murine telomere lengths can be far longer than those from humans. Even worse, there’s a strong cancer connection (making this somewhat an evolutionary theory) - short telomeres are a brute-force way to prevent cells from becoming malignant. Upregulating a telomere-extending enzyme (telomerase) leads to higher cellular susceptibility to tumorigenic transformation. The explanation for this phenomenon is pretty straightforward - stem cells that are designed to renew have mechanisms through which they extend their telomeres already; adding more telomeres to non-stem-cells only leads to an increased risk of cancer.

There is, nonetheless, still an important connection between telomeres and aging. Mice born specifically from stem cells with long telomeres do live longer. There are papers suggesting that overexpression of telomerase can lead to lifespan extension, if administered in a very particular way. Perhaps telomeres are important not because they cap chromosomes, but for another reason altogether - being difficult to repair. Multiple laboratories have shown that DNA damage at telomeres persists across cell generations and leads to a prolonged immune response. Telomerase itself has roles beyond telomere extension, and its reactivation leads to beneficial results in mouse models of neurodegeneration. As a result, the relationship between telomeres and aging continues to be a hot area of research.

Bioelectricity

This is a planarian.

Planaria are incredibly good at regenerating themselves. If you cut them in any direction, they will naturally regrow. In the process of regrowth, they will fully rejuvenate not only the part of their body that they regrew, but the entire body plan. This makes them largely immortal, while they nevertheless accumulate mutations and other forms of cellular damage. A radical interpretation of this finding is that aging has little to do with damage on a cellular level, and everything to do with how an organism is organized. This is the view of well-known scientist Michael Levin, and his outlook on aging is that it’s the breakdown of the overall system (which is regulated through bioelectricity) that leads to aging. The primary outstanding question is the extent to which these findings are relevant to humans.

Information Theory

Harvard professor David Sinclair’s Information Theory is complicated, but it goes something like this:

1. Damage happens to DNA due to metabolic and/or external causes. This damage needs to be resolved by the cell.

2. The cell fixes the DNA using proteins that would normally be regulating DNA elsewhere. There’s a particular focus on sirtuins, which are unique because they are evolutionarily ancient & closely related to metabolic signaling. Sirtuins use NAD+ to carry out their function, so there’s a deep emphasis on NAD+ as well.

3. Since these sirtuins have abandoned their original location on the genome, those areas cease to be regulated properly. This leads to the cell gradually drifting epigenetically, which manifests as aging.

There’s a second corollary that there’s a backup copy of the original epigenetic state hidden somewhere that can be uncovered using epigenetic reprogramming. Epigenetic reprogramming has been covered in rich detail elsewhere, but it entails expressing pluripotency-inducing genes just long enough to get rejuvenation associated with youth without accidentally inducing teratomas. Proponents of the Information Theory of Aging generally tended to be excited about slowing aging via NAD+ restoration and sirtuin activation, but have over time drifted more towards reversing aging with partial reprogramming techniques.

Evolution Drives One Molecular Cause

This is the least populated quadrant, and it is a particularly interesting one because its two entries share the least in common relative to any of the other quadrants. In many ways, this quadrant being true would be the best case for translational potential in aging - if evolution was driving aging through a very small number of biological mechanisms, then we should have a relatively easy time undoing those mechanisms.

Hyperfunction

During early life, there is one central theme: Growth. Organisms need to expand their size many times over, with many of them developing entirely novel capabilities after birth. This growth burst is inherently destabilizing, with high metabolism driving free radical damage, telomere attrition, extracellular disruption, and more. Towards the tail end of the growth phase, the reproductive period starts and then growth programs need to be quickly shut down.

Or do they? It depends on the species, but for many organisms the reproductive period is not very long before death from illness or predation. As a result, there is not a great deal of evolutionary incentive for tuning growth pathways to slow down. Many of them will continue on, gradually leading to dysfunction. This core concept - that continued growth programs lead to aging - is hyperfunction theory.

Hyperfunction theory enjoys a great deal of empirical support. Basic research in worms has shown that growth inhibition can increase lifespan by several-fold. Caloric restriction and rapamycin, both interventions that directly target growth, lead to the largest robust increases in both median lifespan and maximum lifespan recorded4. These interventions also generally work across organisms, from C. elegans all the way to primates (though the jury is still out on humans). Newer interventions that have successfully extended lifespan in mouse models, such as IL-11 antibody inhibition, also rely on inhibition of growth pathways. The biggest question for hyperfunction theory is the specific mechanism - yes, growth accelerates aging, but how?

Pathogen Control

A heated scientific debate in the aging world focused on whether or not there was an actual aging program, i.e. whether or not aging was deliberately controlled. There is evidence in some species that such a program exists - take a look at the queen bee, for instance.

Some queen bees can live many times as long as their worker counterparts, while sharing the same DNA. Similarly, salmon age extremely quickly after spawning and die quickly. However, in most species it has been difficult to identify evidence for such a program, especially on a molecular level.

Peter Lidsky’s pathogen control theory is a callback to the era of programmed aging theories by asserting that evolution actively selected for senescence. The basic premise: aging is a way to prevent chronic diseases from becoming fully embedded in a population. If a population grows old enough, pathogens associated with the long-lived population gain enough time to adapt to the organism and its defenses. The theory strongly emphasizes both immune and growth pathways as being mediators of this programmed aging process.

Entropy Drives Many Causes

This quadrant is one of abstractions, where aging is treated least like a biological phenomenon and most like a physical one. Three of the five theories here were conceived of by physicists and engineers. Metaphors paralleling the aging body to decaying cars are common. The predominant questions being asked are about what and how; why is much more rare. The core intuition that the aging body is akin to a declining mechanical system is ancient, yet this is also the area in modern theoretical aging biology where the most exciting work is happening.

Wear-and-tear

“We do not suddenly fall on death, but advance towards it by slight degrees; we die every day. For every day a little of our life is taken from us; even when we are growing, our life is on the wane….” wrote Seneca, over two thousand years ago. The core intuition that aging is a slow breakdown happening over one’s lifetime - also known as wear-and-tear theory - is, in some ways, the base case assumption for aging. Almost all of the other theories in this article are a response to wear-and-tear, either as an explanation for how it happens (e.g. free radicals), or as a rebuttal explaining why it is wrong. There are two main challenges with wear-and-tear as a framework for aging:

1. We have intrinsic repair capacity. Unlike a car, we are able to self-repair damage we incur. Why is aging different?

2. Why do species age at different rates, and what drives this differential aging?

Despite the relatively ancient nature of these two questions, they both remain conceptually difficult to wrangle to this very day. If even one of these two had a clear solution, the question of why do we age? would be much easier to answer.

SENS

I previously alluded to the idea of using car breakdown as an analogy to aging.

Following that line of thinking for a moment, we do not often think about why cars break down (unless you happen to work for a car manufacturer, in which case I imagine you think about it a great deal). Instead, we solve car breakdowns by identifying the parts that are breaking and fixing them. Aubrey de Grey’s Strategies for engineered negligible senescence (SENS) focuses on exactly this idea by identifying seven types of molecular damage that we undergo and hypothesizing ways to repair them before they become pathological. The primary innovation of SENS was in bringing an engineering mentality into the aging field, making it explicitly clear that the goal of ending aging did not require a perfect underlying theory. This core concept continues to be incredibly influential5, with contemporary conversations often focusing more on how to solve aging rather than on understanding it.

Hallmarks of Aging

The Hallmarks of Aging framework is currently the most popular framing of aging (especially in academic circles). Below is its principal figure:

It is an outline of twelve (originally nine) major changes that happen over time. It does have meaningful theory incorporated into it (you’ll observe the “primary,” “antagonistic,” and “integrative” inner circles), but that tends to be much less discussed than the hallmarks themselves. The Hallmarks framework appears similar to SENS (described above), but in many ways it is the direct opposite - it emphasizes biology rather than engineering, and specifically avoids call-to-arms language in favor of a holistic and descriptive view. The positive view of the Hallmarks is that it recognizes all of the biology that encompasses aging; the critical view6 is that it fails to prioritize the biology that has the highest probability of moving the needle.

Fedichev’s Physics-Based Aging

It is conventional to associate aging with progressive dysfunction and disease, to the extent that some definitions of aging inherently incorporate illness as a component. But many of us know elderly individuals who were remarkably healthy until they died of a sudden illness; likewise there are many individuals who are chronically sick for many decades before death finally takes them7. Peter Fedichev, a physicist by training, emphasizes modeling the relationship between aging, frailty, and death. It involves many equations, but visually it looks like this:

According to this view, individuals have an ability to repair damage from stress and illness (also referred to as resilience). The stressor itself is a fundamentally different thing from resilience. Eventually, resilience becomes low enough that any stress leads to system collapse and subsequent death.

The conceptual achievement of this theory is that it neatly separates aging pathologies - there are those that contribute to stress (inflammation, bacterial infections, etc.) and those that contribute to resilience (a mysterious purer aging process). The former are primarily related to healthspan and can be addressed via engineering strategies. The latter determine maximum lifespan and are much more difficult to tackle, as they’re a proxy for entropy acting on the human body. This distinction should allow for improved clinical targeting of aging therapies.

Saturating Removal

The Saturating Removal (most associated with Uri Alon) view of aging is also based on physics and emphasizes resilience.

There are three processes that are occurring at once in your body:

1. Damage happens. This damage increases over time in a linear fashion.

2. You have processes that clean up the damage that is happening.

3. Some random biological noise happens too.

Over time, damage accumulation saturates available repair mechanisms. When damage hits a certain level, you die. To extend lifespan, organisms have two options: (A) either decrease the amount of damage, or (B) repair better.

This may seem somewhat simplistic, but the basic model has yielded some fairly important insights. One key takeaway is that long-lived species are separated from short-lived species entirely by producing less damage, as opposed to by repairing more effectively. This is somewhat surprising, given literature evidence suggesting long-lived species have more capable damage repair mechanisms. Saturating Removal theory has also provided nuance as to how we age over time - its models, for example, suggest that aging in people aged 20 to 60 is quite different from in those aged 85 and older.

Evolution Drives Many Causes

This quadrant has an incredibly deep legacy, with many authors over the span of centuries generating evolutionary reasons for the existence of aging. Most of these were conceived of before we had even basic molecular biology tools, yet they generally hold up remarkably well. These are theories in the grandest sense of the term - seeking to create grand unifying explanations for why aging exists in the form that it does. They are also quite closely related, each riffing off of one key idea: the selection shadow.

Assuming constant birth rates and death rates, there will definitionally be a greater percentage of the population at earlier years of life than later years of life.

What are the implications? It depends on the theory.

Medawar’s Mutation Accumulation

Medawar’s Mutation Accumulation (somewhat confusingly, ‘mutations’ here refers to genetic variants, not mutations accumulated over one’s lifetime) theory is the neutral case. Evolution cares most about individuals at the earliest stage of their life, because (A) there are a lot of them (they have yet to be eaten) and (B) they are reproductively fit. Therefore, most genetic adaptations exist for a younger generation. Evolution does not optimize for older individuals, because there are fewer of them. As there is no evolutionary force preventing late-acting deleterious traits from piling up, they exist unbothered in the genome and contribute to aging.

Antagonistic Pleiotropy

Antagonistic pleiotropy is the more active variant of Medawar’s mutation accumulation idea. Not only does evolution not optimize for older individuals, it will happily optimize for traits that are useful early in life that are maladaptive later on. Amyloid beta is an example - it’s hypothesized to be a part of the innate immune system and to help kill bacteria, fungi, and viruses. For humans throughout most of our evolutionary lifespan, infectious disease was a major lifespan limiter; Alzheimer’s disease was not. Now, with human lifespans longer than ever, these early-life adaptive traits become lifespan-limiting. If you specifically focus on how growth signaling is a major form of antagonistic pleiotropy, you get hyperfunction theory (see above).

Software Design Flaw

If you take the selection shadow idea and apply seventy years of biological knowledge and a computer scientist lens, you get João Pedro de Magalhães’ Software Design Flaw view of aging. According to this view, aging is a run-on developmental program embedded in our genome and epigenome that was optimized for the first forty years of our lives. Deleterious regulatory networks (subroutines) that are fit for early life are not turned off later in life, and the result is progressive organismal dysfunction. What are promising therapeutic avenues under this framework? ‘Rebooting’ the software through epigenetic reprogramming would be most effective; reversing damage accumulated over the lifetime would do too little.

Disposable Soma

Thomas Kirkwood’s Disposable Soma theory is, in some ways, an economist’s view on the selection shadow. Organisms can only do so many things, and there are inherent resource tradeoffs between growth, reproduction, and maintenance. The more an organism invests into reproduction, the less maintenance it can sustain. There are some interesting studies that seem to reinforce this view, such as findings that male castration leads to a slowing of aging, reinforcing older ideas about eunuchs living longer, healthier lives. The most compelling read of the disposable soma theory is this: longevity does not come for free, and what you gain in repair you will lose somewhere else. Human aging is not akin to a car rusting, because cars need neither reproduce nor grow; all of the resources going into keeping a car afloat can be invested into maintenance (or, perhaps, a good enough warranty).

Conclusion

Which of these theories is most correct?

It’s hard to say. One way to measure ‘correctness’ is by investigating which theories yield the most promising longevity interventions. Two main efforts exist to interrogate this systematically - the National Institute on Aging’s Interventions Testing Program (ITP) and the Longevity Escape Velocity Foundation’s Robust Mouse Rejuvenation (RMR) Study.

The ITP is a program that has been ongoing for over two decades to test putative aging interventions in a reproducible, careful manner at three separate institutions. Their database of tested compounds and resulting lifespan effects can be found here. The main hits that came out of this program are rapamycin and acarbose, which, as key upstream metabolic regulators, perhaps strengthen the hyperfunction framework more than anything else.

Aubrey de Grey’s Robust Mouse Rejuvenation program is an effort to test multiple interventions at a time, validating the possibility of synergistic effects on aging through repairing several aging hallmarks at once (in many ways a test of the SENS framework). Their initial experiment was quite comprehensive - investigating hematopoietic stem cells, telomere extension, rapamycin, and a senolytic. The result of this experiment?

No combination looks a whole lot better than rapamycin by itself.

The alternative approach is to embark on a single-pronged quest to empirically prove one specific framework correct, a la Barry Marshall swallowing a culture of H. pylori to quite definitively prove that gastritis can be caused by bacteria. The vehicle is often either a lab or startup8, with a few fairly well-known examples. The first of these was Geron Corporation, built on the premise that extending telomeres could prolong lifespan9. More recently, David Sinclair’s Life Biosciences has entered clinical trials with an explicit vision of restoring the epigenome, and Peter Fedichev’s Gero is finding ways to slow aging using physics-based models. None of these efforts have borne translational fruit yet, but many of them are still quite early in their journeys.

There is a possibility there is no correct theory of aging. Aging could be a terrible messy web, caused by too many factors to count, manifesting in too complex a manner to understand. But that in itself is a framework10, one with conclusions that are actionable - perhaps implying that the best solution to aging is to use a tool (AI) that can understand the complexity that mortal minds cannot.

Something that is certain: aging is a process that is happening to all of us, driving the probability of life-altering disease higher with every passing year. Through understanding and considering how it operates, we could slow aging down to save an enormous number of lives. Perhaps we could even reverse it…

…but only if you subscribe to certain theories of aging.

Subscribe now

Why? To some extent the answer to this question is unknowable2, but a wide range of explanations have been proposed, from obesity to microplastics to dietary changes. Many of these influence colorectal cancer risk by affecting the microbiome, shifting it away from an enrichment of symbiotic bacteria and towards a higher ratio of aggressive, pro-inflammatory microbes.

But how does changing the microbiome cause cancer? The answer, in many cases, is very direct.

Bacteria can break your DNA

The microbiome is not a static environment. It is an ecosystem containing hundreds of different bacterial species, many of which are directly competing with each other for resources. They are spread out throughout the entire GI tract, sitting above the epithelial cells that form your gut while being policed over by vigilant immune cells. In the same way that a forest ecosystem changes over time, the composition of your particular microbiome is never stagnant. The microbes are not only changing in proportion, but also in character - mutations enable some bacteria to be more efficient in scavenging resources, and others more capable at evading the immune system. Some bacteria have a more direct ability to dominate the microbiome niche: they create toxins that kill other bacteria and any immune cells that might try to stop them.

This brings us to pks+ E. coli. This particular strain of bacteria is among the most problematic, specifically because it creates a toxin that is designed to staple DNA strands together. Bear with me for a moment to discuss the biochemistry, because the mechanism is fascinating.

The toxin in question is colibactin. Below is the chemical structure - I’ve circled the most interesting bits:

It may not look like much. It’s a bit large and unwieldy compared to what most bacterially-derived toxins look like. But looking deeper, there are two key components:

1. On each end, there are chemical structures that are geometrically a bit similar to a nucleotide that you would see in DNA, except with extremely unstable and strained cyclopropane “warheads” (the triangular pieces).

2. In the middle, there’s a flexible linker with aromatic rings every few carbons. This looks a bit like a ladder, with two “steps.”

As it turns out, this structure slots very nicely into DNA. The bond ends up looking like this:

The mechanism for how this molecule causes damage is simple but beautiful. It slides into the minor groove of the DNA helix, while barely able to hold its unstable warhead together. Once it is settled, the strained chemical bonds on colibactin snap and rapidly react with the closest atoms in the vicinity - in this case, those on the DNA. The resulting bond is permanent.

The type of damage this causes is extremely hard for a cell to fix. It does not look anything like the type of damage cells are used to experiencing. The damage that colibactin induces is the equivalent of supergluing together two complementary strands of DNA. The cell is not able to use the DNA for anything - transcription or replication - because that requires the cell to unwind the DNA strands together, which it now cannot do. The best the cell can do is either ignore it, or cut out the region and copy in the correct sequence from the other chromosome (which is hopefully undamaged!). This type of repair has a high probability of inducing a mutation.

How does this tie back to early-onset colorectal cancer? It turns out the type of DNA damage that colibactin induces leaves a particular mutational signature - a DNA scar, if you will. And this signature (labeled SBS88) is highly enriched in colorectal cancers from young patients.

So what can we do about it?

Option 1: Reduce the amount of colibactin-producing bacteria

If the core problem is that we have problematic toxin-producing bacteria in the gut, then the most straightforward approach would be to eliminate them. Unfortunately, this is tricky. Antibiotics are not selective, and some of them can increase the proportions of pathogenic bacteria. On the flip side, drugs that down-regulate the immune system can be paradoxically helpful - E. coli thrive on oxygen-rich environments that immune cells bring - but come with tradeoffs. Probiotics are promising, as introducing symbiotic and commensal bacteria inherently decreases the proportion of problematic pks+ E. coli, but the only study to try this in an animal model introduced, ironically, another bacteria that produces colibactin3. The most promising study to date used putrescine in supplement fashion to decrease the growth of pks+ E. coli, with the general idea of inhibiting the growth of toxin-producing bacteria. In general, the idea of modulating the microbiome has a great deal of promise, but has struggled immensely in the clinic4.

Option 2: Detoxify colibactin

You may be wondering how pks+ E. coli themselves do not fall victim to colibactin, as, after all, they also have DNA that is vulnerable to damage. The answer is that they have evolved their own antidote: an enzyme that breaks the DNA-stapling warhead.

Could we create something similar? Or otherwise create something that would prevent bacteria from producing colibactin in the first place?

The answer is yes! And many of the techniques that researchers have used are quite clever. A group in the Netherlands showed that colibactin-producing bacteria need to stick to epithelial cells in the gut to cause harm. If you block the ability to stick by either knocking out a gene (see figure below), or by adding a drug, the amount of damage is dramatically reduced.

Another study used a boron-containing drug to mimic one of the substrates that bacteria use to create colibactin, which led to a large drop in production. A third paper directly took advantage of the pks+ E. coli detoxifying enzyme by fusing it to the cell surface of a different bacterial strain, generating the biological equivalent of a gut colibactin vacuum cleaner. If you could co-colonize your microbiome with bacteria that break down any colibactin outside of a cell, that should reduce the total gut toxin load significantly. It also points towards a synthetic biology utopian future where one could engineer microbes to detoxify many different toxins. It’s an exciting world of possibilities.

Option 3: Protect DNA directly

Colibactin is a very unique mutagen, but it does have some chemical cousins. One of the closest is mitomycin C, a chemotherapy drug. It looks like this:

It lacks the flat structural piece to bind into the minor groove, but it does resemble guanosine enough to associate with DNA. This, just like with colibactin, leads to the activation of the warhead (same triangular piece, though with a nitrogen this time around) with irreversible crosslinking of DNA. It works as a chemotherapy due to the high division rate in cancerous cells, though it comes with (predictable) negative side effects.

Interestingly, there are chemical compounds that may be able to block damage induced by crosslinking agents. A 2008 study tested administering amifostine prior to mitomycin C exposure, and showed a reduction of DNA damage back to baseline:

The impetus for this study came from wanting to find ways to alleviate DNA damage in patients with Fanconi anemia, a rare disease where the genes that can repair interstrand crosslinks are mutated. If you can protect DNA from agents like mitomycin C, the implications extend well beyond colibactin-producing bacteria.

So where are we now?

Oftentimes, epidemiological problems feel intractable to scientific solutions. Solving the primary driver of lung cancer (smoking) did not come from new drugs that broke the toxins in tobacco, but rather from public-focused anti-smoking campaigns.

In this case, science may actually have the solution. We know the driver of a large fraction of early-onset colorectal cancer cases, and we have scalable ways of alleviating it. We may be early (you’ll note nearly every paper I cited in this piece was released within the last three years), but the proposed interventions do not seem particularly sci-fi. Probiotic solutions, in particular, could move very quickly from being a proof-of-concept to something sitting on a store shelf (or a fridge, in this case).

The other big takeaway is in seeing the medical problem for its molecular drivers. Cardiovascular disease, like early-onset colorectal cancer, is affected by high-level lifestyle components like diet and exercise. And yet, understanding the importance of cholesterol led to the life-saving drug category of statins.

We should be able to intervene in the causes of cancer in a similar way.

Subscribe now

How do you live to be 90?
Eat well.
Exercise.
Sleep.
Avoid alcohol.

How do you live to be 100?
Have parents that lived to be 100.

This speaks to an underlying biological truth - extending lifespan is not just about optimizing healthspan. We likely all know someone who was extremely healthy for their age until they, within a short period of time, died from a surprise illness. The aging vector that ultimately kills you is largely separate from the one that keeps you free from being sick.

A Tale of Two Mice

The mice on the top panel are regular mice at 22 months of ago. The mice on the bottom panel have had all of their senescent cells1 cleared. You’ll observe how much healthier the mice in the bottom panel are, and on average they lived about 25% longer.

But if you look at when the last mouse in the bottom panel died vs. when the last mouse in the top panel died2?

There’s barely a difference:

The average lifespan is increased, likely because removing the senescent cells prevents the mice from getting diseases that would kill them early (such as cancer). But the inevitable march towards death is not slowed, and the luckiest mice from the untreated cohort and the luckiest mice from the treated cohort die at about the same time.

Are there longevity interventions that extend maximum lifespan significantly?

There aren’t many. The clear standout is caloric restriction:

Here, you can see very clearly the last caloric-restricted mouse is dying much later than the mouse eating ad libitum. Their rate of aging is slower, not just their rate of accumulating diseases.

The difference between healthspan and maximum lifespan is not always so clear. If you treat mice with acarbose and rapamycin at 16 months of age, you do see both maximum and median lifespan extension… but the maximum lifespan extension is only readily apparent in females.

Some interventions primarily target healthspan; others primarily maximum lifespan. Some are a combination of both. But these two aspects of aging are distinct.

Is this true in humans?

A Tale of Two Clocks

In 2013, Steve Horvath released his epigenetic clock paper demonstrating that you could train a machine learning model to predict age to within about three years (more on this here). This was a fascinating result, but a number of folks found the idea of being able to predict your age underwhelming - after all, everyone has a sense of how old they are3.

This led to the generation of second-generation epigenetic clocks, which were created to predict health as opposed to age. Two different approaches led to two very different clocks:

1. Morgan Levine’s PhenoAge, which was trained to predict health markers alone and then regressed onto age.

2. Ake Lu’s GrimAge, which was trained to predict time-to-mortality.

If health and mortality were the same, then the two clocks should be redundant as their predictions would be the same. The data show the opposite. Here is a plot showing their correlations:

Quantitative estimates vary, but the correlation here is mediocre, with an R of somewhere between .3 and .5. There is not perfect disentangelment here - a clock trained on mortality data will pick up a lot of healthspan markers, as humans generally do not live to their absolute maximum lifespan potential. In fact, GrimAge was initially built from a composite of healthspan markers that, together, predict mortality. Nevertheless, it’s interesting that these two measures diverge.

Human genetics studies show a similar pattern:

Genetic variants that are linked to extreme longevity are not the same ones that are linked to healthspan. In fact, some genetic variants are connected to improved lifespan but worsened healthspan.

A Tale of Two Worms

For people working on aging research in C. elegans, this concept is not new at all (excellent paper on this topic here). Consider the well-known daf-2 mutant, which nearly doubles worm lifespan:

The effect on worm movement, however, is mostly negative - even at young worm ages.

Not every maximum lifespan-benefiting intervention has this tradeoff. Caloric restriction, which I mentioned earlier as having a major effect on maximum lifespan, seems to generally improve health metrics too (though not if taken to an extreme). But it’s an important parameter to keep in mind.

So what’s the take?

I once made a poorly-received joke to the XPRIZE Healthspan4 competition organizers that I would be interested in proposing giving cocaine and steroids to the elderly in order to reverse their age. Given the therapeutic treatment window only lasts a year, this should be a clear winner!

The deep take is this: maximum lifespan and healthspan are separate; what is good for one may be useless or even detrimental for the other. Beware claims of reversing aging - depending on which aspect of aging is being measured, this may be a Faustian bargain trading health for time. For preclinical and clinical trials focused on aging, it will become increasingly important to ensure that we know whether a therapeutic is improving healthspan, lifespan, or both.

Further Reading

I kept this piece relatively light on aging theory. For folks interested in reading on this more deeply, I’d heavily recommend reading some of the work done by Peter Fedichev and Uri Alon. Physicists, for whichever reason, have picked up on the healthspan vs. maximum lifespan separation most acutely - perhaps it’s because they’re most attuned to the shape of the lifespan curves I shared in the beginning of the piece.

Fifteen years ago, I succeeded in my first negotiation. As a twelve year old child, I would spend nights during the summer with my parents cleaning middle and high schools. It wasn’t a glorious job, but with no knowledge of English it was one of the few that they could find and I had little else to do. At the end of the night, they would give me $10.

After a couple of weeks like this, I picked up on the fact that this was a poor deal for me - I had just helped them out for six to eight hours, putting my hourly wages at a depressing $1.25. When I pointed this out to my mother, she thought for a moment before agreeing to double the amount. I had managed to negotiate a 100% wage increase - an auspicious beginning.

~~

Five years later, cleaning engulfed my life. My father had exited the Picture and my mother did not know how to drive, so scrubbing high school bathrooms with her became my life from the afternoon to the wee hours of the morning. My mother and I became comrades-in-arms, scrambling to figure out taxes, repairs, work and home. I took on a job working at a gas station (I had to pretend to be older to get that job - my alibi was being a Georgia Tech student) 20 hours each weekend to arrange enough money together for my own independence. This job had its own form of beauty: the ultimate exposure to the world - everyone needs gas, rich or poor.

I slept during class.
~~

This past October, I told my mother I was going to start a company. Cancer prevention seemed doable and it was as good a life mission as any, but it’d require a decent amount of cash to get started so I’d embark on a fundraising quest first.

Her first reaction was to relay me a story about how a scientist who had discovered the cure to cancer was shot down over Copenhagen by Them. Who Them was is unclear, but she had always had a bit of a conspiratorial flair.

When I informed her that this was nonsense, she spent a long hour telling me that this was a dumb idea, that it was not worth trying, that this was an endeavor that They would do but not people like Us. In this instance it was more clear who They were - it was those who did not spend their childhoods cleaning bathrooms.

This speech was, if anything, motivating. If They could do it, then so, perhaps, could I.

The journey to becoming a founder

Most of this article will be my path through fundraising, with the hopes that it will be a helpful read for those embarking on a similar journey - particularly those coming from rougher backgrounds. But first, some context:

My mind first turned to startups about a year out of undergrad when I was working for a company working on CAR-T immunotherapies. I loved my time there - it was a wonderful environment of brilliant people working on incredibly challenging scientific problems. Each day I came to work, I had the opportunity to help make life-saving cell therapies for cancer patients a little better.

But these cell therapies cost hundreds of thousands of dollars each, and were largely only relevant to the most terminal of the most terminal cancer patients. Tragically and poetically, this was around the same time that my father underwent multiple circuits of radiotherapy and immunotherapy and steroids (to treat the side effects from the immunotherapy) to treat a melanoma that spread to his brain. He would ultimately succumb to the disease after battling it for three years.

All of this sparked a fire to rethink how we tackle disease. I found the work that I was doing meaningful, but it would have been better to have prevented the cancer from existing in the first place. The main risk factor for cancer, as with most diseases, is aging. Thus, it made sense to go find the causes of aging and slow them down - an idea known as the geroscience hypothesis1.

I wasn’t sure how to effect change in this direction2. I looked for low hanging fruit and found none - scientists haven’t identified any compounds that meaningfully slow the pace of aging except for rapamycin, which was isolated in the 1970s3. My takeaway from this was that the field must be missing something pretty critical, so I opted to invest a few years in thinking very hard on the problem through a PhD. I applied to exactly one PhD program - the Biology of Aging PhD at the Buck Institute - and spent time working in labs that were too big to pay me much attention, allowing me to spend most of my time thinking about what aging is. The PhD path is not in vogue, but for me it felt very intuitive. Aging is the mother of all problems; it should not be surprising that tackling it correctly will take at least a few years of deep focused thinking.

Despite going into it with a hermit-esque “I am here to isolate myself in the mountains and think about aging” philosophy, my PhD experience ended up being fairly community oriented. I am still reflecting on why community was so important on this intellectual journey, but my suspicion was that it is deeply beneficial to have an environment in which people can quickly tell you that your ideas are not very good. There are countless times when I had some concept around aging that ended up being quickly shut down by a friend with a reference citation to some obscure paper from the 1980s. It’s much more difficult to have these rapid creative iteration cycles alone.

I wrapped up the PhD roughly three years after starting it. I had two major takeaways:

1. Aging becomes exponentially more difficult to tackle as we get older, and thus we need to address the root causes of it as early as possible. Ideally, beginning in our 20s.

2. One of the most important of those root causes is genome instability.

Finding ways to make a company out of these takeaways was not easy, but ultimately I arrived at the idea of creating therapeutics to prevent cancer. It sets up the company to directly target improving genome stability, and ‘preventing cancer’ is one of the relatively few reasons young healthy people would take a drug (at least, the FDA-permitted kind). Now it was time to go out and build it.

For this, I needed money.

Strategizing fundraising

Before going out and fundraising, I sat down and reflected deeply on my strengths and weaknesses. There were a number of tailwinds:

1. I knew my stuff. I had made a point of reading three papers every day during my PhD, which came in very handy when I would bludgeon a VC’s diligence team with my encyclopedia’s worth of knowledge on what I was doing.

2. I believed in what I was building and was obsessed about it. One very helpful biotech founder advised me to make it clear to VCs that I was going to be fine without them, to be the “pretty girl at a party.” In my case, this was very much true (the belief part, not being a pretty girl), and I would work on this problem regardless of how much money I’d end up fundraising.

3. I knew the game, kind of. In my time in VC I had the chance to see hundreds of company decks - those that were great and those that were terrible. I knew at the very least that I could field my pitch without missing something fundamental. I also knew that there are certain hard-to-describe perception factors that VCs look for that I was able to check - completing my undergrad at Stanford, a propensity for wearing leather jackets, projecting confidence, etc.

And also a couple of minor headwinds:

1. The market was not in my favor. At the end of 2024, the pre-seed environment for funding biotech startups was horrible and was quickly getting worse.

2. I had literally nothing. I did not have an asset nor a target nor a co-founder nor a team nor previous founding experience nor any preliminary data nor anything else that could be worth anything to anyone.

Taking all of these factors together, it made sense to try to raise $2-3 million dollars.

In case this seems ridiculous, an investor I admire introduced me to the world of VC incentives a few years ago at a retreat organized by the Longevity Biotech Fellowship. He said VCs care about three things:

1. Team. So much of company building is executing properly.

2. Moat or “defensible magic.” If anyone else can copy it, then it is not going to retain exclusive value for very long.

3. Potential to be a billion+ dollar company. Most VC investments fail. To make returns, the few investments that don’t fail need to be massive successes.

I slotted into this framework very nicely. As a lone voice in the desert obsessed about this field, I was one of the very few choices for investors interested in backing it. The technology is easily patentable, and by targeting disrupting all of cancer I chose a pretty large market.

Of course, none of this felt very real to me. I distinctly remember helping my parents with their tax returns (they were illiterate in English) and seeing the annual income of 70+ hour workweeks totaling to a maximum of $40,000 from both of them combined. But I learned quickly to separate myself emotionally from the fundraise amount and treat it as a means to the end goal of improving genome stability - which it is.

In any case, I did need to correct for some of my weaknesses as a founder, which was the first thing I did.

The primary perception flip I needed to do was to convince potential investors that I am not someone who is crazy, but rather someone who is crazy and is on to something big. This could be done by finding advisors. I reached out to about a hundred experts who I saw as potential advisors and had meetings with about half. I offered advisory positions to those who either filled critical gaps in my knowledge or asked particularly insightful questions.

With the company idea in mind, a team of enthusiastic scientific supporters, and a target fundraise amount, I was ready to shoot my shot.

Beginning fundraising

Fundraising for a pre-seed is common enough to where there is a standard playbook. I will describe how most people approach it, and where I deviated.

The usual strategy is to create a massive spreadsheet list of every VC you think could possibly be interested in investing. You rank them in tiers and then ask everyone you know if they could provide you to an introduction to the highest-possible-ranking person at any of those VC funds4. Once you have your list of potential VCs and potential introducers, you give the bugle call to send the introductions and schedule all of the meetings at the same time - starting with the “worst ranked” VCs first. You then tell everyone you have a ton of interest in order to drive momentum and hype.

The core problem that this tackles is the fact that investors have a natural preference to wait. The longer they wait, the more information they have. By scheduling everything all at once, you force them to make decisions quicker as otherwise there’s some risk they’ll miss out on the deal. There’s also a tendency to try to remain as cool and quiet as possible through the entire process to generate excitement and FOMO.

My strategy was much more chaotic. I avoided the tier tactic, mostly because I knew I would tune my pitch depending on feedback and I did not want to over-index on the opinions of folks that I didn’t want to invest anyway. I did schedule every call at once - but then I just kept going, adding on dozens of additional investor meetings week after week after week. I abandoned the “be cool” vibe entirely in favor of the perception that I was manically obsessed with solving a critical problem that everyone else is overlooking, which was much more authentic anyway. I solved for time pressure by initiating fundraising in October and telling everyone I would close the round by the end of the year. This was an arbitrary deadline to investors but a practical deadline for me, as I would run out of personal funds at that point and would probably annoy my housemates by not having money for rent.

Choosing to pitch ad infinitum was tough, but was a battle to which my background prepared me for.

Relentlessly continuing despite all odds and despite repeated failures was a feeling I had become deeply familiar with over the years. In the course of fundraising, I pitched over a hundred times and had dozens of more conversations on top of that. My emotional armor around being told ‘no’ became very refined and practiced.

Early wins

My first few weeks of fundraising went well. This was due to a factor that I had not anticipated: giving a shit.

Most conversations around fundraising - this article included - come at it from the point of view of the founder. This makes it easy to talk about strategy and tactics as though the fundraising process were a wholly impersonal board game. VCs are used to being engaged with on purely transactional terms - as potential sources of funding, irrelevant as soon as they pass on an opportunity.

However, VCs are human beings too [citation needed]. There were a few investors that I had made good relationships with over the years without any expectation of an investment, and reaching out to them led to me raising $1M within a couple of weeks. My days working at a gas station had imprinted on me a sense of trying to see people for who they are instead of for what they can provide for me, and this ironically ended up being very professionally lucrative.

In my brief time on the investing side, I had also created friendships with founders. My biggest surprise in all of fundraising was how much help founders were willing to provide for each other. There were founders who made 4-5 incredibly useful connections even without me prompting. There were founders who sat down with me for three hours over beers teaching me the very basics of company building without any expectation that I would ever return anything back to them. In every environment I had ever operated in, I had experienced some sense of crab mentality - this seems nearly absent among folks building early-stage biotech companies. It’s all camaraderie.

The early honeymoon phase was a good boost of confidence. Not doing the common strategy of prioritizing “lower ranked” VCs meant that I took a few rejections from VCs I respected, but I got very good feedback that was worth its weight in gold.

A few operational tips

This may distract from the overall narrative a bit, but here are some quick bullet points that are likely be useful for founders early on in the fundraising journey:

• I found the 80/20 rule applies to almost everything. Many founders either create and design their own slide decks or pay a designer upwards of $1,500 to make one for them. In my case, I made my own deck and paid someone on Fiverr $150 to make it look aesthetic. I did the same for the logo. I was punished for not putting in the extra 20% effort later, but the time I gained by operating much quicker was worth it.

• I did not worry about confidentiality. If someone wants to go steal my idea and does all the hard work of solving cancer and aging without me, then…

  

• My life became so much less guesswork when I brought on an operational advisor to help sanity check some of my choices. He also unwittingly became a deeply appreciated emotional sounding board.

• Good feedback was critical. I ended up changing the entire narrative of my pitch a month into the fundraise.

A darker period

At some point I hit a dead end and ran out of steam. I had enough VCs hem and haw and signal that they’d be interested for investing some amount, but it was unclear what amount and on which timeline5. Starting fundraising in October was a tactical decision aimed at speeding up the process, but both Thanksgiving and Christmas came and went without a completed process. For New Year’s, I threw a dart at a map on my wall and resolved to go somewhere far away. The dart hit Billings in Montana before falling to the floor, and so off I went.

There were a few assumptions I made that ended up not panning out. Given my scientific background, I expected the have the most traction among biotech investors. What I did not appreciate at the time was that biotech investors have something of a formula: find a particular protein target, then find some disease for that target, then find a drug that will hit this protein. My pitch - let’s target mutations to prevent cancer - fit very poorly into this formula.

I also found myself having to rapidly iterate on not on my pitch but also myself. I was a few months out of academia, and academia strongly frowns upon any sign of ego. I have a funny memory of giving a presentation at the Biomarkers of Aging conference, with my description of the work being ultra-narrowly tailored to the specific plot that I was describing without daring to stray into speculation or even implications. The same evening, I gave a pitch to a VC where my script was effectively “we will literally cure cancer.” It was a wild head-spin.

Eventually, things came together. I resolved to do a last call in January and wrap up fundraising, and this neatly aligned with finding a lead investor. I was not sure how to approach negotiating, but I came into it with two principles:

1. Treat negotiating on terms impersonally. I pushed for the best terms that I thought I could reasonably get, and pushed fairly hard.

2. Making sure everyone gets a fair shake. At the end of the day, the journey is long and it’s good for everyone to feel like they’re in it together like some grand adventure.

These principles led to some perhaps irrational choices. My first ever angel investor (who shifted me from “someone who has not raised capital” to “someone who has raised capital” - a critical binary shift in the eyes of many VCs) committed $500,000 in October on much worse terms than our lead investor negotiated in January. My lawyers informed me that the time gap meant I did not have to do anything here; I chose to harmonize the angel’s terms to match those that the lead negotiated anyway. It was critical for me that the people who took a bet on me knew that I, in turn, would advocate for them.

Some parting thoughts

If you’ve never read Peter Thiel’s Zero to One, I’d highly recommend it. It’s useful, first of all, for understanding how people who have build big things think. More saliently, it begins with this absolute banger of a line:

If you’re trying to start a company, odds are high that you are inherently trying to swim against the current. This is because you’re doing one of the following:

1. Doing something that everyone else thinks is a good idea, in which case everyone else is doing it and you’re competing in a crowded field to be the best.

2. Doing something that few other people think is a good idea, in which case you’re trying to convince the world (and yourself) that you’re not insane.

Both of these require a great deal of bravery. There is skill and intelligence and EQ and knowledge and many other factors that come into the equation, but ultimately I found that the most useful muscle in my fundraising journey was resilience. Fifteen years ago, I negotiated a wage increase from $1.25/hr to $2.50/hr. Two months ago, I negotiated a term sheet to build a company to prevent cancer. The key commonality was believing that I could, and not stopping until I did.

Consider this:

1. An elephant has over a thousand times more cells than a human.

2. Cells accrue mutations independently.

3. Cancer is mostly driven by mutations.

So why do we not get cancer a thousand times more frequently than mice?

This is Peto’s Paradox. There are many proposed “answers” to this paradox, but there’s a major clue found when investigating the number of mutations animals acquire per year. There seems to be a natural limit to the number of mutations each animal acquires per cell on average, and that number is surprisingly small:

3,000.

And it takes mice a little over three years to accumulate these 3,000 mutations, compared to sixty years for an elephant and around eighty for a human. Take a look at the graph below comparing mutation rate per year with an animal’s mass:

from Cagan et al. 2022

This creates a straightforward explanation – larger animals, needing to avoid dying at a young age due to cancer – evolved better mechanisms to protect and repair DNA. This is not the entire story, as better immune surveillance and differences in metabolism contribute as well, but it’s as close to a complete picture as you get in biology.

Entirely separately, longevity researchers have observed a correlation between the size of an animal and its maximum lifespan as well1:

from Kuparninen et al. 2023

Why? It’s not entirely clear. The aging field has assumed that this has something to do with metabolism, as larger animals tend to better pace their energy uptake. But there’s another aspect to the story, in that there’s a strong association between mutational burden and lifespan that dwarfs both the association between mass and mutations and the association between mass and lifespan:

from Cagan et al. 2022, again

In fact, if you account for the association between mutational burden and lifespan, the original link between body mass and lifespan ceases to be significant.

So there are three interesting observations (drawn below):

1. There is a strong association between larger body mass and improved ability to maintain the genome. This is likely to be an evolutionarily-driven adaptation for survival.

2. There is a moderate association between larger body mass and longer lifespan.

3. There is a very strong association between improved ability to maintain the genome and maximum lifespan. This is unlikely to be evolutionarily-driven, as aging is something that tends to occur after an organism’s reproductive window.

A unifying explanation: Extended maximum lifespan in larger animals is an evolutionary accident of requiring better genome stability mechanisms to avoid cancer. Maximum lifespan is a byproduct of solving Peto’s Paradox. If you stop the things that cause you to get cancer, you get extended life for free. Preventing damage to the genome is a path to living longer.

This isn’t the only reason to view genomic instability as a, if not the, major driver of aging. Effectively every single disease in humans that resembles a form of accelerated aging is caused by a mutation in a gene related to genomic stability:

from Worm et al. 2024

The age of menopause in women, which is itself linked to mortality and morbidity, is strongly tied to DNA repair genes. A better version of a DNA maintenance protein is heavily enriched in centenarians. Molecular markers immediately downstream of genomic damage - DNA methylation, mutations - change on a constant basis with age, suggesting they do not participate in the secondary feedback loops that define every other aging hallmark.

Note I emphasize lifespan here instead of healthspan. Healthspan is linked to a separate suite of programs. But that is a story for a different time.

So aging is caused by mutations, then?

No.

There is a rare disease (a mutation in POLE) that leads to an accumulation of mutations across the genome at a frequency up to ten times higher than in the general human population. Unsurprisingly, this leads to a high risk of cancer. Perhaps surprisingly, there does not seem to be much of an effect on aging. The increase is not tied to any particular cell type, either:

from Robinson et al. 2021

It’s also unlikely to be mutations in the mitochondria, as mice have been generated with 500-fold more mitochondrial mutations while not aging a whole lot faster.

To be clear, mutations have been linked to manifestations of aging. The most well-known example is clonal hematopoiesis, where blood stem cells pick up mutations that make them better at dividing which then leads to an increased rate of cancer. There are also a number of studies emerging linking clonal expansion of various cell types in the brain with neurological disorders such as Parkinson’s and Alzheimer’s. The effect sizes for these studies tend to be pretty weak, however. In some cases, mutation-induced clonal expansion has actually been shown to have positive effects, such as by improving outcomes for some cancers.

So what is it?

Here’s a pretty simple model:

I drew this on an iPad on a plane undergoing turbulence

I’m being quite vague here, but this is intentional. The jury is still out on which aspects of genome instability are the most important for aging. David Sinclair’s information theory, likely one of the most well known theories of aging, centers around epigenetics (particularly histones) as the key determinant. 3D genome structure, thus far under-appreciated due to the difficulty of studying it, may play a pivotal role as well. Cell damage signaling is important too, though it may be more tied to healthspan rather than lifespan.

It could also depend on the cell type. About a year ago, I embarked on a quest to map every single disease linked to DNA repair in humans and whether it resembled aging or not. The diagram below is a little inaccurate, but the conclusion I came to was that repair pathways that predominantly impact slow or non-dividing cells are linked to aging, while pathways that impact fast-dividing cells lead to a higher risk of cancer.

Circling back to the main point - if it is true that maximum lifespan is an evolutionary accident related to large animals improving genome stability to avoid dying of cancer, what are the implications?

Two immediately come to mind.

Implication 1: Focus on preventing or reversing genome instability

On the reversal front, the field of epigenetic reprogramming has gained an enormous amount of attention. Other authors have written (one) (two) excellent posts on the subject, so I will leave the reader to peruse those references.

On the prevention side, however, there is remarkably little work being done2. This is surprising, because it seems like a much easier angle (an ounce of prevention is worth a pound of cure, as they say), and because there’s a lot of low hanging fruit. Spermidine, a readily available supplement, has been shown to enhance genome stability (side note: sperm accumulate fewer mutations than other cell types, despite being quite proliferative. spermidine may be why). Deep sea vent bacteria make compounds that protect their DNA. Bowhead whales, which live well past 200 years, have a better version of a key DNA repair enzyme that could have translational potential in humans.

Implication 2: Figure out which pathways drive cancer prevention, and which drive longer lifespan

You may have heard that elephants have many copies of p53, the “guardian of the genome” gene. What you may not know is that mice that have been engineered to have more p53 copies have reduced lifespan and poorer health.

Not every pathway that improves cancer resistance leads to longer lifespan. In many cases, it’s the opposite. For genes related to the cell cycle - such as p53 - having too much “error checking” leads to a failure of stem cells to divide properly, leading to accelerated aging but less cancer. On the flip side, having longer telomeres can be beneficial for aging while also promoting cancer growth. You can imagine a diagram that looks like this:

Note that if the Peto’s Aging theory is correct, then most interventions in genome stability should be somewhere in the middle. This pans out with human data, as diseases resembling accelerated aging tend to come with a higher cancer risk as well. On the flip side, reducing calories both slows down cancer onset and is one of the few interventions that has been shown to extend lifespan. The same is true for rapamycin.

The implication here is pretty profound - if you can prevent cancer, in most cases you should be able to extend lifespan as well. The key will be to prove that this is true by identifying more interventions that can slow the onset of both cancer and death.

Subscribe now