Three Open Problems in Aging
Aging, like many parts of biology, has a surplus of excellent experimentalists but a shortage of good modeling/analysis/theory. LessWrong and the adjacent community have a surplus of people with great modelling/analysis skills, many with an outsider’s interest in aging, so this post outlines a few problems which could fit that skillset. I expect that all of these could be solved to a reasonable degree of certainty with careful modelling and some beyond-high-school-but-not-revolutionary statistics, using already-existing data. For a publication-quality example of the sort of project I have in mind, check out the analysis of Burd et al’s data toward the beginning of this paper, or the modelling in this paper. (I’d also be happy even with something more back-of-the-envelope than these, as long as the approximations made sense and the predictions/assumptions were checked via multiple channels.)
Note that these are all open to the best of my knowledge. I’d be quite surprised to find existing analyses of the sort I have in mind, since it’s not the sort of work academic biology usually incentivizes, but it’s not out of the question.
Atherosclerosis: How Much Oxidation?
For a couple decades now, it’s looked like oxidation of lipids (aka fats) is a mediating upstream cause of atherosclerosis—i.e. whatever the root cause of aging is, its effect on atherosclerosis is mostly mediated by lipid oxidation. There’s also been some mathematical modelling of the lipid oxidation → atherosclerotic plaque formation process.
Obvious next question: what causes the rate of lipid oxidation to increase with age? Obvious hypothesis: reactive oxygen species (ROS), e.g. superoxide and fenton reaction products. ROS play major roles in multiple theories of aging and have confirmed roles in many aging processes, so they’re a natural culprit. But which specific ROS are involved here, and what upstream change causes an increase in concentration of those particular ROS?
I’d like to see someone use models like those linked above to estimate how much lipid oxidation needs to increase to account for age-related development of plaques, then do some back-of-the-envelope calculations on how much we’d expect from specific sources/theories, and compare the two.
For example, one model might be that senescent cells spew ROS, so the age-related increase in senescent cells causes the increase in lipid oxidation. In that case, we’d want to either estimate or find an empirical measurement of ROS production by senescent cells (and which specific ROS are produced), look up the age-related increase in senescent cell count, and then calculate the expected increase in ROS production. (Important note: presumably we want to know about extracellular ROS production, not just ROS production within the cell.) We could also think about variants of this hypothesis—for instance, maybe lipids are oxidized within the senescent cell or its membrane and then exported, rather than being oxidized in the bloodstream, in which case we’d want to find or estimate numbers for that export rate.
Another model might focus more exclusively on fenton chemistry—the iron-catalyzed production of ROS directly from peroxide. Then we’d want to look up estimates of fenton reaction activity in the bloodstream, as well as age-related changes in iron levels and peroxide concentrations, and see how the expected change in ROS production compares to the numbers needed to account for atherosclerosis.
Yet another model would be the mitochondrial free radical (MIFRA) theory. Aubrey de Gray’s book laid out a pathway by which cells with mutant, non-functional mitochondria could continue to live, but would export large amounts of radicals. There is data available on the number of cells with mutant mitochondria (and even on which mutations are present), so presumably one could estimate the expected change in radical production by such a mechanism, and compare it to the numbers needed to account for atherosclerosis.
Finally, changes in antioxidant concentrations could be useful to look at. I doubt that they’re actually the main causal mechanism here, but any age-related change in antioxidant levels would multiply the impact of any other mechanism, so depending on how much precision we can get we might need to account for antioxidant changes in order to make the numbers add up.
Senescence-Induced Senescence
There’s some evidence that senescent cells can induce senescence in healthy neighboring cells (“senescence-induced senescence”). This can’t account for the age-related increase in senescent cells by itself—senescent cells turn over too quickly. If senescent cell counts were in self-reinforcing exponential growth via this mechanism, then aging would take place on a timescale of weeks or months rather than decades. However, senescence-induced senescence could contribute a multiplier effect to some other cause of cellular senescence, one which at first seems too small to account for the age-related increase in senescent cell count.
That said, the multiplier effect can only be so large. The larger the multiplier, the closer the system is to instability (i.e. exponential growth of senescent cells on a timescale of weeks/months). If it’s close to the threshold, then we’d expect to sometimes see a supercritical explosion of senescent cells when some outside stressor happens to push the system past criticality—again, on a timescale much faster than aging. More generally/qualitatively, a system very close to criticality should show a loss of homeostasis on a faster timescale than what we actually see.
Anyway, the main question is, just how big is that multiplier?
Then the next question is: what’s an upper-bound estimate on the fraction of senescent cells in an animal which are long-lived (based on e.g. the data here)? We know (from that linked paper) that senescent cells turn over on an average timescale of weeks, but if a small fraction of them are long-lived and there’s a large multiplier from senescence-induced senescence, then the long-lived senescent cells could be a plausible root cause of the age-related increase in shorter-lived senescent cell counts. Is it at all plausible that some type of long-lived senescent cells can causally account for the number of short-lived senescent cells growing with age, or can we ignore that mechanism in our search for a root cause?
The Mitochondria-Senescence Bistable Switch
Probably the most interesting combo of papers I’ve encountered lately: senescent cells and cells with inefficient mitochondria are probably the same cells. There’s a positive feedback loop at the center of it all, so any of genotoxic stress, mitochondrial inefficiency, double strand breaks/telomere loss, or probably other stressors can induce any others of those. There’s even potential lock-in mechanisms: selection for deficient mitochondria or activation of transposons. The question is, which factors are the chicken and which the egg?
A relevant conceptual model here is the bistable switch. The cell has two states: normal and senescent. Both states are stable: if there’s a small stress to the system, negative feedback will push it back to normal. But if there’s a sufficiently large stress, then a positive feedback loop kicks in, pushing the system into a new state—until eventually the new state’s control system kicks in and everything stabilizes again.
The challenge is that many different factors can be involved in the positive feedback cycle, and a sufficiently large increase in any one of those can cause the switch to flip (i.e. cause a normal cell to senesce). So, if we see an increase in senescent cells, then we’ll see an increase in genotoxic stress and mitochondrial inefficiency and telomere loss and …. If we experimentally induce any one of those, then we’ll find that it causes the switch to flip—any of them could be causal. But it’s difficult to tease out which change occurred first.
Mitochondrial mutations and transposon activation are two factors with slow enough timescales to potentially account for age-related change, but maybe there’s some other we don’t know about, or maybe there’s an intermediate fast-timescale cause and the root cause is even further upstream. Any sort of model which untangled causality in this system would be great. Two possible approaches to narrow it down:
To the extent that senescent cells turn over quickly, the cause of the switch-flip must come from some other cell or extracellular factor. (Though this might include the senescent cell’s parent.)
Whatever the original cause is, we might expect to see it change more even before the switch flips—i.e. at younger ages, or in non-senescent cells.
… or you might have some other clever idea to tease out the first mover. One way or another, we’d need quantitative analysis to inspire any sort of confidence in the answer; it’s the sort of problem which tends to have subtle failure modes.
If someone decided they wanted to fund such a project how much would you estimate it would cost? (Let’s say based on the assumption it didn’t have to be publication quality, just good enough to persuade you it was likely correct)
Depends on the approach.
When writing the post, I was imagining these as mostly-analysis projects, using already-existing data. They shouldn’t be a huge amount of work; the first two in particular could easily be spare-time projects or big class projects, and the third is probably limited more by having the right insight on how to tackle it rather than on time investment. If you wanted to pay someone for that sort of project, basically all the work would be in finding the right person. I could easily imagine paying an undergrad with good analysis and literature search skills $5000 to spend the summer on one of these; that should be enough time to produce a very thoroughly-executed product. I could also imagine paying an undergrad a couple hundred bucks to do one of the first two as a class project, rather than whatever else they might have done as a class project (bearing in mind that it would probably take a bit more background reading and literature search than a typical analysis-focused class project).
(In fact, I originally wrote these up when someone on LW asked me for recommendations for an aging-related project for a probabilistic modelling class. That person is giving the senescence-induced senescence problem a shot, and is welcome to leave a comment here if they want to advertise the project. I could imagine that they might be willing to put in more marginal time/effort for a reasonable number of dollars.)
If you really wanted to throw money at a project, then funding experiments on the same questions might be easier, just because it’s easier to find competent experimentalists already in the field. That would take a lot more money, though; I’d estimate hundreds of thousands. For that approach, we’re talking a lab and full-time researchers.
I’m the person starting to work on the senescence-induced senescence problem. Happy to chat more about current thoughts / plan (I am open to trading marginal time for relatively small amounts of $ but also happy to just talk about what I plan to do anyway). Feel free to DM me.
Frequently you make the claim about time frames and application to aging. While I do think that is something that needs to be looked at I also have a bit of a concern that it might include something of a error. An error along the lines of a fallacy of composition error—all the parts are small so the machine must be small.
In terms of senescence cells I’m not sure I follow you claim about they turn over too quickly. I understood the problem to be that some senescence cells don’t die but sit out there as zombies doing things that are generally not well coordinated with the rest of the system (our body). The cells that reach senescence and then are garbage collected for recycling are not a problem for us.
So as I understand the issue with senescence cells it is the slow accumulation of such cells relative to healthy, normally functioning cells that leads to the problem. That doesn’t seem to be a quickly turning over process. So when you say “senescent cells turn over too quickly” what is the context and what time frames are you considering?
This is exactly the model which the “Senescence-Induced Senescence” section is trying to test.
The data from here shows that the large majority of senescent cells do turn over quickly (~2-3 weeks), and this is true even in old age. The number of senescent cells which turn over quickly increases (slowly) with age, and the large majority of the age-related increase in senescent cell count is an increase in fast-turnover senescent cells. So there is definitely something increasing the production rate of senescent cells in old age. (That paper also estimates the age-related changes in both production and removal rates.)
Now, it’s still possible that accumulation of slow-turnover senescent cells could cause the increased production rate of fast-turnover senescent cells. That model has two key components:
There’s a subpopulation of senescent cells which do not turn over quickly, and accumulate over time
The presence of that increasing subpopulation is enough to cause the increase in fast-turnover senescent cells.
Quantitatively, the first component says there’s an increasing “base count” of slow-turnover cells, and the second component says there’s a multiplier. The same data used to estimate turnover can also be used to upper-bound the fraction of slow-turnover cells, and measurements of senescence-induced senescence should be usable to estimate the multiplier, so we should in principle be able to check whether the (small population of long-lived senescent cells + multiplier) theory is plausible.
Make sense?
Reminds me of this paper, in which they replaced the blood of old rats with a neutral solution (not the blood of young rats), and found large rejuvinative effects. IIRC, they attributed it to knocking the old rats out of some sort of “senescent equilibrium”
The Conboys are looking to start human trials with they neutral blood replacement approach (https://newatlas.com/medical/diluted-blood-plasma-reverse-aging-in-mice/?utm_source=New+Atlas+Subscribers&utm_campaign=9db0c9efb9-EMAIL_CAMPAIGN_2020_06_16_01_29&utm_medium=email&utm_term=0_65b67362bd-9db0c9efb9-92444869).
I didn’t realize they were tying their work to senescent cells though.
More on lipid oxidation: it also depends on the composition (polyunsaturated to saturated fat ratio) of the lipids that compose the cell’s walls. You also need to account for all the other highly reactive cell oxidation byproducts (eg 9-HNE, michael adducts, methylglyoxal, CML) as well as the overall redox potential of the cell (eg glutathionine seems to be an abundant antioxidant that reacts with many highly reactive byproducts..)
Additionally in the case of arteriosclerosis, it also depends on the ratio/supply of oxidized cholesterol (particularly keto-cholesterol). What you described for arteriosclerosis seems generalized for any cell, but it’s clear that there’s something happening on the artery walls at a faster rate that than what’s happening to other cells/tissues and you haven’t fully accounted for the difference yet.
Fenton chemistry has feedback loops with ascorbic acid (higher levels of metals can turn ascorbic acid into a pro-oxidant).
[quantification of AGEs—https://www.fightaging.org/archives/2020/11/towards-better-quantification-of-ages-and-cross-links-in-human-tissues/ ]