Dr. Gordon Lithgow on conflicting results in longevity studies, the origins of the CITP, and more (part I)

Like to share?

Image credit: Buck Institute/Wikimedia Commons

Dr. Gordon Lithgow got in on the ground floor of aging research in the lab that discovered the first anti-aging mutation, age-1, and has spent the 25 years since trying to unravel the mysteries of the aging process. His lab has been at the Buck Institute for Research on Aging since 2001, studying everything from stress response, to protein homeostasis, to antioxidants. Most recently his lab took part in the Caenorhabditis Intervention Testing Program (CITP), a spinoff of the National Institute on Aging’s long-running Intervention Testing Program, which employed a clever study design intended to address the doubts around reproducibility that had long haunted the anti-aging community. Below, he gives us the story behind the study, a possible explanation for the puzzling discrepancies often seen between labs for the same compound, and a closer look at some of the compounds tested.

Dr. Lithgow, can you tell us how the CITP came about?

So back in 2003, the NIA set up a project with three independent labs working in mice to establish a protocol that all three labs would use, and then they would test compounds for effects on lifespan. And they were doing this because there was a feeling that longevity studies, especially in mice, were really not sufficient: there were too few animals, and there wasn’t confidence in the literature. And it was a very successful program, identifying, for example, rapamycin as a compound that slowed aging in all three labs in the consortium.

So then a lot of us were talking with them over the years, saying, “Well this is great, but you could do this so much faster in other systems, and how many compounds do you really want to look at?” And so in 2013 they came out with this call for proposals for another consortium, but this time just with Caenorhabditis. And this is what my lab’s been doing for the last 10 years or more, finding chemical interventions in aging. So of course we applied, and we were one of the three labs lucky enough to get funding. And we did exactly what the mouse people did, we spent a year getting a protocol together that all three labs would then conduct – very, very detailed protocol, very detailed. And we got to work on trying to find compounds that were robust.

I should say, back in 2000 we published a paper, one of the first papers to come out of the Buck, where we showed that you could extend the lifespan of C. elegans by feeding them catalytic antioxidants. So this paper was published in Science, and it got huge media attention. These boxes up here are newspaper clippings from all over the world, and these folders, another set of newspaper reports…

(Pulls a five inch thick binder off the shelf, cracks it open, showing it stuffed with article clippings)

Just crazy coverage, basically saying that this was the first time that anyone had extended the lifespan of an animal with a drug. And that’s not true, but it was true that it was the first one published in a major journal.

Three years later, a good friend who was at graduate school with me, David Gems in London, published a paper saying they didn’t work – they couldn’t see lifespan extension, couldn’t reproduce our data. This has been an issue with the field over the years, that many high profile papers were not reproduced by other labs. And so the CITP was an effort to directly address what was going on with this. We were never able to resolve why David couldn’t reproduce the work, so it was a little bit personal.

Was it difficult to coordinate strict adherence to a single testing protocol?

Yeah, it was really difficult. All three labs were on the phone every week in some form or another, long conversations about minutiae like how you pick up the worm and how you put it down again. These phone calls were really, really important. For example, there was a point in time where one of the labs had a three day lifespan that was longer than the other labs, and we were going over and over, you know, what is it that’s different? All we could think of was the air, everything else in the experiment was controlled. And then in one of these phone calls somebody said, “Wait a minute, did you say you call your adults ‘day 1’?” We said yeah, and they said, “Well, we call our adults ‘day 3’!” And that was it, that was the difference. No one had defined what ‘day 1’ was.

Even though all the trials followed the same protocol, worm lifespans were really variable within each trial. Do you know why that was the case?

That’s a really good question, and I don’t think we really have the answer, but part of it is that aging is very plastic, it’s very responsive to environmental change of every sort. So what might seem like completely insignificant differences in protocols or in air or water or some other basic feature can have dramatic effects on lifespan. So small fluctuations in temperature, for example, could make a big difference between one lab and another, or the way a population of worms will respond to a compound could change.

If you really get down to it, the experiments appear simple. You take worms, they’re on your agar plates, you squirt a compound on and you watch and let them die, but it’s everything from: how’s that compound behaving in storage; how’s it behaving on the plate; is it oxidizing; is it coming out of solution; how is the bacteria metabolizing – because there’s bacteria on there as well – how is the bacteria metabolizing the compound under the particular conditions for this experiment; and then how are the worms metabolizing the metabolizing…

I mean, there’s an endless series of possibilities. We’re not in total control of everything, of all the outcomes. So even under very standard protocols, you can find examples where, wow, it’s just not working, that biological replicate. And there’s no reason to throw it out, you can’t think of a reason why it’s not doing it, but it’s not doing it! The other ten biological replicates are working just fine, thanks very much, but this one’s not. So that’s the practical reason as to why it’s variable. There’s a more theoretical reason, and that is that processes like development and fertility are very much under genetic programming, they’re the outcome of darwinian natural selection. Aging isn’t like that. Aging is not a program, and the effects of genes on aging are not selected, so there’s no selective pressure, and I think that means that things get fuzzy. It doesn’t matter that a gene is going to affect aging late in life in a dramatic way or a small way between individuals or between experiments or between populations, because animals have never lived that long in evolutionary history. So I think that’s a possible explanation as well.

So the variability is not specific to worms, we would expect it in humans.

I would expect variability. We do see variability in humans obviously, and mice in the lab, you get mice that die early. If you look at an old population of mice, they’re really heterogeneous, even if they all have the same genes. Some individuals are aging badly and their fur looks horrible and they’ve got curved spines, and the mouse right next to it, in the same cage, same genes, same environment, is doing fine. So there’s a lot of, I mean, people use the word “stochastic variation”, and I think that just means they don’t have an explanation for where the variation is coming from. And it doesn’t seem to make sense that there should be variation between individuals. It may be that that variation really goes back to early events during development, and that these genetically identical mice are different animals as they come through middle age.

We see this weird thing that we’ve got no explanation for that we published in the paper. The biggest source of variation in the experiments were from experiment to experiment, so there was no systematic variation between the labs, but there was still huge variation. So if you collapse together all the survival curves, all the labs overlay each other. But what we noticed was that over time, the variation was splitting into two groups, it was biphasic. And it’s not happening all the time, but it’s happening frequently, and it’s happening in all three labs. Maybe it’s a trivial problem that we just can’t get a handle on right now, and it won’t be interesting. Or maybe it’s real biology and there’s something real happening there. So we’ll see, but we had absolutely no idea this was going on, until – and I’ve been doing this for 25 years – we had no idea that there was this possibility of one mode or another. We knew that experiments are variable, but nothing like this.

And then when we got around to testing compounds, here’s the best set of data we have for any compound. Thioflavin T was originally published in 2011 in a Nature paper, so we were retesting it in this paper.

Each of the dots here is a biological replicate experiment, and if you’re above the line, it’s a lifespan extension in percentage. So almost everything is long-lived, right? But there’s no explanation for these blue points below the line, and this was in the same lab. I say no explanation, we have some idea of what’s going on, because this compound does become oxidized and can be toxic at high concentrations. So we’ve got some ideas.

But here’s what I think is really important. Here’s an example where the compound’s not doing much at all. Basically all the points are scattered around the line. It’s a sort of normal distribution of replicates. But if you’re in the red lab, you think this compound’s pretty hot shit, and you go and write your Nature paper. If you’re in the blue lab, you think it’s killing worms. We’re doing all of these experiments at exactly the same time, and all the labs are doing roughly the same thing. So there are sources of variation that we just didn’t know about.

And I think this explains what’s happened in the literature. We didn’t know how variable lifespan was, so we would come in and publish our flashy Science paper, and then David would come in, and he samples a different bit of the variance for whatever reason, and he says, “No, these don’t work.” So it’s messy. No one wants to hear this. People want to think you can do an experiment three times and publish and it will be great, but the reality is that we now have to think about doing experiments at scale, and actually having key findings reproduced in other labs.

Do you expect that Thioflavin T or similar compounds will ever be used in humans?

I don’t know, we don’t have a safety profile for Thio T in humans. A compound that’s very similar to Thioflavin T called Pittsburgh compound B actually does exist, and is used in humans to image amyloid in PET scans, so I think that was what attracted some of the media coverage that we originally got. So yeah, it was clearly something that had been used in humans. I think it’s fairly safe and is cleared really rapidly from the brain. Since then I haven’t seen any clinical trials or anything else. As usual with these things, if you can’t secure a patent, it’s difficult to then commercialize or get a clinical trial going. And the patent landscape around Thioflavin T is quite difficult because of Pittsburgh compound B. There are a lot of patents out there on structures very similar to this, so we haven’t been able to obtain a patent, and that limits our ability to go forward.

Was the poor response to the antioxidants a surprise?

I guess it wasn’t a surprise, because there’s been quite a lot of genetic discoveries published suggesting that our understanding of oxygen radicals and aging was too simplistic. Lots of papers were showing that knocking out antioxidant enzyme genes in worms and in mice as well were not affecting lifespan. However, we have seen over the years that antioxidant compounds can be protective, especially against stress, with some of them also extending lifespan in a small way as well. So when we originally published propyl gallate and lipoic acid, both gave small, reproducible effects, generally affecting median lifespan rather than maximum, but we also had other antioxidants that had no effect at all. Propyl gallate did not work in our hands with this protocol that we’re using now, despite the fact that it worked previously for us. So no, I guess we’re not surprised.

There’s no question that oxygen radical damage is a feature of aging. There’s also no question that it’s a feature of almost every age-related chronic disease. We just don’t know how it plays out on lifespan and longevity. I hope that we’ll go back to studying oxygen radicals in aging, because fundamentally I think that’s the chemistry that drives a lot of aging processes, but we just haven’t found a way to prove that yet.

Tegan McCaslin

Tegan is Geroscience's lead editor, and writes on a variety of topics--mainly science, medicine, and humans--here and elsewhere on the web.