Ep 98: There’s a thousand genes for that! (with Nick Barton)
What is the “infinitesimal model”? How has our understanding of complex traits changed recently?
On this episode, we talk with Nick Barton, an evolutionary quantitative geneticist at the Institute of Science and Technology Austria. Quantitative genetics has changed a lot in the past 30 years, driven by massive advances in DNA sequencing power and by new statistical and computational approaches to harnessing the data flood. Nick works at the forefront of the field, developing and testing new theory, and we discuss both his research and his perspectives on these changes. We end by asking Nick about his advice for early career researchers who want to navigate the complex landscape composed of theory, computation, and data.
Cover art by Keating Shahmehri
Donate to our spring fund drive! Or, become a patron! Or both! – hey, we’re not picky!
-
Art Woods 0:00
Hey, Big Biology listeners, we want to alert you that we're doing our spring fund drive. We're not going to torture you with long statements and outtakes from our call center.
Cameron Ghalambor 0:10
We have a call center?
Art Woods 0:11
Nah I was just kidding. But we really are doing a fund drive and we'd encourage you to donate now.
Cameron Ghalambor 0:17
And to sweeten the deal. A superfan of big biology has agreed to provide up to $10,000 in matching funds,
Art Woods 0:24
So please consider making a donation today. There are two main ways you can do this. The first is to make a one time donation at our website big biology.org.
Cameron Ghalambor 0:32
The other way is to sign up to give a monthly donation at our Patreon site, which is that patreon.com/big bio. We'll apply matching funds to the next year of new donations.
Art Woods 0:44
Thanks for listening to Big Biology and for helping us to keep the show on air.
Cameron Ghalambor 0:48
When you picture a scientist, chances are the image in your head doesn't represent most researchers.
Art Woods 0:54
Let's take a moment to rethink our biases and work to shift the narrative within the scientific community into something that reflects reality.
Cameron Ghalambor 1:01
Join scientist and science communicator Anne Chisa, as she chats with African scientists across the globe who are involved in various fields of STEM. The purpose is to amplify Africans in STEM to the world and educate listeners about their research and projects. Find out the "root" of why they got into science and also learn about their other interests.
Art Woods 1:22
In this podcast, airing bi-weekly on Mondays, come learn how diverse the STEM field is, get inspired by the wonderful real life stories of Africans in STEM and hear from today's role models who're motivating the next generation of STEM enthusiasts.
Cameron Ghalambor 1:36
Listen to the show, wherever you get your podcasts. Visit the root of Science Podcast on LinkedIn, Facebook and YouTube or follow at root of sci pod on Instagram, Twitter and Tiktok.
Cameron Ghalambor 1:58
Hey Art, how's your receding hairline doing these days?
Art Woods 2:01
Receding? Ha. I think a better term would be fully receded.
Cameron Ghalambor 2:06
Oh, man, I know what you mean. I'm right there with you. You know, we could get rich if we could find a cure for baldness.
Art Woods 2:12
Hey, that's a good idea. And I recall reading recently that the androgen receptor gene on the X chromosome has been identified as the gene for male pattern baldness. I bet if we could come up with a CRISPR Cas 9 delivery system targeting that gene that we could regain the hair of our youth and become billionaires.
Cameron Ghalambor 2:30
Ah, not so fast. I think there's already a company doing that. And there are lots of other problems. First off, it's not just one gene, over 250 genetic variants are associated with male pattern baldness. And even considering all of them, we still can only explain about 40% of the variation in baldness. Second baldness is a continuously distributed trait from "bald" bald to just a bit of hair loss. And finally a 2018 study led by Chloe Yap found pleiotropic effects between the genes influencing male pattern baldness and other traits, like those associated with the onset of puberty. So messing with these genes might also have some unintended consequences.
Art Woods 3:13
Ugh get rich scheme immediately dashed. You know, it's super frustrating to hear stories in the news about a gene for this and a gene for that. It gives the impression that all traits are discrete and have simple genetic bases, like the way we teach introductory genetics: big A big A gives yellow peas and little a little a gives green peas. But most of the traits we care about and that biologists study are continuously distributed.
Cameron Ghalambor 3:37
This actually touches on a fundamental challenge that had to be resolved in the early days of evolutionary biology. Remember, Darwin and his contemporaries knew that traits were heritable, but they didn't know how. And this led to a big debate between two camps. The biometricians argued that most traits were continuous and that their inheritance could be understood as statistical relationships between parents and offspring. And the biometricians were opposed by the Mendelians who viewed traits as discrete and argued that inheritance followed rules governing the transmission of genes from parents to offspring.
Art Woods 4:12
This debate was eventually resolved in 1918 by Ronald Fisher, who demonstrated that continuous variation can be understood as the product of a large number of genes, each of which makes a small contribution to the trait. In other words, Fisher ended the debate by showing that at one end of the continuum are true Medelian traits that are controlled by a single locus and exhibit discrete variation, and at the other end are quantitative or polygenic traits controlled by large numbers of loci, things like human height.
Cameron Ghalambor 4:42
Fisher referred to this as the infinitesimal model, where an infinitely large number of genes each makes an infinitely small contribution to the trait. And while we know there are not an infinite number of genes, the rise of genome wide association studies has revealed that most complex traits are indeed influenced by many genes of small effect.
Art Woods 5:03
Thinking about complex traits as being continuously distributed and emerging from the influence of many genes falls under the umbrella of what's known as quantitative genetics. This field has been enormously important, serving as the foundation for agricultural breeding programs and guiding how evolutionary biologists study complex traits.
Cameron Ghalambor 5:22
Our guest today is Nick Barton, from the Institute of Science and Technology, Austria near Vienna. Nick is one of the world's leading evolutionary quantitative geneticists and has been at the forefront of developing and testing new theory.
Art Woods 5:34
We talk with Nick about a diverse set of questions from how the field of quantitative genetics has bridged the gap between strictly statistical approaches and the recent avalanche of DNA sequence data. We also talk about how genetic variation is maintained in populations and how natural selection acts on quantitative traits.
Cameron Ghalambor 5:52
Nick argues that the difficulties that biologists have finding the genetic basis of complex traits are actually not that surprising, once we accept that traits are controlled by many genes have small effect.
Art Woods 6:04
We also talk with Nick about his work on hybrid zones and the genetic mechanisms underlying reproductive isolation and speciation.
Cameron Ghalambor 6:11
I'm Cameron Ghalambor
Art Woods 6:12
And I'm Art Woods.
Cameron Ghalambor 6:13
And you're listening to Big Biology.
Cameron Ghalambor 6:27
Thanks so much for joining us today on Big Biology. We're really looking forward to talking to you about your research and perspectives on evolutionary biology. You've worked on some of the most fundamental questions in evolutionary biology and quantitative genetics. And I think today we'll only be focusing on a subset of all of your research. I first became familiar with your research when I was a graduate student. I read your annual review paper that you wrote with Michael Turelli back in 1989 that had this very provocative title, "Evolutionary quantitative genetics, how little do we know?" And as a starting graduate student, I'll have to confess that the paper was really intense. And I didn't probably understand all of the subtleties, but I was really taken, and I guess surprised by, how many fundamental questions or problems that I thought we knew the answer to, still remained unresolved. And I guess I'm just really curious, you know, now looking back 30 years later, what are your general thoughts on how much more, instead of how little, do we know? Compared to then. And are you pleasantly surprised or are you more disappointed, I guess?
Nick Barton 7:36
A bit of both, I have to say. So, on the one hand, we know vastly more about the genetic basis of quantitative traits. So in that article, the '89 article, was written just before the explosion of first QTL, studies where one could map quantitative trait loci, QTL, using the profusion of molecular markers that came with direct DNA-based technologies. And then when we had DNA sequencing, and we had even more markers, people moved on to GWAS, which is genome wide association studies, where one basically simply takes a big population, and tries to associate marker sets with traits. So this has had an enormous lot of funding in the last 20 years, I guess, from the biomedical community in human genetics, with the aim to really try and find the genetic basis of disease traits. And as a side effect, it's been enormously productive for people doing quantitative evolutionary genetics. So yes, we've found far more loci-affected traits that we believed would be found, I think, in the early days.
Nick Barton 8:43
On the other hand, one could say that's, with hindsight, not surprising, because we know that artificial selection works, and we know that, if you select on a population, you will move it many, many standard deviations way outside its original range. And that really can only be explained if there are many, many variants of small effect. And I think we've been astonished at just how many of those there are. So in some ways, it's confirmed what one could have deduced from the success of quantitative genetics. On the other hand, I think it's been very frustrating for the people funding these studies, because they started out doing QTL mapping, thinking there might be 10, 20, loci affecting your trait. And indeed, if you just do a crude mapping of which regions of chromosome will affect a trait, you find not too many, but that's simply the resolution of the technique. The more you work, the more you learn, and people move on to these huge GWAS, on the order of a million people enrolled and sequenced. Those are really drilling down and finding more and more loci of small effect. So that's frustrating because it means, of course, that you can't then do very much with a small effect loci. You know that you've explained maybe a few percent of trait variation, in terms of things you could actually imagine handling, dealing with in the lab, following up. And people always talk about candidate loci, and we have so many candidates loci, what can we do with them? They remain as candidates really. So, on the side of genetic basis of trait variation, we know much more, but we're also still quite frustrated, I think.
Art Woods 10:15
Huh it's interesting, you said the word frustrating and frustrated, I didn't think you're gonna say that I thought you were gonna say something, you know that that's just like a really beautiful result, we're finally sort of getting down to the nuts and bolts of where this genetic variation comes from. And does the frustration come from not being able to analyze or manipulate individual loci anymore, because there's simply too many of them? Is that what you mean?
Nick Barton 10:35
I think there are two aspects to this. The frustration on the part of the people doing most of this GWAS research, funding it, is that they can't find major effective causative alleles, which they can then follow up and you know, do drug design or personalized medicine or whatever. And maybe just following up on that, it's interesting to look at what happened in the animal breeding world, where it was realized pretty early on that you were not going to be able to find the gene for milk yield, and then you know, change the gene for milk yield and get better cows. But what was realized, really, by Mike Goddard, and others, was that actually, you could use DNA sequence data, not to do engineering, but to make a better statistical estimate of breeding value. So what you really wanted to do, particularly in the dairy industry, was decide which bull to breed from. And of course, this isn't obvious because bulls don't produce milk, but you have to infer this from the milk yield of the relatives. And you can do that more efficiently if you know the true relationships by looking at the amount of genome shared. So essentially, this is a statistical exercise where you give up any hope of identifying the actual causal alleles, but you still, in the dairy cattle case, you can double the rate of improvement of milk yield, you know, by a few percent per year, you're doubling the rate of response. And that's worth doing. But it's very far from, I think, what people had imagined when they set out.
Cameron Ghalambor 11:58
Yeah, so I think one area that's kind of related to this that has gotten a lot of attention in the past few years is really the shortcoming of GWAS. And so people talk about the missing heritability associated with traits. And so on the one hand, we have the sequencing technology that allows us to look at all of these markers, and yet our statistical models are still only explaining a fraction of that variation. What are your thoughts on the reasons for that?
Nick Barton 12:31
Okay, so I probably have a very, very boring old fashioned view this, which is that it's completely what you'd expect. And it's become clear that, actually, our statistical models work, it's just that we're dealing with a lot of small effects. In the first few years, when people coined the term missing heritability, they came up with explanations for it, which were involving epistasis, or epigenetics or who knows what. But that isn't really necessary. Because as we do GWAS on larger and larger samples to explain more and more variance, you can actually show pretty conclusively that something like human height, we can explain now most of the variance if we do a big enough study, and the remaining, the missing component is because of things like the SNPs that we are actually observing and not the causal alleles, they're linked to something causal. And if you allow for that, then you can explain the missing heritability. We know that there is of the order of 70, 80% of the variance in human height is genetic, and is mostly additive genetic. And it's actually quite astonishing that we can explain most of that if we work really hard, in terms of associations with markers we can identify. But that's we have tens of thousands of markers. So it's not a puzzle, it's something to be expected if we have a lot of small effect alleles.
Art Woods 13:47
So it's another way of saying that we thought there was missing heritability, but as the power of the studies grows, and we can pick up more alleles, of very small effect, then we sort of fill up and actually come to explain the total amount of variation that we think is there.
Nick Barton 14:01
Yes, exactly. And we're never going to get all of it. But we can get towards it, the more money we spend.
Art Woods 14:06
Yeah, no that makes sense. Let me ask another question about origins of variation, and this has to do with thinking about long genome reads and associating SNPs with traits, so the sort of basic bread and butter of GWAS studies. An increasing number of studies are also starting to look at big kind of structural issues in genome. So structural rearrangements, you know, transversions, chromosomal rearrangements and sort of big things like that, that are contributing, perhaps in some way to heritability of traits. So how is that? Is that accounting for some of the missing heritability? And is there does it require other statistical methods to account for those kinds of mutations?
Nick Barton 14:41
You know, I think it'll be the same statistical methods. These are just variants that segregate in a, essentially, Mendelian way unless we're talking about really huge chromosome rearrangements.
Art Woods 14:50
So SNP or chromosome rearrangement, they're sort of all the same at a statistical level.
Nick Barton 14:54
Well we're talking about small insertions, deletions, inversions, you know, they'll behave in the same way.
Art Woods 14:58
Gotcha.
Nick Barton 15:00
I think that it'll account for some but not a huge amount, because I think the existing SNPs will be tagging much of that. So when you see an association with a SNP or a cluster of SNPs, that you know, you don't know which if any of those is causal. If you do whole genome sequencing, get all the structural variations, then, in effect, you found everything that could be causal, and there's nothing mystical going on. But you don't know which of the variants is really causing it, there may be some combination in some complicated epistatic way, you just can't, you know, all you can do is do the statistics. So I think because the bulk of the variation, or at least more than half will be SNP variation, that would already have tagged the other variation. And, of course, most studies until recently, in human genetics have been using SNP panels, not the whole genome sequence variation. So people pushing towards whole genome, they will therefore pick up more, I think it's an open question, it'll depend on the trait, how much of that is structural, how much it is other stuff.
Cameron Ghalambor 16:01
So this also kind of touches on maybe some of the tension between quantitative genetics, and on the other hand, I don't know, maybe more functional genomics, or thinking maybe just more about mechanism, and the pathways that underlie traits. And so Art and I like to talk a lot about organisms and all the mechanisms that make up organisms and how they're integrated. And on one hand, it seems like the power of quantitative genetics is that you don't really have to worry about all that stuff, you can just assume that, you know, there are many loci of small effect. And, and yet, on the other hand, we have all of this increasing knowledge about biological pathways and patterns of gene expression at specific tissues, and how that underlies different phenotypes. So do you ever envision sort of a future where that kind of more functional perspective, and the more quantitative genetic perspective will merge, and those two sides will be able to talk to each other in a more sort of productive way? Is that too optimistic of a goal?
Nick Barton 17:11
Well, there's two aspects to that. One is sociological really, which is that there's a real gulf between the people who do quantitative genetics and their training, and the people who do functional genomics who really don't learn stuff on a statistical side, and vice versa. I mean, most of the people in the more ecological side of things don't learn so much functional genomics. What happens there is a matter of the education system and how students are trained. And there, I think that the divide has been growing, not shrinking. From the scientific point of view, I think there is a real question about the limits to what we can possibly infer, you know, because we're getting to those limits when we're, you know, things like UK Biobank, where a substantial fraction of the UK population is being, you know, measured from genotype. And there's a limit, you know, we're not going to be getting ancient DNA from all of the ancestors and so on. We have what we have, we have the whole population, whole genome sequence, and even given that it's not easy to determine the functional causes. People have been studying model systems like sticklebacks, and Heliconius, and so on, and really pin down on the functional genetics of a few loci. But it's just impractical to do that on a big scale. And it seems that even doing the sort of statistical systems biology type approach has limits. And Jonathan Pichard, Boyle et al on genetics really was quite provocative in suggesting that, you know, even when we have all the GWAS hits, so we really know which associations there are, we're not necessarily going to be able to identify the pathways, because it may be that most of the variation in a trait is mediated by the very weak effects, at very distant parts of the network. And then you could say, well- why can't we identify gene interactions? Well, that statistically is extremely hard, we can look at marginal effects, we can get to some extent dominance effects, but short of doing a lot of actually artificial crosses, and so on, which is limited in what we can do, it's very hard to identify beyond that pairwise interactions, there's simply too many of them. So it depends on the trait. In some systems, you know, I think the systems we study in ecological genetics, I mentioned the sticklebacks where there's an adaptation, freshwater versus marine environments. There, we know a lot about the traits that are adapting, we can drill into that. But in many cases, and particularly with things like human complex diseases, it may just be impractical to use these statistical approaches.
Art Woods 19:42
Let me ask a follow up that maybe builds on this line of reasoning and this follow up is about these ideas of epistasis and pleiotropy and what the implications are of having lots of loci of small effect for how we think about epistasis and pleiotropy. And just for listeners, epistasis is interaction effects between different loci and pleiotropy would be, you know, effects of individual loci on multiple traits simultaneously. I was really struck thinking about your arguments about, you know, there being hundreds or thousands of loci of small effect that contribute to individual traits. And the fact about what that implies about the pervasiveness of pleiotropy. And that is, there's only, you know, twenty or twenty-five thousand genes, say, in an average mammal. There's many more traits than that that we can think of, maybe not more, more than 25,000, but you know, many more than 20 or 30 traits. And so that implies that almost all of these loci are going to have some sort of pleiotropic effect on more than one trait. So does it mean that the genome is kind of a giant, tangled of pleiotropy? Is that what that implies?
Nick Barton 20:45
Then you wonder how it works at all, because it's such a contrast from the sort of biochemistry textbook, molecular genetics textbooks, where you have clearly defined pathways, and that was, in a way, that's because that's all we can discover, you know, with our methods. And you wonder, I mean, there's two levels to this question. One is what explains the statistical variation in traits we measure. And that doesn't necessarily undermine the sort of traditional classical view of how molecular biology works, because it could be that, yes, there's a few key genes interacting with each other to produce this substance to do this, to do that, to control this developmental stage. But then, all of the others are acting on the traits through general effects on health and metabolism and dynamism. And you know, there are all sorts of general things which will influence many traits. So height is an extreme example, because that's something that's influenced by so many things all the way through development. Of course, if you drill down to maybe fingerprint ridge counts, I'm not sure anyone's ever really worked on them seriously, but those may be affected in a much more simple way. So that's one level. But then at another level, you wonder whether gene regulation really is as simple as these pathway diagrams. And what's done in another group here at IST, which is looking simply at the classical bacterial promoter, which is supposed to be one or two binding sites for a polymerase and a transcription factor, and so on. And it turns out that there are multiple weak binding sites. And the problem is that it takes a lot of work to get into these weaker binding sites. And I think we don't even know theoretically, whether if you're evolving an organism, is there a reason why it would evolve simple pathways that we can understand? Or is it that we have an observer bias, and we tend to miss out on a lot of very diffuse functions, which is really important in aggregate, and yet, which is hard to observe?
Art Woods 22:39
And so the idea is that many of these effects are so small that experimentally, we'll never be able to pick them up?
Nick Barton 22:45
That would be, yeah, if I wanted to be provocative, that would be true, yes. Or we might need a whole different mindset to know how to deal with it, not doing it one by one. You know, I look at, you know, the astonishing activity, molecular biology is building up, an enormous amount of information, gene by gene, pathway by pathway. And you won't know how to put it all together, how do you know when you've understood a system? Because you're never going to be able to put it into a simulation and turn the handle and get an organism out of it, that's not going to happen. But in what sense do you understand things when you have a catalog of detailed information, but you want to know how does this fit together?
Cameron Ghalambor 23:20
I think that's something that a lot of biologists and I think former guests on this show, struggled with quite a bit. And I think that also is some of the motivation behind, you know, a certain group of biologists that would sort of argue that we need this new, a new synthesis, that the traditional theory has limits, and we're not incorporating other levels of complexity, either molecular mechanisms or environmental effects. And I'm curious, what do you think about that? You know, given that we can't just take all that data into our standard evolutionary theory and turn the handle and come out with beautiful answers?
Nick Barton 24:00
Yeah, I must admit, I find it a bit irritating this whole extended synthesis business, because firstly, it's most popular amongst people, I wouldn't say outside biology, but doing history of biology, philosophy of biology. It's really quite a minority of working biologists who are kind of getting on trying to do stuff. And that's not to say that, of course, philosophy doesn't have something to say about how we do our business and whether we could do it better, and so on, but it's not that one has the sense of a field in crisis, because we've missing the missing heritability or whatever. I mean, I think we've worked towards explaining mostly that's successful. But also, I think a lot of phenomena that they're excited about, like epigenetics, like horizontal gene transfer, transposable elements and so on, are actually really interesting to working evolutionary biologists and we do a lot of work on it. And we have models that can explain how selection works across different levels. I mean, transposable elements have had a lot of attention, and have led to, I think, theoretical advances in the way we understand how organisms can coexist with selfish genetic elements, without one or other going extinct. And how it is that a lot of the molecular machinery we look at is actually all about trying to suppress transposable elements. And that's a big area, both on the functional side and on the evolutionary side. And, of course, the standard population genetics textbooks don't include that, because you're trying to teach students without driving them mad. But there are really solid theoretical understandings of our model systems. We really understand how this stuff works.
Art Woods 25:44
I wanted to ask a question about what the consequences are of this sort of view of many, many loci of small effect, what the consequences of that are for how we think about the response of populations to selection, and whether fitness and population changes smoothly or in some sort of, you know, jumpy, non-continuous way. And I think, like, one of the implications of having variation at many loci of small effect is that there's just a lot of standing genetic variation for a lot of traits. And that that could be, that could facilitate a very smooth movement of a population potentially for very many generations. And the opposing way of thinking about it is that populations, in some way, are limited by the origin of relevant mutation, relevant variation by the process of mutation itself. And what I'm thinking about is some of these older studies by Rich Lenski on evolution of populations of E. coli, they saw in some of their experiments, you know, long periods of stasis, and then a jump up in fitness of the population when some new mutation arose and swept very rapidly to fixation. So those seem like two very contrasting ways of thinking about the response of populations to selection, and how do you reconcile those two things?
Nick Barton 26:58
Okay, so it's interesting, I think it's really important to be clear about the difference between asexual and sexual selection, or sexual reproduction anyway. So the Lenski experiments were essentially asexual, and asexual populations, you know, respond to selection, but much more slowly, and they require a much bigger population size. So of course, bacteria have been immensely successful, and they've been around longer than we have, much longer than we have. But actually, when you look at the population genetics of bacteria, they look very different, I think, from these lab experimental evolution situations, because the population sizes are enormous. And there is a significant effect of recombination. And actually, in the long term, you could argue that bacteria have been successful because they've been able to exchange genes by kind of irregular means. And then you look at eukaryotes, which almost always share meiotic sexual reproduction more or less regularly, I mean, if not every generation every so often. And that, one can argue, allows them to use variation and generate variation efficiently, even with a much smaller population size, it allows them to be bigger, allows them to be more settled in the end. And so the smoothness of response to artificial selection in populations of a few hundred, or a few thousand really comes because there's this continual generation of variation by recombination whenever sexual organisms cross.
Nick Barton 27:33
And I find most striking wonderful experiments by Ken Weber in the 90s. And Ken Weber, you know, worked with selection experiments on Drosophila. And he realized that until then, people had never done selection experiments, or more than a few hundred flies, because it drove you may have tried to count hundreds of flies every generation. And so he devised machines like the inebriometer, which was a kind of distillation column, you put flies in and they'd go up until the ethanol vapor was concentrated enough, they'd fall over and so the ones resistant to ethanol will get further up the column. And he had a wind tunnel experiment where you could flies would fly towards the wind and then they'd fall out. So he could select on populations of 10,000. And when you did that, you find astonishing replicability. So you put in, you know, a few thousand flies, you select on them, and you do two independent replicates. And they do the same thing. It's very, very reproducible. And that's really reflecting the sort of store of variance in a sexual population. And it really is an argument for the efficiency of sexual reproduction. We still don't completely know why sex is favored. I think the consensus is that it's favored because it allows selection type sufficiently.
Cameron Ghalambor 28:21
I'd like to kind of follow up on that though, because even in sexual populations, a more population genetic perspective emphasizes a beneficial mutation that then is subjected to a selective sweep that sweeps through the population and increases in frequency. And so even if we survey the kinds of papers that are being published today, on one hand, I, you know, I see these genome scab type up approaches that find evidence for one or two loci that have undergone a selective sweep. And then at the same time, you know, we have all these GWAS studies that are telling us, there are many loci, each one of very small effect. And I'm curious, how do we how do we reconcile those two perspectives?
Nick Barton 30:17
Well, I think we aren't reconciling them at the moment, because it's different people studying different kinds of genetic architecture. So, you know, I can talk about my favorite organism at the moment, the snapdragon, Antirrhinum, which we've been studying populations in the Pyrenees for a long time, where there's a narrow hybrid zone, if you'd like,between populations with magenta flowers, populations with the yellow flowers. And if you do a genome scan, in other words, if you compare sequences between the magenta and the yellow population, and you look for places where there's a sharp transition, or there's an excess divergence relative to diversity, various statistics, you scan along the genome, these statistics pop up, and they show spikes, and we find on the order of ten significant spikes, and they're all flower color genes. So it seems things like flower color, wing pattern of butterflies have a rather simple genetic basis, and we can identify the loci and we can study them, that's all great. But I'm always left frustrated by going to this space and thinking, well, there's a lot of other stuff going on, it's not just flower color. You know, they're adapting to drought, tolerance, shade versus sunlight, high altitude, all kinds of things going on. And we know there's a lot of local adaptation. And a lot of that may just not be showing up in the genome. And people are actually really trying hard at the moment and not yet succeeding in finding statistical ways to detect this kind of polygenic adaptation. We know how to find classical selective sweeps, barriers, which separates hybridizing populations, we can see those and most of the work in ecological genetics is really going on those. But it may not be representative.
Cameron Ghalambor 31:58
Do you think it's partly then I don't want to say maybe biased, but the types of traits we're attracted to in terms of studying. So you mentioned flower colors as an example, so that's something very conspicuous. And so you know, it draws our attention, and perhaps because of that dichotomy, you might have a more sort of Mendelian-like genetic basis, versus something that is more gradual and continuous in its variation that maybe doesn't capture our attention.
Nick Barton 32:30
Yeah, I mean, maybe I'm being a little bit unfair, because I think the flower color example is at one end of an extreme. And if you look at something like the cichlid fish, or the sticklebacks, where there's a huge diversity, many, many different traits in mate preference, in pigmentation, in morphology, in feeding behavior, and so on. So there's clearly a lot of things going on there. And there the issue is perhaps more that there's a bias towards focusing on the things we can study, and so you end up studying the pigmentation difference, and the more polygenic things get left behind. And even if we can map them and do GWAS and find candidate genes, it will be very hard, I mean, in these natural systems to really follow up on those and work out what's going on.
Art Woods 33:15
I had one additional question about population sizes. And this is kind of rewinding about five or seven minutes in the conversation. And it's thinking still about the differences between things like E. coli, which have, you know, orders of magnitude more larger population sizes than do most mammals, including humans. And the effect of that on the patterns of variation in the genome, and E. coli also has many fewer genes than we do it. And I'm going to sort of venture out into a space where I may reveal my profoundly naive way of thinking about population genetics here. But isn't it the case that when you have very large population sizes, that selection can essentially discern smaller and smaller differences among the fitness of individual variants? And shouldn't that mean that there's sort of more purifying selection across the genome in bacteria than there is in things like humans? And so is that a reason why you might see sort of more effective the process of origination of new variants by mutation in a bacterial population, than you do in human populations?
Nick Barton 34:25
I mean, it's true definitely, in principle, that in a large population, very weak selection pressures can be effective, although they take a long time to be effectively many, many generations. And there's an argument that you know, the shared biochemical machinery that we have, which is incredibly sophisticated, the ribosome and so on, and all these other molecular machines, that those were perfected or optimized way, way back in perhaps very large bacterial populations, and essentially all organisms, you know, share these things. But it's hard to get back to what were the selection coefficients involved in substituting this or that sequence in the life cycle of DNA. But on the other hand, I think bacteria they;re more interesting than we imagine from lab population. We don't really know, it's kind of stunning how ignorant we are about where they live outdoors. How does E Coli get from one mammalian host to another? Does it have, you know, adaptations to getting from one mammal to another? Or is it just chance? So there are yeast that people study as a eukaryote, but a microbe that lives outdoors, and they don't really know what it's doing, you know, it's not doing the kind of thing we make it do when we domesticate it. And so this sort of rather rich ecology that bacteria experience may mean, of course, that selection isn't, you know, simple, and that, in fact, you know, one paradox is that if it were the case that single mutations were sweeping through bacterial populations and just fixing, then that would eliminate diversity. So actually, the diversity in bacteria may need an element of ecological heterogeneity to be maintained.
Art Woods 35:58
Okay. So in other words, this is a contrast between the way people typically grow microbes in the lab, under very constant conditions, well-controlled versus the complexity of the real world, and the fluctuating selection pressures that may arise there from.
Nick Barton 36:12
Yeah
Art Woods 36:12
Yeah okay.
Nick Barton 36:13
So, you know, I think I would see, the experimental evolution in microbes, which is wonderful and really instructive, as almost in the same vein as simulating, because you're controlling things very precisely, you're learning new things about the organisms and about the processes, but there's a lot more out in the real world.
Cameron Ghalambor 36:30
Yeah, so that point about ecological heterogeneity, I think is a good segue into thinking about, more generally about environmental effects. And that's something that I'm very interested in in my own research is the role that phenotypic plasticity plays in the evolutionary process. And so, you know, in reading a lot of your papers, you know, I may have missed it, but I haven't seen a lot of papers where you actually talk about what role environmental variation potentially plays, because when we think about sort of many loci of small effect, polygenic traits are also notoriously sensitive to to the environment in terms of the phenotypic expression. And so do you have thoughts on, you know, within a fluctuating environment, and this kind of environmental heterogeneity where the same genotype can produce different kinds of phenotypes, what world do you see that playing in terms of this kind of polygenic selection?
Nick Barton 37:28
Yeah, so I haven't written very much about it, because it's hard enough doing the simple things, I find, so. And I suppose the real question is not so much dealing with the fact that alleles have effects that are conditional on the environment. And there's a G by E interaction, but rather, the extent to which that plasticity or lack of plasticity is adaptive. So on the one hand, you know, it's astonishing that a particular DNA sequence can generate a developmental program that's extremely robust to the environment, so that the effect of things acting in development can be, you know, somewhat independent of biometric conditions. And on the other hand, it may be, you know, that these examples of cycles, examples where the whole phenotype can switch to anything you want that's clearly adaptive. And so in between, the question is, to what extent is the degree of plasticity shaped by selection for either more variance or less variance? And how clever can organisms be in anticipating environmental variations? Because in some ways, they're having to predict from the environment they're experiencing when they're young or as a parent to say, what will be the optimal phenotypes for producing the next generation or the next time step? And I think people vary, it's a psychological thing really, what people vary a lot in how clever, if you like, they think organisms can be in adapting to complicated environmental circumstances. And of course, the answer to that isn't a theoretical answer. It depends also on how predictable the environment is. And what's astonishing to me is that organisms can survive in very unpredictable environments.
Cameron Ghalambor 39:04
Yeah, I guess, you know, for me, I've been kind of disillusioned a little bit by some of the terminology that gets used with regard to plasticity, so genetic assimilation, genetic accommodation. And even though I'm very interested in plasticity, I've stopped using those terms, because it doesn't seem like anybody really agrees on what those things actually mean. And instead, I've kind of been thinking just more about how does the environment alter variation, heritable variation. So we know that because of the plasticity that we were just talking about that that can change the kind of variants that selection sees, but also kind of to your point about how adaptive the plasticity is, the environment can also shift the distribution of phenotypes closer to a new optimum, and that kind of adaptive plasticity should weaken selection. But it can also shift them further away and make selection much stronger. And, in which case, either the population has to rapidly evolve or it goes extinct. And I feel like that's less controversial to kind of put it in those terms than to kind of describe it as: is this genetic assimilation? Is this genetic accommodation? I don't know what those terms mean anymore, so.
Nick Barton 40:17
I mean, we get hung up on these terms, and it's true across the whole of biology. I've just been to a conference on speciation, and, you know, people don't agree on what reproductive isolation exactly means. And in some ways, you know, really arguing through that allows you to understand the subtleties of processes better. On the other hand, it makes dealing with the literature very difficult, because no one's using words in the same way. And I'm not sure what the solution is, because you could say, well, we just have to specify the model. But then, you know, people vary in how quantitative they are. An actually, even if you're quite mathematical, then a model needs a narrative. This was a nice article by Sally Otto, a few years ago, pointing out that the importance of models in biology was really the narrative that links them to the questions we're interested in. But having said that, I don't know how one deals with questions like plasticity, where there are people from very different angles coming in and trying to address the same problem.
Art Woods 41:28
I think it might be a good time to do some things on hybrid zones and speciation Cam, is that, you okay going there?
Cameron Ghalambor 41:33
Yeah. Yeah, we talked a little bit about reproductive isolation, so that's, I think, a good segue
Art Woods 41:38
For much of your career, a good chunk of it, you've worked on a set of things that involve some really interesting evolutionary genetics, in the areas of speciation and hybrid zones, and that is sort of understanding how does speciation happen, and what happens when populations or species come into contact, and there's some reproduction happening in the contact zone and exchange of genes and introgression across those zones. So maybe just, first give us just a little bit of a historical overview of how your thinking has changed about those processes over the past, say 30 years or so. And, you know, what all of this revolution in our ability to see variation has done for thinking about speciation and hybridization.
Nick Barton 42:22
So maybe I'll just come across as being very stuck in my ways by saying, I don't think my thinking has changed that much. In that, you know, I started out working on hybrid zones, and a whole range of things, grasshoppers to start with, and then butterflies, toads and so on. And all of these, you know, I started out trying to understand this very striking phenomenon where you get this very sharp interface between one population as well, but these will be very different populations. And the most divergent I ever worked on was Bombina bombina, this variegata toad named species adapted to very different habitats. And yet these two species of toad meet in a hybrid zone around the Danube basin that's about six kilometers wide. So really very striking. Okay, so you're trying to understand this, and you realize that actually, this was apparent. You know, in the old days of electrophoresis, you didn't need a huge genetic revolution to realize that actually, just because they were producing hybrids and interbreeding, there was a fair exchange of genes. And so at a locus that wasn't maintained different by selection, there wouldn't be a dramatic isolation between these, reproductive isolation, even across these very strikingly distinct forms, is not that strong. And I think what we've discovered with DNA sequencing has been that when you trace deep genealogies, you find that there's a lot of discordance between sister taxa, and that this implies some kind of introgression, even in situations where no one would really have suspected it. So Drosophila persimilis and subobscura, which Mohamed Noor and Jody Hey and others studied way back. Turns out that most of the genome is more or less well mixed, and that the distinction between these two, apparently quite clear species, which very rarely hybridize, less than one in a few 1000 in nature, that's carried by a set of inversions. So part of the genome is reproductively isolated, and most of it isn't. You know, it's very widespread, and people talk about speciation with gene flow, and it's not at all difficult for speciation to happen with gene flow and for gene flow to continue for a long time after you form distinct species.
Art Woods 44:23
So when you were saying that large portions of those genomes of the two Drosophila species are well mixed, that means that there's sort of ongoing hybridization between them, and those sections of the genome act almost as if it's just one giant species.
Nick Barton 44:37
Yeah, yeah. I mean, we know from just looking at the flies that mostly they don't hybridize and the rate of hybridization is very low, one in a few thousand. But from a population genetic point of view, what matters is the effective population size times the rate of hybridization N times n. And since the effective size is, you know, hundreds of thousands of millions, then Nn is large. And so if you trace back the genealogies, over most of the genome, you're going back hundreds of thousands of generations. And over that timescale, they're pretty much well mixed. So that they're, you wouldn't be able to see the proper species phylogeny, if you looked at most of the genes and looked at their ancestor.
Art Woods 45:16
Yeah gotcha yeah
Cameron Ghalambor 45:18
But that's not a perspective that you had going into studying hybrid zones. That's something that must have come out more recently, because the sort of classic way that we learn about speciation is, you know, allopatric, speciation is easy, sympatric speciation with gene flow is really hard. And, you know, I would imagine that's still kind of common in most evolutionary biology textbooks.
Nick Barton 45:40
Yeah, it is true. But I think I got a different perspective through working on hybrid zones in, more or less spatially continuous populations, in parapatry. And realizing that, you know, divergence with gene flow is perfectly straightforward, if you have a big, big range. So I was working on originally, grasshoppers. We've had two different chromosome arrangements in different places, and the interface between those was about a kilometer wide, relatively narrow. And that's, you know, you don't need much selection to do that you need about, you know, half a percent selection will produce this sort of interface. So it's very striking, you know, when you look at the data, but you realize that it just doesn't affect most of the genome, you know, it's a very, there are other things distinguishing these grasshopper taxa, but they still don't really impede the flow of genes. So you can have really striking mosaic patterns, striking divergence. And yet, most of the genes are able to mix and in particular, genes that are advantageous everywhere will just arise here, and they'll spread.
Art Woods 46:38
So what are the loci in these, say, inverted regions that associate with populations or with species? Are there particular sorts of subsets of the loci that account for those differences?
Nick Barton 46:51
Well, I suppose I have to admit that I stopped working on grasshoppers and toads, because they have huge genomes, you can't do the genomics. So, you know, we still don't know in those systems. And now in the snapdragons, of course, we do know that the loci that are clearly maintained distinct, which is a dozen or so loci, those are flower color genes. And the collaborators in Norwich who get at the molecular genetics are digging into, you know, finding out how they affect gene regulation, how they actually affect the distribution of the pigment, or the flower, and so on. Of course, there's a lot of interest, in general, in what it is that generates reproductive isolation, and a lot of interest, more interest now than they used to be, in so called ecological speciation, where you focus on things selected to do different things in different habitats, like the freshwater versus marine in sticklebacks, and so on. There's also a lot of stuff coming out in Drosophila, especially on the role of selfish elements of coevolution, between driving elements and suppressors of drive, which and that arms race plays out differently in different places, and then you can get incompatibilities. So my take on it would just be that a lot of evolution, a lot of selection is happening everywhere, all the time driven by internal arms races, by adaptation to changing environments. And as a side effect of that, you get reproductive isolation. It's not that there are special speciation genes.
Cameron Ghalambor 48:12
Yeah, that's interesting, though, because it seemed like there was a movement to, you know, I saw lots of papers being published, you know, where there was a sort of a search for these speciation genes that are involved in reproductive isolation. Has that sort of line of research kind of died out now a little bit? Are we thinking it's a little more complicated?
Nick Barton 48:33
No, no, it's very active. Yeah, it's very active. And I think, I must admit, when I was starting out, I was obsessed by wanting to count the speciation genes. I didn't necessarily think they'd be special, but I want to know were there ten or a hundred or a thousand, you know, separating these taxa. And that was completely unclear before we had sequenced data, you know, and now it turns out to be a large number. But I think people hope that they will get some kind of consensus view of what kind of genetic differences are separating species, and therefore what kind of processes drive it. Is it sexual selection? Are these mostly mate preference genes? Is it selfish elements? And so you can get some sense of that. At the moment, I think we're still at a stage of accumulating, you know, stories about different systems, and it's hard to generalize, so maybe there won't be a general answer, it will depend on the organism, it will depend on Drosophila. one thing, mammals, another thing.
Cameron Ghalambor 48:33
Do you think that, or at least maybe do you have an impression, if the genes that are associated with this reproductive isolation are confined to certain functions related to say, specifically having to do with like reproduction or sperm binding with eggs? Because it strikes me that like selfish elements can pop up anywhere and they can just kind of build up, but they may or may not necessarily affect a specific kind of function that would be important for actual reproduction.
Nick Barton 50:01
When we found out what these genes do, it's been kind of idiosyncratic and perhaps surprising. So I think the first speciation gene, or one of the first identified was, Daven Presgraves discovered that there was a gene in the nuclear pore, which is responsible for transporting RNA through the nuclear envelope. And that was causing an incompatibility between two sister Drosophila species. Now, you'd think that this kind of nuclear pore is doing an absolutely crucial conserved function, why would it change? And I'm not sure it's been sorted out yet, but it may be because of some sort of coevolution with a viral pathogen, which is getting in through this mechanism. So this is something one couldn't have predicted. You know, I don't know whether we'll eventually be able to put together enough of these stories about enough systems to really understand what's going on.
Cameron Ghalambor 50:50
Yeah, maybe before we leave this topic, we mentioned the sort of challenges of defining reproductive isolation. And you have this recent paper with Anja Westram, where you sort of defined what, or tried to define what you thought was reproductive isolation. And there were a large number of commentaries that sort of chimed in. And, again,here's a term that, you know, I've heard millions of times, and I thought I knew what it meant. Why is reproductive isolation, so difficult to define?
Nick Barton 51:23
Yeah, I sort of had the same reaction, because before we got into writing this, I thought I knew what it meant. And then you start thinking about it more, and you realize it's subtler. And now I also now think I know what it means, but at a different levels. So one reason why I think there's a difference of opinion, and why those commentaries were coming at different angles is that people study it in different ways, using different techniques. So one is people who go and they just look at organisms, they maybe make some crosses, they say, what's the fitness for the F1 and the F2. And they think of reproductive isolation in terms of hybrid fitness, thinking of the first few generations of hybrids. And then other people at the other extreme, try and make inferences from genealogies and ask do the genealogies of different loci stay together, implying some sort of isolation? So looking at a very long timescale, and they do these so-called IM models, isolation-with-migration models, to take the genetic data and say, we can infer the rate of gene flow. Of course, the rates of gene flow there are very, very low, as I was mentioning earlier, you know, it only takes a few exchanges of a few genes, few hybridization events, per generation to produce a lot of homogenization. So they're looking at different timescales, looking at different phenomena. And then I was coming in from somewhere in the middle, working on hybrid zones, where you have a cline, and you can estimate the rate flow genes, one side to the other, over timescales of a few hundred generations. So we were trying to come up with that paper with a definition of reproductive isolation, which was basically that it's the reduction in gene flow, due to genetic differences, as opposed to geographic differences and so on. So we came up with a definition, which we thought applied to all of those three levels. But when you tried to apply it in specific cases, it turns out there are subtleties and pinning it down to a number depends on the context in some circumstances. So I think there is a single concept, reproductive isolation, but people are studying it over at least three different timescales with different sets of techniques.
Cameron Ghalambor 53:22
And so does that mean we have to be a little bit more sort of flexible about the definition, depending on the timescale and context that we're looking at?
Nick Barton 53:32
Yeah, it's difficult, because I don't want it to sound as though it's all completely vague, and you come up with your own definition. I think there is a verbal definition. But it's always true that when you try and make a measurement and quantify something, you have to take into account the context. Is it that there are two genes exchanging at equilibrium? Is it that there's a sort of a continuous cline? Is it that there are genes coming and going over a long period? I mean, all of these things give you a different framework. And you have to then measure the reduction in gene flow and the extent to which that is reduced by genetic differences. It will come out differently. There's no one magic formula, which is right or wrong, I think.
Art Woods 54:08
Well, I wanted to turn at this point to a couple of wrap-up questions sort of getting to the end. And I want to zoom back out a little bit and ask a practical question first. So I would say in biology and in the media, we often hear that modern progress in genetics is going to result in all kinds of miraculous things, including using one's individual genetic profile to get good personalized medicine, cure genetic diseases through gene editing, develop better crops, better animals, and maybe even bring back extinct species. There has been this big splash in the news recently about the dodo. And I guess, given all this that we've talked about, do you share this excitement for these kinds of practical applications? Or does it still seem like a, you know, a distant dream in some way that may never be quite realized?
Nick Barton 54:56
Yeah. I mean, I do this stuff because I'm excited about the basic science and understanding how the world works and understanding how biology works. And I think that's actually important in itself. One shouldn't forget that, because I think it gives us a perspective on where we are in the world, in that, you know, we're on a small and fragile planet, and we shouldn't be messing it up in the way that we are. And I think having an evolutionary perspective makes you realize the richness of diversity and the very fragile and extended process that has led to the accumulation of the diversity we see. So I would say that actually, in terms of the practical problems facing the world, there are very simple things we know how to do-not chopping down the forests, you know, that preserves biodiversity. And so I'm a little bit skeptical about very elaborate technological ways of solving medical or conservation-type problems, because they're often a distraction from actually really doing what we know we ought to do. But, you know, thinking more specifically, then about what the kind of statistical population quantitative genetics, I believe you know, has done, it has actually been useful in animal breeding. And a lot of the stuff actually we do in the basic science came out of statistical methods developed way back in the 50s, and now applied using sequence technology and that really does improve crop yields, and so on. And in human genetics, you know, the sort of GWAS studies are useful. I don't think they're gonna be useful in personalized medicine. I mean, I would not want to go to the doctor and be faced with this enormous spreadsheet of my probability of X, Y, and Z and what do I do with it?
Art Woods 56:28
Ok, here's your GWAS score. And-
Nick Barton 56:30
Yeah exactly. I mean, I know that I should, you know, eat less butter and get more exercise, but whatever. This sort of thing, you know, can be very useful, first of all, in fine understanding heterogeneity in response to drugs, and really interesting cancer genetics, where you can try and produce treatments for particular cancers based on their particular sequence, and so on. And also in optimizing screening programs, you know, it may well be worth, you know, it's not useful to know that this person, this individual has a 50%, or 60, or 70% chance of whatever. But it's useful at a population that will say we should focus our efforts on these people and not these people/ Of course, that requires an economic to the health care system, which is not what we have, at least in the US.
Cameron Ghalambor 57:18
I'm very curious about your path in evolutionary biology, because, you know, people who I interact with are either sort of like empiricists that are out sort of collecting data and analyzing it. There are theoreticians who maybe don't do any empirical work and think about a lot of the models. And there may be sort of computational people who try to work on methods to sort of analyze the data and help test the theory. But I think one thing that really strikes me about your research program is that you've span all of those different kinds of approaches, at least within evolutionary biology. And I'm curious if that was a conscious path that you took that you felt like you needed to touch on all these sorts of different aspects? Or was it more serendipitous? Did it just kind of come about by chance? Or how did you come to do the things that you do?
Nick Barton 58:13
Yeah, so I think I ended up on this path, thanks in large part to a degree in Cambridge, which is still running called Natural Sciences, which is, you know, a degree where you have to study all the sciences. So you go in, I was going to do physics, and then I realized biology was kind of more interesting. I probably didn't do too much physics at school, and then there will this new biology look really amazing. So I sort of switched into biology, but I had the sort of mix training. And there are so many places that do that. The other was that I went out and I started studying this system in the out with the team in with the grasshopper hybrid zone, with Godfrey Hewitt, because it was in the Alps. And it seemed so much more exciting than something in a lab in Cambridge. And then I got there and you see this really striking change. And when you get out there you map, you know, we had a field microscope and you go out, you catch some grasshoppers and measure the frequency of the two chromosome types, and you plot it out. And you see there's a nice sigmoid cline, which is explained with a very simple theoretical model. And so I got into really using the theory to explain, you know, the patterns that I was seeing and the patterns you could see in other data. And that's sort of inspired me throughout. So I published, probably much more theory, but it's often been inspired by projects, you know, field work or real data. And I think it's important to sort of always remember where the data is coming from, what you can measure, what you can't measure, what's causal, and what isn't causal.
Cameron Ghalambor 59:39
Yeah, so we usually like to wrap up and give you the opportunity to cover any material or answer anything that you feel like we didn't quite touch on.
Nick Barton 59:50
I was thinking more about the last question you're asking and thinking about how it is for people coming into the field now. And in some ways, there's quite a diversity of people within evolutionary genetics anyway coming from bioinformatics and computer science, and straight biology, but it's also a field that tends to get a bit fragmented. And it would be really nice if people could see the connections. And if there could be more kind of programs where people could talk to a diverse range of disciplines, rather than separating into their own little specialties.
Cameron Ghalambor 1:00:23
Do you? Would you have any advice for graduate students who might aspire to do that? And how to go about getting that kind of exposure?
Nick Barton 1:00:31
Well, I mean, I could, I should put in an advertisement for our IST Austria graduate program, which is just taking in applications at the moment. But anyway, it is an unusual place in the European scene in that it takes students with a whole range of backgrounds and they can move across fields. But that is true also in the US system. And US graduate programs are, I think, more diverse, and flexible, simply because they give students more time to actually find their own project and find their niche. And so it's maybe a fee; for the space that's for people coming into the field to find where they want to go. Which direction to take.
Cameron Ghalambor 1:01:09
Yeah
Art Woods 1:01:10
Yeah, that seems really important.
Cameron Ghalambor 1:01:11
Great.
Art Woods 1:01:12
All right. Well, thanks so much, Nick, really fun conversation.
Nick Barton 1:01:15
Ok thank you.
Cameron Ghalambor 1:01:29
Thanks for listening to this episode. If you liked what you hear, let us know via Twitter, Facebook, Instagram, or leave a review wherever you get your podcasts. And if you don't like what you hear well, we'd love to know that too. All feedback is good feedback.
Art Woods 1:01:43
Thanks to Steve Lane who manages the website and Ruth Demree for producing the episode.
Cameron Ghalambor 1:01:47
Thanks also two interns Dayna De La Cruz and Kyle Smith for helping produce this episode. Keating Shahmehri produces our awesome cover art.
Art Woods 1:01:56
Thanks to the College of Public Health at the University of South Florida, the College of Humanities and Sciences at the University of Montana, and the National Science Foundation for support.
Cameron Ghalambor 1:02:05
Music on the episode is from Podington Bear and Tieren Costello