 So we're moving out to last talk of the session. Joanna Kelly, she is an associate professor at Washington State University who studies adaptation to extreme environments. And she will be talking about variation in natural systems and how to disentangle complex phenotypes. Thank you all. And it's a pleasure to be here. I really appreciate the invitation. And I am looking forward to hopefully talking to you about my work if we can get this to load. So I leverage natural systems to understand complex phenotypes. And the question that really drives my research and that I think a lot about, and maybe some of you think about this as well, is how do organisms diverge and adapt to the wide range of environments that they encounter? And so here I've just picked pictures of some of my favorite environments. Hopefully you're seeing that it's most environments. And all of them have organisms that have adapted to them. And you can see very organismal, rich jungles, maybe a desert that has fewer organisms. But I think about how do organisms do what they do where they do it? And how did they get there? And so to understand how organisms function in the context of their environment, I really want to think about all of the different hierarchical levels of organization, starting from genetic variation, thinking about how that influences biochemical function, how that then translates to physiological function, organismal performance, which then leads to changes maybe in survival, growth, roommate production, and ultimately fitness. And so it's important to think about all of the different levels of organization and how those maybe phenotypes of interest to each of us. And which phenotype is actually of interest for the question that you're asking about? And I'll return to this because I thought this is a really good framework and a longstanding framework for thinking about genetic variation to fitness. And in terms of the natural environment as well. And we'll throw that in there in a minute. So questions that I ask in terms of comparative genomics in this framework are, how predictable is the evolutionary process? Are there particular genomic architectures that predispose organisms to function better in specific environments? So are there genomic architectures that are somehow facilitating specific clades to expand into specific environments? I'm trying to be vague here. And another one that gets more specific is which gene regulatory networks or components of those networks are possible targets for change and which ones are highly constrained. And I think this relates really nicely to what Emma was talking about in terms of constraint when you're thinking about these very constrained enhancer regions across entire domains versus what things can we tinker with to make subtle or dramatic changes that ultimately influence phenotype, whatever phenotype you're interested in. So I use organisms in extreme environments as the framework for studying evolutionary processes. And the reason that I've turned from actually, my training was in human genomics, from that to extreme environments is that extreme environments are often replicated. So if you think about, I didn't show a desert here because I don't work on deserts, but as you think about deserts, oh, there are lots of deserts. They're also polar environments and alpine environments. And so in addition to being potentially replicated, extreme environments often have strong and constant selective pressures. And the fact that there are strong and constant selective pressures, that allows us to make predictions about how the genome and phenotype may change when organisms adapt to those environments. And along with that, those extreme environments often lead to similar phenotypes across basclades or across different taxonomic levels. And they also sometimes, although not always, provide an opportunity to study closely related, non-adaptive populations. So we can have a comparative framework. That's not always the case. The midges, which are in Antarctica, are fascinating. But they're tens of millions of years separated from the closest thing. So they don't always provide the opportunity to study a closely related non-adaptive population, but often they do. And so the extreme environments that I think about in my lab and the organisms that inhabit them are cold temperature. And I think about fish living in polar seas. Think about bears and hibernation and ice worm that are glacial obligate ice worms. And I also think about hydrogen sulfide as the extreme environment and fish that live in hydrogen sulfide springs. So today, I want to tell you a bit about hydrogen sulfide, but just as the framework for thinking about studying phenotypes and genotypes across these different levels of biological organization. So hydrogen sulfide, if you don't know it, is a very strong and constant selective pressure. It's extreme. In the regions that we're studying, hydrogen sulfide is naturally occurring because of volcanic activity. You can also have volcanic activity as a byproduct of gas extraction of many industrial processes. So keep that in mind. It's acutely toxic in micromolar concentrations. And so here are some ranges, 5 to 40 micromolar. And concentrations in these springs that we're studying can reach up to 1,000 micromolar. That's just to give you a sense of scale. Hydrogen sulfide inhibits oxygen transport and cellular respiration, causes and aggravates hypoxia in aquatic environments. And oxygen is needed for detoxification. So there are two things going on here. There's hydrogen sulfide as a toxicant, as well as hypoxia, because hydrogen sulfide is reacting with oxygen in the water. And lastly, there are biomedical implications and applications of hydrogen sulfide. Obviously, it's toxic. It's also produced endogenously as a byproduct of cysteine catabolism. It's been recognized as a signaling molecule and cardiovascular disease. I'm not going to talk about that at all. So I want to talk about adaptation to hydrogen sulfide and the fish that live there. So I'm going to take you on a journey to southern Mexico where Pesilla, Mexicana, lives. And there are multiple watersheds with pairs of sulfitic and non-sulfitic springs. And here, water is running. Let's see if there is. Does this actually work? Water is running up the page into the Gulf of Mexico. And these are the pairs of springs where we have sulfitic and yellow and non-sulfitic, just marked on the map. And they're replicated natural experiment. And populations of Pesilla have adapted to each of these sulfitic springs. So our natural experiment, perfect. And here on the right is a mitochondrial tree to show that these have independently colonized those sulfitic springs. And that's important when thinking about the genetic and phenotypic changes because these are independent colonizations. So hydrogen sulfide survival varies between ecotypes. Just getting at this idea of fitness, organismal fitness, and hydrogen sulfide varies. So this is a Kaplan-Meier survival curve. These two plots go together. On the x-axis is time in seconds. And this is a hydrogen sulfide spike in experiment, or increasing, right? Hydrogen sulfide here, concentration is on the y-axis. Hydrogen sulfide is increasing in the water, OK? And we can measure cumulative survival. This was done by my close colleague and his group actually before we met. And here's time, right? And all of the non-sulfitic ecotypes, survival goes essentially to zero in 600 seconds. And the sulfitic ecotypes survive a bit longer. This is significantly different. But you see, there's a point at which they also start to reduce their survival. So it's not that they can just survive in any level of hydrogen sulfide. So given that hydrogen sulfide is such a strong selective pressure, we can make predictions about where in the genome we would expect to see changes and then test those. So before we actually did any sequencing, shockingly, we predicted where we thought changes would occur, given what we know about how hydrogen sulfide acts. So here is the electron transport chain. I won't get into all the details, but here's how we can make predictions. We know that hydrogen sulfide, the major site of toxicity, is here, cytochrome c-oxidase. And so we expected there to be changes in cytochrome c-oxidase. We also expected the up-regulation of detoxifying enzymes. Well, they're having to deal with hydrogen sulfide. They have the enzymes to do it. We all do. So we're going to see up-regulation of detoxifying enzymes. And we expected, we hypothesized that there would be a decrease in expression of genes involved in the production of endogenous hydrogen sulfides. Like, you're getting so much from the environment, we're going to see a decrease in those. And there are a bunch of other molecular targets that have been described in the literature. And so we made predictions about those as well. They just don't map really beautifully onto the electron transport chain. So RNA sequencing of sulfitic and non-sulfitic populations showed us that differential expression was true of some of the predicted genes. So as we expected, we saw changes in the genes involved, expression changes in genes involved in detoxification. That's right there on the bottom left. Interestingly, we saw no down-regulation of genes involved in the production of endogenous hydrogen sulfide. And indeed, we actually saw up-regulation of one of those, which is here. So we're like, all right, we need to think about the process by which hydrogen sulfide is used in the body is something that is not being affected by external hydrogen sulfide. And so we were then, this was all in wild individuals. Great, but maybe this is all just a plastic response to the fact that they're living in hydrogen sulfide. And so my collaborator and his graduate student brought the fish into the lab, raised them for several generations, and then asked which changes in gene expression are evolved versus plastic. So here's just the setup of that experiment, the color matching, non-sulfitic and non-sulfitic water. Here's the exposed non-sulfitic fish. So this tells us about something about a plastic response. And then we have sulfitic fish and non-sulfitic water, and then sulfitic and sulfitic. And we can then determine the role of heritable versus evolved differences in gene expression and overlay those onto our wild population. So I just figured out how to make these hatched lines in Illustrator, it's a work in progress here. So here in the purple are just genes that were up-regulated in all of the sulfitic populations that we compared in this pair-wise sulfitic-non-sulfitic framework. And then with this great hatched lines, it's also really hard to find right color palettes for non-color blind. So here in the hatched were up-regulated and had evidence for evolved changes in expression. And so you see that not all genes, and so I'm just showing a snapshot of some of our predicted ones, but this was done genome-wide. Not all genes that showed differences in expression showed evolved differences in expression, and some of those changes in expression were due to constitutive changes in expression, and some were changes in plasticity, either becoming inducible or losing their inducibility. So that's really about expression, but we were also looking at genetic variation. And so that led us to think about the enzymes, and so the first one, we know that cytochrome c-oxidase is the target of hydrogen sulfide toxicity. And so we said, let's look at cytochrome c-oxidase and see what happens in terms of the function of it. So here is just my representation of the toxicity of hydrogen sulfide, and this was actually an assay to measure cytochrome c-oxidase activity with isolated mitochondria from different populations, and so there were two non-sulfidic populations that we looked at and three sulfidic populations. Here on the x-axis is the level of hydrogen sulfide concentration that was measured, and on the y-axis is relative cytochrome c-oxidase activity. And you can see that some of the sulfidic populations have a resistant cytochrome c-oxidase. The level of hydrogen sulfide has no effect on the enzyme. Now, in contrast, you should all notice there's this one sulfidic population. They're living in hydrogen sulfide, and yet they have a sensitive cytochrome c-oxidase. And coupled on top of that, which I'm not going to show you any of the data, so you have to take my word for it, is that these two populations are really two populations and two individuals from one, which is why there are three lines. We saw that there was the convergent evolution of an amino acid change that we think plays a significant role in that lack of sensitivity to hydrogen sulfide, and there was evidence for selection on one of these genes. This was cox one. And these three lineages had no evidence for selection, so coupling both the properties of the enzyme with some of the sequence analysis, which we thought were quite interesting. So another assay that my collaborators' student has been developing for a long time is to measure the sulfur quinine oxo-reductase. So the next one of the really important main detoxifying enzyme, this is the beginning of detoxifying hydrogen sulfide. So circled there. And so that assay has revealed that all of the populations have no effect of hydrogen sulfide on the relative SQR activity. And indeed, there's an increase, a significant increase, in the populations over the non-sulfitic and over the baseline. So again, here on the x-axis, excuse me, is hydrogen sulfide concentration. And on the y-axis is relative SQR activity. And you see that in all of the populations, it goes above the non-sulfitic. And this one that dips down, of course, is that weird one that also has a sensitive cytochrome-seoxidase. So we then wanted to understand what is the genomic context for this SQR and also other enzymatic changes. And so we re-sequenced individuals to high coverage from all of our sampling locations. And I've added an additional drainage up here. So there are actually now four drainages with replicated sulfitic and non-sulfitic populations. And that's because during the course of our studies, that spring was rediscovered. It had been described 100 years ago, was rediscovered. And so then we could also sample it and ask questions about it. So here's the population tree. And really, the take home here that I want to show you, so this is all of the genomic data, re-sequencing of one individual from each of these populations. And you see that there are three sulfitic over there on the left and two sulfitic here. And largely, the sulfitic populations are independently derived. Well, so what about if we look at the genome specifically in the region that contains sulfur, quite an oxidized reductase? So we're just looking at the SQR region of the genome and making the tree of that region. And this is what that tree looks like. All of the sulfitic ecotypes cluster together. And so there's a shared origin for SQR in the sulfitic populations, which suggests that either there's incomplete lineage sorting, migration, or convergence. I think in this case, convergence is very unlikely because it's a large haplotype. So either suggesting that migration is moving alleles via the transporter hypothesis, via the freshwater, or that standing variation in the freshwater population is very high. And then that's, again, being reshuffled into the sulfitic populations. So I put this up here because I think that the genomic context in which you're thinking about phenotypes, even when the population tree, or the species tree, or the mitochondrial tree, tells you that these are independently derived populations, the regions of the genome that may be underlying your phenotype of interest may be more closely related than you expected. And it's interesting to think about and important to think about that population genomic context, as well as the genome. The thing about these fish that I find particularly appealing to work on is that it's not only Sicilian, Mexicana, and southern Mexico that is adapted to hydrogen sulfide rich springs. And there are actually about 11 different species that have adapted to hydrogen sulfide throughout this whole region. They're all in the same family. And so we then went out. I wasn't part of the group. I'm usually on the computer. Some students went out and sampled individual fish from all of the different populations and the closest related non-sulfitic population. And some of them, this looks perfect. It's also a cladogram, so it doesn't have any of the branch links on there. Looks so perfect, but I will point out that there are some lineages of sulfitic fish that don't have a closely related non-sulfitic population. And we did try to sample an ancestral base of that clade. But so went out and sampled just wild caught individuals, put the tissues in RNA later. We can talk about whether or not you like RNA later. But regardless, that's what was done. And then brought those samples back to the lab and sequenced the RNA to look at how expression levels change across all individuals that we're being looked at. And here's one example. This is SQR. And we see this same pattern for many of the genes involved in sulfide detoxification. So we see a consistent shift towards higher expression of genes related to detoxification. And so this was six individuals that were sampled per lineage. And here's just a bar plot for the expression levels. This is the FPKM for SQR. So lessons learned, really, and a few opportunities. So we figured out that, well, detoxification and resistance are important mechanisms for hydrogen sulfide survival. I was not that surprising. What we found was that adaptation to hydrogen sulfide is largely predictable. However, there seem to be multiple ways to solve that problem, at least when looking at specific enzymes. And we do see convergent shifts in expression in all of the sulfide springs. And it seems that expression, potentially more than actual changes in amino acid, are important for this system. Then kind of more generally, I want to just give the plug for natural variation and looking at natural systems, which I think gives us an unprecedented ability to understand complex phenotypes, maybe. They're complex. And that replicated environments or replicated phenotypes across large taxonomic scales really provide a framework for understanding evolutionary processes. So I want to return to this idea of thinking about how organisms function in the context of their environment and which hierarchical level of organization you think about and which phenotypes you're interested in. It's very important also to think about how that translates to the next level of organization if you care. And really thinking deeply about what phenotypes are important to relate and at what level you feel that you have done sufficient work of connecting genomic variation to phenotypic variation. The other part of this, this big white part I love, because we don't talk at all about the environment, because I've said natural systems. And hydrogen sulfide, this is you could put really any stressor here or any selective pressure that influences potentially the abiotic environment, the biotic environment, and all of those things influence different levels of organization here, either directly or indirectly. And depending on what temperature you're at affects enzymes in different ways that may or may not be physiologically relevant depending on where your organism lives. So it's definitely a complex thing. And so just some challenges. I guess I changed this to my challenges slide. So thinking about high throughput phenotyping. And we haven't talked that much about phenotypes or which phenotypes are interesting. But as these take a long time to develop, especially if you are thinking about natural systems and what are the right conditions to actually run those assays, which phenotypes are most relevant or most informative? And obviously that differs depending on what you're thinking about. And if you're thinking about biomedicine, are you thinking about evolution and fitness? The genomic and environmental context matters. The potential lack of genomic resources for natural systems, right? High quality reference genomes matter. Accurate annotations, including genes and regulatory regions. So just to tie that back to the sulfitic fish, we have a very fragmented genome. The genes are relatively well annotated, whatever that means. There are genes. They seem to match to human sometimes. But we really don't have a good sense of what the regulatory regions are. And for us, looking at expression, it seems that expression is really the thing going on here. Having those regulatory regions and being able to identify them and then look at how selection is shaping changes in regulatory regions is incredibly important. And then another challenge when thinking about tying this to human health, because it is NHGRI, NSF, and NIFA sponsored, was I was thinking about one of the challenges of working on hydrogen sulfide. Just as an example, I would be very interested in working with MDs or researchers who think about hydrogen sulfide in a clinical context. But I'm never in the same place as them, ever. So there are challenges in bridging evolutionary or natural system studies with human health and disease. And the three things that I think about with those challenges are finding a common language. How do we speak the same language, or at least know enough about the other person's language to have a conversation that could lead to something productive? The second was attending similar meetings. How do we get to be in the same place together to have those conversations? And then breaking down the research silos, how do we actually bridge these different large areas? And so there's the bridge that I stole from the internet to represent that. That's not built. We've got to build it. So with that, I want to thank all of my collaborators and the funding and this work that I talked about was largely driven in collaboration with Mihi Tobler at Kansas State University and by my graduate student, Anthony Brown. And we've received funding from a variety of sources, mainly the National Science Foundation. And so with that, thank you for your time.