Scientist Stories: Matthew Harris, Genetic Basis of Development

All right, I think we're gonna get started. We are only 1 minute past the, the start time. It is my distinct pleasure to introduce, uh, Dr. Matthew Harris, uh, both colleague and friend, uh, to come give today's lecture. Uh, Matt got his BA at BU, um, with an emphasis on marine sciences, so he did some time down at MBL at that time, right? Then he got his PhD at the University of Wisconsin with John Fallon. So go Badgers, that's for Eric if he's online. And then his postdoc work, he went across the pond to Germany and worked with Dr. Nüßlein-Volhard. And this was sort of at the end stages of the very large early zebrafish screening where most of the focus in the zebrafish field at the time was embryonic development and the genes involved in early pattern formation. Matt was one of the early adopters of starting to look at adult phenotypes in fish and really has led the field in that area since then and really changed the way people have looked at zebrafish as a model, and that it can actually contribute important things to our understanding of biology in adults.

Post-development.

After that, he came back to the US and took a job at Harvard Medical School in the auspicious Department of Genetics there. Many famous scientists have come out of that place. When he was there, he's done a lot of dominant screening in adults looking for skeletal defects. One of the more interesting things he moved into Also, the NIH NHGRI organized an opinion piece earlier this year about the concepts around— we called— termed Darwinian genomics, where we can use the largest mutagenesis screen ever done, which is evolution, to look at how the genome controls morphology and body shape. And so Matt was really very early in adopting this idea and moving into this field. And so we're going to hear some of that today. So please help me welcome Matt, and we look forward to your talk, Matt.

Thanks. So it's a real pleasure to be here and come and visit and give a talk. It's also a real pleasure not to have a centimeter of ice welcome me as I come into Washington. And I realize there's a lot of people on the Zoom audience probably. So I will try to stay and behave myself and stay by the podium and use the computer mouse to actually point at things. I wanted to give a talk today. So I'm also— while I'm at Department of Genetics at Harvard Medical School, I'm also at Boston Children's Hospital. The work we do is pointedly targeted at trying to understand the genetic causes of human disease, especially that affecting children, and to try to understand what the understanding of development can teach us about how to approach, how to alleviate or abrogate a lot of the pathologies associated with a lot of these disorders. When I came from Germany, one of the arguments I were making is that through understanding of development, particularly diversity and development, we can understand what nature has done to try to get at these questions in ways that we cannot do with the limited model systems we're using in the lab. And to sort of highlight the point about this is that just the sheer diversity amongst, for example, even let's just talk about vertebrates. You know, we're talking about ray-finned fishes, which are called teleosts. Teleost fishes have about 30,000 species. And that's more than all the other vertebrates combined. And the different types of forms and functions they have are so diverse that actually trying to understand them and particularly the genetic basis underlying how these different systems work is a ripe area in which to try to do experimentation that would not be possible to do in our existing experimental models. So my lab is particularly interested in the skeleton and how form and function comes about. And one thing you're gonna see is that One thing we focus on is proportion because novelty in, in creating structural elements like the skeleton is something very rare and almost doesn't happen. The novel structures you see are usually formed by elongation or a different shape parameter of co— of bones that already coexisted in the common ancestor that doesn't have that particular trait. So one thing we've been focused on is proportion, and that's something I'll tell you today. But the focus in development is essential for any analysis of evolution, I think, 'cause Development is a translator of the genetic information by which then is the functional aspect by which selection acts. So selection will act on the physiology or the structural component, the phenotype, but that's through development that these things are translated. But part of development is that evolution is incredibly lazy, and this is something we capitalize on all the time. So the first concept is about conservation, and I don't really need to get at this, but that we reuse developmental programs over and over again. Right? So all of our core toolbox developmental genes we like to talk about— Sonic Hedgehog, Notch, Wnt— right? The reason we have 17 Wnts or more is the fact that they've been co-opted and actually then diversified in their new tissues. And so this conservation means that actually, you know, unless you have diversified and create a new ortholog, that mutations or loss of function in these genes have the propensity to be lethal. They usually die during development. But the ones that actually can make it, those are the ones that actually are fertile and actually that selection can act on. And so there's an inherent bias about what types of alleles you might get popping out the other side. And so that's the idea of constraint. There's developmental constraint on the types of alleles you'll see out in populations. And this is strictly due to selection acting with development that give you different type of phenotypic diversity. With this comes the idea that the historical or phylogenetic framework shapes this model of change. And that— so it's that similar to I look like my father and his father before him, is that fishes will have a historical, like, antecedents behind all the phenotypes you're discussing. And it's important to know what those are. Those antecedents might be phenotypic, they might actually be genetic. And understanding, for example, whether there's a duplication event and what that means to particular systems and how they evolve are very important. So the major concept I'll stress today is about the phylogenetic framework leading to experimental design to answer questions. And then lastly is, can these properties all in sum lead to a level of prediction? Can we look at natural populations, find changes there, and then go to patient populations and find answers about how to either shift their phenotype or causes of the phenotype in the first place? So, we get at this primarily for genetic mechanism. So, on the left here, I outline a classic screen in zebrafish, which is an F3 recessive screen for loss of function mutations. This has been done for the last probably since the 1980s, so now 40 years. And this has been intensely powerful to try to understand the syntax, the formation, the structure of how genes lead to development of phenotype. And as Sean nicely said, one of the key things that my lab has done or my work has done is to start trying to focus on the adult phenotype, as you see on the right, which is the zebrafish. This has unique structures that are not seen in the larva that form during metamorphosis or juvenile development of the fish. And these are highly pertinent to adult characteristics you see in— or characteristics you see in newborns, right? These are the structures where if someone's coming to the clinic, they're going to have variation in these sorts of organ systems. And so we were successful, and we've done both recessive As well as now I focus on dominance screens, looking for unique phenotypes to try and understand the logic of how these structures are developing and the genetic basis of this. And in short, why we look at dominance screens is that it's much faster. So original recessive screens, we only screened about 1,000 genomes, meaning about 1,000 loci in the genome. That's about 1/20th. Where in this case, we screened 10,000 genomes, in about 1/10 of the time. So it took about 4 to 5 months to do this. And importantly about these types of mutations, they're not loss of function. So they're not hitting the development with a sledgehammer. Rather, these are nuanced, and they often have a gain of function phenotype, either dominant negative or hypermorphic or neomorphic type changes. And what that means, these are the type of mutations that are pushing development potentially into areas it normally does not go. Rather than just seeing what is necessary, you're seeing what's sufficient to do something. And in that area, we find a lot of very interesting phenotypes, some of which I'll talk about today. But lastly, or conversely, instead of just looking at a forward genetic approach, we'll look at a forward genomic approach, which is called— and we name this phylomapping, which I phrased in my title. And this is looking at the existing diversity in natural populations and their evolutionary relatedness to actually look at this in terms of a natural screen. And what that means is you have to look at how the phylogeny, how they're related, and where characters come and go on this particular tree. And if you treat it like a screen such that a species, a lineage leading to a species that has a fixed, invariant character but has a common ancestor, hopefully recent, that has a more ancestral character, more consistent with the rest of the clade, you can treat this like a mutant. You can actually look at pooled samples from a genome and map it for fixation of particular alleles throughout the genome. And what's nice is not only do you find a single— you can find particular point changes, but also that you get a collection of them, unlike doing a forward mutagenesis screen. So you find sets of genes associating the particular traits. So we've done the opposite, which is to have mutants in zebrafish that we find, go to natural populations, and ask whether they can phenocopy this. And we find that it's to the gene. We have mutations in the zebrafish, we go to natural populations, we get that gene out of the natural population, we find that there's significant loss of function changes in that gene, leading to that phenotype. And this is work we have done in several different species, which I don't have time to talk about today. The phylogenetic structure though provides experimental means for elucidating this change. And what is cool about this and what's been empowered is you can look at gain and loss within a clade using the same dataset. You have power convergence when you get similar phenotypic changes by unrelated lineages to elucidate complex phenotypes to get rid of all the background signal that would normally be there. And that you can compare patterns and trends of radiation amongst completely unrelated lineages in similar environmental situations. And these have been really impactful in the field to actually narrow down all the variations you get with lineages to get a particular signal. So the phylomapping strategy we've used, and we started this about 12 years ago, we decided to take a targeted capture approach at doing this. Instead of sequencing the whole genome, which in those days was short-read, highly complex, annotation was not so good. We decided to actually make predictions about all the coding and conserved non-coding elements we could combine within a clade of the known genomes on the clade and target capture all this orthologous sequence in pooled samples from species lineages we're looking at. And when we pull this out, then we align this to what would be a reference genome. Take those samples and do pairwise analysis genome-wide across non-coding and coding conserved sequences. Now, there's caveats to this, but there's also pluses. Caveats is you're not going to get the intergenic material that is not conserved across populations. But I'll throw at you that we're actually looking at conserved changes so that we would expect them to be in something that is shared amongst those lineages. But by pooling samples, we're actually able to look at analysis of variants within these loci. And that is important to look at fixation. So instead of just taking one individual sequencing it, you don't know what the other individuals have at that particular site. But if you actually sequence 5 or 6, you actually can look for patterns of fixation versus variation and actually track that much better. And that is important for clarifying signal versus noise. And so this is detailed in a 2015 paper from my group. Key to this is the idea of a phylochip. Instead of using just a single reference array that we were using, we actually made a conserved array, a conservome, based on phylogenetic principles of that particular clade. So in this case, we used mostly the sequence we get from lineage A, but we combine sequences that are similar But up to differences about 15%, we've figured out empirically that works and add these sequences as they go up through the lineages to this particular pool of targets that we use. And we have a collection of both microRNA, UC, CNEs, as well as coding sequences that we go and we pull down. Early on, we tested this just on a single case study, a single lineage. On the left, here's an example of where the experimental or the mutant model is a species of the cyprinid, which has lost all the postcranial skeleton of the scales. You can see it's basically naked. Whereas a closely related relative actually has the ancestral trait of all being scaled. You can see here is a zebrafish for comparison. We did this analysis where we pull down genome-wide and look at variation. And we pull out that you get conserved fixed changes in particular genes in different classes. In particular, we pulled out, for example, FGF20A ligand and FGFR1A receptors. And these two genes, actually, we found out that in mutagenesis in the zebrafish are sufficient to phenocopy the scalation and naked phenotype we had in those fish. So we think that we're able to pick out part of the genetic program leading to that. Not everything, but importantly, we're able to get epistasis from this. We get combinations of genes that are required for the phenotype. This is very difficult to do with forward genetics. It takes a long time. But so that led us to the idea that you need, you know, density in your taxonomy to get this. So the higher number of species are more closely related, the more the signal, more noise you can get rid of compared to the signal. And so here's an example of sort of why this works. Here is one of these eye tests you take, and I've taken away the colors because it doesn't matter. Let's imagine here's your set of genes that come out for whatever particular statistics you're running based on fixation or selection of particular lineage. And the size of the dot could either be the number of genes or GO term, or it could be the strength of selection. You can make that up how you like it. Well, this would be fabulous. You get some sort of signal from the noise, 20,000 different loci. You now have something to chase after. Find patterns within it. But there aren't any real patterns you can really deduce from this. Because every part of what makes that species a species is variable in this dataset, right? So all the warts, all the nooks and crannies are all present in this dataset. But if you do it with a sister group that also has the particular trait that's fixed, you can now start to see what's shared between them. And you do that again with another species, and you do that again with another species, Now you can find within there actually distinct networks of genes that are required or correlate or associate with that particular phenotype that has been fixed in those different traits. Other things have altered, right? Metabolism, lifespan, coloration, all this has changed. But particular thing that's consistent between all of them has remained constant. We applied this to large clades. These are the Antarctic notothenioids. These icefish and relatives. We're able to do this on about 45 different species, if I remember the number close enough, and look across the whole clade, including outgroups, about what happened in their radiation in Antarctica and the character changes that have occurred in them. Now, there's natural variation in notothenioids. If you have not heard about this, they have amazing sort of pathological phenotypes that they have evolved to live in the particular cold waters of Antarctica. These things normally would kill any other animal and are hallmark characters for a lot of human diseases. However, these animals are perfectly fine. We're able to do this and we design a new follow chip. We're able to recover high-density coverage throughout these loci. We're able to find traits, for example, here's an example, a character in TRIP11, which relates to osteopenia in the skull. And to actually live in their environment, the history, evolution history of this group is that they were bottom feeders before the polar current formed and the drop of temperature. And then with the influx of all the nutrients, they actually came up in the water column. But they don't have a gas bladder. So to do this, they have to do two things. One is get— well, main thing is get very light. They do this two different ways. One is insert lipids, and the other is lose a lot of bone. And so we're able to see as you go through the phylogeny that actually all of the Elaginopsidae, which contain the icefish and other notothenioid relatives, start to accumulate mutations in this gene TRPL11 that over time start to accumulate more and more. And what you see is that they start to lose more and more bone when this happens. Now, what I don't have time to show you is that mutations in TRPL11, both in humans and in models in mice, lead to a reduction of bone phenotype. And so it's very similar type mutations occurring in this particular protein leading to a skeletal pathology that mirrors that in human disease, but in this case has been selected for. So here's a quote sort of addressing that. And so one would think that we were kind of onto something hot and this is something new and cool. This is from J.B. Sutton from 1890 from a book called Evolution and Disease. The manifestations of disease were regulated by the same laws which govern physiological processes in general, and that many conditions regarded as pathological in one animal are natural in the other. So we're just basically redefining this on a genomic and developmental basis. So the first story I want to tell you today is an intersection between the genetics or genetic approaches and our genomic. And this is about proportion. What you can see up here on the screen I'm showing you is a collection of mutants we got on our dominant screens, in which the wild type form of the zebrafish here, you can see the fins are a particular size. These change in drastic ways in these mutants. For example, all fins here are much larger. Here you have a fin-specific overgrowth. Here you have again much larger in these different mutants. Here you have shorter fins, as well as here. All of these are proportional though. This is not dysmorphic. They pattern themselves just appropriately. So these are genes which relate how proportion is regulated or how size itself is regulated. So the interesting thing about this is the majority of these, we cloned them out, involve bioelectric signaling. So potassium channels or connexins, ion regulators. This was an albatross around our neck because no developmental biologist ever thinks of potassium channels or ions as being a key developmental mediator. But since the fish were telling us something, we decided to chase it down a little bit more. I'm going to summate probably 12 years of work into two panels. What is it about potassium channels? Well, large family of transmembrane channels just allow you to have this sort of sensitivity, I think. Evolution has allowed you to get all these knobs on this stereo equalizer to turn up and turn down. Selective and regulated modulation of ion conductance and membranes. They do this primarily through resting membrane potential, which is the Vmem over here. Our mutations affect channels which can affect not only the signaling component in the cytoplasm, but also the regulation of conductance through. We get different types of channels out, all of which are potassium-mediated. And we know there's intracellular mediators for like calcineurin here, which can affect growth on the downside of this. And that these are present in excitable and non-excitable cells. So, mostly we think of like excitation as neurons, muscles, these sorts of things. But your fibroblasts actually have these cells, have a resting membrane potential, and it controls their differentiation and behavior as well. And we think that what this— way it works is that it's sort of an algebraic formula, that there's additive action of these channels. To integrate environmental and internal state signals into growth patterns. So it gives you coordinated growth of particular structures. And that VMEM serves as a common output for this growth regulation. And finally, that we know that fin growth response to this VMEM is specifically controlled by KCNK5, which actually we have loss-of-function mutants, and you fail to get changes of induction all these particular bioelectric signals. This is what they look like. We know that the channels are sufficient to cause overgrowth. We can cause clones in zebrafish and you get localized overgrowth. You see here the dorsal fin as well as here the caudal fin. We know this is mostly autonomous to structure, but there is some non-autonomous effect in nearby rays. See here in red, this is the marked clone and this is the effect on the non-marked bone of the fin rays themselves. And this is seen not only in one channel, but all the channels pretty much. You overexpress them, they're sufficient to give you this phenotype. Here's two separate channels, one KCNH2, KCNK5. We also know that this function of particular channels are necessary. So in this case, we actually tried this sufficiency experiment for years, and we weren't getting any results. And finally, CRISPR came around. I'm like, let's just do the opposite. Let's see what happens. We know that actually this particular channel is necessary for the actual formation of these structures itself. The normal function of this gene somehow curtails this bioelectric signaling process, and when you remove it, you get a shift probably in the resting membrane potential and you get overgrowth. But here comes the problem. Is bioelectric signals informative? Is it instructional for development or is it just a process of development itself? Byproduct. So here's a quote from Cliff Tabin, actually happens to be an exceedingly productive and well-known developmental biologist and geneticist, but also my chair, but had this wonderful quote in The New Yorker: Electricity might be required to run my vacuum cleaner, doesn't mean there's necessarily an electric code for vacuuming itself. And that is where the field rests right now. Is bioelectricity actually instructing those fins to be particular size, or my arm to be a particular size, or my jaw, or whatever. And that we don't know. But we think we're empowered to do this because we do know by the sufficiency studies that the same cell types were basically affected in both a channel that gives you overgrowth, in which you have basically just a continual addition over time, versus a channel in which you actually have an increased rate of growth rather than— and then stopping at the normal time. So different phenotypes.. But the same cells, just different channels can instruct that by overexpressing. So we're currently using this model right here to get at how these channels are informing the tissues to actually lead to phenotype and whether there's information involved. All right, let me pause a second there. So I'm going to switch gears to the phylogenetic component of it. Can we use nature to actually understand this? And so if you're asking about proportion in fishes, this is one of the best So we'll thank David Attenborough, who is nicely championing—

with an extra thrust from their tails, the flying fish get airborne once more.

So in evolution, as obvious here, probably for an escape mechanism, they can glide— this lineage of fish, flying fishes, have grown proportionality in their fins, not only just their pectoral fins, but also their pelvic fins. And if you see them do this, that's because the ventral lobe of the caudal tail fin is longer and acts like a rudder, right? You kind of see it there. But it's a little takeoff mechanism. So could we go to nature and ask questions about how proportion is regulated? So we went to the group including flying fish, which is called balaeniforms, which also has increased jaw length, which I'm going to talk about at the end. And ask, can we parse out signals for this particular type of phenotype? The flying fish through balaeniforms actually have a phyletic signal of allometry, in which the flying fishes themselves have a much higher allometric growth rate through development than do the halfbeaks and needlefish themselves. So there's an evolutionary signal for this growth. We made a phylochip for this based on this particular phylogeny where we joined in needlefishes, haploids, and flying fish, and used both collection of medaka, molly, and platyfish genomes, which are known at the time, and combined them and asked for things that were different to make a hybrid phylochip to actually what we call in the fishing lingo is seining for their genomes of what's particularly in there. Again, we have both coding and non-coding sequences represented. We went to fabulous places like this is Little Cayman Island. This is lab members going out hunting, and we're able to collect samples live. When we took samples from these, we also are able by this method, because it's not long read sequencing, we can use museum specimens to actually fill out our phylogeny. So from that, we're able to get 43 species of balaeniforms covering both needlefish, sawries, halfbeaks, and flying fishes. And from this we get about 77 to 81% coverage of the array that we put in. So these are a number of loci that we're able to pull back reads from at a depth of about, you know, 34x, around 30x itself. We get about, you know, 4 million total SNPs within each group. And of these, about 600,000 of these are fixed. So it's still a huge dataset within particular groups themselves. Within the different species, it gets reduced, but still within a particular species, you might have 40,000 fixed different SNPs to start weeding through. And this, if you look at this in terms of a heat map, in terms of diversification. So this is a heat map we looked at to look at conserved or accelerated sequences. Accelerated sequences are basically in the lighter blue, conserved are in the darker blue. Each little line here are the average out is a particular locus., and we count down the general variation we see at that locus and whether in this case we get more or less than we'd actually predict for that particular lineage. And you can see, as you expect, like I said, I look like my father, his father before him. There is generational or historical bias in the types of signatures you pull out in this, that in needlefish you pull out large signatures that are unique to this particular clade. Halfbeaks, you pull out for this clade. Flying fish for this clade. But within there you can start to parse this out, and then you can start looking at function. So genes that come up as being differentially selected for in the different groups. And we look at flying fishes, you can see here that you get things like iron transmembrane transport, FGF receptor, calcium ion, ion ion, this sort of stuff. And you get other developmental genes, which is really quite positive for the, the, in the analysis. But you can get even fancier. We look at regulatory signatures. So we use GRADE analysis, where you look at whether structural localization of particular genes to particular CNE and selection at that CNE in the different lineages. And if you, if you model this only at the lineage at the base of the tree, so this includes all the flying fishes compared to those of the halfbeaks, you can find things, for example, that have been enriched, for example, embryonic pectoral fin morphogenesis, digit morphogenesis. We think we have a very rich dataset here to actually start to parse this out. If you look at individual CNEs for variation, you can see here that in the flying fish group, that if you look at percent ID of Medaka, that over the entire basal part of the clade remains up at about 94%. But in each of these particular CNEs, only within the flying fish, you actually see a pretty drastic reduction down to 85-87% here in all these cases. By using what's known about a lot of these developmental genes, you can make hypotheses about what might be happening in these lineages. For example, the SAL genes are necessary for autopod development in the mouse. Therefore, maybe this is regulating the zebrafish about how the shape or the structure of that particular fin might be occurring. The other way is by convergence. We talked about this a little bit earlier. One way to get rid of some of the noise is ask the question about unrelated lineages that also have the phenotype. But you now can subtract all the things that were not related to that phenotype away from those particular gene sets. And here is a case where we look at gliding and some of the halfbeaks, which are the sister group to the flying fish, actually have elongated fins and start to try to jump out. But they don't get very far. They haven't evolved actually a true flight. It hasn't been fixed within the lineages. These have ancestors that do not have the long fins, so it's a unique event. If we scan the genomes for things shared by these and all of the flying fish themselves, then we can find particular hits, for example, unique amino acids that are found, for example, here in Gigi F1 or GSRA, which actually look like very interesting targets because, for example, Gigi F1, is involved in IGF-1. I saw this when I walked in to see Sean's office. I like this cover. As well as GSR, so growth hormone receptor, which is involved in acromegaly and stature. And so by looking at, for example, acromegaly, you could sort of see why this might lead into a common-like phenotype in the plant and fish. But in particular to what we have been studying in terms of biologic signaling, kinase, for example, here in KCNK9, we saw very specific regulation within the gene. Only within the 3' of the sequence, the C-terminal part of the sequence. For example, right when you get past the transmembrane domains of the channel here, sequence divergence varies wildly, whereas the other part, it's actually more constrained. This is the average percent identity between flying fish and the other bilaterians. And if we go through here and look at sequences, you can see— oh, I'm going the wrong way. I've given away the whole talk. There we go. You can see here accumulation of mutations, as you would expect, as you go up through the phylogeny. And actually, some of these are predicted to be highly deleterious. One aspect to this is that if you overexpress KCNK9, it's known by Guo Jinzhan's group with Matthew Seeleck that you overexpress this, you can lead to overgrowth here, which is highlighted here. And we've shown that actually, if you delete the C-terminus of KCNK5B, which is an ortholog of KCNK9, that actually you can get elongation of the fins themselves in the P0s that you've done CRISPR to, so that this C-terminus is necessary for this regulation itself. But getting to some of our models that we have, we actually found that there was fixed mutations in this gene KCNH2 that is compared to the rest of the phylogeny. KCNH2, as I said earlier, leads to elongation of all the fins in in a coordinated fashion. We also find mutations in LAT4A, which in gliders we get specific mutations here. And we actually find in our screen that we get loss-of-function mutations, missense mutations in this particular gene in this mutant NR21, which encodes LAT4A. So short fin mutations, we're finding mutations involved in that too. But being a geneticist, you go, okay, I got gain, I got loss. What do you do? It's like Sesame Street, you put them together, right? It's not all too hard. So we have a short fin and we have a long fin. When you cross them together, you end up getting a structure that looks very much like a phenocopy of flying fish itself. So the pectoral fins here are much larger. They have the long fin phenotype, whereas the medial fins are all suppressed. So you have a wild type now dorsal fin, wild type anal fin, wild type caudal fin. However, you get this little nubbin on the bottom. This little nubbin is exactly like that increased ventral lobe of the flying fish that I talked about. So we think the combination of these two signaling systems doesn't have to be KCNH2, could be another potassium channel, but the upregulation of bioelectric signaling and the suppressive activity of what LAPP4A is involved with can modify this phenotype. And this maybe is not surprising that there's actually independent lineages in the fossil record that actually have the same exact parallel type phenotype that arise. So they have elongated pectoral fins, elongated pelvic fins, and the elongated ventral lobe of the caudal fin here. So that maybe this is like a ball plan, a phenotypic— use a German term— ball plan, that actually there's many ways to get there by altering the genome. Or that there's intense selection that everything else might be lethal. All right, and so I would like to shift gears a little bit to proportion again, but this time a much more complex structure, which is the cranium. And in this, we'll actually use the same data set. So what I show here on the bottom, this is a needlefish that actually, you know, I haven't done this yet for my grants, but I should probably put the sort of the cause or reason for studying this. I think they're, is a massive increase in the number of intrapenetration injuries that occur on the coastlines of people. So both intracranial and intra-abdominal from needlefish as they jump out and attack surfers. I think it's highly important that we address this issue. That was supposed to be a joke, but that's all right. I'm sure the Zoom audience is laughing. But part of this is that there's a historical question about the direction of growth. And this is one that's raised by Gavin De Beer, probably about 150 years ago. This is about terminal addition. You can see here that if you look at the evolution of growth, you look at beginning fish, that they all have this normal fish-like torpedo shape. As they evolve, you start with a half-beak and outgrowth, and then you add on then the upper jaw. Then that would be the phylogeny recapitulates ontogeny argument. Well, you'd look at the balaeniforms and this was what they use as an argument. It looks like it fits perfectly, right? You start with Medaka, you go to flying fishes, halfbeaks here, you get elongated bottom, needlefish stars, you know, both. Maybe we could study that, but evolution plays tricks on you, which actually halfbeaks were the ones that actually developed first. Then you had wasp flying fish as well as increase in the upper jaw. This is not a problem. Actually, when you're given lemons, you make lemonade. So I'll talk about that in a minute. The signatures we looked at actually involved looking at convergence. So there's a difference in developmental heterochrony about formation of the jaw between juveniles and adults. Some lineages are direct. You go straight from both jaws being overgrown to ones being half-beaked to being the both overgrown here. So if you look at that pattern in the genomic sequence, you can get these signatures again, which allow you to discern between specific signals involved in that particular character trait and those of everything else. In those, we find changes specifically within these developmental heterochronic genes. In genes known to be associated with frontal bossing is Martin-Walker syndrome, PZL2, and the fish is PZL2-like, and thrombospondin, which has been linked to all neurocrystopathies. This is a case of domestication of both in pigs as well as in dogs. And what I will tell you about today is a convergent fixed amino acid substitution in AGAP1 locus that we were interested in because actually it leads to microcephaly in patients. On the bottom right is a QTL at the AGAP1 locus. See here that you get statistical association of the particular SNP in the patients. And the function of this gene is really sort of unknown because AGAP1 is really not a developmentally interesting gene. It's one worse than potassium channels. This is like some sort of microtubule stability type structure expressed everywhere. Why is it only giving you necessity in this particular structure? And so they raised the question, which I don't think I really need to dive into too much with you all, is about what is the hypothesis when you see fixation on a particular trait? Well, you're in LD with a large component of the genome— well, not a large component, a footprint in the genome. And so either it could represent that there's changes in function in that coding sequence, even that might be the change you're looking at in the coding sequence, or there's embedded regulatory domains nearby. And that's why you're getting this sort of fixation sweep analysis. Well, we asked the question about whether it might be fixed regulation at AGAP1 that is affecting other places. If you look at genomes across all gnathosomes, all jawed vertebrates, the AGAP1 and GBX2, the gene right next door to it, are in complete synteny. They're always pointed at the direction that they have in every genome, and they're never actually recombined away from each other. This already is like a red, like, look here. In the introns of AGAP1, we're actually able to see there's large UCEs that we pulled out in our study that are conserved amongst all nathosomes. And this is what we call Basically about 340 base pairs are almost identically conserved amongst all ray-finned fishes. That's amazing, that level of conservation. This is right next to a human accelerated region, the HAR, originally found to accumulate mutations through primates up to the human lineage. We'll talk about that in a minute. But importantly, that in gnathans, this UC is not found. That's unique to jawed vertebrates. We can find it in humans and chicken and so forth, and it's about 190 base pairs long, so that there is some sort of degeneration of that UCE in amniotes. So hypothesis that this particular UCE is regulating GBX2 right next door. GBX2 is a hallmark gene that says, well, this looks like this could be functional here. This is the particular expression of the HAR regulation called HANDC1 at the time. You see it's in the branchial arches here as well as the limbs, developing limbs of the mouse. In both chicken and zebrafish, the GBX2 is expressed in the branchial arches you see here as well as here. We think GBX2 might be actually leveraging the phenotypic outcome of this particular enhancer. We made a targeted knock-in of GBX2 into the endogenous locus. In zebrafish, and we can see here is that we see that Gbx2 is expressed very early in development in the forming branchial arches. This will lead to the formation of the jaw, as well as some of the chondrogenic elements of the suspensorium, the feeding apparatus underneath. After larval development, it actually is expressed here in the lateral aspect of the lower jaw, as well as a little bit in the gill filaments, but not in the chondrocytes themselves. But it persists into late development. We have made a transgenic where we've taken this 343CNE driving this reporter and injected them and made full transgenics. Out of about 9 different lineages or lines, each of these, this is representative of all of them, where the 343 is sufficient to actually express into the hinge margin, of the juvenile to adult jaw. And so here is the frontal aspect, and here is the ventral showing on the edges here and the jaw, the sort of hinge region. But is it necessary? Well, we're actually testing this out. We're knocking it out. I don't have that data yet. But one way to ask this is to go back to biology. So I said, when given lemons, make lemonade. So I told you about this switch that you have overgrowth, but it's lost here in the flying fishes. So can we study this loss and ask the question about, is this necessary as a component of that? When we look at the 343 in the flying fish lineages, we see that all of them contain a very large deletion of the most terminal endpoint here in all this, and it's been fixed throughout. Whereas every other vertebrate we see has this particular 343 selected for and conserved. So how do we test this? We can go back to the zebrafish. We make deletions of that particular part of the 343 and ask what it does. So here is the wild type 343 sequence at the hinge margins. Do a 12 base pair deletion of this. We can now see that you're getting encroachment into the midline. You make a 50 base point, base pair deletion in the 343, you now can get expression right in the midline itself. Saying that this normally, the action of this normally represses expression in the midline sequence. And so this gives us a proposed model for this hub, which is that this UCE, this 343, basically acts to normally inhibit GBX2. And when you alter that, it leads to its expression in novel domains. So here, one is the hinge and then up to the midline. Well, as we said, this is normally, we're thinking about proportion or ability to make something. It's not really the ability to make something. In this case, the expression is inhibiting that potential. So it's sort of the inverse of the way we were sort of normally thinking about this. GBX2 is an interesting gene to look at because actually recent work from Ying Zhang's lab in USC is that he looked at Cre marking of these cells in mouse. And you can see that actually by marking these early, that you can watch the lineage migrate. These cells not increase the expression, but these cells are migrating to the distal aspect of the lower jaw. They have a major role in the morphogenesis of the lower jaw itself. Let's say at E9.5, you now have basically labeling of the whole entire jaw. This leads to a broader implication of jaw proportion in all gnathosomes and integrates what is previously known on the HAR2 the accelerated region in primates to a more general regulation hub in AGAP1 to GBX2. That involves this larger 343 domain which we've identified, such that if you mutate this particular HAR domain, you get increased expression in the bronchial arches and the limbs. If you go back to what more is the ancestral sequence, you get much less regulation. We see the same thing in our mutational analysis of 343, where we get increased expression in the jaw. If you then make an evolutionary model about this, it's about 400 million years ago, we had a locus that was more centralized into more constrained ideas of 343. This has been partitioned out into separate different UCs that are maintained in naphthosomes, maybe for partitioning out basically resiliency of this particular expression domain, such that if you mutate one, it's not going to give you effect, but you have to change both. And this not only is important for natural variation, but also variation within humans as well as within primates. And I have 3 minutes left to do, and I just want to say, give you an extension of where we're going with this. And because of the success of this approach at identifying really key loci for understanding variation in phenotype that we want to really start to broaden this, this comparison we're doing. And because actually there's about 14,000 ray-finned fishes themselves, if we go out of that, it even gets more to the 30,000 I talked about earlier. And here is a collection of about 230 species, which I think is the summation totals of what we're doing so far, of the different types of groups. The primitive forms where fishes are zebrafish, planiforms, medaka,— ice fish, salmon, you know, salmon. Pathways have been selected for particular traits they show. Some of this was shown here, like for example, the ice fish, salmon. We're looking at changes in craniofacial growth as well as developmental timing. We have a paper coming out soon in Science Advances about longevity in rockfishes. As well as eelpouts. These sea robins are fantastic. This is actually sits at the base of all of the rockfish. We have study going on in Lake Baikal, which is the deepest lake in Russia, on parallel evolution between the, the, um, uh, the, the Sebastis there compared to the notothinoids in Antarctica, about parallel radiations. So it's been a really powerful method to get at a lot of these questions. That we can bring back to research models to address in the lab as well as clinical samples address what's going on in patients. And so the last thing I'll show about where we're going is what happens if we play the tape twice in evolution. Take independent lineage and ask the questions about proportion. So not only do we have proportion of the fins here, but also of the rostrum itself. And this is a skeleton of a sailfin. And I'm going to go back to David Attenborough because he does it so well. Just to show you the majestic nature of these fish and how they use these traits. This is not a model system in any way. We do not bring this back to the lab. We're not going to do studies with genomic analysis and CRISPR on them, but they're incredibly valuable because evolution has shaped their morphology that we can utilize that by these sorts of methods. And he's very poetic here, so I'm going to stop just a minute. Sailfish live a high-octane life. To survive, they must find prey daily, so their entire existence will be spent on the moon. High octane life. I love it. Does his job well. All right, so where we're at in this is getting tissue samples. Getting tissue samples here, like other ones, it's hard to get really high quality DNA because you get this on a fishing vessel. Often it's a plug of a fin or something that someone's taken, or a piece of a beak that they took home as a souvenir. Currently, we have collected samples of every one of these particular lineages. This is a particular clade that comes next to an outgroup that has Jackson mackerel and barramundi lineages that don't have these traits. These are all new and fixed traits for these particular lineages. This is done in collaboration with Naturalist, been working on this for like 30 years, so it's not me going out there and catching these things. I wish I could act that cool, but these things are like 600 pounds. There's Molly Ludwig, who has been at UMass and Northeastern. Kuei Tam is in Taiwan, and Captain Tony is a fisherman off of South Carolina. But interesting amongst this, they have exactly the same scenario as the flying fish. In terms of the rostrum. Where there's a lineage here, the short-billed spearfish has lost this proportionality. So right now we're looking at several traits. One is fin length in these things, which would be shared primarily about the sunfish. There's several other that have increase in, in fin length, but rostrum length, we're using the fact that it's shared amongst all this group and that here it's now lost as ability to go and test this idea. Now the elephant in the room is, have we tested the 343 in this? And we have, and the 343 is absolutely conserved through all this. However, we can only PCR out a piece of this from the short bill. And so we're going back and now actually doing the genomes to actually assess what's happening at the locus. Not only can we look at actually species level and macroevolution, but there's an interesting story about intraspecific variation. This is actually my lab. It got sent a whole bunch of spears from swordfishes. Swordfish is actually only one species. However, you look at their bills, you put them up to the light, about half of them have what is a completely solid rostrum, all bone, and the other half have these little chambers. The little chambers are square, and being in orthopedics, I'll tell you, bone never makes square. Right? This is something very developmentally patterned and is basically determinative in the developmental of that structure. And so right now we're pulling samples between the fully ossified and the squared versions to actually get at this question. And these are the people involved. This is Kay, this is Jirila, two postdocs involved in the project. You can see some of the swords on the floor. So with that, I'll end. A lot of this work, if not the majority of it, was run by a very talented postdoc named Jake Donne, who's now at the University of Houston. Like all good people, I was scared of all the magnitude of the data he made, so I sent it all with him. So he's currently trying to wade through this in his new lab and having a lot of fun. Part of the phylogenetic group is Steve Treaster, Kate Woronowicz, and Joel Castro is involved in this, and I think I've discussed people who have been involved more peripherally. And just to thank not only the funders but also naturalist groups, for example, Central Marine Institute in the Bahamas, as well as large pelagic organization for looking at the billfishes. So thank you very much for joining both online and in person. I don't know if there are any coming in, but here, I don't know if you have any outside questions. Okay, I'll go. I'll stand back. So super interesting talk. And your focus in this talk on sort of morphometric measures that one can make. And I just wondered, is there also a collaboration with naturalists or biologists to think about social and behavioral aspects of the fish? You sort of hinted at it by having the schools of fish. I mean, how do they collaborate as communities, and is that something that's genetically encoded? Can Zoom people hear that? Do I need to repeat? I'm at a microphone. I don't know if— Anyhow, the question is, can we look at some of the behavioral aspects in some of these lineages or have we done so? The answer is we have not done so in part because behavior is a very complex trait to track. One of the things I learned very early on in my days in Germany doing screens is that the power of your screen is based on the simplicity of your output. And morphometrics is the simplest of ways we can go about that, such that we can go to museums even on something that's dead and measure the proportions. And that is a true phenotype, whereas we'd have to go and have these in our hands to do this behavior. We have published papers on the overlap between our morphology and behavioral phenotypes, like FGFR1 actually turned out to be essential in aggressive boldness behavior. We only found that out when we tried to mate them and the males kept eating the other one. But it's a whole different area and much more complex that the simplest screening tool is often the most empowered is what our argument has been.

That was a great talk. That 343 region looks like it's at least a gene regulatory element, probably an enhancer. And there's a lot of evidence coming from other lineages that some of these gene pairs with an enhancer that's embedded in one or in between them are conserved not only within, say, the, you know, fish arm, but all across vertebrates. Have you come across one of those? The FGFR ones kind of jump out for me, but—

So the answer is yes. Even in this case, we found that the HAR UCE as well as RUCE often are always found together. There's differential shaping of how large the conservation is in different lineages, but they both can be found throughout all jawed vertebrates itself. So somehow, and from things that didn't have jaws, the things that had jaws, all of a sudden it's now fixed there, right? You can't afford to lose it completely. We have not gone and looked at the FGFs. And the FGF receptor regulations itself. In part, it's sometimes you're only going to see what you're ready to see. Because the amount of data in these datasets that you look at anytime you do a genomic study, things pop up even after you publish, after you think you know what's going on. There's more character, more lace edges to the story. But we have not directly addressed the FGF receptors to get more of the direct question you're getting at. And just to follow up, it's how many gene pairs with these HARs are conserved? How many gene pairs? So if you have gene A, gene B, and you have an HAR that works on one or the other. So I currently don't know if anyone's looked at that particularly. I mean, James Noonan's group has done a lot of work on this particular site. And he was one of the first to do this sort of HAR analysis. I would bet that he or somebody in that sort of space actually might have gone and done that work itself. But I currently don't know. That's not— yeah. We have some online. Yeah. So the first question is, do the potassium channels determine a voltage drop across a field which has a regulatory role? And in addition to that, it's, uh, it's still hard to understand how membrane potential relates to size control. Are there various theoretical models that propose potential explanation? So what was the first part of that? I got the second part. So the first part is, do the potassium channels determine a voltage drop across a field which has a regulatory role? I got you. Um, the answer to the first part is There's conceptual models made to actually try to understand how potassium channels might be doing this. Work with Mike Levin's group has done quite a lot of work on that. It's still entirely unclear if they can dissociate shifts in the bioelectric resting membrane potential against a more broad sort of electrical potential of enjoined cells, make a tissue. That actually has a potential within itself. And also, my viewpoint is also the cytoplasmic connection to canonical signaling pathways. We found that actually is essential for the readout. So it is a very intricate sort of switch between the environmental signals, differential signals, and also internal that can regulate basic properties of cell division, cell differentiation. Now, how this works across a tissue, this is entirely unknown in my viewpoint. And it's unknown where this might have come about. One argument that I find very attractive is it came about from bacterial biofilms, which act to signal actually nutritional status from the center to the periphery. They actually use potassium and potassium channels to actually do this. And so some of the things we pulled out, like the LAT4, is actually an amino acid transporter, which actually might signal nutritional information into cells. So, maybe there's a connection between nutritional status, bioelectricity, and growth potential of that particular tissue in a coordinated fashion. But I think it'd be fantastic to look at it. I'm looking at you who you asked the question. Fantastic to ask across a tissue, how is this being regulated in terms of a field? And conversely, is there any integration in terms of regulatory network, but we don't have the tools right now to actually visualize that in a developmental state, either in the embryo or in the adult. To get at your idea about modeling, which is the second question, I'm not aware of any models in an embryogenic standpoint that are sufficient to characterize the output. That would be, in my mind, there are no sufficient models. Great.

Thank— we're right at time, so I'm going to take the last two quick questions. First one is, as prices drop in sequencing, do you see yourself moving to doing full genomes and looking for not only variations in conserved but emergences and segmental duplications and that sort of thing? That's a quick one.

Go ahead. The answer is yes. We stopped making the targeted capture arrays because they cost too much. And so even with the billfishes, we're just doing short reads. We cannot do chromosome level, as I mentioned before about the major thing I view in the future is going to be the limiting reagent for doing genomic analysis is source of quality tissue. And we don't get any, we don't get tissue in a diversity and taxonomic breadth enough to actually do chromosome level analysis. And so yes, we can look at things like segmental duplications and stuff like this, but that's in the future from where we're at.

And my second question is, how far do you think you could back out in terms of closeness of association? I'm wondering, like, could you go take limbed lizard, say lizard clade, go against the fish, and actually start looking back at like the emergence of limbs and at that level, looking for the things that are just fixed in terrestrials versus fixed in fish or such. I mean, how, what do you think that resolution is?

Well, I mean, so there's a lot of just-so stories that you can make, right? People are doing that all the time. And we look at like mutations in amniotes versus fish. The major problem we have is mapping orthology. Back through sort of the gar and coelacanth days of vertebrates, and which, which particular, like, orthologous gene structure maps back to that of fish or vice versa. The other is that we found in our work we're able to actually capture and analyze things back to about 50 to 75 million years, and then we start to get too much noise. So the real sensitivity, the real power comes in looking at clades that are at the, the younger end of that. And that sounds like a limitation, but if you look at the radiation of most ray-finned fishes, they all occurred around the 50 to 75 million year slots, um, such that you look at these amazing diversity and radiations of forms and they all fall within that window of time that you're empowered to do that. If you go back further, there's too much time for drift and too much time for accumulation mutation that it gets foggier and harder to deal with. Having said that, in certain clades they have a slower diversity or slower mutational rate, and you can get signals like some of these, uh, uh, ray-finned fishes are 75 to 90 million years apart, and we're able to actually track back to with some power. But yeah, I would not start with ray-finned fishes and start talking about lizards. I'll go into a lizard clade and start talking about lizards.

All right, I think we're over time, so let's thank Matt one more time.