*Featured image photo credit: Lisandro Milocco

In this episode, Lisandro Milocco (PhD candidate, advised by Dr. Isaac Salazar-Ciudad) shares insights from his new paper, Milocco and Salazar-Ciudad 2022: ‘Evolution of the G matrix under nonlinear genotype-phenotype maps.’ We discuss how Lisandro likes to think about genotype-phenotype maps, how genes control complex developmental processes such as the generation of mammalian teeth, and how the interaction of genes and the developmental processes they control change our expectations for how quantitative phenotypes evolve. Listen to our conversation and then read Lisandro’s full paper here!

Want to develop your understanding even further? Email Lisandro at lisandro.milocco@helsinki.fi.

NATURALIST SELECTIONS IS AN INTERVIEW SERIES PRODUCED BY THE AMERICAN SOCIETY OF NATURALISTS GRADUATE COUNCIL. WE SHOWCASE GRADUATE STUDENT AND POSTDOC AUTHORED WORK IN THE AMERICAN NATURALIST, A PREMIER PEER-REVIEWED JOURNAL FOR ECOLOGY, EVOLUTION, AND ANIMAL BEHAVIOR RESEARCH. CATCH UP ON EXCITING NEW PAPERS YOU MAY HAVE MISSED FROM THE JOURNAL, AND MEET SOME TRULY BRILLIANT EARLY CAREER NATURALISTS!

Listen on Spotify and Apple Podcasts!

Credits

Featured Guest: Lisandro Milocco, University of Helsinki, Finland

Host, Editor, Producer: Sarah McPeek, University of Virginia, US

Original Music: Daniel Nondorf, University of Virginia, US

Transcript:

You’re listening to Naturalist Selections, a science podcast featuring graduate student and postdoc-authored research in The American Naturalist, produced by the American Society of Naturalists Graduate Council. I’m grad council rep Sarah McPeek and today I’m talking with graduate student first-author Lisandro Milocco! Lisandro is finishing up his PhD with Dr. Isaac Salazar-Ciudad at the University of Helsinki, Finland. His paper is called: “Evolution of the G matrix under nonlinear genotype-phenotype maps.” In the field of quantitative genetics, A G matrix is a statistical summary of the additive genetic basis of a phenotype in a population. G matrices are important evolutionary tools for understanding how natural selection has shaped genetic variance in the past and predicting how phenotypes may continue to respond to selection in the future. Typically, geneticists assume a linear relationship between the G matrix and the phenotype it produces, causing G to remain constant for long periods of a population’s evolution. However, this assumption of linearity does not reflect the fact that many of the genes involved in producing a trait control complex developmental processes that ultimately generate the adult phenotype. Lisandro views development as a critical and fundamentally nonlinear intermediary between a population’s G matrix and its average phenotype. Can G change more dramatically if we incorporate an evolutionary developmental approach? In the paper, Lisandro uses his lab’s longstanding simulation model of mammalian tooth development to explore how complex processes during development shift our expectations for the evolution of G. In the model, small groups of genes specify 21 developmental parameters that interactively guide the formation of the cusps and valleys of the adult tooth. For each simulation, he chooses a different phenotypic optimum for tooth shape and allows the population to evolve toward that optimum under selection, tracking how G changes along the way. He finds that the size and shape of G can change rapidly as a result of the complex gene interactions that occur during tooth development. A slight change in the alleles controlling one developmental parameter can have cascading effects on the entire developmental process, resulting in a very distinct tooth morphology, and consequently, a different relationship between phenotypes and the genes that underly their development. His work suggests that predicting phenotypic evolution using quantitative genotype-phenotype maps requires a thorough understanding of how genes control phenotypic development. I spoke to Lisandro to further develop my own understanding of these complex interactions.

Sarah

I feel like every time I think that I’m close to understanding GPM, someone says something that totally throws me off what I thought it was. So I would love it if you would explain a little bit about how you personally think about genotype to phenotype maps.

Lisandro

Yeah, absolutely. So for me, the genotype-phenotype map, in a way, it’s a mapping function that assigns phenotypes to a genotype. But the main thing about it is that it is a representation of the generative process that creates the phenotype. Right. So that process is ultimately development. So basically, the genes, for example, play a very important role in the developmental process. But there’s also interaction with other things such as biophysical properties of the cells and mechanical properties, etc. And that also feed back into the developmental process. So in a way that the kind of development can be seen as a sort of like an origami. When you fold the paper to create that, you start from a flat paper and then you fold it and you create this three dimensional shape. So in a way, for my understanding of the genotype phenotype map is that it summarizes this complex folding process and the creation of the phenotype. And particularly it kind of summarizes it in a way that is compatible with a lot of mathematical formalism. So that’s why it’s particularly useful for me, because it can be incorporated into models and things like that.

Sarah

Oh, I like this folding paper metaphor you bring up. So do you think of the paper itself is the G and then the way it’s folded?

Lisandro

Right. So maybe development I would say that maybe development could be the series of steps that you need to follow. So you need to follow the paper here, and then you need to twist it and follow it there. So in a way, development kind of summarizes all those steps that allow us to produce the final shape in the end? That’s a very complex thing. Of course, there’s allegory with the origami it’s hard to say what are genes and et cetera. But the important thing is that there’s a complex process that can in principle, be summarized by a metaphor such as the genotype-phenotype map.

Sarah

Yeah. That also makes me think of tooth development specifically, because I don’t know a lot about two development, but the way that you described it in the paper sounded somewhat akin to an origami-like process. Would you say that’s accurate?

Lisandro

Yeah, one can say so, of course, in a very simplified way. But yeah, tooth development is an interesting process because it also starts from, like a flat sheet of cells and basically within this sheet of cells, there’s buckling and folding. That kind of ends up resulting in mountains and valleys that then become cusps and valleys of the teeth. So in a way, it is somewhat similar. But the important thing is that there is not a hand that’s holding the paper as in an origami. But this is actually an internal process of the developmental system which makes it even more complex and interesting really. So, for example, with tooth development, all this folding and buckling results from differential growth of the cells. So some cells divide faster than others, and that creates this buckling. Kind of like the idea that the process itself kind of generates the form. I think it’s very important and interesting. Yeah.

Sarah

So there’s the mechanical constraints of how cells can move and which cells can divide when. But then there are also all of the signaling cascades that are going on to program when that happens. And. Oh, that’s complicated.

Lisandro

Exactly. Yeah. And everything is kind of interdependent on each other. For example, during tooth development, there’s a transmission of certain signaling centers that kind of direct surrounding cells to divide faster, but the signal themselves do not divide. And then it’s interesting because the signaling centers also have a certain inhibitory region. So you cannot have two of these too close to each other. Basically, this signaling process is occurring at the same time as the tissue is folding. So then you have even more of a feedback here going on because, of course, if this tissue folds a certain way, then the signaling centers will be distanced in different ways.

Sarah

Wow.

Lisandro

Of course, this type of thing takes years and years of developmental biology research to be able to get into even having candidate genes and how they interact and things like that.

Sarah

So the lab that you worked in for your PhD has done a lot of this work. It looked like it was in ringed seals as the animal model? That’s so cool!

Lisandro

Yeah. It’s very interesting, actually. Here at the University of Helsinki, there’s a very long history of research in teeth. So a lot of what is known about tooth development actually comes from some of the labs here. Then also, it’s been studied through the paleontological perspective and evolutionary perspective. So there’s been a way, a long history here of tooth development research and teeth in general.

Sarah

Wow. I didn’t know that. Teeth have played a really outsized role, I feel like, in our understanding of the evolution of morphology. Right. From a paleontological perspective, it’s one of the things that preserves the most. And when you look at phylogenies of fossils that have teeth, you’re usually looking at tooth morphology as being most of those character states. So it’s definitely a big part of evolution.

Lisandro

Right. Exactly. And it’s a nice system in that sense that you can study from the developmental perspective, but you actually have preserved data to also compare and do interesting things.

Sarah

Yeah. So what got you personally excited to study how this G matrix that underlies this GPM can evolve?

Lisandro

Right. So this interesting process that creates form this development process with the tooth actually leads to a complex relationship between the genotype and the phenotype. Right. And particularly a nonlinear relationship, which basically means that a given perturbation does not necessarily produce a change in the phenotype. So a change in the genome is not proportional necessary to change the phenotype. Right. And that is interesting because most of the research that has been done on G evolution assumes a linear relationship between genotypes and phenotypes. So when people do models, particularly when people do models about G evolution, it is assumed that you have a genotype and each gene contributes an amount, and then you sum those contributions and then you get the phenotype. Right.

Sarah

Yeah.

Lisandro

So basically the phenotype is a linear combination of these genetic values. So then the interest was to see, okay, what happens when you deviate from that assumption and you actually introduce something that results in this relationship not being linear? And how does that affect the evolutionary dynamics and particularly the dynamics of this G matrix?

Sarah

Just to make sure I’m understanding, can we try and talk a little bit about how different changes in G manifest biologically? We talk about the size of G, the G matrix changing. What does that mean in biology speak?

Lisandro

So basically, the size describes, in a way, the total amount of additive genetic variance. So in a way, it is related to kind of a measurement of how much heritable variation there is in the population that you’re studying. Right. So the size directly does not give you any information of how the variation is distributed in the sense that there’s a lot of variation exclusively in a single direction. So then evolution will be quite restrained to occur in that direction according to the theory of quantitative genetics, but kind of the size, it just tells you how much variation there is. So it’s interesting to see if that changes and tells you that at certain points, there is more potential or more total variation than at other points. If you’re setting the variation in a single direction for some of the direction of selection, that’s something that has been called evolvability. So then you can also see how that amount changes, and it tells you, okay, at this time point, for example, the population has this amount of potential to evolve in this direction. But then at this later time point, this amount is less or more.

Sarah

Yeah. That was a really interesting pattern, I thought, from your modeling work, which makes a lot of sense, is that as you move towards your selective peak, you would expect the size of the matrix to shrink because there’s less additive genetic variation if it’s being selected to increase these certain variants and decrease others. But then as you get closer to that peak, it seems that selection weakens. And then there’s a lot more potential for new changes to take over.

Lisandro

Right. And it has to do a lot with this nonlinear relationship and this interesting, complex interaction that actually produces the phenotype. So maybe you would expect that selection should kind of reduce the total amount of variation. But if suddenly you’re in a region of this genotype phenotype map in which small changes can produce a lot of phenotypic, small genetic changes produce a lot of phenotypic changes that will actually increase your variation. The only way that can happen is if you allow the relationship of genotypes to phenotype to change. If you assume that it’s always fixed, that type of thing can never happen.

Sarah

Right. But as we cross this developmental space, we’re having very different interactions among those genes and with the physical environment, and that’s what’s creating this warping of that interaction?

Lisandro

Right. So changing the combination of parameters kind of changes the dynamics of this folding, this origami folding somehow. And then, for example, creates a shape that wouldn’t have been possible for a different combination of folding parameters, let’s say.

Sarah

Got it. Yeah. So I do want to talk a little bit about those development parameters. Specifically, as early on in grad school, I took some courses with my fellow grad students and with faculty where we talked a lot about distinguishing genotype and phenotype and different levels of phenotype. And there are different ways to think about this. But in your model, you kind of treat it as this intermediate between the genotype that creates all of the possible combinations of phenotypes and then the phenotypic outcome that selection sees, which is the physical shape of a tooth. So what would you call these development parameters then? And why did you give them that kind of intermediate role?

Lisandro

Right, absolutely. I would say that intermediate this development parameters, such as, for example, diffusion rate, they are definitely also phenotypes. So, for example, even RNA folding, which is quite the closest you can get to an actual sequence. But the reason, I guess, with the focus here on the kind of higher level phenotypes is first of all, I’d say shapes are more interesting or more attractive to look at because it’s not endless diffusion rates most beautiful. Forms are cooler. So there’s that. And of course, there’s the issue that those are the things that actually interact with environment. So maybe they’re sort of the more interesting for fitness and stuff like that. And also there’s kind of defining this middle level development parameters are also important because mechanistically it is these parameters that really determine the dynamics of development. So the genes, like just the sequence of the gene doesn’t tell you much, but actually, for example, how the protein product, for example, diffuses and actually interacts with other things. That’s what really will determine how this whole thing will happen to produce the shape. So it’s kind of an important middle ground to include.

Sarah

So that makes me think that the targets of selection themselves are not so much the physical shapes of the tooth, but what selection is really shaping, what evolution is shaping is really the interaction of these parameters, which is controlled by the underlying genes themselves.

Lisandro

Right. The way selection was implemented here is through a given shape. So there’s an optimal shape and you have to go there. But of course, how you go. There depends a lot on how all of these genes and the different parameters are interacting with each other to produce the shape.

Sarah

Do you think that that trajectory mostly affects the pace of adaptation, or does it affect other aspects of the adaptive process?

Lisandro

No, definitely. I would say in principle, it could affect everything. For example, there’s the possibility that this development process or this folding origami process cannot produce certain shapes. So, for example, there’s no way that you can shape a paper to produce a certain shape. So then even if that would be the most optimal shape, there is no way that you could actually make evolution go in that direction. So in that way, the actual origami process would completely block the adaptive process in that direction.

Sarah

Yeah.

Lisandro

So it could even have that extreme effect.

Sarah

Yeah, that’s very true. We could cross into an area of space that we can’t come out of.

Lisandro

Right. So you kind of will get stuck somehow in the process. So we definitely would affect the adaptation there.

Sarah

What result from all of this work that you did most surprised you or got you really excited?

Lisandro

It was exciting for me to see all the different ways in which G can change. Particularly, for example, the dimensionality of G can change, which kind of means that as we were talking before, the directions or the number of directions in a way in which there is this additive genetic variation can also change, which kind of tells you that at certain points, maybe the population is being able to produce variation in certain direction, but at other points it’s not. So somehow I think that was particularly interesting because then it’s like in a way that the constraints change. Right. Right. So I think that’s kind of interesting. And also seeing how the G is a sort of a snapshot, the statistical snapshot of the population with a given distribution of genotypes at a given time, and how that snapshot can kind of change in time. I think that’s an interesting thing to see in the simulation sense.

Sarah

Yeah, absolutely. And gives us a lot of insight about some weird empirical results we’ve gotten where G did seem to change really quickly and we couldn’t explain that.

Lisandro

Right. I think there’s a lot of very because in the simulations, the environmental component is not really explored that much. But I think there’s a lot of very interesting research also about how just changing the environment can completely change the G matrix.

Sarah

You mean that if the environment changes, then the adaptive peak might shift.

Lisandro

Right. And even just the G matrix itself can change its shape quite rapidly just by putting the same population in different environments.

Sarah

Oh, really?

Lisandro

Yeah.

Sarah

How does that work?

Lisandro

Well, that’s the thing, because phenotypes are not just the product of genotypes. Right. They are the product of interaction in this complex folding way. So in a way, the G matrix tries to separate how much of the phenotypic variance is associated with this additive genetic effect. But that separation can be problematic when the genes are interacting with the environment. It’s not cut and dry that this is exclusive from genes and this is exclusively from environment. But it will affect whatever you assign to the genotype and whatever you assign to the environment.

Sarah

That makes sense. Yeah. Another thing I thought a lot about was the idea of pleiotropy, not just in the sense that one gene that’s involved in tooth development might impact multiple aspects of tooth development, but that the genes involved in tooth development are likely involved in the development of other structures in the body, like other bones and other morphological features. I don’t know, but I guess that led me to think about how big is G really, just in terms of the phenotypic breadth that a G matrix has to cover in order to be really predictive?

Lisandro

Right. Yeah, I think definitely this idea of pleiotropy is a very important concept, particularly from the evo devo perspective. We know that for the development of many different organs, you kind of end up using the same toolkit of genes. So for example, particularly for teeth, which are ectodermal organs, they share a very similar developmental kind of pathway to, for example, hair and feathers and glands, for example. So in a way, the gene families are basically conserved for this difference. And actually the buckling process is similar to a large extent. So definitely the question of pleiotropy is very important. You meant basically how does this other kind of pleiotropic interactions, how do they affect the G for teeth?

Sarah

Right, right. Exactly. If you’ve got one phenotypic optimum for teeth, but a very different one for another morphological feature that’s also adaptively significant, how are we going to navigate that?

Lisandro

Absolutely. Yeah. There would be a compromise there. For example, I would do the simulations, but there’s also some, for example, stabilizing selection or some other type of selection. Working on something that’s related to my trades in the teeth that I’m focusing on. There’s probably going to be, for example, it’s very possible that this directional selection that I’m applying the teeth isn’t as effective because I have this unaccounted stabilizing selection working on the same genes. Basically, that doesn’t let me move so much in the gene that’s. So there will definitely be kind of more of a constraint maybe, and like more of an arrest than in the simulations where basically in the simulations the individual is the twos. So we remove all other possible traits, which is of course, a specification.

Sarah

It’s necessary. Teeth are complicated enough!

Lisandro

Exactly. Yeah. We need to focus on something. That’s why.

Sarah

But if we’re thinking about using development as a way to understand these genotype to phenotype maps, do you think that that’s an important thing to try and build in in the future, to really predict what the outcomes are for these individual structures but also for the creature as a whole?

Lisandro

Absolutely. And I think that’s really something that it would be important to develop. This kind of way of thinking is to design ways to integrate all of the complexity somehow because that’s the thing in quantitative genetics, for example, by assuming that this relationship is linear, you can do a lot of things that become harder with this other way of thinking, but definitely for them because again, it kind of determines a lot of the actual dynamics of the whole system. So it’s something we should strive towards, I think.

Sarah

Yeah. So there are obviously so many ways that we can complicate the system. What do you think would be the next big question to tackle with this modeling framework or the next thing you’d want to build in to try and gain a little more predictive power?

Lisandro

Right. Well, actually, I think definitely this idea that we’ve been discussing about how this could be incorporated for prediction. So I think that’s something that I think could be very interesting because in a way one thinks about how a classical perspective would be like development introduces certain constraints, but maybe seeing it as the constraint, I think it would be interesting to see it as something that aids predictions somehow. It’s something positive somehow, because if we understand the rules that create variation then we can actually do more.

Sarah

Yeah. I think there’s always been this kind of negative response to anything that constrains adaptation, but, oh good God, if adaptation isn’t constrainable, then we’re never going to understand anything.

Lisandro

Exactly. It helps that things are a bit constrained, like, for example, even in life. Right. When you don’t have so many choices, it’s easier to go in a given direction.

Sarah

Yes!

Lisandro

We would just get stuck lately. So it’s cool to have a bit of, let’s say, structured variation.

Sarah

I remember when I first learned about Sewall Wright’s landscape of infinite genetic potential. I was just like, oh goodness, we’re never going to know anything about what’s going on if there’s this many possibilities. But the idea that there are these processes that send you in certain directions or put you on certain paths is to me really comforting, right? We can actually someday understand how it’s all working.

Lisandro

Exactly.

Sarah

Very cool. Yeah, well, I think this paper is a really important contribution to our understanding of GPMS and incorporating development was a really fascinating perspective. I hadn’t thought about a lot before. Congratulations. It’s a really great piece of work.

Lisandro

Thank you. Absolutely. And I’m happy that maybe if it makes you think about it, maybe in a different perspective. I think that’s the best that one can aim for.

Sarah

Yeah, I think actually it made me understand the GPM probably more than I ever have because it’s such a black box of going from genotype to phenotype and understanding what are the actual mechanistic processes controlling and translating. That relationship was really important I think to wrap my head around what this means. So yeah, you helped a very confused graduate student.

Lisandro

Thank you! Very happy to hear that. Absolutely.