For this whole lesson, we are going to look at some actual empirical data in the same general shape as the simulations you've been running. Our first stop involves the ground finches from Daphne Major, monitored from 1973 to 2012 by the Grants. In the 2000s, genetic sequencing became cheap enough that a lot of the specimens the Grants had been collecting, and especially the new birds they were still monitoring, could be sequenced. The genomes could be analyzed, and the specific alleles being passed on from parent to child could be correlated with the phenotypes that the Grants had been measuring the whole time.
A number of loci were found associated with each of these traits, as you might expect: the characters they were looking at are controlled by a great many loci. However, a few particular loci stuck out. HMGA2 had a strong effect on beak size, depending on which allele you had; ALX1 had a strong effect on beak shape; BMP4 had a strong effect on beak depth; and CALM1 had a strong effect on beak length. Multiple alleles for these and other loci (as, again, each trait was controlled by a large number) were then monitored over time. We're going to track one allele of HMGA2, controlling beak size.
In the simulations below, it's fundamentally the same as what we did before. Rainfall controls how much food is available; how much food is available controls births and deaths. If you don't die, you're a survivor. If you are born, you're a new chick. And the frequency of the big-beak allele is going to be a function of who lives (who survives, who doesn't die), and who is born, and who their parents are.
Where this gets into selection is that beak size is going to have an impact on how rainfall creates food. When the rain falls, seeds are made. All of the birds can eat small seeds, but only certain birds can eat the larger, harder seeds. As a result, how much food is made available to a bird by the falling rain is itself a function of the bird's beak. And so here you're going to run a few simulations to see how the frequency of an allele changes when you can mess with the association of the trait with food acquisition, relative to the amount of births and deaths that are random.
Essentially, in the graph of the model shown below, you're going to be able to influence drift directly, through the role of chance in who lives, who dies and who reproduces; and the role of selection, in terms of how specifically the allele affects food availability as a function of rain, and in what direction. You'll fit this a few times to simulated data, to get a feel for how your manipulations impact it. And then you'll get one chance to try and match the empirical data that was actually seen in the Grants' birds.
In Part A, you were predicting the frequency of an allele at a single locus. A diagram was given showing its context in the genome, amongst many other loci, for that chromosome and several others. Meiosis does not pass on individual alleles; rather, in sexually reproducing diploids, you pass on an entire haploid cell. Recombination scrambles some loci, but genomes are still inherited as chunks instead of individual pieces.
Here we're going to run some simulations where we're tracking the frequency of a new allele, but we're going to be concerned not only with its frequency over time, but also its genomic context. The model here will be extremely simple. We are just looking at the advantage of the allele, and the effects of chance (drift, basically), on the rate at which it spreads. The focus here is on how the rate at which it spreads affects the diversity in other places in the genome: places that have no effect whatsoever on the phenotype of interest.
This is based on the FGFR3 receptor, a protein that is related to fibroblast growth factor. If you look at this particular locus in dachshunds versus non-dachshunds, you'll find that dachshunds have almost zero genetic diversity in a region about 20 megabases, or about 20 million base pairs, on either side of FGFR3. All dachshunds are homozygous for all the same alleles. It's a very, very big region of zero diversity compared to normal dogs in the same area, even though leg length has been important to all dog breeds.
The size of the homozygous region is going to be related to how fast the allele went to fixation: how fast the allele drove all other alleles extinct at that locus. The faster this happens, the less time there is for recombination to separate one locus out from its neighbors, producing a situation where a strongly advantageous allele sweeps the area around it free of diversity, just based on the speed of its wiping out variation around it, relative to the speed of recombination.
Most of the examples we've looked at so far are simple. One locus controls one trait. Different alleles give you different forms of the trait. In lecture, we've talked about how a lot of continuous traits don't work that way: they're polygenic, influenced by many loci at once. Additive genetic variation, where each locus's alleles simply add to the trait, is one example. In Parts A and B, we kind of looked at how one locus can affect the other loci around it, even if those other loci have alleles that don't affect the trait of focus. However, they still have implicitly in them the assumption that a given trait is influenced by a given locus. Unfortunately, that is not usually the case. In most scenarios, we have two related but distinct phenomena happening for a given locus.
Pleiotropy is when one individual locus affects multiple traits. An example of this would be T-box 4, which plays an important role in arm development and in heart development. So if you see a baby born with a messed-up arm, that can be a sign that they also have a heart defect, because the same gene is responsible for multiple traits. That's pleiotropy. Most genes are going to influence more than one trait.
Epistasis is when the effect of one gene on a trait depends on which alleles another gene carries. You can think of this as the difference between a general-purpose tool and a bunch of specialized tools. Most genes are going to produce proteins that are multipurpose, that are going to influence things in a lot of ways. A good example I like is of a kitchen. A high-epistasis kitchen would be one that has a stove and an oven. You can cook basically anything with a stove and an oven: you only need two tools, the stove and the oven, and almost any food is makeable. Low epistasis would be a kitchen that has a bread baker, a rice cooker, a pressure cooker, a microwave, an air fryer, a crock pot, and a stove and oven. There are downsides and upsides to both. In high epistasis, if your oven breaks, you just can't eat, basically. In low epistasis, if your bread baker breaks, you can bake bread in your oven and it's fine. The flip side is that in low epistasis you have a lot of different tools, a lot of specialized tools, and that can present a number of problems.
For this activity, you're going to have a more realistic diagram, though still simplified. There will be two loci and two traits. The two loci will be positioned at some position relative to one another: they could be very close, they could be very far. Each one will have a different impact on a different trait that you can adjust, and these traits will have different impacts on fitness. Drift, as always, will be occurring as well. Your goal in this activity is to explore how having multiple loci, multiple traits, and multiple different directions of impact can change how alleles increase or decrease in frequency. Recombination and drift, again, are always there and always present.
The individual simulations for this one are each going to be based on an actual empirical data set, showing simplified versions of real data, to give you the sort of actual information about trajectories that you can see in real data.
For this final part, we're going to dip our toes into the aspect of selection that, as I said, can make things go a little bit off the rails, which is frequency dependence. In frequency dependence, the fitness of a trait, and therefore any alleles that govern it, is itself influenced by the frequency of that trait, and so of the alleles that govern it. That means that a trait impacts its own fitness based on its frequency in the population.
The example I mentioned in class has to do with side-blotched lizards, where the males come in different color patterns based on, basically, the amount of testosterone they have. So you can think of this as a testosterone-allele sort of situation, though it's somewhat more complicated than that. Ultimately, the lizards are going to have three different traits: they can be orange, yellow, or blue. Orange lizards have the highest testosterone. They grow the largest and are the most aggressive. They control large territories. Yellow lizards have the lowest testosterone. They're the least aggressive, and they actually are hard to distinguish from females. They cannot control a territory, but at a distance they look like females, and so can pass unbothered through other males' territories if they can stay far away. Blue males have an intermediate amount of testosterone. They can control small territories. Because their territories are small, yellow males have a hard time passing through them. But because the individual males are medium-sized, blue males will tend to lose in direct confrontations with orange males.
This is basically a rock, paper, scissors setup that exists in nature. Blue beats yellow, yellow beats orange, orange beats blue. And so how common orange is going to control the fitness of yellow. If there are a lot of orange individuals, being yellow is a good strategy. If there are a lot of blue individuals, being yellow is a bad strategy and being orange is good. If there are a lot of orange individuals, being yellow is good and being blue is bad. This will produce cyclical changes in allele frequency over time. What you're going to do is toggle, basically, the relative strengths of the interactions (to what extent does the rock, paper, scissors exist?) and observe what that does to the frequency of the traits over time: spinning cycles.
Note that these graphs are a little complicated. These are going to be on what is called a ternary diagram, which is basically a triangle, as opposed to the normal sort of rectangular plot. The frequencies of three different traits are shown on the triangle simultaneously, and so a given point within the triangle shows the frequency of all three, blue, yellow, and orange, at the same time. It can be difficult to read, and I would always encourage you to do practice runs before starting the targets, to get a good sense for what the different arrows do and how to read the plots. I would particularly encourage it here, as ternary diagrams tend to be difficult for students at first.