Untitled

6.1 Mendelian Genetics in Populations: Hardy–Weinberg Equilibrium

Most people are susceptible to HIV. Their best hope of avoiding infection is to avoid contact with the virus. There are, however, a few individuals who remain uninfected despite repeated exposure. In 1996, AIDS researchers discovered that at least some of this variation in susceptibility has a genetic basis (see Chapters 1 and 5). The gene responsible encodes a cell-surface protein called CCR5. CCR5 is the handle exploited by most sexually transmitted strains of HIV-1 as a means of binding to white blood cells. A mutant allele of the CCR5 gene, called CCR5-Δ32, has a 32-base-pair deletion that destroys the encoded protein's ability to function. Individuals who inherit two copies of this allele have no CCR5 on the surface of their cells and are therefore resistant to HIV-1. Given that individuals homozygous for CCR5-Δ32 are much less likely to contract HIV, we might ask whether the global AIDS epidemic will cause an increase in the frequency of the Δ32 allele in human populations. If so, how fast will it happen?

Before we can hope to answer such questions, we need to understand how the CCR5-Δ32 allele would behave without the AIDS epidemic. In other words, we need to develop a null model for the behavior of genes in populations. This null model should specify, under the simplest possible assumptions, what will happen across generations to the frequencies of alleles and genotypes. The model should apply not just to humans, but to any population of organisms that are both diploid and sexual. In this first section of the chapter, we develop such a model and explore its implications. In the next section we add natural selection to the model, which will enable us to address our questions about the AIDS epidemic and the CCR5-Δ32 allele.

Population genetics begins with a model of what happens to allele and genotype frequencies in an idealized population. Once we know how Mendelian genes behave in the idealized population, we will be able to explore how they behave in real populations.

We develop our model by scaling Mendelian genetics up from the level of families, where the reader has used it until now, to the level of populations. We illustrate the model with an idealized population of mice (Figure 6.2). A population is a group of interbreeding individuals and their offspring. The crucial events in the life cycle of a population are these: Adults produce gametes, gametes combine to make zygotes, zygotes develop into juveniles, and juveniles grow up to become the next generation of adults. We want to track the fate of Mendelian genes in a population. We want to know whether particular alleles or genotypes become more common or less common across generations, and why.

Figure 6.2 The life cycle of an idealized population

The labels highlight the stages that will be important in our development of population genetics.

Imagine that the mice in Figure 6.2 have in their genome a Mendelian locus, the A locus, with two alleles: A and a. We can begin tracking these alleles at any point in the life cycle. We then follow them through one complete turn of the cycle, from one generation to the next, to see if their frequencies change.

A Simulation

Our task of following alleles around the life cycle will be simplest if we start with the gametes produced by the adults when they mate. We will assume that the adults choose their mates at random. A useful mental trick is to picture random mating happening like this: We take all the eggs and sperm produced by all the adults in the population, dump them together in a barrel, and stir. This barrel is known as the gene pool. Each sperm in the gene pool swims about at random until it collides with an egg, whereupon the egg and sperm fuse to make a zygote. Something rather like this actually happens in sea urchins and other marine creatures that simply release their gametes onto the tide. For other organisms, like mice and humans, this picture is obviously a simplification.

The adults in our mouse population are diploid, so each carries two copies of the A locus. But the adults made their eggs and sperm by meiosis. Following Mendel's law of segregation, each gamete received just one copy of the A locus. Imagine that 60% of the eggs and sperm received a copy of allele A, and 40% received a copy of allele a. That is, the frequency of allele A in the gene pool is 0.6, and the frequency of allele a is 0.4.

Starting with the eggs and sperm that constitute the gene pool, our model tracks alleles through zygotes and adults and into the next generation's gene pool.

What happens when eggs meet sperm? For example, what fraction of the zygotes they produce have genotype AA? And once these zygotes develop into juveniles, grow up, and spawn, what are the frequencies of alleles A and a in the next generation's gene pool?

One way to find out is by simulation. We can close our eyes and put a finger down on Figure 6.3 to choose an egg. Perhaps it carries a copy of allele A. Now we close our eyes and put down a finger to choose a sperm. Perhaps it carries a copy of allele a. If we combine these gametes, we get a zygote with genotype Aa. We encourage the reader to carry out this process to make a large sample of zygotes—50, say, or even 100. We have paused to do so as we write. Among the 100 zygotes we made, 34 had genotype AA, 57 had Aa, and 9 had aa.

Figure 6.3 A gene pool with frequencies of 0.6 for allele A and 0.4 for allele a

Now let us imagine that all these zygotes develop into juveniles, and that all the juveniles survive to adulthood. Imagine, furthermore, that when the adults reproduce, they all donate the same number of gametes to the gene pool. We can choose any number of gametes we like for the standard donation, so we will choose 10 to make the arithmetic easy. We are not worried about whether a particular adult makes eggs or sperm; instead, we are simply counting gametes:

Our 34 AA adults together make 340 gametes: 340 carry allele A; none carry allele a.

Our 57 Aa adults together make 570 gametes: 285 carry allele A; 285 carry allele a.

Our 9 aa adults together make 90 gametes: none carry allele A; 90 carry allele a.

Summing the gametes carrying copies of each allele, we get 625 carrying A and 375 carrying a, for a total of 1,000. The frequency of allele A in the new gene pool is 0.625; the frequency of allele a is 0.375.

We have followed the alleles around one complete turn of the population's life cycle and found that their ending frequencies are somewhat different from their starting frequencies (Figure 6.4). In other words, our population has evolved.

Figure 6.4 Allele and genotype frequencies throughout the life cycle in a numerical simulation

We made the zygotes by picking gametes at random from the gene pool in Figure 6.3 and assumed that all the zygotes survived. The reader's results, on repeating this exercise, will likely be somewhat different.

The genotype frequencies among the zygotes in the reader's sample, and the frequencies of the alleles in the reader's next generation, will almost certainly be somewhat different from ours. Indeed, we carried out the simulation two more times ourselves and got different results each time. In our second simulation, we got zygotes in proportions of 41% AA, 44% Aa, and 15% aa. The allele frequencies in the next generation's gene pool were 0.63 for A and 0.37 for a. In our third simulation, we got zygotes in proportions of 34% AA, 49% Aa, and 17% aa. The allele frequencies in the next generation were 0.585 for A and 0.415 for a.

In simulated populations, allele frequencies change somewhat across generations. This is evolution resulting from blind luck.

Our three results are not wildly divergent, but neither are they identical. In two cases the frequency of allele A rose, whereas in one it fell. We got different results because in each simulation, blind luck in picking gametes from the gene pool gave us a different number of zygotes with each genotype. The fact that blind luck can cause a population to evolve unpredictably is an important result of population genetics. This mechanism of evolution is called genetic drift. (We will return to drift in Chapter 7.) For now, however, we are interested not in whether evolution is sometimes unpredictable, but whether it is ever predictable. We want to know what would have happened in our simulations if chance had played no role.

A Numerical Calculation

We can discover the luck-free result of combining eggs and sperm to make zygotes by using a Punnett square. Punnett squares, invented by Reginald Crundall Punnett, are more typically used in Mendelian genetics to predict the genotypes among the offspring of a particular male and female. Figure 6.5, for example, shows the Punnett square for a mating between an Aa female and an Aa male. We write the genotypes of the eggs made by the female, in the proportions we expect her to make them, along the side of the square. We write the genotypes of the sperm made by the male, in the proportions we expect him to make them, along the top. Then we fill in the boxes in the square to get the genotypes of the zygotes. This Punnett square predicts that among the offspring of an Aa female and an Aa male, one-quarter will be AA, one-half Aa, and one-quarter aa.

Figure 6.5 Punnett square for a cross between two heterozygotes

This device makes accurate predictions about the genotype frequencies among the zygotes because the genotypes of the eggs and sperm are represented in the proportions in which the parents produce them.

We can use the same device to predict the genotypes among the offspring of an entire population (Figure 6.6a). The trick is to write the egg and sperm genotypes along the side and top of the Punnett square in proportions that reflect their frequencies in the gene pool. Sixty percent of the eggs carry copies of allele A and 40% carry copies of allele a, so we have written six A's and four a's along the side of the square. Likewise, for the sperm, we have written six A's and four a's along the top. Filling in the boxes in the square, we find that among 100 zygotes in our population, we can expect 36 AA's, 48 Aa's, and 16 aa's. Note that our population Punnett square has predicted genotype proportions different from the 1:0, 1:1, or 1:2:1 ratios that appear in single-family Punnett squares.

Figure 6.6 When blind luck plays no role, random mating in the gene pool of our model mouse population produces zygotes with predictable genotype frequencies

(a) A Punnett square. The genotypes of the gametes are listed along the left and top edges of the box in proportions that reflect the frequencies of A and a eggs and sperm in the gene pool. The shaded areas inside the box represent the genotypes among 100 zygotes formed by random encounters between gametes in the gene pool. (b) We can also calculate genotype frequencies among the zygotes by multiplying allele frequencies. (See Computing Consequences 6.1.)

The Punnett square in Figure 6.6a suggests that we could also predict the genotype frequencies among the zygotes by multiplying probabilities. Figure 6.6b shows the four possible combinations of egg and sperm, the resulting zygotes, and a calculation specifying the probability of each (see also Computing Consequences 6.1). For example, if we look into the gene pool and pick an egg to watch, there is a 60% chance that it will have genotype A. When a sperm comes along to fertilize the egg, there is a 60% chance that the sperm will have genotype A. The probability that we will witness the production of an AA zygote is therefore

If we watched the formation of all the zygotes, 36% of them would have genotype AA. The calculations in Figure 6.6b show that random mating in the gene pool produces zygotes in the following proportions:

(The Aa category includes heterozygotes produced by combining either an A egg with an a sperm or an a egg with an A sperm.) Notice that

This confirms that we have accounted for all of the zygotes.

Computing Consequences 6.1 Combining probabilities

The combined probability that two independent events will occur together is the product of their individual probabilities. For example, the probability that a tossed penny will come up heads is ½. The probability that a tossed dime will come up heads is also ½. If we toss both together, the outcome for the penny is independent of that for the dime. Thus the probability of getting heads on the penny and heads on the dime is ½ × 2 = ¼.

The combined probability that one or the other of two mutually exclusive events will occur is the sum of their individual probabilities. When rolling a die we can get a one or we can get a two (among other possibilities), but we cannot get both a one and a two at once. The individual probability of each outcome is

. The combined probability of rolling either a one or a two is therefore

+

= .

We now let the zygotes grow to adulthood, and we let the adults produce gametes to make the next generation's gene pool. When chance plays no role, will the frequencies of alleles A and a in the new gene pool change from one generation to the next?

If we assume, as we did before, that 100 adults each make 10 gametes, then

The 36 AA adults together make 360 gametes: 360 carry allele A; none carry allele a.

The 48 Aa adults together make 480 gametes: 240 carry allele A; 240 carry allele a.

The 16 aa adults together make 160 gametes: none carry allele A; 160 carry allele a.

Summing the gametes carrying each allele, we get 600 carrying copies of A and 400 carrying copies of a, for a total of 1,000. The frequency of allele A in the new gene pool is 0.6; the frequency of allele a is 0.4.

As Figure 6.7a shows graphically, we can also calculate the composition of the new gene pool using frequencies. Because adults of genotype AA constitute 36% of the population, they will make 36% of the gametes. All of these gametes carry copies of allele A. Likewise, adults of genotype Aa constitute 48% of the population and will make 48% of the gametes. Half of these gametes carry copies of allele A. So the total fraction of the gametes in the gene pool that carry copies of A is

Figure 6.7 When the adults in our model mouse population make gametes, they produce a gene pool in which the allele frequencies are identical to the ones we started with a generation ago

(a) Calculations using frequencies. (b) A geometrical representation. The area of each box represents the frequency of an adult or gamete genotype. Note that half the gametes produced by Aa adults carry allele A, and half carry allele a.

The figure also shows a calculation establishing that the fraction of the gametes in the gene pool that carry copies of allele a is 0.4. Notice that

This confirms that we have accounted for all of the gametes. Figure 6.7b shows a geometrical representation of the same calculations.

We have come full circle (Figure 6.8). And this time, unlike in our simulations, we have arrived precisely where we began. We started with allele frequencies of 60% for A and 40% for a in our population's gene pool. We followed the alleles through zygotes, juveniles, and adults and into the next generation's gene pool. The allele frequencies in the new gene pool are still 60% and 40%. When blind luck plays no role, the allele frequencies for A and a in our population are in equilibrium: They do not change from one generation to the next. The population does not evolve.

Figure 6.8 When blind luck plays no role in our model population, the allele frequencies do not change from one generation to the next

We made the zygotes with the Punnett square in Figure 6.6 and assumed that all the zygotes survived.

Numerical examples show that when blind luck plays no role, allele frequencies remain constant from one generation to the next.

The first biologist to work a numerical example, tracing the frequencies of Mendelian alleles from one generation to the next in an ideal population, was G. Udny Yule in 1902. He started with a gene pool in which the frequencies of two alleles were 0.5 and 0.5 and showed that in the next generation's gene pool, the allele frequencies were still 0.5 and 0.5. The reader may want to reproduce Yule's calculations as an exercise.

Like us, Yule concluded that the allele frequencies in his imaginary population were in equilibrium. Yule's conclusion was both groundbreaking and correct, but he took it a bit too literally. He had worked only one example, and he believed that allele frequencies of 0.5 and 0.5 represented the only possible equilibrium state for a two-allele system. For example, Yule believed that if a single copy of allele A appeared as a mutation in a population whose gene pool otherwise contained only copies of a, then the A allele would automatically increase in frequency until copies of it constituted one-half of the gene pool. Yule argued this claim during the discussion that followed a talk given in 1908 by none other than Reginald Punnett. Punnett thought that Yule was wrong, but he did not know how to prove it.

We have already demonstrated, of course, that Punnett was correct in rejecting Yule's claim. Our calculations showed that a population with allele frequencies of 0.6 and 0.4 is in equilibrium too. What Punnett wanted, however, is a general proof. This proof should show that any allele frequencies, so long as they sum to 1, will remain unchanged from one generation to the next.

Punnett took the problem to his mathematician friend Godfrey H. Hardy, who produced the proof in short order (Hardy 1908). Hardy simply repeated the calculations that Yule had performed, using variables in place of the specific allele frequencies that Yule had assumed. Hardy's calculation of the general case indeed showed that any allele frequencies can be in equilibrium.

The General Case

For our version of Hardy's general case, we again work with our imaginary mouse population. We are concerned with a single locus with two alleles: A1 and A2. We use capital letters with subscripts because we want our calculation to cover cases in which the alleles are codominant as well as cases in which they are dominant and recessive. The three possible diploid genotypes are A1A1, A1A2, and A2A2.

As in our simulations and numerical example, we will start with the gene pool and follow the alleles through one complete turn of the life cycle. The gene pool will contain some frequency of A1 gametes and some frequency of A2 gametes. We will call the frequency of A1 in the gene pool p and the frequency of A2 in the gene pool q. There are only two alleles in the population, so

The first step is to let the gametes in the gene pool combine to make zygotes. Figure 6.9a shows the four possible combinations of egg and sperm, the zygotes they produce, and a calculation specifying the probability of each. For example, if we pick an egg to watch at random, the chance is p that it will have genotype A1. When a sperm comes along to fertilize the egg, the chance is p that the sperm will have genotype A1. The probability that we will witness the production of an A1A1 zygote is therefore

Figure 6.9 The general case for random mating in our model population

(a) We can predict the genotype frequencies among the zygotes by multiplying the allele frequencies. (b) A Punnett square. The variables along the left and top edges of the box represent the frequencies of A and a eggs and sperm in the gene pool. The expressions inside the box represent the genotype frequencies among zygotes formed by random encounters between gametes in the gene pool.

If we watched the formation of all the zygotes, p2 of them would have genotype A1A1. The calculations in Figure 6.9a show that random mating in our gene pool produces zygotes in the following proportions:

Figure 6.9b shows a Punnett square that yields the same genotype frequencies. The Punnett square also shows geometrically that

This confirms that we have accounted for all the zygotes. The same result can be demonstrated algebraically by substituting (1 − p) for q in the expression p2 + 2pq + q2, then simplifying.

We have gone from the allele frequencies in the gene pool to the genotype frequencies among the zygotes. We now let the zygotes develop into juveniles, let the juveniles grow up to become adults, and let the adults produce gametes to make the next generation's gene pool.

The challenge now is to prove algebraically that there was nothing special about our numerical examples. Any allele frequencies will remain constant from generation to generation.

We can calculate the frequency of allele A1 in the new gene pool as follows. Because adults of genotype A1A1 constitute a proportion p2 of the population, they will make p2 of the gametes. All of these gametes carry copies of allele A1. Likewise, adults of genotype A1A2 constitute a proportion 2pq of the population, and will make 2pq of the gametes. Half of these gametes carry copies of allele A1. So the total fraction of the gametes in the gene pool that carry copies of A1 is

We can simplify the expression on the right by substituting (1 – p) for q. This gives

Figure 6.10 shows this calculation graphically. The figure also shows a calculation establishing that the fraction of the gametes in the gene pool that carry copies of allele A2 is q. We assumed at the outset that p and q sum to 1, so we know that we have accounted for all the gametes.

Figure 6.10 A geometrical representation of the general case for the allele frequencies produced when the adults in our model population make gametes

The area of each box represents the frequency of an adult or gamete genotype.

Our model has shown that our idealized population does not evolve. This conclusion is known as the Hardy–Weinberg equilibrium principle.

Once again, we have come full circle and are back where we started. We started with allele frequencies of p and q in our population's gene pool. We followed the alleles through zygotes and adults and into the next generation's gene pool. The allele frequencies in the new gene pool are still p and q. The allele frequencies p and q can be stable at any values at all between 0 and 1, as long as they sum to 1. In other words, any allele frequencies can be in equilibrium, not just p = q = 0.5 as Yule thought.

This is a profound result. At the beginning of the chapter we defined evolution as change in allele frequencies in populations. The calculations we just performed show, given simple assumptions, that in populations following the rules of Mendelian genetics, allele frequencies do not change.

Computing Consequences 6.2 The Hardy–Weinberg equilibrium principle with more than two alleles

Imagine a single locus with several alleles. We can call the alleles Ai, Aj, Ak, and so on, and we can represent the frequencies of the alleles in the gene pool with the variables pi, pj, pk, and so on. The formation of a zygote with genotype AiAi requires the union of an Ai egg with an Ai sperm. Thus, the frequency of the homozygous genotype AiAi is p2i. The formation of a zygote with genotype AiAj requires either the union of an Ai egg with an Aj sperm, or an Aj egg with an Ai sperm. Thus, the frequency of the heterozygous genotype AiAj is 2pipj.

For example, if there are three alleles with frequencies p1, p2, and p3 such that

then the genotype frequencies are given by

and the allele frequencies do not change from generation to generation.

We have presented this result as the work of Hardy (1908). It was derived independently by Wilhelm Weinberg (1908) and has become known as the Hardy–Weinberg equilibrium principle. Some evolutionary biologists refer to it as the Hardy–Weinberg–Castle equilibrium principle, because William Castle (1903) worked a numerical example and stated the general equilibrium principle nonmathematically five years before Hardy and Weinberg explicitly proved the general case (see Provine 1971). The Hardy–Weinberg equilibrium principle yields two fundamental conclusions:

Conclusion 1: The allele frequencies in a population will not change, generation after generation.

Conclusion 2: If the allele frequencies in a population are given by p and q, the genotype frequencies will be given by p2, 2pq, and q2.

We get an analogous result if we generalize the analysis from the two-allele case to the usual case of a population containing many alleles at a locus (see Computing Consequences 6.2).

What Use Is the Hardy–Weinberg Equilibrium Principle?

It may seem puzzling that in a book about evolution we have devoted so much space to a proof apparently showing that evolution does not happen. Evolution does, of course, happen—we saw it happen in this chapter in our own simulations. What makes the Hardy–Weinberg equilibrium principle useful is that it rests on a specific set of simple assumptions. When one or more of these assumptions is violated, the Hardy–Weinberg conclusions no longer hold.

We left some of the assumptions unstated when we developed our null model of Mendelian alleles in populations. We can now make them explicit. The crucial assumptions are as follows:

There is no selection. All members of our model population survived at equal rates and contributed equal numbers of gametes to the gene pool. When this assumption is violated—when individuals with some genotypes survive and reproduce at higher rates than others—the frequencies of alleles may change from one generation to the next.

There is no mutation. In the model population, no copies of existing alleles were converted by mutation into copies of other existing alleles, and no new alleles were created. When this assumption is violated, and, for example, some alleles have higher mutation rates than others, allele frequencies may change from one generation to the next.

The Hardy–Weinberg equilibrium principle becomes useful when we list the assumptions we made about our idealized population. By providing a set of explicit conditions under which evolution does not happen, the Hardy–Weinberg analysis identifies the mechanisms that can cause evolution in real populations.

There is no migration. No individuals moved into or out of the model population. When this assumption is violated, and individuals carrying some alleles move into or out of the population at higher rates than individuals carrying other alleles, allele frequencies may change from one generation to the next.

There are no chance events that cause individuals with some genotypes to pass more of their alleles to the next generation than others. That is, blind luck plays no role. We saw the influence of blind luck in our simulations. We avoided it in our analysis of the general case by assuming that the eggs and sperm in the gene pool collided with each other at their actual frequencies of p and q and that no deviations were caused by chance. Another way to state this assumption is that the model population was infinitely large. When this assumption is violated, and by chance some individuals contribute more alleles to the next generation than others, allele frequencies may change from one generation to the next. This mechanism of allele frequency change is called, as we said earlier, genetic drift.

Individuals choose their mates at random. We explicitly set up the gene pool to let gametes find each other at random. In contrast to assumptions 1 through 4, when this assumption is violated—when, for example, individuals prefer to mate with other individuals of the same genotype—allele frequencies do not change from one generation to the next. Genotype frequencies may change, however. Such shifts in genotype frequency, in combination with a violation of one of the other four assumptions, can influence the evolution of populations.

By furnishing a list of ideal conditions under which populations will not evolve, the Hardy–Weinberg equilibrium principle identifies the set of events that can cause evolution in the real world (Figure 6.11). This is how the Hardy–Weinberg principle serves as a null model. Biologists can measure allele and genotype frequencies in nature, and determine whether the Hardy–Weinberg conclusions hold. A population in which they hold is said to be in Hardy–Weinberg equilibrium. If a population is not in Hardy–Weinberg equilibrium—if the allele frequencies change from generation to generation or if the genotype frequencies cannot, in fact, be predicted by multiplying the allele frequencies—then one or more of the Hardy–Weinberg model's assumptions are being violated. Such a discovery does not, by itself, tell us which assumptions are being violated, but it indicates that further research may be rewarded with interesting discoveries.

Figure 6.11 Summary of the mechanisms of evolution

Four processes can cause allele frequencies to change from one generation to the next. Selection occurs when individuals with different genotypes survive or make gametes at different rates. Migration occurs when individuals move into or out of the population. Mutation occurs when mistakes during meiosis turn copies of one allele into copies of another. Genetic drift occurs when blind chance allows gametes with some genotypes to participate in more fertilizations than gametes with other genotypes.

In the remaining sections of Chapter 6, we consider how violations of assumptions 1 and 2 affect the two Hardy–Weinberg conclusions, and we explore empirical research on selection and mutation as mechanisms of evolution. (In Chapter 7, we consider violations of assumptions 3, 4, and 5.)

Changes in the Frequency of the CCR5-Δ32 Allele

We began this section by asking whether we can expect the frequency of the CCR5-Δ32 allele to change in human populations. Now that we have developed a null model for how Mendelian alleles behave in populations, we can give a partial answer. As long as individuals of all CCR5 genotypes survive and reproduce at equal rates, as long as no mutations convert some CCR5 alleles into others, as long as no one moves from one population to another, as long as populations are infinitely large, and as long as people choose their mates at random, then no, the frequency of the CCR5-Δ32 allele will not change.

This answer is, of course, thoroughly unsatisfying. It is unsatisfying because none of the assumptions will be true in any real population. We asked the question in the first place precisely because we expect Δ32/Δ32 individuals to survive the AIDS epidemic at higher rates than individuals with either of the other two genotypes. In the next two sections, we will see that our null model, the Hardy–Weinberg equilibrium principle, provides a framework that allows us to assess with precision the importance of differences in survival