Genetic Architecture

Shifting balance also requires multiple selective equilibria, that is, multiple adaptive peaks that are separated by either fitness valleys or fitness ridges in Wright's metaphor. The shape of the adaptive surface depends in part upon the underlying genetic architecture of fitness. Recall from Chapter 10 that the genetic architecture refers to the number of loci and their linkage relationships and the numbers of alleles per locus that contribute to a trait, along with the mapping of genotype onto phenotype (dominance, recessiveness, pleiotropy, epistasis, etc.). The multiple-peak adaptive surface required for shifting balance arises in part from a genetic architecture characterized by strong interactions between genes (either between alleles at the same locus or between alleles at different loci, that is, epistasis) and/or pleiotropy. For example, the two-peak surface shown in Figure 11.8 for the j-Hb A, S, and Cexample in a malarial environment arises in part from the pleiotropic effects associated with these alleles on the traits of malarial resistance and hemolytic anemia and in part from the interactions between alleles (S is dominant to A for the trait of malarial resistance and recessive for the trait of anemia, C is recessive to A for the trait of malarial resistance and codominant with S for the trait of anemia). One implication of a multiple-peak adaptive surface is that the fitness effects of an allele are highly context dependent. For example, in Figure 11.8 the S allele is a beneficial allele in the context of a randomly mating population close to fixation for the A allele but is a deleterious allele in the context of a randomly mating population close to fixation for the C allele.

Fisher rejected Wright's view of genetic architecture and its resulting rugged adaptive landscapes and counteracted it with his own visual metaphor: the adaptive hypersphere. In Fisher's metaphor, the degree of adaptation of a population is represented by its closeness to a fixed point in a multidimensional space that corresponds to a single optimal adaptive type, in contrast to the multipeak adaptive landscapes of Wright. A second point in Fisher's hyperspace represents the population's current average adaptive phenotype. In contrast to the rugged adaptive surfaces of Wright, Fisher envisioned a smooth, continuous relationship between the optimal point and other points in this adaptive space such that the degree of adaptation of any population was a simple decreasing function of its distance from the optimal point. Hence, all the points closer to the optimal type than the current population are more fit than the current population. The points corresponding to better adapted populations relative to the current population are therefore found in a hypersphere whose center is at the optimum point and whose radius is the distance between the optimum and the current

Figure 12.8. Fisher's adaptive target. The point in the center of the circle corresponds to a population that is adapted to the environment in the optimal fashion. The point on the circle corresponds to the level of adaptation of the current population, with increasing distance from the optimal point corresponding to decreased levels of adaptation. The circle encloses the area within which a population would be better adapted than the current population. Fixation for a mutation is indicated by an arrow starting at the current population and ending at a new point in space that is random in direction and magnitude from the current population. On the right is an expanded section of the circle in the vicinity of the current population to illustrate the increased chances of mutations with small phenotypic effect being selectively advantageous.

Figure 12.8. Fisher's adaptive target. The point in the center of the circle corresponds to a population that is adapted to the environment in the optimal fashion. The point on the circle corresponds to the level of adaptation of the current population, with increasing distance from the optimal point corresponding to decreased levels of adaptation. The circle encloses the area within which a population would be better adapted than the current population. Fixation for a mutation is indicated by an arrow starting at the current population and ending at a new point in space that is random in direction and magnitude from the current population. On the right is an expanded section of the circle in the vicinity of the current population to illustrate the increased chances of mutations with small phenotypic effect being selectively advantageous.

population. When dealing with two dimensions, this hypersphere becomes a simple circle or "target" (Figure 12.8). Fisher felt that natural selection would ensure that the population was near the optimal point. The only reason why populations would not be exactly at the optimal point is because the appropriate mutations had not yet occurred. Thus, the random mechanism responsible for exploring Fisher's adaptive space is mutation, with genetic drift playing no role whatsoever.

Fisher represented mutation as a vector coming from the current population that is random in both direction and magnitude. Fixation for this mutation would move the population to the point in the adaptive space indicated by the end of the mutational arrow. Fisher then calculated the probability of such a random vector having its end point land within the adaptive target, that is, closer to the optimum than the starting point. Fisher showed that this probability is zero whenever the magnitude of the vector exceeds the diameter of the hypersphere (see Figure 12.8). As the size of the fitness effect associated with a mutation declines, the probability of a mutation of random direction resulting in a favorable change increases and ultimately reaches a limit of 1 for mutations of very small effect (Figure 12.8). Fisher therefore concluded that mutations of very small effect would provide the primary raw material for adaptive change since they are more likely to be advantageous. Fisher assigned each mutation a single vector to represent its fitness effects. The phenotypic effect is determined at the moment of mutation and remains constant throughout evolutionary history. Crow (1957) provided a rationale for Fisher's assumption by arguing that the genetic background is constantly changing in large, random-mating populations such that those mutations that have a consistent advantageous phenotypic effect regardless of genetic background will be the ones most likely utilized by natural selection to build adaptive traits. Therefore, in the Fisherian model, adaptation occurs by the accumulation of many small mutational steps toward an optimum, with the underlying genetic architecture of adaptive traits being due to a large number of loci with each locus having functional alleles with small, additive phenotypic effects.

Wright agreed with Fisher that adaptive traits generally have a multilocus genetic architecture. Unlike Fisher, Wright did not model the phenotypic effects as intrinsic, constant attributes of a specific mutation, but rather Wright regarded the phenotypic effects of a mutation as highly context dependent because of interallelic interactions, epistasis, and pleiotropy. Indeed, such interactions ensure that only rarely could we regard an allele as being intrinsically of major or minor effect (the critical distinction in Fisher's metaphor). As with the ApoE/LDLR example in Chapter 10, alleles of major or minor effect arise out of contextual interactions and can change dramatically as the genetic background is altered. In contrast to the fixed-length vectors in Fisher's metaphor (Figure 12.8), the magnitude of the phenotypic effects associated with a specific mutation or allele can be dramatically altered in an interaction system (Figures 10.15-10.17). Such interactions, particularly epistasis, make the shifting balance process far more probable and adaptively important (Bergman et al. 1995; Brodie 2000; Goodnight 1995, 2000). Consequently, one critical difference between Fisher and Wright is their opposing views of genetic architecture.

Until recently, we had little insight into the genetic architecture of most quantitative traits. Recall from Chapter 9 that the classical, unmeasured genotype analysis of quantitative genetic traits provides little information about genetic architecture and the role of epistasis and other interaction effects. Indeed, the Fisherian unmeasured genotype analysis is biased against the detection of epistasis. The Fisherian quantitative genetic measures were designed to first fit the "additive" component, then fit the "dominance" component, and finally, attribute what is left over to the "epistasis" component. This creates the artifact that epistasis seems weak compared to additive effects (Cheverud and Routman 1995; Goodnight 1995). For example, recall from Chapter 10 that the extensive epistasis between the ApoE andLDLR loci for the phenotype of serum cholesterol level is translated into "additive variance" at the ApoE locus by the Fisherian quantitative genetic measures (Figure 10.15A).

The dominance of Fisher's quantitative genetic paradigm that places little importance upon epistasis and other interaction effects has more to do with mathematical and statistical convenience than with biological reality. All that we know of biological systems—from the control of gene expression, to biochemical pathways, to developmental processes, to physiological regulation, and so on—indicates that interactions are the norm. Indeed, the idea that a single locus can have a marginal effect that is invariant to the remainder of its biological context is implausible in the extreme, yet Fisher's metaphor is based upon this unlikely premise. Measured genotype approaches are now allowing us to detect epistasis and other interactive effects (Chapter 10), and these interactions are found to be the norm in genetic architecture, not the exception (Templeton 2000). It is therefore not surprising that Frankel and Schork (1996) concluded that "where complex genetic traits loom, epistasis is not far behind."

Indeed, measured genotype approaches are finding extensive epistasis in the genetic architecture of "simple" Mendelian traits that had traditionally been regarded as examples of single-locus genetic architectures. By definition, a single-locus genetic trait is free of epistasis. But are such simple traits truly single-locus traits, and if not, is there any role for epistasis? Consider sickle cell anemia—the workhorse example of a simple Mendelian trait found in most genetic textbooks. Up to now, sickle cell anemia has been treated as a single-locus trait in this book. Indeed, sickle cell anemia is commonly presented as a single-nucleotide trait due to one A-to-T nucleotide change in the second position of the sixth codon of the j chain of hemoglobin (the S allele) that is said to "cause" sickle cell anemia when homozygous. However, when individuals who are homozygous for the S allele are examined, tremendous heterogeneity in clinical severity is revealed (Odenheimer et al. 1983; Sing et al. 1985). Epistasis is a major determinant of this heterogeneity. For example, the S allele is found in high frequency in certain Greek populations, but clinical manifestations are mild. Studies (e.g., Berry et al. 1992) have shown that persistence of fetal hemoglobin into the adult can ameliorate the severity of sickle cell anemia. Fetal hemoglobin is coded for by the tandemly duplicated y loci (called A and G), which are closely linked to the j locus (Figure 2.6). Normally the ygenes are turned off after birth and the j gene is activated, leading to the transition from fetal to adult hemoglobin. Several mutations in or near the y loci can cause persistence of fetal hemoglobin and are found in Greek populations (Berry et al. 1992; Patrinos et al. 1996).

Sickle cell is also common in certain populations in Saudi Arabia that live in historic malarial regions (el-Hazmi and Warsy 1996). An XmnI polymorphic site 5' to the Gy locus (Figure 2.6) and a Hindlll polymorphic site in the Gy locus is associated with persistence of fetal hemoglobin and is in disequilibrium with the sickle cell allele in some Arabian populations. This same haplotype is also found in populations from India, who likewise have mild clinical symptoms with homozygous SS and have persistence of fetal hemoglobin (Ramana et al. 2000). Note that having persistence of fetal hemoglobin at the Gy locus increases the fitness of homozygotes for the S allele at the j-Hb locus because that genetic combination results in sickle cell homozygous individuals that display mild clinical symptoms (el-Hazmi et al. 1992; Ramana et al. 2000). However, persistence of fetal hemoglobin is expected to decrease the ability of the blood to deliver oxygen to the peripheral tissues because fetal hemoglobin has a higher binding affinity for oxygen [this allows the developing fetus to take oxygen from the mother's blood across the placenta (Giblett 1969)]. The disequilibrium found in both Greek and Arab populations therefore favors the high-fitness combination at the expense of the low-fitness combination. This nonrandom association between alleles at different loci is therefore presumably due to natural selection operating upon a multilocus, epistatic genetic architecture. Hence, the sickle cell phenotype is not really a single-locus phenotype but rather in some human populations is a coadapted gene complex, in which the frequencies of alleles at different loci are mutually adjusted with respect to one another by natural selection favoring epistatic combinations with high fitness. In the particular case of the Gy and j -Hb loci in Greek and Arab populations, the physical closeness of these two loci in the genome (see Figure 2.6) means that the disequilibrium built up by natural selection is not readily dissipated by recombination. The resulting genetic stability of this combination of closely linked alleles means that the combination itself approximates an "allele" in its inheritance pattern. Closely linked loci with coadapted combinations of alleles in linkage disequilibrium are called supergenes (Kelly 2000). Accordingly, the DNA region around the j-Hb locus (Figure 2.6) is really a supergene in the selective context of a malarial environment. The behavior of this region as a supergene also means that the adaptive topographies shown in Figures 11.8, 11.9, 12.1, and 12.2 are actually only three-dimensional projections of a higher dimensional adaptive landscape. For example, the frequency of the S allele as shown in these previous adaptive landscapes is actually the frequency of the superallele S that is not linked to the Gy allele that causes persistence of fetal hemoglobin. To accommodate the Greek and Arab populations into the adaptive surface, the adaptive surface would now need a fourth dimension that measures the frequency of the superallele that has S and the Gy allele that causes persistence of fetal hemoglobin on the same chromosome. In other words, at the Gy/j-Hb supergene level there are four alleles: A, C, S-no persistence of fetal hemoglobin, and S-persistence of fetal hemoglobin. By ignoring the epistasis between the Gy and j-Hb loci and focusing only on the j-Hb locus (the norm in most textbooks), we would mistakenly conclude that Greek, Arab, and most Bantu populations living in malarial environments had adapted to malaria by being on the same A/S polymorphic adaptive peak. However, when we consider the four-dimensional adaptive surface associated with the Gy/ j-Hb supergene (which unfortunately cannot be adequately depicted in a two-dimensional figure), we find that many of the Greek and Arab populations that are polymorphic for sickle cell anemia are actually in a different portion of the adaptive surface than the Bantu populations that are polymorphic for sickle cell anemia. Thus, just considering this one small DNA region alone, different human populations have adapted to malaria by evolving toward three different parts of the fitness surface (increasing the frequency of C as in Upper Volta, the A/S-no persistence of fetal hemoglobin polymorphism as in many Bantu populations, and the A/S-persistence of fetal hemoglobin polymorphism as in some Greek and Arab populations). Obviously, there is no single adaptive target for selection to hit, and different human populations have evolved different adaptive strategies in response to a malarial environment.

Even our four-dimensional fitness surface with its added adaptive option is an oversimplification. As pointed out in Chapter 11, several other loci are involved in malarial adaptation, and these and other loci show epistasis with one another and with sickle cell. For example, there is also evidence for an X-linked locus as a contributor to the persistence of fetal hemoglobin in Arabian populations (el-Hazmi et al. 1994a). The protein haptoglobin forms complexes with hemoglobin and is thereby the major determinant of hemoglobin excretion (Giblett 1969). Given this direct physiological interaction, it is not surprising that genetic variants of haptoglobin also display epistasis with the S allele (Giblett 1969). Epistatic interactions for clinical severity of anemia have also been reported between the S allele and «-thalassemia (caused by genetic variation at the a-Hb locus) and G6PD deficiency (Chapter 11) (el-Hazmi et al. 1994b). G6PD deficiency in turn displays epistasis with tha-lassemia (caused by mutations at the a- and j -globin loci) (Siniscalco et al. 1966).

G6PD deficiency achieves malarial resistance by reducing the ability of the malarial parasite to use the oxidative shunt pathway in the parasitized red blood cell (Giblett 1969)— a different molecular mechanism than that associated with the S allele. Accordingly, G6PD deficiency has a different set of clinical effects than the S allele, even though epistasis partially interweaves the two systems. Some bearers of G6PD deficiency are extremely sensitive to environmental oxidizing agents such as fava beans (Chapter 11). Sensitivity to fava beans (favism) can cause death through hemolytic crisis. However, not all carriers of G6PD deficiency are susceptible to favism, and some of this heterogeneity is due to epistasis between the X-linked G6PD locus and the two autosomal globin loci associated with thalassemia (Siniscalco et al. 1966). There is also epistasis for favism between G6PD deficiency and the locus for red cell acid phosphatase (ACP-1) (Bottini et al. 1971; Palmarino etal. 1975; Bottini etal. 1995), yet another locus that has a direct effect on malarial resistance (Bottini et al. 2001a). As a consequence, the joint allele frequencies at these interacting loci are a complex function of the presence of malaria and the ingestion of fava beans (Palmarino et al. 1975).

The ACP-1 locus not only interacts with G6PD for favism and is involved with malarial adaptation but also influences fetal growth, birth weight in human males, and body mass (Amante et al. 1990; Greene et al. 2000). The effects of acid phosphatase on intrauterine growth and neonatal survival themselves are subject to epistatic modification with the adenosine deaminase locus (ADA) (Gloria-Bottini et al. 1989b) and with the maternal-fetal incompatibility reactions modulated by the ABO blood group locus (Lucarini et al. 1995). These epistatic interactions in turn interact with the diabetic status of the individual (Gloria-Bottini et al. 1989a; Bottini et al. 1991). Diabetes is a complex genetic system influenced by many loci with epistasis (Gloria-Bottini et al. 1989a; Bennett and Todd 1996) and also shows interactions with other systemic diseases, such as rheumatoid arthritis [another epistatic trait (Cornelis et al. 1998)] and coronary artery disease (yet another epistatic trait, as shown in Chapter 10). This linking of systems through epistasis could continue, but it should be clear by now that simple Mendelian systems are in reality merely low-resolution projections of complex systems involved in an expanding web of pleiotropic effects and epistatic interactions. This web of epistasis was revealed in the case of sickle cell anemia because it has such strong clinical relevance.

Frankel and Schork (1996) point out that when the tools to investigate epistasis are available and used, epistasis and pleiotropy are almost always found, even for simple Mendelian traits. The Fisherian view of many genes of intrinsic, small, additive effect appears to have little relevancy for any genetic system that is examined in detail. Consequently, when it comes to genetic architecture, Wright was right and Fisher was off target. Computer simulations by Bergman et al. (1995) indicate that the evolutionary advantage of the shifting balance process is an increasing function of the amount of epistasis. Hence, the extensive amount of epistasis that is being found in measured genotype approaches indicates that the conditions for shifting balance are broader than previously thought. Genetic architecture therefore does not appear to be a factor that would limit the operation of Wright's shifting balance process between selection, genetic drift, and gene flow.

Was this article helpful?

0 0

Post a comment