Intraspecific Variation

Natural selection and intraspecific phenotypic variation. Natural selection is a process that follows from the necessary conditions of (a) genetically based pheno-typic variation and (b) differences in reproduction or survival among the variants. With these two necessary conditions, it follows that (1) if the population is out of equilibrium, then the phenotypic distribution of the offspring in a population will differ from that of the parents, or (2) the phenotypic distribution will differ among age classes, beyond that expected from normal ontogenetic change (see Burian 1983; Endler 1986). The logical consequences - 1 and 2 - following conditions a and b constitute a syllogism, not a tautology (Endler 1986). The misconception of a tautology stems from the phrase "survival of the fittest," which ignores the logical structure of conditions and consequences. Natural selection can also be applied to

Figure 3.2. The transition between polymorphism (with clinal variation from A to B), polytypism, and speciation.

Figure 3.2. The transition between polymorphism (with clinal variation from A to B), polytypism, and speciation.

higher levels of the organismal hierarchy (see chapter 1). At the species level, for example, species with their traits constitute the variation, and the analogous measures of reproduction and mortality would be speciation and extinction. The consequence would be changes in the abundance of species. Each hierarchical level must have its own distinct heritable variation and a measure of productivity that could differ among the units. As long as the levels are decomposable in the hierarchical sense, the process of natural selection and even adaptation could be applied to levels above those of the individual.

If the change in phenotypic distribution is consonant with predictions based on a criterion of performance (e.g., owing to natural selection, tolerance to low salinity increases when salinity greatly decreases), then adaptation occurs. Adaptation is the historical process of evolutionary change describing how natural selection interacts with functional and developmental constraints, mutational availability, and random processes. Natural selection is the specific part of adaptation concerned with available variation and the selective environment. By contrast, adaptedness is a static description of the functional superiority of one or another phenotype in a specific environment.

Models of natural selection and adaptation often make the assumption that there are two necessary elements to predict the course of evolution: (1) a predictable description of the relationship between genotype and phenotype and (2) a description of the environment that discriminates among phenotypes. There are good reasons to believe that this is too simplistic. First, genotype-environment interactions are such that one cannot catalogue phenotypes just by knowing the genotype. The environment must often be specified (e.g., Gupta and Lewontin 1983). Our definition of natural selection does not require a separation of adaptedness from historical circumstances, as we require a specification of genotype-environment interaction. To the degree that genotype-environment interactions and environments vary, the outcome of evolution is increasingly complicated.

The course of adaptation may be eccentric, depending on the genetic track taken during evolutionary change. It is well known, for example, that forward and backward selective change of the same morphological trait does not occur at the same rate, despite the imposition of similar selection differentials. Different genes may take a phenotype in the same overall direction, but evolution must work with the complex variability at hand. Differences in environmental history may also result in different adaptive tracks. The problem of history can be seen in a study of mutant strains of Escherichia coli that produce the toxin colicin (Chao and Levin 1981). In a well-mixed 'aquatic culture, the colicinogenic strains are at an advantage only when fairly common. When sensitive bacteria are killed, their death causes release of nutrients at random to both types of bacteria. Colicinogenic strains have lower division rates and therefore lose out unless they overwhelm the system with toxin. By contrast, in agar, colicinogenic strains come to dominate even when initially rare. The latter structured habitat permits the colicinogenic bacteria to create a barren zone, which is rich in nutrients. They then can spread at their "leisure." The fate of the gene is therefore locked up in its historical background.

We cannot overstress the distinction of natural selective forces from the genetic variation present at the time of an evolutionary change. Some have argued that evolution is contingent, mentioning the co-opting of peculiar structures to perform functions (e.g., the panda's thumb used for a sort of grasping). But this argument obscures the presence of a similar selective force that operates on many species with disparate morphologies and genes. A study by Huey, Gilchrist, Carlson, Berrigan, and Serra (2000) proved this point well. Drosophila subob-scura was introduced to North America from Europe, where one could find a stable cline of increasing wing length with increasing latitude. After a mere two decades, a strikingly similar cline had developed in North America, which attests to the predictable power of natural selection. But the exact response of North American populations differed in that different parts of the wing contributed to the size changes, relative to the European populations. Apparently, selection was for overall wing length, but this was achieved with different sources of variability in Europe and America. It would be interesting if the traits that change in the two respective regions correspond to increased heritability of the different specific traits that responded to selection.

Through habitat choice, behavior decisions, and so on, organisms can alter their own environment. This means that to study natural selection properly, we must be able to describe the interaction between organism and environment that determines the actual selective regime (Lewontin 1983b). This often means that history could thwart the prediction of clear evolutionary trajectories.

Fitness is often used interchangeably with adaptedness. Fitness should refer to the relative ability of genotypes to survive and leave offspring. One can also define fitness in terms of alleles at a locus. If we have a locus segregating for two alleles, A1 and A2, let the respective fitnesses be W1 and W2. We can then define a selection coefficient, s, which equals W1/W2 - 1. If p is the frequency of A1 and q is the frequency of A2, then the change in p over one generation will be

Ap = sq w where w is the mean fitness of the entire population. Changes in allele frequencies are therefore associated with a fitness parameter. This expression is oversimplified and applies to haploid organisms. For a more complete discussion of selection in diploid organisms, see Ewens (1969).

The assessment of relative fitness of genotypes at a locus implies a complete randomization of the background genotype (Lewontin 1974). In practice, this is nearly impossible to achieve, given the great difficulty of randomizing the background loci that are tightly linked to the locus in question. This problem is not trivial and is a major source of difficulty in interpreting the meaning of selection experiments. In a crude experiment, selection for variants at an allozyme locus may appear to be intense, simply because the locus marks part of a - or an entire - chromosome (contrast Powell 1971 with Yamazaki et al. 1983). A detailed study of fruit flies, using flanking markers at close map distance, designed to randomize the genetic background, typically requires the counting of tens of thousands of flies (Eanes 1984; Eanes, Bingham, Hey, and Doule 1985). Such studies, rarely done, show that fitness among protein phenotypes is probably much less than usually estimated, when linked loci are factored out (Eanes 1987).

It is difficult to measure with statistical confidence selection among genotypes differing in fitness by as much as 1%. Natural selection involving such levels and less, however, can exert significant evolutionary effects. To demonstrate that recessive lethal chromosomal mutants of Drosophila melanogaster lowered the fitness of heterozygotes by about 1%, Mukai and Yamaguchi (1974) had to score about one million flies. Such Herculean projects have been completed only rarely. One should remember that biologically significant selection need only be s > V2N, and N is often 106.

We shall be mainly concerned with morphological characters whose determination has both a genetic and an environmental contribution. Let us assume further that variation in the phenotype can be arrayed as variation along a single axis of variation. This could apply to both continuously measurable and countable characters. Assume that the population variation follows a normal distribution, with mean = 0 and standard deviation = o. Assume also, following the methods of quantitative genetics (see Falconer 1981), that the phenotypic scale is determined partially by an underlying genotypic scale. We can plot along the genotypic scale a fitness function, which depicts the relative success of different genotypes.

Total phenotypic variance can be partitioned into a variety of genetic and environmental components. For our purposes, assume that all of the loci determining a phenotypic trait are independent, there is no dominance, and that there are no genotype-environment interactions. We can then simply define narrow-sense heritability, h2, as the proportion of the total phenotypic variance, o2a, explainable by between-allele effects, or the additive genetic variance, Ga. Define a selection differential, Sa, which represents the difference in the means of the selected and unselected adults. If our phenotypic scale variable is Za, the change will be

This formulation provides a convenient way of visualizing how selection on a continuous phenotypic character can be related to an underlying genotypic distribution.

The greatest problem with morphological or other complex traits is their probable association with a large number of loci located over widely spread parts of the genome. Quantitative Trait Locus (QTL) Mapping has been developed to establish linkage maps between morphological traits and molecular markers (see Lander and Schork 1994; Lynch and Walsh 1997). It is an extension of traditional linkage mapping of traits, only expanded to larger scales owing to the use of molecular markers, which are practicable when they are highly polymorphic, such as RFLP loci. If a QTL can be mapped to a relatively small chromosomal region, molecular methods might be used to identify specific genes involved in affecting the trait. This approach promises to shed tremendous light on the structure of many morphological structures and their genetic architecture. For example, Zeng, Liu, Stam, Kao, Mercer, and Laurie (2000) found 19 different QTLs underlying trait variation of the posterior lobe of the male genital arch between two species of Drosophila. The differences suggested strong directional selection acting on the trait in each species.

Modes of natural selection. In directional selection (Figure 3.3), the maximum value of the fitness function is shifted away from the mean (we use the right side as a convention). The success of a given phenotype might increase continuously with higher phenotypic value. But directional selection can involve truncation selection of the entire phenotypic distribution past an absolute or relative (e.g., upper 10% of the population) threshold. This might occur when allometric considerations prohibit animals larger than a certain body size to satisfy their maintenance energy requirements, or when animals larger than a threshold size might escape the grasp of a predator.

Complete and careful observations of directional selection in natural populations, where the adaptive significance is clear, are few in number (see Boag and Grant 1981; Endler 1986; Ford 1975; Seeley 1986). This should be no surprise, as selection intensities have to be quite high to show any dramatic change, and such selection will be temporary before a new equilibrium is achieved. A selection intensity among discrete morphs of only a few percent would be effectively invisible, given the swamping effect of collecting difficulties, statistical problems, spatially varying directional selection, and possible differing genotype-environment interactions in different subhabitats. Nevertheless, the outcome of natural selection has been appropriately inferred in a surprisingly large number of cases (see Endler 1986). Consider the following examples.

Over 100 species of insects in British industrial regions blackened by smoke are dark in color, relative to conspecifics in unpolluted areas (Ford 1975, chapter 14). In industrial areas, vegetation is often darkened with smokestack soot, and light-colored substrates, such as lichens, are killed off by pollution. In the moth Biston betu-laria, the black carbonaria variant is dominant over the recessive light-colored morph. Dark-colored morphs are preferentially killed by a variety of insectivorous birds when placed on trees with a normal lichen cover. The light mottled color blends with the background. Birds quickly locate the light morphs against the contrasting background of the blackened vegetation (Kettlewell 1955). Kettlewell's data suggest that the selection coefficient favoring the melanic gene must be about 0.5. This would easily account for the rapid spread of the carbonaria morph since the middle of the nineteenth century, from negligible starting frequencies.

Stabilizing Directional Disruptive

Stabilizing Directional Disruptive

Figure 3.3. Examples of some modes of natural selection on a hypothetical normal distribution of phenotypes. Broken line is the frequency distribution of phenotypes; solid line represents fitness function.

Was this article helpful?

0 0

Post a comment