Populations Population will differentiate differentiation following primarily due genetic drift to selection

Figure 4.13 Genetic differentiation of populations depends on a combination of gene flow, effective population size and natural selection natural selection may mean that populations in one part of a species' range will not fare well if transplanted elsewhere, and this is an important consideration in reintroduction programmes in conservation biology. In practice, however, it is often extremely difficult to obtain estimates of s and Ne in wild populations, meaning that a comparison of these two variables is seldom a practical way of assessing the relative importance of selection. One alternative method for identifying local adaptation is through reciprocal transplant experiments, in which individuals are sampled from one or more populations and then transplanted to other sites of interest. Individuals then are monitored as they develop at both their home site and the sites to which they were relocated. At each population the relative fitness values of native and introduced individuals are compared, and a higher fitness in native plants provides evidence for local adaptation. Reciprocal transplant experiments have been used to great effect in many studies of plants and fungi, although they tend to be impractical when studying more mobile species. Molecular ecologists therefore often rely on other methods for inferring patterns of natural selection and local adaptation, some of which we will outline below.

Patterns of molecular evolution

We know from Chapter 3 that one way to infer selection is to compare sequences from different populations and calculate the proportion of synonymous and non-synonymous substitutions. Synonymous substitutions do not alter the encoded amino acid and are therefore usually neutral, whereas non-synonymous substitutions are much more likely to cause phenotypic change. Patterns of substitution can therefore provide a way of identifying natural selection, because non-synonymous substitutions should be proportionately higher in selected genes.

Table 4.10 Examples of genes in which selection has been inferred from a non-synonymous/ synonymous substitution ratio that is greater than one. Adapted from Ford (2000) and references therein






Resistance to bacterial infection

Humans, mice, fish

Major histocompatibility


complex (MHC)

Flowering plants

The S-locus system



Filamentous fungi

The het-c locus

Regulation of self/non-self-


recognition during


vegetative growth

Marine gastropods

Lysin genes

Proteins used by sperm to create

a hole in the egg vitelline


Sea urchins

Binding genes

Proteins that attach sperm to

eggs during fertilization



Male ejaculate protein

House mouse


Androgen-binding protein


Surface protein genes

Detection of parasites by hosts

Table 4.10 lists some examples in which the ratio of non-synonymous to synonymous substitutions has been used to infer natural selection.

One advantage to inferring selection from substitution ratios is that this method allows us to compare two categories of nucleotide substitutions along a single stretch of DNA, as opposed to comparing substitution rates from two or more gene regions that may have had different evolutionary histories. However, there are also some weaknesses associated with this method. For one thing it is not infallible; patterns of substitution will not reveal all instances of selection, in part because, as we saw in Chapter 1, sometimes only a single nucleotide change can have significant phenotypic consequences. Furthermore, we can use this method only if we know which gene has been selected for, and identifying candidate genes can be difficult. In the next section we will look at how a more general approach based on the comparison of data from multiple markers may give us clues about local adaptation.

Discordant genetic differentiation

Migration and drift are expected to have approximately equal effects on all neutral loci, whereas the effects of selection will vary between neutral and nonneutral loci. We therefore expect all neutral loci to show similar levels of genetic divergence among populations, whereas non-neutral loci (or loci linked to non-neutral loci) are expected to show anomalous levels of divergence. These anomalous levels may be unusually high or unusually low, depending on the type of selection that the relevant genes have been subjected to; directional selection will increase population differentiation if different alleles are selected for in different populations, whereas balancing selection can decrease population differentiation by maintaining the same suite of alleles in multiple populations. By comparing multiple measures of population differentiation that are each based on a different locus, researchers may discover a marker that shows unusual levels of differentiation, and that may therefore indicate a genetic region that is under selection.

There are numerous examples of this in the literature. The dusky grouper (Epinephelus marginatus) inhabits coastal reefs around the Atlantic Ocean and the Mediterranean Sea. Researchers interested in the amount of population differentiation among different sites around the Mediterranean used data from nine allozyme loci to compare three populations. The average FST was 0.214 but this result was influenced strongly by a single locus, ADA (adenosine deaminase), which yielded an FST value of 0.713. This was seven times greater than the next highest FST value, and once ADA was removed from the data set the average FST value fell to 0.060. The high FST value of ADA may be explained by directional selection at that locus (De Innocentiis et al., 2001). In another example of discordant genetic markers in fish, four spawning populations of sockeye salmon (Oncorhynchus nerka) in Alaska were compared using data from allozymes, microsatellites and RAPDs (Allendorf and Seeb, 2000). The FST values were comparable for all allozyme, microsatellite and RAPD loci with the exception of allozyme sAH (aconitate hydratase), which provided an extremely high measure ofFST (Figure 4.14). Once again, this discordant FST value may be explained by natural selection.

Discordant genetic markers are often used to identify seemingly non-neutral genes, although they are seldom interpreted as conclusive evidence for natural

Figure 4.14 The FST values calculated from thirteen allozyme loci, eight microsatellite loci and five RAPD loci among sockeye salmon populations. The outlier is the FST value for the sAH allozyme locus, marked by the arrow, which may be subject to selection (see text). Adapted from Allendorf and Seeb (2000)

Figure 4.14 The FST values calculated from thirteen allozyme loci, eight microsatellite loci and five RAPD loci among sockeye salmon populations. The outlier is the FST value for the sAH allozyme locus, marked by the arrow, which may be subject to selection (see text). Adapted from Allendorf and Seeb (2000)

selection because variation in FST among different loci can result from stochastic variation in genetic drift. This randomly generated variance of FST among loci may be particularly high if population sizes have fluctuated over time, since the loss of rare alleles during recent bottlenecks may have a greater impact on some loci than others. Care also must be taken when comparing FST measures that are based on different types of genetic markers, because markers often mutate at different rates. The effects of mutation on FST should be negligible if populations are in drift-migration equilibrium, but rapidly evolving markers may yield relatively high FST values if new mutations are not dispersed between populations rapidly enough to attain equilibrium between gene flow and genetic drift. When all FST values are higher in one set of markers (e.g. all microsatellites) compared with another set of markers (e.g. all allozymes), discrepancies are more likely to be attributable to different mutation rates than to natural selection.

Clinal variations in allele frequencies

Another way to look for evidence of natural selection is to examine changes in allele frequencies along geographical clines. These are caused by a gradual change in one or more environmental variables, for example a change in photoperiod along a latitudinal gradient. If particular alleles are associated with environmental variables then this may be indicative of selection. Some of the best examples of this have been found in the fruitfly Drosophila melanogaster. One study compared alleles of the 70-kd heat shock protein (hsp70) along a latitudinal gradient in Australia (Bettencourt et al., 2002). Heat shock proteins provide a mechanism that enables organisms to survive extreme temperatures and other environmental stresses. The authors of this study found a significant association between allele frequency and latitude. Since hsp70 is associated with heat resistance, and latitude is associated with both average temperatures and thermal extremes along the sampled transect, it seems possible that changes in temperature are causing natural selection to favour alternative hsp70 alleles. However, it is important to note that before selection can be invoked as an explanation for clinal variation in allele frequencies, the potential role of historical processes such as founder events and bottlenecks must be taken into account because changes in allele frequencies may simply reflect the genotypes of founding individuals.

Stochastic events such as bottlenecks become a less likely explanation for clinal variations in allele frequencies if similar genetic gradients are found in multiple geographically distinct regions. In D. melanogaster, concordant clines of alcohol dehydrogenase (Adh) alleles have been found in several different regions throughout the world. D. melanogaster feeds on rotting fruit, which ferments and produces alcohol. The Adh allele is important in the metabolism of this alcohol, and flies that lack Adh activity are extremely sensitive to the toxic effects of alcohol. A worldwide polymorphism in D. melanogaster Adh activity is attributable to a single amino acid replacement that determines whether or not individuals will have allele F (fast), which confers relatively high Adh activity, or allele S (slow), which confers relatively low Adh activity. The frequency of the F allele increases with latitude in several geographically distant parts of the world, including China (Jiang, Gibson and Chen, 1989), India (Parkash, Shamin and Vashist, 1992) and the Mediterranean (David et al., 1989). This agreement between allele frequencies and latitude in several different areas cannot be easily explained by random neutral processes, therefore associations between latitude and the frequencies of Adh alleles present a compelling case for selection along environmental gradients.

Fst versus QST

Selection can also be inferred from comparisons between FST and variations in quantitative traits, which are traits that are influenced by several different genes. Many important traits are quantitative, including height, weight and measures of reproductive fitness such as clutch size in birds and time to flowering in plants. These contrast with qualitative traits, which are discrete and controlled by one or a few loci, such as the round versus wrinkled seeds that were made famous by Mendel's experiments. Quantitative traits are controlled by a suite of genes known as quantitative trait loci (QTL). They are often influenced strongly by environmental effects, e.g. time to flowering in plants will depend on external factors such as temperature and photoperiod as well as the relevant genes. The variance in quantitative traits can therefore be partitioned into the amount of variation that is due to genetic factors, which is known as the genotypic variance (^2g), and that due to environmental factors, which is known as the environmental variance (c2e).

Differentiating between the contributions made by environmental and genetic factors to the phenotypic variation of quantitative characters can be problematic because many of these characteristics are phenotypically plastic. In laboratory settings, breeding experiments and pedigree analyses can be used to good effect, but these are seldom practical for wild populations. An alternative and often more practical way to estimate the genetic component of a quantitative trait is by calculating its heritability (h2). Heritabilities range from zero to one, with a herita-bility of one meaning that a characteristic is determined purely by genetics, with absolutely no environmental influence. One way to estimate heritability is from parent-offspring regressions, which provide a comparison of parent and offspring phenotypes. If parent and offspring phenotypes consistently show a strong similarity to one another regardless of environmental influences, then genotypic variance (and hence heritability) is high and environmental variance is low. Parent-offspring regressions were used to estimate the heritability of bill depth in the medium ground finch Geospiza fortis on one of the Galapagos Islands. By measuring the phenotypic variance of bill depth in parents and offspring, Boag (1983) calculated an estimated heritability of 0.90 for this trait. Estimates of herit-

Table 4.11 Some examples of heritability that have been obtained through comparisons of the phenotypes of known relatives, e.g. parent-offspring regressions. The extent to which a trait will be influenced by environmental variables is inversely proportional to its heritability



Heritability Reference

Medium ground finch

(Geospiza fortis) European starling

(Sturnus vulgaris) Squinting Bush Brown butterfly (Bicyclus anynana) Common frog tadpoles

(Rana temporaria) Egyptian cotton leafworm (Spodoptera littoralis)

Mouse (Mus musculus) Cotton-top tamarin (Saguinus oedipus) Tree snail (Arianta arbustorum)

Body weight Tarsus length Egg size

Body size

Haemolymph henoloxidase activity (immune response) Life span Body size

Shell width


Boag (1983)

Smith (1993)

Fischer, Zwaan and Brakefield (2004)

Pakkasmaa, Merila and

O'Hara (2003) Cotter and Wilson (2002)

Klebanov et al. (2000) Cheverud et al. (1994)

Cook (1965)

ability can also be obtained by comparing phenotypic variance in full-siblings, i.e. siblings with the same mother and father, or in half-siblings, i.e. siblings with only one parent in common. Further examples of heritability are given in Table 4.11.

Once heritability estimates have been obtained, the genotypic variance of QTLs can be partitioned within and between populations. This is designated as QST and is comparable to FST because it represents the degree to which populations are genetically differentiated. QST is calculated as:

where a^between) is the amount of genotypic variance between populations and a^(within) is the amount of genotypic variance within populations.

We know that FST, when based on neutral genetic markers, estimates the degree to which populations have diverged from one another as a result of gene flow and genetic drift. The QST values should show similar levels of population differentiation if they are based on neutral quantitative traits, but estimates of QST and FST are often discordant. A survey of the literature reveals three possible outcomes in comparisons of QST and FST between populations of the same species. Under the first scenario QST > FST, and this means that quantitative traits have differentiated to a greater extent than would be expected by genetic drift alone. This is often taken as evidence that directional selection is favouring different phenotypes in different populations. In the second scenario QST = FST, which, as noted above, could mean that the quantitative trait is selectively neutral. It is important to note, however, that parity between QST and FST does not necessarily mean that the quantitative trait is neutral, because in these situations we cannot distinguish between the forces of selection and drift. The third possible outcome is QST < FST, which means that population differentiation is less than would be attributable to drift and therefore the same phenotype is being selected for in multiple populations.

To date, few studies have found instances in which QST = FST. QST is occasionally less than FST, one example being an endemic Mediterranean plant species Brassica insularis (Petit et al., 2001). Populations had average FST values of 0.213, whereas overall QST values were only 0.023 in seedlings and 0.087 in adults. These relatively low QST values have been attributed to selection pressures that have decreased the phenotypic variability of this species. Reasons for the discrepant seedling and adult QST values are less clear, although this may be attributable to age-dependent selection. The majority of comparisons between Qst and Fst have found QST > FST (Figure 4.15). An extreme example of this was found in Finnish populations of Scots Pine (Pinus sylvestris) along a latitudinal gradient from 60°N to 70°N. Each year, the timing of bud burst occurs around 21 days earlier in the northernmost populations compared with those further south. This is a quantitative trait with a strong genetic component that provides a QST value of 0.80. In contrast, FST values calculated from allozymes, microsatellites, RAPDs and RFLPs all revealed very low population differentiation (FST <0.02;

Figure 4.15 The relationship between FST and QST in 29 species (21 plants, 5 invertebrates, 3 vertebrates). The dashed line in the middle represents parity between the two measurements; points below this line mean that FST < QST, and points above this line mean that FST > QST. The FST values were based on data from allozymes, microsatellites, RFLPs and RAPDs. The QST values were based on a mixture of morphological and life history traits. After Merila and Crnokrak (2001), McKay and Latta (2002) and references therein

Figure 4.15 The relationship between FST and QST in 29 species (21 plants, 5 invertebrates, 3 vertebrates). The dashed line in the middle represents parity between the two measurements; points below this line mean that FST < QST, and points above this line mean that FST > QST. The FST values were based on data from allozymes, microsatellites, RFLPs and RAPDs. The QST values were based on a mixture of morphological and life history traits. After Merila and Crnokrak (2001), McKay and Latta (2002) and references therein

Karhu et al., 1996). Evidently, directional selection is favouring a relatively early date of bud burst in northern populations.

Was this article helpful?

0 0

Post a comment