You are on page 1of 11

REVIEWS

F u n da M e n ta l c o n c e p t s i n g e n e t i c s

Genetics and the understanding


of selection
Laurence D. Hurst
Abstract | At first sight selection is a simple notion, and some consider it the most
important evolutionary force. But how important is selection, is it really so trivial to
understand and what are the alternatives? Here I discuss how genetics is crucial for
addressing all of these questions: genetics allowed the concept of natural selection to
become viable, it contributed to our understanding of the complexities of selection and it
spurred the development of competing models of evolution. Understanding how and why
selection acts has important potential applications, from understanding the mechanisms
of disease and microbial resistance, to improving the design of transgenes and drugs.

Synonymous mutation
“How extremely stupid not to have thought of that!” was There are still areas of debate and uncertainty asso-
A nucleotide change in an T. H. Huxley’s famous response to the concept of natu- ciated with each of these developments in our under-
exonic DNA sequence that ral selection. The requirements for selection also make standing of selection that are relevant to our current
does not result in a change in it seem inevitable: that there are more individuals born understanding of the role of selection in shaping popu-
the encoded amino acid. A
synonymous mutation that
than can survive and reproduce, that individuals vary in lations and genomes. For example, recent work on gene
reaches fixation becomes a their ability to survive and reproduce and that some of and genome evolution has suggested that selection might
synonymous substitution. this variation is heritable. Partly because selection can reach further than was previously supposed, which has
operate under a wide range of conditions, some sup- practical implications. For example, if there is selection
pose it to be the most powerful explanatory principle1 on gene order 2, it could be important to understand
for understanding the fate of any replicating system. But where a transgene inserts3, and showing that synonymous
is selection inevitable? Can selection act if the selective mutations are under selection might suggest they cause
advantage of an allele is small? Can it even allow deleteri- disease4. Therefore, understanding how and why selec-
ous alleles to invade a population? What limits the action tion acts might also improve genetic intervention5 and
of selection? What are the alternatives to selection? our understanding of disease.
In this Review, I discuss how understanding genetics
is key to answering all of these questions. This discus- Mendelian genetics and the viability of selection
sion is based around three important contributions that Although natural selection that favours the survival
genetics has made to the understanding of selection. of individuals with a fitness advantage might seem
First, Mendelian genetics enabled Darwin’s idea of natu- an obvious concept, this was not the case when it was
ral selection to be accepted and developed by providing originally proposed by Darwin. Natural selection
a mode of inheritance in which selection can operate. was initially not widely accepted owing largely to con-
Second, a fuller appreciation of the complexities of genet- fusion over how inheritance worked. It was commonly
ics allowed a more nuanced understanding of selection. thought that inheritance blended the traits of two indi-
For example, it was initially thought that the fate of an viduals — for example, that the offspring that result from
allele was determined solely by the fitness advantage it a cross between a red flower and a white flower should be
conferred on an individual, but more developed models pink. As blending inheritance removes variation from a
of selection that consider genetics can explain why even population6, it is incompatible with selection because no
Department of Biology and deleterious alleles can spread. Third, genetic observa- individual will have an advantage over another.
Biochemistry, University of tions, such as the unexpectedly high rates of evolution The discovery and development of Mendelian genet-
Bath, Somerset, BA2 7AY, UK. and the common occurrence of polymorphisms, sug- ics resolved the uncertainty over Darwin’s theory of
e‑mail: l.d.hurst@bath.ac.uk
doi:10.1038/nrg2506
gested that selectionist models had limitations, and so natural selection. Under Mendelian inheritance and
Published online spurred the development of competing explanations of with random mating, genotype frequencies after one
2 January 2009 evolution — the neutral and nearly neutral theories. generation do not change. Therefore, unlike blending

NATURe RevIeWS | Genetics volUMe 10 | FeBRUARy 2009 | 83

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

Box 1 | invasion of an allele under Mendelian and non-Mendelian inheritance when the heterozygote is the most fit genotype (a phe-
nomenon known as overdominance; BOX 2), models
Under Mendelian inheritance, the fate of an allele that provides a fitness that ignore genetics cannot explain why homozygotes
advantage can be understood by the formula set out here: are stably maintained in the population because this is a
Suppose that the genotype AA has a small fitness bonus (s) compared with the
consequence of Mendelian segregation.
genotype aa. The genotype Aa has a fitness bonus hs, in which h is a term that
captures the effects of the dominance of A over a with regard to fitness. Assuming
Hardy–Weinberg ratios, the frequency of A in the next generation (p′) is given by the How deleterious alleles can invade
equation p′ = [p2 (1 + s) + pq (1 + hs)]/ω in which ω= p2 (1 + s) + 2 pq (1 + hs) + q2, and is To fully appreciate the role of genetics in selection, it
the weighted mean fitness in the population; p is the frequency of A; q = 1 – p and is necessary to consider that there are different modes
is the frequency of a. Let us now suppose that all the population is type aa and a by which selection can operate (BOX 2) . A number
new A allele enters. The likely fate of this allele is determined by the slope of the of tests have been developed to distinguish these types of
line describing p′ as a function of p when p ≈ 0. If the slope is >1, the allele invades selection and are introduced in BOX 3.
the population. The condition for this to occur is hs >0 — that is, the heterozygote A demonstration of the value of genetics to under-
(Aa) must have higher fitness than the aa homozygote. However, the striking thing standing the complexities of selection is given by exam-
about this solution is what is missing. There is nothing in the solution concerning
ining the reasons for the spread of deleterious alleles.
the mode of inheritance: individual fitness is the sole determinant of the fate
The basic assumption of natural selection is that
of the allele.
The contrast with what is observed under non-Mendelian inheritance is striking, alleles that increase individual fitness invade while del-
as can be demonstrated by considering the invasion of a deleterious allele. eterious alleles are eliminated. However, this assump-
Suppose that heterozygotes transmit the A allele at a rate k and the a allele at a tion was challenged by cases in which deleterious alleles
rate 1 – k. For Mendelian inheritance, these rates would be equal: the unique point did spread. Three extensions of the simple genetic
k = 1 – k = 0.5. Suppose that AA has fitness 1 – t, Aa has fitness 1 – s and aa has framework can explain the spread of deleterious alle-
fitness 1. For the deleterious allele, A, to invade, the inequality 2k (1 – s)>1 must les: non-independence between alleles in a genome
hold105. If k is replaced with 0.5, under Mendelian inheritance, s<0 must hold, and because of linkage; non-Mendelian inheritance; and
thus, as above, the allele is beneficial in heterozygotes. What if k = 1 — that is, non-independence between the same alleles in dif-
Aa × aa gives only Aa offspring? For invasion, s<0.5 must still hold. The allele can
ferent genomes. each of these is considered below.
reduce the fitness of the organism by up to 50% and still spread. Any allele for
The conceptual advances that underlie these insights
which k>0.5, for which invasion is possible despite s<0, is one class of selfish gene
or selfish genetic element23 (not to be confused with the concept of a selfish have provided a more nuanced understanding of how
gene that was defined by R. Dawkins, which describes any allele that can selection operates.
deterministically invade, regardless of the underlying logic). Under Mendelian
conditions, only the fitness of the individual is important, whereas under Linkage, hitchhiking and background selection.
non-Mendelian inheritance, both the rate of transmission and the fitness of the Although the simplest assumption is that each locus
organism determine the fate of the allele. within a genome is independent, this is not realistic.
linkage can result in non-independence between alle-
les, which is quantified as linkage disequilibrium, with the
inheritance, Mendelian inheritance retains variation fate of one allele influencing the fate of others. linkage
and therefore the possibility of selection. This reten- disequilibrium can allow deleterious alleles to spread
tion of variation applies to both discrete alleles (regard- (hitchhiking) and can prevent advantageous alleles from
Hardy–Weinberg ratio less of ploidy) and continuously varying traits (such as invading (thus imposing limits on selection, which is
The binomial distribution of height)7. Importantly, under Mendelian inheritance, the considered in BOX 4).
genotypes in a population, sole determinant of whether the allele will spread when Hitchhiking occurs when the spread of a positively
such that frequencies of the
genotypes AA, Aa and aa will
it enters a population is whether the fitness of hetero- selected, advantageous allele carries linked alleles10 in a
be p2, 2pq and q2, respectively, zygotes is greater than that of wild-type homozygotes selective sweep. If the inheritance is Mendelian, this phe-
in which p is the frequency of (BOX 1). Basic Mendelian genetics therefore showed that nomenon can only occur if the deleterious effects of the
allele A and q is the frequency Darwin’s concept of natural selection is viable and also hitchhiking alleles are not larger than the advantageous
of allele a.
suggested that only fitness, not the mode of inheritance, effects of the positively selected allele. Therefore, hitch-
Evolutionary stable needs to be considered when assessing the action of hiking usually does not cause the spread of highly delete-
strategy theory selection on an allele. rious alleles. evidence of hitchhiking and its concomitant
An approach to mathematically Given that under Mendelian inheritance the likely effects of reducing variation and establishing long-range
modelling evolution that outcome of selection can be unaffected by the mode of haplotype blocks is now commonly used to identify posi-
defines equilibrium positions
as positions at which, if all
inheritance, many researchers (especially those working tive selection in genomes11,12 (BOX 5; reviewed in ref.12).
individuals are using the same on animal behaviour) use mathematical models of selec- For example, analysis of hitchhiking has shown that the
strategy, invasion by rare tion that ignore the underlying genetics and concentrate spread of the caspase 12 pseudogene has been driven by
individuals who adopt a only on the effects of individual fitness. Such models (for positive selection in humans, probably because it confers
different strategy is impossible.
example, the evolutionary stable strategy theory and the protection from sepsis13.
Optimality theory optimality theory ) have the advantage of simplifying Hitchhiking is also key to understanding mutation
A quantitative evolutionary the mathematical analysis of complicated problems. rates and therefore ‘evolvability’. In a sexual species,
model that defines the However, although population genetics provides a an allele that increases the mutation rate (a modifying
maximum or minimum fitness defence for such an approach8, it also suggests there are allele) will generate many deleterious mutations and
values for a given trait under
specified constraints. It is often
limits to the assumption that genetics does not matter should be eliminated from the population. The modi-
used to investigate life-history (the ‘phenotypic gambit’ (ref. 9)). In many circumstances fying allele can only spread if it is in linkage disequilib-
evolution. there is evidence that genetics does matter. For example, rium with a new positively selected allele14. The logical

84 | FeBRUARy 2009 | volUMe 10 www.nature.com/reviews/genetics

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

Box 2 | Modes of selection The converse effect to hitchhiking 15 is background


selection, in which selective elimination of a deleterious
Selection can be subdivided into different types, depending on the evolutionary allele from the population leads to the loss of other alle-
outcome. The main concepts of the different modes of selection are described below. les in linkage disequilibrium (BOX 4). Both hitchhiking
Positive selection and background selection can cause reduced variation
Positive selection (also known as Darwinian selection or directional selection if it in genomic domains with low rates of recombination16–18.
occurs on a quantitative trait) is the spread to fixation of an allele that increases the However, these phenomena may not completely explain
fitness of individuals. Numerous examples of positive selection have been observed — the relationship between genetic variation and recombi-
for example, antibiotic resistance or the increase and decrease in the prevalence of the
nation rate. For example, at least in mammals, there is
melanic form of the peppered moth106. More recently, the impact of humans on positive
selection has been shown by the preference of hunters of bighorn sheep for prey with evidence that recombination might itself be mutagenic
large horns, which led to a shift in horn size and body weight107, and by the high capture owing to errors introduced during double-strand break
rates of cod, which led to selection for smaller fish that develop faster108. In addition to repair 19–21. This is consistent with the positive correlation
these examples at a population level, there are also examples at a molecular level109,110. between the rate of recombination and the rate of sub-
Although Darwin initially suggested that sexual selection and natural selection are stitutions19–21 that is observed, for example, at the mam-
different, sexual selection is now commonly considered a subclass of positive selection. malian pseudoautosomal region, which has a high rate of
Sexual selection is the spread of traits that cause an individual to be more attractive to recombination19.
mates and therefore have more reproductive success. However, why females chose Despite the effects of linkage, single-locus models can
males and whether they gain anything more than attractive sons remains a matter of be defended. Because of recombination, any given allele
debate111. In some cases, female choice leads to traits that are deleterious to viability
is exposed on average to many genetic backgrounds,
(such as large tails) but that enhance fertility by increasing attractiveness to females.
and therefore the average fitness effect of the allele is
Purifying selection what matters, not its current neighbours. More gener-
Purifying selection (also known as negative or stabilizing selection) eliminates
ally, an allele that is not itself subject to selection but that
deleterious mutations. Assuming that the outcome of evolution by natural selection
should be the production of well-adapted organisms, we expect this to be the most
has consequences at other loci (for example, an allele
common mode of selection; random tinkering with a machine that functions well is that changes the rate of recombination) will, under a
likely to cause damage. The assumption that purifying selection is common underpins wide range of conditions and in the absence of linkage
many attempts to find conserved functional motifs in genomic sequences (known as disequilibrium, evolve as if to maximize mean fitness22.
phylogenetic footprinting112,113). However, distinguishing purifying selection and low
mutation rates can be difficult as both processes result in little sequence change (for Non-Mendelian inheritance and selfish elements.
example, compare ref. 88 and ref. 114). Although hitchhiking is normally limited to weak del-
Balancing selection eterious alleles, relaxing the assumption of Mendelian
Balancing selection (also known as disruptive selection) is selection that favours inheritance provides the possibility of the spread of
diversity. Under balancing selection, the spread of an allele never reaches fixation, and highly deleterious alleles (‘selfish genetic elements’).
therefore it can initially seem to be undergoing positive selection, but it then These alleles deterministically invade the population
undergoes negative selection when its frequency is too high. Thus alleles cannot be and simultaneously create the conditions for the spread
classified as advantageous or deleterious. This type of selection can have different
of other alleles at different loci that suppress the selfish
forms, including frequency-dependent selection, which is central to the evolution of
mimicry115, and overdominance (see the main text). Although overdominance is rare, it
element 23. For example, male flies that are heterozygous
can occur when heterozygotes are resistant to a parasite — for example, in sickle-cell for the segregation distorter (Sd+/–) transmit the drive
anaemia and Tay–Sachs disease116. allele (Sd+) almost exclusively 24. This phenomenon in
Diversity might also be selected for in cyclical host–parasite co-evolution (‘Red which one allele is transmitted to more than half the off-
Queen’ selection117), in which there is constant change, with hosts under selection for spring of a heterozygous parent is known as meiotic drive,
resistance and parasites under selection to counteradapt. However, the relevance of and it can allow a strongly deleterious allele to invade
Red Queen cyclical dynamics is debatable: because there is little evidence to support (BOX 1) and permits the spread of unlinked suppressors of
cyclicity and co-evolution118, it has been suggested that it may be only a theoretical meiotic drive.
nicety119. It has conversely been argued that Red Queen selection is hard to prove and Uniparental inheritance of cytoplasmic factors also
that there is some evidence that supports the theory120; for example, from archives of
allows selfish elements to spread. Cytoplasmic sex-ratio
past gene pools in lake sediments121. By contrast, in vitro studies have reported the
distorters are transmitted exclusively by females and distort
presence of stable polymorphisms rather than cyclicity in host–parasite systems122.
Balancing selection is often suggested to be unusual, a theory that is partly supported the sex ratio by favouring females. For example, they kill
by the failure of genome scans to detect balancing selection123. However, this failure males, feminize males and make females asexual, thereby
might reflect a statistical weakness of such scans. With the recent increase in SNP data, leaving only daughters25,26. The male-killing element can
locus-specific analyses are producing more examples than might have been spread if this process redistributes resources to surviving
expected124–127. However, current examples are largely associated with immune sisters, which contain a clonal relative of the killing
functions (as expected from the Red Queen model) and might therefore be allele27. The same logic underpins cytoplasmic male
unrepresentative of the genome as a whole. sterility in plants28.
A further example of how deleterious factors can
spread is provided by the cytoplasmic incompatibility caused
Linkage disequilibrium consequence of this should be that it is easier for dif- by Wolbachia infection in insects29. In males, Wolbachia
A measure of genetic ferent mutation rates to be evolved by asexual species, cause the death of zygotes that lack Wolbachia. In effect,
associations between alleles which lack recombination and therefore link together all Wolbachia transfer a toxic substance into sperm that is
at different loci that indicates
whether certain allelic
alleles, than by sexual species, in which recombination neutralized by an antidote produced by the same bacte-
combinations are more or less limits hitchhiking 14; however, this theoretical prediction ria when the bacteria are maternally transmitted to the
common than expected. remains to be tested. zygote. The death of all of the progeny when the female

NATURe RevIeWS | Genetics volUMe 10 | FeBRUARy 2009 | 85

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

is not infected with Wolbachia ensures an increase in the


Box 3 | diagnosis of the mode of selection from substitution data
frequency of Wolbachia-infected individuals30. Such a
1.7 cytoplasmic incompatibility factor is currently spread-
ing in Californian fruit flies31 and many 32, if not most 33,
1.6
species of insect are probably infected with Wolbachia or
1.5 another cytoplasmic bacterium, although the proportion
1.4 of these bacteria that manipulate their hosts is unknown.
Although selfish elements are important in demon-
1.3
strating how non-Mendelian inheritance can allow the
1.2 spread of deleterious alleles, their relevance in organisms
1.1 remains unclear. A wide range of phenomena have been
suggested to be explained by selfish elements23, includ-
KA/KS

1.0
ing the evolution of sexes, genomic imprinting, polyan-
0.9 dry 34, sex determination and hybrid sterility. Although
0.8
the importance of selfish elements in these examples is
unresolved, selfish elements do explain the otherwise
0.7 paradoxical persistence of sterile individuals in some
0.6 populations; for example, in the cases of segregation dis-
torters in flies and the T complex in mice35, homozygotes
0.5
are sterile but are maintained in the population because
0.4 of meiotic drive in heterozygotes. Whether selfish
0.3 genetic elements are an intellectual curiosity or whether
they drive the evolution of genetic systems or speciation
0.2
remains to be resolved; it will be important to investigate
0.1 how common they are to determine their role.
0
Relatedness, kin selection and inclusive fitness. A fur-
Frequency
ther explanation for how deleterious alleles can spread is
Several tests have been developed to determine the mode of selection by | Genetics
Nature Reviews gained from considering that an allele in one individual
comparing genome sequences from different species, and the principles of these can affect the fortune of the same allele in other indi-
tests are introduced here. If a protein is evolving neutrally, then a mutation that viduals. This has led to the understanding that traits can
alters an amino acid should have the same chance of fixation as a synonymous spread not because they are beneficial to the individual,
mutation. By comparing orthologous genes between any two species, we can but because they promote the fitness of other bearers
partition changes into non-synonymous and synonymous ones. Adjusting for the
of the same allele (relatives). Perhaps the most obvious
fact that most mutations will have been non-synonymous (and for the possibility of
traits that are bad for individual fitness are acts of self-
multiple hidden changes), we can then compare the number of non-synonymous
changes per non-synonymous site (KA) with the corresponding number of sacrifice — dramatic examples include the existence of
synonymous changes per synonymous site (KS). The KA/KS ratio is then a diagnostic sterile castes in social insects and termites that explode.
of how a gene evolves. If this ratio is equal to one, we cannot infer the action of To understand the evolution of such traits, the normal
selection; if it is greater than one we infer positive selection; and if KA/KS << 1 we genetic framework needs to be expanded to consider
infer purifying selection. An example of this type of analysis is shown in the figure: a the impact of an allele on other individuals. In the con-
histogram and box plot of the distribution of the KA/KS ratios for 5,057 mouse–human text of social evolution, we need to consider the fitness
orthologues. These data indicate that few genes have KA/KS >1 , but that most have of the organism that acts on other alleles (the ‘actor’)
KA/KS<<1, which is indicative of the importance of purifying selection. In general, and its relatives that share the same allele (inclusive fit-
many examples of genes with KA/KS > 1 are probably driven by host–parasite ness), rather than the fitness of the actor alone. Selection
coevolution109.
does not need to be good for the individual36, but an
Applied across a whole gene in a pairwise comparison, the KA/KS ratio is a crude
tool, as positive selection might be present in just one lineage or in just a few
individual can promote its own alleles by helping a rela-
codons. The inference of positive selection can also be undermined — for example, tive if b/c>1/r, in which b is the benefit gained by the
by purifying selection on synonymous mutations128. The use of multiple species receiver, c is the cost to the altruist and r is the relat-
alignments allows the form of selection to be resolved129. edness between the two (Hamilton’s rule). This princi-
An extension of this between-species analysis is the McDonald–Kreitman test130, ple, termed kin selection, has robust explanatory power,
which analyses both between- and within-species differences. If the substitutions in and can be used to predict the course of experimental
a protein are due to positive selection, we should see a higher ratio of evolution in microorganisms37.
non-synonymous to synonymous differences between species than within species. If A deeper understanding of kin selection and selfish
the assumption of neutrality of synonymous mutations is correct, then it is possible elements is provided by asking whether a rare allele will
to test whether adaptive evolution has occurred and estimate the proportion of
invade (BOX 1), rather than whether the allele is ‘good for
substitutions that are a consequence of positive adaptive evolution. However, this
method is biased by rare weakly deleterious alleles, and the failure to allow for
the individual’ or ‘good for the group’. This concept is
this limitation might have led to serious underestimation of the proportion of adaptive illustrated by sex-ratio evolution: in a species in which
changes in both flies and bacteria73. An additional method is the Hudson–Kreitman– only females invest in the next generation, making males
Aguadé test131. In the strict sense, this is a test to reject neutrality that compares the is a form of altruism. For example, Fisher 38 identified
rate of polymorphism to divergence for multiple genes. If the ratio varies more among that a sex ratio of 1 to 1 is a stable investment ratio,
genes than is expected from null neutral assumptions, neutrality is rejected. although it does not maximize population fitness, and

86 | FeBRUARy 2009 | volUMe 10 www.nature.com/reviews/genetics

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

Box 4 | linkage and the limits of selection change frequency owing to chance alone (genetic drift).
Recognition of this limitation led to the development of
As well as providing an explanation for how deleterious alleles spread, linkage also new models. The first of these was neutral theory, which
imposes limits on positive selection. Hill and Robertson examined linked alleles in a provided an explanation for evolution that, at least at a
sexual species and showed that selection at one locus can interfere with selection at
molecular level, offered serious competition to natural
a second, beneficial mutation and reduce the probability of its fixation132. This is now
selection. However, as outlined below, this model also
known as Hill–Robertson interference. Crucially, the degree of interference increases
with genetic linkage between the loci under selection. Generally, this means that has many limitations and led to the suggestion of an
linkage between sites under selection will reduce the overall effectiveness of alternative, the nearly neutral theory.
selection in natural populations. Domains of low recombination, and hence Two potential limitations of selection were high-
longer-range linkage disequilibrium, are expected to be those in which the strength lighted by two predictions made by the population
of both positive selection (Hill–Robertson interference) and purifying selection genetics of selection developed by the modern synthesis.
(background selection, see the main text) are limited. The first prediction was that polymorphism should be
The importance of Hill–Robertson interference is still debated132, and interpretation rare. It is difficult to find theoretical conditions in which
of the results is made difficult by a lack of understanding of the confounding alleles should be polymorphic, assuming that balancing
variables. For example, the Hill–Robertson model predicts that the extremities of
selection (BOX 2) occurs rarely. However, protein elec-
exons should be better adapted than cores owing to less interference from introns;
trophoresis and, later, genomic sequencing showed that
this is suggested to be a possible selective advantage for long introns133. However, in
Drosophila melanogaster, the differences in adaptation between the interior and polymorphism is unexpectedly common44–46. The second
exterior of exons is due to the location of splice enhancers, and not necessarily to the prediction, called Haldane’s dilemma47, is that rates of
different selection strengths84. Likewise, although the rates of evolution are adaptive evolution would be limited by population size.
correlated with the local recombination rate as predicted by the Hill–Robertson Haldane argued that for each selective substitution event
model134,135, in yeast (Saccharomyces cerevisiae) this is at least in part due to different there has to be selective death. As many selective events
classes of gene being located in different genomic recombination domains134. require many deaths, the rate of evolution should have
Possibly the most important role of Hill–Robertson interference that has been an upper limit that is determined by the population size.
proposed is in enabling the selection of alleles that affect the recombination rate. Nevertheless, analysis of the rate of evolution of several
These alleles spread by weakening the degree of interference between linked sites136.
proteins suggested that the rate was too high for all of the
This is because background selection against deleterious mutant alleles (see the main
observed evolution to have occurred through selection48.
text) provides a stochastic advantage to recombination, an advantage that increases
with population size. With low levels of recombination, selection at other loci These two apparent problems with selection helped to
reduces the effective population size and genetic variance in fitness at a focal locus. prompt the development of the alternative models based
As a consequence, the population is less able to respond to selection and to rid itself on evolution by genetic drift.
of deleterious mutations. Recombination, by contrast, combines chromosomes of
intermediate fitness to create chromosomes that are free from deleterious The rise and fall of the neutral theory. The first alterna-
mutations. However, this hypothetical role of Hill–Robertson interference remains to tive model based on drift was neutral theory. In strict
be rigorously tested. neutral theory, alleles are modelled as neutrally buoyant
diffusing particles of gas that can go up or down with
equal probability by chance. In this framework, polymor-
Haplotype block Hamilton39 recognized special conditions for biased sex phism is expected to be common and to scale with the
A chromosomal region in which ratios (for a review see ref. 40). These examples can all population size according to equation 1, in which H is
groups of alleles at different
genetic loci are inherited
be explained by asking whether an allele will invade. The heterozygosity, Ne is the effective population size and
together more often than kin-selection framework is valid for assessing sex ratios41, µ is the mutation rate (fIG. 1).
expected by chance. but this framework is not a dynamic model, which might 4Ne µ
reduce its power 42. In addition, the relationship between H= (1)
Pseudoautosomal region
kin selection and the classical multilocus approach is still 1 + 4Ne µ
Any section on a telomeric end
of an X or Y chromosome that debated43, and because it can be difficult to precisely esti- Although this model is qualitatively correct 49, strict
undergoes recombination in mate the core parameters (b, c and r) of Hamilton’s rule it neutral theory has little quantitative explanatory power.
the heterogametic (XY) sex. is difficult to quantitatively test kin selection. For example, the theory predicts large differences in
Meiotic drive
The three concepts of linkage, non-Mendelian inher- heterozygosity between humans (Ne~104) and fruit flies
A distortion of meiotic itance and kin selection described here show that an (Ne~106), but the observed difference is modest 50–52
inheritance such that one understanding of genetics allows selection to explain (fIG. 1). Neutral theory is therefore a useful null hypoth-
allele at a heterozygous site is why deleterious alleles can spread. each of these mech- esis because it makes strong quantitative predictions, but
recovered in more than half of
anisms is also linked to new questions as to how they it is a poor model as these expectations are rarely met 53.
the gametes (in contrast to the
expected 50:50 Mendelian shape evolution in practice, and they provide insights Further evidence against the strict neutral theory
segregation) owing to effects into the reach and role of selection. comes from analysis of the rates of evolution between
that occur before zygote lineages. If molecular evolution is neutral, then we expect
formation. alternatives to selection that, on average, lineages with the same mutation rate will
Cytoplasmic factor The explanations for the spread of deleterious alleles evolve at the same rate. owing to chance, we expect some
A gene present in host show how understanding genetics is important to our lineages to evolve at different rates. The neutral theory
organelles, such as understanding of how selection might explain otherwise predicts that the variation in rates (V) between lineages
mitochondria, or in intracellular unexpected phenomena. However, other genetic discov- should be equal to the mean rate (M), as the process has
parasites, such as Wolbachia,
that is typically passed from
eries highlighted limitations to models of selection. A a Poisson distribution. Consequently, dispersion (D), which
mother to offspring through major limitation of deterministic selectionist models of is equal to V/M, should equal one. Most early analyses
the cytoplasm. evolution is that they miss the possibility that alleles can suggested that D does not equal one for synonymous or

NATURe RevIeWS | Genetics volUMe 10 | FeBRUARy 2009 | 87

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

Cytoplasmic sex-ratio non-synonymous changes49,54. Therefore, neutral theory The nearly neutral theory. The quantitative limitations
distorter became less prominent and is no longer adhered to, of the neutral theory led to the development of a more
A factor inherited in the although its value as a null hypothesis remains. accurate model of drift. This model, known as the nearly
cytoplasm (for example, It is also worth noting that Haldane’s dilemma, which neutral theory, is more than a minor modification of
mitochondria or heritable
bacteria) that modifies the
was originally cited as a reason for needing alternative the neutral theory, not least because the two make
sex ratio, typically causing a models to selection, is now mostly considered not to be different predictions.
female bias. a problem55. Therefore, the cost of substitution suggested Strict neutral theory considers mutations with no
by Haldane cannot be used as a defence of neutral theory. selective effect. This is improbable as all mutations are
Cytoplasmic incompatibility
Haldane’s calculations assume hard selection56, which is likely to have some, albeit potentially small, effect. To
A sperm–egg incompatibility
that is usually associated with
selection that increases the total number of deaths, and address this weakness, the nearly neutral theory 49,59
Wolbachia infection. Wolbachia also assume that the effects of multiple mutations are considers mutations that have a small effect by using
modify the host sperm in the additive. However, selection can operate such that an diffusion equations to evaluate the possibility that
testes, and the same strain of advantageous genotype occupies a higher proportion of particles might not be neutrally buoyant. Some parti-
Wolbachia must be present in
the egg to rescue this
limited niches (soft selection), in which case the amount cles can be slightly buoyant (weakly advantageous) or
modification. Absence of of death is unaffected. Fixation of a selectively advan- slightly more dense (weakly deleterious), but both types
rescue results in incompatibility tageous mutation therefore need not involve selective can be fixed. In this regard, the nearly neutral model
and zygotic lethality. mortality 56. Assuming that each extra mutation has non- focuses on the concept that adaptation may be due not to
additive effects can also mitigate Haldane’s dilemma57. strong selection of rare variants with large effects, but
However, given that positive selection can potentially to weak selection of common variants.
affect non-coding sequences58 and proteins, reexamination The neutral and nearly neutral theories make differ-
of Haldane’s dilemma might be warranted. ent predictions. Strict neutral theory predicts that rates
of evolution should be independent of population size
and simply equal to the mutation rate. By contrast, an
Box 5 | application of intrapopulation data to assay the role of selection
important consequence of the nearly neutral model
Selection leaves its mark not only on the divergence between species (BOX 3), but also is that population size can matter. The nearly neutral
on the patterns within a species. These various footprints of selection have been used model recognizes three classes of deleterious mutations:
to devise tests for selection that rely on different aspects of the effects of selection. those that are effectively neutral (s<<1/2Ne), those that
tests based on population differentiation are slightly deleterious (s~1/2Ne) and those that are del-
If a selective sweep has started in one subpopulation but is yet to reach others, then eterious (s>>1/2Ne), in which s is the fitness cost (fIG. 2).
we should observe extreme levels of population differentiation. The Lewontin– Mutations that are effectively neutral have a substitution
Krakauer137 test takes advantage of this fact and has recently been resurrected in rate equal to that of neutral mutations. The probability
multiple forms138, including genome scans for selection139,140. This approach has been of fixation of slightly deleterious mutations is appreci-
bolstered by the development of a rigorous statistical framework to identify outlier
able, but it is lower than that of neutral mutations. For a
loci141. Further modifications to this test, which aim to discover alleles that have been
nearly fixed within a population by a selective sweep, have identified more than 300
given small value of s, the important prediction of this
strong candidate regions in humans142. model is that the rate of evolution should increase as the
population size decreases (more mutations will be in
tests based on frequency spectra
The neutral theory predicts how common, on average, the most common variant will
the effectively neutral or slightly deleterious class). Hence
be and, more generally, the relative proportions of all variants. This allows us to we expect that in large populations (for example, bacte-
determine the level of homozygosity143 if we know the number of variants. Different ria populations), purifying selection is efficient and few
deviations from neutral expectations are expected for different modes of selection: deleterious mutations fit in the two classes that can lead
purifying selection will tend to increase the fraction of mutations segregating at to fixation. This model correctly predicts that purifying
low frequencies and positive selection will increase the number of alleles seen at high selection should be weaker in species with long generation
frequencies. The central test for examining the expectation of neutrality is Tajima’s D144. times because these species have small populations60.
In this test, the average number of nucleotide differences between pairs of sequences
is compared with the total number of SNPs. Fu and Li’s test145 extends Tajima’s D by The nearly neutral versus selectionist debate. owing to
considering other species so as to polarize the changes (ancestral versus derived
the weakness of the strict neutral model, the neutralist
alleles). These tests and subsequent refinements146,147 are probably the most commonly
applied tests for neutrality and deviation from neutrality.
versus selectionist debate was abandoned and replaced
by the nearly neutral versus selectionist debate. In this
tests based on linkage disequilibrium and haplotype structure debate, the two positions are often hard to discriminate as
Variants that are positively selected shortly after spreading, or alleles under balancing
they make many similar predictions. For example, both
selection, will drag linked alleles with them. This causes an increase in linkage
disequilibrium and extends the span of haplotypes further than expected by the nearly neutral and the selectionist model can explain
chance148. Indeed, two alleles (glucose-6-phosphate dehydrogenase and CD40L) an approximately constant rate of protein evolution49,61,
associated with malaria resistance show such an extended haplotype148. Several tests and both models also claim to explain the insensitivity
have been proposed to formalize this approach to detecting selection (reviewed in of protein polymorphism levels to population size10,49–51.
ref. 149), and recent genome-wide scans that used high-density SNPs have identified Both selection and nearly neutral evolution are also
numerous possible candidates of selection in humans100,150. consistent with high dispersion values (D)62, but the
A concern with all of these tests, unlike the KA/KS and McDonald–Kreitman tests explanations they offer for this observation are differ-
(BOX 3), is that inference of selection or rejection of neutrality tends to be sensitive to ent. Advocates of selection note that high dispersion
assumptions concerning demography101,149, and extensive simulation of alternative
occurs when positive selection on a given protein occurs
possible demographies (see, for example, ref. 100) is necessary to ensure some degree
down some lineages but not others (episodic evolution).
of confidence in the result.
By contrast, nearly neutral models note that if lineages

88 | FeBRUARy 2009 | volUMe 10 www.nature.com/reviews/genetics

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

differ in Ne (possibly owing to speciation bottlenecks), 1.0


dispersion should be high. However, the selectionist the-
ory predicts that the lineages down which selection acts 0.9
are not necessarily the same for different proteins. But if
a species-wide reduction in Ne causes high dispersion, as 0.8
suggested by the nearly neutral model, we expect accel-
erated rates to be a property of the lineage. In the case of 0.7

Heterozygosity (H)
three genes that are involved in circadian clocks, careful
analysis63 showed that selection explains the data better. 0.6
However, the debate between selection and nearly neu-
0.5
tral evolution has recently been reopened by the find-
ings that D might often be equal to one and that prior
0.4
analyses were distorted by unrepresentative genes64.
Given the frustrations in attempting to discrimi-
0.3
nate between the two models, the field has moved on
to address a more direct question: what proportion
0.2
of all substitutions are adaptive? This question can be 0.185
answered by an extension of the MacDonald–Kreitman
0.1
test (BOX 3), under the assumption that positive selec- 0.056
tion leaves an excess of non-synonymous substitutions 0
compared with non-synonymous polymorphisms65–67. 0.001 0.001 0.01 0.1 1 10 100 1,000
T complex Genome-wide application of this method to protein-
A region on chromosome 17 in
0.01483 0.0567 Ne µ
the mouse that is subject to
coding genes in humans suggests that 9% of genes
might be subject to positive selection68. Unfortunately Figure 1 | Heterozygosity predicted by the neutral
meiotic drive. The complex Nature Reviews | Genetics
contains four inversions and such estimates are sensitive to methodological details, hypothesis. A plot of heterozygosity (H) predicted by the
neutral hypothesis as a function of the product of the
occupies approximately half with estimates in flies ranging between 20% and 95%69–71.
of the chromosome. effective population size (Ne) and the mutation rate (µ).
optimization of methods to estimate the proportion of The basis for this pattern is explained in the main text. The
Kin selection substitutions owing to selection, including controlling shaded area indicates the range of heterozygosities
Selection that allows an for confounding effects of demography 72 and rare weakly observed in various animals, including humans, mice and
individual to increase its fitness deleterious alleles73, is an active area of research. In addi- flies. If the neutral theory could explain the observed
by increasing the number of tion, the importance of the assumption of neutrality of heterozygosities, there would be little difference in Ne µ
copies of its genes in the
population by provisioning
synonymous mutations has yet to be evaluated. between these taxa. However, this is not the case, and
benefits to relatives.
so the absurdity of this prediction is accepted as a
the reach and role of selection rejection of the neutral hypothesis52. Figure is modified,
Modern synthesis As the discussion of linkage and weakly selected variants with permission, from ref. 52  (1974) Columbia
The prevailing theoretical University Press.
suggests, selection has limits. For example, selection is
framework of evolution that
resulted from a combination of impeded in domains of low recombination (effectively,
genetics, systematics, domains of low Ne), in species with low Ne, when the
comparative morphology and fitness benefit of a mutation is small and when herit- few examples of strict developmental constraint exist
palaeontology in the 1930s able variation is absent. But in practice what do these and the evolutionary relevance of laboratory-based
and 1940s. It is also known as
evolutionary synthesis or the
limits mean? The roles of selection in domains of low identification of potential constraints is also uncertain.
synthetic theory. recombination, in species with low Ne and when the For example, although artificial breeding experiments
fitness benefit is small are currently being explored in provide little evidence of heritable variation in sex
Effective population size evolutionary genomics, and evolutionary developmental ratios76, sex ratios naturally evolve — social spiders with
(Ne). The size of a population
biology (evo-devo) asks whether the scope of realizable heteromorphic sex chromosomes have independently
measured by the expected
effect (through genetic drift) of heritable variation might determine what phenotypic evolved biased sex ratios at least five times77.
the population size on genetic space is unexplored. A less strict concept of developmental constraint
variability. It is related to, but supposes that the profile of phenotypes is skewed by
never exceeds, the true Does heritable variation limit adaptation? Population the way that development works. To test this hypothesis,
population size (N).
genetics focuses on the fate of mutations. However, evo- artificial selection experiments can be used to find how
Poisson distribution devo researchers emphasize the role of ‘developmental easy it would be for a given trait to evolve78. However,
A discrete frequency constraint’ in shaping variation. In its strict form, devel- there have been few experimental analyses compared
distribution of the number of opmental constraint means that there is no heritable vari- with the number of theoretical studies of developmental
independent events per time
ation for certain developmental traits, as genetic variation constraints. one instructive example is butterfly wing
interval, for which the mean
value is equal to the variance. that affects these crucial aspects of development will patterning. Analyses of patterning show that, despite
result in non-viable individuals or may not be possible. the potential for constraint owing to a developmen-
Dispersion Such strict constraints are suggested to determine unex- tal coupling between different eyespots, independent
The ratio of the variance to plored morphological space74. For example, deleterious change can occur. This flexibility is consistent with the
the mean rate of evolution
observed when orthologous
effects (such as embryonic cancer) of mutations that alter diversity of wing patterns across species and argues that
sequences from at least three the morphology of neck vertebrae75 might explain why natural selection has a dominant role in shaping existing
different taxa are compared. most mammals have seven of these vertebrae. However, variation79.

NATURe RevIeWS | Genetics volUMe 10 | FeBRUARy 2009 | 89

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

1,000 µ specify the most abundant tRNAs — this effect increases


the accuracy and/or rate of translation81. Consequently,
synonymous mutations can be opposed by purifying selec-
s >0 10–4
10–5 tion81. In mammals, evidence for such selection is limited4,
100 µ
but recent evidence from mammals and Drosophila spp.
10–6 shows that selection frequently acts on synonymous muta-
tions that affect exonic splice enhancers82 and that this alters
10 µ 10–7 codon usage83,84. The effect of splicing-associated selection
Fixation rate

is shown by the association of synonymous mutations that


10–8
alter splicing with human disease4. Many other mecha-
nisms by which synonymous mutations affect fitness have
µ s=0
also been identified: they affect mRNA stability 85, miRNA
–10–8 binding 86, protein folding 87 and nucleosome formation88.
The hypothesis that synonymous mutations must be neu-
µ
–10–7 tral therefore lacks an understanding of the complexities
10
of gene expression.
s <0 –10–6
However, explaining how isochores have evolved
requires a more subtle role for selection. Currently, the
µ
100 best explanation for the approximately homogeneous
103 104 105 106 107 108
nucleotide content of these genomic blocks is that iso-
Population size (Ne) chores evolved as a consequence of mutation bias coupled
Figure 2 | effective population size and fixation in the nearly neutral
Nature model.
Reviews The
| Genetics with drift favouring GC to AT mutations. These processes
relationship between effective population size (Ne) and the rate of fixation under are countered in domains of high recombination by biased
the nearly neutral model for differing values of the fitness effect (s). s = 0 for strictly gene conversion that favours GC pairs over AT pairs dur-
neutral alleles; s<0 for deleterious alleles; and s>0 for positively selected alleles. ing the resolution of heteroduplexes89,90. Biased gene
For s = 0, the classic result is derived — the substitution rate is the same as the conversion is similar to meiotic drive because it is a form
mutation rate and is independent of Ne. For deleterious mutations, the important
of selection at the gene level. Such a process can cause
point is that we can divide the plot into the three for any given value of s. First, for
small values of s, there is a domain in which the fixation rate is indistinguishable from genome-wide effects with no substitutional cost, as no
the rate that neutral mutations are fixed. This is the zone of effective neutrality. death of individuals occurs in the process, just ‘death’ (con-
Second, there is a zone around s = 1/2Ne in which fixation is possible but the rate is version) of alleles. Although it is difficult to distinguish
lower than the neutral rate. Third, there is a zone at s>1/2Ne in which fixation is rare biased gene conversion from selection on GC content,
(strongly deleterious alleles). These deleterious alleles can persist in a population the effects of biased gene conversion have only recently
owing to mutation–selection equilibrium, but are unlikely to become fixed variations. begun to be appreciated. Biased gene conversion may even
Figure is modified, with permission, from ref. 151  (1986) W.H. Freeman and Co. give false positive signals of positive selection in protein
coding sequences. This is because non-synonymous
sites tend to be less GC-rich than synonymous sites,
More generally, the history of artificial selection exper- so when biased gene conversion acts it predominantly
iments, practiced by plant and animal breeders, has been accelerates rates of evolution at non-synonymous sites,
important for understanding the relationship between producing a KA/KS ratio (BOX 3) that is greater than one91.
heritable variation and selection. These experiments allow However, genome evolution studies also suggest that,
heritability to be estimated and have driven developments broadly, the nearly neutral theory has great explana-
in quantitative genetics that have been combined with the tory power. lynch and colleagues considered whether
statistical methods currently used to estimate parameters intron size and, more generally, genome size increase as
of selection in the wild80. In general, heritability studies Ne decreases92,93, and found that the predictions of the
have shown that most traits are selectable most of the nearly neutral hypothesis fit the evidence. According
time, so the level of heritable variation does not impose to the nearly neutral model, species with small popula-
Exonic splice enhancer limits on selection. tions (for example, humans) are unable to eliminate
A short (approximately 6 insertions that are mildly deleterious; however, the same
nucleotide) motif that Selection and the genome. The population genetics the- insertion in yeast is easily opposed by purifying selection
is present in exons. It is
necessary for the binding of
ory has shown that there are limitations to the role of (s is <1/2Ne for humans and >1/2Ne for yeast). This model
serine–arginine proteins selection; however, studies of gene and genome evolution suggests that yeast and Escherichia coli have better adapted
that are involved in defining have also shown that selection can occur in unexpected genomes than humans. The same model correctly predicts
exon ends. situations. Genomic research has therefore allowed fur- little variation in 5′ UTR length between species94 but also
ther evaluation of the relative merits of selectionist and predicts greater 5′ UTR length variation as a function of
Isochore
A chromosomal block of nearly neutral models. local GC content within species, which is not observed95.
approximately uniform GC one aspect of genomic research that suggests selection For gene-order evolution in yeast, a simple strictly neutral
nucleotide content. This is may be more widespread than previously thought is the null model96,97 correctly predicts that whether two genes
shown, for example, by a study of synonymous mutations. It has commonly been stay physically linked depends on the size of the intergenic
correlation in nucleotide
content between genes
supposed that synonymous mutations are effectively neu- spacer between them. Nonetheless, there are examples of
when comparing introns tral46, especially in humans (in which Ne is low). However, genes that are grouped2 either as co-expression clusters97,98
and synonymous sites. in many species, highly expressed genes use codons that or as clusters of essential genes99.

90 | FeBRUARy 2009 | volUMe 10 www.nature.com/reviews/genetics

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

How important is selection to gene and genome evolu- Perhaps the most profound conceptual issues have
tion? The examples discussed here show that the answer come from recent genomic discoveries. Genomic
depends on whether the cases in which the assumption features either suggest that genomes are not greatly
that selection must be irrelevant is wrong (as with selection shaped by selection or, as in the case of selection on
on synonymous mutations) are highlighted, or the cases in synonymous mutations, they have shown that there are
which the nearly neutral model is valid (as with genome underexplored complexities of how genomes work. A
size or intron size) are highlighted. However, the exam- number of questions remain unanswered — for exam-
ples of isochores and gene order also show that there are ple: are we still missing the regulatory importance of
circumstances in which neither a basic model of selection some transcribed but untranslated sequences; what
nor the nearly neutral model provide the full picture. proportion of alternative splicing is functional; and is
transcriptional noise a failure of selection or an adapta-
Questions and directions tion to environmental variability? These questions can
This discussion has shown how an understanding all be approached by both comparative and experimen-
of genetics transformed the understanding of selection: tal studies, and the answers have implications for our
genetics rescued selection as a concept, then explained understanding of genomic evolution. For example, if
how seemingly counterintuitive observations make sense many alternative transcripts and non-coding transcripts
and provided a rigorous framework to understand selec- are transcriptional noise, they need not be evolution-
tion and alternative evolutionary models. These events led arily conserved or have repeatable expression profiles,
to the current age in which the development of statistical and exons that are unique to unnecessary isoforms
methods and the availability of new data sets provide an should evolve free from purifying selection. If levels of
improving appreciation of the way in which genes and transcriptional noise are adaptive, engineered strains
genomes evolve. of yeast in which genes that are subject to high levels of
This Review has largely concentrated on the history transcriptional noise are made less noisy (or vice versa)
of conceptual advances. However, the current genomic should be at a growth disadvantage. Again, integration
age is likely to lead to a discovery mode of research, and between molecular biology and molecular evolution is
the development of new genome scans100,101 and other expected to lead the way.
tools will be important. For example, advances in the Why is it important to understand selection? There are
analysis of selective sweeps using SNP data will help to some important practical reasons: what we call purifying
locate genomic domains under selection. However, the selection in the formal literature is disease when the muta-
conservative and error-prone nature of many current tion affects humans. Hence, looking for genomic regions
high-throughput statistical approaches and the difficul- that are under purifying selection and understanding
ties of dealing with demography cannot be understated. the mutations that are subject to purifying selection,
Indeed, tests for selection on synonymous mutations that even beyond those that alter proteins, are important for
begin with a mechanism (for example, binding of splic- identifying disease-causing mutations4 and understand-
ing factors to exonic splice enhancer motifs) have been ing mechanisms of pathogenesis. In addition, identifying
successful in resolving the action of selection; however, sites of positive selection in the genome can help us to
statistical tests that are not based on a mechanism (for understand our interactions with pathogens and how they
example, comparing rates of synonymous site and intronic evade our defences102.
evolution) have proved unhelpful and unconvincing4. The Understanding selection can help us to not only
proliferation of genomic scans based on statistical tests understand disease but also disease prevention or cures.
irrespective of mechanism is probably inevitable, and For example, knowing whether gene order and synony-
these scans can suggest candidate regions of selection for mous sites are under selection is also important to opti-
further scrutiny. However, a fuller understanding of how mize the benefits and minimize the risks of altering genes
genes, proteins and genomes work will be needed to more and genomes, whether for genetically modified crops,
precisely resolve what selection is doing and the under- transgenic models of human disease or gene therapy. The
lying mechanisms, as well as to eliminate the possibility hypothesis that synonymous sites matter has already led
that any signal of selection is not an artefact. Achieving a to new strategies to develop attenuated viruses5. Current
thorough and convincing demonstration of the activity of results also suggest a need for caution: if neighbouring
selection will require ever closer integration of molecular genes can affect each others’ activity and be preserved as a
biology and molecular evolution. pair over a long period of evolutionary time, is it likely that
In addition to these methodological hurdles, many transgenes will affect the expression of genes near their
conceptual questions remain unanswered — for exam- insertion site? Understanding selection might also help
ple: why does the recombination rate positively corre- the introduction of genes into a population: the prospect
late with diversity; and if recombination is mutagenic, of spreading, by hitchhiking, genes that can block trans-
what is the mechanism responsible and does it occur in mission of insect-borne pathogens103, such as Wolbachia,
all species? The relative importance of drift and selec- using the strong selection of selfish elements has
tion also remains unclear: for example, why do species potentially far-reaching consequences.
with a large Ne not have high rates of polymorphism and More generally, where and how selection operates
what proportion of all substitutions is the result of selec- tells us how genes, genomes and organisms work. Indeed,
tion? Developing robust methods to address these issues although evolutionary genetics was once effectively dis-
remains challenging. tinct from molecular genetics, it is now hard to separate

NATURe RevIeWS | Genetics volUMe 10 | FeBRUARy 2009 | 91

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

the two. For example, multisequence alignments, looking geneticists. Beyond the important use of selection as a
for conserved domains and constructing the duplication tool, it is also interesting that genes associated with lan-
history of a gene by phylogenetic analysis are fast becom- guage may be under positive selection104. Such studies
ing standard parts of the analytical tool kit of molecular help us understand what it means to be human.

1. Dennett, D. C. Darwin’s Dangerous Idea (Simon and 25. Hurst, L. D. The incidences, mechanisms and evolution 54. Gillespie, J. H. The Causes of Molecular Evolution
Schuster, New York, 1995). of cytoplasmic sex ratio distorters in animals. Biol. (eds May, R. M. & Harvey, P. H.) (Oxford Univ. Press,
2. Hurst, L. D., Pal, C. & Lercher, M. J. The evolutionary Rev. 68, 121–193 (1993). Oxford, 1991).
dynamics of eukaryotic gene order. Nature Rev. Genet. 26. Hurst, G. D. D., Hurst, L. D. & Majerus, M. E. N. A watershed book that challenged the neutral and
5, 299–310 (2004). in Influential Passengers (eds Hoffmann, A., O’Neill, S. nearly neutral models, arguing that selection is a
3. Gierman, H. J. et al. Domain-wide regulation of gene & Werren, J.) 125–154 (Oxford Univ. Press, Oxford better explanation for polymorphism and patterns
expression in the human genome. Genome Res. 17, 1997). of molecular evolution than drift.
1286–1295 (2007). 27. Hurst, G. D. D. & Majerus, M. E. N. Why do maternally 55. Williams, G. C. Natural Selection: Domains, Levels
4. Chamary, J. V., Parmley, J. L. & Hurst, L. D. Hearing inherited microorganisms kill males? Heredity 71, and Challenges (eds May, R. M. & Harvey, P. H.)
silence: non-neutral evolution at synonymous sites in 81–95 (1993). (Oxford Univ. Press, New York, 1992).
mammals. Nature Rev. Genet. 7, 98–108 (2006). 28. Lewis, D. Male sterility in natural populations of 56. Wallace, B. Hard and soft selection revisited. Evolution
A review that highlights the multiple different hermaphrodite plants. New Phytol. 40, 158–160 29, 465–473 (1975).
mechanisms by which selection can act on (1941). 57. Wallace, B. Genetic Load: its Biological and
synonymous mutations. It also argues that tests 29. Werren, J. H. Biology of Wolbachia. Annu. Rev. Conceptual Aspects (Prentice Hall, New Jersey,
based on mechanism rather than genome-scale Entomol. 42, 587–609 (1997). 1970).
statistical analyses provide the most convincing 30. Fine, P. E. M. On the dynamics of symbiote-dependent 58. Andolfatto, P. Adaptive evolution of non-coding DNA
data. cytoplasmic incompatibility in Culicine mosquitoes. in Drosophila. Nature 437, 1149–1152 (2005).
5. Coleman, J. R. et al. Virus attenuation by J. Invertebr. Pathol. 30, 10–18 (1978). 59. Ohta, T. Slightly deleterious mutant substitutions in
genome-scale changes in codon pair bias. Science 31. Turelli, M. & Hoffmann, A. A. Rapid spread of an evolution. Nature 246, 96–98 (1973).
320, 1784–1787 (2008). inherited incompatibility factor in California 60. Keightley, P. D. & Eyre-Walker, A. Deleterious
6. Bulmer, M. Did Jenkin’s swamping argument Drosophila. Nature 353, 440–442 (1991). mutations and the evolution of sex. Science 290,
invalidate Darwin’s theory of natural selection? 32. Werren, J. H., Windsor, D. & Guo, L. R. Distribution 331–333 (2000).
Brit. J. Hist. Sci. 37, 281–297 (2004). of Wolbachia among neotropical arthropods. 61. Ayala, F. J. Molecular clock mirages. Bioessays 21,
7. Fisher, R. A. The correlation between relatives on the Proc. R. Soc. Lond. B 262, 197–204 (1995). 71–75 (1999).
supposition of Mendelian inheritance. Trans. R. Soc. 33. Hilgenboecker, K., Hammerstein, P., Schlattmann, P., 62. Ohta, T. & Gillespie, J. H. Development of neutral
Edinb. 52, 399–433 (1918). Telschow, A. & Werren, J. H. How many species are and nearly neutral theories. Theor. Popul. Biol. 49,
8. Hammerstein, P. Darwinian adaptation, infected with Wolbachia? — a statistical analysis of 128–142 (1996).
population-genetics and the streetcar theory of current data. FEMS Microbiol. Lett. 281, 215–220 To commemorate the death of Kimura, the chief
evolution. J. Math. Biol. 34, 511–532 (1996). (2008). defender and the harshest critic of the nearly
An important paper that provides a population- 34. Price, T. A., Hodgson, D. J., Lewis, Z., Hurst, G. D. & neutral theory unite to provide a clearly argued
genetics defence for the application of non-genetics Wedell, N. Selfish genetic elements promote polyandry critique of the history and current standing of the
models, such as evolutionary stable strategy in a fly. Science 322, 1241–1243 (2008). neutral and nearly neutral theories.
theory. 35. Lyon, M. F. Male sterility of the mouse t-complex is 63. Rodriguez-Trelles, F., Tarrio, R. & Ayala, F. J.
9. Hadfield, J. D., Nutall, A., Osorio, D. & Owens, I. P. due to homozygosity of the distorter genes. Cell 44, Erratic overdispersion of three molecular clocks:
Testing the phenotypic gambit: phenotypic, genetic 357–363 (1986). GPDH, SOD, and XDH. Proc. Natl Acad. Sci. USA 98,
and environmental correlations of colour. J. Evol. Biol. 36. Hamilton, W. D. The genetical evolution of social 11405–11410 (2001).
20, 549–557 (2007). behaviour I. J. Theor. Biol. 7, 1–16 (1964). 64. Kim, S. H. & Yi, S. V. Mammalian nonsynonymous
10. Maynard Smith, J. & Haigh, J. The hitch-hiking effect The seminal paper that introduced the concept of sites are not overdispersed: comparative genomic
of a favorable gene. Genet. Res. 23, 23–35 (1974). inclusive fitness. analysis of index of dispersion of mammalian proteins.
The seminal paper on hitchhiking. 37. Griffin, A. S., West, S. A. & Buckling, A. Cooperation Mol. Biol. Evol. 25, 634–642 (2008).
11. Harr, B., Kauer, M. & Schlotterer, C. Hitchhiking and competition in pathogenic bacteria. Nature 430, 65. Fay, J. C., Wyckoff, G. J. & Wu, C. I. Positive and
mapping: a population-based fine-mapping strategy 1024–1027 (2004). negative selection on the human genome. Genetics
for adaptive mutations in Drosophila melanogaster. 38. Fisher, R. A. The Genetical Theory of Natural Selection 158, 1227–1234 (2001).
Proc. Natl Acad. Sci. USA 99, 12949–12954 (2002). (Clarendon, Oxford, 1930). 66. Smith, N. G. C. & Eyre-Walker, A. Adaptive protein
12. Nielsen, R., Hellmann, I., Hubisz, M., Bustamante, C. & 39. Hamilton, W. D. Extraordinary sex ratios. Science evolution in Drosophila. Nature 415, 1022–1024
Clark, A. G. Recent and ongoing selection in the human 156, 477–488 (1967). (2002).
genome. Nature Rev. Genet. 8, 857–868 (2007). 40. Bull, J. J. & Charnov, E. L. How fundamental are 67. Fay, J. C., Wyckoff, G. J. & Wu, C. I. Testing the neutral
13. Wang, X., Grus, W. E. & Zhang, J. Gene losses during Fisherian sex ratios? Oxf. Surv. Evol. Biol. 5, 96–135 theory of molecular evolution with genomic data from
human origins. PLoS Biol. 4, e52 (2006). (1988). Drosophila. Nature 415, 1024–1026 (2002).
14. Sniegowski, P. D., Gerrish, P. J., Johnson, T. & 41. Sundstrom, L., Chapuisat, M. & Keller, L. Conditional 68. Bustamante, C. D. et al. Natural selection on
Shaver, A. The evolution of mutation rates: manipulation of sex ratios by ant workers: a test of kin protein-coding genes in the human genome. Nature
separating causes from consequences. Bioessays 22, selection theory. Science 274, 993–995 (1996). 437, 1153–1157 (2005).
1057–1066 (2000). 42. Marrow, P., Johnstone, R. A. & Hurst, L. D. Riding the 69. Bierne, N. & Eyre-Walker, A. The genomic rate of
15. Charlesworth, B. The effect of background selection evolutionary streetcar — where population-genetics adaptive amino acid substitution in Drosophila.
against deleterious mutations on weakly selected, and game-theory meet. Trends Ecol. Evol. 11, Mol. Biol. Evol. 21, 1350–1360 (2004).
linked variants. Genet. Res. 63, 213–227 (1994). 445–446 (1996). 70. Sawyer, S. A., Parsch, J., Zhang, Z. & Hartl, D. L.
16. Nachman, M. W. Single nucleotide polymorphisms 43. Gardner, A., West, S. A. & Barton, N. H. The relation Prevalence of positive selection among nearly neutral
and recombination rate in humans. Trends Genet. 17, between multilocus population genetics and social amino acid replacements in Drosophila. Proc. Natl
481–485 (2001). evolution theory. Am. Nat. 169, 207–226 (2007). Acad. Sci. USA 104, 6504–6510 (2007).
17. Begun, D. J. & Aquadro, C. F. Levels of naturally- 44. Kimura, M. Preponderance of synonymous changes as 71. Shapiro, J. A. et al. Adaptive genic evolution in the
occurring DNA polymorphism correlate with evidence for the neutral theory of molecular evolution. Drosophila genomes. Proc. Natl Acad. Sci. USA 104,
recombination rates in Drosophila melanogaster. Nature 267, 275–276 (1977). 2271–2276 (2007).
Nature 356, 519–520 (1992). 45. Kimura, M. The Neutral Theory of Evolution 72. Williamson, S. H. et al. Simultaneous inference of
18. Andolfatto, P. Adaptive hitchhiking effects on genome (Cambridge Univ. Press, Cambridge, 1983). selection and population growth from patterns of
variability. Curr. Opin. Genet. Dev. 11, 635–641 (2001). 46. King, J. L. & Jukes, T. H. Non-Darwinian evolution. variation in the human genome. Proc. Natl Acad. Sci.
19. Perry, J. & Ashworth, A. Evolutionary rate of a gene Science 164, 788–798 (1969). USA 102, 7882–7887 (2005).
affected by chromosomal position. Curr. Biol. 9, 47. Haldane, J. B. S. The cost of natural selection. 73. Charlesworth, J. & Eyre-Walker, A. The McDonald–
987–989 (1999). J. Genet. 55, 511–524 (1957). Kreitman test and slightly deleterious mutations.
20. Lercher, M. J. & Hurst, L. D. Human SNP variability 48. Kimura, M. Evolutionary rate at the molecular level. Mol. Biol. Evol. 25, 1007–1015 (2008).
and mutation rate are higher in regions of high Nature 217, 624–626 (1968). 74. Chipman, A. D., Arthur, W. & Akam, M. A double
recombination. Trends Genet. 18, 337–340 (2002). 49. Ohta, T. The nearly neutral theory of molecular evolution. segment periodicity underlies segment generation in
21. Hellmann, I., Ebersberger, I., Ptak, S. E., Paabo, S. & Annu. Rev. Ecol. System. 23, 263–286 (1992). centipede development. Curr. Biol. 14, 1250–1255
Przeworski, M. A neutral explanation for the 50. Gillespie, J. H. Genetic drift in an infinite population: (2004).
correlation of diversity with recombination rates in the pseudohitchhiking model. Genetics 155, 75. Galis, F. & Metz, J. A. Anti-cancer selection as a source
humans. Am. J. Hum. Genet. 72, 1527–1535 (2003). 909–919 (2000). of developmental and evolutionary constraints.
22. Karlin, S. & McGregor, J. Towards a theory of the 51. Gillespie, J. H. Is the population size of a species Bioessays 25, 1035–1039 (2003).
evolution of modifier genes. Theor. Popul. Biol. 5, relevant to its evolution? Evolution 55, 2161–2169 76. Toro, M. A. & Charlesworth, B. An attempt to detect
59–103 (1974). (2001). genetic variation in sex ratio of Drosophila
23. Hurst, L. D., Atlan, A. & Bengtsson, B. O. Genetic 52. Lewontin, R. C. The Genetic Basis of Evolutionary melanogaster. Heredity 49, 199–209 (1982).
conflicts. Q. Rev. Biol. 71, 317–364 (1996). Change (Columbia Univ. Press, New York, 1974). 77. Hurst, L. D. & Vollrath, F. Sex ratio adjustment in
24. Lyttle, T. W. Segregation distorters. Annu. Rev. Genet. 53. Kreitman, M. The neutral theory is dead — long live solitary and social spiders. Trends Ecol. Evol. 7,
25, 511–557 (1991). the neutral theory. Bioessays 18, 678–683 (1996). 326–327 (1992).

92 | FeBRUARy 2009 | volUMe 10 www.nature.com/reviews/genetics

© 2009 Macmillan Publishers Limited. All rights reserved


REVIEWS

78. Brakefield, P. M. et al. Development, plasticity and 103. Sinkins, S. P. & Gould, F. Gene drive systems for A provocative paper showing that within genes,
evolution of butterfly eyespot patterns. Nature 384, insect disease vectors. Nature Rev. Genet. 7, more KA/KS>1 peaks are present, owing to
236–242 (1996). 427–435 (2006). unusually low synonymous rates rather than
An important empirical evaluation of the relevance 104. Enard, W. et al. Molecular evolution of FOXP2, positive selection on proteins. Repeatable peaks
of developmental constraint that suggests it is a gene involved in speech and language. Nature 418, tend to be in alternative exons.
largely irrelevant. 869–872 (2002). 129. Yang, Z. PAML: a program package for phylogenetic
79. Beldade, P., Koops, K. & Brakefield, P. M. Developmental 105. Wright, S. Evolution and the Genetics of Populations: analysis by maximum likelihood. CABIOS 13,
constraints versus flexibility in morphological evolution. The Theory of Gene Frequencies (Chicago Univ. Press, 555–556 (1997).
Nature 416, 844–847 (2002). Chicago, 1969). 130. McDonald, J. H. & Kreitman, M. Adaptive protein
80. Kruuk, L. E. B., Slate, J. & Wilson, A. J. New answers 106. Majerus, M. E. & Mundy, N. I. Mammalian melanism: evolution at the Adh locus in Drosophila. Nature 351,
for old questions: the evolutionary quantitative natural selection in black and white. Trends Genet. 19, 652–654 (1991).
genetics of wild animal populations. Annu. Rev. Ecol. 585–588 (2003). The paper that introduced the McDonald–Kreitman
System. 39, 525–548 (2008). 107. Coltman, D. W. et al. Undesirable evolutionary test.
81. Duret, L. Evolution of synonymous codon usage in consequences of trophy hunting. Nature 426, 131. Hudson, R. R., Kreitman, M. & Aguade, M. A test of
metazoans. Curr. Opin. Genet. Dev. 12, 640–649 (2002). 655–658 (2003). neutral molecular evolution based on nucleotide data.
82. Parmley, J. L., Chamary, J. V. & Hurst, L. D. 108. Barot, S., Heino, M., O’Brien, L. & Dieckmann, U. Genetics 116, 153–159 (1987).
Evidence for purifying selection against synonymous Long-term trend in the maturation reaction norm of 132. Comeron, J. M., Williford, A. & Kliman, R. M. The
mutations in mammalian exonic splicing enhancers. two cod stocks. Ecol. Appl. 14, 1257–1271 (2004). Hill–Robertson effect: evolutionary consequences of
Mol. Biol. Evol. 23, 301–309 (2006). 109. Yang, Z. H. & Bielawski, J. P. Statistical methods for weak selection and linkage in finite populations.
83. Parmley, J. L. & Hurst, L. D. Exonic splicing regulatory detecting molecular adaptation. Trends Ecol. Evol. 15, Heredity 100, 19–31 (2008).
elements skew synonymous codon usage near intron– 496–503 (2000). A recent review on Hill–Robertson interference.
exon boundaries in mammals. Mol. Biol. Evol. 24, 110. Zhang, J., Zhang, Y. P. & Rosenberg, H. F. Adaptive 133. Comeron, J. M. & Kreitman, M. Population,
1600–1603 (2007). evolution of a duplicated pancreatic ribonuclease gene evolutionary and genomic consequences of interference
84. Warnecke, T. & Hurst, L. D. Evidence for a trade-off in a leaf-eating monkey. Nature Genet. 30, 411–415 selection. Genetics 161, 389–410 (2002).
between translational efficiency and splicing regulation (2002). 134. Pal, C., Papp, B. & Hurst, L. D. Does the
in determining synonymous codon usage in Drosophila An example of positive selection that uses both recombination rate affect the efficiency of purifying
melanogaster. Mol. Biol. Evol. 24, 2755–2762 (2007). statistical and experimental approaches, and shows selection? The yeast genome provides a partial
85. Nackley, A. G. et al. Human how RNASE1B has evolved rapidly under positive answer. Mol. Biol. Evol. 18, 2323–2326 (2001).
catechol-O-methyltransferase haplotypes modulate selection for enhanced ribonucleolytic activity in an 135. Betancourt, A. J. & Presgraves, D. C. Linkage limits
protein expression by altering mRNA secondary altered microenvironment. the power of natural selection in Drosophila. Proc.
structure. Science 314, 1930–1933 (2006). 111. Andersson, M. & Simmons, L. W. Sexual selection Natl Acad. Sci. USA 99, 13616–13620 (2002).
86. Hurst, L. D. Preliminary assessment of the impact of and mate choice. Trends Ecol. Evol. 21, 296–302 136. Keightley, P. D. & Otto, S. P. Interference among
microRNA-mediated regulation on coding sequence (2006). deleterious mutations favours sex and recombination
evolution in mammals. J. Mol. Evol. 63, 174–182 (2006). 112. Blanchette, M. & Tompa, M. Discovery of regulatory in finite populations. Nature 443, 89–92 (2006).
87. Kimchi-Sarfaty, C. et al. A “silent” polymorphism in the elements by a computational method for phylogenetic 137. Lewontin, R. C. & Krakauer, J. Distribution of gene
MDR1 gene changes substrate specificity. Science footprinting. Genome Res. 12, 739–748 (2002). frequency as a test of the theory of the selective
315, 525–528 (2007). 113. Tagle, D. A. et al. Embryonic ε and γ globin genes of a neutrality of polymorphisms. Genetics 74, 175–195
88. Warnecke, T., Batada, N. N. & Hurst, L. D. The impact prosimian primate (Galago crassicaudatus): (1973).
of the nucleosome code on protein-coding sequence nucleotide and amino acid sequences, developmental 138. Vitalis, R., Dawson, K. & Boursot, P. Interpretation of
evolution in yeast. PLoS Genet. 4, e1000250 (2008). regulation and phylogenetic footprints. J. Mol. Biol. variation across marker loci as evidence of selection.
89. Eyre-Walker, A. & Hurst, L. D. The evolution of 203, 439–455 (1988). Genetics 158, 1811–1823 (2001).
isochores. Nature Rev. Genet. 2, 549–555 (2001). 114. Washietl, S., Machne, R. & Goldman, N. Evolutionary 139. Kayser, M., Brauer, S. & Stoneking, M. A genome scan
90. Galtier, N., Piganeau, G., Mouchiroud, D. & Duret, L. footprints of nucleosome positions in yeast. Trends to detect candidate regions influenced by local natural
GC-content evolution in mammalian genomes: the Genet. 24, 583–587 (2008). selection in human populations. Mol. Biol. Evol. 20,
biased gene conversion hypothesis. Genetics 159, 115. Pfennig, D. W., Harcombe, W. R. & Pfennig, K. S. 893–900 (2003).
907–911 (2001). Frequency-dependent Batesian mimicry. Nature 410, 140. Akey, J. M., Zhang, G., Zhang, K., Jin, L. &
91. Berglund, J., Pollard, K. S. & Webster, M. T. Hotspots 323–323 (2001). Shriver, M. D. Interrogating a high-density SNP map
of biased nucleotide substitutions in humans genes. 116. Gemmell, N. J. & Slate, J. Heterozygote advantage for for signatures of natural selection. Genome Res. 12,
PLoS Biol. (in the press). fecundity. PLoS ONE 1, e125 (2006). 1805–1814 (2002).
This paper shows that putative signals of positive 117. Salathe, M., Kouyos, R. D. & Bonhoeffer, S. The state 141. Beaumont, M. A. & Balding, D. J. Identifying adaptive
selection from KA/KS methodology require more of affairs in the kingdom of the Red Queen. Trends genetic divergence among populations from genome
scrutiny, as they can result from processes such as Ecol. Evol. 23, 439–445 (2008). scans. Mol. Ecol. 13, 969–980 (2004).
biased gene conversion. 118. Woolhouse, M. E., Webster, J. P., Domingo, E., 142. Sabeti, P. C. et al. Genome-wide detection and
92. Lynch, M. The origins of eukaryotic gene structure. Charlesworth, B. & Levin, B. R. Biological and characterization of positive selection in human
Mol. Biol. Evol. 23, 450–468 (2006). biomedical implications of the co-evolution of populations. Nature 449, 913–918 (2007).
93. Lynch, M. & Conery, J. S. The origins of genome pathogens and their hosts. Nature Genet. 32, 143. Watterson, G. A. The homozygosity test of neutrality.
complexity. Science 302, 1401–1404 (2003). 569–577 (2002). Genetics 88, 405–417 (1978).
A wide-ranging analysis arguing that weakened 119. Little, T. J. The evolutionary significance of parasitism: 144. Tajima, F. Statistical method for testing the neutral
selection that is associated with reduced do parasite-driven genetic dynamics occur ex silico? mutation hypothesis by DNA polymorphism. Genetics
population size (hence, the nearly neutral theory) J. Evol. Biol. 15, 1–9 (2002). 123, 585–595 (1989).
can explain the accumulation of genomic elements, 120. Webster, J. P., Shrivastava, J., Johnson, P. J. & Blair, L. 145. Fu, Y. X. & Li, W. H. Statistical tests of neutrality of
such as mobile elements and introns. Is host–schistosome coevolution going anywhere? mutations. Genetics 133, 693–709 (1993).
94. Lynch, M., Scofield, D. G. & Hong, X. The evolution of BMC Evol. Biol. 7, 91 (2007). 146. Fu, Y. X. New statistical tests of neutrality for DNA
transcription-initiation sites. Mol. Biol. Evol. 22, 121. Decaestecker, E. et al. Host–parasite ‘Red Queen’ samples from a population. Genetics 143, 557–570
1137–1146 (2005). dynamics archived in pond sediment. Nature 450, (1996).
95. Reuter, M., Engelstadter, J., Fontanillas, P. & Hurst, L. D. 870–873 (2007). 147. Fu, Y. X. Statistical tests of neutrality of mutations
A test of the null model for 5′ UTR evolution based on 122. Forde, S. E. et al. Understanding the limits to against population growth, hitchhiking and background
GC content. Mol. Biol. Evol. 25, 801–804 (2008). generalizability of experimental evolutionary models. selection. Genetics 147, 915–925 (1997).
96. Poyatos, J. F. & Hurst, L. D. The determinants of gene Nature 455, 220–223 (2008). 148. Sabeti, P. C. et al. Detecting recent positive selection
order conservation in yeasts. Genome Biol. 8, R233 123. Bubb, K. L. et al. Scan of human genome reveals no in the human genome from haplotype structure.
(2007). new loci under ancient balancing selection. Genetics Nature 419, 832–837 (2002).
97. Hurst, L. D., Williams, E. J. & Pal, C. Natural selection 173, 2165–2177 (2006). 149. Nielsen, R. Molecular signatures of natural selection.
promotes the conservation of linkage of co-expressed 124. Baysal, B. E., Lawrence, E. C. & Ferrell, R. E. Sequence Annu. Rev. Genet. 39, 197–218 (2005).
genes. Trends Genet. 18, 604–606 (2002). variation in human succinate dehydrogenase genes: An excellent review of tests for selection.
98. Lercher, M. J., Urrutia, A. O. & Hurst, L. D. Clustering evidence for long-term balancing selection on SDHA. 150. Tang, K., Thornton, K. R. & Stoneking, M. A new approach
of housekeeping genes provides a unified model of BMC Biol. 5, 12 (2007). for using genome scans to detect recent positive selection
gene order in the human genome. Nature Genet. 31, 125. Kawabe, A., Fujimoto, R. & Charlesworth, D. in the human genome. PLoS Biol. 5, e171 (2007).
180–183 (2002). High diversity due to balancing selection in the 151. Crow, J. F. Basic Concepts in Population, Quantitative
99. Pal, C. & Hurst, L. D. Evidence for co-evolution of gene promoter region of the medea gene in Arabidopsis and Evolutionary Genetics (W.H. Freeman and Co.,
order and recombination rate. Nature Genet. 33, lyrata. Curr. Biol. 17, 1885–1889 (2007). New York, 1986).
392–395 (2003). 126. Fumagalli, M. et al. Widespread balancing selection
100. Voight, B. F., Kudaravalli, S., Wen, X. & Pritchard, J. K. and pathogen-driven selection at blood group antigen Acknowledgements
A map of recent positive selection in the human genes. Genome Res. 7 Nov 2008 (doi:10.1101/ The author is a Royal Society Wolfson Research Merit Award
genome. PLoS Biol. 4, e72 (2006). gr.082768.108). holder.
101. Williamson, S. H. et al. Localizing recent adaptive 127. Cagliani, R. et al. The signature of long-standing
evolution in the human genome. PLoS Genet. 3, e90 balancing selection at the human defensin β-1
(2007). promoter. Genome Biol. 9, R143 (2008). FuRtHeR inFoRMation
102. Sawyer, S. L., Wu, L. I., Emerman, M. & Malik, H. S. 128. Parmley, J. L. & Hurst, L. D. How common are Laurence Hurst’s webpage:
Positive selection of primate TRIM5α identifies a intragene windows with KA>KS owing to purifying http://www.bath.ac.uk/bio-sci/research/profiles/hurst-l.html
critical species-specific retroviral restriction domain. selection on synonymous mutations? J. Mol. Evol. 64, All links Are Active in tHe online Pdf
Proc. Natl Acad. Sci. USA 102, 2832–2837 (2005). 646–655 (2007).

NATURe RevIeWS | Genetics volUMe 10 | FeBRUARy 2009 | 93

© 2009 Macmillan Publishers Limited. All rights reserved

You might also like