Download The Limits of Natural Selection in a

TIGS 1258 No. of Pages 10
Opinion
The Limits of Natural Selection
in a Nonequilibrium World
Yaniv Brandvain1 and Stephen I. Wright2,*
Evolutionary theory predicts that factors such as a small population size or low
recombination rate can limit the action of natural selection. The emerging field of
comparative population genomics offers an opportunity to evaluate these
hypotheses. However, classical theoretical predictions assume that populations
are at demographic equilibrium. This assumption is likely to be violated in the
very populations researchers use to evaluate selection's limits: populations that
have experienced a recent shift in population size and/or effective recombination rates. Here we highlight theory and data analyses concerning limitations on
the action of natural selection in nonequilibrial populations and argue that
substantial care is needed to appropriately test whether species and populations show meaningful differences in selection efficacy. A move toward modelbased inferences that explicitly incorporate nonequilibrium dynamics provides
a promising approach to more accurately contrast selection efficacy across
populations and interpret its significance.
Characterizing the Influence of Nonequilibrium Demographic History on the
Efficacy of Natural Selection
Despite the relentless action of natural selection, no organism is perfect. Understanding the
factors that limit natural selection is a fundamental concern of population genetics [1]. These
limits can inform the basis of human genetic diseases [2,3], the extinction of small [4,5], asexual
[6], or selfing populations [7–10], the degradation of sex chromosomes [11], the evolution of
gene [12] and genome structure [13–15], and the consequences of domestication [16–19]. For
example, the history of population bottlenecks associated with range expansion might be
responsible for the excess of potentially damaging and, in some cases, disease-causing
[20,21] variants in historically bottlenecked human populations [21–26]. Similarly, an excess
of putatively deleterious mutations in predominantly self-fertilizing plants suggests that purifying
selection is limited in these taxa [27–30].
Because of its wide-ranging implications for micro- and macroevolution, conservation, plant and
animal breeding, and human health, there has been much research on understanding the limits
of selection. Classic theory predicts that smaller effective population sizes (see Glossary),
higher ploidy levels, and lower effective recombination rates, among other factors, can limit the
action of natural selection, driving populations toward lower mean fitness [1]. Population
genomic datasets allow researchers to precisely identify and quantify putatively deleterious
mutations, providing the opportunity for powerful tests of these hypotheses by investigating
whether populations that differ in one or more of these features result in different patterns of
divergence and variation for putatively deleterious mutations (many are reviewed in Table 1).
However, there is growing recognition that inferring and interpreting differences in the efficacy of
selection based on population genomic data can be more difficult than initially thought. Here we
indicate the challenges in connecting differences in mutational patterns between taxa to differences in the action of natural selection (see Box 1 for a discussion of how researchers attempt to
Trends in Genetics, Month Year, Vol. xx, No. yy
Trends
Next-generation DNA and RNA
sequencing datasets are enabling
powerful genomic tests of the factors
influencing the efficacy of natural selection against deleterious mutations.
Theoretical advances have provided
important insights into the extent to
which demographic history influences
selection strengths and genetic load.
There is growing recognition that
nonequilibrium demographic history
can lead to spurious signals of
increased genetic load.
Model-based methods investigating
the efficacy of selection, incorporating
demographic history and uncertainty in
the dominance of deleterious mutations, are being developed to provide
further insights into the factors limiting
selection in natural populations.
1
Department of Plant Biology,
University of Minnesota, St[1_TD$IF] Paul, MN
55108, USA
2
Department of Ecology and
Evolutionary Biology, University of
Toronto, Toronto, ON M5S 3B2,
Canada
*Correspondence:
[email protected]
(S.I. Wright).
http://dx.doi.org/10.1016/j.tig.2016.01.004
© 2016 Elsevier Ltd. All rights reserved.
1
TIGS 1258 No. of Pages 10
measure selection's limits). In particular, we highlight the complications that arise in attempts to
infer limited efficacy of selection in nonequilibrial populations.
There is growing recognition, particularly in the human population genomics literature, that
nonequilibrium demography can complicate empirical tests for differences in the efficacy of
selection from polymorphism data. This complication has led to considerable debate on the
interpretation of genomic patterns [24,31–33]. For example, following a strong population
bottleneck, sites under purifying selection are expected to recover equilibrium faster than neutral
sites, which can lead to spurious signals of reduced efficacy of selection in bottlenecked
populations [31,34,35]. Therefore, finding strong evidence for limited selection in such cases
is difficult, particularly because theory predicts that population bottlenecks and expansions can
have subtle and complex influences on selection. Although the confounding effects of nonequilibrium demography on our inference of relaxed selection is well appreciated in the human
literature, we argue that such consideration must be extended to studies of other populations
that have experienced a recent bottleneck, including domesticated species, those that have
experienced a shift in mating system, and other candidates for relaxed selection.
We argue that the field has yet to fully address these challenges, meaning that many examples
purporting to provide evidence for selection limits remain ambiguous. Therefore, clarifying what
is robust, genome-wide evidence for differences in selection, and improved understanding of
between-species differences in fitness, will be a major task in the coming years. Below we outline
both the promise and the challenges of using genomic data to test theory about the limits of
natural selection and to investigate what forces are most important in imposing limits, especially
as populations grow, shrink, and spread in our changing world.
Limits to Selection
Theoretical Predictions of the Limits of Natural Selection
The number and frequency of damaging mutations is determined by their introduction into a
population by mutation, their stochastic change by drift, and their removal by selection. For
strongly deleterious mutations in modestly sized populations, selection, mutation, and the mode
of inheritance can describe the expected number and frequency of deleterious mutations, as well
as the mutation load. For large-effect mutations, the predicted genetic load is simply a function
of the mutation rate [36]; however, additional biological reality complicates this prediction [37].
For mutations with more subtle influences on fitness, stochasticity can overwhelm the action of
selection as the number of independent chromosomes in a population decreases. Therefore,
deleterious mutations are predicted to be more common in smaller populations (stronger effects of
drift) and/or populations or genomic regions with lower recombination rates (less independence of
selection on these chromosomes). Another constraint on selection's action is its ability to ‘see’
mutations. In diploids, deleterious variation can be hidden in the recessive state. In species or
genomic regions with higher ploidy, a similar effect can occur and therefore selection may be unable
to eliminate deleterious loss-of-function mutations that are hidden by functional gene copies.
Together, population size, recombination rates, mutation rates, ploidy, and the mode of gene action
interact to set the limits of selection, which we can hope to uncover from genomic data (Table 1).
In this brief treatment, we neglect constraints in organismal design and the mutational landscape
that can also impose limits on selection [1]. Uncovering these limits from genomic data presents
an additional set of challenges orthogonal to the points below.
Estimating Demographic History and Recombination Rates
The basic predictors of the limits of natural selection – recombination rates, mutation rates, and
historical population sizes – can, in principle, be precisely estimated from modern genomic data.
2
Trends in Genetics, Month Year, Vol. xx, No. yy
Glossary
p: the expected number of sequence
differences between two
chromosomes.
pn: the expected number of
sequence differences between two
chromosomes at nonsynonymous
sites within a species or population.
ps: the expected number of
sequence differences between two
chromosomes at synonymous sites
within a species or population.
Distribution of fitness effects
(DFE): the distribution of the
proportion of new mutations in each
class of selection coefficient.
dN: the expected number of
sequence differences between two
chromosomes at nonsynonymous
sites between species.
dS: the expected number of
sequence differences between two
chromosomes at synonymous sites
between species
Effective population size (Ne): the
effective number of breeding
individuals.
Gene surfing: the spread of new
and standing mutations that are on
the wave front of a range expansion;
surfing by neutral or deleterious
mutations mimics positive selection
because they spread and fix over
large areas.
Genetic load: the difference
between the average population
fitness and the fitness of a mutationfree genotype.
TIGS 1258 No. of Pages 10
Table 1. Selected Examples of Genome-Wide Tests of Selection's Limits
Factor
Theoretical
Prediction
Comparison
Metric of Relaxed Selection
Refs
Significant
Effect?
Effective
population
size
Small population
sizes increase
the effects of drift
relative to
selection [68]
Domesticated vs
wild tomato
dN/dS
[69]
+
Domesticated vs
wild sunflowers
pn/ps
[17]
+
Domesticated vs
wild rice
[7_TD$IF]dN/[8_TD$IF]dS
dN/[9_TD$IF]dS
[70]
[16]
+
+
Domesticated vs
wild horses
# Deleterious mutations/ind
[19]
+
Arabidopsis thaliana
populations of
different sizes
pn/ps
[71]
+
Yeast species of
different sizes
Various
[72]
+
Human populations
experiencing
historical differences
in population size
pn/ps (+ related summaries)
pn/ps (+ related summaries)
# Deleterious mutations/ind
# Deleterious mutations/ind
# Deleterious mutations/ind
Predicted individual fitness
[25]
[22]
[31]
[33]
[52]
[51,52]
+
+
-
Low- vs highrecombination
genomic regions
pn/ps, GERP scores
dN/dS
DFE
pn/ps, dN/dS, DFE
[74]
[75]
[76]
[77]
+
Selfing vs
outcrossing species
and populations
pn/ps, dN/dS, DFE
# Stop codons, pn/ps
dN/dS
pn/ps
dN/dS
pn/ps
DFE
DFE
[78]
[28]
[10_TD$IF][28]
[79]
[79]
[30]
[27]
[80]
+
+
dN/dS, pn/ps
pn/ps
[29]
[81]
+
Recombination
rates
Reduced
recombination
will cause higher
selective
interference,
weakening
selection [73]
Asexual vs sexual
populations
+
+
+
+
+
+
+
+
Tests of whether selection is weaker in populations with reduced effective population size or effective recombination rates.
Response variables include the ratio of nonsynonymous to synonymous substitutions (dN/dS), the ratio of nonsynonymous
to synonymous polymorphism pn/ps, and the inferred DFE.
For example, one diploid genome can provide a sketch of an entire population's history [38] and
many sequences allow a fine-grained view of recent population sizes and migration rates
[39–41]. Dense genotype data from many individuals provides precise characterization of
the meiotic recombination landscape and whole-genome sequencing of parent–offspring trios
can reveal the rate of spontaneous mutation [42,43], providing access to additional predictors of
selection's limits. With such information, researchers can explore the relationship between the
action of selection and population parameters of interest [31–33,44].
Quantifying the Limits to Selection
A key measure of interest in evaluating the efficacy of selection is the genetic load – the difference
between achieved fitness and maximum potential fitness. In theory one could test whether two
populations differ in their load by comparing the fitness of individuals from these populations, and
although the genetic load is not an exact measure of the historical action of selection, the efficacy
of selection was likely to be more limited in the population with the larger load. In practice,
however, comparing fitness of individuals adapted to different biotic and abiotic challenges is often
Trends in Genetics, Month Year, Vol. xx, No. yy
3
TIGS 1258 No. of Pages 10
Box 1. How Can We Quantify and Compare the Efficacy of Selection?
Various metrics are applied to compare the efficacy of negative selection, but such comparisons have a number of
challenges in their power and interpretation.
The ratio of nonsynonymous to synonymous substitutions between species (dN/dS) is a common (and crude) measure of
the strength of selection. Elevated dN/dS values are often interpreted as reflecting reduced efficacy of negative selection,
but elevated dN/dS can also indicate elevated positive selection [48,82]. In addition, because new mutations accumulate
slowly, we expect only modest differences in the number of deleterious mutations between recently diverged populations. Because most comparative population genomic studies focus on pairs of closely related populations (Table 1)
between-species contrasts are often underpowered, limiting our ability to contrast contemporary differences in selection.
Therefore, comparisons of intraspecific diversity at selected and neutral sites (e.g., pn/ps) fell into favor as a way to
overcome this limitation. However, nonequilibrium demography can mislead this measure (Box 3).
The use of evolutionary models to infer the DFE of deleterious mutations [62,63] is thus becoming increasingly
widespread (Table 1). By combining information on the reduction in nonsynonymous diversity with the frequency
distribution of nonsynonymous and synonymous mutations, these methods explicitly quantify the strength of selection
against deleterious mutations rather than simply reporting a ratio of diversity levels at functional and neutral sites.
However, cross-species comparisons of the DFE do not always generate straightforward interpretations of differences in
the efficacy of selection. For example, one species can show increased proportions of both effectively neutral and
strongly deleterious mutations compared with another [27].
Additionally, genomic data allow improved assessment of which mutations are likely to have deleterious effects [57].
Researchers can use algorithms (e.g., SIFT [59], MAPP [60]) that combine cross-species sequence conservation and
functional predictions. Since tests of the limits of selection are focused on mutations with fitness effects, an improved
search image for damaging mutations may increase the power of such tests. However, these methods inherently rely on
sites subject to purifying selection over long macroevolutionary history and weakly selected sites that are more likely to be
subject to shifts in the efficacy of selection could remain undetected, and beneficial changes in function can be confused
for deleterious mutations.
unfeasible. As we describe below and in Box 1, annotated population genomic data may provide
an alternative glimpse into the historical action of natural selection, and genome-informed inferences can, in principle, provide information about the genetic load in populations [37].
Such data allow precise estimation of the number and frequency of putatively deleterious
mutations genome wide, potentially facilitating strong hypothesis testing regarding what factors
influence the number and frequency of deleterious variants. The genome contains enough sites
variable within or divergent among populations to investigate whether individuals from a given
population carry more deleterious mutations than individuals in a different population. To test
whether selection's limits differ between populations, species, or genomic regions, researchers
often compare the number or frequency of putatively deleterious variants, perhaps weighted by
their inferred effect on fitness, relative to putatively neutral ones and/or fit a model of demography
and selection to such data (Box 1). The challenge, which we discuss in more detail below, is
relating observed mutational patterns to the goal of quantifying and comparing historical
selection. As we highlight in Box 1, this challenge is not trivial: there is currently no single
agreed on measure of the efficacy of selection.
Evidence of Selection's Limits
Recent analyses of whole-genome data have attempted to test hypotheses of selection's limits
(Table 1). Such tests take species, populations, or genomic regions that differ in some biological
feature such as effective recombination rate, effective population size, and/or ploidy and ask
whether there is a corresponding difference in the action of purifying selection as inferred from
patterns of polymorphism and/or divergence.
The results are somewhat mixed, with some studies purporting to find evidence for reduced
efficacy of selection while others do not. In addition to variable results, the studies differ in how
they test for relaxed selection. Moreover, tests of specific biological hypotheses can be quite
different even when the explanatory variables are superficially similar. For example, both a severe
4
Trends in Genetics, Month Year, Vol. xx, No. yy
TIGS 1258 No. of Pages 10
recent bottleneck and a long-term small census size decrease the effective population size, but
their predicted influence on selection is quite different (see below).
Limits to Selection in Nonequilibrial Populations
Theory in Nonequilibrial Populations
Most results concerning the limits to selection assume demographic equilibrium. However, most
populations are rarely at equilibrium. Many taxa that have been studied to test specific predictions of natural selection's limits have experienced a recent bottleneck (e.g., out-of-African
human populations, domesticated plants and animals, plants that recently evolved a propensity
to self-pollinate; Table 1). In these nonequilibrial populations, the effective population size
estimated by sequence diversity (e.g., pairwise sequence diversity at synonymous sites) does
not adequately summarize the complex predictions of population genetic theory. We summarize
recent developments in population genetic theory addressing the limits of selection in nonequilibrium populations in Box 2. We also highlight the extension of these models to spatially
expanding populations [26,45].
Overall, the theory summarized in Box 2 demonstrates that both a rapid decrease in population
size followed by a fast recovery and a spatial population expansion will have subtle influences on
the action of selection. If mutations have an additive effect on fitness, a recent bottleneck or
population expansion will have a negligible effect on population fitness [31,33,46]. However, the
case becomes more complex for recessive variants [24,45–47], where a short-term bottleneck
can change a population's fitness. Thus, while longer-term reductions in effective population size
are expected to have important effects on population mean fitness, the effects of shorter-term
demographic bottlenecks followed by recovery depend in important ways on the dominance of
deleterious mutations.
Inferring Selection's Limits in Nonequilibrial Populations
A reconsideration of what constitutes empirical evidence of reductions in the efficacy of selection
has accompanied the development of theory described in Box 2. Importantly, one common
measure of weakened selection – the relative ratio of putatively deleterious to putatively neutral
variation within populations (e.g., pn/ps) can be particularly misled by recent bottlenecks. That is,
because neutral variation approaches equilibrium more slowly than deleterious variation (Box 3),
ratios of deleterious to neutral variation will be elevated in bottlenecked populations. Therefore,
Box 2. Theory: Selection's Limits in Nonequilibrial Populations
Most rare mutations are stochastically lost in the contraction phase of a bottleneck. Therefore, a recently bottlenecked
population will retain few segregating deleterious mutations. However, those that survive the contraction are pushed to a
higher frequency, and the average number of deleterious mutations per individual remains roughly constant after a
population contraction [33,46,47]. Providing a bottlenecked population remains small, mildly deleterious mutations can
accumulate; however, this effect is largely minor and transient, especially if the period of small population size is limited.
The impact of a bottleneck on population fitness depends on the mode of gene action. Because the average number of
deleterious mutations per individual is insensitive to recent bottlenecks, the genetic load induced by alleles with additive
effects on fitness is also insensitive to brief bottlenecks [33,46]. By contrast, the higher frequency of alleles surviving a
bottleneck means that the deleterious effects recessive mutations will be expressed, increasing the genetic load [46,47].
However, this burden is temporary: recessive variants return to equilibrium frequencies well before novel deleterious
mutations drift to considerable frequencies [46]. Therefore, theory predicts a deficit in the per-individual number of
deleterious recessive mutations in bottlenecked populations [47].
The impact of population expansions on the action of selection has also received much theoretical interest. Mutations
arising on the geographical edge of an expanding population experience an evolutionary advantage with little regard to
their fitness consequence, a phenomenon known as gene surfing [83–90]. The decline in fitness associated with
deleterious mutations riding the wave front is called the expansion load [26,45]. Standing variants with additive effects on
fitness do not contribute to the expansion load because drift does not influence expected allele frequencies [26,45].
However, new deleterious mutations with additive effects gradually accumulate at the wave front, generating an
expansion load [26]. Additionally, standing recessive variants that ride the wave front increase the expansion load
by exposing their fitness costs [45].
Trends in Genetics, Month Year, Vol. xx, No. yy
5
TIGS 1258 No. of Pages 10
Box 3. Population Contractions Generate a Relative Excess of Deleterious Variation, Regardless of the
Efficacy of Selection
Over time, mutation, drift, and selection generate an equilibrium distribution of allele frequencies across classes of
functional elements in the genome. Major changes in population size shift genetic variation from this equilibrium by
changing the number of chromosomes (in the population) that can be hit by mutation and in the face of population
contractions, removing a large number of segregating mutations.
The influence of demography on genome-wide patterns of variation is well known. However, the fact that this equilibrium
is reached more quickly for deleterious mutations than neutral variants (Figure I) (see also [34,35,49]) has not been
appreciated in our interpretation of the relative ratios of synonymous and nonsynonymous variation. In other words,
segregating deleterious mutations are at a lower frequency at equilibrium than are weakly selected mutations and so
deleterious mutations recover to their pre-bottleneck equilibrium levels faster than weakly selected mutations simply
because their frequency was not high before the bottleneck.
Genec diversity X me, by strength of purifying selecon,
starng with one chromsome
Key:
Neutral (s=0)
s = 0.001
s = 0.01
s = 0.1
0.000
0.002
0.004
π%
0.006
0.008
0.010
The implications of this result for inferring reductions in the efficacy of selection are serious. We often evaluate evidence for
differences in selection in recently bottlenecked populations using polymorphism patterns. The more rapid return of
equilibrium variation at nonsynonymous than synonymous sites means that we could be misled to infer our hypothesized
prediction; that is, we will infer that selection is relaxed following a bottleneck (elevated pn/ps) regardless of the actual
process. Thus, a direct estimate of the number of putatively deleterious mutations can provide a stronger test of
predictions of relaxed selection [31,33]. Alternatively, methods that estimate the distribution of fitness effects while
explicitly controlling for nonequilibrium population history can be implemented [62,63,91–94].
0
1000
2000
3000
4000
5000
Gen
Figure I. pn Reaches Equilibrium Faster Than ps after a Population Bottleneck. Pairwise sequence
diversity for a population of a constant size that begins with no variation (e.g., a population that has
undergone[2_TD$IF] a contraction [3_TD$IF]that [4_TD$IF]wiped [5_TD$IF]out [6_TD$IF]all genetic diversity and then experiences an instantaneous
expansion) with differing strengths of purifying selection. The blue or green lines are proxies for pn
and the red and black lines are proxies for ps. At less than 2000 generations into the expansion, for
example, pn/ps calculated using the green line (pn) to the red line (ps) must be interpreted with the
caveat that the red line has not reached equilibrium. Our cartoon corresponds directly to the finding
of Pennings et al. [35]: after a genome-wide sweep associated with drug resistance in HIV, the ratio
of nonsynonymous to synonymous diversity (pn/ps) is high and decreases with time. This result
reflects the more rapid approach to equilibrium under negative selection (pn) than with drift (ps) [34],
presumably because deleterious mutations rise to a lower frequency.
6
Trends in Genetics, Month Year, Vol. xx, No. yy
TIGS 1258 No. of Pages 10
pn/ps measures in bottlenecked populations can generate statistical artifacts that mimic expectations of relaxed selection [31,33–35,48,49].
Simons et al. [33] and Do et al. [31] have therefore suggested that an increased number of
deleterious mutations per individual, rather than the relative frequency of segregating variants,
should be used to estimate selection's limits. However, while the number of putatively deleterious mutations is robust to artifacts induced by nonequilibrium demography, this measure
neglects the frequency of mutations. Therefore, simply counting the number of deleterious
mutations per individual does not reliably estimate the genetic load when deleterious mutations
are partially recessive [24,33]. Because deleterious mutations are likely to be often recessive [50],
and because theoretical results concerning selection in nonequilibrial populations depend on the
mode of gene action, some authors have found the per-individual number of deleterious
mutations measure to be an inadequate summary of the genetic load [24,45,51,52].
The challenges in estimating the genetic load has intensified debate on the extent to which
selection has been limited in populations experiencing recent demographic shifts and particularly
in out-of-African human populations [24,26,31–33,53]. Out-of-African human populations contain proportionately more putatively deleterious variation relative to neutral genetic diversity than
African populations [25,26,54,55]. Counting the per-individual number of putatively deleterious
derived mutations reveals subtle [52,56] or no [31,33] difference among populations. However,
on average out-of-African populations are more homozygous for the derived, putatively deleterious mutations they do carry, and if these mutations are on average recessive, out-of-African
populations may have an elevated mutation load [24,32,52]. More generally, most populations
with recent dips in effective population size carry proportionally more putatively deleterious
variation than closely related populations with large effective population sizes (Table 1). The
observed increase in the proportion of putatively deleterious variation is consistent with expectations that selection is limited in populations with lower effective population sizes but can
plausibly be induced by the artifacts discussed above. Accounting for nonequilibrium while
considering the effects of dominance is therefore critical to our interpretation of the extent to
which selection has been relaxed in bottlenecked populations.
Promising Directions
As a field, we have not yet identified how to best infer selection's limits. One creative approach
weights genotypes by an estimate of their fitness costs under alternative dominance coefficients
[52]. However, such attempts must admit the considerable uncertainty in estimates of both the
mode of gene action and the fitness effect of mutations, which are usually inferred using crossspecies sequence conservation and/or functional predictions [57–60] and are not an actual
fitness estimate.
Similar challenges lie in the interpretation of other comparisons of selection efficacy in Table 1.
Selfing populations, for example, rarely show evidence for higher rates of substitution at selected
sites between species (using dN/dS), but more often show higher selected relative to neutral
variation segregating within populations (see [61] and examples in Table 1). Domesticated
organisms have similarly shown elevated proportions of selected variation compared with wild
relatives. The extent to which these polymorphism patterns are a simple consequence of
nonequilibrium demography without a corresponding increase in genetic load remains uncertain; however, recent studies in dogs [18] and horses [19] suggest that an increase in the genetic
load has accompanied domestication. This supports the hypothesis that a higher genetic load is
a ‘cost of domestication’ [16].
Model-based inference provides an alternative approach to comparing the efficacy of selection
across populations and genomic regions. Of particular note is a popular set of models that
Trends in Genetics, Month Year, Vol. xx, No. yy
7
TIGS 1258 No. of Pages 10
estimate the distribution of fitness effects (DFE) of deleterious mutations while explicitly
incorporating simple models of nonequilibrium demography [62,63]. The DFE approach has been
recently harnessed to infer both an accumulation of modestly deleterious mutations and the purging
of severely damaging mutations following the transition to self-fertilization [27]. However, DFE is
improperly calibrated under complex demographic histories [62] and with recurrent selection [64].
Moreover, current DFE models assume an additive model of gene action and random mating,
which can complicate interpretations of fitness differences across species in the presence of
recessive mutations, especially across species that differ in ploidy and selfing rate [27,65].
Outstanding Questions
More explicit modeling of the joint effects of dominance, demography, and nonequilibrium
transitions in population size and mating system will enable a more informed interpretation of
selection from population genomic data. With realistic forward simulations incorporating nonequilibrium, mating system, and our uncertainty in selection and dominance parameters, we
should be able to gain better insights into the scale and significance of selection differences.
These comparisons in combination with model fitting may also in turn inform us about the
distribution of dominance coefficients.
How rapidly can changes in population
size and recombination drive major
shifts in mean fitness and mutation load?
We caution, however, that the interpretation of results from model-based inference should be
treated with care, as model misspecification can mislead inference. For example, some increase
in the genetic load of out-of-African human populations is to likely reflect the introgression of
deleterious alleles from archaic human populations [66,67] in which selection was likely to have
been relaxed [31] and not accounting for this history of archaic introgression would seriously
mislead models that aim to infer the extent to which population expansion has limited natural
selection as some members of our species left Africa. Another possibility is that colonization
bottlenecks and transitions in mating system are associated with reduced selection coefficients
themselves, as shifts in environment and selective pressures free populations from historical
selection pressures [30]. Distinguishing the role of effective population size change and differences in selection strengths adds a further challenge in interpretation.
Concluding Remarks
As large-scale genomic datasets accumulate, there are exciting possibilities for investigating the
extent to which populations and genomic regions differ in the strength of selection and how this
affects the evolutionary process. Such analyses, however, highlight the difficulties in quantifying
selection's limits, especially in nonequilibrial populations. In parallel with increasingly sophisticated model-based approaches to investigate population history using neutral diversity, it is
becoming clear that simple comparisons of summaries of selected variation can give a limited or
inaccurate picture of differences in selection. Future research addressing selection's limits will
improve with explicit consideration of nonequilibrium conditions and mode of gene action; this
will inform better understanding of the extent to which species differ in deleterious mutation
accumulation (see Outstanding Questions). Meanwhile, we encourage researchers evaluating
evidence for relaxed selection to present multiple analyses of their data and explicit modeling of
demographic and mutational effects on fitness and to jointly consider the benefits and limitations
of these approaches when interpreting their results.
Acknowledgments
The authors thank Suzanne McGaugh and Josh Schraiber for their comments and suggestions on the manuscript and Brian
Charlesworth for helpful discussion.
References
1. Barton, N. and Partridge, L. (2000) Limits to natural selection.
Bioessays 22, 1075–1084
3. Pritchard, J.K. (2001) Are rare variants responsible for susceptibility to complex diseases? Am. J. Hum. Genet. 69, 124–137
2. Eyre-Walker, Y.C. and Eyre-Walker, A. (2014) The role of mutation
rate variation and genetic diversity in the architecture of human
disease. PLoS ONE 9, e90166
4. Lande, R. (1998) Risk of population extinction from fixation
of deleterious and reverse mutations. Genetica 102–103,
21–27
8
Trends in Genetics, Month Year, Vol. xx, No. yy
How can we develop models that
inform the historical action of selection
in nonequilibrial populations?
What is the distribution of dominance
coefficients of deleterious mutations
and how does this influence differences
among populations and species in the
efficacy of selection?
Did domestication bottlenecks, transitions to self-fertilization, out-of-African
bottlenecks, and northward expansion
following the retreat of the glaciers drive
an increase in mutation load and concomitant reduction in mean fitness?
TIGS 1258 No. of Pages 10
5. Lynch, M. et al. (1995) Mutation accumulation and the extinction of
small populations. Am. Nat. 146, 489–518
6. Lynch, M. et al. (1993) The mutational meltdown in asexual populations. J. Hered. 84, 339–344
31. Do, R. et al. (2015) No evidence that selection has been less
effective at removing deleterious mutations in Europeans than in
Africans. Nat. Genet. 47, 126–131
7. Goldberg, E.E. et al. (2010) Species selection maintains selfincompatibility. Science 330, 493–495
32. Lohmueller, K.E. (2014) The impact of population demography
and selection on the genetic architecture of complex traits. PLoS
Genet. 10, e1004379
8. Igic, B. and Busch, J.W. (2013) Is self-fertilization an evolutionary
dead end? New Phytol. 198, 386–397
33. Simons, Y.B. et al. (2014) The deleterious mutation load is insensitive to recent population history. Nat. Genet. 46, 220–224
9. Takebayashi, N. and Morrell, P.L. (2001) Is self-fertilization an
evolutionary dead end? Revisiting an old hypothesis with genetic
theories and a macroevolutionary approach. Am. J. Bot. 88,
1143–1150
34. Gordo, I. and Dionisio, F. (2005) Nonequilibrium model for estimating parameters of deleterious mutations. Phys. Rev. E: Stat.
Nonlin. Soft Matter Phys. 71, 031907
10. Wright, S.I. et al. (2013) Evolutionary consequences of self-fertilization in plants. Proc. Biol. Sci. 280, 20130133
11. Charlesworth, B. and Charlesworth, D. (2000) The degeneration of
Y chromosomes. Philos. Trans. R. Soc. Lond. B: Biol. Sci. 355,
1563–1572
12. Lynch, M. (2006) The origins of eukaryotic gene structure. Mol.
Biol. Evol. 23, 450–468
13. Lynch, M. (2007) The Origins of Genome Architecture, Sinauer
Associates
14. Lynch, M. and Conery, J.S. (2003) The origins of genome complexity. Science 302, 1401–1404
35. Pennings, P.S. et al. (2014) Loss and recovery of genetic diversity
in adapting populations of HIV. PLoS Genet. 10, e1004000
36. Haldane, J.B.S. (1937) The effect of variation on fitness. Am. Nat.
71, 337–349
37. Agrawal, A.F. and Whitlock, M.C. (2012) Mutation load: the fitness
of individuals in populations where deleterious alleles are abundant. Annu. Rev. Ecol. Evol. 43, 115–135
38. Li, H. and Durbin, R. (2011) Inference of human population history
from individual whole-genome sequences. Nature 475, 493–496
39. Rasmussen, M.D. et al. (2014) Genome-wide inference of ancestral recombination graphs. PLoS Genet. 10, e1004342
15. Lynch, M. et al. (2006) Mutation pressure and the evolution of
organelle genomic architecture. Science 311, 1727–1730
40. Schiffels, S. and Durbin, R. (2014) Inferring human population size
and separation history from multiple genome sequences. Nat.
Genet. 46, 919–925
16. Lu, J. et al. (2006) The accumulation of deleterious mutations in
rice genomes: a hypothesis on the cost of domestication. Trends
Genet. 22, 126–131
41. Sheehan, S. et al. (2013) Estimating variable effective population
sizes from multiple genomes: a sequentially Markov conditional
sampling distribution approach. Genetics 194, 647–662
17. Renaut, S. and Rieseberg, L.H. (2015) The accumulation of deleterious mutations as a consequence of domestication and
improvement in sunflowers and other Compositae crops. Mol.
Biol. Evol. 32, 2273–2283
42. Conrad, D.F. et al. (2011) Variation in genome-wide mutation rates
within and between human families. Nat. Genet. 43, 712–714
18. Marsden, C.D. et al. (2016) Bottlenecks and selective sweeps
during domestication have increased deleterious genetic variation
in dogs. Proc. Natl. Acad. Sci. U.S.A. 113, 152–157
44. Gravel, S. (2014) When is selection effective? bioRxiv Published
online October 30, 2014. http://dx.doi.org/10.1101/010934
19. Schubert, M. et al. (2014) Prehistoric genomes reveal the genetic
foundation and cost of horse domestication. Proc. Natl. Acad. Sci.
U.S.A. 111, E5661–E5669
20. Laberge, A.M. et al. (2005) Population history and its impact on
medical genetics in Quebec. Clin. Genet. 68, 287–301
21. Polvi, A. et al. (2013) The Finnish Disease Heritage Database
(FinDis) update – a database for the genes mutated in the Finnish
disease heritage brought to the next-generation sequencing era.
Hum. Mutat. 34, 1458–1466
22. Casals, F. et al. (2013) Whole-exome sequencing reveals a rapid
change in the frequency of rare functional variants in a founding
population of humans. PLoS Genet. 9, e1003845
23. Lim, E.T. et al. (2014) Distribution and medical impact of loss-offunction variants in the Finnish founder population. PLoS Genet.
10, e1004494
24. Lohmueller, K.E. (2014) The distribution of deleterious genetic
variation in human populations. Curr. Opin. Genet. Dev. 29,
139–146
25. Lohmueller, K.E. et al. (2008) Proportionally more deleterious
genetic variation in European than in African populations. Nature
451, 994–995
26. Peischl, S. et al. (2013) On the accumulation of deleterious mutations during range expansions. Mol. Ecol. 22, 5972–5982
27. Arunkumar, R. et al. (2015) The evolution of selfing is accompanied
by reduced efficacy of selection and purging of deleterious mutations. Genetics 199, 817–829
28. Brandvain, Y. et al. (2014) Speciation and introgression between
Mimulus nasutus and Mimulus guttatus. PLoS Genet. 10,
e1004410
29. Hollister, J.D. et al. (2015) Recurrent loss of sex is associated with
accumulation of deleterious mutations in Oenothera. Mol. Biol.
Evol. 32, 896–905
30. Slotte, T. et al. (2013) The Capsella rubella genome and the
genomic consequences of rapid mating system evolution. Nat.
Genet. 45, 831–835
43. Kong, A. et al. (2012) Rate of de novo mutations and the importance of father's age to disease risk. Nature 488, 471–475
45. Peischl, S. and Excoffier, L. (2015) Expansion load: recessive
mutations and the role of standing genetic variation. Mol. Ecol.
24, 2084–2094
46. Kirkpatrick, M. and Jarne, P. (2000) The effects of a bottleneck
on inbreeding depression and the genetic load. Am. Nat. 155,
154–167
47. Balick, D.J. et al. (2015) Dominance of deleterious alleles controls
the response to a population bottleneck. PLoS Genet. 11,
e1005436
48. Jensen, J.D. and Bachtrog, D. (2011) Characterizing the influence
of effective population size on the rate of adaptation: Gillespie's
Darwin domain. Genome Biol. Evol. 3, 687–701
49. Song, Y.S. and Steinrucken, M. (2012) A simple method for finding
explicit analytic transition densities of diffusion processes with
general diploid selection. Genetics 190, 1117–1129
50. Agrawal, A.F. and Whitlock, M.C. (2011) Inferences about the
distribution of dominance drawn from yeast gene knockout data.
Genetics 187, 553–566
51. Henn, B.M. et al. (2015) Estimating the mutation load in human
genomes. Nat. Rev. Genet. 16, 333–343
52. Henn, B.M. et al. (2016) Distance from sub-Saharan Africa predicts mutational load in diverse human genomes. Proc. Natl. Acad.
Sci. U.S.A. 113, E440–E449
53. Sousa, V. et al. (2014) Impact of range expansions on current
human genomic diversity. Curr. Opin. Genet. Dev. 29, 22–30
54. Fu, W.Q. et al. (2013) Analysis of 6,515 exomes reveals the
recent origin of most human protein-coding variants. Nature
493, 216–220
55. Tennessen, J.A. et al. (2012) Evolution and functional impact of
rare coding variation from deep sequencing of human exomes.
Science 337, 64–69
56. Fu, W. et al. (2014) Characteristics of neutral and deleterious
protein-coding variation among individuals and populations.
Am. J. Hum. Genet. 95, 421–436
57. Chun, S. and Fay, J.C. (2009) Identification of deleterious mutations within three human genomes. Genome Res. 19, 1553–1561
Trends in Genetics, Month Year, Vol. xx, No. yy
9
TIGS 1258 No. of Pages 10
58. Davydov, E.V. et al. (2010) Identifying a high fraction of the human
genome to be under selective constraint using GERP++. PLoS
Comput. Biol. 6, e1001025
76. Gossmann, T. et al. (2014) Highly variable recombinational landscape modulates efficacy of natural selection in birds. Genome
Biol. Evol. 6, 2061–2075
59. Ng, P.C. and Henikoff, S. (2003) SIFT: predicting amino acid
changes that affect protein function. Nucleic Acids Res. 31,
3812–3814
77. Campos, J.L. et al. (2014) The relation between recombination
rate and patterns of molecular evolution and variation in Drosophila
melanogaster. Mol. Biol. Evol. 31, 1010–1028
60. Stone, E.A. and Sidow, A. (2005) Physicochemical constraint
violation by missense substitutions mediates impairment of protein
function and disease severity. Genome Res. 15, 978–986
78. Burgarella, C. et al. (2015) Molecular evolution of freshwater snails
with contrasting mating systems. Mol. Biol. Evol. 32, 2403–2416
61. Glemin, S. and Galtier, N. (2012) Genome evolution in outcrossing
versus selfing versus asexual species. Methods Mol. Biol. 855,
311–335
62. Keightley, P.D. and Eyre-Walker, A. (2007) Joint inference of the
distribution of fitness effects of deleterious mutations and population demography based on nucleotide polymorphism frequencies.
Genetics 177, 2251–2261
63. Boyko, A.R. et al. (2008) Assessing the evolutionary impact of
amino acid mutations in the human genome. PLoS Genet. 4,
e1000083
64. Messer, P.W. and Petrov, D.A. (2013) Frequent adaptation and the
McDonald–Kreitman test. Proc. Natl. Acad. Sci. U.S.A. 110,
8615–8620
65. Douglas, G.M. et al. (2015) Hybrid origins and the earliest
stages of diploidization in the highly successful recent polyploid
Capsella bursa-pastoris. Proc. Natl. Acad. Sci. U.S.A. 112,
2806–2811
66. Harris, K. and Nielsen, R. (2015) The genetic cost of Neanderthal
introgression. bioRxiv Published online October 30, 2015. http://
dx.doi.org/10.1101/030387
67. Juric, I. et al. (2015) The strength of selection against Neanderthal
introgression. bioRxiv Published online November 21, 2015.
http://dx.doi.org/10.1101/030148
68. Akashi, H. et al. (2012) Weak selection and protein evolution.
Genetics 192, 15–31
69. Koenig, D. et al. (2013) Comparative transcriptomics reveals patterns of selection in domesticated and wild tomato. Proc. Natl.
Acad. Sci. U.S.A. 110, E2655–E2662
70. Nabholz, B. et al. (2014) Transcriptome population genomics
reveals severe bottleneck and domestication cost in the African
rice (Oryza glaberrima). Mol. Ecol. 23, 2210–2227
71. Cao, J. et al. (2011) Whole-genome sequencing of multiple Arabidopsis thaliana populations. Nat. Genet. 43, 956–963
72. Elyashiv, E. et al. (2010) Shifts in the intensity of purifying selection:
an analysis of genome-wide polymorphism data from two closely
related yeast species. Genome Res. 20, 1558–1573
79. Brandvain, Y. et al. (2013) Genomic identification of founding
haplotypes reveals the history of the selfing species, Capsella
rubella. PLoS Genet. 9, e1003754
80. Qiu, S. et al. (2011) Reduced efficacy of natural selection on codon
usage bias in selfing Arabidopsis and Capsella species. Genome
Biol. Evol. 3, 868–880
81. Pellino, M. et al. (2013) Asexual genome evolution in the apomictic
Ranunculus auricomus complex: examining the effects of hybridization and mutation accumulation. Mol. Ecol. 22, 5908–5921
82. Mank, J.E. et al. (2010) Effective population size and the Faster-X
effect: empirical results and their interpretation. Evolution 64,
663–674
83. Edmonds, C.A. et al. (2004) Mutations arising in the wave front
of an expanding population. Proc. Natl. Acad. Sci. U.S.A. 101,
975–979
84. Excoffier, L. and Ray, N. (2008) Surfing during population expansions promotes genetic revolutions and structuration. Trends Ecol.
Evol. 23, 347–351
85. Hallatschek, O. and Nelson, D.R. (2010) Life at the front of an
expanding population. Evolution 64, 193–206
86. Klopfstein, S. et al. (2006) The fate of mutations surfing on the
wave of a range expansion. Mol. Biol. Evol. 23, 482–490
87. Lehe, R. et al. (2012) The rate of beneficial mutations surfing on the
wave of a range expansion. PLoS Comput. Biol. 8, e1002447
88. Moreau, C. et al. (2011) Deep human genealogies reveal a selective advantage to be on an expanding wave front. Science 334,
1148–1150
89. Slatkin, M. and Excoffier, L. (2012) Serial founder effects during
range expansion: a spatial analog of genetic drift. Genetics 191,
171–181
90. Travis, J.M.J. et al. (2007) Deleterious mutations can surf to high
densities on the wave front of an expanding population. Mol. Biol.
Evol. 24, 2334–2343
91. Eyre-Walker, A. and Keightley, P.D. (2007) The distribution of
fitness effects of new mutations. Nat. Rev. Genet. 8, 610–618
73. Hill, W.G. and Robertson, A. (1966) The effect of linkage on limits
to artificial selection. Genet. Res. 8, 269–294
92. Eyre-Walker, A. and Keightley, P.D. (2009) Estimating the rate of
adaptive molecular evolution in the presence of slightly deleterious mutations and population size change. Mol. Biol. Evol. 26,
2097–2108
74. Hussin, J.G. et al. (2015) Recombination affects accumulation of
damaging and disease-associated mutations in human populations. Nat. Genet. 47, 400–404
93. Eyre-Walker, A. et al. (2006) The distribution of fitness effects of
new deleterious amino acid mutations in humans. Genetics 173,
891–900
75. Bullaughey, K. et al. (2008) No effect of recombination on
the efficacy of natural selection in primates. Genome Res. 18,
544–554
94. Kousathanas, A. and Keightley, P.D. (2013) A comparison of
models to infer the distribution of fitness effects of new mutations.
Genetics 193, 1197–1208
10
Trends in Genetics, Month Year, Vol. xx, No. yy