Review Can Population Genetics Adapt to Rapid...

ReviewCan Population GeneticsAdapt to Rapid Evolution?Philipp W. Messer,1,* Stephen P. Ellner,2 andNelson G. Hairston Jr.2

Population genetics largely rests on a ‘standardmodel’ in which randomgeneticdrift is the dominant force, selective sweeps occur infrequently, and deleteriousmutations are purged from the population by purifying selection. Studies ofphenotypic evolution in nature reveal a very different picture, with strongselection and rapid heritable trait changes being common. The time-rate scalingof phenotypic evolution suggests that selection on phenotypes is often fluctu-ating in direction, allowing phenotypes to respond rapidly to environmentalfluctuations while remaining within relatively constant bounds over longer peri-ods. Whether such rapid phenotypic evolution undermines the standard modelwill depend on how many genomic loci typically contribute to strongly selectedtraits and how phenotypic evolution impacts the dynamics of genetic variation ina population. Population-level sequencing will allow us to dissect the geneticbasis of phenotypic evolution and study the evolutionary dynamics of geneticvariation through direct measurement of polymorphism trajectories over time.

The Standard Model of Population GeneticsCharles Darwin thought of evolution as an innately slow process driven by small incrementalchanges that lead to noticeable differences between species only because they can accumu-late over long periods of time. This view still runs deep in modern population genetics. We tendto assume that selective sweeps are rare in most natural populations and that most commongenetic variation remains largely unaffected by such events. Catching a beneficial mutation onthe fly should be extremely unlikely. Deleterious mutations can occur in individuals and some ofthem may have strong effects on fitness, yet purifying selection will prevent them from everbecoming prevalent in a population. The general conclusion in modern population genetics hasbeen that the vast majority of common genetic variation should have only small effects onfitness, if any. Summarizing this view, John Gillespie wrote ‘the forces that alter the geneticstructure of populations tend to be very weak, operating on time scales of thousands tomillionsof years’ [1].

The paradigm of slow evolution by small changes has convenient implications for both theoryand practice. The few selective sweeps that occur should leave characteristic signatures in thesurrounding patterns of diversity that stand out against the prevailing neutral background.Searching for such signatures in genomic data then allows us to identify recent adaptive events.Adaptation to environmental changes should proceed via well-separated successive fixations ofnew mutations in a process called an adaptive walk, which can be studied in theoretical models[2,3]. The dominant process in the evolution of common genetic polymorphisms should berandom genetic drift, which we understandmathematically and can incorporate rather easily intoour theoretical models and computational tools. Figure 1 provides an illustration of evolutionarydynamics under these assumptions, which we will hereafter refer to as the ‘standard model’ ofpopulation genetics.

TrendsStudies of evolution in action show thatphenotypic traits can often change dra-matically over the course of just a fewgenerations.

This differs markedly from the para-digm of slow molecular evolution com-monly adopted in our populationgenetic models.

To assess whether our models are stillappropriate, we need to better under-stand how rapid phenotypic evolutionimpacts the trajectories of genetic var-iation in a population.

New population genomic datasets willmake it possible to directly observe howpolymorphism frequencies change in apopulation over time, allowing us to testand refine our population geneticmodels.

1Department of Biological Statisticsand Computational Biology, CornellUniversity, Ithaca, NY 14853, USA2Department of Ecology andEvolutionary Biology, CornellUniversity, Ithaca, NY 14853, USA

*Correspondence: [email protected](P.W. Messer).

TIGS 1279 No. of Pages 11

Trends in Genetics, Month Year, Vol. xx, No. yy http://dx.doi.org/10.1016/j.tig.2016.04.005 1© 2016 Elsevier Ltd. All rights reserved.


The standard model's drift-dominant view has enabled the development of a rapidly growingtoolbox for inference of selection, demography, and population structure from genomic data[4,5]. It is also consistent with estimates of the molecular rate of adaptive evolution in largeorganisms, such as humans, as long as we assume that evolution generally occurs by randomgenetic drift and selection on consistently beneficial or deleterious mutations. For example,recent estimates suggest that on the order of 10 000 adaptive amino acid changes haveoccurred between humans and chimpanzees [6–8]–about one adaptive amino acid substitutionevery 50 generations. If those substitutions proceeded by selective sweeps with moderatefitness benefits of 1%, and using an average recombination rate of 1 cM/Mbp, each such eventshould have affected a surrounding genomic region of approximately 100 kbp. Assuming thatsweeps occur uniformly across the genome, a random genomic position in the 3-Gbp humangenome should therefore be affected only about once every 1 500 000 generations–muchlonger than the average lifespan of a neutral polymorphism.

Examples of Rapid Phenotypic Evolution in Nature and ExperimentStudies of ‘evolution in action’ paint a markedly different picture from the paradigm of slowmolecular evolution commonly adopted in our population genetic models. These studies showthatphenotypic traits canoftenchangedramatically over the courseof just a fewgenerations.Peterand Rosemary Grant's classic studies of rapid evolution in Darwin's finches [9–11] are well knownby both scientists and non-scientists, but many other studies have now demonstrated rapidchange in heritable traits in other organisms. For example, Koskinen et al. observed that naturalselection led to diversifying evolution in life-history traits in grayling over the course of only 10–20generations after a small number of fish were introduced to a new lake [12]. Cichlid fish rapidlyevolved the same elongated-body phenotype in isolatedNicaraguan crater lakes [13] and similarlyhigh rates of phenotypic evolution have been observed in berry bug beak length [14,15], colorationand life-history traits in guppies [16,17], and rainbow trout [18]. One of the most rapid examplesinvolves diapause date (a mechanism for avoiding fish predation) in the freshwater copepod

0 1000 2000 30000

1

Genera!ons

Freq

uenc

y

BeneficialDeleteriousNeutral

Figure 1. Polymorphism Dynamics in the Standard Model. The figure shows several thousand mutation trajectoriesin a population of 1000 haploid individuals simulated under a Wright–Fisher model with selection and drift. All mutations areassumed to evolve independently of each other without any genetic linkage between them. Gray trajectories depict neutralmutations that arise in a single copy and then evolve by drift. Some of them can eventually become fixed in the populationand neutral mutations thus contribute to both polymorphism and divergence. Red trajectories show deleterious mutationswith a selective disadvantage of 5% occurring at the same rate as the neutral mutations in our model. Purifying selectionprevents these mutations from reaching higher frequencies and thus contributing to divergence or common polymorphism.The blue trajectory depicts a single beneficial mutation with a selective advantage of 5% that overcomes drift and quicklysweeps to fixation. Beneficial mutations contribute to divergence, but we should be unlikely to catch them in a populationsample taken at a random point in time.

2 Trends in Genetics, Month Year, Vol. xx, No. yy


Onychodiaptomus sanguineus, which changed at an average rate of roughly one standarddeviation per selected generation following complete fish removal by drought [19]. Excellentreviews of the growing list of examples of rapid contemporary evolution are provided in [20,21].

Rapid phenotypic evolution is also commonly observed in evolution experiments. Conover and[3_TD$DIFF]Munch documented changes in body mass in a marine fish of both a 45% increase (highselection) and a 25% decrease (low selection) over the course of only four generations of artificialselection [22]. Anolis lizards repeatedly evolved larger toe pads over just 20 generations(Figure 2A), allowing them to reach higher perch heights in response to invasion by a congener[23]. In laboratory microcosms of rotifer–algal dynamics, the alga Chlamydomonas evolvedcoloniality as a defense trait within a single population cycle of high rotifer density [24,25]. In soilmites, a 76% increase in age to maturity was observed over a few generations after a wild

SPC3 (1)

YUR1 (11)DOT6 (2)PCT1 (1)

WHI2 (7)STE4 (6)

HXT14 (1)Intergenic

YPR117W (1)

ROT2 (11)WHI2 (7)PMD1 (2)YPR196W (1)2 Intergenic

0 200 400 600 1,000800

Genera!on

(A)

(B)

Indi

vidu

als

Perc

h he

ight

(cm

)(m

ean

+ 1

SE)

A. ca

rolin

ensis

toe

pad

area

(cm

)(m

ean

+ 1

SE)

0 –0.2

–0.1

0.0

0.1

50

100

Un-invadedislands

Invadedislands

Un-invadedislands

A.carolinensisA.sagrei

Invadedislands

150

200

Figure 2. Rapid Evolution of Phenotypes and Genotypes. (A) Anolis carolinesis lizards on small islands in Floridarepeatedly moved to higher perches within just a few generations after invasion by their competitor Anolis sagrei [23]. Theseadaptations were associated with rapid, heritable changes in toe pad area that occurred at a rate of roughly 0.1 standarddeviations per generation. (B) Evolutionary dynamics observed in an initially clonal Saccharomyces cerevisiae populationgrowing in rich medium for 1000 generations [42]. Different colors specify the six main ‘cohorts’ observed in the populationover the course of the experiment. Even beneficial driver mutations must often hitchhike with functionally unrelated co-drivers in their cohort because of clonal interference. The numbers in parentheses indicate how often mutations wereobserved in the specific gene across 40 replicate evolution experiments.

Trends in Genetics, Month Year, Vol. xx, No. yy 3


population was transferred into a novel laboratory environment [26]. Similarly dramatic changesin life-history strategies were observed in seed beetles after a host-shift [27].

These abundant examples of rapid phenotypic evolution are not automatically inconsistent withthe paradigm that molecular evolution in nature should generally remain slow. For instance, thereis probably ascertainment bias towards reporting examples of rapid phenotypic evolution inthe literature, compared with studies of populations in which nothing happened [28]. Anotherissue is knowing the extent to which contemporary evolution is driven by environments changingmuch faster than ‘natural’ due to human activities. Classic examples of such human-inducedadaptations include the rapid evolution of industrial melanism in the peppered moth [29] and theincreased rates of phenotypic change in human-harvested organisms [30].

Even if rapid phenotypic evolution is common in nature, evolutionary dynamics on the molecularlevel could remain practically indistinguishable from random genetic drift if rapidly evolving traitstend to be highly polygenic [31,32]. In that case, their selective responses could be due to subtlefrequency changes of genetic polymorphisms at a large number of loci. This so-called infinitesi-mal model of adaptation [3] has a long tradition in population genetics and is widely used foranalyzing and predicting how quantitative traits respond both to natural selection and to artificialselection in applied animal and plant breeding experiments [33].

However, it is also possible that molecular adaptation could be abundant despite a slow rate ofphenotypic changes if molecular changes occur to maintain a relatively constant phenotype inthe face of frequent environmental change [34–36]. Thus, without better knowledge about theactual rates of phenotypic change and how this affects the trajectories of genetic variation in apopulation, it is difficult to say whether the standard model is still appropriate.

Examples of Rapid Molecular Evolution at Individual LociImportant additional clues are provided by sequencing studies that increasingly allow us todissect the genomic basis of phenotypic evolution. Such studies have revealed many examplesof rapid phenotypic adaptation that were associated with extensive allele-frequency changes atindividual genetic loci. One prominent example is the adaptation of marine sticklebacks tofreshwater environments, which is driven by only a small set of genomic loci [37] yet has occurredrepeatedly over just 50 generations after an earthquake created multiple freshwater ponds [38].Similarly, wild crickets have rapidly evolved a specific wing phenotype that, while lowering thepotential of males to attract females, protects them from a parasitoid fly, and this single-locusMendelian trait has evolve repeatedly on neighboring Hawaiian islands [39].

Extensive frequency changes of molecular variants are also commonly observed in laboratoryevolution. A particularly striking example was provided by Levy et al., who evolved an initiallyclonal population of 108 [2_TD$DIFF] asexual yeast cells under glucose-limited conditions for 168 generations[40]. Using a clever barcoding technology that allowed them to track mutations at extremely lowpopulation frequencies, they detected thousands of de novo beneficial mutations that competedwith each other such that only a few eventually became prevalent in the population. This type ofadaptive dynamics is commonly referred to as clonal interference (Figure 2B) and has beenobserved in other evolution experiments [41–43]. It is very different from the adaptive walk modelof successive selective sweeps assumed by the standard model. These experiments alsoshowcase the ample adaptive potential of microbial populations even when they are initiallyclonal, providing an explanation for why such populations can often respond so surprisinglyquickly to environmental challenges [44].

Microbial populations tend to be large and adaptive genetic variants can therefore arise readilyby de novomutation. Yet even in experiments with much smaller populations, artificial selection



often produces extensive allele-frequency changes at many genetic loci. For example, Burkeet al. exposed a laboratory population of Drosophila melanogaster to artificial selection foraccelerated development and observed dozens of genomic regions that showed strong allele-frequency differentiation over the course of the experiment [45]. Similar dynamics have beenobserved in selection experiments for other traits, including body size [46], temperatureresponse [47,48], and pathogen resistance [49,50].

The adaptive dynamics observed in these experiments were, again, very different from theclassic selective sweep model, while also differing from the clonal interference dynamicsobserved in microbial experiments. Because of the typically small population sizes, adaptationwas likely to have proceeded from standing genetic variation already present in the population.Under such circumstances, adaptation is expected to produce so-called soft selective sweeps,where multiple alleles at the same locus rise in frequency simultaneously [51]. Interestingly,fixation of a selected allele was rarely observed in these experiments [45], suggesting that manyof the responding alleles may not be unconditionally beneficial.

Are Short-Term and Long-Term Evolutionary Rates Different?If rapid phenotypic evolution is indeed common in nature and often associated with extensivefrequency changes of molecular variants at many loci, why do we not observe higher levels ofmolecular divergence between extant species? A possible explanation is that selection may notalways be as static as assumed by the standard model. If a considerable fraction of geneticvariants has selection coefficients that vary in sign over space and time, being advantageous atsome times and locations and disadvantageous at others, these variants could enable rapidevolutionary responses without being driven towards fixation or loss in the population. Molecularevolution would therefore appear slow over the timescales of extant species despite pervasiveadaptive changes at any point in time.

In the ecological literature, temporally fluctuating selection on phenotypic traits is generallyassumed to be common and has been invoked to explain the discrepancy between long-termand short-term rates of phenotypic evolution [52–57]. Figure 3 summarizes empirical estimatesof rates of phenotypic evolution for studies spanning a few generations to the macroevolutionaryscale of extant species lifetimes, where trait changes can be estimated over scales of tens ofmillions of years. In both cases, rates of phenotypic evolution are inversely proportional to thetimescale of measurement. That is, they tend to be fast when estimated over short timescales

0 2 4 6 8

−10

−8

−6

−4

−2

0

log10(Years)

log1

0(Ra

te)

Hendry et al. 2008 (slope = −1.03)

Key:

Quintero and Wiens 2013 (slope = −0.92)

Figure 3. An Example of How Mea-sured Evolutionary Rates (TraitChange Divided by Elapsed Time inYears) Scale with the Time PeriodOver Which the Trait Change WasMeasured. Triangles (blue) are rates froman expanded version of the database ofstudies on contemporary rapid trait evolu-tion provided by Hendry and Kinnison[53,100]. Trait change is calculated asthe absolute change in log trait values.Circles (black) are based on estimateddivergences in thermal optimum betweenancestral and extant vertebrate speciesand estimated divergence times [101].Trait change is calculated as the absolutechange divided by the mean, becauseinitial and final values were not tabulated.Unbroken blue and [1_TD$DIFF]broken black lines arelinear regressions fitted to the rates fromthe two datasets.



but become increasingly slower as the time intervals expand. This scaling arises naturally ifphenotypic traits can change rapidly in response to environmental fluctuations but remain withinrelatively constant bounds over longer periods [28,44,52,53].

All natural environments fluctuate, often dramatically, with frequent changes in intensity andreversals of direction. Day–night cycles of light and temperature lead to changes in microbialpopulations. Seasonality is a ubiquitous driver of fluctuating selection in organismswith generationtimes of a month or less. At longer timescales, many populations cycle in abundance every fewgenerations, often as a result of interactions with other species [58,59]. Changes due to climatevariation can be extreme, with warm years and cold years, wet years and dry years, and periods ofhigh fire frequency and years of low all affecting local populations, although often driven by globalprocesses such as the El Nĩno Southern Oscillation, the Pacific Decadal Oscillation, and the NorthAtlantic Oscillation. Recruitment variation between years is frequently on the order of 100-fold to10 000-fold for a variety of organisms, includingplants, invertebrates, andvertebrates fromdiversehabitats [60]. Birth and death rates in plankton populations can vary as much as tenfold or morewithin a season [61]. Biotic factors such as resource quality [9], predation intensity [62], and rapidevolution of interacting species [35] are also potential sources of temporally fluctuating selection.While some populations evolve adaptations to tolerate or even take advantage of such changes inthe conditions they experience, such as bet-hedging strategies and adaptive phenotypic plasticity[63], others track these fluctuations with heritable trait changes [9,62].

Population Genetics Might Be Underestimating the Role of TemporallyFluctuating SelectionWhile it is widely acknowledged that spatially varying selection can play an important role in thedynamics and maintenance of molecular variation [64], population genetics has remained ratherskeptical regarding the significance of temporally fluctuating selection. This view traces back totheoretical arguments showing that, in standard models with non-overlapping generations,temporally fluctuating selection cannot maintain a polymorphism unless the heterozygote hashigher geometric mean fitness than either homozygote [1,65,66]. As a result, temporallyfluctuating selection–and any other forms of balancing selection–are seldom incorporated inour population genetic models. Importantly, this theoretical argument does not preclude thatfluctuating selection could strongly impact the frequency dynamics and persistence time ofgenetic polymorphisms; it says only that temporally fluctuating selection on its own will notmaintain these polymorphisms for very long.

In ecology, the allegation that temporal fluctuations cannot promote diversity is widely con-tested, growing from the discovery of the storage effect mechanism for coexistence of com-peting species in fluctuating environments [67]. The storage effect is driven by density-dependent changes in covariance between environment quality and competition. When aspecies is common, it cannot take full advantage of a year when conditions are good forreproduction because high offspring production will lead to intense competition, limiting popu-lation growth. A rare species, by contrast, may not have this problem because high reproductionby a few individuals may have little effect on the overall level of crowding in the population. In ayear that is good for the rare species but bad for others, it can have high per capita reproductivesuccess. However, these gains by the rare come to nothing if they are wiped out in bad years.The storage effect can therefore maintain coexistence only if species are buffered againstsudden rapid declines. One natural way for this to occur is if generations overlap and establishedindividuals are immune to the causes of temporal variation (e.g., viability selection on offspring,no selection on adults).

The population genetic analog to the storage effect is that an allele close to fixation cannotincrease in frequency by much regardless of how strongly it is favored by selection at the time. A



rare allele, by contrast, has the potential for rapid increase when it is favored by selection. Oncethe assumption of non-overlapping generations is dropped, this mechanism can alreadymaintain a genetic polymorphism in a model of stabilizing selection with a randomly fluctuatingtrait optimum [68,69]. Precisely this kind of dynamic was observed in a population of freshwatercopepods subject to interannual fluctuations in selection for a life-history trait and in which only afraction of the diapausing eggs produced each year hatched the following year [62,70]. Yi andDean demonstrated a similar outcome in a chemostat model of microbial competition whereoverall population abundance was bounded [71].

Another type of genomic storage effect can promote genetic polymorphism in fluctuatingenvironments through recombination [72,73]. The way this mechanism works is that a targetallele could have its deleterious effects buffered in detrimental environments if it becomesgenetically linked to a modifier allele. Once the environment changes and the target allelebecomes advantageous without the modifier, it can escape the modifier background throughrecombination and enjoy full selective advantage [73]. Importantly, this mechanism does notrequire any form of age-specific or spatial heterogeneity in selection.

Heterozygote advantage itself could play a more prominent role in adaptive dynamics andmaintenance of genetic diversity than previously recognized. Sellis et al. showed that adaptivewalks in diploids should often proceed via successive balanced states (Figure 4A), whereinvading alleles are beneficial in heterozygotes but overshoot the fitness optimum in homo-zygotes [74]. The resulting short-lived balanced states allow diploid populations to adapt morerapidly to fluctuating environments by maintaining phenotypically consequential variation in apopulation, although these ephemeral balanced polymorphisms will remain mostly invisible tocurrent scans for long-term balancing selection.

In addition to these theoretical arguments, there is accumulating experimental evidence thatbalancing and fluctuating selection on the molecular level may be more widespread thancurrently accommodated in our population genomic models [75–83]. A particularly striking

(A)

(B)

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

Alle

le f

requ

ency

Spring Fall Fall FallSpring Spring2009 2010 2011

Alle

le f

requ

ency

Time →

Figure 4. Evolutionary DynamicsBeyond the Standard Model. (A)Allele-frequency trajectories at a singlelocus in a diploid population during simu-lated adaptive walks in Fisher's geometricmodel of adaptation [74]. Different colorsindicate the different alleles that wereobserved in the population. Most beneficialalleles do not reach 100% frequency butare maintained for some time at intermedi-ate frequencies due to heterozygoteadvantage. They are then displaced bynew beneficial alleles that themselves areoften balanced. (B) Allele-frequency trajec-tories at 1750 SNPs in the Drosophila mel-anogaster genome that were observed tofluctuate between seasons in a NorthAmerican orchard [81]. Allele frequenciesare polarized such that they are higher inthe spring than in the fall.



example of pervasive fluctuating selection has recently been provided by Bergland et al., whostudied fruit fly samples collected in a temperate orchard in Pennsylvania and identified hundredsof polymorphisms that showed dramatic and repeatable oscillations in allele frequenciesbetween seasons (Figure 4B) [81].

Population Genetics Beyond the Standard ModelIn the standard model we tend to assume that the selection coefficients of mutations remainconstant over time and space. If instead selection coefficients often vary, evolutionary dynamicscould be quite different. In this case, selection could play a much more important role among theprocesses that cause alleles to change in frequency over time. Classic selective sweeps, however,would remain rare, as alleles would usually not be driven all the way to fixation or loss. Instead, weshould more often observe incomplete sweeps, soft sweeps, and alleles that rapidly change infrequency but in an inconsistent manner. Identifying selection from patterns of genetic diversitywouldbemuchmoredifficult in this caseandscans for selective sweepswouldprovideuswithonlya very limited picture of the overall contribution of selection to evolutionary dynamics.

It would also no longer hold that functional genetic variants should necessarily have small effectson fitness if they are common in a population. Some of those common variants could be quiteconsequential, just not consistently good or bad. Indeed, these variants would be the very onesmost likely to play a key role in providing the potential for the rapid evolutionary responses toenvironmental change. An extreme example of this is mutations conferring resistance to drugs orpesticides, which can provide tremendous fitness benefits in the exposed parts of a populationwhere they would sweep rapidly but typically carry high costs in untreated populations wherethey hence remain rare.

The assumption that each mutation can be assigned a fixed selection coefficient lies at thefoundation of many key concepts in population genetics, including mutation–selection balanceand the Poisson random field [84]. These concepts, in turn, form the basis for approaches widelyused to infer evolutionary parameters from population genomic data, such as McDonald–Kreitman-type tests for estimating rates of adaptive substitution [85–87] and inference of thedistribution of fitness effects of new mutations [6,88–90]. It is unclear to what extent theseapproaches would be affected by frequent fluctuating selection, except that they would almostcertainly underestimate the prevalence and strength of selection.

Finally, if fluctuating selection is common, even neutral genetic variation might not be subject torandom genetic drift alone. Linked selection could strongly affect the dynamics of neutralpolymorphisms similarly to so-called genetic draft, which describes the effects of linked selectionon neutral polymorphisms under frequent classic selective sweeps [91,92]. However, whilegenetic draft is expected to leave characteristic signatures in the frequency spectra of neutraldiversity, such as an excess of high- and low-frequency polymorphisms [93–95], incomplete andsoft sweeps can leave these spectra relatively undisturbed [91]. Thus, the stochastic dynamicsof neutral polymorphisms could be fundamentally different from random genetic drift, with mostof their frequency changes over time being driven by linked selection, but it would be difficult todiscern these dynamics from drift-dominated evolution based on patterns of neutral geneticdiversity in a single population sample.

If linked selection plays an important role in the dynamics of neutral polymorphism, this couldhave profound implications on population genetic inferences, given that basically all of ourmethods for inferring demography and population structure from population genomic data arebased on the drift-dominant paradigm. We already know that genetic draft can severely misleadthese approaches [7,96] and it is reasonable to assume that linked selection from incompleteand soft sweeps will cause similar problems.



Population Genetics Should Embrace Rapid EvolutionThe appeal of modern population genetics stems in no small part from the elegance andsimplicity of its underlying theoretical models. These models were largely devised in times whendata were limited to measurements of molecular divergences between species and roughestimates of the genetic diversity within populations. Kimura's neutral theory of molecularevolution [97] provided a convincing explanation for the patterns in these data that did notrequire processes more complicated than random genetic drift, purifying selection, and theoccasional selective sweep.

Yet as we have presented in this review, there is accumulating evidence that the standard modelis insufficient to describe the rapid phenotypic evolution often observed in nature and experi-ments. The time-rate scaling of phenotypic evolution strongly suggests that the selection drivingthese phenotypic changes is fluctuating, so that rapid short-term changes do not accumulateinto large long-term changes. The genes underlying these changes then do not fit into any ofthe categories in the standard model (Figure 1), unless we are truly in the limiting situation of theinfinitesimal model, where substantial, ecologically important changes in traits occur withoutsubstantial frequency changes at any polymorphic site in the genome.

Whether most rapidly evolving traits truly are polygenic, so that the infinitesimal model applies, isan open question. We have described examples where the answer is ‘no’–a few genes of largeeffect are responsible, but these may be unrepresentative. There are very few such traits forwhich we know much about the phenotype–genotype map, and understanding phenotype–genotype maps is a difficult and longstanding problem unlikely to see a quick resolution.

Concluding RemarksWith genome sequencing becoming easier and cheaper, we have the opportunity to directlyobserve the essence of evolution: how allele frequencies change over time within a population,as in Figure 4B. With such data we can finally test the key assumptions of our population geneticmodels and study the processes that govern the patterns and dynamics of genetic variation inpopulations (see Outstanding Questions).

Key to achieving this goal will be extensive population sampling across time and space, alongwith measurements of environmental factors, abiotic and biotic, that are potential sources offluctuating selection. These samples need to be taken frequently enough to observe short-termgenetic changes before they are reversed when selection changes in direction. Identifyingthe relevant timescale of genetic change must come from an ecological understanding ofthe selective forces acting on the population and the timescales of variation in the factors thatgive rise to directional selection and rapid evolutionary responses. The first computationalapproaches that will allow us to test directly for variable selection, given temporally high-resolution population genomic data, are already being developed [33,98,99]. We are aboutto enter exciting times for population genetics.

AcknowledgmentsThe authors thank Alan Bergland, Dmitri Petrov, and two anonymous reviewers for critically reading the manuscript and

providing valuable feedback. They also thank Andrew Hendry for sharing the data compilation used in Figure 3. N.G.H. and

S.P.E. were supported by NSF grant DEB-1256719. S.P.E. was also supported by NSF DEB-1353039.

References1. Gillespie, J.H. (1991) The Causes of Molecular Evolution, Oxford

University Press

2. Orr, H.A. (1998) The population genetics of adaptation: thedistribution of factors fixed during adaptive evolution. Evolution52, 935–945

3. Orr, H.A. (2005) The genetic theory of adaptation: a brief history.Nat. Rev. Genet. 6, 119–127

4. Vitti, J.J. et al. (2013) Detecting natural selection in genomic data.Annu. Rev. Genet. 47, 97–120

Outstanding QuestionsHow common is rapid evolution innature and how does this impact thedynamics of genetic variation in apopulation?

Is rapid evolution typically underlain bysubtle frequency changes of manypolymorphisms or do large-effect lociplay an important role as well?

How realistic are the assumptions thatwe can assign a fixed selection coeffi-cient to each mutation that remainsconstant across time and that potentialepistatic interactions can be ignored?

What roles do fluctuating and balanc-ing selection play in the evolutionarydynamics of genetic polymorphisms?

Is random genetic drift truly the domi-nant process in the evolution of com-mon genetic variation or could linkedselection play an important role aswell?

If drift is not the dominant process, howwill this affect our population geneticmachinery for inference of selection,demography, and population structurefrom population genomic data?

How can we leverage the new popula-tion-level datasets to answer thesequestions by studying directly howpolymorphisms change in frequencyover time?



5. Schraiber, J.G. and Akey, J.M. (2015) Methods and models forunravelling human evolutionary history. Nat. Rev. Genet. 16,727–740

6. Boyko, A.R. et al. (2008) Assessing the evolutionary impact ofamino acid mutations in the human genome. PLoS Genet. 4,e1000083

7. Messer, P.W. and Petrov, D.A. (2013) Frequent adaptation andthe McDonald–Kreitman test. Proc. Natl. Acad. Sci. U.S.A. 110,8615–8620

8. Enard, D. et al. (2014) Genome-wide signals of positive selectionin human evolution. Genome Res. 24, 885–895

9. Grant, P.R. and Grant, B.R. (2002) Unpredictable evolution in a30-year study of Darwin's finches. Science 296, 707–711

10. Grant, P.R. and Grant, B.R. (2006) Evolution of character dis-placement in Darwin's finches. Science 313, 224–226

11. Grant, P.R. and Grant, B.R. (2011) Causes of lifetime fitness ofDarwin's finches in a fluctuating environment. Proc. Natl. Acad.Sci. U.S.A. 108, 674–679

12. Koskinen, M.T. et al. (2002) Contemporary fisherian life-historyevolution in small salmonid populations. Nature 419, 826–830

13. Elmer, K.R. et al. (2014) Parallel evolution of Nicaraguan craterlake cichlid fishes via non-parallel routes. Nat. Commun. 5, 5168

14. Carroll, S.P. and Boyd, C. (1992) Host race radiation in thesoapberry bug: natural history with the history. Evolution 46,1052–1069

15. Carroll, S.P. et al. (2001) Genetic architecture of adaptive differ-entiation in evolving host races of the soapberry bug, Jaderahaematoloma. Genetica 112-113, 257–272

16. Reznick, D. et al. (1997) Evaluation of the rate of evolution innatural populations of guppies (Poecilia reticulata). Science 275,1934–1937

17. Gordon, S.P. et al. (2015) Selection analysis on the rapid evo-lution of a secondary sexual trait. Proc. Biol. Sci. 282,20151244

18. Pearse, D.E. et al. (2009) Over the falls? Rapid evolution ofecotypic differentiation in steelhead/rainbow trout (Oncorhyn-chus mykiss). J. Hered. 100, 515–525

19. Hairston, N.G. and Walton, W.E. (1986) Rapid evolution of a lifehistory trait. Proc. Natl. Acad. Sci. U.S.A. 83, 4831–4833

20. Thompson, J.N. (2013) Relentless Evolution, University of Chi-cago Press

21. Hendry, A.P. (2016) Eco-Evolutionary Dynamics, Princeton Uni-versity Press

22. Conover, D.O. andMunch, S.B. (2002) Sustaining fisheries yieldsover evolutionary time scales. Science 297, 94–96

23. Stuart, Y.E. et al. (2014) Rapid evolution of a native speciesfollowing invasion by a congener. Science 346, 463–466

24. Becks, L. et al. (2010) Reduction of adaptive genetic diversityradically alters eco-evolutionary community dynamics. Ecol. Lett.13, 989–997

25. Becks, L. et al. (2012) The functional genomics of an eco-evolu-tionary feedback loop: linking gene expression, trait evolution,and community dynamics. Ecol. Lett. 15, 492–501

26. Cameron, T.C. et al. (2013) Eco-evolutionary dynamics inresponse to selection on life-history. Ecol. Lett. 16, 754–763

27. Savkovi!c, U. et al. (2016) Experimentally induced host-shiftchanges life-history strategy in a seed beetle. J. Evol. Biol. 29,837–847

28. Hairston, N.G. et al. (2005) Rapid evolution and the convergenceof ecological and evolutionary time. Ecol. Lett. 8, 1114–1127

29. Cook, L.M. and Saccheri, I.J. (2013) The peppered moth andindustrial melanism: evolution of a natural selection case study.Heredity 110, 207–212

30. Darimont, C.T. et al. (2009) Human predators outpace otheragents of trait change in the wild. Proc. Natl. Acad. Sci. U.S.A. 106, 952–954

31. Pritchard, J.K. and Di Rienzo, A. (2010) Adaptation–not bysweeps alone. Nat. Rev. Genet. 11, 665–667

32. Rockman, M.V. (2012) The QTN program and the alleles thatmatter for evolution: all that's gold does not glitter. Evolution 66,1–17

33. Wellenreuther, M. and Hansson, B. (2016) Detecting polygenicevolution: problems, pitfalls, and promises. Trends Genet. 32,155–164

34. Hairston, N.G. et al. (2001) Natural selection for grazer resistanceto toxic cyanobacteria: evolution of phenotypic plasticity? Evo-lution 55, 2203–2214

35. Ellner, S.P. et al. (2011) Does rapid evolution matter? Measuringthe rate of contemporary evolution and its impacts on ecologicaldynamics. Ecol. Lett. 14, 603–614

36. Conover, D.O. and Schultz, E.T. (1995) Phenotypic similarity andthe evolutionary significance of countergradient variation. TrendsEcol. Evol. 10, 248–252

37. Jones, F.C. et al. (2012) The genomic basis of adaptive evolutionin threespine sticklebacks. Nature 484, 55–61

38. Lescak, E.A. et al. (2015) Evolution of stickleback in 50 years onearthquake-uplifted islands. Proc. Natl. Acad. Sci. U.S.A. 112,E7204–E7212

39. Pascoal, S. et al. (2014) Rapid convergent evolution in wildcrickets. Curr. Biol. 24, 1369–1374

40. Levy, S.F. et al. (2015) Quantitative evolutionary dynamics usinghigh-resolution lineage tracking. Nature 519, 181–186

41. Kao, K.C. and Sherlock, G. (2008) Molecular characterization ofclonal interference during adaptive evolution in asexual popula-tions of Saccharomyces cerevisiae. Nat. Genet. 40, 1499–1504

42. Lang, G.I. et al. (2013) Pervasive genetic hitchhiking and clonalinterference in forty evolving yeast populations. Nature 500,571–574

43. McDonald, M.J. et al. (2016) Sex speeds adaptation by alteringthe dynamics of molecular evolution. Nature 531, 233–236

44. Bell, G. and Gonzalez, A. (2009) Evolutionary rescue can preventextinction following environmental change.Ecol. Lett. 12, 942–948

45. Burke, M.K. et al. (2010) Genome-wide analysis of a long-termevolution experiment with Drosophila. Nature 467, 587–590

46. Turner, T.L. et al. (2011) Population-based resequencing ofexperimentally evolved populations reveals the genetic basis ofbody size variation in Drosophila melanogaster. PLoS Genet. 7,e1001336

47. Orozco-terWengel, P. et al. (2012) Adaptation of Drosophila to anovel laboratory environment reveals temporally heterogeneoustrajectories of selected alleles. Mol. Ecol. 21, 4931–4941

48. Franssen, S.U. et al. (2015) Patterns of linkage disequilibrium andlong range hitchhiking in evolving experimental Drosophila mel-anogaster populations. Mol. Biol. Evol. 32, 495–509

49. Jalvingh, K.M. et al. (2014) Genomic changes under rapid evo-lution: selection for parasitoid resistance. Proc. Biol. Sci. 281,20132303

50. Martins, N.E. et al. (2014) Host adaptation to viruses relies on fewgenes with different cross-resistance properties. Proc. Natl.Acad. Sci. U.S.A. 111, 5938–5943

51. Messer, P.W. and Petrov, D.A. (2013) Population genomics ofrapid adaptation by soft selective sweeps. Trends Ecol. Evol. 28,659–669

52. Gingerich, P.D. (1983) Rates of evolution: effects of time andtemporal scaling. Science 222, 159–161

53. Hendry, A.P. and Kinnison, M.T. (1999) Perspective: the pace ofmodern life: measuring rates of contemporary microevolution.Evolution 53, 1637–1653

54. Gingerich, P.D. (2001) Rates of evolution on the time scale of theevolutionary process. Genetica 112-113, 127–144

55. Bell, G. (2010) Fluctuating selection: the perpetual renewal ofadaptation in variable environments. Philos. Trans. R. Soc. Lond.B Biol. Sci. 365, 87–97

56. Uyeda, J.C. et al. (2011) The million-year wait for macroevolu-tionary bursts. Proc. Natl. Acad. Sci. U.S.A. 108, 15908–15913

57. Kingsolver, J.G. et al. (2012) Synthetic analyses of phenotypicselection in natural populations: lessons, limitations and futuredirections. Evol. Ecol. 26, 1101–1118

58. Bulmer, M.G. (1974) A statistical analysis of the 10-year cycle inCanada. J. Anim. Ecol. 43, 701–718

59. Turchin, P. (2003) Complex Population Dynamics: A Theoretical/Empirical Synthesis, Princeton University Press



60. Hairston, N.G. et al. (1996) Overlapping generations: the storageeffect and the maintenance of biotic diversity. In In PopulationDynamics in Ecological Space and Time (Rhodes, O.E. et al.,eds), pp. 109–145, University of Chicago Press

61. Caceres, C.E. (1998) Seasonal dynamics and interspecific com-petition in Oneida Lake Daphnia. Oecologia 115, 233–244

62. Hairston, N.G. and Dillon, T.A. (1990) Fluctuating selection andresponse in a population of freshwater copepods. Evolution 44,1796–1805

63. Seger, J. and Brockmann, H.J. (1987) What is bet-hedging?Oxford Surv. Evol. Biol. 4, 182–211

64. Felsenstein, J. (1976) The theoretical population genetics ofvariable selection and migration. Annu. Rev. Genet. 10, 253–280

65. Dempster, E.R. (1955) Maintenance of genetic heterogeneity.Cold Spring Harb. Symp. Quant. Biol. 20, 25–32

66. Barton, N.H. and Turelli, M. (1989) Evolutionary quantitativegenetics: how little do we know? Annu. Rev. Genet. 23, 337–370

67. Chesson, P.L. and Warner, R.R. (1981) Environmental variabilitypromotes coexistence in lottery competitive-systems. Am. Nat.117, 923–943

68. Ellner, S. and Hairston, N.G. (1994) Role of overlapping gener-ations in maintaining genetic variation in a fluctuating environ-ment. Am. Nat. 143, 403–417

69. Hedrick, P.W. (1995) Genetic polymorphism in a temporallyvarying environment: effects of delayed germination or diapause.Heredity 75, 164–170

70. Ellner, S.P. et al. (1999) The roles of fluctuating selection andlong-term diapause in microevolution of diapause timing in afreshwater copepode. Evolution 53, 111–122

71. Yi, X. and Dean, A.M. (2013) Bounded population sizes, fluctu-ating selection and the tempo and mode of coexistence. Proc.Natl. Acad. Sci. U.S.A. 110, 16945–16950

72. Gulisija, D. and Kim, Y. (2015) Emergence of long-term balancedpolymorphism under cyclic selection of spatially variable magni-tude. Evolution 69, 979–992

73. Gulisija, D. et al. (2016) Phenotypic plasticity promotes balancedpolymorphism in periodic environments by a genomic storageeffect. bioRxiv http://dx.doi.org/10.1101/038497

74. Sellis, D. et al. (2011) Heterozygote advantage as a naturalconsequence of adaptation in diploids. Proc. Natl. Acad. Sci.U.S.A. 108, 20666–20671

75. Ghosh, R. et al. (2012) Natural variation in a chloride channelsubunit confers avermectin resistance in C. elegans. Science335, 574–578

76. Leffler, E.M. et al. (2012) Revisiting an old riddle: what determinesgenetic diversity levels within species? PLoS Biol. 10, e1001388

77. Leffler, E.M. et al. (2013) Multiple instances of ancient balancingselection shared between humans and chimpanzees. Science339, 1578–1582

78. Lenz, T.L. et al. (2013) Evaluating patterns of convergent evolu-tion and trans-species polymorphism at MHC immunogenes intwo sympatric stickleback species. Evolution 67, 2400–2412

79. Rodríguez-Trelles, F. et al. (2013) Genome-wide evolutionaryresponse to a heat wave in Drosophila. Biol. Lett. 9, 20130228

80. Pujolar, J.M. et al. (2014) Genome-wide single-generation sig-natures of local selection in the panmictic European eel. Mol.Ecol. 23, 2514–2528

81. Bergland, A.O. et al. (2014) Genomic evidence of rapid andstable adaptive oscillations over seasonal time scales in Dro-sophila. PLoS Genet. 10, e1004775

82. Charlesworth, B. (2015) Causes of natural variation in fitness:evidence from studies of Drosophila populations. Proc. Natl.Acad. Sci. U.S.A. 112, 1662–1669

83. Chakraborty, M. and Fry, J.D. (2015) Evidence that environmen-tal heterogeneity maintains a detoxifying enzyme polymorphismin Drosophila melanogaster. Curr. Biol. 26, 219–223

84. Sawyer, S.A. and Hartl, D.L. (1992) Population genetics of poly-morphism and divergence. Genetics 132, 1161–1176

85. McDonald, J.H. and Kreitman, M. (1991) Adaptive protein evo-lution at the Adh locus in Drosophila. Nature 351, 652–654

86. Charlesworth, J. and Eyre-Walker, A. (2008) The McDonald–Kreitman test and slightly deleterious mutations. Mol. Biol. Evol.25, 1007–1015

87. Fay, J.C. (2011) Weighing the evidence for adaptation at themolecular level. Trends Genet. 27, 343–349

88. Bustamante, C.D. et al. (2005) Natural selection on protein-coding genes in the human genome. Nature 437, 1153–1157

89. Keightley, P.D. and Eyre-Walker, A. (2007) Joint inference of thedistribution of fitness effects of deleterious mutations and popu-lation demography based on nucleotide polymorphism frequen-cies. Genetics 177, 2251–2261

90. Eyre-Walker, A. and Keightley, P.D. (2009) Estimating the rate ofadaptive molecular evolution in the presence of slightly deleteri-ous mutations and population size change. Mol. Biol. Evol. 26,2097–2108

91. Coop, G. and Ralph, P. (2012) Patterns of neutral diversity undergeneral models of selective sweeps. Genetics 192, 205–224

92. Garud, N.R. et al. (2015) Recent selective sweeps in NorthAmerican Drosophila melanogaster show signatures of softsweeps. PLoS Genet. 11, e1005004

93. Gillespie, J.H. (2000) Genetic drift in an infinite population. Thepseudohitchhiking model. Genetics 155, 909–919

94. Neher, R.A. and Shraiman, B.I. (2011) Genetic draft and quasi-neutrality in large facultatively sexual populations. Genetics 188,975–996

95. Neher, R.A. (2013) Genetic draft, delective interference, andpopulation genetics of rapid adaptation. Annu. Rev. Ecol. Evol.Syst. 44, 195–215

96. Schrider, D. et al. (2016) Effects of Linked Selective Sweeps onDemographic Inference and Model Selection, Cold Spring Har-bor Laboratories

97. Kimura, M. (1983) The Neutral Theory of Molecular Evolution,Cambridge University Press

98. Berg, J.J. and Coop, G. (2014) A population genetic signal ofpolygenic adaptation. PLoS Genet. 10, e1004412

99. Gompert, Z. (2015) Bayesian inference of selection in a hetero-geneous environment from genetic time-series data. Mol. Ecol.25, 121–134

100. Hendry, A.P. et al. (2008) Human influences on rates of pheno-typic change in wild animal populations. Mol. Ecol. 17, 20–29

101. Quintero, I. and Wiens, J.J. (2013) Rates of projected climatechange dramatically exceed past rates of climatic niche evolutionamong vertebrate species. Ecol. Lett. 16, 1095–1103


Review Can Population Genetics Adapt to Rapid...

Documents

Transcript of Review Can Population Genetics Adapt to Rapid...