Entertainment & Media


In addition to the well-studied evolutionary parameters of (1) phenotype-fitness covariance and (2) the genetic basis of phenotypic variation, adaptive evolution by natural selection requires that (3) fitness variation is effected by heritable
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  ORIGINAL ARTICLE doi:10.1111/j.1558-5646.2010.01200.x A TEST FOR THE GENETIC BASIS OF NATURALSELECTION: AN INDIVIDUAL-BASEDLONGITUDINAL STUDY IN ASTREAM-DWELLING FISH Michael B. Morrissey 1 , 2 , 3 and Moira M. Ferguson 1 , 4 1 Department of Integrative Biology, University of Guelph, Guelph, Ontario N1G 2W1, Canada  2 Corresponding author, current address: Institute of Evolutionary Biology, University of Edinburgh EH9 3JT, United Kingdom 3 E-mail: michael.morrissey@ed.ac.uk  4 E-mail: mmfergus@uoguelph.ca Received June 11, 2010Accepted October 20, 2010 In addition to the well-studied evolutionary parameters of (1) phenotype-fitness covariance and (2) the genetic basis of phenotypicvariation, adaptive evolution by natural selection requires that (3) fitness variation is effected by heritable genetic differencesamongindividualsand(4)phenotype-fitnesscovariancesmustbe,atleastinpart,underlainbygeneticcovariances.Theselattertworequirements for adaptive evolutionary change are relatively unstudied in natural populations. Absence of the latter requirementscould explain stasis of apparently directionally selected heritable traits. We provide complementary analyses of selection andvariation at phenotypic and genetic levels for juvenile growth rate in brook charr  Salvelinus fontinalis  in Freshwater River,Newfoundland, Canada. Contrary to the vast majority of reports in fish, we found very little viability selection of juvenile bodysize. Large body size appears nonetheless to be selectively advantageous via a relationship with early maturity. Genetic patternsin evolutionary parameters largely reflected phenotypic patterns. We have provided inference of selection based on longitudinaldata, which are uncommon in high fecundity organisms. Furthermore we have provided a practicable framework for furtherstudies of the genetic basis of natural selection. KEY WORDS:  Fitness, genetic covariance, Robertson-Price identity, selection response. Adaptive evolution by natural selection is expected to occur whenfitness covaries with genetically based phenotypic differencesamong individuals. Empirical studies of phenotypic microevo-lution have determined that traits that may be under selectionare indeed generally underlain at least in part by geneticallybased differences among individuals (Mousseau and Roff 1987;Roff 2002) and that traits that we expect might be under se-lection do in fact covary systematically with fitness (Kingsolveret al. 2001; Siepielski et al. 2009). Thus, two critical require-ments for adaptive microevolution are very often met in naturalpopulations. Two other critical requirements for adaptive evolu-tionary responses to selection are (1) adaptive evolution cannotoccur unless fitness variation is underlain by genetic differencesamongindividualsand(2)ageneticcovariancemustunderliephe-notypic relationships between traits and fitness (Morrissey et al.2010). Ultimately, evolutionary change will occur if the distri-bution of genotypes is nonrandom with respect to fitness, andsuch a pattern can be directly assessed by the genetic covarianceof a trait with (relative) fitness (Robertson 1966; Price 1970).Such a genetic covariance is necessarily zero unless genetically 1037 C   2010 The Author(s).  Evolution  C   2010 The Society for the Study of Evolution.  Evolution 65-4: 1037–1047   M. B. MORRISSEY AND M. M. FERGUSON basedvariationforfitnesssegregatesinapopulation,andthusthisquantity is equally critical to adaptive evolution. These latter crit-ical conditions for adaptive evolution have been far less studiedin nature (Morrissey et al. 2010).The genetic basis of fitness variation, and consequently theopportunity for genetic covariances to exist among potentially se-lected traits and fitness, has been far less studied than the geneticbasis of other traits. Thus, whether the absence (or presence) of heritable genetic variation in fitness may typically constrain (orfacilitate) adaptive evolution in natural populations is unclear.Nonetheless, some studies have reported substantial additive ge-netic variances of fitness-related traits such as life-history traits(Meril¨a and Sheldon 1999, 2000). Heritabilities of these traits aregenerally low, but because of large environmentally induced vari-ance rather than a lack of genetically based variation (Price andSchluter 1991; Meril¨a and Sheldon 1999, 2000; Coltman et al.2005; but see Teplitsky et al. 2009).Quantitatively, the genetic covariance between fitness anda trait hypothesized to be under selection can summarize thepotential for adaptive phenotypic evolution by natural selection(Robertson 1966), without bias from environmental covarianceor unmeasured traits (Morrissey et al. 2010). This genetic covari-ance will be zero unless additive genetic variation exists both forfitness and for the traits upon which we hypothesize selectionacts. Furthermore, modern methods of estimating quantitative ge-netic parameters such as genetic covariances (Kruuk 2004) arenot expected to be biased by environmental sources of covari-ance. Genetic covariances are notoriously difficult to estimatewith precision (Lynch and Walsh 1998), especially when envi-ronmental variances are large, as will typically be the case forfitness (Meril¨a and Sheldon 1999). Nonetheless this covariance isthe critical parameter in adaptive microevolution and we shouldtherefore attempt its estimation. We refer to an analysis of thegenetic covariance between a trait and fitness as a test for the“genetic basis of selection.” Of course selection is a phenotypicphenomenon only. We perpetuate this terminology (from the titleof Lerner’s seminal 1958 book in animal breeding) as a meansof emphasizing the fact that a shift in the distribution of pheno-typic values in a population will only be transmitted to futuregenerations if it is underlain by a shift in breeding values.The quantification of the genetic basis of trait–fitness rela-tionships provides an alternate methodology to the well-knownbreeder’s equation for predicting the course of evolution in theshort term. The breeder’s equation (  R = h 2 S  , where  R  is the re-sponse to selection,  h 2 is the heritability or proportion of traitvariance that is attributable to heritable genetic effects, and  S   isthe selection differential) predicts evolutionary change withoutdirectly assaying whether fitness and genetic differences amongindividuals covary, by assuming that all covariance of a trait withfitnessiscausedentirelybythattrait(Queller1992;Rausher1992;Hadfield 2008), or in multivariate analyses, that all covariance-inducing traits or environmental factors have been accountedfor. Analyses of heritable and apparently selected traits in long-runningbirdandlargemammalsystems(Kruuketal.2000,2002),and in controlled experimental plant systems (Stinchcombe et al.2002), have revealed that such traits do not necessarily covarygenetically with fitness. More specifically, these examples revealthat the breeder’s equation’s assumption that all sources of co-variance between a focal trait and fitness are adequately modeledwas inappropriate, at least for these traits and populations. Un-fortunately, the generality of this finding is currently not testabledue to a paucity of assays of the genetic covariance of traits withfitness.The body size of fish provides an ideal trait for joint phe-notypic and genetic analysis of selection and prediction of evo-lutionary change. The indeterminate growth of fish provides sub-stantial opportunity for environmental influences on body size,but nonetheless, size and other traits are very often heritable(e.g., Serbezov et al. 2010; Wilson et al. 2003a, and reviewed byCarlsonandSeamons2008),andalsoveryoftenselected(Hendryet al. 2003; Carlson et al. 2004, 2008). Within fish, stream-dwellingsalmonidsareparticularlyuseful,becauseintensivesam-pling of some populations is possible (e.g., Carlson et al. 2008),and so robust studies can be designed. In particular, brook charr, Salvelinusfontinalis ,inFreshwaterRiver,Newfoundland,Canada(Fig. 1) provide a promising study system in which to test forwhether apparent patterns of natural selection have a genetic ba-sis. Freshwater River is free of anthropogenic influences such asexploitation or interactions with introduced species. FreshwaterRiver is furthermore a simple system containing only one fishspecies and is small enough to allow intensive sampling. PreviousworkonFreshwaterRiverbrookcharrhassuggestedthatlargeju-venilesizeconfersaviabilityadvantage(Wilsonetal.2003b),andthatlargebodysizeisassociatedwithearlymaturity(Wilsonetal.2003b;MorrisseyandFerguson2009),whichisalsoahighfitnessstate(MorrisseyandFerguson2009).However,thepreviousstudyof selection (Wilson et al. 2003b) reported an inconsistent patternof co-occurrence of selection and genetic variation for body sizeatdifferentages.Thereforethepotentialforadaptiveevolutionre-mains unclear. Freshwater River brook charr thus provide a studysystem in which evaluation of the genetic relationship betweenbody size and fitness proxies will provide a better description of the potential for adaptive evolution of body size.Here, we test for genetically based variation in alevin and juvenile body size and in fitness proxies (age at first maturationand estimates of survival to adulthood) in Freshwater River brook charr. We test for both phenotypic and genetically based covaria-tionbetweenbodysizeandfitnessproxiestodeterminethepoten-tial for adaptive evolution of alevin and juvenile body size in thispopulation. We use estimates of the genetic covariances between 1038  EVOLUTION  APRIL 2011  GENETIC BASIS OF SELECTION 6664626058565452454749515355Longitude ( °  West)    L  a   t   i   t  u   d  e   (             °     N  o  r   t   h   ) N0m500m1000m 0m 500m 1000m 1500m 2000m Figure 1.  The location of the in Freshwater River, Cape Race,Newfoundland, showing sampling locations (gray). Diagonal hashmarks indicate distance from the lower-most point in the studyarea. The inset map denotes the location of Freshwater Riverwithin eastern Canada with an arrow (UTM North 5168220 East329991). bodysizeandproxiesoffitnesstoinferpatternsofselectionatthegenetic level in this natural population. Importantly, we conductallofouranalysesonindividual-basedlongitudinaldata.Analysisof this kind of data, although particularly challenging to collect innonterrestrial organisms, avoids a number of technical and inter-pretive limitations, especially uncertainties in back-calculationsof size from annual growth rings in hard structures (Wilson et al.2009) and assumptions that multiple cohorts experience similarenvironmental conditions (Endler 1986). Methods SAMPLING We sampled the 2002 cohort of Freshwater River brook charrseven times by single-pass electrofishing between June 2003 andSeptember 2006. We concentrated our sampling effort on thepreviously studied lower 2 km of the river, (e.g., Wilson et al.2003a,2004;HutchingsandGerber2002),butexcludingasectionof fast-flowing water that contains very few fish (Fig. 1). At eachsampling occasion, we sampled between 541 and 860 fish witha range of fork lengths so as to target the entire size distributionof brook trout from the 2002 cohort. This approach necessarilyresulted in the inclusion of individuals from cohorts from otheryears, especially in the later samples, but would not have causedconfusion between individuals of different life cycle stages. Therange of fork length that we targeted was based on length-at-agekeys in Hutchings (1993) and Wilson et al. (2003b). We recordedeach fish’s body size (fork length, FL, in millimeter), sex andlocation in the river.We obtained tissue samples for molecular analyses for allsampled fish. Samples were genotyped at up to 19 polymorphicmicrosatellite loci to obtain molecular data with which we couldidentify individuals and investigate patterns of reproductive suc-cess. Polymorphism in the marker set ranged from number of observed alleles between 2 and 15 and with expected heterozy-gosities between 0.41 and 0.82. We tracked individual identity(and pedigree relationships, see below) using these molecularmarker data. Briefly, all assignments were made based on geno-typic identities of 10 − 6 or lower. See Morrissey and Ferguson(2009) for specific details of our individual assignmentprocedure. PEDIGREE RECONSTRUCTION We used the Markov Chain Monte Carlo (MCMC)-based sib-ship partitioning algorithm implemented in the program  COLONY (Wang 2004) to reconstruct the most-likely sibship partition of the individuals that we captured in the first three samples of the2002 cohort. Briefly, this program uses an MCMC approach toiteratively produce increases in the likelihood of a sibship parti-tion;thealgorithmisrunformany(millionsof)iterationsuntilthelikelihood cannot be further increased. To allow convergence inour large dataset, we modified  COLONY  on the author’s advice (J.Wang,pers.Comm.)torunfor10 4 iterationsforeachof150steps,where in each step the annealing temperature was reduced by 1%.This resulted in approximately 100-fold more time for the algo-rithmtolocatesibshipsthatmaximizetheoveralllikelihoodofthesibship partition, relative to default parameters. Despite the factthat half-sib structure likely exists within cohorts of brook charr(e.g., Blanchfield et al. 2003), we partitioned our sample of the2002FreshwaterRivercohortintofullsibships.Thissimplerpedi-gree structure has proven useful for quantitative genetic analysesin several previous studies (Wilson et al. 2003a,b; Th´eriault et al.2007), and more broadly, it is clear that modest rates of pedigreeerror do not necessarily hinder the interpretation of quantitativegeneticparameters(Morrisseyetal.2007;DiBattistaetal.2009). PHENOTYPIC SELECTION We treated the entire study area in Freshwater River as a singlepopulation in analyses of selection. This treatment was appropri-ate because we have estimated that at least 20% of individualsmove between the upper and lower portions of the study area EVOLUTION  APRIL 2011  1039  M. B. MORRISSEY AND M. M. FERGUSON (Morrissey and Ferguson 2010) during their lives, and thus manyindividuals will experience selection in both areas. We tested forselection on the mean and variance of age-specific alevin and juvenile sizes, as well as standardized juvenile size, followingEndler (1986). We determined statistical significance of each se-lection coefficient using randomization tests. For selection onlength-at-age during the first through third intervals of the studywe estimated coefficients of selection on the mean from the re-capture data as i  =  (¯  z  x  + t   − ¯  z  x  ) · r  (  z  x  ,  z  x  + t  )( n  x  + t  /  n  x  + t  ) ,  (1)where the summations are from the interval of the study underconsideration through the third interval, where ¯  z  x   represents themean size at age  x  , where ¯  z  x  + t   represents the mean body size atage  x  ofindividualscapturedattime  x  + t  ,andequivalently,wherevalues of   n  are the sample sizes of individuals recaptured at times  x  + t  , and where  r  ( u  x  ,  u  x  + t  ) is the correlation between size at ageobserved at ages  x   and at  x   +  t  . This formula thus corrects forthe potential for selection on correlated size-at-age traits to biasestimatesofselectiononsize-at-ageduringearlierintervals,whileretaining the statistical power of multiple samples collected overseveral intervals. Estimates of   j  (i.e., of selection on the variance)were calculated equivalently to the calculation of   i  presented inequation (1), above. We used generalized additive model (GAM)-based nonparametric visualizations of the relationships betweenphenotypic body size and fitness-related traits to further explorethe form of selection on body size. All GAMs were fitted usingthe mgcv function in R (Wood 2006). In the univariate analysesapplied here, these GAMs are equivalent to the spline-regressionmethodologies advocated by Schluter (1988) and Brodie et al.(1995). GENETIC VARIATION IN BODY SIZE ANDFITNESS-RELATED TRAITS To test for additive genetic variances, we fitted univariate animalmodels of alevin body size, juvenile body sizes at ages 1 and 1.67(assessing age in years from fertilization; that is, size in October2003 and June 2004), and standardized juvenile body size. Wecalculated standardized juvenile body size by standardizing bodysizes at ages 1 and 1.67 to means of zero and variances of 1 andthenaveragingstandardizedlengthforthoseindividualsthatweremeasuredatbothages.Wefitted similar univariateanimal modelsfor each of the six fitness-related traits. The univariate modelstook the form y = u + p + Za + e ,  (2)where  y  is a vector of individual phenotypes,  u  contains the pop-ulation mean trait value,  p  contains the effect of location withinthe sampling area (i.e., upper or lower),  Z  is an incidence ma-trix relating phenotypic values in  y  to individual additive geneticeffects in  a , and  e  is a vector of random errors.We tested for the most parsimonious model to describe theontogeny of genetic variation in body size by fitting four compet-ing models. This exercise allowed us to provide the best descrip-tion of additive genetic effects on body size over time using thefewest necessary parameters. We first fitted three random regres-sion models: one where individual additive genetic effects werefixed over time, one where additive genetic effects varied linearlywith age, and the third where individual additive genetic effectsvaried linearly with the phenotypic standard deviation of bodysize. The random regression models took the form y t  = u t + tZa + e ,  (3)where  y t  are phenotypes indexed by time  t ,  u t  are time-specificmean phenotypic values, and  t  relates additive genetic effects tothe times at which phenotypic records were taken.Finally, we fit a model with independent but correlated addi-tive effects for each size-at-age trait. This full multivariate modeltook the form of equation (2), above, except that the vectors inequation(2)nowbecamen × 3matricestoaccommodatethethreelength-at-age phenotypes. In all models we fit a covariance struc-turethatdidnotimposeanypatternoneithertheresidualvarianceor the residual covariance among sizes-at-age. All animal modelswere fitted using ASRML (Gilmour et al. 2002). GENETIC COVARIANCES BETWEEN BODY SIZE ANDFITNESS COMPONENTS We tested the genetic covariances between body size and fitnesscomponents by adding the fitness components to the most par-simonious random regression model of the ontogeny of additivegenetic effects on body size. We again constrained neither theresidual variance nor the covariance between and among the fit-ness and size-at-age traits. We compared the models to reducedmodels that constrained the genetic covariances to zero, wherebyproviding quantitative tests of the hypotheses that a genetic co-variance exists between body size and the fitness proxies.We used GAMs to test for potential nonlinear relationshipsbetween genetic dispositions for body size and fitness traits. Weobtained best linear unbiased predictions of breeding values foralevinandjuvenilebodysizefromthemostparsimoniousrandom-regression model of the ontogeny of additive genetic effects onbody size. We predicted each individual breeding value from ananimal model analysis from which that individual’s phenotypicrecord had been deleted. Individuals belonging to sibships of sizeone were excluded from the GAM regression analyses, becausein the absence of information from the phenotypes of relatives,unbiased prediction of the breeding values of such individualsis not possible. We applied GAM regressions to determine the 1040  EVOLUTION  APRIL 2011  GENETIC BASIS OF SELECTION Table 1.  Growth rate statistics of Freshwater River brook charr from the 2002 cohort. Summary statistics for (A) all individuals collectedin each sample, (B) size-at-recapture of individuals that were captured in the first sample and thus certainly belonged to the 2002 cohort,and (C) those individuals used in subsequent analyses. Maximum values in (C) were determined by comparison of parts (A) and (B). Sample 1 2 3 4 5 6 7age 0.67 1 1.67 2 2.67 3 4(A) Total capturesMinimum (mm) 22 37 46 63 70 79 95Maximum (mm) 38 74 82 106 108 123 203Mean (mm) 31.4 56.3 67.7 87.0 89.9 101.9 121.6 sd   2.8 5.5 6.6 9.2 8.3 10.4 17.8 n  764 768 860 787 586 606 541(B) Recaptures from sample 1Minimum (mm) na 46 51 73 84 87 100Maximum (mm) na 67 78 100 99 114 182Mean (mm) na 56.6 68.4 84.3 91.4 101.1 122.6 sd   na 4.1 5.0 6.5 3.9 8.0 25.1 n  na 70 68 43 17 15 10(C) Individuals included in subsequent analysesMinimum (mm) 22 37 46 - - - -Maximum (mm) 38 71 79 - - - -Mean (mm) 31.4 56.1 67.6 - - - - sd   2.8 5.2 6.8 - - - - n  764 757 856 - - - -relationshipbetweenthesepredictedgeneticdispositionsforbodysize and fitness proxies. Results SUMMARY STATISTICS OF RECAPTURE DATA ANDTHE RECONSTRUCTED PEDIGREE We sampled and obtained genotypic data from 764 alevin (age0.67 years from fertilization, developmental terminology follow-ing Balon 1975) brook charr in June 2003. We sampled 757 and856 juveniles with body lengths that allowed us to ascertain thatthey belonged to the 2002 cohort (see length-at-age distributionsin Table 1) in October of 2003 (age 1) and June of 2004 (age1.67), respectively. We captured 182 individuals as alevins thatwere subsequently recaptured in the second through fourth sam-plesofthe2002cohort,allowingustomeasureselectiononalevinbody size. We were able to measure selection on juvenile bodysize based on recaptures of 226 and 118 individuals that hadbeen captured as juveniles in October of 2003 and June of 2004,respectively.Of individuals that we captured either as alevins or juveniles,we recaptured 175 in October 2004, allowing us to determinewhether they first matured at age 2. In 2004, 2005, or 2006 werecaptured 162 individuals in a sexually mature state that we hadpreviously captured as alevins or juveniles. We recaptured 349individuals from the first three samples of the cohort as adults(Table 2).The most likely sibship partition of the 2005 unique individ-uals sampled as alevins and juveniles contained 639 full sibships.Sibship sizes ranged from one to nine individuals with a meansize of 3.1 individuals. PHENOTYPIC SELECTION We did not detect any significant associations between alevinor juvenile body sizes and survival during the first three in-tervals of the study (Table 3). Relatively large age 1.67 juve-niles (i.e., fast-growing) were significantly more likely than small Table 2.  Counts of data for fitness traits, where  n t   is the totalnumber of individuals phenotyped and  n  s  is the number of indi-vidualsknowntobelongtothehigh-fitnessclass(i.e.,reproducingearly, reproducing, or surviving to adulthood). Full studyareaLowersectionUppersectionFitness trait n t   n s  n t   n s  n t   n s Survival 1-2 687 67 397 47 290 20Survival 2-3 759 151 440 92 319 59Survival 3-4 836 114 439 67 397 47Early maturation 175 73 110 37 65 36Maturation 2005 162 1098 85 903 77Long-term survival 2005 349 1098 203 903 146 EVOLUTION  APRIL 2011  1041
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks