Comparing Mutational Variabilities
Total Page:16
File Type:pdf, Size:1020Kb
Copyright 0 1996 hv the Genetics Sorirty of America Comparing Mutational Variabilities David Houle, * Bob Morikawa"J and Michael Lyncht *Department of Zoology, University of Toronto, Toronto, Ontario M5S 3G5, Canada and tDepartment of Biology, University of Oregon, Eugene, Oregon 97403-1 210 Manuscript received November 21, 1995 Accepted for publication April 15, 1996 ABSTRACT We have reviewed the available data on I&, the amount of genetic variation in phenotypic traits produced each generation by mutation. We use these data to make several qualitative tests of' the mutation-selection balance hypothesis for the maintenance of genetic variance (MSB) . To compare VM values, we use three dimensionless quantities: mutational heritability, vtf/&..;the mutational coefficient of variation, Cytl;and the ratio of the standinggenetic variance to Ytf, VJ V,,.Since genetic coefficients of variation for life history traits are larger than those for morphological traits, we predict that under MSB, life history traits should also have larger CV,,. This is confirmed; life history traits have a median CV,, value more than six times higher than that formorphological traits. V,:/V, approximates the persistence time of mutations under MSB in an infinite population. In order for MSB to hold, VJV, must be small, substantially less than 1000, and life history traits should have smaller values than morpho- logical traits. V(;/ averages about 50 generations for life history traits and 100 generations for morpho- logical traits. These observations are all consistent with the predictions of a mutation-selection balance model UTATION, as the source of all genetic variation, maintains standing variation. Under pleiotropic muta- M is ultimately responsible for both variation and tion-selection balance models, much of the variation adaptation. A long-standing, fundamentaldebate in segregating may be unconditionally deleterious ( KON- evolutionary genetics concerns the strengthof the rela- DRASHOV and TURELLI1992), andtherefore not avail- tionship between mutation and variation. The two most able to promote adaptation. Underbalancing selection, plausible mechanisms for the maintenance of variation the same processes that maintain variation may retard are mutation-selection balance and various models of the use of that variation in promoting adaptation. An balancing selection (BARTON1990) . With mutation-se- understanding of mutation is therefore required to an- lection balance, a steady input of mutation is necessary swer both the question of what maintains genetic vari- to maintain genetic variance, so we expect a positive ance and the question of what determines the rate of correlation between mutation and variation. With bal- response to selection. ancing selection, mutation need only produce alleles For quantitative traits the parameter V,, the increase leading to such polymorphisms infrequently, so muta- in genetic variance due to a single generation of muta- tion and variation may be only weakly related. tion, is important in models of both response to selec- An equally long-standing question concerns the rela- tion and maintenance of genetic variation (LYNCH tionship between mutation and adaptation. On the one 1988; BARTONand TIJRELLI1989). Even if all the varia- hand, a popular model of adaptation assumes that the tion is unconstrained by conflicting selection pressures, standing variance in a populationis the principal source further response will be limited by Vw if directional of the response to selection (e.g., LANDE 19'79). Under selection is strong and prolonged for more than about this assumption, it is the amountof variation that limits N,. generations, where N,. is the effective population size the rate of adaptation. This justifies the widespread use (HILL 1982).I& thus may be of particular importance of quantitative genetics in evolutionary biology. The as human-mediated changes in the environment chal- extreme alternative view is that the alleles that poten- lenge a wide variety of species, particularly those with tially give rise to adaptations do notnormally segregate the smallest populations ( LYNCHand LANDE 1992). In in populations. In that case, the mutational processes addition, v,determines the rate of divergence in neu- that give rise to advantageous genotypes would limit tral models of phenotypic evolution ( LANDE 1976b; the rate of evolution, regardless of the mechanism that CHAKRABORTYand NEI 1982;LYNCH and HILL 1986; LYNCH1994) . Corresponding author: David Houle, Department of Zoology, Univer- To compare thevariability of different traits, previous sity of Toronto, 25 Harhord St., Toronto, OntarioM5S 3G5, Canada. E-mail: [email protected] reviews of have standardized estimates with the envi- ' Present oddress: CAP-HAFF, 74 MFI, P.O. Box 15665, West Palm ronmental variance of the trait, V, ( LANDE 1976a; HILL Beach, FL 33416. 1982; LYNCH1988). V,/V,: is the rate of increase in Grnrtics 143 1467-1483 (July. 1996) 1468 D. Houle, B. Morikawa and M. Lynch heritability in an initially homozygous population, and explain the higher CVAs of life history traits (HOULE therefore is called the mutational heritability. LYNCH'S 1992) . First, a larger proportion of the genome could (1988) review confirmed previous generalizations that affect the average life history trait than the average mor- the average V,/ value is about 1 x 10 -'. LYNCH also phological trait. Every locus in the genomemust poten- identified considerable variation about this figure, al- tially affect fitness, and life history traits that summarize though much of this may be due to sampling error. major components of fitness must therefore also be Mutational heritability is widelyused in models of quan- affected by large numbers of loci ( HOULE1991 ) . Sec- titative traits under stabilizing selection to eliminate V' ond, loci witheffects on life historytraits may be particu- as a free parameter. larly likely to have balanced polymorphisms, for exam- An alternative class of measures of mutational vari- ple, because of genotype-environment interactions or ability are those where VMvalues are standardized by antagonistic pleiotropy (ROSE1982; GILLESPIEand TUR- the trait mean, such as the mutational coefficient of ELLI 1989). Third, selection directly on trait variance variation, CV, = 100 X @,/x, where x is the trait could favor modifiers that reduce the variance in mor- mean. These measures are valuable because an intu- phological traits or increase the variance in life history itively reasonable way of standardizing the potential re- traits. Variance in traits under stabilizing selection is sponse to selection is relative to the trait mean (W selected against; if the fitness function is locally concave DANE 1949).A rate of response to selection of 10% per upward (both first and second derivatives positive) , an generation would always be regarded as high, and a increase in variance is favored, for agiven mean ( LANDE rate of 0.01% low. Genetic coefficients of variation, in- 1980). This argument is plausible because fitness is by cluding CV,, are correlated with the potential propor- definition under linear directionalselection so variance of fitness itself is a neutral trait; morphological traits tional response to selection (BURTON1952; JOHNSON et al. 1955; CHARLESWORTH1984; HOULE1992). In ad- will usually be subject to stabilizing selection and be dition, when fitness components are modeled,it is con- selected for decreasedvariance. However, all of the life venient to think in terms of mean standardized values, history traits with genetic data areat best fitness compo- nents rather than measures of fitness itself. The condi- as standardized variances are readily converted to vari- tions under which variance in a fitness component will ance in relative fitness (CROW 1958) or to selection be neutralor favored are complex, once potentialtrade- coefficients. offs are taken into account (D. HOULE andL. ROWE, Both the mutational heritability and coefficient of unpublished data). variation are thus useful in specific models of quantita- Under the first hypothesis, that life history traits are tive traits. However, previous reviewshave depended a larger mutational target, we predict that VMshould exclusively on VM/&as a basis for summarizing and also be higher forthese traits. This would be true which- comparing V,values. This can obscure important infor- ever process maintains genetic variance. However, we mation. For example, comparisons reveal that life his- cannot explain high CV, for life history traits by muta- tory traits have lower standing heritabilities than mor- tion-selection balance unless this expectation is met. phological traits ( MOUSSEAUand ROFF1987; ROFF and Under mutation-selection balance, genetic variance will MOUSSEAU1987). This can be due either to smaller be negatively correlated with the average selection coef- additive genetic variances or larger residual variances, ficient against mutant alleles, and positively correlated where residual variance (V,) is the difference between with I&. The higher CVAs of life history traits ( HOULE phenotypic and additive genetic variance. Until re- 1992) run counterto the expected negative correlation cently, the differences in heritability between life his- of variation with the strength of selection, since life tory and morphological traits were usually assumed to history traits are often under strong directional selec- be due to lower additive genetic variance in life history tion. Given this, if the