<<

Copyright 0 1994 by the Society of America

The Effects of Spontaneous on Quantitative Traits. I. Variances and Covariances of History Traits

David Houle,' Kimberly A. Hughes,* Debra K. Hoffmaster: Jeff Ihara,' Stavroula Assimacopoulos, Darlene Canada and Department of and , University of Chicago, Chicago, Illinois 60637-1573 Manuscript received July 16, 1993 Accepted for publicationJuly 2'7, 1994

ABSTRACT We have accumulated spontaneous in the absenceof in mela- nogaster by backcrossing 200 heterozygous replicates of a single high second to a balancer stock for 44 generations. At generations 33 and 44 of accumulation, we extracted samples of and assayed their homozygous performance for female fecundity early inand adult late life, male and female longevity, male mating ability early and late in adult life, productivity (a measure of fecundity times viability) and body weight. The variance among lines increased significantly for all traits except male mating ability and weight. The rate of increase in variance was similar to that found in previous studies of egg-teadultviability,when calculated relative to trait means. The mutational correlations among traits were all strongly positive. Many correlations were significantly different from 0, while none was significantly different from 1. These data suggest that the mutation-accumulation hypothesis is not a sufficient explanation for the evolutionof senescence inD. melanogaster. Mutation-selection balance does seem adequate to explain a substantial proportion of the additive genetic variance for fecundity and longevity.

UTATION is fundamentally important for evolu- significant amount of additive genetic variance in fitness M tion as the ultimate source of adaptive variation. components (MUM et al. 1974; MUKAI1988). This ad- On the other hand, averagethe new mutation clearly has ditive genetic variance may provide a basis for selection deleterious effects on fitness. There is ample evidence in favor of choice of mates by females (CHARLESWORTH that deleteriousalleles are maintained in populationsof 1987; POMIANKOWSKI1988; RICE 1988). In addition,pleio- higher as a result of mutation pressure, and tropic effects ofthese probably contribute to the contribute to a significant reduction in the average fit- maintenance of for other quantitative ness ofindividuals in the population comparedwith that traits (ROBERTSON1967; KEIGHTLEYand HILL1988; BAR- of mutation-free individuals (KONDRASHOV1988; CROW TON 1990; KONDRASHOVand TURELLI1992). Second, the 1993a). Despite the low mutation rates at individual loci, partially recessive of most deleterious mutations this high geneticload is sustained because the total num- (CROWand SIMMONS1983) means that they contribute to ber of loci in a higher is very large. Direct depression, which creates a selective pres- estimates of the genomic in Drosophila sure in favor of outbreeding devices (LANDE and SCHEM- melanogaster (MUKAI1964; MUKAIet al. 1972; OHNISHI SKE 1985; CHARLESWORTHet al. 1990). Third, mutation 1977b; CROWand SIMMONS1983), and indirect estimates load can provide an evolutionary advantage to sexual from selfing plants (CHARLESWORTHet al. 1990) suggest and (CROW1970; that there is likely to be an average of at least one new FELDMANet al. 1980; KONDRASHOV1988; CHARLESWORTH deleterious mutation per individual per generation in 1990a). In addition, thestochastic decline in fitness un- these . The per genomemutation rate in - der mutation pressure in non-recombining finite popu- isms withlarger sizes, such as mammals, is likely lations by the process known as Muller's ratchet (FELSEN- to be even greater, although no direct estimates are STEIN 1974) may cause the of asexual available (KONDRASHOV1988; CROW1993b; KONDRASHOV populations (HAIGH 1978; LYNCH and GABRIEL1990; and CROW1993). CHARLESWORTHet al. 1993),and lead to the degeneration This constant flux of mutations has many important of Y chromosomes (CHARLESWORTH1978). implications for evolution. First, mutation maintains a Finally, theory suggests that accumulation of delete- rious mutations may cause the evolution of senescence ' Present address: Departmentof , University of Toronto, Toronto, (MEDAWAR1952; CHARLESWORTH1994). All evolutionary Ontario M5S 1A1, Canada. *Present address: Chicago Zoological Society, Brookfield, Illinois 60513. theories of senescence depend on the fact that the ef- ' Present address: Divisionof Computers, Math and Science, Siena Heights fects of alleles on fitness early in life are more strongly College, Adrian, Michigan 49221. ' Presentaddress: Department of Biology,Miracosta College, 1 Barnard selected than effects late in life (HAMILTON 1966;CHARLES- Drive, Oceanside, California 92506. WORTH 1990b; ROSE1991; PARTRIDGEand BARTON1993).

Genetics 138 773-785 (November, 1994) 774 D. Houle et al.

If there is substantial mutation pressure for age-specific theory for the evolution of senescence, this work bears deleterious effects, then alleles with deleterious conse- on the otherissues raised above. For example, in spite quences late in life will reach a higher equilibrium fre- of the importance of the rate at which mutation supplies quency than those with effects early in life, thus leading new genetic variation and covariation for theories of to a disproportionate decrease in late-life fitness com- mutation-selection balance (BARTONand TURELLI1989; ponents. As mutation pressure depresses late-life per- CHARLESWORTH1990b; HOULE1991; KONDRASHOVand formance, this further decreases the effects of subse- TURELLI1992), relatively few estimates are available for quent late-acting on relative fitness, life-history traits (LYNCH1988). accelerating the rate of decline. If mutation pressure is strong enough,it can lead to a truncation of later parts MATERIALS AND METHODS of the life cycle (CHARLESWORTH1990b, 1994; PARTRIDGE conditions: were reared on a standard brew- and BARTON1993). The other major hypothesis for the er's -cornmeal-sucrose-agar medium, with propionic acid evolution of senescence, the antagonistic added to inhibit growthof .No live yeast was theory, notes that alleles which increase fitness early in added to the medium, although yeastwas usually carried from life, but have deleterious effects late in life are often vial to vial by the flies. Antibiotics were occasionally added to the medium to eliminate (ASHBURNER and THOMPSON favored by selection (WILLIAMS1957; HAMILTON 1966; 1978).Flies were maintained in temperature-controlled incu- CHARLESWORTH1994). If alleles with such effects have bators at 25" during experiments, and at 18" at otherAll times. often been fixed over evolutionary time, this hypothesis incubators were maintained on a 12:12 hr light dark cycle. would explain the prevalence of senescence; if such al- Experimental flies were reared and maintained Xin 90-mm 25 leles are maintained in a polymorphic state they would shell vials closed with rayon plugs.Other stocks were reared in 60 X 135-mm bottles. Flies were handledunder CO, anesthesia both cause senescence, and allow a response to selection when necessary. for increased life span. ZVpopulation: The base population from which wild chro- The mutation-accumulation hypothesis depends on mosomeswere drawn to initiatethese experiments was two strong testable assumptions about mutation (PAR- founded from approximately 400 iso-female lines captured in Amherst, Massachusetts, by P. T. IVES in 1975, and reconsti- TRIDGE and BARTON1993). First, the rate of production tuted from21 inversion-free iso-female lines in 197'7(CHARLES of deleterious mutations must be large,and, second, the WORTH and CHARLESWORTH1985). This population has been age-specific effects of these mutations must besubstan- maintained in 10 half-pint milk bottles at 25" and a 1212 hr tially uncorrelated. If mutations have similar effects at all light dark cycle on standard medium since 1977. Paper towels adult ages, then selection for early lifeperformance will are addedto increase the carrying capacity of the bottles. Flies are transferred every 14 days, using a regular scheme where also increase performance late in life. While we cannot each new bottle is founded by mixing flies from two bottles know howfar frommaximal the survivorship of a popu- from the previous generation. lation is, we can observe drastic declines in early-life per- Balancer stock: We used the standard balancer stockSMl, formance when late-life performance is selected (ROSE Cy/bd. SMl is a multiplyinverted second chromosome, and CHARLESWORTH1981b; LUCKINBILLet al. 1984; ROSE marked with the dominant mutationCurly wing, and bdis a dominant of the brown eye locus. Both of these chro- 1984b; SERVICE1993). If the mutation-accumulation hy- mosomes are lethalwhen homozygous. To allowdetection of pothesis is a sufficient explanation for the evolution of contamination, we introduced the fourth chromosome reces- senescence, similar declines in early-life performance sive eye mutation sparkling-poliert (spa!'"') into this stock. X should be observable in the absence of any selection and third chromosomes from the ZV population were intro- duced into thisstock by repeated backcrosses of femalesto ZV, (CHARLESWORTH1984). comparable strong assump- No and selection of the visible markers. More completedescrip tions about mutational variance or covariance are nec- tions of these chromosomes and mutationsmay be found in essary for the antagonistic pleiotropy hypothesis. LINDSLEYand ZIMM (1992). Senescence-increasing alleles may be fured by selection, Selection of the stem chromosome: In 1988, 50 wild-type or maintained as balanced polymorphisms under this second chromosomes were extracted from the ZVpopulation. In the course of this, each chromosome stock was also ren- hypothesis, so that there is no requirement that such dered homozygous for spu~"'.Each chromosome was assayed alleles arise frequentlythrough mutation (PARTRIDGE for homozygous egg-to-adult viability relative toSMl/+. The and BARTON1993). 10 chromosomes with the highest viability were then assayed We have studied the variance and covariance of life for fitness in competition withSMI in population cages (SVED history traits, due to spontaneous mutations accumu- 1971; HOULEet al. 1992), and forlongevity and female fecun- dity of homozygotes. Considerable variation between chromo- lated inthe virtual absence of natural selection. In order somes in these traitswas detected, despite similar egg-to-adult to do this, we have adopted the experimental techniques viabilities. One chromosome (line 55), identified as having of Mum (1964; MUKAIet al. 1972) for studying muta- good overall performance in these measures, including rela- tions arising on the second chromosome of D. melano- tively high survival and reproductive performancelate in life, was chosen foruse in the mutation-accumulation experiment. gaster. We report here onstudies through generation44 The remaining lines were pooled to form theZV; spaPo' stock. of mutation-accumulation for longevity, and for fecun- The line 55 and SMl /bd stocks were found to be of P/I cy- dity and male mating success at two ages. In addition to totype, thus precluding dysgenesis on crossing with P/Z providing data relevant to the mutation-accumulation males (ASHBURNER 1989). Life HistoryMutation and Life 775

Mutation accumulation: Two hundred replicate mutation- licates of this “control” population were contaminated by an- accumulation (MA) lines were founded from line 55, by back- other second chromosome (S. NUZHDINand T. MACKAY,per- crossing a single male SMl/+,, to three virgin SMl/bd sonal communication). The contamination of both replicates females. In each subsequent generation, two backcross vials suggests that oneof the generation 33 chromosomes may have were set up within each line from a single SMl/+ male and been the source of the contamination, thus making it likely three virgin SMl/bd females. One of these two crosses was that the control stock was contaminated during the generation then chosen arbitrarily to supply malesfor the following gen- 44 assays. This population was used as a control in a previous eration. If this cross failed produceto offspring, offspringwere publication (HOULEet al. 1992), but wenow regard results taken from the second vial. First vials failed to produce off- involving this stock as suspect. Consequently,we have no con- spring less than 1 % of the time. Thisprocedure has two effects: trol population which would allowus to compare the perform- first, the MA chromosomes were never in homozygous con- ance of the original line 55 and the MA lines. dition, so recessive effects werenot subject to natural selection. Subsequent analyses of sites Second, each MA chromosome forms a population with an on generation 62 MA lines have not revealed any contamina- effective size of 1/2. The combined result is that fixation of tion of the MA lines themselves. We examined insertion sites new mutations with small fitness effects within lines occurs for the elements 297 and roo (LINDSLEY and ZIMM 1992), close to the mutation rate. Selection would only be effective following the methods of MONTGOMERYet al. (1987) and against alleles with very large heterozygous effects. CHARLESWORTHet al. (1992). Element sites were scored overthe In generations 11, 22, 33 and 44, chromosomes were ex- entire second chromosome, with the exception of the proxi- tracted from a sample of lines and checked for lethality. Lethal mal portion of 2L. A total of six MA lines were examined, and lines werediscarded. In the first 62 generations, an additional for two of them we examined both replicate extractions. For 20 lines were accidentally lost.At generations 33 and 44 other each extraction, we examined one to three slides with an av- fitness components were studied in the homozygous lines.To erage of 2.33. Testing took place approximately22 generations control for potential effects of variation in genetic back- after extraction, and thus over 80 generations since mutation- ground, as well ascross-generation environmental effects, two accumulation began. For 297, we detected a total of six in- independent extractions of each line were camed outfor each variant sites,and two apparent element insertions. Forroo, we experimental line. In addition to being isogenic for the second detected 20 invariant sites,and three element insertions. Each chromosome, replicate extractions have a coancestry of 0.042 insertion was only found in a single line. an independent for the third and fourth chromosomes, and 0.021 for the X, As check, S. NUZHDINkindly examined a different lines for 297 due to the fact that males used to initiate the replicate ex- six only, finding 11 sites across the whole chromosome, with no tractions were half sibs2/3 of the time, and full sibs 1/3 of the gains or losses. This sampleincluded the lines with the highest time. and lowest fitnesses ina Sved cage testof 40 lines at generation Expression of Cy: In generation 11 we discovered that in (D. HOULEand unpublished). These are the lines our genetic background, the Cy marker on SMl and related 62 E. BROWN, balancers had incomplete . It is likely that, in se- most likely tobe contaminated. The number of element gains lecting for a stem chromosome with high viability and fitness, is consistent with previous estimates of transposition rates, we inadvertently selected for modifiers of Cy. Problems with while the lack of any element losses strongly argues against the expression of Cy have been noted previously (NOZAWA contamination of any of these lines. While it seems likelythat 1956). In all lines, Cy was expressed in the majority of indi- at least one MA line was contaminated, based on the fact that viduals, so the chief diffxculties this caused wereuncertainty in both replicates of the control stock were contaminated, the the estimation of egg-to-adult viability, detection of recessive proportion of lines affectedwas probably very small. Sinceno sterile mutants, and in obtaining homozygous second chro- lines had anomalously high estimates for any fitness compo- mosome stocks. During extractions, stocks were raised at 27” nent, the effect of such contamination on among line variance during the late pupal period critical for expression of Cy. This is probably small. improved penetrance, but did not entirely correct the prob Production of assay flies: Assays were carried out in three lem. To counteract this, and allow detection of lethal muta- experimental blocks at generation 33, and two blocks at gen- tions, replicate single pair crosses of flies with wild-type wings eration 44. Starting two generations before each block’s assays, were setup, and their offspring examined for Cy. Only crosses vials were set up with 4 female and 4 male flies between3 and where no Cy offspring were detected were retained for sub 7 days posteclosion. These were allowed to lay eggs for 2-4 sequent analyses. Chromosomes were classifiedas lethal if ap days, depending on the experiment. The number of parental proximately 10 such single pair matings failedeliminate to the vials set up per lineextraction combination varied from 5 to Cy marker. In practice, this only occurred in lines where the 10, depending on the block. In blocktwo of generation 33 and of wild-type wings was already low. in both generation 44 blocks, each group of four parental flies Lack of a control population: We originally formed a con- was allowed to lay eggs in two vials. Flies were collected for trol population isogenic for the line 55 chromosome, and longevity assays from the first, or L, laying, and for fecundity minimized the accumulation of mutations in this population and male mating from the second, or F, laying. by maintaining it at large effective size. However,in the course Fecundity: We assayed female fecundity on days 5 and 6, of attempts to select thispopulation for increased expression and days 27 and 28 posteclosion. Early fecundity is the sum of of an introduced Cy marker, this population was reduced to eggs laidby an individualon days 5 and 6, while late fecundity a census size on the order of 20 for a number of generations. is the sum of eggs laid on days 27 and 28. Females usedin early Previous results make it likely that this is small enough that fecundity experiments were always drawn from flies eclosing substantial mutation-accumulation could have occurred (Mu- on the same day. For the late fecundity experiment, we ob KAI et al. 1972). Following generation 33 assays, a new popu- tained all flies from the target eclosion date, where possible, lation was formed by pooling the six MA lines with the highest although we frequently used some flies up to 2 days older or average rank for the fecundity and longevity traits. Replicate younger than this. Subsequent analyses indicated that there extractions of each line were pooled separately, yielding two were no significant differences in fecundity of flies within this “control” populations. Following generation 62, analyses of age spread (data not shown). Flies used in the early fecundity transposable element positions have revealed that both rep experiment were never used for late fecundity assays. 776 D. Houle et al.

Prior to assay, females were mated with males ofthe outbred in vials with 10 virgin W' females without anesthesia.The age ZV; spup"' stock. In most experiments, no special attempt was of wa males usedwas chosen so that approximately equal pro- made to collect virgin females(although most probablywere), portions of each type of mating were obtained. After approxi- on the assumption that any effectsof early matingswith males mately 2 hr, the flies were anesthetized and the males and of their own genotype would be negligible followingremating females separated. Females were then placed individually in with ZV; spap"' males. For both early and late fecundityassays, vials, and their offspring checkedfor the expression of w". The flies wereheld in groups of 15-20 females per vial, witha com- proportion of mated females producing wild-type offspring parable number of IV; spup"' males. Mating with ZV; spaPo' was used as the index of male mating success. males always took place on day 2 posteclosion for early fe- Fitness: We have previously reported the results of fitness males, and between 2 and 5 days after the target date for late assays of lines at generation 44 (HOULEet al. 1992). Briefly, we females. Flies held for late fecundity weretransferred to fresh constructed a population consisting of a test chromosome, or food every 2-3 days for the first 2 weeks, and every 3-4 days sample of chromosomes, and a balancer chromosome (bw"') thereafter. carrying a recessive lethal. The equilibrium frequency in Three days before a fecundity assay, flies weretransferred to adults was then used to obtain a measure of the relative fitness fresh vials, seeded with a live yeast solution. In the late fecun- of the test chromosome(s). dity assay, the ZV; spaPo' males were replaced with males be- Productivity and weight: To control for potential effects of tween 3 and 5 days old atthe same time.On the first assay day, rearing density on othertraits, we counted the number of pu- flies wereanesthetized and sorted into male-female pairs,and paria on the side of each vial after we had collected flies for each pair placed on fresh yeasted standard medium in dis other assays. We refer to this character as productivity. In ad- posable plastic shell vials witha diameter of 2.5 cm. Green food dition to reflecting larvalenvironment, productivity is also an coloring was added to the medium to make the eggs more index of parental fecundity and offspring viability, and so is visible. After 24 hr, flies were transferred to fresh medium itself a life historytrait. For blocks wheretwo layings wereused, without anesthetization. After another 24 hr, the flies were two measures of productivity are available. As a direct measure discarded. Vials in which the female died were discarded.Fol- of size, we weighed groups of approximately 10 live males at lowing removal of the adults, vials were frozen to await egg least 5 days posteclosion, whenever sufficient flieswere avail- counting. able. A total of nine observers counted eggs from fecundityassays, Analyses showedthat the environmental correlations involv- and the counts were assigned so that vials with eggs from the ing these traitswere small, with72 values lessthan 5%,and the same female were generally counted by different observers. signs of the correlations do not correspond well with the ex- Some eggs were difficultto observe on the surface of the food pectations for environmental correlations based on previous for several reasons,including hatching of some larvae before studies (ROBERTSON1963; PROUTand MCCHESNEY1985). Con- the vials were frozen and residual yeast on the surface. Con- sequently, all analysesreported are not adjusted for produc- sequently it proveddifficult to getobservers to agree on tivity or weight. counts. Although the means obtained by each counter are Statistical analyses: Standard statistical analyses were car- within 10% of each other, analyses ofvariance show significant ried out with the SAS statistical package, version 6.03 (SAS counter effects. To adjust for these effects, we assumed that Institute, 1988a,b). Before further analyses, the data were ex- counters tended to over- or underestimate counts by a constant amined for outliers. For both early and late fecundity the data proportion. Counts were log-transformed and residuals ob- show two modes, one at zero fecundity, so ANOVA residuals tained from analyses of variance (ANOVAs) for each genera- were very far from normal. There was no detectable genetic tion, with blocks, day of egg layingand identity of counter as component to the probability of these very lowfecundities. We effects. Following this,the least squares means for day of laying eliminated the data in this low mode, in order to render the were added to the residuals, the data back-transformed, and distribution more normal. Early fecundities with less than 10 the two adjusted counts added together to obtain the fecun- and 5 eggs were eliminated from generations 33 and 44, re- dity. This procedure removed most of the effect of counters, spectively; all late fecundities of 0 were dropped in both gen- as well asdifferences among blocks. Tables2 and 3 give the raw erations. For the other each block was analyzed means for each block. Becauseof the counter adjustment pro- separately, and the residuals examined for outliers, which were cedure, the block means differfrom those inthe data used for subjected to Grubb's test (SOKALand ROHLF1981) and dis- further analyses. carded if significant at P < 0.05. When samplesizes exceeded Longevity: Male and female survival was assessed in single the maximum tabledvalue of 150, observations whose residu- groups of 20 virgin flies. Flies eclosing on a single target als were separated from the main body of residuals, and were date were used where possible,although variation among lines more than 4 SD from the mean were discarded. This resulted necessitated the use of flies up to 4 days older in some cases. in the removal of three male longevity vial means, two female Longevity statistics were calculated from the mean ages of longevities,two male weightsand two vialproductivities in gen- adults in a vial. Once a week, flies wereanesthetized, and sur- eration 33. In generation 44, one fitness cage, two early and viving MA flies counted. Dead flies werereplaced with virgin one late mating trial, eight male weights and six vial produc- flies froma stock bearing the eye mutation -apricot ( w') , tivities were removed. so that the density of flies in a vial was always close to 20. The Bootstrap analyses were carried out in several steps. First, 20 flies were then placed on fresh unyeasted food. data were corrected for counter effects and outliers were re- Male mating ability: We assayed early malemating abilityon moved, as indicated by the above analyses. Second, the cor- day 3 and late mating ability on day 21 posteclosion. Males rected data were resampled at all levels of the design, except used in the early matingassay were always collected from flies for the extraction level. Since there were only two replicate eclosing on the same day. For the late mating experiment, extractions, resampling can only reduce the variance at this most flies could beobtained from the target date, although we level. At each level, sampling was always carried out with re- occasionally used malesup to 2 days older or younger than the placement. To resample the data from each generation, a line target date. Prior to assay, males wereheld in single sexgroups was first chosen at random, then from within each block- of up to 25 flies. Late flies were transferred weekly to fresh extraction combination individual observationsfor each trait food. To assay mating ability,we placed 10 MA and 10 wamales were again drawn, up to the actual sample size of that . If M utation and Life History Life and Mutation 777

TABLE 1 Trait means by block for generations 33 and 44

Generation 33 at block Generationblock: 44 at

Phenotype 3 1 2 1 2

Early fecundity Mean fecundity Early 81.10 61.56 100.29 90.56 80.65 SE 0.75 0.98 0.790.72 0.94 700 764 73 N 764 652 700 1 741 Late fecundity Mean fecundity Late 35.98 47.56 66.02 45.88 61.60 SE 0.70 0.81 0.75 0.76 0.95 759 722 618 671 618 N 722 604 759 M ale longevity (days) Mean (days) longevity Male 31.07 29.07 39.70 31.47 31.57 0.64 0.69 0.56 SE 0.69 0.63 0.64 0.70 N 73 84 79 63 74 Female longevity (days) Mean (days) longevity Female 56.70 52.09 61.06 53.95 58.60 SE 0.57 0.73 0.85 0.68 1.09 85 79 65 N 79 69 85 67 Early male mating Mean mating male Early 0.570 0.668 SE 0.064 0.053 N 71 60 Late male mating Mean mating male Late 0.461 0.301 SE 0.082 0.088 N 76 56 Fitness Mean 0.462 0.313 SE 0.021 0.018 N 77 80

the line drawn was not assayed in a particular block, no data 113 tested after 22 generations; 7/56 tested after 33 were included for that block-line combination. Similarly,if a generations; and 5/13 tested only after 44 generations. given extraction was missing for that block-line combination, this generateda missing cell inthe resampled data. This proc- After correcting for chromosomeswith more than one ess was repeated until thesame number of lines were resam- lethal (MUKAIet al. 1972), lethals were estimated to oc- pled as were assayed in the set of experiments for that gen- cur ata rate of 41.6/6017 = 0.0069 per generation. Al- eration. The datawere always somewhat unbalanced, and the though this is slightly higher than many previous esti- resampling scheme preserved this lack of balance. mates (SIMMONSand CROW1977), it is not significantly Each trait necessitated a slightlydifferent resampling scheme. Fecundities were resampledat the level of 2day egg different from the rateof 36.0/6000 estimated by MUKAI counts, longevities as individual dates of , and produc- et al. (1972) (G = 0.39, 1 d.f., P>0.5). tivities at the vial level. For the male matings, samples were Distribution of MA line means: The block means for drawn from the binomial distribution with the estimatedpro- all traits are given in Tables 1 and 2. Forty-two lines were portion and samplesize as parameters. For fitness and weight, only a single observation was made in all (fitness) or many assayed for life history traits at generation 33, and 43 at (weight) block-lineextraction combinations,so observations generation 44. The generation 44 lines were among the were sampledby adding a normal deviate with variance equal 47 lines whose fitness was the subject of a previous report to the variance to the observation. Since fitness is con- (HOULEet al. 1992). The lines were sampled at random strained to be between 0 and 1, observations were arc-sin trans- formed, resampled on this scale, then back-transformed. To from thenon-lethal lines remaining at each generation, yield a uniform type of analysis, block-lineextraction means so 14 of the lines were studied at both generations 33 from the resamples were analyzed.The resultsof each analysis and 44. The distribution of MA line means was always were saved, and after1,000 bootstrap resamples, the quantiles unimodal, and generally close to normal, althougha few of the desired statistics were examined to determine the pa- rameter estimates and confidence intervals. lines usually showed markedly lower performance. This is consistent with data on viability in previous mutation- RESULTS accumulation experiments (MUKAI1964; MUKAIet al. Lethals: A random sample of MA lines were extracted 1972). In theseprevious experiments, an arbitrary cutoff and assayed for lethality in generations 22, 33 and 44. point was chosen, and lines with viabilities below the With random samplingof chromosomes, many chromo- cutoff were assumed to carry a severely deleterious mu- somes were tested more than once,so lethality tests were tation. We tested atypical lines using Grubb’s test (Sow made after variable numbers of generations of muta- and ROHLF1981), andlines were identified as probable tion. We identified a total of 38 lethals: 9 in 101 chro- carriers of severely deleterious mutations if they were mosomes tested 11 generations after the last assay; 17/ significant outliers at theP< 0.05 level. In generation33, 778 D. Houle et al.

TABLE 2 Block means of productivity and male weight

Productivity Male Productivity weight (mg)

GenerationMean Block SE N Mean SE N

33 1 136.70 2.43 177 0.722 0.012 88 33 2L 52.09 1.71 240 33 2F 103.83 2.29 321 0.773 0.012 170 33 3 131.02 1.73 612 0.763 0.026 56 44 1L 30.90 0.83 337 0.813 0.037 39 44 1F 40577.36 1.81 0.732 0.010 186 44 1L 60.22 1.39 396 0.817 0.015 83 44 1F 80.50 1.66 391 0.758 0.013 71

two lines were significant outliers for both male and fe- responding GLM analysis. Bootstrap resampling was car- male longevity. In generation 44, one of these same lines ried out as described in MATERIALS AND METHODS, yielding was again an outlier for longevity and fitness, and an- a sample of lineextraction means for each . other line was also an outlier for fitness. All of these These were analyzed as a two-wayANOVA with tray severely deleterious lines also showed below averageper- (where appropriate) and lines as main effects, and the formance for other life-history traits. The remaining line effect variance component estimated by a method lines were considered to have “quasi-normal” perform- similar to the SAS type I method. The Table shows me- ance. Subsequent analyses werecarried out bothinclud- dians of 100 bootstrap samples, with the significance ing and excluding the severely deleterious lines for all level calculated from the proportion of bootstrap esti- traits, to determine what influence these few large mu- mates greater than0. With few exceptions, the bootstrap tations have on the results. and ANOVA results are rather close, indicating little Mutational variances: Conventional variance compo- bias. However, the significance levels of the bootstrap nent analyses withingenerations were carried out using samples are usually more conservative than those from the SAS GLM procedure (SAS Institute, 1988b) with ANOVAs, taking into account that the highest signifi- lines, extractions nested within lines, and block by line cance level from the bootstrap sample is 0.01. interactions as random effects, and block as a fixed ef- A slightly different structurewas used for analyses of fect. For longevity, flies werehandled in trays, and tray productivity and male weight because of the availability effects were alsoincluded in the analyses for these traits. of data from two layings in some blocks. In generation Every phenotypeexcept the male mating abilities 33, flies were onlyweighed from thefecundity assay vials, showed significant genetic effects, either as line effects while flies wereweighed from both layings ingeneration or block by line interactions (at P < 0.05; results not 44. These laying effects are highly significant for pro- shown) in one generation or the other. Block by line ductivity, so that each laying was analyzed separately, as interactions were significant, or nearly so for all of the shown in Table 3. Analyses of weightsfrom block 1 gen- fecundity and longevity traits in generation 33, making eration 33, and from laying L, block 1, generation 44 are it difficult to interpret variance components from the not included, as the samples were small and highly un- full analyses. Instead, we assume that the genetic vari- balanced. ances in each block are drawn from the same distribu- Severely deleterious lines were detected for longevity tion, even though therelative performance of lines may phenotypes, so removing these lines has a relatively large vary among blocks. Analysis eachof block separately also impact on the results for these traits. Male and female facilitates bootstrap resampling of thedata, as such longevity data were also analyzed using the Gompertz analyses are muchsimpler in structure and moresimilar regression model, where log mortality is partitioned into across phenotypes than the full analyses. “intrinsic” component which is assumed constant The variance components from these single block throughout life, and a “senescent” component which is analyses are shown in Table 3. The variables were not assumed to increase linearlywith age (FINCHet al. 1990; transformed before analyses, so the variance compo- HUGHESand CHARLESWORTH1994). Neither component nents aregiven on theoriginal measurement scale. The showed significant genetic variance for either sex (re- ANOVAvariance components were obtained by the type sults not shown). I method in theSASVarcomp procedure (SAS Institute, To combine variances into a single estimate of the per 198813). Type I estimates were chosen because they are generation mutational variance ( VM)we used the slopes not constrained to be greater thanor equal to 0, which of regressions of genetic variance on generation. The would bias the regression analyses below. The signifi- regressions were forced to go through the origin, since cance levels shown are for the lineeffect from the cor- it is reasonable to assume that no genetic differences Mutation and Life History 779

TABLE 3

Among line variance components from single block analyses for all phenotypes

~ All lines Quasi-normals

Phenotype GenerationVARCOMP Block Bootstrap VARCOMP Bootstrap

Early fecundity 33 1 10.384 13.196 10.384 7.966 2 67.295* 80.608* 51.137* 62.843* 3 2.469 0.624 2.810 2.180 44 1 -4.663 -9.669 -9.791 -11.921 2 53.649** 56.692* 51.519** 52.090** Late fecundity 33 1 0.802 20.645 0.802 14.009 2 32.772* 45.655* 28.004 25.949t 3 -0.616 16.104 -1.471 12.902 44 1 60.785** 60.452* 60.239** 58.675* 2 91.932**** 103.089** 77.628*** 79.313* Male longevity 33 1 9.606** 3.977 9.606** 4.952 2 21.448**** 20.291* 7.900** 7.136t 3 16.316*** 7.971 16.716*** 8.958 44 1 -4.387 0.257 -4.387 0.039 2 0.365 2.864 -1.849 2.583 Female longevity 33 1 2.831 6.622 2.831 5.350 2 33.079**** 29.345** 7.654* 7.181 3 0.538 -4.408 0.599 -6.113 44 1 1.029 1.239 1.029 4.183 2 24.987* 27.735t -1.395 5.005 ~arlymale mating X lo3 44 1 -1.94 -8.395 -2.795 -8.434 2 5.6711. 6.037 5.853t 7.027 Late male mating X lo3 44 1 16.05t 11.743 15.534t 11.391 2 11.96 27.816t 13.14 22.921 Male weight X lo3 33 2 -0.085 0.096 -0.088 0.220 3 0.730 1.458 0.796 1.385 44 1F -0.231 -0.250 -0.205 -0.358 2L 0.259 0.350 0.272 0.240 2F -0.288 -0.186 -0.323 -0.258 Productivity 33 1 297.989** 325.484** 297.989** 274.542* 2L 27.097 44.774 -4.755 -133.083 2F 245.439t 176.028 117.726 95.492 3 314.479t 302.092* 260.378 240.028* 44 1L 22.419 24.002 23.037 30.444t IF -22.026 6.881 -49.419 -25.456 2L 115.554t 160.875t 113.1771. 146.539* 2F 97.705 157.655-t 93.658 114.389 t P < 0.10; * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001. Significance levels not adjusted for multiple comparisons. Values are given on the measurement scale. existed ingeneration 0. Median slopesfor regressions of variances and correlationsamong traits. Covariances trait variances on generation from 1,000 bootstrap were calculated fromthe variance of sums (KEMFTHORNE samples are shown in Table4. To place the estimates for 1957, p. 265). A single best estimate of the increase in different blocks and traits on the same scale, the data each covariance was calculated from the slope of co- within each block were standardized by the mean for variance vs. generation. Corresponding V, values were that block. Significance levels were againobtained from calculated using only data from block-line-extraction the quantiles of the bootstrap sample of estimates.The combinations where both traits had been measured. significance levels indicated following the median esti- Table 5 gives the genetic correlationsfor traits with sig- mates are for one-sided testsfor differences from0, with nificant increases in V, in Table 4. Since correlations upper and lower two-sided 95% confidence limitsgiven cannot be calculated when a variance component is below. V, is significant for all traits exceptfor male mat- negative, the confidence limitson the valid correlations ing and weight when all lines are analyzed. Only the do not necessarily agree with the significance levels from fecundity traits and productivity are significant in the the covariance test. Inaddition the confidence limitsare quasi-normal sample. two-sided, whilethe covariance has been subjected toa Mutational covariances: The bootstrapresampling one-sided test. procedure used inthe last sectionwas also used to obtain The overall pattern of these correlationsis quite strik- estimates and confidence limits for the mutational co- ing; every correlation is positive. Theyrange from 0.41 780 D. Houle et al.

TABLE 4 the D. melanogaster genome. Therefore, our estimates Rate of increase in line variance (VM X lo4) per generation due to of V, in Table 4 should be multiplied by 2.5 to extra- mutation, estimated from combined data, standardized to a polate them to the whole genome. On the other hand, block mean of 1.0 we have estimated the homozygous effects of new mu- tations. Previous experiments have shown that muta- Trait All lines Quasi-normals

~ tions affecting life history traits are overwhelmingly Early fecundity Med. 1.192* 1.025* deleterious in their effects (CROWand SIMMONS1983; L 0.069 -0.190 U 2.574 2.304 LYNCH1988), so that thehomozygous effects themselves are of little interest in random mating populations. Late fecundity Med. 5.252** 4.312** L 1.832 1.199 Spontaneous mutations affecting viabilityhave been U 9.100 8.537 shown to be nearly additive in their effects. Using a Male longevity Med. 1.798* 0.927 model where the relative phenotypes of the original L 0.135 -0.556 homozygote, heterozygote and mutanthomozy- U 4.137 2.588 gote are 1, 1 - hs and 1 - s, respectively, experiments Female longevity Med. 1.141* 0.311 L -0.013 -0.425 suggest that for new mutations affecting viability h, the U 2.710 1.057 coefficient, is approximately 0.4 (MUKAI Early mating Med. -1.629 -1.697 et al. 1965; MUM and YAMAZAK~1968; OHNISHI1977a; L -12.488 -13.670 SIMMONSand CROW1977). Assuming that these estimates U 8.810 9.170 are applicable to our life historytraits, the incrementin Late mating Med. 33.311t 33.995t L -11.708 -17.605 additive variance due to mutation over the whole ge- U 96.024 93.288 nome should be 2(1/0.4) h2 = 0.8 times the values in Productivity Med. 5.001* 4.141* Table 4. L 0.565 -0.315 Our estimates of the genetic variances and covari- U 9.551 8.942 ances are biasedslightly, as the replicate extractions Male weight Med. -0.018 -0.027 L -0.221 -0.248 have an average coancestry of 0.042for thirdand fourth U 0.194 0.166 chromosomes and 0.021 for the . The Data are quantiles from 1,000 bootstrap resamplings of the entire shared segments in all cases are derived from the bal- data set. For each character, the median estimate is presented in the ancer stock used during mutation-accumulation. Using first row. The second and third lines give the 2.5 and 97.5% percen- standard estimates of relative chromosome sizes we es- tiles. timate that the among line variance would be biased to 0.93 for all lines, and from 0.238 to over 1 for the upwards by an amountbetween 0.021 times the additive quasi-normal sample. While many of the covariances genetic variance in the balancer stock and 0.021 times and correlations are significantly different from 0,none the homozygous genetic variance. Since our V,,, esti- of the correlations are significantly different from 1.0. mates are calculated on a per generation basis, the ex- This suggests that most new mutations affect all com- pected bias is the average of 0.021/33 and 0.021/44, or ponents of these life-history traits in similar ways. This about 0.0006, times the appropriate genetic variance. pattern is not caused by the severely deleterious lines, as Several factors argue that thebiases are unlikely to be the mean correlation is 0.68for all lines, and 0.62 for the large in our experiments. Following the initial genera- quasi-normals. tions of extraction, each extraction was maintained un- der crowded conditions in single vials, slowing further DISCUSSION inbreeding. This suggests that the bias will be much We have studied the homozygous effectsof mutations closer to the additive variance than thehomozygousvari- accumulated on the second chromosome of D. mela- ance. Our mean-standardized estimates of V,, scaled to nogaster, in the virtual absence of natural selection. additive effectson thewhole genome, as above, are com- We obtained strongevidence for the importanceof mu- parable to IA (the ratioof additivegenetic variance to the tation for fecundity, longevity, productivity (a measure square of the trait mean) values for fecundity and lon- of fecundity times viability) and a measure of fitness, gevity reviewed by HOULE(1992). For these traits the and somewhat weaker evidence for male mating ability ratio V,/VA for all lines ranges from 0.03 to 0.006, at late in life. We found no evidence for mutation- least an order of magnitude larger than the 0.0006 pos- accumulation in male mating ability early in life, and sible from background additive variance. Furthermore, very weak evidence for weight. The most striking results the populationsize followingextraction would also allow of this study are the high, positive, mutational correla- some selection, thereby reducing the frequenciesof any tions among traits. shared segments with large effects on fitness,which Several factors must be borne in mind when exam- would be likely to contribute a large proportion of the ining our results. First of all, we studied effects accu- background variance. The balancer stock is also unlikely mulated only on thesecond chromosome, about40% of to have levels of genetic variance as high as those in an Mutation and Life History 781

TABLE 5

Genetic correlations betweentraits with signifcant increases in variance

Early Late Male Fem. Produc- Fem. Male Late Early Fitness fecundity fecundity longevity longevity longevity fecundityTrait fecundity Fitness tivity

Fitness Med. Fitness 0.690t 85 0.691** 99 0.664t 61 0.875* 90 0.605* 81 L -1.434 0.116 -0.350 -0.033 0.046 U 2.953 1.265 2.984 3.540 2.137 Early fecundityMed. 0.571t 77 0.666** 97 0.680t 95 0.636 97 0.7561. 94 L -0.446 0.203 -0.409 -0.285 -0.105 U 2.280 1.682 2.288 1.820 1.961 Late fecundityMed. 0.736* 100 0.591* 92 0.5621. 95 0.817** 99 0.448t 94 L 0.010 0.085 -0.200 0.155 -0.172 U 1.758 2.523 1.305 1.673 1.240 Male longevityMed. 0.818 57 0.5551. 88 0.399 89 0.780* 97 0.607t 93 L -0.400 -0.245 -0.320 0.232 -0.324 U 2.964 3.880 1.715 1.877 1.813 Female longevityMed. 1.022t 73 0.854 67 1.192** 77 0.666 66 0.606* 93 L -0.787 -1.007 0.378 -5.680 0.028 U 3.53 26.503 4.949 3.093 1.698 Product- ivity Med. 0.641* 76 0.588 84 0.334 81 0.367 78 0.697 66 L -0.090 -0.517 -0.574 -1.670 -1.151 U 1.642 4.113 1.478 3.181 3.748

~~~~~~ ~ ~ ~ Correlations among all lines above the diagonal, quasi-normal lines only below. For each trait, the median correlation is in bold type, and followed by the significance level for a 1-tailed test that the covarianceis greater than 0, and by the number of estimates of the correlation where both variances were greater than 0. The second and third lines give the lower and upper95% confidence limits, respectively, for the correlation. Significance valueswere not adjusted for multiple comparisons. t P < 0.10; * P < 0.05; ** P < 0.01. outbred population. The balancer stock’s genetic back- stitute an extremely competitive, if physically benign, ground was initially derived from the Npopulation, but environment. Flies used in the longevity assays were un- during the course of these experiments the stock was mated. Rather different estimates might have been ob- expanded from small numbers of flies on several occa- tained if, for example, longevity could be assayed within sions, probably reducing the background variation in the crowded cages where fitness was assayed. Consistent the process. with these arguments, A. S. KONDRASHOVand D. HOULE Aside from these factors which are expectedto reduce (submitted for publication) have shown that the differ- the background variance, the data themselves do not ences in relative fitness between mutation-accumulation suggest an important role forbias from this source. Our lines drawn from generation 62 of this experiment and variance estimates for cage fitnesses are not affected by a line subjected to natural selection are greatly magni- background variation, as fitnesses are calculated relative fied under more competitive conditions. to the performance of a genotype within the same cage, Our relative lack of success in detecting mutational and therefore with the same genetic background. If the variance in male mating ability and weight is not unex- background effect was large, we would expect to see a pected. Our assay of male mating ability is the propor- discrepancy between the variances and correlations in- tion of matings obtained by test males in competition for volving fitness, and those which do notinvolve fitness. In 10 females, and only about 130 such tests were per- particular the variance of the non-fitness components formed with MA lines at each age. Since the error vari- would be biased upward, while the covariance would ance foreach assay is approximately binomial, it is clear not, leadingto low estimates of the genetic correlations. that our power to detect mutational effects was limited. In fact all of the estimates seem consistent, as well as We measured weight asan indicator of density effects on consistent with previous studies of variation in viability other traits, rather than as a trait of primary interest in (see below). this study. We expect that careful control of larval den- Our estimates of genetic correlations may be biased sities might reveal genetic variation in size. downward, as the environments flies experienced dur- Comparisons to previousmutation studies: Our ing each assay were quite different. Our fecundity and mean standardized results are consistent with measures longevity assays were carried out in benign environ- of egg-to-adult viability and development rate from pre- ments with little competition, but thefitness cages con- vious mutation-accumulation experiments on the sec- 782 D. Houle et al. ond chromosome of D. melanogaster. In a large series lection does tend to purge those alleles contributing of experiments, MUKAIet al. (1972) estimated a stan- positive covariance between life historytraits. The result dardized mutational variance of 3.3 X for non-lethal is that correlations due to standing variation are con- chromosomes. Very similar estimates were alsoobtained siderably lower than those among unselected mutants, for development rate (MUKAIand YAMAZAKI19’71; YOSHI- although they do notnecessarily become negative. Early MARU and Mum 1985). OHNISHI(197%) also obtained and late female fecundity have been shown to have a estimates of mutational parameters forviability in simi- small negative additive genetic correlation in the same lar experiments, obtaining anestimate for non-lethals of IV base population used in ourexperiments (ROSEand V, = 2.1 X For our study, comparable values for CHARLESWORTH1981a). YOSHIMARUand MUKAI(1985) the five fitness components with significant variance found that the correlation of development rate and vi- among all lines average V, = 2.88 X Forcage fit- ability of chromosomes from a natural population in nesses, the comparably standardized estimate is 17.2 X both homozygous and heterozygous condition was with a lower 95% confidence limit of 6.1 X about 0.25, ascompared to thecorrelation of 0.9among (HOULEet al. 1992). Mum’s viability estimates differ new mutants. SIMMONSet al. (1980) found evidence for slightly in that they are standardized to a premutation a slight negative correlation between viability and fitness mean of 1.0, while we have standardized to a mutant of heterozygotes from a natural population. On the mean of 1.0. other hand,when genotypes from outbred populations The high, positive mutational correlations in our data are inbred, they show strong positive correlations be- have precedents in other mutation-accumulation ex- tween fitness components (HIRAIZUMI 1961; ROSE 1984a; periments. Development rate and viability are strongly MACKAY 1986), consistent with the inefficiency of selec- positively correlated forunselected mutations, with most tion in removing recessive deleterious variation. estimates being 0.9 or greater (Mum and YAMAZAKI The similarity of our estimates of V, to those for vi- 1971; YOSHIMARUand MUM 1985). Unpublishedobser- ability, (see above) and the strong positive mutational vations of 0. OHNISHIon EMSinduced mutations also correlations found inthis and otherstudies suggest that show positive correlations (CROWand SIMMONS1983). the genomic mutation rate of our life history traits is SIMMONSet aZ. (1980) compared viability and fitness of similar to that for viability as well. Previous minimum heterozygotes from fourdifferent sets of mutation- estimates of total mutation rate for viability on the sec- accumulation lines, and foundweaker evidence for posi- ond chromosome are between 0.07 and 0.15 (CROWand tive correlations. Experimental difficulties may have af- SIMMONS1983). The consistent estimates by MUKAIfavor fected the results of this study, as the genetic variances the higher figure (Mum 1964; MUKAIet aZ. 1972). Ex- amonggroups of lines donot correspond to the trapolating to the entirehaploid genome a minimal es- amounts of mutation-accumulation they hadunder- timate of the haploid genomic mutation rate is about gone, and one ofset lines gave strikinglydifferent results 0.38/generation. This estimate is derived by assuming than theothers. The authors noted a high frequency of that thereis no variation in the effects of newmutations, hybrid dysgenesis in crosses similar to those used to es- which is clearly not true. Assuming reasonable levels of timate viability and fitness. A small mutation- variation in allelic effects, wouldrequire increasing his accumulation experiment in Daphnia showed a variety estimate by a factor of two or more (CROWand SIMMONS of correlations among fitness components, although the 1983). only correlations which were significantlydifferent from Implications for the evolution of senescence: The 0 were consistent withpositive correlations near 1 mutation-accumulation hypothesis for the evolution of (LYNCH1985). senescence depends directly on two assumptions about The many high positive correlations among life his- new mutations (CHARLESWORTH1984; PARTRIDGEand BAR- tory traits suggest that most unselected mutations will TON 1993). The first of these is that mutation pressure have deleterious pleiotropic effects on all components decreasing mean performancemust be large. A series of of fitness. The fact that mutational variance for fitness is experiments selecting for late life performance show higher than those for fitness components is consistent that early fecundity declines at about 2.5% in such ex- with this interpretation. We have previously shown that periments (ROSEand CHARLESWORTH1981b; LUCKINBILL mutation load can have a significant impact on genetic et al. 1984; ROSE1984b; CLAREand LUCKINBILL1985), covariances of life-history traits under the assumption while male mating ability declines at nearly 1%/ that mutation-selection balance maintains genetic vari- generation (SERVICE1993). The second assumption is ance (CHARLESWORTH1990b; HOULE1991). This is par- that mutations must be substantially independent in ticularly so when the mutational covariance is very dif- their effects on early and late life performance. If this ferent from that predictedby optimality models (HOULE were not the case, then, in the experimentscited above, 1991), as we have shown here. changing the age where selection acts most strongly Comparisons of these mutation studies with the cor- would have little effect on early life performance, as al- relations found in outbred populationssuggests that se- leles normally kept atlow frequencies by selection early Life HistoryMutation and Life 783

in life would still be keptat low frequencies by selection can be favored by selection, their occurrence atany time acting only late in life. The relatively high mutational in the evolutionary history of a can increase its correlation of 0.6 we found between early and late fe- rate of senescence (PARTRIDGEand BARTON1993). To cundity suggests that mutation-accumulation is unlikely explain the correlated decreases in early fecundity un- to explain the experimental results. In order to rescue der late-life selection outlined above, it is necessary that the mutation-accumulation hypothesis with such a high alleles with antagonistic effects favoring late-life fitness correlation, one would have to assume that the muta- and depressing early-life fitness be segregating in the tional pressure decreasing early fecundity is many times selected populations. As outlined above, there is sub- larger than that affecting viability in previous mutation- stantial evidence that selection decreases the covariance, accumulation experiments. Even assuming age- consistent with the existence of such alleles. In the ex- independent effects, the rate of viability decline due to treme, alleles with antagonistic effects may form pro- mutation is not as high as the decline in early fecundity tected polymorphisms, and so maintain noted above (CHARLESWORTH1984). indefinitely in large outbreeding populations (ROSE A number of other lines of evidence have suggested 1982). Thus, mutational data cannotreject the antago- that mutation-accumulation may play a role in deter- nistic pleiotropy hypothesis. mining D. melanogaster life span, and our data do not Implicationsfor the maintenanceof genetic variance: rule this out. SERVICEet al. (1988) showed that reversal V, is an essential parameter in mutation-selection bal- of selection for late life performance did not reverse ance models for the maintenance of genetic variance gains in ethanol and desiccation resistance, although (BARTONand TURELLI1989), and the adequacy of the early fecundity and starvation resistance did return to mutation-selection balance model is also important to their original levels. This suggests that alleles whichlead models of inbreeding depression, mate choice, and the to ethanol and dessication resistance do have late-life evolution of outbreeding devices (see Introduction). specificeffects, HUGHESand CHARLESWORTH(1994) One important indication of the strengthof mutation in found that additive genetic variance for survival prob- promoting variation is the ratioV,/ V, with V, adjusted ability and male mating ability increase markedly with to heterozygous effects on the whole genome, as out- age in males in the Npopulation, as predicted under lined above. Underthe mutation-selection balance the mutation-accumulation hypothesis (CHARLESWORTHmodel V,/V, is the inverse of the average time that a 1994). Alternative explanations of these data are pos- deleterious allele would haveto persist in the population sible. For example, the average effects of alleles could to explain the observed level of V, (CROW1993b). Our increase with age, without the presenceof mutants with values of V,/V, (above) suggest short persistence times age specific effects. Our mutational data provide some of 33-167 generations, which are consistent with the ex- weak evidence that either the number or theeffects of pected persistence times for spontaneous mutations af- alleles are magnified late in life, as both fecundity and fecting viability (CROW199313). Therefore, it appears male mating have higher V, values than their early life likely that mutation-selection balance can explain a sub- counterparts. In neither case is this difference signifi- stantial proportion of the genetic variance in fecundity cant. A recent series of selection experiments shows and longevity. V, is also likely to be a useful predictor weak evidence for tradeoffs between early and late fe- of the rate at which populations can respond to long cundity, and has also been interpreted as providing term selection pressures (HILL1982). some evidence for the mutation-accumulation hypoth- Comparingmutational variances: Previous authors esis (PARTRIDGEand FOWLER 1992; ROPERet al. 1993). As summarizing V, data have standardized their estimates the authors point out,many of their results can also be by the environmental variance ( V,), yielding the incre- ascribed to unintended selection, due to thefact that the ment in heritability due to a single generation of mu- base populations for these experiments is maintained tation (LYNCH1988). For fitness components, the envi- on a continuous schedule, while both selection regimes ronmental variance is irrelevant to the response to are carried out on a discrete generation schedule. On selection, making it unlikely that V,/ V, is an appropri- the other handthese experiments suggest that tradeoffs ate quantity with which to compare levels of variation did take place between larval and adult fitness. Under (HOULE1992). However, for comparative purposes we the antagonistic pleiotropy hypothesis, it would not be have calculated V,/V‘ from our data by averaging esti- surprising if different pairs of traits are involvedin mates of V, from the residual variance from the single tradeoffs in different populations. block ANOVAs of mean standardized data. To make our The high mutational correlations among traits may estimates comparable to those in LINCH (1988), we ad- seem to pose a challenge to the antagonistic pleiotropy justed our estimates of V, to heterozygous effects on the theory of senescence as well; however, the pleiotropy whole genome as described above. Our estimates of VM/ theory does not depend oncontinual input of mutations V,, calculated from all lines, are 1.8 X lo-’ for early with antagonistic effects. Since mutations whichin- fecundity, 3.7 X lo-’ for late fecundity, 0.8 X lo-’ for crease early-life fitness at the expenseof late-& fitness male longevity, and 1.2 X lo-’ for female longevity. 784 D. Houle et al.

These estimates cluster around the value of 1 X lo-’ Mutation-accumulation in finite outbreeding and inbreeding popu- usually assumedto be typical of quantitative traits, based lations. Genet Res. 61: 39-56. CLARE,M. J., and L. S. LUCKINBILL,1985 The effects of gene- on extensive studies of variation in bristle numbers in environment interaction on the expression of longevity. Drosophila. In fact there is substantial variation around 55 19-29. this figure, although the medianover all traits approxi-is CROW,J. F., 1970 Genetic loads and the cost of natural selection, pp. 128-177 in Mathematical Topics in , ed- mately 1 X lo-’ (LYNCH1988). This similarity of V,/V, ited by K.-I. KOJIMA.Springer Verlag, New York. estimates probably is an artifact of the choice of V, to CROW,J. F., 1993a How much do we know about spontaneous standardize the estimates. Data from outbred popula- mutation rates? Env. Mol. . 21: 122-129. CROW,J. F., 1993b Mutation, mean fitness, and . Oxford tions shows a positive correlation between V, and V,, Surv. Evol. BIOI.9: 3-42. when standardized by trait means (HOULE1992). Thus, CROW,J. F., and M. J. SIMMONS,1983 The mutation load in Drosophila, a likelyexplanation for the clusteringof V,/ V, values is pp. 1-35 in The Genetics and Biology ofDrosophila,Vol.3c, edited by M. ASHBURNER,H. L. CARSON andJ. N. THOMPSON,JR. Academic a similar positive relationship between V, and V,. On a Press, London. mean standardized scale, V, tends to be higher for life FELDMAN,M. V., F. B. CHRISTIANSENand L. D. BROOKS, 1980 Evolution history traits than for other traits in outbred popula- of recombination in a constant environment. Proc. Natl. Acad. Sci. USA 77: 4824-4827. tions, eventhough heritabilities for life historytraits are FELSENSTEIN,J., 1974 The evolutionary advantage of recombination. lower, implying that life history traits have dispropor- Genetics 78: 737-756. tionately large environmental variances. SimilarV,/ V, FINCH,C. E., M. C. PIKEand M. WIITEN,1990 Slow mortality rate accelerations during aging in some animals approximate that of values may conceal considerable variation in mutational . Science 249: 902-905. parameters. HAIGH,J. 1978 The accumulation of deleterious in a popula- tion. Theor. Popul. Biol. 14: 251-267. This work was supported by National Institutes of Health (NIH) HAMILTON,W. D., 1966 The moulding of senescence by natural se- grant AGO7748 to B.C. The preparation of this manuscript was sup lection. J. Theor. Biol. 12: 12-45. ported in part by National Science Foundation grant BIR-9014265 HILL,W. G., 1982 Rates of change in quantitative traits from fixation and NIH grant GM36827 to M. LYNCH.We thankJ. CROW,M. LYNCH, of new mutations. Proc. Natl. Acad. Sci. USA 79 142-145. T. MACKAY,and two reviewers for comments on the manuscript. HIRAIZUMI,Y., 1961Negative correlation between rate of develop Thanks to I. ADDIS,I. CHEN,J. CHUN,E. HAN,J. HOULE,D. MATZKOW, ment and female fertility in . Genetics 46 615-624. T. NORMAN,A. PRINGLE, REYNOLDS,T. ROBBINS SCHUMACHER, S. S. HOULE,D., 1991 Genetic covariance of fitness correlates: what K. UZENDOSKIand L. Yu for their assistance in performing these genetic correlations are made of and why it matters. Evolution 45 experiments. 630-648. HOULE,D., 1992 Comparing and variability of quantita- LITERATURE CITED tive traits. Genetics 130 195-204. HOULE,D., D. K. HOFFMASTER,S. A~SIMACOPOULOS B. CHARLESWORTH, and ASHBURNER,M. 1989 Drosophila. ALaboratory Handbook. Cold 1992 The genomic mutation rate for fitnessin Drosophila. Spring Harbor Laboratory, Cold Spring Harbor, N.Y. Nature 359: 58-60. ASHBURNER,M., and J. N. THOMPSON,1978 The laboratory culture of HUGHES,K A., and B. CHARLESWORTH,1994 A of se- Drosophila, pp. 1-109 in The Genetics and Biology ofDrosophila, nescence in Drosophila. Nature 367: 64-66. Vol. 2A, edited by M. ASHBURNER and T. R. F. WRIGHT.Academic KEIGHTLEY,P. D., and W. G.HILL, 1988 Quantitativegeneticvariability Press, London. maintained by mutation-stabilizing selection balance in finite BARTON,N., 1990 Pleiotropic models of quantitative variation. Ge- populations. Genet. Res. 52: 33-43. netics 124: 773-782. KEMPTHORNE,K., 1957 AnIntroduction to GeneticStatistics. Wiley, BARTON,N. H., and M. TURELLI,1989 Evolutionary quantitative New York. genetics: how little do we know? Annu. Rev. Genet. 23: 337-370. KONDRASHOV,A. S., 1988 Deleterious mutations and the evolution of CHARLESWORTH,B., 1978 A model for the evolution of Y chrome . Nature 336 435-440. somes and dosage compensation. Proc.Natl. Acad. Sci. USA KONDRASHOV,A. S., and J. F. CROW,1993 A molecular approach to 75: 5618-5622. estimating the human deleterious mutation rate. Hum. Mutat. 2: CHARLESWORTH,B., 1984 The evolutionary genetics of life histories, 229-234. pp. 117-133 in , edited byB. SHORROCKS. KONDRASHOV,A. S., and M. TURELLI,1992 Deleterious mutations, ap Blackwell, Oxford. parent stabilizing selection and the maintenance of quantitative CHARLESWORTH,1987B., The heritability of fitness, pp. 21-40 in variation. Genetics 132: 603-618. : Testing the Alternatives, edited by J. BRADBURY LANDE, R., and D. W. SCHEMSKE,1985 The evolution of self- and M. B. ANDERSSON.John Wiley & Sons, London. fertilization and inbreeding depression in plants. I. Genetic mod- CHARLESWORTH,B.,199Oa Mutation-selection balance and the evo- els. Evolution 39 8-23. lutionary advantage of sex and recombination. Genet. Res. 55: LINDSLEY,D. L., and G. ZIMM,1992 The Genome of Drosophila mela- 199-221. nogaster. Academic Press, San Diego. CHARLESWORTH,B., 1990b Optimization models, quantitative genet- LUCKINBILL, L. S., R. ARKING, M. G.CLARE, w. c.CIROCCO and s.A. BUCK, ics, and mutation. Evolution 44: 520-538. 1984 Selection for delayed senescence in Drosophilamelano- CHARLESWORTH,B., 1994 Evolution in Age-Structured Populations, gaster. Evolution 38: 996-1003. Ed. 2. Cambridge University Press, Cambridge. LYNCH,M., 1985 Spontaneous mutations for life-history characters in CHARLESWORTH,B., and D. CHARLESWORTH, 1985Genetic variation in an obligate parthenogen. Evolution 39: 804-818. recombination in Drosophila. I. Responses to selection and pre- LWCH,M., 1988 The rate of polygenic mutation. Genet. Res. 51: liminary genetic analysis. Heredity 54: 71-83. 137-148. CHARLESWORTH,B., D. CHARLESWORTHand M. T. MORGAN,1990 Ge- LWICH,M., and W. GABRIEL, 1990Mutation load and the survival of netic loads and estimates of mutation rates in highly inbred plant small populations. Evolution 44: 1725-1737. populations. Nature 347: 380-382. MACKAY,T. F. C., 1986 A quantitative genetic analysis of fitness CHARLESWORTH,B., A. LAPID and D. CANADA, 1992 The distribution of and its components in Drosophila melanogaster. Genet. Res. 47: transposable elements within and between chromosomes in a 59-70. population of Drosophila melanogaster. I. Element frequencies MEDAWAR, P. B.,1952 An Unsolved Problem in Biology. H. K. Lewis, and distribution. Genet. Res. 60: 103-114. London. CHARLESWORTH,D., M. T. MORGANand B. CHARLESWORTH,1993 MONTGOMERY,E. A., B. CHARLESWORTHand C. H. LANGLEY,1987 A test Mutation and Life History 785

for the role of natural selection in thestabilization of transposable adaptive female choice: the effect of mutation-selection balance. element copy number in a population of Drosophila melano- Evolution 42 817-820. gaster. Genet. Res. 49: 31-41. ROBERTSON,A., 1967 The nature of quantitative genetic variation, Mm,T., 1964 The genetic structure of natural populations of pp 265-280 in Heritage from Mendel,edited by R A. BRINK.Uni- Drosophila melanogaster. I. Spontaneous mutation rate of poly- versity of Wisconsin, Madison. genes controlling viability. Genetics 50 1-19. ROBERTSON,F.W., 1963 The of growth in MUKAI,T., 1988 Genotype-environment interaction in relation the Drosophila: 6. The genetic correlation between the duration of the maintenance of in populations of Drosoph- the lmal period and body size in relation to larval diet. Genet. ila melanogaster, pp. 21-31 in Proceedings of the Second Inter- Res. 4: 74-92. national Conferenceon QuantitativeGenetics, edited byB. s. ROPER,C., P. PIGNATELLIand L. PARTRIDGE,1993 Evolutionary effects WEIR,E. J. EISEN,M. J. GOODMAN and G. NAMKOONG.Sinauer, of selection on age at reproduction in larval and adult Drosophila Sunderland, Mass. melanogaster. Evolution 47: 445-455. MUKAI,T., and T. YAMAZAKI,1968 The genetic structure of natural ROSE,M., 1982 Antagonistic pleiotropy, dominance, and genetic populations of Drosophila melanogaster. V. Coupling repulsion variation. Heredity 48: 63-78. effect of spontaneous mutant polygenes controlling viability. ROSE,M. R., 1984a Genetic covariation in Drosophila lifehistory: Genetics 59 513-535. untangling the data. Am. Nat. 123 565-569. MUKAI,T., and T. YAMAZAKI,1971 The genetic structure of natural ROSE,M. R., 1984b Laboratory evolution of delayed senescence in populations of Drosophila melanogaster. X. Developmental time Drosophila melanogaster. Evolution 38: 1004-1010. and viability. Genetics 69 385-398. ROSE, M. R.,1991 of Aging. Oxford, New York. Mum, T., S. CHIGUSAand I. YOSHIKAWA,1965 The genetic structure ROSE,M. R., and B. CHARLESWORTH,1981a Genetics of life-history in of natural populations of Drosophila melanogaster. 111. Domi- Drosophila melanogaster. I. Sib analysis ofadult females. Genetics nance effect of spontaneous mutant polygenes controlling 97: 173-186. viability in heterozygous genetic backgrounds. Genetics 52: ROSE,M. R., and B. CHARLESWORTX,1981b Genetics of life-history in 493-501. Drosophila melanogaster. 11. Exploratory selection experiments. MUKAI,T., S. I. CHIGUSA,L. E. MEITLERand J. F. CROW,1972 Mutation Genetics 97: 187-196. rate and dominance of genes affecting viability in Drosophila melanogaster. Genetics 72 335-355. SAS Institute, Inc. 1988a SAS ProceduresGuide, Release 6.03 MUKAI,T., R. CARDELLINO,K WATANABE T. and J. F. CROW,1974 The Edition. SAS Institute, Inc., Cary, N.C. genetic variance for viability and its components in a local popu- SAS Institute, Inc., 1988b SAS/STAT User’s Guide,Release 6.03Edi- lation of Drosophila melanogaster. Genetics 78: 1195-1208. tion. SAS Institute, Inc., Cary, N.C. NOZAWA,IC, 1956 The effects of the environmental conditions on SERVICE, M. P. 1993 Laboratory evolution of longevity and reproduc- Curly in Drosophila melanogaster. Jpn. J. Genet. 31: tive fitness components in male fruit flies: mating ability. Evolu- 165-171. tion 47: 387-399. OHNISHI,O., 1977a Spontaneous and - SERVICE,P.M., E. W. HUTCHINSONand M. R.ROSE, 1988 Multiple induced mutations controlling viability in Drosophilamelano- genetic mechanisms for the evolution of senescence in Drosophila gaster. 111. Heterozygous effectof polygenic muta6ons. Genetics melanogaster. Evolution 42 708-716. 87: 547-556. SIMMONS,M. J., and J. F. CROW,1977 Mutations affecting fitness in OHNISHI,O., 197% Spontaneous and ethyl methanesulfonate- Drosophila populations. Annu. Rev. Genet. 11: 49-78. induced mutations controlling viability in Drosophilamelano- SIMMONS,M. J., C.R. PRESTON and W. R. ENGELS,1980 Pleiotropic gaster. 11. Homozygous effects of polygenic mutations. Genetics effects on fitness of mutations affecting viability in Drosophila 87: 547-556. melanogaster. Genetics 94: 467-475. PARTRIDGE,L., and N. H. BARTON,1993 Optimality, mutation and the SOW, R. R., and F. J. ROHLF,1981 Biometry, Ed. 2. W. H. Freeman, evolution of . Nature 362: 304-31 1. San Francisco. PARTRIDGE,L., and K FOWLER,1992 Direct and correlated responses S~D,J. A., 1971 An estimate of heterosis in Drosophila melanogaster. to selection on age at reproduction in Drosophila melanogaster. Genet. Res. 18: 97-105. Evolution 46 76-91. WILLIAMS,G. C., 1957 Pleiotropy, natural selection, and the evolution POMIANKOWSKI,A. N., 1988 The evolution of mate preferences for of senescence. Evolution 11: 398-411. male genetic quality. Oxford Sum. Evol. Biol. 5: 136-184. YOSHIMARU, H., and T. MUM, 1985 Relationships between the poly- PROUT, T.,and F. MCCHESNEY,1985 Competition among immatures genes affecting the rate of development and viability in Drosophila affects their fertility when they grow up: population dynamics. melanogaster. Jpn. J. Genet. 60: 307-334. Am. Nat. 126 521-558. RICE,W. R., 1988 Heritable variation in fitness as a prerequisite for Communicating editor: T. F. C. MACKAY