Copyright 0 1996 by the Genetics Society of America
Mapping Quantitative Trait Loci Affecting LifeHistory Traits in the Nematode Caenorhabditis elegans David R. Shook,*’+Anne Brooks*9sand Thomas E.Johnson*3 *Institute forBehavioral Genetics, +Department of Environmental, Population, and Organismic Biology, and :Department of Psychology, University of Colorado, Boulder, Colorado 80309 Manuscript received December 1, 1993 Accepted for publication November 17, 1995 ABSTRACT Wehave identified chromosomal regions containing quantitative trait loci (QTLs) specifylnglife history traits in recombinant-inbred strains of the nematode Caenorhabditis elegans. This approach also allows us to examine epistatic interactions between loci and pleiotropic effects on different traits at specific loci. QTLs for mean life span were identified on chromosomes ZZ (near stPlOl), N (stP5) and the X (stP61), and QTLs for fertility were identified on ZZ (maPI), ZZZ (stPl9) and N (stP5I).The QTLs for mean life span accounted for 90% of the genetic component of variance. The loci for mean fertility accounted for 88% of the genetic component of variance. Additional QTLs for temperature-sensitive fertility [ZZ(stP36) and V (stPG)] and internal hatching [ N (stP5)l were also mapped in these crosses. We found evidence for epistatic effects on mean life span between maPl and bP1 (V), and for epistatic effects on mean fertility between stP36 and stP6, between stP98 (ZI) and stP192 (V), between maPl and stPl27 (ZZI), between maPl and stPlO3 (X),and between stP5 and stP6. Negatively correlated, pleiotropic effects on mean life span and internal hatching were found linked to stP5.
B
ECAUSE lifehistory traits are complex and are probably specified by several quantitative trait loci (QTLs) (FALCONER 1989), we have used QTL mapping strategies (LANDER and BOTSTEIN1989; HALEY and W o n 1992; TANKSLEY 1993) to localize loci influencing such traits in the nematode Caenmhabditis elegans. In this study, QTLs for mean and maximum survival, mean and maximum fertility, temperature-sensitive fertility and internal hatching of progeny (bagging) were mapped. In addition tobeing a first step toward identifying genes controlling these traits, this approach also allows us to look at the effects of interactions among loci on individual traits and to find pleiotropic effects of individual loci on different traits (CHEVERUDand ROUTMAN1993). Inferences can also be made about the role of mutation in the evolution of these loci. The accumulation of empirical data such as this should help to resolve questions regarding the importance of such phenomena in the evolution of life history traits. The underlying genetics of phenotypes plays an important role in their evolutionary dynamics (BARTONand TuRELLI 1989) and may be important for the maintenance of genetic variation (TURELLI1988). Epistatic effects among coadapted gene complexes have been postulated to be importantin foundereffect speciation ( ~ Y 1963; R TEMPLETON 1980). C. ekgans is a useful model system for the study of coadapted gene Curresponding authm: Thomas E. Johnson, Campus Box 447, Institute for Behavioral Genetics, University of Colorado, Boulder, CO 80309-0447. E-mail:
[email protected]
Genetics 142 801-817 (March, 1996)
complexes, since it is normally inbreeding but can also outcross. Epistasis has also been postulated to be important in shifting balance theories of evolution (WRIGHT1931) and specifically in fitness peak shifts coupled with reproductive isolation (WAGNERet al. 1994) and as a mechanism for the maintenance of genetic variation (KONDRASHOVand TURELLI1992; GAVRILETS and DE JONG 1993). Mutation has been suggested to be involved both in maintaining genetic variation (LANDE 1975; for current perspectives on this issue, see: BULMER1989; KONDRASHOV and TURELLI 1992; GAVRILETS and DE JONG1993) and in the evolution of genes limiting life span (ROSE 1991; PARTRIDGEand BARTON 1993). However, evidence from HOULEet al. (1994) argues against the sufficiency of mutation as a mechanism for the evolution of senescence in Drosophila mlanogaster, Pleiotropic effects of genes may be involved in the maintenance of genetic variation by mutation (BARTON1990; KONDRASHOV and TURELLI1992). It is a general assumption (e.g., BARTON1990) that pleiotropy is wide spread, withmost genes affecting many different traits. Antagonistic pleiotropy has been suggested as a mechanism for maintaining genetic variation (ROSE1982,1985; but see CURTSINGER et al. 1994) and also as a mechanism for theevolution of genes that limit life span, trading off survival with some early fitness trait (MEDAWAR 1952; WILLIAMS 1957; ROSE1991). We have started our analysisoflifehistory traits by focusing on mean lifespan and fertility, traits commonly postulated to show genetic trade offs (ROSE 1991).
D. R. Shook, A. Brooks T. and
802
While previous studies have sometimes demonstrated trade offs, forexamplein Drosophila (ROSE and CHARLESWORTH 1981a,b; LUCKINBILL et al. 1984; ROSE 1984; but see PARTRIDGE and FOWLER 1992) or in opossums (ALISTAD 1993), these have all been at the level of the whole genome. Evidence that a single genetic locus shows negatively correlated phenotypic variation for two different lifehistory or fitness traits, such as survival and fertility, would be strong supportfor antagonistic pleiotropy as a mechanism for the evolution of senescence; however, there are numerous reasons why a negative genetic correlationmight not be found, even if a tradeoff existed (e.g., VANNOORDWIJKand DE JONC 1986; HOULE 1991). Numerous QTL analyses of life historytraits of commercial interest have been performed in plants (e.g., MANSURet al. 1993; BRADSHAWand STETTLER 1995; KHAVKIN and COE1995; LI et al. 1995), butwe are aware of only one other study mapping QTLs specifying lifehistory traits in animals: EBERTet al. (1993) used segregating populations of C. elegans to localize QTLs for longevity by genotyping the longest-lived members of the populations. Their study identified genetic regions that were associated with increased survival but did not examine life history traits other than survival. We mapped QTLs using recombinant inbred ( R I ) strains that have been described previously (JOHNSON and WOOD 1982;JOHNSON 1987; BROOKS andJOHNSON 1991). The study is facilitated by a multiplex PCR method for marker assessment (WILLIAMS et al. 1992) based on different numbers of Tcl transposable elements between the N2 and Bergerac wild-type strains (EMMONSet al. 1983; LIAOet al. 1983); Bergerac has several hundred Tcls not foundin N2. The use of RIs has the advantage of stable unchanging genotypes that can be examined for multiple distinct phenotypes over a period of months or years. Indeed, studies on behavioral traits in these strains are already underway (B. VANSWINDEREN, personal communication; P. PHILLIPS, personal communication). This study provides a first step toward determining the particular genes that are involved in specifying life history traits and in elucidating the mechanisms underlying the interrelationships among these traits. In conjunction with selection experiments, this type of study will contribute to our understanding of the coordinated evolution oflifehistory traits and the genetic constraints that are placed on evolution. MATERIALSAND
METHODS
General methods and media: Standard nematode culture techniques were used (BRENNER 1974; SULSTON and HODGKIN 1988). Survival media consists of S basal solution (SULSTON and HODGKIN 1988), 10 pg/ml cholesterol and lo9 of Escherichia coli strain OP50 per ml. Fertility assays were done on NGM agar plates (SULSTON and HODCKIN1988) spotted with
E. Johnson
OP50. Stocks were maintained at 20°, and all manipulations were at room temperature unless otherwise indicated. Construction of RI strains: Two wild-type C. elegans varieties were the progenitors of the RI strains used in this analysis: Bristol (N2) and Bergerac (NIGON1949; BRENNER 1974; EMMONS et al. 1979). These strains are roughly 0.2% divergent in nucleotide sequence based on frequency of detectable restriction fragment lengthpolymorphisms (G.J.LITHGOUJ, personal communication). The methods used to produce the RI strains used in these studies were originally describedin JOHNSON and WOOD (1982). These RI strains were constructed at two different times: set 1 consists of 34 RI strains (JOHNSON and woo^ 1982), andset 2 consists of 47 RI strains (E. HUTCHINSON and T.JOIINSON, unpublished).setIn2, potential bias in selecting offspring was prevented by permitting each RI adult to lay eggs for 3-5 days and then picking the first individual larva or adultseen under the dissecting microscope. The sets were tested for differences in each of the traits being measured and for genotype. Assay of survival and bagging: The 81 RI strains plus N2 and Bergerac were assayed for survival as previously described (JOIINSON and WOOD 1982) in two replicate samples of 15 worms each; two strains that became contaminated with fungus or bacteria were not included in the analysis. Transfers were done by mouth pipetting. At each time point the replicate plates were assayed for survival in a "blind" manner by two individuals working coordinately. Worms were scored as alive, deadorcensored;the censored category included worms that were killed by premature hatching of eggs within the body of the hermaphrodite (bagging). The frequency of bagging was also analyzed as an independentlife history trait. Assay of hermaphroditeself-fertility: Self-fertility wasassayed following the methodof FRIEDMAN andJOHNsON (1988) with slight modification. The worms were axenized (made free of bacterial and fungal contamination) by treatment with a 1% NaOCl, 0.25 N KOH solution(EMMONS and YESNER 1984). From the second generation after axenization, 10 L4 hermaphrodites from each RI strain were transferred to individual NGM plates at 20"; the 10 plates were divided into two replicate samples, and 48 and 96 hr later, adults were transferred to new NGM plates. The offspring produced during each 48-hr period were allowed to mature for 24 hr at 20" and then either counted immediately or the plate was placed at 2" until the offspring could be counted (refrigeration prevents further growth and reproduction). Replicate 1 was counted by three different people while replicate 2 was counted only byA.B. One strain was not includedin the analysis due to fungal contamination. Assay of temperature-sensitive fertility: Temperature sensitivitywas determined by placing a replicate set of strains, derived from the same parents as those used for the survival assay, at 25" and observing the number of progeny on the plate at the end of the fertile period. Temperature-sensitive phenotypes for a strain replicate were scored qualitatively as 0 if the worms produced no progeny, 1 if they produced only a few progeny, or 2 if worms produced the usual number of progeny. Assay of molecular markers: The genetic markers used in this study are the Tcl elements used by WILLIAMS et al. (1992). The primer sequences and the expected band sizes can be et al. (1992) with the exception that polyfound in WILLIAMS morphism stP192 has an expected product size of 216 bp (B. WIILIAMS, personal communication). In the final analysis, 34 of the 40 markers described were used. The average spacing of markers was -4.5 cM [based on marker positions in the C. elegans data base, ACEDB (DURBIN and MIEG, 1991)1, with
QTLs for Life 803History Traits threetonine markers on every chromosome, covering roughly 55-60% of the genetic map, assuming coverage of 5 cM on eitherside of each marker. The presence of individual Tcl elementswere assessed via PCR. The PCR primers consist of an oligomer homologous to a unique sequence thatflanks the marker Tcl-site and a primer homologous to one end of the Tcl element (ROSENZWEIG et al. 1983; ROSE et al. 1985). Two additional primers (ATC TTA GGA GCATACATGAGC A and ATA TCC GGG TAG CTT CGG ATC C, generously provided by M. PERRY and B. ROBERTSON) that amplify a 479-bp region of the her-I gene (PERRYet al. 1993) were used as an internal positive control for each reaction. DNA forthe PCRassaywas prepared using methods adapted from BARSTEAD et al. (1991). Twenty to 30 worms from each strain were placed in 30 pl of worm lysis buffer [60 pg/ml proteinase K in 10 mM Tris (pH 8.3),50 mM KCl, and 2.5 mM MgC12] in a 0.2 ml tube suitable for PCR. These lysis reactions were frozen at -70" for 2 1 0 min, incubated at 60" for 1 hr, at 95" for 15 min, and finally cooled to 4" until assay. The PCR was performed following the methods in WIL LIAMS et al. (1992) with some modification. A standard reaction master mix contained the following: PCR reaction buffer [ l o mM Tris (pH 8.3), 50 mM KCI, and 1.5 mM MgCI,]; 0.2 mM each dATP, dCTP, dCTP, anddTTP; 6.25 pmol of appropriate flanking sequence primers for each marker being assayed; 6.25 pmol of Tcl primer per marker, 6.25 pmol each of two her-I specific oligomers that served as an internal positive control; and 1.0 unit Taq polymerase (AmpliTaq, Perkin Elmer). Five microliters of target DNA were added to 45 pl of the above reaction master mix for each individual reaction. The 50-pl reaction was quickly heated to 95" for 1 min, then cycled 30 times at 94" for 30 sec, 58" for 1 min, and 72" for 1 min. After cycling, the reaction was held at 72" for 10 min and then put at 2" until the amplified productscouldbe examined. N2 (negative control) and Bergerac (positive control) DNAs were included as targets for each PCR reaction mix containing a given set of primers. Amplification products were separated on 3% NuSieve 3:l (FMC) agarose gels (SAMBROOK et al. 1989) and visualized with ethidium bromide. Distinct DNA fragments of defined size are easily detectable. Markers were typed as either 0 (N2 type) or 1 (Bergerac type). These data were entered intoMap 1994). After all the datawere Manager (MANLY and CUDMORE tabulated, apparent double crossover events were identified as an indication that incorrect scoring may have occurred. In this case additional analyses were performed until the issue was resolved. Statistical analyses: The phenotypic values for each strain replicate were calculated as follows. Mean life span was calculated from life span determined as the date of egg lay to the midpoint between the last date scored alive and the date of death; only senescent deathswere included in this calculation. Maximum life span was that of the longest-lived worm. Bagging was the fraction of worms that died from matricide due to internal egg hatch. Mean fertility was that of an individual replicate of five worms. Maximum fertility was that of the most fertile worm in a replicate. Both means and maximums were used, despite obvious statistical problems with maxima, because we wished to explore the possibility that mean and maxima are specified by different genetic mechanisms. All statistical analyses were performed using SPSS 4.0 (SPSS Inc. 1990a,b). TheBonferroni correction procedure (GROVE and A N D W E N 1982) was applied toa in all multiple comparisons such that a / N i s the corrected significance level, and N is the numberof multiple comparisons.Following the method of BELKNAP (1992),we have Bonferronicorrected for all multiple tests of linkage by determining the number of sets of
marker loci that are not significantly correlated with each other at a = 0.05 for each chromosome, and then summing across all chromosomes, yielding an estimate of 10 nonredundant linkage groups. This number is likely an overestimate, since some of the "independent" markers on each chromosome are still positively correlated with each other (BEL.KNAP 1992). We have used a = 0.1 (a/10 = 0.01) for alltestsof significance regarding QTL linkage to markers or epistasis among markers, since this study is intended to be a preliminary screen for interesting loci, and since the Bonferroni correction is considered quite conservative. An a of 0.05 was used for all other tests of significance. The proportion of Vpdue to genetic causes was estimated from analysis of variance (ANOVA) of individual worm values for traits as V J ( V,; + I+), where V,; ( &) is the genetic variance (the component of variance due to genetic differences between RI strains) (HEGMANN and POSSIDENTE 1981) and V, is the environmental variance (&), the component of variance between individuals within strains. The standard error of V,/ VI. was calculated as the square root of the sampling variance of V,;/ &; (2(1
+ (k
-
l)t)'(l
-
t)'/k(k
-
1 ) ( N - 1))"'
(FALCONER 1989, pg. 182), where t is the intrastrain correlation and is equal to VJ &,for RI strains (J. DEFRIES, personal communication). N is the number of strains and k is the geometric average of individuals per strain. RI strains are derived from multiple generations of breeding, therefore the recombination fraction, r, is calculated as r = R/ (2(1 - R ) ) ,where R is the observed frequency of recombination between two loci in selfed RI strains (HALDANE and WADDINGTON 1931; MANLYand CUDMORE 1994). ris the recombination frequency that would be found in a standard backcross with one generationof recombination. We use r to compare our results with the positions of the markersin ACEDB (DURBIN and MIEC 1991). The QTL maps based on thecomparison ofN2 us. Bergerac alleles are presented in terms of the Tvalues for each marker obtained from the regression analysis. The regression equation can be written as below yt
=p
+ P A 4 + P,S + P,\,,YV(MYSJ + E , ,
where y, is the trait value for the zth individual, p is the overall trait mean, M, is a dummy variable for marker (coded - 1 for N2, 1 for Bergerac),S, is a dummy variable for set (coded - 1 for set 1, 1 for set 2), and theP s are the respective regression coefficients, and is the error term. The T statistic for the regression coefficient for a given marker is positive if the mean of strains carrying the Bergerac genotype is higher at that marker and negative if the mean of strains carrying the N2 genotype is higher. Each marker was tested to determine if that marker was associated with a significant difference in the distribution of life span, using the Lee-Desu D statistic (LEE andDEW 1972; SPSS INC. 1990b) for comparisons of survival function. Rather than using individual life spans, which would be inappropriate, the strain mean for life span was taken as the "time of death" for that strain; while this may result in some loss of power, it should not result in an increase of false positives. This test compares the distribution of survival times and so has the potential to be more powerful than simple comparisons of mean life spans. Significant differences in survival function are indicated on the linkage map displaying QTLs for mean life span. The proportionof K, explained by each QTL was calculated as the percentage of total phenotypic variance due to marker
56
804
D. R. Shook, A. Brooks and T. E. Johnson
TABLE 1 Correlation between replicates
Correlation coefficient Trait Mean life span Maximum life span
Mean fertility Maximum fertility Bagging Temperature
