Fly stocks:

Our mapping population consisted of a set of 92 RIL (Nuzhdinet al. 1997) derived from a cross between two isogenic strains, the Oregon-R (Ore) (Lindsley and Zimm 1992) and the Russian 2b strain, an isogenic line derived from a population of flies selected for decreased male sexual activity (Pasyukova and Nuzhdin 1993). The procedure used to construct the RIL has been previously described (Nuzhdinet al. 1997) and is summarized here. F₁ offspring of the cross between the Ore and 2b strains were backcrossed to the 2b strain and then randomly mated for four generations. After the last generation of random mating, 200 male-female pairs were used to create sublines from this population by carrying out brother-sister matings for 25 generations. Ninety-eight RIL were subsequently created from these sublines. Of the original 98 RIL, 92 were extant at the time this experiment was carried out.

Early and late-age fecundity assay:

The total number of eggs laid by single females over a 2-day period was used as an estimate of fecundity at 1 and 4 weeks of age. Two replicate sets of flies of each genotype were set up simultaneously so that early and late-age fecundity could be measured on different females from each RIL. We measured early and late-age fecundity on different females to minimize the effects of handling at early age on late-age fecundity and to decrease the influence of nongenetic phenotypic correlations between early and late-age fecundity. One limitation of this design, however, is that because fecundity was measured on different females at early and late age we could not calculate the phenotypic correlation between early and late-age fecundity.

To produce females for each fecundity estimate, 15–20 pairs of flies from the Ore and 2b strains and each of the 92 RIL were allowed to lay eggs over a 4-day period in egg-collecting chambers (described below). From these chambers, 50 first or second instar larvae of each genotype were collected and placed in vials containing 5 ml of standard cornmeal/agar/molasses medium to standardize larval density. This procedure was repeated with five replicate vials per line. Larval density was controlled in this manner to reduce the influence of variation in larval density on female size at eclosion, a trait that is positively correlated with fecundity (Robertson 1957; Tantawy and Rahka 1964; Partridgeet al. 1986; Nunney 1996; Zwaanet al. 1995). From each replicate vial, three virgin females that emerged on the same day were collected and placed in a vial containing standard fly food. To provide mates, six young males (<1 week old) of the Samarkand (Sam) strain were added to each vial (two males:one female per vial). Sam (an unrelated isogenic laboratory stock) males were used as mates to standardize the potential effects of male genotype on female fecundity.

Age-specific fecundity of individual females was measured at young (4–5 days) and old (28 days) age in a fashion similar to that of Houleet al. (1994) and Service (2000). These dates were chosen because reproductive output typically peaks in D. melanogaster at ∼6–12 days posteclosion and declines beyond that point (Houleet al. 1994; Tataret al. 1996; Gasseret al. 2000). At each age, single females were removed from each vial and placed in an egg-laying chamber with a single Sam male. Egg-laying chambers consisted of standard 10-ml fly vials, containing 1 ml of hardened 2% agar to provide a source of moisture and 1 ml of fly food to provide a site for oviposition. Food was placed on the flat side of a hardened foam plug (which was also used to cap the vial). The vial was then inverted so that the plug of food was on the bottom of the vial. Females were allowed to lay eggs on food plugs for 24 hr, after which each plug was removed to count eggs. A fresh food-containing plug was replaced in the vials and the above process was repeated for a second 24-hr period. After this second day of egg laying, all females were discarded.

At each age, fecundity estimates were made on 15 females/line (five replicate vials containing 3 females/vial/line) for a total of 2760 observations (1380 young flies and 1380 old flies).

Statistical analyses:

We tested for differences in age-specific fecundity between the inbred parental strains (Ore and 2b) and the RIL in separate analyses. Differences in fecundity between the parental strains were tested using a mixed-model ANOVA according to the model y = μ + A + S + (A × S) + R(A × S) + error, where μ is the overall mean, A is the fixed effect of age (1 or 4 weeks), S is the fixed effect of parental strain (Ore or 2b), and R is the random effect of the replicate nested within age and strain. Flies in a particular replicate were those that had shared a vial until the age that we measured fecundity.

To assess genetically based differences in fecundity among the RIL, we used three separate analyses. The first two analyses used a random-effects ANOVA to test for genetic differences among lines at 1 and 4 weeks of age and provided estimates for the among-line variance components for fecundity at each age. The model partitioned the random effects of line (L) and replicate (R) within line and residual error according to the model y = μ + L + R(L) + error for flies in each age group. In the third model we used a mixed-model ANOVA on the entire data set to examine the potential for a line-by-age interaction that would in essence tell us if the effect of age on fecundity was similar among lines. For this analysis we used the model y = μ + A + L + (A × L) + R(A × L) + error, where A is the fixed effect of age on fecundity and all other effects are random.

All statistical analyses were carried out using SAS V.9.1. The PROC GLM and VARCOMP procedures were used for the analyses of variance on RIL within each age and for estimating variance components within each age. The PROC MIXED procedure was used for both of the mixed-model analyses and the significance of random effects was determined using likelihood-ratio tests (Littellet al. 2002). Phenotypic data were ln-transformed to meet the assumptions of ANOVA.

Variance components from the random-effects analyses above were used to calculate the proportion of the total phenotypic variation in fecundity explained by genetic differences among lines at each age (also separately estimated as the coefficient of genetic variation) and the genetic correlation of fecundity at young and old ages. The coefficient of genetic variation (CV_G) was calculated at each age as CV_G = 100(V_L)^1/2/, where V_L is the among-line variance component and is the overall mean fecundity (Houle 1992). The genetic correlation across ages (r_GA) was computed as cov₁₂/(σ_L1σ_L2) (Robertson 1959), where cov₁₂ is the covariance among-line means at 1 and 4 weeks of age and σ_L1 and σ_L2 are the square roots of the among-line variance components of fecundity at 1 and 4 weeks of age from the reduced model analyses.

QTL mapping:

Molecular markers used to determine the genotype of the RIL were the cytological insertion sites of the roo transposable element (Nuzhdinet al. 1997). Eighty-one informative markers were used (Nuzhdinet al. 1997; Leips and Mackay 2000) with an average spacing between markers of 7.9 cM. Spacing between markers was estimated from the observed recombination (r) frequencies between pairs of markers using the Kosambi map function 100d_M = 0.25 ln[(1 + 2r)/(1 − 2r)]. The distance between markers in this study is slightly greater than that of previous studies using these RIL (Nuzhdinet al. 1997; Wayneet al. 2001; Leips and Mackay 2002) because the loss of 6 of the original 98 lines reduced the number of observed recombination events in the mapping population.

QTL mapping was done using composite interval mapping (Zeng 1994) in QTL Cartographer (Version 1.14) and as outlined in Leips and Mackay (2002). This mapping procedure tests the hypothesis that an interval between adjacent markers contains a QTL affecting the quantitative trait, while controlling for the effects of linked QTL outside of the test interval. Markers on which the QTL analyses were conditioned were based on a forward-backward elimination stepwise regression analysis. Because the results of each analysis can be sensitive to the conditioning window used around each test interval, we tested a range of window sizes (5, 10, 15, and 20 cM) to evaluate the effect of window size on the likelihood ratios for each QTL. On the basis of the results from this set of analyses we used a window size of 10 cM because QTL identified with this window size were also determined to be significant in all analyses regardless of window size and so represent a conservative choice. The significance level for each QTL analysis was determined by randomly permuting the fecundity data 1000 times and calculating the maximum-likelihood ratio statistic across all test intervals for each permutation. LR statistics from the original data that were exceeded by the permutation maximum LR statistics <50 times were considered significant at ∝ = 0.05 (Churchill and Doerge 1994; Doerge and Churchill 1996).

We used ANOVA (PROC GLM in SAS V.9.1) to test for epistasis first by looking for significant pairwise interactions between QTL that had significant additive effects on fecundity on the basis of our mapping analysis. For each interaction, the genotype of each marker (homozygous for either the Ore or the 2b allele) closest to each significant QTL peak was used to evaluate the significance of marker interactions on fecundity. Because epistasis may also occur between QTL without main effects on the trait (Mackayet al. 2005), we performed a whole-genome screen for pairwise interactions between all possible pairs of markers using a two-way ANOVA where y = μ + M_i + M_j + (M_i × M_j) + error, where M is the genotype of each marker at positions i and j in each line. Using 81 markers means that we tested 3240 possible interactions. As such, we expect 162, 32.4, 3.24, and 0.324 significant interactions by chance alone at P < 0.05, < 0.01, < 0.001, and < 0.0001, respectively.

Genetic variation in age-specific fecundity (parental strains):

The average fecundity of the Ore strain (±1 standard error) at 1 and 4 weeks of age was 19.5 ± 3.0 and 10.06 ± 2.7 eggs, respectively. The average fecundity of the 2b strain at 1 and 4 weeks was 10.87 ± 2.6 and 8.53 ± 1.4 eggs, respectively. Despite these differences there was no significant difference in fecundity between the parental strains when averaged across both ages (F_1,16 = 2.51, P = 0.13). There was also no significant effect of age (F_1,16 = 3.36 P = 0.08) nor was there a significant age by parental strain interaction (F_1,16 = 1.23, P = 0.28).

Genetic variation in age-specific fecundity (RIL):

Fecundity at 1 and 4 weeks averaged over all RIL was 14.8 and 15.2, respectively (Table 1) and the range of fecundity among the RIL at 1 and 4 weeks of age was similar (week 1: 5–23 eggs/female; week 4: 3–27 eggs/female). Thus, fecundity changed very little between weeks 1 and 4 when averaged over all lines.

On the basis of the ANOVA at each age, however, we found significant differences in fecundity among the RIL at both 1 (F_91,368 = 2.03, P < 0.0001) and 4 (F_91,368 = 2.44, P < 0.0001) weeks of age. Notably, the genetic component of the total variation in fecundity at 4 weeks was almost twice what it was at 1 week (Table 1). This is not because there was less phenotypic variation in older aged individuals. In fact, the amount of residual variance in fecundity was 30% higher in the analysis of 4-week-old females compared to that of 1-week-old females. This increase in the environmental component of variance with age is consistent with other studies of fecundity (Rose and Charlesworth 1981) and longevity (Charlesworth and Hughes 1996) in Drosophila and suggests that older individuals may be more sensitive to environmental variation than younger flies (Charlesworth and Hughes 1996). Another possible explanation is that because older flies have experienced a greater range of environments the cumulative effect of this variation results in greater phenotypic variation among older aged individuals.

Although the RIL differed in fecundity at each age, the effect of age on fecundity differed dramatically among lines (Figure 1). The line-by-age interaction term was significant (χ²₍₁₎ = 23.4, P < 0.0001) and the genetic correlation of fecundity at 1 and 4 weeks of age was not significantly different from zero (Table 1). Together these results suggest that genes contributing to the variation in early age fecundity are distinct from those producing variation in fecundity at 4 weeks and/or that the allelic effects of loci that contribute to genetic variation in fecundity vary with age.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.—

The average fecundity of each line at weeks 1 and 4. Fecundity was measured as the number of eggs laid by a single female over a 2-day period. Fecundity counts within lines were measured for different females at each age.

Age-specific QTL for fecundity:

We identified two or three QTL affecting fecundity at 1 week of age, one on the second chromosome and another one or two on the third chromosome (Figure 2). Given the proximity of the two QTL on the third chromosome and the confidence intervals around each, it is not clear if these are distinct QTL. The most likely position of the QTL on the second chromosome is at cytological position 34E with the 2-LOD support interval (Lynch and Walsh 1998) extending from cytological positions 30D to 38A. Variation at this QTL explained 10% of the variation in fecundity among lines at 1 week of age. The additive effect of the Ore allele at this locus increased the fecundity of females by 0.78 eggs/day compared to the effects of the 2b allele. One of the QTL on the third chromosome has the highest likelihood of being at position at 85F (2-LOD support interval 73D–87E). Variation at this locus explained 14% of the genetic variation in fecundity, and the Ore allele at this locus also increased the fecundity by 0.74 eggs/day compared to the effect of the 2b allele. The third QTL appears at position 87B (2-LOD support interval also ranges from 73D to 87E) and explains an additional 10% of the genetic variation. At this locus, the Ore allele increases fecundity by 0.67 eggs/day relative to the 2b allele. Interestingly, the direction of effects of the 2b alleles is consistent at both sites and may reflect the fact that this line is derived from a population selected for decreased male mating activity. However, the 2b and Ore parental strains do not differ genetically for fitness (Wayneet al. 2001) so it is unlikely that the effects seen here are due to a general effect of 2b alleles in reducing overall fitness.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.—

Composite interval mapping results indicating the positions of QTL affecting fecundity at each age. The threshold value for significance at each age is given by the horizontal lines. Positions of informative markers are denoted by triangles on the x-axes. Lines above peaks indicate 2-LOD support intervals for the location of the QTL.

Unexpectedly, we found no QTL that influenced fecundity at 4 weeks of age, despite the fact that the ANOVA indicated a higher degree of genetically based variation in late-age fecundity than in early age fecundity. This suggests that genetic variation in fecundity among lines at the older age is due to alleles with small late-age specific effects. Our observation that there was no genetic correlation of fecundity between young and old ages is thus explained by the diminution of the effects of QTL affecting early age fecundity.

Effects of epistasis among marker loci on fecundity:

None of the pairwise tests for epistasis among the QTL with main effects on fecundity were significant. In the global test for pairwise interactions among loci at early age, a total of 97 markers exhibited significant epistatic effects on fecundity (P < 0.05). Of these, 83 interactions had P-values between 0.01 and ≤0.05, 13 had P-values between 0.001 and <0.01, and only one interaction had a P-value of <0.001. In a comparable test using late-age fecundity a total of 98 markers exhibited significant epistatic effects on fecundity (P < 0.05). Of these, 92 interactions had P-values between 0.01 and ≤0.05, and 6 had P-values between 0.001 and <0.01. Given 3240 possible pairwise interactions, the number of significant interactions that we found were well within the numbers of interactions expected by chance alone (expected: 162 P ≤ 0.05, 32.4 P < 0.01, and 3.2 P < 0.001).

Age-specific effects of QTL:

Our results show that the relative influence of genes regulating the age schedule of reproduction changes with age. This interpretation is supported by the high degree of variation among lines in the effect of age on fecundity and the lack of a genetic correlation between early and late-age fecundity in the RIL. Combining the QTL results with those from the analyses of genetic components of variation leads to the interpretation that variation in fecundity at the younger age was determined at least in part by a few genes of moderate effect but fecundity at old age was determined by many more loci of smaller effect. While it is unclear how many actual genes within the QTL regions contribute to the variation in fecundity at the early age, it is clear that their relative effect on fecundity diminishes greatly with age.

Genes with age-specific effects on other fitness components have been implicated in many other studies of D. melanogaster (Kosuda 1985; Engstromet al. 1989; Hughes and Charlesworth 1994; Hughes 1995; Charlesworth and Hughes 1996; Promislowet al. 1996; Tataret al. 1996; Pletcheret al. 1998, 1999; Macket al. 2000; Yampolskyet al. 2000; Curtsinger and Khazaeli 2002; Hugheset al. 2002; Snoke and Promislow 2003) and a few studies have begun to map the QTL underlying these age-specific effects on mortality rates (Curtsinger and Khazaeli 2002; Nuzhdinet al. 2005) and metabolic rates (Khazaeliet al. 2005). Our findings of the age-specific effects of QTL are similar to those seen by Khazaeliet al. (2005). In their study, metabolic rates at days 16 and 29 posteclosion and life span appeared to be affected by the same QTL, but these QTL did not affect metabolic rates at the youngest and oldest ages examined in their study. The results from Nuzhdinet al. (2005) are also consistent with ours in that no QTL identified in their study had a significant influence on mortality rate at all ages studied. Even more intriguing, their results suggested that at two of the QTL, the allelic effects on mortality rate were negatively correlated across different ages, with the same allele having either a positive or a negative effect on mortality, depending on age.

Given the transient nature of the effects of QTL on senescent phenotypes, an important goal for future studies will be to identify the factors that give rise to age-specific genetic effects. One possibility is that age-specific mutational effects result from genotype-by-environment interactions. Under this scenario, genetic influences on the phenotype depend on the internal physiological conditions; age-related changes in the physiology of the organism modulate the effects of these loci such that they have a notable phenotypic effect only within a certain range of conditions. In addition, age-specific changes in gene expression may give rise to age-specific effects. Numerous studies have demonstrated age-dependent changes in gene expression (Roginaet al. 1998; Jinet al. 2001; Weindruchet al. 2001; Pletcheret al. 2002; Seroudeet al. 2002; McCarrollet al. 2004; Kimet al. 2005) and it may be that the age at which the effects of allelic variation are notable coincides with the ages of peak expression of these loci. Along these same lines, a growing body of evidence implicates age-related changes in chromatin structure, which directly regulates gene expression, as a mechanism for regulating aging (Chang and Min 2002; Roginaet al. 2002; Tissenbaum and Guarente 2002; Issa 2003; Jaenisch and Bird 2003). It is possible that genetic variation in the enzymes controlling age-related changes in chromatin remodeling produces the age-specific effects of particular loci on senescence.

Gene action:

We found little compelling evidence for the influence of epistasis on fecundity among our RIL at either age. Admittedly, the method used to correct for the number of expected false positives is conservative and so does not preclude the potential importance of epistasis on fecundity in our mapping population. A more sophisticated statistical approach that has greater power to detect epistasis might have been more useful in this regard (Kaoet al. 1999) but such methods require much larger sample sizes than we had in this study. If our interpretation about epistasis is correct—that there are few if any significant epistatic interactions among loci that affect fecundity—then this strengthens our interpretation of the QTL results. This is because the composite interval mapping method used produces biased estimates of the position and marginal effects of QTL, given any amount of epistasis and linkage between epistatic QTL (Kao and Zeng 2002).

The lack of epistatic interactions affecting fecundity is in stark contrast with previous work on these lines, which found extensive epistasis among QTL affecting virgin life span (Leips and Mackay 2000; Mackayet al. 2005). Interestingly, when these lines were measured for mated longevity, only a single pair of markers appeared to interact epistatically. Thus, differences in the mating status appeared to influence the degree to which epistasis affected longevity. More work is necessary to evaluate the extent to which epistasis is dependent on the environmental/physiological condition and whether different traits are more or less affected by epistatic interactions.

Implications for evolutionary theories of aging:

Many studies on the genetic basis of senescence have used age-specific variance components to test the predictions of the MA and AP theories of aging. These studies typically use breeding designs that allow estimation of the additive and dominance genetic components of variation in a trait with age that can then be evaluated in light of the predictions of each theory. While increases in the additive component of variation (V_A) with age can be expected from either model of aging (Charlesworth and Hughes 1996), a unique prediction of the MA theory is that the genetic component of variation among homozygous lines will increase with age. Using this metric, the results of our study lend support for MA producing variation among lines in reproductive senescence. Within the age span covered in this experiment (ages 1 to 4 weeks) our results are in agreement with most studies of age-specific genetic effects on a number of traits, including fecundity (Engstromet al. 1989; Tataret al. 1996), age-specific mortality (Hughes and Charlesworth 1994; Charlesworth and Hughes 1996; Promislowet al. 1996; Tataret al. 1996; Hugheset al. 2002; Snoke and Promislow 2003), and aspects of male mating success (Kosuda 1985; Hughes 1995; Charlesworth and Hughes 1996; Hugheset al. 2002; Snoke and Promislow 2003). The lack of a significant genetic correlation between early and late-age fecundity adds additional support for the MA theory because this model assumes that the effects of alleles on fitness early in life are uncorrelated with allelic effects on fitness later in life (Partridge and Barton 1993). Under the AP model we would expect to see a negative genetic correlation between early and late-age fecundity. On the basis of these summary statistics we can conclude that MA contributes to the variation in age-specific fecundity in our mapping population.

To examine evidence supporting the AP theory, we compared the results of this study with our earlier study mapping QTL affecting the life span of mated males and females in the same population of RIL (Leips and Mackay 2002). The earlier experiment differed from the current one in that the life-span measurements were made on the offspring of the cross between each RIL and the inbred parental strains, Ore and 2b. Thus, the allelic effects of QTL on life span were estimated in different genetic backgrounds. However, there was only one instance in which life-span QTL were shown to interact epistatically in that earlier study and neither of the two QTL involved were those identified as fecundity QTL. Therefore, comparison of the additive allelic effects of fecundity QTL on life span using these two studies is appropriate in looking for evidence supporting the AP theory. While the antagonistic relationships assumed by AP could presumably exist between any number of traits that influence fitness at early age and that are negatively correlated with longevity (Leroiet al. 2005), one of the most commonly observed trade-offs that support the AP theory is the trade-off between early age fecundity and life span (Rose and Charlesworth 1981; Rose 1991; Mardenet al. 2003; Leroiet al. 2005). Application of this theory to our QTL mapping studies would predict a negative correlation between early age fecundity and life span among the RIL. At the QTL level we would expect that QTL affecting life span and early age fecundity should colocalize and that the allelic effects at these loci should have opposite effects on these two traits. To investigate these possibilities, we first calculated the correlation between the average mated life span of males and females from our earlier study with early and late-age fecundity. As predicted by AP, we did find a significant negative correlation (r = −0.43, P < 0.0001) but interestingly this correlation was between early female fecundity and the life span of males from the RIL × Ore cross. At the QTL level, both of the early age fecundity QTL colocalize with QTL affecting life span and alleles at these loci do indeed have antagonistic effects. Oddly enough and in accordance with the correlation analysis described above, it is the life span of males and not females that exhibits antagonistic allelic effects with female fecundity at early age. At the QTL on the second chromosome the Ore allele increases fecundity by 0.78 eggs/day compared to the 2b allele but decreases male life span by 4 days. The allelic effects at the QTL on the third chromosome are similar to those at the QTL on the second; the Ore allele increases early age fecundity by 0.74 eggs/day but decreases male life span by 5 days. Thus, it appears that the allelic effects of these QTL may exhibit sexual antagonism with alleles having an advantageous effect on one sex but a deleterious effect on the other. Sexual antagonism in QTL studies is not unusual (Nuzhdinet al. 1997; Leips and Mackay 2000; Vieiraet al. 2000; Wayneet al. 2001) although neither is it universal (Curtsinger and Khazaeli 2002). Another possible explanation is that male longevity is genetically correlated with male fecundity (which was not measured) and so would indicate a positive correlation between male and female fitness. It should be noted here that the allelic effects at these QTL potentially represent the combined effects of many genes within the QTL region. Therefore, the actual genes that contribute to the variation in fecundity may be in the same region as, but distinct from, those affecting male life span. Only by identifying the actual loci underlying the variation in these traits can this issue be resolved. Also, genotype-by-environment interactions can influence the sign and magnitude of the allelic effects of QTL (Leips and Mackay 2000; Vieiraet al. 2000) and it may well be that under different environmental conditions a trade-off between fecundity and female life span would be evident. Mardenet al. (2003) found just such a situation in studying the combined influence of the Indy mutation on life span and fecundity. A trade-off between these traits was evident only when flies were reared on a calorically restricted diet. Confirmation of the degree to which particular alleles contribute to trade-offs among traits will require that we not only identify the loci that affect these traits, but also observe the allelic effects on all traits affected in a range of ecologically relevant environments.

From QTL to gene:

Identification of QTL with age-specific effects on life-history traits represents the first of many steps toward understanding the complexities of the genetic basis of variation in these traits. Identification of the actual genes that contribute to the variation in fecundity identified in this study will require fine-scaled mapping of the QTL regions using crosses to deficiency strains and complementation tests to candidate genes within refined QTL regions (e.g., Pasyukova et al. 2000). Once candidates are identified, testing for the effects of naturally segregating variation on fecundity can be accomplished by association mapping studies (e.g., De Lucaet al. 2003). The QTL regions identified in this study contain many candidate genes that are involved in some aspect of reproduction/oogenesis and might contribute to the variation in this study. These genes include daughterless, rho-6, zucchini, kekkon-1, vasa, cactus, and kelch on the second chromosome and maelstrom, jim, rpk, bicoid, poached, and squid on the third chromosome. As there are hundreds of genes within each of these QTL regions, most of unknown function, resolving the actual loci contributing to variation in fecundity will require additional mapping efforts.

One limitation of this study, which is indeed a limitation of QTL studies in general, is that only a limited sample of genetic diversity is represented in our lines. Also, the mapping population used was derived from two inbred laboratory strains and not isolates from a natural population. As such, the generality of our results needs to be tested by repeating this study with independent lines ideally constructed from a natural population. The fact that the parental strains did not differ in fecundity but we were still able to map QTL in the RIL derived from them is not unusual (e.g., Leips and Mackay 2000; Vieiraet al. 2000). These results suggest that QTL with positive and negative effects on fecundity that were fixed for each parental strain were revealed when they appeared in different combinations in the RIL.

Assuming that we can make use of the rapidly developing technological tools to identify and characterize the genetic architecture of life-history traits at the molecular genetic level, many questions will remain after we have the loci in hand. For example, given genetic variation in the age-specific expression of a trait, does variation result from the action of a different subset of genes acting on the trait at different ages or does variation arise from differences in the influence of particular alleles at the same genes with age? Do physiological changes with age modulate allelic effects on traits (in a fashion similar to genotype-by-environment interactions)? And if so, what are the relevant physiological changes that alter these allelic effects? Such questions represent a few of the many challenges that remain in understanding the genetic basis of life-history variation.