Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

Transcription

1 The Evolution of Phenotypic Life-History Trade-Offs: An Experimental Study Using Drosophila melanogaster Author(s): Armand M. Leroi, Sung B. Kim, Michael R. Rose Source: The American Naturalist, Vol. 144, No. 4 (Oct., 1994), pp Published by: The University of Chicago Press for The American Society of Naturalists Stable URL: Accessed: 31/07/ :25 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. The University of Chicago Press and The American Society of Naturalists are collaborating with JSTOR to digitize, preserve and extend access to The American Naturalist.

2 Vol. 144, No. 4 The American Naturalist October 1994 THE EVOLUTION OF PHENOTYPIC LIFE-HISTORY TRADE-OFFS: AN EXPERIMENTAL STUDY USING DROSOPHILA MELANOGASTER ARMAND M. LEROI,* SUNG B. KIM, AND MICHAEL R. RoSE Department of Ecology and Evolutionary Biology, University of California, Irvine, California Submitted Janiuaty 13, 1993; Revised August 27, 1993; Accepted September 4, 1993 Abstract.-Experimental manipulation of an organism's environment can reveal trade-offs among life-history and physiological traits such as fecundity and energy reserves. Such phenotypic trade-offs have been suggested to reflect evolutionary constraints. We test this idea by asking whether the selection responses of laboratory populations of Drosophila melanogastet can be predicted from a trade-off between fecundity and starvation resistance revealed by manipulating levels of dietary yeast. Fecundity and starvation resistance vary inversely over levels of dietary yeast: at high yeast levels, females have a high fecundity and low starvation resistance; at low yeast levels, females have a low fecundity and high starvation resistance. We examine the role of this trade-off in two independent sets of populations that have been selected for increased starvation resistance. We find that the joint selection responses of mean fecundity and starvation resistance cannot be predicted from the phenotypic trade-off in either the ancestral or control populations. This result implies that at least some of the genetic variation and covariation on which selection acted originated in physiological pathways not involved in the phenotypic trade-off. We also find that selection has altered the slope of the phenotypic trade-off such that, across yeast levels, the reproductive cost of a unit gain in starvation resistance is less in the starvation-resistant populations than in the controls. Reproduction is often thoughto incur physiological costs manifest as decreases in other life-history traitsuch as adult survival (Williams 1966; Partridge and Harvey 1985; Reznick 1985; Bell and Koufopanou 1986; Parker and Maynard Smith 1990; Lessells 1991; Smith 1991; Roff 1992; Stearns 1992). The physiological relations that give rise to such costs are referred to as "trade-offs," especially when the costs are a consequence of different fitness componentsharing common material resources. Such physiological trade-offs are interesting for at least two reasons. First, they can cause alleles to have antagonistic pleiotropic effects, and such alleles can give rise to negative geneti correlations among life-history traits (Rose 1982, 1985; but see Charlesworth 1990). For this reason, physiological trade-offs are often suggested to "constrain" the evolutionary trajectories of populations, that is, give rise to evolutionary trade-offs (Loeschke 1987; Charnov 1989; Stearns 1989; 1992, p. 76; Partridge and Sibly 1991). Second, physiological trade-offs may determine the plastic response of an or- * To whom correspondence should be sent at present address: Department of Molecular Genetics, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Bronx, New York Am. Nat Vol. 144, pp ? 1994 by The University of Chicago /94/ $ All rights reserved.

3 662 THE AMERICAN NATURALIST ganism to its environment (Reznick 1985; Bell and Koufopanou 1986; Stearns et al. 1991), that is, give rise to phenotypic trade-offs. Indeed, the plastic responses of life-history traits to experimental conditionsuch as exposure to gamma radiation, and the deprivation of mates, oviposition sites, and dietary components, have been used for decades to reveal their physiological relations (recent reviews in Bell and Koufopanou 1986; Roff 1992; Stearns 1992). Since Williams (1966), many have attempted to uncover the trade-offs that shape the evolution of life histories, but few have agreed as to how this might best be done (Bell and Koufopanou 1986). Most recently, opinion has been divided between those who claim that evolutionary important trade-offs are revealed by phenotypic manipulations and those who claim that some sort of genetic analysis is required. Advocates of the former position (Partridge and Harvey 1985, 1988; Parker and Maynard Smith 1990; Lessells 1991; Smith 1991; Partridge 1992) assume that environmental manipulations will generally have the same phenotypic effects as the genetic variation that ultimately constrains the evolution of life-history traits. They also often argue that experimental manipulations can reveal parts of the global evolutionary "options set" that local patterns of genetic variation and covariation cannot (M0ller et al. 1989a, 1989b; Smith 1991), especially since such patterns are often capricious, being easily influenced by population structure (Rose 1984a) and environment (Service and Rose 1985; Clark 1987; Scheiner et al. 1989; de Jong 1990). Advocates of genetic analysis, in turn, point out that there is no compelling reason to think that trade-offs revealed by phenotypic manipulations should correspond to allelic effects, much less patterns of genetic variation and covariation, and that it is such patterns that shape the short-term evolution of populations (Reznick 1985, 1992a, 1992b; Rose et al. 1987; Charlesworth 1990). Under this view, knowledge of evolutionary trade-offs can only come from directly estimating geneticorrelations in populations selective equilibrium. Because few populations have been studied by means of both-phenotypic manipulations and the family or selection experiments required to detect genetic correlations, the degree to which these methods concur in their identification of trade-offs remains a largely open question (but see M0ller et al. 1989a, 1989b). Ultimately, however, the critical issue is not whether phenotypic manipulations reflect geneti correlations but whether they can be used to predict evolutionary trajectories. Only if this is so do we have grounds for believing that phenotypic trade-offs are of consequence to the evolution of an organism's life history. EXPERIMENTAL OVERVIEW We have previously documented the existence of a nongenetic trade-off between fecundity and starvation resistance in Drosophila melanogaster by means of phenotypic manipulation (Chippindalet al. 1993). Reducirig the quantities of yeast available to adult females, against a constant background of abundant carbohydrates, results in a decrease in mean daily fecundity as well as a concomitant increase in mean longevity (Partridget al. 1987; Chippindalet al. 1993). This purely phenotypic response mimics the response of longevity and fecundity

4 EVOLUTION OF PHENOTYPIC TRADE-OFFS 663 to selection, for it is known that populationselected for late-life reproductive success usually have a genetically reduced early life fecundity and increased mean longevity (Rose and Charlesworth 1981; Rose 1984b). Such populations also evolve an increased starvation resistance (Service et al. 1985), which is consistent with the existence of a negative genetic correlation between early fecundity and starvation resistance in outbred populations of D. melanogaster (Service and Rose 1985). In this study, we examine the evolution of the phenotypic trade-off between starvation resistance and fecundity in two sets of populationselected for stress resistance and their two sets of controls. At the time of assay, these populations, called SO and SB, had been under selection for starvation resistance for over 2 yr, approximately 40 generations, whereas their controls, called CO and CB, respectively, had not (Rose et al. 1992). For the sake of comparison, we also reanalyze data previously presented (Chippindalet al. 1993) on the ancestors of the starvation-selected stocks, the B and 0 stocks that were respectively selected for early- and late-life reproductive success. We address two questions. First, do phenotypic trade-offs determine evolutionary trajectories? If the trade-off between fecundity and starvation resistance identified by manipulating yeast determines the pattern of evolution, populationselected for starvation resistance should be expected to evolve up a trajectory identical to this trade-off function as shown in figure 1A. Here, the ancestral and derived populations depicted differ only in their position on a trade-off function defined by manipulating yeast levels; that is, the evolutionary change in mean starvation resistance and fecundity, considered over all manipulation treatments, is supposed to be the same as the slope of the ancestor's phenotypic trade-off. Populations need not, however, be constrained to evolve along an ancestral phenotypic trade-off. For example, in figure 1B, the means of starvation resistance and fecundity do not change, but there is a considerable change in the slope of the trade-offunction. Actual populations may evolve in either of the ways depicted in figures IA and B or in some combination thereof (e.g., fig. 1C). We determine the trade-off functions for the CB, SB, CO, and SO stocks in order to discriminate among the alternatives presented above. Second, can phenotypic trade-offs evolve? The slope of a phenotypic trade-off reflects a physiological relationship between the traits in question. Are such relationships immutable? Figure lb shows how starvation selection might alter the slope of a phenotypic trade-off. Since the slope of the relationship represents a minimal estimate of the cost, in terms of daily fecundity, of investment starvation resistance, change in the slope represents the evolution of this cost. MATERIAL AND METHODS Populations Employed The populations employed in this study all derive ultimately from a common ancestral population called IV (for details see Rose 1984b). In 1980, 10 populations were founded from IV. Five of these populations, called Bs, were main-

5 (n 0.o Fecundity el (eggs Ifemale I24 hrs) 0 a L B > H (n unl 1*D : X Fecundity (eggs / female I 24 hrs) 0 > R ~~~H; Fecundity (eggs I female I 24 hrs) FIG. 1.-Predicting evolutionary trade-offs from phenotypic manipulations: hypothetical outcomes. Phenotypic trade-off functions are postulated for an ancestral (open circles) and starvation-selected (solid circles) population. The trade-off function is supposed to have been determined from the fecundity and starvation resistance of females conditioned at six different yeast levels: low (L) to high (H). The bivariate mean (asterisk) is the fecundity and starvation of resistance of a population averaged over six yeast levels. A, The ancestor's phenotypic trade-off predicts the evolutionary trajectory of the starvation-selected population: the evolutionary change in the bivariate mean of the selected population relative to its ancestor's is equal to the slope of the ancestor's phenotypic trade-off. B, The slope of the starvation-selected population evolves such that it becomes shallower than its ancestor's, but the bivariate mean remains constant. Over yeast levels, the selected population incurs a smaller fecundity cost per unit of starvation resistance gained than does the ancestor. C, Both the slope of the trade-off function and the bivariate mean have evolved in the selected population relative to its ancestor.

6 EVOLUTION OF PHENOTYPIC TRADE-OFFS 665 tained on a 2-wk generation; the other five, called Os, were selected for late-life reproductive success by maintaining them on a 10-wk generation cycle (Rose 1984b). In the course of 10 yr of laboratory evolution, the Bs and Os diverged substantially in fitness-related traits, such that the Os now have higher fecundity, longevity, and starvation resistance than the Bs (Service et al. 1985, 1988; Leroi et al., in press a). The B and 0 populations are, in turn, the ancestors of all the lines used in this study. In 1989, two populations were derived from each of the B and 0 lines, which made 20 new populations in total. One of each pair of populations was selected for starvation resistance, while the other was not (Rose et al. 1992). Thus, the 20 new populations are divided into four treatments of five replicate populations each, in which the treatments are distinguished by ancestry (B vs. 0) and current selection regime (S vs. C). The selection regime is as follows (see Rose et al. 1992): Each generation, flies from all populations are raised in 8-dr vials at densities of flies/vial. At day 14 after oviposition, when all the flies have eclosed, they are placed into Plexiglas cages. Whereas the control populations are given an abundance of food, the starvation-selected populations are only given moist agar, which provides water but no nutrition. The SOs and SBs experience a low-yeast environment in their culture vials immediately prior to selection. Selection is halted by providing the flies with food; eggs are collected for the following generation simultaneously from all selected and control populations. As a consequence of selection, the SB and SO populations now have a higher starvation resistance and, under some assay conditions, lower fecundity relative to their controls (Rose et al. 1992; Leroi et al., in press b). The starvation-selected stocks and their controls currently have a generation time on the order of 5 wk. Assay Procedures In order to eliminate nongenetic parental effects, all assayed populations were standardized under a common rearing environment for two generations prior to the experimental generation. Experimental flies were raised under controlled temperatures (25?C) and larval densities (60-80/vial) (Rose 1984b). Age was standardized by discarding early-eclosing flies and collecting adults for the assay over the course of the subsequent 48 h. Flies of each assayed population were then randomly selected from culture vials and placed into six different concentrations of live yeast against a carbohydrate-rich banana food medium. These flies were "conditioned" on the dietary treatments by placing four males and four females into each vial and maintaining them on a given yeast concentration for 4 d. Ten conditioning vials per treatment per population were used; 20 populations were assayed (10 x 6 x 20 = 1,200 vials); 4,708 flies were assayed in total. Yeast concentration was varied by serially diluting a mixture consisting of 5.0 g of live Fleishmann's bakers yeast, 2.0 ml of 1% acetic acid, and 40 ml of distilled water, then pipetting a constant volume of yeast/water mixture into each vial to give six yeast levels: 0.195, 0.391, 0.781, 1.562, 3.125, and 12.5 mg per vial. After the conditioning phase, one male and one female from each conditioning vial were allocated to a fecundity assay in

7 666 THE AMERICAN NATURALIST which the female was allowed to lay eggs for 24 h. Another male and female from each conditioning vial were starved in the presence of atmospheric water provided by means of a cotton ball soaked with water separated from the flies by a sponge (see Service et al. 1985), and their time of death was recorded. Experimental Design and Statistical Analysis All hypothesis testing was done using replicate population means as observed variates. All treatment means and standard errors are based on the means of five replicate populations. Because of the small number of populations, the distribution of population statistics (means and slopes) cannot be usefully tested for normality. The phenotypically plastic trade-off function between starvation resistance and fecundity, as well as the responses of individual traits to yeast level, was determined for each population by means of Model II reduced major-axis regression (Sokal and Rohlf 1981, pp ). The responses of each trait to varying yeast levels were described by linear regression in which the independent variable was loglo (yeast level). This transformation was used to make the plastic responses linear. Since each selected population bears a unique relation to its control by virtue of being derived from a common ancestral population, paired t-tests were used in all comparisons. In comparisons of the B and 0 populations, no one of which bears a unique historical relation to any other, paired tests were also used because each B population was paired with an 0 population in a blocked design. All statistical work was carried out using STATVIEW on Macintosh computers. RESULTS Evolution of the Trade-Off between Fecundity and Starvation Resistance The trade-off function of any one population is characterized by measuring the fecundity and starvation resistance of females that have been conditioned at one of each of the six yeast levels. The trade-off function, then, can be described by the slope of the fecundity-starvation resistance relationship shown by these females across yeast levels, as well as the bivariate mean of these traits, also across yeast levels. The mean "across yeast levels" or "environments" is the fecundity or starvation resistance of a population averaged across six yeast levels. Figure 2 shows such trade-off functions for four sets of populations: SB, CB, SO, and 0. Figure 2A shows that the trade-offs the five SB populations have evolved relative to those of their controls, the five CBs. Figure 2B shows that the tradeoffs of the five SO populations have evolved in a similar fashion relative to their controls, the five COs. The trade-off function has, in both sets of starvationresistant populations, become steeper after about 40 generations of selection (as described by the differences in slopes; table 1). Furthermore, trade-offs both sets of starvation-resistant populations have become displaced such that they no longer intersecthe curves of the control populations over the range of manipulations (yeast levels) studied. The trade-off functions of the unselected

8 X225 s~~~~~~~ A a)~ ~ S CM 150 S CB 0 (1) B o 150 s 0 O 75 O.n Fecundity (eggs / female / 24 hrs) FIG. 2.-Evolution of the trade-off in starvation-selected populations and their controls. For each stock, the trade-offunction represents the relationship between starvation resistance and fecundity shown by females conditioned at six different yeast levels: low (L) to high (H). For each stock, the bivariate mean (asterisk) is the fecundity and starvation of resistance averaged over six yeast levels. The trade-offunction and bivariate means shown here are estimated from the individual replicate populations (table 1). A, Populations derived from the Bs (SB; solid circles) relative to their controls (CB; open circles). B, Populations derived from the Os (SO; solid circles) relative to their controls (CO; open circles). While the relative positions of the bivariate means show some decline of fecundity in both the SBs and SOs as they gained starvation resistance, their evolution clearly cannot be predicted by the trade-off revealed by phenotypic manipulation in their respective controls. Furthermore, in both the SBs and SOs, the phenotypic trade-off has become steeper than that of their respective controls. That is, the starvation-resistant stocks incur a smaller cost in fecundity per unit of starvation resistance gained over yeast levels than their controls. Statistical analysis is provided in tables 1, 2, and 3. Error bars represent SEs about the mean based on five replicate populations.

9 668 THE AMERICAN NATURALIST TABLE 1 THE EVOLUTION OF A PHENOTYPIC TRADE-OFF IN THE ANCESTRAL B AND 0 STOCKS AND ITS RESPONSE TO SELECTION FOR STARVATION RESISTANCE SLOPE POPULATION B 0 CB SB CO SO Mean SE NOTE.-The phenotypic trade-off of each population represents the joint response of female fecundity and starvation resistance to yeast level. The functional relationship between these traits is described for each population by a Model II standard major-axis regression. Each slope was estimated from the measurement of approximately 240 individuals. B vs. 0: P >.1, NS; CB vs. SB: P <.001; CO vs. SO: P <.01. In all comparisons, P is the two-tailed probability of rejecting the null hypothesis of no difference among stocks due to the effects of chance alone and is based on a t distribution with n - 1 = 4 df. B and 0 data are from Chippindale et al control populations, CO and CB, are very similar, in both slopes and means, to the ancestral B and 0 populations that were assayed approximately 20 mo (or 40 generations in the case of the Bs and eight generations in the case of the Os) after the derivation of the starvation-selected stocks and controls. Since the B and 0 stocks were assayed in a different experiment than the CB and CO stocks, we did not apply formal tests to the comparison. Evolution of Phenotypic Plasticity in Fecundity and Starvation Resistance As discussed above, the trade-off function for any one population describes the effect of yeast level on the joint response of female fecundity and starvation resistance. This relationship can be decomposed into two plastic responses: that of fecundity to varying yeast level and of starvation resistance to varying yeast level. These are described by the slope of the regression of phenotype on yeast level and the mean phenotype across yeast levels. Table 2 shows the plastic response of starvation resistance in the starvation-selected stocks and their controls. As shown by the differences in mean, the starvation resistance of the SO and SB populations is generally greater than that of their controls, CO and CB (fig. 3A,C). However, as reflected in the slopes of each stock, this difference is even greater at low yeast levels than at high yeast levels. Thus, not only is starvation resistance a phenotypically plastic trait, but there appears also to be selectable genetic variation for the magnitude of the plastic response within these populations. Table 3 shows the plastic response of fecundity in the starvation-selected stocks and their controls. As shown by the differences in mean, the fecundity of the SOs is generally and significantly lower than that of their controls, the COs (fig. 3D; table 3). As reflected in the slopes of the SOs, this difference is even greater

10 EVOLUTION OF PHENOTYPIC TRADE-OFFS 669 TABLE 2 THE PLASTIC RESPONSE OF STARVATION RESISTANCE IN THE ANCESTRAL B AND 0 STOCKS AND ITS EVOLUTION IN RESPONSE TO SELECTION FOR STARVATION RESISTANCE CONTROL SELECTED COMPARISON Mean SE Mean SE P Slope: B VS <.01 CO VS. SO <.01 CB VS. SB <.01 Mean across environments (survival in hours): B VS <.001 CO VS. SO <.001 CB vs. SB <.01 NOTE.-The response to varying yeast levels of a stock's starvation resistance is described by its average slope (Model I linear regression) and mean across yeast levels. All means and SEs are based on five replicate populations. P gives the two-tailed probability of rejecting the null hypothesis of no difference among stocks due to the effects of chance alone and is based on a t distribution with n - 1 = 4 df. B and 0 data are from Chippindale et al. 1993; by convention Bs are control, and Os are selected. at high yeast levels than at low yeast levels. Thus, in the SOs, the phenotypic plastic response of fecundity appears to have evolved, relative to the controls, as a correlated response to selection for starvation resistance. Not all aspects of this correlated response of fecundity are, however, seen in the SBs relative to their controls. While the mean of fecundity across environments is significantly lower in the SBs than in the CBs, the slope has remained unchanged (fig. 3B; table 3). DISCUSSION The phenotypic trade-off between starvation resistance and fecundity does not predicthe evolutionary trajectories of populationselected for increased starvation resistance. The slopes of the trade-off in ancestral (B and 0) and control (CB and CO) stocks range from to and are all significantly greater than -1. In these stocks, each hour of starvation resistance gained by a female under manipulation entailed a cost of eggs/d. Is this purely phenotypic fecundity cost similar to the cost incurred in evolving a higher mean starvation resistance? It appears not. Considering mean starvation resistance across all yeast environments (table 2), we observed that the SBs had gained ( 9.91) h and SOs ( 12.40) h, relative to their respective controls (mean and standard errors). The evolved loss in mean fecundity across all yeast levels (table 3), however, was far less dramatic than that predicted by the phenotypic trade-off: SBs, 2.77 ( ) eggs/d, and SOs, 5.26 (? 1.19) eggs/d relative to their respective controls. In other words, over all environments, each hour of starvation resistance gained by selection entailed a cost of a mere eggs/d (SBs) or eggs/d (SOs) in reduced fecundity. Considering the alternative scenarios

11 670 THE AMERICAN NATURALIST 240 A 100- B X,160- * SB - 75 U)~~~~~~~~~~~~~c -160 a C 1'50 o-d c,80 o., c AASB C 100 D O,~~~~~~L 20 o-,0, 25_, 80 doj C:160 6 U) cn 0) 0~ ~ co ~~~~~~~~~~~~~~~~~~~) Q 40 /S cu80 ~ 0 C /- 0 20~~~/ Logi o (yeast level) FIG. 3.-Evolution of plastic responses of fecundity and starvation resistance in the starvation-selected stocks (solid circles) and their controls (open circles). The plastic response of each trait is described by the slope of its relationship when regressed on yeast level. A, SB and CB starvation resistance; B, SB and CB fecundity; C, SO and OC starvation resistance; D, SO and CO fecundity. Statistical analysis is provided in tables 2 and 3. for the joint evolution of the environmental responses of fecundity and starvation resistance outlined in figure 1, it appears that the actual evolution of starvationselected stocks, relative to their controls (fig. 2A,B), partakesomewhat of each of the firstwo cases. As in figure 1A, mean fecundity has declined somewhat as mean starvation resistance increased; as in figure ib, the slope of the trade-off has changed, specifically becoming less negative; together these changes result in a selection response similar to that depicted in figure 1C. Nonetheless, the decline of mean fecundity in populationselected for increased starvation resistance is consistent with the negative genetic correlation known to exist among these traits (Service and Rose 1985). It is also consistent with the notion that the physiological mechanism(s) indicated by manipulating yeast levels are of evolutionary import-presumably by virtue of some role in shaping the genetic variance/covariance structures of the selected populations.

12 EVOLUTION OF PHENOTYPIC TRADE-OFFS 671 TABLE 3 THE PLASTIC RESPONSE OF FECUNDITY IN THE ANCESTRAL B AND 0 STOCKS AND ITS EVOLUTION IN RESPONSE TO SELECTION FOR STARVATION RESISTANCE CONTROL SELECTED COMPARISON Mean SE Mean SE P Slope: B VS <.01 CO vs. SO <.05 CB VS. SB NS Mean across environments (eggs/female/24 h): B VS <.05 CO vs. SO <.001 CB VS. SB <.05 NOTE.-The response to varying yeast levels of a stock's fecundity is described by its average slope (Model I linear regression) and mean across yeast levels. All means and SEs are based on five replicate populations. P gives the two-tailed probability of rejecting the null hypothesis of no difference among stocks due to the effects of chance alone and is based on a t distribution with n - 1 = 4 df. B and 0 data are from Chippindale et al. 1993; by convention Bs are control, and Os are selected. Candidate physiological mechanisms that might have such a role are those responsible for allocating lipids between body fat and ovaries; such mechanisms are known to be important in selectively increased starvation resistance (Service 1987). They are also thoughto be involved in the diet-mediated trade-off (Chippindal et al. 1993), and preliminary results from histological studies (A. M. Leroi, unpublishedata) suppor this idea. It is clear, however, that even if the loci involved in the selection response partly influence the same physiological mechanisms that dietar yeast does, such loci must influence other mechanisms besides. These other, unknown physiological mechanisms apparently increase starvation resistance without incurring a decline in fecundity, although they may well trade off with other unmeasured life-history attributes. Reznick (1992a) has discussed other instances in which phenotypic manipulations influenced traits in a manner inconsistent with their evolved distribution among populations. As in this study, a common finding has been that the tradeoffs revealed by any given phenotypic manipulation can account for some, but not all, of the differentiation among populations (see, e.g., Luckinbill et al. 1988; Service 1989; Sinervo 1990). We believe that this is because differences among populations are strongly influenced by geneticorrelations that, in turn, describe the patterns of genetic variance and covariance associated with loci over the entire genome. It is inevitable, then, that geneticorrelations reflecthe net effect of many more physiological pathways than do most phenotypic manipulations. This will be especially true for quantitative genetic traitsuch as fecundity, body size, survival, and so forth, which may be influenced by tens or hundreds of loci (Falconer 1989) and on which nearly all physiological processes impinge. Phenotypic manipulations probably reveal the action of only a few of the many

13 672 THE AMERICAN NATURALIST trade-offs that may influence the evolution of populations. Even if manipulations were to reveal all the evolutionarily important physiological mechanisms, they could not be used to determine the influence of those mechanisms on the patterns of genetic variance and covariance that necessarily constrain at least the short-term evolution of populations (Charlesworth 1990; Houle 1991; Reznick 1992a, 1992b). Evolution of a Physiological Trade-Off Although phenotypic manipulations do not tell us much about evolution, they may reveal physiology (Reznick 1992a). The evolution of the slopes of the phenotypic trade-offs in the SO and SB stocks relative to their controls (table 1; fig. 2A,B) suggests that the physiological relationship between starvation resistance and fecundity in the selected stocks has been altered. Specifically, selection for starvation resistance has lessened at least one component of the fitness cost of starvation resistance. As discussed above, the B, 0, CB, and CO stocks gain an hour of starvation resistance at the expense of eggs/d; the SB and SO stocks, on the other hand, gain an hour of starvation resistance at the expense of only 0.68 and 0.63 eggs/d, respectively, under dietary manipulation. A number of physiological scenarios could explain the evolution of a steeper trade-off function in the starvation resistance stocks. In a previous article (Chippindale et al. 1993) we argued that the inverse phenotypic response of fecundity and starvation resistance to varying levels of dietar yeast-found in all stocks thus far examined-could be most simply explained by a Y model (Sheridan and Barker 1974; van Noordwijk and de Jong 1986) in which some constant and limiting resource, possibly lipids, is allocated largely to fecundity at high yeast levels and largely to metabolism (starvation resistance) at low yeast levels. Given such a pathway, steeper trade-off curves could arise if the starvation-selected stocks allocated especially great amounts of thei resources to maintenance when kept at low yeast levels. Alternatively, the starvation-selected stocks might use lipids with especial efficiency in the course of starving, after having been kept at low yeast levels (as in Hoffmann and Parsons 1991, p. 176; Boggs 1992). These explanations differ in their predictions for the environment-specific evolution of starvation resistance and fecundity. The environment-specific allocation model predicts that the environment in which SB and SO starvation resistance gain was greatest should be that in which the fecundity decline was greatest. The environment-specific efficiency model does not require a loss of fecundity in any environment. The evolution of environment-specific starvation resistance and fecundity can be seen by comparing the slopes of the plastic responses of these traits for the selected and control stocks (fig. 3; tables 2 and 3). Both starvationselected stocks are disproportionately starvation resistant at low yeast levels; the slopes of their plastic responses for this trait, initially negative, have become more so (fig. 3A,C; table 2). But contrary to the prediction of the allocation model, the fecundity slopes, initially positive, have not become appropriately more so. In the SB populations, the fecundity slope does not differ significantly from that of the controls, the CBs (fig. 3B; table 3). In the SO populations, the fecundity slope has become significantly less positive as the mean across environments has declined (fig. 3D; table 3). This is because most of the decline

14 EVOLUTION OF PHENOTYPIC TRADE-OFFS 673 in SO fecundity occurred at high rather than low yeast levels. The physiological basis of the evolution of the phenotypic trade-off remains, then, obscure. Implications for the Study of Life-History Evolution The trade-off between starvation resistance and fecundity is thoughto be one component of the trade-off between longevity and fecundity (Service et al. 1988; Chippindalet al. 1993). Maynard Smith (1989, p. 72) wrote that this trade-off is "well established experimentally, and... would probably be hard to alter by selection." Contrary to this assertion, our finding of an evolved reduction in the reproductive cost of starvation resistance suggests that this particular physiological trade-off is changed rather easily by selection. Analogously, Kaitala (1991) has shown that, while northern European populations of the waterstrider Gerris thoracicus exhibit a diet-mediated trade-off between fecundity and longevity, central European populations of the same species do not; this is but a more extremexample of the type of evolution seen here. Taken together, these studies cause us to suspect that physiological trade-offs are far more malleable over the course of evolutionary time than is currently perceived. While many authors have explicitly distinguished between physiological and evolutionary trade-offs (see, e.g., Roff 1992; Stearns 1992), othersee physiological trade-offs (those revealed by manipulation of an organism's phenotype) as reflecting the evolutionary option set available to an organism (see, e.g., Partridge and Sibly 1991; Smith 1991; Partridge 1992). The lack of quantitative agreement found here between the trade-offs revealed by dietary yeast manipulations and those revealed by selection suggests that the latter view is incorrect, at least for these populations. To the degree that this reflects the case of natural populations, we concur with Reznick (1992b) that if the object of a study is to reveal trade-offs of evolutionary importance, then the use of environmental manipulations as a substitute for genetic analysis seems imprudent. A possible objection to this generalization might be that phenotypic trade-offs should only reflect evolutionary ones when populations are located on an evolutionary "constraint surface" (see Levins 1968; Charnov 1989) and that, in contras to natural populations, which are indeed so constrained, the laboratory populations studied here were located at some distance from such a surface. While we cannot know anything about the case for natural populations (Partridge and Sibly 1991), it can be argued that the populationstudied here are likely to be close to such a surface, for they are descended from outbred populations (IV, B, and 0) that have adapted to a relatively constant laboratory environment over the course of hundreds of generations (Rose 1984b; Rose et al. 1992; Leroi 1993). Of course, the above argument presupposes that the notion of evolutionary constraint surfaces is, indeed, a coherent one. Yet there would seem to be no general theory that connects phenotypic responses to evolutionary trajectories. A particular model might be erected in which it is supposed, for example, that alterations of input nutrients have some effect on the shift of allocation between components of a trade-off. One mighthen add to this assumption the additional assumption that allelic effects on this same allocation are to have the same functional form, and this model mighthen be further tuned to produce an exact

15 674 THE AMERICAN NATURALIST correspondence between phenotypic and evolutionary dynamics. At the end of this process, the hypothesis assumed at the outset would be mathematically illustrated, but it would not be supported in terms of general, well-attested theory. Furthermore, the central assumption of constraint-based notions of life-history evolution, that the physiological relationships that cause life-history trade-offs evolve more slowly than the traits themselves, may simply be wrong. In showing that phenotypic trade-offs can evolve, we have shown that the physiological relationship between starvation resistance and fecundity can change, and it can do so even as it influences (if not determines) the evolution of mean starvation resistance and fecundity. If phenotypic manipulations are held to represent evolutionary surfaces at all, they cannot be held to represent static ones. These concerns notwithstanding, we do believe that phenotypic manipulations can be useful in the study of life-history evolution, for they reveal the plastic responses of organisms to environmental variation (see, e.g., Partridge et al. 1987; Service 1987; M0ller et al. 1989b; Chippindalet al. 1993), and this remains an important and largely unexplored area of life-history study (Lessells 1991; Stearns 1992). ACKNOWLEDGMENTS We thank A. R. Bennett, A. K. Chippindale, A. Gibbs, L. D. Mueller, M. Rausher, D. Reznick, A. E. Weis, and two anonymous referees for their comments on the manuscript, as well as R. Chen, H. Saing, and J. Shiotsugu for assisting in the laboratory. This research was supported by United States Public Health Service grants AG06346 and AG09970 to M.R.R. from the National Institute on Aging. LITERATURE CITED Bell, G., and V. Koufopanou The cost of reproduction. Pages In R. Dawkins and M. Ridley, eds. Oxford surveys in evolutionary biology. Vol. 3. Oxford University Press, Oxford. Boggs, C. L Resource allocation: exploring connections between foraging and life history. Functional Ecology 6: Charlesworth, B Optimization models, quantitative genetics, and mutation. Evolution 44: Charnov, E. L Phenotypic evolution under Fisher's fundamental theorem of natural selection. Heredity 62: Chippindale, A. K., A. M. Leroi, S. B. Kim, and M. R. Rose Phenotypic plasticity and selection in Drosophila life-history evolution. Nutrition and the cost of reproduction. Journal of Evolutionary Biology 6: Clark, A. G Senescence and the geneticorrelation hang-up. American Naturalist 129: de Jong, G Quantitative genetics of reactionorms. Journal of Evolutionary Biology 3: Falconer, D. S Introduction to quantitative genetics. 3d ed. Longman, London. Hoffmann, A. A., and P. A. Parsons Evolutionary genetics and environmental stress. Oxford University Press, New York. Houle, D Geneti covariance of fitness correlates: what geneticorrelations are made of and why it matters. Evolution 45: Kaitala, A Phenotypic plasticity in reproductive behaviour of waterstriders: trade-offs between reproduction and longevity during food stress. Functional Ecology 5:12-18.

16 EVOLUTION OF PHENOTYPIC TRADE-OFFS 675 Leroi, A. M The origin and evolution of life-history trade-offs. Ph.D. diss. University of California, Irvine. Leroi, A. M., A. K. Chippendale, and M. R. Rose. In press a. Long-term laboratory evolution of a genetic life-history trade-off in Drosophila melanogaster. I. The role of genotype-byenvironment interaction. Evolution. Leroi, A. M., W. R. Chen, and M. R. Rose. In press b. Long-term laboratory evolution of a genetic life-history trade-off in Drosophila melanogaster. II. Stability of geneticorrelations. Evolution. Lessells, C. M The evolution of life histories. Pages in J. R. Krebs and N. B. Davies, eds. Behavioural ecology. 3d ed. Blackwell Scientific, Oxford. Levins, R Evolution in changing environments. Princeton University Press, Princeton, N.J. Loeschke, V., ed Geneti constraints adaptiv evolution. Springer, Berlin. Luckinbill, L. S., J. L. Graves, A. Tomkiw, and 0. Sowirka A qualitative analysis of life history characters in Drosophila melanogaster. Evolutionary Ecology 3: Maynard Smith, J Phenotypic models of evolution. Pages in W. G. Hill and T. F. C. Mackay, eds. Evolution and animal breeding. C. A. B. International, Wallingford. M0ller, H., R. H. Smith, and R. M. Sibly. 1989a. Evolutionary demography of a bruchid beetle. I. Quantitative genetical analysis of the female life-history. Functional Ecology 3: b. Evolutionary demography of a bruchid beetle. II. Physiological manipulations. Functional Ecology 3: Parker, G. A., and J. Maynard Smith Optimality theory in evolutionary biology. Nature (London) 348: Partridge, L Measuring reproductive costs. Trends in Ecology & Evolution 7: Partridge, L., and P. H. Harvey The costs of reproduction. Nature (London) 316: The ecological context of life history evolution. Science (Washington, D.C.) 241: Partridge, L., and R. Sibly Constraints the evolution of life histories. Philosophical Transactions of the Royal Society of London B, Biological Sciences 332:3-13. Partridge, L., A. Green, and K. Fowler Effects of egg-production and of exposure to males on female survival in Drosophila melanogaster. Journal of Insect Physiology 33: Reznick, D Costs of reproduction: an evaluation of the empirical evidence. Oikos 44: a. Measuring the costs of reproduction. Trends in Ecology & Evolution 7: b. Measuring reproductive costs: response to Partridge. Trends in Ecology & Evolution 7:134. Roff, D The evolution of life histories. Chapman & Hall, London. Rose, M. R Antagonistic pleiotropy, dominance, and genetic variation. Heredity 48: a. Geneticovariation in Drosophila life history: untangling the data. American Naturalist 123 : b. Laboratory evolution of postponed senescence in Drosophila melanogaster. Evolution 38: Life-history evolution with antagonistic pleiotropy and overlappingenerations. Theoretical Population Biology 28: Rose, M. R., and B. Charlesworth Genetics of life-history in Drosophila melanogaster. II. Exploratory selection experiments. Genetics 97: Rose, M. R., P. M. Service, and E. W. Hutchinson Three approaches to trade-offs in life-history evolution. Pages in V. Loeschke, ed. Genetic constraints adaptive evolution. Springer, Berlin. Rose, M. R., L. N. Vu, S. U. Park, and J. L. Graves Selection for stress resistance increases longevity in Drosophila melanogaster. Experimental Gerontology 27: Scheiner, S. M., R. L. Caplan, and R. F. Lyman A search for trade-offs among life-history traits in Drosophila melanogaster. Evolutionary Ecology 3: Service, P. M Physiological mechanisms of increased stress resistance in Drosophila melanogaster selected for postponed senescence. Physiological Zoology 60: The effect of mating status on life-span, egg laying, and starvation resistance in Droso-

17 676 THE AMERICAN NATURALIST phila melanogaster in relation to selection on longevity. Journal of Insect Physiology 35: Service, P. M., and M. R. Rose Genetico-variation among life history components: the effects of novel environments. Evolution 39: Service, P. M., E. W. Hutchinson, M. D. MacKinley, and M. R. Rose Resistance to environmental stress in Drosophila melanogaster selected for postponed senescence. Physiological Zoology 58: Service, P. M., E. W. Hutchinson, and M. R. Rose Multiple genetic mechanisms for the evolution of senescence in Drosophila melanogaster. Evolution 42: Sheridan, A. K., and J. S. F. Barker Two-trait selection and geneticorrelation. II. Changes in the geneticorrelation during two-trait selection. Australian Journal of Biological Sciences 27: Sinervo, B The evolution of maternal investment lizards: an experimental and comparative analysis of egg size and its effects on offspring performance. Evolution 44: Smith, R. H Genetic and phenotypic aspects of life-history evolution in animals. Advances in Ecological Research 21: Sokal, R. R., and F. J. Rohlf Biometry. 2d ed. W. H. Freeman, New York. Stearns, S. C Trade-offs in life history evolution. Functional Ecology 3: The evolution of life histories. Oxford University Press, Oxford. Stearns, S. C., G. de Jong, and G. Newman The effects of phenotypic plasticity on genetic correlations. Trends in Ecology & Evolution 6: van Noordwijk, A. J., and G. de Jong Acquisition and allocation of resources: their influence on variation in life-history tactics. American Naturalist 128: Williams, G. C Natural selection, the costs of reproduction and a refinement of Lack's principle. American Naturalist 100: Editor: Mark D. Rausher