TREE vol. 7, no. 2, February

15 Hanski, I. (1983) Ecology 64, 493-500 16 Taylor, A.D. (1990) Ecology 71, 429-433 17 Roughgarden, J. and Iwasa, Y. (19861 Theor. Popul. Biol. 29, 235-261 I8 Harrison, S. and Quinn, j.F. 11989) Oikos 56, 293-298 I9 Slatkin, M. (1974) Ecology 55, 128-134 20 Pulliam. R. (1988) Am. Nat. 132, 652-661 21 Vance, R.R. (1984) Am. Nat. 123, 230-254 22 Fahrig, L. and Merriam, G. (1985) Ecology 66, 1762-l 768 23 Weiner, J. and Conte, P.T. (I981 I Ecol.

Model. 13, 131-147 24 van Tongeren, 0. and Prentice, I.C. (1986) Vegetatio 65, 163-173 25 Crawley, M.J. and May, R.M. (1987) 1. Theor. Biol. 125, 475-489 26 CdrZm, T. ( 19891 Coenoses 4, I 13-l 20 27 Green, D.G. II9891 Vegetatio 82, 139-153 28 Weiner, 1. (19821 Ecology 63, 1237-1241 29 Hara. T. ( 19881 Trends Ecol. Evol. 3, 129-133 30 Slatkin, M. and Anderson, D.I. (19841

7992

Ecology 65, 1840-1845 31 Walker, I., Sharpe, P.I.H.. Penridge, L.K. and Wu, H. (1989) Vegetatio 83, 81-95 32 Pacala, S.W. and Silander, ].A., Jr (1985) Am. Nat. 125. 38541 I 33 Pacala, S.W. ( 1986) Theor. Popul. Biol. 29, 262-292 34 CzArAn, T. and Bartha, S. ( 1989) Vegetatio 83, 229-239 35 Kenkel, N.C. (1990) Coenoses 5, 149-158 36 Shugart, H.H. (1984) A Theory of Forest Dynamics, Springer-Verlag

Measuringthe Costsof Reproduction David Reznick The measurement of costs of reproduction is of interest because such costs are generally assumed 6g life history theory. There is some controversy concerning how to measure costs: common methods include experimental manipulations of life history, such as preventing some individuals from reproducing, or estimates of genetic correlations. These two methods often yield similar results, suggesting that one can serve as a substitute for the other. There are now experiments which demonstrate that there are different mechanisms underlying the response to an experimental manipulation versus a genetic correlation, so the two methods ure nof equivalent in estimating costs. Costs of reproduction are tradeoffs among different components of an organism’s life history. Costs are believed to be fundamental constraints on the evolution of life history patterns or the evolution of traits associated with propagation, such as age at maturity, offspring number and size, and frequency of reproduction. For this reason, measuring and characterizing these costs has been a subdiscipline in the study of life history evolution for almost two decades. In spite of the intense interest, we still have not resolved many details about the nature and prevalence of costs, or even how to measure them experimentally. Recently, new contributions in this field have given rise to some debate. In reviewing this debate, 1 will summarize a specific controversy, then discuss its general implications. Note that an earlier

David Reznickis at the Dept of Biology, University of California, Riverside, CA 92521, USA.

review in TREE’ dealt largely with practical problems in estimating the costs of reproduction using clutchsize manipulations in birds. First, consider in more detail why we measure costs. Theories of life history evolution view the life history as the competitive allocation of resources to growth, maintenance and reproduction. If resources are limiting, then an increase in the allocation to one function necessarily results in a decrease to other functions. It is this competition for limiting resources that underlies one concept of costs, often referred to as physiological costs. It is important to note that theories predict the evolutionary response to selection and hence assume that the costs have a genetic basis. One way to test a theory is to evaluate its assumptions. This, then, is one motive for measuring costs. A second type of cost (ecological costs) does not necessarily involve the competitive allocation of resources but instead involves interactions with the external environment. For example, activities associated with reproduction potentially expose the individual to risks such as disease, injury or predation. Costs of reproduction are divided into two main categories, although there are a variety of more specific ways of evaluating costs (see Ref. 2 for a detailed discussion of costs). The first major class is survival costs, where current reproductive effort influences an individual’s probability of future survival. The second is fecundity costs, where current reproductive effort influences an individual’s capacity to reproduce in the future.

The controversy Rose and Charlesworth’s ‘natural selection’ experiment on agespecific reproduction was crucial for its characterization of a cost of reproduction. As a sequel to describing the variance-covariance matrix for life history characters in Drosophila melanogaster4, these authors selected for production of successful offspring either early or late in life. ‘B’ lines, flies that were only allowed to reproduce successfully early in life, were compared with ‘0’ lines, flies which were only allowed to reproduce successfully late in life. After I2 generations of selection, O-line flies laid fewer eggs early in life, but had significantly longer lifespans and laid more eggs late in life than B-line flies. Rose5 and Luckinbill et aL6 demonstrated the repeatability of this result. This remains one of the best examples of the use of selection experiments to estimate the costs of reproduction, with the cost being represented by the inverse relationship between high fecundity early in life and longevity or reproduction late in life. It is also a key result supporting the idea that senescence is caused in part by antagonistic pleiotropy, or a negative genetic correlation between reproductive performance early in life and longevity. Charlesworth notes that an accumulation of mutations that reduce fecundity early in life could also explain the results for the 0 lines; however, the rate of change in fecundity appears too high. In addition, the associated observations of negative genetic correlations between early and late fecundity4 and changes in longevity? argue for antagonistic pleiotropy as the cause of these results. 0

1992.

Elsev~er

Science

Publishers

Ltd (UK)

TREE vol. 7, no. 2, February

1992

Partridge8,9 questioned the conclusion that this result revealed antagonistic pleiotropy. She argued instead that the result was a consequence of selection for a change in mating behavior. Partridge and her colleagues have shown that it is possible to increase lifespan by artificially decreasing encounters with potential mates+‘3. They suggest that the increased lifespan observed by Rose and Charlesworth was therefore the result of indirect selection for a change in mating behavior and perhaps a direct, uncorrelated response in age-specific fecundity. Luckinbill et aLI4 and ServiceI have responded to Partridge’s challenge with experiments. Both studies compared the longevities of the lines selected for early or late reproductive success when flies were either mated or unmated. If Rose and Charlesworth’s result for longevity was caused by changes in mating behavior, then the differences should be present when mated lines are compared but absent when virgin, unmated lines are compared (Fig. la). ServiceI worked with Rose’s B and 0 lines while Luckinbill and hiscolleagues14 worked with an independently selected series of lines. In three of four comparisons (males and females in ServiceI and males in Luckinbill et al.14) they found that ‘O’-type flies outlived ‘B’-type flies and that unmated flies outlived mated flies, but found no interaction between these independent variables (example in Fig. lb). This result indicates that the genetic differences among selected lines are independent of the response of each line to being mated or remaining celibate. The one exception to this additivity was seen in the evaluation by Luckinbill et aLI of the influence of mating on the females from the long-lived lines (Fig. Ic). They found that the longevity of these females was not significantly reduced by mating. In this case, it appears that part of being long-lived is a reduced sensitivity to the costs of encounters with members of the opposite sex. The important point is that, in all cases, the longevity differences are present whether or not the flies are mated. The genetic mechanism under-

:k

lying the response in Rose and Charlesworth’s selection experiment is therefore different from the phenotypic plasticity in lifespan revealed by Partridge and her colleagues’ manipulations of mating frequency. It thus appears that Rose and Charlesworth’s original result was indeed an example of antagonistic pleiotropy, rather than selection for a change in mating behavior. Methods for estimating costs of reproduction This exchange addresses a more general controversy over what methods are most appropriate for estimating costs of reproduction. The different approaches for measuring these costs include (after Ref. 16): (1) Phenotypic correlations, or the estimation of statistical correlations between life history traits measured on a series of individuals, populations or related species. (2) Manipulations of the life history (e.g. Refs 9-l 3). (3) Genetic correlations among different components of the life history, generally with formal quantitative genetics experiments (e.g. Ref. 3). (4) Responses to selection. This method can evaluate the predictions of the previous method. The most common approach here is to select for age-specific reproduction (e.g. Ref. 4). The four methods differ in their assessment of genetic and environmental sources of cost (Fig. 2). The first method, phenotypic correlations, considers the phenotypes alone, and hence incorporates both environmental and genetic effects. The second method, manipulation of reproductive investment, represents a fixed environmental effect and is independent of the genetic background. Methods 3 and 4 both measure the genetic contribution to costs, first via static genetic correlations, then by the correlated response to selection. iI6 argued that, if one’s goal in estimating costs of reproduction is to evaluate assumptions of evolutionary theories, then only methods 3 and 4 yield the appropriate information. It does not foollow that the other methods are not worth pursuing (see below). Partridge and Harvey17,18 and Bell

0

6

I

g

g5-(c)

E! 85El 75s: 265-

-~

Flg. 1. la) One prediction implicit in PartridgeV9 challenge of Rose and Charlesworth’s’ result: there should be no difference in the longevities of B and 0 lines when they are not mated, but greater longevity in 0 lines when they are mated. (b) A sample of the results from ServiceI (his Table II. These data are for females that were transferred to new vials daily. fc) Results from Luckinbill et aLI (their Table I I for female longevity.

and Koufopanou’9 argued that manipulations were a valid and useful way of evaluating these costs, in part to fulfill the same goal. Partridge and HarveyI pointed out that phenotypic manipulations are much easier to do and can be done on a much larger number of organisms than genetic correlations. Bell and Koufopanou19 argued that it is possible to determine something about the genetic correlation, such as the sign, from phenotypic correlations. Finally, there isalready extensive evidence that the results of studies using method 2 are in accord with our expectations of costs, as in the research by Partridge and her colleaguesg-i3. If the mechanisms underlying these phenotypic responses are the same as those revealed by methods 3 and 43

TREE vol. 7, no. 2, February

I

I

Method

1992

2

Environment

Method

1

/-&q

1

1

klethods

3,4

Fig. 2. A symbolic representation of the combined role of the environment and genotype in determining the phenotype of an organism, with an indication of the level of analysis addressed by different methods for evaluating the costs of reproduction. Method I, which deals with phenotypic correlations, measures the combined influences of the environment and the genotype; the two cannot be distinguished by this method. Method 2, which involves a manipulation of the organism, such aspreventingitfrom reproducing, isa fixed manipulation of the environment alone. Methods 3 and 4 (quantitative genetics and selection experiments) deal exclusively with the influence of the genotype on the phenotype land on the estimated costs of reproduction, in this context).

4, then as much is revealed with a much smaller investment of effort. These same arguments underlie Partridge’s8,9 original challenge of Rose and Charlesworth’s results. The conviction that phenotypic manipulations reveal the same mechanisms as genetic variation suggested that her results for the manipulation of mating frequency can explain Rose and Charlesworth’s results from a selection experiment. The missing link here is experiments that evaluate the difference between genetically and phenotypically based costs. The studies by Luckinbill et a/.14 and ServiceI represent the first such experiments and they show unambiguously that, even though the two approaches have the same impact on the phenotypic expression of costs, the underlying mechanisms are different. One such result does not prove that phenotypic manipulations are inappropriate for measuring the sorts of costs assumed by life history theory, but it certainly does not encourage the practice. Additional evidence An additional study highlights the difference between the information yielded by phenotypic manipulations versus genetically based trade-offs. Izo addressed the question ‘is there a trade-off between growth and reproduction?’ with two experiments on guppies (Poecilia Such a trade-off reprereticulata). sents an indirect assessment of fecundity costs because fecundity is positively correlated with size, and slowed growth translates into a

reduction in the future capacity to reproduce. In the first experiment, I manipulated whether or not an individual reproduced and I looked for compensatory changes in growth. In the second experiment, 1 analysed the energetics of growth and reproduction in guppies derived from four localities for which I had previously found differences I determined in life histories. whether there were complementary differences in the energy devoted to growth and reproduction. In the first experiment, I found no change in body size or the quantity of calories in the soma. Most of the calories saved by not reproducing disappeared. There is, thus, little evidence for a cost of reproduction here. In the second experiment, a relative increase in the energy allocated to reproduction in one population was matched by a decrease in the energy allocated to growth. The implication is that evolving a change in either growth or reproduction results in a complementary change in

the other variable. This difference in results suggests that the mechanism underly ing the response in the first experiment (method 2) is not the same as the one in the second experiment (method 4). Guppies may lack the mechanism to recoup the energy that would otherwise have been devoted to reproduction when reproduction is prevented, but can still evolve a change in the allocation of resources to reproductive versus somatic tissues. The mechanism that underlies the response to a phenotypic manipulation may be different from the organism’s capacity for evolutionary change. It is this difference in the mechanisms underlying the response to phenotypic manipulations versus evolutionary responses that is crucial. The difference is conspicuous in the above example, where the responses in the two experiments differed so dramatically, but such differences may also exist when the manipulation mimics the observed response to selection, as revealed by Luckinbill eta1.14 and Service15. The virtues of phenotypicmanipulations Whether or not phenotypic manipulations reveal evolutionary

potential, they can reveal a great deal about how an organism works and, hence, remain an important experimental tool. Some examples illustrate this point. One such method is to pollinate flowers artificially to increase seed production; if production can be increased, pollinator activity may normally limit fruit set. If artificial pollination increases fruit set, then it also increases the amount of resources allocated to reproduction. Recent investigators have followed the plants over subsequent years to see if this increase in fruit production influences future growth and reproduction. Montalvo and Ackerman*’ found that subsequent costs exceeded the immediate gains from artificial pollination, suggesting that resource limitation may be of overriding importance in structuring reproductive investment, even if pollinators are limiting within a year. Paige and Whitham** used similar manipulations to study plasticity in the life history of the scarlet gilia (Ipomopsis aggregata). This species is normally semelparous (monocarpit), but the authors found that high-elevation plants may also be iteroparous (polycarpic). lteroparity appeared to be a facultative switch which became more likely as the season progressed and as the abundance of pollinators declined. Paige and Whitham significantly increased the probability of iteroparity by excluding pollinators or removing flower buds early in the season. The alternative life history therefore represents a form of plasticity, which gives the individual a second chance at reproduction if the first attempt fails. Both examples illustrate that much can be learned from manipulations of reproductive investment, even if the technique is inappropriate for evaluating the assumptions of life history theory. This message is important if one considers the many recent papers that measure the costs of reproduction using phenotypic manipulations with the goal of characterizing these assumptions, while often ignoring the general worth of the results. Manipulations may also help one to evaluate ecological costs of reproduction (which involve interactions with the external environment1 such

TREE vol. 7, no. 2, February

1992

as increased susceptibility to predation, parasitism or disease caused by increased reproductive effort, decreased maintenance, decreased locomotor abilities (e.g. in pregnancy) or increased activity (e.g. foraging to feed young). Manipulations allow one to generate more variation in reproductive effort than is likely to be available naturally among different genetic lines. This variation can, in turn, be used to evaluate possible interactions with the external environment, as already practiced by some investigators (e.g. Ref. 23). the As a final example, recent work by Sinervo and his colleagues24-26 illustrates both the benefits and the limitations of phenotypic manipulations for studying life history evolution. These investigators can reduce offspring size by removing yolk after eggs are laid24,25, or can simultaneously increase egg size and reduce clutch size by destroying developing follicles26. The benefits of their methods are well described by TREE Bernardo27 in a recent article. Bernard0 argues that these methods are ideal for evaluating maternal (environmental) effect on egg size, or the influence of hatchling size on offspring fitness independently of the genetic background. The limits would arise if one interpreted these methods as an evaluation of the evolutionary trade-off between clutch size and egg size. For example, egg/hatchling size alone accounts for all of the differences among populations in sprint speed, but only some of the differences in stamina or incubation time, and none of the differences in leg dimensions. The manipulation, therefore, serves as a partial surrogate for interpopulation (evolved) differences in egg size, but cannot account for all the correlated differences among populations. These unexplained differences may be attributable either to independent genetic changes, or to changes that would normally be correlated with the evolution of egg size but are not caused by a process like yolk removal; it is not possible to tell the difference.

The limitations of genetic methods The chief message of these arguments is that, if you want to study the

trade-offs assumed in theories of life history evolution, environmental manipulations of the life history are likely to be inappropriate; they may mimic in part the effects of a response to selection, but the underlying mechanisms are likely to be different. A manipulation isolates the effects of one component of the life history, which is experimentally attractive, but is not likely to incorporate the correlated changes that may accompany an evolved reponse, such as pleiotropic effects. The implicit message is that quantitative genetics and selection experiments remain the best approach; however, these methods are limited in their applicability, as discussed by Partridge and Harvey I8. For example, only some organisms are amenable to quantitative genetic analyses of the life history. In addition to their arguments, a new and more sobering limitation has been presented by CharlesworthZs and Hou1e29. Both authors demonstrate that, in the evolution of a multivariate set of characters subject to constraints or trade-offs, it is possible for the genetic correlation between any pair of these characters to be weakly negative, or even positive (see also Ref. 30). Finding positive genetic correlations therefore does not disprove the existence of costs, although Charlesworth argues that significant negative correlations provide support for them. In spite of the difficulties in inferring constraints from correlations, Charleswotth28points out that variance-covariance the genetic matrix still predicts the short-term response to selection, and hence has some utility. He also argues that these difficulties do not justify the use of manipulationsas a substitute. Conclusions using The argument against environmental manipulations to evaluate the costs of reproduction assumed by life history theory is stronger than before; the argument in favor of quantitative genetics is weaker. As emphasized by Stephen Stearns (Zoological Institute, Base], Switzerland) at a Royal Society (London) meeting on life history evolution in October 1990, the evaluation of costs of reproduction

remains a crucial, unresolved component of the study of life history evolution. Acknowledgements I thank Joe Graves, Steve Morey, Linda Partridge, Mike Rose, Helen Rodd, Phil Service and four anonymous reviewers for their helpful discussions and comments on an earlier draft of the manuscript. 1 was supported by NSF Grant BSR-8818071 while preparing this manuscript.

References 1 Linden, M. and Moller, A.P. (1989) Trends Ecol. Evol. 4, 367-371 2 Stearns, SC. (1989) Funct. Eco/. 3, 259-268 3 Rose, M.R. and Charlesworth, B. ( 1981 I Genetics 97, 187-196 4 Rose, M.R. and Charlesworth. B. 11981) Genetics 97, I 73-l 86 38, 1004-1010 5 Rose, M.R. 11984) Evolution 6 Luckinbill, L.S., Arking, R., Clare, M.1.. Cirocco, W.C. and Buck, S.A. (1984) Evolution 38, 996-1003 7 Charlesworth, B. (1982) in Evolutionary Ecology (Shorrocks, B., ed.), pp. 117-133,

Blackwell Scientific Publications 8 Partridge, L. and Andrews, R. (1985) I. Insect Physiol. 3 I, 393-395 9 Partridge, L. ( 1989) in Towards a More Exact Ecology (Grubb, P.J. and Whittaker, LB.. edsl, pp. 231-246, Oxford University Press IO Partridge, L. and Farquhar, M. i 1981I Nature 294, 580-582 II Partridge, L., Fowler, K., Trevitt, S. and Sharp, W. ( 1986) 1. insect Physiol. 32, 925-929 12 Partridge, L.. Green, A. and Fowler, K. 11987) 1. Insect Physiol. 33, 745-749 I3 Fowler, K. and Partridge, L. (1989) Nature 338, 760-761

I4 Luckinbill, L.S., Graves, I.L., Tomkin, A. and Sowirka, 0. ( 1988) Evol. Ecol. 2, 85-94 I5 Service, P.M. (1989) /. insect Physiol. 35, 447-452 16 Reznick, D.N. ( 1985) Oikos 44, 257-267

I7 Partridge, L. and Harvey, P. II9851 Nature 316, 20-21 18 Partridge, L. and Harvey, P. (1988) Science 2 14, 1449-1455 I9 Bell, G. and Koufopanou, V. (1986) Oxf Sun/. Evol. Biol. 3, 83-l 3 I 20 Reznick, D.N. II9831 Ecology 64, 862-873 21 Montalvo, A.M. and Ackerman, I.D. ( 1987) Biotropica 19. 24-3 I 22 Paige, K.N. and Whitham, T.G. (1987) Ecology 68, I69 I-I 695 23 Shine, R. (1980) Oecologia 46, 92-100 24 Sinervo, B. (1990) Evolution 44, 279-294 25 Sinervo, B. and Huey,R.B. II9901 Science 248, I lO&-l 109 26 Sinervo, B. and Licht, P. (1991 I Science 252, 1300-1303 27 Bernardo, J. ( I991 ) Trends Ecol. Evol. 6, l-2 28 Charlesworth, 520-538

B. (1990) Evolution

44,

29 Houle, D. I I99 I) Evolution 45, 630-648 30 Pease, CM. and Bull, I.J. II9881 /. Evol. Biol. I, 293-303

45