The Evolution of Virulence When Parasites Cause Host Castration and Gigantism

The American Society of Naturalists

7KH(YROXWLRQRI9LUXOHQFH:KHQ3DUDVLWHV&DXVH+RVW&DVWUDWLRQDQG*LJDQWLVP $XWKRU V 'LHWHU(EHUW+DQV-RDFKLP&DULXV7RP/LWWOH(OOHQ'HFDHVWHFNHU 6RXUFH7KH$PHULFDQ1DWXUDOLVW9RO1R(FRORJ\DQG(YROXWLRQRI+RVW3DWKRJHQ ,QWHUDFWLRQVLQ1DWXUDO3RSXODWLRQV 1RY SS66 3XEOLVKHGE\7KH8QLYHUVLW\RI&KLFDJR3UHVVIRU7KH$PHULFDQ6RFLHW\RI1DWXUDOLVWV 6WDEOH85/http://www.jstor.org/stable/3473219 $FFHVVHG

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. 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 http://www.jstor.org/action/showPublisher?publisherCode=ucpress.

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 [email protected].

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.

http://www.jstor.org VOL. 164, SUPPLEMENT THE AMERICAN NATURALIST NOVEMBER 2004

The Evolution of Virulence When Parasites Cause Host Castration and Gigantism

Dieter Ebert,',2,* Hans Joachim Carius,1Tom Little,1'3and Ellen Decaesteckerl'4

1. Universitit Basel, Zoologisches Institut, Vesalgasse 1, 4051 Basel, Keywords:evolution of virulence, parasitic castration, infectious dis- Switzerland; ease, Daphnia, bacteria, Pasteuria. 2. Universit&de Fribourg, D~partement de Biologie, Ecologie et Evolution, Chemin du Mus&e 10, 1700 Fribourg, Switzerland; 3. Institute for Cell, Animal and Population Biology, University of Models on the evolution of the virulence of infectious West Mains EH9 Edinburgh, Kings Buildings, Road, Edinburgh diseases are largely built on two assumptions: first, that 3JT, Scotland; virulence is an unavoidableby-product of parasiterepro- 4. Laboratory of Aquatic Ecology, Catholic University of Leuven, duction (Bull 1994) and second, that it is in the pathogen's Chemin de B~riotstraat 32, 3000 Leuven, Belgium interest to avoid unnecessaryhost mortality because host death may curtail the parasite'slifetime transmissionsuc- cess. Consequently, pathogens are expected to evolve a balance between their need to reproduceand the costs of ABSTRACT: It has been suggested that the harm parasites cause to harming the host. Parasiticcastration, here defined as se- their hosts is an unavoidable of consequence parasite reproduction vere parasite-inducedreduction in host fecundity,has been with costs not only for the host but also for the parasite. Castrating suggested as an alternativestrategy for the evolution of parasites are thought to minimize their costs by reducing host fe- virulence. Certain transmitted cundity, which may minimize the chances of killing both host and horizontally parasites (in- parasite prematurely. We conducted a series of experiments to un- cluding pathogens) specificallyreduce or eliminate host derstand the evolution of virulence of a castrating bacterium in the reproductivefunction. Reproduction draws energy away planktonic crustacean Daphnia magna. By manipulating food levels from survival,so by lessening host reproductionparasites during the infection of D. magna with the bacterium Pasteuria ra- can keep their host alive longer, thereby reducing costs mosa, we showed that both antagonists are resource-limited and that associatedwith early host death (Baudoin 1975; Obrebski a negative correlation between host and parasite reproduction exists, 1975). Further,when parasitesconsume host reproductive indicating resource competition among the antagonists. Pasteuria tissue, which is rich in nutrients and energy content, they ramosa also induces enhanced growth of its hosts (gigantism), which increasetheir own or survival (Jo- we found to be negatively correlated with host fecundity but posi- may directly fecundity kela et al. tively correlated with parasite reproduction. Because infected hosts 1993). never recovered from infections, we concluded that gigantism is ben- Models of this process suggest that the optimal degree eficial only for the parasite. Hosts, however, have evolved counter- of virulenceis total host castration(Obrebski 1975; Jaenike adaptations. We showed that infected hosts have enhanced repro- 1996; O'Keefe and Antonovics 2002). The apparent ad- duction before castration. This shift to earlier reproduction increases vantages of castration for the parasite are so strong that overall host fecundity and compromises parasite reproduction. Fi- imperfect castrationhas been cited as an example of sub- we showed that this resource conflict is to nally, subject genetic optimal parasitevirulence (Jaenike1996), and it has been variation among host and parasite genotypes within a population asked why all parasitesdo not castratetheir hosts (Ebert and is therefore likely to be an important force in the coevolution and Herre Models on the evolution of of virulence in this system. A verbal model is presented and suggests 1996). parasitic that the adaptive value of gigantism is to store host resources, which castrationare based on two key assumptions:first, parasite are liberated after parasitic castration for later use by the growing growth and reproductionare limited by host resource,and parasite. This hypothesis assumes that infections are long lasting, that second, limited resourcescause a negative correlationbe- is, that they have a high life expectancy. tween host fecundityand the production of parasitetrans- mission stages.Testing these assumptionsrequires a system * Correspondingauthor; e-mail: [email protected]. in which variationin host and pathogen reproductivesuc-

Am. Nat. 2004. Vol. 164, pp. S19-S32. ? 2004 by The University of Chicago. cess can be assessed independently.The first aim of this 0003-0147/2004/1640S5-40121$15.00. All rights reserved. study was to test these two assumptions, which are key S20 The AmericanNaturalist for the understanding of the evolution of virulence of subject of both laboratoryand field studies documenting castratingparasites (Hurd 2001). the potential for reciprocalnatural selection (Ebert et al. Parasite-inducedcastration is often associatedwith en- 1998;Little and Ebert2000; Cariuset al. 2001). Following hanced body growth of the host (gigantism), a trait that infection, typicallyall hosts are castratedby their parasite. is among the most puzzling parasite-relatedchanges in This system has a number of features that makes it well host life history (Mouritsen and Jensen 1994; Gorbushin suited for studying coevolution and testing assumptions 1997; Sorensen and Minchella 1998; Gorbushin and Le- and predictions of mathematical models. First, Daphnia vakin 1999; Moore 2002). Parasite-inducedgigantism has can reproduce via apomictic parthenogenesis (sexual re- been observed in diverse taxa, including mollusc, crus- production is possible but can be controlled in the lab- tacean, vertebrate, and plant hosts and bacterial, fungal, oratory), which permits the separation of genetic from and helminth parasites (Ebert et al. 1996; Arnott et al. nongenetic effects. Second, host and parasite lifetime re- 2000; Krist 2000; Pan and Clay 2002). Because body size productive success can be measured independently and is often correlatedwith fitness, it has been suggested that can be related to each other and to other traits such as gigantism may be a host adaptation (Minchella 1985; host size or age. Third, fecundity reduction and gigantism Ballabeni1995). Long-livedhosts that are preventedfrom due to P. ramosavary across host and parasitegenotypes, reproducingby the parasite transferthe energy normally enabling the covariance of these traits to be measured allocated for reproduction to enhanced growth. Larger with other traits of interest, which is important for un- hosts may have better survival,be more competitive, and derstanding their evolution. Fourth, host castration by have a higher fecundity if they outlive the infection. This P. ramosa is reversible.After treatment with antibiotics, hypothesis requires that infected hosts have a reasonable castratedDaphnia resume reproduction (Little and Ebert chance to recover from the infection and resume repro- 2000), suggesting that the parasite does not destroy the duction. An older and as yet untested hypothesis is that reproductivemachinery of its hosts but may instead use host gigantism is beneficial only for the parasite(Baudoin chemical means (e.g., hormonal control) to castratetheir 1975; Dawkins 1982; Sousa 1983). Under this scenario, hosts. This is important when considering potential ben- the parasite suppresses host reproduction to make re- efits to the host because hosts could gain an advantage sources availablefor itself that the host would have used by resuming reproductionafter they outlive the infection. for reproduction.Gigantism is then a by-product of more energy being released by castration than the parasitecan use at that time. Later in the infection, the parasite may Material and Methods be able to use the resources stored in the host's body The Host tissue. These two hypotheses about gigantism make con- trasting predictions:The host-benefit hypothesis predicts Daphnia magna Straus is a planktonic freshwater crus- that host size will correlatewith the lifetime reproductive tacean usually found in eutrophic shallow ponds. It is success of the infected host. In contrast, the parasite- attacked by a variety of bacterial, microsporidial, and benefit hypothesis predicts a positive correlationbetween fungal parasites (Green 1974; Stirnadel and Ebert 1997; host size and parasite lifetime reproductive success. An- Little and Ebert 1999; Ebert et al. 2001). Prevalence of other alternative hypothesis is that gigantism is a non- Daphnia parasites can be high (up to 98%), and both adaptive side effect of parasitic castration (Wright 1971; field studies and laboratory experiments have demon- Minchella et al. 1985; Keas and Esch 1997; Probst and strated that these parasites typically have a large impact Kube 1999) and that it benefits neither the host nor on Daphnia fitness (Green 1974; Stirnadel and Ebert the parasite. Finally, a parasite-induced shift in a host 1997; Little and Ebert 1999; Ebert et al. 2001). Daphnia trait may benefit both antagonists (Karbanand English- are iteroparous and have indeterminate growth, which is Loeb 1997), although this idea has not been explored for stepwise because a change in body length occurs only parasite-inducedgigantism. The second aim of this study when the old carapax is shed at molting. Juveniles go was therefore to test for correlations between host body through a series of moltings (instars) before reaching size and the reproductivesuccesses of hosts and parasites maturity. At 200C and given 5 x 106 cells green algae in order to distinguish among the different hypotheses (Scenedesmussp.) food per day, the first young are re- for gigantism. leased about 10 days after birth. Juveniles are released We conducted a series of experiments using the plank- with every adult instar, which is about every 3-4 days. tonic crustacean Daphnia magna and the castrating and The first clutch contains about 10 juveniles, whereas up gigantism-inducing bacterium Pasteuria ramosa to test to 30 juveniles are produced in later clutches. Uninfected hypotheses on the adaptive significance of castrationand hosts live for >60 days under laboratory conditions. gigantism. Daphnia magna and P. ramosahave been the In all experimentsdescribed here, D. magnawere kept Evolutionof ParasiticCastration S21 under standardizedlaboratory conditions with artificial day. On the third day, they were challenged with 105 culturemedium (Ebertet al. 1998), a temperatureof 200C, spores of P. ramosa(Gaarzerfeld strain) or with a placebo and a 16L:8D cycle. If not mentioned otherwise, indi- solution. After five more days each of the two treatment vidual Daphnia were kept in 100 mL medium and were groups was split in half; one portion received a high food fed daily with 5 x 106 cells of the green algae Scendesmus level (5 x 106 cells/day), and the other a low food level sp. grown in continuous chemostat cultures. (106 cells/day). We checked all the females every 12 h until they reached maturity to determine age at first re- From each of these four treatment The Parasite production. groups, 15 females were checked daily for reproduction and sur- Pasteuriaramosa Metchnikoff 1888 is a bacterialobligate vival until 5 days after the last infected hosts had died. endoparasiteof Daphnia (Ebert et al. 1996). It has been From the other females, on days 24, 30, 36, and 42, we found in prevalences up to 50% in natural populations collected three females from each of the four treatment (Stirnadeland Ebert 1997;Little and Ebert 1999). Infection groups (total n = 48) and measured their body length. takes place when a host comes in contact with waterborne All host females used in this experiment were ground up spores or with spores in pond sediments; the likelihood to assess infection status and, if infected, the number of of infection depends on the dose of spores (Regoes et al. parasite spores produced. All controls and some of the 2003). The bacteriumgrows in the body cavity of its host; spore-treated females were uninfected. Obtaining unbi- in the final state of infection a single host contains several ased estimates of host length after day 42 was not possible million endospores that fill the entire body cavity.At this because of increasingmortality among the infected hosts. point, the infection can be easily recognizedby the naked We further counted P. ramosa spores in all females that eye. The fitness costs for the host are high because all died naturally from the infections. infections lead to castration.Parasite transmission requires host death because spores are only released from the de- The Four-CloneExperiment caying cadaver.Thus, it is possible to estimate the pathogen's lifetime reproductivesuccess by counting the trans- This experiment assessed the covariance among host fe- mission stages in the dead hosts. The large endospores cundity, host body size, and age at first reproductionand (diameter about 5 xm) can be easily counted with a he- parasite spore production. In addition to using four dif- mocytometer using phase contrast microscopy. Pasteuria ferent host clones, we used differentspore dose treatments ramosa spores can be stored at -200C for several years to test whetherincreased spore dose gave the parasitemore without significantloss in viability. control over their hosts. The four clones, from D. magna To infect Daphnia we added a suspension of P. ramosa populations in southern Finland, North and South Ger- spores to the water. The parasite spore suspensions used many, and southernEngland, were raisedin the laboratory. here were produced by grinding up heavily infected hosts To randomize maternaland grand-maternalenvironmen- around the time we expected them to die from the infec- tal effects and to minimize environmentaleffects, we kept tion. Spores were counted and then diluted with a sus- 30 uninfected lines from each clone for three generations pension of ground, uninfected hosts such that each sus- under standardizedenvironmental conditions. From each pension contained the same amount of macerated host of the (30 x 4 = ) 120 lines, four newborn Daphniafrom tissue but different amounts of parasite spores. This was one clutch (second, third, or fourth clutch) were placed important because ground Daphnia tissue might have a individuallyin 20 mL medium (split-brood design). Each nutritional value for the filter feeding hosts. Placebo sus- of these four newborn Daphnia received a differenttreat- pensions contained no spores but contained the same ment: 105, 104, 103 or 0 P. ramosa spores (Gaarzerfeld amount of maceratedhost tissue from uninfectedDaphnia. strain) per milliliter of medium, administeredat day 3 of their life. On day 6, these females were placed in 100 mL medium and checked daily for offspring production and The Two-Food-LevelExperiment survival. Medium was replaced every adult instar (about This experiment tested for the effect of resource limi- 3-4 days). All animals were fed 2 x 106 cells algae/day. tation on both antagonists and its consequence for host On day 32 we measured host body length, tested each growth and castration. For this experiment we used one female for the presenceof Pasteuriainfection, and counted D. magna clone isolated from a pond near Gaarzerfeld the number of parasite spores in the ground-up bodies. in North Germany. Two hundred newborn D. magna, Relativehost body length (length infected/lengthcontrol) born within a period of 12 h in four 1.5-L cultures with and shift in age at first reproduction (age control - age 20 females each, were placed individually in 100-mL jars infected female) were calculatedfor pairs of females com- under standardized conditions with 2 x 106 cells algae/ ing from the same clutch. S22 The AmericanNaturalist

GeneticVariation among Hosts and ParasiteIsolates StatisticalAnalysis For all statisticalanalysis, the datawere tested for departure Here we report on a further analysis of an experiment from normality and transformed if necessary, or non- published earlier (Carius et al. 2001). Daphnia magnain- parametrictests were used. In most cases, the statistical dividualsthat were infected with P. ramosawere collected tests are mentioned in "Results"and are not discussed from a pond in Gaarzerfeld,North Germany,in August here. 1997. They were brought to the lab and placed singly in The genetic correlationsin the four-clone experiment jars filled with 100 ml medium. Nine infected D. magna were calculatedwith a nested ANCOVAcorrelation (Fal- individualsproduced viable offspringbefore parasiticcas- coner and MacKay1996). The same analysiswas used for tration was complete. The offspring,which are genetically the test for genetic correlationbetween host fecundityand identical to their mother but uninfected, were collected spore production from the experiment with nine host and maintainedas single female lines in the lab. The nine clones and nine parasiteisolates. However, this correlation infected mothers of these clones were the source of the is equivalent but not the same as a genetic correlation parasiteisolates. The females were kept until their death, (Falconerand MacKay1996) becauseevery host clone was tested in combination with and the parasite spores were propagatedby infecting 30 several parasite isolates and further hosts from the same host clone. In the final ex- vice versa. The correlation is here presented to illustrate that the conflict over resources is also visible the periment, two spore doses were used (0.2 x 106 and using means of unique genetic combinations of hosts and 1 x 106 spores per host), but here only the data from the parasite genotypes. high dose treatment are presented because the low dose treatment resulted in too few infections. The experiment was a complete cross-infectionexperiment. With nine D. Results clones and nine isolates, there were 81 magna parasite The Two-Food-LevelExperiment combinations (plus nine uninfectedtreatments) with nine replicateseach. Replicatelines had been kept under stan- Parasitespore production increasedwith host age and for dardized conditions for at least three generations before a given host age was higherin well-fed hosts than in poorly start of the experiment.For the experiment,juveniles were fed hosts (fig. 1A; ANCOVA: food [main effect] F = placed singly in 100-mL jars filled with 20 mL medium. 111.64, df = 1,49, P<.0001; age [covariable], F = = On day 1, the spore solution was added. The daily food 100.42,df 1,49, P< .0001, r2 = 0.82). HealthyDaphnia released their first at around 10-12 old and supply until day 5 was 2 x 106 algae cells. After 5 days, young days then a clutch of 3-4 Infected the individuals were transferredinto jars filled with 100 produced eggs every days. Daphnia magna releasedtheir first young about 1.5 days mL medium and fed daily 5 x 106 algaecells. The medium earlier than the controls (fig. 1B; Wilcoxon two-sample was changedwith every clutch, and Daphnia that stopped test (normal approximation): high food, Z = 2.491, due to infection received fresh me- reproduction parasite P = .012; low food, Z = 4.203, P< .0001). Note that to dium every third Infections and the number of day. avoid a food effect during the infection procedure, the clutches were recorded. On 30, all infected day Daphnia food treatmentwas only applied when the hosts were al- were and later the number of mature frozen, on, parasite ready8 days old, that is, 5 days afterthey had been exposed was cadavers were and the spores counted; ground up, to the parasitespores. Therefore,there was no food effect spore counts were determined in a bacteria counting on age at firstreproduction. Infected hosts producedmany chamber (0.1-mm depth; Neubauer ruling) under a light fewer offspringthan controls in both food levels (fig. 1C; microscope with x 600 magnification. In total, host fe- two-wayANOVA with log-transformedoffspring numbers cundity and parasitespore productionwere assessedin the as dependent variable:food, F = 378.7, df = 1,56, P< (9 clones x [9 parasite isolates + 1 control] x 9 repli- .0001; infection, F = 699.6, df = 1,56, P<.0001; cates =) 810 females. At least one female was infected in food x parasite interaction, F = 283.3, df = 1,56, P< 54 of the 81 parasite challenged combinations (in total .0001; r2 = 0.96). These data showed that high food con- 309 infected females). Only the infected femaleswere used ditions were beneficial for both hosts and parasites, in- in the analysis presented here. Furtherdetails of this ex- dicating that both antagonists suffer from resource periment have been published elsewhere (Carius et al. limitation. 2001), where the results of two populations and two dose Infected hosts grew to be larger than uninfected hosts levels are described.Here we analyzedonly the high dose (gigantism;fig. 1D; two-way ANCOVA:age (covariable), treatment from the German population because too few F = 177.8, df = 1,43, P<.0001; infection, F = 126.8, hosts were infected in all other combinations. df = 1,43, P<.0001; food, F = 825.3, df = 1,43, P< Evolution of Parasitic Castration S23

A B 12

A A c 60 .1a a "-a 60 AA oAA A"A A o AAo _-: 40- 20 30A AA40 5i 60 70Cnto nfce Hotag das C2 20 AAA Q_ N510

20 30 40 50 60 70 Control Infected Host age (days) C ~- 300 E g fo r t i

S10 Co WA c 30 o osI Bn U)o

- Cro) 5. 3- - 10 Control Infected 20 25 30 35 40 45 0.2 CoO.E A HostuLwfo age (days)ifce

2 0.8 " - * Highfood, control -ag)0.6 ?-' A High food, infected Co 0.4 ,oCo Low food, control S0.2 A Low food infected S0o 0 10 20 30 40 50 60 70 Host age at death (days)

Figure 1: Host and parasite life-history traits in relation to food level (high food = black symbols,low food = white symbols)and parasitismfrom the two-food-level experiment. A, Parasitespore production plotted against host age (= time past infection + 3 days). B, Age at first reproduction in days in control and infected hosts in two food levels. Note that to avoid food effects during the infection procedure, the food treatment was applied only when the hosts were 8 days old. Therefore,there is no food effect on age at first reproduction.C, Host fecundity (offspringper female) in control and infected hosts in two food levels. D, Host body length (mm) in control and infected hosts in two food levels in relation to host age. E, Survivalof hosts (proportion of host surviving) in control and infected hosts in two food levels.

.0001; interactions were not significant, r2 = 0.96). The all controls were alive when the last infected host died differencein body length between infected and uninfected (Fisher exact test: P< .0001), and under low food con- hosts translatedinto a differenceof about 20%-25% bio- ditions, 80% of the controls survivedthe last infectedhosts mass (using a body-length dry-weight conversion for D. (Fisher exact test: P < .0001). None of the infected hosts magna [Yampolskyand Ebert 1994]). Infected hosts died were able to clear the infection aftercastration had started, earlierthan controls (fig. 1E). Under high food conditions, and none were able to reproducelater in life. S24 The AmericanNaturalist

The Four-CloneExperiment F = 4.17, df = 2,91, P = .018; clone effect, F = 16.62, df = 3,91, P < .0001), and lower host fecundity (fig. 3C; The results from this experiment are presented in two two-way ANOVA as before: dose effect, F = 3.43, df = parts. Because a substantial number of females became 2,97, P = .036; clone effect, F = 56.13, df = 3,97, P< infected only in the highest spore dose treatment,we used .0001). Increasingthe spore dose to even higher levelswill, only the data from the highest dose treatmentin the first however,reduce spore counts due to the parasite'sdensity- part of the analysis(fig. 2 and the correspondingstatistical dependentwithin-host growth (Ebertet al. 2000a). As seen analysis in table 1). In the second part, we specifically in the two-food-level experiment,infected hosts matured address the effect of spore dose; therefore, all data are earlier than the uninfected controls in all three exposure included. regimes of the four-clone experiment,but this effect was Among the infected hosts, significantclone effectswere not significant (paired t-tests in each of the three dose presentfor all traitsanalyzed (one-way ANOVAwith clone treatmentswere not significantafter Bonferroni correction as main effect (only highest spore dose): number of par- for three tests; fig. 3D). Furthermore,there was no dose asite spores, F = 10.36, df = 3,53, P< .0001; total host and clone effect (difference in age at maturity: dose, fecundity, F = 45.79, df = 3,54, P< .0001; number of F = 0.09, df = 2,88, P = .90; clone, F = 2.03, df = host F = df = P < at ma- clutches, 55.56, 3, 54, .0001;age 3, 88, P = .11). None of the infected hosts in this exper- turity, F = 24.73, df = 3, 54, P < .0001;host body length iment were able to clear the infection after castrationhad at age 32, F = 9.75, df = 3,53, P< .0001; relative host started, and none were able to reproducebefore the end body length at age 32, F = 2.83, df = 3, 53, P = .04). We of the experiment. found evidence for a conflict between host and parasite in the form of a negative correlation between total host fecundity and parasite spore production (fig. 2A shows data aftercorrecting for the clone effect,i.e., residualsfrom one-way ANOVAwith clone as main effect;table 1). Fur- ther, the quickerthe parasitecastrated its host (hosts pro- GeneticVariation among Hosts and ParasiteIsolates duced fewer clutches), the more spores the parasitepro- duced (fig. 2B; table 1), suggestingthat host castrationis We sought to test for genetic covariationwith respect to advantageous for the parasite. The negative correlation the conflict over resources between hosts and parasites between parasitespore production and host fecundity re- within a single population.We analyzeddata from all com- mained significant even after correcting for host body binations of nine D. clones and nine ramosa length (partialcorrelation: r = -0.48, P < .001, n = 51). magna P. isolates (nine combination) in which at least Infected hosts that matured at an earlierage were able to replicatesper one female was infected (54 out of 81 combinations).Par- produce more clutches than hosts maturinglater (fig. 2C; asite and host were table 1), indicating a benefit to infected hosts for shifting spore production fecundity negatively correlated across all data Pearson's correlation: maturationto an earlier age. (fig. 4A; = = During the experiment nearly all infected hosts were total phenotypic correlation, r -0.455, df 307, P < found to be larger than the uninfected controls. Larger .0001). A linear regressionwith all data in fig. 4A revealed absoluteand relativehost body length correlatedpositively a slope parameter of -4.66 x 106, meaning that one with parasite spore production and negatively with host clutch of host eggs was equivalentto 4.66 million parasite fecundity (fig. 2D-2G), suggesting that parasite-induced spores. The same correlation but using the unweighted gigantismin the D. magna-Pasteuriaramosa system is ben- means of each of the 54 host clone-parasiteisolate com- eficial for the parasite and disadvantageousfor the host. bination revealed a negative covariance as well (fig. 4B; The three dose levels used in the four-clone experiment Pearson'scorrelation: r = -0.366, df = 52, P = .0065). allowed us to test whether parasitemanipulation of hosts Note that the genetic correlationis equivalentbut not the was dose dependent, that is, whether a higher parasite same as a genetic correlationas usually discussedin quan- biomass during early infection gave the parasite a head titativegenetics texts (Falconerand MacKay1996) because start in its aim to monopolize host resources,as suggested every host clone was tested in combination with several by Sorensen and Minchella (2001). We found that higher parasite isolates and vice versa. A nested ANCOVA re- exposuredose resultedin strongergigantism (fig. 3A;two- vealed an environmental correlation of r = -0.487 way ANOVA:relative host length, dose effect, F = 8.92, (df = 287, P < .001). None of the infected hosts in this df = 2,96, P = .0003; relative host length, clone effect, experimentwere able to clear the infection aftercastration F = 2.01 df = 3,96, P = .12), higher parasite spore had started, and none were able to reproducebefore the counts (fig. 3B; two-way ANOVA as before: dose effect, end of the experiment. A D F 20 A

-~A Ai A o A oooooooooo- A A -l) ,A - ' 0 0 • -5-20 U) I I I 1 I I I I I I I______-15 -10 -5 0 5 10 15 20 U) -0.3 -0.2 -0.1 0 0.1 0.2 0.3 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 o U) > Host fecundity (residuals) Host length (residuals) Host length (residuals) a. Gk20) B 5000000 4 E 5000000 A AA 10 o A I 3 A A o An ,• ,A, -5000000 A -5000000 - -10 AA

A, A -20 -1 0 -0. 0 0.05 0 0.05 0.1 -0.5 0 0.5 1 -5 -0.1 0.1 -0.1 0..05 Clutches per host (residuals) Relativehost length (residuals) Relativehost length(residuals)

C - A 0 1-

AA -o 0-0 AAtAAAAAAk A _c:'•oAA , O -1

-2 -1 0 1 2 Age at first reproduction (residuals)

Figure 2: Scatterplot of host and parasite life-history traits in the four-clone experiment after correcting for clone effects (residuals shown). A, Number of parasite spores per host plotted against host lifetime fecundity. B, Parasite spore numbers per host plotted against number of host clutches. C, Number of host clutches plotted against age at first reproduction. Note that some infected females became castrated before maturation and therefore no data for age at first reproduction are available. D, Parasite spores plotted against host body length. E, Parasite spores plotted against relative host body length (=length infected female/length control female). F, Host fecundity (total number of offspring) plotted against host body length. G, Host fecundity (total number of offspring) plotted against relative host body length. Only data from the infected females of the highest dose treatment are shown. S26 The AmericanNaturalist

Table1: Phenotypic,genetic, and environmentalcorrelations for host and parasite fitnesscomponents from the four-cloneexperiment Correlationbetween traits Phenotypic Genetic Environmental Parasitespores/host fecundity -.52*** -.37 NS -.50*** Parasitespores/number of host clutches -.52*** -.23 NS -.49*** Numberof host clutches/ageat maturity -.76*** -.96 NS -.38** Parasitespores/host body length .27* -.12 NS .29* Parasitespores/relative host bodylength .11 NS -.54 NS .32* Host fecundity/hostbody length -.64*** -.87 NS -.45*** Host fecundity/relativehost body length -.35** -.48 NS -.53***

Note: In this analysis, only the data from the highest spore dose are included. Host fecundity is the total number of offspring of an infected female, and parasite spores are the total number of spores per infected host. Degrees of freedom are 56, 3, and 47 for the phenotypic, genetic, and environmental correlations, except for the correlation between number of host clutches and age at maturity (df = 51, 3, and 48, respectively). NS = not significant. Note that a substantial number of females became infected only in the highest dose treatment of the four-clone experiment. Therefore, we used only the data from the infected females of the highest dose treatment in the statistical analysis in this table. * P < .05. ** P < .01. *** P < .001.

Discussion servation that spore yield increasedwith host age (fig. 1). Killingthe host prematurelyappears costly for the parasite Castrationand the Coevolution Virulence of unless this benefit is discounted by high host adult mor- Adaptive explanations about the evolution of virulence tality. Thus, early castrationfollowed by a comparatively suggest that virulence is an unavoidableside effect of the long life span of the infected hosts is beneficial for the parasite's attempts to achieve transmission. It has been energy-demandingP. ramosa.Other parasitesof Daphnia suggested that host castration is adaptive for pathogens do not show such a temporalseparation between castrating because it allows the pathogen to exploit hosts by mini- and killing their host, do not induce gigantism,and appear mizing the harmfulside effects of killing the host and thus to have much lower energydemands than P. ramosa(Ebert the parasiteprematurely. Our study showed that a path- et al. 2000b; Bittner et al. 2002). ogen directly benefits from castratingits host. This was Our analysis revealedthat the negative correlationbe- indicated by two findings. First, because food limitation tween parasite and host reproduction is not only visible harmed parasitesas well as hosts (fig. 1) by reducingboth when correcting for genetic effects (environmental cor- host and parasitereproductive success, it suggeststhat the relation) but also when using the means of the unique bacterial parasite Pasteuria ramosa has high energy re- genetic combinations of host clone and parasite isolates quirements.A similar result has been observed for a snail (fig. 4). This finding goes hand in hand with the earlier infectedwith a castratingtrematode (Keas and Esch 1997). demonstration that there are strong host-parasiteinter- Second, as a possible consequenceof resourcecompetition, actions for host and parasite fitness components in this host and parasitefecundity were negativelycorrelated with system (Carius et al. 2001), indicating the possibility for each other (figs. 2, 4). This has been postulated to be a antagonisticcoevolution. driving force for the evolution of parasiticcastration, es- Theoreticalmodels about the evolution of parasiticcas- pecially when castrator biomass represents a substantial tration agree that from the parasite'sperspective, the op- portion of host biomass (Baudoin 1975; Sousa 1983), timal level of castrationis total castration(Obrebski 1975; which is typical for castrators(Kuris 1974). The total bio- Jaenike1996; O'Keefeand Antonovics 2002). The key as- mass of P. ramosaspores in a host around time of death sumption of these models-the negative correlation be- can make up more than 10% of the host biomass (D. tween host and parasite reproductivesuccess-was well Ebert, unpublished data). To produce such high parasite supported by our study. Nevertheless, castration in the biomass without killing the host may only be possible Daphnia-Pasteuriasystem is far from being total. Some because P. ramosacan use the resources the host would hosts produced up to five clutches before being totally otherwise invest in reproduction.These resourcescan be castrated(but no hosts escaped castrationonce they were the equivalent of five to 15 clutches of parthenogenetic infected). Our data suggest that one reason P. ramosamay eggs that a Daphnia magnafemale would produce were it not achieveperfect host castrationis becauseinfected hosts not infected. Consistentwith this supposition was the ob- reproduceearlier than uninfected controls, perhapsto ac- Evolutionof ParasiticCastration S27

4.66 million. Thus, it is for P. ramosato castrate A B adaptive its host earlybut to kill it late. Futuregenerations of math- 1.1 -25 ematicalmodels may include host evolution to understand and predict the consequences of the conflict over resources. 1.08 o 20 . , x In summary,the evolution of parasiticcastration in the U) Daphnia-Pasteuriasystem seems to be driven by the par- - 15 asite's needs for resources.The parasite races against the o-1.06 0o host's attemptsto reproducebefore castrationis complete, - while the host tries to secure at least some resourcesbefore . 1.04 10 U U) its reproductive death. Thus, the evolution of virulence - seems to be the result of a tied coevolution between both 5 rr1.02 antagonists,a process often suggestedto maintain genetic . variation for the traits involved (Clarke 1976; Hamilton 1 0 1980; Barrett 1988). Genetic variation for castration and 4 5 6 4 5 6 gigantism may have contributed to the variableoutcome of studies on host life histories in the presenceof parasites Parasite spore dose (Log10[spores / host]) (Ballabeni1995; Sorensenand Minchella 1998; Loot et al. C D 2002). 50 "0.8 The Evolutionof Gigantism >" 4Q _ In the Daphnia-Pasteuriasystem, gigantism seems not to benefit the host. In our experiments,none of the infected V -o 0.46 hosts were able to clear the infection or after 30 0, reproduce castrationwas their life --0 complete, despite long expectancy after castration (four to five times the age at first repro- = 220 duction). The host's failureto reproduceis not due to the 0.2 destruction of the ovaries (castrationis reversiblein this "o o U, system through antibiotic treatment [Little and Ebert I a 2000]); rather, it indicates that the parasite may control 0 456 its host through chemical manipulation. Thus, although it has been that is for the C4SC 4 565 6 4 5 6 suggested gigantism adaptive hosts that are able to reproduce later in life (Minchella Controls (C) and 3 parasite 1985), this conclusion does not seem to apply to our sys- doses spore (Log10[spores /host]) tem. Largerinfected hosts had lower fecunditythan smaller infected hosts, as has also been observed in a trematode snail system (Gorbushin 1997). Figure 3: Host and parasitetraits in the four-clone experimentin relation Gigantism appears to benefit the parasite. Our results to the dose to which hosts were Each bar shows parasitespore exposed. showed that the degree of gigantism and castrationwere the means across the four host clones used in this experiment.Treated but uninfected females were not included. A, Relative host length both positively correlatedwith parasitelifetime reproduc- (= length infected female/length control female). B, Parasitespores per tive success. Gigantismprobably benefits the parasitebe- host female. C, Total host fecundity until the end of the experiment (32 cause larger hosts are a larger resource and may be used days). At this time all infected hosts were totally castrated,while the by the parasitefor increasedspore production. Indeed, in Shift in at of controls continued to produce eggs. D, age maturity (age the final of infection, P. ramosahas such re- controls - of infected stage high age females). source demands that it consumes all available host biomass, filling the entire body cavity of the host with spores. cess resources before the parasite gains control over re- In addition, largerDaphnia are more efficientfilter feeders source allocation. This earlierhost reproductionincreases (Lampert 1987) and are therefore able to acquire more the numbers of clutches relativeto those hosts that do not resources per unit time, which may be ultimately con- shift age at maturity,indicating that this shift is beneficial verted into more parasite spores. However, as gigantism for the host and costly for the parasite. On average,one occurred even when the absolute amount of food was clutch more for the host reduced parasitespore counts by limited (fig. 1), increasingfeeding efficiency alone cannot S28 The AmericanNaturalist

of infection that a of 60 quence dynamics, is, consequence A the growingparasite's changes in energyrequirements. We call this the temporal storage hypothesis. It is illustrated A50 A schematicallyin figure 5. To infect a Daphnia, only very o40 A few P. ramosa spores are necessary (as few as 10 spores I. can lead to infections [Regoeset al. 2003]), suggestingthat A bacterial biomass is very small during early infections. 3)0 Thus, in absolute terms the parasite'sinitial growth re- oo A A few resources. P. ramosa 0, 2030 A AAI I quires only very Although may A benefit by inducing castration early and preventing the host from using valuable resources for reproduction, its o2 , small biomass has no use for the amounts of re- , , large sourcesliberated through castrationduring the earlyphase 0 1 2 3 4 5 of an infection. Resource allocation studies have shown Host clutches before castration that the amount of biomass allocated into Daphnia reproduction is substantial (>60% of daily biomass pro- 45 duction; Lynch et al. 1986), which may be much more B than the parasitemight need at this stage.Thus, a temporal o 40 A storageof these resourcesin form of host body mass could X be adaptive.Consistent with the suggestion that liberated A resources are allocated into host was our C/) 35 A growth finding that the degree of gigantism and the degree of castration AA kA j oL.O 303 tAAA were positively correlatedwith each other. Whether the qI) A shift in resourceallocation was a parasitestrategy or a side " A AAA A A L, 25 A product of host physiology of the castratedhosts is not A clear. However, even if it were a side product of host A AAI AA S20 A A A physiology, it is beneficial for P. ramosa, and because it has no costs for the alreadycastrated (reproductively dead) 15 host, it is not counterselectedby the host. The parasite 0 1 2 3 4 grows, at least initially,with a higher growth rate than its Host clutches before castration host, thus increasingin resourceneeds relativeto the needs of its host. In later stages of its growth, the parasite (in case of a microparasite,the parasitepopulation) needs, in Figure 4: Relationship between parasite spore counts per host female total, largeramounts of resourcesthat the larger(gigantic) ( x 106) and the number of clutches an infected host is able to produce host body may have to offer over a normal-sized host. before castration (i.e., before reproductionceased). Parasitespores were Thus, the temporal storage hypothesis is both a mecha- counted in all females at 30 All nine host clones and nine age days. nistic and an evolutionary explanation for the evolution parasite isolates used were collected from the same population. A, All infected females (Pearson's correlation: r = -0.455, P< .0001, n = of parasiticcastration and gigantism.It is possibleto derive 309); B, means of each host clone-parasite isolate combination (r = a number of testable predictions from this model. -0.366, P = .0065, n = 54) in which at least one host was infected. First, temporal storage of resourcesyields benefits only for long-lasting infections and under conditions of low explain gigantism. The hypothesis that larger hosts may host (and thus parasite)mortality rates. If the parasitewere provide an advantageto the parasitein form of increased likely to be cleared by the host immune defense or if the host survival (e.g., due to reduced predation and com- host were to die (parasiteinduced or due to other reasons), petition; Baudoin 1975;Dawkins 1982;Arnott et al. 2000) it would not be adaptivefor resourcesto be stored. Thus, cannot explain the correlation between parasite spore castrationand gigantism should be found predominantly counts and host size in our experimentsbecause all ani- in systems with long-lasting (chronic) infections (fig. 5) mals were kept separatelyand without predators.In nat- and low host (and parasite) mortality. ural situations, gigantism may, however, provide this ad- Second, gigantism should be induced only when the ditional advantagefor the parasite. total energy demands of the growing parasite during the The mechanism and evolution of gigantism is poorly time of castrationplus the energy costs of the host's im- understood. We agree with earlier authors (Sousa 1983; mune defense are substantiallylower than the resources Sorensen and Minchella 2001) that it may be a conse- liberatedby castration.The growingparasite's low resource Evolutionof ParasiticCastration S29

UNINFECTED HOSTS INFECTED HOSTS Cumulative Castration resources investedinto aturity Maturity host reproduction

Enhancedhost Gain in host Enhancedhost body mass Cumulative body growth due to resources parasitic investedinto Birth Infection castration host body (= gigantism) mass

Time \Time Resource requirements Resource requirementsof of slow growing parasite rapidly growing parasite

Figure 5: Schematic illustration of the temporal storage hypothesis for the evolution of parasiticcastration and gigantism.The two graphs on the left show the cumulative biomass of a healthy host for reproduction (top) and growth (=increase in host body mass; bottom). The host has indeterminategrowth. On the right, the host is infected with a castratingparasite during its juvenile phase. Some time after infection, the parasite castratesthe host and thus prevents it from investing more resources into reproduction (upper rightgraph). As the parasiteneeds little resources during its early growth phase (solid line in bottom right graph), the liberated resources are invested into host body growth (see increase in the cumulative resources for host body mass). Later,the now larger parasite can make use of the additional resourcespresent in the largerhost body. Earlyon, a fast-growingparasite (stippledline in bottom rightgraph) would need much of the host's resourcesto grow and would deplete the host early. requirementsmay be a consequence of low parasitebio- which is consistent with comparativeevidence (Baudoin mass during the initial stage of the infection or of slow 1975; Moore 2002). parasite growth (but see next prediction). The latter may A third prediction related to the one above is that par- be a strategy to avoid the host immune defense, which asite biomass increasesat a higher rate than host biomass would keep the total resource costs (parasitegrowth plus until resource depletion slows down parasitegrowth (fig. host defense) of the infection initially low. The second 5). For microparasitesthis can be measured as the pop- prediction may further narrow down why parasitic cas- ulation growth rate, while for macroparasitesthis is in- tration and gigantism are found only in certain host- dividual body increase. It is important here that growth parasitesystems. Parasites with largetotal energydemands is slowed down by resourcedepletion and not by the host early during an infection (due to parasitegrowth or host immune defense. As a consequence, it is expected that defense; compare solid and stippled lines for parasite parasite biomass reaches a substantialproportion of the growth in the lower right graph of fig. 5) may not allow host biomass, as has been observedbefore (Baudoin 1975; excess resourcesto build up even if the parasite castrates Sousa 1983). its host. In these cases, resource limitation would occur Fourth, enhanced host growth relative to uninfected so earlyduring an infection that the resource-depletedhost controls should be observed only during early phases of would also be killed much sooner and the host's fecundity infection, when the parasite biomass is still small. Host reduction and death would not be separatedin time. If, growth may be stunted in late phases of infection when however, castration and host death are clearly separated the parasiteneeds more resources.This fourth prediction in time, it is often associatedwith the finding of gigantism, is consistent with resultsfrom earlierstudies on gigantism S30 The AmericanNaturalist

(Minchellaet al. 1985; Gerardand Theron 1997) and may upper hand, indicatingthat reciprocalselection is ongoing. explain some of the variation in host growth across in- In such an arms race, adaptationsof high value today may dividualsseen in field observationsin other castratorsys- be of low value in the future. tems where the age of the infections was not known (Gor- bushin Sorensen and Minchella 1997; 2001). Acknowledgments Fifth,parasites may achievehigher levels of control over the host or gain more rapid control over the host when We thank J. Hottinger for laboratoryassistance and K.-H. more individualsenter a host (higherexposure dose). Our Jensenfor helpful discussions.S. Zweizigimproved earlier results suggest that with higher spore doses, parasitescas- versions of this manuscript.This study was supportedby trate their hosts more effectively,induce stronger gigan- the Swiss Nationalfonds. tism, and produce more spores (fig. 3). Similarresults have been reportedfrom other castratorsystems (Zakikhani and LiteratureCited Rau 1999; Sorensen and Minchella 2001). In an earlier study that included 13 dose levels, we found that P. ramosa Arnott, S. A., I. Barber, and F. A. Huntingford. 2000. producedmore spores as dose increased.However, beyond Parasite-associated growth enhancement in a fish- a certain dose level, spore production rapidly declined, cestode system.Proceedings of the RoyalSociety of Lon- presumablybecause within-host competition among par- don B 267:657-663. asites (density dependence)became so strong that parasite Ballabeni,P. 1995. Parasite-inducedgigantism in a snail: development within the host was stunted (Ebert et al. a host adaptation?Functional Ecology 9:887-893. 2000a). This suggeststhat higher exposuredoses may help Barrett, J. A. 1988. Frequency-dependentselection in the parasiteto gain control over the host but at the same plant-fungalinteractions. Philosophical Transactions of time increase density dependence and thus reduce spore the Royal Society of London B 319:473-483. yield. We predict that gigantism will also peak at inter- Baudoin, M. 1975. Host castrationas a parasiticstrategy. mediate dose levels. Consistent with this prediction, the Evolution 29:335-352. highest lifetime cercariaeproduction and largestadult size Bittner,K., K. O. Rothhaupt,and D. Ebert.2002. Ecolog- has been found at intermediatedose levels in a study of ical interactions of the microparasiteCaullerya mesnili a digenean parasite of a snail (Zakikhaniand Rau 1999). and its host Daphnia galeata. Limnology and Ocean- ography 47:300-305. 1994. Virulence. Evolution 48:1423-1437. Conclusions Bull, J. J. Carius, H. J., T. J. Little, and D. Ebert. 2001. Genetic Our study suggeststhat castrationand gigantismare adap- variationin a host-parasiteassociation: potential for co- tive for the parasiteP. ramosainfecting D. magnabecause evolution and frequency-dependentselection. Evolution they are linked with greaterproduction of parasitespores. 55:1136-1145. More generally,we suggest that castrationand gigantism Clarke, B. C. 1976. The ecological relationship of host- are adaptivefor parasitesthat have a high life expectancy parasite relationships.Pages 87-104 in A. E. R. Taylor (chronic infections) and that have low resource require- and R. M. Muller, eds. Genetic aspects of host-parasite ments during the initial stage of an infection but high relationships.Blackwell, Oxford. requirementslater on when resources can be mobilized Dawkins, R. 1982. The extended phenotype. Oxford Uni- from the enlarged host body. The host body serves as a versity Press, Oxford. temporaryresource storage unit. Ebert,D., and E. A. Herre. 1996. The evolution of parasitic Our study supports the idea that antagonistsmay com- diseases. ParasitologyToday 12:96-100. pete for resourcesand that this conflict is subjectto genetic Ebert, D., P. Rainey,T. M. Embley,and D. Scholz. 1996. variation among both host and parasite genotypes. The Development,life cycle, ultrastructureand phylogenetic traitsinvolved on both sides may be of quantitativegenetic position of Pasteuria ramosaMetchnikoff 1888: redis- nature rather than single gene effects as is often assumed covery of an obligate endoparasiteof Daphnia magna in coevolutionary models (Barrett 1988; Gandon et al. Straus. PhilosophicalTransactions of the Royal Society 1996). Whether the parasite or the host is ahead in the of London B 351:1689-1701. "armsrace" may depend on the relativeevolutionary speed Ebert, D., C. D. Zschokke-Rohringer,and H. J. Carius. with which the antagonistsadapt. This in turn may depend 1998. Within and between population variation for re- on the rates at which genetic variationis createdthrough sistanceof Daphnia magnato the bacterialendoparasite mutations, sexual recombination, and migration (Ham- Pasteuriaramosa. Proceedings of the Royal Society of ilton 1980; Lively 1999). In the D. magna-P. ramosasys- London B 265:2127-2134. tem, neither antagonistappears to have solidly gained the - . 2000a. Dose effects and density-dependentreg- Evolutionof ParasiticCastration S31

ulation of two microparasitesof Daphnia magna.Oeco- Kuris,A. M. 1974. Trophicinteractions: similarity of par- logia (Berlin) 122:200-209. asitic castratorsto parasitoids.Quarterly Review of Bi- Ebert,D., M. Lipsitch,and K. L. Mangin. 2000b. The effect ology 49:129-148. of parasiteson host population density and extinction: Lampert,W. 1987. Feedingand nutrition in Daphnia.Me- experimental epidemiology with Daphnia and six mi- morie dell'IstitutoItaliano di Idrobiologia45:143-192. croparasites.American Naturalist 156:459-477. Little, T. J., and D. Ebert. 1999. Associationsbetween par- Ebert, D., J. W. Hottinger, and V. I. Pajunen. 2001. Tem- asitism and host genotype in natural populations of poral and spatial dynamics of parasites in a Daphnia Daphnia (Crustacea: Cladocera). Journal of Animal metapopulation:which factorsexplain parasite richness? Ecology 68:134-149. Ecology 82:3417-3434. - . 2000. The cause of parasiticinfection in natural Falconer, D. S., and T. MacKay. 1996. Introduction to populations of Daphnia (Crustacea:Cladocera): the role quantitativegenetics. 4th ed. Longman, Harlow. of host genetics. Proceedings of the Royal Society of Gandon, S., Y. Capowiez,Y. Dubois, Y. Michalakis,and I. London B 267:2037-2042. Olivieri. 1996. Local adaptation and gene-for-gene co- Lively, C. M. 1999. Migration, virulence, and the geo- evolution in a metapopulation model. Proceedings of graphic mosaic of adaptation by parasites. American the Royal Society of London B 263:1003-1009. Naturalist 153(suppl.):S34-S47. Gerard,C., and A. Theron. 1997. Age/size- and time-spe- Loot, G., R. Poulin, S. Lek, and J. FEGuegan. 2002. The cific effects of Schistosomamansoni on energyallocation differentialeffects of Ligulaintestinalis (L.) plerocercoids patterns of its snail host Biomphalariaglabrata. Oeco- on host growth in three natural populations of roach, logia (Berlin) 112:447-452. Rutilus rutilus(L.). Ecology of FreshwaterFish 11:168- Gorbushin, A. M. 1997. Field evidence of trematode- 177. induced gigantism in Hydrobiaspp. (Gastropoda:Pro- Lynch,M., L. J. Weider,and W. Lampert.1986. Measure- sobranchia). Journalof the Marine Biological Associa- ment of the carbon balancein Daphnia.Limnology and tion of the United Kingdom 77:785-800. Oceanography31:17-33. Gorbushin, A. M., and I. A. Levakin. 1999. The effect of Minchella, D. J. 1985. Host life-history variation in re- trematode parthenitaeon the growth of Onobaaculeus, sponse to parasitism.Parasitology 90:205-216. Littorinasaxatilis and L. obtusata(Gastropoda: Proso- Minchella, D. J., B. K. Leathers,K. M. Brown, and J. N. branchia).Journal of the Marine BiologicalAssociation McNair. 1985. Host and parasitecounteradaptations: an of the United Kingdom 79:273-279. example from a freshwatersnail. American Naturalist Green,J. 1974. Parasitesand epibionts of Cladocera.Trans- 126:843-854. actions of the ZoologicalSociety of London 32:417-515. Moore, J. 2002. Parasites and the behavior of animals. Hamilton, W. D. 1980. Sex versus non-sex versus parasite. Oxford University Press, Oxford. Oikos 35:282-290. Mouritsen, K. N., and K. T. Jensen. 1994. The enigma of Hurd, H. 2001. Host fecundity reduction: a strategy for gigantism: effect of larval trematodes on growth, fe- damage limitation?Trends in Parasitology17:363-368. cundity, egestion and locomotion in Hydrobia ulvae Jaenike, J. 1996. Suboptimal virulence of an insect- (Pennant) (Gastropoda,Prosobranchia). Journal of Ex- parasiticnematode. Evolution 50:2241-2247. perimentalMarine Biology and Ecology 181:53-66. Jokela,J., L. Uotila, and J. Taskinen. 1993. Effects of cas- Obrebski,S. 1975. Parasitereproductive strategy and evo- trating trematode parasite Rhipidocotylefennica on en- lution of castration of hosts by parasites.Science 188: ergy allocation of fresh-waterclam Anodontapiscinalis. 1314-1316. Functional Ecology 7:332-338. O'Keefe,K. J., and J.Antonovics. 2002. Playingby different Karban, R., and G. English-Loeb. 1997. Tachinid para- rules:the evolution of virulencein sterilizingpathogens. sitoids affecthost plant choice by caterpillarsto increase American Naturalist 159:597-605. survival. Ecology 78:603-611. Pan, J. J., and K. Clay. 2002. Infection by the systemic Keas, B. E., and G. W. Esch. 1997. The effect of diet and fungus Epichloeglyceriae and clonal growth of its host reproductivematurity on the growth and reproduction grass Glyceriastriata. Oikos 98:37-46. of Helisoma anceps (Pulmonata) infected by Halipegus Probst, S., and J. Kube. 1999. Histopathologicaleffects of occidualis(Trematoda). Journal of Parasitology83:96- larval trematode infections in mudsnails and their im- 104. pact on host growth:what causes gigantismin Hydrobia Krist, A. C. 2000. Effect of the digenean parasite Proter- ventrosa(Gastropoda: Prosobranchia)? Journal of Ex- ometra macrostoma on host morphology in the fresh- perimental Marine Biology and Ecology 238:49-68. water snail Elimia livescens.Journal of Parasitology86: Regoes, R. R., J. W. Hottinger,L. Sygnarski,and D. Ebert. 262-267. 2003. The infection rate of Daphnia magnaby Pasteuria S32 The AmericanNaturalist

ramosa conforms with the mass-action principle. Epi- Stirnadel,H. A., and D. Ebert.1997. Prevalence,host spec- demiology and Infection 131:957-966. ificity and impact on host fecundity of microparasites Sorensen, R. E., and D. J. Minchella. 1998. Parasiteinflu- and epibionts in three sympatricDaphnia species. Jour- ences on host life history: Echinostomarevolutum par- nal of Animal Ecology 66:212-222. asitism of Lymnaeaelodes snails. Oecologia (Berlin) 115: Wright, C. A. 1971. Flukes and snails. Allen & Unwin, 188-195. Plymouth. . 2001. Snail-trematode life history interactions: Yampolsky,L. Y., and D. Ebert. 1994. Variationand plas- past trends and future directions. Parasitology ticity ofbiomass allocationin Daphnia.Functional Ecol- 123(suppl.):3-18. ogy 8:435-440. Sousa, W. P. 1983. Host life-history and the effect of par- Zakikhani,M., and M. E. Rau. 1999. Plagiorchiselegans asitic castration on growth: a field-study of Cerithidea (Digenea:Plagiorchiidae) infections in Stagnicolaelodes californicaHaldeman (Gastropoda,Prosobranchia) and (Pulmonata: Lymnaeidae):host susceptibility,growth, its trematodeparasites. Journal of ExperimentalMarine reproduction,mortality, and cercarialproduction. Jour- Biology and Ecology 73:273-296. nal of Parasitology85:454-463.