Recent evolution of extreme cestode growth suppression by a … · elminth parasites (trematodes, nematodes, and cestodes) currently infect ∼24% of humanity (1), undermine agricul-tural

Recent evolution of extreme cestode growthsuppression by a vertebrate hostJesse N. Webera,b,1,2,3, Natalie C. Steinela,b,c,1, Kum Chuan Shima,b, and Daniel I. Bolnicka,b

aDepartment of Integrative Biology, The University of Texas at Austin, Austin, TX 78712; bHoward Hughes Medical Institute, The University of Texas atAustin, Austin, TX 78712; and cDepartment of Medical Education, Dell Medical School, The University of Texas at Austin, Austin, TX 78712

Edited by Edward J. Pearce, Max Planck Institute of Immunobiology and Epigenetics, Freiburg, Germany, and accepted by Editorial Board Member Stephen M.Beverley May 5, 2017 (received for review December 6, 2016)

Parasites can be a major cause of natural selection on hosts, whichconsequently evolve a variety of strategies to avoid, eliminate, ortolerate infection. When ecologically similar host populationspresent disparate infection loads, this natural variation can revealimmunological strategies underlying adaptation to infection andpopulation divergence. For instance, the tapeworm Schistocephalussolidus persistently infects 0–80% of threespine stickleback(Gasterosteus aculeatus) in lakes on Vancouver Island. To testwhether these heterogeneous infection rates result from evolveddifferences in immunity, we experimentally exposed laboratory-reared fish from ecologically similar high-infection and no-infectionpopulations to controlled doses of Schistocephalus. We observedheritable between-population differences in several immune traits:Fish from the naturally uninfected population initiated a strongergranulocyte response to Schistocephalus infection, and their gran-ulocytes constitutively generate threefold more reactive oxygenspecies in cell culture. Despite these immunological differences,Schistocephalus was equally successful at establishing initial infec-tions in both host populations. However, the no-infection fish dra-matically suppressed tapeworm growth relative to high-infection fish,and parasite size was intermediate in F1 hybrid hosts. Our resultsshow that stickleback recently evolved heritable variation in theircapacity to suppress helminth growth by two orders of magnitude.Data frommany natural populations indicate that growth suppressionis widespread but not universal and, when present, is associated withreduced infection prevalence. Host suppression of helminth somaticgrowth may be an important immune strategy that aids in parasiteclearance or in mitigating the fitness costs of persistent infection.

immune evolution | helminth | resistance | Schistocephalus solidus |threespine stickleback

Helminth parasites (trematodes, nematodes, and cestodes)currently infect ∼24% of humanity (1), undermine agricul-

tural productivity (2, 3), and threaten conservation of some wildpopulations (4). These social, economic, and environmental costsmotivate substantial interest in how vertebrate hosts combat hel-minth infection. Helminths often severely reduce their hosts’ fitnessand therefore represent a strong source of natural selection inhost populations (5–7). In response, vertebrates have evolved acomplex repertoire of innate and adaptive immune responsesthat serve to detect and eliminate infections or to promote tol-erance of successful infections.Despite vertebrates’ sophisticated immune systems, helminths

remain common and persistent, often establishing infections thatlast months or years. Parasites’ continued success may be attributedto (i) hosts’ failure to evolve effective resistance because of in-sufficient time or genetic diversity (8), (ii) trade-offs that constrainhosts’ ability to evolve costly immune traits (9–12), or (iii) parasites’counter adaptations to undermine host immunity (13). These fac-tors may apply unequally across host populations because ofdifferences in host–parasite encounter rates, host genetic diversity,coevolutionary time, ecological costs of immunity, or parasite gen-otypes (14). As a result, some hosts may be substantially moreresistant or tolerant than others.

Natural variation in host immunity provides an opportunity toelucidate vertebrates’ evolving strategies to limit helminth in-fections. This evaluation can be done by surveying diverse nat-ural populations to identify instances of exceptionally high or lowparasite prevalence and testing whether these differently in-fected populations vary with respect to immune phenotypes.Finally, we can evaluate whether genetic differences in resistanceare responsible for the natural variation in infection rates. Thisapproach is exciting, because it can identify immune strategiesthat natural selection has favored to mitigate parasite infections.Unlike traditional immunogenetics, which uses mutagenesis tocreate phenotypic aberrations [frequently involving loss-of-functionmutations (15)] or forward genetic approaches to map causative loci(16), natural selection scans entire populations for many genera-tions to find beneficial resistance traits.Here, we document wide-ranging variation in parasite abun-

dance (from 0 to 80% prevalence) in wild populations and showthat this variation is associated with heritable differences in im-mune response and the recent evolutionary gain of an under-appreciated form of resistance: suppression of parasite size.Threespine stickleback (Gasterosteus aculeatus) inhabit brackishand freshwater habitats throughout coastal north temperate re-gions. After the Pleistocene deglaciation (∼12,000 y ago), marinestickleback colonized many replicate freshwater lakes on Vancouver

Significance

Large parasites are a persistent source of morbidity andmortality in humans, domesticated animals, and wildlife. Hostsare subject to strong natural selection to eliminate or toler-ate these parasite infections. Here, we document the recentevolution of a striking form of resistance by a vertebratehost (threespine stickleback) against its cestode parasite(Schistocephalus solidus). After the Pleistocene glacial retreat,marine stickleback colonized freshwater lakes, encounteredSchistocephalus, and evolved varying levels of resistance to it.We show that heavily and rarely infected populations ofstickleback can similarly resist Schistocephalus colonization,but rarely infected fish suppress parasite growth by orders ofmagnitude. These populations represent ends of a naturalcontinuum of cestode growth suppression which is associatedwith reduced infection prevalence.

Author contributions: J.N.W., N.C.S., and D.I.B. designed research; J.N.W., N.C.S., K.C.S.,and D.I.B. performed research; J.N.W., N.C.S., and D.I.B. analyzed data; and J.N.W., N.C.S.,and D.I.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. E.J.P. is a guest editor invited by the EditorialBoard.

Data deposition: Data reported in this paper are available on the Open Science Frame-work repository, https://osf.io/e53y4.1J.N.W. and N.C.S. contributed equally to this work.2Present address: Division of Biological Sciences, University of Montana, Missoula,MT 59812.

3To whom correspondence should be addressed. Email: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1620095114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1620095114 PNAS | June 20, 2017 | vol. 114 | no. 25 | 6575–6580

EVOLU

TION

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

0

Island and elsewhere. This colonization brought marine fish intocontact with a freshwater parasite, the cestode Schistocephalussolidus. Because S. solidus eggs do not hatch in brackish water (17),marine stickleback are rarely infected by this cestode and thereforehave not evolved effective resistance (18). Newly establishedfreshwater populations encountered S. solidus at a higher rate,resulting in reduced survival and fecundity (19, 20) and selection forincreased resistance. Independently colonized lake populationsevolved parallel gain of resistance (18). Prior studies showed thisresistance entails both reduced infection establishment rates (18,21) and slight suppression of parasite growth (21, 22). Here, weshow that this parallel evolution is incomplete: Lake populationsdiffer with respect to immune responses and parasite resistance,because some populations evolved to suppress cestode growth bytwo orders of magnitude.

ResultsInfection Prevalence Differs Among Natural Stickleback Populations.Samples of stickleback from 50 lakes on Vancouver Island (Ta-ble S1) revealed among-population variation in S. solidus in-fection prevalence spanning at least two orders of magnitude andpersisting for a decade (χ2 = 4,629, df = 49, P < 0.0001) (Fig. 1).To test experimentally for variation in stickleback resistance toS. solidus, we focused on two lakes that bracket the range ofinfection prevalence. Gosling Lake stickleback (GOS) had thethird highest infection prevalence in our sample (50–80% acrossyears). In contrast, a decade of sampling revealed no infectionsin Roberts Lake stickleback (ROB) (Fig. 1; n = 1,480 fish).Ecological differences between the two lakes are minimal.

Gosling and Roberts Lakes are only 17 km apart, both aremoderately large (60 and 160 ha, respectively) and are similarlydeep (maximum depth, 40 and 53 m, respectively). Both lakescontain large pelagic zones in which stickleback forage on co-pepods (S. solidus’ first host) at similar rates [averaging 0.444 and0.457 cyclopoid copepods per fish per day, respectively [Poissongeneralized linear model (GLM) P = 0.9707 for n = 27 and35 fish] (23). Both lakes host at least one breeding pair of loons(a terminal host for S. solidus) per summer. The Roberts Lakeloons make regular foraging trips to nearby lakes where S. solidusis common. Consequently, there is no currently known differencein stickleback’s risk of exposure to S. solidus in Gosling andRoberts Lakes.

Divergence in Immune Responses. We used a common-gardenbreeding experiment to test whether GOS and ROB sticklebackevolved divergent immune phenotypes. We crossed wild-caughtadults from GOS and ROB to generate 51 families, which in-cluded both pure-population fish and F1 hybrids with GOS orROB dams (G×R or R×G, respectively). The eggs were hatchedand reared to adulthood in laboratory aquaria. We also bred

S. solidus plerocercoids obtained from infected fish from threelakes in British Columbia. Approximately five fish per familywere fed copepods containing infective S. solidus procercoids(Fig. S1). One to two siblings per family were fed uninfectedcopepods as a control (Fig. S1). Approximately 42 d postexposure,we measured host immune phenotypes to test for constitutive andinfection-induced differences between stickleback populations.Here, we focus on two traits previously associated with defenseagainst macroparasites: the ratio of granulocytes to lymphocytesin head kidney (HK) primary cell cultures and the magnitude ofreactive oxygen species (ROS) produced by granulocytes (9,24–28).The relative abundance of granulocytes differed between the

sexes (4.7% less in males; t = 4.03, P = 8.4e-5) (Fig. S2) and alsovaried among stickleback families within the GOS and ROBpopulations (random family effect, Table S2). This family-levelvariation could be caused either by segregating genetic variationwithin populations or by transgeneration effects of environmentaldifferences among the natural-caught parents (29). Granulocyterelative abundance was independent of cestode genotype (TableS2) and showed no statistically significant main effect of exper-imental exposure to or infection by S. solidus (Fig. S3).After accounting for family effects and averaging over sex,

granulocytes constituted a lower proportion of HK cells in ROBfish (5.4% less than in GOS fish; t = 2.06, P = 0.05). Despite thislower initial frequency, ROB fish responded to S. solidus infectionby increasing granulocyte proportions (16.3% higher than un-infected ROB fish; t = 5.13, P < 0.0001) (Fig. 2A). Infection had noeffect on GOS fish, resulting in an interaction between fish pop-ulation and infection status (Table S2). An interaction also oc-curred in F1 hybrids: Infection did not affect granulocyte frequencyin G×R hybrids, but granulocytes increased 11.4% in infectedR×G fish (t = 2.17, P = 0.03) (Fig. 2A). This asymmetry in re-sponse to infection by reciprocal F1 hybrids strongly suggests that amaternal effect in ROB stickleback drives the observed phenotype.The production of ROS was quantified by measuring the

median fluorescence intensity (MFI) of HK cells stained withDihydrorhodamine-123 (DHR). DHR is a cell-permeable fluo-rophore that fluoresces in the presence of peroxide and perox-ynitrite (two forms of ROS). For each fish we measured(i) baseline ROS production in unstimulated granulocytes and(ii) ROS production of cells exposed to Phorbol 12-myristate 13-acetate (PMA), which is a potent granulocyte stimulant. Here,we focus on the second metric (see Fig. S4 for baseline ROS).Similar to granulocyte frequency, ROS production varied amongsibships (random family effect, Table S2). Neither fish sex norcestode genotype had detectable effects on granulocyte ROSproduction. After accounting for family differences, PMA-stimulated ROB cells generated 2.4-fold greater MFI than GOScells (post hoc pairwise Tukey tests on ln-transformed MFI, t = 7.89,P < 0.0001) (Fig. 2B). ROS production by F1 hybrids’ granulocytes

Pro

porti

on o

f fis

h in

fect

ed w

ithS

chis

toce

phal

us s

olid

us

Lakes on Vancouver Island

0.0

0.2

0.4

0.6

0.8

Year

0.0

0.4

0.8

2004 2008 2012 2016Roberts Lake

GoslingLake

Fig. 1. Natural variation in S. solidus prevalence among 50 lakes on Van-couver Island, British Columbia. Lake names, sample year, sample size, andlocation details are given in Table S1. Error bars indicate SEs of each sample.The Inset presents data from two focal populations, Roberts and GoslingLakes, over a roughly 10-y time span.

Per

cent

Gra

nulo

cyte

s25

3035

4045

GOS GxR RxG ROB

InfectedUninfected

A

MFI

of P

MA

+ G

ranu

locy

tes

13.4

13.8

14.2

14.6

GOS GxR RxG ROB

B

Fig. 2. Variation in stickleback immunity, as measured by (A) the percent-age of granulocytes in HK primary cell culture and (B) the MFI of PMA-stimulated granulocytes. Error bars indicate one SE interval above and belowthe mean. Sample sizes of uninfected and infected fish in each population: GOS,95 and 16; G×R, 43 and 10; R×G, 57 and 7; ROB, 105 and 9, respectively.

6576 | www.pnas.org/cgi/doi/10.1073/pnas.1620095114 Weber et al.

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

0

was intermediate between ROB and GOS stickleback (Fig. 2B).G×R and R×G cells generated equivalent ROS levels (P = 0.988),less ROS than ROB fish (P = 0.032 and 0.061, respectively), andmore ROS than GOS fish (both P ≤ 0.0034). This trait did notdiffer between infected and uninfected (control or experimen-tal) stickleback, nor was there an interaction between fishpopulation and infection status (Table S2). Because ROS wasunaffected by infection status, we infer that ROS productiondiffers constitutively between GOS and ROB fish. Overall, weconclude that GOS and ROB stickleback display heritabledifferences in immune responses. Some differences (i.e., ROSlevels) are constitutive and are consistent with genetic effects,whereas others (i.e., the percentage of granulocytes) are in-duced by infection and could be caused by either environmentalor genetic mechanisms.

Recently Evolved Differences in Resistance. GOS and ROB stick-leback did not differ significantly in the frequency or intensity oflaboratory infections (Fig. 3 A and B). Infection rates in GOSand ROB fish were 20.0 and 11.4% (with SEs of 4.5 and 3.6%),respectively (binomial GLM, Z = −1.47, P = 0.14). Includingrandom effects of host family did not improve model fit [Akaikeinformation criterion (AIC) = 142.3 and 140.4 with and withoutfamily, respectively]. Although parasite genotype did not affectthe infection success rate significantly (P = 0.81), there was amarginal interaction between host and parasite genotype (P =0.09) (Fig. S5). Infected GOS and ROB fish carried equivalentnumbers of cestodes (1.75 versus 1.78 out of five cestodes pro-vided per fish; binomial GLM, Z = 0.06, P = 0.95). We concludethat the absence of S. solidus from wild ROB fish is not attrib-utable to any greater resistance to colonization by S. solidus. Fishfrom both populations raised and infected in a common-gardenlaboratory setting have higher resistance than laboratory-raisedmarine fish (the putative ancestral form), indicating that re-sistance is heritable and recently evolved. This conclusion impliesthat the separately colonized lake populations independentlyevolved quantitatively similar resistance to S. solidus establishment,relative to their highly susceptible marine ancestors (18). Thus weprovide a rarely documented instance of convergently evolved hel-minth immunity in a wild vertebrate.Although infection rate and intensity were equivalent, age-

matched cestodes grew dramatically larger in laboratory-raisedGOS than in ROB stickleback (post hoc test, t = 5.79, P < 0.0001)

(Fig. 3C). After controlling for effects of fish family and hostmass on cestode size (parasite mass increased by ∼0.2 mg forevery 0.1-g increase in host mass; t = 2.81, P = 0.007), parasitesfrom GOS fish were on average 34-fold larger than those fromROB fish. The mean mass of each S. solidus removed from thehigh-infection population (GOS) was 10.2 mg, whereas cestodesfrom ROB fish averaged 0.3 mg. There was no effect of cestodeorigin (Table S2). After controlling for host mass, cestodes grewto intermediate size in G×R and R×G hybrids (means of 2.1 mgand 3.8 mg, respectively). Although S. solidus sizes did not differbetween the reciprocal hybrids (P = 0.88), these cestodes diddiffer from those removed from the pure host populations.Cestodes from GOS fish were 5.3-fold larger than those fromG×R hybrids (P = 0.11) and were 2.9-fold larger than those fromR×G hybrids (P = 0.53). Conversely, cestodes from R×G hybridswere 12.7-fold larger (P = 0.006) and those from G×R hybridswere sevenfold larger (P = 0.02) than those from pure ROB fish.Cestode mass was independent of host length, condition (mass/total length), sex, and infection intensity (Table S2). Note,however, that the influence of host size may change in oldercestodes that are larger and more space-limited (30, 31). Theseresults demonstrate there are heritable and thus evolved differencesbetween the ROB and GOS stickleback, given differences inlaboratory-raised sticklebacks’ ability to suppress cestode growth.The intermediate size of parasites in hybrid fish also strongly sug-gests an additive genetic basis for this evolutionary difference.Because most S. solidus growth occurs in stickleback, egg

production is correlated with the adult parasite’s dry weight, andbecause S. solidus must reach a minimum size (∼10–50 mg) toestablish infections in its terminal host (32, 33), growth sup-pression effectively constrains the parasite’s reproductive po-tential. Arguably, this measure of parasite success represents anextended phenotype (34) of the stickleback, in the sense that thehost’s genotype alters the parasite’s phenotype.To determine whether reduced parasite growth is an ancestral

or derived trait, we measured S. solidus growth in susceptiblemarine fish. These cestodes grew quickly, reaching up to 320 mgin 2.5 mo (mean = 130 mg) (Fig. S6). This size is 4.5-fold largerthan the mean size of S. solidus in GOS fish, possibly because thecestodes from marine fish were older (75–83 d for marine, 41–51 dfor GOS). We then compared cestodes from pure marine fish withcestodes from marine × freshwater F1 hybrids (ROB×M andM×GOS, all assayed 2.5 mo postexposure). Consistent with theresults reported above, cestodes from M×GOS fish (n = 2) were77.7-fold heavier than cestodes from the ROB×M fish (n = 3;t test, P = 0.001; Wilcoxon signed-rank test, P = 0.2). Althoughthis sample is small, the effect size is massive and is quantitativelyconsistent with the fold-difference between cestodes from pureGOS and ROB fish. These data demonstrate that the growth-suppressive phenotype observed in ROB fish is an evolutionarygain of immune function from a susceptible and growth-permissiveancestral marine population, which has evolved rapidly duringthe ∼12,000 y since Roberts Lake was likely colonized by marinefish. The magnitude of cestode growth suppression by ROB fishis also extreme compared with previous stickleback–S. solidusstudies. Kalbe et al. (21) found that S. solidus from Germanygrew to a mean mass of 144 mg in local stickleback from abrackish habitat (infection duration of 84 d). Notably, this par-asite mass is very similar to the mass we observed in marine fishover the same infection duration. In contrast, the same Germanparasites only grew to ∼50 mg in a population of Norwegianfreshwater stickleback (21). Our study differed from the lastexperiment in several respects, making it difficult to compareresults precisely. However, quadrupling the mean mass of ROBparasites (i.e., to 1.2 mg) should adequately account for differ-ences in infection duration between the two studies (especiallyconsidering that ROB fish were approximately twofold largerthan the Norwegian fish). These results indicate that the mag-nitude of cestode growth suppression by ROB stickleback ismore than 40-fold greater than previously reported.

A

C D

B

Fig. 3. Infection success as measured by (A) infection frequency (the pro-portion of stickleback with S. solidus), (B) mean abundance of S. solidus perfish, and (C) log-transformed cestode mass (averaged across all cestodes ineach fish). Error bars indicate SEs. (D) Full-sibling cestodes removed fromGOS and ROB stickleback 42 d postexposure. Fig. S7 presents cestode massesfrom marine genotypes and F1 crosses between marine fish and GOS or ROBfish. Number of experimental fish per population inA: GOS, 80; G×R, 40; R×G, 45;ROB, 79. The number of infected fish per population is noted in Fig. 2.

Weber et al. PNAS | June 20, 2017 | vol. 114 | no. 25 | 6577

EVOLU

TION

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

0

Growth Suppression in Wild Populations. We hypothesize that theextreme growth suppression observed in laboratory-raised ROBfish may explain the absence of observed infections in wild ROBfish. To test this hypothesis, we predicted that stickleback pop-ulations that suppress S. solidus growth more effectively wouldhave correspondingly fewer observable infections. As expected,we found a positive correlation across populations between themean size of S. solidus and the parasite’s abundance. Focusingfirst on variation among fish within populations, we confirmed awell-established crowding effect (30, 31). Controlling for a ran-dom effect of lake, individual stickleback with more S. solidushave correspondingly smaller S. solidus (mixed-model linear re-gression P < 0.0001, controlling for host mass) (Fig. 4A). How-ever, we observed a counter gradient trend across populations:Lakes with higher prevalence of S. solidus on average had largercestodes (P = 0.0058) (Fig. 4B). GOS and ROB stickleback sit atopposite ends of this continuum, ROB growth suppression beingso extreme that we could not locate infections in wild fish tomeasure cestode mass. This counter gradient trend lends addi-tional support to our inference that stickleback populationsdiffer in their ability to suppress parasite growth and that growthsuppression is a repeatedly evolved trait that reduces S. solidusprevalence. An important caveat, however, is that, because theseare natural rather than controlled infections, the age of eachparasite is a potentially important but unknown covariate.

DiscussionStickleback are well known for their parallel evolution of adap-tations to freshwater habitats (35, 36). These transitions includerepeated independent evolution of increased resistance to freshwater-specialized parasites (18). However, as shown here, this immuneevolution is not entirely parallel: ROB and GOS stickleback livein ecologically similar lakes, but, unlike GOS stickleback, ROBstickleback evolved an ability to suppress S. solidus growth by twoorders of magnitude. We infer that growth suppression by ROBfish explains the apparent absence of S. solidus in wild-caughtfish from Roberts Lake. In laboratory trials, these fish are just assusceptible to infection as GOS fish. However, growth-suppressedcestodes can be so small that extant infections in wild-caught fishcould easily be overlooked. Our survey of wild-caught fish sug-gests that growth suppression may be a widespread phenomenon;lakes with smaller S. solidus tended to also have lower parasiteabundance.This work identifies a number of heritable differences between

stickleback populations related to parasite resistance and im-munity, despite parallel evolution of reduced infection rates

relative to marine ancestors. However, why the lake populationsevolved such divergent phenotypes is unknown. Evolutionarytheory suggests that selection should optimize the intensity ofimmune responses to balance marginal costs and benefits (37),which depend on ecological variables such as parasite exposurerates, nutritional state, and predation risk, as well as behavioraldifferences among populations of stickleback (38–41). The popu-lations studied here consume copepods (the first host of S. solidus)at roughly comparable rates (23) and thus should have similarexposure risks. However, these populations’ diets differ in otherrespects. Future ecological measurements (e.g., natural rates ofcopepod infections and exposure rates) and experiments (e.g.,transplants and perturbation of immune traits) are necessary toaddress this issue.An important caveat is that we exposed stickleback to a one-

time dose of five tapeworms. This exposure is probably lowerthan natural exposure rates in some populations, which may bedrawn out over an extended period (we observed some wild fishinfected with >20 S. solidus). Nevertheless, we believe our resultsare ecologically relevant. We found large tapeworms not only inlaboratory-reared GOS fish with low parasite exposures andloads but also in naturally caught GOS fish with high parasiteloads. Fast growth of cestodes in GOS fish means that wild in-dividuals will tend to harbor at least some large and successfulcestodes, regardless of what happens to later exposures. Howthese fast-growing cestodes impact later arrivals through directcompetition for host resources or through indirect effects viahost immune suppression or activation is an interesting question.Negligible growth of cestodes in our laboratory-raised ROB fishsuggests that infections will typically fail (or go undetected).Even if wild ROB fish experience high exposure rates, the ab-sence of detectable cestodes indicates that this exposure does notinfluence parasite size or establishment. Finally, although para-site load was not a significant predictor in any of our statisticalmodels, high rates or multiple bouts of exposure may lead tomore complex immune dynamics than those reported here.Previous laboratory-based infection experiments found differ-

ences in tapeworm size among stickleback populations (21, 22).However, the extreme magnitude of growth suppression (>40-foldstronger than previous reports), ancestral context for divergence,rapid evolution, and tremendous natural variation in parasite sizethat we report here considerably extend the implications of thisextended phenotype. Several lines of evidence suggest that growthsuppression may be common in other vertebrates as well. Thereis pervasive heritability for resistance-related traits in ruminantlivestock (42). Most livestock studies measure helminth burdenand resistance using noninvasive fecal egg counts (FECs).However, helminth length has very high heritability among hostsand likely contributes to variation in FEC (43, 44). Even humansharbor heritable variation in helminth resistance and size (45).Notably, a study on hookworm weight in Papua New Guineavillagers suggests that humans vary in their ability to regulateparasite size (46).Bishop elegantly defined resistance as a host’s ability “to exert

control over the parasite life cycle,” even in persistent infections(e.g., heritable variation for lower FEC) (47, 48). Using thisdefinition, stickleback control of cestode growth represents aneffective form of resistance. In populations such as GOS thatlack growth suppression, individual cestodes can grow to be up to72% of the parasite-free mass of their host. Such large parasitescan impose severe fitness costs by limiting a host’s ability toforage, grow, disperse, evade predators, and breed (19, 20, 49).To the extent that these effects depend on parasite size, growthsuppression may partially or wholly rescue the fitness of ROBfish by facilitating survival or reproduction despite infection.Although tapeworm growth suppression could also be viewed

as a form of tolerance, this interpretation conflicts with both the-oretical expectations and our natural data. Tolerance is expected toincrease infection prevalence within populations and to alleviateantagonistic host–parasite interactions (50). However, we find thatinfection prevalence is lowest in populations with small parasites.

A B

Fig. 4. Cestode mass is correlated with infection abundance in the wild, butthe effect direction is reversed when comparing individual fish within eachof many populations (A) versus averages across populations (B). In A, pointsrepresent individuals within each of many populations. Within-populationtrend lines are shown in black, and the consensus effect is shown in blue(from a mixed-model GLM). In B, we plot population mean cestode abun-dance versus mean cestode mass with lakes as the level of replication.

6578 | www.pnas.org/cgi/doi/10.1073/pnas.1620095114 Weber et al.

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

0

Moreover, size reduction is almost certainly costly to the parasiteand is likely to have ancillary effects on parasite epidemiology.Smaller cestodes should be less fecund (33) and also may be lesseffective at manipulating host behavior in ways that may facili-tate transfer to the terminal bird host (51, 52). Last, suppressingcestode growth may help the host clear the parasite infection.We occasionally observed small S. solidus encased in cysts (Fig.S7) within ROB (but not GOS) hosts. Some of these cysts con-tained degraded cestodes. We infer that growth suppression mayaid in cyst formation (smaller parasites being easier to envelop),which in turn facilitates cestode killing.We also found heritable between-population differences in

constitutive and infection-induced immune traits (granulocyteROS production and granulocyte frequency, respectively). Bothtraits have been linked to macroparasite killing by vertebrates(9, 24–28). However, we were unable to find reports of theirinfluence on macroparasite growth, and our study is not designedto test this association formally. The infection-associated in-crease in HK granulocytes in ROB and R×G fish is consistentwith the eosinophilia observed with helminthiasis in other ver-tebrate species (53). However, two observations suggest this in-crease is insufficient to explain variation in cestode growth. First,R×G hybrids had larger cestodes than ROB fish despite exhib-iting comparable granulocyte abundance. Second, R×G hybridshad slightly larger parasites than G×R hybrids, despite the for-mer’s stronger granulocyte response. Although granulocyte ROSlevels negatively correlated with cestode size (Figs. 2B and 3C),this trend is confounded with overall population effects and doesnot hold among families within populations. The interactionsbetween granulocyte-generated ROS and cestode growth war-rant further investigation.Although the proximate cause of cestode growth suppression

remains unknown, our results offer several insights into stickle-back immune evolution. First, even though GOS and ROB fishevolved to resist initial establishment by S. solidus equally, otherimmune traits exhibited nonparallel evolution. Relatively little isknown about the extent of parallel (or nonparallel) evolution ofimmune phenotypes across evolutionary replicate populations, incontrast to more easily measured morphological traits. We sug-gest that parallel and nonparallel changes in immunity will proveto be useful in studying the context dependency of immuneevolution and redundancy in immune function.Second, we found that ROS production (both pre- and post-

PMA stimulation) was largely unresponsive to S. solidus infection.This result is contrary to several previous studies, which are them-selves conflicting. One study suggested that ROS response increasesin early-stage S. solidus infections but is eventually suppressed(27), and another suggested that granulocyte proportions fluc-tuate throughout infection but ROS levels increase in only late-stage infection (54). These different results may be a function ofthe particular populations studied (here Canadian populationsversus the European stickleback and parasites used in priorstudies) or of the variation in postexposure time points of im-mune assays. Given the differences in granulocyte response be-tween geographically nearby lakes (Gosling and Roberts Lakes),it is quite plausible that host–parasite interactions on differentcontinents may follow very different kinetics. Differences in ROSmeasurement and in vitro assays may also play a role. For ex-ample, infection-associated ROS differences may be hard todetect if only a small subset of cells contacts or indirectly re-sponds to parasite presence. This possibility could be addressedby experimentally exposing cell cultures to parasite antigensbefore assaying ROS (55).Finally, the maternally inherited increase in relative gran-

ulocyte abundance is an intriguing response that warrants addi-tional attention. ROB but not GOS fish initiated a strong shifttoward granulocytes following S. solidus infection, and F1 hy-brids matched the response of their maternal parent. A recentstickleback study also found that establishment of S. solidusseems to depend on an interaction between maternal effect andparasite origin (21). However, we found no significant maternal

effects on cestode success in our study. Many immunologicalmaternal effects in birds, fish, and mammals are related to antibodytransmission between mother and offspring (56). These antibody-based effects tend to be relatively short-lived. Because fish in thepresent study were not exposed to S. solidus until adulthood, it ispossible that a different mechanism may be involved.Debate continues about the heritability of human immune

responses (57, 58), but our results show that genetic variation forimmunity segregates both within and among wild populationsand that immune differences influence parasite infection. Iden-tifying the underlying mechanisms of variation in natural in-fection will provide important contributions to this debate.Natural selection in disparate wild populations provides a pow-erful genetic scan to locate adaptive genetic variation involved inparasite resistance and immune function (59).

Materials and MethodsEstimating Natural Infection Prevalence. We used unbaited minnow traps tosample threespine stickleback (Scientific Fish Collection permit NA12-77018 and NA12-84188) (Table S1). The majority of sites were sampled oncein 2009 or in 2013, but some were sampled repeatedly over multiple years(Table S1). All animals were killed by immersion in MS-222, fixed in 10%neutral-buffered formalin, and then dissected to evaluate the presence ofS. solidus.

Breeding Wild-Caught Stickleback. We generated crosses between wild-caught anadromous marine stickleback from Sayward Estuary to generatepure marine families (M genotype). We also generated one M×R cross andone G×M cross. All crosses were performed in June 2012, and fertilized eggswere sent to The University of Texas at Austin (Transfer License NA12-76852), where all animals were reared in freshwater conditions andhoused with full-siblings (9–39 animals per family) in either 40-L or 10-Ltanks. This experiment was approved by The University of Texas InstitutionalAnimal Care and Use Committee as part of AUP-2010-00024.

S. solidus Collection and Experimental Exposures. S. solidus were harvestedfrom stickleback captured at three sites in British Columbia, Canada (GoslingLake, Echo Lake, and Jericho Pond). The parasites were bred to obtain eggsfrom full-sibling families, using medium and methods described in refs. 18and 60. Juvenile copepods (Macrocyclops albidus) were exposed to one ortwo freshly hatched tapeworm coracidia in 24-well plates. Two weeks afterexposure we screened copepods to identify singly and doubly infected in-dividuals, which were fed to fish between 21–26 d postexposure [whenS. solidus coracidia are developmentally capable of infecting stickleback(61)]. Similarly aged uninfected copepods were used for control exposures.

Stickleback exposures began by isolating five to seven fish per family andclipping either the first or second dorsal spine to distinguish control (one ortwo fish per family) from experimental animals. At least 12 d after clipping,we moved individual fish into 1-L containers and withheld food for 24 h.Experimental and control fish were fed four or five infected or uninfectedcopepods, respectively. The following day, after confirming that all copepodswere consumed, we combined control and experimental animals from eachfamily into a single tank. We used one tank per family. At 41–51 d post-exposure (41–44 d for all but two families), we dissected fish, recorded fishmetrics (length, mass, sex), noted the presence or absence of tapeworminfection, and performed immune assays. All fish exposed to S. solidus werebetween 15 and 23 mo of age.

Measurement of Immune Traits via Flow Cytometry. We dissected HKs fromcontrol and experimental fish and immediately placed organs in cold HKmedium [0.9× Roswell Park Memorial Institute (RPMI) medium containing10% FCS, 100 μM nonessential amino acids (NEAA), 100 U/mL penicillin,100 μg/mL streptomycin, and 55 μM β-mercaptoethanol]. HK were manuallydisrupted and filtered by grinding against a cell strainer (35-μmmesh; Falcon352235) and rinsing with 4 mL of cold HK medium. Cell suspensions werecentrifuged at 300 × g, 4 °C for 10 min; then the supernatant was removed,and the cell pellet was resuspended in the remaining medium (∼200 μL). LiveHK cells were counted using a hemacytometer (Hausser Scientific 3520)based on Trypan blue exclusion (Corning 25-900-CI). For each fish, HK cellswere divided into three treatment groups: (i) medium-only control, (ii) DHR-123–stained, and (iii) DHR-123–stained and PMA-stimulated.

For each treatment group 2 × 105 HK cells were plated in 200 μL of HKmedium (group 1) or HK medium containing DHR-123 (2 μg/mL) (SigmaD1054) (groups 2 and 3). All samples were incubated in a 96-well plate at

Weber et al. PNAS | June 20, 2017 | vol. 114 | no. 25 | 6579

EVOLU

TION

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

0

18 °C, 3% CO2 for 10 min. Next, PMA (130 ng/mL; Sigma P8139) was added togroup 3, and groups 1 and 2 received equivalent volumes of plain HK me-dium. After cells were incubated for an additional 20 min at 18 °C, 3% CO2,all samples were run on an Accuri C6 Sampler and analyzed using FlowJo(TreeStar). Granulocyte:lymphocyte ratios were calculated from the medium-only controls (Fig. S8). The magnitude of ROS production was determined bycomparing the median fluorescence of unstimulated and PMA-stimulated cells.Fig. S8 contains additional information on gating strategies.

Statistical Analyses. All statistical tests were performed in the R softwareenvironment (V3.3-1) (62). For models without random effects, we performed

Tukey’s honestly significant difference (HSD)–corrected post hoc compari-sons using the multcomp package (V1.4-5) (63). We used the lme4 package(V1.1-12) (64) to construct random effect models and used maximum likeli-hood to calculate AIC scores for each model (Table S1). We used the lsmeanspackage (V2.23-5) (65) to perform Tukey’s HSD-corrected post hoc compar-isons on mixed models.

ACKNOWLEDGMENTS. We thank L. Ma for assistance with laboratory work.Wenfei Tong produced the illustrations in Fig. S1. Douglas Emlen providedcomments on earlier drafts. This research was funded by the Howard HughesMedical Institute.

1. World Health Organization (2017) Fact sheet: Soil-transmitted helminth infections.Available at www.who.int/mediacentre/factsheets/fs366/en/. Accessed May 23, 2017.

2. Charlier J, et al. (2016) Decision making on helminths in cattle: Diagnostics, economicsand human behaviour. Ir Vet J 69:14.

3. Perry BD, Randolph TF (1999) Improving the assessment of the economic impact ofparasitic diseases and of their control in production animals. Vet Parasitol 84:145–168.

4. Thompson RC, Lymbery AJ, Smith A (2010) Parasites, emerging disease and wildlifeconservation. Int J Parasitol 40:1163–1170.

5. Fumagalli M, et al. (2010) The landscape of human genes involved in the immuneresponse to parasitic worms. BMC Evol Biol 10:264.

6. Fumagalli M, et al. (2009) Parasites represent a major selective force for interleukin genesand shape the genetic predisposition to autoimmune conditions. J ExpMed 206:1395–1408.

7. Hayward AD, et al. (2014) Natural selection on individual variation in tolerance ofgastrointestinal nematode infection. PLoS Biol 12:e1001917.

8. King KC, Lively CM (2012) Does genetic diversity limit disease spread in natural hostpopulations? Heredity (Edinb) 109:199–203.

9. Ashley NT, Weil ZM, Nelson RJ (2012) Inflammation: Mechanisms, costs, and naturalvariation. Annu Rev Ecol Evol Syst 43:385–406.

10. Auld SK, et al. (2013) Variation in costs of parasite resistance among natural hostpopulations. J Evol Biol 26:2479–2486.

11. Sandland GJ, Minchella DJ (2003) Costs of immune defense: An enigma wrapped in anenvironmental cloak? Trends Parasitol 19:571–574.

12. Sorci G, Faivre B (2009) Inflammation and oxidative stress in vertebrate host-parasitesystems. Philos Trans R Soc Lond B Biol Sci 364:71–83.

13. Maizels RM, et al. (2004) Helminth parasites–masters of regulation. Immunol Rev 201:89–116.14. Johnson PTJ, de Roode JC, Fenton A (2015) Why infectious disease research needs

community ecology. Science 349:1259504.15. Housden BE, et al. (2017) Loss-of-function genetic tools for animal models: Cross-

species and cross-platform differences. Nat Rev Genet 18:24–40.16. Appleby MW, Ramsdell F (2003) A forward-genetic approach for analysis of the im-

mune system. Nat Rev Immunol 3:463–471.17. Simmonds NE, Barber I (2016) The effect of salinity on egg development and viability

of Schistocephalus solidus (Cestoda: Diphyllobothriidea). J Parasitol 102:42–46.18. Weber JN, et al. (2017) Resist globally, infect locally: A transcontinental test of ad-

aptation by stickleback and their tapeworm parasite. Am Nat 189:43–57.19. Bagamian KH, Heins DC, Baker JA (2004) Body condition and reproductive capacity of

three-spined stickleback infected with the cestode Schistocephalus solidus. J Fish Biol64:1568–1576.

20. Heins DC, Baker JA (2014) Fecundity compensation and fecundity reduction amongpopulations of the three-spined stickleback infected by Schistocephalus solidus inAlaska. Parasitology 141:1088–1096.

21. Kalbe M, Eizaguirre C, Scharsack JP, Jakobsen PJ (2016) Reciprocal cross infection of stick-lebacks with the diphyllobothriidean cestode Schistocephalus solidus reveals consistentpopulation differences in parasite growth and host resistance. Parasit Vectors 9:130.

22. Kurtz J, et al. (2004) Major histocompatibility complex diversity influences parasiteresistance and innate immunity in sticklebacks. Proc Biol Sci 271:197–204.

23. Snowberg LK, Hendrix KM, Bolnick DI (2015) Covarying variances: More morpho-logically variable populations also exhibit more diet variation. Oecologia 178:89–101.

24. Allen JE, Maizels RM (2011) Diversity and dialogue in immunity to helminths. Nat RevImmunol 11:375–388.

25. Maizels RM, Hewitson JP, Smith KA (2012) Susceptibility and immunity to helminthparasites. Curr Opin Immunol 24:459–466.

26. Moreau E, Chauvin A (2010) Immunity against helminths: Interactions with the hostand the intercurrent infections. J Biomed Biotechnol 2010:428593.

27. Scharsack JP, Kalbe M, Derner R, Kurtz J, Milinski M (2004) Modulation of granulocyteresponses in three-spined sticklebacks Gasterosteus aculeatus infected with thetapeworm Schistocephalus solidus. Dis Aquat Organ 59:141–150.

28. Alvarez-Pellitero P (2008) Fish immunity and parasite infections: From innate immu-nity to immunoprophylactic prospects. Vet Immunol Immunopathol 126:171–198.

29. Kaufmann J, Lenz TL, Milinski M, Eizaguirre C (2014) Experimental parasite infectionreveals costs and benefits of paternal effects. Ecol Lett 17:1409–1417.

30. Heins DC, Baker JA, Martin HC (2002) The “crowding effect” in the cestodeSchistocephalus solidus: Density-dependent effects on plerocercoid size and in-fectivity. J Parasitol 88:302–307.

31. Read CP (1951) The “crowding effect” in tapeworm infections. J Parasitol 37:174–178.32. Hopkins CA, McCAIG MLO (1963) Studies on Schistocephalus solidus. I. The correlation of de-

velopment in the plerocercoid with infectivity to the definitive host. Exp Parasitol 13:235–243.33. Tierney JF, Crompton DWT (1992) Infectivity of plerocercoids of Schistocephalus sol-

idus (Cestoda: Ligulidae) and fecundity of the adults in an experimental definitivehost, Gallus gallus. J Parasitol 78:1049–1054.

34. Dawkins R (1999) The Extended Phenotype: The Long Reach of the Gene (Oxford UnivPress, New York) Revised edition, p 313.

35. Colosimo PF, et al. (2005) Widespread parallel evolution in sticklebacks by repeatedfixation of Ectodysplasin alleles. Science 307:1928–1933.

36. Jones FC, et al.; Broad Institute Genome Sequencing Platform and Whole GenomeAssembly Team (2012) The genomic basis of adaptive evolution in threespine stick-lebacks. Nature 484:55–61.

37. Viney ME, Riley EM, Buchanan KL (2005) Optimal immune responses: Immunocom-petence revisited. Trends Ecol Evol 20:665–669.

38. Bell AM, Sih A (2007) Exposure to predation generates personality in threespinedsticklebacks (Gasterosteus aculeatus). Ecol Lett 10:828–834.

39. Ingram T, et al. (2012) Intraguild predation drives evolutionary niche shift inthreespine stickleback. Evolution 66:1819–1832.

40. Jolles JW, Manica A, Boogert NJ (2016) Food intake rates of inactive fish are positively linkedto boldness in three-spined sticklebacks Gasterosteus aculeatus. J Fish Biol 88:1661–1668.

41. Svanbäck R, Bolnick DI (2007) Intraspecific competition drives increased resource usediversity within a natural population. Proc Biol Sci 274:839–844.

42. Bishop SC, Morris CA (2007) Genetics of disease resistance in sheep and goats. SmallRumin Res 70:48–59.

43. Stear MJ, et al. (1997) How hosts control worms. Nature 389:27.44. Stear MJ, Boag B, Cattadori I, Murphy L (2009) Genetic variation in resistance to

mixed, predominantly Teladorsagia circumcincta nematode infections of sheep: Fromheritabilities to gene identification. Parasite Immunol 31:274–282.

45. Quinnell RJ (2003) Genetics of susceptibility to human helminth infection. Int JParasitol 33:1219–1231.

46. Quinnell RJ, Griffin J, Nowell MA, Raiko A, Pritchard DI (2001) Predisposition tohookworm infection in Papua New Guinea. Trans R Soc Trop Med Hyg 95:139–142.

47. Bishop SC (2012) A consideration of resistance and tolerance for ruminant nematodeinfections. Front Genet 3:168.

48. Bishop SC, Stear MJ (2003) Modeling of host genetics and resistance to infectious diseases:Understanding and controlling nematode infections. Vet Parasitol 115:147–166.

49. Barber I, Huntingford FA (1995) The effect of Schistocephalus solidus (Cestoda:Pseudophyllidea) on the foraging and shoaling behaviour of three-spined stickle-backs, Gasterosteus aculeatus. Behaviour 132:1223–1240.

50. Best A, White A, Boots M (2014) The coevolutionary implications of host tolerance.Evolution 68:1426–1435.

51. Giles N (1983) Behavioural effects of the parasite Schistocephalus solidus (Cestoda) onan intermediate host, the three-spined stickleback, Gasterosteus aculeatus L. AnimBehav 31:1192–1194.

52. Milinski M (1985) Risk of predation of parasitized sticklebacks (Gasterosteus aculeatusL.) under competition for food. Behaviour 93:203–216.

53. Shin MH, Lee YA, Min D-Y (2009) Eosinophil-mediated tissue inflammatory responsesin helminth infection. Korean J Parasitol 47:S125–S131.

54. Scharsack JP, Koch K, Hammerschmidt K (2007) Who is in control of the sticklebackimmune system: Interactions between Schistocephalus solidus and its specific verte-brate host. Proc Biol Sci 274:3151–3158.

55. Franke F, et al. (2014) In vitro leukocyte response of three-spined sticklebacks (Gasterosteusaculeatus) to helminth parasite antigens. Fish Shellfish Immunol 36:130–140.

56. Grindstaff JL, Brodie ED, 3rd, Ketterson ED (2003) Immune function across genera-tions: Integrating mechanism and evolutionary process in maternal antibody trans-mission. Proc Biol Sci 270:2309–2319.

57. Brodin P, et al. (2015) Variation in the human immune system is largely driven by non-heritable influences. Cell 160:37–47.

58. Li Y, et al. (2016) Inter-individual variability and genetic influences on cytokine re-sponses to bacteria and fungi. Nat Med 22:952–960.

59. Downs CJ, Adelman JS, Demas GE (2014) Mechanisms and methods in ecoimmunology: In-tegrating within-organism and between-organism processes. Integr Comp Biol 54:340–352.

60. Jakobsen PJ, et al. (2012) In vitro transition of Schistocephalus solidus (Cestoda) from cor-acidium to procercoid and from procercoid to plerocercoid. Exp Parasitol 130:267–273.

61. Benesh DP, Hafer N (2012) Growth and ontogeny of the tapeworm Schistocephalussolidus in its copepod first host affects performance in its stickleback second in-termediate host. Parasit Vectors 5:90.

62. R Core Team (2016) R: A Language and Environment for Statistical Computing (RFoundation for Statistical Computing, Vienna).

63. Hothorn T, Bretz F, Westfall P (2008) Simultaneous inference in general parametricmodels. Biom J 50:346–363.

64. Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects modelsusing lme4. J Stat Softw 67:1–48.

65. Lenth RV (2016) Least-squares means: The R package lsmeans. J Stat Softw 69:1–33.

6580 | www.pnas.org/cgi/doi/10.1073/pnas.1620095114 Weber et al.

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

0