Daphnia Magna

Received: 23 May 2019 | Revised: 17 September 2019 | Accepted: 27 September 2019 DOI: 10.1111/mec.15260

ORIGINAL ARTICLE

11. Hernadi, E., Rohaeti, E., Rafi, M., Wahyuni, W. T., Putri, S. P., and Fukusaki, E. (2019). HPLC fingerprinting coupled with linear discriminant analysis for the detection of adulteration in Orthosiphon aristatus. Journal of Liquid Chromatography & Related Parasite‐mediated selection in a natural metapopulation of Technologies 42: 513-520. Daphnia magna

12. Wisman, A. P., Tamada, Y., Hirohata, S., Gomi, K., Fukusaki, E., and Shimma, S. (2020). Mapping haze-komi on rice koji grains using β-glucuronidase expressing Andrea P. Cabalzar1 | Peter D. Fields1 | Yasuhiko Kato2 | Hajime Watanabe2 | Aspergillus oryzae and mass spectrometry imaging. Journal of Bioscience and 1,3 Bioengineering 129: 296-301. Dieter Ebert

13. Heyman, B., Tulke, H., Putri, S. P., Fukusaki, E., and Büchs, J. (2020). Online 1Department of Environmental Sciences, monitoring of the respiratory quotient reveals metabolic phases during microaerobic Zoology, University of Basel, Basel, Abstract Switzerland Parasite‐mediated selection varying across time and space in metapopulations is ex‐ 2,3‐butanediol production with Bacillus licheniformis. Engineering in Life Sciences 2 Department of Biotechnology, pected to result in host local adaptation and the maintenance of genetic diversity in 20: 133-144. Division of Advance Science and Biotechnology, Graduate School of disease‐related traits. However, nonadaptive processes like migration and extinction‐ 14. Umakoshi, Y., Nakano, Y., Fukuda, K., Watanabe, K., Miyawaki, I., and Fukusaki, E. Engineering, Osaka University, Suita, Japan (re)colonization dynamics might interfere with adaptive evolution. Understanding 3 (2019). Automatic switching valve system to minimize variation of liquid Tvärminne Zoological Station, Tvärminne, Finland how adaptive and nonadaptive processes interact to shape genetic variability in chromatography-tandem mass spectrometry-based chiral amino acid profiling. life‐history and disease‐related traits can provide important insights into their evolu‐ Journal of Bioscience and Bioengineering 128: 773-779. Correspondence Andrea P. Cabalzar, Department of tion in subdivided populations. Here we investigate signatures of spatially fluctuat‐ Environmental Sciences, Zoology, University ing, parasite‐mediated selection in a natural metapopulation of Daphnia magna. Host 15. Septiana, S., Yuliana, N. D., Bachtiar, B. M., Putri, S. P., Fukusaki, E., Laviña, W. A., of Basel, Basel, Switzerland. genotypes from infected and uninfected populations were genotyped at microsatel‐ and Wijaya, C. H. (2020). Metabolomics approach for determining potential Email: [email protected] lite markers, and phenotyped for life‐history and disease traits in common garden metabolites correlated with sensory attributes of Melaleuca cajuputi essential oil, a Funding information experiments. Combining phenotypic and genotypic data a Q –F ‐like analysis was promising flavor ingredient. Journal of Bioscience and Bioengineering 129: 581-587. Schweizerischer Nationalfonds zur ST ST Förderung der Wissenschaftlichen conducted to test for signatures of parasite mediated selection. We observed high Forschung, Grant/Award Number: 16. Claes, B. S. R., Takeo, E., Fukusaki, E., Shimma, S., Ron M. A., and Heeren, R. M. 310030B_166677 variation within and among populations for phenotypic traits, but neither an indica‐ A. (2019). Imaging isomers on a biological surface: A review. Mass Spectrometry 8: tion of host local adaptation nor a cost of resistance. Infected populations have a A0078. higher gene diversity (Hs) than uninfected populations and Hs is strongly positively correlated with fitness. These results suggest a strong parasite effect on reducing 17. Nakano, Y., Taniguchi, M., Umakoshi, Y., Watai, D., and Fukusaki, E. (2019). High- population level inbreeding. We discuss how stochastic processes related to frequent Throughput LC-MS/MS Method for Chiral Amino Acid Analysis Without extinction‐(re)colonization dynamics as well as host and parasite migration impede Derivatization. Methods Mol. Biol. 2019. the evolution of resistance in the infected populations. We suggest that the genetic

and phenotypic patterns of variation are a product of dynamic changes in the host gene pool caused by the interaction of colonization bottlenecks, inbreeding, immi‐ gration, hybrid vigor, rare host genotype advantage and parasitism. Our study high‐ lights the effect of the parasite in ameliorating the negative fitness consequences caused by the high drift load in this metapopulation.

KEYWORDS

Daphnia, local adaptation, metapopulation, Microsporidia, neutral evolution, QST–FST analysis, resistance

90 91 CABALZAR et AL. | 4771

1 | INTRODUCTION Kawecki, 2004). In addition, variation in parasite presence or abun‐ dance can be quantified and is ubiquitous in natural populations, Environmental heterogeneity among subpopulations of the same even at small geographic scales (Goren & Ben‐Ami, 2013; Laine, species may cause divergent selection leading to patterns of local 2006; Roy & St‐Louis, 2017). In parasite‐infected populations, hosts adaptation. The ability of populations to adapt locally to prevailing are expected to evolve defence mechanisms (resistance) to avoid, environmental conditions depends on a complex interplay between reduce or tolerate parasite infection over time. However, the evo‐ directional selection, migration, and random genetic drift (Blanquart, lution of resistance is thought to come with a cost, as immune re‐ Gandon, & Nuismer, 2012; Lenormand, 2002; Pfeifer et al., 2018; sponses are energetically expensive (Carton, Nappi, & Poirie, 2005). Yeaman & Otto, 2011). Indeed, geographic scale can greatly in‐ Thus, resistance only confers a fitness advantage in the presence fluence the potential of local adaptation. Migration rate often de‐ of infection (Little & Killick, 2007; Webster & Woolhouse, 1999; creases as a function of geographic distance; thus, local adaptation Wielgoss, Bergmiller, Bischofberger, & Hall, 2016), and selection for is more expected at larger geographic scales where the distances resistance may vary over space, depending on the local risk of in‐ between local populations exceed the species' ability to disperse, fection. This would promote local adaptation at a microgeographic although fine‐scale evolution is increasingly appreciated as a source scale (Kawecki & Ebert, 2004; Lazzaro & Little, 2009), with selec‐ of phenotypic differentiation (Hämälä, Mattila, & Savolainen, 2018; tion favouring more resistant genotypes in infected populations but Hendrick et al., 2016; Moody et al., 2015; for review see Richardson, disfavouring them in uninfected populations. Urban, Bolnick, & Skelly, 2014). At smaller geographic scales, where Studies have also shown that parasites can generate and main‐ local populations are more strongly connected by migration, the tain genetic diversity in host populations, either by promoting sexual outcome of heterogeneous selection pressures is more complex reproduction via negative frequency dependent selection (Jokela, (Blanquart, Kaltz, Nuismer, & Gandon, 2013; Kawecki & Ebert, 2004; Dybdahl, & Lively, 2009), by heterozygote advantage of resistance Richardson et al., 2014). alleles (Penn, Damjanovich, & Potts, 2002), or by enhancing selec‐ In a metapopulation, local populations are subject not only to tion against genome‐wide homozygosity of recessive, deleterious gene flow but also to high population turnover. Extinction‐(re)col‐ mutations (inbreeding depression; Coltman, Pilkington, Smith, & onization dynamics often cause strong founder effects, resulting in Pemberton, 1999). This latter case, where parasites are predicted to small effective subpopulation sizes with a high sensitivity to genetic select against homozygotes, particularly affects small populations, drift and inbreeding (De Kort, Vandepitte, & Honnay, 2013; Hanski where inbreeding is high and genetic drift can lead to the random fix‐ & Gaggiotti, 2004; Lande, 1976; Slatkin, 1977; Wade & McCauley, ation of deleterious mutations (Coltman et al., 1999; Reid, Arcese, & 1988). In these cases, demographic and stochastic processes, rather Keller, 2003). While the link between inbreeding and parasite suscep‐ than adaptive evolution, might cause the spatial differentiation in tibility is well explored (King & Lively, 2012), only a few studies actually functional traits. For example, migration and genetic drift may in‐ quantify the link between infection, inbreeding, and fitness in natural terfere with adaptive divergence by either homogenizing local allele populations (but see: Coltman et al., 1999; Keller & Waller, 2002 for frequencies or by eroding genetic variation, respectively (Blanquart a review). Combining population genetic data with fitness estimates et al., 2012; Gandon, Capowiez, Dubois, Michalakis, & Olivieri, 1996; in the presence and absence of parasites facilitates insight into how Lenormand, 2002). Given this complex interplay of demographic, parasites maintain genetic variation in immune related traits as well as ecological and evolutionary processes, the study of fine‐scale genome‐wide heterozygosity in spatially structured populations. local adaptation must be an integrative one, combining population This type of study must be done in metapopulations, where and quantitative genetic approaches along with experiments. One we can define distinct subpopulations connected by gene flow method that accomplishes this is the QST–FST comparison (Merilä & and quantify the environmental variation between populations. As Crnokrak, 2001; Spitze, 1993), which compares neutral divergence pond‐dwelling organisms offer ideal systems to study evolution at

(FST, null expectation; Wright, 1951) with the divergence of a phe‐ a microgeographic scale (they have clearly‐defined populations, and notypic (quantitative) trait of interest (QST; Blanquart et al., 2013; environmental factors can vary markedly between ponds, even at Leinonen, McCairns, O'Hara, & Merilä, 2013). Traits which will be small spatial scales), we chose for this study the Holarctic‐distrib‐ the focus of a QST–FST comparison are best measured in a common uted, cyclically parthenogenetic crustacean Daphnia magna and environment (common garden experiment), so that environmental its microsporidian parasite Hamiltosporidium tvaerminnensis (Haag, influences can be standardized and genetic and environmental con‐ Larsson, Refardt, & Ebert, 2011). Quantitative variation in resistance tributions to the phenotype can be distinguished. Such an approach is common in this system and has been shown to have a genetic basis requires knowing the environmental factor that acts as a selective (Ebert, 2008; Krebs, Routtu, & Ebert, 2017; Routtu & Ebert, 2015). agent and being able to experimentally manipulate it. Evidence from experimental evolution in both natural and semi‐nat‐ Host‐parasite systems are ideal for investigating local adaptation ural conditions indicates that H. tvaerminnensis has the potential to at the metapopulation level because parasites exert strong selec‐ drive rapid changes in the genetic composition of host populations tion on their hosts (Barber & Dingemanse, 2010; Cornetti, Hilfiker, (Haag & Ebert, 2004; Zbinden, Haag, & Ebert, 2008) and that para‐ Lemoine, & Tschirren, 2018; Haldane, 1949), which can in turn site‐mediated selection leads to rapid host adaptation to the parasite lead to local adaptation (Karvonen & Seehausen, 2012; Spichtig & (Zbinden et al., 2008).

92 93 CABALZAR et AL. | 4771 4772 | CABALZAR et AL.

1 | INTRODUCTION Kawecki, 2004). In addition, variation in parasite presence or abun‐ FIGURE 1 Geographic location of dance can be quantified and is ubiquitous in natural populations, sampled rock pool populations used in this study. The age of the population Environmental heterogeneity among subpopulations of the same even at small geographic scales (Goren & Ben‐Ami, 2013; Laine, is given in years in brackets beside the species may cause divergent selection leading to patterns of local 2006; Roy & St‐Louis, 2017). In parasite‐infected populations, hosts population name. The inset map shows adaptation. The ability of populations to adapt locally to prevailing are expected to evolve defence mechanisms (resistance) to avoid, Finland with the location of the study site environmental conditions depends on a complex interplay between reduce or tolerate parasite infection over time. However, the evo‐ indicated [Colour figure can be viewed at directional selection, migration, and random genetic drift (Blanquart, lution of resistance is thought to come with a cost, as immune re‐ wileyonlinelibrary.com] Gandon, & Nuismer, 2012; Lenormand, 2002; Pfeifer et al., 2018; sponses are energetically expensive (Carton, Nappi, & Poirie, 2005). Yeaman & Otto, 2011). Indeed, geographic scale can greatly in‐ Thus, resistance only confers a fitness advantage in the presence fluence the potential of local adaptation. Migration rate often de‐ of infection (Little & Killick, 2007; Webster & Woolhouse, 1999; creases as a function of geographic distance; thus, local adaptation Wielgoss, Bergmiller, Bischofberger, & Hall, 2016), and selection for is more expected at larger geographic scales where the distances resistance may vary over space, depending on the local risk of in‐ between local populations exceed the species' ability to disperse, fection. This would promote local adaptation at a microgeographic although fine‐scale evolution is increasingly appreciated as a source scale (Kawecki & Ebert, 2004; Lazzaro & Little, 2009), with selec‐ of phenotypic differentiation (Hämälä, Mattila, & Savolainen, 2018; tion favouring more resistant genotypes in infected populations but Hendrick et al., 2016; Moody et al., 2015; for review see Richardson, disfavouring them in uninfected populations. Urban, Bolnick, & Skelly, 2014). At smaller geographic scales, where Studies have also shown that parasites can generate and main‐ local populations are more strongly connected by migration, the tain genetic diversity in host populations, either by promoting sexual outcome of heterogeneous selection pressures is more complex reproduction via negative frequency dependent selection (Jokela, (Blanquart, Kaltz, Nuismer, & Gandon, 2013; Kawecki & Ebert, 2004; Dybdahl, & Lively, 2009), by heterozygote advantage of resistance Richardson et al., 2014). alleles (Penn, Damjanovich, & Potts, 2002), or by enhancing selec‐ In a metapopulation, local populations are subject not only to tion against genome‐wide homozygosity of recessive, deleterious gene flow but also to high population turnover. Extinction‐(re)col‐ mutations (inbreeding depression; Coltman, Pilkington, Smith, & onization dynamics often cause strong founder effects, resulting in Pemberton, 1999). This latter case, where parasites are predicted to small effective subpopulation sizes with a high sensitivity to genetic select against homozygotes, particularly affects small populations, drift and inbreeding (De Kort, Vandepitte, & Honnay, 2013; Hanski where inbreeding is high and genetic drift can lead to the random fix‐ & Gaggiotti, 2004; Lande, 1976; Slatkin, 1977; Wade & McCauley, ation of deleterious mutations (Coltman et al., 1999; Reid, Arcese, & We investigated whether spatial variation in parasite‐mediated se‐ 23°15′E). This metapopulation has been the subject of biannual 1988). In these cases, demographic and stochastic processes, rather Keller, 2003). While the link between inbreeding and parasite suscep‐ lection explains spatial variation in host resistance and life‐history traits censuses for 33 years, providing data on population age, popu‐ than adaptive evolution, might cause the spatial differentiation in tibility is well explored (King & Lively, 2012), only a few studies actually in a natural metapopulation of D. magna and its parasite H. tvaermin‐ lation extinction, and recolonization events (Pajunen & Pajunen, functional traits. For example, migration and genetic drift may in‐ quantify the link between infection, inbreeding, and fitness in natural nensis. Using 77 host genotypes obtained from six infected and five un‐ 2003), as well as on the presence and absence of different parasite terfere with adaptive divergence by either homogenizing local allele populations (but see: Coltman et al., 1999; Keller & Waller, 2002 for infected Daphnia rock pool populations, we quantified host resistance species (since 2009). The microsporidian parasite Hamiltosporidium frequencies or by eroding genetic variation, respectively (Blanquart a review). Combining population genetic data with fitness estimates and estimated host fitness in the presence and absence of the parasite. tvaerminnensis (previously Octosporea bayerii; Corradi, Haag, et al., 2012; Gandon, Capowiez, Dubois, Michalakis, & Olivieri, 1996; in the presence and absence of parasites facilitates insight into how Hosts were genotyped at 24 microsatellite markers to estimate popu‐ Pombert, Ebert, & Keeling, 2009; Haag et al., 2011) is the most Lenormand, 2002). Given this complex interplay of demographic, parasites maintain genetic variation in immune related traits as well as lation heterozygosity and neutral genetic divergence, and in order to prevalent parasite in the Finnish D. magna metapopulation, infect‐ ecological and evolutionary processes, the study of fine‐scale genome‐wide heterozygosity in spatially structured populations. conduct a QST–FST‐like comparison (Karhunen, Merilä, Leinonen, Cano, ing about half of all populations. H. tvaerminnensis transmits hori‐ local adaptation must be an integrative one, combining population This type of study must be done in metapopulations, where & Ovaskainen, 2013; Karhunen, Ovaskainen, Herczeg, & Merilä, 2014; zontally from dead individuals, as well as vertically from mother to and quantitative genetic approaches along with experiments. One we can define distinct subpopulations connected by gene flow Ovaskainen, Karhunen, Zheng, Arias, & Merilä, 2011). We sought to parthenogenetic and sexual host offspring; therefore, the parasite method that accomplishes this is the Q –F comparison (Merilä & and quantify the environmental variation between populations. As ST ST address four specific questions: (a) Are individuals from infected pop‐ reaches prevalence as high as 100% within a population by the end Crnokrak, 2001; Spitze, 1993), which compares neutral divergence pond‐dwelling organisms offer ideal systems to study evolution at ulations on average more resistant to the parasite H. tvaerminnensis of the season (Lass & Ebert, 2006). By transmitting vertically to (FST, null expectation; Wright, 1951) with the divergence of a phe‐ a microgeographic scale (they have clearly‐defined populations, and than individuals from uninfected populations? (b) Does the QST–FST‐ the host's resting eggs, H. tvaerminnensis is able to disperse along notypic (quantitative) trait of interest (Q ; Blanquart et al., 2013; environmental factors can vary markedly between ponds, even at ST like analysis show signatures of adaptive divergence between naturally with its host to new populations and endure harsh environmental Leinonen, McCairns, O'Hara, & Merilä, 2013). Traits which will be small spatial scales), we chose for this study the Holarctic‐distrib‐ infected and uninfected populations? (c) Is resistance to H. tvaermin‐ conditions such as summer droughts and winter freezing (Vizoso, the focus of a Q –F comparison are best measured in a common uted, cyclically parthenogenetic crustacean Daphnia magna and ST ST nensis costly? (d) Does host fitness correlate with within‐population Lass, & Ebert, 2005). environment (common garden experiment), so that environmental its microsporidian parasite Hamiltosporidium tvaerminnensis (Haag, gene diversity? We collected female D. magna from 11 different populations influences can be standardized and genetic and environmental con‐ Larsson, Refardt, & Ebert, 2011). Quantitative variation in resistance (G‐13, G‐2, K‐10, LA‐29, M‐64, N‐27, N‐41, N‐62, SK‐1, SK‐16, SK‐58; tributions to the phenotype can be distinguished. Such an approach is common in this system and has been shown to have a genetic basis Figure 1). Of these, five were parasite free and six had been infected requires knowing the environmental factor that acts as a selective (Ebert, 2008; Krebs, Routtu, & Ebert, 2017; Routtu & Ebert, 2015). 2 | MATERIALS AND METHODS with H. tvaerminnensis for at least three years. The age of the sampled agent and being able to experimentally manipulate it. Evidence from experimental evolution in both natural and semi‐nat‐ populations ranged from 5 to 31 years, with most populations over Host‐parasite systems are ideal for investigating local adaptation ural conditions indicates that H. tvaerminnensis has the potential to 2.1 | The study system 10 years old (Figure 1). To ensure that every female represented a at the metapopulation level because parasites exert strong selec‐ drive rapid changes in the genetic composition of host populations Our focal study system is a natural metapopulation of the plank‐ unique genotype, we collected samples in May 2014 after they had tion on their hosts (Barber & Dingemanse, 2010; Cornetti, Hilfiker, (Haag & Ebert, 2004; Zbinden, Haag, & Ebert, 2008) and that para‐ tonic crustacean Daphnia magna found in seasonal freshwater rock freshly hatched from sexually produced resting eggs. These samples Lemoine, & Tschirren, 2018; Haldane, 1949), which can in turn site‐mediated selection leads to rapid host adaptation to the parasite pools on archipelago islands in southwestern Finland (59°50′N, were propagated clonally in the laboratory for all further experiments, lead to local adaptation (Karvonen & Seehausen, 2012; Spichtig & (Zbinden et al., 2008).

92 93 CABALZAR et AL. | 4773 so we will refer to these genotypes as clones from now on. Seven (first generation) or singly (second and third generation) for three clones per populations were used for all experiments. Since most ani‐ generations under standard laboratory conditions. Females from the mals sampled from naturally infected populations carried the parasite, third generation were the mothers of the experimental animals. First we treated all collected clones with Fumagilin‐B (Medivet) following clutch offspring were avoided because they are generally smaller. the protocol of Zbinden, Lass, Refardt, Hottinger, and Ebert (2005) Because Finnish D. magna frequently produce male offspring in later to obtain parasite‐free lines for each clone. Prior to the experiments, clutches even under good conditions (Roth, Ebert, Vizoso, Bieger, & all animals were kept in monoclonal populations in artificial medium Lass, 2008), we minimized the chances of having only male offspring (ADaM; Klüttgen, Dülmer, Engels, & Ratte, 1994) under constant labo‐ by collecting the experimental animals from the second, third, or ratory conditions at 20°C with a dark: light cycle of 8:16 hr for several fourth clutch over a time period of 12 days. generations. They were fed green algae (Scenedesmus sp.) grown in The experimental animals were followed daily until they released chemostats. These laboratory conditions were applied to all the exper‐ their third clutch. At this time, the following data were collected: iments described in the section below. To avoid positional effects, jars birthdate, age and size at maturity (when the first eggs appear in were rearranged twice a week in all experiments. the brood chamber), survival until the third clutch was released, off‐ spring production, date when clutches were released, and size after the third clutch was released. Based on these data we calculated 2.2 | Experimental design the net reproductive rate (R0 = ∑mx*lx, where mx is the number of

This study consisted of three independent common garden experi‐ offspring to an individual female for a specific time x, and lx is the ments: two to test whether a difference in fitness exists between probability of that individual surviving to time x), as well as the in‐ D. magna individuals from naturally infected and uninfected popula‐ trinsic growth rate (Malthusian r) and the adult somatic growth rate tions, and one to assess whether infected and uninfected D. magna (size after third clutch released − size at maturity) for each individual. populations differ in resistance to H. tvaerminnensis and whether re‐ The intrinsic growth rate was iteratively calculated using a custom sistance is costly. First, we conducted a life‐history experiment to script in the r statistical software (script provided on: https ://doi. estimate the net reproductive rate (R0) and the intrinsic growth rate org/10.5061/dryad.3r228 0gbp; R Core Team, 2016) with the Euler‐ (Malthusian r) of each clone in the absence of the parasite. These are Lotka equation: important fitness components at the beginning of the season, when a 1 = e−rxI m X X D. magna population emerges from its resting eggs, and growth is not limited by competition over resources. Second, we measured relative Our life table calculation of r assumes that all offspring are fe‐ fitness in a competition experiment, where every clone competed males, which is often not the case. However, we believe that this against a fixed tester clone, both in the presence and in the absence of effect is small, as first clutch offspring, which result in the biggest the parasite. The competitive ability of a D. magna clone is an impor‐ contribution to r, are typically all female. The size of this effect for tant fitness proxy during midseason, when population density is high males being produced in later clutches was not assessed. (Capaul & Ebert, 2003). Third, we conducted an infection experiment Age at maturity and age at the release of the first offspring are measuring parasite growth in monoclonal host populations. Previous highly correlated. To reduce inaccuracies that can occur when ani‐ studies in the D. magna–H. tvaerminnensis system have shown that mals are checked only once every 24 hr, these data were averaged the amount of parasite spores per individual after a defined time post and called mid‐primiparae age. infection is a good measure of host resistance (Ebert, 2008). We then combined the phenotypic data with the population genetic data de‐ 2.4 | Competition experiment rived from microsatellites to test whether fitness is correlated with within‐population gene diversity and to conduct a QST–FST‐like analy‐ The competition experiment was conducted as described by Zbinden sis to test for signatures of parasite mediated selection. et al. (2008). In short, the experiment was set up in four blocks with two replicates per clone and parasite treatment. Experimental clones (infected or uninfected, depending on the parasite treatment) 2.3 | Life‐history experiment competed against a GFP tester clone (GFP‐positive, Kato, Matsuura, The life‐history experiment was conducted with five replicates per & Watanabe, 2012) for 28 days, with an initial frequency of the two clone, seven clones per population (except for one population that clones of 50%. The GFP tester clone is resistant against the parasite had only six clones) and 11 populations. Both before and during the and was therefore uninfected in both parasite treatments. experiment, all clones were kept in 100 ml jars filled with 80 ml of At day 28, 71–96 individuals per jar were randomly collected ADaM under controlled laboratory conditions and fed daily between and put in 96 well plates filled with 30% EtOH to induce death. five and 20 Mio cells of algae depending on their age. Animals were The plates were then examined under a fluorescence microscope, transferred to new ADaM either after they released a clutch or every and the number of GFP positive and GFP negative individuals were 3 days. counted. Only very few males were observed while assessing the To standardize the experimental animals and reduce variation relative proportion in these population. We therefore did not at‐ in maternal effects, mature females were kept in groups of three tempt to account for males. Relative fitness of the experimental

94 95 CABALZAR et AL. | 4773 4774 | CABALZAR et AL. so we will refer to these genotypes as clones from now on. Seven (first generation) or singly (second and third generation) for three clone (ln(w)) was calculated as a function of the change in fre‐ as a fixed effect, while population nested within infection status and clones per populations were used for all experiments. Since most ani‐ generations under standard laboratory conditions. Females from the quency during clonal competition (Hartl & Clark, 1997) using the clone nested within population were fitted as random factors. In the mals sampled from naturally infected populations carried the parasite, third generation were the mothers of the experimental animals. First formula ln(w) = ln(A/B), where A and B were the frequencies of case of relative fitness, parasite presence/absence (parasite treatment) we treated all collected clones with Fumagilin‐B (Medivet) following clutch offspring were avoided because they are generally smaller. the experimental and tester clones, respectively, at day 28. Since during the experiment was fitted as a fixed factor. For experiments the protocol of Zbinden, Lass, Refardt, Hottinger, and Ebert (2005) Because Finnish D. magna frequently produce male offspring in later the GFP‐tester clone is not susceptible to H. tvaerminnensis, it is conducted over several blocks, the block variable was also included in to obtain parasite‐free lines for each clone. Prior to the experiments, clutches even under good conditions (Roth, Ebert, Vizoso, Bieger, & not expected to suffer a fitness loss such as infected animals do. the model as a random factor to control for block effects. To assess the all animals were kept in monoclonal populations in artificial medium Lass, 2008), we minimized the chances of having only male offspring Hence, the mean relative fitness of the experimental clones in the significance of each variable, models were compared using paramet‐ (ADaM; Klüttgen, Dülmer, Engels, & Ratte, 1994) under constant labo‐ by collecting the experimental animals from the second, third, or presence and absence of the parasite can be directly compared, ric bootstrapping, with 10,000 simulations of the PBmodcomp func‐ ratory conditions at 20°C with a dark: light cycle of 8:16 hr for several fourth clutch over a time period of 12 days. and the difference can be attributed to the parasite's effect on the tion from the pbkrTeST package (Halekoh & Højsgaard, 2014). Variance generations. They were fed green algae (Scenedesmus sp.) grown in The experimental animals were followed daily until they released experimental clone. components of the random effects were calculated as described in chemostats. These laboratory conditions were applied to all the exper‐ their third clutch. At this time, the following data were collected: Crawley (2007). All statistical analyses were conducted in r (version iments described in the section below. To avoid positional effects, jars birthdate, age and size at maturity (when the first eggs appear in 3.3.2; R Core Team, 2016), and mixed models were fitted using the 2.5 | Infection experiment were rearranged twice a week in all experiments. the brood chamber), survival until the third clutch was released, off‐ package lme4 (version 1.1‐12; Bates, Mächler, Bolker, & Walker, 2015). spring production, date when clutches were released, and size after In the third experiment, we determined host resistance by meas‐ Because relative fitness only showed signals of divergent selection in the third clutch was released. Based on these data we calculated uring spore load per individual after a defined time post infection the Q –F ‐like analysis when parasites were absent (see Sections 2.8 2.2 | Experimental design ST ST the net reproductive rate (R0 = ∑mx*lx, where mx is the number of (Ebert, 2008; Gandon & Michalakis, 2000). We set up three rep‐ and 3), we also fitted a mixed model for relative fitness in the absence

This study consisted of three independent common garden experi‐ offspring to an individual female for a specific time x, and lx is the licated monoclonal populations for each clone in 1.5 L jars filled and the presence of the parasite separately. Plots were drawn using ments: two to test whether a difference in fitness exists between probability of that individual surviving to time x), as well as the in‐ with 1.3 L ADaM (11 populations with seven clones each = 231 the R package GGplOT2 (Wickham, 2016). D. magna individuals from naturally infected and uninfected popula‐ trinsic growth rate (Malthusian r) and the adult somatic growth rate jars). We then added 1 ml of a spore suspension (made by ho‐ tions, and one to assess whether infected and uninfected D. magna (size after third clutch released − size at maturity) for each individual. mogenizing 17 adult, infected D. magna females from each of the 2.7 | Microsatellite genotyping and analysis populations differ in resistance to H. tvaerminnensis and whether re‐ The intrinsic growth rate was iteratively calculated using a custom six naturally infected populations—102 animals in total) to each sistance is costly. First, we conducted a life‐history experiment to script in the r statistical software (script provided on: https ://doi. population. Due to low genetic diversity and asexual reproduc‐ We genotyped all D. magna clones at 24 microsatellite markers grouped estimate the net reproductive rate (R0) and the intrinsic growth rate org/10.5061/dryad.3r228 0gbp; R Core Team, 2016) with the Euler‐ tion of H. tvaerminnensis it is unlikely that parasite isolates from in five multiplex sets (Agostini et al., 2011; Colson, Du Pasquier, & (Malthusian r) of each clone in the absence of the parasite. These are Lotka equation: different hosts strongly differ in their fitness (Haag, Traunecker, Ebert, 2009; Appendix S2). Genomic DNA was extracted from a single important fitness components at the beginning of the season, when a & Ebert, 2013). However, we chose to use a mix of parasites ob‐ individual from each clone following a protocol adapted from Edwards, 1 = e−rxI m X X D. magna population emerges from its resting eggs, and growth is not tained from all our infected populations to balance out possible Johnstone, and Thompson (1991; see Appendix S3 for details). For limited by competition over resources. Second, we measured relative Our life table calculation of r assumes that all offspring are fe‐ effects of parasite adaptation to the host. The spore suspension fragment amplification, PCR reactions of 5 µl were set up using the fitness in a competition experiment, where every clone competed males, which is often not the case. However, we believe that this had a concentration of 2.5 Mio H. Tvaerminnensis spores per ml. QIAGEN Multiplex PCR kit (see Appendix S3 for details). Genotyping against a fixed tester clone, both in the presence and in the absence of effect is small, as first clutch offspring, which result in the biggest This is about the number of parasite spores an infected animal has was done on an AB 3130xl Genetic Analyzer (Applied Biosystems) using the parasite. The competitive ability of a D. magna clone is an impor‐ contribution to r, are typically all female. The size of this effect for at death, allowing our experiment to mimic the invasion of an un‐ GeneScan 500 LIZ size standard (Applied Biosystems). Microsatellite al‐ tant fitness proxy during midseason, when population density is high males being produced in later clutches was not assessed. infected host population by a single infected animal. The parasite leles were scored using Genemapper Software version 4.0 (ABI Prism) (Capaul & Ebert, 2003). Third, we conducted an infection experiment Age at maturity and age at the release of the first offspring are was then allowed to spread in the host population for 4 weeks. and binned with the software Tandem v. 1.09 (Matschiner & Salzburger, measuring parasite growth in monoclonal host populations. Previous highly correlated. To reduce inaccuracies that can occur when ani‐ After that time, we randomly collected 80 individuals per popula‐ 2009). We obtained a complete set of genotypes for all 77 D. magna studies in the D. magna–H. tvaerminnensis system have shown that mals are checked only once every 24 hr, these data were averaged tion and froze them for later spore counting. clones. We used the software micrO‐checker v. 2.2.3 (Van Oosterhout, the amount of parasite spores per individual after a defined time post and called mid‐primiparae age. Due to logistical constraints, the procedure was split into three Hutchinson, Wills, & Shipley, 2004) to assess the microsatellite data infection is a good measure of host resistance (Ebert, 2008). We then blocks, representing the three replicates per clone. These blocks for the presence of null‐alleles, large allele dropout, and stutter peaks, combined the phenotypic data with the population genetic data de‐ were run in parallel but were set up and sampled on three consec‐ as well as for typographic errors. We additionally conducted further 2.4 | Competition experiment rived from microsatellites to test whether fitness is correlated with utive days. Because freezing might affect the parasites' infectivity, assessment for the presence of null alleles following the recommen‐ within‐population gene diversity and to conduct a QST–FST‐like analy‐ The competition experiment was conducted as described by Zbinden spore suspensions were produced fresh for each block. dations of Manangwa et al. (2019; see Appendix S3 for details). The sis to test for signatures of parasite mediated selection. et al. (2008). In short, the experiment was set up in four blocks with We then pooled and homogenized the 80 collected D. magna, fSTaT v. 2.9.3.2 program (Goudet, 2001) was used to test for deviations two replicates per clone and parasite treatment. Experimental labelled the spores with a fluorescent dye (Calcofluor White), and from the Hardy–Weinberg Equilibrium (HWE) within populations and clones (infected or uninfected, depending on the parasite treatment) counted them under a fluorescence microscope using a Neubauer to estimate F between infected and uninfected populations, F 2.3 | Life‐history experiment ST ST competed against a GFP tester clone (GFP‐positive, Kato, Matsuura, improved cell counting chamber (Superior Marienfeld). Each sample and FIS over all loci and subpopulations and Nei's estimator of within‐ The life‐history experiment was conducted with five replicates per & Watanabe, 2012) for 28 days, with an initial frequency of the two was counted twice, and the average was taken to calculate spore population gene diversity (Hs; Nei, 1987). We used a Monte‐Carlo ap‐ clone, seven clones per population (except for one population that clones of 50%. The GFP tester clone is resistant against the parasite load per individual. Spore counts were log10 transformed to approx‐ proximation of Fisher's exact test implemented in arlequin (version 3.5; had only six clones) and 11 populations. Both before and during the and was therefore uninfected in both parasite treatments. imate normality for the QST–FST‐like analysis. Excoffier & Lischer, 2010) to test for linkage disequilibrium between all experiment, all clones were kept in 100 ml jars filled with 80 ml of At day 28, 71–96 individuals per jar were randomly collected pairs of markers. ADaM under controlled laboratory conditions and fed daily between and put in 96 well plates filled with 30% EtOH to induce death. 2.6 | Statistical and population genetic analysis five and 20 Mio cells of algae depending on their age. Animals were The plates were then examined under a fluorescence microscope, 2.8 | Associations between genetic, transferred to new ADaM either after they released a clutch or every and the number of GFP positive and GFP negative individuals were To analyze how the original population's infection status affected rela‐ environmental, and phenotypic data 3 days. counted. Only very few males were observed while assessing the tive fitness and life‐history traits, we fitted a linear mixed model for each

To standardize the experimental animals and reduce variation relative proportion in these population. We therefore did not at‐ trait separately. Spore load was fitted with a Poisson error distribution Using a Mantel test in r (R Core Team, 2016) from the R package ecO‐ in maternal effects, mature females were kept in groups of three tempt to account for males. Relative fitness of the experimental and a log link function. In all models natural infection status was fitted diST (Goslee & Urban, 2007), we compared the pairwise geographic

94 95 CABALZAR et AL. | 4775

QST–FST‐like analysis proposed by Ovaskainen et al. (2011), which and higher intrinsic growth rate (Malthusian r; 0.255 ± 0.029 vs. is implemented in the R package drifTSel (Karhunen et al., 2013). 0.224 ± 0.042), than clones from uninfected populations (Figure 2 This method uses a Bayesian framework of the animal model and and Table 1). Population explained between 9% and 31.1% of the an extended F‐model to obtain the joint posterior distribution of variance and significant clone effects were found for all life‐history the mean additive genotypes of the local populations, the additive traits, explaining between 14.6% and 36.1% of the total variation genetic variance of the ancestral population, and a refined version (Table 1). of the population‐level coancestry matrix ϴP from neutral genetic In the competition experiment, both parasite treatment (infected markers and phenotypic data. Whereas ϴP is first estimated from the or uninfected during competition) and natural infection status had genetic data by applying the admixture F‐model implemented in the a significant effect on the relative fitness of a clone. As expected,

R package rafm (Karhunen & Ovaskainen, 2012) and used as a prior clones had a higher relative fitness when they competed against the for fitting the animal model as described by Ovaskainen et al. (2011; GFP‐tester clone in the absence of the parasite than in the pres‐

MH() function in drifTSel). ence of the parasite (−0.422 ± 1.263 vs. −2.149 ± 1.446, p < .0001,

Compared to traditional QST–FST methods (Rogers, 1986; Spitze, Figure 3a,b). Similar to the life‐history experiment, we found that

1993; Wright, 1951), drifTSel does not directly compare QST to FST naturally infected populations had higher relative fitness than nat‐ but describes the likelihood of the observed pattern of population urally uninfected populations in both parasite treatments (para‐ divergence under genetic drift alone (neutrality) (S.test in drifTSel). site absent: 0.236 ± 0.782 vs. −1.202 ± 1.288; parasite present: This neutrality test can also incorporate habitat information to test −1.543 ± 1.463 vs. −2.877 ± 1.048; Table 1 and Figure 3a,b). There whether populations in similar habitats are more similar to each was no significant interaction between parasite treatment (competi‐ other than expected under neutrality (H.test in drifTSel; Karhunen tion in the presence/absence of the parasite) and the natural infec‐ et al., 2014). The H values yielded by this test range from 0 to 1, tion status of a population; hence, this experiment does not provide high values indicate that populations are adapted to their environ‐ evidence for a cost of resistance. Both population and clone signifi‐ ment (e.g., H > 0.95 indicates adaptation to the environment at a 95% cantly influenced relative fitness, with population explaining 10.4% incredibility level) whereas low values imply that the environment of the variance, and clone explaining 13.3% (Table 1). is not important regarding adaptation. This approach allowed us to The analysis of relative fitness in the presence/absence of the parasite showed that in the absence of the parasite most variation

96 97 CABALZAR et AL. | 4775 4776 | CABALZAR et AL.

distance matrix to the matrix of pairwise FST to test for isolation‐by‐ estimate signals of selection with respect to the natural infection distance (IBD). We conducted 100,000 permutation tests to statis‐ status of our populations. tically compare the observed matrix to the null hypothesis that the We computed the population‐level coancestry matrix ϴP in the estimated Mantel r is significantly > 0 (one‐sided, significantly posi‐ R package rafm (Karhunen & Ovaskainen, 2012) by running 200,000 tive). To visualize the relationship between genetic and geographic iterations with a burnin of 10,000 iterations and a thinning inter‐ distance, linearized FST values (FST/(1−FST)) were plotted against the val of 100. The MH() function of the R package drifTSel (Karhunen logarithm of pairwise geographic distance (Rousset, 1997). Because et al., 2013; modified for clonal data structure provided by Otso a number of recent publications have questioned the accuracy of Ovaskainen, see De Kort et al., 2016) was executed using 50,000 the Mantel test for spatial analysis, we also applied the recommen‐ iterations with a burnin of 10,000 iterations and a thinning inter‐ dations of Legendre, Fortin, and Borcard (2015), wherein our ma‐ val of 50. All traits were analyzed individually. For relative fitness trix of pairwise genetic distances was decomposed into principle we included experimental block as a random factor in the model components using the R base stats function prcomp and used as to correct for a block effect in the data. We then used the H.test a response variable in a distance‐based Moran's eigenvector map in drifTSel to estimate signals of selection with respect to the natural (dbMEM) analysis by redundancy analysis (RDA). Our pairwise geo‐ infection status of the population. For every analysis three parallel graphic distance matrix was used to compute a dbMEM using the R chains were run to check convergence, and the variance within and package adeSpaTial function dbmem (Dray et al., 2019). RDA analysis between chains was compared using the Gelman‐Rubin diagnostics was conducted using the R package veGan function rda (Oksanen implemented in the R package cOda (Plummer, Best, Cowles, & Vines, et al., 2019), with significance assessed with 1,000 permutations. 2006). Mean within‐population gene diversity (Hs) between natu‐ rally infected and uninfected populations was compared in FSTAT 3 | RESULTS using 10,000 permutations. We used Spearman's rank correlations to quantify the relationships between Hs and mean population fit‐ 3.1 | Life‐history, competition, and infection ness (Malthusian r [see life‐history experiment], and Hs and relative experiment fitness in the absence/presence of the parasite [see Section 2.4]), FIGURE 2 Boxplots of the five life‐history traits for each population and grouped by the natural infection status. Boxplots show as well as Hs and population age. All statistical analyses were con‐ Clones from naturally infected populations matured and re‐ median, upper and lower quartiles, with the dots representing outliers. Notches extend to ±1.58 IQR/sqrt(n) [Colour figure can be viewed at wileyonlinelibrary.com] ducted in r v. 3.3.2 (R Core Team, 2016). produced significantly earlier (mean 10.25 days ± SD 1.00 vs. To test for signatures of parasite‐mediated divergent selection 10.92 days ± 1.36), had higher adult somatic growth (1.05 mm ± 0.098 was explained by the different populations, whereas in the presence rather than by physical linkage of the markers. Within population between naturally infected and uninfected populations we used the vs. 0.89 mm ± 0.19), higher R (48.27 ± 9,93 vs. 36.36 ± 11.70), 0 of the parasite most variation was explained by the different clones. gene diversity (Hs) ranged from 0.0238 (M‐64) to 0.2671 (N‐27), QST–FST‐like analysis proposed by Ovaskainen et al. (2011), which and higher intrinsic growth rate (Malthusian r; 0.255 ± 0.029 vs. Natural infection status had a significant effect on relative fitness in with a mean Hs of 0.1423 (Figure 4a). FST among uninfected popu‐ is implemented in the R package drifTSel (Karhunen et al., 2013). 0.224 ± 0.042), than clones from uninfected populations (Figure 2 both analyses (Table S2 in Appendix S1). lations (0.726) was significantly higher than F among infected This method uses a Bayesian framework of the animal model and and Table 1). Population explained between 9% and 31.1% of the ST The infection experiment revealed no clear evidence that in‐ populations (0.458, p = .038), F over all loci and subpopulations an extended F‐model to obtain the joint posterior distribution of variance and significant clone effects were found for all life‐history ST dividuals from naturally infected populations have a lower spore was 0.568 (p = .0001). Computed over all populations per‐marker the mean additive genotypes of the local populations, the additive traits, explaining between 14.6% and 36.1% of the total variation load than clones from naturally uninfected population (12,425 F and F ranged from 0.167 (B008) to 0.829 (A003) and from genetic variance of the ancestral population, and a refined version (Table 1). ST IS spores ± 6,090 vs. 13,899 spores ± 7,943; Table 1 and Figure 3c,d). –0.465 (93) to 0.438 (173), respectively (see Appendix S1 for per of the population‐level coancestry matrix ϴP from neutral genetic In the competition experiment, both parasite treatment (infected Spore load per individual significantly varied between populations locus F and F computed for all populations and for infected and markers and phenotypic data. Whereas ϴP is first estimated from the or uninfected during competition) and natural infection status had ST IS and clones indicating genetic variation for host resistance (Table 1). uninfected populations separately). IBD analysis showed a signifi‐ genetic data by applying the admixture F‐model implemented in the a significant effect on the relative fitness of a clone. As expected, cant positive correlation of pairwise genetic distance with geo‐ R package rafm (Karhunen & Ovaskainen, 2012) and used as a prior clones had a higher relative fitness when they competed against the graphic distance between populations, with an estimated Mantel for fitting the animal model as described by Ovaskainen et al. (2011; GFP‐tester clone in the absence of the parasite than in the pres‐ 3.2 | Population genetic analysis R of 0.267 (p [R ≤ 0] = 0.034, with a 95% confidence interval of MH() function in drifTSel). ence of the parasite (−0.422 ± 1.263 vs. −2.149 ± 1.446, p < .0001, All 77 individuals were successfully genotyped at the 24 micros‐ [0.089, 0.408], based on 100,000 permutation tests; Figure 5). Compared to traditional Q –F methods (Rogers, 1986; Spitze, Figure 3a,b). Similar to the life‐history experiment, we found that ST ST atellite markers. Four of these markers were monomorphic (B045, Additionally, our bMEM analysis showed that geographic distance 1993; Wright, 1951), drifTSel does not directly compare Q to F naturally infected populations had higher relative fitness than nat‐ ST ST B065, B033 and B050) across all tested populations. The poly‐ had a significant effect on pairwise genetic distance (p = .004; but describes the likelihood of the observed pattern of population urally uninfected populations in both parasite treatments (para‐ morphic loci showed between two and five alleles. Analysis with R2 = 0.44). divergence under genetic drift alone (neutrality) (S.test in drifTSel). site absent: 0.236 ± 0.782 vs. −1.202 ± 1.288; parasite present: the micrO‐checker program indicated excess of homozygosity due This neutrality test can also incorporate habitat information to test −1.543 ± 1.463 vs. −2.877 ± 1.048; Table 1 and Figure 3a,b). There to null alleles or stuttering for marker B074 in eight populations, whether populations in similar habitats are more similar to each was no significant interaction between parasite treatment (competi‐ 3.3 | Associations between genetic, so we excluded this marker from further analysis, leaving 19 poly‐ other than expected under neutrality (H.test in drifTSel; Karhunen tion in the presence/absence of the parasite) and the natural infec‐ environmental, and phenotypic data morphic and four monomorphic markers. There was no significant et al., 2014). The H values yielded by this test range from 0 to 1, tion status of a population; hence, this experiment does not provide deviation from HWE within populations (p = .97 based on 10,000 The gene diversity (Hs) of naturally infected populations was signifi‐ high values indicate that populations are adapted to their environ‐ evidence for a cost of resistance. Both population and clone signifi‐ randomizations). Significant linkage disequilibrium was found for cantly higher than of naturally uninfected populations (0.193 ± 0.080 ment (e.g., H > 0.95 indicates adaptation to the environment at a 95% cantly influenced relative fitness, with population explaining 10.4% 101 out 171 marker pairs (Appendix S4). Since our markers are ran‐ vs. 0.082 ± 0.085, p = .020; Figure 4b). Hs showed strong positive incredibility level) whereas low values imply that the environment of the variance, and clone explaining 13.3% (Table 1). domly drawn from the genome, linkage is most likely caused by the correlations with Malthusian r (Rho = 0.85, n = 11, p = .002) and rela‐ is not important regarding adaptation. This approach allowed us to The analysis of relative fitness in the presence/absence of the strong genetic structure of the focal metapopulation (see below) tive fitness in the absence (Rho = 0.91, n = 11, p < .001; Figure 6a,b) parasite showed that in the absence of the parasite most variation

96 97 CABALZAR et AL. | 4777 and the presence of the parasite (Rho = 0.74, n = 11, p = .013). There selection. All other estimates of H changed only marginally. Without was no significant association between population age and Hs the population M‐64 the signal for isolation by distance was still sig‐ (Rho = 0.096, n = 11, p = .78). nificant in both the mantel test (Matel R = 0.33, p (R ≤ 0) = 0.021, 2 The QST‐FST‐like analysis including habitat information (H.test in 95% CI: 0.10–0.55) and the bdMEM anlysis (p = .014; R = 0.46). As drifTSel) showed evidence for divergent selection between infected the exclusion of population M‐64 did not change our general conclu‐ and uninfected populations for the traits R0 (H = 0.98) and relative sions, we will discuss the results including all 11 populations. fitness in the absence of the parasite (H = 0.94) The other traits showed no signal for parasite‐mediated divergent selection, with H ranging from 0.44 to 0.87 (Table 2). 4 | DISCUSSION

TABLE 1 Statistics for the effect of infection status, population and clone for all traits

Fixed factor effects Random factor effects

Infection status Population Clone Residual effects

Variance com‐ Variance compo‐ Variance component Trait p‐Value p‐Value ponent (%) p‐Value nent (%) (%)

Mid‐primiparae age .015 .08 — .0002 26.5 73.5 Size at maturity .412 .002 12.7 .0001 20.6 66.7 Adult growth .042 .0002 31.1 .0001 14.6 54.3

R0 .021 .0002 17.1 .0002 30.9 52.1 Malthusian r .041 .007 11.5 .0002 36.1 52.3 Spore load .778 .0026 9.0 .0001 30.0 NA

Fixed factor effects Random factor effects

Infection Parasite status treatmenta Population Clone Residual effects

Variance Variance compo‐ Trait p‐Value p‐Value p‐Value component p‐Value nent Variance component

Relative fitness .0025 >.0001 .0008 10.4 .0029 13.3 70.5

Note: The models for spore load and relative fitness include random block effects with variance components of 61.0% and 5.7%, respectively. aParasite treatment = presence or absence of the parasite during the competition experiment.

98 99 CABALZAR et AL. | 4777 4778 | CABALZAR et AL. and the presence of the parasite (Rho = 0.74, n = 11, p = .013). There selection. All other estimates of H changed only marginally. Without (a) (c) was no significant association between population age and Hs the population M‐64 the signal for isolation by distance was still sig‐ (Rho = 0.096, n = 11, p = .78). nificant in both the mantel test (Matel R = 0.33, p (R ≤ 0) = 0.021, 2 The QST‐FST‐like analysis including habitat information (H.test in 95% CI: 0.10–0.55) and the bdMEM anlysis (p = .014; R = 0.46). As drifTSel) showed evidence for divergent selection between infected the exclusion of population M‐64 did not change our general conclu‐ and uninfected populations for the traits R0 (H = 0.98) and relative sions, we will discuss the results including all 11 populations. fitness in the absence of the parasite (H = 0.94) The other traits showed no signal for parasite‐mediated divergent selection, with H ranging from 0.44 to 0.87 (Table 2). 4 | DISCUSSION

Parasites exert strong selection on their hosts, which can lead to the 3.4 | Extreme population structure rapid evolution of host defences such as resistance. Spatial variation Since the genetic structure of metapopulations is strongly influenced in parasite abundance may constitute an important mechanism for (b) (d) by dynamic extinction and recolonization processes as well as by mi‐ maintaining genetic variation in resistance across populations. Here, gration rates, isolated populations might have low genetic variation we used an integrative approach to test whether spatial variation due to strong founder effects and low rates of immigration. In our in parasite presence explains variation in resistance and life‐history study, population M‐64 seemed the most extreme case. This popu‐ traits in a natural metapopulation of D. magna. We sought to address lation is isolated from other D. magna populations, exhibits very low four specific questions: (a) Are individuals from infected populations genetic diversity (only one polymorphic marker out of 24 markers), on average more resistant to H. tvaerminnensis than individuals from and turned out to be an outlier in almost all of the phenotypic assays. uninfected populations? (b) Does the QST–FST‐like analysis show sig‐ Therefore, we repeated all analyses without population M‐64 to as‐ natures of adaptive divergence between naturally infected and un‐ sess whether our results were overly influenced by this exceptional infected populations? (c) Is resistance to H. tvaerminnensis costly? (d) population. As one would expect when an extreme value is removed, Does host fitness correlate with within‐population gene diversity? omitting M‐64 from the linear mixed model weakened the signifi‐ cant variation among populations for most life‐history traits; how‐ 4.1 | No evidence for host local adaptation and ever, clone effects remained significant for all traits. The posterior FIGURE 3 (a) Boxplots of relative fitness tested in the presence and absence of the parasite grouped by population and (b) grouped by differential selection for resistance probability for divergent selection of adult somatic growth between natural infection status. (c) Boxplots of spore load per host individual for every population and (d) grouped by the natural infection status. Boxplots show median, upper and lower quartiles, with the dots representing outliers. Notches extend to ±1.58 IQR/sqrt(n) [Colour figure naturally infected and uninfected populations (H) changed from Although we found strong variation in resistance between popula‐ can be viewed at wileyonlinelibrary.com] 0.8 to 0.6, indicating neutral evolution rather than a weak signal of tions and clones, the natural infection status of a population had no effect on resistance against H. tvaerminnensis (Figure 3c,d). Also, the bill shape in island populations of Berthelot's pipit was shown to be TABLE 1 Statistics for the effect of infection status, population and clone for all traits QST–FST analysis for resistance (spore load) showed no signals of di‐ driven by the colonization history rather than by natural selection Fixed factor vergent selection (Table 2). Thus, we have no evidence for host local (Spurgin, Illera, Jorgensen, Dawson, & Richardson, 2014). Likewise, effects Random factor effects adaptation to control parasite proliferation. Demographic processes, a study of local adaptation of resting stage production in the Finnish such as founder effects and migration, might contribute to these find‐ D. magna metapopulation (Roulin et al., 2015) has found no evidence Infection status Population Clone Residual effects ings by overriding the effects of adaptive evolution at the subpopula‐ for local adaptation, despite high variation in dormancy‐related traits Variance com‐ Variance compo‐ Variance component tion level. Our microsatellite data showed that populations are highly within and between populations. Trait p‐Value p‐Value ponent (%) p‐Value nent (%) (%) differentiated from each other and exhibit very low genetic diversity In addition, strong founder effects result in low effective pop‐ Mid‐primiparae age .015 .08 — .0002 26.5 73.5 (Figure 4a), even though long‐term data on this metapopulation has ulation size (Hartl & Clark, 1997), rendering selection ineffective, Size at maturity .412 .002 12.7 .0001 20.6 66.7 revealed continuous extinction and recolonization, indicating migra‐ even though the parasite is highly prevalent and virulent (Bieger & Adult growth .042 .0002 31.1 .0001 14.6 54.3 tion and gene flow (Ebert, Hottinger, & Pajunen, 2013; Pajunen & Ebert, 2009; Lass & Ebert, 2006; Vizoso & Ebert, 2005). Although

R0 .021 .0002 17.1 .0002 30.9 52.1 Pajunen, 2003; Robinson, Hall, & Wares, 2013). As expected, both two experimental evolution studies reported adaptation to H. tvaer‐ Malthusian r .041 .007 11.5 .0002 36.1 52.3 the Mantel test and a dbMEM analysis showed a pattern of isola‐ minnensis (Haag & Ebert, 2004; Zbinden et al., 2008), these studies Spore load .778 .0026 9.0 .0001 30.0 NA tion‐by‐distance (IBD; Figure 5). High levels of population differ‐ were both set up artificially using populations of genetically variable entiation despite high dispersal capabilities are commonly found in clones. Therefore, these experiments excluded stochastic and de‐ Fixed factor effects Random factor effects populations of passively dispersed invertebrates (Boileau, Hebert, & mographic variation among populations that might have limited our Infection Parasite Schwartz, 1992; Makino & Tanabe, 2009; Mills, Lunt, & Gómez, 2007; power to uncover adaptive evolution in the natural metapopulation. a status treatment Population Clone Residual effects Muñoz et al., 2008; Walser & Haag, 2012). This paradox is probably This discrepancy highlights the importance of population history in Variance caused by persistent founder effects (i.e., strong bottlenecks during determining the adaptive potential of a population. Variance compo‐ colonization by one or few individuals) followed by rapid population Trait p‐Value p‐Value p‐Value component p‐Value nent Variance component expansion and inbreeding (Haag, Riek, Hottinger, Pajunen, & Ebert, 4.2 | No cost of resistance Relative fitness .0025 >.0001 .0008 10.4 .0029 13.3 70.5 2006; Wade & McCauley, 1988). These persistent founder effects

Note: The models for spore load and relative fitness include random block effects with variance components of 61.0% and 5.7%, respectively. probably contribute to the observed variation in resistance and other Our finding that host clones that had lower spore load (indicating aParasite treatment = presence or absence of the parasite during the competition experiment. traits among populations. For example, the substantial variation in higher resistance) did not have a higher fitness in the absence of

98 99 CABALZAR et AL. | 4779

FIGURE 4 (a) Within‐population gene (a) diversity (Hs) based on 23 microsatellite markers and (b) grouped by the natural infection status. Boxplots show median and upper and lower quartiles, whiskers extend to ±1.5 interquartile range [Colour figure can be viewed at wileyonlinelibrary. com]

(b)

the parasite, neither at the subpopulation level nor at the meta‐ population level, suggests that there are no costs of resistance. This result might reflect the strongly varying heterozygosity of the subpopulations, with more heterozygote clones being fitter under both conditions (as discussed below). An earlier experiment con‐ ducted under natural outdoor conditions and using clones from the same metapopulation also found no evidence of a cost of resistance to H. tvaerminnensis (Altermatt & Ebert, 2007). These findings do not, however, exclude the possibility that under other, untested, experimental conditions, a cost would be detected (McKean, Yourth, Lazzaro, & Clark, 2008; Meaden, Paszkiewicz, & Koskella, 2015). Nevertheless, a cost of resistance trade‐off does not appear to be an important mechanism for maintaining varia‐ tion in resistance in this system.

FIGURE 5 Isolation by distance: FST/(1−FST) plotted against the logarithm of geographic distance between each population pair 4.3 | Fitness correlates with population (p = .034, Mantel r = .267). Grey area shows the 95% confidence heterozygosity interval level

We found strong positive correlations between our three fit‐ between homozygosity and fitness (inbreeding depression) are well ness measures (relative fitness, net reproductive rate (R0, data not supported by theory and empirical studies (Charlesworth & Willis, shown), and intrinsic growth rate (Malthusian r)) and the within‐ 2009; Hedrick & Garcia‐Dorado, 2016; Keller & Waller, 2002) and population gene diversity (Hs; Figure 6a,b). Negative correlations are thought to be caused by the presence of deleterious, recessive

100 101 CABALZAR et AL. | 4779 4780 | CABALZAR et AL.

FIGURE 4 (a) Within‐population gene (a) (a) TABLE 2 Posterior probabilities for divergent selection (H value) diversity (Hs) based on 23 microsatellite with respect to the natural infection status for all traits markers and (b) grouped by the natural infection status. Boxplots show median Trait H value (min.–max.) and upper and lower quartiles, whiskers Life‐history experiment extend to ±1.5 interquartile range [Colour Malthusian r 0.44 (0.420–0.456) figure can be viewed at wileyonlinelibrary. com] R0 0.98 (0.979–0.993) Mid‐primiparae age 0.46 (0.428–0.473) Size at maturity 0.16 (0.151–0.181) Adult somatic growth 0.87 (0.866–0.883) Competition experiment

(b) Relative fitness with H.t. 0.50 (0.570–0.511) Relative fitness w/o H.t. 0.94 (0.931–0.953) Resistance experiment Spore load 0.60 (0.570–0.613) (b) Note: Minimum and maximum values of H refer to the three parallel chains, which were run for every trait using the MH() function of the R package drifTSel. Abbreviation: H.t., Hamiltosporidium tvaerminnensis.

evolved in the presence or absence of H. tvaerminnensis (Zbinden et al., 2008). An explanation for this result might be the combi‐ nation of two effects: the older a population gets, the higher its FIGURE 6 Correlation between within population gene chance of becoming infected (Ebert, Hottinger, & Pajunen, 2001), diversity and (a) intrinsic growth rate (Spearman Rho = 0.85, n = 11, p = .002), as well as (b) relative fitness in the absence of the parasite and the more time it has to accumulate genetic variation due to (Spearman Rho = 0.91, n = 11, p < .001). Grey area shows the 95% immigration (Haag, Riek, Hottinger, Pajunen, & Ebert, 2005). This confidence interval level reasoning, however, cannot explain the results of the experiment by Zbinden et al., (2008), which ruled out population age effects. mutations (genetic load; Hedrick, 1994). Fixation of such mutations Also, in our study we did not find any statistically significant ef‐ is expected to be common in populations with small effective popu‐ fect of population age on either fitness traits or Hs. An alternative explanation is that H. tvaerminnensis selects for higher levels of lation size (Ne), where drift is strong and the effect of purging weak, the parasite, neither at the subpopulation level nor at the meta‐ resulting in a high drift load (Lynch, Conery, & Burger, 1995). Strong heterozygosity. A mechanistic link for this is suggested in different population level, suggests that there are no costs of resistance. host‐parasite systems, where parasitism was shown to be asso‐ founder effects result in low Ne and a higher probability of inbreed‐ This result might reflect the strongly varying heterozygosity of the ing, according to this inbreeding depression is frequently reported in ciated with host inbreeding, i.e., inbred hosts had increased sus‐ subpopulations, with more heterozygote clones being fitter under populations that were founded by only a few individuals (Billing et ceptibility to parasites, higher parasite load, and suffered higher both conditions (as discussed below). An earlier experiment con‐ al., 2012; Gelatt et al., 2010; Mattila et al., 2012). In D. magna rock mortality upon infection (Acevedo‐Whitehouse, Gulland, Greig, & ducted under natural outdoor conditions and using clones from Amos, 2003; Coltman et al., 1999; Ellison, Cable, & Consuegra, pool populations, the effect of low genetic diversity (a proxy for Ne) the same metapopulation also found no evidence of a cost of on genetic load and inbreeding depression has been demonstrated 2011; MacDougall‐Shackleton, Derryberry, Foufopoulos, Dobson, resistance to H. tvaerminnensis (Altermatt & Ebert, 2007). These by Lohr and Haag (2015). Assessment of fitness of selfed and out‐ & Hahn, 2005). However, two experimental studies of the findings do not, however, exclude the possibility that under other, crossed D. magna clones from populations with strong differences in D. magna–Hamiltosporidium system with material from the same untested, experimental conditions, a cost would be detected genetic diversity indicated that many mildly deleterious mutations metapopulation found no evidence of a synergistic effect of in‐ (McKean, Yourth, Lazzaro, & Clark, 2008; Meaden, Paszkiewicz, & are fixed in D. magna rock pool populations (Lohr & Haag, 2015), breeding and infection on relative fitness (Haag, Sakwińska, & Koskella, 2015). Nevertheless, a cost of resistance trade‐off does aligning well with our finding of a positive relationship between fit‐ Ebert, 2003; Salathé & Ebert, 2003). not appear to be an important mechanism for maintaining varia‐ ness and genetic diversity (Figure 6a,b). Although no link between parasitism and homozygosity has tion in resistance in this system. been found on an individual level, different studies suggest that

FIGURE 5 Isolation by distance: FST/(1−FST) plotted against the such a link can arise on the population level. First, the immigration 4.4 | Link between parasitism and heterozygosity logarithm of geographic distance between each population pair success of uninfected host genotypes increases with the prevalence 4.3 | Fitness correlates with population (p = .034, Mantel r = .267). Grey area shows the 95% confidence of the parasite in the invaded population (Altermatt, Hottinger, & heterozygosity interval level We further found that naturally infected populations have a higher Ebert, 2007). Thus, high prevalence of H. tvaerminnensis promotes Hs (Figure 4b), a lower FST and a higher fitness, in both the com‐ We found strong positive correlations between our three fit‐ between homozygosity and fitness (inbreeding depression) are well petition experiment (relative fitness; Figure 3d) and the life‐his‐ effective host gene flow, enhancing genetic diversity within in‐ ness measures (relative fitness, net reproductive rate (R , data not fected populations and decreasing genetic differentiation among 0 supported by theory and empirical studies (Charlesworth & Willis, tory experiment (R0 and Malthusian r; Figure 2), than individuals shown), and intrinsic growth rate (Malthusian r)) and the within‐ 2009; Hedrick & Garcia‐Dorado, 2016; Keller & Waller, 2002) and from uninfected populations. A similar result was observed in them. Altermatt et al. (2007) further found that parasites reduce population gene diversity (Hs; Figure 6a,b). Negative correlations are thought to be caused by the presence of deleterious, recessive an experimental evolution study, where replicated populations the relative fitness of allopatric hosts less than that of resident

100 101 CABALZAR et AL. | 4781 hosts, suggesting that parasites adapt to specific host lines, ele‐ fitness only in the absence of the parasite. Several other studies vating the immigrant's invasion success even more. Second, it has indicate that the interaction between infection and inbreeding is been shown that H. tvaerminnensis spreads faster in experimental complex and highly dependent on the host's genetic background populations with low genetic diversity than in populations with (Haag et al., 2003; Ouborg, Biere, & Mudde, 2000; Stevens, Yan, & high genetic diversity (the monoculture effect; Altermatt & Ebert, Pray, 1997; Strauss & Karban, 1994). Similarly, our separate mixed 2008), and, therefore, that it potentially selects for outcrossing in model analysis for relative fitness in the presence/absence of the natural host populations. The authors hypothesize that parasite parasite revealed that in the absence of the parasite, most vari‐ adaptation to specific host lineages might be responsible for this ation in relative fitness is explained by the different populations result. When a parasite is adapted to a certain host genotype, it is (Table S2 in Appendix S1), as reflected by the different levels of thought to have an impaired ability to infect and grow within an‐ Hs. But in the presence of the parasite, most variation in relative other host genotype (Kaltz & Shykoff, 1998). Thus, parasites spread fitness is explained by host clones within populations, and the more slowly in genetically diverse host populations, due to the population factor loses its significant effect (Table S2 in Appendix lower density of susceptible genotypes (dilution effect; Schmidt & S1). Hence, infection by H. tvaerminnensis seems to alter the pat‐ Ostfeld, 2001). Rare host genotypes thus gain an advantage, result‐ tern of relative fitness in a way that is highly specific to the host ing in increased genetic diversity in the host population. Third, to genotype, diminishing the effects observed at the population level. survive the winter, host and parasite must both undergo diapause. Again, these results confirm a lack of local adaptation of D. magna The parasite does so by transmitting vertically into the resting eggs to H. tvaerminnensis, with naturally infected populations showing of the host. However, H. tvaerminnensis' success of vertical trans‐ no optimized fitness in the presence of the parasite, despite plenty mission is lower to D. magna's outbred sexual eggs (85.2%) than to of variation in resistance on which selection could act. its inbred sexual eggs (98.9%; Ebert, Altermatt, & Lass, 2007). This In conclusion, by using an integrative approach, our study result is probably due to lineage‐specific parasite adaptation, giving shows how the interplay of genetic drift, ecology, and parasite‐ further support for the rare host genotype advantage. Fourth, the mediated selection molds the genetic and phenotypic structure of host's diapausing stages also act as dispersal stages. As the para‐ a metapopulation. Subpopulations experience high rates of popu‐ site is vertically transmitted to these stages, it codisperses along lation turnover, seen as high extinction‐(re)colonization dynamics with the host. Thus, pools that receive more immigrants will have from an ecological perspective and seen as founder effects, in‐ not only a higher host genetic diversity but also a higher likelihood breeding, and gene flow from a population genetic perspective. of infection. Taken together, these factors might explain the meta‐ These processes lead to dynamic changes of the subpopulation's population wide pattern of an association between parasitism and genetic composition and prevent them from reaching genetic host genetic diversity. equilibrium (Haag et al., 2005; Pajunen & Pajunen, 2003). The ge‐ netic structure of newly colonized populations is impacted by the strong founder effects, with often only a single clone recolonizing 4.5 | Differential selection for relative fitness only a rock pool. Therefore, the subsequent evolution of these popula‐ in the absence of infection tions is dominated by chance events, impeding subpopulation spe‐

Our QST–FST analysis showed differing results for relative fit‐ cific local adaptation. Then, as the population ages, environmental ness in the absence and presence of the parasite (Table 2). When effects (host and parasite immigration) and parasite‐mediated se‐ measured in the absence of the parasite, the differences in rela‐ lection (effective rate of gene flow, rare‐genotype advantage, and tive fitness between naturally infected and uninfected populations monoculture effect) increasingly contribute to the evolution of seems to be due to differential selection, whereas relative fitness local gene pools. These selective processes create strong genetic variation in the presence of the parasite does not differ from the variation among subpopulations and the accumulation of genetic expectation of random drift. This result might be linked to the ob‐ diversity is faster in infected than in uninfected populations. Our servation that naturally infected populations have a significantly results, once more, demonstrate that adaptation at the metapop‐ higher mean Hs than naturally uninfected populations and that Hs ulation level is complex and strongly dependent on the size and is strongly correlated with relative fitness. Although many studies the demographic history of the local populations, as well as on the have demonstrated that low levels of heterozygosity compromise turnover rates of subpopulations. Therefore, it is vital to combine fitness more markedly under stress (like parasitism) than under be‐ population genetic and phenotypic data of spatially structured, nign conditions (Ilmonen et al., 2008; King & Lively, 2012; Pearman natural populations to better understand how adaptive and non‐ & Garner, 2005; Smallbone, van Oosterhout, & Cable, 2016), no adaptive processes interact to shape spatial variation in functional general pattern seems to exist for how inbreeding affects fitness traits. More general, a deeper knowledge of the factors that fa‐ in the presence of infection or stress in general (Keller & Waller, vour or disfavour microgeographic adaptation is important to gain 2002). As discussed above, there is no evidence that inbreed‐ knowledge about the evolutionary potential of fragmented popu‐ ing and infection synergistically affected relative fitness in the lations. In addition to this, our study highlights the role of parasit‐ D. magna–H. tvaerminnensis system. The absence of such an effect ism in this complex interplay. A next step towards understanding might explain why we found signals of positive selection on relative this process would be to conduct time‐series monitoring of the

102 103 CABALZAR et AL. | 4781 4782 | CABALZAR et AL. hosts, suggesting that parasites adapt to specific host lines, ele‐ fitness only in the absence of the parasite. Several other studies dynamics of local populations, including host and parasite migra‐ Barber, I., & Dingemanse, N. J. (2010). Parasitism and the evolution‐ vating the immigrant's invasion success even more. Second, it has indicate that the interaction between infection and inbreeding is tion and the evolution of the host gene pool. ary ecology of animal personality. Philosophical Transactions of the Royal Society B: Biological Sciences, 365, 4077–4088. https ://doi. been shown that H. tvaerminnensis spreads faster in experimental complex and highly dependent on the host's genetic background org/10.1098/rstb.2010.0182 populations with low genetic diversity than in populations with (Haag et al., 2003; Ouborg, Biere, & Mudde, 2000; Stevens, Yan, & ACKNOWLEDGEMENTS Bates, D., Mächler, M., Bolker, B. M., & Walker, S. C. (2015). Fitting linear high genetic diversity (the monoculture effect; Altermatt & Ebert, Pray, 1997; Strauss & Karban, 1994). Similarly, our separate mixed mixed‐effects models using lme4. Journal of Statistical Software, 67, 2008), and, therefore, that it potentially selects for outcrossing in model analysis for relative fitness in the presence/absence of the We would like to thank Kristina Müller for help in the laboratory and 1–48. https ://doi.org/10.18637/ jss.v067.i01 Bieger, A., & Ebert, D. (2009). Expression of parasite virulence at dif‐ natural host populations. The authors hypothesize that parasite parasite revealed that in the absence of the parasite, most vari‐ Urs Stiefel, Jürgen Hottinger, Daniel Lüscher, Nicolas Boileau, and the ferent host population densities under natural conditions. Oecologia, adaptation to specific host lineages might be responsible for this ation in relative fitness is explained by the different populations staff of Tvärminne Biological Station for technical support. Further we 160, 247–255. https ://doi.org/10.1007/s00442‐009‐1297‐x result. When a parasite is adapted to a certain host genotype, it is (Table S2 in Appendix S1), as reflected by the different levels of thank Otso Ovaskainen for providing a modified version of the R pack‐ Billing, A. M., Lee, A. M., Skjelseth, S., Borg, Å. A., Hale, M. C., thought to have an impaired ability to infect and grow within an‐ Hs. But in the presence of the parasite, most variation in relative age drifTSel. We are also grateful to Suzanne Zweizig for proofreading Slate, J., … Jensen, H. (2012). Evidence of inbreeding depres‐ sion but not inbreeding avoidance in a natural house spar‐ other host genotype (Kaltz & Shykoff, 1998). Thus, parasites spread fitness is explained by host clones within populations, and the the manuscript and to all members of the Ebert research group and row population. Molecular Ecology, 21, 1487–1499. https ://doi. more slowly in genetically diverse host populations, due to the population factor loses its significant effect (Table S2 in Appendix three anonymous reviewers for helpful discussions and comments on org/10.1111/j.1365‐294X.2012.05490.x lower density of susceptible genotypes (dilution effect; Schmidt & S1). Hence, infection by H. tvaerminnensis seems to alter the pat‐ the manuscript. This work was supported by a Swiss National Science Blanquart, F., Gandon, S., & Nuismer, S. L. (2012). The effects of mi‐ Ostfeld, 2001). Rare host genotypes thus gain an advantage, result‐ tern of relative fitness in a way that is highly specific to the host Foundation grant to DE (Nr. 310030B_166677). gration and drift on local adaptation to a heterogeneous environ‐ ing in increased genetic diversity in the host population. Third, to genotype, diminishing the effects observed at the population level. ment. Journal of Evolutionary Biology, 25, 1351–1363. https ://doi. org/10.1111/j.1420‐9101.2012.02524.x survive the winter, host and parasite must both undergo diapause. Again, these results confirm a lack of local adaptation of D. magna AUTHOR CONTRIBUTIONS Blanquart, F., Kaltz, O., Nuismer, S. L., & Gandon, S. (2013). A practical The parasite does so by transmitting vertically into the resting eggs to H. tvaerminnensis, with naturally infected populations showing guide to measuring local adaptation. Ecology Letters, 16, 1195–1205. of the host. However, H. tvaerminnensis' success of vertical trans‐ no optimized fitness in the presence of the parasite, despite plenty A.P.C., and D.E. conceived the study, designed the experiments and https ://doi.org/10.1111/ele.12150 mission is lower to D. magna's outbred sexual eggs (85.2%) than to of variation in resistance on which selection could act. collected the animals from the metapopulation. A.P.C. performed Boileau, M. G., Hebert, P. D. N., & Schwartz, S. S. (1992). Non‐equilibrium gene frequency divergence: Persistent founder effects in natural its inbred sexual eggs (98.9%; Ebert, Altermatt, & Lass, 2007). This In conclusion, by using an integrative approach, our study the experiments and with the help of P.D.F. analyzed the data. A.P.C. populations. Journal of Evolutionary Biology, 5(1), 25–39. https ://doi. result is probably due to lineage‐specific parasite adaptation, giving shows how the interplay of genetic drift, ecology, and parasite‐ wrote the article with the input of all authors. Y.K., and H.W. pro‐ org/10.1046/j.1420‐9101.1992.50100 25.x further support for the rare host genotype advantage. Fourth, the mediated selection molds the genetic and phenotypic structure of vided the H2B‐GFP transgenic Daphnia magna clone. Capaul, M., & Ebert, D. (2003). Parasite‐mediated selection in experi‐ host's diapausing stages also act as dispersal stages. As the para‐ a metapopulation. Subpopulations experience high rates of popu‐ mental Daphnia magna populations. Evolution, 57, 249–260. https :// doi.org/10.1111/j.0014‐3820.2003.tb002 60.x site is vertically transmitted to these stages, it codisperses along lation turnover, seen as high extinction‐(re)colonization dynamics ORCID Carton, Y., Nappi, A. J., & Poirie, M. (2005). Genetics of anti‐parasite resis‐ with the host. Thus, pools that receive more immigrants will have from an ecological perspective and seen as founder effects, in‐ tance in invertebrates. Developmental and Comparative Immunology, not only a higher host genetic diversity but also a higher likelihood breeding, and gene flow from a population genetic perspective. Andrea P. Cabalzar https://orcid.org/0000‐0001‐7626‐2769 29, 9–32. https ://doi.org/10.1016/j.dci.2004.05.004 Charlesworth, D., & Willis, J. H. (2009). The genetics of inbreeding of infection. Taken together, these factors might explain the meta‐ These processes lead to dynamic changes of the subpopulation's Peter D. Fields https://orcid.org/0000‐0003‐2959‐2524 depression. Nature Reviews Genetics, 10, 783–796. https ://doi. population wide pattern of an association between parasitism and genetic composition and prevent them from reaching genetic Hajime Watanabe https://orcid.org/0000‐0001‐9657‐6554 org/10.1038/nrg2664 host genetic diversity. equilibrium (Haag et al., 2005; Pajunen & Pajunen, 2003). The ge‐ Colson, I., Du Pasquier, L., & Ebert, D. (2009). Intragenic tandem repeats Dieter Ebert https://orcid.org/0000‐0003‐2653‐3772 netic structure of newly colonized populations is impacted by the in Daphnia magna: Structure, function and distribution. BMC Research strong founder effects, with often only a single clone recolonizing Notes, 2, 206. https ://doi.org/10.1186/1756‐0500‐2‐206 4.5 | Differential selection for relative fitness only Coltman, D. W., Pilkington, J. G., Smith, J. A., & Pemberton, J. M. a rock pool. Therefore, the subsequent evolution of these popula‐ in the absence of infection DATA AVAILABILITY STATEMENT (1999). Parasite‐mediated selection against inbred Soay sheep in a tions is dominated by chance events, impeding subpopulation spe‐ free‐living, island population. Evolution, 53, 1259–1267. https ://doi. The data presented in this study are available on Dryad, https ://doi. Our QST–FST analysis showed differing results for relative fit‐ cific local adaptation. Then, as the population ages, environmental org/10.2307/2640828 ness in the absence and presence of the parasite (Table 2). When effects (host and parasite immigration) and parasite‐mediated se‐ org/10.5061/dryad.3r228 0gbp. Cornetti, L., Hilfiker, D., Lemoine, M., & Tschirren, B. (2018). Small‐scale measured in the absence of the parasite, the differences in rela‐ lection (effective rate of gene flow, rare‐genotype advantage, and spatial variation in infection risk shapes the evolution of a Borrelia resistance gene in wild rodents. Molecular Ecology, 27, 3515–3524. tive fitness between naturally infected and uninfected populations monoculture effect) increasingly contribute to the evolution of REFERENCES https ://doi.org/10.1111/mec.14812 seems to be due to differential selection, whereas relative fitness local gene pools. These selective processes create strong genetic Corradi, N., Haag, K. L., Pombert, J. F., Ebert, D., & Keeling, P. J. (2009). variation in the presence of the parasite does not differ from the variation among subpopulations and the accumulation of genetic Acevedo‐Whitehouse, K., Gulland, F., Greig, D., & Amos, W. (2003). Draft genome sequence of the Daphnia pathogen Octosporea bay‐ expectation of random drift. This result might be linked to the ob‐ diversity is faster in infected than in uninfected populations. Our Disease susceptibility in California sea lions. Nature, 422, 35. https :// eri: Insights into the gene content of a large microsporidian genome doi.org/10.1038/422035a and a model for host‐parasite interactions. Genome Biology, 10, R106. servation that naturally infected populations have a significantly results, once more, demonstrate that adaptation at the metapop‐ Agostini, C., Agudelo, P. A., Bâ, K., Barber, P. A., Bisol, P. M., Brouat, https ://doi.org/10.1186/gb‐2009‐10‐10‐r106 higher mean Hs than naturally uninfected populations and that Hs ulation level is complex and strongly dependent on the size and C., & Zulaiha, A. R. (2011). Permanent genetic resources added Crawley, M. J. (2007). The R book. Chichester, UK: John Wiley & Sons. is strongly correlated with relative fitness. Although many studies the demographic history of the local populations, as well as on the to Molecular Ecology Resources Database 1 October 2010–30 De Kort, H., Vandepitte, K., & Honnay, O. (2013). A meta‐analysis of have demonstrated that low levels of heterozygosity compromise turnover rates of subpopulations. Therefore, it is vital to combine November 2010. Molecular Ecology Resources, 11, 418–421. https :// the effects of plant traits and geographical scale on the magnitude doi.org/10.1111/j.1755‐0998.2010.02970.x of adaptive differentiation as measured by the difference between fitness more markedly under stress (like parasitism) than under be‐ population genetic and phenotypic data of spatially structured, Altermatt, F., & Ebert, D. (2007). The genotype specific competitive QST and FST. Evolutionary Ecology, 27, 1081–1097. https ://doi. nign conditions (Ilmonen et al., 2008; King & Lively, 2012; Pearman natural populations to better understand how adaptive and non‐ ability does not correlate with infection in natural Daphnia magna org/10.1007/s10682‐012‐9624‐9 & Garner, 2005; Smallbone, van Oosterhout, & Cable, 2016), no adaptive processes interact to shape spatial variation in functional populations. PLoS ONE, 2(12), e1280. https ://doi.org/10.1371/journ De Kort, H., Vander Mijnsbrugge, K., Vandepitte, K., Mergeay, J., general pattern seems to exist for how inbreeding affects fitness traits. More general, a deeper knowledge of the factors that fa‐ al.pone.0001280 Ovaskainen, O., & Honnay, O. (2016). Evolution, plasticity and Altermatt, F., & Ebert, D. (2008). Genetic diversity of Daphnia magna evolving plasticity of phenology in the tree species Alnus glutinosa. in the presence of infection or stress in general (Keller & Waller, vour or disfavour microgeographic adaptation is important to gain populations enhances resistance to parasites. Ecology Letters, 11, Journal of Evolutionary Biology, 29, 253–264. https ://doi.org/10.1111/ 2002). As discussed above, there is no evidence that inbreed‐ knowledge about the evolutionary potential of fragmented popu‐ 918–928. https ://doi.org/10.1111/j.1461‐0248.2008.01203.x jeb.12777 ing and infection synergistically affected relative fitness in the lations. In addition to this, our study highlights the role of parasit‐ Altermatt, F., Hottinger, J., & Ebert, D. (2007). Parasites promote host Dray, S., Bauman, D., Blanchet, G., Borcard, D., Clappe, S., Guenard, G., & D. magna–H. tvaerminnensis system. The absence of such an effect ism in this complex interplay. A next step towards understanding gene flow in a metapopulation. Evolutionary Ecology, 21, 561–575. Wagner, H. (2019). adeSpaTial: Multivariate multiscale spatial analysis. https ://doi.org/10.1007/s10682‐006‐9136‐6 might explain why we found signals of positive selection on relative this process would be to conduct time‐series monitoring of the

102 103 CABALZAR et AL. | 4783

R package version 0.3‐4. Retrieved from https ://CRAN.R‐proje ct.org/ et sp. nov., a microsporidian parasite of Daphnia magna, and estab‐ packa ge=adesp atial lishment of Hamiltosporidium magnivora comb. nov. Parasitology, 138, Ebert, D. (2008). Host‐parasite coevolution: Insights from the Daphnia‐ 447–462. https ://doi.org/10.1017/S0031 18201 0001393 parasite model system. Current Opinion in Microbiology, 11, 290–301. Haag, K. L., Traunecker, E., & Ebert, D. (2013). Single‐nucleotide poly‐ https ://doi.org/10.1016/j.mib.2008.05.012 morphisms of two closely related microsporidian parasites suggest Ebert, D., Altermatt, F., & Lass, S. (2007). A short term benefit for out‐ a clonal population expansion after the last glaciation. Molecular crossing in a Daphnia metapopulation in relation to parasitism. Journal Ecology, 22, 314–326. https ://doi.org/10.1111/mec.12126 of the Royal Society Interface, 4, 777–785. https ://doi.org/10.1098/ Haldane, J. B. S. (1949). Disease and evolution. La Ricerca Scientifica rsif.2007.0232 (Supplement), 19, 68–76. Ebert, D., Hottinger, J. W., & Pajunen, V. I. (2001). Temporal and Halekoh, U., & Højsgaard, S. (2014). A Kenward‐Roger approximation and spatial dynamics of parasite richness in a Daphnia metapopu‐ parametric bootstrap methods for tests in linear mixed models‐the R lation. Ecology, 82, 3417–3434. https ://doi.org/10.1890/0012‐ package pbkrtest. Journal of Statistical Software, 59, 1–32. https ://doi. 9658(2001)082[3417:TASDO P]2.0.CO;2 org/10.18637/ jss.v059.i09 Ebert, D., Hottinger, J. W., & Pajunen, V. I. (2013). Unsuitable habitat Hämälä, T., Mattila, T. M., & Savolainen, O. (2018). Local adaptation and patches lead to severe underestimation of dynamics and gene flow ecological differentiation under selection, migration, and drift in in a zooplankton metapopulation. Journal of Animal Ecology, 82, 759– Arabidopsis lyrata. Evolution, 72, 1373–1386. https ://doi.org/10.1111/ 769. https ://doi.org/10.1111/1365‐2656.12044 evo.13502 Edwards, K., Johnstone, C., & Thompson, C. (1991). A simple and rapid Hanski, I., & Gaggiotti, O. E. (2004). Ecology, genetics and evolution of method for the preparation of plant genomic DNA for PCR analysis. metapopulations. Amsterdam, The Netherlands: Elsevier Academic Nucleic Acids Research, 19, 1349. https ://doi.org/10.1093/nar/19.6.1349 Press. Ellison, A., Cable, J., & Consuegra, S. (2011). Best of both worlds? Association Hartl, D. L., & Clark, A. G. (1997). Principles of population genetics. between outcrossing and parasite loads in a selfing fish. Evolution, 65, Sunderland, MA: Sinauer Associates. 3021–3026. https ://doi.org/10.1111/j.1558‐5646.2011.01354.x Hedrick, P. W. (1994). Purging inbreeding depression and the probabil‐ Excoffier, L., & Lischer, H. E. L. (2010). arlequin suite ver 3.5: A new series ity of extinction: Full‐sib mating. Heredity, 73, 363–372. https ://doi. of programs to perform population genetics analyses under Linux org/10.1038/hdy.1994.183 and Windows. Molecular Ecology Resources, 10, 564–567. https ://doi. Hedrick, P. W., & Garcia‐Dorado, A. (2016). Understanding inbreed‐ org/10.1111/j.1755‐0998.2010.02847.x ing depression, purging, and genetic rescue. Trends in Ecology and Gandon, S., Capowiez, Y., Dubois, Y., Michalakis, Y., & Olivieri, I. (1996). Evolution, 31, 940–952. https ://doi.org/10.1016/j.tree.2016.09.005 Local adaptation and gene‐for‐gene coevolution in a metapopulation Hendrick, M. F., Finseth, F. R., Mathiasson, M. E., Palmer, K. A., Broder, E. model. Proceedings of the Royal Society B: Biological Sciences, 263, M., Breigenzer, P., & Fishman, L. (2016). The genetics of extreme micro‐ 1003–1009. https ://doi.org/10.1098/rspb.1996.0148 geographic adaptation: An integrated approach identifies a major gene Gandon, S., & Michalakis, Y. (2000). Evolution of parasite virulence underlying leaf trichome divergence in Yellowstone Mimulus guttatus. against qualitative or quantitative host resistance. Proceedings of Molecular Ecology, 25, 5647–5662. https ://doi.org/10.1111/mec.13753 the Royal Society B: Biological Sciences, 267, 985–990. https ://doi. Ilmonen, P., Penn, D. J., Damjanovich, K., Clarke, J., Lamborn, D., org/10.1098/rspb.2000.1100 Morrison, L., … Potts, W. K. (2008). Experimental infection magnifies Gelatt, T. S., Davis, C. S., Stirling, I., Siniff, D. B., Strobeck, C., & Delisle, inbreeding depression in house mice. Journal of Evolutionary Biology, I. (2010). History and fate of a small isolated population of Weddell 21, 834–841. https ://doi.org/10.1111/j.1420‐9101.2008.01510.x seals at White Island, Antarctica. Conservation Genetics, 11, 721–735. Jokela, J., Dybdahl, M. F., & Lively, C. M. (2009). The maintenance of sex, https ://doi.org/10.1007/s10592‐009‐9856‐6 clonal dynamics, and host‐parasite coevolution in a mixed population Goren, L., & Ben‐Ami, F. (2013). Ecological correlates between cladocer‐ of sexual and asexual snails. The American Naturalist, 174, S43–S53. ans and their endoparasites from permanent and rain pools: Patterns https ://doi.org/10.1086/599080 in community composition and diversity. Hydrobiologia, 701, 13–23. Kaltz, O., & Shykoff, J. A. (1998). Local adaptation in host‐parasite systems. https ://doi.org/10.1007/s10750‐012‐1243‐5 Heredity, 81, 361–370. https ://doi.org/10.1038/sj.hdy.6884350 Goslee, S. C., & Urban, D. L. (2007). The ecOdiST package for dissimilarity‐ Karhunen, M., Merilä, J., Leinonen, T., Cano, J. M., & Ovaskainen, O. based analysis of ecological data. Journal of Statistical Software, 22, (2013). drifTSel: An R package for detecting signals of natural selec‐ 1–19. https ://doi.org/10.18637/ jss.v022.i07 tion in quantitative traits. Molecular Ecology Resources, 13, 746–754. Goudet, J. (2001). fSTaT, a program to estimate and test gene diversities https ://doi.org/10.1111/1755‐0998.12111 and fixation indices (version 2.9.3). Journal of Heredity, 86, 485–486. Karhunen, M., & Ovaskainen, O. (2012). Estimating population‐level Haag, C. R., & Ebert, D. (2004). Parasite‐mediated selection in exper‐ coancestry coefficients by an admixture F model. Genetics, 192(2), imental metapopulations of Daphnia magna. Proceedings of the 609–617. https ://doi.org/10.1534/genet ics.112.140871 Royal Society B: Biological Sciences, 271, 2149–2155. https ://doi. Karhunen, M., Ovaskainen, O., Herczeg, G., & Merilä, J. (2014). Bringing org/10.1098/rspb.2004.2841 habitat information into statistical tests of local adaptation in quan‐ Haag, C. R., Riek, M., Hottinger, J. W., Pajunen, V. I., & Ebert, D. (2005). titative traits: A case study of nine‐spined sticklebacks. Evolution, 68, Genetic diversity and genetic differentiation in Daphnia metapopula‐ 559–568. https ://doi.org/10.1111/evo.12268 tions with subpopulations of known age. Genetics, 170, 1809–1820. Karvonen, A., & Seehausen, O. (2012). The role of parasitism in adap‐ https ://doi.org/10.1534/genet ics.104.036814 tive radiations‐when might parasites promote and when might they Haag, C. R., Riek, M., Hottinger, J. W., Pajunen, V. I., & Ebert, D. (2006). constrain ecological speciation? International Journal of Ecology, 2012, Founder events as determinants of within‐island and among‐island 1–20. https ://doi.org/10.1155/2012/280169 genetic structure of Daphnia metapopulations. Heredity, 96, 150– Kato, Y., Matsuura, T., & Watanabe, H. (2012). Genomic integration and 158. https ://doi.org/10.1038/sj.hdy.6800774 germline transmission of plasmid injected into Crustacean Daphnia Haag, C. R., Sakwińska, O., & Ebert, D. (2003). Test of synergistic interac‐ magna eggs. PLoS ONE, 7, e45318. https ://doi.org/10.1371/journ tion between infection and inbreeding in Daphnia magna. Evolution, al.pone.0045318 57, 777–783. https ://doi.org/10.1111/j.0014‐3820.2003.tb002 89.x Kawecki, T. J., & Ebert, D. (2004). Conceptual issues in local Haag, K. L., Larsson, J. I. R., Refardt, D., & Ebert, D. (2011). Cytological adaptation. Ecology Letters, 7, 1225–1241. https ://doi. and molecular description of Hamiltosporidium tvaerminnensis gen. org/10.1111/j.1461‐0248.2004.00684.x

104 105 CABALZAR et AL. | 4783 4784 | CABALZAR et AL.

R package version 0.3‐4. Retrieved from https ://CRAN.R‐proje ct.org/ et sp. nov., a microsporidian parasite of Daphnia magna, and estab‐ Keller, L. F., & Waller, D. M. (2002). Inbreeding effects in wild popu‐ Mattila, A. L. K., Duplouy, A., Kirjokangas, M., Lehtonen, R., Rastas, P., & packa ge=adesp atial lishment of Hamiltosporidium magnivora comb. nov. Parasitology, 138, lations. Trends in Ecology and Evolution, 17, 230–241. https ://doi. Hanski, I. (2012). High genetic load in an old isolated butterfly pop‐ Ebert, D. (2008). Host‐parasite coevolution: Insights from the Daphnia‐ 447–462. https ://doi.org/10.1017/S0031 18201 0001393 org/10.1016/S0169‐5347(02)02489‐8 ulation. Proceedings of the National Academy of Sciences of the United parasite model system. Current Opinion in Microbiology, 11, 290–301. Haag, K. L., Traunecker, E., & Ebert, D. (2013). Single‐nucleotide poly‐ King, K. C., & Lively, C. M. (2012). Does genetic diversity limit disease States of America, 109, E2496–E2505. https ://doi.org/10.1073/ https ://doi.org/10.1016/j.mib.2008.05.012 morphisms of two closely related microsporidian parasites suggest spread in natural host populations. Heredity, 109, 199–203. https :// pnas.12057 89109 Ebert, D., Altermatt, F., & Lass, S. (2007). A short term benefit for out‐ a clonal population expansion after the last glaciation. Molecular doi.org/10.1038/hdy.2012.33 McKean, K. A., Yourth, C. P., Lazzaro, B. P., & Clark, A. G. (2008). crossing in a Daphnia metapopulation in relation to parasitism. Journal Ecology, 22, 314–326. https ://doi.org/10.1111/mec.12126 Klüttgen, B., Dülmer, U., Engels, M., & Ratte, H. T. (1994). ADaM, an arti‐ The evolutionary costs of immunological maintenance and of the Royal Society Interface, 4, 777–785. https ://doi.org/10.1098/ Haldane, J. B. S. (1949). Disease and evolution. La Ricerca Scientifica ficial freshwater for the culture of zooplankton. Water Research, 28, deployment. BMC Evolutionary Biology, 8(1), 76. https ://doi. rsif.2007.0232 (Supplement), 19, 68–76. 743–746. https ://doi.org/10.1016/0043‐1354(94)90157‐0 org/10.1186/1471‐2148‐8‐76 Ebert, D., Hottinger, J. W., & Pajunen, V. I. (2001). Temporal and Halekoh, U., & Højsgaard, S. (2014). A Kenward‐Roger approximation and Krebs, M., Routtu, J., & Ebert, D. (2017). QTL mapping of a natural ge‐ Meaden, S., Paszkiewicz, K., & Koskella, B. (2015). The cost of phage spatial dynamics of parasite richness in a Daphnia metapopu‐ parametric bootstrap methods for tests in linear mixed models‐the R netic polymorphism for long‐term parasite persistence in Daphnia resistance in a plant pathogenic bacterium is context‐dependent. lation. Ecology, 82, 3417–3434. https ://doi.org/10.1890/0012‐ package pbkrtest. Journal of Statistical Software, 59, 1–32. https ://doi. populations. Parasitology, 144, 1686–1694. https ://doi.org/10.1017/ Evolution, 69, 1321–1328. https ://doi.org/10.1111/evo.12652 9658(2001)082[3417:TASDO P]2.0.CO;2 org/10.18637/ jss.v059.i09 S0031 18201 7001032 Merilä, J., & Crnokrak, P. (2001). Comparison of genetic differentiation at Ebert, D., Hottinger, J. W., & Pajunen, V. I. (2013). Unsuitable habitat Hämälä, T., Mattila, T. M., & Savolainen, O. (2018). Local adaptation and Laine, A. L. (2006). Evolution of host resistance: Looking for coevolu‐ marker loci and quantitative traits. Journal of Evolutionary Biology, 14, patches lead to severe underestimation of dynamics and gene flow ecological differentiation under selection, migration, and drift in tionary hotspots at small spatial scales. Proceedings of the Royal 892–903. https ://doi.org/10.1046/j.1420‐9101.2001.00348.x in a zooplankton metapopulation. Journal of Animal Ecology, 82, 759– Arabidopsis lyrata. Evolution, 72, 1373–1386. https ://doi.org/10.1111/ Society B: Biological Sciences, 273, 267–273. https ://doi.org/10.1098/ Mills, S., Lunt, D. H., & Gómez, A. (2007). Global isolation by dis‐ 769. https ://doi.org/10.1111/1365‐2656.12044 evo.13502 rspb.2005.3303 tance despite strong regional phylogeography in a small Edwards, K., Johnstone, C., & Thompson, C. (1991). A simple and rapid Hanski, I., & Gaggiotti, O. E. (2004). Ecology, genetics and evolution of Lande, R. (1976). Natural selection and random genetic drift in metazoan. BMC Evolutionary Biology, 7(1), 225. https ://doi. method for the preparation of plant genomic DNA for PCR analysis. metapopulations. Amsterdam, The Netherlands: Elsevier Academic phenotypic evolution. Evolution, 30(2), 314–344. https ://doi. org/10.1186/1471‐2148‐7‐225 Nucleic Acids Research, 19, 1349. https ://doi.org/10.1093/nar/19.6.1349 Press. org/10.2307/2407703 Moody, K. N., Hunter, S. N., Childress, M. J., Blob, R. W., Schoenfuss, Ellison, A., Cable, J., & Consuegra, S. (2011). Best of both worlds? Association Hartl, D. L., & Clark, A. G. (1997). Principles of population genetics. Lass, S., & Ebert, D. (2006). Apparent seasonality of parasite dynam‐ H. L., Blum, M. J., & Ptacek, M. B. (2015). Local adaptation despite between outcrossing and parasite loads in a selfing fish. Evolution, 65, Sunderland, MA: Sinauer Associates. ics: Analysis of cyclic prevalence patterns. Proceedings of the Royal high gene flow in the waterfall‐climbing Hawaiian goby, Sicyopterus 3021–3026. https ://doi.org/10.1111/j.1558‐5646.2011.01354.x Hedrick, P. W. (1994). Purging inbreeding depression and the probabil‐ Society B: Biological Sciences, 273, 199–206. https ://doi.org/10.1098/ stimpsoni. Molecular Ecology, 24, 545–563. https ://doi.org/10.1111/ Excoffier, L., & Lischer, H. E. L. (2010). arlequin suite ver 3.5: A new series ity of extinction: Full‐sib mating. Heredity, 73, 363–372. https ://doi. rspb.2005.3310 mec.13016 of programs to perform population genetics analyses under Linux org/10.1038/hdy.1994.183 Lazzaro, B. P., & Little, T. J. (2009). Immunity in a variable world. Muñoz, J., Gómez, A., Green, A. J., Figuerola, J., Amat, F., & Rico, and Windows. Molecular Ecology Resources, 10, 564–567. https ://doi. Hedrick, P. W., & Garcia‐Dorado, A. (2016). Understanding inbreed‐ Philosophical Transactions of the Royal Society B: Biological Sciences, C. (2008). Phylogeography and local endemism of the native org/10.1111/j.1755‐0998.2010.02847.x ing depression, purging, and genetic rescue. Trends in Ecology and 364, 15–26. https ://doi.org/10.1098/rstb.2008.0141 Mediterranean brine shrimp Artemia salina (Branchiopoda: Gandon, S., Capowiez, Y., Dubois, Y., Michalakis, Y., & Olivieri, I. (1996). Evolution, 31, 940–952. https ://doi.org/10.1016/j.tree.2016.09.005 Legendre, P., Fortin, M. J., & Borcard, D. (2015). Should the Mantel test Anostraca). Molecular Ecology, 17, 3160–3177. https ://doi. Local adaptation and gene‐for‐gene coevolution in a metapopulation Hendrick, M. F., Finseth, F. R., Mathiasson, M. E., Palmer, K. A., Broder, E. be used in spatial analysis? Methods in Ecology and Evolution, 6, 1239– org/10.1111/j.1365‐294X.2008.03818.x model. Proceedings of the Royal Society B: Biological Sciences, 263, M., Breigenzer, P., & Fishman, L. (2016). The genetics of extreme micro‐ 1247. https ://doi.org/10.1111/2041‐210X.12425 Nei, M. (1987). Molecular evolutionary genetics. New York, NY: Columbia

1003–1009. https ://doi.org/10.1098/rspb.1996.0148 geographic adaptation: An integrated approach identifies a major gene Leinonen, T., McCairns, R. J. S., O'Hara, R. B., & Merilä, J. (2013). QST–FST University Press. Gandon, S., & Michalakis, Y. (2000). Evolution of parasite virulence underlying leaf trichome divergence in Yellowstone Mimulus guttatus. comparisons: Evolutionary and ecological insights from genomic Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, against qualitative or quantitative host resistance. Proceedings of Molecular Ecology, 25, 5647–5662. https ://doi.org/10.1111/mec.13753 heterogeneity. Nature Reviews Genetics, 14, 179–190. https ://doi. D., & Wagner, H. (2019). veGan: Community ecology package. R pack‐ the Royal Society B: Biological Sciences, 267, 985–990. https ://doi. Ilmonen, P., Penn, D. J., Damjanovich, K., Clarke, J., Lamborn, D., org/10.1038/nrg3395 age version 2.5‐4. Retrieved from https ://CRAN.R‐proje ct.org/packa org/10.1098/rspb.2000.1100 Morrison, L., … Potts, W. K. (2008). Experimental infection magnifies Lenormand, T. (2002). Gene flow and the limits to natural selection. ge=vegan Gelatt, T. S., Davis, C. S., Stirling, I., Siniff, D. B., Strobeck, C., & Delisle, inbreeding depression in house mice. Journal of Evolutionary Biology, Trends in Ecology and Evolution, 17, 183–189. https ://doi.org/10.1016/ Ouborg, N. J., Biere, A., & Mudde, C. L. (2000). Inbreeding effects on I. (2010). History and fate of a small isolated population of Weddell 21, 834–841. https ://doi.org/10.1111/j.1420‐9101.2008.01510.x S0169‐5347(02)02497‐7 resistance and transmission‐related traits in the silene‐microbotryum seals at White Island, Antarctica. Conservation Genetics, 11, 721–735. Jokela, J., Dybdahl, M. F., & Lively, C. M. (2009). The maintenance of sex, Little, T. J., & Killick, S. C. (2007). Evidence for a cost of immunity when pathosystem. Ecology, 81, 520–531. https ://doi.org/10.1890/0012‐ https ://doi.org/10.1007/s10592‐009‐9856‐6 clonal dynamics, and host‐parasite coevolution in a mixed population the crustacean Daphnia magna is exposed to the bacterial pathogen 9658(2000)081[0520:IEORA T]2.0.CO;2 Goren, L., & Ben‐Ami, F. (2013). Ecological correlates between cladocer‐ of sexual and asexual snails. The American Naturalist, 174, S43–S53. Pasteuria ramosa. Journal of Animal Ecology, 76, 1202–1207. https :// Ovaskainen, O., Karhunen, M., Zheng, C., Arias, J. M. C., & Merilä, J. ans and their endoparasites from permanent and rain pools: Patterns https ://doi.org/10.1086/599080 doi.org/10.1111/j.1365‐2656.2007.01290.x (2011). A new method to uncover signatures of divergent and stabi‐ in community composition and diversity. Hydrobiologia, 701, 13–23. Kaltz, O., & Shykoff, J. A. (1998). Local adaptation in host‐parasite systems. Lohr, J. N., & Haag, C. R. (2015). Genetic load, inbreeding depression, lizing selection in quantitative traits. Genetics, 189, 621–632. https :// https ://doi.org/10.1007/s10750‐012‐1243‐5 Heredity, 81, 361–370. https ://doi.org/10.1038/sj.hdy.6884350 and hybrid vigor covary with population size: An empirical evalua‐ doi.org/10.1534/genet ics.111.129387 Goslee, S. C., & Urban, D. L. (2007). The ecOdiST package for dissimilarity‐ Karhunen, M., Merilä, J., Leinonen, T., Cano, J. M., & Ovaskainen, O. tion of theoretical predictions. Evolution, 69, 3109–3122. https ://doi. Pajunen, V. I., & Pajunen, I. (2003). Long‐term dynamics in rock pool based analysis of ecological data. Journal of Statistical Software, 22, (2013). drifTSel: An R package for detecting signals of natural selec‐ org/10.1111/evo.12802 Daphnia metapopulations. Ecography, 26, 731–738. https ://doi. 1–19. https ://doi.org/10.18637/ jss.v022.i07 tion in quantitative traits. Molecular Ecology Resources, 13, 746–754. Lynch, M., Conery, J., & Burger, R. (1995). Mutation accumulation and org/10.1111/j.0906‐7590.2003.03542.x Goudet, J. (2001). fSTaT, a program to estimate and test gene diversities https ://doi.org/10.1111/1755‐0998.12111 the extinction of small populations. The American Naturalist, 146, Pearman, P. B., & Garner, T. W. J. (2005). Susceptibility of Italian agile and fixation indices (version 2.9.3). Journal of Heredity, 86, 485–486. Karhunen, M., & Ovaskainen, O. (2012). Estimating population‐level 489–518. https ://doi.org/10.1086/285812 frog populations to an emerging strain of Ranavirus parallels pop‐ Haag, C. R., & Ebert, D. (2004). Parasite‐mediated selection in exper‐ coancestry coefficients by an admixture F model. Genetics, 192(2), MacDougall‐Shackleton, E. A., Derryberry, E. P., Foufopoulos, J., Dobson, ulation genetic diversity. Ecology Letters, 8, 401–408. https ://doi. imental metapopulations of Daphnia magna. Proceedings of the 609–617. https ://doi.org/10.1534/genet ics.112.140871 A. P., & Hahn, T. P. (2005). Parasite‐mediated heterozygote advan‐ org/10.1111/j.1461‐0248.2005.00735.x Royal Society B: Biological Sciences, 271, 2149–2155. https ://doi. Karhunen, M., Ovaskainen, O., Herczeg, G., & Merilä, J. (2014). Bringing tage in an outbred songbird population. Biology Letters, 1, 105–107. Penn, D. J., Damjanovich, K., & Potts, W. K. (2002). MHC heterozygos‐ org/10.1098/rspb.2004.2841 habitat information into statistical tests of local adaptation in quan‐ https ://doi.org/10.1098/rsbl.2004.0264 ity confers a selective advantage against multiple‐strain infections. Haag, C. R., Riek, M., Hottinger, J. W., Pajunen, V. I., & Ebert, D. (2005). titative traits: A case study of nine‐spined sticklebacks. Evolution, 68, Makino, W., & Tanabe, A . S. (20 09). Extreme population genetic differentia‐ Proceedings of the National Academy of Sciences of the United States Genetic diversity and genetic differentiation in Daphnia metapopula‐ 559–568. https ://doi.org/10.1111/evo.12268 tion and secondary contact in the freshwater copepod Acanthodiaptomus of America, 99, 11260–11264. https ://doi.org/10.1073/pnas.16200 tions with subpopulations of known age. Genetics, 170, 1809–1820. Karvonen, A., & Seehausen, O. (2012). The role of parasitism in adap‐ pacificus in the Japanese Archipelago. Molecular Ecology, 18, 3699– 6499 https ://doi.org/10.1534/genet ics.104.036814 tive radiations‐when might parasites promote and when might they 3713. https ://doi.org/10.1111/j.1365‐294X.2009.04307.x Pfeifer, S. P., Laurent, S., Sousa, V. C., Linnen, C. R., Foll, M., Excoffier, Haag, C. R., Riek, M., Hottinger, J. W., Pajunen, V. I., & Ebert, D. (2006). constrain ecological speciation? International Journal of Ecology, 2012, Manangwa, O., De Meeûs, T., Grébaut, P., Ségard, A., Byamungu, L., … Jensen, J. D. (2018). The evolutionary history of Nebraska deer Founder events as determinants of within‐island and among‐island 1–20. https ://doi.org/10.1155/2012/280169 M., & Ravel, S. (2019). Detecting Wahlund effects together mice: Local adaptation in the face of strong gene flow. Molecular genetic structure of Daphnia metapopulations. Heredity, 96, 150– Kato, Y., Matsuura, T., & Watanabe, H. (2012). Genomic integration and with amplification problems: Cryptic species, null alleles and Biology and Evolution, 35, 792–806. https ://doi.org/10.1093/molbe 158. https ://doi.org/10.1038/sj.hdy.6800774 germline transmission of plasmid injected into Crustacean Daphnia short allele dominance in Glossina pallidipes populations from v/msy004 Haag, C. R., Sakwińska, O., & Ebert, D. (2003). Test of synergistic interac‐ magna eggs. PLoS ONE, 7, e45318. https ://doi.org/10.1371/journ Tanzania. Molecular Ecology Resources, 19, 757–772. https ://doi. Plummer, M., Best, N., Cowles, K., & Vines, K. (2006). cOda: Convergence tion between infection and inbreeding in Daphnia magna. Evolution, al.pone.0045318 org/10.1111/1755‐0998.12989 diagnosis and output analysis for MCMC. R News, 6, 7–11. 57, 777–783. https ://doi.org/10.1111/j.0014‐3820.2003.tb002 89.x Kawecki, T. J., & Ebert, D. (2004). Conceptual issues in local Matschiner, M., & Salzburger, W. (2009). Tandem: Integrating automated R Core Team (2016). R: A language and environment for statistical com‐ Haag, K. L., Larsson, J. I. R., Refardt, D., & Ebert, D. (2011). Cytological adaptation. Ecology Letters, 7, 1225–1241. https ://doi. allele binning into genetics and genomics workflows. Bioinformatics, puting. Vienna, Austria: R Foundation for Statistical Computing. and molecular description of Hamiltosporidium tvaerminnensis gen. org/10.1111/j.1461‐0248.2004.00684.x 25, 1982–1983. https ://doi.org/10.1093/bioin forma tics/btp303 Retrieved from http://www.R‐proje ct.org

104 105 CABALZAR et AL. | 4785

Reid, J. M., Arcese, P., & Keller, L. F. (2003). Inbreeding depresses immune escape form herbivores adapted to the parent plant? Evolution, 48, response in song sparrows (Melospiza melodia): Direct and inter‐ 454–464. https ://doi.org/10.2307/2410104 generational effects. Proceedings of the Royal Society B: Biological Van Oosterhout, C., Hutchinson, W. F., Wills, D. P. M., & Shipley, P. Sciences, 270, 2151–2157. (2004). micrO‐checker: Software for identifying and correcting geno‐ Richardson, J. L., Urban, M. C., Bolnick, D. I., & Skelly, D. K. (2014). typing errors in microsatellite data. Molecular Ecology Notes, 4, 535– Microgeographic adaptation and the spatial scale of evolution. Trends 538. https ://doi.org/10.1111/j.1471‐8286.2004.00684.x in Ecology and Evolution, 29, 165–176. https ://doi.org/10.1016/j. Vizoso, D. B., & Ebert, D. (2005). Phenotypic plasticity of host‐ tree.2014.01.002 parasite interactions in response to the route of infec‐ Robinson, J. D., Hall, D. W., & Wares, J. P. (2013). Approximate Bayesian tion. Journal of Evolutionary Biology, 18, 911–921. https ://doi. estimation of extinction rate in the Finnish Daphnia magna metapop‐ org/10.1111/j.1420‐9101.2005.00920.x ulation. Molecular Ecology, 22, 2627–2639. https ://doi.org/10.1111/ Vizoso, D. B., Lass, S., & Ebert, D. (2005). Different mechanisms of mec.12283 transmission of the microsporidium Octosporea bayeri: A cocktail of Rogers, A. R. (1986). Population differences in quantitative characters as solutions for the problem of parasite permanence. Parasitology, 130, opposed to gene frequencies. The American Naturalist, 127, 729–730. 501–509. https ://doi.org/10.1017/S0031 18200 4006699 https ://doi.org/10.1086/284519 Wade, M. J., & McCauley, D. E. (1988). Extinction and recolonization: Roth, O., Ebert, D., Vizoso, D. B., Bieger, A., & Lass, S. (2008). Male‐bi‐ Their effects on the genetic differentiation of local populations. ased sex‐ratio distortion caused by Octosporea bayeri, a vertically Evolution, 42, 995. https ://doi.org/10.2307/2408915 and horizontally‐transmitted parasite of Daphnia magna. International Walser, B., & Haag, C. R. (2012). Strong intraspecific variation in genetic Journal for Parasitology, 38, 969–979. https ://doi.org/10.1016/j. diversity and genetic differentiation in Daphnia magna: The effects ijpara.2007.11.009 of population turnover and population size. Molecular Ecology, 21, Roulin, A. C., Mariadassou, M., Hall, M. D., Walser, J. C., Haag, C., & 851–861. https ://doi.org/10.1111/j.1365‐294X.2011.05416.x Ebert, D. (2015). High genetic variation in resting‐stage production Webster, J. P., & Woolhouse, M. E. J. (1999). Cost of resistance: in a metapopulation: Is there evidence for local adaptation? Evolution, Relationship between reduced fertility and increased resistance in 69, 2747–2756. https ://doi.org/10.1111/evo.12770 a snail schistosome host‐parasite system. Proceedings of the Royal Rousset, F. (1997). Genetic differentiation and estimation of gene Society B: Biological Sciences, 266, 391–396. https ://doi.org/10.1098/ flow from F‐statistics under isolation by distance. Genetics, 145, rspb.1999.0650 1219–1228. Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis. New Routtu, J., & Ebert, D. (2015). Genetic architecture of resistance in York, NY: Springer Verlag. Daphnia hosts against two species of host‐specific parasites. Wielgoss, S., Bergmiller, T., Bischofberger, A. M., & Hall, A. R. (2016). Heredity, 114, 241–248. https ://doi.org/10.1038/hdy.2014.97 Adaptation to parasites and costs of parasite resistance in mutator Roy, C. L., & St‐Louis, V. (2017). Spatio‐temporal variation in prevalence and nonmutator bacteria. Molecular Biology and Evolution, 33, 770– and intensity of trematodes responsible for waterfowl die‐offs in 782. https ://doi.org/10.1093/molbe v/msv270 faucet snail‐infested waterbodies of Minnesota, USA. International Wright, S. (1951). The genetical structure of populations. Annals of Journal for Parasitology: Parasites and Wildlife, 6, 162–176. https ://doi. Eugenics, 15, 323–354. https ://doi.org/10.1111/j.1469‐1809.1949. org/10.1016/j.ijppaw.2017.07.003 tb024 51.x Salathé, P., & Ebert, D. (2003). The effects of parasitism and inbreeding Yeaman, S., & Otto, S. P. (2011). Establishment and maintenance of adap‐ on the competitive ability in Daphnia magna: Evidence for synergis‐ tive genetic divergence under migration, selection, and drift. Evolution, tic epistasis. Journal of Evolutionary Biology, 16, 976–985. https ://doi. 65, 2123–2129. https ://doi.org/10.1111/j.1558‐5646.2011.01277.x org/10.1046/j.1420‐9101.2003.00582.x Zbinden, M., Haag, C. R., & Ebert, D. (2008). Experimental evolution of Schmidt, K. A., & Ostfeld, R. S. (2001). Biodiversity and the dilu‐ field populations of Daphnia magna in response to parasite treat‐ tion effect in disease ecology. Ecology, 82, 609–619. https ://doi. ment. Journal of Evolutionary Biology, 21, 1068–1078. https ://doi. org/10.1890/0012‐9658(2001)082[0609:BATDE I]2.0.CO;2 org/10.1111/j.1420‐9101.2008.01541.x Slatkin, M. (1977). Gene flow and genetic drift in a species subject to fre‐ Zbinden, M., Lass, S., Refardt, D., Hottinger, J., & Ebert, D. (2005). quent local extinctions. Theoretical Population Biology, 12, 253–262. Octosporea bayeri: Fumidil B inhibits vertical transmission in https ://doi.org/10.1016/0040‐5809(77)90045‐4 Daphnia magna. Experimental Parasitology, 109, 58–61. https ://doi. Smallbone, W., van Oosterhout, C., & Cable, J. (2016). The effects of org/10.1016/j.exppa ra.2004.11.002 inbreeding on disease susceptibility: Gyrodactylus turnbulli infection of guppies, Poecilia reticulata. Experimental Parasitology, 167, 32–37. https ://doi.org/10.1016/j.exppa ra.2016.04.018 Spichtig, M., & Kawecki, T. J. (2004). The maintenance (or not) of poly‐ SUPPORTING INFORMATION genic variation by soft selection in heterogeneous environments. The American Naturalist, 164, 70–84. https ://doi.org/10.1086/421335 Additional supporting information may be found online in the Spitze, K. (1993). Population structure in Daphnia obtusa: Quantitative Supporting Information section. genetic and allozymic variation. Genetics, 135, 367–374. Spurgin, L. G., Illera, J. C., Jorgensen, T. H., Dawson, D. A., & Richardson, D. S. (2014). Genetic and phenotypic divergence in an island bird: How to cite this article: Cabalzar AP, Fields PD, Kato Y, Isolation by distance, by colonization or by adaptation? Molecular Ecology, 23, 1028–1039. https ://doi.org/10.1111/mec.12672 Watanabe H, Ebert D. Parasite‐mediated selection in a Stevens, L., Yan, G., & Pray, L. A. (1997). Consequences of inbreeding on natural metapopulation of Daphnia magna. Mol Ecol. invertebrate host susceptibility to parasitic infection. Evolution, 51, 2019;28:4770–4785. https ://doi.org/10.1111/mec.15260 2032–2039. https ://doi.org/10.2307/2411025 Strauss, S. Y., & Karban, R. (1994). The significance of outcrossing in an intimate plant‐herbivore relationship. I. Does outcrossing provide an

106 107