Parasite survey of a Daphnia hybrid complex: host-specificity and environment determine infection



    1. Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland, and Institute of Integrative Biology, ETH Zurich, 8092 Zürich, Switzerland,
    Search for more papers by this author
    • Present address: Indiana University, Department of Biology, 1001 East 3rd Street, Bloomington, Indiana 47405, USA


    1. Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland, and Institute of Integrative Biology, ETH Zurich, 8092 Zürich, Switzerland,
    Search for more papers by this author

    1. CNR Institute for Ecosystem Studies (ISE), 28922 Verbania Pallanza, Italy
    Search for more papers by this author

    1. Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland, and Institute of Integrative Biology, ETH Zurich, 8092 Zürich, Switzerland,
    Search for more papers by this author

Justyna Wolinska, Department of Biology, Indiana University, 1001 East 3rd Street, Bloomington, Indiana 47405, USA. E-mail:


  • 1Hybridization between species is a common phenomenon in plants and animals. If parasite prevalence differs for hybrids and parental species (i.e. taxa) there may be considerable consequences for relative hybrid fitness. Some studies have investigated hybrid complexes for infection, and complex-specific differences in parasite prevalence have been detected.
  • 2Based on the results of a field study on a hybridizing Daphnia population from a single lake, it has been hypothesized that permanently over- or under-infected hybrids do not exist. The observed field-patterns can only be temporal because taxa, in addition to single genotypes, might be the subject of parasite driven host frequency-dependent selection. Thus, parasites will track any common taxon within a hybrid complex.
  • 3In the present study, hybridizing Daphnia populations from 43 lakes were screened for parasite infections to obtain indirect evidence for coevolutionary cycles. It was hypothesized that, due to time lags between the evolution of resistance in host populations and the evolution of the parasite towards tracking of a common host taxon, the same Daphnia taxon will be over-infected in some lakes, while being under-infected in others.
  • 4Two of the four parasite species were specialists: their prevalence differed among coexisting Daphnia taxa. The varying infection patterns detected across spatially segregated hybridizing Daphnia populations are consistent with theoretical predictions for coevolutionary cycles. Thus the infection patterns, as observed under natural conditions, are temporal and unstable.
  • 5Additionally, the spatial distribution of the four parasite species was analysed with respect to habitat differences. The results show that the presence of a particular parasite on a host taxon was determined not only by the host-specificity of the parasite, but also by host-habitat relations.


Hybridization between species is common in plants and animals: at least 25% of plant and 10% of animal species are involved in hybridization (Mallet 2005). Parasite infection patterns, however, have been investigated for only a small fraction of hybridizing species. In two studied animal systems, hybrids were more resistant than parental species to parasites, whereas in seven others, hybrids were either the intermediately or the most infected group (reviewed in Moulia 1999; Jackson & Tinsley 2003; Parris 2004). These differences in infection levels can be incorporated into general theories, dealing with hybrid and parental species’ fitness differences. There are two groups of models. If hybrids are the most infected group, the ‘tension zone models’ are supported (Barton & Hewitt 1985), which generally presuppose low hybrid fitness compared with parental species. On the other hand, when hybrids are less infected than either of their parental species, the ‘bounded hybrid superiority models’ are supported (Moore 1977). The latter models assume that, under specific conditions, hybrids have a higher fitness. Recently, we have criticised such generalizations (Wolinska et al. 2006) because the conclusion of over- or under-infection is often based on data from a single field season (e.g. Coustau et al. 1991). During a four-year field survey of parasite distributions among parental species and hybrids (hereafter taxa) in a Daphnia (waterflea) hybrid complex, we found that over- or under-infection of one taxon may be temporary, thus suggesting that parasites can evolve to infect a previously under-infected taxon (Wolinska et al. 2006).

Although hybridizing taxa from the Daphnia galeata/hyalina/cucullata complex are the most common daphnids in European lakes (reviewed in Schwenk & Spaak 1997), there is a lack of parasite data from this system. In general, studies of parasitism in natural Daphnia populations have been biased towards small and often fishless water bodies (reviewed in Ebert 2005), which are inhabited by nonhybridizing species (D. magna and D. pulex). It has been assumed (Ebert, Payne & Weisser 1997; Ebert 2005) that pond species have more parasites than species living in lakes with strong fish predation pressure (e.g. Daphnia galeata/hyalina/cucullata complex). Parasite spread may be largely limited in such habitats, because infected daphnids are selectively eaten by planktivorous fish, due to their increased opacity (Duffy et al. 2005). Although one can clearly expect different intensities of infection and even different parasite species in Daphnia populations coexisting with planktivorous fish (after Ebert 2005), hybridizing daphnids have only been screened for infection in two lakes (Bittner 2001; Wolinska et al. 2004, 2006).

The importance of parasites in altering the competitive abilities of Daphnia hybrids and parental taxa has been shown for one specific parasite and host population (Wolinska et al. 2006). It is unknown, however, whether this generally applies to various lakes and/or various parasites. For this to be true, at least two assumptions have to be met: the parasite must infect one taxon more than others, and the fitness of infected Daphnia must be substantially reduced. Further, it is unclear if host–parasite interactions in Daphnia hybridizing systems are always dynamic, as has been shown by a four-year field survey of one parasite and host population (Wolinska et al. 2006). Frank (1991) proposed that in addition to a longitudinal survey of one population, a survey of multiple populations sampled simultaneously can be a test for coevolutionary oscillations. If host-parasite coevolutionary cycles operate, one should find the same taxon over-infected in some populations while under-infected or as infected as the remaining taxa in others, because single populations might be by chance in different parts of the oscillatory cycle (see also Lively 1999; Gandon 2002).

In addition to host-parasite specificity, environmental conditions might also alter the outcome of infection. For example, host populations covering larger geographical areas are predicted (Anderson & May 1978, 1979) and shown (e.g. Gregory 1990; Ebert, Hottinger & Pajunen 2001) to harbour more parasite species than ones covering small geographical regions. Specifically for daphnids, parameters like temperature (e.g. Ebert 1995), food quantity (e.g. Ebert, Zschokke Rohringer & Carius 2000) and host density (e.g. Bittner, Rothhaupt & Ebert 2002) have been shown to influence parasite occurrence and spread under controlled laboratory conditions. However, the extent to which these and other conditions matter in natural populations remains unknown.

The main purpose of this study was to determine if the over- or under-infection of hybrids compared with parental taxa (as shown for one parasite and host population, Wolinska et al. 2006) is a common phenomenon in hybridizing Daphnia systems. First, we determined the prevalence of different parasite species in hybridizing Daphnia populations from 43 lakes to investigate if parasites are widespread in this system. Second, we tested if parasites have infection preferences, specifically if any taxon is more infected than would be expected if infections were random. Third, we hypothesized that if host–parasite associations are dynamic, then simultaneous screening across lakes will show that the same taxon is over-infected in some lakes, while under-infected or as infected as the remaining taxa in other lakes. Fourth, to investigate if taxon-specific parasite attacks might alter the outcome of host competition, we determined the fecundity reduction. Finally, we tested which environmental conditions besides the presence of susceptible hosts might influence the distribution of parasites.

Material and Methods

sampling and data collection

This study was carried out in 17 lakes south (Northern Italy) and 26 lakes north (Switzerland) of the Alps (Fig. 1). All lakes are at an altitude below 900 m asl, except for four which are situated in an inner alpine valley (CHAM, MURE, SEGL, SILV; for lake abbreviations see Table 1), above 1760 m asl. To the best of our knowledge, all 43 lakes are inhabited by planktivorous fish. From each lake we collected the following environmental parameters from several published sources: (1) volume (2) maximum depth (3) surface area (4) altitude, and (5) total phosphorus concentration, Ptot (Table 1). In lakes with no published phosphorus concentration, the Ptot was determined from water samples collected by integrative sampling of the upper 20 m. The lakes were sampled for zooplankton both in late spring (Northern Italy: May 2004, Switzerland: May–June 2003) and in autumn (Northern Italy: September 2004, Switzerland: August–October 2003 and August–October 2004). Three lakes were studied more intensively (PFAF – monthly for one year, GREI – biweekly for four years, BRIE – monthly for two years). A 250 µm net (132 µm for Italian lakes) was hauled through the whole water column within a deep basin of each lake to collect a representative zooplankton sample. Samples were stored on ice and analyses were carried out within 24 h to avoid selective mortality of infected individuals.

Figure 1.

Distributions of four Daphnia parasites across 43 lakes inhabited by Daphnia galeata/hyalina/cucullata complex. Pie diagrams indicate presence/absence of the parasites. Numbers refer to lake names (see Table 1).

Table 1.  List with lake abbreviations and corresponding numbers for Fig. 1, and the values of five environmental variables used for PCA analyses. Data sources: (a) Federal Office for the Environment, FOEN; (b); (c) Keller (2003); (d) Liechti (1994); (e) Osservatorio dei Osservatorio dei Laghi Lombardi 2005); (f) Ohlendorf (1998); (g) Elber, Hürlimann & Niederberger (2001); (h) Ludovisi, Pandolfi & Taticchi (2004); (i) Ribi, Bührer & Ambühl (2001); (j) field sampling June 2003, upper 20 m; (k) Nordostschweizerische Kraftwerke, NOK
AbbreviationsCorresponding no (Fig. 1)Lake namePtot (µg L−1)Volume (106 m3)Maximum depth (m)Surface area (km2)Altitude (m asl)Ref. Ptot/hydrography
AGER 1Ägerisee 10  353 83  7·3 724c/d
ALSE 2Lago d’Alserio 26    7  8  1·2 260e, h
BALD 3Baldeggersee 52  173 66  5·2 463a/d
BIEL 4Bielersee 13 1240 74 39·8 429a/d
BRIE 5Brienzesee  3 5170261 29·8 564a/d
CHAM 6Lej da Champfèr 11    8 33  0·51791j/f
COMA 7Lago di Comabbio 72   16  8  3·6 243e/e, h
COMO 8Lago di Como 3522500410145·0 198e/e, h
ENDI 9Lago d’Endine 17   12  9  2·1 334e/e, h
GREI10Greifensee 63  148 32  8·4 435a/d
HALL11Hallwillersee 45  280 47 10·0 449a/d
IDRO12Lago d’Idro 24  684122 11·4 370b/e, h
ISEO13Lago d’Iseo 17 7600251 61·0 186e/e, h
KLON14Klöntalersee 12   56 47  3·3 847j/k
LUGA15Lago di Lugano 47 5860288 48·9 271e/e, h
LUNG16Lungenersee 10   66 68  2·0 689j/i
MAGG17Lago Maggiore 1137500370213·0 194a/e, h
MERG18Lago di Mergozzo  1   83 73  1·8 194b/b, h
MONA19Lago di Monate  5   45 34  2·5 266e/e, h
MONT20Lago di Montorfano  8    2  7  0·5 397e/e, h
MORO21Lago Moro  8    4 42  0·2 389b/e, h
MURE22Lej da S. Murezzan 43   20 44  0·81768f
MURT23Murtensee 27  550 45 22·8 429a/d
NEUE24Neuenburgersee 1413980152217·9 429a/d
ORTA25Lago d’Orta  4 1300143 18·2 290b/b, h
PFAF26Pfäffikersee 22   58 35  3·3 537a/b
PUSI27Lago di Pusiano 74   69 24  4·9 259e/e, h
SARN28Sarnersee  5  244 52  7·6 469a/d
SEGL29Lej da Segel 11  137 71  4·11797f
SEGR30Lago del Segrino 12    1  9  0·4 374b/b, h
SEMP31Sempachersee 34  639 87 14·4 504a/d
SIHL32Sihlsee 13   96 23 10·8 889j/i
SILV33Lej da Silvaplauna 13  126 77  2·71791f
SIRI34Lago di Sirio 24    5 43  0·3 271b/b, h
THUN35Thunersee  3 6470217 48·4 558a/d
TURL36Türlersee 15    6 22  0·5 643g/i
VARE37Lago di Varese 82  160 26 14·8 238e/e, h
VWS38Vierwaldstättersee  711907214113·6 434a/d
WAGI39Wägitalersee 19  149 65  4·2 900j/k
WALE40Walensee  2 3180145 24·1 419a/d
ZH_O41Zürichsee, Obersee 14  467 48 20·3 406a/d
ZH_U42Zürichsee, Untersee 25 3300136 68·1 406a/d
ZUGE43Zugersee114 3174198 38·3 413a/d, i

From each zooplankton sample, 70–100 Daphnia (asexual females) were randomly selected. Only females which were larger than 1·0 mm (measured from the top of the eyespot to the base of the tail spine) were chosen to ensure that the sample was representative of the adult population of all taxa (size at maturity for daphnids from the D. galeata × D. hyalina complex is greater than 1·1 mm, regardless of food conditions Weider 1993). For individuals with the characteristic D. cucullata shape, this lower size range was set at 0·8 mm (Spaak & Hoekstra 1995). Selected Daphnia were screened using a microscope at × 50 magnification for external signs of parasite infection. This ‘nondissecting’ method allows for processing of large samples; however, some parasite species (e.g. microsporidia Ebert 2005) and the earliest stages of infections could have been overlooked. An additional subsample of infected animals (20–80 individuals) was taken to get an adequate number of infected daphnids for statistical tests. If parasites could not be identified to the species level, we classified them into functional groups according to their pathology and morphological traits. Additionally, the presence and number of eggs (i.e. clutch size) were recorded. The animals were frozen in −20 °C for later allozyme analyses.

genetic markers for the host

Daphnia were assayed for four polymorphic allozyme marker loci: phosphoglucose isomerase (PGI, enzyme commission number: EC 5·3·1·9), phosphoglucomutase (PGM, EC 5·4·2·1), aldehyde oxidase (AO, EC 1·2.3·1) and aspartate amino transferase (AAT, EC 2·6·1·1). AO and AAT are diagnostic markers for D. galeata, D. hyalina and D. cucullata (Wolf & Mort 1986; Gießler 1997). The computer program of Anderson & Thompson (2002) was used to calculate the probability that an individual belongs to parental species or various hybrid categories (all four polymorphic loci were used). Daphnia were classified into four groups (taxa): D. galeata, D. hyalina, D. cucullata (i.e. parental species) and the hybrids, which consisted of first and further generations of hybrids pooled with nonidentifiable individuals (identification probability was set at 95% in the program).

data analysis

(i) Distribution of Daphnia parasites

A principal components analysis (PCA) of five log-transformed environmental variables (volume, maximum depth, surface area, altitude, and total phosphorus concentration, ptot, Table 1) was used to relate parasite occurrence to combinations of these variables. To investigate if infected populations were nonrandomly distributed along each component of the PCA, a one-way anova was used with the presence/absence of each parasite as a main factor (as in Bengtsson & Ebert 1998). To prevent bias, anovas were repeated excluding the high altitude lakes (CHAM, MURE, SEGL, SILV).

(ii) comparison of infection level among taxa

Each parasite species was treated separately and the genetic profile (i.e. taxa composition) of infected daphnids was compared to that of noninfected ones: significant differences were determined with an R × C-test of independence (Sokal & Rohlf 1995) for specific lake/date combination (hereafter populations). For two lakes (GREI and PFAF) there were many such defined populations; therefore, a sample taken at a similar time as samples from the other lakes and having the largest number of analysed animals was selected for this and following tests. Within each parasite group, sequential Bonferroni correction (Rice 1989) was used to calculate significance levels for simultaneous statistical tests.

(iii) influence of parasite on host fitness

For all the following tests, individuals with multiple infections were excluded and only populations with at least 20 females per analysed group were considered. Only females larger or equal in size to the smallest gravid female were analysed to ensure that lack of eggs or small population mean body size was not caused by including immature stages. The proportion of gravid females was compared between the infected and noninfected groups. The binomial trait (‘gravid’ or ‘nongravid’) was analysed with a generalized linear model, assuming a binomial error distribution and a logit link-function. Additionally, if a substantial number of the infected females were gravid, the difference in mean clutch size between the infected and noninfected group was analysed with an anova, where infection status (‘infected’ and ‘noninfected’) and population were the main effects. The data were log-transformed to generate a homogenous distribution. All analyses were performed using STATISTICA for Windows, release 6·0 (StatSoft, Inc.).


(i) Distribution of Daphnia parasites

In the 43 lakes studied, we found four common Daphnia parasites (one fungus, one protozoan and two bacteria, see Table 2). The fungus (FUNG) most probably belongs to the genus Saprolegnia but, as this is still under investigation, we will refer to it as unidentified fungus. This parasite infects only the brood of its host: the infected brood is visible as a brown spot in the dorsal region of its host. Caullerya mesnili (CAUL) was previously described by Green (1974) and Bittner et al. (2002). The spore clusters are found inside the gut epithelium. Two bacteria (BACT W and BACT Y) grow in the body cavity of their host and strongly increase host opacity: infected daphnids appear whitish or yellowish, respectively. BACT W was described by Bittner (2001) and Wolinska et al. (2004), whereas BACT Y might be Jahnel's bacterium which was described in Daphnia longispina from the Lunzersee (Austria) (see Green 1974).

Table 2.  Characteristics and prevalence of parasites across all 43 analysed lakes. aVertical transmission was experimentally excluded, but horizontal transmission was not achieved, Tellenbach, unpublished data; bBittner, Rothhaupt & Ebert (2002); cTellenbach, unpublished data
ParasiteAbbreviationTaxonic groupSite of infectionTransmission modeNumber of infected lakesMaximum prevalence (%)
Unidentified fungusFUNGfungusbroodunknowna1530·0
Caullerya mesniliCAULprotozoagut wallhorizontalb1641·0
Unidentified bacteriumBACT Wbacteriabody cavityhorizontalc2262·5
Unidentified bacteriumBACT Ybacteriabody cavityunknown 3 6·3

Parasite distributions across the lakes are shown in Fig. 1. With the exception of BACT Y, which was found in only three lakes, the other parasites were present in more than one third of the studied lakes. BACT W was the most common parasite; it was found in 22 lakes and reached the highest prevalence (62·5%, Table 2). Within any given lake, FUNG, BACT W and BACT Y infections were detected either in late spring or in autumn whereas CAUL infected daphnids were found only in autumn samples. Different parasites often co-occurred in time. While co-occurrence was found in 38% of samples from lakes where at least two parasites were ever detected, multiple infections of single individuals were rare (only in 1·3% of all infected individuals at the time of parasite co-occurrence).

Two main principal components were extracted from the original five variables. The first component PC1 (explaining 57% of the variation) mainly reflected a gradient in lake size, with larger lakes scoring more negative values. The second principal component PC2 (explaining 23% of the variation) was mainly an axis reflecting trophic status (higher Ptot – more negative values) and altitude (higher altitude, more positive values). The size of the habitats (PC1) seems not to have any influence on parasite abundance (Table 3, Fig. 2). However, lakes with or without parasites differed in their distributions along variables associated with PC2, and the differences were significant for FUNG and BACT W parasites. Lakes with these parasites are on average more eutrophic and are situated at lower altitudes (Table 3, Fig. 2). When four high altitude lakes (see Fig. 2) were excluded from the tests, differences in distributions along PC2 axis remained significant for FUNG (Table 3).

Table 3.  One-way anova of the effects of principal components (PC1 and PC2) on the presence or absence of Daphnia parasites in 43 studied lakes. ‘Mean infected’ and ‘mean noninfected’– mean values of principal components for lakes where the parasite was present or absent, respectively. *P < 0·05, **P < 0·01. aAnalyses repeated without the four high altitude lakes (see text and Fig. 2)
ParasiteNumber of lakesVariableFMean (SE)
 PC29·17**−0·62 (0·25)+ 0·32 (0·18)
 PC2a6·10*−0·62 (0·25)+ 0·15 (0·19)
BACT W2221PC10·57  
 PC26·35*−0·38 (0·21)+ 0·39 (0·22)
 PC2a2·82−0·38 (0·21)+ 0·15 (0·24)
BACT Y 340PC10·11  
Figure 2.

Distributions of four Daphnia parasites across 43 lakes, plotted along two principal components extracted from the original five variables. Black dots: lakes with parasite; white dots: lakes without parasite; white squares: high altitude lakes, all without parasite (see Table 3).

(ii) comparison of infection level among taxa

In some populations parasite prevalence was too low to collect a sufficient number of infected individuals: data of taxa distributions among infected and noninfected groups were collected for 8 (FUNG), 11 (CAUL), 7 (BACT W) and 3 (BACT Y) infected Daphnia populations (Fig. 3). For some of these populations (0, 3, 1 and 1, respectively, Fig. 3) among taxa differences in the infection level could not be assessed because the entire population (i.e. random sample) consisted of predominantly one taxon (> 95%). For BACT W and BACT Y infected populations, taxa were homogenously distributed between infected and noninfected groups (Fig. 3). Significant (after sequential Bonferroni correction) differences in genetic profile between the infected and noninfected groups were found for 3 (FUNG) and 5 (CAUL) analysed populations. The infection model differed among the studied populations. For example, hybrids were two times over-infected by CAUL, three times under-infected, and three times as infected as their parentals (Fig. 3). With respect to infection patterns of CAUL in Greifensee (GREI), detailed data have been published elsewhere (Wolinska et al. 2004, 2006).

Figure 3.

Differences in genetic profile between the infected (left bar) and noninfected (right bar) part of the Daphnia population for lakes in which substantial infections were found. Groups were compared with R × C-test, separately for each parasite and for each lake/date combination. Star (*) indicates a significantly over-infected taxon. If differences in genetic profile remained significant after sequential Bonferroni correction (done within each parasite group), over-infected taxa are denoted as (**). Cases which are strike-through could not be analysed because the entire population (i.e. random sample) consisted of predominantly one taxon (> 95%). aSampling dates were randomly selected from a larger data set (see text).

(iii) influence of parasite on host fitness

All parasites significantly reduced fecundity of Daphnia hosts (Table 4), and only females infected with FUNG are capable of recovering from infection and producing healthy clutches thereafter (Tellenbach, unpublished data). The reduction in fecundity was consistent among populations, except BACT Y, for which a significant infection × population interaction was detected (Table 4). The clutch size of the infected and noninfected gravid females could be compared only for BACT W because for the other parasites, too few infected females were gravid. The clutch sizes of BACT W infected gravid females were reduced by 25% compared with those of noninfected females (Table 4).

Table 4.  Influence of parasite on host fitness. The ‘egg presence’ variable was analysed using a binomial error distribution. The cell entries show Wald χ2-values and significance levels. ‘Clutch size’ was analysed with anova: F-values and significance levels are shown. D.f. for infection × population interactions are provided. FUNG infected daphnids were excluded from both tests because, in contrast to other parasites, they can recover and produce healthy clutches thereafter (see text). Analysis of ‘clutch size’ reduction was only possible for BACT W, because only small fractions of females infected by other parasites were gravid. *P < 0·05, **P < 0·01
ParasiteTraitInfectionPopulationInfection × populationD.f.% with eggs% of trait reduction
CAULegg presence252·90**11·3910·5913 3·768·294·6
BACT W  63·04**95·63**10·84 948·776·436·2
BACT Y  35·40** 0·52 6·20* 110·786·287·6
      mean clutch size 
BACT Wclutch size 33·08** 9·70** 1·68 7 1·95 2·6025·0


Daphnia parasites were found in 30 out of 43 hybridizing Daphnia populations (Fig. 1). Three of four detected parasites were common and found at high prevalence (Table 2) which contrasts with the hypothesis that parasite prevalence in Daphnia populations coexisting with planktivorous fish is limited (Ebert et al. 1997; Ebert 2005). Moreover, three parasites were observed that substantially increase the opacity of their host (FUNG, BACT W and BACT Y), contrasting with the belief that parasites that persist in lakes with visually hunting fish, should not make their host more conspicuous (Ebert 2005). The persistence of such parasites might be possible because infected Daphnia hide from fish in the dark hypolimnion during the day (see Fels, Lee & Ebert 2004; Hall et al. 2005). All four parasites significantly reduced fecundity of infected Daphnia (Table 4). BACT W, the least harmful but most common parasite, reached the highest prevalence (Table 2) and, in contrast to the second common one (CAUL), was detected consistently in both late spring and autumn samples. This finding is in agreement with a parasite survey in Daphnia pond populations where the least harmful parasites were the most prevalent and had the highest persistence time (Decaestecker et al. 2005). Also, epidemiological models predict that low virulence infections are more likely to reach high prevalence and widespread distributions (Anderson & May 1979; Gandon, Agnew & Michalakis 2002).

For each of the four parasites, the genetic profile was compared between infected and noninfected daphnids to test the null hypothesis that Daphnia parasites are generalists (i.e. all taxa are infected at random). Looking across the analysed lakes, all parasites were able to infect all Daphnia taxa (Fig. 3). However, when we take the spatial aspect into account, our results show clearly that FUNG and CAUL are specialists in the sense that in a given lake they do not infect all Daphnia taxa to the same extent (Fig. 3). Given the high number of populations analysed with BACT W, this parasite clearly appears as a generalist (Fig. 3). Although one could argue that classifying parasite species based on pathology and morphology may not be sufficient, the effects of the parasites were consistent enough among studied populations to lead us to believe that we were indeed dealing with the same parasite species across all lakes. The only significant infection × population interaction was detected for BACT Y (Table 4), but this parasite was found in too few populations to speculate about its specificity pattern.

Across lakes where FUNG or CAUL were present, the same taxon (e.g. hybrids) was over-infected in some lakes, while it was under-infected, or infected to the same degree as the remaining taxa in others (Fig. 3). Moreover, we found significantly over-infected taxa to cover the full range of frequencies (Fig. 3). A similar pattern, but on a single-genotype level, was previously observed in a longitudinal survey of pond Daphnia populations. Little & Ebert (1999) found that genotypes oscillated between over- and under-infection and the outcome of infection was independent of the frequency of a particular clone at a particular time. The authors argued that this inconsistency is due to the time lag (Clarke 1976) between the evolution of resistance in hosts and counter-adaptation in parasites. Theoretical predictions are that a diverse pattern may arise across spatially segregated populations because single populations are, by chance, in different parts of the oscillation cycle (Frank 1991; Lively 1999; Gandon 2002). Although these models consider the fluctuations of genotypes within population of single species, there is evidence that parasites may maintain species diversity within a community due to similar processes (reviewed in Clay, in press).

An alternative interpretation of the obtained pattern is while the same taxon is over-infected under some conditions, different environmental settings may lead to its under-infection (genotype–by–environment interactions, e.g. Mitchell et al. 2005). Since the studied lakes have different environmental conditions (Table 1, Fig. 2), it is difficult to be certain that coevolution is solely responsible for the observed spatial variation. However, the infection patterns were observed to change over time within the same population (PFAF unpublished data, GREI Wolinska et al. 2006), and these changes occurred regardless of the lakes’ seasonal cycles. We believe therefore that the infection patterns we examined over several populations (across space), might be evidence for taxa variation emerging over several generations in a single population (across time). The environment might additionally produce asynchrony in coevolutionary cycles by causing phase differences between the trajectories of host and parasite gene frequencies. For example, parasite virulence and transmission rate, which are key parameters in coevolutionary models (e.g. Anderson et al. 1979; Gandon 2002) are considerably altered by minute changes in temperature (reviewed in Thomas & Blanford 2003).

The outcome of coevolution might additionally depend on the scale to which both players migrate between populations. If the subdivided populations exchange genes in heterogeneous environments, this may lead to complex interactions between intra- and interpopulation processes over space and time (the geographical mosaic theory by Thompson 1994). Thus, asynchrony can be maintained in the long term only if the homogenizing effect of migration is decoupled (e.g. Gandon 2002). Particularly in our system, the metapopulation dynamics should not interfere strongly with local selection processes. It is known that Daphnia adaptations to local conditions strongly reduce effective gene flow (reviewed in De Meester et al. 2002). Moreover, the dispersal rate of parasites seems to be limited because none of them are vertically transmitted and therefore cannot disperse through daphnids diapausing eggs (i.e. the most dispersal Daphnia stage, De Meester et al. 2002). Even if parasite cysts are directly carried by insects or birds, such transmission paths are rather unlikely to provide high infective doses, which are often necessary for successful transmissions (e.g. Ebert et al. 2000).

In contrast to the opinion of Price (1990) that the distribution of parasites is mainly determined by their host ranges, parasites were not detected in all of the lakes inhabited by potential hosts. The geographical distribution of CAUL and BACT Y were not explained by any of the studied variables, whereas FUNG and BACT W occurred more often in eutrophic and low altitude lakes (PC2: Table 3, Fig. 2). In oligotrophic lakes, food is often a limiting factor (Lampert & Sommer 1999) and growth of Daphnia parasites is often reduced in hosts kept on low-food diets (e.g. Ebert et al. 2000; Bittner et al. 2002). Furthermore, starvation induces a high mortality among heavily infected daphnids (Pulkkinen & Ebert 2004). Food stress of the host can inhibit parasite populations from spreading and can jeopardise their persistence. Lake trophy is also a good predictor of Daphnia density (e.g. Keller et al. 2002), and both epidemiological models (e.g. Anderson & May 1978) and empirical data from various systems (e.g. Arneberg et al. 1998; Takemoto et al. 2005) have shown a positive relationship between host density and the abundance of directly transmitted parasites. When Daphnia density is low, as found under low trophic conditions (Keller et al. 2002), the contact rate between infected and noninfected individuals is limited and thus, the impact of disease is expected to be diminished. This theory is strongly supported by the fact that over a two-year observation period, we have not found any parasites in Brienzersee (BRIE) (Fig. 1), an ultra-oligotrophic lake (Table 1) with extremely low Daphnia densities (Rellstab, unpublished data). An additional determinant of parasite geographical distributions could be water temperature as higher temperatures are often related to higher infection levels (reviewed in Mouritsen & Poulin 2002). At low temperatures, filtration rate of cladocerans is weakened (Mourelatos & Lacroix 1990), which might in turn reduce the probability of parasite uptake, spread, and persistence in the case of waterborne diseases (discussed in Ebert 2005). For example, the growth and transmission of a Daphnia microsporidium parasite is greatly impaired below 12 °C (Ebert 1995). This seems to be a general pattern because FUNG and BACT W parasites were more abundant in lower altitude lakes (PC2: Table 3, Fig. 2) and thus at higher water temperatures (habitat altitude is negatively correlated with water temperature, e.g. Livingstone & Lotter 1998). It is striking that no parasites were found in the four high altitude lakes (Fig. 2). For FUNG, the changes in altitude were even important in a range of midland lakes; differences in the distribution of infected and noninfected lakes along PC2 axis remained significant after removing the four alpine lakes from the test (Table 3).

In contrast to predictions that host populations covering larger geographical areas harbour more parasite species than ones covering small geographical regions (Anderson et al. 1978, 1979), Daphnia parasites were abundant across a wide range of lake sizes (PC1: Table 3, Fig. 2). In a comparable study of Daphnia from rockpool habitats, the presence of parasites was correlated positively with pool volume (Ebert et al. 2001). However, these rockpools were much smaller (0·006–27·9 m3) and did not differ in trophic status (Ebert et al. 2001). Since the present survey included large, but oligotrophic (e.g. VWS, THUN, BRIE), and small, but eutrophic, lakes (e.g. COMA, MURE, PUSI, see Table 1) the number of potential hosts cannot be simply correlated with lake size. Rather, the trophic status seems more important and diminishes any potential effect of habitat area.


We investigated the effect of parasites on hybridizing Daphnia populations from 43 lakes. At the genetic level, we confirmed that nonrandom infections of coexisting Daphnia taxa are not restricted to one particular parasite or lake (Wolinska et al. 2004, 2006), but are indeed a common phenomenon. Our spatial survey showed that the same taxon was differentially infected across lakes, which is consistent with theoretical predictions about the outcome of an infection if host and parasites are in coevolutionary cycles. Until now, parasite driven time-lagged negative frequency-dependent selection was considered to be operating on a single-genotype level, whereas our results show that dynamic interactions might function also for taxa. Thus, infection levels of different taxa in the Daphnia galeata/hyalina/cucullata species complex cannot be generalized based on the data from individual host populations sampled at single time periods. The relationship between resistance and the response to selection may be additionally altered by environmental parameters of the habitat itself. Host–parasite interactions in hybridizing systems were so far thought to be stable as hybrids were seen to be less, more or intermediately susceptible compared to their parental taxa (reviewed in Moulia 1999), whereas our results show that host–parasite associations in a Daphnia hybridizing system might be only temporal due to the evolution of the parasite as well as to variable environments. It could be that such a dynamic scenario may be valid for other hybridizing systems, especially if parental taxa and hybrids coexist in sympatry, and parasites potentially have equal access to all taxa; however, this hypothesis has yet to be tested.


We thank Christian Rellstab for providing us with zooplankton samples from Brienzersee and Esther Keller and Andrea Ferrari for collecting zooplankton samples from all other lakes. Colleen Durkin, Piotr Madej and Christoph Tellenbach helped us run the allozyme electrophoresis. Dieter Ebert, Britt Koskella and two anonymous reviewers provided valuable comments which helped us to improve this manuscript. We thank Larry Weider and Kayla King for linguistic help.