Experimental evolution of local parasite maladaptation


  • S. ADIBA,

    1. UPMC Univ. Paris 06, Laboratoire de Fonctionnement et Evolution des Systèmes Ecologiques, CNRS-UMR 7625, Paris, France
    Search for more papers by this author
  • M. HUET,

    1. UPMC Univ. Paris 06, Laboratoire de Fonctionnement et Evolution des Systèmes Ecologiques, CNRS-UMR 7625, Paris, France
    Search for more papers by this author
    • 1

      Present address: Station d’Ecologie Experimentale CNRS USR 2936, 09200 Moulis, France.

  • O. KALTZ

    1. UPMC Univ. Paris 06, Laboratoire de Parasitologie Evolutive, CNRS-UMR 7103, Paris, France
    Search for more papers by this author

Oliver Kaltz, Institut des Sciences de l’Evolution, UMR 5554, Université Montpellier 2, Place Eugène Bataillon, 34095 Montpellier Cedex 05, France.
Tel.: 33 04 67 14 40 63; fax: 33 04 67 14 40 61;
e-mail: oliver.kaltz@univ-montp2.fr


Sign and magnitude of local adaptation in host–parasite systems may vary with ecological, epidemiological or genetic parameters. To investigate the role of host genetic background, we established long-term experimental populations of different genotypes of the protozoan Paramecium caudatum, infected with the bacterial parasite Holospora undulata. We observed the evolution of an overall pattern of parasite local maladaptation for infectivity, indicating a general coevolutionary disadvantage of this parasite. Maladaptation extended to host populations with the same genetic background, similar to extending from the local to a higher regional level in natural populations. Patterns for virulence were qualitatively similar, but with less statistical support. A nonsignificant correlation with levels of (mal)adaptation for infectivity suggests independent evolution of these traits. Our results indicate similar (co)evolutionary trajectories in populations with different genetic backgrounds. Nonetheless, the correlated clines of genetic distance and parasite performance illustrate how genetic background can shape spatial gradients of local adaptation.


Coevolution between hosts and parasites can shape the spatial distribution of genetic variation within and among natural populations (Hamilton, 1980; Bell, 1982; Thompson, 2006). A prominent spatial pattern is that of local adaptation of parasites to their hosts (or vice versa). This pattern can be the signature of ongoing antagonistic coevolution (Bell & Maynard Smith, 1987; Nee, 1989; Woolhouse et al., 2002; Gandon et al., 2008), and it arises if one of the two players evolves more rapidly than the other. It is commonly assumed that parasites have a higher evolutionary potential because of shorter generation time and larger population sizes, allowing them to adapt to locally most common host genotypes (Haldane, 1949; Hamilton et al., 1990; Chaboudez & Burdon, 1995; Lively & Dybdahl, 2000). Local parasite adaptation can thus be defined as parasite genotypes on their local (sympatric) hosts having a higher fitness than genotypes on foreign (allopatric) hosts (Ebert & Hamilton, 1996; Gandon & Van Zandt, 1998; Kaltz & Shykoff, 1998; Mopper & Strauss, 1998).

Over the past years, the issue of local adaptation has been addressed in theoretical (Morand et al., 1996; Lively, 1999; Gandon, 2002; Gandon & Michalakis, 2002; Nuismer, 2006; Gandon & Nuismer, 2009) and empirical (Kaltz & Shykoff, 1998; Lajeunesse & Forbes, 2002; Greischar & Koskella, 2007; Hoeksema & Forde, 2008) studies. Consistent with conventional wisdom, meta-analyses revealed a general trend of local parasite adaptation over various host–parasite systems (Greischar & Koskella, 2007; Hoeksema & Forde, 2008). However, it has also been shown that sign and magnitude of this pattern vary with relative host and parasite migration rates (Greischar & Koskella, 2007; Hoeksema & Forde, 2008), host range (Lajeunesse & Forbes, 2002) or host breeding system (Van Zandt & Mopper, 1998). Further, local adaptation is more frequent for traits describing the probability of infection (infectivity, resistance) than for traits related to parasite proliferation or host damage (infection intensity or virulence) (Greischar & Koskella, 2007; Hoeksema & Forde, 2008), possibly because the two types of traits are under different modes of selection (Dybdahl & Storfer, 2003). Thus, parasite local adaptation is not the universal rule and life history, demographic and ecological factors as well as the genetic architecture of the underlying traits may change patterns of local adaptation from system to system.

Variation in these parameters may not only explain differences among host–parasite systems, but also produce variable outcomes within a given system. In simple formalizations of local adaptation, populations harbour identical sets of host and parasite alleles, but at different relative frequencies, because of asynchrony in the coevolutionary dynamics (Kaltz & Shykoff, 1998; Gandon, 2002). More realistically, however, spatially structured populations may historically differ in genetic background or experience variation in environmental conditions, stochasticity and gene flow. Such contingencies may alter fitness functions determining the intensity of reciprocal selection and thereby generate a geographic mosaic of coevolutionary outcomes (Thrall & Burdon, 1997; Hochberg & van Baalen, 1998; Thompson, 1999; Gomulkiewicz et al., 2007). Thus, over the geographic range of the interacting species, parasites may be locally adapted in some areas, but not in others (Nuismer, 2006; Gandon & Nuismer, 2009). With spatial structure, we may further expect clines of parasite performance on hosts from increasingly distant (i.e. genetically dissimilar) populations (Ebert & Hamilton, 1996; Kaltz et al., 1999; Hoeksema & Thompson, 2007); depending on the precise shape of these clines, local adaptation may only be detectable at certain spatial scales (population, region, etc.) (Thrall et al., 2002; Laine, 2005).

Microbial host–parasite systems can be used to reproduce coevolution in the laboratory and to test ideas about the geographic mosaic model of coevolution. Experimental manipulation of environmental and demographic factors has been shown to generate variable (co)evolutionary responses in host and parasite defence traits and differences in patterns of local adaptation (Forde et al., 2004, 2007; Morgan et al., 2005; Lopez-Pascua & Buckling, 2008; Gallet et al., 2009). Typically, these studies are controlled for genetic background in that replicate populations initially have the same genetic composition. To our knowledge, there are no studies comparing the patterns arising from coevolving populations with different initial genetic backgrounds. From a geographic mosaic point of view, we may consider different genetic backgrounds as spatially isolated and genetically distinct subpopulations. We can then ask two questions. First, are patterns of local adaptation repeatable for different (initial) genetic backgrounds of parasite and/or host? Second, does genetic background shape the spatial structure of local adaptation, by producing clines of parasite performance on hosts from populations with the same or different genetic background, analogous to adaptation at a higher regional level?

We investigated theses questions, using the freshwater protozoan Paramecium caudatum and its bacterial parasite Holospora undulata. Replicate populations of different host genetic backgrounds (i.e. host clones) were allowed to freely coevolve with the parasite for 4 years (≥ 400 generations) in the laboratory. Previous experiments showed evidence for reciprocal coevolution and for parasite local maladaptation (Lohse et al., 2006; Nidelet & Kaltz, 2007). However, it remained an open question whether this parasite disadvantage was repeatable over the different host genetic backgrounds. Therefore, a first goal of the present study was to test for the generality of local parasite maladaptation, by using a larger number of replicate populations and by replicating the question of maladaptation at the level of host genetic background. If (co)evolutionary trajectories depend on the starting point, i.e. the identity of the host clones, patterns of local (mal)adaptation may arise for some host clones, but not for others. Second, we tested whether genetic background generated a cline of (mal)adaptation for infectivity, from local populations (sympatric combinations of host and parasite) to foreign populations with the same host genetic background (near-allopatric combinations) to foreign populations with different backgrounds (far-allopatric combinations). Such clines of adaptation are expected if (co)evolutionary trajectories are more similar for host populations with the same genetic background. Third, we also measured local adaptation in virulence. This parasite is transmitted both vertically and horizontally, and thus measuring only the horizontal component of adaptation (infectivity) may provide an incomplete picture. We were particularly interested in the correlation of local adaptation in these two traits.

Materials and methods

Study organisms

Like all ciliates, paramecia have two nuclei: the diploid micronucleus and the highly polyploid macronucleus (Wichtermann 1986; Görtz 1988). The micronucleus is capable of mitosis and meiosis; it is mainly active during the sexual part of the life cycle. During the asexual part of the cycle, reproduction occurs through mitotic cell division, and most gene expression occurs in the macronucleus. Our long-term selection lines were kept as asexual clones and thus responses to selection likely occurred in the macronucleus.

The micronucleus-specific H. undulata is a gram-negative alpha-proteobacterium (Görtz & Brigge, 1998; Fokin, 2004; Fujishima, 2009). Infection can occur when paramecia ingest immobile infectious forms (15–20 μm) of the parasite during food uptake. The parasite then mediates its transport from the digestive vacuole to the micronucleus. Once in the micronucleus, the infectious form differentiates into the short reproductive forms (5–10 μm). After 7–10 days of multiplication of reproductive forms inside the micronucleus, reproductive forms begin to differentiate into infectious forms, which are released into the medium upon cell division of the host or when the host dies. Thus, infectious forms are the agents of horizontal transmission, whereas reproductive forms are vertically transmitted at each mitotic division of the host. Infection reduces host division and survival, in particular when infectious forms accumulate in the micronucleus (Restif & Kaltz, 2006).

Long-term experiment

In 2002, we used mass cultures of nine paramecium clones (genotypes) to establish infected and uninfected replicate populations. The majority of these clones (all with a ‘K’-suffix) had been established from matings between two uninfected stocks of opposite mating type, provided by T. Watanabe (Tohoku University, Japan). Other clones were of German origin and provided by H.-D. Görtz (Universität Stuttgart, Germany). To establish infections, parasites were isolated from a highly infected mass culture of a single host clone (provided by H.-D. Görtz, Universität Stuttgart, Germany); thus genetic variation may have existed in the parasite founder population.

The replicate populations were kept in 50 mL culture medium in 50-mL Falcon tubes (Fisher Scientific, Elancourt, France) at 23 °C. The culture medium was based on organically grown lettuce, inoculated with the bacterium Serratia marcescens (Strain A173; Institut Pasteur, Paris, France) as a food resource for the paramecia (Lohse et al., 2006). In 2-week intervals, we applied 20% nonselective mortality, by replacing 10 mL from each tube (medium and paramecia) with fresh medium, but otherwise did not interfere with any selective process. This protocol maintained cultures almost constantly near carrying capacity, at densities of 50–150 individuals per mL. We do not know the exact population turn-over rate under these conditions; we assume that the average individual produces 2–3 divisions per week, which gives approximately 400 generations over the course of the long-term experiment.

We had started 3–7 replicate populations per paramecium clone and infection status, but some populations went extinct over the years or lost the parasite. Thus, for the local adaptation assays in 2006, we used 24 infected replicate populations from seven host clones, with 2–4 replicate populations per clone. Evidence for local maladaptation was found in a previous assay on four replicate populations, each with a different clonal background (Lohse et al., 2006).

Local adaptation assay: infectivity

Experimental design

For a cross-infection experiment, we isolated up to four uninfected paramecia from each long-term replicate population and grew these up as clonal populations (referred to as ‘subclones’, hereafter, Fig. 1). Parasite inocula prepared from each replicate population were then confronted with replicates of the subclones from their own population (sympatric hosts), with subclones from foreign replicate populations with the same clonal origin (K2, K3, etc.) as the sympatric population (near-allopatric hosts) or with subclones from foreign populations from different clones (far-allopatric hosts; Fig. 2). Thus, near-allopatric hosts should be genetically more similar to sympatric hosts than far-allopatric hosts. A given parasite population was tested against one subclone per near-allopatric host population and one subclone per far-allopatric host clonal background (Fig. 1).

Figure 1.

 Experimental design. (a) Preparation of host and parasite material for the cross-infection experiment. From each long-term replicate population, up to four uninfected individuals were isolated and grown up as new clonal mass cultures (‘subclones’); individuals from these subclone cultures were used to set up assay tubes in the cross-infection experiment. Infectious forms were isolated from each long-term population and used to inoculate the assay tubes. (b) Details of the experimental setup at the replicate population/subclone level. A given parasite population (here illustrated for K3-A), was tested against all available sympatric subclones (K3-A). It was further tested against all near-allopatric populations (K3, B-D), but only against one arbitrarily chosen subclone per population. Finally, it was tested against one arbitrarily chosen subclone from one far-allopatric population per host clone background (K6-B, K8-C, etc.). Overall, the different subclones from the same host population, as well as the different host populations with the same clonal background, were used equally frequently. Note that not for all host populations four subclones could be established (see main text and Fig. 2).

Figure 2.

 Sympatric and allopatric combinations host and parasite in the cross-infection experiment. Parasite replicate populations coevolving with hosts of different clonal origin (K2, K3, etc) were confronted with hosts from their own population (sympatric combinations, black squares), with hosts from foreign populations with the same genetic origin (near-allopatric combinations, dark grey) and with hosts from foreign populations with foreign genetic origin (far-allopatric combinations, light grey). Numbers show the replicates for a given combination. Only the parasite populations used for analysis are represented. Six parasite populations were omitted because of missing data (i.e. data were not available for all sympatric/allopatric categories).

Inoculation protocol

A sample from each infected replicate population was filtered through one layer of medical gauze and then concentrated by centrifugation at 2500 g (30 min). Dimethyl-sulfoxide (DMSO; concentration: 5%) was added to kill the paramecia and release the infectious forms of the parasite. In two rounds of centrifugation, inocula were washed with sterile Volvic mineral water (Danone, Puy de Dome, France) to dilute the DMSO. An assay tube in the cross-infection experiment consisted of a 2-mL plastic tube, containing 100 individuals of a given host subclone in 300 μL of medium, to which 300 μL of freshly prepared inoculum was added (≈104 infectious forms). Two days after inoculation, 40–50 individuals were fixed with lacto-aceto-orcein (Görtz & Dieckmann, 1980), and the proportion of infected individuals was determined at 1000× magnification (phase contrast).

The experiment was spread out over three different days (= experimental blocks), with the different clonal origins of host and parasite populations distributed equally over blocks.

Local adaptation assay: virulence

For measurements of virulence, we recorded growth curves of small clonal populations started from single infected individuals. One month after inoculation, infected individuals from the cross-infection assay tubes were isolated under a dissecting microscope (25×) and placed singly in 200 μL of growth medium on 96-well plates (Nunc™; Fisher Scientific). Cell density in each well was recorded on day 2, 7, 10, 14 and 22. Uninfected control cell lines were established and followed in parallel. Low infection success in the cross-infection experiment limited replication in this new experiment. We succeeded in isolating 5–8 individuals per parasite replicate population and type of combination (sympatric, near-, far-allopatric). On average, we set up 3.6 (±0.4) uninfected individuals per subclone and 10.6 (±1.3) per host replicate population. This virulence assay was also split into three experimental blocks.

Data analysis


Infectivity was taken as the proportion of infected hosts on day 2 after inoculation (e.g. Lohse et al., 2006; Nidelet & Kaltz, 2007). Only 60 of the 96 host subclones envisaged for the full experimental design produced sufficient numbers of individuals for inoculation. Therefore, certain host–parasite combinations could not be tested. To obtain a fully balanced data set for analysis, we proceeded as follows. First, we only considered parasite replicate populations that were tested against all three types of hosts (sympatric, near-allopatric, far-allopatric). This reduced data set represented 18 parasite replicate populations from six clonal origins and paramecium subclones from 21 replicate populations and seven clonal origins (total N = 215 assay tubes; Fig. 2). Second, we combined the proportion of infected hosts per parasite replicate population and type of combination (sympatric, near- or allopatric; N = 18 × 3 = 54). We carried out an Analysis of Deviance to test for effects of type of combination (sympatric, near-allopatric, far-allopatric) and parasite origin on infectivity. This analysis is based on a logistic regression, with the number of infected individuals as response variable and total number of individuals as binomial denominator to correct for variation in sample size.


For each individual cell line, starting from a single individual, we used its growth curve (i.e. the number of individuals over time) to calculate the area under the curve (AUC), an integrated estimate of fitness reflecting division and death rates in the cell line (Capaul & Ebert, 2003). There was a consistent difference in (log-transformed) AUC between experimental blocks (one-way anova: F2,457 = 98.51, P < 0.0001). To take into account this variation, we used the residuals from this analysis for subsequent calculations.

As for infectivity, a balanced data set was created for statistical analysis, based on observations for parasites from 13 replicate populations and five clonal origins, as well as hosts from 19 replicate populations and seven clonal origins (total N = 249 cell lines). We calculated the mean residual AUC for each combination of parasite and host replicate population; from this mean, we subtracted the corresponding mean of the uninfected cell lines. Thus, negative values indicate a negative effect of the parasite on host fitness. This standardized AUC was used as response variable in an anova, carried out on the means per parasite replicate population and type of combination (sympatric, near- or far-allopatric; N = 13 × 3 = 39), as in the analysis of infectivity. Analyses were carried out with the sas (SAS, 1996) and JMP statistical packages (SAS, 2003).



Infection was detected in 78% (168/215) of the inoculated tubes. Infectivity (= proportion of infected individuals) ranged from 0 to 0.47 (mean: 0.10 ± 0.01 SE).

We observed significant differences in infectivity among parasite replicate populations, but no significant overall effect of parasite origin (Table 1). Further, infectivity differed between sympatric and allopatric combinations of host and parasite. For all six parasite origins, infectivity was lower on sympatric than on far-allopatric hosts (Fig. 3a), indicating a general pattern of local parasite maladaptation. Infectivity on near-allopatric hosts was more variable, but in all but one cases, also lower than on far-allopatric hosts (Fig. 3a). On average, near-allopatric infectivity was intermediate between sympatric and far-allopatric infectivity (multiple contrasts: S ≤ NA; NA ≤ FA; S < FA; Fig. 3b). The parasite origin × host type interaction was not statistically significant (Table 1).

Table 1.   Analysis of Deviance of infectivity in sympatric (S) vs. near-allopatric (NA) vs. far-allopatric (FA) combinations, for different parasite origins.
Sourced.f.Mean devianceFP
  1. Analogous to mean sums of squares in an anova, mean deviances (= 2 × log-likelihood ratio/degrees of freedom) were used to perform F-tests. S-NA-FA was considered a fixed effect, the other factors as random effects. S-NA-FA was tested over the S-NA-FA × parasite origin interaction, parasite origin over replicate population (origin). Because of overdispersion, deviances were scaled so that Residual deviance/d.f. = 1.

Parasite origin52.6890.630.6807
S-NA-FA × parasite origin100.9130.910.5371
Replicate population (origin)124.2744.270.0012
Residual (scaled)241.0  
Figure 3.

 Infectivity in sympatric (S), near-allopatric (NA) and far-allopatric (FA) combinations of parasite and host. (a) Means for six different parasite origins (K2, K3, etc); (b) overall means (±SE), averaged over the means per parasite origin. Different letters (a, b) indicate significant difference in multiple contrasts.

There was a significant positive correlation between sympatric and near-allopatric infectivity (Pearson correlation coefficient, r = 0.55, n = 18, P = 0.0185; Fig. 4a), indicating a considerable degree of similarity of parasite performance on hosts with the same genetic background. Sympatric infectivity was only weakly correlated with that on far-allopatric hosts (r = 0.26, n = 18, P = 0.2961; Fig. 4b). Most (16/18) parasite selection lines were less infectious on sympatric hosts than on far-allopatric hosts (t17 = 3.12, P = 0.0062; Fig. 4b), confirming the overall pattern of local parasite maladaptation detected in the anova. Excluding the one German host clonal background (O6; all other backgrounds were of the same Japanese ‘K’ origin; Fig. 2) from analyses did not strongly influence the main results: Levels of local maladaptation (sympatric minus allopatric difference) calculated with or without O6 were highly positively correlated (r = 0.90, n = 18, P < 0.0001).

Figure 4.

 Correlations between sympatric and near-allopatric (NA) or far-allopatric (FA) infection success (a, b) and virulence (c, d). Each point represents the mean for a given parasite population, the different symbols indicate the host clone origin (K2, K3, etc., see Fig. 3a), on which the parasites (co)evolved. To illustrate the shape of the correlations, a bivariate normal density ellipse is imposed on each scatterplot, enclosing ≥ 95% of the points. Points located above the diagonal have higher allopatric than sympatric values, and thus (potentially) indicate local parasite maladaptation.


Paramecia in the singleton experiment produced up to five divisions during the first 7–10 days (mean: 2.9 ± 0.04, median: 2.8). Because no further resources were supplied, densities began to decline in the second week, and by the end of the third week, 95% of the cell lines were dead. The overall mean standardized AUC was negative (−0.10 ± 0.05 SE; t84 = −2.16, P = 0.0333, based on the means per combination of host and parasite replicate populations), indicating a negative effect of infection on host fitness.

Analysis of the standardized AUC (Table 2) did not reveal significant differences in virulence among sympatric and allopatric combinations of host and parasite (Fig. 5a). Qualitatively, the overall patterns resembled those for infectivity: average virulence tended to increase from far-allopatric to near-allopatric to sympatric combinations (Fig. 5b).

Table 2.   Analysis of Variance of virulence in sympatric (S) vs. near-allopatric (NA) vs. far-allopatric (FA) combinations of parasite and host, for different parasite origins.
Sourced.f.Mean squaresFP
  1. For details of hypothesis testing, see Table 1.

Parasite origin40.22120.980.4698
S-NA-FA × parasite origin80.08180.560.7925
Replicate population (origin)80.22661.560.2130
Figure 5.

 Virulence in sympatric (S), near-allopatric (NA) and far-allopatric (FA) combinations of parasite and host. (a) Means for five different parasite origins (K3, K4, etc.); (b) overall means (±SE), averaged over the means per parasite origin. Values represent the area under the curve (AUC) of infected host cell density, relative to uninfected AUC; negative values indicate a negative effect of infection on host density.

As for infectivity, we obtained a larger effect size for the correlation between virulence on sympatric and near-allopatric hosts (r = 0.46, n = 13, P = 0.1147; Fig. 4c) than for the correlation between sympatric and far-allopatric hosts (r = −0.02, n = 13, P = 9416; Fig. 4d). Nine of the 13 parasite replicate populations were more virulent on sympatric than on far-allopatric hosts, but this difference was not statistically significant (t12 = 0.70, P = 0.4958; Fig. 4d).

Correlations between infectivity and virulence

There was little evidence for correlations between infectivity and virulence across the 13 parasite replicate populations. We obtained only small effect sizes for these correlations, when analysing them separately for sympatric and allopatric host–parasite combinations, with correlation coefficients ranging from 0.26 in sympatric to 0.22 in near-allopatric to −0.05 in far-allopatric combinations (all P > 0.3862). Furthermore, there only was a weak correlation between the strength of local adaptation for the two traits (r = −0.15, n = 13, P = 0.6105; Fig. 6), and thus levels of local (mal)adaptation for infectivity were largely unrelated to those for virulence.

Figure 6.

 Correlation between degrees of local (mal)adaptation for infectivity and for virulence [sympatric (S) minus far-allopatric (FA) differences]. Each point represents the mean for a given parasite population, the different symbols indicate the clonal origin (K3, K4, etc., see Fig. 5a), on which the parasites (co)evolved. The 95% bivariate normal density ellipse illustrates the shape of the correlation (see Fig. 4).


Local maladaptation for parasite infectivity

We found a general pattern of local parasite maladaptation for infectivity: on average, sympatric combinations of parasite and host produced fewer infections than did far-allopatric combinations. This confirms results for a smaller set of populations at an earlier time point in their coevolutionary history (Lohse et al., 2006).

Patterns of local host or parasite adaptation have two implications. First, they indicate specificity in attack and defence, possibly governed by relatively simple gene-for-gene interactions (Thompson & Burdon, 1992; Agrawal & Lively, 2002). In the Paramecium-Holospora system, specificity exists at genotype or species levels (Görtz & Brigge, 1998; Fokin, 2004; Fujishima, 2009), but the precise underlying mechanism or genetic basis are largely unknown. It is conceivable that compatibility between different host and parasite genotypes results from differential gene expression and direct cell–cell interactions during early stages of infection, for example when the membrane of infectious form of the parasite comes into contact with the host’s membrane system (vacuole, transport vesicle) to mediate its transfer to the micronucleus (Fujishima, 2009).

The second implication concerns the coevolutionary potential of the interacting players. Many natural host–parasite systems show local parasite adaptation (Greischar & Koskella, 2007; Hoeksema & Forde, 2008), indicating higher evolutionary rates of the parasite (Kaltz & Shykoff, 1998; Gandon et al., 2008). Shorter generation times, larger population sizes or higher mutation rates of parasites have often been invoked as explanations for this evolutionary advantage (Hamilton, 1980; Price, 1980; Ebert & Hamilton, 1996). In our experimental populations, paramecia may divide 2–3 times per week, whereas parasite within-host populations can achieve 1–2 doublings per day. Further, total bacterial population size easily exceeds that of the host by a factor 100, even at low prevalences. This suggests a moderate generation time and population size advantage, or at least no disadvantage of our parasite.

What then caused local maladaptation of the parasite? One possible answer points to differences in genome size and organization. In asexual populations of paramecia, responses to selection occur in the highly polyploid macronucleus (genes in the diploid micronucleus are not expressed). High copy number [>> 100n (Prescott, 1994)] may outweigh the host’s population size disadvantage and increase the number of gene copies in the population over that of the parasite. Holospora, in turn, has a rather small genome [1.7 Mbp (Lang et al., 2005)], which, on its own, may already constrain the evolutionary potential. Smaller genome size may also cause local parasite maladaptation in experimentally coevolving populations of Pseudomonas fluorescens and a bacteriophage (Brockhurst et al., 2007).

Alternatively, parasite maladaptation may arise if infectivity is not under selection or if it trades off with other fitness components, for example, the capacity of vertical transmission. However, this is an unlikely explanation. Under the conditions of this long-term experiment, relative low rates of host division provide few opportunities for vertical transmission, and the parasite shows massive accumulation of horizontal transmission stages (Kaltz & Koella, 2003; Restif & Kaltz, 2006), suggesting that horizontal fitness components, such as infectivity, are main targets of selection.

The role of genetic background: repeatability of local maladaptation and clines of genetic distance

In studies on host specialization, parasites often specialize on some host species or genotypes, but not on others (Gould, 1979; Agrawal, 2000; Turner & Elena, 2000; Little et al., 2006; Magalhaes et al., 2009), suggesting that (co)evolutionary outcomes vary with the identity of the interacting genotypes or species. We investigated the repeatability of parasite specialization to different host clones (genotypes) that vary in life-history traits and resistance (Fels & Kaltz, 2006; Restif & Kaltz, 2006); a previous experiment had also suggested variable costs of adaptation to these clones (Nidelet & Kaltz, 2007). Nonetheless, we found that parasite local maladaptation was consistent over host genetic backgrounds, indicating a general coevolutionary disadvantage of the parasite. On the one hand, this similarity of outcomes is plausible, as most of the clones were genetically related. On the other hand, host clonal origin nevertheless structured our results. Average infectivity in near-allopatric combinations was intermediate between that in sympatric and far-allopatric combinations. In addition, sympatric infectivity was more strongly correlated with near-allopatric than with far-allopatric infectivity, indicating similar performance on hosts with the same genetic background.

From a geographical perspective, these results illustrate the potential role of genetic distance in shaping patterns of local adaptation. Indeed, in spatially structured populations, isolation by geographical distance can produce correlated clines of genetic distance and local adaptation, with decreasing performance on more distant and thus more dissimilar hosts (Ebert, 1994; Kaltz et al., 1999; Thrall et al., 2002; Laine, 2005; Hoeksema & Thompson, 2007; Sicard et al., 2007). For our virtual meta-population of spatially isolated replicate tubes, we conclude that parasite local maladaptation extends to a higher ‘regional’ level, including other, genetically related host populations.

Local adaptation for virulence?

Patterns found for virulence resembled those for infectivity, at least qualitatively: Virulence in sympatric combinations was more similar to that in near-allopatric than far-allopatric combinations, and there was a decreasing cline in mean virulence from sympatric to far-allopatric combinations of host and parasite. However, these trends had no strong statistical support. In part, this may be a sample size problem, as fewer parasite replicate populations were available for analyses of virulence (13 vs. 18 populations for infectivity). For example, the relatively large, but nonsignificant, effect size of the correlation between sympatric and near-allopatric virulence would be marginally significant for a sample size of 18 populations (P = 0.05). In contrast, the effect size for the sympatric–allopatric comparison of virulence in the anova [following Cooper & Hedges (1994): [F/(+ denominator d.f.)]0.5 = 0.18 for the S-NA-FA term in Table 2) was three times smaller than that for infectivity (0.55, S-NA-FA in Table 1). Therefore, lack of statistical power is unlikely to explain the difference between virulence and infectivity.

Regardless of the statistical power, infectivity and virulence are not necessarily expected to show the same patterns of local adaptation (Dybdahl & Storfer, 2003). Traits like infection intensity or virulence may be under balancing selection, because of trade-offs between the benefits from transmission and the cost of damage to the host (Frank, 1996). If fitness on sympatric hosts is maximized at intermediate values of virulence, allopatric performance can be more as well as less virulent than sympatric performance. On average, allopatric virulence may therefore not differ from sympatric virulence. Indeed, of studies on local adaptation in natural populations measuring both infectivity and virulence (infection intensity), more than half reported local adaptation for infectivity, but only one-third for virulence (Table 1 in Greischar & Koskella (2007)).

Moreover, even if there is a difference between sympatric and allopatric virulence, its interpretation depends on the relationship between virulence and fitness (Ebert, 1994). In our system, reduced host division and survival is associated with higher parasite loads (Restif & Kaltz, 2006) and therefore potentially higher rates of horizontal transmission. If horizontal transmission is the predominant transmission pathway in our populations, higher sympatric virulence may reflect local adaptation rather than maladaptation of the parasite.

Correlations between local adaptation for infectivity and for virulence

Selection on infectivity genes and on virulence genes may be linked through epidemiological feedbacks (Gandon et al., 2002; Kirchner & Roy, 2002); and in certain systems, genetic links (trade-offs) exist between infectivity and virulence (Ebert & Mangin, 1997; Thrall & Burdon, 2003). This may lead to correlations between the patterns of local adaptation for the two traits. Our study provides little evidence for such an association. The general picture emerging is that of a parasite less able to infect local hosts, and upon infection, tending to be more harmful to them. However, there was no significant correlation between sympatric–allopatric differences for infectivity and for virulence; if anything, the relationship was negative, rather than positive. Furthermore, infectivity and virulence were not significantly correlated across parasite selection lines (see also Nidelet et al., 2009), consistent with results for individual parasite isolates (J. Brusini, unpublished). Thus, ‘interaction genes’ determining infection success seem to evolve independently of genes determining parasite growth or host exploitation.

Our results are in line with those from studies on natural populations, reporting no significant correlations between the degree of local adaptation for infectivity and parasite development or virulence (Kaltz & Shykoff, 2002; Sicard et al., 2007). Similarly, across studies, the sign of local adaptation for infectivity is not significantly associated with that of local adaptation for virulence [χ2(2) = 0.94, n.s., analysed from Table 1 in Greischar & Koskella (2007)], nor is there a significant quantitative correlation between levels of local adaptation for the two traits (r = −0.29, n = 11, P = 0.3947, analysed from Table 1 in Hoeksema & Forde (2008)).


Our results add to other examples of local parasite maladaptation, showing that the conventional wisdom of parasite local adaptation is by no means the general rule (Greischar & Koskella, 2007; Hoeksema & Forde, 2008). Causes of maladaptation may differ among systems. In natural associations, meta-population dynamics, namely migration, seem to be crucial (Greischar & Koskella, 2007; Hoeksema & Forde, 2008); studies on experimental populations, excluding migration, suggest that genetic architecture (genome size, ploidy level, mutator capacity) is perhaps a more important determinant of the coevolutionary potential than, for example, generation time (Gandon & Michalakis, 2002; Morgan & Buckling, 2006).

This study is among the first to test for variation in the outcome of local adaptation in populations with different initial genetic backgrounds. It appears that these populations followed similar (co)evolutionary trajectories and that, independent of genetic origin, the host generally holds the upper hand in the coevolutionary race. Nonetheless, the similarity of parasite performance on hosts with similar genetic background suggests that genetic origin can play an important role in structuring geographic mosaics of adaptation (Thompson, 2006), even after many generations of independent coevolution.

Finally, our findings confirm the idea that local adaptation is more easily detected for infectivity than for virulence. It is, however, still largely unclear how host–parasite coevolution shapes geographic patterns of variation in virulence and how this relates to coevolution of other traits relevant to the interaction (Dybdahl & Storfer, 2003). This requires more theoretical work, but also more empirical studies examining, for example, the correlations of local adaptation in these traits.


We thank Alison Duncan and Simon Fellous for discussion, and Rhonda Snook and an anonymous referee for comments on an earlier version. This study was financed by a grant ‘ACI Jeunes Chercheurs’ (no. 035167) to O.K.