SEARCH

SEARCH BY CITATION

Keywords:

  • adaptive evolution;
  • experimental evolution;
  • host–parasite coevolution;
  • Red Queen hypothesis;
  • resistance

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Host–parasite coevolution can lead to a variety of outcomes, but whereas experimental studies on clonal populations have taken prominence over the last years, experimental studies on obligately out-crossing organisms are virtually absent so far. Therefore, we set up a coevolution experiment using four genetically distinct lines of Tribolium castaneum and its natural obligately killing microsporidian parasite, Nosema whitei. After 13 generations of experimental coevolution, we employed a time-shift experiment infecting hosts from the current generation with parasites from nine different time points in coevolutionary history. Although initially parasite-induced mortality showed synchronized fluctuations across lines, a general decrease over time was observed, potentially reflecting evolution towards optimal levels of virulence or a failure to adapt to coevolving sexual hosts.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The antagonistic nature of host–parasite interactions should lead to reciprocal (genetic) changes in key host and parasite fitness traits, such as resistance and infectivity, respectively (Woolhouse et al., 2002). Under certain conditions, this is expected to lead to cycles of adaptation and maladaptation through negative frequency-dependent selection (Peters & Lively, 1999; Kouyos et al., 2007). This is at the very basis of the Red Queen Hypothesis, a theory that explains how sex and meiotic recombination can allow host populations to escape their parasites by creating offspring with rare and potentially resistant genotypes (Hamilton, 1980). Indeed, there is a substantial body of work demonstrating coevolutionary changes in host–parasite interactions, ranging from evidence for parasite-mediated selection for resistance in coevolving hosts (Buckling & Rainey, 2002), parasite adaptation to its contemporary hosts (Decaestecker et al., 2007), fluctuating parasite infectivity in time (Forde et al., 2004; Decaestecker et al., 2007), decrease in frequency of common clones (Koskella & Lively, 2009), and the maintenance of sexual reproduction (Jokela et al., 2009) and genetic variation (Schulte et al., 2010). However, most of these studies have used organisms that either completely or predominantly reproduce by clonal propagation, and little evidence exists for temporal dynamics of parasite infectivity in systems with an obligatory outcrossing host (Little, 2002).

It is likely that evolutionary trajectories may be strikingly different in sexual host populations. In a general context, selection in asexual populations acts on co-adapted complexes, due to the tight linkage between genes, and in the absence of mutations, response to selection is limited by the best-existing genotype. Adaptation in sexual populations should be more rapid, as meiotic recombination allows advantageous alleles to spread faster through a population (Fisher, 1930; Crow & Kimura, 1965).

More specific to the realm of host–parasite interactions, there is accumulating evidence for a polygenic inheritance of resistance (Wilfert & Schmid-Hempel, 2008), with resistance loci often being found on separate linkage groups. In theory, the magnitude of parasite adaptation decreases with increasing number of loci involved in resistance (Ridenhour & Nuismer, 2007), meaning that local adaptation of parasites is not necessarily an expected outcome of coevolution (Greischar & Koskella, 2007). Similarly, all other things being equal, sexual populations should demonstrate larger genotypic heterogeneity than asexual populations. Both theory and experiments show that this can select for generalist parasites, thus preventing adaptation to their hosts (Kopp & Gavrilets, 2006; Legros & Koella, 2010).

The model system of the Red flour beetle, Tribolium castaneum, and its natural obligately killing parasite, the microsporidian Nosema whitei (Milner, 1972a,b), is ideal to study the temporal dynamics of host–parasite interactions in outcrossing hosts. Tribolium castaneum can be easily maintained in laboratory conditions and has a rapid development time (<30 days at 33 °C, 70% humidity) and high reproductive output (females can lay up to 10 eggs a day and can live up to 2 years) (Sokoloff, 1974). Nosema whitei infects internal organs, such as the fat body, and generally causes the host to die in the late larval or early pupal stage (Dunn & Smith, 2001; Blaser & Schmid-Hempel, 2005). N. whitei reproduces strictly asexually (Taylor et al., 1999) and is considered an obligate killer, because transmission occurs after host death, when spores are released into the medium due to the decay of carcasses or when the dead infected hosts are cannibalized by larvae. In fact, dead hosts are highly enriched with spores (Blaser & Schmid-Hempel, 2005), which can be stored at 4 °C for years without losing any infectivity (Milner, 1972c). This allows the execution of time-shift experiments (Gaba & Ebert, 2009) where the infectivity of parasites from several time steps in evolutionary history is assayed on one fixed host. Such a set-up can reveal changes in the parasite populations, which would otherwise be masked by rapid reciprocal changes in the host populations, and could very well lead to an overall pattern where no change in average resistance or infectivity is observed (Gaba & Ebert, 2009). By applying experimental coevolution, we previously found that parasite selection maintains host genetic variability (Berenos et al., 2011) and that both parasites and hosts responded to selection during coevolution by a decrease in parasite-induced host mortality or an increase in resistance, respectively (Berenos et al., 2009). Capitalizing on our previous results of selective responses of both hosts and parasites, we now determine the detailed underlying time course of parasite-induced host mortality, using a time-shift experiment (Gaba & Ebert, 2009). We here specifically focus on the dynamics of the coevolving antagonists as (i) our previous study found that that virulence is expressed rather specifically in relation to the evolutionary history of the host lines (Berenos et al., 2009) and (ii) cross-infection between isolated coevolving populations can lead to a wild mosaic of patterns that have little relationship with the coevolutionary dynamics within coevolving populations (Nuismer, 2006).

With this set-up, we would like to address two possible outcomes of coevolution. For one, coevolution can result in directional selection if evolutionary potential of one party is significantly larger than the other. This could lead to an increase in parasite-induced mortality during coevolution, similar to results obtained with serial passage experiments (Ebert, 1998) or to a decrease in parasite-induced mortality if coevolving hosts can prevent parasite adaptation by creating heterogeneity among their offspring (Agrawal & Otto, 2006; Legros & Koella, 2010). The other conceivable outcome of coevolution is the presence of cyclical dynamics where periods of high mortality are followed by periods of low mortality reflecting the rapid process of adaptation and counter-adaptation. Our time-shift experiment offers a resolution fine enough to differentiate between both outcomes and can therefore reveal the underlying evolutionary dynamics in a host–parasite system with obligate sexual reproduction of the host.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

All hosts and parasites used in this study originated from a long-term coevolution experiment using T. castaneum and its natural parasite N. whitei as described in more detail in Berenos et al., 2009. All lines were started at the same time and were kept at standard environmental conditions (24-h dark, 33 °C, 70% humidity) on standard medium (type 550 “Knospe” organic flour containing 5% dried yeast) throughout the experiment. At the start of the coevolution experiment, all lines were infected with the same mixture composed of equal spore numbers from eight different N. whitei isolates to ensure ample standing genetic variation within the parasite population. Host generations were discrete, and population size was constant as 500 unsexed surviving adult beetles were used as breeders for the next generation. For the parasite population, we collected the dead infected larvae that are enriched with spores and kept them at 4 °C. Dead larvae were collected in a flexible time window between 35 and 50 days post egg-laying, matching the intrinsic differences in development time between the lines, thereby balancing the relative time of death across experimental lines. The collected larvae were ground and sieved (through a 510-μm mesh) to obtain a homogenous spore-containing powder, which was added to the medium of the new host generation to obtain a final concentration of 2 × 104 spores per gram of medium. The remaining spore-containing powder was kept at 4 °C where spores can be stored for extended periods of time without losing infectivity (Milner, 1972c). A random subset of four of the eight beetle lines used in Berenos et al., 2009 was used for the current time-shift infection experiment representing lines nos. 1, 2, 7 and 8 in that former study.

Infection and survival experiment

After 12 generations of coevolution, we relaxed selection on the host lines by rearing them on parasite-free medium to avoid direct trans-generational effects (Sadd et al., 2005; Roth et al., 2009). In the following generation 13, we collected 100 unsexed beetles from each of these experimental lines to produce offspring that were subsequently used in the time-shift experiment. The resulting offspring were subjected to the following treatments. (i) Exposure to spores from the ancestral parasites (i.e. spores used to infect all lines in generation zero), (ii) Exposure to spores collected from eight different time points during the coevolution experiment. Specifically, we used parasites that had been allowed to coevolve with each respective host for different lengths of time, that is, 2, 3, 5, 7, 9, 11, 12 and 13 generations of coevolution, respectively, (iii) Controls that were not exposed to parasite spores. The ancestral parasite source was the same for all lines (a mixture of eight N. whitei isolates that have been cultured on distinct T. castaneum stock lines), whereas the coevolved spores always originated from the same host line for which the respective test was conducted. Freshly hatched larvae (1–2 days old) were randomly collected from a single jar for each host line and were subsequently placed individually into glass vials (13 × 40 mm, VWR Switzerland), containing 0.1 g of either parasite-inoculated flour or parasite-free medium. A total of ca 1’070 larvae were used (four replicate lines × 10 treatments × 25–30 larvae each), and distribution of larvae was executed on a single day to avoid random temporal differences. In the infection treatment, a dosage of 5*104 spores gram−1 of flour was used for all parasite sources. After their assignment to a treatment, the larvae were kept under standard environmental conditions (24-h dark, 33 °C, 70% humidity). The vials were checked for survival three times a week until 42 days after their placement and once a week thereafter for a total of 55 days since the start of the experiment. T. castaneum is a long-lived insect, and unmated individuals can live up to 2 years (Sokoloff, 1974). Mortality due to nonparasite-related effects, such as potential damaging effects of distributing young larvae, is likely to be very low in the chosen experimental conditions. The average development time under control conditions (measured as time to adult emergence since distribution) in our experiment was 21.3 days.

Statistical analysis

To test whether exposure to parasites causes higher host mortality than when hosts are reared under control conditions, we used a generalized linear mixed model (GLMM) (using the glmmPQL function from the MASS package for R), with presence or absence of parasites in the medium as a fixed factor (exposed/nonexposed), individual mortality as a response variable and host line as a random factor. Total host mortality was analysed using a GLMM with binomial error distribution. Time-shift window was kept as a fixed factor, whereas host line was treated as a random factor. Multiple contrast analysis was used to compare means of total mortality between sequential time-shifts. For a correlation between the number of generations of coevolution (i.e. time-shift window) and host mortality, a mixed model ancova was used with generations of coevolution as a numeric independent variable and average mortality per time-shift window for each line as response variable. Host line was treated as a random factor. All statistical analyses were conducted with the R statistical package (R Development Core Team, 2010).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

To begin with, we found that host mortality was – not surprisingly – significantly higher when exposed to parasites than under control (nonexposed) conditions (34.4% vs. 6.1%, GLMM: F1,1061 = 24.15, P < 0.001). When exposed to parasites, host mortality varied among time-shifted parasites (Fig. 1, Table 1). Contrast analysis between sequential time-shifts showed that initially, time-shifts differed significantly in mean mortality and also synchronously changed direction in all lines (see changing signs of parameter estimates in Table 1). However, after five generations of coevolution, mean mortality did not differ significantly between any sequential time-shifts (Fig. 1). Over the whole time-series, host mortality decreased and was lower when exposed to spores from recent time-shift windows than when exposed to spores from the more distant past in coevolutionary history (Fig. 1).

image

Figure 1.  Host mortality when exposed to parasites from their own coevolutionary trajectory. The black solid symbols indicate the mean mortality for all four host lines under the respective parasite conditions (±SE). The grey open symbols represent mortality of each of the replicate host lines (connecting lines added to aid visualization). Total host mortality varied among time-shift windows (Table 1), and parasite-induced mortality was lower when exposed to spores from recent time-shift windows than when exposed to spores from the more distant past in coevolutionary history (dashed line, regression R2 = 0.55, F1,31 = 15.316, P < 0.001). Time-shifts that showed a significantly different mean mortality than the immediately following time-shift are indicated by asterisks (***P < 0.001, **P < 0.01, ns P > 0.05, more statistical details can be found in Table 1).

Download figure to PowerPoint

Table 1.   Post hoc tests (sequential) of a generalized linear mixed model (GLMM) of mortality of coevolved beetles when exposed to time-shifted parasites. Mortality varied significantly between time-shifts (F8, 956 = 8.08, P < 0.001).
IntervalEstimateSEZ valueP value
G0–G2−1.5580.306−5.085<0.001
G2–G31.0120.3023.3480.006
G3–G5−0.9950.295−3.3740.006
G5−G70.1320.3050.4320.999
G7–G90.3690.2941.2540.791
G9–G11−0.6530.299−2.1790.191
G11–G12−0.0850.319−0.2660.999
G12–G130.1360.3250.4180.999

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Here, we present a negative correlation between generations of host–parasite coevolution and parasite-induced mortality, a result that is in concordance with a previous independent long-term selection experiment using the same system (Fischer, 2006). Additionally, an experiment using lines from the same coevolution experiment showed that, after 11 generations, parasite-induced mortality decreased both on coevolved and on naïve host lines (Berenos et al., 2009).

The observed long-term evolution towards lower parasite-induced mortality can potentially be attributed to three mechanisms. First, the decrease in parasite-induced mortality may be the result of trade-offs between virulence and transmission, even in obligate killers such as N. whitei (Jensen et al., 2006; Berenos et al., 2009). In our previous results, we found optimal virulence expressed on coevolved hosts to cause approximately 60% mortality. This is substantially higher than the host mortality estimate for generation 13 of approximately 35%. As most host individuals surviving past 55 days are usually uninfected, it seems unlikely that such low mortality can be a result of adaptive evolution towards optimal virulence.

The second possibility is that the fully out-crossing T. castaneum might escape by faster evolution of resistance than its clonal parasite can adapt. This is in line with previous findings where T. castaneum has been shown to evolve higher recombination when coevolving with N. whitei (Fischer & Schmid-Hempel, 2005). Elevated recombination rates may prevent parasite adaptation by creating genetic heterogeneity among offspring (Legros & Koella, 2010) and potentially novel resistant genotypes (Salathe et al., 2008), even in the absence of influx of new genetic material through migration (Gandon & Michalakis, 2002). As a result, this amounts to the creation of few new susceptible genotypes, leading to selection on rare parasite genotypes. If input of new genetic material by mutation in the clonal parasite is comparatively low, selection imposed by a few susceptible host genotypes will act mainly on standing genetic variation among different clonal lineages. As only successful infection will lead to propagation into the next round of coevolution, selection will be strong and quickly erode standing genetic variability of the parasite, thus limiting its adaptive potential.

The third explanation could arise from an interaction between evolutionary change of hosts and parasites. Because we did not experimentally control for host evolution, evolved host resistance alone cannot explain the observed changes in parasite-induced mortality. If only host resistance evolved and the parasite populations did not change, parasites from all time-shifts should show equal parasite-induced mortality. However, given that coevolving antagonists have changed phenotypically (Berenos et al., 2009) and/or genotypically (Berenos et al., 2011), hosts from recent generations may have lost adaptation to parasites from earlier time points, if such evolutionary change followed a directional trajectory. This would give rise to the observed pattern, as parasites from recent generations should cause lower mortality than parasites from earlier time points.

Similar to our findings, experimental coevolution of the facultatively sexual Paramecium caudatum with Holospora undulata did not lead to an increase in infectivity (Lohse et al., 2006) but rather parasite maladaptation (Adiba et al., 2010), undermining the common view that parasites have an inherent evolutionary advantage. Interestingly, a time-shift using the predominantly clonal Daphnia magna and its bacterial endoparasite Pasteuria ramosa showed that the current parasites were more infective than parasites from the past or the future (Decaestecker et al., 2007). In contrast to the Daphnia–Pasteuria system, which shows clear genotype x genotype interactions (Carius et al., 2001), there is evidence for a complex genetic architecture of resistance against N. whitei (Wegner et al., 2008, 2009), which is predicted to restrict the magnitude of parasite adaptation (Ridenhour & Nuismer, 2007). Consequently, under such a scenario, and as observed in our experiment, easily interpretable patterns of local adaptation are not an inevitable outcome of coevolution.

In addition to a long-term directional decrease, mortality initially showed synchronous fluctuations in all independent host backgrounds (Fig. 1). Experimental evidence for temporal variation over the coevolutionary history of interacting host–parasite populations, such as presented here, is scarce (Decaestecker et al., 2007; Koskella & Lively, 2007, 2009). Interestingly, while strong and highly synchronized in the beginning, the fluctuations disappear over time, as after five generations of coevolution, there was no significant difference in mortality between consecutive time-shifts. There are several explanations for this observation. First, the introduction of parasites in novel host populations as an emerging disease resembles traditional biological invasions (Mack et al., 2000). In such cases, epidemiology upon invasion may initially show unpredictable and highly dynamic patterns during the course of reciprocal adaptation. For example, introduction of the myxoma virus to control invasive rabbit populations in Australia led to a rapid decrease in virus fitness, which subsequently stabilized (Fenner & Cairns, 1959). Second, on theoretical grounds, rapid dynamics as observed here are plausible in a system with high selective pressures on infectivity and resistance (Peters & Lively, 1999), which is the case when hosts coevolve with obligate killers such as N. whitei (Blaser & Schmid-Hempel, 2005). Theoretical models show that depending on the initial conditions, and the strength of selection, coevolution may move towards equilibria while maintaining host genetic variation (Kopp & Gavrilets, 2006), a pattern that is somewhat similar to our experiment (Fig. 1).

In conclusion, our time-shift experiment is among the few experiments that study the evolution of parasite fitness using multiple time points when hosts are allowed to adapt. Our understanding of parasite adaptation is largely based on serial passage experiments (Muskett et al., 1985; Ebert, 1998; Yourth & Schmid-Hempel, 2006) that generally show an increase in parasite virulence or infectivity, or coevolution in systems with clonal hosts where parasites can adapt by tracking common host genotypes leading to negative frequency-dependent selection (Ebert, 1994; Decaestecker et al., 2007; Koskella & Lively, 2009). Our data of coevolution with a sexual host rather support theoretical models predicting variable or even chaotic interactions (Kopp & Gavrilets, 2006), and preclusion of parasite adaptation due to complex genetic architecture of host resistance (Ridenhour & Nuismer, 2007).

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The authors thank Igor Vuillez, Tilmann Silber, Daniel Trujillo-Vallegas, Natasha Rossel and Miguel Jales Hon for help at various stages during the experiment. We thank Andrew Read and five anonymous reviewers for their helpful comments. Supported by the Genetic Diversity Center of ETH Zurich (GDC) and CCES. Financially supported by SNF grant 31-120451 to KMW and ETH grant number TH-09 60-1 to PSH.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References