Materials and methods 1. Origin of hosts and parasites

with development time

2. Materials and methods 1. Origin of hosts and parasites

Gammarus pulex individuals were collected between November 2006 and October 2007 in two different rivers located in Burgundy, eastern France. Gammarids from the first population, Val-Suzon, were collected in a small tributary of the Suzon River (N 47°

24’12.6’’ ; E 4° 52’58.2’’). In surveys over 10 years, P. laevis parasites have never been found in this population (L. Bollache, unpublished data), which can therefore be considered naïve for the parasite (see Franceschi et al. 2008, 2009). On the contrary, the second population, from the Ouche River (N 47° 17’52.91’’ ; E 5° 02’27.28’’), a tributary of the Saône River, is naturally-infected by P. laevis (e.g. Perrot-Minnot 2004). In the laboratory, all gammarids were acclimated for 4 weeks prior to infection experiments, in well-aerated aquaria of 37 ¯ 55 ¯ 10 cm, containing dechlorinated, UV-treated, tap water at 15 ± 1°C and elm leaves for food, under a 12:12 h light:dark cycle. Only males were used for experimental infections.

Parasite eggs were taken from naturally-parasitized chubs (Leuciscus cephalus) sampled by electrofishing in the Ouche River. Fish were anaesthetised, killed and dissected within 24h after collection. Adult parasites were immediately collected from the fish intestines and eggs were obtained by dissecting female worms. Eggs were placed in 400 µL of water (each clutch in a separate tube) and parasite tissues were preserved in 300 µL of alcohol for species molecular identification.

2.2. Parasite molecular identification

In Burgundy, gammarids and fish may be infected by two closely-related species of acanthocephalan parasites, P. laevis and Pomphorhynchus tereticollis. These two species cannot be reliably distinguished based on morphology, and thus a molecular method was used for parasite identification (see details in Franceschi et al. 2008).

2.3. Infection procedure

Because of laboratory constraints, experimental infections were carried out in three steps, in December 2006, May 2007 and October 2007, following the same protocol, detailed below. For the first infection (winter 2006), we only used gammarids from the Ouche population, and for the last one (winter 2007), we used gammarids from the Val-Suzon population. For the experiment in the spring 2007, we used gammarids from these two populations simultaneously, which were infected with the same parasite clutches. In the laboratory, the gammarids were acclimated for four weeks prior to infection experiments, by groups of 250 individuals, in well-aerated aquaria of 37 ¯ 55 ¯ 10 cm containing water at 15

± 1°C and elm leaves for food, under 12h:12h light:dark cycle.

Parasite eggs from each female were examined under a Nikon microscope (20 x) to evaluate their maturity (mature eggs contain a developed larval stage called acanthor, see Crompton & Nickol 1985). Eggs were counted in three viewing areas, in order to carry out

the experiments with clutches having approximately the same number of mature eggs. Ten clutches of P. laevis, from ten different fish, were selected for each infection run (December 2006, May 2007 and October 2007). We considered each clutch as a “parasite sibship”, since all eggs are assumed to come from one female and only one male, thus making them full-sibs. The number of eggs was then estimated by averaging the counts made under a microscope in 10 samples of 1 μL. After dilution with water, suitable egg exposure doses were obtained. We finally used 10 different parasite families during the winter 2006 experiment, 9 in the spring 2007, and 10 in the winter 2007.

Prior to infection, gammarids were deprived of food for 24 h. The infection was then carried out as described in Franceschi et al. (2008, 2009). Two gammarids were placed in a dish of 6 cm diameter, filled with water at 15 ± 1 °C, and the egg suspension at a concentration of 100 parasite eggs per gammarid (this dose provided an acceptable ratio of infection success versus multiple infection rate, see Franceschi et al. 2008) was deposited on a 1cm² dry elm leaf placed in the dish. The gammarids were allowed to feed on it for 48 h, while uninfected leaves were provided to control groups. For each treatment (one treatment corresponding to an infection with one parasite sibship), 108 male gammarids were used.

At the end of the exposure period, the gammarids were rinsed and placed in 0.5 L aquaria by groups of eighteen individuals. Thus, each treatment was carried out in six replicates of eighteen individuals randomly assigned to each aquarium, themselves randomly distributed on shelves in a single room. They were then maintained in standard conditions (they all received the same oxygenated water at 15 ± 1 °C, automatically replaced 6 times during the day, and were under a 12:12 h light:dark cycle).

Survival was checked every week and, from the sixth week, all gammarids were inspected once a week under a binocular microscope to detect the presence of parasites (see Franceschi et al. 2008). As soon as a parasite larval stage was detected through the cuticle, the gammarid was isolated in a 0.2 L plastic dish filled with water at 15 ± 1 °C. At the same

time, uninfected individuals from the control treatment were also isolated. Parasite development was then followed once a week, and the date on which the parasites reached the cystacanth stage (infective for the definitive host) was noted. The prevalence (number of infected hosts/total number of surviving gammarids) was calculated 94 days post-exposure in December 2006 experiment, 90 days post-exposure in the May 2007 experiment and 66 days post-exposure in the October 2007 experiment (because of a more rapid development of parasites, see results). In the Ouche population, natural infections may appear in the course of the experiment, despite the four weeks of quarantine before experimental infection. Since the average time of development in this population was around ten weeks, the exposed G.

pulex which developed an infection before the sixth week post-exposure were discarded from the analysis (n = 2 in December 2006 and n = 1 in Spring 2007). The intensity of infection (number of parasites per infected host) was verified at the end of the experiment by dissecting all animals.

2.4. Behavioural measurements

The phototaxis of each individual was measured the day after the parasite reached the cystacanth stage (called ‘young cystacanth’ stage hereafter), and then two weeks later (called

‘old cystacanth’ stage hereafter). The reaction to light of isolated individuals was measured as described in Franceschi et al. (2007). A single gammarid was introduced into a horizontal tube filled with well-aerated water, comprising a dark zone and a light zone of equal size.

After a 5 min period of acclimatisation, the position of the gammarid was recorded each 30 s for 5 min. At each observation, a score of 0 was given if the individual was located in the dark area and a score of 1 was given if it resided in the lighted area. At the end of each trial, summed scores ranged from 0 (always in the dark, strongly photophobic) to 10 (always in the light, strongly photophilic).

At the end of the experiments, all individuals were measured (body height at the level of the fourth coxal plate basis) using a Nikon SMZ 1500 stereoscopic microscope and Lucia G 4.81 software. They were dissected and the intensity of infection was recorded.

2.5 Data analyses

All tests were performed using JMP 6.0 Software (SAS Institute Inc.) and were two-tailed. P values < 0.05 were considered significant.

Survival data were analysed using the Cox regression method. The analysis was performed during the growth phase of the parasite, i.e. before the visual detection of the parasite. We first tested the effect of the different parasite families on the survival of the gammarids, independently for the two host populations, and we also took into account an

‘aquarium’ factor nested within the ‘parasite family’ factor (in order to account for the variability between aquaria within each treatment). We then compared the survival of the two host populations infected simultaneously by the same parasite families, during the spring 2007 experiment.

The infection success of the different parasite families was analysed independently for the two host populations, using a logistic regression. We included the following factors:

exposure date (winter 2006/spring2007/winter2007), and a random ‘parasite family’ factor nested within the ‘exposure date’ factor. We also analysed the effect of the different parasite families on the proportion of multi-infections using a logistic regression with the same factors. We then investigated the effect of the number of parasites per host (we created three categories to describe the intensity: infection by one, two, or more than two parasites, see Franceschi et al. 2008) on phototaxis scores and on parasite development time, using Kruskal-Wallis tests. The development time of the parasite was estimated as the time lapse between the day of exposure and the day when the parasite reached the ‘young cystacanth’

stage.

To compare infection success in the two host populations during the spring experiment, we used another logistic regression including the following factors: host population (Ouche/Val-Suzon), parasite family, and the interaction between host population and parasite family. Finally, we tested the correlation between the infection successes of the different parasite families in each host population using a non-parametric Spearman correlation.

The effect of the different parasite families on development time and phototaxis scores were analysed using non-parametric Wilcoxon or Kruskal-Wallis tests, since the data did not meet either normality or homoscedasticity conditions, even after transformation.

To analyse the relationship between phototaxis scores, infection success and development time, we used non-parametric Spearman correlations (because of the non parametric distribution of behavioural data). Two types of correlations were investigated.

The first took into account the average values per sibship, which could be seen as a genetic correlation in a broad sense, including additive, non-additive and interaction genetic covariance plus maternal covariance (Roff 1997, Hammerschmidt & Kurtz 2005); the second took into account individual parasite values.

3. Results

3.1. Host survival

When infecting hosts from the Ouche population, the different parasite families did not produce significant differences in terms of host survival (winter 2006 experiment: Whole model: χ²71 = 126.63, p < 0.0001; parasite family: χ²10 = 13.47, p = 0.19; aquarium [parasite family]: χ²61 = 113.64, p < 0.0001; spring 2007 experiment: Whole model: χ²34 = 43.44, p = 0,13).

Similarly, with hosts from the Val-Suzon population, we did not find any difference in virulence between the different P. laevis families, either in the spring 2007 experiment

(Whole model testing the parasite families effect: χ²34 = 39.34, p = 0.24) or in the winter 2007 one (Whole model: χ²62 = 75.89, p = 0.11).

When we simultaneously analysed the two host populations infected by the same parasite families, during the spring 2007, we found a significant difference in survival between the two populations (survival decreased more rapidly in gammarids from the Ouche than in gammarids from the Val-Suzon), but still no effect of parasite family and no significant interaction between host population and parasite family were detected (Whole model: χ²35 = 60.39, p = 0.005; host population: χ²1 = 15.27, p < 0.0001; parasite family: χ²8

= 11.38, p = 0.18; aquarium [parasite family]: χ²18 = 24.47, p = 0.14; host population*parasite family: χ²8 = 11.23, p = 0,19).

3.2. Infection success and development time

With gammarids from the Ouche population, the prevalence was very low, ranging from 0% to 7.14% for the first experiment and from 0% to 8.33% for the second one (Fig 1).

None of the factors (exposure date and parasite family) included in the logistic regression analysing infection success was significant (Whole Model: χ²86 = 62.56, p = 0.97). The parasite families did not significantly differ in their ability to infect gammarids. Due to the very low infection rate, we were not able to analyse the proportion of multi-infections in this population.

Analysing gammarids from the Val-Suzon population, the main difference with those from the Ouche population was a much higher prevalence, ranging from 4.55% to 83.33%

for the first experiment and from 55.74% to 83.61% for the second one (Fig 1). Both parasite family and exposure date influenced infection success (Whole Model: χ²18 = 174.95, p <

0.0001; exposure date: χ²1 = 35.98, p < 0.0001; parasite family [exposure date]: χ²17 = 149.32, p < 0.0001). The proportion of multi-infections was also influenced by parasite

family, but only weakly by exposure date (Whole Model: χ²18 = 42.60, p = 0.0009; exposure date: χ²1 = 5.21, p = 0.02; parasite family [exposure date]: χ²17 = 41.48, p = 0.0008).

In the spring 2007 experiment, infectivity was compared between the two host populations using a logistic regression. Infection success was affected by host population and parasite family, but not by the interaction between these two parameters (Whole Model: χ²17

= 433.62, p < 0.0001; host population: χ²1 = 103.87, p < 0.0001; parasite family: χ²8 = 38.58, p < 0.0001; host population*parasite family: χ²8 = 10.28, p = 0.24; Fig 1). In addition, we found no significant correlation between the infection rates in the Ouche population and in the Val-Suzon population (Spearman ρ = 0.35, p = 0.34). Thus, the infection success of a parasite family in a given host population cannot predict its infection success in the other population.

Since prevalence was too low in the Ouche host population, development time of parasites was analysed for the Val-Suzon only. Multi-infections did not affect development time (Table 1), and we thus analysed all infected individuals (singly- and multi-infected ones) together.Development time to the ‘young cystacanth’ stage was influenced by the date of exposure (Wilcoxon test: Z = 21.17, p < 0.0001, with a median development time of 12.3 weeks in the spring 2007 experiment, and 8.7 weeks in the winter 2007 one) but not by parasite family (Kruskal-Wallis test: spring 2007: χ²8 = 14.00, p = 0.08; winter 2007: χ²9 = 13.57, p = 0.13). In the winter 2007 experiment, parasites were also more synchronised in their development than in the spring 2007 (O’Brien test for equal variances: F1,589 = 50.27, p

< 0.0001).

3.3. Behavioural manipulation

Phototaxis scores were analysed for the Val-Suzon population only, as the infection rate in the Ouche population was too low. Since multi-infections did not affect phototaxis scores either at the ‘young cystacanth’ or at the ‘old cystacanth’ stage, either in the spring

2007 or in the winter 2007 (Kruskal-Wallis tests, Table 1), we analysed all infected individuals together.

We first analysed phototaxis scores at the ‘young cystacanth’ stage (Fig 2a). The global level of manipulation was much lower in the winter experiment than in the spring one (Wilcoxon test: Z = 9.44, p < 0.0001), with a median of 1 (N = 362) and 4 (N = 229) respectively. In light of this difference, we then analysed the two experiments separately. In spring, the variation between parasite families in their ability to manipulate the behaviour of their intermediate host was significant (Kruskal-Wallis test: χ²8 = 19.43, p = 0.01; Fig 2a). In addition, the phototaxis scores of infected gammarids were significantly higher than the scores of control individuals (Wilcoxon tests, all p < 0.02; the family H was not analysed, because only one individual was infected). In winter, there was no significant difference between parasite families (Kruskal-Wallis test: χ²9 = 9.87, p = 0.36; Fig 7a) and the scores of infected individuals were not different from those of the control gammarids (Wilcoxon tests, all p > 0.18).

At the ‘old cystacanth’ stage, we found no difference in phototaxis scores between the two experiments (Wilcoxon test: Z = -0.21, p = 0.83, N = 181 in spring 2007 and N = 286 in winter 2007). We thus analysed the two experiments together, and found no significant variation between parasite families for the phototaxis scores recorded at this stage of development (Kruskal-Wallis test: χ²18 = 26.94, p = 0.08; Fig 2b). Moreover, the phototaxis scores of infected gammarids from all families significantly differed from the phototaxis scores of their respective control groups (Wilcoxon tests, all p < 0.02).

In both spring and winter, we thus observed a significant increase in phototaxis scores between the ‘young cystacanth’ and the ‘old cystacanth’ stage (Wilcoxon Signed-Rank test:

spring: p < 0.0001, N = 181; winter: p < 0.0001, N = 286). This increase was thus much higher for parasites in the winter experiment than for parasites in the spring experiment (Wilcoxon test: Z = -6.02, p < 0.0001). We found no difference between families for this

increase in phototaxis scores, either for the spring experiment (Kruskal-Wallis test: χ²8 = 7.81, p = 0.45) or for the winter one (Kruskal-Wallis test: χ²9 = 7.86, p = 0.54).

3.4. Correlations between infection parameters and phototaxis scores

Since we found no significant variation between parasite families in either development time or phototaxis scores in winter, we only analysed data from the spring experiment.

At the ‘young cystacanth’ stage, the genetic correlation between family average development time and family median phototaxis score was not significant (Spearman ρ = 0.55, p = 0.12, Figure 3a, the family H removed because only one gammarid was infected).

However, at the individual level, we found a significant positive correlation between development time and phototaxis score (ρ = 0.31, p < 0.0001, Fig 3b). The same positive correlation was revealed in the various categories of parasite intensities: in hosts harbouring a single parasite (ρ = 0.26, p = 0.02, n = 82), hosts harbouring two parasites (ρ = 0.34, p = 0.02, n = 47) or more than two parasites (ρ = 0.35, p = 0.0004, n = 97). No significant correlation was found at the ‘old cystacanth’ stage (all p > 0.06).

The correlations between infection success and phototaxis scores were not significant, either at family level (ρ = -0.47, p = 0.19), or at individual level (all p > 0.72) at the ‘young cystacanth’ stage. The same results were found at the ‘old cystacanth’ stage (ρ = -0.15, p = 0.71 for family level, and all p > 0.17 at individual levels)

Finally, there was no significant correlation between infection success and development time (all p > 0.46).

Discussion

This work first revealed that the two host populations under investigation have different resistance levels to P. laevis infection. However, since we found no correlation

between the infection success of any parasite family in either of the two gammarid populations, this finding suggests that there are not universally “good” and “bad” parasites sibships having a high or a low infection potential in any host population. Rather, the infection success of a particular parasite family seems to depend on the host population it infects. Nonetheless, the sympatric host population (Ouche) was much more resistant to infection than the naïve one (Val-Suzon). Taken together, these results are similar to those obtained in a previous experiment, but extend them to the intra-population parasite level. In the previous experiment, cross-infections between host and parasite populations revealed that parasites were more infective in a host population which was naïve for P. laevis infection (Franceschi et al. 2009), which confirms the findings obtained in another acanthocephalan-crustacean system (Hasu et al. 2009). However, since the infection rate was very low in the sympatric Ouche host population, we were unable to compare behavioural manipulation between the two host populations. We have therefore limited the following discussion to the results obtained with hosts from Val-Suzon.

Our study showed that three main parameters may be evoked to explain the parasite intra-population variation in behavioural manipulation: parasite sibships, season, and an individual trade-off between manipulation and parasite development time.

There were differences in terms of infection success, proportion of multi-infections and behavioural manipulation among the different parasite families, although parasite virulence was not different. These differences among parasite sibships could be due to genetic variations. However, our experimental design did not allow us to differentiate genetic effects from maternal effects. Neither can we ignore the possibility of an “environmental” effect of the definitive host from which female parasites were sampled (see also discussion below for the observed variation between seasons). In numerous host-parasite systems, within population variations in infection success have been reported and linked to differences in parasite genotypes (e.g. Incani et al. 2001, Carius et al. 2001, Kaltz & Shykoff 2002), but

very little is known about behavioural manipulation. Only the results of Leung et al.

(submitted manuscript) suggest that there could be a variation between cestode clones from the same natural population in their ability to reach the host tissues where they can express host manipulation. However, in this case, the presence of other parasites seems to be a stronger factor explaining parasite distribution than parasite genotype (Leung et al.

submitted manuscript). Even if our results are to be taken cautiously because of several confounding parameters, we suggest that the observed differences between parasite sibships may be linked to genetic differences.

The variation between parasite sibships was nevertheless observed in only one of our experiments, in the spring replicate. In the winter replicate, no such variation was observed, and most P. laevis sibships were poor manipulators at the ‘young cystacanth’ stage. This does not mean that they were intrinsically bad manipulators because, when they had grown to the ‘old cystacanth’ stage, their manipulation capacity was the same as that of the spring parasites. Parasite development time was also much faster and less variable in the winter experiment than in the spring one. This finding strongly suggests a trade-off between development time and behavioural manipulation (see discussion below), the winter parasites being unable to both develop rapidly and at the same time induce behavioural manipulation when they reach the infective stage. Environmental parameters such as temperature and resource availability can affect parasite traits such as growth, or the outcome of parasite transmission by modulating the exposure of the host to their pathogens (Ebert et al. 2000, Fels 2005, Jokela et al. 2005, Fels & Kaltz 2006). A single parasite genotype can produce different phenotypes across a range of environmental conditions and environmental

Dans le document Université de Bourgogne UMR CNRS 5561 Biogéosciences Equipe Ecologie Evolutive THESE (Page 72-97)