Host manipulation revisited: no evidence for a causal link between altered photophobia and increased trophic transmission of amphipods infected with acanthocephalans

Authors


Correspondence author. E-mail: mjperrot@u-bourgogne.fr

Summary

1.Many parasites with complex life cycles critically rely on trophic transmission to pass from an intermediate host to a final host. Parasite-induced behavioural alterations in intermediate hosts are often supposed to be adaptive through increasing the susceptibility of intermediate hosts to predation by final hosts. However, the evidence is so far only correlational, and direct evidence for a causal link between one single behavioural alteration and increased trophic transmission is still missing.

2.Here, we addressed, for the first time, the relationship between increased vulnerability to fish predation and altered photophobia in an amphipod, Gammarus pulex, infected with a fish acanthocephalan, Pomphorhynchus tereticollis.

3.In microcosms, naturally-infected amphipods were significantly more vulnerable to fish predation than uninfected ones at two different light intensities. However, although variation in illumination significantly affected the extent of difference in photophobia between infected and uninfected individuals, it had no effect on predation bias towards infected amphipods.

4.In addition, although an injection of a mixture of serotonin and fluoxetine in uninfected amphipods mimicked the parasite-induced decreased photophobia, injected individuals were not more vulnerable to fish predation that uninfected ones injected with a control solution.

5.Overall, our results indicate that the decreased photophobia in infected intermediate hosts does not play in itself a causal role in the trophic transmission of the parasite to its final host. The actual role a parasite-induced behavioural alteration plays in trophic transmission should be carefully assessed before an adaptive interpretation is given.

Introduction

According to the concept of ‘host manipulation’, some parasite species have developed, through natural selection, an ability to modify the phenotype of their hosts in ways which are beneficial to the parasite's fitness and detrimental to that of the host (Moore 2002; Thomas, Adamo & Moore 2005; Poulin 2007). Phenotypic alterations brought about by such ‘manipulative parasites’ to their hosts (Moore 2002; Thomas, Adamo & Moore 2005) are traditionally regarded as perfect illustrations of the concept of extended phenotype (Dawkins 1982). The phenomenon is particularly well documented in parasites with complex life cycles, where the phenotypic alterations observed in infected intermediate hosts often appear to contribute to increase trophic transmission to final hosts (Moore 2002).

Although the manipulation of host behaviour by parasites with complex life cycles can take many forms (Moore 2002; Poulin 2007), the inversion of photophobia (avoiding light) or phototaxis (movement in reaction to light) has been reported in several different host-parasite associations (see references in Moore 2002; as well as Bauer et al. 2000; Cézilly, Grégoire & Bertin 2000; Park et al. 2002; MacNeil et al. 2003; Perrot-Minnot 2004; Benesh, Duclos & Nickol 2005; Franceschi et al. 2007; Sanchez, Georgiev & Green 2007; Coats, Poulin & Nakagawa 2010; Ponton et al. 2011; Rauque et al. 2011). Whereas uninfected intermediate hosts tend to be strongly photophobic and avoid lighted areas, infected ones become attracted to or insensitive to light. From a physiological point of view, this behavioural alteration has been shown to coincide with an increase in the immunoreactivity of serotonergic neurons in the nervous system of amphipod crustaceans infected with helminths (Maynard, DeMartini & Wright 1996; Helluy & Thomas 2003; Tain, Perrot-Minnot & Cézilly 2006, 2007).

Bethel & Holmes (1973, 1977) were the first to produce evidence that amphipods infected with acanthocephalans showed both altered reaction to light and increased vulnerability to predation by appropriate final hosts. Since then, a causal link between altered phototaxis or photophobia in intermediate hosts infected with helminths and increased trophic transmission to final hosts has been repeatedly advocated (e.g. Moore 1983; Helluy 1984; Bakker, Mazzi & Zala 1997; Bauer et al. 2000; Cézilly, Grégoire & Bertin 2000; Sanchez, Georgiev & Green 2007; Benesh et al. 2008; Guler & Ford 2010). However, the evidence remains essentially correlational (Cézilly et al. 2010), and there is, to date, no direct proof that the alteration of phototaxis co-occurring with infection contributes effectively to increase trophic transmission of the parasite. Alternatively, infected intermediate hosts might be more vulnerable to predation simply because they are weaker and less efficient at escaping from predator attacks (Cézilly et al. 2010). Indeed, several studies indicate that prey infected with parasites show reduced ability to escape from predators (Libersat & Moore 1999; Alzaga et al. 2008; Goodman & Johnson 2011; Luong, Hudson & Braithwaite 2011; but see Frank & Kurtz 2002).Thus, the adaptiveness of the alteration of reaction to light in hosts infected by parasites with trophic transmission remains to be demonstrated convincingly (Poulin 2007).

The purpose of this study was to assess to what extent altered reaction to light in the amphipod Gammarus pulex infected with the fish acanthocephalan, Pomphorhynchus tereticollis (Tain, Perrot-Minnot & Cézilly 2006), is influencing vulnerability to a fish predator. To address this question, we relied on experimental modulation of photophobia and on predation tests in microcosms offering to a goldfish, a mixture of amphipods differing in their reaction to light. In a first experiment, the extent of difference in photophobia between infected and uninfected hosts was controlled through manipulating light intensity during predation tests. In a second experiment, we altered photophobia in uninfected amphipods by injecting a neuromodulator (serotonin) known to reverse natural photophobia in G. pulex (Tain, Perrot-Minnot & Cézilly 2006). Such phenotypic engineering aimed at mimicking the decreased photophobia of P. tereticollis-infected amphipods in uninfected ones prior to predation tests. Following the host manipulation hypothesis, (1) the lower the extent of difference in photophobia between infected and uninfected individuals is the lower the predation bias towards infected amphipods should be and (2) uninfected serotonin-injected amphipods should be more vulnerable to predation than uninfected amphipods injected with the vehicle solution. Alternatively, if the increased susceptibility of infected amphipods is attributable to their reduced stamina, and not to altered reaction to light, differences in photophobia should have no effect on predation bias.

Material and methods

Predator-prey model

We collected G. pulex from two distinct populations located in Burgundy, eastern France, one located in the river Suzon (47°41N, 04°09E), where no infection with P. tereticollis has ever been detected, and the other located in the river Cbiron (45°47N, 05°31E), where the prevalence of infection with fish acanthocephalan P. tereticollis varies between 5% and 30%, depending on time of the year (M. J. Perrot-Minnot, unpublished data). Gammarids were acclimated in the laboratory for at least 5 days prior to experiments. They were maintained at 15 °C in large tanks filled with oxygenated dechlorinated UV-treated water, and fed with elm leaves.

Previous studies using familiar fish predators have shown that a fish odour is repulsive to uninfected amphipods, but attractive to amphipods infected with fish acanthocephalans (Baldauf et al. 2007; Perrot-Minnot, Kaldonski & Cézilly 2007). Therefore, to assess solely the effect of decreased photophobia on the vulnerability of the amphipod G. pulex to fish predation, it was first necessary to control for the potential effect of predator olfactory cues. As recognition of predator odour is expected to be a learned response (Wisenden, Cline & Sparkes 1999), we used the exotic fish, Carassius auratus, as a predator to which the amphipods were presumably naïve. Indeed, in both rivers where the amphipods were sampled, the natural fish predators of G. pulex were predominantly Cyprinidae, Cottidae and Salmonidae (M.-J. Perrot-Minnot, unpublished data). Goldfish used in the predation experiments were between 12 and 15 cm in length. They were purchased from a pet shop and were maintained in the laboratory in large tanks filled with oxygenated dechlorinated UV-treated water and fed with TetraCichlid granules. Each individual fish was recognizable by a distinctive phenotypic trait or by a small notch on the caudal fin.

The lack of anti-predatory response towards chemical cues from goldfish was verified using a choice test in a Y-maze olfactometer, following Perrot-Minnot, Kaldonski & Cézilly (2007). Fish-conditioned water was prepared by leaving six goldfish for 24 h in 48 L of oxygenated dechlorinated tap water, giving a ratio of about 9 g L−1. This ratio of fish biomass to water volume was comparable to previous studies, addressing the effect of fish chemical cues on invertebrate behaviour (Åbjörnsson et al. 1998; Pennuto & Keppler 2008). When given the choice between fish-conditioned water and control water, P. tereticollis-infected gammarids spent an equal amount of time in each arm of the Y-maze (median time in the scented arm: 40·3%, quartiles: 11·8–72·7%, = 18; Wilcoxon test against the value of 50%: = −22·5, = 0·35). The same was true of uninfected gammarids (time in the scented arm: 59·5%, 30·2–91·4%, = 20; Wilcoxon test against the value of 50%: = 29·5, = 0·28). In addition, parasitized and uninfected G. pulex did not differ in the percentage of time spent in the scented arm of the Y-maze (Wilcoxon–Mann–Whitney test: = −1·58, d.f. = 1, = 0·11). Overall, then, olfactory cues from goldfish were neither attractive nor repulsive to infected or uninfected amphipods.

Photophobia and vulnerability to predation of P. tereticollis-infected gammarids at two light intensities

Reaction to light was quantified in uninfected and infected amphipods collected in the river Cbiron through measuring photophobia in a tube following the procedure described in Perrot-Minnot (2004). Each individual amphipod was given a choice between a light zone and a dark zone of identical volumes, in one of 10 glass tubes. Illumination was provided by 36 W solar spectra fluorescent tubes (OSRAM Lumilux-865, Molsheim, France) placed above the set of 10 experimental glass tubes. Scan sampling of the gammarids' position was performed every 30 s during 5 min, after an initial acclimatization period of 5 min. At each step in time, each individual was scored as either 1 (present on the light side) or 0 (not present on light side). Individual score thus varied from 0 (strongly photophobic) to 10 (strongly photophilic), with a score of 5 indicating indifference to light. For each individual, photophobia was measured at a 24-h interval at two different light intensities, 20 and 1500 lux. These light intensities are in the range of variation in the natural environment of amphipods at the time they were collected, i.e. under weak illumination at night and under a cloudy sky during the day, respectively (Lagrue et al. 2007). Light intensity was controlled by varying the height of the tubes above the test apparatus and checked to the nearest 1 lux with a light meter (HI97500; Hanna Inst., Tanneries, France). Half of the amphipods were first tested at the lower light intensity while the other half was first tested at the higher one.

Following Kaldonski, Perrot-Minnot & Cézilly (2007) and Perrot-Minnot, Kaldonski & Cézilly (2007), we assessed the differential vulnerability of uninfected vs. infected amphipods collected in the river Cbiron to predation by goldfish in microcosms. The test aquarium (60 × 25 × 35 cm) was divided in two unequal parts (one-third and two-thirds) with a perforated plastic partition, allowing chemical and visual cues but no physical contact between predator and prey. It was surrounded with opaque screens and illuminated by overhead solar spectra fluorescent tubes at 20 or 1500 lux intensities. In the large part, two pieces of airbrick (22 × 5·3 × 5·3 cm) were provided as a refuge. The aquarium was filled with 30 L of oxygenated UV-treated and dechlorinated water, which was renewed after each trial. A single goldfish, previously food-deprived for 24 h, was introduced in the smaller part of the aquarium, whereas 30 infected gammarids and 60 uninfected ones were introduced in the larger one. The plastic partition was removed after a 1-h acclimatization period, and the goldfish was allowed to feed on gammarids for 30 min. This length of time was determined from preliminary experiments, to ensure that the total number of prey captured would be lower than the initial number of infected prey. At the end of the predation period, the fish was removed, and remaining gammarids of each type (infected or not) were counted. Each fish (= 9) was tested twice for selectivity in predation, once at 20 lx and once at 1500 lx, in a randomized running order.

Photophobia and vulnerability to predation of uninfected G. pulex injected with 5-HT+ fluoxetine

The contribution of decreased photophobia in G. pulex to its vulnerability to predation was also assessed by engineering their reaction to light pharmacologically prior to predation trials in microcosms. Previous studies (Helluy & Holmes 1990; Tain, Perrot-Minnot & Cézilly 2006) have shown that the injection of 5-HT in amphipods results in decreased photophobia, thus mimicking the behaviour of acanthocephalan-infected individuals. In the present study, uninfected gammarids were injected with 1 μL of a mixture of serotonin (5-HT) and fluoxetine, a serotonin selective reuptake inhibitor at a 2 : 1 ratio (5-HT: fluoxetine; thereafter referred to as 5-HTF), in a vehicle solution (crustacean ringer). Fluoxetine was added to 5-HT to ensure that the effect of 5-HT on reaction to light would last for at least 1·5 h (M.-J. Perrot-Minnot, unpublished data). The dose of neuromodulator injected to amphipods was adjusted as to mimic the decreased photophobia observed in G. pulex infected with P. tereticollis, i.e. 3·8 μg of serotonin mixed with 1·9 μg of fluoxetine. It was delivered through a single injection using a Hamilton syringe (701RN) with a fine needle (RN33/51/3), following Tain, Perrot-Minnot & Cézilly (2006). Two controls were run simultaneously: gammarids injected with the vehicle saline solution and non-injected gammarids. Photophobia was recorded one hour after injection, as described above.

Predation trials on 5-HTF-injected and ringer-injected gammarids were performed in the same way as the predation experiment at two light intensities, except that only 30 gammarids were offered (10 injected with 5-HTF, 20 with ringer), the acclimatization and predation periods lasted 45 min and 15 min, respectively, and a single brick was provided as a refuge. Control trials were run with 10 ringer-injected gammarids mixed with 20 non-injected ones. The former trial was testing the effect of 5-HTF injection on vulnerability to predation (thereafter referred to as 5-HTF trial), while the latter was testing the effect of injection per se on vulnerability to predation (thereafter referred to as control trial). All tests were performed at the 1500 lux illumination. At the end of each predation test, the fish was removed and the remaining gammarids were sorted by type (injected with 5-HTF or ringer; ringer-injected or non-injected). Injected amphipods were easily recognizable by the melanization of the injection point, a few hours after injection. Ringer and 5-HTF solutions were injected either on the left or on the right side of amphipods, alternating from one trial to the other. This experiment was performed using the gammarid population from the river Cbiron, as in the first experiment. It was replicated at the same time on another population from river Suzon, known to be highly responsive to 5-HT injection (M.-J. Perrot-Minnot, unpublished data).

Statistical analysis

We relied on nonparametric statistics (Siegel & Castellan 1988) to analyse photophobia. Scores in reaction to light were compared between groups using Wilcoxon–Mann–Whitney test, or, when appropriate, the Wilcoxon signed-rank test for tied ranks on paired data (Siegel & Castellan 1988). Following Kaldonski, Perrot-Minnot & Cézilly (2007) and Perrot-Minnot, Kaldonski & Cézilly (2007), selectivity during predation test was assessed using Manly's index (Manly 1974), which allows for the depletion of prey during the course of the trial, and thus for changes in the proportions of available prey classes as prey are eaten. We calculated Manly's alpha for variable prey population using the equation

display math

where αi is the Manly's alpha (preference index) for prey type I, pi and pj are the proportion of prey i or j remaining at the end of the trial and m is the number of prey types. Manly's selectivity index ranges from 0 (when only control prey are eaten: uninfected or ringer-injected) to 1 (when only focal prey are eaten: infected, 5-HTF-injected or ringer-injected). A value of 0·5 indicates an absence of preference. Observed values of αi were compared with a situation of equal vulnerability using the Wilcoxon signed-rank test to a hypothesized value (αi = 0·5) for small sample size (Siegel & Castellan 1988). Comparisons of preference indexes between treatments were carried out using the Wilcoxon signed-rank test on paired data, as each individual fish was used twice (at the two light intensities in the first experiment, and for both 5-HTF and control trials in the second experiment). All tests were bilateral and considered significant at the 5% level.

Results

Photophobia and vulnerability to predation of P. tereticollis-infected gammarids at two light intensities

Pomphorhynchus tereticollis-infected G. pulex exhibited reduced photophobia compared to uninfected individuals at both light intensities (Wilcoxon–Mann–Whitney test: 1500 lux, = −3·44, = 0·0006; 20 lux, = −3·00, = 0·0026; Fig. 1). Indeed, uninfected amphipods were significantly photophobic at 1500 lux and 20 lux (Wilcoxon signed-rank test for large sample and tied ranks against the hypothesized value of 5 : 1500 lux, = 38, T+ = 22, < 0·0001; 20 lux, = 34, T+ = 68, < 0·0001), whereas P. tereticollis-infected ones were indifferent to light both at 1500 lux and 20 lux (= 36, T+ = 262·5, = 0·98; and = 37, T+ = 343, = 1, respectively) (Fig. 1). The photophobia of uninfected gammarids was, however, significantly stronger at 1500 lux than at 20 lux (Wilcoxon signed-rank test for large sample and tied ranks: = 31, T+ = 105, = 0·0003), whereas no difference in reaction to light was observed in P. tereticollis-infected individuals between the two light intensities (= 36, T+ = 282, = 0·39). Individual photophobia showed consistency across light intensities in uninfected amphipods (Spearman correlation test, = 39, Rs = 0·38, = 0·016), but not in P. tereticollis-infected ones (Spearman correlation test, = 40, Rs = −0·19, = 0·282). Overall, the difference between the median scores of uninfected and infected individuals was two times more pronounced at high compared to low light intensity.

Figure 1.

Phototactic behaviour in Pomphorhynchus tereticollis-infected (P) and uninfected (U) Gammarus pulex under 1500 lux and 20 lux illumination. Figures show median phototactism scores, interquartile range and sample size.

We thus relied on this contrasted magnitude of differences in photophobia between infected and uninfected amphipod at 1500 lux and 20 lux to estimate the contribution of decreased photophobia to trophic transmission. Thanks to a large number of prey initially offered, parasitized prey were never totally depleted, with at least 28% per cent of the initial number of infected gammarids remaining at the end of the test. Therefore, all trials were kept for analysis. The predation bias towards P. tereticollis-infected prey was significant at both light intensities, as assessed from Manly's αi preference index (Wilcoxon signed-rank test to the hypothesized value of 0·5: = 9, T+ = 48, < 0·004 at each light intensity) (Fig. 2). Predation bias towards infected G. pulex was very high and did not differ significantly between the two light intensities (Wilcoxon signed-rank test: = 9, T+ = 28, = 0·57). On average, 48·5% of infected gammarids and only 10·2% of uninfected ones were caught by goldfish at 1500 lux. Similarly, 50% of infected gammarids and only 13% of uninfected ones were caught by goldfish at 20 lux. The total number of prey caught did not differ significantly between 1500 lux and 20 lux (paired t-test: = −0·74, d.f. = 8, = 0·48).

Figure 2.

Preference index expressing selectivity of goldfish on Pomphorhynchus tereticollis-infected vs. uninfected Gammarus pulex at two light intensities (box-plot showing median, interquartiles and min–max values; paired trials, N = 9). Values above the dotted line indicate overconsumption of infected prey. Asterisks above the bars indicate significant selectivity for infected prey over uninfected ones (Wilcoxon signed-rank test to the hypothesized value of 0·5, < 0·004, at each light intensity). Manly's preference index was not significant between the two light intensities (Wilcoxon matched pairs test, = 0·57).

Photophobia and vulnerability to predation of uninfected G. pulex injected with 5-HT+ fluoxetine

In both populations, ringer-injected uninfected gammarids were as photophobic as non-injected ones (Wilcoxon–Mann–Whitney test; river Cbiron: = 35 and = 39, = 1·05, P = 0·29; river Suzon: = 28 and = 30, = −0·57, P = 0·57, Fig. 3). Therefore, the two controls were pooled for further analysis. Significant differences in photophobia were found between P. tereticollis-infected G. pulex, G. pulex injected with 5-HTF and control ones, in amphipods collected from the river Cbiron (Kruskal–Wallis test: = 30·23, d.f. = 2, < 0·0001; Fig. 3). The injection of 5-HTF significantly decreased the photophobia of G. pulex (a posteriori multiple comparison test to uninfected controls: < 0·01), at a level not significantly different from that of G. pulex infected with P. tereticollis (a posteriori multiple comparison test: > 0·05; Fig. 3). Thus, the injection was successful in mimicking the effect of infection. However, while infected amphipods were indifferent to light (Wilcoxon signed-rank test for large sample and tied ranks against the value of 5: = 31, T+ = 302·5, = 0·16), 5-HTF-injected ones were slightly photophobic (= 38, T+ = 273, = 0·03). Controls were strongly photophobic (= 71, T+ = 197·5, < 0·0001).

Figure 3.

Phototactic behaviour in Gammarus pulex from two rivers (Cbiron, Suzon): uninfected gammarids injected with serotonin (5-HT), injected with the vehicle solution (R) or not injected (U) and gammarids infected by Pomphorhynchus tereticollis (P). Figures show median phototactism scores, interquartile range and sample size.

Gammarus pulex from the river Suzon injected with 5-HTF also showed decreased photophobia compared to control individuals (Wilcoxon–Mann–Whitney test: = 28 and = 58, respectively, = 4·81, < 0·0001; Fig. 3). Uninfected and ringer-injected gammarids from the river Suzon were photophobic (Wilcoxon signed-rank test with large sample and tied ranks against the value of 5: = 51, T+ = 324, < 0·0001), whereas 5-HTF-injected gammarids were highly photophilic (= 27, T+ = 329·5, < 0·0001) (Fig. 3).

Injected gammarids were never depleted in any of the predation trials, with at least 30% of the initial number of 5-HTF- or ringer-injected gammarids left at the end of a trial. Therefore, trials for which at least five preys had been eaten were kept for analysis. No selective predation was detected on gammarids injected with 5-HTF in both populations (Fig. 4), as the preference index in 5-HTF trials did not differ from random expectation (Wilcoxon signed-rank test against the hypothesized value of 0·5: Cbiron, = 9, T+ = 23, > 0·50; Suzon, T+ = 37, = 0·10). In both populations, control trials were also consistent with a random capture of ringer-injected and non-injected gammarids (river Cbiron: = 7, T+ = 23, > 0·50; river Suzon: = 9, T+ = 23, > 0·50). The Manly's index of selectivity in 5-HTF trials was not significantly different from that of control trials (Wilcoxon matched pairs test: River Cbiron: = 9,T+ = 27, = 0·65; River Suzon : = 9, T+ = 29, = 0·50).On average, 31·1% and 41·1% of 5-HT-injected gammarids and 32·2% and 29·4% of ringer-injected ones were caught by goldfish in 5-HTF trials, in the river Cbiron and the river Suzon, respectively. Similarly, 32·8% and 28·9% of ringer-injected gammarids and 42·2% and 27·8% of uninfected ones were caught by goldfish in control trials, in the river Cbiron and the river Suzon, respectively.

Figure 4.

Preference index expressing selectivity of goldfish on Gammarus pulex injected with serotonin vs. controls injected with ringer vehicle solution [5-HT: fluoxetine (5-HTF)/R] and on G. pulex injected with ringer vs. non-injected controls (R/C) (box-plot showing median, interquartiles and min–max values; paired trials, = 9). The dotted line indicates equal consumption of both types of prey. Selectivity was not significant in any trial (Wilcoxon signed-rank test to the hypothesized value of 0·5: > 0·05).

Although 5-HTF-injected gammarids from river Cbiron did not differ from P. tereticollis-infected ones in their reaction to light, predation bias towards the latter in the first experiment at 1500 lux was significantly higher than predation bias towards the former in the second experiment (Wilcoxon–Mann–Whitney test, N = 8, = 10, = 41, = 0·0015). Conversely, the magnitude of predation bias (as assessed from Manly's index of selectivity) in the 5-HTF trials did not differ between the two populations tested (Wilcoxon matched pairs test: = 6, T+ = 8, = 0·68), despite the strong photophily of 5-HTF-injected gammarids from river Suzon compared to that of 5-HTF-injected gammarids from river Cbiron (Wilcoxon matched pairs test, = 28, = 40, Z = 3·87, = 0·0001).

Discussion

According to the ‘parasite manipulation hypothesis’, decreased photophobia of intermediate host acts to increase their vulnerability to a fish predator, and, hence trophic transmission to definitive hosts. We were able to test this hypothesis in two different ways: (i) the difference in photophobia between P. tereticollis-infected and uninfected gammarids was lower at weak light intensity and (ii) photophobia of uninfected gammarids from two populations was pharmacologically decreased or even reversed, by injecting a mixture of 5-HT and fluoxetine.

Amphipods infected by P. tereticollis were more vulnerable to predation by fish than uninfected ones, thus confirming the results of previous studies (Bakker, Mazzi & Zala 1997; Kaldonski, Perrot-Minnot & Cézilly 2007; Lagrue et al. 2007; Perrot-Minnot, Kaldonski & Cézilly 2007; Kaldonski et al. 2009; Cézilly et al. 2010). However, the magnitude of selective predation on naturally-infected amphipods did not vary in relation to variation in the difference in reaction to light between uninfected and infected individuals. In addition, following injection with 5-HTF, amphipods were not more vulnerable to predation that control individuals, although they were less photophobic compared to ringer-injected ones. Overall, our results indicate that altered photophobia in itself does not enhance parasite transmission.

Taken together, our results indicate that, contrary to previous claims (Bethel & Holmes 1973, 1977; Helluy 1984; Bakker, Mazzi & Zala 1997; Bauer et al. 2000; Cézilly, Grégoire & Bertin 2000; Benesh et al. 2008), a decrease in photophobia is not the main causative agent of the increased vulnerability of infected amphipods to predation by fish. The observed higher vulnerability of infected amphipods may thus be due either to other specific phenotypic alterations brought about by the parasites (but see Kaldonski et al. 2009), or, more simply, to their reduced stamina (as evidenced in crustacean intermediate hosts infected with acanthocephalan parasites at both individual and population levels, Lettini & Sukhdeo 2010), making them less efficient at escaping from predator attacks (Cézilly et al. 2010). The strong bias towards infected amphipods in predation by goldfish, together with the absence of reaction of infected and uninfected gammarids to chemical cues from that predator, suggests, however, that the second interpretation might be more parsimonious. Alternatively, several behavioural changes, photophobia included, might act synergistically to enhance trophic transmission while not contributing significantly when expressed singly. Such a multidimensional approach to parasite manipulation has been recently advocated to address the adaptiveness of parasite-induced alterations (Cézilly et al. 2010; Thomas, Poulin & Brodeur 2010). However, evidence for adaptive multidimensionality in parasitic manipulation is still missing (Cézilly et al. 2010; Thomas, Poulin & Brodeur 2010).

Our results have several important implications for the study of host manipulation by parasites. Several criteria have been advanced to assess the adaptiveness of parasite-induced phenotypic alterations in intermediate hosts (Poulin 1995, 2007). According to the argument from ‘purposive design’ (Poulin 1995, 2007), a close fit between the presumed function of a parasite-induced behavioural alteration and the parasite's life cycle would provide reasonable evidence for adaptive manipulation (Poulin 2007). However, the adaptive consequences of parasite-induced phenotypic alterations may not necessarily be so obvious that it would be unnecessary to question the validity of the host manipulation hypothesis. For instance, although tenebrionid beetles infected with the cestode Hymenolepis diminuta display behavioural alterations (reduced concealment) that would logically increase the likelihood of the intermediate host being predated by the Rattus sp. definitive host, no difference in predation rates has been observed between infected and uninfected beetles (Webster et al. 2000). The present study goes further through showing that even a close association between increased trophic transmission and purposive design in behavioural alteration is no evidence for a causal relationship.

Another criterion advanced to assess the adaptiveness of parasite-induced behavioural alteration is the timing of appearance of the behavioural alteration in relation to the developmental stage of the parasite (Poulin 2007). Indeed, acanthocephalan parasites are known to alter reaction to light in their intermediate hosts only when reaching the cystacanth stage which is infective for the definitive host (Bethel & Holmes 1974; Franceschi et al. 2008). However, relying on this criterion to infer adaptive manipulation would obviously be misleading in the present case, where the trait under study, although altered only after the parasite has reached the infective stage, is not in itself enhancing parasite transmission.

Our results also cast doubt on the validity of the interpretation of some aspects of parasite-induced behavioural alterations. For instance, Dianne et al. (2010) interpreted the lower decrease in photophobia of G. pulex harbouring both infective and non-infective larvae of a fish acanthocephalan, as evidence for competitive interaction between different developmental stages with opposite interests. Similarly, Cézilly, Grégoire & Bertin (2000); (see also Rauque et al. 2011) interpreted the level of altered photophobia in G. pulex co-infected with both a fish and a bird acanthocephalan as evidence for conflict in manipulation. However, in the absence of evidence for a direct effect of decreased photophobia on trophic transmission or of evidence for correlated behaviours together acting synergistically on trophic transmission, such an interpretation is not appropriate.

More generally, our study indicates that the criterion of paramount importance in assessing the adaptiveness of parasite-induced behavioural alteration should be its effective contribution to increased trophic transmission. As noted by Poulin (2007), several ‘classical’ examples of host manipulation still await confirmation that the observed phenotypic alterations are indeed enhancing trophic transmission, possibly owing to the difficulty in isolating the contribution to parasite transmission of a single altered behaviour. Relying on phenotypic engineering and predation tests in controlled environments may help to solve the issue. Until then, some caution should be exerted when resorting to the concept of host manipulation to interpret the effects of parasites on the phenotype of their hosts, in order to avoid the pitfalls of extreme adaptationism (see Cézilly & Perrot-Minnot 2010).

Acknowledgments

We thank Quentin Mori for technical assistance, and the Institut Universitaire de France for financial support.

Ancillary