INTRODUCTION
Multiple species of parasites rely on trophic transmission (consumption of infected intermediate hosts by an appropriate definitive-host predator) to complete their life cycle (Parker et al. Reference Parker, Chubb, Michaud and Milinski2003). Such strategy creates strong selective pressures on these parasites to increase the probability of intermediate hosts predation by final hosts (Lafferty, Reference Lafferty1999; Moore, Reference Moore2002). Some trophically transmitted parasites have evolved the ability to alter intermediate host phenotypes in ways that should increase transmission probabilities to definitive hosts (Moore, Reference Moore2002; Thomas et al. Reference Thomas, Adamo and Moore2005; Kaldonski et al. Reference Kaldonski, Perrot-Minnot and Cézilly2007). Acanthocephalan parasites are known to manipulate several behavioural traits of crustacean hosts, increasing their vulnerability to predation by definitive host predators (Poulin, Reference Poulin1995; Lafferty, Reference Lafferty1999; Cézilly et al. Reference Cézilly, Gregoire and Bertin2000; Baldauf et al. Reference Baldauf, Thünken, Frommen, Bakker, Heuprl and Kullmann2007; Médoc and Beisel, Reference Médoc and Beisel2011). Some species have been documented to alter habitat selection by intermediate hosts, infected individuals showing spatially divergent distributions compared to uninfected ones, resulting in patches of high parasite prevalence (MacNeil et al. Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003; Wellnitz et al. Reference Wellnitz, Giari, Maynard and Dezfuli2003; Médoc et al. Reference Médoc and Beisel2009). For example, Gammarus pulex amphipods infected with Echinorhynchus truttae aggregated in fast-flowing, shallower stretches of river, areas below which drift-feeding fish, the parasite definitive hosts, usually congregate (MacNeil et al. Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003). Changes in the micro-distribution of intermediate hosts are often interpreted as a consequence of host behavioural manipulation, an adaptive strategy of the parasite to increase transmission probabilities. Infected hosts move in or closer to definitive hosts’ habitats and/or feeding areas, thus increasing their vulnerability to predation and parasite transmission probabilities (MacNeil et al. Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003; Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007; Médoc and Beisel, Reference Médoc and Beisel2009). However, the direct link between infected intermediate host altered spatial distribution and parasite transmission to definitive hosts is almost never tested.
Alterations of intermediate host distributions could also expose infected hosts to predation by other predators, non-definitive hosts (i.e. dead-end predators that parasites cannot infect; resulting in parasite death after intermediate host consumption). This may strongly reduce the advantage of host behavioural manipulation (Marriott et al. Reference Marriott, Collins, Paris, Gudgin, Barnard, McGregor, Gilbert, Hartley and Behnke1989; Mouritsen and Poulin, Reference Mouritsen and Poulin2003), although not necessarily eliminating all benefits of such strategy (Seppälä and Jokela, Reference Seppälä and Jokela2008). Infected host distribution may be different in response to the presence of suitable hosts compared to other predators, preferentially exposing infected hosts to predation by the appropriate definitive host and/or decreasing its vulnerability to predation by non-host predators (Levri, Reference Levri1998; Médoc et al. 2006, 2009). Polymorphus minutus-infected amphipods aggregate onto floating material while their uninfected conspecifics preferentially shelter in river-bottom substrate. Such parasite-induced alterations in host micro-distribution patterns are thought to move infected individuals closer to P. minutus definitive bird hosts and away from dead-end predators like fish and invertebrates, thus potentially increasing the parasite's transmission probabilities while avoiding non-host predator species (Médoc and Beisel, Reference Médoc and Beisel2009; Médoc et al. Reference Médoc, Rigaud, Bollache and Beisel2009). However, counter examples exist where P. minutus made their crustacean hosts significantly more vulnerable to consumption by fish (Marriott et al. Reference Marriott, Collins, Paris, Gudgin, Barnard, McGregor, Gilbert, Hartley and Behnke1989). Polymorphus minutus-infected G. pulex vulnerability to dead-end predators also varies depending on habitat complexity (i.e. refuge availability) and predator type (i.e. sit-and-wait or active hunter), potentially eroding host manipulation benefits for the parasite (Kaldonski et al. Reference Kaldonski, Perrot-Minnot, Motreuil and Cézilly2008; Médoc and Beisel, Reference Médoc and Beisel2008). Also, most studies looking at infected host vulnerability to definitive host and/or non-host predators are performed under laboratory conditions rather than in situ, and usually focused on a single parasite-predator species association (Médoc and Beisel, Reference Médoc and Beisel2009).
Overall, field data on links between possible parasite-induced changes in infected intermediate host distribution, host and non-host predator presence, and predation rates on infected hosts are generally lacking. In particular, little is known about the specificity of parasite-induced habitat shifts in intermediate hosts (Médoc and Beisel, Reference Médoc and Beisel2009). The present study investigated the micro-distribution of acanthocephalan-infected intermediate hosts in response to the absence or presence of predators, either suitable definitive hosts or non-host predators. First, we examined the prevalence of 2 acanthocephalan parasites, Echinorhynchus truttae and Polymorphus minutus, in their common intermediate host (Gammarus pulex), in the absence or immediate vicinity of both suitable definitive hosts and non-host predators in the field. Second, we assessed whether infected intermediate host micro-distributions resulted in increased parasite transmission to the definitive host and/or avoidance of non-host predators; these two aspects of parasite transmission being often the main characteristics required to qualify host behavioural manipulation as adaptive (Poulin, Reference Poulin1995; Cézilly and Perrot-Minnot, Reference Cézilly and Perrot-Minnot2005; Seppälä and Jokela, Reference Seppälä and Jokela2008; Médoc and Beisel, Reference Médoc and Beisel2011).
MATERIALS AND METHODS
Study site and animal material
Field experiments were carried out in spring 2011 (25 March to 25 April) in a second-order permanent karst spring, tributary of the Suzon River (47°24′14.45″N, 4°53′1.46″E), about 20 km north-west of Dijon, Eastern France. It is characterized by very stable water levels and temperatures, due to subterranean karstic origins, and a succession of small riffle-pool structures. The stream contains high densities of Gammarus pulex amphipod crustacean. Low densities of bullhead (Cottus gobio) can be found all year, while juvenile brown trout (Salmo trutta) are observed in spring and adults in late autumn during spawning season (Klemetsen et al. Reference Klemetsen, Amundsen, Dempson, Jonsson, Jonsson, O'Connell and Mortensen2003); the stream being used as a spawning ground by adult S. trutta from the larger, adjacent Suzon River. Adult S. trutta, migrating in the stream to spawn, are likely to be the main source of E. truttae infection to G. pulex as the highest prevalence of mature cystacanths are observed in spring, 3–4 months after trout spawning, the time necessary for E. truttae to reach the infective cystacanth stage (Awachie, Reference Awachie1966). Field experiments were thus run at pick prevalences of mature E. truttae cystacanths and at a time when host behavioural manipulation should be maximal (Benesh et al. Reference Benesh, Hasu, Seppälä and Valtonen2009).
Gammarus pulex amphipods present in the stream are intermediate hosts to 2 species of acanthocephalan parasites, Echinorhynchus truttae and Polymorphus minutus. Echinorhynchus truttae uses both C. gobio and S. trutta as definitive hosts (Awachie, Reference Awachie1973; Okada and Koura, Reference Okada and Koura2000). Adult P. minutus are bird parasites (Holmes and Bethel, Reference Holmes and Bethel1972), mainly infecting mallard duck (Anas platyrhynchos) and white-throated dipper (Cinclus cinclus) in our study stream. These acanthocephalan parasites are characterized by a complex, 2-host life-cycle. They mature and sexually reproduce in vertebrate definitive hosts. Eggs are then released in the water with host feces and must be consumed by an amphipod crustacean intermediate host. The life cycle is completed when the infected crustacean is eaten by the appropriate definitive host (Crompton and Nickol, Reference Crompton and Nickol1985).
Predators used in our study were bullheads (C. gobio), being suitable definitive hosts to E. truttae, and the noble crayfish (Astacus astacus) as the non-host predator to both E. truttae and P. minutus. Both species are present in the Suzon river catchment and are known to feed on invertebrates, including amphipods (Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007; Kaldonski et al. Reference Kaldonski, Perrot-Minnot, Motreuil and Cézilly2008; Haddaway et al. Reference Haddaway, Wilcox, Heptonstall, Griffiths, Mortimer, Christmas and Dunn2012). Bullheads were captured by electro-fishing in an upstream stretch of the Suzon River where the parasites are absent. Crayfish were bought from a fish farm (Loro's Farm, Loromontzey, North-Eastern France; 48°26′38.42″N, 6°22′1.95″E) where A. astacus is raised in large ponds, feeding on naturally occurring prey, including amphipods.
Host density and parasite prevalence and distribution
Field experiments consisted of 5 blocks of 3 cages (15 in total), each cage haphazardly assigned to one of 3 treatments: control, crayfish or bullhead. Blocks 1 to 5 consisted in structurally similar riffles evenly spaced (20 m) along a 100 m reach of the stream. Cages were cylindrical (Ø = 30 cm and height = 50 cm), made of 10 mm steel mesh and designed such that predator (bullhead and crayfish) foraging behaviour and access to prey would not be restricted. Mesh aperture and cage height allowed all prey movements across cage boundaries and over the entire water column (15–30 cm water depth) while preventing other naturally occurring fish predators from accessing amphipod traps. Cages were imbedded 5 cm deep into the streambed and anchored to the river bottom using iron rods. Gravel was then added so that substratum surfaces inside and outside the cages were levelled. Control treatment cages contained no predator, each crayfish cage contained 1 adult individual of the noble crayfish (Astacus astacus; total length >10 cm) and each fish cage contained 3 adult bullheads (C. gobio; mean total length±s.e. = 8·1 ± 0·3 cm). Cages within each block were at a distance of 2 m to avoid possible chemical or visual interactions between predators. Predator numbers were such that there was an equivalent predator biomass between crayfish and bullhead treatments and reflected natural predator densities. Crayfish were selected as natural predators of amphipods and unsuitable hosts for E. truttae and P. minutus; Cottus gobio being a definitive host for E. truttae but not P. minutus.
Autumn-shed leaves of field elm [Ulmus minor (Mill.)] were used as food/habitat in the amphipod traps set out in treatment cages. We used air-dried leaves to make standardized leaf packs to capture G. pulex amphipods. After wetting, leaves were enclosed in 10 mm steel mesh cages (15 × 10 × 1 cm) to constitute a trap and assigned to one of the 3 treatments (5 control, 5 crayfish and 5 bullhead, respectively); the large mesh size and low trap height (1 cm) allowed both bullheads and crayfish to access and predate on amphipods present within the traps or migrating towards and from these. One trap was then added in each treatment cage, maintained flat against the substrate using iron rods and left in the cage for 10 days before retrieval. Elm leaf mass used in each trap allowed amphipods to feed without noticeably impacting the resource; food never ran out between sampling dates. Traps were designed to allow immigration and emigration of amphipods colonizing elm leaves so that samples reflected an amphipod distribution equilibrium, and not amphipod accumulation, in each trap at each sampling date. Amphipods were sampled 3 times during the experiment (4, 14 and 25 April, respectively), allowing 10 days for amphipod colonization before each sampling date. Traps were swiftly recovered from treatment cages and put into a plastic tray to capture all amphipods contained within. Amphipods were sorted from elm leaves, collected in a 0·5 mm mesh sieve and preserved in 70% ethanol before laboratory analyses. Leaf packs were replaced by fresh ones within each trap after each sampling date. In the laboratory, all amphipods were counted and G. pulex densities expressed as captures per unit effort (i.e. CPUE), corresponding to the number of G. pulex individuals per amphipod trap. Individuals were measured (total body length to the closest 0·5 mm) and, when possible, sexed from the shape and size of segment 6 (propodus) of gnathopods 1 and 2 (Bollache and Cézilly, Reference Bollache and Cézilly2004). Amphipods were then dissected under a dissecting microscope to record and identify possible acanthocephalan parasites. Only mature cystacanths, developmental stage infective to definitive hosts, were recorded and considered in our analyses (Awachie, Reference Awachie1966); host manipulation and thus infected amphipod aggregation around predator hosts being expected only for mature, infective acanthocephalan cystacanths (Dianne et al. Reference Dianne, Perrot-Minnot, Bauer, Gaillard, Leger and Rigaud2011). Prevalence of each parasite in each treatment could thus be determined and compared.
Potential differences in amphipod densities between treatments, sampling dates and blocks were tested using main-effects ANOVAs and trap/cage as the sampling unit. Amphipod numbers (CPUE) were log-transformed before analyses to normalize the data. Generalized Linear Models (GLMs) were used to analyse the effects of the different factors on the presence or absence of acanthocephalan parasites in amphipods, hereafter referred to as parasite prevalence (proportion of infected hosts). We examined the effects of treatment (control, crayfish or bullhead), amphipod host sex (male or female), sampling date (4, 14 or 25 of April) and block (1 to 5) as fixed factors and amphipod host size as a covariate, on acanthocephalan parasite prevalence (dependent variable), using individual amphipods as the sampling unit. Two GLMs were run, one for each parasite species, E. truttae and P. minutus. Finally, linear regressions between amphipod CPUE and parasite prevalence in each block were used to test for a potential relationship between G. pulex density and E. truttae or P. minutus prevalences and thus possible differences in micro-distribution of infected and uninfected hosts.
Fish diet and parasite transmission
Fish diet was analysed in situ by collecting stomach contents of C. gobio maintained in the 5 bullhead treatment cages described above, allowing comparisons between parasite prevalences recorded in the bullhead treatment (i.e. amphipod traps) and actual bullhead diet. Because C. gobio do not have pharyngeal teeth, their diet can be analysed by collecting bolus, using stomach-flushing methods (Lagrue and Bollache, Reference Lagrue and Bollache2006; Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007). Fish stomach contents were collected every 5 days for a total of 9 sampling dates using a simplified stomach flushing technique from Gaudin et al. (Reference Gaudin, Martin and Caillere1981; see also Lagrue and Bollache, Reference Lagrue and Bollache2006). Amphipods recovered from fish stomachs were preserved in 70% ethanol before analyses. In the laboratory, intact amphipods, i.e. defined as the carcass possessing a head still connected to the pereon (MacNeil et al. Reference MacNeil, Elwood and Dick2001; Lagrue and Bollache, Reference Lagrue and Bollache2006), were sexed, measured and dissected as described above. Other amphipod remains were disregarded since they could not be measured and/or positively categorized as infected or not. As for amphipods collected in traps, prevalences of both acanthocephalan parasites were determined.
Echinorhynchus truttae and P. minutus prevalences (proportions of infected amphipods) in fish diet were compared with parasite prevalences in amphipod traps from the bullhead treatment using Fisher's exact tests. Possible differences in amphipod size between treatments were tested using a non-parametric test (Mann-Whitney U test). We performed all tests with a 5% type I error risk, using STATISTICA Software 6.0 (StatSoft Inc., France).
RESULTS
Parasite prevalence and distribution
In total, 8919 amphipods were captured in leaf traps, 2430 males, 3089 females and 3400 juveniles. Amphipods qualified as ‘juveniles’ were immature individuals that could not be sexed and measured less than 5 mm in body length. Juvenile amphipods were never infected by acanthocephalan parasites or predated upon by C. gobio and were therefore discarded from the following analyses. Of the 5519 adult amphipods collected, 1787 were captured in the control treatment, 1757 in the crayfish and 1975 in the fish treatment. At the trap/cage level, treatment had no effect on amphipod densities (119·1 ± 16·5, 117·5 ± 19·0 and 131·7 ± 17·2 mean CPUE±s.e. for control, crayfish and fish treatments, respectively; ANOVA, F2,42 = 0·452, P = 0·640). However, G. pulex CPUE (i.e. mean number of amphipods per trap) was significantly different between sampling blocks (156·6 ± 25·3, 114·8 ± 27·2, 87·4 ± 17·8, 107·4 ± 14·0 and 149·9 ± 21·1 for blocks 1 to 5, respectively; ANOVA, F4,40 = 2·921, P = 0·034). CPUE also decreased between sampling dates (178·2 ± 17·5, 106·1 ± 13·5 and 78·3 ± 8·3 for 4, 14 and 25 of April, respectively; ANOVA, F2,42 = 11·04, P = 0·0002).
Generally, parasite prevalences did not vary between sampling dates (Table 1). More importantly, treatment had no overall effect on either E. truttae or P. minutus (Table 1 and Fig. 1). Echinorhynchus truttae prevalence was very similar between control, crayfish and bullhead treatments (2·07, 2·16 and 2·28%, respectively; Fig. 1) while P. minutus prevalence was slightly lower in the bullhead treatment (0·66%) compared to the two others (1·18 and 1·08% in control and crayfish treatments, respectively; Fig. 1). Sampling blocks did not significantly influence P. minutus infection (Table 1B); prevalence varying between 0·64 and 1·14% among blocks. Contrastingly, E. truttae infection was variable across sampling blocks. Observed prevalences were highly dependent upon blocks (Table 1A). This pattern was mainly due to variations in overall amphipod densities since E. truttae infected G. pulex numbers did not vary between sampling blocks (2·22 ± 0·76, 2·56 ± 0·65, 3·56 ± 0·94, 2·33 ± 0·60 and 2·67 ± 0·73 CPUE±s.e. for blocks 1–5, respectively; Kruskal-Wallis ANOVA, H4,45 = 1·51, P = 0·826).
Fig. 1. Prevalences of Echinorhynchus truttae (□) and Polymorphus minutus (■) in Gammarus pulex individuals captured in amphipod traps in the different treatments (control, crayfish and fish, respectively) and recovered in Cottus gobio stomach samples (fish diet). Sample size (i.e. number of G. pulex amphipods) for each treatment is indicated above bars.
Table 1. Generalized Linear Models testing for the effects of treatment (control, crayfish and fish), sampling date (4, 14 and 25 April 2011), sampling block (1–5), and amphipod host sex and size (total body length) on the prevalence of (A) Echinorhynchus truttae and (B) Polymorphus minutus in Gammarus pulex captured in amphipod traps
Individual amphipod size significantly influenced the probability of infection by both parasite species (Table 1 and Fig. 2). Echinorhynchus truttae-infected individuals were larger (mean body length±s.e. = 8·75 ± 0·14 mm) than uninfected amphipods (8·29 ± 0·02 mm, Fig. 2). The trend was opposite for P. minutus with infected individuals being smaller (7·72 ± 0·16 mm) than their uninfected conspecifics (8·30 ± 0·02 mm, Fig. 2). However, in both species, the highest mean parasite prevalence occurred in amphipod hosts of intermediate size (Fig. 2), a pattern classically used as an estimate of parasite-induced mortality in intermediate hosts (Thomas et al. Reference Thomas, Renaud, Rousset, Cézilly and de Meeüs1995; Rousset et al. Reference Rousset, Thomas, de Meeüs and Renaud1996). Host sex also significantly influenced E. truttae infection but had no effect on P. minutus (Table 1 and Fig. 2). Male amphipods were 2·2 times more likely to be infected by E. truttae than females (3·13 and 1·42% prevalence, respectively) and seemed to carry heavier parasite loads (Fig. 2). No significant interaction was detected between the different factors included in the models.
Fig. 2. Mean prevalences (±s.e.) of Echinorhynchus truttae (top graphs) and Polymorphus minutus (bottom) in relation to host size (amphipod length class) in male (left graphs) and female (right) Gammarus pulex captured in amphipod traps. Sample size (i.e. number of G. pulex amphipods) for each host size class is indicated above each data-point.
Finally, we found a significant relationship between G. pulex densities (CPUE) and E. truttae prevalence (r = 0·933, n = 5, P = 0·021); prevalence being higher in sampling blocks with lower densities of amphipod hosts (Fig. 3A). Contrastingly, P. minutus prevalence seemed to increase with host densities although the trend was not significant (r = 0·589, n = 5, P = 0·296; Fig. 3B).
Fig. 3. Relationship between G. pulex amphipod density (CPUE±s.e.) and parasite prevalence for (A) Echinorhynchus truttae and (B) Polymorphus minutus across sampling blocks (1–5). Lines of best fit and coefficients of determination are shown on each figure. Block number is indicated next to each data-point.
Fish diet and parasite transmission
A total of 39 intact amphipods were recovered in C. gobio stomach samples (6, 8, 6, 8 and 11 from blocks 1–5 respectively); 12 were infected by E. truttae (2, 2, 2, 2 and 4 from blocks 1–5 respectively) and zero contained P. minutus. Because of the small sample size, Amphipods from all individual fish and all sampling blocks were grouped before analyses.
Host size was not significantly different between amphipods recovered in fish stomachs (mean body length±s.e. = 7·91 ± 0·30 mm) and individuals captured in the fish treatment traps (8·28 ± 0·04 mm; Mann-Whitney U test, Z = −1·367, P = 0·172). Again, E. truttae-infected amphipods tended to be larger than their uninfected conspecifics, both in C. gobio diet (8·38 ± 0·47 and 7·70 ± 0·39 mm, respectively; Mann-Whitney U test, Z = −1·297, P = 0·195) and fish treatment (8·27 ± 0·04 and 8·74 ± 0·27 mm, respectively; Mann-Whitney U test, Z = −2·179, P = 0·029); although the difference was not significant in C. gobio diet. Prevalence of E. truttae was significantly higher in C. gobio diet (30·77%) than in the fish treatment (2·28%; Fisher's exact test, χ 2 = 112·88, P < 0·0001; Fig. 1). Gammarus pulex hosts recovered in fish stomach contents were 13·5 times more likely to be infected than individuals captured in amphipod traps from the fish treatment (Fig. 1). Finally, although P. minutus-infected amphipod were absent from C. gobio stomach contents, no significant difference in P. minutus prevalence was detected between amphipod hosts in fish diet and fish treatment traps (0 and 0·66%, respectively; Fisher's exact test, χ 2 = 0·26, P = 0·611; Fig. 1).
DISCUSSION
Uninfected G. pulex densities were variable across sampling blocks, but unaffected by predator presence or species (fish or crayfish) within blocks. Additionally, E. truttae- and P. minutus-infected amphipods showed homogeneous distributions (i.e. similar densities) across sampling blocks. These contrasting patterns induce variations in E. truttae prevalences, apparently negatively correlated with amphipod densities. Furthermore, E. truttae infected G. pulex micro-distribution (i.e. within blocks) was not linked to predator presence and/or type (i.e. suitable host or dead-end predator). Overall, amphipod distribution differed between infected and uninfected individuals but was not affected by predators. Although G. pulex is known to be able to use chemical cues from predators to adjust habitat use and efficiently avoid predator proximity (Dahl et al. Reference Dahl, Anders Nilsson and Pettersson1998), this capacity did not influence G. pulex distribution in our study. Heterogeneity in amphipod density was detected across sampling blocks but within a block G. pulex micro-distribution did not depend on predator presence, in contrast to suggestions from MacNeil et al. (Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003). Overall, in our study, uninfected G. pulex did not seem to avoid habitat patches occupied by predators. Micro-distribution of E. truttae and P. minutus infected individuals was not influenced by suitable definitive or non-host predator presence although these parasites could still alter host spatial distribution in other ways that may be beneficial (MacNeil et al. Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003; Médoc and Beisel, Reference Médoc and Beisel2009, Reference Médoc and Beisel2011).
Nevertheless, evidence exists that crustacean amphipods can use chemical cues from predators and develop adaptive anti-predator responses (Mathis and Hoback, Reference Mathis and Hoback1997; Wudkevich et al. Reference Wudkevich, Wisenden, Chivers and Smith1997; Åbjörnsson et al. Reference Åbjörnsson, Dahl, Nyström and Brönmark2000). In particular, amphipods tend to reduce activity and increase refuge use upon chemical detection of predators (Kaldonski et al. Reference Kaldonski, Perrot-Minnot and Cézilly2007; Perrot-Minnot et al. Reference Perrot-Minnot, Kaldonski and Cézilly2007; Benesh et al. Reference Benesh, Kitchen, Pulkkinen, Hakala and Valtonen2008a). While our results suggest that such anti-predator behaviours do not affect amphipod distribution, they still constitute ideal targets for manipulation by acanthocephalans to increase transmission to definitive hosts. For example, G. pulex infected with the fish acanthocephalans, Pomphorhynchus laevis or P. tereticollis, showed increased activity levels, decreased use of shelter and/or reversed reactions to fish chemical cues, being attracted to it while uninfected individuals were clearly repulsed (Baldauf et al. Reference Baldauf, Thünken, Frommen, Bakker, Heuprl and Kullmann2007; Perrot-Minnot et al. Reference Perrot-Minnot, Kaldonski and Cézilly2007). Similarly, E. truttae-infected G. pulex are more active and less photophobic than uninfected individuals (MacNeil et al. Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003). Movement being a major stimulus eliciting predator attack on prey, E. truttae-infected amphipods should be more likely to be seen and consumed by predators than their hidden and motionless uninfected conspecifics (Poulin, Reference Poulin1995; MacNeil et al. Reference MacNeil, Elwood and Dick1999, Reference MacNeil, Fielding, Hume, Dick, Elwood, Hatcher and Dunn2003). This is clearly suggested by our results; E. truttae-infected G. pulex were 13·5 times more likely to be consumed by C. gobio, the parasite's definitive host, than uninfected individuals, even though E. truttae did not influence amphipod distribution. Parasitized host vulnerability to fish predation could thus be increased regardless of infected host micro-distribution. Our results suggest that, even though the parasite does not seem to alter amphipod hosts micro-distribution into areas with fish definitive hosts, E. truttae-induced behavioural manipulation may still operate at very close range, in the immediate vicinity of definitive host predators. However, E. truttae-induced changes in G. pulex anti-predatory responses remain to be tested and characterized under controlled, laboratory conditions if we are to understand fully how the parasite is able to increase its transmission at such close range.
Upon detection of a specific predator, prey should adjust their behaviour so that they are less likely to encounter, be detected and captured by that particular predator (Lima, Reference Lima1998). Parasites could thus increase transmission probabilities accordingly through manipulation of specific intermediate host anti-predator behaviours, thus combining predation enhancement by suitable hosts and predation suppression by non-host predators (Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007; Médoc et al. Reference Médoc, Rigaud, Bollache and Beisel2009; Médoc and Beisel, Reference Médoc and Beisel2011). Such multidimensionality in host manipulation means that a single parasite may alter more than one behavioural trait in its intermediate host in order to target a unique definitive host species and avoid dead-end predators (Benesh et al. Reference Benesh, Valtonen and Seppälä2008b; Cézilly and Perrot-Minnot, Reference Cézilly and Perrot-Minnot2010; Thomas et al. Reference Thomas, Poulin and Brodeur2010). Recent empirical evidence suggests that acanthocephalans may be able to reduce intermediate host predation risk by non-host predators through parasite-induced increase in specific anti-predator responses (Médoc and Beisel, Reference Médoc and Beisel2011), although several counter-examples exist (Holmes and Bethel, Reference Holmes and Bethel1977; Marriott et al. Reference Marriott, Collins, Paris, Gudgin, Barnard, McGregor, Gilbert, Hartley and Behnke1989; Kaldonski et al. Reference Kaldonski, Perrot-Minnot, Motreuil and Cézilly2008; Seppälä et al. Reference Seppälä, Valtonen and Benesh2008). For example, the amphipod Gammarus roeseli reacts to predator chemical cues by increasing refuge use, regardless of infection status. However, amphipods infected with P. minutus, a bird parasite, showed increased shelter use and tend to exploit refuges close to the surface, away from non-host benthic predators (Médoc et al. Reference Médoc, Bollache and Beisel2006, Reference Médoc, Rigaud, Bollache and Beisel2009). Infected amphipods were consequently less preyed upon by fish predators (Médoc et al. Reference Médoc, Rigaud, Bollache and Beisel2009). Whether this is the case with the crayfish predator could not be determined from our study since crayfish stomach contents cannot be sampled on live animals. However, results show that, while E. truttae infected G. pulex were significantly over-predated by C. gobio, P. minutus-infected individuals were never found in fish diet. Results suggest that different acanthocephalans render common intermediate host species specifically vulnerable to particular predators, thus increasing parasite transmission to appropriate final hosts only. While P. minutus did not seem to suffer from increased predation by dead-end fish predators, we cannot conclude from our results whether infected amphipods are capable of actively avoiding predation by fish, thus increasing P. minutus survival when faced with non-host predators.
Overall, our study showed that, although parasites did not influence amphipod host micro-distribution, intermediate host manipulation by acanthocephalans may still increase parasite transmission to the definitive hosts in the field by operating at very close range. Results from fish stomach samples also point to potential parasite species-specific behavioural alterations in G. pulex. Echinorhynchus truttae induced host behaviour alterations increased parasite transmission probabilities through over-predation of infected hosts. Contrastingly, P. minutus is unlikely to suffer from increased predation by non-host fish predators. However, whether this manipulative parasite is capable of enhancing G. pulex anti-predator responses, when facing dead-end predators, would require further investigations.
ACKNOWLEDGEMENTS
We thank three anonymous reviewers for very constructive comments on a previous version of the manuscript.
