INTRODUCTION
While the majority of parasites are known to exploit multiple host species, either sequentially or because they have a range of suitable hosts for the same stage of their cycle (Ruiz-González et al. Reference Ruiz-González, Bryden, Moret, Reber-Funk, Schmid-Hempel and Brown2012), host-parasite interactions are usually studied in simplified one-to-one relations, disconnected from the real-life complex systems (Rigaud et al. Reference Rigaud, Perrot-Minnot and Brown2010). Multi-host parasites may use host species differing in abundance, exposure and susceptibility, and thus unlikely to contribute equally to parasite transmission and fitness. The ‘key hosts’ are those contributing significantly more to the completion of the parasite life cycle (Streicker et al. Reference Streicker, Fenton and Pedersen2013). Three non-exclusive processes serve to identify a host as a key species, contributing disproportionately to parasite transmission: high host abundance, high exposure and/or susceptibility to infection, and/or large number of infective stages produced per infected individual (Streicker et al. Reference Streicker, Fenton and Pedersen2013).
Parasites with complex life cycles are, by definition, multi-host parasites because they require at least two successive host species to achieve their development. However, they may also use several different host species at any stage of their cycle. Such parasites may show weak specificity when infecting the intermediate host, or sometimes even the definitive host, although there is great interspecific variation in these traits (Combes, Reference Combes2001). Numerous parasites with a complex life cycle have evolved the ability to modify several aspects of the phenotype of their intermediate hosts, concomitantly increasing the probability of transmission to their definitive hosts (reviewed in Poulin, Reference Poulin, Mitani, Brockmann, Roper, Naguib and Wynne-Edwards2010). Many trophically transmitted parasites can even modify certain behaviours of their intermediate hosts (Thomas et al. Reference Thomas, Adamo and Moore2005; Perrot-Minnot et al. Reference Perrot-Minnot, Sanchez-Thirion and Cézilly2014). Modification of a number of anti-predatory behaviours is directly linked to the modulation of predation rates in intermediate hosts, either increasing for infected vs non-infected hosts (Kaldonski et al. Reference Kaldonski, Perrot-Minnot and Cézilly2007; Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007), or decreasing when the parasites are not yet infective for the definitive host (Dianne et al. Reference Dianne, Perrot-Minnot, Bauer, Gaillard, Léger and Rigaud2011; Weinreich et al. Reference Weinreich, Benesh and Milinski2013). These behavioural changes have been referred to as ‘host manipulation’ because parasites alter the phenotype of their hosts in ways that enhance their own fitness at the expense of that of infected hosts (Thomas et al. Reference Thomas, Adamo and Moore2005; Cézilly et al. Reference Cézilly, Thomas, Médoc and Perrot-Minnot2010). For these parasites, the sensitivity of the host to manipulation should be included to determine key host species, because of its implication in parasite transmission.
Acanthocephala are trophically transmitted parasites for which the ability to modify host phenotype is ubiquitous, possibly having evolved in the common ancestor of the group (Moore, Reference Moore1984). They all use at least two hosts to complete their cycle, whether for intermediate, definitive or paratenic hosts, with different degrees of fitness depending on the hosts and/or spatial distribution of these hosts (see Kennedy, Reference Kennedy2006 for an overview). Pomphorhynchus laevis have been extensively studied in the contexts of host manipulation and ecology (Kennedy, Reference Kennedy2006). They infect several freshwater gammarid amphipod species as intermediate hosts, and several freshwater fish species as definitive or paratenic hosts (Kennedy, Reference Kennedy2006; Médoc et al. Reference Médoc, Rigaud, Motreuil, Perrot-Minnot and Bollache2011). In central and eastern France, the cryptic Gammarus pulex and Gammarus fossarum species (Lagrue et al. Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014) are resident intermediate host species, while Gammarus roeseli is a relatively recent colonizer from Southern Central Europe (Jazdzewski, Reference Jazdzewski1980). These gammarids are often found in sympatry (Chovet and Lécureuil, Reference Chovet and Lécureuil1994) and infected by P. laevis in these sympatric sites (e.g. Bauer et al. Reference Bauer, Trouvé, Grégoire, Bollache and Cézilly2000; Rigaud and Moret, Reference Rigaud and Moret2003; Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007). Prevalence and infection intensity are usually higher in G. roeseli than in G. pulex (Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007, Lagrue, unpublished data), despite the fact that the latter is generally more abundant than the former when present in sympatry (e.g. Lagrue et al. Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007). It would therefore seem logical for P. laevis to rely more on G. roeseli than on G. pulex for its transmission. However, several elements indicate that exactly the opposite situation could be the rule. Crude prevalence is not an accurate measure to quantify the abundance of a manipulative parasite, since observed prevalence diminishes as infected intermediate hosts are preferentially preyed upon by the next host(s), rather than uninfected hosts (Lafferty, Reference Lafferty1992; Rousset et al. Reference Rousset, Thomas, De Meeûs and Renaud1996). Lagrue et al. (Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007) showed that the prevalence of P. laevis in G. pulex was low in the river benthos but high in the definitive host's stomach, whereas prevalence in G. roeseli was higher in the field and lower in the stomach of the definitive host. In addition, by analysing the distribution of parasite intensity, they showed that parasites accumulate in older G. roeseli, but not in older G. pulex, confirming a higher death rate of infected G. pulex compared with infected G. roeseli. This result is consistent with the fact that infected G. roeseli is known to be less strongly manipulated than G. pulex by P. laevis (Bauer et al. Reference Bauer, Trouvé, Grégoire, Bollache and Cézilly2000). Furthermore, uninfected G. roeseli has been found to be less sensitive to predation by trout (Bollache et al. Reference Bollache, Kaldonski, Troussard, Lagrue and Rigaud2006) or bullhead (Kaldonski et al. Reference Kaldonski, Lagrue, Motreuil, Rigaud and Bollache2008) than uninfected G. pulex, because of more efficient anti-predatory defences. The combination of all these factors provides reasonable evidence of a predation differential between infected animals of each species, and so G. roeseli can reasonably be considered a lower quality host for P. laevis transmission.
However, the relative susceptibility of the two amphipod species to infection by P. laevis remains undetermined. Yet this information is crucial to assess the relative importance of the two concurrent hosts in the P. laevis life cycle. If G. pulex is more susceptible to infection than G. roeseli, then both susceptibility and behavioural manipulation would act in synergy, making this host a true key host for transmission. If, conversely, G. roeseli is more susceptible than G. pulex, then P. laevis transmission would be ‘diluted’ by the presence of this host, because of its inefficiency in transmitting the parasite, and could potentially impact the epidemiology of the infection (see Hall et al. Reference Hall, Becker, Simonis and Duffy2009; Johnson et al. Reference Johnson, Lund, Hartson and Yoshino2009, for examples). We conducted a laboratory infection experiment by submitting both species to the same dose of P. laevis eggs to measure the susceptibility of these sympatric gammarid species to P. laevis. To assess the impact of predation, we compared prevalence in two contrasted amphipod collections from the field: animals directly collected from rivers (i.e. previously exposed to natural predation), and animals collected from the same rivers, but then maintained for several weeks in the laboratory (i.e. in the absence of any fish predation pressure).
METHODS
Amphipod collection and prevalence in the field
Since field prevalence may be variable between populations, two rivers were chosen, where G. fossarum and G. roeseli live in sympatry and are naturally infected by P. laevis. Amphipods from the Albane River, in Trochères (47°20′34″N, 5°18′21·8″E), and the Meuzin River, near Villy-le-Moutier (47°2′7·71″N, 4°59′53·87″E), were sampled between September and October 2013.
Amphipods (G. roeseli and G. fossarum) were captured using kick nets. All potential habitats present at each site were sampled, and the collected animals were randomly divided into three groups, each maintained in a container with aerated water from the river.
The first group was used to estimate the ‘field/direct’ prevalence. Animals from this group were kept in well aerated aquaria at 15 °C and all checked for parasite presence within 2 days after capture. Infected individuals were dissected to confirm parasite species. Larval parasites can be detected through the host cuticle, either at the late acanthella stage of their development (translucent light orange, shapeless larval stage) or at cystacanth stage (bright yellow-orange, spherical larval stage). Earlier acanthella stages (where parasites are small and translucent) can only be detected after dissection. Preliminary investigation showed that acanthella detection could only be certified after 40 days (without microscope and staining), so that all prevalence reported in the following experiments is prevalence for P. laevis of more than 40 days old (Labaude et al. submitted).
Gammarids from the second group were kept individually in the laboratory, in cups of c.a. 50 mL at 15 °C for 96 days. All gammarids where infection was detectable by eye were removed from this group so that, at the beginning of this survey, the remaining animals were classified as ‘uninfected’. However, as previously stated, younger acanthella stages are too small to be detected through host cuticule, so some of these isolated gammarids may have already been infected in the field in the days preceding their capture. It is the prevalence of these undetected infections that was recorded during this survey. Animals dying during this period were dissected the day after their death, and all living animals were checked and dissected 96 days post isolation, a delay long enough to ensure that all parasites could be detected. This survey therefore allowed prevalence to be estimated in gammarids not exposed to predation during parasite development (hereafter called ‘field/protected’ prevalence). All infected G. fossarum were kept in ethanol for genetic analysis (see above).
A third group of gammarids was used for experimental infections (see below).
Experimental infection
Before being isolated for the experiment, all gammarids were inspected under a dissecting microscope to remove naturally infected animals. The remaining gammarids were kept in quarantine for 30 days, to distinguish any further natural infection (by parasites too young to be detected) from experimental infection. Some additional G. pulex were also collected in a small tributary of the Suzon River at Val-Suzon (47°4′12·6″N; 4°52′58·2″E). Given that the G. pulex from Val-Suzon are particularly sensitive to experimental infection by P. laevis (Franceschi et al. Reference Franceschi, Cornet, Bollache, Dechaume-Moncharmont, Bauer, Motreuil and Rigaud2010), they were used to confirm the success and timing of experimental infection.
Gravid P. laevis females were collected from the intestines of chubs (Leuciscus cephalus), from naturally infected fish caught in September 2013 in the Vouge River (Burgundy, Eastern France: 47°9′34·36″N; 5°9′2·50″E). A foreign parasite population was chosen to avoid potential local adaptation in our two gammarid populations (Franceschi et al. Reference Franceschi, Cornet, Bollache, Dechaume-Moncharmont, Bauer, Motreuil and Rigaud2010), so that it was possible to estimate gammarid sensitivity to parasite strains with which they had not evolved. Molecular identification of parasites and exposure of gammarids to parasite eggs followed the procedure described in Franceschi et al. (Reference Franceschi, Bauer, Bollache and Rigaud2008). Gammarus, in cups filled with c.a. 50 mL of aerated water, were allowed to feed for 48 h on a 1 cm2 piece of elm leaf, on which a suspension of 100 mature eggs per gammarid had been deposited (see detailed procedure in Franceschi et al. Reference Franceschi, Bauer, Bollache and Rigaud2008). Food was then removed, and gammarids were maintained at 15 °C for 3 months. The field/protected group described above was used as control. Individuals from this group were treated and maintained under the same conditions as exposed gammarids but were unexposed to parasite eggs. A total of 615 G. fossarum (162 males and 109 females from Albane, 214 males and 130 females from Meuzin) and 440 G. roeseli (157 males and 102 females from Albane, 121 males and 60 females from Meuzin) were exposed to parasite eggs, as were the G. pulex (155 males from Val-Suzon). 308 G. fossarum (104 males and 61 females from Albane, 89 males and 54 females from Meuzin) and 324 G. roeseli (102 males and 67 females from Albane, 104 males and 51 females from Meuzin) were used as control individuals. All infected G. fossarum, along with 100 individuals from the control group, were kept in ethanol for genetic investigation (see below).
The water of each dish was completely renewed every 2 weeks with aerated water from the river, and water levels were restored to original levels twice a week. The amphipods were fed ad libitum with elm leaves, and their diet was enriched with a chironomid larva twice a month. A daily mortality survey was carried out, and animals were dissected the day after their death to detect young acanthella stages. From the sixth week post-exposure, living gammarids were inspected every week under a dissecting microscope to detect the presence of parasites. Infected animals were examined every 2 days after detection to estimate the date when the cystacanth stage was reached. Gammarids from Val-Suzon (where P. laevis is absent) were a control group for the timing and success of experimental infection. Previous studies revealed that P. laevis reaches cystacanth stage in about 80–120 days in laboratory conditions (Franceschi et al. Reference Franceschi, Bauer, Bollache and Rigaud2008, Reference Franceschi, Cornet, Bollache, Dechaume-Moncharmont, Bauer, Motreuil and Rigaud2010). In gammarids from the Meuzin and Albane rivers, even after a quarantine of 30 days before exposure, parasites from the wild can develop. Therefore, if P. laevis were detected before the first signs of infection in animals from Val-Suzon, individuals were removed from the analysis to avoid any potential confounding effect.
Gammarid genotyping
Because of the recently discovered cryptic genetic diversity within the G. fossarum-pulex species complexes (e.g. Lagrue et al. Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014), there is a need to examine the patterns of infection in the light of this diversity (see Westram et al. Reference Westram, Baumgartner, Keller and Jokela2011a , Reference Westram, Jokela, Baumgartner and Keller b ). Such a study is not necessary for G. roeseli because no cryptic diversity has been detected in Western and Central Europe (Moret et al. Reference Moret, Bollache, Wattier and Rigaud2007). Genetic diversity was assessed in these two rivers using the amplification of part of the mtDNA cytochrome c oxidase subunit 1 (CO1) by polymerase chain reactions (PCR) and a subsequent restriction fragment length polymorphism (RFLP) procedure (Lagrue et al. Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014). Only G. fossarum belonging to one group were known to occur at the Meuzin site (GfI, see Lagrue et al. Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014), while genetic diversity for the Albane River had not previously been estimated. All infected G. fossarum and G. pulex from each river were preserved in pure ethanol after death, for subsequent DNA extraction. In addition, 100 uninfected animals randomly sampled from each site were also preserved. Gammarid DNA was extracted from two pereopods (‘walking legs’ in amphipods), following the standard chelex method (Lagrue et al. Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014). The DNA was then amplified for CO1 using universal primers (LCO1490 and HCO2198; Folmer et al. Reference Folmer, Black, Hoeh, Lutz and Vrijenhoek1994). The PCR were performed using Qiagen Multiplex DNA polymerase kits (Qiagen Inc, Düsseldorf, Germany), as in Lagrue et al. (Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014). The PCR-amplified DNA products were then digested overnight using the appropriate reaction buffer and restriction endonuclease(s), following manufacturer's instructions (New England Biolabs, Ipswich, Massachusetts, USA). The resulting fragments were separated by gel electrophoresis in a 1·5% agarose gel. Restriction enzyme profiles were used to assign each individual amphipod to its respective genetic group (see Lagrue et al. (Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014) for the detailed procedure and the specific digestion enzymes for each gammarid genetic group).
Statistical analyses
All statistical analyses were performed using R software or JMP software (version 10·0·0).
For natural infections, a binomial logistic regression was performed to analyse prevalence, with the following potential explanatory factors: site (Albane River vs Meuzin River), Gammarus species (G. roeseli vs G. fossarum), Gammarus sex (males vs females), experiment (field/direct: natural infection from the field sample vs field/protected: natural infection after maintenance in the laboratory), and their second-order interactions.
For experimental infections, a binomial logistic regression was performed to analyse prevalence, with site, species and sex, and their second-order interactions, as potential explanatory factors.
All possible models were compared using the Akaike Information Criterion (AICc). The models presented are those minimizing the AICc.
RESULTS
Genetic diversity among G. fossarum-like gammarids
For the gammarids from the Albane River, PCR-RFLP revealed 87% of G. fossarum and 13% of G. pulex in the 50 randomly sampled, uninfected animals, with 82% of G. fossarum and 18% G. pulex in the 68 infected animals. The species ratios in infected and uninfected groups were not significantly different (χ 2 = 0·2438, P = 0·6215). As we detected no difference in sensitivity to infection between G. pulex and G. fossarum, and since the majority of the gammarids, even at the Albane site, are G. fossarum, this term is used to encompass all G. fossarum-like gammarids.
Natural infection: direct field prevalence vs field prevalence protected from predation
Prevalence of P. laevis was higher in G. roeseli than in G. fossarum in direct field prevalence, at both sites, whereas reverse relative prevalence was observed when measured after keeping putative uninfected animals in the laboratory, where they were preserved from predation (Table 1, Fig. 1).
Fig. 1. Prevalence levels for Gammarus fossarum (Gf) and G. roeseli (Gr) in the two populations, for all experiments (field/direct: prevalence in natura; field/protected: prevalence in gammarids kept in the laboratory, i.e. protected from predation; experiment: experimental infection). Number in bars are sample size.
Table 1. Logistic regression testing for the effects of site (river), Gammarus species and experiment (direct field prevalence or protected field prevalence) on the field prevalence of P. laevis
The model initially included sex of gammarids, and other interactions. After removing these non-significant factors, the model presented now minimizes the Akaike Information Criterion (AICc).
Global model: LR χ 2 = 15·4448, 5 d.f., P = 0·0086; n = 1787.
Experimental infection
The first observations of acanthellae through the host cuticle occurred 60 days post-exposure for the control Val-Suzon gammarids, as was the case for gammarids of both species from the Albane and Meuzin rivers. The cystacanth stage was achieved 82 ± 10 days post-exposure of the control Val-Suzon group, after 80 ± 6 days for G. fossarum, and after 83 ± 3 days for G. roeseli.
We found a strong effect of river origin on infection (Table 2, Fig. 1), with gammarids from the Albane River being three times more sensitive to infection. The difference in prevalence between species, with G. fossarum being approximately twice as infected as G. roeseli, was nevertheless not strong enough to be fully supported statistically (Table 2, Fig. 1).
Table 2. Logistic regression testing for the effects of site (river) and Gammarus species on the prevalence of P. laevis after experimental infection by parasites from the Ouche River
The model initially included sex of gammarids and interactions. After removing these non-significant factors, the model presented now minimizes the Akaike Information Criterion (AICc).
Global model: LR χ 2 = 19·9606, 2 d.f., P < 0·0001; n = 807.
DISCUSSION
Our data initially showed that the crude P. laevis prevalence is higher in G. roeseli than in G. fossarum, confirming results of Lagrue et al. (Reference Lagrue, Kaldonski, Perrot-Minnot, Motreuil and Bollache2007) for another site. In the ‘field/protected’ experiment, the prevalence was reversed, and was higher in G. fossarum for the two populations investigated. In addition, prevalence in G. fossarum was approximately twice that in G. roeseli in both populations after experimental infection by a non-coevolved parasite population, even though this result was not fully supported statistically (probably due to the stronger population effect). Prevalence observed in the field is therefore not a reliable measure of the actual parasite burden for this manipulative trophically transmitted parasite. Differences in the duration of parasite development could possibly have explained the differences in prevalence observed between the two Gammarus species. However, parasite growth was synchronous for all hosts during the laboratory infection experiment.
As the two hosts have similar lifespans, parasites developing in G. roeseli have a lower probability of completing their life cycle, both due to reduced natural predation by fish compared with G. pulex (Bollache et al. Reference Bollache, Kaldonski, Troussard, Lagrue and Rigaud2006; Kaldonski et al. Reference Kaldonski, Lagrue, Motreuil, Rigaud and Bollache2008) and lower manipulation levels for infected individuals (Bauer et al. Reference Bauer, Trouvé, Grégoire, Bollache and Cézilly2000). Therefore, G. roeseli seems to ‘dilute’ P. laevis transmission when this host is sympatric with G. fossarum. However, as shown here, G. roeseli is not more susceptible than G. fossarum to infection by P. laevis, so the dilution effect is not as strong as previously thought when natural prevalence alone was considered. Lower infection success in G. roeseli counterbalances the low predation rate, limiting the ‘sink effect’ for the parasite. As G. fossarum is first more susceptible to infection and then more predated, our data confirm this species as a key host for P. laevis.
Our results also have implications in explaining the role of parasites in the success of biological invasions. Gammarus roeseli is a species that colonized Western Europe during the 20th century (Chovet and Lécureuil, Reference Chovet and Lécureuil1994). Parasitism may play a role in the coexistence of native and introduced (or invasive) host species. Some studies support the ‘enemy release’ hypothesis, in which invaders are no longer exposed to their original parasites, but also less susceptible to infection by native parasites, providing invasive hosts with a competitive advantage (Dunn and Dick, Reference Dunn and Dick1998; Kopp and Jokela, Reference Kopp and Jokela2007). In contrast, other studies show a decrease in prevalence in native species by the dilution effect, both experimentally (Kopp and Jokela, Reference Kopp and Jokela2007) and in natura (Telfer et al. Reference Telfer, Bown, Sekules, Begon, Hayden and Birtles2005). The invader acts in that case as a dead-end sink for the parasite. G. roeseli being less susceptible to both infection (this study) and to behavioural changes induced by P. laevis (Bauer et al. Reference Bauer, Trouvé, Grégoire, Bollache and Cézilly2000; Moret et al. Reference Moret, Bollache, Wattier and Rigaud2007), our results are in line with the ennemy realese hypothesis.
Our results completely strengthen the hypothesis that sympatric G. roeseli and G. fossarum are not hosts of the same quality for acanthocephalan parasites. Should this assumption be extended to all gammarid hosts of freshwater acanthocephalans? Because of the high level of cryptic speciation in the G. pulex/fossarum group (e.g. Westram et al. Reference Westram, Jokela, Baumgartner and Keller2011b ; Lagrue et al. Reference Lagrue, Wattier, Galipaud, Gauthey, Rullmann, Dubreuil, Rigaud and Bollache2014), the situation will probably be quite complex to study. Westram et al. (Reference Westram, Baumgartner, Keller and Jokela2011a ), coupling natural prevalence estimations and field infection experiments, also showed differences in susceptibility between Gammarus species to infection by the acanthocephalan Pomphorhynchus tereticollis, with G. pulex being less infected than G. fossarum. Differences within the G. fossarum group, while less marked, were also detected. However, in Switzerland, where the study was carried out, different species (and/or cryptic species) are rarely found in sympatry, each stream or river harbouring a single gammarid species, so there is confusion between host species and the sites where the host-parasite couple is living, with the potential for local adaptation confounding the results of host specificity (Franceschi et al. Reference Franceschi, Cornet, Bollache, Dechaume-Moncharmont, Bauer, Motreuil and Rigaud2010). Apart from our case-study of the G. roeseli/G. fossarum system, no clear data are available yet on infectivity and behavioural changes induced by the same local parasite strains on two sympatric species. In the present study, we found no significant difference in prevalence between sympatric G. pulex and G. fossarum from the Albane River. However, this result should be replicated in other rivers, with more individuals and more species tested. Behavioural modifications should also be measured to confirm this apparent lack of specificity.
ACKNOWLEDGEMENTS
We thank Aude Balourdet, Sophie Labaude and Sébastien Motreuil for their valuable field and experimental assistance, and Carmela Chateau for English corrections. We thank the anonymous referee for valuable suggestions and comments.
FINANCIAL SUPPORT
This study was supported by a grant from the Agence Nationale de la Recherche (grant # ANR-13-BSV7-0004-01).
