INTRODUCTION
Climate change is predicted to alter host–parasite relationships during this century, especially where warming combines with other anthropogenic disturbances (Rohr et al. Reference Rohr, Dobson, Johnson, Kilpatrick, Paull, Raffel, Ruiz-Moreno and Thomas2011; Paull et al. Reference Paull, LaFonte and Johnson2012; Lõhmus and Björklund, Reference Lõhmus and Björklund2015). In northern latitudes, where climatic factors are important regulators of host–parasite population dynamics and parasite occurrence, and transmission is regulated by seasonal temperature changes, shortened winter periods could alter host–parasite relationships via alterations in host susceptibility and parasite virulence (Hakalahti et al. Reference Hakalahti, Karvonen and Valtonen2006; Lõhmus and Björklund, Reference Lõhmus and Björklund2015). Should growth rates of the hosts and parasites be altered by temperature changes then pathology and transmission rates could also be affected (Raffel et al. Reference Raffel, Rohr, Kiesecker and Hudson2006; Lafferty, Reference Lafferty2009). Consequently, predictions tend to be for warming to increase the prevalence of parasites at higher latitudes (e.g. Marcogliese, Reference Marcogliese2001, Reference Marcogliese2008; Harvell et al. Reference Harvell, Mitchell, Ward, Altizer, Dobson, Ostfeld and Samuel2002), although there is limited empirical evidence to support this at present (Bentley and Burgner, Reference Bentley and Burgner2011; Lõhmus and Björklund, Reference Lõhmus and Björklund2015).
An understanding of how host–parasite interactions will shift under the effects of warming, and the consequences for host populations and their communities, is thus an important aspect of environmental management (Lafferty, Reference Lafferty2009; Macnab and Barber, Reference Macnab and Barber2012). Integral to this is developing understandings of how elevated temperatures affect host susceptibility to infection vs their effects on parasite virulence and life cycle completion rates (Harvell et al. Reference Harvell, Mitchell, Ward, Altizer, Dobson, Ostfeld and Samuel2002; Altizer et al. Reference Altizer, Ostfield, Johnson, Kutz and Harvell2013). The susceptibility of hosts to infection could increase through, for example, thermal stress that leads to reduced immune-competency (Weyts et al. Reference Weyts, Cohen, Flik and Verburg-Van Kemenade1999; Nikoskelainen et al. Reference Nikoskelainen, Bylund and Lilius2004) and enhanced consumption rates of prey that leads to increased parasite exposure via intermediate hosts (Toscano et al. Reference Toscano, Newsome and Griffen2014). Parasite fitness and transmission rates could be enhanced by warming through positive effects on their metabolism, resulting in higher numbers of transmission stages being produced, with their rate of development and growth within hosts also accelerated (Paull and Johnson, Reference Paull and Johnson2011; Callaway et al. Reference Callaway, Shinn, Grenfell, Bron, Burnell, Cook and Shields2012). However, should warming result in the temperature optimum for the parasite being exceeded, then their decreased prevalence in host populations might result, with suggestions that increased parasite prevalence due to warming will only occur for a proportion of fish pathogens (Karvonen et al. Reference Karvonen, Rintamaki, Jokela and Valtonen2010). Consequently, there is an outstanding requirement to derive enhanced understandings of how warming will affect host–parasite dynamics, particularly the decoupling of the underlying mechanisms involved, i.e. the effects of warming on host susceptibility vs on parasite transmission and virulence.
The aim of this study was thus to test how elevated temperature affected host susceptibility to infection under different parasite exposure levels and how this affected parasite prevalence and intensity. Objectives were to quantify how temperature and parasite challenges affected: (i) infection outcomes (as parasite prevalence), (ii) host infection parameters (as parasite abundance, mean individual weight and burden); and (iii) host growth rates. Outcomes were assessed in relation to the effects of temperature elevation on the host–parasite relationship and the potential mechanisms involved. The model parasite was Pomphorhynchus laevis (Müller), an acanthocephalan with a complex life cycle whose final hosts are a wide range of fishes (Nedeva et al. Reference Nedeva, Atanassov, Karaivanova, Cakic and Lenghardt2003). The model final host was chub Squalius cephalus (Linnaeus), a preferred freshwater fish host of P. laevis (Hine and Kennedy, Reference Hine and Kennedy1974). This parasite uses the freshwater shrimp Gammarus pulex (Linnaeus) as its intermediate host. It is also a conspicuous orange-yellow parasite that is visible through the transparent cuticle of G. pulex (Bakker et al. Reference Bakker, Mazzi and Zala1997). This enables individual G. pulex to be identified for both their parasite status (infected/uninfected) and the number of parasites it is infected by. Transmission to fish hosts is via consumption of infected G. pulex, with some evidence that the parasite manipulates the behaviour of G. pulex to increase their probability of being predated upon and so enabling the parasite to be transmitted to its final host (e.g. Franceschi et al. Reference Franceschi, Bauer, Bollache and Rigaud2008; Dianne et al. Reference Dianne, Perrot-Minnot, Bauer, Gaillard, Léger and Rigaud2011; Labaude et al. Reference Labaude, Rigaud and Cézilly2015).
MATERIALS AND METHODS
Ethical approval
All experimental procedures used in the study were approved by the Ethical Review Committee of Bournemouth University and completed under UK Government Home Office project licence 30/3094. Where fish were euthanized, the procedure followed the UK Government Home Office Schedule 1 regulations.
Experimental design and pre-experiment data collection
The S. cephalus used in the experiment were all between 69 and 89 mm starting length (mean 80·8 ± 0·8 mm) and age 1+ years. They were sourced from an aquaculture site in Southern England in August 2014. Although they had not been exposed to the parasite during their lifetime, they were produced from broodstock that had originally been collected from a river where P. laevis was present naturally. On the aquaculture site, the fish were reared in outdoor ponds (approximate water temperatures at the time of collection: 15–19 °C), with some supplementary feeding with pelletized fishmeal. On arrival to the laboratory, the fish were tagged with passive integrated transponder tags (PIT tags), so that individual fish could be tracked through the experiment. Concomitantly, they were measured (fork length, nearest mm) and weighed (W, nearest 0·1 g). They were then allowed to recover and acclimate to laboratory conditions by being held in tanks held at 18 °C for 14 days on a 16:8 h light: dark cycle. In addition, a sub-sample of 5 fish was removed from the sample on arrival to the laboratory. These were euthanized and dissected to check for the presence of P. laevis. None of these fish were infected. Infections of other parasites were very light and considered part of the natural parasite fauna of the fish in Southern England and were recorded at levels that were not considered high enough to cause clinical pathology (Hoole et al. Reference Hoole, Bucke, Burgess and Wellby2001).
Parasite exposure
The S. cephalus were challenged by P. laevis through exposing individuals to known numbers of infected G. pulex. These were collected from a local river, the Hampshire Avon (latitude, 50·8865; longitude, −1·7883), when water temperatures were approximately 18 °C. These were then held in laboratory conditions at 18 °C for 96 h, with infectious individuals then identified visually (Bakker et al. Reference Bakker, Mazzi and Zala1997; Bauer and Rigaud, Reference Bauer and Rigaud2015), with a subset confirmed by dissection. As multiple infections were identifiable in the G. pulex (Bakker et al. Reference Bakker, Mazzi and Zala1997), then individuals were only used here that were host to one parasite. Exposure of the fish to the parasite was done individually, with the fish transferred to 10 L tanks containing dechlorinated water with supplementary oxygenation provided via an air stone and pump, and at a water temperature of 18 °C. Prior to parasite exposure, the fish were held in the tanks for 24 h with no feeding to ensure standardized levels of hunger.
Each individual fish was then exposed to a specific number of infected G. pulex from the following options: 0 (as a control), 5, 10, 20, 40 and 60. There were ten fish used at each level of exposure. After 24 h, the fish were removed from the tanks, with confirmation that all the G. pulex had been consumed. For each exposure level, the fish were then split randomly into two groups of five and transferred into 45 L tank aquaria at either 18 or 22 °C. These tank aquaria were arranged on a flow-through system using recirculated water (originally dechlorinated tap water), with a different system used for each temperature. Across the two flow-through systems used, the tanks were identical in dimensions, the water was taken from the same original source, and the tanks contained identical environmental enrichment for the fish in terms of refugia (lengths of plastic pipe of 65 mm diameter) and cover (artificial macrophytes).
Post-experiment data collection and analysis
Following their exposure to P. laevis, the fish were held in their tanks for 60 days under a 16:8 h light: dark regime, with feeding daily using crushed pelletized fishmeal (approximately 2% starting body mass/day). At the end of this period, the fish were removed from their tanks, euthanized, scanned for their PIT tag, re-measured and weighed. They were then dissected, with intestinal examinations to identify individuals in which infections by P. laevis had developed. For infected fish, parasites were removed, counted and weighed (mg).
These data enabled parasite prevalence to be assessed as the proportion of infected fish per temperature/exposure treatment. The effects of temperature (T) and parasite exposure (PE; as the number of consumed intermediate hosts) on prevalence were then tested using a probability of infection (PoI) model using binary logistic regression and the equation PoI = e (a+bT+cPE)/1 + e (a+bT+cPE), where a, b and c were binary logistic regression coefficients. This also provided the significance of both variables on parasite prevalence. As the tank conditions were identical across the individual fish, with only water temperature and levels of exposure to the parasite via intermediate hosts being different, and then the model did not take account of the fish being within different tanks per temperature treatment. Thus, the individual fish were being treated as the replicate unit in the model.
The following infection parameters were then calculated from the data of the infected fish. Parasite abundance was determined as the total number of parasites per host and the total mass of parasites per host, and enabled calculation of the mean parasite weight per host. Parasite burden was calculated as the proportion of the body weight of each host comprising P. laevis (Pegg et al. Reference Pegg, Andreou, Williams and Britton2015). Differences in these infection parameters, plus parasite prevalence, between temperatures were tested using generalized linear models (GLM), with parasite exposure level as the covariate. In all models, data on uninfected fish were not included as their inclusion in the models would introduce a bias in outputs, given the higher numbers of uninfected fish at the lower temperature/levels of parasite exposure. For parasite number, a Poisson log-linear model was used as the data represented parasite counts. As with the binary logistic regression model, in these models, the data for individual fish were used as the replicate units due to the identical conditions the fish were in, i.e. this was not considered as artificially inflating the number of degrees of freedom in the models that would otherwise result in pseudo-replication. The reported model outputs then included the mean value of the infection parameters per temperature treatment (as estimated marginal means, with the effects of parasite exposure as the covariate controlled in the model) and their standard error (s.e.). To identify if differences between these mean values were significant, linearly independent pairwise comparisons were used with Bonferroni adjustment for multiple comparisons. Differences in infection parameters were then tested between the exposure levels using the same process, except temperature was used as the covariate in these models.
Finally, to determine if infection influenced the growth rate of the fish, specific growth rate (SGR) was calculated as the change in body mass of the fish over the experimental period, from [lnW t+1–lnW t]/t × 100, where W t = starting weight, W t+1 = finishing weight, and t = number of days between W t and W t+1. Differences in specific growth rates of fish between temperatures and parasite exposure levels were then tested using GLMs as described above, with multiple linear regression analysis then used to test the influence of the infection parameters, temperature and parasite exposure on SGR. This provided the significance of the predictor variables and their standardized beta coefficients (β). Variables with the highest β value had the strongest singular contribution to the model.
Results
Probability of infection
At the conclusion of the 60 days after parasite exposure, there were considerable differences in infection levels apparent between temperatures and exposure levels (Fig. 1). The logistic regression model revealed both temperature and exposure level had significant effects on parasite prevalence (Fig. 1; Table 1). At 18 °C, infection required higher parasite exposure levels compared with 22 °C, with 50% prevalence requiring exposure to six intermediate hosts at 22 °C, but 26 at 18 °C (Fig. 1).
Fig. 1. Parasite exposure vs (i) proportion (0–1) of infected fish at 18° (filled circles) and 22 °C (open circles) and (ii) PoI (0–1 scale) according to binary logistic regression [cf. Table 1, equation (1)] at 18 °C (solid line) and 22 °C (dashed line).
Table 1. Binary logistic regression coefficients [Equation (1)] and their statistical significance, for the PoI of Squalius cephalus by Pomphorhynchus laevis according to temperature and parasite exposure.
Infection parameters
The GLM testing the effect of temperature on the parasite abundance of the infected fish revealed that there were significant differences in the mean numbers of parasites between the two treatments (Wald λ 2 = 4·23, P = 0·04), with mean parasite number significantly higher at 22 than 18 °C (P < 0·01; Fig. 2a). The effect of exposure on parasite abundance also revealed significant differences in mean number (Wald λ 2 = 20·46, P < 0·01), with significantly higher numbers of parasites per infected fish at exposure to 40 intermediate hosts (mean number: 7·80 ± 0·98) than at all than other exposure levels (mean numbers: 2·42 to 3·46; P < 0·01 in all cases; Fig. 2b). In both GLMs, the effect of the covariate was also significant (P < 0·05).
Fig. 2. Mean adjusted parasite number and mass, and mean parasite weight per fish (from generalized linear models) according to temperature (a, c, e), where parasite exposure was the model covariate, and parasite exposure (b, d, f), where temperature was the covariate. Error bars represent the s.e.
Temperature was not a significant predictor of parasite abundance when it was measured as the total parasite mass in the infected fish (Wald λ 2 = 0·01, P = 0·92; Fig. 2c), but parasite exposure was (Wald λ 2 = 13·10, P = 0·01). Mean total parasite mass was higher at 40 intermediate hosts (mean parasite mass: 24·23 ± 3·06 mg) than all other exposure levels (mean parasite mass range: 9·02 to 17·12 mg), although the differences were only significant between 40 and 60 hosts (difference 15·20 ± 4·35 mg; P < 0·01) (Fig. 2d).
The mean weight of individual parasites in the infected fish was significantly influenced by temperature (Wald λ 2 = 9·48, P < 0·01), being higher at 18 than 22 °C (P < 0·01; Fig. 2e). The effect of parasite exposure on the mean weight of individual parasites was also significant (Wald λ 2 = 13·29, P < 0·01), with higher means at lower exposure levels (Fig. 2f). The effect of temperature on parasite burden was significant (Wald λ 2 = 15·37, P < 0·01), with significantly higher burdens at 18 (0·23 ± 0·03%) than 22 °C (0·06 ± 0·03%) (P < 0·01). The effect of exposure on parasite burden was, however, not significant (Wald λ 2 = 7·63, P = 0·11).
Fish growth
Mean fish weight at the start of the experiment was 5·20 ± 0·16 g and at the end was 7·89 ± 0·31 g. The effect of temperature and parasite exposure on fish growth (as SGR) was not significant in either GLM (Wald λ 2 = 0·01, P = 0·91; Wald λ 2 = 5·01, P = 0·28, respectively). Multiple regression revealed the effects on SGR of all infection parameters, exposure and temperature were not significant (R 2 = 0·11; F 5,23 = 0·77, P = 0·56), with no significant predictors (all P > 0·05).
Discussion
Elevated water temperature had a significant and positive effect on parasite prevalence, with parasite infections developing from exposure to lower numbers of intermediate hosts in the warmer water. Despite these clear differences in prevalences, the effects of temperature and parasite exposure on the infection parameters of the individual hosts were relatively complex. Although elevated temperature resulted in increased parasite number in hosts, this involved a trade-off with their mass, with significantly smaller parasites present in hosts held at higher temperatures and resulting in significantly lower parasite burdens. These outputs on the infection parameters are a contrast to Macnab and Barber (Reference Macnab and Barber2012), who revealed that elevated temperature increased the growth rates of the parasite Schistocephalus solidus (Müller) in three-spined stickleback Gasterosteus aculeatus Linnaeus.
A major challenge in understanding how warming will affect host–parasite interactions is decoupling the individual effects of warming on the susceptibility of hosts to infection from the effects on parasite virulence. Here, the collection and holding of the parasite intermediate hosts, and the holding of the fish and their exposure to the parasite, was all completed at 18 °C, an ambient temperature representative of temperate freshwaters in the late summer period (Britton, Reference Britton2007). The exposed fish were then held at this ambient temperature and an elevated temperature (+4 °C) for the experimental period. With the initial parasite exposure all being completed at ambient temperature, it is suggested that the effect of the sudden temperature elevation in the treatment altered the susceptibility of the fish hosts to infection (Hakalahti et al. Reference Hakalahti, Karvonen and Valtonen2006), rather than it affecting the parasite virulence (Lõhmus and Björklund, Reference Lõhmus and Björklund2015). The sudden increase in temperature for this fish meant it was not possible to decouple the effect of the temperature effect on susceptibility per se from the specific effect of the rapid temperature increase. Nevertheless, that the net effect of the elevated temperature increased host susceptibility to infection was supported by evidence from other studies that suggest it often results in substantial negative consequences for fish immuno-competence (Dittmar et al. Reference Dittmar, Janssen, Kuske, Kurtz, Scharsack and Ardia2014), as it potentially shifts energy allocation from immunological processes (Poisot et al. Reference Poisot, Šimková, Hyršl and Mor2009) and/or acts as an additional stressor that compromises the immune response (Cramp et al. Reference Cramp, Reid, Seebacher and Franklin2014).
The complex effects of both temperature and parasite exposure on the infection parameters within the hosts were related to either temperature impacting the development rate of parasites or the increased parasite number in hosts at elevated temperatures resulting in marked density-dependent effects, resulting in relatively high densities of parasites with relatively small body sizes (Luong et al. Reference Luong, Vigliotti and Hudson2011). It is suggested that the latter explanation was more consistent with the outcomes of the experiment, given that these revealed fish exposed to high numbers of intermediate hosts at the ambient temperature resulted in low parasite numbers compared with the elevated temperature, but with these parasites being substantially larger, resulting in significantly higher parasite burdens.
Notwithstanding, as elevated temperatures can have both marked effects on the development rates of parasites in temperate regions (Tinsley et al. Reference Tinsley, York, Stott, Averard, Chapple and Tinsley2011) and on fish immune function, disease resistance and fitness (Cramp et al. Reference Cramp, Reid, Seebacher and Franklin2014), then it is remains difficult to definitively decouple the effects of warming on these aspects of the infection dynamics from these data. It is thus recommended that these outputs serve as an initial assessment of the effects of warming temperatures and parasite exposure levels on these host–parasite dynamics, enabling the design of subsequent experiments of greater complexity that should enable, for example, greater assessment of how warming affects the development rate the parasite within hosts, such as their maturity (e.g. Altizer et al. Reference Altizer, Ostfield, Johnson, Kutz and Harvell2013), how temperature affects the immune response of hosts (e.g. Nikoskelainen et al. Reference Nikoskelainen, Bylund and Lilius2004), and how parasite virulence is affected by the interactions of warming with other environmental variables, and the influence of this on selection (e.g. Wolinska and King, Reference Wolinska and King2009). Given the ease at which fish final hosts, such as S. cephalus, can be infected experimentally with known numbers of P. laevis via G. pulex intermediate hosts, then this host–parasite model would provide a strong model host–parasite system to answer these questions in both controlled and semi-controlled conditions. For example, to decouple the effects of host susceptibility from parasite virulence across different temperatures could utilize experiments where the fish and intermediate hosts are held at the different temperatures prior to exposure (unlike here, where they were all initially held at 18 °C) and then used in the experimental design used here. Parasites from these initial experiments could then be harvested and used to produce laboratory grown parasites in G. pulex that are raised across the different temperatures. Their subsequent exposure to the fish would then be completed in a fully factorial experimental design that enables quantification of differences in virulence and hosts susceptibility across the different generations and rearing temperatures of both G. pulex and the host fish.
Despite the strong effect of temperature on parasite prevalence and development, there were no measureable consequences for the hosts, with no differences in the specific growth rates of the fish between the controls, temperature and exposure treatments. Studies have suggested that P. laevis is a relatively benign parasite in temperate European fluvial fishes (Hine and Kennedy, Reference Hine and Kennedy1974), with the effects of ancanthocephalan parasites generally being more related to the consequences of their pathology rather than their loading (Latham and Poulin, Reference Latham and Poulin2002). Thus, it is suggested that the effect of elevated temperature on this host–parasite system was primarily in relation to altering host susceptibility to infection, with this then influencing parasite development and dynamics via density-dependent mechanisms within hosts. Consequently, the importance of these findings are that they indicate that warming could result in substantial shifts in disease progression via altered host susceptibility, but potentially with concomitant changes in parasite infectivity and development.
FINANCIAL SUPPORT
The study was supported by the Natural Environment Research Council (NERC research grant reference number NE/J016675/1 to D. J. S.).
