Experiments were conducted in climate-controlled facilities at Queens University Belfast (UK) and McGill University (Canada) to ensure environmental conditions were constant throughout trials. In the summers of 2016 and 2017, F. limosus were collected from two native populations (hereafter designated by N; Quinebaug River, Massachusetts: 42°06'32"N, 72°07'25"W; Panther Pond, Maine: 43°54'04"N, 70°27'55"W) and four invasive populations (hereafter designated by I; St Lawrence River, Quebec: 46°09'22.81"N, 72°59'54.85"W; St Croix River, New Brunswick: 45°37'01"N, 67°25'35"W; Lake Müggelsee, Germany: 52°26'54"N, 13°38'55"E; Albert Canal, Belgium: 50°56'34"N, 5°29'27"E). Crayfish collected from European sites were transported overnight by courier to Queen’s University Belfast. North American populations were transported by research vehicle from the field site to McGill University within 2–48 h of collection. The population of F. limosus from the St Lawrence River (I) was used first in distributed experiments and then in competitor-signal experiments three months later. Individuals of F. virilis used in competitor-signal experiments were collected from Blue Chalk Lake (N) in Dorset, Ontario (45°11'55"N, 78°56'20"W). For competitor-signal experiments, subjects were held in communal aquaria with up to five other individuals for three months prior to experiments. All crayfish collected from their invasive range were done so in areas where no other crayfish species currently co-exist. In contrast, those collected in their native ranges were from sites with sympatric crayfish species.

Crayfish were introduced to holding tanks at 18 °C immediately upon arrival and allowed to acclimate for at least one week prior to the start of experiments. Individuals were housed at low densities with ample shelter to mitigate territorial and aggressive behaviour (Reynolds et al. 2013) and they were fed a diet of sinking shrimp pellets. Aquaria maintenance included weekly water changes (25–50%, depending on tank size and crayfish density) and biweekly water quality testing. Given that Faxonius spp. are more active at night (personal observation), experimental animals were acclimated to a reversal of their usual 12:12 hour light:dark regime over a period of 96 h, following Alexander et al. (2012).

Low sample sizes of F. limosus obtained in the St Lawrence River (I) required that some individuals from this population be used in more than one trial, but each individual was tested only once at each density and in a maximum of three trials. To track individual identity, each crayfish was tagged with visible implant elastomer tags – a method that has been shown to have high tag retention rates and no influence on crayfish growth rates or mortality (Clark and Kershner 2006; Buřič et al. 2008).

Distributed experiments

All experiments were completed by the same researcher to minimize handler variation that often occurs in spatially distributed experiments coordinated among multiple research groups (Fraser et al. 2013; Dick et al. 2013). FR trials were conducted at ambient temperatures of 18 °C in opaque experimental tanks (61 cm × 35.5 cm filled with 10.8 L of water, and no substrate), and provided with a single PVC pipe as shelter (~12 cm length, 5 cm diameter). Experiments allowed for prey depletion, potentially underestimating the attack rate. However, the type of FR curve, handling time and maximum feeding rates remain uninfluenced by this procedure (Alexander et al. 2012).

Owing to natural variation in body size (carapace length) among populations of F. limosus crayfish and the low sample sizes available, no attempt was made to size-match individuals; instead, crayfish representing the estimated median size of individuals within each source population were used (see Suppl. material 1). Crayfish were introduced into experimental tanks 24 h prior to the beginning of experiments and were not fed during this period to standardize hunger levels. Following acclimation, each individual was randomly assigned one of 11 prey density treatments (3, 4, 5, 7, 10, 15, 20, 25, 30, 40 or 50 gammarid amphipod individuals of Gammarus pulex in the UK and Gammarus fasciatus in North America). In addition, one control treatment at each prey density was conducted in the absence of a crayfish to account for potential mechanisms of prey death other than crayfish predation. Experiments were run for 6 h in the dark, after which the crayfish were removed from experimental tanks and the remaining gammarids counted to determine the number of prey consumed. This procedure was replicated three times for each crayfish population (n = 33, plus controls for each population). Following each trial, crayfish were blot-dried and weighed, and their carapace length was measured.

Amphipod prey activity levels

We did not have access to a gammarid prey species common to both the UK and eastern North America (NA); therefore, it was necessary to account for differences in body size (length) and activity levels of a subsample (n = 30) of Gammarus spp. from each region. Activity level was measured at 18 °C by placing an individual amphipod into a petri dish filled with 1 cm of dechlorinated tap water, allowing the individual to acclimate for 90 s, and then counting the number of times it crossed the center of the dish in 60 s (Maynard et al. 1998).

Competitor-signal experiments

This second set of FR experiments took place between February 10 and April 18, 2017, and consisted of six experimental treatments using the two crayfish species in a full factorial design, plus predator-free controls (Table 2). Experiments featured ‘focal’ and ‘competitor’ crayfish(es) that occupied a shared experimental arena but had minimal physical contact. Experimental conditions were identical to those of the distributed FR experiments, except that the addition of a lidded container with 1cm holes in each side was included in the experimental arena. The focal crayfish wandered freely in the experimental chamber and could access the shelter, while the competitor crayfish was confined to the secondary container. The container holes were large enough to allow crayfish antennae and claws to pass through, but the competitor was unable to access the prey.

The focal and competitor crayfish were introduced to the experimental chamber simultaneously, 24 h before the beginning of the trial. The beginning of the trial was signalled by the introduction of defrosted bloodworms (Diptera, Chironomidae, Chironomus) to the experimental chamber, in each of the following prey densities: 15, 20, 25, 30, 40, 50, 60, 70, 80, 100, and 120 individuals. Trials lasted 6 h in the dark and allowed for prey depletion. Each of the 11 densities in each treatment was replicated in triplicate (n = 33 for each of the six experimental treatments, plus one replicate at each density as a predator-free control). After each trial, crayfish were blot-dried and weighed, and their carapace length measured. The remaining prey were counted to determine the number attacked during the trial. Prey were scored as ‘attacked’ if at least part of the worm had been eaten (determined by fragmented worms and loss of colour caused by draining of hemolymph). Owing to insufficient numbers of experimental animals, individual crayfish were re-used in trials up to 10 times, but never re-used twice at the same density, regardless of treatment.

Model selection and fitting

All analyses were completed using R (version 3.2.4). As is appropriate for prey non-replacement designs, FR was modeled using the Random Predator Equation (Rogers 1972) for Type II curves and Hassell’s equation (Hassell et al. 1977) for Type III curves, following Alexander et al. (2012) and Iacarella et al. (2015a). Model selection was conducted using three methods outlined by Pritchard et al. (2017). The best fit model was selected for each population (distributed experiments) and experimental treatment (competitor-signal experiments), based on consensus of selection methods. Best-fit models were fit to FR data via maximum likelihood estimation (see Suppl. material 2).

Model comparisons

To compare the fitted FR curves among populations and experimental treatments, the data were bootstrapped (n = 999) to produce 95% confidence intervals (CI) on the fit. Using this method, we may statistically compare models between populations by simply observing the overlap, or lack of, between model CIs (Pritchard et al. 2017). In addition, the frair_compare function of the FRAIR package was used to compare the difference in handling time (h) estimates between populations, but is limited to comparing those that are fit by the same model type (Pritchard et al. 2017). An overall assessment of differences between model parameters can also be achieved by observing overlap between 95% CIs, as with full model comparisons. The effect of crayfish size (carapace length and weight) on maximum feeding rate and proportion of prey killed was tested using linear models (linear regression for each population). Sex-related effects were also assessed for each population using t-tests to compare the proportion of prey consumed between sexes.

Functional responses differed among populations by maximum feeding rate and curve type. The responses of populations from Lake Müggelsee (I), Albert Canal (I), and the Quinebaug River (N) were best fitted by a Type II curve, while those of remaining populations were best fitted by a Type III curve (Fig. 1). Bootstrapped 95% confidence intervals indicated significant overlap in model fits and maximum feeding rates among the populations, regardless of whether the population was native or invasive (Fig. 1); however, parameter comparisons using the frair_compare function indicate a significantly greater maximum feeding rate (lower handling time, h) for the population of F. limosus from the St Croix River (I) than either the Lake Müggelsee (I) or Albert Canal (I) populations. Similarly, the Panther Pond (N) population exhibited a greater maximum feeding rate than the St Lawrence River (I) population (Table 1).

Figure 1.

Functional response curves with bootstrapped 95% confidence intervals (shaded regions) for F. limosus from native and invasive populations. Lines represent the best fit model for each population (Type II or Type III). n = 33 for each population.

Crayfish predation was the principal source of prey death in experimental trials, as indicated by high survival rates in control treatments (across all populations, controls exhibited 99.98% survival of prey during 6-hour experimental trials). Overall, maximum feeding rates declined with mean crayfish size and weight (linear models; carapace length: F_1,4 = 10.83, p = 0.030, weight: F_1,4 = 10.40, p = 0.032), but the size effects on prey consumption varied among populations. The proportion of prey consumed increased with crayfish size for the Panther Pond population (N) (linear models; carapace length: F_1,31 = 4.36, p=0.048, adjusted r² = 0.09, weight: F_1,31 = 6.91, p = 0.013, adjusted r² = 0.16), but decreased with crayfish weight for the Albert Canal population (I) (linear model; F_1,31 = 4.68, p = 0.038, adjusted r² = 0.10). Female crayfish from the Quinebaug River population consumed a marginally greater proportion of prey (t-test; t₂₈ = 2.45, p = 0.021) than males. No differences in prey consumption were detected between crayfish sexes in other populations.

Amphipods used as prey in FR experiments in the UK (G. pulex; mean size ± SE = 6.67 mm ± 0.50) and North America (G. fasciatus; 6.10 mm ± 0.48) did not differ in size (Mann-Whitney U Test; W = 560.5, p = 0.091). However, North American prey were significantly more active than UK prey, crossing the centre of the disk an average of 5.7 times per minute (SE = 1.8) while G. pulex in the UK crossed an average of 3.7 times per minute (SE = 1.1) (Mann-Whitney U Test; W= 624, p = 0.0098).

The maximum feeding rate of F. limosus was suppressed in the presence of conspecific and heterospecific competitors, whereas the handling time (and thus, maximum feeding rate) of F. virilis was unaffected (Fig. 2, Table 3). However, despite apparent inhibition in the presence of competitors, F. limosus consistently exhibited a significantly higher maximum feeding rate than that of F. virilis (Fig. 2). All treatments were best fitted by Type II curves.

Despite significant size differences between F. limosus and F. virilis, carapace length and crayfish weight were not significant predictors of maximum feeding rate (linear models; carapace length: F_1,4 = 0.055, p = 0.83; weight: F_1,4 = 0.059, p = 0.82).

Figure 2.

Maximum feeding rate (MFR) calculated for each treatment in the competitor-signal experiments (1/hT, where h is estimated handling time and T is experimental duration). Treatment codes represent the focal and competitor crayfish species in each treatment – L = F. limosus alone, LL = F. limosus with an F. limosus competitor, LV = F. limosus with an F. virilis competitor, V = F. virilis alone, VL = F. virilis with an F. limosus competitor, and VV = F. virilis with an F. virilis competitor. F. limosus used in these experiments came from an invasive population, whereas F. virilis came from a native population. Bars indicate the standard errors of the MFR calculated by propagating the model fit standard error given for h for each treatment. Differences in letters above error bars indicate significant differences (α = 0.05) between treatments. n = 33 for each treatment.

Our competitor-signal experiments tested the influence of biotic context on per capita effects of invasive species and found that closely related species differ in their response to the presence of competitor signals. Invasive F. limosus adjusted its feeding behaviour in the presence of conspecific and heterospecific competitor signals, whereas native F. virilis did not (Fig. 2). However, even where the maximum feeding rate of F. limosus is suppressed, it still exceeds that of the native, consistent with the resource consumption hypothesis (Dick et al. 2013; Ricciardi et al. 2013). This result suggests that F. limosus can have a greater impact on prey populations than F. virilis where the two species are sympatric; but as demonstrated, impact differs between F. limosus populations. Therefore, conclusions on the potential species impacts cannot be reliably drawn from estimates from only one population. We tested individuals from a population of F. virilis that was sympatric with a native congener (F. propinquus), whereas our F. limosus population was allopatric. We hypothesize that the suppression of feeding rates resulted from intimidation by adversarial crayfish, the effect of which could dampen over time as dominance hierarchies are established and agonistic interactions decline (Goessmann et al. 2000). Thus, a likely explanation for the unaffected feeding rate of F. virilis in the presence of a competitor is that individuals from this population are accustomed to the presence of other crayfishes.

For F. limosus, the presence of congeners can trigger individuals to trade off foraging effort with shelter protection. Access to suitable shelters is crucial for crayfish survival by enhancing predator avoidance, facilitating successful moulting, and reducing the frequency and intensity of agnostic interactions with other individuals (Nyström 2005, Hirsh et al. 2016, Lele and Pârvulescu 2017). Previous studies have demonstrated the superior ability of invasive crayfishes to evict natives and co-opt shelters for themselves (Lele and Pârvulescu 2017), aided by increased aggression (Klocker and Strayer 2004). Therefore, a reduced maximum feeding rate could have been the result of F. limosus sacrificing foraging opportunities to maintain possession of shelter in the presence of a perceived competitor. A competitor signal might have triggered increased aggressive or territorial behaviour in F. limosus at the expense of sacrificing its foraging effort. We did not make observations throughout the trial and so cannot confirm these behaviours. In contrast, the lack of a behavioural change in F. virilis in the presence of competitors might also be explained by shelter usage; Garvey et al. (1994) examined interspecific competition for shelter among congeneric Faxonius crayfishes and determined that among the three species studied, F. virilis was most often found actively mobile outside of shelters even in the presence of predators. Bold behaviour and naivety to competition from non-native species could prevent crayfishes from trading off feeding intensity for shelter or territorial protection.