Superior predatory ability and abundance predicts potential ecological impact towards early-stage anurans by invasive ‘Killer Shrimp’ (Dikerogammarus villosus)

Invasive alien species negatively impact upon biodiversity and generate significant economic costs worldwide. Globally, amphibians have suffered considerable losses, with a key driver being predation by large invasive invertebrate and vertebrate predators. However, there is no research regarding the potential ecological impact of small invertebrate invaders. The invasive freshwater amphipod Dikerogammarus villosus can act as a top predator capable of displacing native amphipods and preying heavily upon a range of native species. Listed as one of Europe’s top 100 worst invaders, D. villosus has significantly restructured freshwater communities across western Europe and is expected to invade North America in the near future. Here we explore the ecological impact of invasive D. villosus upon UK native and invasive amphibians (Rana temporaria and Xenopus laevis respectively) using the “Relative Impact Potential” (RIP) metric. By combining estimations of per capita effects (i.e. functional response; FR) and relative field abundances, we apply the RIP metric to quantify the potential ecological impact of invasive D. villosus upon embryonic and larval amphibian prey, compared to the native amphipod Gammarus pulex. Both native and invasive amphipods consumed early-stage amphibians and exhibited potentially destabilising Type II FRs. However, larger body size in invasive D. villosus translated into a superior FR through significantly lower handling times and subsequently higher maximum feeding rates—up to seven times greater than native G. pulex. Higher invader abundance also drove elevated RIP scores for invasive D. villosus, with potential impact scores predicted up to 15.4 times greater than native G. pulex. Overall, D. villosus is predicted to have a greater predatory impact upon amphibian populations than G. pulex, due primarily to its larger body size and superior field abundance, potentially reducing amphibian recruitment within invaded regions.

The RIP metric returned a substantially greater impact potential in invasive D. villosus, relative to native G. pulex ( Table 3). The RIP metric indicated that considerably higher RIP scores estimated for large and intermediate D. villosus, compared to G. pulex, were driven by different biological characteristics. Despite the non-significant differences in FR parameters between size-matched amphipods (Table 1), significantly superior field abundance estimates recorded for intermediate D. villosus generated a greater RIP score, than large G. pulex ( Table 3). The RIP biplot illustrates this, with differential field abundance estimates generating a substantial vertical shift for intermediate D. villosus in comparison with the RIP for large G. pulex (Fig. 3). Large D. villosus displayed similar field abundances, when compared to large G. pulex (Table 3). However, significantly lower handling times, and subsequently higher maximum feeding rates, resulted in a superior RIP scores, with the RIP biplot highlighting a substantial shift to the right when compared to G. pulex (Fig. 3).

Discussion
Over the past 20 years, D. villosus has spread rapidly throughout Europe 15 . Within invaded communities, D. villosus has a significant ecological impact, with considerable declines in native macroinvertebrate populations and altered ecosystem functioning ascribed to its competitive and predatory capabilities 14,15,[42][43][44][45][46] . Whilst previous evidence concerning the predation of early stage amphibians by amphipods is scarce (see Fries & Tesch 47 ; Hudgens & Harbert 48 ), we provide the first empirical evidence that amphipods can kill and consume both embryonic and larval amphibians. Greater per capita feedings rates combined with higher densities in the field lead us to predict that invasive D. villosus might also exert a population-level effect upon amphibians within invaded communities.
Predation of invasive X. laevis embryos was observed, with D. villosus consuming a significantly larger number of embryos -increasing with invader body size-whilst predation by native G. pulex was negligible. When presented with native R. temporaria embryos, both native and invasive amphipods appeared incapable of consuming these larger embryos, although there was evidence of attempted predation. Predation of native R. temporaria embryos by invasive amphipods was only observed when embryos were damaged prior to exposure. Rana temporaria embryos are surrounded by a comparatively thick vitelline jelly capsule 49 , which protects the eggs from some predators 50 . Rana temporaria embryos may be susceptible to predators with piercing, sucking mouthparts, yet reasonable invulnerable to predators which possess chewing, biting mouthparts 51 , such as amphipods, thereby escaping potential predatory pressures exerted by invasive amphipods upon embryonic amphibians.
Invasive X. laevis embryos are comparably smaller than those generated by native R. temporaria (2.19 ± 0.02 mm and 7.83 ± 0.16 mm, respectively; pers. obs.), and other anurans 52 . Our findings suggest that D. villosus may present a predatory threat to other native amphibian species with relatively small embryos, such as the great crested newt (Triturus cristatus; embryo diameter = ~ 4.5 mm), the smooth newt and the palmate newt (Lissotriton vulgaris and L. helveticus; embryo diameter = ~ 3 mm 53 ); of which several species have been recorded in regions invaded by D. villosus (e.g. Grafham Water and Pitsford Reservoir; Anglian Water, pers. comm; The Wildlife Trust for Bedfordshire, Cambridgeshire & Northamptonshire, pers. comm). Our findings also retain ecological relevance with regards to invasive X. laevis populations, given the previous existence of several feral populations in Great Britain 54 . www.nature.com/scientificreports/ Native and invasive amphipods readily preyed upon R. temporaria larvae. Large D. villosus expressed significantly lower handling times, consuming early-stage amphibians at a greater rate than smaller amphipods, which translated into substantially higher maximum feeding rates, seven times greater than large G. pulex. As such, the higher per capita prey intake observed in significantly larger invaders is likely explained by the naturally larger body size of D. villosus, rather than any interspecific differences in innate predatory ability 16,24 . Our observation of higher consumption by larger amphipods is consistent with previous studies 16,24 , and by extension, general biological theory 55 . Metabolic theory predicts that metabolic rate typically exhibits positive allometric scaling relative to size 56 , with greater metabolic demand requiring high resource intake to maintain fundamental biological processes 57 .
Superior consumptions rates by larger amphipods may be facilitated by larger mouthparts and gnathopods, allowing individuals to capture and consume a wide range of prey 58 , or a larger gut capacity required to efficiently digest food items 59 . The presence of large glycogen reserves in D. villosus, compared to other amphipod species, allow this invader to overcome various anti-predatory evasion behaviours demonstrated by larval amphibians, such as burst-swimming 60 ; a trait which appears superior in ranids, compared to other anurans (e.g. bufonids 61 ). Given that R. temporaria larvae remain vulnerable to predation until complete metamorphosis 62 , we posit that the potential impact of invasive D. villosus could persist throughout development, with prolonged predation on growing larvae in the field continuing until they achieve a size or developmental stage which is invulnerable to D. villosus.
The comparative FR approach revealed that, when compared to native G. pulex, invasive D. villosus generally exhibited a higher Type II FR. This differential predatory response became more apparent in larger invaders, with significantly higher FRs exhibited by large D. villosus feeding upon embryonic and larval amphibians. Separation between FR curves generated for large invaders and size-matched amphipods across both prey systems would imply the potential for D. villosus to impose a greater predatory impact upon native amphibian populations, compared to native G. pulex.
Type II FRs are indicative of potentially unstable predator-prey interactions 63 . At higher prey densities, per capita predation rates decelerates to an asymptote as consumption become limited by consumer handling times 63 . As a result, an unstable equilibrium is attained, centred on the asymptotic point 63 . If predation levels exceed the growth capacity of prey populations existing at densities below the established equilibrium point, predator-prey dynamics may destabilise resulting in the extirpation of affected prey populations 63,64 . Differential Type II FRs identified in comparative laboratory-based studies of predation on macroinvertebrate prey are consistent with observed field patterns of reduced macroinvertebrate diversity and abundance 15,43 . Our findings suggest a similar threat may extend to amphibians in the field.
When compared to native amphipods, D. villosus exhibits notably higher fecundity and a short interbrood interval 18 , allowing this invader to rapidly form "super-abundant" populations in invaded regions 14 . The RIP metric highlighted a significantly greater impact potential by invasive D. villosus than by native G. pulex, driven by both larger body size and greater abundance of this invasive species. Large D. villosus exhibited a higher per capita impact than native G. pulex, translating into an RIP score 6.4 times stronger than native G. pulex. A superior RIP score was also predicted for intermediate D. villosus, when compared to G. pulex. Whilst the comparative FR approach identified no significant difference between per capita effects recorded in size-matched native and invasive amphipods, inclusion of field abundance estimates into the RIP metric detected a substantially higher impact potential for intermediate D. villosus, with significantly higher field densities (4.8 times higher than G. pulex) generating a predicted impact score approximately 15 times greater than its native counterpart.
The RIP metric considers the effect of differential field abundances of natives and invaders, but assumes that consumer interactions are explicitly advantageous 8 . In reality, interactions between consumers may be additive 65 , synergistic 66 , or antagonistic 67 . By incorporating such context dependencies into FR models, we might further refine predictions. Nevertheless, the RIP metric has proven to be an effective predictive tool when applied to previous literature. Estimations of invader RIP support alternative impact measurements (e.g. Laverty et al. 68 ) and correspond to observed field impacts 8 . As such, the RIP metric has formed the foundation for several alternative quantitative metrics (see Dickey et al. 69 ).
In the current study, the RIP metric highlighted a significantly greater impact potential by invasive D. villosus towards native early-stage amphibians, when compared to native G. pulex. These findings are consistent with similar magnitudinal patterns of differential impacts identified in D. villosus towards other freshwater organisms (see Dick et al. 8 ). However, our estimates for the abundance of D. villosus in Grafham Water Reservoir were considerably lower than those recorded in other European and UK localities in which R. temporaria have been reported. The potential for D. villosus to reach higher densities indicate that the potential impact of this invader upon early-stage amphibians may be even stronger in other invaded regions.

Conclusions
This is the first empirical evidence of predation of early-stage amphibians by freshwater amphipods. The invasive D. villosus exhibited consistently higher per capita predation rates upon invasive amphibian embryos and native amphibian larvae, with predation increasing as a function of invader body size. The detection of Type II FRs, significantly higher in large-bodied invaders, are indicative of the potential ecological impact of D. villosus, with higher predation rates predictive of a depletive, potentially destabilising effect upon amphibian populations, through the consumption of vulnerable embryos and larvae. This higher ecological impact, predicted for invasive D. villosus, is further intensified when the higher field abundances of this invasive amphipod are considered.
Large-bodied invasive predators are acknowledged as primary drivers of global amphibian declines 33,34 . With evidence of potential co-occurrence between D. villosus and native amphibians, recorded in both UK freshwaters (Anglian Water, pers. comm.; Cardiff Harbour Authority, pers. comm.) and in mainland Europe 36 www.nature.com/scientificreports/ findings of the current study predict that the highly predacious 'killer shrimp' may further contribute to declining amphibian populations through the predation of early life-stages. Given the projected expansion of D. villosus in British freshwaters 70,71 , we might predict the introduction of D. villosus into amphibian-rich areas in the near future, with consequences for amphibian populations expected to follow. However, further research is required to determine the suitability of different freshwater habitats which are typically used by breeding amphibians.

Materials and methods
We compared the predatory impacts of invasive D. villosus, and the British-native amphipod, G. pulex, upon the early life-stages of amphibians. Initial experiments used invasive X. laevis embryos as a focal prey organism and established the potential for native and invasive amphipods to predate upon early-stage anurans. Therefore, experiments proceeded utilising the embryonic and larval forms of native R. temporaria as focal prey types. Rana temporaria have been recorded occupying the same habitat as D. villosus, both in UK invaded sites 36 (The Wildlife Trust for Bedfordshire, Cambridgeshire & Northamptonshire, pers. comm), and in mainland Europe 37,38 . We compared size-matched amphipods to examine intrinsic differences between species, as well as significantly larger D. villosus to reflect natural differences in amphipod size 16,24 . A comparative functional response (FR) approach was utilised to quantify amphipod predation upon invasive and native amphibian embryos and larvae. FRs are a fundamental measure of resource use frequently applied in invasion ecological research as a metric to assess trophic interactions; quantifying the relationship between per capita predation rate and prey abundance (i.e. FR). By comparing FRs of IAS and native analogues predictions can be made as to how differential predator behaviours might impact upon prey populations in the field 7,8 .
When considering the absolute ecological impact of invasive predators, total invader impact should consider predatory capability, relative to native analogues, as well as relative abundance 8 . Based on the classic total response equation (Total Response = Functional Response x Numerical Response), the Relative Impact Potential (RIP) metric has recently been developed and incorporates relative consumer abundance or biomass-a proxy for numerical response-as a means of scaling relative per capita effects (i.e. FR) to predict the relative impact of an invasive predator in comparison with a native analogue (RIP = FR x Abundance 8 ).
Experimental organisms. Ethical consent was obtained from Natural England, the Home Office and the University of Leeds Ethics Committee. The use of freshly hatched, pre-feeding R. temporaria larvae for experimentation fell outside the remit of the Wildlife and Countryside Act 1981 (section 9.5; protected against sale only), and the UK Animal Scientific Procedures Act 1986 (ASPA; section 1.4.2). Animals were maintained in compliance with guidelines stated in the Code of Practice for the Housing and Care of Animals Bred, Supplied or Used for Scientific Purposes (section 3, chapter 11). All experimental work was conducted in accordance with relevant guidelines and regulations, including the maintenance, use and termination of study organisms.
Xenopus laevis embryos were sourced from adult females, commercially reared by the European Xenopus Resource Centre (EXRC, University of Portsmouth). In December 2016, embryos were transported to the laboratory, stored in isotonic 1 × Modified Barth's Saline (MBS) solution (Supplementary Materials Table S1). Upon receipt, embryos were gradually transferred into aerated dechlorinated tap water over the course of several hours and kept at 14.0 ± 0.1 °C under a 12:12 h light:dark regime, as recommended by the EXRC.
Freshly deposited native R. temporaria embryos (approximately 36 h post-fertilisation) were collected between February and March 2017 from several freshwater sites around West Yorkshire (Supplementary Materials Table S2). Embryos were removed as whole clutches and approximately halved, with half of the clutch transferred to 2L sterile storage containers with site-sourced water, and half of the clutch returned to the site. Harvested embryos accounted for less than 10% of the total population of embryos present at each site.
Embryos were transported to the laboratory in insulated boxes and stored as individual half-clutches in aerated aquaria with dechlorinated tap water at 4.0 ± 0.2 °C under a 12:12 h light:dark regime. By maintaining the embryos at 4 °C, the rate of embryonic development was reduced considerably (~ 30 days to hatching), maximising the potential experimental period whilst enabling greater control over developmental progress.
To obtain R. temporaria larvae for experimentation, embryos were reared to early-stage larvae. When larvae, still encapsulated in vitelline jelly, began to develop external gill filaments and exhibited neuromuscular reflex responses (i.e. Gosner stage or G 18-19; see Gosner 72 ), they were transferred to 14 °C (consistent with ambient temperatures recorded during field sampling) in preparation for hatching. Transference to the higher temperature regime, conducted gradually over the course of 24 h, accelerated development, with hatching typically occurring within approximately 24 h of changing temperature regimes. Shifting temperature regimes also allowed larvae to acclimatise to warmer conditions prior to experimentation. Conditions were sufficient to produce high rates of larval hatching (> 75%) and survival (~ 70%). Embryonic and larval stock tanks were cleaned twice weekly. Only recently hatched, pre-feeding larvae (i.e. G. 19-20 72 ), lacking any obvious functioning mouthparts and relying solely on the yolk sac for nutrition 73 , were used for experimentation.
Male amphipods were used in experimental treatments as females may exhibit variations in predatory behaviour 76  www.nature.com/scientificreports/ the presence of genital papillae, and the absence of oostegites (i.e. brood plates). Amphipods exhibiting visible parasitic infections were excluded from experimentation, controlling for potential variations in behaviour caused by infections 77,78 . Amphipods were kept in sex-specific communal tanks for at least 24 h prior to their use in experimental trials, and were only used once in each experimental treatment. Given the significantly larger natural body size of D. villosus when compared to other European gammaridean amphipods 21 , amphipods were categorised into three size groups; large G. pulex, intermediate D. villosus and large D. villosus. Controlling for amphipod body size enabled fundamental comparisons of inherent differences in predatory impact between size-matched native (large G. pulex) and invasive (intermediate D. villosus) amphipod groups, whilst also considering the predicted amplificatory effect that larger natural body size in D. villosus may have on maximal predatory impact 16 .
Amphipods were blotted dry, weighed and photographed in a resting curved state, with measurements taken approximately 2 h prior to the starvation of amphipods in preparation of experimental trials. Body length was measured as a curved line from the rostrum tip to urosome base in Image J (http://rsbwe b.nih.gov/ij/). Rarefaction of datasets using size parameters recorded for amphipod groups used across all experimental systems indicated appropriate size-matching between large G. pulex (mean ± standard error (SE), length = 16 Table S3).

Functional response (FR) experiments. Experimental design.
To compare predatory FRs of native and invasive amphipods against amphibian prey, three independent experiments were conducted in which amphipods were presented with amphibian prey in varying densities. The first experiment compared FRs between native and invasive amphipods towards invasive X. laevis embryos. The second experiment compared amphipod FRs towards native R. temporaria embryos. The third experiment assessed amphipod FRs towards R. temporaria larvae.
Prior to experimentation, individual amphipods were placed in clear plastic arenas (90 mm diameter, 50 mm height) with 250 ml of dechlorinated tap water, and starved for 24 h. A single glass bead (20 mm diameter, 9 mm height) was placed in arenas as substrate, providing amphipods with shelter and to prevent continuous swimming. Amphipods were then transferred to experimental arenas, identical to those described above, containing a known number of invasive X. laevis embryos (2,4,6,8,10,15,20,40 Table S4). Prey were situated in arenas two hours prior to the introduction of amphipod predators and the commencement of trials.
Experimental trials began with the introduction of a single amphipod predator and were conducted at 14 ± 0.1 °C under a 12:12 h light:dark regime. Trials continued for 24 h (X. laevis/R. temporaria embryos) or 48 h (R. temporaria larvae), without replacing consumed prey. Trials concluded with the removal of amphipod predators and the enumeration of alive, dead or consumed prey. Dead prey which did not exhibit signs of predation were assumed to reflect background mortality (< 1.24% in all experiments). After terminating experimental trials, amphipods were monitored for a further 24 h. Amphipods that moulted or died were excluded from analysis. Following rarefaction to ensure appropriate size-matching, data pertaining to embryonic prey treatments was retained for five replicates, whilst the larval prey treatment comprised up to eleven replicates at all prey densities. Controls consisted of five (embryos) or eleven (larvae) replicates of each prey density, without amphipod predators present.
Field sampling: estimating amphipod abundance. In November 2017, field sampling was undertaken at several un/invaded freshwater sites within Great Britain to estimate field abundances of native G. pulex and invasive D. villosus within these regions. Dikerogammarus villosus were sampled from six sites situated around the perimeter of Grafham Water, Cambridgeshire (Supplementary Materials Table S5). Sampling was conducted approximately 2 m from the shoreline within a 50 × 50 cm area (0.25m 2 ), using a modified, bottomless receptacle (50 cm diameter, 65 cm height, 80L volume) which allowed access to the underlying substrate. Over a five-minute period, the substrate was agitated and netted, followed by two minutes of netting through the water column.
Gammarus pulex were sampled from five sites along Adel Beck and Meanwood Beck, West Yorkshire (Supplementary Materials S5). Sampling was conducted in the centre of these lotic systems, within a 50 × 50 cm area (0.25 m 2 ). The substratum was agitated for 5 min and any amphipods dislodged were collected in a surber sampler. A further two minutes were spent hand sampling larger rocks present within the sampling area.
Amphipod specimens were stored in 70% ethanol. In the laboratory, amphipod specimens were sorted into size categories, matching those selected for FR trials, and enumerated.
Abundance data was supplemented using estimates reported in previously published literature; recorded for G. pulex within native ranges [78][79][80][81][82][83][84][85] , and D. villosus within invaded ranges 39,81,84,[85][86][87][88][89][90][91][92][93] . Data was taken from studies of amphipod abundance at sites where the presence of native R. temporaria, and other European amphibian species, had also been recorded (amphibian occurrence taken from recording databases 36,94,95  Statistical analysis. Statistical analyses were performed in R studio, version 3.3.2 96 . FR analyses were conducted using an integrated package for functional response analysis in R (frair, version 0.5.100 97  www.nature.com/scientificreports/ Functional response analysis. Phenomenological functional response analysis. A phenomenological approach was applied to each experimental combination (amphipod x prey type) to determine FR type (I, II, III) based on the general shape of the response curve. For each amphipod x prey type combination, logistic regressions of proportional prey consumption against prey density were performed, fitted with a quasibinomial error distribution to account for overdispersion. A significant negative first-order term was indicative of a Type II FR, whilst a significant positive first order term, superseded by a significant negative second order term denoted a Type III FR 64 .
Mechanistic functional response analysis. Where analyses suggested that Type II FRs were appropriate, FRs were modelled using the Rogers' random predator equation (Eq. 1) 98 . This model accounts for the depletion and non-replacement of prey 64 .
where N e is the number of prey consumed, N O is the initial density of prey, a and h represent the mechanistically explicable coefficients for attack coefficient ( a ) and handling time ( h ), and T is the total experimental period in days. Using these parameters, maximum feeding rate was calculated as 1/Th . FR models were fitted using the frair_fit function, which utilises maximum likelihood estimations (bbmle::mle2, version 1.0.20 99 ), and a modified version of Eq. (1), incorporating the Lambert W function to resolve the presence of N e on either side of the equation (Eq. 2).
Comparisons of attack coefficient ( a ) and handling time ( h ) were conducted between amphipod groups (within each prey type) using an 'indicator variable' approach to explicitly model differences in the parameter estimates for each amphipod group (frair_compare function; see Juliano 64 , Pritchard et al. 97

, Taylor & Dunn 24 ).
Each fitted FR model was non-parametrically bootstrapped (n = 2000) to generate 95% confidence intervals, thereby visualising model variability. Additional non-parametric bootstrapping (n = 30) was applied to models, allowing multiple estimates of handling time ( h ), and thus maximum feeding rate ( 1/Th ) to be calculated 16,100,101 . This generated mean (± SE) estimated maximum feeding rates for RIP calculations.
In FR experiments which utilised invasive X. laevis embryos as focal prey, negligible predation recorded in native G. pulex prevented comparisons of FR curves and parameters between native and invasive amphipods. Instead, Chi-square (χ 2 ) tests were applied to compare the frequency of predation (i.e. proportion of individuals that consumed embryos) recorded between amphipods. Chi-squared tests were also conducted for FR experiments with native R. temporaria larvae, given the relatively low incidence of predation recorded in size-matched amphipod groups, compared to large D. villosus. In FR experiments with native R. temporaria embryos, a complete absence of predation prevented statistical analyses.
Amphipod field abundance estimates. Field abundance estimates for native and invasive amphipods were compared using a generalised linear model (GLM), fitted with a quasipoisson error distribution to account for overdispersion. A post-hoc Tukey HSD test (α = 0.05; multcomp::glht, version 1.4-8 102 ), was subsequently conducted to compare field abundance estimations between amphipods.
Relative impact potential (RIP) analysis. Mean (± SE) estimates for maximum feeding rates (i.e. FR), generated from bootstrapped models, and field abundances were incorporated into the RIP metric, enabling pairwise comparisons of relative impact potential between invasive and native amphipods. This allowed RIP probabilities and confidence intervals to be generated for invasive amphipods, when compared to native G. pulex. Due to the potential ecological significance of invader predation upon native amphibian species, RIP analyses focussed on FR models pertaining to the predation of native R. temporaria larvae. 'RIP biplots' were generated, presenting the RIP values of the three amphipod groups using field abundance estimates as a proxy for numerical response (see Laverty et al. 100 ; Cuthbert et al. 101 ).