A novel metric reveals biotic resistance potential and informs predictions of invasion success

Invasive species continue to proliferate and detrimentally impact ecosystems on a global scale. Whilst impacts are well-documented for many invaders, we lack tools to predict biotic resistance and invasion success. Biotic resistance from communities may be a particularly important determinant of the success of invaders. The present study develops traditional ecological concepts to better understand and quantify biotic resistance. We quantified predation towards the highly invasive Asian tiger mosquito Aedes albopictus and a representative native mosquito Culex pipiens by three native and widespread cyclopoid copepods, using functional response and prey switching experiments. All copepods demonstrated higher magnitude type II functional responses towards the invasive prey over the analogous native prey, aligned with higher attack and maximum feeding rates. All predators exhibited significant, frequency-independent prey preferences for the invader. With these results, we developed a novel metric for biotic resistance which integrates predator numerical response proxies, revealing differential biotic resistance potential among predators. Our results are consistent with field patterns of biotic resistance and invasion success, illustrating the predictive capacity of our methods. We thus propose the further development of traditional ecological concepts, such as functional responses, numerical responses and prey switching, in the evaluation of biotic resistance and invasion success.


Results
In all experiments, 100% of control prey survived and thus experimental deaths were directly attributable to predation by copepods, which was also evidenced by partially consumed prey remaining post-experiment. In the functional response experiment, support for raw consumption models containing prey species, predator species and prey density received substantial support (Table S1a). Significantly greater numbers of invasive A. albopictus were consumed as compared to native C. pipiens overall (χ 2 = 9.14, df = 1, p = 0.003; Fig. 1). Consumption differed significantly across predator species (χ 2 = 16.11, df = 2, p < 0.001), owing to significantly greater consumption by M. fuscus than M. albidus (p < 0.001) and M. viridis (p = 0.02). However, consumption towards A. albopictus was higher for all predators given a statistically unclear 'prey species × predator species' interaction (χ 2 = 4.19, df = 2, p = 0.12). Consumption was also significantly greater under increasing prey densities (χ 2 = 73.66, df = 4, p < 0.001).
All predatory copepods exhibited type II functional responses towards both C. pipiens and A. albopictus, owing to significantly negative first order linear coefficients ( Fig. 1; Table 1). Attack rates tended to be higher towards A. albopictus than C. pipiens prey by M. albidus and M. viridis, whilst handling times were generally shorter for A. albopictus across all predator species (Table 1; Fig. 1). Maximum feeding rates were thus higher   None of the focal copepod predators exhibited a prey switching propensity between invasive A. albopictus and native C. pipiens (Table 2; Fig. 2). Instead, preferential selection towards the invader was exhibited across all proportional availabilities, as evidenced by recurrently high preference indices (a i > 0.5; Table 2). Preference index type (predicted/observed) and proportional availability were identified as important model components (Table S1c). Preference indices towards A. albopictus were significantly higher than expected under conditions of no preference (χ 2 = 16.33, df = 1, p < 0.001). However, the strength of preference towards A. albopictus interacted with its proportional availability (χ 2 = 17.10, df = 4, p = 0.002). Here, overall, preferences towards A. albopictus were particularly stronger than expected under higher proportional availabilities (0.5, 0.75, 0.9) (all p ≤ 0.01).
Owing to greater relative FRR between invasive/native prey species, higher fecundity and strong preferences towards invasive A. albopictus prey, Relative Biotic Resistance (RBR) was greater by M. albidus as compared to M. fuscus (Table 3). However, in turn, RBR scores of both M. albidus and M. fuscus were lower than M. viridis. This was driven by particularly marked relative FRR of M. viridis towards A. albopictus prey, coupled with higher levels of fecundity in this copepod (Table 3; Fig. 3). Accordingly, M. viridis is expected to exert the greatest degree of biotic resistance towards the invasive prey.

Discussion
Invasive alien species continue to spread, establish and reproduce in novel environments 1 , yet we have a distinct lack of methodologies to predict invasion success 3,4 . Biotic resistance may be a key mechanism which controls the success of invasive species. In the present study, we develop novel measures of biotic resistance driven by resident consumers towards invasive prey through the integration of functional responses, numerical response proxies and prey switching propensities. Our model species Aedes albopictus is a highly invasive vector mosquito, known to have superior competitive abilities for resources over analogous native mosquitoes, however often fails to displace these natives 24,27,29,37 . Culex pipiens is a native and widespread mosquito in our study region (United Kingdom), whilst A. albopictus has only recently been detected in the United Kingdom, following numerous successful invasions across Europe 38 . However, our results suggest that differential biotic resistance may limit the invasion success of this species, and cyclopoid copepods are known to be important mosquito predators 35,36 .  www.nature.com/scientificreports www.nature.com/scientificreports/ Indeed, given that all native copepods exhibited greater interaction strengths and strong preferences towards the invasive A. albopictus over the native C. pipiens, resident predator communities may limit levels of invader success and impact. Copepods are particularly well-adapted to thrive in ephemeral aquatic habitats which these mosquitoes colonise, owing to their ability to enter dormant life history stages and spread via zoochorous dispersal by vectors such as birds and insects, or by wind 33,34,39,40 . Whilst human-mediated copepod introductions into minute container-style habitats have been required for effective mosquito control at community-scales 35 , even in these instances, our results provide evidence for the efficacy of copepods in reducing target invader populations whilst alleviating native species from predatory impact. Nevertheless, mechanistic interpretation of our laboratory experimental results should be cautioned in the context of invasion success, with further field-based validation required that incorporates additional context-dependencies, such as emergent effects from interactions with other predator types and habitat complexities.
In the present study, irrespective of copepod species, we consistently demonstrate higher magnitude per capita ecological impacts towards invasive A. albopictus prey as compared to analogous native C. pipiens prey using a comparative functional response approach. We subsequently demonstrate clear consumptive preferences towards invasive over native mosquito prey by all focal consumers regardless of proportional prey species availability, conducive with a lack of prey switching. Then, we integrate fecundity estimations as a proxy for the numerical response to quantify and compare the potential biotic resistance of resident consumers towards this invasive prey species. Whilst drawing parallels between laboratory-based studies and field observations should viewed with caution, our results align with field patterns of coexistence, wherein invasive Aedes mosquitoes have repeatedly been shown to coexist with native competitors, despite their clear competitive advantage [23][24][25] . We propose that the combination of functional and numerical responses, alongside examinations of prey switching propensities, may help to predict the occurrence of such field patterns in relation to biotic resistance and invasion success. Although we applied our metrics to a model copepod-mosquito predator-prey system, our approaches are equally applicable to other consumer-resource systems where an invasive species suffers from biotic resistance by resident consumers (e.g. predators). Thus, our metrics can, at least theoretically, be applied across multiple habitat types and taxonomic groups for predictions of invasion success and quantifications of biotic resistance. Yet, given numerous additional context-dependencies are known to alter levels of biotic resistance (e.g. habitat complexity 6,18 ), these effects should be considered in future studies to better-reflect real systems.
Macrocyclops albidus, M. fuscus and M. viridis exhibited type II functional responses towards both prey types, characterised by high rates of mosquito prey consumption at low densities. This finding aligns with copepod functional responses forms reported in other studies 18 . However, functional response attack rates tended to be considerably higher, and handling times lower, towards invasive A. albopictus as compared to native C. pipiens. Given attack rates correspond to impacts at low prey densities whilst handling times reciprocate asymptotic maximum feeding rates, per capita predatory impacts towards A. albopictus remain higher than analogous native prey irrespective of prey density. The particularly high per capita impact of copepods towards A. albopictus also aligns with the documented ability of copepod biocontrol agents to be especially efficacious in the suppression of Aedes mosquitoes as compared to Culex 36 . Furthermore, our holistic derivations of per capita impact through coupling www.nature.com/scientificreports www.nature.com/scientificreports/ of attack rate and handling time into the functional response ratio (FRR: a/h) demonstrate greater predatory impacts by all focal copepods towards the invasive mosquito prey as compared to the native. This differential impact was particularly pronounced for M. viridis. The FRR metric has recently been developed for invasion scientists and practitioners and balances information from both key functional response parameters 41 . Invasive consumers have been shown to exhibit consistently higher FRRs as compared to native comparators across multiple study systems and taxonomic groups 41 . In a similar vein, we suggest that the FRR metric can be applied in quantifications of biotic resistance across study systems, as it can negate contradictory impact predictions for natural enemies based on one functional response parameter over the other (i.e. attack rate, handling time).
None of the copepod predators examined in the present study exhibited a prey switching propensity away from the invasive prey. That is, relative to proportional abundances, invasive A. albopictus were disproportionately selected over native C. pipiens across all availabilities. Prey switching facilitates patterns of coexistence in ecosystems through a form of frequency-dependent predation characterised by low density refuge effects 4,15 . If our laboratory-based results persist in the wild, it is likely that the sustained preferential selection towards A. albopictus would permit patterns of coexistence between these prey species even in light of the superior competitive capability of A. albopictus over analogous native mosquito species 24,[27][28][29]37,42,43 . As such, this consumptive preference may offset competitive replacement of the native by the invader. Our results exemplify the potential power of our metrics for quantifications of biotic resistance which may mediate levels of invasion success, and future work should ground-truth these concepts across other invasive species study systems using field-based observations.
Behavioural responses to predator cues are likely key drivers of such differential biotic resistance between invasive and native prey. In particular, naïveté to unfamiliar predators in novel ecosystems can influence interaction strengths and further impede invasion success 20,44 . Invasive Aedes mosquitoes have been shown to be less responsive to predation risk and exhibit higher incidences of behaviours which make them more apparent and vulnerable to predators (e.g. thrashing, browsing) 24,25 . Indeed, whilst Culex mosquitoes are filter feeders which spend most time at the water surface, Aedes are browsers which spend more time thrashing below the surface 45 . Given these substantial behavioural differences, it is plausible that the predatory patterns exhibited by copepods in the present study extend to other aquatic predator groups, owing to potentially higher encounter rates with the more motile Aedes prey. Furthermore, Aedes mosquitoes have been shown to be attracted to predatory copepods when ovipositing, whilst Culex mosquitoes are evasive of these cues 18,46 . Such behavioural factors could be major drivers in limiting the success of invasive species via biotic resistance and, in our system, may help to regulate disease risk in the context of invasive vector mosquito species.
The present study integrated estimates of fecundity as proxies for numerical responses of copepods 47 . The use of such proxies has proven robust in derivations of ecological impacts of invasive species and biocontrol agents 3,48 , and, here, high fecundity could facilitate rapid population-level responses to increases in resource availability following natural enemy inoculation. However, importantly, the present study did not compare fecundities of predators fed on the focal invasive/native prey, which may have altered estimates given differences in nutritional values between prey species. Nonetheless, whilst M. fuscus exhibited high magnitude functional responses towards invasive prey and strong selective tendencies, the fecundity of this copepod is substantially lower than both M. albidus and M. viridis 47 . Accordingly, in this study, our novel metric identified M. viridis as a particularly efficacious predator towards invasive A. albopictus prey, owing to high per capita impacts towards the invader, strong selectivity traits and relatively marked fecundity. For management, our predictive metrics suggest augmentative releases of native copepods such as M. viridis for biocontrol could be especially efficacious in the suppression of invasive mosquito species, in light of favourable consumptive and reproductive traits.
In conclusion, we propose that the integration of traditional ecological concepts that have been neglected by invasion scientists could enhance predictions of biotic resistance and help to inform invasion success. In turn, such predictions of biotic resistance directly inform management strategies for pests, vectors and invasive species via biocontrol. Biotic resistance from predators can be a key mechanism which controls invasion success, and these predator-prey interactions can be robustly quantified in controlled laboratory conditions. We show that the assimilation of functional responses, numerical response proxies and prey switching propensities enables more holistic derivations of potential biotic resistance towards invasive species at the population-level. Our results are consistent with empirical patterns, whereby the invasive mosquito A. albopictus is capable of outcompeting native mosquito species in a laboratory setting 27,37 , but has not been able to displace native mosquitoes in a similar fashion in the field 24,49 . Yet, further field-based studies are required to validate the predation patterns documented in the present study, and link them to invasion success. Nevertheless, we propose that biotic resistance is an important factor in regulating the invasion process, and can be quantified using metrics grounded in classical ecological concepts. For practitioners, use of these concepts could enable relatively rapid comparisons of biological control agents prior to release, by quantifying and comparing per capita agent effects and preferences alongside population-level responses. In turn, this could improve the efficiencies associated with natural enemy introductions. Future research should also seek to ascertain the context-dependency of these approaches in predicting the success or failure of invasions across a multitude of study systems, alongside implications for the efficacy of biocontrol agents. Moreover, quantifications of predatory efficacies across a full spectrum of life history stages would provide a more holistic account of biotic resistance levels.
Experimental design. Adult female predatory M. albidus, M. fuscus and M. viridis (respective mean total lengths excluding caudal setae ± SD: 1.70 ± 0.09 mm; 1.81 ± 0.11 mm; 1.83 ± 0.17 mm) were selected for experiments and separately starved for 24 h prior to feeding. Recently hatched, size-matched first instar A. albopictus (mean ± SD: 1.40 ± 0.12 mm) and C. pipiens (mean ± SD: 1.32 ± 0.11 mm) larvae were used as prey. Copepods are known to be most efficient in consumption of early instar mosquito prey 36 . Experiments were undertaken in 20 mL arenas of 42 mm dia. containing dechlorinated tapwater from an aerated source during daylight. We employed a phenomenological experimental approach to compare biotic resistance towards mosquitoes factorially in a replicated laboratory design. Accordingly, our design does not seek to mechanistically replicate natural systems (see 9 ). Indeed, mechanistic interpretation of such experiments must be approached with caution, or supported with further empirical parameter estimates [50][51][52] . Nevertheless, phenomenological designs, such as ours, are useful for comparative purposes in factorial experiments to examine differences in predator-prey interactions under controlled conditions 53 .
For the functional response experiment, prey species were introduced separately at each of five densities into arenas (2, 4, 7, 10 or 15; n = 5 per experimental group). For the prey switching experiment, prey species were introduced in combination at each of five ratios (2:18, 5:15, 10:10, 15:5 or 18:2; n = 3 per experimental group). Experiments were conducted in a completely randomised array to eliminate positional effects. After addition, prey were allowed to settle for 2 h prior to the beginning of the experiments via predator introduction. Once individual copepod predators were introduced, they were allowed to feed for 6 h, after which the predators were removed and remaining live larval mosquito prey counted and identified to quantify numbers killed. Controls in each experiment consisted of a replicate of each prey treatment in the absence of predators.
Statistical analyses. Data were analysed using R v 3.5.1 54 . In the functional response experiment, raw numbers of prey consumed were examined with respect to prey species, predator species and starting prey densities in a factorial generalised linear model (GLM). All interaction terms were included in the initial model. A Poisson error distribution with log link was employed. We used second-order derivations of Akaike's Information Criterion (AICc) and model averaging to identify the best-supported model using the 'MuMIn' package 55,56 . Here, all possible models were identified and ranked based on AICc (lower values indicate a better fit). Model comparisons used ∆AICc, comprising the difference between the AICc of candidate models and the best-supported model. Akaike model weights (w i ) were additionally used to probabilistically identify the best model, wherein predictor variables with good support yielded high cumulative w i values (near 1). Post-hoc Tukey tests were performed using 'lsmeans' where a factor yielded significance at the 95% confidence interval 57 .
The 'frair' package was used to perform functional response analyses 58 . Logistic regression considering the proportion of prey consumed with respect to initial prey density was used to identify functional response types. Categorically, a type II functional response is inferred where a significantly negative first-order term results, whilst a significantly positive first order term followed by a significantly negative second-order term indicates a type III functional response 59 . To account for prey depletion over the experimental period, we fit Rogers' random predator equation for the non-replacement of prey 59,60 : where N e is the number of prey eaten, N 0 is the initial density of prey, a is the attack rate, h is the handling time and T is the total experimental period. The random predator equation was fit for each predator and prey treatment group using maximum likelihood estimation, with the Lambert W function implemented to make the equation solvable 61 . Functional response models were non-parametrically bootstrapped 2000 times to generate 95% confidence intervals around starting estimations. Using the handling time (h) parameter, maximum feeding rate estimates (1/h) were additionally calculated.
We subsequently applied a new overall measure of per capita impact towards both prey types for each predator, by combining attack rates (a, functional response initial slope) and handling times (h, functional response asymptote) into the functional response ratio, which captures both parameters 41 : where FRR is the attack rate a divided by the handling time h. This solves the problem of which parameter to choose for comparisons, as a large a combined with a small h gives a large value (and hence quantifies a large per capita effect), while a low a and a high h gives a low value (and hence quantifies a low per capita effect). We denote FRR i as towards invasive A. albopictus and FRR n as towards native C. pipiens.
In the prey switching experiment, numbers of prey consumed were analysed using generalised linear mixed models (GLMM) with Poisson error distribution and log link using the 'lme4' package 62 . Here, consumption was modelled with prey species, predator species and proportion available, alongside their interactions, as fixed effects, and with a random effects structure to account for repeated measures of prey types within each experimental replicate. Model averaging based on AICc was, again, implemented to select predictors which minimised information loss 55 , and post-hoc comparisons were performed using pairwise Tukey tests 57 .
Manly's selectivity index was then used to quantify preferences for invasive A. albopictus prey by each predator species across proportions available, with adjustments for non-replacement of prey 63  where a i is Manly's selectivity index for invasive A. albopictus, n i0 is the number of the invader available at the start of the experiment, r i is the number of the invader consumed, m the number of prey types, n n0 the number of native C. pipiens available at the start of the experiment and r n is the number of native prey consumed. Resulting indices range from 0 to 1, wherein 0 indicates complete avoidance and 1 indicates complete preference. In our two-prey system, values of 0.5 are indicative of neutral selectivity by predators between prey types. Prior to formal analysis, we transformed resulting a i values to account for extreme data points (0, 1) 65 : where α t is the transformation and n is the sample size. Beta regression using the 'betareg' package was used to compare indices towards A. albopictus with those expected under null preference (0.5) with respect to predator species and prey proportion available 66 . Model averaging based on AICc was used in model selection as before, and post-hoc comparisons were undertaken using Tukey tests 55,57 .
Combining the above results, we then quantified Biotic Resistance (BR) towards invasive prey using relative FRRs between invasive A. albopictus and native C. pipiens prey (FRR i /FRR n ; Eq. 2), reproductive effort as a numerical response proxy (clutch weight produced per female body weight per day 47 ) and mean invasive prey preferences (Eq. 3) for each predator species: where the Biotic Resistance (BR) of a predator towards invasive prey is a product of the relative FRR between invasive and native prey (FRR i /FRR n ), predator numerical response proxy reproductive effort (fecundity, FE) and the mean preference index towards the invasive prey (α i ). We selected fecundity as a suitable numerical response proxy given its importance for the proliferation of natural enemies following changes in resource availability, and because reproductive effort estimates for the focal predator species were readily available in the literature. Relative Biotic Resistance (RBR) was then developed and used to compare among the three different predator species: = RBR BR1/BR2 (6) where BR1 and BR2 are Biotic Resistance for predator 1 and predator 2, respectively. Here, values of 1 indicate equivalence in biotic resistance between the two predators and values >1 indicate greater biotic resistance by predator 1 as compared to predator 2. Conversely, RBR values <1 indicate lesser biotic resistance by predator 1 compared to predator 2. We produced triplots to further illustrate differences 16 . We thus first quantified and compared functional responses by three native predators towards native and invasive prey when presented separately. Second, we examined prey preferences of the same predator species towards the two prey species when both are present simultaneously at different relative proportions. Thirdly, we used a predator numerical response proxy (fecundity), alongside functional responses and prey preferences, to predict which resident predator is likely to exert the greatest degree of biotic resistance towards the focal invasive species.

Data availability
Underlying functional response and prey switching data are available in the online supporting information.