Introduction

Chemical cues are considered the main sensory modality in both freshwater (Tollrian & Harvell, 1999; Brönmark & Hansson, 2012) and marine (Hay, 2009) environments, affecting individual morphology, behavior (Kats & Dill, 1998) and metabolic responses (Barry & Syal, 2013), population-level processes (Takken & Dicke, 2006), community structure and ecosystem functions (Hay & Kubanek, 2002; Pohnert et al., 2007). Water-borne predator (kairomones) and predation (pre- and post-consumption) cues can inform potential prey species on the occurrence of cryptic or non-visible potential predators and, in many species, have been demonstrated to allow the accurate assessment of predation risk (Lima & Dill, 1990). To recognize a potential predator, prey species associate their peculiar odor, probably a mix of volatile molecules produced by the predator itself and its environment, with previous threat-related experiences. This recognition occurs either through learning (ontogeny) or adaptation (innate recognition) (Wyatt, 2014). The ability of prey to discriminate cues and assessing risk is pivotal for survival. Failing to respond to a predator may lead to death, whereas responding to non-predatory species can incur costs detrimental to an individual’s fitness, such as impaired feeding, growth, and development (Kavaliers & Choleris, 2001).

Lack of recognition and effective defensive responses are expected when predators are voluntary or accidentally introduced into novel habitats (Callaway & Aschehoug, 2000; Shea & Chesson, 2002): in the absence of any prior experience or shared evolutionary history, native prey often fail to detect alien predators. Consequently, the impact of these predators on prey populations is typically more severe than that of native predators, potentially leading to decline and extinction of native species (Vitousek et al., 1997; Salo et al., 2007; Arribas et al., 2014). The greatest loss of biodiversity due to introduced species is likely occurring in freshwater ecosystems (Olsen et al., 1991; Kolar & Lodge, 2000), which, being isolated and “island-like”, are more sensitive to variations in community composition than terrestrial systems (Cox & Lima, 2006). Introductions are implicated in the general decline of amphibians (Kiesecker et al., 2003), and the impact of alien predators has been shown to cause the local extinction of amphibians with complex life cycles, aquatic eggs and larvae being particularly vulnerable to alien aquatic predators (Kats & Ferrer, 2003).

Many studies have demonstrated that amphibian larvae respond to the presence of chemical stimuli from various predators by varying either their morphology or behavior as to lower predation risk (Lima & Dill, 1990; Kats & Dill, 1998; Hossie & Murray, 2012; Maher et al., 2013). However, introduced species are often not recognized as potential predators, due to the absence of a shared evolutionary history (Kiesecker & Blaustein, 1997; Marquis et al., 2004; Polo-Cavia et al., 2010; Hettyey et al., 2015). In this study, we investigated the defensive responses of anuran larvae from three different brown frog species (Rana dalmatina Bonaparte, 1838, Rana latastei Boulenger, 1879 and Rana temporaria Linnaeus, 1758) to the chemical cues of two widespread alien predator species: the red swamp crayfish (Procambarus clarkii Girard, 1852) and the pond slider (Trachemys scripta Schoepff, 1792). The red swamp crayfish is native to north-eastern Mexico and southern USA. It has been introduced worldwide, except for Antarctica and Oceania (Loureiro et al., 2015). In Italy, this species was introduced in 1977 and first recorded in the wild in 1989 in the River Banna, a tributary of the River Po, after the escape of a few individuals from an experimental farm (Delmastro, 1992). Currently, it occurs with extremely abundant populations in a wide range of lentic, lotic, and brackish environments (Fea et al., 2006). The red swamp crayfish have been reported to affect the abundance and distribution of many amphibians (Cruz & Rebelo, 2005; Cruz et al., 2008; Lo Parrino et al., 2020), including Rana latastei (Manenti et al., 2020).

The pond slider is native to the United States, from southeastern Virginia to northern Florida (Ernst & Lovich, 2009). This turtle species has been widely introduced, mostly because of pet trade, in every continent except Antarctica (Kraus, 2009). Hundreds of thousands of individuals of the subspecies T. scripta elegans have been imported to Europe (Scalera 2009; Luiselli et al., 2016), with breeding populations established in many countries, including Italy (Ficetola et al., 2002, 2012), where the first reports date back to1972 (Ferri & Soccini, 2000). Pond sliders are opportunistic omnivores, and their diet includes the eggs, larvae and adults of several amphibians (Ernst & Lovich, 2009). Currently both alien species are widespread in northern Italy: The red swamp crayfish occurs in 93.6% of the 45 provinces (Lo Parrino et al., 2020); while, the pond slider is particularly widespread in Lombardy, where its range overlaps almost entirely with those of the two agile frogs (Macchi et al., 2020).

All the three brown frog species occur in northern Italy: The Italian agile frog (Rana latastei) is an endemic species, occurring in the residual lowland woods of the River Po plain; while, the agile frog (Rana dalmatina) is widespread, occurring also in cultivated and hilly areas; the common frog (Rana temporaria) mainly occurs in the Alps and Prealps (Sindaco et al., 2006). As a result of each species’ habitat preferences, the range overlap between Rana dalmatina and the other two frogs lies between 20.2% (Rana latastei) and 27.2% (Rana temporaria), while the latter two share only 7% of their overall range (as assessed from data in Sindaco et al., 2006). Their tadpoles have been reported to show behavioral defensive responses when exposed to the chemical cues of native predators (e.g.: Van Buskirk, 2001; Gazzola et al., 2018a, 2018b, 2021). To test whether the three Rana species assess predation risk by native and exotic predators differently, we also exposed tadpoles to the olfactory cues of the aquatic larvae of a widespread native predator, the Southern hawker (Aeshna cyanea Muller, 1764). Based on previous research, we expected all tadpoles to reduce their activity levels when exposed to native predator’s cues and exhibit negligible or weak responses to alien crayfish and turtles.

Methods

Collection and rearing of animals

In March–April 2023, we collected aliquots of approximately 50 eggs from six freshly laid egg clutches of each frog species. The collection site for Rana latastei was a small fragment of floodplain forest including an elliptical oxbow of the River Ticino and many small wet areas connected by narrow canals (Integral Natural Reserve “Bosco Siro Negri”, 45°21′ N, 9°06′ E). Rana dalmatina eggs were collected from a pond located inside a protected area (“Bosco del Vignolo”, 45°13′ N, 8°56′ E), while those of Rana temporaria were collected from a pre-alpine lake within the “Campo dei Fiori” Regional Natural Park (“Laghetto della Motta d’Oro”, 45°84′ N, 8°73′ E, 470 m a.s.l.). All sampling sites are in Lombardy, northern Italy. Both alien predators have been recorded in the first two sites, while none was reported in the pre-alpine lake.

The eggs were immediately transported to the laboratory and individually placed in 20 l tanks filled with 8 l of aged tap water. Upon hatching, the tadpoles were transferred to 50 L tanks (one per clutch). These were kept in an unheated room under natural light conditions and fed ad libitum with rabbit chow. Water temperature ranged between 17 and 22 °C throughout the study period, and approximately 50% of the water was changed every two days. Late instar Aeshna cyanea larvae (n = 15; total length = 35–41 mm) were collected from an artificial pond in the Botanical Garden of Pavia (Northern Italy, 45°11′N, 9°10′E), using dip nets. The predators were individually housed in 500 ml plastic tubs filled with 250 ml of aged tap water. Each tub was equipped with a small net to serve as perching site.

Two adults Trachemys scripta (carapace length = 185–193 mm) were captured by dip netting from a pond within the University Campus of Pavia and then housed individually in plastic containers (60 × 40 × 30 cm) filled with 10 l of aged tap water. Additionally, six adult red swamp crayfish (total length = 97–113 mm) were collected using dip nets and accommodated in three plastic boxes (two individuals per box), each filled with 2 l of aged tap water. Brick fragments were provided in each box as shelters. All predators were fed with gammarids. The use of different amounts of water for each individual predator aimed to control for size variation among predator species when collecting chemical cues (see below). The experiment was performed when tadpoles reached Gosner’s developmental stage 27–29 (Gosner, 1960).

Preparation of predators’ olfactory cues

The experiment involved four olfactory treatments: (1) aged tap water (serving as the control); (2) cues (kairomones) from fasting dragonfly larvae; (3) fasting pond sliders; and (4) fasting red swamp crayfish. To collect dragonfly cues, 100 ml of water was gathered each test day from five randomly chosen rearing tubs and combined in a separate container. For turtle cues, 250 ml of water were extracted from each container just before the experiment and mixed. The same protocol was applied to collect crayfish cues, taking 150 ml of water from each container. The day before collecting the olfactory cues, water was totally changed in each predator container, and no food was provided, ensuring a 24-h fasting period (Polo-Cavia et al., 2010).

Experimental procedure

To assess tadpole activity levels before and after the introduction of predatory cues, each tadpole was placed in an opaque plastic tub (15 × 10.5 × 5 cm) containing 250 ml of aged tap water, where it was allowed to acclimate for 15 min. Each trial involved a 10-min pre-stimulus period (prior to infusion) and a 10-min post-stimulus period (following infusion). Considering the time required to administer the stimulus, the total duration of each trial was approximately 36 min. To minimize disturbance, the stimulus (2 ml) was carefully injected using an 8 ml disposable syringe. The concentration of the odorous stimulus during the trials was 1:125, as for previous studies (e.g., Gomez-Mestre & Diaz-Paniagua, 2011; Gazzola et al., 2018a, b; Scribano et al., 2020). Tadpoles were exposed to cues 1–4 h after their collection from predator tubs. Given that cues have been documented to elicit strong behavioral responses after 36–48 h of aging in well water (Peacor, 2006; Van Buskirk et al., 2014), we were confident in their effectiveness.

For each testing session (trial), a grid of 12 tubs was arranged, enabling the simultaneous observation and recording of as many individuals (Scribano et al., 2020; Guadin et al., 2021). All trials were conducted indoors, and video recorded by a digital camera (Canon Legria, 1080-pixel resolution, 25 frames per second) suspended 1 m above the setup. The arena was shielded using opaque panels and uniformly illuminated by spotlights. Operators were present in the room solely for the duration of the stimulus injection. We tested 18 tadpoles (three from each clutch) for each stimulus (n = 4) and anuran species (n = 3), for a total of 216 recorded individuals. To take account of intra-population genetic diversity, three tadpoles from each clutch were assigned to each level of both factors (stimulus and species). Each tadpole was tested only once. All trials were carried out over three days, between 10 am and 1 pm.

All videos were analyzed using ToxTrac, a source executable software designed for image-based tracking (Rodriguez et al., 2018). This software provided detailed locomotor information by recording the x and y coordinates of the central point of each tadpole at intervals of 0.04 s (Fig. 1). To prevent the risk of bias, a blind approach was employed; the operator analyzing the recordings was unaware of the specific chemical stimulus received by each tadpole within the experimental arena.

Fig. 1
figure 1

Trajectories followed by an agile frog tadpole before (pre-stimulus) and after (post-stimulus) exposure to either turtle cues or water (control)

Full size image

Statistical analysis of behavioral data

To assess tadpole defensive behavior, we evaluated two behavioral responses: (i) the proportion of time tadpoles were active during the 10 min following cue injection (“activity level”; Van Buskirk & McCollum, 2000; Altwegg, 2003; Gazzola et al. 2022), and (ii) the total number of freezing events. Tadpoles were considered as “frozen” when they moved less than 5 mm in 3 s. For analyzing behavioral variables, we applied Bayesian Generalized Linear Mixed Models (BGLMMs). Priors were determined using the get_prior function from the brms package (Bürkner, 2017) and incorporated into the models. The same package was also used to run the models (5000 iterations across four chains, of which 1000 as warm-up). Model convergence was visually inspected using trace plots, and inferences were based on posterior means and their associated 95% credibility intervals (CrIs). Each model included, as fixed effects, the behavioral variable (either the proportion of time active or the number of freezing events) recorded during the 10 min before stimulus injection (pre-stimulus behavior). Additional fixed effects were the type of stimulus (a 4-level factor), anuran species (a 3-level factor), and their interaction (stimulus × species). The trial was treated as a random factor. A beta family distribution with a logit link function was used for the activity level, while freezing events were analyzed using a negative binomial distribution and log link function. The interaction between stimulus and species was explored using Watanabe–Akaike Information Criterion (WAIC) (Gelman et al., 2014), a Bayesian extension of the Akaike Information Criterion (AIC). The models incorporating the interaction performed significantly better (ΔWAIC =  − 79.3 and − 23.8 for activity level and freezing events respectively) than those without. A model was also run to explore the activity levels of tadpoles during the pre-stimulus period, including the species as the sole predictor and trial as a random effect, and maintaining the same conditions as the previously mentioned model for activity level (chain, iterations, family, and link function). We considered the estimates of any effect and pairwise comparisons between treatments to be significant if their 95% CrIs did not include zero. For post hoc analysis, we used the emmeans function of the same package (Lenth, 2023). The estimates of posterior means and emmeans contrasts are presented along with their 95% CrIs (Highest Posterior Density intervals, HPD).

Results

Pre-stimulus activity levels varied among the anuran species (Fig. 2). Rana temporaria (RT) tadpoles were more active than both Rana dalmatina (RD; estimated difference: − 0.43, CI − 0.48 to − 0.37) and Rana latastei (RL; − 0.32, CI − 0.37 to − 0.26) tadpoles. Additionally, also RL and RD differed in their levels of activity (− 0.113, CI − 0.16 to − 0.05).

Fig. 2
figure 2

Predicted posterior means from the Beta BGLMM for the proportion of time active recorded during the pre-stimulus period. Black thick lines show 66 and 95% credibility intervals (n = 216)

Full size image

All anuran species showed a clear defensive response to odonate predator cues, reducing the proportion of time spent active compared to controls (Fig. 3). The responses of RD and RL followed a similar pattern, with the strongest reduction in activity being elicited by turtle cues and the weakest, although still significant, by crayfish. In contrast, odonate cues were perceived as the most dangerous by RT; while, turtle cues elicited the weakest response, resulting from a few tadpoles significantly lowering their activity level respect to controls (− 0.13, CI − 0.22 to − 0.03; Fig. 3). Pairwise post hoc tests confirmed the differences for all comparisons, except for the response of RD to odonate vs turtle cues (Table 1).

Fig. 3
figure 3

Predicted posterior means (left) and effects (right), from the Beta BGLMM for the proportion of time active recorded after stimulus exposure. Black thick lines show 66 and 95% credibility intervals. The right plot shows the estimated comparisons of each olfactory cue with the control group (n = 216). Thick lines that overlap with dashed lines indicate no significant difference

Full size image
Table 1 Post hoc, Tukey comparisons, within each anuran species, for the proportion of time active (Beta BGLMM) recorded after exposure to alien or native predators’ cues
Full size table

The native predator odor highly lowered the total number of freezing events for all anuran species with respect to controls (Fig. 4). Turtle-elicited responses were similar to those induced by odonates in both RD and RL; while, they did not differ significantly from controls in RT (Fig. 4). Crayfish cues did not affect the number of freezing events in any frog species (Fig. 4). Post hoc tests showed no significant difference between the responses of both agile frogs toward odonate and turtle cues; while, RT showed a significant higher number of freezing events when exposed to turtle cues (Table 2). Finally, crayfish cues elicited fewer freezing events than the other predators’ in all anuran species, except for turtle cues in RT.

Fig. 4
figure 4

Predicted posterior means (left) and effects (right), from the binomial BGLMM, for the number of freezing events recorded after stimulus exposure. Black thick lines show 66 and 95% credibility intervals. The right plot shows the estimated comparisons of each olfactory cue with the control group (n = 216). Thick lines that overlap with dashed lines indicate no significant difference

Full size image
Table 2 Post hoc, Tukey comparisons, within each anuran species, for the number of freezing events (Binomial BGLMM) recorded after exposure to alien or native predators’ cues
Full size table

Discussion

Since the 1950s, the rate of species introductions on a global scale has increased exponentially (Seebens et al., 2017), and the number of invasive alien species is expected to double in the next decade (Mormul et al., 2022). In this scenario, understanding the potential impact of alien predators on native assemblages is crucial for conservation. Despite its importance, there is currently limited knowledge about the ability of aquatic anuran larvae, including endemic and threatened species, to detect and cope with unfamiliar predators. We investigated the behavioral responses of three sympatric, closely related Rana species. Two of these species, Rana dalmatina and Rana latastei, coexist in the residual forested areas of the intensively cultivated Po River plain, while the third species, Rana temporaria, is widespread in hilly and mountainous regions.

The behavioral responses of the two agile frog species were strikingly similar. Both species showed a relatively weak response to the kairomones of the red swamp crayfish, aligning with previous findings that suggest such kairomones only trigger effective defensive behaviors when combined with the alarm cues of conspecific tadpoles (Gazzola et al., 2018a, b, 2021). In contrast, the strong response toward the pond slider, as high as that elicited by the cue of native dragonfly larvae, was unexpected, as four species belonging to as many anuran families (Ranidae, Pelobatidae, Bufonidae and Hylidae) had been previously reported to ignore its cues (Polo Cavia et al., 2010). Differently, crayfish elicited a clear response in the common frog, although less prominent than that toward dragonfly larvae; while, the response to turtle cues was negligible.

To date, the impact of alien predators’ cues on the behavior of common frogs has not been extensively studied and the existing research has yielded conflicting results. Marquis et al. (2004) observed no behavioral changes in tadpoles exposed to the recently introduced Turkish crayfish (Astacus leptodactylus Eschscholtz, 1823). Conversely, recent studies in the Czech Republic have demonstrated that the presence of alien pond sliders not only affects the behavior of the common frog but also influences its life history traits (Berec et al., 2016; Vodražkova et al., 2022). Although the occurrence of native and alien predators in freshwater systems can vary at both locally and regionally, providing spatial heterogeneity in predation risk and affecting the chances for either co-evolution with archetypes (sensu Cox & Lima, 2006) or coexistence with aliens (since their introduction into local communities), our results suggest that both the “similarity” of introduced and native predators (‘cue similarity’ hypothesis, Sih et al., 2010) and “time from invasion” may play a role in shaping the two observed patterns (agile frogs vs. common frog).

The native, white-clawed crayfish (Austropotamobius pallipes Lereboullet, 1858) and the European pond turtle (Emys orbicularis Linnaeus, 1758) are likely candidates for the role of predator archetypes, that is taxonomically related, native predators that may share both predatory adaptations and chemical signals with alien species and with which tadpoles could have co-evolved (Cox & Lima, 2006). Despite their severely declining numbers (Füreder et al., 2010), native crayfish are primarily found in the hilly, wooded habitats of the Pre-Alps, largely overlapping with the common frog’s range. Meanwhile, the European pond turtle, currently listed as endangered by IUCN (Rondinini et al., 2022), was once widespread in the Po plain, where it co-evolved with the two agile frog species. Secondly, since the mid-1980s, pond sliders have established themselves in peri-urban and agricultural areas of northern Italy, predominantly inhabiting the intensively cultivated lowlands (Sindaco et al., 2006); while, the red swamp crayfish began spreading in the same region since the beginning of this century (Lo Parrino et al., 2020). Given the absence of sufficiently similar pre-existing native predators, as reported by previous studies on Perez’s frog (Pelophylax perezi López-Seoane, 1885) (Gomez-Mestre & Díaz-Paniagua, 2011; Nunes et al., 2014), the time from invasion of the red swamp crayfish may be too short for its cue eliciting defensive behaviors in agile frogs. The opposite can be expected for the common frog, the previous occurrence of native crayfish having been claimed to enhance the co-option of alien crayfish cues (Pujol-Buxó et al., 2013; Polo-Cavia & Gomez-Mestre, 2014). In turn, long-lasting coexistence with either native turtles or pond sliders may explain the strength of agile frogs’ response to predation threat, in agreement with the ‘cue similarity’ hypothesis (Sih et al., 2010). Co-existence in evolutionary times may also explain common frog behavior in the Czech Republic, where Emys orbicularis was widespread throughout the Holocene (its last record dates back to 2000; Široký et al., 2004).

We acknowledge that, dealing with pond-breeding species, range overlap at regional level is only a rough approximation of co-occurrence, and that long-term information on the distribution and abundance of all relevant species would be necessary to better understand predator–prey relationships. Unfortunately, such data are rarely available (but see Ficetola et al., 2018). Notwithstanding, alien species have been recorded to affect the meta-population dynamics of their prey even in predator-free patches (Trekels & Vanschoenwinkel, 2019; Manenti et al., 2020), supporting the hypothesis that “sympatry” with predators or their archetypes may have shaped the defensive responses of Rana frogs.

Regarding the slightest responses, specifically those elicited by red swamp crayfish in agile frogs and pond sliders in the common frog, they might be attributed to a “precautionary behavior” in response to a generic unknown chemical. This behavior is typically adopted by less bold tadpoles, as indicated by the larger variation in response strength compared to reactions to odonate cues. In the case of both agile frogs, the observed decrease in the number of freezing events indicates that the average time spent motionless increased when tadpoles were exposed to odonate and turtle cues. As both predators rely on ambush tactics and are dependent on prey movement (Corbet, 1980; Alcott et al., 2020), freezing can be supposed as an effective strategy to prevent detection and attacks. The same strategy was not adopted by common frogs when threatened by crayfish, which actively explore the substrate and move toward odor sources detected by their maxillipeds (Kreider and Watts, 1998), making freezing a less effective defense.

While available information on the response of anuran larvae to alien predators does not allow to draw general conclusions, local and evolutionary historical factors affecting interactions at community level, the role played by co-evolution in shaping current cue–response patterns asks for the application of a phylogenetic approach to test whether clades are more prone to show similar defensive behaviors. This approach has been rarely used, but phylogeny has been reported to shape the response of Gryllid crickets to cues of predator presence (Dalos et al., 2022). Agile frogs Rana dalmatina and Rana latastei diverged from the “Rana temporaria group” in the early Pliocene (Veith et al., 2003); thus, phylogenetic constraints might have contributed to shape the two distinct behavioral patterns observed, a hypothesis that warrants further investigation. Moreover, while agile frogs occur almost exclusively in lowlands, the Italian range of the common frog covers a wide altitudinal gradient, up to 2760 m above sea level, where predation risk, both by native and alien predators, can be predicted to be lower (Laurila et al., 2008). Hence common frogs offer the opportunity of testing intra-specific variation in defensive responses along an environmental gradient.