The relationships between toad behaviour, antipredator defences, and spatial and sexual variation in predation pressure

Study species

Epidalea calamita is a bufonid toad that thrives in diverse habitats, including unaltered as well as human-modified systems, in extensive areas in central and western Europe (Gomez-Mestre, 2014). Owing to the variable climatic conditions throughout this vast area, the phenology of this species is asynchronous, with aestivation being common in hot regions and hibernation occurring in cold climates (Gomez-Mestre, 2014). This species is primarily nocturnal, and its activity and reproduction take place during wet and not excessively cold weather, which happens during winters in warmer regions and in the spring in colder regions (Gomez-Mestre, 2014). Under adverse circumstances, they rest under rocks or logs, or in dens they burrow in loose soils, safe from predators (Gomez-Mestre, 2014). These toads are potential prey of a wide array of predators, including snakes (e.g., Natrix maura and Natrix astreptophora), birds (e.g., Larus ridibundus and Pica pica) and mammals (e.g., Meles meles), among others (see Gomez-Mestre, 2014). When under attack, toads use intermittent runs to flee (Zamora-Camacho, 2018). When escape is not possible, however, they commonly arch their loins and exhibit their parotoid glands, which can release great amounts of toxins (Stawikowski & Lüddecke, 2019).

Animal capture and management

Toads were captured in the pine grove “Pinares de Cartaya” (SW Spain: 37°20′N, 7°09′W) and in the agrosystem nearby. The forest is an 11,000-ha extension dominated by Pinus pinea and an undergrowth of Mediterranean bushes such as Pistacea lentiscus, Cistus ladanifer and Rosmarinus officinalis. Although this plant assemblage could be considered autochthonous or introduced in this region, its predominance dates back at least 4,000 years (Martínez & Montero, 2004), and as such it is deemed a natural habitat for the purposes of this study. The agrosystem is about 5 km away from the pine grove, and is a 2,800-ha agricultural area where extensive vegetable crops have gradually given way to intensive orange tree, blueberry, and strawberry fields (among others) throughout the last few decades. In these croplands, landowners apply fertilizers, fungicides, herbicides, and pesticides at their discretion, and artificial watering softens the three-to-four-month-long summer droughts. Animal capture and management was according to permits by the Junta de Andalucía government (Reference AWG/mgd GB-369-20).

Due to the mild local climate, E. calamita breeds in the winter there. Accordingly, toad capture was conducted from December 2018 to March 2019. I caught 22 females and 20 males in the agrosystem, plus 21 females and 25 males in the pine grove. Toads were captured by hand while active in nights of suitable weather, then transported to the laboratory in plastic buckets with well-ventilated lids and a substrate of humid earth. When they were in the laboratory, I used their sexual dimorphism in coloration (females have browner backs and greyish throats, whereas males have greener backs and purplish or pinkish throats; Zamora-Camacho & Comas, 2019) and the presence of blackish nuptial pads on male forelimbs (Gomez-Mestre, 2014), to sex them. Next, I allocated them to individual plastic terraria (20 × 13 × 9 cm) with wet peat as a substrate and an opaque plastic shelter. Toads were undisturbed in these terraria at all times, except during the trials. Photos were taken approximately 24 h after capture (see below). Then, 24 h after the photos, toad activity trials were recorded (see below). Finally, 24 h after the activity trials, sprint speed tests were performed (see below). Toads were released at their capture sites shortly afterward.

Measurements of coloration and morphology

I used a ruler to measure toad snout-vent length (hereafter, SVL) to the nearest mm, and a scale (model CDS-100) to weigh them to the nearest 0.01 g. No later than 24 h after capture, I orthogonally photographed each toad’s back using a photo camera Canon EOS 550D, set at 18 megapixels of resolution, F10 of shutter-aperture, and a focal length fixed at 53 mm. Only exposure time was allowed to be automatically adjusted by the device, to optimize sharpness in each individual photo. The camera was secured to a tripod, which guaranteed perpendicularity, steadiness, and a constant distance of 40 cm from the lens to the photographed area. This area was a square (30 cm side), white piece of paper that lay horizontal. On both lateral and the rear sides (considering that the tripod was located opposite to the front side), three square (30 cm side), white pieces of white polyester sat vertically, conforming an incomplete cube which lacked the front (allowing toad handling) and the upper sides (allowing photograph taking). In order to avoid all parasitic lights (i.e., any uncontrolled source of light), photos were taken at night in a completely closed room, where the only sources of light were two 80W white-light bulbs, one next to each lateral side of the cube, externally to it, at a height of 20 cm, so that shades on the photographed area were prevented and the white polyester of the lateral, vertical squares filtered the light. This setting is depicted in Fig. S1. Immediately prior to taking the photos, once the set was fixed as described, I calibrated white balance to a spotless piece of paper, after which I added a standardized colour chart (IMAGE Photographic) for digital calibration of white balance, and a piece of graph paper to calibrate length. Any remainder of humidity and dirt was gently removed from the toads’ skins with a disposable napkin before each photo.

Afterwards, these photos were processed with the software Adobe Photoshop CS5. Firstly, I calibrated white balance one more time in each photo by using the tool eyedropper in the white calibration function on the colour chart. Furthermore, colour mode was set to the L*a*b* colour space preconized by the Commision Internationale d’Eclairage (CIE) (Montgomerie, 2006). This is a three-dimensional colour space were L* quantifies lightness, and varies from 0 (pure black) to 100 (pure white); a* quantifies the green-red axis (positive values represent red and negative values represent green); and b* quantifies the blue-yellow axis (positive values represent yellow and negative values represent blue). I calibrated length using the piece of graph paper, and manually outlined both parotoid glands making use of the lasso tool, which allowed me to calculate the sum of the areas of each. After this, parotoid gland relative area was calculated as the residuals of regression of parotoid gland area against SVL. Once the parotoid glands were outlined, I calculated their average colour and, with the histogram tool, retrieved their average values of L*, a*, and b*. Lastly, I followed the same steps to trace the dorsum (excluding the parotoid glands and the limbs) and retrieve its average L*, a*, and b* values. The average L*, a*, and b* parotoid gland and dorsum values were used to calculate parotoid gland contrast (ΔE*) as in the CIE formula to assess difference in colour: ΔE* = (ΔL*² + Δa*² + Δb*²)^1/2 (Nguyen, Nol & Abraham, 2007; Moreno-Rueda et al., 2019).

Measurements of activity and sprint speed

Starting 24 h after the photos were taken, toads were recorded for activity and sprint speed trials (see details below), in this order. Videos were filmed with a camera Canon EOS 550D, at 25 frames per second. The camera was attached to a 2.5 m high tripod, with a 90° angle, at all times. In both trials, only one individual was recorded at a time. To remove the effect that temperature may have on amphibian activity (Muller, Cade & Schwarzkopf, 2018) and locomotion (Preest & Pough, 2003), the room was at approximately 19 °C at all times. Light conditions were standardized, as the only light source during all trials was a 60 W white light bulb 2.5 m high at the centre of the container where the toad was performing the trial in question (see below). All videos were recorded at night (approximately between 21:00 and 02:00, local time), when these toads are naturally active (Gomez-Mestre, 2014).

Measurement of activity

Activity trials were recorded while these toads were freely moving in a plastic arena (54 × 27 × 40 cm). A grid of 9 cm side squares was painted with non-toxic ink on the bottom of this arena. Prior to the recordings, toads were placed at the centre of the arena, enclosed within a vertical hollow cylinder (50 cm high, 15 cm diameter) open at its lower end. The cylinder was built with a metal mesh (5 mm light), which allowed acclimation to the experimental setting. After 2 min, the cylinder was gently removed in the vertical plane, and the toad’s activity was recorded for 10 min (Chajma, Kopecký & Vojar, 2020).

Videos were then analysed with the program Tracker v. 4.92. I measured several variables as surrogates of different traits of alleged relevance in the behaviour of animals in general (Réale et al., 2007) and of amphibians in particular (Kelleher, Silla & Byrne, 2018). Activity time was the amount of time (s) the toad spent moving (Chajma, Kopecký & Vojar, 2020). Exploration behaviour was estimated as the number of squares visited (excluding squares that had been visited before; Chajma, Kopecký & Vojar, 2020) and the number of square visits (counting the number of times any square was visited, including repeated visits; Carlson & Langkilde, 2013). These measures differ in the fact that the former assumes that the individual distinguishes and keeps track of the areas that have already been visited, whereas the latter assumes the opposite (Carlson & Langkilde, 2013). Time until the first move (i.e., latency) was also recorded, as a surrogate of the shyness/boldness gradient, as bolder individuals are expected to start moving sooner (Chajma, Kopecký & Vojar, 2020). The use of space is also widely considered a surrogate of the shyness/boldness gradient, in amphibians (Réale et al., 2007; Carlson & Langkilde, 2013; Chajma, Kopecký & Vojar, 2020) and other taxa (Burns, 2008; Harris, D’Eath & Healy, 2009). Specifically, thigmotaxis, the tendency for some individuals to remain in the periphery of their enclosures next to the walls rather than in the open areas, has been regarded as an anxiety-like, predator-avoidance behaviour as opposed to the boldness subjacent to the use of open areas (Harris, D’Eath & Healy, 2009; Carlson & Langkilde, 2013; Chajma, Kopecký & Vojar, 2020). Therefore, I also estimated the shyness/boldness gradient by counting independently the number of external and internal squares visited (excluding squares that had been visited before) and the number of external and internal square visits (counting the number of times squares of these types were visited, including repeated visits). Then, I divided the number of external squares visited by the total number of squares visited (external squares visited ratio), as well as the number of external square visits by the total number of square visits (external square visit ratio). Both ratios have a direct relationship with boldness. These measurements are relevant in an ecological context, as laboratory surrogates of animal behaviour mirror actual behaviour in the wild (Herborn et al., 2010).

Measurement of sprint speed

Prior to conducting and recording the sprint speed trials, I emptied toad bladders by firmly-but gently-pressing their lower abdomens, which eliminates any potential effect of different bladder water burden by reducing it to zero (Preest & Pough, 1989; Walvoord, 2003; Prates et al., 2013). Next, I allowed toads to rest in their terraria for 1 h. After that, they were recorded (with the same camera already described) while running along a brownish cardboard linear raceway (200 × 15 × 15 cm). On its bottom, one transversal white stripe of insulating tape was placed every 10 cm, so that the raceway was longitudinally divided into 10-cm stretches delimited by contrasting-colour stripes easy to visualize in the videos. Locomotor performance may depend on the substrate where the race takes place (Vanhooydonck et al., 2015): cardboard provided a surface rough enough to facilitate an appropriate traction. I also set a dark background at one end of the raceway, which could be viewed as a shelter and encourage toads’ moving forward (Zamora-Camacho, 2018; González-Morales et al., 2021). Individuals were placed at the opposite end of the raceway, and continuously pursued as a way of encouraging running, until they covered the raceway. Once these trials were completed, toads were released at their capture sites within 24 h. No visible damage was inflicted on toads because of this investigation.

The footages produced were analysed with the program Tracker v. 4.92, which allows frame-by-frame video handling. For each toad, I registered the time (to the nearest 0.01 s) needed to cover each stretch in the raceway, which equals the time elapsed between the moments when the snout of a toad surpassed two consecutive white stripes (Martín & López, 2001; Zamora-Camacho et al., 2014; Zamora-Camacho, 2018). As the distance covered was 10 cm in all cases, I calculated the speed (in cm/s) reached in each stretch by dividing 10 cm by the time (s) it took for the toad to cover it. I considered each individual’s sprint speed as its highest speed value. Finally, relative speed was calculated as the residuals of regression of sprint speed against SVL.

Statistics

Firstly, I built two correlation matrices, one including the behavioural traits measured (number of squares visited, number of square visits, external squares visited ratio, external square visits ratio, activity time, and time until the first move) and another including the antipredator defences measured (body mass, parotoid gland contrast, parotoid gland relative area, and relative sprint speed). The aim of these matrices was to detect collinearity between both sets of data. Most behavioural traits measured were highly correlated, except for time until the first move (Table S1). On the contrary, the antipredator defences measured were mostly uncorrelated, except for relative sprint speed and parotoid gland relative area, which were positively and significantly correlated (Table S2).

Then, to condense the correlated variables into fewer, uncorrelated variables, and solve the limitation caused by the high collinearity among the variables measured, I conducted a Principal Component Analysis (PCA; Jongman, Braak & Tongeren, 1995) including the behavioural traits measured that were correlated (number of squares visited, number of square visits, external squares visited ratio, external square visits ratio, and activity time; Table S3a) and another PCA including the antipredator defences measured that were correlated (parotoid gland relative area and relative sprint speed; Table S3b). In both cases, only Principal Components (PC) with an eigenvalue greater than 1 were selected, according to the Guttmann-Kaiser Criterion (Yeomans & Golder, 1982).

Then, I conducted three separate ANCOVAs, where habitat, sex and their interaction were included as factors, and all PC with an eigenvalue greater than 1 in the second PCA (namely, PC1, see Results below), as well as body mass and parotoid gland contrast, were included as covariates. The response variable of the first ANCOVA was time until the first move. In the second and the third ANCOVAs, the variable responses were each PC with an eigenvalue greater than 1 in the first PCA (namely, PCa and PCb, see Results below). Stepwise backward selection was applied to these ANCOVAs. Tests were conducted with the package “nlme” (Pinheiro et al., 2012) in the software R (R Development Core Team, 2012). Before conducting parametric statistical analyses, I checked that the data met the criteria of homoscedasticity and residual normality (Quinn & Keough, 2002). Since no transformation could make body mass homoscedastic, I implemented the function “varIdent” (Zuur et al., 2009).

Variable grouping according to PCAs

In the PCA including the behavioural traits, the only two PC with eigenvalues greater than 1, named PCa and PCb, explained jointly 88.99% of the total variance. PCa was strongly and positively correlated with the number of squares visited, the number of square visits, and activity time, whereas its correlations with the external squares visited ratio and the external square visits ratio were strong and negative (Table 1). Therefore, PCa was positively correlated with exploration behaviour, activity, and boldness. In turn, PCb was negatively correlated with all behavioural traits measured (Table 1). Therefore, PCb was negatively correlated with exploration behaviour and activity, and positively correlated with boldness.

In the PCA including parotoid gland relative area and relative sprint speed, the only PC with an eigenvalue greater than 1, named PC1, explained 68.20% of the total variance. PC1 was strongly and positively correlated with relative sprint speed and relative parotoid gland area (Table 2).

ANCOVAs

After stepwise backward selection was applied to the ANCOVA including PCa as the response variable, sex had a significant effect, with PCa being greater in females than in males (Mean ± SE; females: 0.387 ± 0.220; males: −0.349 ± 0.300; ${\chi }^2$_{1, 85} = 3.910; P = 0.048; Fig. 1). Moreover, body mass had a negative, significant relationship with PCa ( ${\chi }^2$_{1, 85} = 8.122; β = −0.045; P = 0.004).

After stepwise backward selection was applied to the two ANCOVAs including either PCb or time until the first move as response variables, no factor or covariate appeared significant in either case.