Toxicity
To test for toxicity of our study tadpoles to potential vertebrate predators, we conducted automated behavior analysis (Amézquita et al. 2017) on mice injected with methanolic extracts of skins (Daly et al. 1997) or with saline solution, as control for the effects of manipulation and injection. Mice are a widely used model system in studies of pharmacology and toxicology (Shimizu 2004), and also specifically in studies with poison frogs (Darst 2006; Darst and Cummings 2006). We first collected 45 larvae of P. vaillanti , 24 of P. bicolor, and eight of C. tomopterna of variable size. The skins were preserved in 96% methanol until arrival to the laboratory. There, each skin was cut and processed with a PRO® 200 homogenizer during 3 min using methanol as a solvent. The resulting extract was filtered using a vacuum filtration funnel with a glass nail and a small piece of filter paper. Using a syringe to induce vacuum. The methanol was afterward evaporated at a temperature of 38° C in a JOUAN RC®1010 vacuum centrifuge for 5 h. The dry pellet was finally resuspended in 0.2 ml of saline solution.
We used 40 x 20 x 30 cm (W x D x H) glass arenas to film the behavior of each experimental or control mouse; they were adult male Mus musculus of the Carworth Farms Swiss Webster (CFW) strain. Four arenas were arranged to be adjacent and filmed from about 80 cm above with a single video camera. Filming was thus conducted in blocks where three arenas contained mice injected with skin extracts and one contained a mouse injected with saline solution. We impeded visual detection among mice by adhering opaque cardboard to the external glass surfaces of the arenas. And we ruled out a potential effect of past chemical cues on mice behavior by thoroughly washing the internal glass surfaces between two consecutive experiments. Mice were given 15 minutes acclimation period prior to the trial start.
To estimate baseline levels of mice behavior, we first filmed them during 5 min in the experimental arenas. Each mouse was then injected with either 0.2 mL of one tadpole’s skin extract resuspended in saline solution (experimental treatment) or 0.2 mL of only saline solution (control treatment). The mouse was immediately returned to the arena and filming continued for up to 30 min. At the end, mice were taken to their original cages and provided with food and water for recovery.
To quantify changes in mice behavior and to estimate a toxicity score we analyzed the video recordings on the software Ethovision XT8 (Noldus 2001) following the procedures described elsewhere (Amézquita et al. 2017). Briefly, the automated analysis extracted 68 behavioral variables describing mice mobility (three variables), directionality (three variables), body stretching (two variables), and space use (seven variables). In addition, we directly observed the videos to register the occurrence and frequency of thoracic or limb spasms (four variables). To reduce redundancy among the extracted variables, we conducted a principal component analysis (PCA), which rendered 19 orthogonal PCs that worked as our descriptors of behavior.