We collected about 200 eggs from four I. nebulifer egg strands (800 eggs total) in San Marcos, TX (long: −97.94285, lat: 29.90115) on 26 March 2020 (one strand) and 7 April 2020 (three strands). All eggs were at the same stage of development when collected and the site was a natural area and considered to have little anthropogenic alterations. After collection, we transported the eggs to our lab at Texas State University and we divided each egg strand among three 5.7 L polypropylene containers filled with 3 L aged tap water (12 containers total). Eggs hatched 1–2 days after collection and were free-swimming after 1 week. Once the tadpoles were free swimming, they were fed a food pellet consisting of a mixture of spirulina powder and Tetramin fish flakes in an agar base ad libitum. After the first feeding, we took 30 tadpoles from each container that were seen feeding (360 total) and mixed them in a single tank to be used for the experiment.

We randomly divided tadpoles among 56 containers (six tadpoles per container) and each container was exposed to one of four treatments: (1) 3 L of 23°C water (i.e., low urbanization alterations), (2) 1.5 L of 23°C water (i.e., moderate urbanization alterations), (3) 3 L of 32°C water (i.e., moderate urbanization alterations), and (4) 1.5 L of 32°C water (i.e., high urbanization alterations). We used four strands to increase the genetic diversity of our sample but differences among egg strands were not the focus of this study so strand ID was not maintained. Temperatures were selected to be ecologically relevant and close to developmental thermal tolerances. The 23°C water temperature was based on the average water temperature measured at the site of collection during the time of collection (March 25, 26, 27, and April 7, 2020) which was 21.7°C. The 32°C water temperature was based on the average water temperature measurements taken toward the end of the rainy season when surface water is present but water levels are low (June 28, 29, July 5, and 6) which was 34.2°C. Developmental thermal tolerances for I. nebulifer are between a low of 18–22°C and a high of 34–38°C (Ballinger and McKinney, 1966). We chose to have two fixed water levels rather than a fixed and a reducing water level to focus on the environmental stimulus of water level rather than the stimulus of drying. We heated 32°C water treatments with 100 W submersible All-Aquarium heaters and supplied all containers with a foam float for toadlets to climb out on after metamorphosis. We maintained all containers on a 12:12 light/dark cycle and performed partial water changes bi-weekly. Containers were randomly shuffled among spaces weekly to prevent any shelf effects.

Once metamorphosed, we kept toadlets individually in 500 mL deli cups with perforated lids, enough water to cover part of the bottom of the cup, and a shelter made from a cardboard egg crate. Five days post metamorphosis we fed toadlets seven wingless fruit flies (Drosophila melanogaster; determined by preliminary feeding experiments). Toadlets that ate were weighed and photographed for total length measurements and then randomly divided into two treatments: (1) RIFA exposure and (2) control groups. Two toadlets from each tank were used in the RIFA and control treatments. However, mortality reduced some tanks to one or zero toadlets (final n = 150). Toadlets were fed seven wingless fruit flies every other day for the duration of the experiment. We refreshed the water and recorded the number of flies eaten every other day. Cardboard hides were replaced bi-weekly or as needed. RIFA exposure consisted of removing the water and hide from the deli cup and introducing three fire ants for 10 min. After the 10 min exposure, we removed RIFA with a gloved hand and forceps, added clean water, replaced the hide, and fed the toadlets. RIFA were not prevented from attacking toadlets but the number of fire ants presented could not subdue the toadlets and no mortality as a direct result of the RIFA exposure was observed. The control group had the same treatment but without the introduction of RIFA (water and hide removed for 10 min and gloved hand and forceps placed in the deli cup as was done to remove RIFA). Wingless fruit flies and RIFA were sedated using CO₂ for easy handling but were allowed time to recover before use. Toadlet treatments lasted 30 days from treatment assignment.

After 30 days of treatment exposure, we measured water-borne corticosterone profiles: baseline, agitation, and recovery release rates following an established protocol (Gabor et al., 2013; Forsburg et al., 2019) with a minor modification for toadlets. Briefly, we placed toadlets in glass petri dishes with lids and 15 mL spring water (enough to cover the bottom 1/3 of each toadlet) for water-borne hormone collection. Separate petri dishes were used for baseline, agitation, and recovery and toadlets were transferred with a gloved hand. Toadlets were agitated by placing petri dishes on a supply cart and gently swirling and rocking the supply cart for 1 min every 3 min for 60 min. Water was collected from petri dishes into sterile falcon tubes and stored at −20°C until processing. Hormone collection was done in the same room the treatments were performed where the temperature was being maintained at 23°C.

Hormones were extracted from the water using C18 solid-phase extraction columns (SepPak Vac3 cc/500 mg; Waters, Inc., Milford, MA, United States) under vacuum pressure (following Forsburg et al., 2019). We primed the columns with 4 mL methanol and 4 mL distilled water. After extraction, we eluted columns with 4 mL methanol into borosilicate vials. The methanol was evaporated under a gentle stream of nitrogen gas while in a 37°C water bath. We re-suspended the residue in 500 μL of 5% ethanol diluted with enzyme-immunoassay (EIA) buffer (provided by Cayman Chemicals Inc., Ann Arbor, MI, United States). After resuspension, we shook samples for 2 h to fully resuspend the residue. The resuspension volume ensured that sample values were within the assay range of the EIA kits. Corticosterone release rates were measured in duplicate for all samples with an EIA kit (Cayman Chemicals Inc.) on a microplate absorbance reader (BioTek ELX800) set to 405 nm. Inter-plate coefficient of variation was 12.50% (11 plates; range: 1.01–12.03%).

To quantify hop performance, toadlets were placed on a flat surface covered with damp white paper towels to help maintain body moisture, provide a textured surface, and give a contrasting background for video recording. A video camera was secured to a tripod to provide an overhead view of hop performance trials. Toadlets were placed under a 118 mL cup for a 5 min acclimation period. After the acclimation period, the cup was lifted and recording of the trial started. If the toadlet did not hop, a gentle touch of the hind legs was used to encourage hopping. The first three sequential hops that were uninterrupted and in the forward direction (toadlet was not hopping sideways) were analyzed. To quantify hop performance, we measured the three hops using Kinovea¹ (Puig-Diví et al., 2019). Hop distance was determined by marking the location of the urostyle at the start and end of the hop and then measuring the distance between the two locations. The longest of the three hops was used as our maximum hop performance and the average of the three hops was used for average hop performance (Moreno-Rueda et al., 2020). Both measures were interpreted as measures of predator avoidance response. Hop performance trials were performed in the same room the toadlets were housed where the temperature was maintained at 23°C. After the hop performance trials, toadlets were euthanized in an ice bath to prevent contamination of bufadienolides. After euthanasia, we weighed toadlets and photographed their dorsal side to measure snout-urostyle length (SUL) using imageJ (version 1.51w). Maximum and average hop performance were SUL corrected before analysis. Euthanized toadlets were placed in 2 mL microcentrifuge tubes with 1 mL HPLC-grade methanol and stored at room temperature for preservation and bufadienolide extraction until analysis.

After preservation in HPLC-grade methanol (utilized throughout the bufadienolide analysis), we removed toadlets and evaporated the methanol under a gentle stream of nitrogen gas while in a 37°C water bath for shipping. Samples were then reconstituted in 0.7 mL of methanol and vortexed for 45 s to solubilize the rinse. Solubilized rinses were subsequently filtered through 13 mm syringe filters (0.2 μm PTFE membrane) into clean, labeled centrifuge tubes. Another 0.7 mL of methanol was added to the syringe and filtered to remove any residual extract adhered to the membrane. We then evaporated the filtered extracts in vacuo (without heat) on a DNA SpeedVac. After drying, the extracts were resuspended in 50 μL of methanol and centrifuged. The supernatants were placed into vials for LC–MS/MS analysis.

We purchased eight bufadienolide standards (telocinobufagin, arenobufagin, cinobufotalin, cinobufagin, bufalin, marinobufagenin, resibufogenin, and gamabufotalin) from Cayman Chemicals Inc. and suspended them in methanol. Bufadienolide standards and reconstituted toadlet filtrates were then analyzed on an Exion HPLC equipped with a PDA detector and Sciex TripleTOF 5,600+ mass spectrometer in ESI positive mode with a calibrant delivery system. All samples were analyzed via reverse-phase HPLC utilizing an Agilent C18 column (4.6 × 150 mm; 3.5 μm particle size). The mobile phase consisted of two solvents: acidified water (0.1% formic acid, v/v; Solvent A) and acidified methanol (0.1% formic acid, v/v; Solvent B). We separated the extracts using the following gradient developed for the separation of the eight bufadienolides: 10% solvent B held for 2 min then stepped to 55% solvent B. The gradient was then increased from 55–65% solvent B in 5 min. Next, the gradient was increased from 65–74% solvent B over 9 min then stepped to 100% solvent B for 7 min. Finally, solvent B was stepped to initial conditions for 6 min to re-equilibrate the column. The separation was performed at a flow rate of 1.00 mL/min with the LC column oven set to 40°C. A 30 μL sample of each reconstituted toadlet rinse was injected on-column.

Within the TOFMS method developed for bufadienolide relative quantification, the curtain gas was set to 30 PSI with source gases 1 & 2 set to 60 PSI. Additionally, a source temperature of 550°C was utilized with a needle spray voltage of 5,500 V. The declustering potential and collision energy were set to 70 V and 10 V, respectively. The collision energy was kept at 10 V throughout the TOFMS scan (50–1,000 Da) to attempt to retain the molecular ion of bufadienolides for mass accuracy calculations. The mass spectrometer was calibrated every 5 samples, utilizing ESI positive calibration solution and the calibrant delivery system.

For the relative quantitation of the seven bufadienolides and absolute quantitation of bufalin, we prepared a serial dilution of the bufalin standard to generate a series of standard curves via LC–MS/MS. Due to structural similarities in the backbone of bufalin compared to the other bufadienolides being quantified (Figure 1), we assumed these compounds have relatively equivalent ionization allowing for their quantitation based on bufalin measurements. As a result of bufadienolide standards’ high cost and sporadic availability, we were unable to make standard curves for each bufadienolide. Bufalin standards ranging from 0.00819 μg/mL to 6.55 μg/mL were analyzed via LC–MS/MS (Supplementary Table S1) from low to high concentration with blanks between each standard to generate standard curves for bufalin. Three full bufalin standard curves were analyzed on three separate days. To account for concentration ranges in which instrument response was non-linear, a weighted linear regression was utilized. Two of the three calibration curves were split for linear equations to be generated for high and low concentrations (high spine/low spine linear equations; Supplementary Table S2). Sample concentrations calculated from the curve generated closest to when samples were analyzed to properly control for instrument variability (Supplementary Tables S3, S4). Four system suitability bufalin standards of 0.0164 μg/mL, 0.328 μg/mL, 1.64 μg/mL, and 6.55 μg/mL were analyzed prior, mid, and post sample batch acquisition to assess and account for instrument variability throughout the duration samples were analyzed.

We used PeakView™ software for the evaluation of data. The protonated exact mass of the eight bufadienolide standards was extracted from the total ion chromatogram (TIC) for peak area quantitation. The mass accuracy of the measured versus actual bufadienolide protonated exact masses were calculated using the formula:

Compounds with a mass accuracy less than 5 ppm were considered identical to bufadienolide standards. For secondary confirmation of the identification of the bufadienolide, retention time was utilized.

After removal of toadlet remains from the methanol for bufadienolide analysis, we dried the toadlets at 60°C until they were a constant weight (36–48 h). We then placed the toadlet remains in a lipid free filter paper (Whatman™ Grade 602H) envelope made by cutting the filter paper into a 4 × 6 cm rectangle. The short sides of the rectangle were pre-folded into the middle and the long sides of the rectangle were pre-folded into thirds to create guidelines for sample placement. Toadlet remains were then placed in the 2 × 2 cm center square made by the pre-folding guidelines. We repeated the folding, the short sides folded followed by the long sides, and stapled the end of the resulting 2 × 2 cm square envelope to seal it. Preliminary trials showed that staples did not change weight during the extraction process and encased remains were sufficiently retained. Because of the small size of the toadlet remains, we used the intact toadlets to prevent any loss of material that could occur if homogenized.

We labeled the filter paper envelopes using a graphite pencil and placed the filter paper envelopes into a Soxhlet extraction setup. For efficiency, we processed samples in batches with ~12 samples per setup. We used petroleum ether as our extraction solvent and heated it to 45°C. This resulted in a cycle time of about 15 min. We extracted the samples for 6 h, allowed the remaining petroleum ether to evaporate off the samples, and placed the samples in a 60°C dryer for 1 h.

We weighed the toadlet remains after initial drying (S), after being placed in the filter paper envelope, and after the final drying (T). We determined the total lipid content as the percent of total body weight lost during the extraction process and used the following formula:

Where FP is the weight of the filter paper determined by subtracting the initial dried weight from the weight of the sample in the filter paper envelope.

As additional measurements of energy allocation tradeoffs we measured tadpole development time, tadpole growth, and toadlet growth. Tadpole development time was measured as days from collection to complete tail reabsorption at metamorphosis. Tadpole growth was estimated from toadlet mass after metamorphosis but before first feeding. Tadpoles were too small at the start of the experiment to get an accurate starting weight (<10 mg) without excessive handling to dry tadpoles. Differences in size immediately following metamorphosis were assumed to represent differences in growth as tadpoles. Toadlet growth was measured as the difference in weight from complete tail reabsorption to weight at euthanasia. Tadpole and toadlet growth were log₁₀ transformed.

We performed all statistics in RStudio (version 1.4.1106) using the R statistical software (version 4.0.5). All response variables were analyzed with linear mixed-effects models using the packages “lme4” and “lmerTest” with water temperature, water level and RIFA exposure as fixed effects and container as a random effect. Mass or SUL was used to control for size for corticosterone, bufadienolide, lipid, and hop performance measurements. Toad stress response was analyzed at absolute release rates (baseline, stress induced, and recovery release rates) and relative release rates (magnitude of stress response and negative feedback). Corticosterone absolute release rates were quantified as the amount of water-borne corticosterone measured over 1 h, divided by body mass (pg/g/h), and log base-10 transformed. The magnitude of stress response was quantified using the following formula (Lattin and Kelly, 2019; Bókony et al., 2021):

Similarly, negative feedback was quantified using the following formula:

A type-2 analysis-of-deviance table was used to pairwise compare the three absolute concentrations across the treatment groups using the “Anova” function in the “car” package. The significance level was corrected using the false discovery rate (FDR) method in the package “emmeans.” To parse out any three-way interactions we calculated the estimated marginal means using the “emmeans” package and plotted the results with 95% confidence intervals. Correlations between corticosterone measurements and total measured bufadieniolide amount, hop performance, and lipid levels were performed using Spearman’s rank correlation. Because of the inherent variability of these response variables, we considered a value of p threshold of 0.06 to be significant rather than the arbitrary convention of 0.05.

The main treatment effects significantly changed baseline tadpole corticosterone release rates. Warmer water increased corticosterone release rates and reduced water level also increased corticosterone release rates (temperature: estimate = 0.82, df = 51.9, t = 3.4, p = 0.001; water level: estimate = 0.45, df = 51.9, t = 1.9, p = 0.06). Pairwise comparisons of toadlet corticosterone release rates across all treatments showed a significant effect of sample type indicating that our hormone sampling methods were effective in detecting an active HPI axis (Table 1).

Baseline toadlet corticosterone release rate showed a significant three-way interaction with water temperature, water level, and exposure to RIFA (Table 2). Baseline corticosterone release rates are higher when exposed to RIFA and 32°C water or low water level and RIFA, but when exposed to 32°C water, low water level, and RIFA baseline corticosterone release rates are similar to other treatments although more variable (Figure 2).

The magnitude of the stress response showed a significant three-way interaction as well (Table 2). Low water level, 32°C water, and exposure to RIFA increased the magnitude of stress response, while the same larval treatment that was not exposed to RIFA reduced the magnitude of stress response (Figure 3). The treatments showed no significant effect on the other measures of the corticosterone profile (negative feedback efficiency, stress induced release rates or recovery release rates).

Maximum and average hop performance were SUL corrected to account for the different sizes of individuals. Our treatments had no effect on maximum hop performance. Still, average hop performance was reduced in toadlets exposed to RIFA (Table 3).

The limit of detection (LOD) and limit of quantitation (LOQ) were calculated for the three bufalin standard curves (Supplementary Tables S4, S5). Most samples contained bufadienolides above the limit of quantitation for their respective calibration curves, but all calculated bufadienolides concentrations were above the limit of detection (Supplementary Table S3). Of the eight bufadienolide compounds we analyzed for in the reconstituted toadlet filtrates via LC–MS/MS, only three bufadienolides (arenobufagin, marinobufagin, and telecinobufagin) were detected (Supplementary Table S3). Since no bufadienolides without a standard were included in our analysis, we were unable to conclude if alternative bufadienolides were present in our samples.

We found a significant interaction between the temperature treatment and the RIFA exposure treatment on arenobufagin concentration. Toadlets reared in 32°C water had reduced concentrations of arenobufagin when not exposed to RIFA but the concentration found in toadlets across temperature treatments did not significantly differ when exposed to RIFA (Figure 4). We found a marginal effect of the temperature treatment on marinobufagin concentration as well with the 32°C water treatment again reducing the concentration of marinobufagin. We did not find any significant effect of our treatments on telecinobufagin concentration, the total amount of bufadienolides, or bufadienolide diversity (Table 4).

We found a significant interaction between water level and RIFA exposure. Low water levels were associated with reduced lipid levels regardless of RIFA exposure but those toadlets that were reared in low water levels and not exposed to RIFA had slightly increased lipid levels and those toadlets that were reared in sufficient water levels and not exposed to RIFA had decreased lipid levels. We also found an effect of temperature where 32°C water increased lipid levels (Table 5).

Tadpoles from 32°C water also developed quicker and were smaller in mass after metamorphosis than tadpoles from 23°C water. However, toadlets from reduced water level treatments as tadpoles had slower growth rates (Table 6).

We found that total measured bufadienolides were negatively correlated with magnitude of stress response and average (and marginally maximum) hop performance was positively correlated with baseline corticosterone levels. Interestingly, lipid levels were not correlated with any measure of corticosterone (Table 7).