Materials and Methods
Sampling. We collected adult salamanders of A. barbouriand A. texanum from Ohio, Indiana, and Kentucky. Study sites were selected from outside the proposed zone of sympatry, from typicalA. barbouri and A. texanum breeding habitats, i.e., stream and vernal pool, respectively, and previously confirmed as harboringA. barbouri or A. texanum mtDNA, respectively (Figure 1). We collected a total of 18 males and 18 females of A. barbourifrom 3 localities and a total of 18 males and 18 females of A. texanum were collected from 4 localities (Figure 1). For both species, only unspent adults in breeding condition were selected (i.e., females with swollen abdomens and males with swollen cloacas). We also considered direction of movement. If an adult was seen moving away from a stream or vernal pool, they were typically not collected. Assessing movement direction was especially important for A. barbouri , as they are not explosive breeders, i.e., their breeding season for a particular population, may last months, whereas A. texanum may only last a few days, and the risk of collecting a spent adult was greater than that of A. texanum (Petranka 1984). We began collection for both species on the same night, but techniques differed between A. barbouri and A. texanum sites. For A. texanum , 30.5 m x 0.6 m sections of tarp drift fences were installed along vernal pools, with pitfall traps placed every 6 m. Conversely, we collected A. barbouri crossing roads or trails running parallel to streams. We transported animals to an environmental chamber and placed with either a conspecific or a heterospecific individual of the opposite sex within 48 hours.
Breeding and Egg development. We used a nested experimental design, considering genotype crossings as a fixed factor, and resulting eggs/larval clutches from each pair as a nested random factor within each crossing. Four treatments of genotype crossings were made for adult pairs (Figure 2). Each pair was randomly assigned to a 90 L aquarium, resulting in nine pairs per treatment, for a total of 36 aquaria. Prior to salamander introductions, we filled aquaria with 54.5 L of water, with 3.78 L coming from each field site and the rest being dechlorinated water. Field site water inoculations occurred to spur natural algal growth in aquaria, as algae have been shown to have a mutualistic relationship with some Ambystoma species (Graham et al. 2014). Additionally, field site water might be important for scent-cued site fidelity (Gamble et al. 2007). Light cycle was set to a 12hr/12hr day/night cycle. We placed two large, flat limestone rocks in each tank to provide typical oviposition location of A. barbouri . Along with the rocks, we anchored plastic aquatic plants to the bottom of each tank to mimic typical A. texanum oviposition location of grasses or plants. Once pairings were placed in aquaria, we checked each pair multiple times per day for deposited spermatophores, active spermatophore uptake by females, and oviposition. Oviposition was considered an indication of successful breeding, and the effect of the four pairing treatments on breeding proportion was determined using Fisher’s exact test. We recorded oviposition location, which included plants, rocks, or scattered throughout aquaria. We also used Fisher’s exact test to determine oviposition location differences between the pairings. After six days, we removed adult pairings from the aquaria.
We reared the eggs in the breeding experiment aquaria, with no water change, filtration, or cleaning of aquaria occurring prior to or during egg stage to avoid disruption (personal communication, Ruth Marcec Greaves). Oxygen was provided using air pumps and air stones placed in the corner of each aquarium. The temperature throughout the egg stage increased from 12°C to 15°C, approximating average outdoor temperatures at collection localities. We calculated egg development times (i.e., the difference in days between oviposition date and hatch date) using the median hatch date of a clutch, as single clutches typically took up to 10 days to complete hatching. Given the difficulty of photographing all eggs without disruption of development, the maximum value of either egg number or larval number was used to measure clutch size.
We collected larvae from breeding tanks for 10 consecutive days following the first hatching and placed them into 13.6 L holding tubs filled with dechlorinated water and separated by clutch. At the end of the 10-day collection period, larvae in holding tubs were photographed and counted using the OpenCV python package and ImageJ software (Bradski 2000, Schneider et al. 2012). For larval length measurements, at the end of the 10-day collection period, clutches were photographed with a ruler as a reference. Three clutches of each genotype crossing were randomly selected, and within each clutch, 15 random larvae were selected by using the random point selector tool, and their length measured using ImageJ. Statistical effects of pairing groups on egg development time, larval count, and initial larval sizes, were determined using one-way ANOVAs.
Pigmentation. We measured natural pigmentation in larvae of each genotype early in development. Due to different hatching times of each genotype, hatching times of clutches within genotypes were averaged and pigmentation was measured 27 days post-average hatching date. We randomly selected three clutches from each genotype and took a random sample of six larvae per clutch taken, for a total sample size of 18 larvae per genotype (Figure 2). We placed each clutch sample into a glass petri dish, with enough water to cover each larva. The dishes were set onto stainless steel and covered with aluminum foil to prevent pigmentation change due to background color and laboratory lights (Garcia et al. 2003, Garcia and Sih 2003). We held larvae in these conditions for 24 hours prior to imaging (Garcia et al. 2003, Garcia and Sih 2003). Photographs were taken with a Zeiss Discovery V12 microscope installed with an AxioCam MRc5, under LED light, with 10 ms of exposure and 9.8x zoom. Images were exported as uncompressed BigTiff files and cropped into 250x420 pixel rectangles using Adobe Illustrator 2022. RGB values for each pixel were extracted and then averaged to produce a single grayscale value for each larva. This grayscale value was then divided by 3 to create a grayscale range between zero and one, zero being black and one being white. The effect of genotype on pigmentation (grayscale values) was determined using one-way ANOVA, with clutch included as a nested variable within each genotype.
Laboratory and Mesocosm Fitness Experiments. The same experimental design used in the breeding tanks was applied to our laboratory and mesocosm settings, using genotype crossing as fixed factor and clutch as a random factor nested within genotype (Figure 2). We transferred larvae from laboratory holding tubs to mesocosm or laboratory enclosures for fitness proxy experiments. Three clutches were randomly selected from each genotype, and six random samples of 15 larvae were taken from each clutch. From each clutch, three replicates were assigned to 90 L aquaria and the other three to mesocosms. In summary, we used 36 aquaria and 36 mesocosms, with three 15-larvae replicates per clutch, and three clutches per genotype crossing (Figure 2). Experiments taking place in both laboratory and mesocosm were identical in concept and experimental design, but differed in environmental conditions, larval density, and prey availability and type. Larvae of each genotype were then reared to metamorphosis and number surviving to metamorphosis, time to metamorphosis, and size at metamorphosis were measured as proxies for fitness (Earl and Whiteman 2015, Semlitsch et al. 1988).
The laboratory setup was similar to the breeding experiment. Thirty-six 90 L aquaria were filled with 54.5 L of dechlorinated water and placed in an environmental chamber with 12 hr/12 hr day/night cycle, and a gradual increase in temperature from 15-22°C to mimic natural environmental conditions. Laboratory husbandry included daily feedings of ad libitum brine shrimp. The mesocosm experiment arrangement consisted of 36 polyethylene stock tanks (1.85 m in diameter; 1,480-L volume), filled with water and 1 kg of deciduous leaves and inoculated with zooplankton. To deter predation upon larvae in mesocosms, we fastened a 0.15-cm fiberglass screen-mesh lid on the top of each mesocosm. We visually confirmed the presence of zooplankton in all experimental mesocosms prior to larval inoculation. Twice a week, tanks were checked until the first metamorph was seen. After the first metamorphosis, daily checks of mesocosms occurred each morning to protect metamorphs from desiccation. Upon metamorphosis, which was defined by a full reabsorption of the gills, larvae were removed from enclosures by netting, the date was recorded, individual photographs were taken using a Sony DSCW800 camera, and SVL measured using ImageJ software. Genotype effect on size at metamorphosis, time to metamorphosis, and survival to metamorphosis of larvae raised in laboratory aquaria and mesocosms were determined using one-way ANOVA, where clutch was a nested, random factor. All statistical analyses were conducted using R software version 4.0.2 (R Core Team, 2020). Post-hoc t-tests were conducted for each one-way ANOVA. To correct for type I error inflation associated with multiple comparisons, all post-hoc t-test p-values were first run without correction, combined into a vector, and then corrected together using the Benjamini-Hochberg procedure (Supp Table 1.).