Injury Prevalence in Ambystoma
All statistical analyses were conducted using R, version 2.15 (R Core Team 2013). Non-lethal injury prevalence among all focal species was monitored throughout the larval period in 14 ephemeral ponds in the Shawnee National Forest (SNF), southern Illinois, from 2005 to 2008. Each month following hatching of A. opacum and until all focal species had metamorphosed, 30 larvae of each species were inspected for injuries to the front and rear limbs, gills, and tail and then released. We assigned individuals to injury categories of: (1) uninjured = no visible injuries; (2) tail injured = missing all or portions of the tail; (3) gill injured = missing one or multiple gill filaments; (4) limb/digit injured = missing digits, feet, or entire limbs; or (5) multiple injury = exhibiting two or more different injuries. Due to extreme interannual variation in dates of hatching and metamorphosis within species, often by several months, we were unable to utilize repeated measures analyses due to an abundance of missing values that, when omitted following listwise deletion, would reduce the total data set by over 50%. We therefore assessed seasonal changes in injury frequency (% larvae injured/pond) by species, month, and pond using ANOVA. We initially analyzed each year separately to account for yearly variation in hatching phenology; however, based on similar results for all years, we subsequently pooled data for all years. Significant pairwise differences by month and species were identified using Tukey's HSD Tests.
Correlations between Injury Prevalence, Body Size, and Behavior in A. opacum
The relationship between injury, larval body size (i.e. snout-vent length), and microhabitat association of A. opacum was assessed in three ponds total between 2006 and 2008 in SNF and the Touch of Nature Environmental Education Center (TONEC), Carbondale, Illinois. Ponds were selected based on the absence (1 pond) or extremely low density (2 ponds) of large, predatory macroinvertebrates that could cause injuries similar to those inflicted by salamander larvae. The average density among all three ponds was 0.13 predators/m2 as determined by ten sediment core samples per pond (Mott 2010; Mott & Sparling 2010), while observed densities of larval odonate and dytiscid predators commonly exceed 50/m2 (Benke & Benke 1975; Van Buskirk 1989; Pearman 1995). Between April and May, we examined larvae from four microhabitats within ephemeral ponds: (1) benthos = > 1 m from pond edge on or among the leaf litter at the pond bottom; (2) littoral = < 1 m from pond edge in or among leaf litter; (3) water column = > 1 m from pond edge, > 10 cm above pond benthos and > 10 cm below the pond surface; and (4) surface = > 1 m from pond edge, head at pond surface. Microhabitats were designated based on the different dominant prey types in these pond regions (C. L. Mott, unpubl. data), which were examined in a subsequent analysis. A minimum of 35 larval A. opacum were collected from each region in each pond, and body sizes were measured using ImageJ (Abramoff et al. 2004). Large ponds were selected so that sampling of one region of the pond would not disturb subsequent sampling in additional regions, and sampling was conducted after larvae had stratified in the water column at night (Anderson & Graham 1967; Branch & Altig 1981).
We assigned each larva to one of five injury categories: (1) no injury; (2) one minor injury (a single injury consisting of a missing gill, digit, limb, or < 10% tail loss); (3) two minor injuries; (4) three minor injuries; or (5) major injury (any combination of missing gills, digits, or limbs with > 20% tail loss). These categories differed slightly from those utilized for identifying aforementioned seasonal trends in larval injury because the former were a component of a previous study (Mott 2010), while the latter were developed specifically for this study. Injury categories were based on observed patterns and the assumption that tail injuries would most severely limit foraging abilities (Hassinger et al. 1970).
Although we collected data on injuries and body size from the same individuals, body sizes were determined ex situ via digital photographs. However, we could not assign both injury categories and body sizes simultaneously to individual larvae as intended due to difficulties in diagnosing which injuries were observed among which larvae from photographs due to low camera resolution. Larvae were also photographed in groups of 2–4 individuals, such that we could obtain the injury scores and body sizes within these small groups, but not for any single larva, thereby preventing us from stating whether smaller larvae in particular exhibited the most severe injuries. Therefore, we conducted separate one-way ANOVAs for the effect of microhabitat and pond on (1) body size and (2) average injury level, as such an approach pools data for each microhabitat among larvae and does not require matching size and injury data for each larva individually. Essentially, the first analyses determined whether larval injury categories decreased with size and the second whether larval injury categories are associated with microhabitat selection. Given that this analysis may be inaccurate due to our inability to directly link injury with body size, we also measured body sizes and diagnosed injuries for 162 larvae preserved for gut content analyses (see below). The effects of pond and injury category on body size were assessed for preserved specimens using ANOVA, with significant pairwise differences within each analysis identified using Tukey's HSD Tests.
To determine the impact of injuries on prey selection by IG predators, we examined gut contents from 162 larval A. opacum (< 5% of each population; Mott 2010); samples sizes for each injury category were uneven due to the rarity of some injury combinations (N for injury categories 1–5 = 28, 40, 51, 12, and 31, respectively). Larvae were euthanized using a 250 mg/l aqueous solution of benzocaine and preserved in 10% buffered formalin. Stomachs were excised and examined under 10–40 × magnification, and larvae and prey items were dried at 70°C for 24 h and weighed. Micro- (< 1 mm) and macroinvertebrate (> 1 mm) prey were enumerated and identified to order and family, respectively. To account for variation in prey capture based on predator size, we utilized predator body size as a covariate in our analysis. The influence of larval injury category on diet was analyzed using MANCOVA, with response variables of prey diversity, number of prey items, and prey mass, with random and fixed factors of pond and injury category, respectively. Prey diversity was the combined numbers of orders and families for micro- and macroinvertebrates, respectively.
We assigned prey to one of three mobility categories, hypothesizing that predators with severe tail injuries would be unable to capture highly mobile prey. Prey categories were as follows: (1) high mobility = dipteran larvae, corixids, gerrids, coleopterans, (2) medium mobility = ostracoda, cladocera, copepoda, hydrachnida, nematoda; (3) low mobility or immobile = amphibian eggs, rotifera, oligochaeta, mollusca. Proportions of dietary items from each category were calculated for A. opacum, and the effects of injury on dietary proportions of each category were examined via MANOVA with random and fixed factors of pond and injury category, respectively.
We also identified pond zones in which prey were likely captured by predators based on typical patterns of prey microhabitat use. Prey items were defined as: (1) benthic = rotifera, chironomidae, oligochaeta, nematoda, physidae; (2) water column = ostracoda, cladocera, copepoda, chaoboridae, coleopteran larvae; or (3) surface = culicidae, amphibian eggs, gerridae. To link predator injury with changes in microhabitat and associated prey choice, proportions of dietary items from each microhabitat were calculated for injured and uninjured A. opacum. Response variables of prey proportions from each microhabitat were quantified with respect to larval injury categories via MANOVA using random and fixed factors of pond and injury category, respectively.
To identify size-based foraging trade-offs associated with injury, we also categorized prey by size, with the hypothesis that injured predators consume large numbers of smaller prey that are easily captured. Dietary items were assigned to size categories of small (< 1 mm; ostracoda, cladocera, copepoda, nematode, rotifera) or large (> 1 mm; amphibian eggs, chaoboridae, chironomidae, coleopteran larvae, culicidae, gerridae, oligochaeta, physidae), with proportions of prey items from each size category recorded. The effects of injury category on dietary proportion of large prey size were assessed via ANOVA with random and fixed factors of pond and injury category, respectively.
Agonistic Behavior between Injured and Uninjured A. opacum
Ambystoma opacum consume conspecifics in addition to IG and basal prey (Petranka 1998); therefore, we utilized laboratory observations of agonistic behavior to determine how previous injury affects subsequent intraspecific interactions. Injured and uninjured A. opacum were collected from SNF and TONEC in May 2008 and transported to the animal care facilities at the Cooperative Wildlife Research Laboratory, Southern Illinois University Carbondale. Larvae were housed individually in 15 × 12 × 5 cm plastic containers filled with 600 mL of reconstituted water under a 12:12 L:D photoperiod. Ambient temperature ranged from 16 to 19°C, larvae were fed daily with the amount of Artemia spp. nauplii hatching from 0.015 g of eggs, and all individuals remained under laboratory conditions for at least 1 wk prior to behavioral observations.
Larval head width was measured using ImageJ (Abramoff et al. 2004), and larvae were assigned to conspecific pairs based on equal head width because size asymmetry can skew the results of behavioral observations (Mott & Maret 2011). Uninjured larvae were assigned to competitors consisting of either another uninjured larva or a naturally injured larva, the latter of which exhibited 2–3 minor injuries including missing gills, feet, or tail tips. After pairwise assignment, larvae were returned to individual containers and denied access to food for 24 h to standardize hunger levels.
Nocturnal behavioral observations were video-recorded in a circular arena (diameter = 19 cm, depth = 3.5 cm) using infrared lighting. For each larval pair, individuals were placed in separate plastic cylinders, acclimatized for 5 min, and released. For both larvae over the following 30 min, we calculated combined totals of the behaviors: MOVE TOWARD – one salamander moves toward in the direction of another such that continued movement would result in contact; LUNGE – one salamander rapidly and abruptly moves toward another but does not intersect with any part of the other individual's body; BITE – one salamander grabs another with the mouth (definitions from Walls & Jaeger 1987) and MUTUAL APPROACH – ‘move toward’ performed by both larvae simultaneously (Mott & Sparling 2009), all of which served as indicators of ‘individual aggression’. For both control (two uninjured larvae) and treatment (one uninjured and one injured larva) conditions, we conducted 25 behavioral trials, and each larva was utilized for a single trial. ‘Total aggression’ for uninjured:uninjured and uninjured:injured larval pairs were calculated as the summed values of ‘individual aggression’ for both larvae in a trial, and these values were square-root-transformed to satisfy assumptions of normality. We subsequently analyzed ‘total aggression’ exhibited by control vs. treatment trials using a t-test. To determine whether differences in aggression observed between control and treatments larval pairs were due to increased aggression directed specifically toward injured larvae in treatment trials, we also assessed differences in ‘individual aggression’ between injured and uninjured larvae strictly from treatment trials using a paired t-test.