Selective Predation by Pond-Breeding Salamanders in Ephemeral Wetlands of Ohio and Illinois Open Access

Amphibians are key consumers that comprise a primary component of wetland ecosystems. Amphibians can occur in high densities in ephemeral wetlands and are a large constituent of wetland biodiversity, especially in their larval stage (Lannoo, 2005; Gibbons et al., 2006). Therefore, amphibian larvae have significant ecological impacts within these systems through a variety of processes (Hocking and Babbitt, 2014). For example, larval amphibians cycle nutrients in aquatic systems that provide available nutrients for other aquatic organisms (Seale, 1980; Seale and Beckvar, 1980; Whiles et al., 2013) and provide pathways for energy transfer between aquatic and terrestrial ecosystems (Regester et al., 2008; Earl et al., 2014; Schriever et al., 2014; Capps et al., 2015). Larval amphibians also influence aquatic communities through competition and consumption (Walls, 1995; Ranvestel et al., 2004; Benoy, 2008). Multiple studies have found larval pond-breeding salamanders to have both direct and indirect effects on food webs within their respective ecosystems, and studies suggest these organisms may be keystone species in lentic aquatic habitats (Holomuzki et al., 1994; reviewed in Davic and Welsh, 2004; Benoy, 2008; Regester et al., 2008). Yet, a deeper understanding about how larval salamander predation shapes wetland communities is largely unknown. Thus, examining how amphibian larvae predate aquatic invertebrate (i.e., macroinvertebrates and zooplankton) communities could shed light on how larval predatory amphibians shape aquatic communities.

Predatory pond-breeding salamanders (genus Ambystoma) are abundant in many fishless ephemeral wetlands in the midwestern United States (Lannoo, 2005), where densities of larvae can vary widely depending on species and wetland characteristics (Werner et al., 2007). For example, Van Buskirk and Smith (1991) reported densities of Ambystoma laterale (Blue-spotted Salamanders) ranging from <20 larvae/m² to 158 larvae/m² on Isle Royale in Lake Superior, Michigan, USA. In addition to occurring in potentially high densities within ephemeral wetlands, predatory salamander larvae can consume large numbers of invertebrates throughout their development (Whiles et al., 2004; DuRant and Hopkins, 2008; Chaparro-Herrera et al., 2011). The abundance of larval pond-breeding salamanders and the high volume of prey they consume suggest these salamanders could significantly alter the abundance of prey in ephemeral wetlands.

Aquatic amphibian predation on zooplankton and macroinvertebrates (hereafter aquatic invertebrates) has been found to significantly alter the demography of aquatic invertebrate communities as well as influence prey behavior (Dodson, 1974; Morin, 1981; Holomuzki et al., 1994; Wissinger et al., 1999). Yet the mechanisms for how amphibians shape aquatic invertebrate communities through predation are ill-defined. Specifically, it is unknown if larval salamanders affect aquatic invertebrate communities by disproportionately consuming certain prey (i.e., selective predation). Selective predation by larval salamanders could be the mechanism by which they alter the demography of aquatic invertebrate communities, similar to the effect selective predation by eastern newts (Notophthalmus viridescens) had on amphibian communities as evidenced by Morin (1981). If selective predation by larval pond-breeding salamanders occurs, considering the ecological function (i.e., functional feeding groups) of their prey could also provide a mechanism for the larger, indirect effects observed by others (e.g., Holomuzki et al., 1994). Other research has shown disproportionate predation by salamanders on functional feeding groups of aquatic invertebrates in Appalachian streams (Keitzer and Goforth, 2013). Therefore, an investigation of prey selection and selection of functional feeding groups by larval pond-breeding salamanders is warranted.

Here we investigate the diet composition of larval pond-breeding salamanders (hereafter salamanders) and community composition (i.e., density, functional feeding groups, richness) of aquatic invertebrates across a series of ephemeral wetlands in northeast Illinois and southwest Ohio. We tested whether the composition of larval salamander stomach contents differed significantly from the community composition of wetland water columns.

We investigated three species of salamanders across two sampling locations in the midwestern United States; Ambystoma tigrinum tigrinum (Eastern Tiger Salamanders) and Ambystoma laterale (Blue-spotted Salamanders) in Illinois, and Ambystoma jeffersonianum (Jefferson's Salamanders) in Ohio—the most abundant species present in our sampling. Ambystoma tigrinum is the largest bodied of the three species, typically grows to nearly 20 cm as an adult, breeds in a variety of water bodies from permanent to ephemeral, and occurs throughout the eastern United States from Florida to Minnesota and from the Mississippi River Valley east to the Atlantic Coast (Lannoo, 2005). Ambystoma laterale and A. jeffersonianum have similar life histories and ecologies and are known to hybridize where sympatric (Lannoo, 2005). These species are smaller than A. tigrinum and reach 14–18 cm as adults. Ambystoma laterale and A. jeffersonianum are distributed from southern Canada south to Kentucky and Virginia and east to Maine, with A. laterale representing the northern half and A. jeffersonianum representing the southern half of that distribution, with an area of hybridization in the middle (Lannoo, 2005:fig. 1).

We sampled five wetlands in Lake County, Illinois for larval salamanders and water column invertebrates 13 May to 3 September 2015 (Table 1). Wetlands ranged in size from 376 m² to 1,538 m² and were surrounded by a matrix of forest, agriculture, and urbanization. All wetlands were fishless; however, each wetland contained at least one species of larval anuran (Bullfrogs; Lithobates catesbeiana, Green Frogs; Lithobates clamitans, and Chorus Frogs; Pseudacris triseriata) and one or both species of salamanders, A. tigrinum or A. laterale.

We quantitatively sampled amphibians biweekly during the study period. Each sampling period consisted of 1) three consecutive nights of trapping using 15 to 25 Promar minnow traps (46 × 25 cm) with trapping density dictated by wetland area, and 2) “pipe sampling” at the time of each visit using methods proposed by Werner et al. (2007), where a 52 × 27-cm galvanized steel cylinder was used to sample a 0.2-m² area of the sediments and associated water column. Pipe samples were chosen by selecting locations evenly distributed throughout the wetland each sampling period. Samples were taken by carefully approaching the wetland and then quickly pushing the cylinder into the substratum. Nets (23 × 15 cm) were employed to remove all animals within the cylinder with circular sweeps until 10 consecutive empty sweeps were taken (Mullins et al., 2004). We took five pipe samples per wetland; however, if the wetland area was reduced because of drying, and samples could not be taken at least 2 m apart, the number of samples was reduced accordingly. Similarly if, because of drying, a wetland area did not allow for 25 minnow traps spaced evenly at a minimum of 2 m apart, trap density was reduced to 15 to maintain spacing. Density of larval salamanders and nonpredatory amphibians (i.e., anuran larvae) were quantified by assuming that one minnow trap (Ousterhout et al., 2015) and/or five pipe samples sampled approximately 1 m² of wetland.

For diet analysis, once per month we collected five to 10 larval A. tigrinum (50 total) and A. laterale (47 total) across all wetlands. Specimens were sacrificed using an Orajel® solution (Cecala et al., 2007) immediately after capture and were fixed within 5 min of sacrifice, using 10% neutral buffer formalin for 24 h, and stored in 70% EtOH (ethyl alcohol). Specimens were later measured (snout–vent length [SVL], anterior portion of the snout to the posterior portion of the vent; 0.1 mm), weighed (wet mass; ±0.01 g), and dissected to remove the stomach and intestinal tracts. Stomach and intestinal tracts were preserved in 70% EtOH and contents were identified to the lowest possible taxon using Merritt et al. (2008) and Thorp and Covich (2001).

Water column invertebrates were sampled with 80-μm Wisconsin plankton nets (20-cm diameter, 37-cm length). Plankton nets were attached to a 1.5-m wooden handle and swept within a 1-m² area to sample water column invertebrates at three randomly selected replicate locations within each wetland once per month. Five total plankton net sweeps per replicate were taken in the water column. Invertebrate samples were preserved in 70% EtOH. Each replicate sample was subsampled to process samples more efficiently. Water column invertebrates were sampled on the same day larval salamanders were sacrificed. All organisms were identified to family, and genus when possible, using Merritt et al. (2008) and Thorp and Covich (2001).

We sampled six wetlands located in Hamilton, Butler, and Clermont Counties, Ohio from 29 March to 24 July 2012. Wetlands ranged in size from 21 m² to 3,749 m² (Table 1). This region contains several ephemeral and permanent wetlands and is surrounded by a matrix of urban, agriculture, and forested land. All wetlands were fishless and contained larval A. jeffersonianum. Each wetland contained one to several species of larval anurans including Anaxyrus americanus (American Toads), Acris crepitans (Northern Cricket Frogs), Hyla chrysoscelis (Cope's Gray Treefrogs), L. catesbeiana, L. clamitans, and P. crucifer (Spring Peepers).

Amphibians were quantitatively sampled using the “pipe sampling” technique (Werner et al., 2007) as described above. Five to 10 pipe samples were taken per wetland, depending upon wetland size. Five samples were collected from wetlands <300 m² and 10 samples were collected from wetlands ≥300m². Samples were randomly located at least 2 m from adjacent samples. If the wetland area was reduced because of drying, and samples could not be taken at least 2 m apart, the number of samples was reduced accordingly. Density (number of individuals/m²) was calculated as the number of individuals captured within five pipe samples (0.2 m² each). For diet analysis, we collected a series of larval A. jeffersonianum from six wetlands (n = 20 to 30 per wetland; 325 total). Within 30 min of collection, all salamanders were euthanized by immersion in a 0.1% solution of neutral pH-buffered MS-222 (ethyl m-amino-benzoate methanesulfonate; Gentz, 2007). Samples were then processed as above and stomach contents were identified to the lowest taxon possible.

Water column invertebrates were sampled using a dipper consisting of a plastic container (11-cm diameter; 350-mL volume) that sampled the upper water column and surface. The number of samples collected was standardized based on site size. For example, 15 (for wetlands <300 m²) or 30 dips (wetlands ≥300 m²) were taken at each wetland. All dipper samples from within a wetland were combined and rinsed through a 63-μm sieve and preserved in 70% EtOH. These semiquantitative dip samples included mainly small macroinvertebrates and vegetation, if present. Dipper samples were taken on the same days as larval salamanders were sacrificed. In the laboratory all organisms were sorted and identified to family level using Merritt et al. (2008) and Thorp and Covich (2001).

We used nonmetric multidimensional scaling (NMDS) ordinations using Bray-Curtis dissimilarities to identify if diet items of salamanders were consumed in different relative proportions to water column invertebrate communities within wetlands. We used analysis of similarities (ANOSIM) post hoc tests to identify significant dissimilarities between diet items and water column invertebrates. We further investigated differences in salamander diets and available water column invertebrates with Ivlev's Electivity Index to determine selective predation of salamanders on certain taxa and functional feeding groups. This index uses relative abundances of diet items and water column invertebrates, where positive electivity values indicate taxa low in relative abundance in the water column, yet high in diets, suggesting preferred selection of those taxa, whereas negative electivity values indicate taxa with high relative abundance in the water column, yet low in diets, suggesting avoidance of those taxa. All percentage data (including relative abundances) were square-root transformed. All analyses were performed in Statistica 12.0 (Statsoft, Inc., Tulsa, Oklahoma) with the exception of NMDS and ANOSIM, which were performed in PRIMER 6 (PRIMER-E Ltd., Plymouth).

Wetlands in both Illinois and Ohio differed in surface area and days with water across our sampling months (Table 1). All wetlands in Illinois dried completely by the end of the study period and four wetlands in Ohio never dried (Table 1). Densities of salamanders differed across wetlands with higher densities in Ohio wetlands compared to Illinois (Table 1).

We collected 1,072 A. tigrinum and 814 A. laterale in Illinois wetlands during the study. Mean (±standard deviation [SD]) densities were 0.5 (±0.7) A. tigrinum/m² and ranged from 0 to 5.3/m² and 0.3 (±0.5) A. laterale/m², with a range of 0 to 2.88/m² across all wetlands (Table 1). Other larval amphibian densities ranged from 0 to 1.5/m². We identified 28,009 individual invertebrates from the water column representing 42 taxa across Illinois wetlands. Invertebrate communities were dominated by Cladocera (Daphnia), Copepoda, and Diptera (Chaoborus; Fig. 2A; Appendix 1). A total of 4,829 invertebrates representing 30 taxa were found in stomachs of salamander larvae in Illinois wetlands (Fig. 3B; Appendix 1). Diets of both species of salamander consisted mostly of Cladocera, Copepoda, Ostracoda, Chironomidae, Amphipoda, and smaller numbers of other macroinvertebrates (Fig. 3B). Ambystoma tigrinum and A. laterale consumed different relative proportions of prey. For example, A. laterale diets consisted of 41% Cladocera, 27% Copepoda, and 19% Ostracoda, whereas A. tigrinum consumed 74% Cladocera, with the next highest taxon group being Ostracoda, comprising only 5% of diets (Fig. 3B). Per capita consumption was estimated to be 24.78 invertebrates/day (average number of individual taxa found in each stomach) by A. laterale and 104.29 invertebrates/day by A. tigrinum.

We collected 554 A. jeffersonianum during the study. Larval A. jeffersonianum density averaged 6.7 (±21.2)/m² across all wetlands, with the highest densities found in March. Larval anuran density averaged 25.1 (±55.45)/m² across all wetlands. In total, we quantified 21,776 invertebrates representing 48 taxa across Ohio wetlands (Fig. 2A; Appendix 2). Water column invertebrate communities were dominated by Cladocera followed by Gastropoda, Copepoda, and Diptera larvae (Chaoboridae, Culcidae, and Chironomidae; Fig. 3A; Appendix 2). Diets of A. jeffersonianum were dominated by Cladocera (83%) and Chironomidae (10%), with much smaller proportions of other taxa such as Copepoda and Ostracoda (Fig. 3B; Appendix 3). A total of 15,882 invertebrates representing 25 taxa were consumed by A. jeffersonianum throughout our sampling period and per capita consumption was, on average, 47.32 invertebrates/day.

NMDS ordination showed diet items in larval salamanders were distinct from water column aquatic invertebrate taxa (Fig. 4) in which diet items and water column invertebrate relative abundances were dissimilar in Illinois (P < 0.001, Global R = 0.626) and Ohio (P = 0.022, Global R = 0.159). Thus, the taxa larval salamanders consumed were distinctly different than what was available to them in the water column. Although salamander selectivity for water column invertebrates differed among species, there were a few similarities (Fig. 5). All salamander species had positive electivity values for Cladocera and Chironomidae (Diptera) and had negative electivity values for Copepoda. Copepods had high relative abundances in the water column but low relative abundances in salamander diets, whereas Cladocera and Chironomidae (more so for Chironomidae) had relatively low abundances in the water column but were represented in much higher relative abundances in diets. All Ambystoma species had positive electivity values for collector–filterers and collector–gatherers, and negative electivity values for predators (Fig. 5).

We provide evidence that pond-breeding salamander larvae feed selectively from the water column, with a disproportionately high amount of cladocerans and a disproportionately low amount of copepods in salamander diets (Fig. 4). These results are consistent with other studies that have shown cladocerans as a major component of larval salamander diets (Holomuzki et al., 1994, Whiles et al., 2004, Bardwell et al., 2007). For example, Whiles et al. (2004) found Ambystoma cingulatum diets dominated by Cladocera in South Carolina and Florida wetlands. Selection for cladocerans over copepods has been shown in other planktivores such as fish (Fink et al., 2012) as well as in other pond-breeding salamander species (Ghioca-Robrecht and Smith, 2008). This may be a function of the ability of copepods to evade predation compared to other, more slow-moving prey such as cladocerans (Amundsen et al., 2009; Peterka and Matena, 2009), making cladoceran prey items more easily captured by larval salamanders. Other research suggests selection for prey items may be related to the predator's perception of prey size. For example, O'Brien et al. (1976) suggested predators' perceived overall size of prey based on the size of their prey's eye spot. This may partly explain the selection for cladocerans and against copepods in our study, as Daphnia species (the most abundant cladocerans in our study wetlands) have large eye spots whereas many copepods have small or no eye spots. Larval salamanders may also select cladocerans such as Daphnia for their high nutritional value compared to the lower nutritional value of copepods (Andersen and Hessen, 1991). The ability to select more-nutritious prey has been exhibited by fish, but evidence salamanders exhibit this behavior is limited (Reissig et al., 2015).

Larval salamanders selected for Chironomidae, which represented a relatively low percentage of diet items but some of the highest electivity values. Yet, larval forms of chironomids in this study region generally inhabit benthic habitats rather than open water columns. Chironomidae occur in high abundance and are an important diet item for other aquatic taxa (i.e., fish; Anderson et al., 2012) and may be important to salamander species across a wide variety of aquatic habitats and geographic locations because they are one of the most ubiquitous taxonomic groups of aquatic insects (Merritt et al., 2008). Chironomids are a common constituent in the diets of pond-breeding salamanders, as observed in other studies (Holomuzki et al., 1994; Benoy et al., 2002; Whiles et al., 2004). Regester et al. (2008) found Chironomidae to be a significant contributor to larval production of pond-breeding salamanders, but it is not known if salamanders readily identify prey that is more nutritious. The disparity between the abundance of chironomid larvae in water column samples and salamander stomachs could be the result of salamanders feeding in benthic habitats in addition to the water column. However, the majority of the remaining taxa found as salamander prey, except Amphipoda, are typical water column inhabitants. Therefore, it is difficult to determine whether chironomid larvae entered the water column and then were readily eaten by salamanders, or if salamanders sought prey associated with wetland sediments. In shallow, depressional wetlands, the distance between the water column and benthic habitats is likely a short transition, which may allow either salamanders or aquatic invertebrates to readily move from one to the other. Additionally, chironomid larvae were found in water column samples, suggesting that some did move from the bottom of wetlands into the water column.

Larval salamanders selected against predatory invertebrates while selecting for all other functional feeding groups, which may have consequences for ecosystem processes. The variation in selectivity of particular functional feeding groups may partly explain the observations by Holomuzki et al. (1994) that salamander larvae (genus Ambystoma) had no significant effect on invertebrate densities directly, but had a significant impact on primary production of wetlands. A similar effect has been observed when comparing aquatic systems with and without vertebrate predators (e.g., Herbst et al., 2009). Selection against predators may be shifting the proportion of functional feeding groups in these systems, effectively enhancing the number of predators and thereby increasing primary production through a trophic cascade. This cascading effect was observed by Herbst et al. (2009) when comparing trout stream to fishless stream communities. Indirect effects of selective predation have been observed in other studies as well (Turner and Mittelbach, 1990). However, it is difficult to attribute any one variable or species to ecological processes within dynamic food web systems, as their interactions are complex and interwoven. This was not the initial objective of our study and, therefore, our data collection is poorly aimed at determining if trophic cascading is occurring as a result of larval salamander selective predation. Our data indicate selective predation exists and provide a basis for further investigation into the effects this may have on whole ecosystem processes within ephemeral wetlands.

Further understanding of how larval salamanders may shape ephemeral wetland food webs may underscore their importance across the landscape and increase conservation efforts. This would be especially useful in the midwestern United States, where habitat loss and fragmentation have contributed to the decline of salamander populations. Therefore, we suggest that further investigation of ephemeral wetland food webs and the impact of larval salamanders within ephemeral wetlands are needed. We also suggest that including functional feeding groups in further research may provide interesting avenues for investigation. For example, altered ratios in wetland functional feeding groups may unveil indirect effects of salamander predation. Studies centered on larval salamander selection of functional feeding groups, branching from the research we present here, may provide insight into the full ecological significance of larval predatory pond-breeding salamanders in isolated wetlands.

We thank J. Carr, C. Dobbs, K. Perkins, M. Radanovic, and N. Rastegar for field assistance. This research was conducted under US Environmental Protection Agency (EPA) Institutional Animal Care and Use Committee (IACUC) Protocol 11-01 and Loyola University of Chicago IACUC Protocol 83. The Hamilton County Park District, Cincinnati Parks, and Cincinnati Nature Center provided permission to access sites. This research was conducted under Ohio Department of Natural Resources Permit 14-2, Illinois Department of Natural Resources Permit NH16.5883, Illinois Nature Preserve Commission Permit: Joseph Milanovich, Lake County Ambystoma, and Lake County Forest Preserve District Permit 2015-34. A portion of this research was conducted while JRM held a National Research Council Associateship Award at the US EPA. The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the US EPA.