Introduction
Phenotypic changes in organisms can be induced by changes in environmental variables (Van Buskirk and Relyea, 1998” Langerhans et al., 2007) and selected phenotypes are those that confer increased fitness within the habitat (Langerhans et al., 2007). Thus, habitat heterogeneity (i.e. spatial variance of environmental variables) often becomes a strong promoter of morphological variation, and an association between phenotype and environment is expected in organisms that live in changeable habitats (Langerhans et al., 2007” Michel, 2011).
Tadpoles can be found occupying several types of habitats and microhabitats, such as temporary or permanent ponds, streams and phytotelmatas (Inger, 1986; Altig & McDiarmid 1999a,b). Differences in the physical (e.g. size, width, depth” Eterovick, 2003), chemical (e.g. pH, temperature” Noland & Ultsch, 1981), landscape (e.g. matrix type, conservation level of native vegetation” Souza”Queiroz et al,. 2015) and biotic traits (e.g. presence of competitors and predators” Azevedo”Ramos et al,. 1999) can affect the rate of development and the predation risk, for example, and regulate the tadpole diversity. However, tadpoles are generally restricted to the site selected by the adults during oviposition (Wells, 2007), making microhabitat selection important to tadpole survival, wherein they can select the most appropriate patch of microhabitat that provides both an enemy free space and feeding opportunities. Additionally, tadpoles can change their morphology as a response to changes in the environment and to the presence of predators and competitors, showing high phenotypic plasticity (i.e. the capacity of a genotype to display different phenotypes as a response to changes in the environment” Miner et al., 2005; Gianoli & Valladares, 2012).
Morphological variability allows tadpoles to fine tune their morphology to the microhabitat, resulting in significant morphological variation among ponds (Langerhans et al. 2007, Michel 2011, Relyea 2002). Thus, we can expect a phenotype”environment association for tadpoles, influenced not only by historical factors, such as phylogenetic conservatism, but also by contemporaneous effects, such as predation and competition, resulting in the mixed influence of adaptive convergence and common evolutionary history (Marques & Nomura, 2015). Therefore, we can observe a considerable amount of morphological diversity in tadpoles, as tadpoles of different anuran species have different morphological adaptations to different habitats or microhabitats (Orton, 1953, 1957” Altig & Johnston, 1989).
In this study, our objectives were (i) to investigate which set of environmental variables best explains the occurrence of tadpoles in a particular assemblage and (ii) to determine how environmental heterogeneity can affect tadpole morphology. Although environmental heterogeneity is an important predictor for both assemblage composition (e.g. Pfennig, 1990; Eterovick & Barata, 2006) and tadpole morphology (e.g. Van Buskirk et al., 1997; Van Buskirk, 2002), we predict that the perceived and discrete environmental components will affect differentially the tadpole assemblage. Our hypothesis was that environmental heterogeneity at the pond’s margin would affect the occurrence of adults and also the assemblage composition, but would not interact with tadpole morphology. Thus, we expected that tadpole distribution would be mainly driven by adult choice of breeding sites. Conversely, because environmental heterogeneity is scale”dependent (i.e. environmental variables exhibit different levels of heterogeneity according to the measured scale” Michel, 2011), we predicted that environmental variables such as aquatic vegetation inside the pond and the pond bottom substrate would have the strongest association with the tadpole morphological traits, once we expected that changes in the perceived environment would be compensated by morphological adjustments.
Materials and methods
Study area
Brazilian Cerrado is the second largest biome in Brazil and is formed by a set of different patches of dry forests and gallery forests surrounded by savannah and grasslands (Ribeiro & Walter, 2008). There are two seasons; the wet season lasts from October to March and the dry season from April to September. Mean temperatures range from 22 to 27”C (Ribeiro & Walter, 2008). This singular myriad of vegetation types results in the high environmental heterogeneity of the Cerrado biome, and in equally diverse biota (Nogueira et al., 2009), making this biome one of the world biodiversity hotspots and a priority area for conservation efforts (Mittermeier et al., 2004). Currently, 209 species of amphibians are known in the Cerrado, with 51% of them endemic species (Valdujo et al., 2012).
We sampled 86 ponds over a period of four years (2010 ‘ 2013) in Cerrado areas from the state of Goi”s (Brazil) (Fig. 1). All ponds were randomly selected and sampled once, during the wet season. Pond length ranged from 1.5 to 400 m (average 89.19 m) and pond width from 1 to 212 m (average 45.31 m), with a depth between 0.03 and 5 m (average 0.92 m). Most ponds were permanent, lentic, and located in open areas.
Data sampling
Tadpoles
We sampled 86 ponds, throughout four years (2010’2013) in State of Goi”s, Brazil (Tab.S1). The tadpoles were captured using a wired dip net with a mesh of 3 mm and a diameter of 300 mm that was swiped during one hour in the pond margins, clumps of aquatic vegetation and water of 0.9 m depth in average. The tadpoles collected were anesthetized, killed, and identified in the laboratory using taxonomic keys (e.g. Rossa”Feres & Nomura, 2006), and deposited in the Zoological Collection of the Federal University of Goi”s (ZUFG 534-628; ZUFG 927-978, ZUFG 1056-1084, ZUFG 1192-1215, ZUFG 1281-1312, ZUFG 1335-1376, ZUFG 1779-1803, ZUFG 1928-1940, ZUFG 1962, ZUFG 2583). After the tadpole identification, we organized the presence”absence information of each species found in each sampled pond in a presence-absence (occurrence) matrix (Tab. S2).
Habitat descriptors
We described the ponds where the tadpoles were sampled using a detailed exclusive standardized table to visually describe the heterogeneity of the sampled ponds which include local descriptors: pond dimensions (greatest length, greatest width and maximum sampled depth)” pond margin type (cliffed, plane, sloped, excavated)” pond bottom substrate (rock, stones, coarse gravel, gravel, sand, clay, mud and leaf litter)” percentage of each vegetation type inside the pond (none, submerged, floating, upright herbaceous, shrub, tree and Typha domingenses)” percentage of each vegetation type at the pond margins (none, herbaceous undergrowth, erect herbaceous, shrub, arboreous and Typha domingenses) and landscape descriptor: percentage of land use within 500 m of the pond perimeter (natural vegetation, pasture and short crops). All traits were estimated as a percentage and transformed into a semi”qualitative variable to reduce the variation caused by the estimation of the observer as follows: 0%=0” 1’20%=0.1” 21’40%=0.3” 41’60%=0.5” 61’80%=0.7” 81’100%=0.9. These environmental variables reflect the pond heterogeneity, the availability of microhabitats and differences in land use.
We created five independent indices (Tab. S1) using the environmental descriptors based on the Nessiman et al. (2008) approach, and adapted for tadpoles (see Costa et al. 2017), as a measure of environmental heterogeneity. To calculate the indices for each pond trait (i.e. pond marginal type, pond bottom substrate, percentage of each vegetation type inside pond, percentage of each vegetation type at the pond margins and percentage of land use within 500m of the pond perimeter) we assigned a score from 1 to 5 for each category (0.1=1; 0.3=2; 0.5=3; 0.7=4; 0.9=5). The scores were then weighted to the maximum scores (i.e. we divided each score of the trait for the maximum score obtained for the category, then we add all scores and divided for the number of traits of the category), in this way, all features has the same weight in the analysis. Finally, we obtained an average of all the analyzed parameters for each pond trait, resulting in a value that range between 0 to 1, where 0 represents the less value of environmental heterogeneity and 1 the higher value of environmental heterogeneity. In this way, we created five indexes for each pond that reflect the individual contributions of each pond trait to the total heterogeneity. Pond dimension were calculated as a result of length x width x depth.
Morphological matrix
To extract morphological information from tadpoles, we photographed all individuals between Gosners’ stages 35 and 42 in lateral view. We positioned the tadpoles in a petri dish, with ultrasound gel, in lateral view and took photographs using a camera positioned with a tripod, standardizing the distance from the camera to the tadpole at 20 cm. The pictures obtained allowed us to extract the landmark features of all individuals. The landmarks provide anatomical references that represent the shape of the organisms and, ideally, should be recognizable in all individuals, as they are homologous and have biological correspondences (Sneath & Sokal, 1973). We defined 23 landmarks that describe tadpole morphology in lateral view (modified from Van Buskirk, 2009) and used them to perform a geometric morphometric analysis. In the geometric morphometric, each landmark is represented by x and y coordinates in a Cartesian plane and we use a set of landmarks to represent the organism shape, after excluding the effects of size, position, and rotation (Zelditch et al., 2012). We transformed our landmark coordinates using the Procrustes method (Rohlf, 1990) that reduced the differences between landmarks configurations from the centroid of the average shape (i.e. the average distance among all landmarks from the gravity center” Zelditch et al,. 2012), and obtained the partial warp scores that represent deformation in the external morphology of the tadpoles. To summarize the deformation information, we performed a principal component analysis (PCA), extracted the two first eigenvectors (PW1 and PW2) and used them in a posterior statistical analysis. The landmarks were obtained using tpsUtil (Rohlf, 2013) and tpsDig2 (Rohlf, 2013b) software.
Data analysis
We evaluated which set of environmental variables best explained tadpole occurrence and the significance of trait”environment relationships using a generalized linear mixed model ‘ GLMM ‘ with binomial family and logit error approach (Jamil et al., 2013). The GLMM procedure described by Jamil et al. (2013) is an alternative method to RLQ (Dol”dec et al. 1996) and Fourth-Corner (Legendre et al. 1997) and resulted in parsimonious models. Thus, the GLMM allowed us to identify the set of environmental variables that account for the species occurrence data, and provides a measure of the significance of morphological trait”environment relationships (Jamil et al., 2013).
The GLMM described by Jamil et al. (2013) use the tiered forward selection approach on multiple traits and multiple environmental variables to choose the best model to predict species occurrence (Jamil et al., 2012). The tiered forward selection approach has three tiers: in the first tier the random factors were selected” in the second tier the fixed effects were selected” in the third tier nonsignificant terms were removed using a modified Akaike information criterion (SigAIC” Jamil et al., 2012). Using this approach, environmental variables were added to the model as random and fixed terms, and, in our study, morphological traits were added only as fixed terms (see Table 1 for a list of random and fixed effects used in model; Jamil et al., 2012). To describe species occurrence and evaluate the significance of trait”environment relationships, we selected models with the lowest SigAIC values. All analyses were performed using R software (version 3.2) with the package lme4 (version 1.1” Bates et al., 2015).
Results
A total of 486 individuals of 41 species were collected in 86 ponds. The most abundant species were Physalaemus cuvieri (138 individuals) and Scinax fuscomarginatus (59 individuals).
We used 23 pairs of partial-warps to describe the total morphological variation of individuals. According to our PCA results, the first partial warp (PW1) explained 45.5% of the general shape variation and mostly reflects variation in the tadpole body shape, tail shape, eye deviation, and nostril and mouth positions (Fig. 2). The second partial warp (PW2) explained 17.6% of the general shape variation and mostly reflects variation in tadpole tail and body shapes (Fig. 2).
The species occurrence was best explained by a multi-interaction model. In the first tier were chosen the random terms of the model and included all environmental variables (Tab.1). In the second tier, environmental variables entered in the model as random and fixed terms, then morphological traits were added to account morphology:environment interaction. A total of eight interactions was added to the model, and one interaction (pond bottom substrate:PW2) was subsequently deleted in the third tier. The model that best explained the occurrence of tadpoles in the Brazilian Cerrado included all environmental variables (tier one) and seven morphology”environment interactions (tier two” Tab.1).
Vegetation type at the pond margins, pond dimensions and land use were positively associated with PW1, while vegetation type inside the pond and the pond margin type were negatively associated with PW1. For PW2, pond dimensions and land use were negatively associated (Tab.2). Figure 2 shows general tadpole shape change in morphological space and how it responds to the environmental traits according to our best model. Thus, in ponds with large amounts of vegetation at the pond margins tadpoles developed relative small bodies, relative larger tails, and deviations in the eye, nostril and mouth positions. In ponds with less vegetation inside the pond tadpoles developed larger bodies, small tails, and deviations in the eye, nostril and mouth positions. Pond dimensions and land use were both positively correlated with PW1 and negatively with PW2. Thus, tadpoles in larger and small ponds with more intensive human activity developed relatively larger bodies and larger tails, but we could observe deviations in eye, nostril and mouth positions only in larger ponds.
Discussion
Environmental variables linked to ponds characteristics and land use influenced the occurrence of tadpoles in the Brazilian Cerrado, with a strong phenotype”environment association in tadpoles. The most common tadpole morphological change was in body and tail shape, and in some cases we detected deviations in eye, nostril and mouth positions.
It is known that physical environment can exert an inductive and selective pressure on tadpole morphology (Van Buskirk and Relyea, 1998” Langerhans et al., 2007). Variation in morphology is known to be an important strategy for organisms to cope with diverse environmental conditions (Stearns, 1989” Scheiner, 1993). Our results highlight this interaction, demonstrating that some changes in tadpoles’ morphology are induced by certain environmental variables, such as presence of vegetation at the pond margins and land use change.
Pond dimensions and land use were environmental variables that were important in explaining the variation in morphological traits (PW1 and PW2). Tadpoles in larger ponds and with more intensive human activity had larger deviations in their eye, nostril and mouth positions. The presence of vegetation at the pond margins and inside ponds can provide different microhabitat for tadpoles and may be a structural factor that promotes niche partition among tadpoles of different anuran species. Microhabitat selection by tadpoles differs not only according to the position of the water column that they occupy (Eterovick & Barata, 2006), but also according to their association with different substrates. Thus, our results showed that in Brazilian Cerrado, higher levels of vegetation at pond margins favor tadpoles with low fins, as benthic species. Low levels of vegetation inside the pond margins favor tadpoles with high tail fins. Complex habitats with a lot of vegetation can be a challenge to tadpole locomotion and aquatic vegetation can favor ambush by invertebrate predators (e.g. larval dragonflies and water bugs” Nomura et al., 2011).
The percentage of land use within 500m of the pond perimeter provides an estimative of the human disturbance and conversion of native vegetation into pasture or agriculture. Moreover, intense use of pesticides can contaminate water pools through overspray in aerial application, lixiviation and overland flow (Goldsborough & Beck, 1989” Queiroz et al., 2011). Anthropogenic stressors can have adverse effects on organisms through effects on various individual traits (e.g. behavior and morphology” Teplitsky et al., 2005). Tadpoles are highly susceptible to contamination and their morphological traits could be affected by contaminants such as Glyphosate (Relyea, 2005” Costa & Nomura, 2016). Costa and Nomura (2016) evaluated the impacts of an agrochemical in tadpoles of Physalaemus cuvieri and found deviations in nostril”snout distance and eye width, which are morphological traits associated with tadpole sensory capabilities. Our results show deviations in eye, nostril and mouth positions, which may be due to anthropogenic stressors, and could affect an individual’s survival by reducing their competitive potential and increasing predation risks (Costa & Nomura, 2016).
Environmental variables are known to influence the occurrence of tadpoles (Van Buskirk, 2005). Vegetation types inside the pond, the pond dimensions and the pond bottom substrate reflect the availability of different microhabitats. Different types of vegetation and bottom substrate can offer site refuges and food for tadpoles, reducing the predation and competition rates (e.g. Rozas & Odum, 1988” Kopp et al., 2006). Moreover, vegetation inside the pond and vegetation at the pond margin can provide moisture, shelter and calling sites for adults. As tadpoles have little control of the habitat type that they will endure during metamorphosis (i.e. it is the adult that chooses the oviposition site) and their dispersion is limited (Altwegg & Reyer, 2003” Gr”zinger et al., 2012” Stein & Blaustein, 2015), the factors which influence tadpoles and adult distribution should be different. Adult distribution is affected by the spatial and terrestrial landscape (Resetarits et al., 2005” Wells, 2007), while tadpole distribution is affected by pond variables (e.g. size, vegetation, presence of predators” Ultsch et al., 1999). However, tadpole metacommunity structure can be influenced on a large scale by the physical characteristics of ponds (Provete et al., 2014). This suggests that adult reproductive behavior drives the tadpole occurrence dynamic, but on a local scale, microhabitat availability influences tadpole survival.
There is a lack of knowledge about amphibian diversity, abundance and occurrence in the Brazilian Cerrado (Valdujo et al. 2012), and our study provides additional information about environmental factors that contribute to the patterns of species occurrence. Furthermore, linking morphological and environmental traits help us to understand some ecological and evolutionary processes (e.g. niche partitioning, predation, competition) which may play important roles in tadpole assemblage structure. The Brazilian Cerrado is one of the global biodiversity hotspots and recent expansion of soybean culture and cattle ranching (Spera et al 2016) is threatening this biodiversity (Ribeiro & Walter, 2008). Our results indicate that there is more than one source of morphological variation in tadpoles, and should cause concern about the extensive use of agrochemicals linked to agricultural expansion. These chemicals can threaten tadpoles by changing their morphology, probably affecting their sensory capacity.