Diet composition of the Karpathos marsh frog ( Pelophylax cerigensis ) : what does the most endangered frog in Europe eat ?

Diet composition of the Karpathos marsh frog (Pelophylax cerigensis): what does the most endangered frog in Europe eat? The Karpathos marsh frog (Pelophylax cerigensis) is considered the most endangered frog in Europe. Here we assess its feeding ecology and examine 76 individuals from the two known populations using the stomach flushing method. We also measured body weight, snout–vent length, mouth width and prey width and length. Pelophylax cerigensis follows the feeding pattern of green frogs of the adjacent areas, with Coleoptera, Araneae, Isopoda and Hymenoptera being the main prey groups. The two populations differed in body size but had similar values of prey abundance and frequency. It seems that P. cerigensis follows a strict feeding strategy. Further research on prey availability in its habitats will provide valuable insight.


Introduction
The extensive anthropogenic activity in the last century has changed natural ecosystems and put many species at stake (Wilson, 2002;Pimm et al., 2014).Successful protection and effective conservation of threatened species require good knowledge of their overall biology (Tracy et al., 2002;Wikelski and Cooke, 2006;Bertolero and Oro, 2009).Feeding ecology is one of the most important biological parameters as it shapes numerous aspects of animal life (Vervust et al., 2010;Brown et al., 2017;Olsen, 2017).Assessing the dietary regimes of endangered animals provides important insight into the identification of critical food resources and may contribute to integrated conservation plans for many animal taxa (Palazón et al., 2008;Pagani-Núñez et al., 2011;Butler et al., 2012).
The Karpathos marsh frog (Pelophylax cerigensis) (Beerli et al., 1994) is endemic to the island of Karpathos, south Aegean Sea, Greece (Valakos et al., 2008).Categorized as Critically Endangered by the IUCN (Beerli et al., 2009), it is considered the most endangered anuran amphibian in Europe because its range is restricted to y two small rivers (rivulets or brooks in reality) in the north part of the island (Temple and Cox, 2009).Such small, insular wetlands are nowadays considered the most endangered ecosystems in the Mediterranean Sea, representing isolated oases for birds, amphibians, aquatic reptiles and invertebrates (Cuttelod et al., 2008).The general biology of P. cerigensis is largely understudied since the first description of the species (Beerli et al., 1994).Feeding ecology is a classic scientific topic in frog studies as food quality and availability reveal the important position of the group in food webs (Duellman and Trueb, 1994).Frogs represent a considerable portion of the riparian biomass and serve as energy redirectors to higher trophic levels (Burton and Likens, 1975).Thus it is important to understand where they stand in food webs and to unravel how habitat global energy fuels riparian communities (Çiçek, 2011;Bogdan et al., 2013).
Pelophylax cerigensis remains an unknown animal in terms of ecology, and as such, no specific protection measures have been taken so far.Here we studied the two known populations of P. cerigensis.We aimed to (1) assess the diet of the species for the first time, (2) examine possible differences between the two populations, and (3) compare trophic niches and food composition to those of other frogs in the Balkans.

Material and methods
We sampled both river sites at Argoni (35.6948º N, 27.1523º E) andNati (35.7018º N, 27.1786º E) in the northern part of Karpathos Island.Frogs were collected during late spring (last week of May 2015, 2016 and 2017; average temperature and rainfall did not differ between the three years) from small ponds along riverbeds.Fieldwork was carried out in May, as this is the time of the year when the frogs are easier to observe and capture on Karpathos Island.During summer and early fall, the rivulets dry out and frogs resort to the nearby vegetation, being hard to catch.On the contrary, during late fall and winter, the rivulets turn to torrents, and frogs, once more, avoid them.The landscape in general is characterized by sparse and low vegetation in a rocky background with high erosion.The vegetation around the rivers consists of tall nerium shrubs (Nerium oleander), sparse pine trees (Pinus brutia) and low shrubs such as spiny rush (Juncus articulatus) and thyme (Thymbra capitata).
Seventy-six individuals (51 from Argoni and 25 from Nati) were captured by net or hand and were anesthetized with an MS-222 solution.To anaesthetize the frogs we followed the instructions of the Amphibian Research and Monitoring Initiative (U.S. Geological Survey) (ARMI SOP No. 104-Standard Operating Procedure) for safe anesthesia (Downes, 1995).We placed the animals in a plastic water bath (2 cm deep) containing a tricaine methane sulfonate solution (50 mg/L) for 15 min.After this period, we rinsed its skin with fresh water to avoid deeper levels of anesthesia.Frogs started to recover after 10 min.For each frog we took the basic morphometric measurements: body weight (W) with a digital scale (i500 Backlit Display, My Weight, accurate to 0.1 g) and snout-vent length (SVL) and mouth width (MW) with a digital caliper (Silverline 380244, accurate to 0.01 mm).To remove stomach content we used the stomach flushing method (Solé et al., 2005).Besides being simple and effective, this method provides high quality results without sacrificing animals as it can be applied in live individuals.It is the most widely used, non-invasive technique in frogs (e.g.Lamoureux et al., 2002;Çiçek, 2011;Rebouças et al., 2013;Bogdan et el., 2013;Plitsi et al., 2016) with significantly less impact than other methods (Bondi et al., 2015) and thus it can be used even in endangered species (Bower et al., 2014;Watson et al., 2017).Holding the animal with one hand, we gently opened the mouth with a spatula and then carefully introduced the infusion tube (made of supple silicon to avoid perforations of oesophagus) of a 20 ml syringe that contained water from the pond where the frogs were captured.We flushed the content of the syringe into the stomach forcing out the consumed prey items till no more stomach content appeared.The water with stomach content was stored in a plastic glass and then decanted into a sieve.Prey items were collected with forceps and preserved in 70 % alcohol in small eppendorf tubes.
After measuring the frogs and collecting stomach contents, we waited for the captured individuals to recover.We kept the frogs into a plastic bin for 30 min to ensure that all of them were in good condition.None of the individuals died during this procedure and all of them were released in their habitat after fully recovering.The personnel of the Management Body of Karpathos that regularly patrolled the river did not encounter any dead individual during the days following the measurements.
Stomach contents were preserved in 70 % alcohol and then transported to the lab (Dept. of Biology, National and Kapodistrian University of Athens) where they were identified to order with a stereomicroscope (Wild Heerburg M38).Prey item width (W) and length (L) were measured to the nearest 0.1 mm.Following Dunham (1983), prey volume (V) was calculated using the ellipsoid volumetric formula: At this point, we should point out that although this is the typical method used in similar studies, the volume of prey groups such as Diplopoda or Formicidae calculated with this approach is rather unrealistic.However, for the sake of comparison with other studies on frog feeding ecology, we apply it here as well.
For each individual whose stomach contained prey items (contrary to empty stomachs), we calculated the minimum, mean and maximum prey item width, length and volume, while also counting the total number of prey items.For every identified prey category we calculated its relative abundance (%A), frequency of occurrence (%F) and relative volume (%V), per population as well as overall.
We assessed food niche breadth using Levin's standardized index Bi (Levins, 1968), where p i is the relative abundance of every prey category in each population: Niche overlap (O) was evaluated with Pianka's (1973) index, where p i is the relative proportion of prey category i in each of two populations A and B: We used t-tests to compare frog and prey morphometric variables between the two populations, unless the assumptions of normality (Shapiro-Wilk test, p < 0.05) and homogeneity of variances (Levene's test, p < 0.05) were violated, in which case the non-parametric Mann-Whitney test was used instead.A Mann-Whitney test was used to compare the number of prey items between the two populations.The relationship between frog body size and prey size was assessed with Pearson's product-moment correlation.In order to examine the similarity of diet composition between the two populations, x 2 -tests of independence were performed on absolute and relative abundance (N, %A) and frequency of occurrence (F, %F) of the five most common overall prey categories respectively.All statistical analyses were performed according to Zar (2010) using R 3.4.2(R Core Team, 2017).
Stomach analyses yielded 296 prey items (199 from Argoni and 97 from Nati).Fifteen frogs, nine from Argoni and six from Nati, had empty stomachs.The two populations did not differ in the number of prey items consumed (Mann-Whitney U = 640, p = 0.98).A frog from the Argoni River had the highest number of prey items in a single individual (11).
The main prey category was insects, with an overall value between the two populations reaching 64.65 %.Coleoptera and Araneae topped the list of the most abundant and the most frequently consumed prey taxa (table 2).Finally, Coleoptera occupied the largest global relative volume in frog stomachs (35 %).
Eleven prey categories were present in both sites, while six taxa were found only at Argoni and three only at Nati.These unique prey groups, however, were of minor importance as they were represented by low relative abundance and frequency in the stomachs examined (table 2).
Argoni frogs ate larger prey items than the Nati population (mean length: Mann-Whitney U = 703.5,p < 0.0001; mean width: Mann-Whitney U = 696, p < 0.0001) (table 3).When we additionally compared the mean minimum and mean maximum prey item lengths and widths, we found that the two populations differed considerably in all prey size variables (Mann-Whitney U, p < 0.05 for all comparisons).As a consequence, prey volume was also higher in the Table 1.Values for snout-vent length (SVL) and mouth width (MW) (in cm) and body weight (BW) (in g): means ± standard deviation; range (between brackets): N, sample size.

Site
The most abundant prey taxa and the most frequently eaten prey items were similar in both populations (table 2).The two populations differed significantly when absolute abundance of the five most common prey groups was tested (x 2 = 17.004, df = 4, p = 0.0019).However, relative abundance differences were not statistically significant (x 2 = 0.13, df = 4, p = 0.998).No statistical differences were found between the two populations, either in absolute (x 2 = 1.91, df = 4, p = 0.753) or in relative (x 2 = 0.07, df = 4, p = 0.999) frequencies of occurrence.
Finally, a notable deviation was observed in the relative volume between the two populations: while Coleoptera were the dominant group in frog stomachs at the Argoni River (38.19 %), they reached only a small percentage (5.09%) at Nati, where Isopoda was the first pray taxon in volume (33.92 %) (table 2).

Discussion
Karpathos is one of the oldest islands in the eastern Mediterranean (around 8 my).This long isolation is reflected in the two endemic amphibians occurring on the island, the Karpathos Lycian salamander (Lyciasalamandra helverseni) and P. cerigensis.However, although L. helverseni maintains dense populations, P. cerigensis is known from only two sparse populations in the north of Karpathos (Lymberakis et al., 2018).Climate change strongly affects the east Mediterranean, with increasingly fewer rainfalls (Giorgi and Lionello, 2008), further exacerbating the general water scarcity on Karpathos.Thus, and also due to its restricted range, P. cerigensis faces severe risks in the immediate future.Here we present the first data on its diet.These findings should be taken into account to effectively protect the species.We should point out, however, that the most important protection task undertaken on the island should be to conserve the Karpathos environment as a whole.The Karpathos marsh frog adopted a rather simple diet, comprising fewer prey taxa than other ranid frogs in the area (Cogălniceanu et al., 2000;Mollov et al., 2006;Bisa et al., 2007;Bogdan et al., 2013).The dominant prey groups were insects, spiders and isopods.Frogs from the two examined populations prey on the same invertebrate taxa (11 common groups), with limited differentiation between them (table 2).
A clear difference arose from the body size measurements: the frogs from Argoni were larger and heavier, and had a larger mouth width than their Nati peers (table 1).As a consequence, Argoni frogs ate larger food items despite the fact that the number of prey items per stomach was similar (Cogălniceanu et al., 2000) (table 3).Intrapopulation analysis revealed that body size also affected diet: larger frogs consumed larger prey in both populations.Argoni and Nati share the same ecological and abiotic parameters (vegetation, substrate, slope, exposure to winds and sunlight), and hence it is difficult to identify the reason underlying discrepancies in their diets.The strong relation between body and prey size may account for the differences in the taxonomy (limited though, as mentioned above) and size of the prey consumed.
Diet compositions were quite similar between the two populations, suggesting a high dietary/niche overlap (Pianka's index,O = 0.840).The reason for such a high dietary overlap should be sought in habitat similarity: the two focal populations share the same vegetation type and main abiotic characteristics.The values of Levin's index (Bi) fall within the range of other Balkan frogs (Covaciu-Marcov et al., 2010;Çiçek, 2011;Cicort-Lucaciu et al., 2011;Bogdan et al., 2012), indicating a rather broad feeding niche for the species.
The two populations did not differ in the frequency of consumed prey.Argoni frogs primarily consumed Coleoptera (76.19 %), whereas Araneae was the most frequent prey taxon in Nati (table 2).Our findings confirm previous research reporting that Coleoptera and Araneae are a typical food resource for Pelophylax frogs (Sas et al., 2007;Balint et al., 2010;Mollov et al., 2010;Bogdan et al., 2012;Plitsi et al., 2016).Frequency, taken together with relative abundance in stomach content, provides a reliable evaluation of feeding homogeneity (Cogălniceanu et al., 2000).In both populations, frequency and relative abundance received high values (table 2).
It is worth highlighting two particularities in the diet of P. cerigensis: the small percentage of aquatic taxa and the finding of a lizard tail in a single stomach from a frog in Nati.The limited consumption of aquatic prey is not a surprising feature among ranid frogs that typically seek their food in the banks of their habitats (Çiçek and Mermer, 2007;Mollov, 2008;Covaciu-Marcov et al., 2010;Çiçek, 2011).However, the particularly low portion in the P. cerigensis stomach content should be attributed to the physical water scarcity on the island.From April to October the rivers practically survive by small, shallow pools that cannot support a typical aquatic invertebrate fauna.Thus, frogs largely consume terrestrial and flying prey, and may include unusual food, such as the observed lizard tail.The latter belongs to the European copper skink (Ablepharus kitaibelii), a small lizard of the Scincidae family.Though the consumption of vertebrate prey is not uncommon within the genus Pelophylax (Ruchin and Ryzhov, 2002;Covaciu-Marcov et al., 2005;Çiçek and Mermer, 2006), this is only the second time an incident of saurophagy is reported (Nicolaou et al., 2014).Mediterranean insular herpetofaunas are Table 3.The number of examined individuals, the total number of prey items, mean dimensions of the consumed prey, and the feeding diversity (Levin's, B) and niche overlap estimation (Pianka) indices.

Argoni Nati
No  (Castilla et al., 2009;Brock et al., 2014;Cooper et al., 2015).Our finding may echo such behaviors but it is probably of an accidental nature.
We stress here that our results refer only to spring.Frog diet is known to change through the seasons, depending on food availability and microhabitat use (Das, 1996;Covaciu-Marcov et al., 2005).To assess the annual diet of the species, similar work should be carried out even during the unfavorable periods of the year that were mentioned earlier.
The dramatic changes in the Miocene eastern Mediterranean and their impact on Pelophylax phylogenetic history have been well studied (Lymberakis et al., 2007;Akin et al., 2010;Plötner et al., 2010;Poulakakis et al., 2015).Though the feeding ecology of the mainland Pelophylax frogs is also well studied (Çiçek andMermer, 2006, 2007;Sas et al., 2007;Mollov, 2008;Sas et al., 2009;Mollov et al., 2010;Bogdan et al., 2012;Plitsi et al., 2016), in striking contrast there is an intense lack of general biology studies in the case of insular species and populations.To the best of our knowledge, this is the first systematic work on the diet of an island frog in the eastern Mediterranean.The significance of the dietary profile of species that requires protection has been highlighted in a comparative frame (Pope and Matthews, 2002;Fisher and Owens, 2004;Wiens et al., 2010).Our results stress the importance of beetles and spiders as primary food sources in the focal habitats.Measurements that will attract these two prey taxa (e.g.specific plants, suitable microhabitats) will ensure the smooth energy flow to the frogs.Fresh water is a rare commodity on Mediterranean islands, directly affecting frogs' distribution and future survival (Vervust et al., 2013).The conservation of these unique populations is a demanding task of the highest priority.

Table 2 .
The relative abundance (A%), frequency of occurrence (F%) and relative volume (V%) of prey consumed by the two P. cerigensis populations.