Feeding ecology of Erythrolamprus jaegeri jaegeri ( Günter , 1858 ) and Erythrolamprus poecilogyrus sublineatus ( Cope , 1860 ) in the coastal zone of Subtropical Brazil ( Serpentes , Dipsadidae )

The snakes Erythrolamprus jaegeri jaegeri and Erythrolamprus poecilogyrus sublineatus are sympatric and syntopic in the coastal region of southern Brazil. Herein, we analyzed the diet composition to evaluate the niche breadth and the prey selection by both species. We examined 192 specimens, and analysis of stomach contents revealed that both species predominantly consume anurans. However, the diet of E. j. jaegeri consists mainly of fi sh and amphibians, whereas that of E. p. sublineatus is broader, including fi sh, amphibians, reptiles and mammals. The Standardized Levins Index presented lower values for E. j. jaegeri (BA = 0.17) than for E. p. sublineatus (BA = 0.61), evidencing specialist and generalist strategies for each species, respectively. Regarding prey selection, E. p. sublineatus presented a larger snout-vent length, head, mouth and lower jaw than E. j. jaegeri and fed on larger prey. In addition, positive correlations between the size and weight of predators and prey were confi rmed in both species. The results show the development of different mechanisms for co-occurrence of the two species, such as prey selection by size, such that the size of the predator is related to the size of their prey, or by developing different strategies to decrease niche overlap between species.


INTRODUCTION
The ecological niche, defi ned as the set of physical and environmental variables in multi-dimensional space (Hutchinson 1957), encompasses three dimensions: spatial niche, feeding niche and temporal niche (Pianka 1982).The feeding niche is one of the most important dimensions among snakes and may infl uence biological and ecological features such as use of habitat, predatory behavior and period of activity (Toft 1985).
By the snakes eating vertebrate and invertebrate animals, we can call them carnivorous anyway (Mushinsky 1987, Greene 1997).This diversity is possible due to the development of morphological and biochemical mechanisms that facilitate snakes' detection of prey (Pough et al. 2008).Researchers have identifi ed the following among the evolutionary events that infl uenced the foraging success of snakes: cranial and muscular adaptations that allow the ingestion of prey wider than the predator itself (Pough et al. 2008, Greene 1997); the development of different mechanisms for prey detection (chemoreception, accurate DANIELE N. CORRÊA, FERNANDO M. QUINTELA and DANIEL LOEBMANN vision, heat orientation) and subjugation (strike, constriction, poisoning) (Greene 1997, Burghardt and Krause 1999, Pough et al. 2008, De Fraga et al. 2013); and occupation of different habitats (Quintela andLoebmann 2009, De Fraga et al. 2013).
Erythrolamprus jaegeri jaegeri (Günter, 1858) and Erythrolamprus poecilogyrus sublineatus (Cope, 1860) are semi-aquatic dipsadid snakes, sympatric in southernmost Brazil (Quintela et al. 2006, Quintela and Loebmann 2009, Santos et al. 2012).Erythrolamprus j. jaegeri can reach 550 mm in total length (Giraudo 2001) and is distributed in southeastern and southern Brazil and Uruguay (Dixon 1989).Over its southern distribution, it has been found mainly associated with wet habitats (Carreira et al. 2005, Quintela andLoebmann 2009), where underwater foraging and inactivity are documented (De Lema 2002).The adaptations for semiaquatic life observed in the species include the presence of a vestibule preventing the entry of water into the nasal cavity and the fenestra narina in dorsal position (Schmitt and Deiques 2009).Erythrolamprus p. sublineatus grows up to 700 mm in total length (Giraudo 2001) and is distributed throughout Argentina, Uruguay and the Rio Grande do Sul state of southernmost Brazil (Dixon and Markezick 1992), where it inhabits such varied habitats as grasslands, wetlands forests and coastal dunes (Carreira et al. 2005, Winck et al. 2007, Quintela and Loebmann 2009).
Studies have demonstrated that the competition between congeneric species is intense due to the sharing of habitat and resources (Darwin 1872, Losos 1994, Steen et al. 2014).Thus, different strategies emerge from phylogenetically close species to decrease competition and favor cooccurrence in a specifi c habitat (Pianka 1973).As a result, the level of resources shared by sympatric species is quantifi ed by the analysis of niche breadth and overlap (Hurlbert 1978).
Given the sympatric condition of E. j. jaegeri and E. p. sublineatus in southernmost coastal Brazil, we investigated the diet composition of both species aiming to answer two questions: (1) Does E. p. sublineatus feed on larger prey, seeing that larger predators forage larger prey (Shine 1991)?(2) Considering the high association of E. j. jaegeri with wet habitats (De Lema 2002, Quintela andLoebmann 2009), does this species feed on more prey of aquatic origin compared with E. p. sublineatus?
The southern RS coastal plain is characterized by low altitudes (generally under 3 m above sea level) and the predominance of open physiognomies such as grasslands, savannic grasslands, sandy fi elds and dunes (Vieira 1984).Forest formations are restricted to restinga patches, generally associated with water courses (riparian forest) and elevated groundwater areas with sandy or muddy soil (sandy and peat forests, respectively) (Waechter 1985, Marchiori 2004).Major hydrographical elements include the Patos-Mirim lagoon complex, coastal lagoons, wetlands and coastal streams (Vieira 1984).
DATA SAMPLING Specimens were collected in different habitats of the study area (e.g., wetlands, grasslands, restinga forests) and were euthanized promptly after capture.From these, we obtained the following biometric measurements using a flexible ruler to the nearest 1.0 mm and a digital caliper to the nearest 0.01 mm: snout-vent length (SVL), head length (HL), head width (HW), maximum head height (HH), mouth length (ML), mouth width (MW), mandible length (IML) and wet weight (W).Specimens examined in this study are deposited in the Herpetological Collection of the Universidade Federal do Rio Grande (CHFURG), Rio Grande, Brazil (Appendix).
The digestive tracts of the analyzed specimens were removed through incisures made from esophagus to cloaca.All of the digestive tract contents were extracted, and the fresh and partially digested items were selected.The total length (TL) and weight (W) of fresh items were measured.The digested items for which total length was possible to obtain also had their body mass estimated by comparison with biometric data from specimens stored in the herpetological and ichthyological collections of FURG.For this, three collection specimens with equal length were selected, and the mean of their body masses was used as an estimated value for the prey body mass.In the case of items at an advanced stage of digestion, well-preserved structures (e.g., femur) were used to estimate total length and body mass.For this, the means of total length and body mass of three collection specimens with equal dimensions of the selected structure were used as estimated values for the prey total length and body mass.The direction of ingestion (headfi rst or tail-fi rst) was recorded whenever possible.The food items were identifi ed to the lowest possible taxonomic category.All items were conserved in 70% alcohol and stored in the Laboratory of Vertebrates, Biological Sciences Institute, Universidade Federal do Rio Grande (FURG), Rio Grande, Brazil.All procedures were carried out according to the international practices for animal use and approved by the internal committee of ethics for animal use of the Universidade Federal do Rio Grande, Brazil.

DATA ANALYSIS
Qualitative and quantitative methods were applied to our database as proposed by Hyslop (1980).The qualitative method consisted of the analysis of the frequency of occurrence (FO%) of each identifi ed taxon, calculated by the rate of number of stomachs containing such taxon over the total number of analyzed stomachs.
The fi rst quantitative method applied was the numeric abundance index (N%) of each identifi ed taxon, obtained by calculating the rate of the absolute number of prey of such taxon over the sum of prey of all identifi ed taxa.The following method consisted of determining the percentage weight (W%) of each identifi ed taxon, given by the rate of the sum of the weights of prey of such taxon over the sum of the weight of prey of all identifi ed taxa (Hynes 1950).The Index of Relative Importance (IRI) [Hacunda (1981), modified from Pinkas et al. (1971)] of each identifi ed taxa was obtained by applying the following formula: IRI = FO%(N% + W%).
To verify the niche breadth of the two species, we applied the Levins index (Krebs 1999), given by BA = [(1/Σpi2)-1]/n-1, where BA is the standard, pi is the frequency of occurrence of the taxon i in the whole sample, and n is the number of all taxa identifi ed in the whole sample.The obtained values can vary from zero to one, such that a value close to one implies a well-distributed diet (item of the distinct taxa consumed in equivalent proportions; generalist dietary pattern), whereas a value close to zero implies that items of few taxa were consumed in high proportion and items of most taxa where consumed in low proportion (specialist dietary pattern).
The existence of significant differences between E. p. sublineatus and E. j. jaegeri SVL means and the means of total length and weight of prey categories "fi shes" and "amphibians" were examined with a Student t test.Similarly, we tested the difference between measurements of HL, HW, HH, ML, MW and IML between the species.
A linear regression method was used to verify the correlation between the log-transformed weights of prey and its respective predators.The level of correlation was measured using the Spearman Correlation Coeffi cient.DANIELE N. CORRÊA, FERNANDO M. QUINTELA and DANIEL LOEBMANN
The anurans L. latrans and P. gracilis were the items with the highest abundance, weight, frequency of occurrence and IRI in the E. j. jaegeri diet.All of the remaining prey taxa showed comparatively low values for all indexes (Table II).
The most abundant item in the contents of E. p. sublineatus stomachs was the poeciliid P. caudimaculatus, which presented low values of weight, frequency of occurrence and IRI.Leptodactylus latrans was the second-most abundant item and showed high values of weight, frequency of occurrence and IRI.The third most abundant taxon, O. maisuma, also presented high values for weight, frequency of occurrence and IRI (Table III).The niche breadth measured by the Levins index was 0.17 for E. j. jaegeri and 0.61 for E. p. sublineatus.The direction of ingestion was possible to verify in 20 prey (71.4%) of E. j. jaegeri; 18 of them (90%) showed headfi rst ingestion, and only two prey (10%) were ingested tail-fi rst.Prey total length ranged from 4 to 14% of predators' SVL (Table IV).
We could determine the direction of ingestion in 34 prey (30.6%) of E. p. sublineatus; 29 of them (85.3%) were ingested headfi rst and fi ve tail-fi rst.
Prey length varied from 2 to 17% of predators' SVL (Table V).No signifi cant difference was found between the mean weight of E. j. jaegeri and E. p. sublineatus (p=0.l4), and the SVL means between the two species differed significantly (p=0.03).Erythrolamprus j. jaegeri SVL ranged from 127 to 453 mm, and E. p. sublineatus presented SVL ranging from 153 to 534 mm.Graphic bar representations for weight and SVL of both species are shown in Figures 1a and 1b, respectively.Signifi cant differences were found between the mean weight (p=0.002) and total length (p=0.05) of prey of the two snakes species, with higher values observed in E. p. sublineatus prey.Graphic bar representations for mean weight and total length of prey of both snake species are shown in Figures 1c  and 1d, respectively.
The regression analysis indicated a moderate positive linear correlation between the weights of E. j. jaegeri specimens and their respective prey (Pearson Coefficient r= 0.5048) (Figure 2a).A moderate positive linear correlation was also observed between the SVL of E. j. jaegeri specimens and the total length of their respective prey (r= 0.5463) (Figure 2c).
A positive moderate linear correlation (r=0.6050;p<0.000) was observed between the weights of E. p. sublineatus specimens and their respective prey (Figure 2b).A weak positive linear correlation (r=0.2658;p=0.008) was detected between the SVL of E. p. sublineatus specimens and the total length of their respective prey (Figure 2d).
All graphic bars of the averages generated for measures of head (HL, HW, HH, ML, MW and IML) indicated a significant difference between species.For all variables, the highest values obtained corresponded to E. p. sublineatus (Figure 3).2011).Erythrolamprus p. sublineatus, in contrast, showed a more diversifi ed diet, preying on almost all groups of vertebrates: fi shes, amphibians, reptiles and mammals.Regarding the diet composition, we verifi ed a higher abundance of aquatic origin items, contradicting our initial hypothesis of higher intake of terrestrial prey by this species when compared to E. j. jaegeri.However, even though the fi sh P. caudimaculatus represented the most abundant item, the intake of fish was absolutely casual (present in a single stomach) and resulted in low values of IRI when compared with the amphibian contribution to the E. p. sublineatus diet.The same was verifi ed for E. j. jaegeri.

DISCUSSION
The effectiveness of E. j. jaegeri's fi sh predation was already documented (Santos et al. 2010), and, according to De Lema ( 2002), E. p. sublineatus preys on the fi shes of the genera Phallocerus and Jenynsia, which was verifi ed in the present study.However, the event here recorded may indicate an opportunist action, wherein the predator may have taken advantage of the high prey abundance while foraging.It should also be noted that the effort expended in aquatic foraging is greater when compared with terrestrial foraging due to the higher density and viscosity of the aquatic environment (Ricklefs 2003).Thus, it is fundamental that the energy obtained from food exceeds the energy expenditure of foraging (Gerking 1994, Ricklefs 2003).In this case, the unfavorably small size of the consumed fi shes was compensated for by the intake of a high number of individuals (Table I), providing a satisfactory energy gain.It is also possible that the consumed fi sh were under easy-capture conditions, such as water bodies in drought, which enhanced their vulnerability to predation.
The prey with the highest importance, occurrence and contribution to weight in the E. p. sublineatus diet was Leptodactylus latrans, a large-sized frog species (Achaval and Olmos 2007) commonly found in grasslands, riparian forests, wetlands and even in urbanized environments (Maneyro and Carreira 2012).This anuran was also the second-most important and frequent item in the E. j. jaegeri diet.The dominance of this item in the E. p. sublineatus diet and its relevancy to the E. j. jaegeri diet could be related to the fact that foraging of large prey is advantageous, providing higher energy gain (Krebs and Davies 1996).The other fact that may have contributed to L. latrans' representativeness in the diet of both snake species  Cienc (2016) 88 (1) FEEDING ECOLOGY OF E. j. jaegeri AND E. p. sublineatus 303 could simply be its high abundance in the study area.

An Acad Bras
Another relevant prey item in the E. p. sublineatus diet was Odontophrynus maisuma, a small sized but robust anuran with digging habits that occurs in terrestrial and periurban environments (Maneyro and Carreira 2012).The third-most important item in the E. p. sublineatus diet was medium-sized toads of the Rhinella granulosa group, which in the study area may correspond to R. dorbigny and/or R. fernandezae.Both species are robust toads that use from wet to severely dry environments (Narvaes and Rodrigues 2009).In sum, we can observe that the most relevant prey items for E. p. sublineatus showed robust size, indicating a strategy of foraging prey for potential energy gain (Pianka 1982).
The remaining amphibian species present in the E. p. sublineatus diet were, in order of importance, Leptodactylus gracilis, Hypsiboas pulchellus, Elachistocleis bicolor and P. gracilis, all of which showed IRI values lower than 15.Although they represent common species in the study area (Loebmann 2005), all of these anurans are characterized by small and slender bodies when compared with L. latrans, O. maisuma and Rhinella gr.granulosa, and their representation in the E. p. sublineatus diet may be related to the already discussed apparent preference for larger prey.
The item of highest importance, frequency of occurrence and weight in the E. j. jaegeri diet was the medium-sized anuran P. gracilis, an inhabitant of both natural and anthropic environments (Maneyro and Carreira 2012).Along with L. latrans, these two species confi rmed the importance of leptodactylids in the diets of both studied snake species.The sharing of leptodactylids by the two Erythrolamprus species under verifi ed syntopic conditions could be explained by the high abundance of the frog species in the studied area (Loebmann 2005).The interspecifi c competition here observed could be attenuated by the high availability of prey, resulting in resource partitioning (Toft 1985) or different feeding strategies adopted by the snakes (Mori and Vincent 2008).
The occurrence of an exclusively terrestrial item, the small lizard Cercosaura schreibersii, was recorded only in the E. p. sublineatus diet.Cercosaura schreibersii is an abundant lizard in grasslands and dunes of the study area (Quintela and Loebmann 2009), has cryptozoic habits (Deiques et al. 2007, Quintela andLoebmann 2009) and has been already found in the E. poecilogyrus diet in Uruguay (Carreira-Vidal 2002) and northeastern Argentina (Prieto et al. 2012).The reptiles here recorded, despite their abundance (Quintela and Loebmann 2009), represented sporadic predation events, and do not seem to consist of a usual E. p. sublineatus prey item.
The hairs of an unidentifi ed small mammal were found in a single stomach.At least 15 species of small mammals, including sigmodontinae and caviid rodents and didelphid marsupials, occur in the study area (Bonvicino et al. 2008, Quintela et al. 2012, 2013, Sponchiado et al. 2012), and the rat species of genus Oligoryzomys seem to be the most common (Quintela et al. 2012, 2013, Sponchiado et al. 2012).The single record of small mammal predation indicates that this may represent an uncommon E. p. sublineatus prey item.Records of mammalian items in the E. p. sublineatus diet are restricted to one case in Uruguay (Carrera-Vidal 2002).Snakes that commonly feed on mammals present solenoglyphous dentition or kill by constriction and generally immobilize or kill the prey before intake (De Fraga et al. 2013).
Our data on the composition of the E. p. sublineatus diet corroborate the data obtained by Carrera-Vidal (2002) in Uruguay, where anurans (genera Hypsiboas, Physalaemus and Rhinella) appeared as the most-consumed item, followed by lizards (including C. schreibersii) and mammals in low proportions.The composition of the diet in our study area was also very similar to the L. poecilogyrus diet in the wetlands of northeastern Argentina (Prieto et al. 2012), where the same anuran families were recorded, with the exception of the lizard C. schreibersii.Erythrolamprus p. poecilogyrus, in contrast, presented a strictly anurophagous diet in southeastern Brazil (Pinto and Fernandes 2004), composed of bufonids and hylids.Microhylids were also recorded in the diet of E. p. schotti (Cacciali and Motte 2010).Fishes, absent in our samples, have already been recorded in the E. poecilogyrus diet (Skuk 1985, Palmuti et al. 2009).
Regarding E. j. jaegeri, only anurans and fi shes were found by J.G. da Frota (unpublished data) in the stomach contents of specimens along the whole distribution of the subspecies, indicating that these may represent the most utilized resources.Insects were sporadically recorded (Michaud andDixon 1989, Achaval andOlmos 2007) and may represent occasional prey.

NICHE BREADTH AND PREY SELECTION
The Levins index indicated distinct feeding strategies between the two Erythrolamprus species.Although the value of 0.61 obtained for E. p. sublineatus does not correspond to a completely generalist diet (indicated by values very close to 1), this contrasts with the value of 0.17 obtained for E. j. jaegeri, indicative of a high level of feeding specialization.The prey availability is a factor that contributes to the differences in diet compositions of the congeneric species (Teixeira and Fonseca 2003).However, considering that the analyzed species under sympatric conditions have access to the same resources, it is possible that the capability to exploit a greater diversity of habitats and niches by E. p. sublineatus is the main factor determining such a discrepancy.The broader feeding niche of E. p. sublineatus in relation to E. j. jaegeri, was therefore verifi ed.
The direction of prey ingestion was verifi ed as anteroposterior in most cases.This was expected; anteroposterior ingestion reduces the risk of injuries caused by prey limbs and resistance to ingestion, reducing the time and energy spent during this process (Greene 1976, De Fraga et al. 2013).
Data on the rate of prey TL over predator SVL indicated that E. p. sublineatus feed on larger prey when compared with E. j. jaegeri.Such results provide support for the experimental hypothesis that E. p. sublineatus, due to its larger size, feeds on larger prey.A possible explanation could be that the optimal prey size for maximum energy gain in this subspecies could be larger, considering its larger size (Pitcher and Hart 1982).It is also possible that E. j. jaegeri, due to its smaller size, encounters diffi culties in the capture and subjugation of larger prey (Shine 1991), limiting its foraging to smaller prey.It is noteworthy that the larger E. j. jaegeri prey recorded in our study, a 32 mm P. gracilis, corresponded to less than half the length of the larger E. p. sublineatus prey, a 68 mm L. latrans.
The regression analysis indicated a moderate positive correlation between the weights of both snakes and their prey and between E. j. jaegeri SVL and length of respective prey.Positive correlation between weights of predator and prey is well documented in snakes (Mushinsky 1987, Shine 1991, Bryant et al. 2012).Juveniles generally ingest smaller prey to avoid the risks of injuries and even death caused by a poor assessment of prey (Garland and Arnold 1983, Sazima and Martins 1990, Nogueira et al. 2013).Throughout development, individuals increase their capacity to ingest larger prey, as well as the range of the consumed items (Arnold 1993).The higher coeffi cient of correlation generated for E. p. sublineatus in relation to E. j. jaegeri in weight analysis could also have been infl uenced by the higher number of prey found in the fi rst subspecies (111 compared with 28 in E. j. jaegeri), inducing a better fi t of the model.However, the low coeffi cient observed between the length of E. p. sublineatus and its prey may have been generated by the high variability found in prey size, resulting in outliers.
The differentiated head morphology between the subspecies was statically signifi cant.Erythrolamprus p. sublineatus has a larger head, considering the length and the height dimensions (Figure 3).Larger preys were also observed for this subspecies, and this may be related to the larger head structures.It is known that gape is a limiting factor for prey selection (Greene 1997, Arnold 1993).Thus, larger head, mouth and mandible dimensions, as verified for E. p. sublineatus, could favor the intake of larger items (Shine 1991, Arnold 1993, Webb and Shine 1993).Similarly, the shorter head structures of E. j. jaegeri may limit its foraging to smaller prey (Webb and Shine 1993).The overlap of prey species that differ only in size, such as L. latrans, could likely be explained by the differences in head, mouth and mandible structure sizes between E. p. sublineatus and E. j. jaegeri.Considering the condition of sympatry, high abundance and partial dietary superimposition of the subspecies herein studied, it is probable that the asymmetric competition could be attenuating the resource partitioning, allowing co-occurrence.This type of competition acts by the utilization of different strategies, microhabitats or any other mechanism that reduces the level of competition between the involved taxa (Ricklefs 2011).
In the present study, we verifi ed a divergence in the niche breadth between the two analyzed subspecies.The specialist E. j. jaegeri occupies a narrow niche, and the generalist E. p. sublineatus exhibits exclusive habits and forages on larger prey, which attenuates the competition between these congeneric sympatric forms.

Figure 1 -
Figure 1 -Means ± standard deviation of weight (in grams) (a) and snout-vent length (SVL) (in millimeters) (b) of Erythrolamprus jaegeri jaegeri and Erythrolamprus poecilogyrus sublineatus in the present study; means ± standard deviation of weights (in grams) and total length (in millimeters) of prey found in digestive tracts of the snakes are presented in the fi gures c and d, respectively.