INTRODUCTION
Trophic ecology is one of the outstanding features of several organisms, including marsupials’ biology, and it has major implications in the understanding of their natural history and on the functional role played by these animals (Castilheiro & Dos Santos Filho 2013; Camargo et al. 2014). Several didelphid species participate in mutualistic relationships in the Neotropics, since they feed on and disperse intact seeds of a wide variety of fleshy fruits (Cáceres 2002; Lessa et al. 2013). Furthermore, evidences suggest that didelphids’ feeding habits can change depending on age, sex, reproductive activity and season (Martins et al. 2006; Casella & Cáceres 2006; Santori et al. 2012; Lessa & Geise 2014a). Nevertheless, the knowledge about the diet and food strategies of most species belonging to the genus Monodelphis remains poor: only 29% of the :60 Brazilian didelphid marsupials had their diets investigated in specific studies (Lessa & Geise 2010; Santori et al. 2012).
Short-tailed opossum (genus Monodelphis Burnett 1830, Didelphidae) is the most speciose genus of didelphid marsupials; it has 22 currently recognized species (Pavan & Voss 2016). Monodelphis domestica (Wagner, 1842) is a terrestrial, diurnal and crepuscular didelphid and its representatives have the widest geographic distribution in the genus (Streilein 1982; Emmons & Feer 1997; Smith 2008). They are found in open dry and high-altitude rocky habitats (campos rupestres) in Central, Northeastern and Eastern Brazil, mainly in seasonal savanna-like biomes (Cerrado and Caatinga) (Macrini 2004; Melo & Sponchiado 2012).
Monodelphis domestica shows sexual size dimorphism with males being larger than adult females (mean body mass 71.4 g; range, 55 - 95 g) (Eisenberg & Redford 1999; Macrini 2004), this difference in size can result in different energetic demands, influencing the diet composition (Camargo et al. 2014; Melo et al. 2018). Like other Neotropical marsupials, M. domestica also shows a seasonal pattern of reproduction occurring in the beginning of the warm-wet season (Bergallo & Cerqueira 1994; Macrini 2004). During this period, males feed more intensively on insects because of higher energy demands related to mates searching, whereas females have a large energetic demand because they are pregnant or lactating (Martins et al. 2006; Camargo et al. 2014; Melo et al. 2018).
In addition, seasonal environments, such as the Cerrado (a savanna-like ecosystem), can also influence the species diet because the food availability (specially insects and fruits) can change during the year (Martins et al. 2006; Camargo et al. 2014; Lessa & Geise 2014a). Therefore, some didelphid species can adapt their diet in order to consume resources according to their availability in the environment (Lessa & Geise 2014a; Melo et al. 2018).
Monodelphis domestica, and Monodelphis species more generally, have traditionally been treated as insectivores (Paglia et al. 2012), although information about its food habits also reports the consumption of small vertebrates (rodents, lizards, frogs and snakes), fruits and carrion (Streilein 1982; Emmons & Feer 1997; Hume 1999; Neto & Dos Santos 2012). However, most detailed information about the diet of this species was reported by cafeteria experiments, in which M. domestica individuals were fed in captivity (Daniels et al. 2005; Constantino 2015). Recent studies on the diet of Monodelphis spp., based on fecal analysis and on stomach contents, showed that insects were the most frequently consumed resource (Casella & Cáceres 2006; Pinotti et al. 2011; Castilheiro & Dos Santos Filho 2013). These findings suggest that M. domestica may also be primarily insectivorous, although the available data are very limited. So far, we found no systematized and detailed studies about the composition of M. domestica natural diet.
This article reports the feeding habits of M. domestica specimens living in a seasonal high-altitude rocky habitat in southeastern Brazil. The aims of our study were to describe the species natural diet in the assessed area and to evaluate whether there are diet changes due to individual body mass, sex (male vs. female) and sampling time (dry vs. rainy season). We expect the following: (1) That M. domestica feeds more intensively on insects (mainly termites, ants and beetles) compared to fruits (see Casella & Cáceres 2006; Castilheiro & Dos Santos Filho 2013); (2) that vertebrate consumption would be positively influenced by M. domestica body size. Body size could be more associated with predation among didelphids than could be explained just by chance (see Santori et al. 1997; Macedo et al. 2010); (3) there are seasonal variations in the pattern of food consumption by M. domestica related to sex. The sexual dimorphism in size, associated with different energetic demands over the year, may exert influence on food consumption by M. domestica, reflected in differential diets for males and females (as suggested by other studies with small Neotropical didelphids in seasonal environments, see Martins et al. 2006; Camargo et al. 2014; Melo et al. 2018).
MATERIALS AND METHODS
Study site
We conducted the study in a high-altitude rocky habitat (campo rupestre) (18°14’S; 43°36’W; 1 387 m altitude) in the northern portion of Espinhaço mountain range, Diamantina County, Minas Gerais State, Brazil. Campos rupestres are ecotonal highland habitats, located between the Cerrado and Atlantic Forest domains. These rocky formations are influenced by a whole diversity of factors, including water availability, sun exposure, topography, as well as the type of soil. (Giulietti & Pirani 1988).
Based on Köppen classification, the climate in the site is of the Cwb type (mesothermal): mild and wet summers (October to March) and dry and cold winters (April to September) (Neves et al. 2005). However, rainfall rates throughout the sampling period did not match the expected standard for the study site; thus, we adopted the accumulated rainfall and evapotranspiration values (recorded during the sampling months) as the criterion to define the dry and rainy periods. We took into account the months when rainfall rates were higher than the evapotranspiration ones (November and December- 2015 and January, February, March, October, November and December – 2016) as rainy periods. Dry months were those when rainfall rates were lower than the evapotranspiration ones (October - 2015 and April, May, June, July, August and September - 2016). Rainfall rates are available in the National Institute of Meteorology – INMET (http://www.inmet.gov.br), data were collected in an automatic station in Diamantina County, MG (Station 83538, Latitude: 18.23, Longitude: 43.64, altitude: 1 296.12 m).
Sampling design and data collection
Opossum specimens were captured during five consecutive nights in a monthly basis. We used 100 Tomahawk traps (300 x 160 x 160 mm), which were set on the ground based on the capture-mark-recapture method from October 2015 to December 2016. As bait we used a mixture of banana, sardine oil, corn meal and oat grains. Captured opossums were identified, weighted, marked with numbered ear tags (Zootech® ) and released in the same location. Feces of each specimen were collected during manipulation procedures or inside the traps. Samples were stored in paper envelopes and preserved at -10°C in order to avoid fungi infestation.
The collected material was washed in metal mesh sieve (0.1 mm), separated and identified through stereomicroscopy in laboratory. Items were identified at the lowest taxonomic level possible and separated into 5 categories:
i) seeds, ii) flowers, iii) vegetative parts (leaves or stems), iv) arthropods, and v) vertebrates. The consumption frequency of each food category (expressed in percentage) was based on the number of samples presenting the category in question. Each fecal sample encompassed all feces produced by a single animal overnight.
Invertebrates in the samples were identified according to the specialized literature (Triplehorn & Johnson 2011; Constantino 2012; 2015) and through the comparison to an arthropod collection and to seeds collected in the same study site. Lizard and snake scales found in the feces samples were morphologically (number and position of mucros and sensorial holes) and morphometrically (total length, total width, mucro length and mucro-base width) evaluated. Next, they were compared to voucher specimens deposited in a scientific collection (Coleção Herpetológica do Semiárido – CHSA) and to information available in the literature (Rodrigues 1986; Franco & Ferreira 2002; Franco et al. 2017). We asked specialists to identify Isoptera and Hymenoptera fragments (Thiago Santos, Federal University of Vale do Jequitinhonha and Mucuri - UFVJM).
The project was approved by the Committee on Ethics in the Use of Animals - CEUA\UFVJM (protocol n. 12\2016) and by Chico Mendes Institute for Biodiversity Conservation (Instituto Chico Mendes de Conservação da Biodiversidade - ICMBIO), which issued permission to capture and handle opossums (license n. 52836-1). The trapping and handling procedures complied to the guidelines sanctioned by the American Society of Mammalogists (Sikes 2016).
Statistical analyses
We used the Relative Frequency of Occurrence (FO), expressed as the number of samples where an item was found (n) divided by the total number of samples and multiplied by 100, to determine the contribution of each item in the diet of M. domestica (Korschgen 1987).
We used the Levins’s standardized index (Krebs 1998) to calculate dietary niche breadth. This index ranges from 0 to 1: lower values indicate diets dominated by few prey items (specialist predators) and the higher ones appoint out more generalist diets (Hurlbert 1978).
Wherein, BA = Levins’ standard measure (from 0 to 1); B= Levins’ Niche breadth measurement (estimated through B= 1/ Ep2j; pj proportion of food item jin the diet; and n=number of consumed resources.
We used a Mann-Whitney test to evaluate possible differences in the body mass between individuals who had consumed or not vertebrate items. We tested the null hypothesis that medians body mass did not differ between these groups. For these analyses we used the software BioEstat 5.3 (Ayres et al. 2007). Although the age of captured individuals was not identified through the tooth eruption sequence (see Quental et al. 2001), median weight was the reference to define the adult specimens, since adult females weigh from 80 to 110 g and males from 80 to 150 g (Fadem & Rayve 1985).
We used the G-test (Zar 2010) to evaluate possible diet differences between the dry and rainy months and between males and females. These analyses were conducted in the BioEstat 5.3 software (Ayres et al. 2007). We used the Shannon-Wiener index (H’) to calculate food–item diversity in samples subjected to each treatment, i.e., dry and rainy months, and sex (males versus females). The Hutcheson’s t-test (Zar 2010) was used to evaluate the significance of differences recorded through the ShannonWiener index. It tested the null hypothesis that diet diversity did not differ between treatments. These analyses were conducted in the PAST software version 2.17 c. Significance level 5% was adopted for all analyses. Feces samples of recaptured specimens were excluded from the analyses in order to avoid pseudo-replications (Hurlbert 1984).
RESULTS
The 6,700 trap nights (900 trap nights in 2015 and 5800 in 2016) resulted in 1.33% capture success. We collected 70 fecal samples of 29 M. domestica specimens, 33 samples in the dry months and 37 in the rainy ones. Forty-one (41) samples belonged to male specimens (23 collected in the dry months and 18 in the rainy ones) and 28 belonged to females (10 collected in the dry months and 18 in the rainy ones), being one sample excluded because it had no sex identification. Males (n = 12; average weight ± sd = 89.33 ± 35.05 g) were heavier than females (n = 10; average weight ± sd = 48.70 ± 25.32 g) (t = -3.06; df = 20; p = 0.0062), being seven samples excluded from this analysis from those individuals which body mass was not recorded. Monodelphis domestica specimens presented specialized diet (BA = 0.24); they mainly fed on arthropods, which was the most frequent food category, since they were founded in 100% of the collected samples (Table 1). Hymenoptera (ants and bees), Isoptera (termites belonging to genera Nasutitermes and Neocapritermes) and Coleoptera (beetles) were the most frequent orders of arthropods. Diplopoda had an intermediate frequency, being followed by Blattodea, Aranae, Hemiptera, Lepidoptera, Odonata and Orthoptera, which was recorded lower frequencies (Table 1). Seeds from five plant families were recorded in 31.4% of the samples; among them, Asteraceae was the most frequent one (Table 1). The samples also had plant vegetative parts (leaves and stems) recorded in intermediate (22.9%) and flowers recorded in low frequencies (7.1%), respectively (Table 1).
Vertebrates were found in 14% of the samples. They were identified through the presence of scales, fragmented bones and eggs in the collected feces (Table 1). Scales belonged to the lizard species Eurolophosaurus nanuzae Rodrigues 1981 and Tropidurus montanus Rodrigues 1987 and to a small snake belonging to genus Thamnody nastes Wagler 1830. Monodelphis domestica samples also presented avian bones and egg fragments, which could not be identified under low taxonomic level because fragments were broken into small pieces that hampered identification. There was significant difference between the median body-mass (U = 18.5; p = 0.023) of individuals who consumed vertebrates (n = 6, median = 102 g, 5 males and 1 female) and that of the ones who did not consume them (n = 17, median = 45 g, 8 males, 8 females and 1 individual not sexed). Only 23 individuals who had their weight recorded were used in this analysis.
Diet significantly changed between climatic periods (G = 96.9361; df = 26; p < 0.0001); animals fed on a wider diversity of food items during the rainy months (H’dry = 2.09; H’rainy = 2.32; mboxtextitt = -2.64; df = 651.94; p = 0.0085). Some groups of items were only consumed during the rainy season: Lepidoptera, Odonata, Orthoptera, as well as some seeds such as Miconia sp. (Melastomataceae) and Solanum buddleiaefoliu (Solanaceae) (Fig. 1). Vertebrates were more frequent in the samples collected throughout the rainy season (21.62%) than in samples from the dry months (6.06%). On the other hand, Asteraceae seeds were more frequent (27.27%) in the dry months (Table 1).
The species also showed differences in the frequency of food items consumed by each sex (G = 119.25; df = 26; p < 0.0001): males had more diversified diet than females (H’males = 2.34; H’females = 2.02; t = 3.85; df = 655.14; p = 0.0001). Food items such as Solanum buddleiaefolium (Solanaceae), Blattodea, Hemiptera, Lepidoptera, Odonata were consumed only by males, while Miconia sp. (Melastomataceae) and Orthoptera were consumed only by females (Table 1). Vertebrates were more consumed by males (19.51%) than females (7.14%) (Fig. 2).
The diet of M. domestica also showed differences in diet composition (FO) between sexes in dry (G =155.11; df =15; p < 0.0001) and rainy seasons (G =165.41; df = 24; p < 0.0001). Males presented a more diverse diet than females during the dry (H’males = 2.05; H’females = 1.82; t = 3.20; ; p = 0.0015) and during the rainy season (H’males =2.49; H’females = 2.02; t = 6.45; df = 724.76; p < 0.0001).
DISCUSSION
In our study, M. domestica specimens presented a specialized diet feeding mainly on animal matter, including a wide variety of arthropods (e.g. Hymenoptera, Isoptera and Coleoptera), and small vertebrates such as lizards (e.g. E. nanuzae, T. montanus), snakes (Thamnodynastes sp.) and birds (unidentified). Overall, Neotropical didelphidmarsupials have been classfied as omnivorous andgeneralists, since they consume a wide variety offood items (Paglia et al. 2012). However, recent studies have shown that small-bodied open-habitat species (e.g. Gracilinanus agilis, Thylamys macrurus and M. domestica) tend to have more insectivorous and/or carnivorous diets (Camargo et al. 2014; Lessa& Geise 2014b; Melo et al. 2018, this study).
Monodelphis domestica inhabits seasonal savannalike biomes (Caatinga and Cerrado) presenting irregular rainfall in arid and semi-arid regions (Eisenberg & Redford 1999; Macrini 2004). In this sense, M. domestica appears to be ecologically comparable to equivalent insectivorous/carnivorous species from Australia’s xeric areas (Morton 1980) in its ability to maintain water balance in the production of metabolic water from a high-protein animal-matter diet (Christian 1983).
The presence in the diet of scales that belonged to small vertebrates such as the lizards (E. nanuzae and T. montanus) as well the snake Thamnodynastes sp., suggest their predation by larger-bodied M. domestica individuals, probably adults. In this sense, small vertebrates would only be accessible to adult short-tailed opossums, since according to Streilein (1982), captive and wild M. domestica can only capture preys who are almost similar to their size and their own body mass. Such result corroborates other studies, which reported that small vertebrates are more often eaten by adult than younger marsupials (Santori et al. 1997; Macedo et al. 2010; Santori et al. 2012), which would be expected since young individuals would not be able to subdue lizards or snakes bigger than themselves.
Food availability can explain the variations inthe diet composition of some didelphids (Cáceres 2002; Lessa & Geise 2010; 2014a; Melo et al. 2018),which are mainly opportunistic preferring items that are available over time and accessible in space (Santori et al. 2012). According to our results,Monodelphis domestica presented seasonal changes in diet, feeding more on arthropods, small vertebratesand on plant vegetative parts during therainy months. Seasonal shift in diet were foundin studies with other primarily insectivorous didelphid species distributed in savanna-like environments, such as Metachirus nudicaudatus (Lessa &Geise 2014a), Gracilinanus agilis (Camargo et al. 2014; Lessa & Geise 2010; 2014b), and Thylamys macrurus (Melo et al. 2018). Hymenoptera (ants), Isoptera (termites) and Coleoptera (beetles) are the most common insect orders distributed in Cerrado (Pinheiro et al. 2002), and these orders were also the most frequent found in M. domestica fecal samples throughout the year. This evidence suggests that diet of M. domestica in campos rupestres depends, at least partially, on the environmental availability of invertebrate preys. In fact, many other didelphids, known by their opportunistic feeding strategies, presented similar behavior (Streilein 1982; Cáceres 2002; Albanese et al. 2012; Lessa & Geise 2014a; Martins et al. 2006; Melo et al. 2018).
During reproduction, male mammals usually face high energetic costs in searching for a mate, while females usually face high energetic costs during pregnancy and lactation (Martins et al. 2006). In M. domestica, males usually are heavier than females (Macrini 2004) and increases home range during the breeding season in order to maximize the chances of finding and obtaining mates and food (Bergallo & Cerqueira 1994; Smith 2008). In turn, females are pregnant/lactating during the wet season (Bergallo & Cerqueira 1994; Cáceres & Graipel 2012), fact that increases their energetic demand during this time. According to Martins et al. (2006), is expected that males increased their food consumption at the end of the cool-dry season to store reserves to be used in searching for females and in mating. In Females of small mammals, higher energetic demand are associated with lactation (Thompson 1992 apud Martins et al. 2006), so that, females of M. domestica are expected to have an increased food consumption in the warm-wet season. However, our results do not support this pattern, since the diversity of food items detected in the diet of females was smaller than that detected in the diet of males in both seasons.
A possible explanation for this difference in diet between males and females could be explained by three hypotheses. First, it may be related to differences in energy demand over the year due to intrinsic reproductive cycle of each sex. Second, M. domestica is an opportunistic predator, so the spatial and temporal availability of food items in the environmental may reflect this variation in the diet. Third, as some food items were identified into wide taxonomic categories (mostly orders or families), it is possible that subtle differences in diet between sexes have been masked by the taxonomic categories used. Although we corroborate our hypothesis that diet differs between sexes, an analysis that measure the availability of food items in the environment at each season as well as the caloric content of the food resources consumed, would be needed to confirm their influence on energetic demand between males and females of M. domestica.
The herein assessed M. domestica population, which was distributed in Brazilian mountain rocky grasslands, presented primarily insectivorous/carnivorous diet. Our results demonstrated that body size may influenced diet composition (larger individuals were more likely to eat vertebrates than the smaller ones), diet changed throughout the year (it was more diverse in the rainy season), and feeding habits changed depending on sex (males had more generalist diets than females). This is the first study addressing the diet of didelphid species living in high-altitude rocky habitats, and our findings contribute to a better understanding about the food ecology of free-ranging short-tailed opossums.