First records of Brazilian three-banded armadillo (Tolypeutes tricinctus, Mammalia, Cingulata, Chlamyphoridae) predation by jaguar (Panthera onca, Mammalia, Carnivora, Felidae)

The jaguar Panthera onca and the Brazilian three-banded armadillo Tolypeutes tricinctus are two threatened mammals that coexist in the Caatinga dry forests and the Cerrado savannas of Brazil. Yet, to the best of our knowledge, interactions between these species have not been reported in the literature. Here, we present the first records of P.  onca predation on T. tricinctus from two different areas in the Caatinga in northeastern Brazil. We showed that P. onca can pierce the hard carapace of T. tricinctus, which may be possible due to its distinctly strong bite and associated predation behavior. We argue that P. onca may be the most adapted non-human predator to feed on T. tricinctus, and that the smaller body sizes of P. onca individuals in the Caatinga may increase their likelihood to feed on smaller prey, including T. tricinctus. Thus, the originality of our records is probably more related to insufficient research in the areas where these species coexist than to the rarity of this interaction.


INTRODUCTION
The jaguar Panthera onca (Linnaeus, 1758) and the Brazilian three-banded armadillo Tolypeutes tricinctus (Linnaeus, 1758) are two threatened mammals that coexist in the Caatinga and Cerrado domains in Brazil (Bocchiglieri et al., 2010;Astete et al., 2017). T. tricinctus is one of the two species of the genus Tolypeutes, along with Tolypeutes matacus (Desmarest, 1804) (Wetzel et al., 2008). These species are among the smallest Brazilian armadillos and are unique in presenting the ability to roll up their bodies into a ball (Wetzel et al., 2008). When rolled up, Tolypeutes protect their softer tissues with their hard, ossified tegument, thereby providing protection against predation (Wetzel et al., 2008;Superina & Loughry, 2012).
Tolypeutes tricinctus is the only armadillo endemic to Brazil and is restricted to the Caatinga and Cerrado domains (Feijó et al., 2015). The species is facing a significant decline in its populations due to hunting and habitat loss, which led it to be classified as Endangered in Brazil (Reis et al., 2015). Nevertheless, little is still known about its biology (Superina et al., 2014;Reis et al., 2015). For instance, as far as we know, there is no published study that presented evidence of T. tricinctus predation by other native predators, although it has been considered a potential prey for larger felids (Astete et al., 2017). This scenario led T. tricinctus to be classified as a top-priority armadillo species for research and conservation initiatives (Superina et al., 2014).
Panthera onca is the largest felid in America, occurring from south to north of the continent, where it inhabits numerous ecosystems (Sanderson et al., 2002;Morato et al., 2013). In Brazil, P. onca is found in almost all domains, including the Caatinga, where the species is classified as Critically Endangered as a result of habitat loss and hunting (Morato et al., 2013).

MATERIAL AND METHODS
The records presented here were obtained in two areas situated in the Caatinga, a semiarid phytogeographical domain predominantly covered by Seasonally Dry Tropical Forests (SDTF) distributed throughout northeast Brazil (Silva et al., 2017). We recorded the interaction once in the municipality of Brotas de Macaúbas (hereaf-ter, BdM), state of Bahia ( Fig. 1), while monitoring threatened felids between August and September 2017. At the occasion, we placed 20 camera traps in pairs at 10 sites throughout a 53-km² area composed of shrubby-arboreal SDTF, former croplands in regeneration covered by ruderal vegetation, roads, and a complex of three adjacent wind farms. The sampling lasted 15 days, and we set cameras to operate 24 hours a day, with one-minute intervals between consecutive 15-second-long videos.
The other record was opportunistically obtained at the Serra da Capivara National Park (hereafter, SCNP), a 1,291-km² Protected Area located in the state of Piauí, Brazil ( Fig. 1), during field activities to document the distribution of armadillos. The SCNP houses an important remnant of conserved Caatinga, comprising from sparse, bushy formations rich in cacti and bromeliads to dense arboreal SDTF (Olmos, 1992).

RESULTS AND DISCUSSION
The record from BdM was obtained on September 1 st , 2017, at 06:49 pm (Fig. 2). The camera trap site was in a  Different factors led to the differential diagnosis for P. onca predation for the damaged carapace. Some scutes of the scapular shield were broken (Fig. 3B), and they fit the canines of P. onca (Fig. 3C). When the carapace was collected in the field, these scutes constituted the only damaged region of the scapular shield; the other missing scutes were probably accidentally removed by posterior manipulation and transport or deterioration of the material. More distinct, the perforation in the pelvic shield fits P. onca canines (Fig. 3D), although some scutes are also missing. When the carapace is positioned half-rolled (simulating the defense mechanism) against a P. onca skull, the jaws fit both damaged areas while respecting the maximum mandibular opening (Fig. 3E).
Furthermore, P. onca has one of the strongest bites and the most robust canine teeth among the extant felids for their body sizes (Christiansen & Adolfssen, 2005;Christiansen, 2007;Meachen-Samuels & van Valkenburgh, 2009). These features are considered adaptations to prey on larger animals (Hayward et al., 2016), but also to attack by biting hard tissues, such as by braincase piercing or by cracking open the shells of tortoises (Meachen-Samuels & van Valkenburgh, 2009). Hence, they allow P. onca to prey on species that rely on armor as a defense mechanism (Miranda et al., 2016), which is the case of chelonians (Brito et al., 2018) and Tolypeutes spp. Consequently, P. onca is the most probable predator that could use a bite on the hard carapace of these armadillos to feed on them. In other words, while the hardness of Tolypeutes carapace may be efficient to avoid predation by other predators, it should be less efficient in avoiding predation by P. onca.
Such armor inefficiency is considered a possible reason why armadillos can be avoided by Puma concolor (Linnaeus, 1771), while not by Panthera onca where these predators coexist (Novack et al., 2005), which is the case of SCNP (Astete et al., 2017). Besides, armadillos are significantly more consumed by Panthera onca than by Puma concolor, probably for the same reason (Miranda et al., 2016). Nevertheless, T. matacus remnants have already been found in Puma concolor scats (Taber et al., 1997). Therefore, even though predation by other animals might occur, Panthera onca seems to be the most adapted predator to feed on Tolypeutes spp. and the most probable predator responsible for the perforations in the carapace that we found at SCNP (Fig. 3).
The two records we presented are complementary. The T. tricinctus individual in the video obtained in BdM was completely rolled up and thus, it may still be alive. Although this record does not show how and whether the predator fed on the prey or not, it demonstrates that P. onca exerts predation risk for T. tricinctus. The effects of predation risk should affect T. tricinctus population dynamics through the costs of antipredator behavioral responses (or risk effects), which can result in reduced survival, growth, or reproduction (Creel & Christianson, 2008). The record from SCNP presents the actual removal of T. tricinctus individuals from its population, and the record from BdM indicates that P. onca may capture T. tricinctus individuals at one place to feed on them at another.
The consumption of Tolypeutes by P. onca has already been reported through the presence of T. matacus remnants in P. onca scats (Taber et al., 1997;McBride et al., 2010). Thus, the records we obtained add to previously available information on this interaction, by demonstrating that it also occurs with T. tricinctus and how it may occur for both Tolypeutes species. We believe that the P. onca predation behavior we reported here should be valid for both T. tricinctus and T. matacus, as these armadillos share similar biology (Medri et al., 2011).
Panthera onca individuals tend to present smaller body sizes in areas where the mean weight of its consumed prey is lower (Hoogesteijnt & Mondollf, 1996). This appears to be the case of the Caatinga, where P. onca individuals present lower body mass than in floodplains and rainforests (Cláudia Bueno de Campos, personal communication; Hoogesteijnt & Mondollf, 1996), as it is the mean weight of their consumed prey . Although P. onca tends to avoid species as small as Tolypeutes spp. (Hayward et al., 2016), the smaller sizes of P. onca individuals in the Caatinga should reduce the individua's energy requirement, allowing them to feed more frequently (but not solely) on smaller species, especially if larger prey are less available . Those individuals might also be less capable of successfully slaughtering larger prey.
Armadillos are shown to be consumed by P. onca according to their abundance (or availability) (Hayward et al., 2016). This might explain why T. matacus has already been found in higher frequency than D. novemcinctus and E. sexcinctus in P. onca scats (Taber et al., 1997; McBride   et al., 2010). But, while it could be due to the higher availability of T. matacus, it could also be a consequence of the easiness for P. onca to prey on Tolypeutes spp., which is derived from its capacity to prey on armored animals (Meachen-Samuels & van Valkenburgh, 2009;Miranda et al., 2016;Brito et al., 2018).
The inefficiency of Tolypeutes defense mechanism against P. onca predation might also make slaughtering these armadillos less energetically costly for P. onca individuals than other armadillos that rely on fleeing and digging burrows to escape predation. Depending on the relative availability of these prey, P. onca predation on Tolypeutes spp. could be expected not to be uncommon, and especially on T. tricinctus in the Caatinga.
In conclusion, the originality of the records we obtained is more likely to be a result of the negligence-driven scarceness of scientific research in the Caatinga (Santos et al., 2011) as well as in areas of the Cerrado where P. onca and T. tricinctus coexist than the rarity of this interaction. However, different factors may influence its occurrence, such as the population densities of these species, which may decrease due to overexploitation and habitat loss, even leading to local extinctions (Morato et al., 2013;Reis et al., 2015). Overall, our findings allow this interaction to be assessed in future research on the ecology of P. onca and T. tricinctus without further uncertainty about its existence.