The materials used in this study come from Lower Permian deposits from Brazil (Iratí Formation) and Uruguay (Mangrullo Formation; Figure 1). The specimens from the Mangrullo Formation were found at the Paso del Cuello locality, Tacuarembó County, northern Uruguay, and are housed at the Vertebrate Collection of the Facultad de Ciencias (FC-DPV), Universidad de la República Oriental del Uruguay (Figures 1A–C,E–G), while the samples from the Iratí Formation were collected from outcrops in the municipality of Perolândia (Goiás State, Brazil) and are deposited in the Museu de Ciências Naturais of the Fundação Zoobotânica do Rio Grande do Sul, Paleovertebrate Collection (MCN-PV; Figures 1D,H,I). The specimens (Figures 2–9) consist of: (1) partially articulated mesosaurid individuals preserving gastric and gut contents, plus associated coprolites (Figure 2), (2) partially articulated individuals preserving only gastric contents (Figure 6), (3) isolated coprolites (Figures 3,4,7,8), and (4) regurgitalites (Figure 9). Some other mesosaurid materials from São Paulo State, belonging to the fossil collection of the Instituto de Geociências, Universidade de São Paulo, Brazil (GP-2E) were also studied by GP. The complete list of the examined 136 specimens is provided as Supplementary Information. Skeletons from Brazil have been mostly preserved as original bones, in place or slightly displaced. Partial skeletons from the Mangrullo Formation are mostly preserved as molds.

All identifiable skeletal remains from Brazil here studied were attributed to Brazilosaurus sanpauloensis by Raimundo-Silva et al. (1997) and Raimundo-Silva (1999) based on the criteria established by Oelofsen and Aráujo (1983); Oelofsen and Araújo (1987): adults developing less pachyostotic ribs, and hemal arches which are U-shaped due to their marked pachyostosis. However, these diagnostic characters turned out not to be useful, considering that the ribs are in fact wider than the radius, ulna, tibia, or fibula in specimens assigned to Brazilosaurus, and that U-shaped hemal arches can be present along with V-shaped ones in the tail of a single individual (see Piñeiro et al., 2015a for a revision of the Brazilosaurus status and general mesosaurid diversity). Most of the materials were collected ex-situ in abandoned and active quarries, when rock debris and blocks formed as the result of mining activities. There are also specimens that were collected in situ. The materials collected from the debris of quarries often represent single blocks fragmented into several parts, which are here referred to as portions A, B, C of the same specimen. Moreover, most of the studied specimens have split into part and counterpart. They consist of two complementary plates that will be referred to as A/A′, B/B′ and so on. For instance, in MCN-PV 2254, a partially articulated adult individual, part, and counterpart (MCN-PV 2254 A and A′; Figure 2) are differentiated in that one (the part) shows bones and the other (the counterpart) just molds. Other specimens consist of molds in both slabs due to loss by weathering.

The isolated bones, and the dispersed material around the adult individuals preserving gut contents, were measured and compared to the bones of the specimens studied by Sedor (1994) and Piñeiro et al. (2016) in order to determine their relative age and the minimum number of juvenile individuals associated with the adults in each of the samples (from Brazil and from Uruguay).

Determination of the crustacean species in the gut contents and inside the coprolites was difficult because of their weathering and high degree of fragmentation, but they might be juvenile pygaspids (Dr. Irajá Damiani Pinto, personal communication).

The methods for a more profound study of the gut and coprolite contents were restricted to scanning electron microscopy and the X-ray diffraction. Further, both gut and coprolite contents were compared under normal light microscopy, stereoscopic microscopy, and scanning electron microscopy. Thin sections were made at the laboratory of the Instituto de Geociências, Universidade Federal do Rio Grande do Sul (UFRGS), and the Paleontological Laboratory at the Museu de Ciências e Tecnologia, Pontifícia Universidade do Rio Grande do Sul (PUCRS) and photographed under a Zeiss microscope, at magnifications between 80 and 500 times. The thin sections were made following standard histological techniques (e.g., Ferigolo, 1985).

For study under scanning electron microscopy, cololite/coprolite contents were removed from the slabs using electronic drills, dental probes, stylets, and tweezers. Some samples were etched in 5% hydrochloric acid for 10 min, whereas others were treated with phosphoric acid for a variable time. After this, the samples were placed in an oven at 60°C for 24 h, in order to remove any moisture. The samples were glued on stubs with silver paste, metalized with 24 carat gold, and examined and photographed under a Jeol Scanning Electron Microscope (SEM), JSM-5200, under 25 kV with variable magnification. Other samples, mostly those specimens (coprolites and gastric contents) from the Uruguayan Mangrullo Formation, were also analyzed under SEM, but they remained uncoated.

All measurements are given in millimeters. The measurements taken were anteroposterior length and dorsoventral diameter of the cololites, and minimum and maximum diameter of the coprolites. They were taken using a Mitutoyo caliper with a span of 15 cm opening and accuracy of 0.05 mm. Bones were measured following Sedor (1994) and fragmented bones were measured by combining the measurements taken from parts and counterparts (A, A′).

Most of the specimens have an elongate ellipsoid, oval, or slightly curved shape (see Figures 3A,B,D,E, 5A). The specimen MCN-PV 2226 has a more rounded shape (Figure 3C). As a rule, coprolites have a reddish-brown color, and most of the exceptions have a lighter shade (Figure 4). Some, possibly due to diagenesis, have a slightly bluish (e.g., MCN-PV 2212 A/A′; MCN-PV 2220 A/A′) or brown-black hue (e.g., MCN-PV 2208, MCN-PV 2218 A/A′/B) and some show small bone inclusions (Figure 4A).

The coprolites are usually dispersed, or sometimes agglomerated, in slightly different planes. In the specimens in which there are a larger number of coprolites most are more or less parallel to each other, while some are oriented obliquely to the others (Figures 3D, 5A). Some specimens show coprolites overlapping each other (Figure 4B).

Some coprolites are complete (e.g., MCN-PV 2158 A/A′; MCN-PV 2254 A/A′), but some are broken and incomplete (MCN-PV 2158; MCN-PV 2210 A/A′; MCN-PV 2254). Most of them were split lengthwise by the fracture between the part and counterpart slabs (Figure 3). Some complete coprolites are well-preserved and visible on both slabs (e.g., MCN-PV 2102; MCN-PV 2231; MCN-PV 2254 A/A′). Others have only the thin outer layer preserved, and still others were completely lost so that just a slight halo remains. A few of them were in an initial stage of decomposition when fossilized, because their content is dispersed (Figures 3C, 4C).

Of the 220 coprolites studied, it was possible to establish both diameters of 107; only the largest diameter of a further six; and only the smallest diameter of the remaining 107. In the 113 coprolites for which the largest diameter could be measured, this parameter ranged from 9.8 to 32.5 mm (average largest diameter: 18.12 mm). In the smallest 214 coprolites for which the minimum diameter could be measured, this parameter ranged from 2.3 to 9.9 mm (average smallest diameter: 5.15 mm). Most coprolites have a largest diameter of 15.0–17.0 mm and a smallest diameter of 4.0–6.0 mm.

There was variation in content of the casting, not only between the different coprolites but between these preserved in the same slab. Four basic types were established: Type 1, only amorphous content; Type 2, with some crustacean fragments; Type 3, predominance of minuscule crustacean fragments; Type 4, crustacean and bone fragments and/or teeth (Figures 3–5).

MCN-PV 2220 is a coprolite containing an incomplete mesosaur tooth with a greatly altered enamel layer, which is opaque due to decalcification by the digestive process (Figures 5B,C).

Under microscopic stereoscopy, sometimes, the bone fragments appear very similar to fragmentary crustaceans, due to digestion. However, generally they are distinguished based on morphology, as the crustaceans are preserved as elongated, compact, fractured carapaces in continuity with each other. Other bone fragments are small polygonal pieces with rounded angles, due to digestion. Some of the best preserved of these were selected for further study under scanning electron microscopy, such as the specimen MCN-PV 2227 (Figure 5D).

Mesosaur bromalites tell us quite much about the ecological (trophic) interactions between the components of the peculiar “Mesosaur community” (Piñeiro et al., 2012b). As expected, mesosaurs fed on pygocephalomorph crustaceans, possibly as their preferred prey. They appear to have been size-selective, eating crustacean prey not exceeding 20 mm in length (Ramos, 2015). This can be asserted from almost all the coprolites and from all the studied gastric contents. We also show that small scattered mesosaur bones are present in the cololites, and exceptionally in coprolites (e.g., teeth), suggesting that these skeletal pieces must have been ingested. However, there is no evidence that large mesosaurs preyed on small ones, even though mesosaur bones are common in stomach contents and coprolites (see Table 4). Schwimmer et al. (2015) described partially articulated vertebrae of a baby turtle in a spiral bromalite (coprolite or cololite) of a shark. In contrast, bone elements present within cololites and coprolites in the material from the Mangrullo and Iratí Formations show strong acid etching and many fractures, and they are never articulated. In a hypothetical scenario of predation, we would expect to find more or less articulated skeletons in the gastric contents (where the digestive processes would still allow for a general identification of the bones). Mesosaur teeth seem not to be adapted to powerful biting (although see Pretto et al., 2014) and the jaw aperture in an average-sized mesosaur is much too small to allow even small newborn mesosaurs to be swallowed whole (see Figure 12). Thus, we support instead the hypothesis that mesosaurs were scavengers and fed on soft tissues of smaller decaying individuals. Some bones could have been accidentally ingested along with soft tissue. This could be why some of them are so eroded that they are difficult to identify. As the soft tissues are partially decomposing, the teeth need not be so strong and fixed to the jaws, explaining the high number of them preserved separately from the rest of the skeleton and the frequent occurrence of toothless skulls. Another hypothesis to explain the presence of mesosaur bones in the gastric contents and coprolites is that they were eaten accidentally, during the capture of pygocephalomorph crustaceans that were scavenging on the carrion accumulated on the bottom.

Teeth are among the skeletal elements that are most resistant to dissolution, although extant crocodiles normally decalcify ingested teeth (Fisher, 1981a,b). This may explain why the teeth found in the regurgitalites and gastric contents of mesosaurs tend to be better-preserved than bones. Interestingly, if tooth fragments found in the samples derive from prey, mesosaurs must have swallowed at least some cranial elements during acts of predation. Such a behavior is known in modern crocodiles, although very large prey needs to be dismembered first (Grigg and Gans, 1993; Webb and Manolis, 2009). But, in the case of mesosaurs, the evidence suggests that partial decomposition of the consumed carrion facilitated the presence of teeth in gastric contents and coprolites. Alternately, the mesosaurs might have sometimes swallowed their own teeth when they broke during feeding. Such tooth damage is known in some archosaurs, which are often characterized by fast tooth replacement (Grigg and Gans, 1993) as mesosaurs were Pretto et al. (2014). Stone et al. (2000) described from the Upper Jurassic of the Morrison Formation a large bone-rich phosphatic mass interpreted as a coprolite, which contained a broken distal end of an Allosaurus tooth. The authors suggested that the tooth broke and was accidentally ingested when the theropod used its jaws to break bones into smaller portions (Stone et al., 2000). Some mesosaur coprolites were found to contain teeth that could have either belonged to very young individuals or been recently erupted replacement teeth of adults, which are also small and short.

Based on the analyzed material, it is hard to determine whether the mesosaurids consumed carrion of the same or another species. Distinguishing between mesosaur species is a very complex issue, even in the case of well-preserved individuals (see Piñeiro et al., 2015a). Therefore, we prefer to identify mesosaurs as scavengers rather than cannibals. The largest crustacean remains (e.g., abdomens and some thoracic segments) often appear to be more weathered than expected, and thus it seems possible that they were also scavenged from the bottom. Moreover, some possible coprolites/regurgitalites resemble the crustaceans in shape and some aspects of internal structure (Figure 3D). It would be conceivable that crustaceans larger than 20 mm in length were swallowed whole, but were soon regurgitated in an incompletely digested state because of their large size. The fossil record contains some previously documented examples of that kind of inappropriate predator-prey interaction, in which a predator and an excessively large prey animal both died after the former swallowed the latter (see Bieńkowska-Wasiluk, 2010). Alternatively, the crustaceans may have been expelled as feces (see Figure 13).

The presence of apparent bone inclusions in the coprolites (Figure 4A) is intriguing. Reptiles are characterized by a long period of digestion and, for example, crocodiles digest the bones they ingest practically completely (Milàn, 2012; see Owocki et al., 2012). However, mesosaurs were fairly small; while large snakes may defecate even less than once a year, gracile arboreal snake species, which have low body mass, sometimes defecate within 24–48 h of ingesting a meal (Lillywhite et al., 2002). Relatively large bones are seen also in the putative regurgitalites, which also contain teeth of varied size and fragmentary bones that are difficult to identify, although they could be ribs or other long bones (Figure 9). Mesosaurs probably regurgitated bone fragments, perhaps consumed accidentally, which were too large to pass through the gastrointestinal tract. For example, raptor birds (e.g., Bond, 1936; Dauphin et al., 2003; Terry, 2004) and crocodiles (e.g., Grigg and Gans, 1993; Webb and Manolis, 2009; Semeniuk et al., 2011), as well as probably ichthyosaurs (Doyle and MacDonald, 1993; Nature News, 2002; Thies and Hauff, 2013), regurgitate most of the indigestible or hard-to-digest remains. Regurgitation might also have been caused by some kind of environmental stress, as suggested by the conditions under which the Mangrullo Formation is thought to have been deposited. Because digestion efficiency depends on body temperature in extant reptiles, undigested remnants of prey may be regurgitated during periods of unfavorable environmental temperature, as exemplified by crocodiles (Verdade et al., 1994) and snakes (Hailey and Davies, 1987), and regurgitation may also be caused by disease (Lock et al., 2003). Also, the presence of poorly digested remains in feces, caused by a short food-retention time in the gastrointestinal tract, may have to do with fluctuating food availability (see Chin et al., 2003).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.