Neogene sharks and rays from the Brazilian ‘Blue Amazon’

The lower Miocene Pirabas Formation in the North of Brazil was deposited under influence of the proto-Amazon River and is characterized by large changes in the ecological niches from the early Miocene onwards. To evaluate these ecological changes, the elasmobranch fauna of the fully marine, carbonate-rich beds was investigated. A diverse fauna with 24 taxa of sharks and rays was identified with the dominant groups being carcharhiniforms and myliobatiforms. This faunal composition is similar to other early Miocene assemblages from the proto-Carribbean bioprovince. However, the Pirabas Formation has unique features compared to the other localities; being the only Neogene fossil fish assemblage described from the Atlantic coast of Tropical Americas. Phosphate oxygen isotope composition of elasmobranch teeth served as proxies for paleotemperatures and paleoecology. The data are compatible with a predominantly tropical marine setting with recognized inshore and offshore habitats with some probable depth preferences (e.g., Aetomylaeus groups). Paleohabitat of taxa particularly found in the Neogene of the Americas (†Carcharhinus ackermannii, †Aetomylaeus cubensis) are estimated to have been principally coastal and shallow waters. Larger variation among the few analyzed modern selachians reflects a larger range for the isotopic composition of recent seawater compared to the early Miocene. This probably links to an increased influence of the Amazon River in the coastal regions during the Holocene.

Large specimens were collected directly from the outcrops, following the classical stratigraphic successions of the Pirabas Formation presented previously in several works [6,19,27,[53][54][55][56][57]. In addition, 30 kg of sediments were collected in the Atalaia section, screenwashed and sieved with 0.5, 1.0 and 2.0 mm open mesh-size, dried and picked under a stereomicroscope to examine the presence and relative abundance of microdental elements.
The fossiliferous localities of Sitio da Olaria, Sitio Pedro Teixeira and B-11 and B-5 quarries (Capanema Municipality) were destroyed by industrial mining activity, agriculture and urban development. As a consequence, only the specimens collected in the 1940s and 1950s were studied from the collections at the Museu de Ciências da Terra from the Companhia de Pesquisa de Recursos Minerais (CPRM) and in the Museu Nacional at Universidade Federal do Rio de Janeiro (MN UFRJ). All necessary permits for fieldwork, laboratory analyzes and descriptions conducted by the team from the Museum Paraense Emilio Goeldi and the Universidade Federal do Pará were provided by the Departamento Nacional de Produção Mineral (DNPM), which complied with all relevant regulations.
All specimens collected during this project are housed in the paleontological collection of Museu Paraense Emilio Goeldi (MPEG-V), Brazil. Specimen numbers are provided in the supplementary appendix with repository information of studied species (S1 Appendix). All specimens from the studies of Santos and Travassos [23], Reis [25], and Costa et al. [26], were reviewed and included in our study. Elasmobranch taxonomic classification follows Compagno [58,59] and Cappetta [42]; terminology is based on Cappetta [42]. Taxonomic identifications are based on an extensive literature review (e.g. [23,25,26,42,) and comparative analyses between fossil and extant specimens from the following collections: Departamento 52 selected fossil teeth of 10 selachian taxa were used for isotope analyses (δ 18 O PO4 ). The taxa and their isotopic values are shown in Table 1. To complement the study, fossil shark teeth (10 specimens of †H. serra) from proto-Caribbean Neogene deposits were also analyzed, serving as an additional comparative basis of prevalent tropical settings [50,51].
Teeth (n = 10) of the modern bullshark Carcharhinus leucas Müller and Henle 1839 [84] from the inner shelf of the Bragantina coast in the Pará state, were also analyzed ( Table 2). This species was selected due to its known long-term migratory habitat into estuarine river systems [85] and hence can be compared to the Amazonian fossils in terms of freshwater influence on marine waters.
Stable isotope analyses of the shark and ray teeth (n = 72, Tab. 1, 2) were done at the Stable Isotope Laboratory of the University of Lausanne, Switzerland. The focus was on the more resistant phosphate derived δ 18 O PO4 , however, the isotopic composition of the structural carbonate in apatite (δ 18 O CO3 & δ 13 C) was also analyzed. All teeth were cleaned in Milli-Q water in an ultrasonic bath to reduce sedimentary contamination. Preferentially shark tooth enameloid was sampled, but some amount of dentine could have remained in some fossil samples where the tip (apex of crown) of the small teeth was taken. The relative proportion of dentine

Geological setting (Pirabas Formation)
The stratigraphic sections (Fig 2) are characterized by massive mudstone with trace fossils, bioturbation, plant remains, pyrite concretions, and massive to laminated wackestones with Oxygen isotopic composition of extant shark from the Amazon Delta region (n = 10).
plant fragments. The wackestones and packstones/grainstones have low angle of cross-stratification; the hardground is rich in bryozoans, rudstones and contain broken or well-preserved invertebrate fossils. These facies and micro-facies were interpreted in the general context of representing marginal lagoon/mangrove, tidal inlets, and bioclastic bars/platform paleoenvironments. The marginal lagoon/mangrove consists of mudstone with pyrite concretions and plant remains with a thickness of about 80 cm (more restricted occurrence). Thalassinoides and Gyrolithes ichnofossils were found in the mudstone layer in a thickness level of about 50 cm. Massive wackstone with wavy laminations yielded both well-preserved and broken invertebrate fossils. The thickness of these layers ranged between 2 to 5 m. These facies associations were deposited in a paleoenvironment with low energy deposition by suspension in the limit between the oxic-anoxic zones, in agreement with the presence of pyrite.
The tidal inlet deposits are characterized by recurrent bioturbation in the first meter, wackestone with wavy lamination with up to 2 m; the fossiliferous packstones/grainstones has whole and fragmented invertebrate fossils arranged in recurrent beds with up to 2.5 m of thickness. This facies association represents moderate to high energy channels, dominated by ebb and flood tidal currents that were reworked continuously, sporadically sands were transported by currents and deposition by suspension occurred during low level stands.
The bioclastic bars/platform deposits are characterized by 70 cm-thick low angle crossstratified wackestones. Fossiliferous packstones/grainstones contain fragments or entire invertebrates in layers with up to 4 m of thickness. Grainstones and hardgrounds with 50 cm thickness exhibit abundant bryozoans and the rudstone beds, with up to 3 m thick, exhibit wellpreserved invertebrate fossils. This facies association represents moderate to high energy setting frequently reworked by oscillatory flow.

Fish assemblage
355 elasmobranch fossil teeth attributable to 24 taxa are identified (Figs 3-7, Table 3). The shark fauna is dominated by representatives of the Carcharhinidae Jordan and Evermann 1896 [87] (62.7% of the total assemblage), which are associated mainly with shallow water and nearshore environments. This family includes: Galeocerdo Müller and Henle 1837 [88] (Fig 3M-3Z), the latter being the most abundant taxon ( Table 3).
Tropical America analyzed in this work and from the literature, average δ 18 O PO4 values and triangle representing age; (c) Geographical map with fossiliferous units and locations that correspond to box numbers found at the side of each dataset: 1, 2 -Brazil, 3 -Venezuela, 4 -Colombia, 5 -Panama, 6 -Peru, 7 -Ecuador.
https://doi.org/10.1371/journal.pone.0182740.g008 The carbonate data clearly show discrimination related to different tissues analyzed, however this cannot be said for the more resistant δ 18 O PO4 data as shark teeth with or without some dentine content have similar average isotopic compositions. In this regard and as observed in previous researches, tissue discrimination is a stronger factor to influence carbonate isotopic compositions in phosphate than analyzing different taxa [29,37]. Moreover, the oxygen compositions of carbonate and phosphate are not correlated (Fig 10c). Considering the consistency of δ 18 O PO4 values checked by statistical tests, these data are considered for further ecological and paleoenvironmental discussions. The complementary dataset of South American sharks provided isotopic compositions slightly enriched ranging from 19.8 ‰ to 20.8 ‰ (n = 10), overlapping against the upper limit of δ 18 O PO4 values found for Pirabas fossil elasmobranchs. The following terms will be used in the discussion: "Pirabbean group" (Pirabas Formation elasmobranchs), "fossil shark group" (Pirabas Formation sharks), "fossil rays group" (Pirabas Formation rays) and "Recent shark group" (Recent Amazonian sharks).
Some species such as cf. Chiloscyllium Müller and Henle 1837 [88] (Fig 3A-3C), Nebrius Rüppell 1837 [137] (Fig 3D-3G), and Rhynchobatus Müller and Henle 1837 [88] (Fig 5S-5V), which are present in our fossil fauna, only have living counterparts in the eastern Atlantic and Indo-West Pacific (e.g. [136]). The presence of cf. Chiloscyllium sp. in the Pirabas Formation represents the first Neogene fossil record of this taxon in the Americas, as it was previously recorded from the Upper Cretaceous of North America and Trinidad [42,138]. The presence of Nebrius, Rhynchobatus [83], and now the cf. Chiloscyllium in the Miocene sediments of the Americas confirms that these taxa became extinct in the Western Atlantic and Eastern Pacific, possibly as environmental changes occurred after the definitive closure of the Isthmus of Panama (e.g. [83,[139][140][141]). With the exception of cf. Chiloscyllium sp., the remaining elasmobranch taxa of the Pirabas Formation has been found in other Neogene marine deposits of the Americas (e.g. [60,73,77,78,81,83]). This taxonomic commonality of the Pirabas Formation (Table 3) is better expressed by the nearby early Miocene assemblages from the Gatunian/ proto Caribbean bioprovince [83].
Within the prospections realized so far in the Pirabas assemblage, †C. ackermannii and †G. mayumbensis are the most abundant shark taxa ( Table 3). The fossil record for †C. ackermannii is restricted exclusively to a few full-marine early Miocene units of Brazil (Fig 11, S1 Fig) and Venezuela [83]. The fossil record of †C. ackermannii unknown in other Neogene units outside Tropical America would suggest that this species was endemic in the region during the early Miocene.
In contrast, †G. mayumbensis has been reported in the scientific literature from a few Miocene localities of Africa, Asia, North America and South America [42, 61,83,149,150]. The known fossil record of †G. mayumbensis [42, 61,77,83,149,150] suggests that this was a coastalpelagic species, with a widespread distribution in tropical environments and probably restricted to the early to middle Miocene.

Pre-Amazon delta
The shallow water Oligocene-Miocene platform of North Brazil was dominated by benthic carbonate producers, such as coralline red algae, bryozoans, crinoids, echinoids, mollusks and fishes [53]. A complex of faunal assemblages of marine micro invertebrates (e. g. foraminifera and ostracods), macro invertebrates (e. g. mollusks, echinoids, crustaceans) and vertebrates (fishes, reptiles and mammals) represented an area of high productivity in rocky reef-fringing reef complexes along the North and Northeastern Atlantic coast (Fig 12). The Amazon shelf, incised by a canyon during early to late Miocene, was favorable for the paleo-Amazon fan siliciclastic deposition [15]. The first Amazon fan may have covered an area of about 330,000 km 2 and the sediment depths accumulated may have approached 9 km [13]. Therefore, this siliciclastic input into the Atlantic coastal zone may have had a significant influence around the river mouth, causing the demise of the carbonate platform during the Plio-Pleistocene. The Pirabas carbonate platform was not exclusively affected by the first Amazon fan dynamics because the deposition area is further away from the mouth of the Amazon River. However, the siliciclastic Barreiras progradation during the middle Miocene to Pliocene (Barreiras Formation) progressively replaced the Pirabas carbonate platform [5].

A regional stable isotope signal?
The δ 18 O PO4 values of elasmobranch teeth represent an instantaneous record of water parameters in which the biogenic apatite was formed. Most biological groups that synthesize phosphate biominerals have a controlled mechanism with specialized proteins capturing ions rapidly, and chemical exchange of phosphate ions is negligible through inorganic process at low temperatures [151][152][153]. In a pioneer study, Longinelli and Nuti [154,155] and Kolodny et al. [151] noted that δ 18 O PO4 values of ectothermic fishes were correlated with ambient water isotopic composition (δ 18 O w ) and temperature. Since sharks and rays synthesize many teeth during their life, the δ 18 O PO4 value of each tooth should correspond to conditions of the aqueous environment where they lived at the given period of tooth formation [29]. Most shark and ray species commonly migrate at least short distances throughout their life, but even long ranging species tend to return or stay within their home areas, natal (birth) sites or other adopted localities [156][157][158][159][160][161][162][163][164] for extended periods prior to migration. It is possible that some of the analyzed specimens were regionally 'Pirabbean' given the abundance of nutrients and the presence of sheltered environments (shallow bays, river mouth regions) within the area that could support this hypothesis [27,53,55]. However, such high productive settings are also observed in fossil assemblages from adjacent regions (e. g., proto-Caribbean [60,[81][82][83]). A larger variation in δ 18 O PO4 values could be expected if these selachians migrated regionally (e. g., [48,50,51]), with higher values reflecting cooler waters while lower ones recorded warmer, tropical rather than sub-tropical waters. Data from other South American localities generally have higher, more positive values compared to those studied here (Fig 8b, [50,51]). This subtropical/temperate characteristic observed in Pisco (Peru) and Canoa (Ecuador) formations could be derived both from some transient taxa used in the analyzes (e. g. Carcharocles relatives), but as well due to a distinct global forcing influencing the specific climatic conditions between the Pirabbean and non-Pirabbean elasmobranchs. The Pirabas setting is typical for shallow and warm water masses with very little influence of deep-cold currents [27]. Meanwhile localities that are closer to the Pacific Ocean may have been subjected to important upwelling [50], and these cooler deeper waters may have spilled over into the proto-Caribbean until the Central Panamanian Seaway (CAS) closure (Fig 11, [28, 71,141,142,165]). Interoceanic (proto-Caribbean) Miocene †H. serra teeth from Venezuela (Cantaure, Caujaurao), Colombia (Jimol, Castilletes) and Panama (Chagres) deposits analyzed in parallel with Pirabbean samples again have higher δ 18 O PO4 values (mean: 20.1 ±0.4 ‰, n = 10, Table 1). Last but not least, inter-specific variability of δ 18 O PO4 values in extant specimens in South Africa [29] were up to 2.5‰, twice the values obtained here for the fossil specimens. Therefore, the δ 18 O PO4 values of Pirabbean samples correspond at least to a typical equatorial signal of paleoceanographic condition without upwelling influence integrated over 3 to 4 Ma.
The oxygen isotope data were converted to temperature using the equation of Lécuyer et al. [133] [T (˚C) = 117.4 − 4.5× (δ 18 O phosphateδ 18 O water )]. Seawater isotopic composition (δ 18 O w ) of -0.5 ‰ was used for the early Miocene samples [134] and 0.5 ‰ for the Recent samples [135]. The combined isotope and calculated temperature data are shown in Fig 8. The water column profile from the Amazon delta described by Moura et al. [166] shows nonplume and plume profiles, with consistent surface temperatures of about 28˚C, and below 90 m depth about 25˚C. The values obtained from the fossil and recent specimens here match well with temperatures observed in extant and fossil rays from low to mid-latitude waters [30,48,158,160,161,167]. Batoids had a slightly higher δ 18 O PO4 value (e. g., cooler temperature), that may be attributed to the demersal behaviour of these individuals, as recent relatives of the sampled specimens usually forage near the bottom for benthic invertebrates such as mollusks, as the most common prey in their diet [168][169][170][171]. Hence, the isotopic values of rays could reflect their ecology in inhabiting not only surface but also bottom water, with temperatures characterizing middle to lower limits of the Pirabas' waters.
Regarding the sharks, it appears that these still maintain the environmental preferences reflected in the paleontological record. Their δ 18 O PO4 values suggest paleotemperatures of 22˚C to 32˚C also noted for extant and fossil euryhaline sharks [29,33,38,48,172,173]. The higher variation present in the recent group may be attributed to the change in the regional hydrological system after the establishment of the Amazon delta fan. Karr and Showers [135] studied the oxygen isotopic composition of the open ocean Amazon shelf waters and found a large variation of up to 3 ‰ (-1 ‰ to 2 ‰) reflecting changes in seasonal runoff. As such the variations in the seawater isotope values are likely to be reflected in the δ 18 O PO4 values of the bioapatites [133,155]. Yet, in this study the ecosystem appears to be distinct from that of the Recent conditions. Only from the Plio-Pleistocene onwards an increased influence of the Amazon may have affected the inner shelf waters imposing a larger variation in the δ 18 O w driven by also seasonal cycle [13,135]. Nevertheless, the isotherms in Fig 8 still support that the δ 18 O PO4 values of all the elasmobranch groups are still well characteristic of marine ecosystems.

Ecological traits of Amazonian cartilaginous fishes based on stable isotope measurements
Although the average isotopic compositions of fossil shark are not significantly different, two end-members can be proposed, when compared pair-wised: †C. chubutensis and †C. ackermannii (t-test: t(7) = 2.42, p<0.045). Similarly, significantly different end-members can be recognized among the rays, on the same genus: Aetomylaeus. The end-members within the batoids include †Aetomylaeus cubensis and the other unassigned individuals of Aetomylaeus (t-test: t(11) = 2.81, p<0.016). This is possibly due to the different ecological niches (inshore vs offshore, Fig 9) occupied by these species. Furthermore, most genera overlap in their isotopic values indicating more generalist patterns like the tiger shark †G. mayumbensis, while others have a more specialized behavior or at least a preference to restricted niches (e. g., Rhinoptera). Therefore, small nuances measured in the δ 18 O PO4 values could be related to the ecological characteristics of the elasmobranchs.
Among the studied taxa, †C. ackermannii and †A. cubensis are probably representatives of an inshore/warmer predilection. Both have relatively low average δ 18 O PO4 values with low variance. It can be proposed that such sharks inhabited preferentially warm and coastal waters within a restricted habitat range, but still migrating occasionally as they also occur in other Neogene units of the Americas [42]. This behavior would be similar to extant Carcharhinus porosus Ranzani, 1839 [174] individuals, a small and short ranging shark very common in many coastal areas of tropical and subtropical waters in the Western Atlantic [136,[175][176][177]. Equivalent considerations can be said about †A. cubensis species, a taxon first observed in Central America by Iturralde-Vinent et al. [99]. The four tooth plates from this group have minor differences from the Aetomylaeus sp. group (n = 9). While the former have a lower variance and also mean δ 18 O PO4 value, the latter group recorded a higher average δ 18 O PO4 value (see Fig 9). Consequently, †A. cubensis could have had a peculiar shallower-inshore behavior, while the other group probably lived in colder or deeper waters. Two hypotheses may explain why the mentioned set of samples presented divergences. The first compares different species: extant Aetomylaeus usually occur in nearshore waters but are also present in variable bathymetric ranges, some preferring shallower intervals (e. g. Aetomylaeus maculatus Gray 1834 [178]), while others may occur in offshore settings up to depths of about 150 m (e. g., Aetomylaeus bovinus Geoffroy Saint-Hillare 1817 [179]) [180][181][182][183][184]. However, it is difficult to confirm this based on isolated teeth of the unassigned specimens, and precise identification would require tooth plates similar to †A. cubensis. In contrast, it is also possible that we sampled the same species but in different stages of life. No study is available referring specifically to dentition vs animal size for Aetomylaeus, however, taking into consideration comparisons of closely related myliobatoid crushing-like teeth vs adult size. †A. cubensis tooth plates are very large and probably reflect adult individuals of at least 1.5 m in total length (Fig 7C-7H) [42, 185,186]. The teeth of the other unspecified Aetomylaeus vary in sizes; generally not being as large compared to a single tooth from the plates of the other taxa and therefore could belong to smaller specimens or younger individuals (Fig 7I-7N). Given all these reasons it is possible that these larger rays were able to forage in shallower waters more frequently, being less susceptible of predation by sharks because of their size and therefore recorded lower isotopic compositions (e. g., warmer paleotemperature).
To represent the offshore predilection earlier proposed, Carcharocles transient shark has the highest mean value of the fossil shark group, which was expected considering the nature of the extant analogous species Carcharodon carcharias (or great white shark). They can occur at shallow inshore waters but are more common in the outer part of the continental shelf and remote oceanic islands. Moreover, these are one of the most wide-ranging fishes, migrating over thousands of kilometers through the ocean [177,187,188]. While migrating, long periods are spent in the pelagic habitat travelling across the ocean at depths down to about 1300 m, therefore the teeth analyzed may well have been formed in colder/deeper waters, providing higher mean δ 18 O PO4 values compared to other fossil shark taxa. Still, their isotopic values are within the total range of other resident selachian results (Table 1), and there is a high degree of site-fidelity in great white sharks and low interchange between populations aggregated at different coastal zones, even if their migration areas overlap for this species [177,189,190].
On that basis, we estimate that if elasmobranchian groups were not using the Pirabbean coastal waters as a protected site to give birth, the 'Blue Amazon' was still a valuable habitat for many species of this fishes' group. These inferences still need further investigation using statistical tests on larger datasets and estimating species size on the available groups; nevertheless, movement patterns and ecological characteristics of sharks can be applied to understand the nature of isotopic variations [33-37,86].

Conclusions
Taxonomic characteristics and oxygen isotope compositions of 72 teeth of sharks and rays were examined for sediments from the Pirabas Formation, Eastern Amazon, Brazil. A total of 24 taxa of sharks and rays were identified including a new fossil record for the American Neogene: cf. Chiloscyllium sp. Based on the phosphate bound δ 18 O PO4 values of biogenic apatites in many elasmobranch taxa three distinct groups were separated: a fossil shark group, a fossil ray group, and a group representing Recent sharks. Comparison between the fossil and Recent isotopic compositions led to interesting paleoecological propositions. Before the establishment of the Amazon fan, inner shelf water habitats are reflected by a smaller isotopic variation compared to the Carcharhinus leucas values. This divergency between isotopic compositions could be due to the coastal re-configuration with the contribution of Amazon River runoff to the Atlantic Ocean, imposing a higher outflow of 18 O-depleted water at the river mouth. The oxygen isotope approach used allowed the ecological traits between the investigated chondrichthyans to be divided into inshore or offshore habitat preferences. This approach suggests a shallow-water predilection for †Carcharhinus ackermannii and †Aetomylaeus cubensis, species known (so far) from the Neogene of Tropical America. Further work dealing with larger datasets for recent and fossil specimens can help to refine the proposed hypotheses. Nonetheless, the information presented here underlines the importance of a multidisciplinary approach to help understand past ecological dynamics of fishes.

Acknowledgments
We thank the Departamento Nacional de Produção Mineral (DNPM) authorities for permission to conduct the field trip. We are also grateful to Wolmar B. Wosiacki and Izaura Maschio for all assistance provided with the recent material for geochemical investigation. Thanks also to Charlie Underwood, Jan Fischer, Paola Rachello and an anonymous reviewer for the contributions to improve the manuscript. Special thanks to the Senior Editor Michelle Dohm for the manuscript handling.