Specimens

The specimen DGM 921-R consists of a right partial maxillary preantorbital body with teeth in loci (Figs 2, 3) and a right posterior region of a dentary (Fig. 4). As recorded in the catalogue book at Museu de Ciências da Terra (MCT) in Rio de Janeiro, both elements are considered to belong to a single individual, having been found closely associated in the same locality and horizon. The studied material had originally been housed at the MCT, currently part of the Serviço Geológico do Brasil–CPRM, and was on loan to the Museu Nacional-Universidade Federal do Rio de Janeiro (MN-UFRJ) for study. After the great fire that destroyed the main building of the MN-UFRJ in September 2018 (e.g. Kellner et al. 2019a), however, the specimen was not recovered and is here considered as lost.

Nomenclatural acts

The fossil material designated as the name-bearing type of a new genus and species is considered lost. Despite being uncommon and previously unpermitted, designation of lost specimens as name-bearing types is currently permitted by the ICZN under specific circumstances and is not unprecedented.

Our act is in accordance with Declaration 45 of the ICZN (2017), which allows the designation of a lost specimen as a name-bearing type if it had been, originally, adequately collected and housed in a collection and only then lost; and if the reasons for accidental specimen loss are properly justified. This was the case, for example, of the sauropod Maraapunisaurus fragillimus, whose name-bearing type specimen is the famous holotype of Amphicoelias fragillimus, described by Cope (1878) and subsequently lost. Recently, reassessment of the available data on this specimen led Carpenter (2018) to name it as the new rebbachisaurid genus Maraapunisaurus, invoking Declaration 45 of the ICZN (2017). Similarly, Apesteguía (2007) named the rebbachisaurid Nopcsaspondylus alarconensis when the holotype specimen was already lost, subsequent to its first description (without naming) by Nopcsa (1902). It is important to note that Declaration 45 of the ICZN (2017) does not permit the naming of new taxa based on photographs of specimens that have never been safely stored in a public scientific collection. In addition, the specimen described here (DGM 921-R) was also under study at the time of the MN-UFRJ fire (Holgado et al. 2018), and the manuscript was about to be submitted when the accident occurred.

Anatomical nomenclature and measurements

We follow Hendrickx & Mateus (2014) for general anatomical maxillary nomenclature and measurements, and Hendrickx et al. (2015) for dental nomenclature and measurements. The measurements of Buitreraptor gonzalezorum and Austroraptor cabazai were done using ImageJ (Schneider et al. 2012), from images available in the literature (Gianechini et al. 2011; Gianechini & Apesteguía 2011; Gianechini et al. 2017).

Heuristic phylogenetic tree search

The phylogenetic analysis was performed based on the data matrix from Cau et al. (2017), with the inclusion of the specimen DGM 921-R, a new character, and a few codification changes (Brum et al. 2021). The phylogenetic analysis was performed using TNT version 1.5 (Goloboff & Catalano 2016). The memory available for the analysis was 1000 MB, which enabled the program to hold 600 000 trees. A search for the minimum-length trees (MLTs) was conducted using the TNT algorithms Sectorial Search, Ratchet (parameters: 25 substitutions made, or 99% swapping completed, six up-weighted probabilities, six down-weighted probabilities, and total number of iterations = 10) and Tree-fusing in a driven search (initial addseq = 15, find the minimum tree length 50 times), with random seed equal to 0, and without collapsing the trees after the search. The aim of the first analysis was to recover the island with the most MLTs. Subsequently, the recovered island was used as the starting point for a traditional search (tree bisection and reconnection (TBR) swapping algorithm; starting with trees held in the RAM) and without collapsing the trees after the search. The aim of this second analysis was to recover the maximum number of MLTs from the island recovered in the first analysis. The MLTs were collapsed using ‘condense 1; collapse;’ and the script ‘STATS.RUN’ was used to retrieve the retention and consistency indexes (RI and CI, respectively). A strict consensus was then generated for the recovered MLTs. It is important to note that as a consequence of the size of the matrix and memory limitation of the computer, the generated consensus is not the most complete (i.e. it was generated based on all MLTs available for the best recovered island) for this data matrix, but it is still more exhaustive than previous analyses. All characters were assigned equal weights and no additive characters were considered.

Tooth assignment

We conducted analyses of the teeth of DGM 921-R to obtain more details on the tooth morphotypes in theropods and to evaluate the morphological and morphometric details. Such analyses followed the protocols commonly used in the assignment of isolated teeth (e.g. Delcourt & Grillo 2018; Young et al. 2019; Delcourt et al. 2020; Hendrickx et al. 2020).

Parsimony analysis of endemicity

To discuss the biogeographic relationships between the geological formations and their fauna, we carried out a parsimony analysis of endemicity (PAE; Rosen 1988; Morrone 1994), based on the cladistic analysis of presence and absence data matrices of supraspecies taxa (Morrone 2009; Morales-Guerrero et al. 2017), which allow the identification of biogeographical assemblages and their hypothetical relationships. These biotic components are graphically represented as areas of endemism (Morrone 2009).

We considered only Upper Cretaceous formations to establish the study areas in this work. The poorly known and/or poorly sampled taxa were grouped into the most inclusive taxonomic unit (e.g. Podocnemididae, Squamata). We considered only tetrapod vertebrates for analysis, and excluded the teeth and ichnological data. In addition, solely osteological records confirmed in the literature were chosen for the analysis (e.g. we excluded the dubious assignment of the turtle record to the Santo Anastácio Formation; Menegazzo et al. 2015; Romano 2016). We also considered only the formations that crop out (i.e. subsurface formations such as Santo Anastácio were not considered). The data matrix (composed of one hypothetical outgroup area, 14 formations and 25 taxon characters; see Brum et al. 2021) was edited using Mesquite version 3.6, and the PAEs were conducted using TNT version 1.5 (Goloboff & Catalano 2016).

Description

Comparisons

The specimen has a thecodont ziphodont dentition, which is found in theropods, sphenacodontids, basal archosaurs and crurotarsians (see Hendrickx et al. 2015). However, the occurrence of unserrated ziphodont thecondont teeth is exclusive to some small theropods (e.g. unenlagiines, compsognathids). The mesial shelf perpendicular to the maxilla, as well as the dorsoventrally high maxilla, also indicate that DGM 921-R belonged to a theropod. The medial surface of the maxilla DGM 921-R is vertically oriented, implying the presence of a dorsally recessed palate relative to the horizontal plane of the tooth line, as commonly found in theropods (e.g. Tsuihiji et al. 2014), and in contrast to crocodilians and pterosaurs, in which the palate is level with the tooth line (e.g. Witton 2013; Pol et al. 2014). Therefore, based on these dentition characteristics, our cladistics, and morphometric analyses (see further), we propose that these bones belong to a small-sized theropod and unenlagiine. We compared the preserved portion of the maxilla and dentition of DGM 921-R with A. cabazai and B. gonzalezorum, the only unenlagiine specimens to have preserved cranial remains, and with other small-sized theropods (Fig. 5). The proportions of both the maxilla and dentary (e.g. the height and thickness) are consistent with what is found in unenlagiines.

The maxilla in the studied specimen is dorsoventrally low. Although the maxillary preantorbital body is not preserved in B. gonzalezorum, the low anterior margin of the antorbital fossa suggests that it is also dorsoventrally low, unlike that observed in A. cabazai. Although Gianechini et al. (2017) reported wide spacing between the teeth in the middle of the maxillary tooth row, which is not preserved in DGM 921-R, we could assume that it was present in DGM 921-R because of the wide spacing observed in the preantorbital body fragment of DGM 921-R. In both DGM 921-R and B. gonzalezorum, the spacing between alveoli/width of alveoli ratio is ≥1, which also occurs in Archaeopteryx and Anchiornis, but not in A. cabazai (Table 3). The lateral surface of this part of the maxilla is flat, which is also found in A. cabazai. The lateral surface of the maxillary preantorbital body in DGM 921-R lacks numerous neurovascular grooves, a lack that is also commonly observed in dromaeosaurids (e.g. Lü et al. 2010; Evans et al. 2013). Such a feature is hard to observe in the lateral surface of A. cabazai due to the poor preservation of the bone surface.

Based on the diagnosis provided by Gianechini et al. (2017), focusing on the cranial elements of B. gonzalezorum, DGM 921-R and B. gonzalezorum are similar in their small teeth, which lack carinae and denticles and the constriction between the crown and root, and which have crowns exhibiting parallel longitudinal striae (Gianechini et al. 2011). However, DGM 921-R differs from B. gonzalezorum in that the teeth of DGM are less compressed (DGM 921-R CBR between 0.7 and 1, B. gonzalezorum CBR <0.5). Small teeth also occur in A. cabazai, and this is considered a homoplastic condition that is present in alverezsaurids, ornithomimosaurs, troodontids, and basal avialians such as Archaeopteryx and Jeholornis. The absence of denticles in the teeth is synapomorphic for Unenlagiinia, while the presence of two or more longitudinal grooves is synapomorphic for Unenlagiinae (Fig. 3; Gianechini et al. 2011). The absence of carinae in the teeth is also shared by A. cabazai, as is the fluted condition (Novas et al. 2009; Gianechini et al. 2011). However, the ziphodont condition resembles that observed in B. gonzalezorum rather than in A. cabazai, which presents a conidont condition.

The absence of denticles on both mesial and distal edges is also seen in specimens attributed to the Laurasian wastebasket genus Paronychodon, such as AMNH 27127B, NMMNH P-30218, NMMNH P-30233, RTMP 2000.19.2 and RTMP 88.211.66 (Sankey et al. 2002; Hwang 2005; Williamson & Brusatte 2014). Notwithstanding, AMNH 27127B and RTMP 2000.19.2 differ from DGM 921-R and unenlagiines in exhibiting a constriction between the crown and root, typical of troodontid teeth (Makovicky et al. 2003; Hwang 2005). Furthermore, NMMNH P-30218 and NMMNH P-30233 differ from unenlagiines in exhibiting both mesial and distal carinae (Williamson & Brusatte 2014). Finally, even though RTMP 88.211.66 presents lateral longitudinal grooves on the crown, it differs from unenlagiines in tooth morphology, density and location (Sankey et al. 2002; Gianechini & Apesteguía 2011).

The dentary of DGM 921-R shows a constriction of the alveolar margin, with a marked Meckelian cavity. The development of the Meckelian groove is similar to that observed in the referred dentary material of Rahonavis ostromi Forster et al., 1998 (see Forster et al. 2020) and the derived paravian Archaeopteryx and Anchiornis (Pei et al. 2017), in which the groove runs conspicuously along the length of the medial surface of the dentary. It differs from that observed in A. cabazai by being wide and by beginning in the middle portion of the dentary. In DGM 921-R the Meckelian groove is more constricted. Therefore, the constriction of the alveolar margin with the Meckelian cavity indicates that it is the posterior portion of a dentary, posterior to the antorbital fenestra. The lateral surface of the dentary lacks any sign of a longitudinal groove or a nutrient groove, the lack of which is also observed in B. gonzalezorum and A. cabazai, and which has been previously recovered as synapomorphic to Troodontidae (Gianechini et al. 2017). However, the dorsal half of the dentary is not preserved and we are therefore unable to assert that this feature is lacking. The ventral margin is thin and straight, and forms a triangular cross-section, as in the paravians R. ostromi, Archaeopteryx and Anchiornis. Although the posterior portion of B. gonzalezorum is not preserved, the preserved regions indicate that the ventral margin is straight. The dentary of DGM 921-R lacks any signs of interdental plates. This feature was recovered as synapomorphic for the Maniraptoriformes, but these structures are present in A. cabazai (Gianechini et al. 2017). Among the unenlagiines, the absence of dentary interdental plates resembles the condition observed in B. gonzalezorum (Gianechini et al. 2017).

Tooth assignment

Heuristic search

The heuristic search resulted in MLTs with tree length = 1072, CI = 0.246 and RI = 0.588. Although we obtained distinct topological resolutions, in both the TBR + TNT algorithms and the TBR heuristic searches Ypupiara was recovered as a sister-group of Buitreraptor. The assignment of Ypupiara + Buitreraptor was also stable (Fig. 7; for full topologies, see Brum et al. 2021). The TBR + TNT algorithms analysis resulted in 33 520 MLTs, while the TBR analysis resulted in 360 MLTs. No autapomorphies were recognized for Ypupiara. The TBR + TNT algorithms analysis was marked by poor resolution, and resulted in a polytomy with 61 theropod taxa and 8 groups (including the group Ypupiara + Buitreraptor). The TBR synapomorphies resulted in a well-resolved topology.

The characters that group Ypupiara + Buitreraptor in the TBR + TNT algorithms analysis are as follows (* indicates that they were recovered only in some trees; characters are based on Hendrickx et al. 2020): mesiodistal curvature of the labial surface of the crown at 1/3 of the central region of the crown surface, with the labial depression restricted to the crown base (char. 74(0 > 3*)); teeth with flutes present on the labial surface or on both the labial and lingual surfaces (homoplastic with Sinornithosaurus and spinosaurids; char. 111(0 > 2)); single longitudinal groove centrally positioned on the labial and/or lingual surface of the crown (also occurs as a polymorphism in Byronosaurus; char. 119(0 > 1)); single ridge centrally positioned (differing from flutes) on the lingual surface of the crown (also occurs as a polymorphism in Byronosaurus; char. 120(0 > 1)). The synapomorphies that group Ypupiara + Buitreraptor within the major group are: CH < 10 mm (also present in Eoraptor and in the group formed by Jianchangosaurus, Erlikosaurus, Falcarius, Eshanosaurus, Zanabazar, Byronosaurus, Sinusonasus, Almas, MCPD 1001128, Mononykus, Shuvuuia and Nqwebasaurus; char. 69 (01 > 0*)); absence of mesial carina (also occurs in Orkoraptor, Halszkaraptor and in some trees in the group formed by Zanabazar, Byronosaurus, Sinusonasus, Almas, MCPD 1001128, Mononykus, Shuvuuia and Nqwebasaurus; char. 77(01 > 1*)); absence of distal carina (also occurs in Halszkaraptor and in some trees in the group Suchomimus and Baryonyx; char. 79(0 > 1*)); texture of the enamel surface smooth or irregular (present in some trees in the group Majungasaurus, Rugops and Kryptops, as in the Eoraptor + some maniraptorans; char. 121(01 > 0*)); lateral view of the root showing convex margins, with a root significantly larger than crown base (also present in Dilong, Troodon, Jianchangosaurus + Erlikosaurus + Falcarius + Eshanosaurus and Almas + MCPD 1001128; char. 143(0 > 1*)).

The characters that grouped Ypupiara + Buitreraptor in the TBR analysis were: CHR < 2 in mesial teeth (char. 42(2 > 3) is not scored in Ypupiara); and the mesiodistal curvature of the labial surface of the crown at 1/3 of the central region of the crown surface, with the labial depression restricted to the crown base (char. 74(0 > 2*)). This analysis recovered less support for the group Ypupiara + Buitreraptor and one synapomorphy is not scored for Ypupiara: that character 74(0 > 2) is scored as 74:3 for both Ypupiara and Buitreraptor. Therefore, we consider that the result recovered in the first analysis better represents what is observed in both genera. Regarding the plesiomorphic characters that grouped both taxa in the major recovered group (Fig. 7), they consist of characters 69(1 > 0); 77(0 > 1*); and 143(0 > 1). We note that the characters recovered for this group were considered as more derived in the TBR + TNT algorithms analysis.

Linear morphometric analyses

The first LDA conducted on the whole dataset showed that the teeth of Ypupiara, as well as those of the Bauru dromaeosaurid specimens MPMA-12-00D13-97 and MPMA-12-00D14-97, occupy the morphospace of small-sized theropods (LD1 (first linear discriminant axis), 51.93%; LD2, 18.21%; see Brum et al. 2021). In the LDA classification the Ypupiara teeth are classified as Therizinosauria and Spinosauridae, while the Bauru specimens occupy the Neovenatoridae morphospace (Brum et al. 2021). However, the morphospaces of these specimens overlap, therefore we performed a second LDA of taxa with small crowns (CH < 20 mm). Although both Ypupiara and the Bauru specimens occupied morphospaces in which we observe troodontids, dromaeosaurids, therizinosaurs, Coelophysis, Megaraptor and non-tyrannosaurid Tyrannosauroidea, the Ypupiara and MPMA specimens are distinct from each other (LD1, 28.92%; LD2, 20.03%; Fig. 8). The teeth of Ypupiara are in a region of overlap of the morphospaces of dromaeosaurids (except Buitreraptor), troodontids, noasaurids and non-tyrannosaurid Tyrannosauroidea. Regarding the MPMA specimens, they were located in the morphospace of non-tyrannosaurid Tyrannosauroidea. The classification assigns the teeth of Ypupiara to Masiakasaurus, while MPMA-12-00D13-97 is assigned to Liliensternus and MPMA-12-00D14-97 to Gualong (success rate of 60.78%). The PCA (see Brum et al. 2021) shows that the main measurements that influence the variance observed in the genus morphospaces are CBL, CBW, CH and AL.

The classical hierarchical and the joining clustering analyses produced similar morphoclades for the placement of the teeth of the Ypupiara and MPMA specimens (Brum et al. 2021). Two teeth of Ypupiara occupy a morphoclade that is mostly composed of troodontids, but which contains some dromaeosaurids. One tooth of Ypupiara, which is the most anterior, was different to the others and occupies the morphoclade shared with Eoraptor, Dilong and the dromaeosaurid Saurornitholestes. Regarding the MPMA specimens, they occupy a different morphoclade to that of Ypupiara, with both MPMA specimens in a morphoclade that mostly includes dromaeosaurids, Megaraptor, non-tyrannosaurid Tyrannosauroidea, Masiakasaurus and basal theropods.

Results of the parsimony analysis of endemicity

An exhaustive search was able to be performed, due to the small taxa–character data matrix, and the final cladogram is a consensus of nine trees found with 60 steps (Fig. 9). As already established in the literature, when taxa are coded as zero this implies that such taxa are absent and excludes dispersal a priori (Morrone 2014).

Then the cladogram divides into two major branches, with the Argentine formations in one and the Brazilian formations in the other (Fig. 9). The exception is the Bajo de la Carpa Formation (Santonian), which forms a polytomy with the Adamantina (Coniacian to late Maastrichtian, sensu Castro et al. 2018) and Marília (Maastrichtian) formations.

In the two resulting MLTs, the relationships in the clade (Marília + Adamantina + Bajo de la Carpa formations) are sustained by the presence of the advanced notosuchian, non-Unenlagiinae maniraptorans and avialans. The presence of Unenlagiinae taxa is recovered as a synapomorphy in the clade (Candeleros + Portezuelo formations), and as an autapomorphy (homoplasy) in the Allen Formation. Despite this, the Unenlagiinae presence in the Marília and Adamantina formations was not detected as an apomorphy (see the output file in Brum et al. 2021).

Unenlagiinae affinities

The specimen DGM 921-R shares with the Unenlagiinae the secondary loss of both teeth carinae, as well as the fluted teeth and the presence of two or more longitudinal grooves on tooth surfaces. The dentition of Ypupiara lopai shows mixed features with unenlagiines. Among unenlagiines, Y. lopai is more similar to Buitreraptor gonzalezorum, due to the wide space between the maxillary tooth alveoli, the ziphodont tooth condition and the lack of interdental plates in the dentary, which are present in Austroraptor cabazai. Although not conidont, Y. lopai was grouped as the sister taxon of A. cabazai based on the similar CBR ratio. Therefore, the morphology is similar to that observed in B. gonzalezorum, but the proportion resembles A. cabazai. This unique combination of features, in addition to the maxillary preantorbital body with the restricted number of neurovascular foramina and the rectangular interdental plate, indicate that the new specimen represents a new species of Unenlagiinae, probably closely related to the contemporaneous A. cabazai.

According to the heuristic search analysis, the teeth of Y. lopai are markedly similar to those of B. gonzalezorum due to the mesiodistal curvature of the teeth, the occurrence of flutes and the longitudinal central groove with ridges. They share with other small theropods the small crown (CH < 10 mm), the lack of carinae, a smooth enamel surface and roots significantly larger than the crown base. Although they are morphologically similar, they occupy distinct morphospaces and morphoclades in multivariate analysis, with Y. lopai falling in the overlap between dromaeosaurids and troodontids. We interpret the overlap of Y. lopai with Spinosauridae and Masiakasaurus in the LDA analyses as being due to the occurrence of flutes in the teeth crown, which are present in B. gonzalezorum, but which were not included in the dataset (see further in Brum et al. 2021). Regarding the overlap of Y. lopai with dromaeosaurids and troodontids, as well as the marked separation of the morphoclades of Y. lopai and B. gonzalezorum, this could be related to the occurrence of different ecological niches, as suggested by the differences in the teeth of B. gonzalezorum and A. cabazai. We also highlight the marked variation between the proportions of B. gonzalezorum and the morphospaces occupied by most dromaeosaurids, troodontids and Y. lopai. We also note that the low number of flutes in the teeth of B. gonzalezorum could also contribute to the observed difference to Y. lopai, although most of the variation relates to the dental crown proportions. Instead, we can first conclude that Y. lopai is a dromaeosaurid based on both a heuristic search and multivariate analyses of the teeth. We can also conclude, as a result of the teeth data, that Y. lopai is an unenlagiine based on phylogeny and the morphological similarity to B. gonzalezorum.

The record of deinonychosaurs from Brazil

The presence of small maniraptorans and small deinonychosaurs was previously reported in Brazil as a result of teeth and isolated bone remains (Franco-Rosas 2002; Novas et al. 2005; Machado et al. 2008; Ghilardi & Fernandes 2011; Delcourt & Grillo 2014; Tavares et al. 2014). However, only one dorsal vertebra from the Campanian–Maastrichtian of the Adamantina Formation has been tentatively assigned to Unenlagiinae (Candeiro et al. 2012). As expected for the theropods of the Bauru Group (Bandeira et al. 2018), DGM 921-R is incomplete, although it is the third specimen from the Bauru Group to be found with some associated material. This pattern of poor preservation of theropod fossils indicates a preservation bias in comparison with baurusuchids (for a detailed discussion, see Bandeira et al. 2018).

A record of a partial right metatarsus also exists from the Marília Formation in the Ponto 1 do Price site. This specimen disappeared after Price’s death and is only recognized by an unfinished plate (Fig. 10), indicated as ‘Lopasaurus’. The plate showed an articulated metatarsus, with a complete metatarsal II and distal portions of metatarsals III and IV (Fig. 10). Based on this illustration, the material could be assigned to Dromaeosauridae based on the ginglymoid distal end of metatarsals II and III (Turner et al. 2012). We could assign this illustrated specimen to Unenlagiinae by the following shared similarities with B. gonzalezorum, Neuquenraptor argentinus Novas & Pol, 2005 and Pamparaptor micros Porfiri et al., 2011 in the distal portion of the metatarsus: the anterior surface of metatarsal III transversally expanded; the expansion of the posterolateral region of metatarsal II; and the expansion of the posteromedial region of metatarsal IV. In posterior view, the expansions observed in metatarsals II and IV overlap metatarsal III. However, such missing material was not associated with Y. lopai in the geological horizon. This illustration indicates the potential diversity of the maniraptorans in the Bauru Group. Further fieldwork in Peirópolis could provide new data on this theropod group.

The dromaeosaurid tooth record from the Bauru Group is still scarce, with most of them exhibiting carinae, as well as denticles (Franco-Rosas 2002; Tavares et al. 2014), although a small number of reports note that the dromaeosaurid teeth lack carinae (Ghilardi & Fernandes 2011). Regarding the MPMA specimens from Tavares et al. (2014), and reanalysed here with Y. lopai, their assignment based only on morphometry is still unclear due to the overlap and the lack of some morphometric data (e.g. the number of denticles per mm in each carina). The morphology is also distinct from Y. lopai. However, we could not rule out that the MPMA specimens are not dromaeosaurid. Further studies should focus on updating the teeth assignments based on recent approaches, as observed for the abelisaurids in the Bauru Group (e.g. Delcourt & Grillo 2018; Delcourt et al. 2020). However, the MPMA specimens (which may be dromaeosaurids) and Y. lopai, together with noasaurids, indicate a hitherto hidden diversity of small-sized theropods in the Bauru Group, which reinforces the niche-partitioning hypothesis of Bandeira et al. (2018).

The presence of unenlagiine remains could indicate that: (1) unenlagiines were the part of the maniraptoran diversity of the Bauru Group; and (2) the unenlagiines and dromaeosaurids of the Bauru Group exhibited a diversity of forms during the Campanian–Maastrichtian. The recovery of Y. lopai, with cranial bones and associated teeth, could prove to be an important comparative key for the assignment of isolated teeth from the Bauru Group and to clarify aspects of the distribution of the group.

Comments on Unenlagiinia

Our analysis provides support for the Unenlagiinia, with the Halszkaraptorinae as a sister-group within the Unenlagiinia (as independently found by Brownstein 2019, Hartman et al. 2019 and Cau 2020) instead of falling at the base of the Dromaeosauridae (as in Cau et al. 2017). Five unambiguous synapomorphies support Unenlagiinia (see Systematic Palaeontology, above). Furthermore, another four ambiguous synapomorphies were recovered: more than 75 cheek teeth with relatively small alveoli (also recovered by Cau 2020); presence of the sixth sacral vertebra (also recovered by Brownstein 2019 and Cau 2020); transition point in the tail extending across more than three caudal vertebrae; and flexor tubercle of the pedal unguals III and IV poorly developed.

In this way, it is possible that up to nine synapomorphies could support the Unenlagiinia. Even though the understanding of the phylogenetic relationships and evolutionary history of deinonychosaurs has been thoroughly investigated recently (e.g. Cau et al. 2017; Hartman et al. 2019; Brownstein 2019; Cau 2020; Currie & Evans 2020), further work is needed. Of particular note are the results presented by Hartman et al. (2019), which indicate a large Unenlagiinae clade involving the Laurasian taxa Ornithodesmus, Pyroraptor and Dakotaraptor; and a clade ‘Unenlagiidae’ (Unenlagiinae + Halszkaraptorinae) falling outside of the Dromaeosauridae (Hartman et al. 2019). In the present work, we preferred to use a dataset modified from Cau et al. (2017) because it is has a more comprehensive character list. We note that both Pyroraptor and Dakotaraptor have serrated teeth (Allain & Taquet 2000; DePalma et al. 2015), which would exclude these taxa from the Unenlagiinae and the Unenlagiinia as a whole. Dakotaraptor was originally recovered as a eudromaeosaurian (DePalma et al. 2015), and this result was independently replicated recently by Currie & Evans (2020). Ornithodesmus is an elusive taxon represented exclusively by a sacrum, which deserves further investigation.

Finally, it is interesting to note that the current ecological interpretations of unenlagiines and halszkaraptorines are consistent with their close relationship and the monophyly of Unenlagiinia as herein defined. Unenlagiines have been interpreted as piscivores (Gianechini et al. 2011; Cau et al. 2017; Cau 2020 contra Brownstein 2019), which is here corroborated (see below). Piscivory implies the foraging for fishes in aquatic environments, irrespective of whether this involves swimming or standing on foot (e.g. Souza et al. in press). Halszkaraptorines, in turn, have been interpreted as semi-aquatic dromaeosaurids with some level of swimming abilities (Cau et al. 2017; Cau 2020; but see Brownstein 2019). If both interpretations are correct, then the most parsimonious conclusion is that some degree of preference for foraging in aquatic environments was present at the base of the Unenlagiinia; thus characterizing their evolutionary history as intimately linked to some level of semi-aquatic habits.

Feeding mode of Unenlagiinia

The anatomy of fossil species is the key to understanding their behaviour and environmental interactions, and there are several difficulties involved in this kind of study (e.g. Kellner et al. 2019b). Here we discuss the anatomical evidence related to Unenlagiinia feeding behaviour. Buitreraptor gonzalezorum and A. cabazai exhibit tooth crowns that lack carinae, lack any serrations and bear longitudinal ridges and flutes. Furthermore, the tooth crowns are conical in Austroraptor and only slightly laterally compressed in Buitreraptor, rather than strongly compressed pattern that is most common among theropods (e.g. Hendrickx et al. 2015). The dentition of Austroraptor is thus conidont (sensu Hendrickx et al. 2015, approximately conical, recurved teeth without carinae or serrations), while that of Ypupiara can be considered as somewhat intermediate between conidont and ziphodont, which is more common throughout dromaeosaurids and theropods overall (e.g. Hendrickx et al. 2015).

The conidont condition of some unenlagiines and the oval to circular cross-section of the tooth crown is typical of piscivorous reptiles (Souza et al. in press). Regarding the grooves in the surface of the teeth, this feature is present in purported piscivorous tetrapods, which supports a piscivorous habit for unenlagiines (Gianechini et al. 2011). We highlight that, even though inferred piscivorous archosauromorphs have a plethora of bauplans (as disparate, for example, as gavialoid crocodylians, spinosaurid theropods and anhanguerid pterosaurs), two of the anatomical features that are common to most toothed piscivorous archosauromorphs are a conidont dentition (or at least a rounded cross-section of the crown), and an extensive tooth count, both of which are related to an inertial mode of feeding (i.e. using teeth to hold the prey and then swallow it whole), which is typical of, although not exclusive to, piscivorous reptiles (Souza et al. in press). We thus find it plausible that unenlagiines consumed fish at least as part of their diet, and potentially as the main source of food at least in Austroraptor due to its conidont dentition.

Still, Gianechini et al. (2020) recently proposed that unenlagiines were not mainly piscivorous (contra Gianechini et al. 2011). According to Gianechini et al. (2020), tooth morphology is not unequivocal evidence for piscivory; and other evidence must be considered (to which we fully agree; Souza et al. in press). Furthermore, through morphometric analyses and morphofunctional comparisons, Gianechini et al. (2020) showed that the hindlimbs of unenlagiines could have performed cursorially even better than other dromaeosaurids, being similar to ground birds that feed on terrestrial prey, such as Cariama, implying that unenlagiines too could have foraged on the ground for terrestrial prey. Based on the faunal record of La Buitrera, Gianechini et al. (2020) inferred that small tetrapods, such as sphenodonts, crocodyliforms and mammals, were the potential prey of Buitreraptor. However, Gianechini et al. (2020) did not include in their analyses ground birds that forage for fish (e.g. ardeids, ciconiids), which could be appropriate analogues for possible piscivorous unenlagiines such as we describe here. Furthermore, as mentioned above, although not unequivocal, tooth morphology is still strongly associated with piscivory.

Nonetheless, as observed by Gianechini et al. (2020), Buitreraptor could have used inertial feeding for small prey items other than fish, such as invertebrates and small terrestrial vertebrates. A somewhat generalist diet for Buitreraptor as suggested by Gianechini et al. (2020) seems plausible if this taxon comes from aeolian strata (Apesteguía et al. 2016), and not fluviolacustrine ones, as is the case for Austroraptor and Ypupiara. Furthermore, with all of this taken into account, we find it plausible that at least Austroraptor and Ypupiara would have consumed fish as a considerable part of their diet (Fig. 11), and that Buitreraptor would have been more of a generalist due to its drier environment, as well as its labiolingually compressed teeth. Although the morphology of the maxillary teeth in Ypupiara lopai is remarkably similar to that observed in B. gonzalezorum, the morphometric analyses indicate some differences in their proportions, and that they fell into separate morphospaces and morphoclades. Such a difference could be related to their distinct ecological traits and feeding habits.

Within dromaeosaurids, there are two other cases of proposed piscivorous habits. A specimen of Microraptor gui Xu et al., 2000 has fish remains in its abdominal cavity, and this was noted as the first definitive evidence of piscivory in a non-avian maniraptoran (Xing et al. 2013). In M. gui, serration on the mesial carinae is absent, which Xing et al. (2013) have interpreted as supporting evidence for a diet inclined towards piscivory. However, the dentition of M. gui still presents distal serrations and lateral compression. Furthermore, another specimen of M. gui had an enantiornithean bird in its gut (O’Connor et al. 2011), showing that this taxon could have in fact been a generalist. More recently, ecomorphological analyses of the Mongolian dromaeosaurid Halszkaraptor indicated a probable semi-aquatic lifestyle (Cau et al. 2017; Cau 2020). Its dentition is similar to that of unenlagiines, in that it lacks serrations and the anterior teeth are less compressed than ziphodont teeth (Cau et al. 2017). Considering that there is evidence of fish-eating elsewhere in the Dromaeosauridae, it seems not unreasonable to hypothesize that some dromaeosaurids consumed fish as a considerable part of their diet.

Distribution of the unenlagiines

In the PAE analysis, the synapomorphies in the generated consensus area cladogram are traditionally interpreted as vicariance events (e.g. Morrone 1994, 2014). In this case, the distribution of Unenlagiinae is a result of vicariance events only in the clade that involves the Candeleros and Portezuelo formations. The presence of unenlagiines in the formations studied here suggests dispersal events or extinction events (Morrone 2014), especially given the systematic affinities of Ypupiara.

The spatial reunion of populations of Unenlagiinae from the Bauru Group and the Neuquén Basin formations (as well as the other tetrapods analysed here) may have been stimulated after the Cenomanian–Turonian interval (known as the Cretaceous Thermal Maximum, CTM; Rothmann 2002; McInerney & Wing 2011). During this interval, the temperatures started to drop and many humid palaeoenvironments started to form (Rothmann 2002; McInerney & Wing 2011). The Bauru Group was surrounded by the Caiuá desert, which is primarily Aptian (sensu Pinheiro et al. 2018), and which separates the Bauru Group from the Neuquén Basin. The Caiuá palaeodesert, from which an interesting fauna of reptiles is beginning to emerge (e.g. Kellner et al. 2019b) may have acted as a geographical barrier, preventing the dispersion of some taxa between the Neuquén Basin and the Bauru Group. This barrier could have isolated the Unenlagiinae in the Neuquén Basin, until the Coniacian–Santonian. Later, with the posterior regression of the Caiuá desert and a change to a more humid palaeoenvironment (e.g. Pinheiro et al. 2018) in the Coniacian–Santonian, as suggested by the Adamantina Formation (Pinheiro et al. 2018; Castro et al. 2018), the establishment of palaeo-lakes, and subsequently of rivers, may have allowed these animals to disperse to these palaeoenvironments.

At the same time, the differentiation of the palaeoenvironment over time may have been a limiting factor for some taxa. The stratigraphic units of the Bauru Group congregate in general deposits of aeolian, lake and fluvial origin (with meandering and interlaced channels; Fernandes 2004), suggesting a warm seasonal climate. The arid conditions intensified towards the top of the Bauru Group (e.g. Marília Formation, sensu Bittencourt & Langer 2011), which may have been a limiting factor for dispersion throughout higher latitudes. Further evidence may facilitate the understanding of the faunal similarities and historical (evolutionary and biogeographical) relationship between the geological units discussed here.

Traditionally, the tetrapod taxa of the Bauru Group have been considered endemic (e.g. Bertini et al. 1993; Santucci & Bertini 2001), especially the small-bodied animals (Santucci & Bertini 2001). More recently this endemism has been contested (e.g. Bittencourt & Langer 2011) due to some tetrapod genera (e.g. titanosaurs, baurusuchids, peirosaurids, etc.) being recorded in Argentinian and Brazilian strata. We found a closer ecological relationship between the Bauru Group formations, indicating that the sharing of similar taxa in Argentina and Brazil may not reflect historically structured biotic assemblages.

The record of B. gonzalezorum (Cenomanian; Corbella et al. 2004) and that of A. cabazai (Maastrichtian) are, respectively, the earliest and latest records of unenlagiines (Fig. 8). Specimen DGM 921-R is the sister-group of A. cabazai, being coeval with this taxon. The similarity between Y. lopai and B. gonzalezorum indicates a more recent divergence from the others. The species U. comahuensis and U. paynemili are interpreted here as syntopic (e.g. Hartman et al. 2019), that is, a type of sympatry in which sister-taxa occur in the same habitat within the same region, because they could coexist in the same habitat without outcompeting each other. Deeper ecological affinities of Neuquenraptor and other taxa need to be investigated in the future.

Unenlagiines are also referred to as widespread through South America, with Argentinean records also accompanied by Brazilian and Bolivian records, albeit fragmentary (Gianechini & Apesteguía 2011). However, the incompleteness of the fossil material means that it is still not confirmed that the group was indeed widely distributed throughout South America.