The phylogenetic affinities and morphological peculiarities of the bird-like dinosaur Borogovia gracilicrus from the Upper Cretaceous of Mongolia

Introduction

Troodontidae is a species-rich clade of bird-like dinosaurs with unsettled phylogenetic ties and pliable ingroup relationships. Troodontids are traditionally associated with dromaeosaurids within a broader Deinonychosauria (see, e.g., detailed review by Turner, Makovicky & Norell, 2012; Agnolín et al., 2019; Pittman et al., 2020). The name Deinonychosauria, anchored on the dromaeosaurid taxon Deinonychus, explicitly refers to the ‘δɛινóς őνυξ’ (Ostrom, 1969), the specialized second toe bearing a large, sickle-shaped, trenchant ungual, variably present in dromaeosaurids and troodontids (Turner, Makovicky & Norell, 2012). Paul (1988a) extended the meaning of the morphological term ‘sickle-claw’ using it as a synecdoche for ‘dromaeosaurid’ (Paul, 1988a). The same author then introduced ‘sickle-clawed’ (Paul, 1988b) as an informal term for dromaeosaurids, troodontids and other close relatives of birds, taxa that are now all inferred to be paravians (Sereno, 1997; Sereno, 1998).

Nevertheless, the phylogenetic hypotheses comprising rootward paravian interrelationships remain conflicting and particularly sensitive to taxon and character samplings (see e.g., Turner, Makovicky & Norell, 2012; Agnolín & Novas, 2013; Brusatte et al., 2014; Cau, Brougham & Naish, 2015; Cau et al., 2017; Cau, 2018; Hartman et al., 2019; Pei et al., 2020; Pittman et al., 2020). As such, they represent a subject for further studies exploring, among other things, assessments of character distribution and use of a wider array of analytical methods.

Here we provide new information on the peculiar morphology of Borogovia gracilicrus Osmólska, 1987, a troodontid from the Maastrichtian (Upper Cretaceous) Nemegt Formation of southern Mongolia. The only known specimen of B. gracilicrus is a partial skeleton, represented exclusively by distal hindlimb elements (Figs. 1–5). The material was obtained during the 1971 Polish-Mongolian paleontological expedition to the Nemegt Basin. It was discovered at the base of the Altan Uul sections, at Altan Uul IV, which exposes the lower Nemegt Formation (Eberth, 2018). Osmólska (1987) referred Borogovia to Troodontidae based on the morphology of the metatarsus, which shares with other troodontids, among other features, the combination of metatarsal III being pinched proximally and wedged between the adjacent metatarsals, and a transversely robust metatarsal IV which forms more than half the width of the metatarsus.

The troodontid affinities of Borogovia have never been questioned; despite having been tested in only a few numerical analyses. Although these analyses supported the original taxonomic attribution of Borogovia, they did not resolve the precise placement of the taxon among troodontids (e.g., extended data figure 9 in Cau et al., 2017; supplementary figure 1 in Hartman et al., 2019). Indeed, even though the combination of apomorphies in the metatarsus of Borogovia strongly supports referral to Troodontidae, this taxon differs from all other troodontids and paravians in having an unusual set of features present in the toes (Osmólska, 1987), including the non-falciform shape of the second ungual.

The specialized second toe, bearing a sickle-like ungual, which is absent in Borogovia, has frequently been considered a key synapomorphy of a clade including troodontids and dromaeosaurids and usually referred to as Deinonychosauria (Colbert & Russell, 1969; Russell, 1969; Sereno, 1998; Turner, Makovicky & Norell, 2012). Osmólska (1987) compared the feet of dromaeosaurids and troodontids, concluded that the specialized second toe might have developed convergently in the two groups, and dismissed the possibility of a monophyletic Deinonychosauria. The hypothesis that the specialized second toe in troodontids evolved independently from the condition in dromaeosaurids was further discussed by Currie & Peng (1993). Yet, after the introduction of phylogenetic methods in theropod systematics, Deinonychosauria has been frequently reconstructed as a clade (see review in Turner, Makovicky & Norell, 2012), with its monophyly supported by synapomorphies pertaining to the skull and axial skeleton, rather than only by the shared presence of the ‘sickle-clawed’ second toe (Turner, Makovicky & Norell, 2012; Brusatte et al., 2014; Hartman et al., 2019; Pei et al., 2020; Pittman et al., 2020). Alternatively, this grouping has been considered as a paraphyletic series of successive sister taxa to avialan (bird-line) paravians, and the evolution (and secondary loss) of the specialized second toe has been interpreted to be homoplastic within Paraves (e.g., Agnolín & Novas, 2013; Cau, Brougham & Naish, 2015; Cau et al., 2017). These alternative results suggest that a specialized second toe bearing a falciform ungual (Ostrom, 1969), here termed the ‘falciphoran condition’ (from Latin ‘falcis’, sickle, and Greek ‘φoρɛας’, bearer), did not necessarily originate only once and was not necessarily associated exclusively with Deinonychosauria (here meant as the troodontid-dromaeosaurid grouping, and excluding avialans and other theropods; following Sereno, 1998; Turner, Makovicky & Norell, 2012).

Owing to the incomplete knowledge of the morphological features present in Borogovia gracilicrus, the primary aims of this study are (1) to supplement the original description of Osmólska (1987), in part by providing updated discussion of some morphological peculiarities observable in B. gracilicrus and comparisons with the numerous paravian taxa erected over the last three decades, and (2) to test and discuss the phylogenetic affinities of this unusual taxon.

Material and Methods

Systematic Paleontology

Type locality. Altan Uul IV, lower Nemegt Formation; Maastrichtian, Upper Cretaceous; Mongolia.

Holotype. ZPAL MgD-I/174, both tibiae missing most of their proximal ends (Figs. 1 and 2), fragment of proximal fibula, almost complete astragalocalcanea fused to the tibiae (Figs. 1 and 2), both feet incompletely preserved (of the left pes: distal parts of metatarsals II and IV, second toe with phalanx II-2 damaged, third toe including only distal end of phalanx III-1 and slightly damaged phalanx III-3, almost complete fourth toe with proximal part of ungual; of the right pes: distal portions of articulated metatarsals II, III and IV, complete second toe, but phalanx II-2 lacking ventrodistal ‘heel’, third toe with preserved distal end of phalanx III-1, phalanx III-2 dorsodistally with some damage and phalanx III-3 ventrodistally with some damage, ungual with damaged dorsal edge, complete fourth toe with ungual somewhat damaged at the dorsal edge; Osmólska, 1987) (Figs. 3–5).

Diagnosis (emended from Osmólska, 1987). Troodontid theropod with the following combination of features (autapomorphies marked by *): distinct semilunate fossa at base of ascending process of astragalus; lateral condyle of astragalocalcaneum extends farther proximally across extensor surface of bone than medial condyle; extensor surface of the distal end of the first phalanx of second toe moderately developed, not significantly expanded dorsally (along the flexor-extensor axis) relative to the shaft; second phalanx of second toe extremely abbreviated (length subequal to proximal depth) and lacking a distinct constricted diaphysis between the articular ends; penultimate phalanx of third toe subequal (96%) in length to preceding phalanx; phalanges of fourth toe more robust than those of third toe*; ungual of second toe low, unrecurved and non-falciform*; and ventral margins of unguals II-IV straight in lateral/medial view.

Differential diagnosis (with respect to other Nemegt Formation troodontids). Borogovia gracilicrus differs from Tochisaurus nemegtensis (Kurzanov & Osmólska, 1991) in that the distal end of metatarsal II is subequal in width to the distal ends of metatarsals III and IV (in Tochisaurus, the distal end of metatarsal II is 75% as wide as the distal ends of metatarsals III and IV; see table 1 in Kurzanov & Osmólska, 1991). Furthermore, the shallow proximodistal sulcus running along the flexor surface of the distal end of metatarsal III of Tochisaurus nemegtensis (Kurzanov & Osmólska, 1991) is absent in B. gracilicrus. Borogovia differs from Zanabazar junior (Norell et al., 2009) in the following combination of features: the ascending process of the astragalus lacks a distinct bump in the middle of the extensor surface (present in Zanabazar, Norell et al., 2009); the medial condyle of the astragalus is not buttressed by a swollen area (present in Zanabazar, Norell et al., 2009); the long axis of the medial fossa of the medial astragalar condyle forms a wider angle with the proximodistal axis of the tibia (about 60°, compared to an angle of less than 30° in Zanabazar, see fig. 33B in Norell et al., 2009); in extensor view, the lateral condyle of the astragalus is more extended proximally than the medial condyle (in Zanabazar, the two condyles are comparably extended proximally).

Remarks. Borogovia shows a unique combination of features which supports the validity of this taxon and distinguishes it from other Cretaceous theropods. Although the majority of these features also supports a paravian and troodontid placement for this taxon, some are unusual for a paravian, and are shared with other maniraptoriform clades.

Osmólska (1987) listed the very slender and long tibiotarsus among the diagnostic features of Borogovia gracilicrus. Although we agree that the tibiotarsus of this theropod is noteworthy for its elongation and slenderness, the diagnostic status of this feature is problematic. The tibiotarsi of B. gracilicrus are both incomplete, and their actual length could only be inferred to be at least ten times their distal mediolateral width (Figs. 1 and 2). This ratio is expected to be ontogenetically variable and allometrically-controlled, making any explicit definition of that feature that might be used in the species diagnosis questionable without a sufficiently large sample. Furthermore, the lack of complete tibiotarsi in several troodontids, including the other two Nemegt taxa (e.g., Kurzanov & Osmólska, 1991; Norell et al., 2009; Zanno et al., 2011) prevents any unambiguous differentiation between the peculiar elongation of the tibiotarsi in B. gracilicrus and the condition seen in its closest relatives.

In Borogovia, the suture between the astragalus and calcaneum is obliterated, and the two bones are tightly sutured to the tibia (Figs. 1 and 2). Yet, the ascending process of the astragalus can be distinguished from the overlapped tibia along its medial margin, and can be seen to stand out distinctly from the tibial shaft in medial view. This combination of features has been considered a paravian synapomorphy (Turner, Makovicky & Norell, 2012). In the majority of other coelurosaurs, the suture between the astragalus and calcaneum is not obliterated, and the ascending process is not tightly sutured to the tibia (e.g., Gallimimus, Osmólska, Roniewicz & Barsbold, 1972; Tyrannosaurus, Brochu, 2003; Khaan, Balanoff & Norell, 2012). In many avialans, the degree of coossification of the tibiotarsal elements is more extreme than in other paravians, including Borogovia, and the ascending process of the astragalus can barely, if at all, be distinguished from the tibia in adult individuals (e.g., Hollanda, Bell et al., 2010; Qiliania, Ji et al., 2011).

The ascending process of the astragalus of Borogovia is expanded proximally, and overlaps the whole extensor surface of the distal end of the tibia (Figs. 1 and 2). An ascending process of the astragalus which is more than twice as long proximodistally than wide at its base, covering the whole distal shaft of the tibia, is a condition widespread among maniraptoriforms (e.g., Velociraptor, Norell & Makovicky, 1999). This feature distinguishes Borogovia from some parvicursorine alvarezsaurids, in which the astragalus bears a proportionally narrower ascending process of the astragalus, restricted to the medial half of the tibiotarsal complex and having a concave lateral margin (e.g., Mononykus, Perle et al., 1994).

The distal end of the tibiotarsus of Borogovia bears prominent condyles which describe a broad arch in both the extensor and distal directions. In distal view, the flexor margin of the astragalocalcaneum describes a broad concavity (extension of the intercondylar sulcus onto the flexor surface), bordered by the projections of the condyles (Figs. 1E, 2E). In most theropods, the flexor margin of the distal end of the astragalocalcaneum is more flattened, lacking both a distinct flexor projection of the condyles and an extension of the intercondylar sulcus onto the distal end of the flexor surface (e.g., Osmólska, Roniewicz & Barsbold, 1972; Ostrom, 1969; Brochu, 2003). A condition comparable to Borogovia is present in avialans (e.g., Bell et al., 2010).

As preserved, the shafts of both metatarsals II in Borogovia are markedly compressed transversely, making them less than half as wide as the distal articular surfaces of the same bones (Figs. 3 and 4). This condition differs from the majority of coelurosaurs, regardless of their body size, which usually have a more proportionally robust diaphysis of metatarsal II (e.g., Deinonychus, Ostrom, 1969; Tyrannosaurus, Brochu, 2003, Hulsanpes, Cau & Madzia, 2018). The diaphysis of metatarsal II is significantly narrower than the distal end of the bone in Late Cretaceous troodontids (e.g., Gobivenator, Tsuihiji et al., 2014; Philovenator, Currie & Peng, 1993); Talos, Zanno et al., 2011).

The distal end of metatarsal II of Borogovia is uniformly convex in extensor view, because the flexor sulcus does not extend onto the distal and extensor surfaces of the bone (Figs. 3, 4). This condition is widespread among coelurosaurs (e.g., Gallimimus, Osmólska, Roniewicz & Barsbold, 1972; Talos, Zanno et al., 2011; Tyrannosaurus, Brochu, 2003) and differs from most dromaeosaurids (e.g., Velociraptor, Norell & Makovicky, 1999) and some avialans (e.g., Yungavolucris, Chiappe, 1993), which show a ginglymous distal end of metatarsal II and a distinct sulcus running along the extensor surface of the distal end of the bone.

When articulated, the distal ends of metatarsals II to IV of Borogovia closely fit together along well-defined surfaces. In particular, the distal shaft of metatarsal III is triangular in cross section, being tightly wedged between articular facets on the shafts of the other two metatarsals, with the latter broadly overlapped by the former in extensor view (Holtz, 1994; Snively, Russell & Powell, 2004). This feature is one of the components of the ‘arctometatarsalian condition’ (Holtz, 1994), and is widespread among Late Cretaceous coelurosaurs, including tyrannosaurids (Snively, Russell & Powell, 2004), ornithomimids (Holtz, 1994; Snively, Russell & Powell, 2004), parvicursorine alvarezsaurids (Perle et al., 1994), some oviraptorosaurs (i.e., caenagnathids and avimimids; Snively, Russell & Powell, 2004; Funston et al., 2016), and late-diverging troodontids (Kurzanov & Osmólska, 1991; Snively, Russell & Powell, 2004; Norell et al., 2009; Tsuihiji et al., 2014). Among the gracile-limbed theropods from the Upper Cretaceous of Mongolia, the metatarsal III morphology of Borogovia distinguishes this taxon from eudromaeosaurs (Ostrom, 1969; Holtz, 1994) and halszkaraptorines (Cau & Madzia, 2018), in both of which the distal shaft has a quadrangular cross section and is not wedged between the adjacent metatarsals. In those dromaeosaurids, metatarsal III barely, if at all, overlaps the distal shafts of the other metatarsals in extensor view (e.g., Snively, Russell & Powell, 2004; Cau & Madzia, 2018).

Related to the previous feature, although not strictly co-variant with it, is the presence of flanges along the dorsal margins of the distal shaft of metatarsal III which overlap the adjacent metatarsals (Snively, Russell & Powell, 2004). This feature is particularly evident in ornithomimids (Osmólska, Roniewicz & Barsbold, 1972; Holtz, 1994; Snively, Russell & Powell, 2004), avimimids (Funston et al., 2016) and parvicursorines (Perle et al., 1994; Turner, Nesbitt & Norell, 2009), which show a distinct and symmetrical development of these flanges over both metatarsals II and IV. Borogovia differs from the aforementioned taxa in that the flanges of metatarsal III are less symmetric and more poorly developed. In particular, the lateral margin of the distal half of metatarsal III of Borogovia, adjacent to metatarsal IV, describes a straight line in extensor view, parallel to the proximodistal axis of the bone (Fig. 3E). This feature distinguishes troodontids from other arctometarsalian taxa (e.g., ornithomimids, Snively, Russell & Powell, 2004).

In the majority of theropods, the trochlear distal articular surface of metatarsal III extends onto the flexor surface and bears a longitudinal sulcus (e.g., Gallimimus, Osmólska, Roniewicz & Barsbold, 1972; Mononykus, Perle et al., 1994). The third metatarsal of Borogovia shares a peculiar condition with Sinornithoides (Russell & Dong, 1993) and a specimen referred to Stenonychosaurus (Russell, 1969), in that the flexor surface of the distal end is subtriangular in flexor view and lacks a distinct longitudinal sulcus (Fig. 3F). This ‘tongue-like’ plantar extension of the articular surface, devoid of a longitudinal sulcus, is only incipiently developed in Tochisaurus (Kurzanov & Osmólska, 1991) and absent in Philovenator (Currie & Peng, 1993).

In most theropods, including the majority of paravians, the shaft of metatarsal IV is not significantly wider than those of metatarsals II and III (e.g., Deinonychus, Ostrom, 1969; Gallimimus, Osmólska, Roniewicz & Barsbold, 1972; Mononykus, Perle et al., 1994; Sinovenator, Xu et al., 2002; Tyrannosaurus, Brochus, 2003; Avimimus Funston et al., 2016). Borogovia shares with the Late Cretaceous troodontids a notably wide metatarsal IV, which - partly owing to the relative narrowness of the shafts of metatarsals II and III - forms at least half of the mid-shaft width of the metatarsus (e.g., Russell, 1969; Kurzanov & Osmólska, 1991; Currie & Peng, 1993).

The second toe of Borogovia is unique among non-avian theropods in showing a peculiar subset of the usual falciphoran features (Ostrom, 1969; Turner, Makovicky & Norell, 2012). Phalanx II-2 is very abbreviated, being less than half as long as phalanx II-1 (Figs. 5A–5D). As in other troodontids (e.g., Philovenator, Xu et al., 2012; Talos, Zanno et al., 2011), phalanx II-2 bears a prominent ventral process (heel) extending proximally from the ventral margin of the proximal facet, which as a result is significantly larger than the distal articular surface of the same bone. Furthermore, that phalanx lacks a clear distinction between diaphysis and articular ends, a condition otherwise seen in some Late Cretaceous troodontids (e.g., Gobivenator, Tsuihiji et al., 2014) and, in a less extreme form, in Halszkaraptorinae (e.g., Cau et al., 2017). The corresponding ungual is low and robust with a flattened ventral surface bounded proximally by a low yet well-defined flexor tubercle (Figs. 5F–5H). Such morphology differs radically from the falciform and mediolaterally-compressed unguals that are widespread among paravians, including most troodontids (e.g., Deinonychus, Ostrom et al., 1969; Talos, Zanno et al., 2011; Balaur, Brusatte et al., 2013; Halszkaraptor, Cau et al., 2017; Rahonavis, Forster et al., 2020). Although divergent from the morphology widespread among other ‘deinonychosaur-grade’ paravians, the second toe of Borogovia is overall similar to that of the extant bird Casuarius (e.g., Saber & Hassanin, 2014).

Borogovia differs from almost all other theropods in that the phalanges of the third toe are distinctly less robust than those of the fourth toe (Osmólska, 1987). Although this feature might be a by-product of the homologous enlargement of the fourth metatarsal shared with other troodontids, the fourth toe is less robust in these taxa than in B. gracilicrus (e.g., Talos, Zanno et al., 2011; Philovenator, Xu et al., 2012). A similar apomorphic condition is present in the parvicursorine Mononykus (although this taxon lacks a comparably robust metatarsal IV; see fig. 18 in Perle et al., 1994) but not in other alvarezsaurids, in which the third toe is the most robust in the foot (e.g., Turner, Nesbitt & Norell, 2009).

The extensor surfaces of the phalanges in the fourth toe of Borogovia bear relatively shallow extensor pits with poorly defined margins (Figs. 5Q–5X). This condition is shared with the majority of coelurosaurs (e.g., Ostrom, 1969; Brochu, 2003), but not with most alvarezsauroids, in which the fourth toe phalanges bear deeper extensor pits that are usually sharply bordered by the distal condyles (e.g., Monoykus, Perle et al., 1994).

Phylogenetic analysis

Results of the unweighted parsimony analysis

In all shortest trees found during the preliminary analysis, the very fragmentary OTU ‘Troodon’ was consistently reconstructed within the clade including all taxa traditionally treated as troodontids (e.g., Currie, 1987; Kurzanov & Osmólska, 1991; Russell & Dong, 1993; Norell et al., 2009; Pei et al., 2021), a result that validates the use of the name Troodontidae for this branch of Paraves. In particular, our analysis supports the hypothesis that Troodon is a member of the most inclusive clade containing Late Cretaceous troodontids but excluding Jinfengopteryx and Sinovenator. Exclusion of the ‘Troodon’ OTU from the analysis does not alter significantly the tree topology, except by improving resolution among the late-diverging troodontids. The main analysis reconstructed 50,000 shortest trees of 5805 steps each (Consistency Index excluding uninformative characters = 0.2252; Retention Index = 0.5754). In both analyses, Avialae is reconstructed as the sister taxon of Troodontidae (Fig. 6). The troodontid clade shows a relatively well-resolved topology, with only a few fragmentary OTUs (e.g., Tochisaurus, Urbacodon) acting as ‘wildcards’. The analysis reconstructed an early-diverging jinfengopterygine-sinovenatorine lineage, including Jinfengopteryx (the sole jinfengopterygine representative) and Mei, Ningyuansaurus (Ji et al., 2012) and Sinovenator (Sinovenatorinae). All shortest trees found in this analysis place Borogovia as the sister taxon to Troodontinae, which is the least inclusive group containing Troodon (when included), Gobivenator, and Zanabazar. Despite being relatively well-resolved, this topology is only weakly supported, which is likely due to the inclusion of very fragmentary OTUs in the analysis.

We replicated the analysis enforcing three alternative topologies reflecting the frameworks of some recent analyses of paravian relationships (Agnolín et al., 2019; Hartman et al., 2019; Pei et al., 2020). Enforcing the framework of Agnolín et al. (2019), i.e., troodontids, eudromaeosaurs, microraptorines, and unenlagiines successively closer to avialans, with halszkaraptorines—not discussed in that study—set as ‘floaters’), the shortest trees reconstructed were 26 steps longer than the shortest unenforced topologies (halszkaraptorines were reconstructed as unenlagiids). Enforcing the framework of Hartman et al. (2019): i.e., troodontids closer to eudromaeosaurs than the ‘Halszkaraptorinae + Unenlagiinae’ clade), the shortest trees reconstructed were 16 steps longer than the shortest unenforced topologies. Enforcing the framework of Pei et al. (2020: i.e., troodontids in Deinonychosauria; Balaur in Velociraptorinae; Rahonavis in Unenlagiinae) the shortest trees reconstructed were 50 steps longer than the unenforced shortest topologies. When the latter analysis was replicated setting Balaur and Rahonavis as ‘floaters’, the shortest trees reconstructed were 12 steps longer than the unenforced shortests topologies (Balaur and Rahonavis were reconstructed as avialans).

Discussion

Conclusions

Borogovia gracilicrus is a troodontid dinosaur from the Nemegt Formation of Mongolia. Although overlapping material with the other Nemegt trooodontids (Tochisaurus nemegtensis and Zanabazar junior) is limited, Borogovia can be differentiated from these taxa based on the unique combination of features in its tibiotarsus and metatarsus.

The phylogenetic position of B. gracilicrus among Maniraptora is investigated in detail for the first time. This taxon is robustly supported as a member of Troodontidae, and reconstructed among non-troodontine troodontids. These results are not biased by a priori assumptions on character weighting.

The foot of Borogovia is autapomorphic and combines troodontid synapomorphies (e.g., the arctometatarsus) and the secondary loss of some elements of the falciphoran condition widespread among paravians. In particular, the morphology of the second toe of Borogovia might indicate an ecological niche distinct from those of other ‘deinonychosaur-grade’ paravians. Such a scenario could account for the co-occurrence of several troodontids in the paleofauna of the Nemegt Formation.

Supplemental Information