Revised morphology of Pycnonemosaurus nevesi Kellner & Campos, 2002 (Theropoda: Abelisauridae) and its phylogenetic relationships Delcourt, Rafael Zootaxa 2017 4276 1 1 45 Kellner & Campos, 2002 Kellner & Campos 2002 [151,773,504,531] Reptilia Abelisauridae Pycnonemosaurus GBIF Animalia Saurischia 1 2 Chordata species nevesi   Holotype. The holotypeconsists of an incomplete post-cranial skeleton (  DGM859-R) housed in the Museu de Ciências da Terra( Earth Sciences Museum), Rio de Janeiro, Brazil. Itcomprises: two incomplete caudal vertebrae, four caudal centra, three caudal transverse processes, rib remains, a right proximal pubis, right tibia, right distal fibula and unidentified elements.  Kellner& Campos(2002)did not recognize the three transverse processes as belonging to  Pycnonemosaurus. Nevertheless, additional material, including a distal caudal vertebra, one isolated transverse process, an isolated neural spine and some unrecognized elements were allocated to the same accession number as the holotype. Thus, here I recognized the transverse processes as belonging to  Pycnonemosaurus nevesi(see description and discussion). In the original description, Kellner & Campos (2002)assigned five incomplete teeth to the holotype (DGM 859-R) of  Pycnonemosauruswith only a few features described and the authors wrote that the “detail description (with illustration) of this material will be presented elsewhere”. In the same month of  Pycnonemosaurus’s publication Bittencourt & Kellner (2002)published a detailed description of five teeth assigned to DGM 859-R. According to these authors there are nine teeth, four that are very fragmentary and consist of just the root and five others that are better preserved and only lacking portions of the enamel. However, these teeth were not assigned as  Pycnonemosaurus nevesiby Bittencourt & Kellner (2002)and this assumption was made in 2011 when Bittencourt & Langer mentioned materials of Parecis Group’s abelisaur. As the teeth were found isolated from any dentigerous element (i.e. premaxilla, maxilla and dentary) and the  Pycnonemosaurusholotype is based on caudal vertebrae and hind limb elements, it is thus difficult to assign them to the holotype. For these reasons, I consider the DGM 859-R teeth as belonging to an indeterminate abelisaurid instead of  Pycnonemosaurus nevesi.    TypeLocality.  Pycnonemosaurus nevesiwas collected in the Fazenda Roncador, Mato Grosso, Brazil, and according to  Kellner& Campos(2002), this locality is in the Bauru Group. However, recent work has shown that this locality is actually part of the Parecis Groupand that it is of Campanian-Masstrichitianage (  Weska, 2006;  Bittencourt& Langer, 2011). Inthis geological context then, many sauropods and theropods remains have been discovered over the last years (see  Bittencourt& Langer, 2011for a wide revision).   Revised Diagnosis.  Pycnonemosaurus nevesican be distinguished from other abelisauroids on the basis of the following autapomorphies: 1) a pubis with small rounded foot and a ventrally bowed anterodistal end, 2) the posterior caudal transverse process is hook-shaped with the anterodistal expansion being short and bowed (modified from Kellner & Campos, 2002) and 3) a well-developed lateral malleolus of tibia that is ventrally expanded.  Kellner & Campos (2002)diagnosed  Pycnonemosaurusas having “tibia with hatchet-shaped cnemial crest” and “caudal vertebrae showing moderate distal expansion of the transverse process”. Nevertheless, as seen in the description section, these features are shared among other abelisaurids species, thereby requiring the amended diagnosis above.   Description and Comparisons. Axial skeleton.The six caudal vertebrae of  Pycnonemosaurus nevesiare assigned to the caudal series due to the absence of pneumaticity, as observed in other abelisaurids ( Méndez, 2014) (Measurements in Table 1). In addition, just two vertebrae preserve the neural arch ( Figs. 1–2) and similarity of size between them suggests that all vertebrae belong to the anterior position in the tail. All caudal centra of  Pycnonemosaurusare larger than other known abelisaurids ( Figs. 1–3) ( Bonaparte et al., 1990; O’Connor, 2007; Grillo & Delcourt 2017) and the articular facets are round like the those in  Ceratosaurus( Madsen & Welles, 2000),  Eoabelisaurus,  Masiakasaurus,  Carnotaurus, Aucasaurus, Ekrixinatosaurus, Ilokelesiaand Skorpiovenator. Bycontrast,  Majungasaurus, Arcovenatorand Rajasaurushave articular facets with an elliptical shape ( Wilson et al., 2003; O’Connor, 2007; Tortosa et al., 2014). The caudal centra are spool-shaped and without pleurocoels, as in other abelisaurids. However, the lateral surface just ventral to the neural arch suture is slightly depressed with foramina. These depressions are present in  Masiakasaurus(FMNH PR2133),  Carnotaurus, Aucasaurus, Ekrixinatosaurusand probably in  Skorpiovenator(poor preservation in the latter prevents a more definitive statement regarding this character). Some foramina are large, being located on the neural arch suture near the articular surface. These may be associated with the attachment of the caudofemoralis muscle (Persons & Currie, 2011). The centra are amphicoelous as in other abelisaurids ( Méndez, 2014),  Masiakasaurusand  Ceratosaurus( Madsen & Welles, 2000, plate 17) with the anterior surfaces slightly more concave than the posterior ones. The chevron facets are ventrally prominent and the ventral surfaces have a shallow longitudinal sulcus, as seen in  Masiakasaurus, Eoabelisaurus, Aucasaurus,  Carnotaurus, Ilokelesia, Viavenator( Filippi et al., 2016)and  Majungasaurus( O’Connor, 2007)whereas  Ekrixinatosaurusand  Rajasaurushave a low ventral keel. The neural arches generally resemble those of other South-American abelisaurids with a dorsally inclined transverse process and an anterior distal projection. The pre- and post-spinozygapophyseal lamina limit the pre- and post-spinal fossae anteriorly and posteriorly, respectively. The spinoprezygapophyseal fossa of  Pycnonemosaurusresembles that of the sixth caudal vertebra in  Carnotaurus, being larger and wider than the condition in  Eoabelisaurus, Skorpiovenatorand Aucasaurus(particularly if we consider the more incomplete vertebra). The prezygapophyses form a right angle with one another and are inclined nearly 45° from vertical, as in  Carnotaurus(from the third vertebra),  Eoabelisaurus, Viavenator( Filippi et al., 2016: fig. 7A) and  Majungasaurus( O'Connor, 2007: fig. 15). More acute angles can be seen in the caudal vertebrae of  Ekrixinatosaurus, Aucasaurus, Skorpivenator and Arcovenator( Tortosa et al., 2014: fig. 5). In  Pycnonemosaurus nevesi,the prezygapophysis is slightly more elongated than in  Carnotaurus, Aucasaurusand Skorpiovenator, being also similar in length to those in  Eoabelisaurus,  Majungasaurus( O'Connor, 2007: fig. 15) and  Arcovenator( Tortosa et al., 2014: fig. 5), exceeding the limit of vertebral centrum such as in  Majungasaurus, Eoabelisaurus, Viavenator( Filippi et al., 2016: fig. 7C) and  Arcovenator. In  Skorpiovenator, Aucasaurusand  Carnotaurus, the extent of the prezygapophysis reaches the margin of the vertebral centrum. In  Pycnonemosaurus,the more incomplete neural arch of the prezygapophysis are very abraded and does not show this condition. In addition, the postzygapophyses are spaced, as seen in the first and second caudal vertebrae of  Carnotaurus, in the sixth caudal vertebra of  Skorpiovenatorand in the fifth caudal vertebra of  Eoabelisaurus.However, this structure differs from that of  Aucasaurus, in which all caudal postzygapophyses are near to each other. The hyposphene-hypantrum articulations are present in  Pycnonemosaurus, similar to the condition in  Ceratosaurus( Madsen & Welles, 2000),  Masiakasaurus, Eoabelisaurus, Viavenator( Filippi et al., 2016),  Carnotaurus, Aucasaurus, Ekrixinatosaurusand Skorpiovenator, but differing from that in  Majungasaurus( O’Connor, 2007)and  Arcovenator( Tortosa et al., 2014).   TABLE 1. Measurements from the appendicular and vertebral elements of  Pycnonemosaurus nevesiin mm. Abbreviations: FDW, fibula distal width; H, height; L, length; PBL, pubic boot length; TDW, tibia distal width; TSAD, tibia shaft anteroposterior diameter; TSW, tibia distal width; W, width; ¢, incomplete material; *, Figure 1; #, Figure 2; ε, Figure 3.      P. nevesi Tibia Pubis Fibula εCA1 ε CA2 ε CA3 ε CA4 *CA5 #CA6  L 870 776¢ - 155 154 149 155 148 144  W - - - 151 148 142 136 112 107  H - - - 174 165 151 146 148 123  PBL - 243 - - - - - - -  TDW 235 - - - - - - - -  TSW 124 - - - - - - - -  TSAD 99 - - - - - - - -  FDW - - 94.6 - - - - - -   FIGURE 1. Anterior and most complete caudal vertebra of  Pycnonemosaurus nevesi(DGM 859-R) in anterior ( A), posterior ( B), right lateral ( C), dorsal ( D) and left lateral ( E) views. Abbreviations: hyf, hyposphene; hyp, hypanthrum; nc, neural canal; ns, neural spine; poz, postzygapophysis; prz, prezygapophysis; tp, tranverse process. Scale bar equals 100 mm.   FIGURE 2. Anterior and less complete caudal vertebra of  Pycnonemosaurus nevesi(DGM 859-R) in anterior ( A), left lateral ( B), posterior ( C), right lateral ( D) and dorsal ( E) views. hyf, hyposphene; nc, neural canal; ns, neural spine; poz, postzygapophysis; tp, tranverse process Scale bar equals 100 mm.   FIGURE 3. Four anterior caudal centra of  Pycnonemosaurus nevesi(DGM 859-R) in left lateral ( A1-D1), anterior ( A2-D2) and posterior ( A3-D3) views. Scale bar equals 100 mm. Both neural spines are incomplete, lacking the distal portion and located dorsal to the posterior half of the centrum. Nevertheless, the anteroposterior proportion of the neural spine in the articulated vertebrae is similar to anterior caudal of  Carnotaurusand  Majungasaurus, suggesting a rod-like condition for  Pycnonemosaurusas in other abelisaurids ( Rauhut, 2003). The ventral half of the neural spine is directed posteriorly as in other abelisaurids. However, due to the fragmentary condition it is impossible to infer exact degree of inclination. Additionally, there are three isolated transverse processes ( Figs 4-6). The dorsal inclination of the articulated processes is similar to those of furileuraurian abelisaurids, such as  Carnotaurus, Aucasaurusand Skorpiovenatorand in contrast to the condition in  Majungasaurus, Arcovenatorand Eoabelisaurus. Thelength ratio of the transverse process to the length of vertebral centrum is greater than 1.3 inabelisaurids ( Rauhut, 2003). Due to incomplete preservation (i.e. loss of the distal end of the transverse process), it is not possible to assess this state in  Pycnonemosaurus. The three isolated transverse processes in  Pycnonemosaurus nevesiexhibit an anterior projection at the distal end, similar to furileuraurian abelisaurids ( Filippi et al., 2016). The smaller transverse process is hook-shaped with a short and bowed tip ( Fig. 6). This condition is different from  Aucasaurus, Ilokelesia, Ekrixinatosaurusand Skorpiovenatorin which the distal end of the posterior caudal transverse process is more elongated and straight. The ventral surface of the transverse process is convex and has anterior and posterior centrodiapophyseal lamina, similar to other abelisaurids. These laminae delimit the three fossae on the lateral surface; the prezygapophyseal centrodiapophyseal fossa, the centrodiapophyseal fossa and the postzygapophyseal centrodiapophyseal fossa ( sensu Wilson et al., 2011). Interestingly, the left side of the most complete vertebra, the postzygapophyseal centrodiapophyseal fossa has three small hollows unlike the right side, which only has two hollows ( Fig 1C). This condition is unique among abelisaurids and probably is an individual condition of this specimen of  Pycnonemosaurus nevesi.  Appendicular skeleton.The appendicular elements of  Pycnonemosaurusare represented by a right pubis, right tibia and the distal end of the right fibula (Measurements in Table 1).  Pubis.The pubis of  Pycnonemosaurusresembles the non-abelisaurinae abelisaurs with only the right shaft and the distal pubic foot preserved, whereas the proximal articulation has been lost ( Fig 7). The shaft is proximally sub-circular and becomes more compressed anteroposteriorly toward the distal end. There is an anteroposterior expansion immediately distal to the first constriction of the shaft, and then the shaft becomes compressed again. In the anterior view, the shaft is laterally bowed and the distal portion is rectangular shaped. The shaft shows a medial dorsoventral shelf that becomes the pubic apron and extends at the pubic foot as seen in  Masiakasaurus(Carrano et al., 2002), MCF-PVPH-237 (Coria et al., 2006) and  Carnotaurus. This shelf is quite rugose, particularly at the caudal end and suggests a strong contact with the left pubis, however, it clearly was not fused as exhibited by  Masiakasaurus( Carrano et al., 2011)and  Aucasaurus. The lateral surface of the shaft is flat and has scars that were interpreted as belonging to the muscle ambiens insertion. The distal foot is not elongated anteroposteriorly as in  Carnotaurusand  Aucasaurus, and the anterior end is anteroventrally inclined with a flat dorsal surface, whereas the posterior end is dorsally inclined. The distal pubis of  Eoabelisaurusis short as in  Pycnonemosaurus, however, the former has a straight foot base. The foot base is thick and robust as generally seen in other abelisaurids.   FIGURE 4. Largest isolated transverse process of  Pycnonemosaurus nevesi(DGM 859-R) in dorsal ( A) and ventral ( B) views. Scale bar equals 100 mm.  Tibia.The tibia of  Pycnonemosaurus nevesihas a strong, nearly straight shaft and subcircular transverse section this is slightly compressed anteroposteriorly ( Fig 8). There are some incomplete parts with some surfaces covered with plaster. It is very large and thick, rivalling in size with those of large theropods such as  Mapusaurus(Coria & Currie, 2006),  Allosaurus(Madsen, 1976)and  Tarbosaurus( Christiansen & Fariña, 2004)and it is larger than in other abelisaurids ( Grillo & Delcourt, 2017). Apart from its unusual size, the tibia has many abelisaurian characteristics. The shaft of the tibia of  Pycnonemosaurusis gently curved laterally in lateral view, with the curvature being almost inconspicuous as in  Ekrixinatosaurus, Quilmesaurusand Rahiolisaurusand distinguishing it from the tibial shaft of MCT 1783-R (Cambará taxon),  Skorpiovenator, Xenotarsosaurus, Aucasaurusand Arcovenatorin which the curvature is much more pronounced. In the dorsal view, the lateral condyle is subcircular and robust. This shape is also present in  Carnotaurus, Aucasaurus, Skorpiovenator, Ekrixinatosaursand Arcovenator( Tortosa et al., 2014: fig. 6) whereas in  Quilmesaurusthe lateral condyle is straight and in  Eoabelisaurus,  Majungasaurus, Cambarátaxon and  Xenotarsosaurus,it is square-shaped. The latter characteristic is even more pronounced in MTC 1783-R and  Eoabelisaurus. The intercondylar groove is deep as in Cambará taxon ( Machado et al., 2013) and  Aucasauruscompared with other taxa in which it is shallower. The cnemial crest of  Pycnonemosaurusis strongly pronounced anteriorly, inclined further proximally than the condyles with a dorsoventral constriction, giving it a hatchet-shaped similar to the condition in  Aucasaurus. Its internal (i.e. lateral) surface is concave with an anteroposterior bulge in the dorsal portion, making an expanded dorsoventrally lateral fossa like in  Ekrixinatosaurus, Aucasaurusand Quilmesaurus( Fig. 8B). The distal border is thicker than the body of the crest, suggesting strong insertion of muscles. The fibular crest of  Pycnonemosaurusis incomplete and filled with plaster, but appears to reach the ventralmost level of the cnemial crest and ending approximately at the first one-quarter of the bone’s shaft as seen in  Majungasaurus( Carrano, 2007). The preserved portion is oriented vertically as in  Masiakasaurus, Velocisaurusand Eoabelisaurus, whereas in  Carnotaurus, Aucasaurus, Skorpiovenatorand Xenotarsaurusthe fibular crest is slightly anteroposteriorly inclined.   FIGURE 5. Medium-sized isolated transverse process of  Pycnonemosaurus nevesi(DGM 859-R) in dorsal ( A) and ventral ( B) views. Scale bar equals 100 mm.   FIGURE 6. Smallest isolated transverse process of  Pycnonemosaurus nevesi(DGM 859-R) in dorsal ( A) and ventral ( B) views. Scale bar equals 100 mm. The lateral malleolus is well developed and ventrally expanded with an inclination of nearly 45° to the horizontal axis. In  Pycnonemosaurusthis structure is more pronounced than in the other abelisaurs.  Quilmesaurus, Ekrixinatosaurus,  Majungasaurusand  Skorpiovenatoralso have ventral expansion but to a lesser degree than  Pycnonemosaurus. The articular facet of the astragalus of  Pycnonemosaurus( Fig. 9) is tongue-shaped and is not fused with the tibia as seen in the unfused astragalus of  Quilmesaurus, Cambarátaxon ( Machado et al., 2013),  Rajasaurus( Wilson et al., 2003)and  Majungasaurus( Carrano, 2007). Its surface is narrower than figured by Kellner & Campos (2002)and the scars suggest a small ascending process with subparallel margins like in  Ekrixinatosaurus( Calvo et al., 2004; Rauhut, 2012). This condition is quite different to  Rajasaurusand  Quilmesauruswho have the articular facet broader and triangular.   FIGURE 7. Right pubis of  Pycnonemosaurus nevesi(DGM 859-R) in anterior ( A), lateral ( B), posterior ( C) and medial ( D) views. Abbreviations: dpf, distal pubic foot. Scale bar equals 100 mm.   FIGURE 8. Right tibia from  Pycnonemosaurus nevesi(DGM 859-R) in anterior ( A), lateral ( B), posterior ( C), medial ( D), dorsal ( E) and ventral ( F) views. Abbreviations: cn, cnemial crest; fcr, fibular crest; lct, lateral condyle; lf, lateral fossa; lm, lateral malleolus; mm, medial malleolus. Scale bar equals 100 mm.   FIGURE 9. Distal end of the tibia of  Pycnonemosaurus nevesi(DGM 859-R) in anterior view. The ellipse shows the scar of ascending process of the astragalus. Without scale bar.  Fibula. The right fibula is incomplete preserving only the distal portion. The shaft is flattened mediolaterally, gently posteriorly bowed and distally expanded, with a slight twist and many foramina on its surface ( Fig 10). In anterior view the distal end is marked with a medial dorsoventral bulge as seen in the fibulae of  Aucasaurus, Skorpiovenatorand  Masiakasaurus(Carrano et al., 2002: fig. 15A). In Xenotarosaurusand  Eoabelisaurusthis structure is not preserved. The articular surface is semi-circular-shaped and robust, as seen in  Skorpiovenator, Xenotarsosaurus, Aucasaurusand Eoabelisaurus. However, in Eoabelisaurus, the distal articulation has a cranial projection on the distal end. The distal end of the fibula of  Masiakasaurusis square-shaped (Carrano et al., 2002: fig. 15B) unlike the one in  Pycnonemosaurus nevesi,in which the distal margin is rounded.  Other DGM 859-R elements.The  Pycnonemosaurus nevesiholotype described by Kellner & Campos (2002)did not include the isolated neural spine, the three transverse processes, a distal caudal centrum and an unidentified incomplete bone with the same number (DGM 859-R). The authors refer to these materials as “several unidentified incomplete bones”. Here, I have identified three transverse processes as belonging to an abelisaurid dinosaur, as described above. The isolated neural spine and distal caudal centrum as belonging to a sauropod dinosaur, the incomplete bone perhaps pertaining to a sauropod with the other isolated transverse process being unidentified taxa. Here, I described these specimens as the Parecis taxon.   FIGURE 10. Right fibula of  Pycnonemosaurus nevesi(DGM 859-R) in anterior ( A), lateral ( B), posterior ( C) and medial ( D) views. Scale bar equals 100 mm. The neural spine ( Fig 11A) differs from sauropod titanosaurs, such as  Trigonosaurus pricei,  Baurutitan britoiand  Uberabatitan ribeiroi. In these taxa the neural spines are distally broad in the mid-caudal vertebrae ( Salgado & Carvalho, 2008: fig. 14). However, the proportion of the neural spine present in Parecis taxon is similar to the mid and posterior caudal vertebrae of  Aeolosaurus maximus( Santucci & Arruda-Campos, 2011: fig. 4). In these neural spines, the proximal bases are broad, whereas the distal ends are laterally narrow. In the posterior view, the isolated neural spine is wide, whreas in the anterior view it is narrow as similarly seen in  Aeolosaurus maximus( Santucci & Arruda-Campos, 2011: fig. 4). These proportions are clearly different from those of  Carnotaurusand  Aucasaurus. Therefore, this neural spine is assigned to an indeterminate sauropod and not to an unidentified bone of  Pycnonemosaurus nevesi. The distal caudal centrum ( Fig 11B) is abraded and is procoelous as seen in titanosauran sauropods (Novas, 2009), differing from the amphicoelous condition of  Majungasaurus( O’Connor, 2007),  Carnotaurus, Aucasaurus, Ilokelesiaand Ekrixinatosaurus. Thiscondition suggests that the Parecis taxon distal caudal belongs to a titanosaur sauropod. The isolated transverse process ( Fig 11C) belongs to unidentified taxa. It differs from the other three, that have straight borders. This element is very abraded and too incomplete to assign to another taxon with certainty. The unidentified bone is a large element with a tip marked with sulci. This bone exhibits pneumaticity, but it is not possible to assign it to a theropod. Pneumaticity in theropods (e.g.  Aerosteon; Sereno et al., 2009), is gracile with narrow laminae, whereas in Parecis taxon it is robust with wide laminae. Neither abelisaurid theropod shows pneumatisation in the post-cranial skeleton as observed in this unidentified bone. For this reason, I cannot assign this bone to  Pycnonemosaurus.   FIGURE 11. Parecis taxa elements undescribed by Kellner & Campos (2002), and identified here as an isolated neural spine belonging to a sauropod dinosaur in right lateral view ( A); a distal caudal centrum belonging to a sauropod dinosaur in left lateral view ( B) and an isolated transverse process of an unidentified taxon in dorsal view ( C). Although labelled as DGM 859 R, they are not belonging to the holotype of  Pycnonemosaurus nevesi. Scale bar equals 100 mm.   FIGURE 12. Cladogram based on modified dataset of Filippi et al. (2016). The topology represents the strict consensus of all the 125 recovered MPCs with a length of 851 steps (CI.: 0.61; R.I.: 0.744). Numbers at the branches represent Bremer support. All analyses were performed using TNT (Goloboff et al., 2008). Abbreviation: Majun., Majungasaurinae. FIGURE 13. Hypothetical reconstruction of the Brazilian Upper Cretaceous fauna.  Pycnonemosaurus nevesi(left side); a titanosaurian sauropod (centre-right on side); a medium-sized maniraptoran (near the neck of the sauropod) and a megaraptorid (behind the sauropod). Art by Pedro Rodrigues Busana. 1563457668 DGM Museu de Ciencias da Terra & It & Nevertheless & Thus Brazil Campos Earth Sciences Museum Kellner 1 2 DGM 859 2 Rio de Janeiro holotype 1563457669 Brazil Langer Fazenda Roncador Bittencourt 1 2 1 Mato Grosso holotype