Hudiesaurus sinojapanorum, Dong, 1997

Upchurch, Paul, Mannion, Philip D., Xu, Xing & Barrett, Paul M., 2021, Re-assessment of the Late Jurassic eusauropod dinosaur Hudiesaurus sinojapanorum Dong, 1997, from the Turpan Basin, China, and the evolution of hyper-robust antebrachia in sauropods, Journal of Vertebrate Paleontology (e 1994414) 41 (4), pp. 1-31 : 3-9

HUDIESAURUS SINOJAPANORUM Dong, 1997

( Figs. 2–4 )

Original Diagnosis — Re-written from Dong (1997:102): (1) top of neural spine of anterior dorsal vertebra forms a ‘U’-shaped shallow cleft; (2) wing-like process between bases of postzygapophyses and lateral margin of neural spine; (3) anteriorly directed laterally compressed ‘sword-like’ process on anterior face of neural spine; (4) deep pleurocoels on lateral faces of the centrum; (5) midline keel on the ventral surface of the centrum.

Comments on Original Diagnosis — The original diagnosis provided by Dong (1997) can now be shown to be inadequate. Putative autapomorphies 1, 4, and 5 are present in several other sauropod genera. For example, shallow ‘U’-shaped bifurcation of the posterior cervical and anterior dorsal neural spines also occurs in Mamenchisaurus ( Young and Chao, 1972) , Klamelisaurus ( Zhao, 1993; Moore et al., 2020), Euhelopus ( Wiman, 1929; Wilson and Upchurch, 2009 ), several turiasaurians ( Royo-Torres et al., 2006, 2017; Britt et al., 2017), Camarasaurus ( Osborn and Mook, 1921; Gilmore, 1925), and Opisthocoelicaudia (Borsuk-Białynicka, 1977), among others. Deep lateral pneumatic openings (= ‘pleurocoels’) are widespread in the presacral centra of many eusauropods ( Upchurch et al., 2004a), and a ventral keel is also present in the cervicodorsal region of several other taxa, including Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ. 2010), Klamelisaurus ( Moore et al., 2020), and Euhelopus ( Wilson and Upchurch, 2009) . It is not entirely clear what Dong (1997) meant by the ‘wing-like’ processes (putative autapomorphy ‘2’), as their location was neither fully described nor annotated in his figures. However, it seems likely that these are merely the typical posterolateral projection of the postzygapophyses, rather than unusual processes. Finally, the ‘sword-like’ anterior process is not part of a novel articulation with the hyposphene of a preceding vertebra (contra Dong, 1997: see Description, below); rather, it appears to be a transversely compressed sheet of ossified intervertebral ligament. Ossification of such ligaments and tendons is rare, but not unheard of, among sauropods (e.g., Camarasaurus [= ‘ Cathetosaurus ’] lewisi [ Jensen, 1988]; Diplodocus [ USNM 10865; Gilmore, 1932; PU pers. observ., 1991]; see also Cerda, 2009; Klein et al., 2012; Cerda et al., 2015). Thus, the presence of such a feature is more likely to represent individual variation, pathology, and/or unusual preservation, rather than an autapomorphy. If this feature is to be accepted as having some diagnostic value, this must wait until it is found repeatedly in other individuals of Hudiesaurus.

Revised Diagnosis —Hudiesaurus can be diagnosed on the basis of the following autapomorphies: (1) small projection on neurocentral junction above lateral pneumatic opening; (2) ACDL splits into upper and lower branches (the former extends to anterodorsal margin of the diapophysis, and the latter to posteroventral margin of the diapophysis, where it meets the anterior end of the PCDL); (3) approximately transverse row of 5–6 small coels on dorsal surface of prezygapophyseal process, immediately posterior to articular facet; (4) SPRLs bifurcate close to the base of the metapophysis, with one branch extending up anterior surface and fading out before reaching the summit, and the other branch forming a thin sheet that extends along the anterolateral margin of the metapophysis to the summit; and (5) SPOL bifurcates into two distinct ridges immediately above postzygapophysis (or this could be described as a short lamina extending dorsomedially from the PODL to the SPOL). N.B., portions of the PRDLs and diapophyses have been heavily restored with plaster, so autapomorphy 2 should be treated with caution.

Holotype — A nearly complete vertebra from the cervicodorsal region (estimated to be the last cervical vertebra; IVPP V11120 ) ( Figs. 2–4 ; Table 1 ). N.B., Dong (1997) identified this specimen as an anterior dorsal vertebra, but we regard it as being more probably a posterior cervical vertebra (see below).

Locality and Horizon — Lower part of the Kalazha Formation (Upper Jurassic: upper Kimmeridgian–Tithonian) of Qiketai , Shanshan County, Turpan Basin, Xinjiang Uyghur Autonomous Region, China ( Dong, 1997; Deng et al., 2015; Fang et al., 2016; Fig. 1 ).

Description and Comparisons

Dong (1997) identified the holotype of Hudiesaurus as an anterior dorsal vertebra; however, it also resembles a posteriormost cervical vertebra in several features. Even with well-preserved presacral series, it is often difficult to define the point where the neck meets the trunk in sauropods: this is because the morphology of the posterior cervical vertebrae gradually transforms into that of the most anterior dorsal vertebrae ( Wilson and Upchurch, 2009 ; Moore et al., 2020). Despite some occasional doubts and apparent inconsistencies, we have generally accepted the identifications of the cervical-dorsal junction proposed by previous workers for other taxa. However, in the case of Mamenchisaurus hochuanensis (CCG V 20401 ), we note that the suggested 19 cervical and 12 dorsal vertebrae ( Young and Chao, 1972) is likely to be incorrect. This is because ‘Dv2’ possesses a hyposphene (PU and PMB pers. observ., 2010), which would be atypical for such an anterior dorsal vertebra: a hyposphene does not usually appear until Dv3 or Dv4 in sauropods ( Upchurch et al., 2004a). We therefore propose provisionally that Mamenchisaurus hochuanensis had 18 cervical and 13 dorsal vertebrae. Given the difficulties of pinpointing the cervical-dorsal junction in even well preserved and complete presacral series, identifying the precise position of an isolated vertebra (such as Hudiesaurus) is even more problematic. Below, we compare the Hudiesaurus vertebra with both the posterior cervical and anterior dorsal vertebrae of other sauropods. The majority of features support a position as either the last cervical or the first dorsal vertebra, with the former being more probable based on some features that are uniquely shared by Hudiesaurus and the last cervical vertebra (Cv18) of Xinjiangtitan. This identification, of course, depends on the assumption that Zhang et al. (2020) were correct when they placed the cervical-dorsal junction of Xinjiangtitan between the 18th and 19th presacral vertebrae (counting from the head).

The Hudiesaurus vertebra is relatively complete, although the PRDLs and transverse processes have been partly reconstructed (see also Dong, 1997). As in the cervical and anterior dorsal vertebrae of most eusauropods, it has a strongly opisthocoelous centrum ( Dong, 1997) ( Fig. 2 ), differing from the amphiplatyan/amphicoelous presacral vertebrae of most non-gravisaurian sauropodomorphs ( Upchurch, 1995 ; Wilson, 2002; Upchurch et al., 2007a; Yates, 2007; Allain and Aquesbi, 2008; McPhee et al., 2014). In anterior or posterior view, the centrum is subcircular in outline, being slightly wider transversely than dorsoventrally ( Table 1 ), as is typical for the cervicodorsal vertebrae of neosauropods ( Mannion et al., 2019a) and some earlier-branching forms such as Qijianglong, Mamenchisaurus youngi , and Bellusaurus ( Moore et al., 2020 and references therein). This contrasts with the transversely compressed middle–posterior cervical centra of many other East Asian eusauropods, including Shunosaurus , Erketu , Euhelopus , Mamenchisaurus hochuanensis (CCG V 20401 ), and Xinjiangtitan ( Upchurch, 1998 ; Mannion et al., 2013; Moore et al., 2020; Zhang et al., 2020; PU and PMB pers. observ., 2010), as well as most rebbachisaurids ( Mannion et al., 2019a). The Functional (i.e., excluding the anterior convexity) Average Elongation Index (FAEI) is 1.0 in the Hudiesaurus vertebra. FAEIs tend to decrease towards the cervical-dorsal junction compared with those for middle cervical vertebrae, and a value close to 1.0 is compatible with a position either as the last cervical or one of the first two dorsal vertebrae of a non-diplodocine sauropod ( Table S1 in Supplemental Data 1). As in Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010), Klamelisaurus ( Moore et al., 2020; contra Zhao, 1993), Euhelopus ( Wilson and Upchurch, 2009) , and many flagellicaudatans ( Upchurch et al., 2004a), the ventral surface of the Hudiesaurus centrum is strongly concave transversely as well as anteroposteriorly over its whole length, and is bounded by ventrolaterally directed ridges ( Dong, 1997). A prominent midline ridge is present within the ventral concavity, as also found in dicraeosaurids ( Upchurch, 1998 ; Wilson, 2002), Cv17–Dv1 of Euhelopus ( Wilson and Upchurch, 2009) , posterior cervicals to Dv2 in Klamelisaurus ( Moore et al., 2020), Cv13–18 in Xinjiangtitan ( Zhang et al., 2020), and Dv1 (= ‘Cv19’) in Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010).

The parapophysis is located at the anteroventral corner of the lateral surface of the centrum ( Fig. 2 ). This position is typical for sauropod cervical vertebrae, although it also occurs in Dv1 in most taxa ( Upchurch et al., 2004a), including Klamelisaurus ( Moore et al., 2020), Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010), and Xinjiangtitan ( Zhang et al., 2020), and in Dv1 and 2 in Euhelopus ( Wilson and Upchurch, 2009) and Apatosaurus ajax ( Upchurch et al., 2004b). In Hudiesaurus, there is no indication that the shallowly concave articular surface of the parapophysis was fused to a rib: this is more consistent with this specimen being a dorsal, rather than cervical, vertebra ( Hatcher, 1901; Gilmore, 1936; McIntosh, 1990; Upchurch, 1998 ; Upchurch et al., 2004a; Zhang et al., 2020). However, rib–vertebra fusion is not an infallible indicator that a vertebra is a cervical ( Moore et al., 2020): for example, the ribs of Cv17 and 18 of Mamenchisaurus hochuanensis (CCG V 20401 ) are not fused to the parapophyses (PU and PMB pers. observ., 2010). The dorsal surface of the parapophysis is excavated in Hudiesaurus, and this depression is continuous with the lateral pneumatic opening, as seen in the cervical vertebrae of many non-neosauropod eusauropods, such as Cetiosaurus and Chebsaurus ( Upchurch and Martin, 2002, 2003; Upchurch et al., 2004a; Mahammed et al., 2005). Many neosauropods also have dorsally excavated cervical parapophyses, but such taxa typically possess a ridge that divides this depression from the lateral pneumatic opening ( Upchurch, 1998 ; Upchurch and Martin, 2002, 2003). The lateral pneumatic opening of Hudiesaurus is small and deep, with a rounded, wide anterior margin that is positioned dorsal to the parapophysis ( Fig. 2 ). Posteriorly, this opening is bounded dorsally by a sharp ridge that runs posteroventrally, giving the posterior margin an acute profile. Such a ridge is unusual in sauropods, only being reported previously in Cv17 and 18 of Xinjiangtitan ( Zhang et al., 2020:figs. 15, 16, and 18), and confirmed as absent in Mamenchisaurus youngi by the latter study. Dorsal vertebrae 1 and 2 of Apatosaurus ajax have a ridge bounding the lateral pneumatic opening dorsally ( Upchurch et al., 2004b), but this differs from the condition in Hudiesaurus and Xinjiangtitan by extending further anteriorly (i.e., to the anterior end of the opening) and being horizontal rather than posteroventrally inclined. In Hudiesaurus, this ridge merges into the centrum-arch junction, where there is a small, laterally extending projection on each side ( Fig. 2 ): the latter is unique and is regarded as an autapomorphy. The presence of lateral pneumatic openings with oval outlines (i.e., strongly rounded and dorsoventrally wide anterior margins and acute posterior ends) in anterior dorsal vertebrae has frequently been regarded as a derived character state uniting Macronaria or a slightly less inclusive clade (e.g., Upchurch, 1998 ; Mannion et al., 2013). However, they are also seen in Dv1 and 2 of Klamelisaurus ( Moore et al., 2020), the anterior dorsal vertebrae of Bellusaurus and Haplocanthosaurus priscus ( Mannion et al., 2019a), and indeterminate cervicodorsal vertebrae from the Late Jurassic Shishugou Formation of China ( Moore et al., 2020). In Hudiesaurus, the lateral pneumatic opening is not as elongate as those found in either the cervical centra of Cetiosaurus ( Upchurch and Martin, 2002) or several Jurassic Chinese taxa (such as Dashanpusaurus and Daanosaurus ; Peng et al., 2005; Ye et al., 2005). Indeed, Hudiesaurus possesses a lateral pneumatic opening that is largely restricted to the anterior two-thirds of the centrum (excluding the anterior articular convexity), a derived condition seen in the cervical vertebrae of many CMTs (e.g., Klamelisaurus, Mamenchisaurus youngi, Qijianglong, Xinjiangtitan ), Euhelopus , and several titanosauriforms ( Whitlock, 2011; Moore et al., 2020). However, the relatively small size and anterior location of the lateral pneumatic opening is also consistent with the Hudiesaurus vertebra being from the anterior dorsal region. The oblique accessory lamina that divides the lateral pneumatic opening into anterior and posterior sections in the cervical vertebrae of several non-neosauropod eusauropods (e.g., Mamenchisaurus, Klamelisaurus, Xinjiangtitan ) and many neosauropods ( Wilson, 2002; Upchurch et al., 2004a; Moore et al., 2020) is not present in Hudiesaurus ( Fig. 2 ). While its absence is more compatible with an identification of the Hudiesaurus specimen as being an anterior dorsal vertebra, the oblique lamina is also sometimes absent in posterior-most cervical vertebrae, such as Cv18 of Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010), Cv17 and 18 of Xinjiangtitan ( Zhang et al., 2020), and Cv17 of Euhelopus ( Wilson and Upchurch, 2009) . The lateral pneumatic opening becomes shallower posteriorly in Hudiesaurus, as is typical for most sauropod cervical vertebrae (e.g., Cetiosaurus , Patagosaurus , and the CCG V 20401 specimen of Mamenchisaurus hochuanensis: Bonaparte, 1986 ; Upchurch and Martin, 2002, 2003; PU and PMB pers. observ., 2010).

Measured on the anterior surface, the ratio of the dorsoventral height of the neural arch (from the dorsal surface of the centrum to the ventromedial tips of the prezygapophyses) to centrum height is low (∼0.35) in Hudiesaurus. With the exception of comparably low neural arches in some somphospondylans and Omeisaurus tianfuensis, this ratio is ≥0.5 in the posterior cervical vertebrae of other eusauropods ( Bonaparte et al., 2006; Mannion et al., 2013). In Hudiesaurus, the prezygapophyses project forward to a point beyond the anterior end of the condyle ( Fig. 2 ). Such projection is typical for the posterior cervical and anterior dorsal vertebrae of many sauropods: for example, in Klamelisaurus it is only posterior to Dv5 that the prezygapophyses no longer project beyond the anterior articulation of the centrum ( Moore et al., 2020). However, this contrasts with the condition in taxa like Apatosaurus ajax, where the prezygapophyses no longer project beyond the anterior end of the centrum from Cv12 rearwards ( Upchurch et al., 2004b). In Hudiesaurus, the prezygapophyses are large and broad, with transversely convex articular surfaces ( Fig. 3A ). Sauropods typically have flat prezygapophyseal articular surfaces plesiomorphically, but the derived, strongly convex condition is also present in the cervical vertebrae of diplodocines ( Upchurch, 1995 ; Tschopp et al., 2015a) and the CMTs Klamelisaurus ( Moore et al., 2020) and Xinjiangtitan ( Zhang et al., 2020), as well as the anterior dorsal vertebrae of Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010). The zygapophyses have several small, irregularly shaped coels on their dorsal surfaces ( Dong, 1997). In the case of the prezygapophyses, these coels form a line of 5–6 adjacent pits, separated from each other by small anteroposteriorly directed ridges, located immediately posterior to the articular facet ( Fig. 3A ). These might represent a pneumatized internal tissue structure that has been revealed by erosion of the surface bone: however, their presence in the same position on both prezygapophyses suggests that they are not taphonomic artifacts. We therefore regard these coels as external pneumatic features and as autapomorphic for Hudiesaurus. The thin, medial edges of the prezygapophyses descend steeply to meet each other on the midline and form a single lamina extending down to the top of the small, subcircular neural canal ( Fig. 2C ); this is probably the “well developed medial lamina” of Dong (1997:103), here termed the interprezygapophyseal lamina (TPRL) according to a revised version of Wilson’ s (1999) system (see Tschopp and Mateus, 2013). This TPRL partially subdivides the centroprezygapophyseal fossa (CPRF) into left and right subfossae. A TPRL is absent from the posterior cervical vertebrae of Euhelopus ( Wilson and Upchurch, 2009) and Xinjiangtitan ( Zhang et al., 2020), and the anterior dorsal vertebrae of Klamelisaurus and Mamenchisaurus youngi ( Moore et al., 2020) , although it is present in several other sauropods (e.g., there is a short, stout version on the posterior cervical vertebrae of Apatosaurus ajax; Upchurch et al., 2004b). The centroprezygapophyseal laminae (CPRLs) of Hudiesaurus are large and stout (as in Cetiosaurus ; Upchurch and Martin, 2003) and do not bifurcate at their dorsal ends, unlike those of the cervical vertebrae of several diplodocids ( Upchurch, 1998 ) and many non-neosauropod eusauropods ( Moore et al., 2020), such as those on Cv18 in Xinjiangtitan ( Zhang et al., 2020). The stout, single CPRLs of Hudiesaurus more closely resemble those of anterior dorsal vertebrae in taxa such as Klamelisaurus, although the former lacks the accessory laminae seen in the PRCDF of the latter taxon ( Moore et al., 2020). In lateral view, the CPRLs slope anterodorsally and are subparallel with the PCDLs ( Fig. 2A, B ), a configuration also seen in the cervical and anterior-most dorsal vertebrae (i.e., Dv1 and 2) of many sauropods. By contrast, in Dv3 and 4 of most taxa, these laminae become more vertical, and are fully vertical from around Dv5 onwards, as seen in Klamelisaurus ( Moore et al., 2020). Thus, the orientation of the CPRLs further supports the view that the Hudiesaurus vertebra is either a cervical or one of the most anterior dorsal vertebrae. As in the cervical vertebrae of some non-neosauropod eusauropods (including Shunosaurus , Omeisaurus tianfuensis, Chuanjiesaurus, and Cetiosaurus ) and many diplodocoids, pre-epipophyses are absent in Hudiesaurus. This contrasts with most CMTs, such as Klamelisaurus and Mamenchisaurus youngi , as well as Bellusaurus , Euhelopus , and many other neosauropods, in which these projections are welldeveloped ( Wilson and Upchurch, 2009 ; Mannion et al., 2013, 2019a; Moore et al., 2020). However, pre-epipophyses are typically absent in the dorsal vertebrae of sauropods ( Wilson and Upchurch, 2009 ), so the condition in Hudiesaurus might merely reflect a location in the anterior dorsal series.

The transverse processes are short and project laterally and slightly ventrally ( Dong, 1997), although it is difficult to ascertain how genuine this morphology is, given the degree of plaster restoration. If the transverse processes are truly pendant, then this is consistent with this specimen being either a cervical or very anterior dorsal vertebra ( Upchurch et al., 2004a). For example, the shift from pendant to horizontal transverse processes occurs between Cv18 and Dv2 in Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010), from Cv17 to Dv2 in Euhelopus ( Wilson and Upchurch, 2009) , and more abruptly between Cv18 and Dv1 in Xinjiangtitan ( Wu et al., 2013; Zhang et al., 2020). In Hudiesaurus, the transverse process lies some distance below the level of the zygapophyses ( Fig. 2 ), as is typical for posterior cervical and the most anterior dorsal vertebrae ( Moore et al., 2020). Prominent anterior and posterior centrodiapophyseal laminae (ACDLs, PCDLs) extend anteroventrally and posteroventrally, respectively, at approximately 45° to the horizontal ( Fig. 2 ). The presence of an ACDL is consistent with this specimen being either a cervical or anterior dorsal vertebra: for example, in Klamelisaurus, the ACDL is present in Dv1 and 2 as a separate lamina, and in Dv3 and 4 merges into the CPRL ( Moore et al., 2020; see also Wilson, 1999). As the ACDL approaches the transverse process in Hudiesaurus, it bifurcates to form two laminae that extend along the ventral and anterior surfaces of the transverse process (potentially as far as the distal articular end) ( Fig. 3B ). The more posterior of these laminae merges into the posteroventral margin of the transverse process, where it meets the anterodorsal end of the PCDL. This posteriorly bifurcate ACDL appears to be unique to Hudiesaurus. The relatively steeply inclined PCDL is consistent with the identification of the Hudiesaurus vertebra as a posterior-most cervical or an anterior dorsal vertebra: this lamina is typically close to horizontal in cervical vertebrae but tends to become more steeply inclined in the cervicodorsal region ( Wilson and Upchurch, 2009 ). Sauropods display some variation in this regard, although this might also reflect inconsistent identification of the cervical-dorsal junction. For example, PCDLs remain shallowly inclined even in the most posterior cervical vertebrae of Qijianglong ( Xing et al., 2015:fig. 12F), but they become increasingly steep from Cv16 to 18 in Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010). In Hudiesaurus, the prezygodiapophyseal lamina (PRDL) extends anterodorsally from the transverse process to the prezygapophysis at a moderate angle (c. 30°) to the horizontal, whereas the postzygodiapophyseal lamina (PODL) is nearly vertical ( Fig. 2 ). The anterior margin of the PRDL forms a convex projection or ‘kink’ ( Figs. 2 , 3 ) that is potentially homologous with the apomorphically convex ventral margin seen in the middle and posterior cervical vertebrae of several CMTs ( Moore et al., 2020). Unlike the condition in the cervicodorsal vertebrae of Euhelopus, Klamelisaurus , and some additional CMT specimens ( Moore et al., 2020), the PODL is not bifid ventrally.

The posterior margins of the postzygapophyses terminate some distance anterior to the posterior margin of the centrum ( Fig. 2 ). This condition is a derived state when it occurs in posterior cervical vertebrae, which is seen in several non-neosauropod eusauropods (e.g., Omeisaurus tianfuensis — He et al., 1988:fig. 23; Mamenchisaurus youngi — Ouyang and Ye, 2002:fig. 18C; Chuanjiesaurus— Sekiya, 2011:fig. 14; Qijianglong— Xing et al., 2015:fig. 12F; Xinjiangtitan— Zhang et al., 2020:figs. 15 and 16; Jobaria— Mannion et al., 2017), and early diverging macronarians (e.g., Camarasaurus ; Osborn and Mook, 1921:pl. LXVII), but is typically absent in many diplodocoids, including Apatosaurus ajax ( Upchurch et al., 2004b), Dicraeosaurus ( Janensch, 1929, 1936:table I, fig. 11a), and Limaysaurus ( Calvo and Salgado, 1995:fig. 8B) (see also Tschopp and Mateus, 2013; Tschopp et al., 2015a; Poropat et al., 2016). Epipophyses are greatly reduced or absent in Hudiesaurus, perhaps being represented by small tab-like processes above the postzygapophyses ( Fig. 3 ). Such a condition is typical for the posterior-most cervical vertebrae of sauropods, except Euhelopus ( Wilson and Upchurch, 2009) , Jobaria (MNN specimens; PDM pers. observ., 2012), Nigersaurus (MNN specimens; PDM pers. observ., 2010), and diplodocines ( Tschopp and Mateus, 2013). For example, epipophyses are present in Cv2–16 in Xinjiangtitan, but are absent in Cv17 and 18 ( Zhang et al., 2020). Their absence is also consistent with the Hudiesaurus vertebra being an anterior dorsal, since it is even rarer for well-developed epipophyses to be present on such vertebrae (to date they have only been reported in anterior dorsal vertebrae of some turiasaurians ( Britt et al., 2017; Mannion, 2019; Mannion et al., 2019a), although they can be traced into the dorsal series as the homologs of the tips of the aliform processes in Euhelopus ( Wilson and Upchurch, 2009) . Given the uncertainty in the position of the Hudiesaurus vertebra, and the subtlety of its putative epipophyses, we score this character (i.e., presence/absence of epipophyses) as a ‘?’ in our phylogenetic data matrices. The postzygapophyses of Hudiesaurus are relatively large, with concave articular surfaces facing downwards and outwards ( Fig. 2D ). Their ventral margins merge into the dorsal parts of well-developed centropostzygapophyseal laminae (CPOLs) that descend separately without meeting on the midline; however, the detailed anatomy of this region is obscured by damage and reconstruction. Nevertheless, despite Dong’ s (1997) assertion of its presence, there is no hyposphene-hypantrum articulation (see above). On the left side at least, and possibly also the right, the CPOLs bifurcate dorsally, creating a small subtriangular fossa that faces mainly posteriorly ( Fig. 2D ). A dorsally bifurcated CPOL is sporadically present in the middle and posterior cervical vertebrae of eusauropods (e.g., Cetiosaurus , Patagosaurus , Camarasaurus , Giraffatitan , Rapetosaurus , and some flagellicaudatans), and is generally absent in CMTs apart from the ‘Phu Kradung taxon’ ( Tschopp et al., 2015a; Carballido et al., 2017; Moore et al., 2020). However, the medial branch of the bifid CPOL of Hudiesaurus supports the postzygapophysis rather than curving medially to meet its partner on the midline as occurs in other taxa. Similarly, no single vertical midline interpostzygapophyseal lamina (TPOL) can be observed, although it is not clear whether this represents genuine absence or the effects of poor preservation.

The spinoprezygapophyseal laminae (SPRLs) are low ridges that extend medially from the middle of the posterior margins of the prezygapophyses to the anterior bases of the metapophyses ( Figs. 2 , 4 ). At this point, each SPRL autapomorphically splits into two branches: one ascends the anterior surface of the metapophysis and fades out at about midheight; the other becomes a thin flange-like ridge that extends along the anterolateral margin of the metapophysis and reaches the summit. These anterolateral flanges are potentially homologous with the ‘scabrous’ projections observed in the middle–posterior cervical vertebrae of Klamelisaurus (which become less ‘ragged’ in the most posterior cervical vertebrae), and the dorsolaterally flattened SPRLs seen in the middle and posterior cervical vertebrae of Bellusaurus ( Moore et al., 2020) . In Hudiesaurus, there is a large flat space on the anterior surface of the neural spine between the SPRLs and below the bifurcated summit. Near the top of this area, along the midline, is the base of a transversely compressed process ( Figs. 2 , 4 ): this is the feature that Dong (1997) described as an 84 mm long, anteriorly directed, ‘swordlike’ process (for which he used the term ‘prepophysis’). We observed this process in our first examination of this specimen in 1995, but by our second examination, in 2007, we found that the process had been broken and lost, so that now only its base is preserved. Dong (1997) suggested that this structure might be for the insertion of muscles, or for articulation with the hyposphene of the preceding vertebra; however, the latter proposal would seem to be impossible because the location of the process on the spine means that it would project into the spinopostzygapophyseal fossa (SPOF: = postspinal fossa) of the preceding vertebra. Moreover, hyposphene-hypantrum articulations have not been observed in the posterior cervical or anterior-most dorsal vertebrae of any sauropod: such structures are restricted to middle and posterior dorsal vertebrae ( Upchurch et al., 2004a). We instead interpret this structure to be part of an ossified ligament (see above).

The posterior margin of the neural spine slopes strongly forward in lateral view, and the spine is slightly anterodorsally directed (though not to the same extent as in Dicraeosaurus ; Janensch, 1929). The neural spine of Cv16 in Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010) has a nearly vertical anterior margin and gently sloping posterior one, resembling that of Hudiesaurus. This contrasts with the posterior-most cervical vertebrae of some taxa, such as Qijianglong ( Xing et al., 2015:fig. 12E, F), in which the neural spine has a fairly symmetrical lateral profile, with posterodorsally sloping anterior and anterodorsally sloping posterior margins. As in the cervicodorsal vertebrae of CMTs, turiasaurians, Camarasaurus , and some titanosaurs, the neural spine is relatively low in Hudiesaurus, projecting only slightly above the level of the postzygapophyses ( Mannion et al., 2019a; Moore et al., 2020). The neural spine is bifurcated ( Fig. 2C, D ), as in the presacral vertebrae of numerous other eusauropods (Klamelisaurus, Mamenchisaurus, Qijianglong , some turiasaurians, flagellicaudatans, Camarasaurus , Euhelopus , and several somphospondylans; Wiman, 1929; Young, 1954; Borsuk-Białynicka, 1977; Zhao, 1993; Wilson, 2002; Harris and Dodson, 2004; Upchurch et al., 2004a; Royo-Torres et al., 2006; Ksepka and Norell, 2006; D’ Emic et al., 2013; Mannion et al., 2019a; Moore et al., 2020). In anterior and posterior views ( Fig. 2C, D ), the metapophyses are divergent, as in diplodocids and most other taxa with bifid neural spines, but unlike the derived condition seen in dicraeosaurids, in which these structures are subparallel or converge towards their summits ( Rauhut et al., 2005; Xu et al., 2018). In Hudiesaurus, the notch between the metapophyses is moderately deep and ‘U’- shaped, with a median tubercle at its base ( Fig. 2C, D ). Such a tubercle is variably present in other sauropods with bifid presacral spines: for example, it occurs in the last two cervical vertebrae and Dv1–4 of Euhelopus , where it is drawn out into a large process that is as prominent as the metapophyses ( Wilson and Upchurch, 2009 ); it is present as a low rounded process in the last two cervical vertebrae and Dv1–3 of Barosaurus ( Zhang et al., 2020) ; it is a small bump on the posterodorsal margin of the notch in Klamelisaurus ( Moore et al., 2020); it is variably absent/present in specimens of Camarasaurus ( Tsuihiji, 2004) ; and it is absent in Mamenchisaurus, Qijianglong, Suuwassea , and Amargasaurus ( Wilson, 2002; Harris and Dodson, 2004; Xing et al., 2015). The metapophyses of Hudiesaurus are knob-like and subtriangular in dorsal view, robust rather than compressed transversely, and relatively short dorsoventrally (not elongated as in derived dicraeosaurids: Janensch, 1929; Xu et al., 2018).

The spinodiapophyseal fossa (SDF), posterior to the SPRL and anterior to the SPOL, is divided into three subtriangular coels by two accessory laminae or ridges ( Fig. 4 ). Dong (1997:103) described these structures as forming “a V-shaped posterolaterally projecting lamina”: in lateral view, the two laminae meet each other at their posterior ends and diverge anteriorly. This ‘V’ is created from a lower horizontal lamina that extends from the PODL to the base of the SPRL, and an upper anterodorsally directed lamina that extends from the posterior end of the horizontal lamina to the posterior margin of the anterolateral branch of the SPRL (see above). Although both of these ridges are found separately on the presacral vertebrae of many sauropods (see below), the presence of both of them in this ‘V’-shaped configuration is only known in Cv18 of Xinjiangtitan ( Zhang et al., 2020:figs. 16A, 17B) and Hudiesaurus. The lower, horizontal, lamina is reminiscent of the ‘epipophysealprezygapophyseal lamina’ (EPRL) that occurs in the cervical vertebrae of several sauropods, such as Nigersaurus ( Sereno et al., 2007) and Euhelopus ( Wilson and Upchurch, 2009) , as well as some other dinosaurs ( Moore et al., 2020). Occasionally, this structure can also occur in the anterior-most dorsal vertebrae, such as Dv1 and 2 in Euhelopus , where it partially divides the SDF into lower and upper portions ( Wilson and Upchurch, 2009 ), and Dv1 of Klamelisaurus ( Moore et al., 2020). However, Moore et al. (2020) demonstrated that simply identifying this structure as the EPRL is problematic because it can be formed by either one or both of two separate components. One component is a more anteriorly placed ridge (termed the horizontal accessory lamina) that lies fully within the SDF and was probably formed by pneumatization. The other component is a more posteriorly placed ‘anterior epipophyseal’ epaxial muscle scar that lies on the lateral surface of the postzygapophyseal process and may project anteriorly into the posterior part of the SDF. Here, we identify the lower strut in Hudiesaurus as the horizontal accessory lamina formed by pneumatization. Moore et al.’ s (2020) survey of these structures among sauropods suggests that, when considering just posterior cervical vertebrae, the pneumatic strut is currently only known in rebbachisaurids (e.g., Nigersaurus, Limaysaurus ), Euhelopus (where it lies below, and separate from, the anterior epipophyseal muscle scar), and some CMTs such as Klamelisaurus and Mamenchisaurus hochuanensis (CCG V 20401 ; PU and PMB pers. observ., 2010). It can be confirmed as being absent in the posterior cervical vertebrae of some non-neosauropods such as Mamenchisaurus youngi (where it only occurs in middle cervical vertebrae: Zhang et al., 2020), as well as several macronarians in which it has previously been identified, including Camarasaurus lewisi, Europasaurus , Giraffatitan , and Uberabatitan . The anterodorsally directed ridge within the SDFs of Hudiesaurus and Xinjiangtitan is potentially a SPDL, though it contacts the PODL rather than the diapophysis directly. The SPDLs in Dv4 of Klamelisaurus and Dv3 of Euhelopus resemble this anterodorsal lamina, but no such structure occurs in the more anterior dorsal or posterior cervical vertebrae of these taxa ( Wilson and Upchurch, 2009 ; Moore et al., 2020). Despite the presence of two ridges produced by pneumatization within the SDF (i.e., the?SPDL and horizontal accessory lamina), Hudiesaurus lacks the 3–4 irregular coels in this region seen in several early-branching titanosauriforms and many CMTs ( Mannion et al., 2017; Moore et al., 2020). In Hudiesaurus, the SDF is not roofed dorsally by a horizontal rugose line of epaxial muscle scars immediately below the spine summit, unlike the condition in some non-neosauropod sauropods (e.g., Klamelisaurus, Jobaria, Mierasaurus, and Moabosaurus ), as well as most diplodocids and many non-titanosaurian macronarians ( Tschopp and Mateus, 2013; Mannion et al., 2019a; Moore et al., 2020). The prominent SPOLs of Hudiesaurus extend anteromedially and dorsally to the summit of each metapophysis ( Fig. 2 ). At its posteroventral end (above the postzygapophysis), the SPOL splits into two ridges, with a small subtriangular fossa (SPOL-F) between them ( Fig. 4 ). Such a bifurcated SPOL and cavity is not known in the posterior cervical vertebrae of other sauropods, but SPOL bifurcation in dorsal vertebrae has been listed as a synapomorphy of a clade of eusauropods comprising Barapasaurus , Omeisaurus , Mamenchisaurus , Patagosaurus, Jobaria , and neosauropods ( Wilson, 2002). However, the SPOL bifurcation noted by Wilson typically occurs in the middle and posterior dorsal vertebrae and has a very different structure. In the Barapasaurus + Neosauropoda clade, each SPOL is a single structure close to the postzygapophysis and then bifurcates into a lateral SPOL (which usually merges with the SPDL) and a medial SPOL (which usually meets its partner on the midline within the SPOF: Wilson, 1999, 2002). Aside from occurring in a more anteriorly placed presacral vertebra, the condition in Hudiesaurus also differs from other eusauropods in that the SPOL is single over most of the spine length and then bifurcates as it approaches the postzygapophysis. As such, irrespective of whether the Hudiesaurus specimen is a posterior cervical or anterior dorsal vertebra, it appears to possess an autapomorphic condition with regard to its SPOL bifurcation. The SPOF is large, ‘U’-shaped in transverse cross-section, and opens posterodorsally.

We could not observe the internal tissue structure of the vertebra. As such, we cannot determine whether the vertebra is camerate, as is the case in most eusauropods ( Wedel, 2003), or pneumatized by camellae, which characterizes the presacral vertebrae of titanosauriforms ( Wilson, 2002; Wedel, 2003) and many CMTs ( Young and Chao, 1972; Moore et al., 2020).

Figures

FIGURE 1. Map showing Xinjiang Autonomous Region in China, with a magnified inset showing the approximate location of the Hudiesaurus specimens within Shanshan County.

FIGURE 2. Posterior cervical vertebra of Hudiesaurus sinojapanorum (IVPP V11120; holotype). A, right lateral view; B, left lateral view; C, anterior view; D, posterior view. Abbreviations: acc.proc, accessory process; ACDL, anterior centrodiapophyseal lamina; CPOF, centropostzygapophyseal fossa; CPOL, centropostzygapophyseal lamina; CPRF, centroprezygapophyseal fossa; CPRL, centroprezygapophyseal lamina; dia, diapophysis; lig, ossified intervertebral ligament; mp, metapophysis; mt, median tubercle; PCDL, posterior centrodiapophyseal lamina; POCDF, postzygapophyseal centrodiapophyseal fossa; PODL, postzygodiapophyseal lamina; poz, postzygapophysis; pp, parapophysis; PRCDF, prezygocentrodiapophyseal fossa; PRDL, prezygodiapophyseal lamina; PRDL.k, kink in PRDL; prz, prezygapophysis; SDF, spinodiapophyseal fossa; SPOF, spinopostzygapophyseal fossa; SPOL, spinopostzygapophyseal lamina; SPRL, spinoprezygapophyseal lamina; TPOL, interpostzygapophyseal lamina; TPRL, interprezygapophyseal lamina. Scale bars equal 100 mm.

FIGURE 3. Posterior cervical vertebra of Hudiesaurus sinojapanorum (IVPP V11120; holotype). A, dorsal view; B, close up on anterior vertebral laminae supporting the diapophysis in right lateral view (not to scale). Abbreviations: ACDL, anterior centrodiapophyseal lamina; d.ACDL, dorsal branch of ACDL; l.ACDL, lateral branch of ACDL; dia, diapophysis;?epi, epipophysis; poz, postzygapophysis; PRDL.k, kink in PRDL; prz.p, pits on dorsal surface of prezygapophysis; SDF.c, pneumatic coel within spinodiapophyseal fossa; SPOL, spinopostzygapophyseal lamina. Scale bar equals 100 mm.

FIGURE 4. Posterior cervical vertebra of Hudiesaurus sinojapanorum (IVPP V11120; holotype). Close-up on the right lateral side of the neural spine in dorsolateral view to show pneumatic coels and accessory laminae within the spinodiapophyseal fossa (not to scale). Abbreviations: AHL, accessory horizontal lamina; lig, ossified intervertebral ligament; PODL, postzygodiapophyseal lamina; SPOL, spinopostzygapophyseal lamina; SPRL, spinoprezygapophyseal lamina.

Tables

TABLE 1. Measurements of the posterior cervical vertebra of Hudiesaurus sinojapanorum (IVPP V11120). All measurements in mm.

Description	Measurement
Anteroposterior length of centrum (including condyle)	466
Anteroposterior length of centrum (excluding condyle)	376
Dorsoventral height of centrum (posterior surface)	351
Transverse width of centrum (posterior surface)	398
Maximum transverse width across prezygapophyses	539
Maximum transverse width across postzygapophyses	439
Height of postzygapophyses above dorsal margin of	295
centrum
Transverse width across prezygapophyseal articular	178
surface
Transverse width across metapophyses	177