Rhomaleopakhus turpanensis Upchurch, Mannion, Xu, and Barrett 2021

(Figs. 6 – 10; Tables 3 and 4) Nomenclatural Acts — The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ‘ http: // zoobank. org /. ’ The LSID for this publication is: urn: lsid: zoobank. org: pub: A 42348 FE-ECE 6 - 4524 - B 536 - 857 AFFD 22 DB 2. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: CLOCKSS.

Description and Comparisons Humerus — The right humerus is nearly complete, apart from a portion of the proximomedial expansion (Dong, 1997) and a small part of the proximolateral corner (Figs. 6, 7 A, 8 A). The posterior surface of this element could not be examined fully due to its large size and storage within a protective cradle. It is a relatively robust element, with an estimated Humeral Robusticity Index (sensu Wilson and Upchurch, 2003) of 0.35, similar to those of other heavily built taxa such as Mamenchisaurus youngi, Apatosaurus, dicraeosaurids, and Opisthocoelicaudia (Upchurch et al., 2015: table 2). Proximally, the humerus expands laterally relative to the shaft, giving it an hourglass-shaped outline in anterior view; this is the plesiomorphic sauropod condition, contrasting with the more asymmetrical humeri of most titanosauriforms and turiasaurians (Tschopp et al., 2015 a; Poropat et al., 2016). The anterior surface of the humerus is too damaged proximally to determine whether a tuberosity for the attachment of the M. coracobrachialis was present. The deltopectoral crest of Rhomaleopakhus is more prominent than those of most sauropods and is similar to those in Turiasaurus (Royo-Torres et al., 2006) and brachiosaurids (Wilson and Sereno, 1998). The crest lies entirely on the anterolateral margin of the humeral shaft: it does not expand or project medially across the anterior surface (Fig. 7 A), unlike those in many titanosauriforms (Wilson, 2002; Mannion et al., 2013). It terminates at ∼ 44 % of humerus length from the proximal end: by comparison, values among other sauropods range between 35 – 50 % (Upchurch et al., 2015: table 2). In this respect, Rhomaleopakhus is almost identical to several other CMTs: for example, these values are 44 % in Anhuilong and Omeisaurus tianfuensis, and 43 % in Huangshanlong (Ren et al., 2018). In anterior view, the anterolateral margin of the deltopectoral crest has a sigmoid profile and is relatively narrow throughout its length. One unusual feature of the deltopectoral crest is that its distal terminus forms a narrow ridge that is offset medially and laterally from the rest of the crest surface by deep, dorsoventrally oriented grooves or breaks-in-slope: this is provisionally regarded as autapomorphic. Rhomaleopakhus lacks prominent ridges or bulges on the posterolateral surface of the shaft, at the level of the deltopectoral crest. Such projections occur in many titanosaurs, including Alamosaurus, Opisthocoelicaudia, Patagotitan, and Saltasaurus, and have been interpreted as the insertion sites of a number of muscles, including the M. latissimus dorsi, M. scapulohumeralis anterior, and M. deltoideus clavicularis, although these interpretations are debated (e. g., Borsuk-Białynicka, 1977; Otero, 2010, 2018; Upchurch et al., 2015; Moore et al., 2020; Otero et al., 2020; Voegele et al., 2020). In Rhomaleopakhus, as in most sauropods (Wilson, 2002; Mannion et al., 2013; Upchurch et al., 2015), the humeral shaft is wider transversely than anteroposteriorly, producing an elliptical horizontal cross-section at midlength. The transverse width of the shaft at midlength to proximodistal length ratio is estimated at 0.17 – 0.18. There is a small amount of torsion in the shaft, such that the long-axes of the proximal and distal end surfaces are slightly rotated relative to each other, but Rhomaleopakhus lacks the marked torsion (c. 40 °) seen in many diplodocids (Tschopp et al., 2015 a) and some CMTs (e. g., at least 30 ° in Klamelisaurus [Moore et al., 2020] and 25 ° in Huangshanlong [Huang et al., 2014] and Anhuilong (Ren et al., 2018]). Huang et al. (2014) regarded such humeral torsion as a synapomorphy of Mamenchisauridae, but there is clearly some variation among CMTs and homoplasy within Sauropoda, especially given that a strong degree of torsion of the humeral shaft is the plesiomorphic sauropodomorph condition that is lost in early sauropods (e. g., Yates, 2007; McPhee et al., 2014). The distal end of the humerus is relatively wide transversely compared with the width of the shaft at midlength, largely because it projects a considerable distance medially (Fig. 7 A). The ratio of distal end transverse width to humerus proximodistal length is 0.38, which is equaled or exceeded only by Apatosaurus and a few titanosaurs (Poropat et al., 2016; Table S 2 in Supplemental Data 1). Distally, the anterior surface of the humerus is flat, apart from the relatively large lateral and medial anterodistal processes (sensu Upchurch et al., 2015) (Fig. 8 B). Although the relative size of these anterodistal processes is difficult to quantify, they are very reduced or absent in Chubutisaurus and titanosaurs (D’ Emic, 2012), and are particularly large in several CMTs (Remes, 2008), such as Chuanjiesaurus (Sekiya, 2011) and Huangshanlong (Huang et al., 2014). Enlarged (Huang et al., 2014) and / or anteriorly directed (Ren et al., 2018) anterodistal processes have been regarded as a synapomorphy of Mamenchisauridae: however, reduction and loss of these processes appears to be the derived state (D’ Emic, 2012), and increased process size requires quantification and more comparative work before it can provide support for mamenchisaurid affinities. In Rhomaleopakhus, the distal articular surface is rugose and does not expand up onto the anterior face of the shaft, unlike the humeri of some titanosaurs (Wilson and Carrano, 1999; Wilson, 2002). The ulnar and radial condyles are not strongly divided from each other, and the former is somewhat larger than the latter. Remes (2008) suggested that mamenchisaurids possess a unique distal humeral configuration. In Klamelisaurus, Omeisaurus tianfuensis, and Mamenchisaurus youngi, the lateral condyle (which Remes [2008] termed the ‘ radial’ condyle, but which has become the ulnar condyle in sauropods because of the rotation of the antebrachium [Bonnan, 2003]), is larger than the radial one. Moreover, the ulnar and radial condylar surfaces have long axes that are at ∼ 90 ° to each other in distal end view, with the former directed anterolaterally. This results in the lateral part of the distal end having a distinct subtriangular profile, formed by fairly straight anterolateral and posterolateral margins that meet each other at an acute angle (e. g., He et al., 1988: fig. 44 B; Ouyang and Ye, 2002: fig. 35 F; Sekiya, 2011: figs. 38 C, 39 C). In many other sauropods, this lateral portion is more semicircular or subquadrate in distal view (see Upchurch et al., 2015: fig. 4; N. B., Upchurch et al. ’ s fig. 4 A shows the distal end profile of the right humerus of Mamenchisaurus youngi incorrectly labelled as the left). Rhomaleopakhus possesses the same distal end profile seen in other CMTs (Fig. 8 B): however, several non-CMTs also possess this state and, in any case, it is potentially the plesiomorphic eusauropod condition (Mannion et al., 2019 a). In Rhomaleopakhus, the lateral third of the flat distal end surface is quite strongly beveled (∼ 30 ° relative to the plane lying perpendicular to the proximodistal long-axis of the humerus) (Fig. 7 A): as a result, it faces laterodistally. This feature, however, does not seem to have a clear phylogenetic significance; it occurs sporadically in distantly related taxa such as Amargasaurus, Anhuilong, Haestasaurus, Limaysaurus, Mamenchisaurus youngi, and Saltasaurus (Ouyang and Ye, 2002; Upchurch et al., 2015; Ren et al., 2018; Mannion et al., 2019 a). The supracondylar (= olecranon or cuboid) fossa, and the medial and lateral ridges that bound it on the distal part of the posterior surface of the shaft, are partially obscured by the packing material upon which the humerus rests (Fig. 8 B). However, this fossa is not deep, unlike those of Giraffatitan and several somphospondylans (Upchurch et al., 2004 a, 2015; D’ Emic, 2012), and the associated ridges are broadly rounded transversely rather than acute. Ulna — The ulna is complete apart from a small amount of material missing from the proximal end (Figs. 6, 9 A – F). It is extremely robust, with one of the highest proximal end maximum width to proximodistal length ratios (0.50) of any sauropod, although Opisthocoelicaudia has a ratio of 0.51 (Table S 2 in Supplemental Data 1). The expanded proximal end is triradiate because of the presence of well-developed anterolateral, anteromedial, and posteromedial processes. As in other sauropods, the anterolateral and anteromedial processes define a deep concavity that receives the proximal end of the radius (Wilson and Sereno, 1998). In proximal view (Fig. 9 E), the ulna of Rhomaleopakhus has a ‘ V’ - shaped profile, rather than the ‘ T’ - shape seen in several somphospondylans (Upchurch et al., 2015). The angle between the anteromedial and anterolateral processes is ∼ 70 °, which is the derived state (i. e., less than 80 °) that occurs in most sauropods (including Chuanjiesaurus, Mamenchisaurus youngi, and Klamelisaurus), except some nonneosauropods, such as Shunosaurus, Omeisaurus tianfuensis, Anhuilong, Huangshanlong, Bellusaurus, and Cetiosaurus, as well as several titanosaurs, in which this angle is greater than 80 ° and often approaches 90 ° (Huang et al., 2014; Tschopp et al., 2015 a; Poropat et al., 2016; Ren et al., 2018; Moore et al., 2020). In Rhomaleopakhus, the anteromedial to anterolateral process length ratio (sensu Upchurch et al., 2015) is 1.72 (N. B., the measurements in Table 3 give a ratio of 1.25, but these are the maximum lengths of the processes, not their lengths measured to the intersection of process long-axes, as defined by Upchurch et al. [2015: fig. 13 A]). This ratio typically ranges between 1.6 – 1.8 in non-neosauropod eusauropods (e. g., Vulcanodon, Cetiosauriscus, Ferganasaurus), 1.0 – 1.3 in most diplodocoids and non-titanosauriform macronarians, and> 1.5 in titanosauriforms (with values> 1.6 in titanosaurs such as Opisthocoelicaudia and ≥ 2.0 in Epachthosaurus and Cedarosaurus) (Upchurch et al., 2015: table 2). The anteromedial process of the proximal end of the Rhomaleopakhus ulna has a strongly concave articular surface (Fig. 9 A – D), as also occurs in many titanosaurs (Upchurch, 1995, 1998), several non-neosauropod eusauropods such as Janenschia and Haestasaurus (Bonaparte et al., 2000; Upchurch et al., 2015; Mannion et al., 2019 a), and in a more shallowly concave form in Chuanjiesaurus (Sekiya, 2011). Dong (1997) stated that the olecranon process is relatively low in Rhomaleopakhus, although this region is moderately projected, which is emphasized by the concave proximal surface of the anteromedial process. Similarly developed olecranon processes are seen in Mamenchisaurus youngi (Ouyang and Ye, 2002: fig. 36), Chuanjiesaurus (Sekiya, 2011: fig. 40), Haestasaurus (Upchurch et al., 2015), Janenschia (Bonaparte et al., 2000; Mannion et al., 2019 a), and several titanosaurs (Upchurch, 1995; Wilson and Carrano, 1999; Upchurch et al., 2004 a). In Rhomaleopakhus, the posteromedially directed process of the proximal end creates a concavity on the posteromedial surface that does not fade out until approximately the midlength of the element, whereas the lateral surface is flat or slightly convex anteroposteriorly. In horizontal cross-section, the proximal portion of the ulna retains the triradiate configuration, but by midlength it is elliptical, with the long-axis of this ellipse oriented anteromedially. There is a prominent ridge for a ligamentous attachment to the radius, located on the anteromedial surface of the shaft at ∼ 100 mm above the distal end. The distal end of the ulna is expanded both anteroposteriorly and transversely relative to the shaft. In distal view (Fig. 9 F), the margins of this surface are strongly convex laterally and posteriorly, but slightly concave anteromedially, resulting in a comma-shaped distal profile, as is typical for most non-titanosaurian sauropods (Upchurch et al., 2015). The distal articular surface is mildly convex anteroposteriorly and transversely. Radius — The radius is complete and is 63 % of the length of the humerus. This is broadly similar to the condition in many other sauropods, which tend to have values ≥ 65 % (Yates and Kitching, 2003; Mannion et al., 2013). For example, this value is ∼ 66 % in Mamenchisaurus youngi (Ouyang and Ye, 2002) and ranges from 65 – 76 % in specimens referred to Omeisaurus (He et al., 1988; Ren et al., 2018). By contrast, this ratio is reduced in titanosauriforms (Mannion et al., 2013) and many CMTs (Ren et al., 2018), with particularly low values of 58 % and 50 % in Huangshanlong and Anhuilong, respectively (Huang et al., 2014; Ren et al., 2018). The radius of Rhomaleopakhus is a robust element with expanded proximal and distal ends relative to the shaft (Dong, 1997) (Fig. 9 G – J). The maximum widths of the proximal and distal ends are subequal, the proximal end transverse width to radius proximodistal length ratio is 0.31, and the distal end is ∼ 1.3 times as wide as the shaft at its midlength (Table 3). The proximal end surface is flat, with a central shallow concavity and a slightly convex portion around both its anterior and lateral margins. In proximal view (Fig. 9 K), the radius has a ‘ D’ - shaped profile, comprising a straight posterior margin (that becomes mildly concave towards the medial corner), and strongly convex anterior and lateral margins. This proximal profile appears to be plesiomorphic for sauropods, contrasting with the derived subtriangular profile with pointed medial process seen in many titanosauriforms (Upchurch et al., 2015: fig. 9), and the anteroposteriorly narrow morphology that characterizes some turiasaurians (Mateus et al., 2014). Approximately 100 mm below the mildly concave posteromedial margin of the proximal end, on the posterior surface, there is a prominent 100 mm long ridge that projects posteromedially. Titanosaurs, such as Epachthosaurus, Rapetosaurus, and Saltasaurus, usually have a ridge on the posterior surface of the radius that extends along much of the element’ s length (Curry Rogers, 2005, 2009; Mannion et al., 2013), and Ren et al. (2018: fig. 4 C) described a ‘ lateral ridge’ (‘ lr’) on the proximal part of the Anhuilong radius. However, the morphology and position of the short, prominent and posteromedially directed ridge seen in Rhomaleopakhus appears to be unique and is provisionally regarded as an autapomorphy. The radius is twisted along its length such that the long-axis of the proximal articular surface is set at about 90 ° to that of the distal end. As a result, the posterior surface of the shaft turns to face laterally as it approaches the distal end. Such torsion of the radius is rare among sauropods (Mannion et al., 2013), although it has also been observed in the somphospondylan Huabeisaurus (D’ Emic et al., 2013) and a few titanosaurs (e. g., Epachthosaurus – Poropat et al., 2016; Malawisaurus – Gomani, 2005; Rapetosaurus – Curry Rogers, 2009). At midlength, the cross-section through the shaft is elliptical in Rhomaleopakhus, with the radius being wider transversely than anteroposteriorly. There is a prominent vertical ridge on the posterolateral surface, located at approximately onefifth of element length from the distal end. This matches the prominent ridge on the anteromedial surface of the shaft of the ulna, close to the distal end, suggesting that these two ridges marked the location of a strong interosseous ligament (Upchurch et al., 2004 a). In medial view (Fig. 9 J), the distal end surface is set at an oblique angle to the long axis of the shaft such that it slopes anteroproximally (N. B., this would be proximolateral beveling of the distal end, in anterior view, if the radius was not twisted through 90 ° along its length). As a result, the distal end surface is set at ∼ 15 ° to the plane perpendicular to the proximodistal longaxis of the radius. Non-neosauropod eusauropods (such as Shunosaurus and Mamenchisaurus), and at least some rebbachisaurids, display no such beveling of the distal radius, whereas turiasaurians and several titanosaurs have angles of ∼ 25 ° or higher (Wilson, 2002; Mannion et al., 2019 a). The degree of distal radial beveling in Rhomaleopakhus is similar to that seen in several nonneosauropod eusauropods, including Omeisaurus tianfuensis, Chuanjiesaurus, and Jobaria, as well as some neosauropods such as Diplodocus and Giraffatitan (Mannion et al., 2019 a). In Rhomaleopakhus, beveling of the distal end extends uniformly across the entire articular surface, as occurs in some titanosaurs such as Opisthocoelicaudia and Saltasaurus (Wilson, 2002; Mannion et al., 2013; Upchurch et al., 2015). This contrasts with the more typical form of distal beveling in other sauropods, in which the medial half of the distal end surface is perpendicular to the long-axis of the shaft, such that the beveled section is limited to the lateral half (Mannion et al., 2013; Upchurch et al., 2015). The distal end has a ‘ D’ - shaped outline (Fig. 9 L), with the derived, nearly straight posterior (= lateral because of shaft torsion) margin observed in other sauropod radii, rather than the plesiomorphic convex margin that occurs in non-sauropod sauropodomorphs (Wilson & Sereno, 1998). In fact, this posterior distal margin is mildly concave between the posterolateral and posteromedial ‘ condyles. ’ Such distal radial condyles were first discussed by D’ Emic (2012, 2013), and their wider distribution among sauropods was further investigated by Upchurch et al. (2015). According to the latter, such condyles tend to occur in neosauropods, but with several reversals in, for example, some titanosaurs. Laterally, the distal surface of the Rhomaleopakhus radius is mildly convex, whereas the central and medial portions are markedly concave: this contrasts with the uniformly convex distal surfaces seen in nearly all other sauropods (Janensch, 1961; Upchurch et al., 2004 a). Ren et al. (2018) described the distal end surface of the radius of Anhuilong as also being flat over most of its extent, with a convex area placed posteriorly and medially. Thus, while Rhomaleopakhus and Anhuilong potentially share the unusual flattening of the distal articular surface, the location of the residual convex area differs. Consequently, this concavity is regarded as an autapomorphy of Rhomaleopakhus. Manus — The right manus is virtually complete, including one carpal element, five metacarpals, and two phalanges per digit except for digit V (see below) (Fig. 10). These elements are preserved in articulation, but many details are obscured by matrix (especially the ‘ palmar’ surfaces of the metacarpals – see below for definitions of the orientations of the latter). A large, flat, block-like carpal is situated above metacarpals I and II (Fig. 10 A, D) (N. B., Dong [1997] stated that this element also articulated with metacarpal III, but this is not supported by our observations of the specimen). Possession of block-like carpals is a synapomorphy of Eusauropoda according to Wilson and Sereno (1998), contrasting with the carpals of nonsauropod sauropodomorphs, which tend to have proximodistally more rounded margins, and proximal and distal surfaces that are less parallel (Yates, 2007). Sauropods have often been interpreted as possessing ossified distal carpals only (e. g., Gauthier, 1986; Wilson and Sereno, 1998; Upchurch et al., 2004 a), although an ossified proximal carpal is probably present in at least ‘ Bothriospondylus madagascariensis ’ and Apatosaurus (Läng and Goussard, 2007; Tschopp et al., 2015 b). The Rhomaleopakhus carpal resembles the ‘ medial distal carpal’ in Camarasaurus (Tschopp et al., 2015 b). With the exception of Apatosaurus (Hatcher, 1902; Gilmore, 1936), the largest carpal in the sauropod wrist is generally placed over metacarpals I and II and articulates closely with them. This element could represent: a single enlarged distal carpal I; a fusion of distal carpals I and II; or the fusion of the intermedium, one or two centrales, and distal carpal I (as proposed for ‘ Bothriospondylus madagascariensis ’ by Läng and Goussard, 2007). If the latter interpretation is correct, then we cannot regard the carpal of Rhomaleopakhus as being either a proximal or distal carpal since it would be a composite with contributions from each of the three rows of carpals found in the plesiomorphic archosaurian wrist. The margins of the Rhomaleopakhus carpal are damaged, such that its outline can only be estimated as subcircular to elliptical, with the long axis running transversely. The approximate transverse: anteroposterior width ratio is 1.23, similar to the values seen in several non-neosauropod eusauropods such as Shunosaurus and turiasaurians, but differing from the higher values (> 1.4) observed in many neosauropods (Royo-Torres et al., 2014; Mannion et al., 2017). The proximal surface of the carpal is irregularly flat, with a slight convexity near the posterior and lateral margins. The posterolateral edge has a small vertical groove, suggesting that this portion is possibly a small medial part of a more lateral carpal, perhaps supporting the view that this large medial element is a composite structure (Läng and Goussard, 2007). The distal surface of the carpal cannot be examined because of the presence of matrix and the proximal ends of the metacarpals. The true number of ossified carpals in Rhomaleopakhus cannot be determined. Sauropods appear to show a trend towards loss and / or fusion of carpals through their evolutionary history, with five and three-to-four elements in the early-diverging taxa ‘ Bothriospondylus madagascariensis ’ and Shunosaurus, respectively, two in non-neosauropod eusauropods and nontitanosauriform macronarians, one in diplodocids (such as Apatosaurus and Diplodocus) and Giraffatitan, and complete loss in some titanosaurs such as Alamosaurus and Opisthocoelicaudia (Janensch, 1961; Upchurch, 1998; Upchurch et al., 2004 a; Apesteguía, 2005; Remes, 2008; Tschopp et al., 2015 b). The single carpal in Apatosaurus (Gilmore, 1936; Bonnan, 2003) is placed centrally over metacarpals II – IV and has a proximal surface that conforms closely to the distal ends of the ulna and radius (Tschopp et al., 2015 b). Although it is possible that Rhomaleopakhus only possessed one carpal and that this taxon differed from Apatosaurus in having this placed medially over metacarpals I and II, we consider it more likely that there was at least one additional (lateral) carpal placed over metacarpal III (as in Mamenchisaurus youngi: Ouyang and Ye, 2002) or metacarpal V (as in Camarasaurus, Atlasaurus, and possibly Argyrosaurus: Apesteguía, 2005; Tschopp et al., 2015 b). This view is supported by the possible presence of a small portion of a more lateral carpal (as described above) which, if correctly identified, would suggest that the wrist of Rhomaleopakhus most closely resembled that of Mamenchisaurus youngi (Ouyang and Ye, 2002). The stout metacarpals have a semicircular or horseshoeshaped arrangement with their long axes oriented vertically (Fig. 10); this is a eusauropod synapomorphy (Upchurch, 1995, 1998; Yates, 2007; McPhee et al., 2014; Apaldetti et al., 2018). The arc of a circle covered by this metacarpal arcade is ∼ 270 °, as occurs in neosauropods (Upchurch, 1998; Wilson and Sereno, 1998; Bonnan, 2003; Apesteguía, 2005; Remes, 2008) and several taxa close to the neosauropod radiation, such as Mamenchisaurus youngi (Ouyang and Ye, 2002) and ‘ Bothriospondylus madagascariensis ’ (Läng and Goussard, 2007). This contrasts with the apparently less strongly curved arcades (∼ 90 – 180 °) seen in other non-neosauropod eusauropods, such as Omeisaurus tianfuensis (Bonnan, 2003), Shunosaurus (ZDM T 5402; PU pers. observ., 1995), and possibly Ferganasaurus (Alifanov and Averianov, 2003) (N. B., we are skeptical about the accuracy of the reconstruction of the manus of the latter based on, for example, an anomalous arrangement of the metacarpals as reconstructed in distal end view: see Alifanov and Averianov, 2003: fig. 9 C). The vertically oriented metacarpals, in a ‘ tubular colonnade, ’ make conventional directional anatomical terms ambiguous unless care is taken to define them (e. g., see Upchurch, 1994). Here, we treat the metacarpals as if they were laid on a flat surface side-by-side. As such, ‘ lateral, ’ ‘ medial, ’ ‘ dorsal, ’ and ‘ ventral’ refer to surfaces on the shafts of the metacarpals, rather than how these surfaces would face in the articulated manus. As a result, the dorsal surfaces face outwards, ventral surfaces face towards the center of the tubular colonnade, and metacarpals typically contact each other via portions of their lateral and medial surfaces. In correct articulation, the phalanges are placed in a more conventional orientation, with their ventral surfaces facing approximately downwards. Therefore, no additional definitions are required for phalanges, although it should be borne in mind that, for example, the medial surface of the pollex claw would have faced posteriorly or posteromedially in life with respect to the sagittal plane of the animal (Fig. 10). The proximal ends of metacarpals I and II in Rhomaleopakhus are obscured by the overlying carpal. In anterior view (Fig. 10 A), the proximal ends of metacarpals I – III are level with each other, whereas that of metacarpal IV is displaced distally. The proximal end of metacarpal V is, in turn, displaced distally with respect to metacarpal IV. These displacements of metacarpals IV and V are presumably the result of post-mortem distortion rather than an unusual morphology possessed by the living animal. In metacarpals III – V, the exposed proximal end surfaces are generally flat and mildly rugose. Metacarpal I is short compared with the other metacarpals (e. g., it is only 0.67 of the averaged length of metacarpals II and III: Table 4) and shorter than the ungual on digit I. Such a relatively short metacarpal I is the plesiomorphic state that occurs in non-sauropod sauropodomorphs, non-neosauropod eusauropods (such as Shunosaurus, Omeisaurus tianfuensis, and Mamenchisaurus youngi), and, to a lesser extent, in diplodocines (Table S 2 in Supplemental Data 1). In Rhomaleopakhus, metacarpal I is substantially longer along its medial margin than on its lateral one (Table 4): this reflects the beveling of the distal end relative to the long-axis of the shaft. This condition is a derived state that occurs in most eusauropods except Shunosaurus, with a reversal to the plesiomorphic state in most titanosauriforms (Wilson, 2002; Mannion et al., 2013). As in Chuanjiesaurus (Sekiya, 2011), Turiasaurus (CPT- 1195 - 1210; PU and PDM pers. observ., 2009), and many neosauropods (Wilson, 2002), the distal end of metacarpal I is not divided into two distinct condyles. In dorsal view, the proximal end of metacarpal II is strongly expanded to overhang the medial surface of its shaft (Fig. 10 A, C). This feature is absent in taxa such as Mamenchisaurus youngi (Ouyang and Ye, 2002: fig. 38 B), Apatosaurus ajax (Upchurch et al., 2004 b: pl. 8, fig. D), and Camarasaurus (Tschopp et al., 2015 b: fig. 11), but a medial process appears to be developed to some extent in Ferganasaurus (Alifanov and Averianov, 2003: figs. 9, 10), Giraffatitan (Janensch, 1961: 194, fig. 1 a), and Alamosaurus (Gilmore, 1946: fig. 10). A minimum shaft width to proximodistal length ratio of <0.2 in metacarpal II was proposed as a diagnostic character of Chuanjiesaurus by Sekiya (2011); however, this ratio is 0.19 in Rhomaleopakhus, similar to those of several other non-neosauropod eusauropods, such as Omeisaurus tianfuensis, Mamenchisaurus youngi, and Turiasaurus (Poropat et al., 2016). The proximal articular surface of metacarpal III is subtriangular in outline (Fig. 10 D). This element is the longest of the five metacarpals, as is the case in most eusauropods (Poropat et al., 2015 a), although it only slightly exceeds the length of metacarpal II (Table 4). The length of metacarpal III is 0.42 of radius length, similar to the condition in taxa such as Mamenchisaurus youngi and Apatosaurus, but lower than the derived 0.45 ratio employed as a synapomorphy of Macronaria by Wilson and Sereno (1998; Table S 2 in Supplemental Data 1). Its proximal end lacks the mediolaterally expanded morphology that characterizes brachiosaurids, as well as Atlasaurus and Jobaria (Mannion et al., 2017). Metacarpal IV also has a subtriangular proximal end but differs from metacarpal III by possessing a ventromedially directed palmar process (Fig. 10 D). Unlike the metacarpal IVs of several brachiosaurids and a few titanosaurs, that of Rhomaleopakhus lacks the chevron-shaped proximal end profile that wraps around the proximal end of metacarpal V (D’ Emic, 2012; Mannion et al., 2013). The proximal end of metacarpal V is semicircular to slightly subrectangular in outline, with a flattened medial surface that articulates with metacarpal IV (Fig. 10 D). Metacarpal V is twisted along its length such that the long-axes of its proximal and distal ends lie at ∼ 90 ° to each other, and this degree of twisting has also been reported in Ferganasaurus (Alifanov and Averianov, 2003). Some torsion of metacarpal V also occurs in neosauropods but is less extreme than in Rhomaleopakhus and Ferganasaurus (Apesteguía, 2005; Bedwell and Trexler, 2005; Tschopp et al., 2015 b). For example, in Camarasaurus and Diplodocus the amount of torsion is ∼ 25 – 30 ° (Bedwell and Trexler, 2005; Tschopp et al., 2015 b), and in the titanosaur Epachthosaurus it is ∼ 45 ° (UNPSJB-PV 920; PU and PDM pers. observ., 2013). The phalanges are hyper-extended such that they lie on the dorsodistal parts of each metacarpal, except in metacarpal I where the phalanx obscures the distal end (resulting in the distal end surfaces being visible in metacarpals II – V) (Fig. 10 E). In general, the distal articular surfaces of the metacarpals are expanded dorsoventrally, and especially transversely, and have a rounded subrectangular outline. These surfaces are gently saddle-shaped, with mild midline grooves between slightly expanded lateral and medial condyles. The ventral portions of the distal ends are flattened and have a rugose texture. Generally, the distal articular surfaces do not extend onto the dorsal surfaces of the shafts: this is a derived state seen in titanosauriforms (Gimenez, 1992; Salgado et al., 1997; Apesteguía, 2005; D’ Emic, 2012; Mannion et al., 2013) that also occurs convergently in rebbachisaurids (Mannion et al., 2019 a). Rhomaleopakhus lacks the additional flanges, close to the distal ends of the metacarpals, that helped bind them together in some titanosaurs (Apesteguía, 2005). Dong (1997) stated that IVPP V 11121 - 1 has a phalangeal formula of 2 - 2 - 2 - 1 - 1; however, it is actually 2 - 2 - 2 - 2 - 1 (Fig. 10 E). Retention of two phalanges on manual digit IV occurs in early-branching sauropods such as Shunosaurus, but in most neosauropods the phalangeal formula has been reduced to 2 - 2 - 2 - 1 - 1, 2 - 2 - 1 - 1 - 1, or 2 - 1 - 1 - 1 - 1 (in diplodocoids and early-diverging macronarians), or the phalanges are completely lost (apart from a rudimentary phalanx IV- 1) in titanosaurs such as Epachthosaurus, Alamosaurus, and Opisthocoelicaudia (Gilmore, 1946; Borsuk-Białynicka, 1977; Salgado et al., 1997; Bonnan, 2003; Martínez et al., 2004; Upchurch et al., 2004 a, b; Mannion et al., 2013; Poropat et al., 2015 b; Tschopp et al., 2015 b). The phalanges (except for the ungual of digit I) of Rhomaleopakhus are wider transversely than they are proximodistally, which is a eusauropod synapomorphy (Wilson, 2002; Upchurch et al., 2004 a, 2007 b; Yates, 2007). The phalanges in the proximal row have flattened or mildly concave ventral surfaces. These phalanges are also expanded transversely at their distal ends, so that they are wider at this point than they are at midlength. Phalanx I- 1 is subrectangular in dorsal view, decreasing only slightly in proximodistal length towards its medial margin. Similar subrectangular manual phalanx I- 1 s are seen in several other non-neosauropod eusauropods, such as Ferganasaurus (Alifanov and Averianov, 2003: fig. 11) and Omeisaurus tianfuensis (He et al., 1988: pl. XIV, fig. 6), as well as the titanosauriform Giraffatitan (Janensch, 1961). Thus, Rhomaleopakhus retains the plesiomorphic manual phalanx I- 1 dorsal profile, rather than the derived trapezoidal outline seen in Turiasaurus (Mannion et al., 2019 a) and Jobaria (Läng and Goussard, 2007), or the even more strongly wedge-shaped outline seen in several diplodocids and the non-titanosauriform eusauropod specimen MfN MB. R. 2093 (previously referred to Janenschia but removed from that genus by Mannion et al. [2019 a]) (Upchurch et al., 2004 a; Tschopp et al., 2015 b). The proximal and distal ends of phalanx I- 1 are obscured by the metacarpal and ungual respectively, but the general outline of the transverse cross-section is an irregular ‘ D’ - shape, with rounded medial, dorsal, and lateral surfaces, and a flattened ventral surface. There is no lappet-like projection from the proximodorsal margin. Such a lappet occurs as the plesiomorphic condition in early-branching eusauropods such as Shunosaurus, Omeisaurus tianfuensis, Turiasaurus, and Zby, but is absent in most neosauropods (Mannion et al., 2019 a). Distally, the phalanx terminates in well-developed, rounded lateral and medial condyles. Phalanx I- 2 is a large, robust ungual that is transversely compressed. As in most other sauropods, this ungual is much longer than phalanx I- 1 (Fig. 10 E), whereas in Giraffatitan the two elements are subequal in length (Janensch, 1922). In dorsal view, the proximal articular surface of the Rhomaleopakhus ungual is approximately perpendicular to the long axis of the claw: this is the plesiomorphic state, whereas in neosauropods (e. g., Apatosaurus — Upchurch et al., 2004 b; Camarasaurus — Tschopp et al., 2015 b; Giraffatitan — Janensch, 1961) this surface is set at an oblique angle to the long-axis such that it faces proximolaterally. The Rhomaleopakhus ungual bears a groove on each of the lateral and medial surfaces, with the former being positioned lower than the latter. The ventral side merges smoothly into the medial surface but meets the lateral surface at a sharper edge. Phalanx II- 1 is subrectangular in dorsal view. The medial, lateral, and dorsal surfaces round smoothly into each other, although the medial edge meets the ventral surface at a slightly more acute angle than the lateral edge. The ventral surface is nearly flat. Phalanx II- 2 is larger than phalanx II- 1 (Table 4) (contra Dong, 1997) but seems to have a pathological distal termination. It appears damaged and ends irregularly, with a cavity running down the central part of its ventral surface (Fig. 10 E). Phalanx III- 1 is large and dorsoventrally compressed, with two distinct distal condyles. Whereas the dorsal surface meets the proximal and distal end surfaces at an obtuse angle in lateral or medial views, the articular surfaces expand ventrally to make the ventral surface concave proximodistally. In dorsal view, this element narrows slightly in transverse width towards its distal end. Phalanx III- 2 is similar to phalanx III- 1, but is slightly smaller, with its distal end rounding transversely in dorsal view so that it curves into the corners of the proximal end. It is therefore more semicircular, rather than rectangular, in dorsal profile. This element is also bowed upwards in distal end view. Phalanx IV- 1 is large, dorsoventrally compressed, and subrectangular in dorsal outline. The medial condyle is large and rounded, and projects more distally than the lateral one. In dorsal view, the medial margin is mildly concave, whereas the lateral one is straighter. This element tapers slightly transversely towards the distal end in dorsal view. Phalanx IV- 2 is a very small, flattened hemisphere of bone that sits in the intercondylar groove on the distal end of phalanx IV- 1. The dorsal and ventral surfaces are slightly concave longitudinally because of the expansion of both ends. The lateral condyle of the distal end is enlarged dorsoventrally, but the medial condyle is indistinct. Phalanx V- 1 is large, subrectangular, and dorsoventrally compressed. The dorsal and ventral surfaces are slightly concave longitudinally because of the expansion of the proximal end. The element tapers in dorsoventral thickness towards its distal end. The distal surface is generally convex both dorsoventrally and transversely, with little division into two separate condyles. Thus, this phalanx in Rhomaleopakhus still resembles the other proximal phalanges, as it does in several other sauropods such as Apatosaurus (Upchurch et al., 2004 b): this contrasts with phalanx V- 1 of Camarasaurus, which is very irregular and rather different from the other proximal phalanges (Tschopp et al., 2015 b).

Species Diagnosis — Rhomaleopakhus turpanensis is diagnosed on the basis of three autapomorphies: (1) humeral deltopectoral crest terminates distally in a transversely narrow ridge that is separated from the main body of the crest by distinct lateral and medial grooves; (2) prominent (100 mm long) ridge, projecting posteromedially, on posterior surface of radial shaft, a short distance below the proximal end; and (3) radial distal articular surface markedly concave in central and medial portions. In addition, Rhomaleopakhus turpanensis possesses one of the most robust ulnae of any known sauropod (maximum proximal end width to proximodistal length ratio is 0.50; Table S 2 in Supplemental Data 1), and is currently the only known non-somphospondylan eusauropod with the long-axes of the proximal and distal surfaces of the radius twisted through ∼ 90 ° with respect to each other.

Etymology — Rhomaleos (ancient Greek, masculine) equals ‘ robust’ (pertaining to the body), and pakhus (ancient Greek, masculine) equals ‘ forearm. ’ The species name refers to the Turpan Basin, China, where the holotype was found.

Holotype — A right forelimb, IVPP V 11121 - 1 (Figs. 6 – 10; Tables 3 and 4), consisting of the humerus, ulna, radius, one carpal, and virtually complete manus of a single individual.

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).